Nobody likes uncertainty. Everybody on the receiving end of risk communications prefers those communications to be definitive, not tentative.

But the painful truth is that the concept of “risk” is intrinsically uncertain, and risk communicators have no choice but to address the uncertainty, whether they do so well or badly.

This column will concentrate on one complicated example of uncertainty communication, the severe outbreak of E. coli food poisoning that preoccupied Germany and much of Europe from late May until early July 2011. We will focus our lens on the most important aspect of uncertainty about this outbreak: uncertainty about what food was contaminated, how it got contaminated, and where.

We’ll ignore a number of other uncertainties that were raised by the outbreak – whether the source might turn out to be bioterrorism; whether this previously rare strain of E. coli is likely to become common now; whether the new strain was really unusually virulent or just looked that way because a lot of milder cases went untested; whether the benefits of eating suspect raw vegetables outweighed the risks even during this serious outbreak; etc.

As we watched the E. coli outbreak, we flagged several other “uncertainty risk communication” stories as potential case studies. Among them:

  • The World Health Organization’s International Agency for Research on Cancer decided to classify mobile telephones as a “possible human carcinogen,” a super-broad category that includes everything from chloroform and DDT to pickled vegetables and sand.
  • The threat of floods in the U.S. Midwest forced officials to decide how much water to release from dams and levees, intentionally flooding low-population farmlands in designated floodways in order to protect more densely populated towns and cities … risking upstream drought if they released too much.
  • An Italian judge announced that seven members of a committee responsible for assessing natural disaster risks would be tried for manslaughter for having made over-reassuring statements about earth tremors in March 2009, failing to anticipate the earthquake that devastated the town of L’Aquila in April.
  • A hotel housekeeper in New York City accused high-ranking French politician Dominique Strauss-Kahn of sexually assaulting her in his suite, but prosecutors’ early overconfidence in the complainant’s credibility gave way to uncertainty as some parts of her story fell apart.
  • The debate over the safety of shale gas extraction continued to heat up, focusing on conflicting claims and uncertain data about the safety of “fracking” – underground high-pressure injection of a mixture of water, sand, and chemicals in order to fracture (“frack”) the shale and release the gas.

We settled on the German E. coli outbreak because it is a particularly rich “bad example” of uncertainty risk communication. But the issues it raises are generic. Before launching into the case study, we want to lay out some of these generic issues – starting with the fact that nobody likes uncertainty.

The Dilemma of Uncertainty Communication

If you’re warning us that X is dangerous, we want you to be sure that it’s dangerous, not just worried that it might be. And we want you to be sure that the danger is coming from X, not from Y or Z. We’re reluctant to bother to take precautions that could turn out unnecessary or misplaced, precautions against an uncertain risk from an uncertain source.

And if you’re reassuring us that X isn’t dangerous, we want you to be sure about that too. “X might turn out to be dangerous, but I don’t think so” isn’t a particularly reassuring message. People who are convinced that X is dangerous will pounce on your uncertainty, pointing out that even you admit it might be. People who are worried and looking for a reason to relax will stay worried and wish you’d said something more clearly reassuring. And people who are confident that X is safe will resent the half-heartededness of your support.

A Response from the BfR

On November 24, 2011, Germany’s Federal Institute for Risk Assessment (BfR) emailed us a response to this column. As the BfR requested, we have posted its response on our website. Our point-by-point reply is also posted.

The BfR’s most significant point is that it posted an “opinion” on its website on May 26, the same day the Hamburg Institute announced that it had found Spanish cucumbers contaminated with E. coli and that the cucumbers were probably the source of the outbreak. Although the BfR opinion focused on what foods consumers should avoid, it did contain the information that there was not yet any evidence linking the outbreak to the Spanish cucumbers – which undermines our contention in this column that German and European authorities unwisely let the Hamburg Institute’s overconfident claim stand unrebutted for five crucial days.

On the other hand, the major media in Germany and elsewhere also missed the May 26 BfR Comment, as did officials throughout Europe. Neither the BfR nor any other agency (outside of Spain) issued a news release or media statement indicating any dissent from the Hamburg Institute’s claim, which the media were depicting as the universally shared view of the German authorities.

The universal human preference for certainty in risk communications leads us to put considerable pressure on risk communicators to tell us they’re certain, even if they’re not. It also leads us to hear, remember, and report what they say as more certain than it was.

At every stage in the risk communication process, in fact, the pressure is toward certainty. The audience wants you to sound certain. Your employer wants you to sound certain. Journalists want you to sound certain. You’d probably rather sound certain too. So maybes tend to get lost in the shuffle.

That’s half the dilemma of uncertainty communication: Everybody wants you to sound certain. Here’s the other half: Our understandable yearning for certainty in risk communications doesn’t keep most of us from punishing communicators who sounded certain and turned out wrong. Overconfident false alarms lead to lost credibility and considerable derision. Overconfident false reassurances lead to lost lives and a devastated reputation.

Nor can you avoid the dilemma by shutting up until you’re certain (as critics often claim you should). When imperfect, tentative information is all you have, then imperfect, tentative information is what you must give people so they can decide how best to cope. If your preliminary data suggest it, you must warn people that so far the hurricane looks bad and seems to be headed their way – before you’re sure that it won’t lose force or change direction. If your preliminary data suggest it, you must reassure people that so far the pandemic looks mild and seems to be winding down – before you’re sure that there won’t be a much more severe second wave.

Officials charged with warning publics about possible risks cannot afford to wait until they’re sure. How long to wait is a tough judgment call, but “sooner rather than later” is almost always the answer.

Imagine an engineer getting evidence that a bridge might collapse, and deciding to let people keep crossing the bridge because the evidence is uncertain and the bridge might be safe after all. Or imagine a doctor diagnosing that the patient might have cancer, but doing nothing because the diagnosis is uncertain and it might turn out not to be cancer in the end. Imagine the engineer and the doctor not even mentioning the possibility that the bridge might collapse or the patient might have cancer. This is not treating people like grown-ups.

Of course if there’s one more test that can resolve the question in a matter of minutes, and if there’s very little chance of things getting worse (the bridge collapsing, for example) during those few minutes, and if it would do irreparable harm to take any action or even say anything until you’re sure, then any reasonable engineer or doctor would decide to wait.

But you’re not likely to be tempted to jump the gun in these situations. What gets risk communicators into trouble is succumbing to the opposite temptation, the temptation to wait too long. When it comes to warning people before you’re sure, it’s not damned if you do and damned if you don’t. It’s darned if you do and damned if you don’t. The price of your failure to warn if a disaster happens is far higher (usually) than the price of issuing a warning that turns out inaccurate or unnecessary.

That said, what can be done to reduce the price of issuing a warning that turns out inaccurate or unnecessary? This is the question that will preoccupy us as we examine uncertainty risk communication during Europe’s 2011 E. coli outbreak.

Bottom line: The wisest, most responsible course is usually to talk before you want to – that is, before you’re sure. Many of the most important, actionable risk communication messages are uncertain messages. (Telling people what foods they should avoid while you try to figure out where the contamination is coming from is such a message.) If you’re going to communicate about risk, you will need the courage to talk when your information is uncertain. And you will need the skill to express uncertainty in ways that guide your audience’s decisions and minimize the cost (to you and your audience both) if you turn out mistaken. The communication of uncertainty is a central risk communication capability.

In fact, risk communication is about uncertainty, the uncertainty of bad outcomes. The phrase “uncertain risk” is as redundant as “unexpected surprise” or “free gift.” If you know X isn’t going to happen, then obviously there’s zero risk of X. If you know X is going to happen, it’s not a risk; it’s the future. Of course there are different degrees and different kinds of uncertainty. Believing that there’s a 30% probability of X is different from believing that there’s a 70% probability of X, and both are different from not having a clue what the probability of X is. And an expert consensus that the probability of X is 30% (or that it’s 70%, or completely unknown) is different from believing that the probability is 30% but knowing there are other experts who believe it’s 70%.

Readers of our work know we think the most important variable in risk communication is outrage, not hazard. To our minds, warning people is saying things that will increase their outrage (fear, worry, concern, anger) about the risky situation; which increases their hazard perception; which increases their inclination to take, demand, or accept precautions. Reassuring people is saying things that will decrease their outrage about the situation; which decreases their hazard perception; which decreases their inclination to take, demand, or accept precautions.

Risk communicators should try to get people either more upset or less upset, in order to motivate them to consider the situation more or less serious than they did before, in order to increase or decrease their support for precautions. Ethical risk communicators aim to produce the level of outrage in their audience that matches their own current best assessment of the actual hazard.

Whether you’re trying to increase or decrease your audience’s level of outrage, explaining uncertainty is at the very heart of risk communication:

  • Explaining that the risk is uncertain.
  • Explaining what you think the probabilities are, and how much confidence you have in your opinion.
  • Explaining what evidence you have to support your view, and what additional evidence you wish you had but don’t.
  • Explaining what other experts think, the extent to which they agree or disagree, and the basis for any disagreement.
  • Explaining what you are doing to reduce your uncertainty, and when you might have more information.

And the soul of risk communication is explaining uncertainty humanly, empathically, and respectfully, in a way that shows you share the universal yearning for a certainty you are unable to provide.

Merely explaining uncertainty isn’t enough, however. Because you will tend to soft-pedal your uncertainty and sound more confident than you are; because journalists will tend to drop your maybes and on-the-other-hands and quote the most confident-sounding things you say; because the certainty-seeking audience will tend to overlook the signals of uncertainty that have managed to wriggle through the journalistic filter – for all these reasons, “explaining” uncertainty isn’t enough. You need to “proclaim” uncertainty. You need to sell it, insist on it … or it will get lost.

In 2004, one of us (Peter) wrote a column entitled “Acknowledging Uncertainty.” It has two main sections: a brief list of tips on how to sound uncertain, and a more complex protocol for “being precise about uncertainty” – for deciding how uncertain you want to sound and then finding words that sound that way. The 2004 column goes into detail about the biases and pressures that lead risk communicators to sound overconfident. But we now think it greatly underestimates the difficulty of successfully communicating uncertainty, as opposed to merely “acknowledging” it – which may get it onto the record but doesn’t necessarily get it into the minds of the audience, where it needs to be.

This column is about the need to proclaim uncertainty, not just acknowledge it.

What does proclaiming uncertainty sound like? Let’s start with a homey example. Your best friend’s outdoor wedding is scheduled for the day after tomorrow, and meteorologists are predicting a 30% chance of rain – not just showers, but a strong, heavy rain. An alternative indoor venue is available only if you act now. Obviously, “it’s going to rain” is an excessively confident warning, and “it’s not going to rain” is an excessively confident reassurance. “I think we should move indoors because it might rain” does a better job of acknowledging uncertainty. So does “it probably won’t rain so I think we should stick with the outdoor venue.”

But neither proclaims the uncertainty emphatically enough to get it across – emphatically enough so that, among other benefits, nobody will misremember afterwards (when you turn out wrong) that you were confident it would or wouldn’t rain. Here’s how to proclaim uncertainty:

I think a 30% chance of a downpour is too big a risk to take. So we should move the wedding indoors. We’ll feel foolish if it’s a beautiful day and we moved indoors for nothing. But if we stay outdoors and the storm comes, we’ll feel worse than foolish; we’ll have ruined the wedding! And we have to decide now. So let’s move to the indoor venue. We’ll still be hoping for a beautiful day, but we won’t be counting on it.

Or on the other hand:

I think we should take the chance. We knew when we chose an outdoor venue in the first place that there was a risk of rain; now we have a number, 30%. We all wish it were zero – but it’s still more than twice as likely not to rain (70%) as it is to rain (30%). And even if it rains, we might get time for the ceremony between downpours! Passing up the chance to move the wedding indoors is a gamble, but I think it’s a gamble we should take.

Yes, we know – a reporter may still seize on the first statement and write, “Palace officials announced today that the Prince’s wedding will be held indoors because of rain.” And a headline writer may “misoversimplify” even further, yielding “Rain cancels outdoor wedding plans” – written in the present tense, two days before the wedding! But by proclaiming your uncertainty, you have reduced the odds of these distortions. And you have reduced the odds of a sarcastic commentary two days later, when the sun is shining, shouting, “Remember when they said it was gonna rain for the wedding?”

Our all-time favorite example of a public health official superbly warning the public about uncertainty comes from Dr. Richard Besser, acting head of the U.S. Centers for Disease Control and Prevention at the start of the H1N1 flu pandemic (and now chief health and medical editor for ABC News). On April 24, 2009, at the very start of the pandemic, here’s what Dr. Besser said to reporters:

Before I talk about the cases and specific actions, I want to recognize some initial guiding concepts. First I want to recognize that people are concerned about this situation. We hear from the public and from others about their concern, and we are worried, as well. Our concern has grown since yesterday in light of what we’ve learned since then.
I want to acknowledge the importance of uncertainty. At the early stages of an outbreak, there’s much uncertainty, and probably more than everyone would like. Our guidelines and advice are likely to be interim and fluid, subject to change as we learn more. We’re moving quickly to learn as much as possible and working with many local, state and international partners to do so.
I want to recognize that while we’re moving fast, it’s very likely that this will be more of a marathon than a sprint. I want to acknowledge change. Our recommendations, advice, approaches will likely change as we learn more about the virus and we learn more about its transmission.
I want to acknowledge that we’re likely to see local approaches to controlling the spread of this virus, and that’s important; that can be beneficial; that can teach us things that we want to use in other parts of the country and that other people in other places may find useful. Because things are changing, because flu viruses are unpredictable and because there will be local adaptation, it’s likely that any given moment there will be confusing – or may be confusing or conflicting information available. We are very committed to minimizing and that where we find that, clearing up any of that misconception.

This isn’t easily condensed into a sound bite – and given the hard news Dr. Besser had to impart that day, it is understandable that few if any reporters actually covered his warning about uncertainty. It’s not clear reporters even heard it. Perhaps if he (and his successor, Dr. Thomas R. Frieden) had repeated it and embroidered it more often in the months that followed, the public might have come to understand that the severity of the swine flu pandemic was a moving target. On April 24, Dr. Besser did a superb job of proclaiming uncertainty. But once wasn’t enough. It needed to become a mantra, and it didn’t.

In addition to repetition and emphasis, another important way to proclaim uncertainty is to respond openly when journalists (or others) have missed your earlier uncertainty claims. We don’t mean that you should blame the media for getting it wrong – the messenger (you) should always take responsibility for all misunderstandings. It’s easy and straightforward to say something like this:

Yesterday we tried to convey how uncertain we were about [whatever]. Obviously we weren’t clear enough, because many media reports sounded like we were sure. So today I want to emphasize that, no, we wish we were sure, but in fact we are still very uncertain.

Odds are this passage won’t end up in the coverage – not if you have hard news to communicate as well. But it stands a good chance of taking up residence in reporters’ minds, so that they capture more of your uncertainty in the coverage to come.

The rest of this column will dissect a specific, complicated example of the need to proclaim uncertainty: a foodborne E. coli outbreak in Europe (mostly in Germany). The outbreak appears to be over now (though whether related outbreaks will occur is among the uncertainties), but it will be remembered as one of the developed world’s most severe foodborne illness outbreaks in modern times. As of July 21, very near the end of the outbreak, the World Health Organization tally of confirmed and probable cases stood at 4,075 illnesses (including over 900 cases with severe kidney damage) and 50 deaths in 16 countries in Europe and North America.

Introduction to E. coli and the Epidemiology of Foodborne Disease Outbreaks

Escherichia coli (E. coli for short) is a bacterium commonly found in the lower intestine and feces of warm-blooded animals, including humans. E. coli can easily get into the food supply, directly or indirectly, from the feces of an infected animal. Most strains of E. coli are harmless, even beneficial. But some strains can make you sick, and a few can make you very sick.

The most common health effect of E. coli is a minor gastrointestinal infection, resulting in a few days of diarrhea. The most serious effect is hemolytic-uremic syndrome (HUS), which leads to kidney failure and is deadly in about 10% of cases (mostly young children and the elderly). The kinds of E. coli that can cause HUS are referred to as shiga toxin-producing E. coli (STEC) or enterohemorrhagic E. coli (EHEC). One reason EHEC is so dangerous is that doctors don’t dare treat it with antibiotics, which can stimulate production and release of the toxin, and thus increase the risk of HUS.

The best known EHEC is E. coli O157:H7, the serotype (strain) responsible for the Jack-in-the-Box food poisoning outbreak of 1993. (Peter worked on some of the litigation that came out of the Jack-in-the-Box case, and emerged with a wonderful example of the risk communication seesaw.) That time the offending food was quickly identified: It was undercooked hamburger from a fast-food chain called Jack-in-the-Box. Healthy cattle harbor E. coli O157:H7 without getting sick themselves, and it’s all too easy for the bacteria to migrate from a slaughtered animal’s intestines or feces into the meat. But O157:H7 and other EHECs can get into vegetables as well (for example, if the vegetables were irrigated or washed with water contaminated with animal feces, then eaten raw). O157:H7 was rare when the Jack-in-the-Box crisis occurred. It is much more common today.

The EHEC serotype responsible for the outbreak we’re going to examine is O104:H4. Until the 2011 outbreak, only a handful of humans had been infected with E. coli O104:H4. There was one case each in Finland (2010), South Korea (2005), and France (2004), and two in Germany (2001). According to the European Food Safety Authority, “The Finnish case was travel-related with infection acquired in Egypt” – a fact we mention only because Egypt also turned out to be the source of the 2011 outbreak, and because Egypt keeps insisting that serotype has never been found in Egypt.

Nobody knows yet whether E. coli O104:H4 will remain rare or will become a widespread source of foodborne illness the way O157:H7 has done.

Tracking down the source of a foodborne disease outbreak is important for obvious reasons. You want to know the source so you can decide which foods to take off the market or warn people not to eat. And you want to know the source so you can figure out how the contamination happened, and which procedures (and whose procedures) need to be reassessed and, if possible, improved.

But sometimes you never figure out the source. Some foodborne disease outbreaks remain mysteries forever. Other times you think you know what food made everybody sick, but you’re not absolutely sure; maybe you’re pretty sure, and maybe you just have grounds for suspicion. And of course you can know what food (probably) made people sick and still not know which batch and where it came from; and you can know all that and still not know where and how it got contaminated.

To find out as much as you can about the source of the contamination, two main evidence pathways are simultaneously pursued. The first is the search for actual food contaminated with the outbreak strain. That’s microbiological evidence. The second is the search for commonalities among victims: For instance, most of the people who got sick had eaten X, and very few people who hadn’t eaten X got sick. That’s epidemiological evidence. Microbiological evidence is forensic; epidemiological evidence is circumstantial.

Ideally you get both. You figure out what food the outbreak victims have in common, and you find the outbreak strain in that food in at least one victim’s kitchen. You also try to trace it back to a single source – batch, market, distribution center, processing center, farm, seed producer, etc. – to figure out where and how the food got contaminated and where else it might have got to.

Microbiological evidence on its own isn’t ironclad. Suppose thousands of people in Germany are afflicted with EHEC and you find a radish (for instance) somewhere in Germany – but not a victim’s kitchen – that’s contaminated with the same strain of EHEC. Unless you can find evidence that the victims all ate contaminated radishes, or that they all ate radishes from the same batch as your contaminated radish, or at least that they all ate radishes, there’s a good chance you may have stumbled on an unconnected instance of contamination, a coincidence. Of course the odds against a coincidence are high if you’re dealing with a rare strain like O104:H4, and they’re higher still if you dig even deeper and find identical genetic markers.

At the very least, microbiological evidence helps tell you where to look for circumstantial evidence: See if the victims ate radishes from the same source as your contaminated samples. Or see if the victims ate some other food product that was produced on the same farm or transported by the same wholesaler or sold in the same market – all places where cross-contamination might have occurred (maybe from radishes to some other food; maybe from some other food to radishes).

A suggestive but not definitive piece of microbiological evidence – not a radish, by the way – played a role in the German E. coli O104:H4 outbreak we’ll be dissecting soon.

Sometimes no microbiological evidence can be found. Maybe you’re too late. Produce that hasn’t been eaten gets thrown out, and can no longer be retrieved, cultured, and tested. (Frozen foods, canned foods, seeds, and the like last longer.) Maybe the food is still there to be tested if you can find it, but you can’t find it. You’re looking for a needle in a haystack. There are lots of radishes out there and most of them aren’t contaminated.

Epidemiological evidence isn’t a sure thing either. Identifying a food that 100% of the victims ate is too much to expect, especially since some may have contracted the disease from a secondary source, such as a different food contaminated by the original source or a family member who’s already sick. Identifying a single food that many of the victims ate from a single source (one batch, farm, wedding, restaurant, or market, for instance) is as good as it gets.

But there may be several foods that many of the victims ate, and you don’t know which one is responsible. This is especially likely if what the victims ate in common is a dish with many ingredients, like a salad. Or there may be no food that many of the victims ate, or at least that they know they ate (some ingredients aren’t necessarily obvious) and remember eating (you may be asking weeks later). Worse still, there may be a food they remember eating – and a different food they were unaware of that’s your actual culprit.

For sure, epidemiological evidence tells you where to look for microbiological evidence: Once you’ve identified a food that many victims have in common, and you’ve tracked the food to a common source, looking for contamination in that food from that source is obviously high-priority. But if you have a strong epidemiological case, you may be pretty sure (though not certain) that you know which food was contaminated, even if you can’t find any microbiological evidence. Of course you still need to try to discover where and how it got contaminated.

Epidemiologists trust strong epidemiological evidence. Even though it’s circumstantial, they know it’s often the best evidence available. They’re pretty comfortable warning the public not to eat radishes based on a solid epidemiological (circumstantial) case that eating radishes seems to be what many victims have in common.

But the public – including journalists and politicians – tends to favor microbiological evidence. We are like juries that hate to convict based on circumstantial evidence alone, preferring to rely on DNA or other forensic proof (or eyewitnesses – even though they’re notoriously unreliable). And of course an industry that’s the target of an epidemiology-based warning is bound to complain loudly and bitterly that its sales have been unfairly devastated when nobody has found a single contaminated radish!

Sometimes the epidemiological evidence is near-certain, even though it’s circumstantial. Sometimes it’s far-from-certain – not just circumstantial but really, really tentative, though still convincing enough to justify alerting the public so people can take action to protect themselves from the possible culprit. The same is true of microbiological evidence. You may be sure that you have found a contaminated radish without being at all sure that your radish is connected to the outbreak you’re investigating.

So part of proclaiming uncertainty is proclaiming how uncertain you are, and why. If you sounded excessively confident when you were seriously uncertain early in an outbreak, it’s going to be much harder to sound credibly near-certain later when the evidence has improved … especially if it’s better evidence incriminating a completely different culprit.

Our case study includes an important example from the German O104:H4 outbreak, in which unduly confident-sounding statements about a false lead early in the outbreak set the stage for a growing perception of general incompetence, even as the outbreak investigators gradually built a solid case against the real source of the contamination.

It is not rare for investigators to get the source of a foodborne outbreak wrong before they get it right. In 2008, for example, a rare (“Saintpaul”) serotype of Salmonella enterica caused at least 1,329 cases of food poisoning in the U.S., the worst U.S. salmonellosis outbreak since 1985. The CDC and the FDA initially attributed the outbreak to raw tomatoes imported from Mexico. Then they decided it might be “food items that are commonly consumed with tomatoes.” Among the possibilities they publicly mentioned were fresh Serrano peppers, fresh cilantro, fresh bulb onions, fresh scallions, and fresh jalapeño peppers.

They finally settled on jalapeño peppers as the principal culprit, though they continued to think that tomatoes and Serrano papers were probably responsible for some cases as well. The zigzagging investigation was grounded mostly in epidemiological tracebacks, though along the way a few peppers (mostly but not exclusively jalapeño peppers) were found to be contaminated with Salmonella Saintpaul.

Europe’s O104:H4 story is much the same. After a few false starts, investigators eventually determined with justified confidence but not-quite-certainty that Europe’s O104:H4 outbreak came from fenugreek seeds grown in Egypt. As we shall see, they developed a strong epidemiological case with some fairly minor supporting microbiological evidence.

That’s the end of the story. Here’s what came before, as seen through the lens of uncertainty risk communication.

Act One: Spanish Cucumbers

The outbreak apparently started in Northern Germany in early May 2011. The first death, an 83-year-old woman, happened on May 21 and was reported on May 24, by which time at least 140 people had fallen ill. Sample headline: “Deadly E. coli outbreak claims victims in Germany.” One unusual feature of the outbreak was already visible: Most of the victims were adults, not children or the elderly. Another unusual feature, a preponderance of women, would be reported a few days later. Up to this point no specific E. coli serotype had been mentioned.

On May 25, the next day, the German agriculture ministry announced to the media that initial epidemiological investigations by the prestigious Robert Koch Institute (RKI) had shown “a high probability” that the infections were “caused by consumption of raw tomatoes, cucumbers or green salads.” The common denominator, the food that most of the victims ate, was pretty obviously something commonly found in salad. In the end the contaminated salad ingredient turned out to be sprouts – but they didn’t know that on May 25.

Also on May 25, German officials first announced that a very rare E. coli strain, O104, was found in five of the sick patients. As Science reported:

“I hope that more patient samples will be identified today, but it certainly looks as if an E. coli O104 is the cause of these infections,” says Klaus Stark, head of the division of gastrointestinal infections, zoonoses, and tropical infections at Robert Koch Institute (RKI) here….
The source of infection in Germany is still unclear, and the rare serotype could make it harder for authorities to find it. “O104 is very hard to distinguish from normal, non-pathogenic E. coli,” says Lothar Beutin, head of the National Reference Laboratory for E. coli at the Federal Institute for Risk Assessment here. He and other researchers are now trying to develop a more rapid and specific test for the rare serotype.

The news focus changed radically one day later, when the Hamburg Institute for Hygiene and Environment – a German state-level agency, not a national agency – announced that it had zeroed in on Spanish cucumbers. Here’s a translation of the first two paragraphs of the May 26 Hamburg Institute news release:

The Hamburg Institute for Hygiene and Environment (HU) of the Board of Health and Consumer Protection (BGV) has clearly identified a cucumber from Spain as a carrier of Enterohemorrhagic Escherichia coli (EHEC). Samples from two other Spanish cucumbers and a cucumber of unknown origin also tested positive for EHEC.
“The suspicion regarding the trigger of the disease is now targeted towards cucumber,” said Health Inspector Senator Cornelia Prüfer-Storcks. “Information on the origin and further details are now being compiled. These results are a big success of our extensive search and further investigation will hopefully bring a big step forward.”

This May 26 release does not say whether the Hamburg Institute had identified the serotype of the EHEC strain on the Spanish cucumbers, nor does it say whether the Institute was in the process of identifying the serotype. It just says “positive for EHEC.”

Five crucial days later, on May 31, the Hamburg Institute itself would announce that the EHEC serotype in its Spanish cucumbers was not O104 – that is, not the serotype responsible for the outbreak.

But between May 26 and May 31, the “hooray for us” tone of the May 26 release left the firm impression that the Hamburg Institute had found Spanish cucumbers contaminated with the rare, deadly EHEC strain that the Robert Koch Institute had identified in sick humans on May 25. That impression was capably and dramatically passed along by the media.

Note that the Hamburg Institute announcement was based on very weak microbiological evidence. Preliminary epidemiology at the Robert Koch Institute and elsewhere had suggested that investigators should look closely at salad ingredients, so cucumbers were a reasonable hypothesis. They were already on the May 25 list of raw salad ingredients to avoid eating. But there was no epidemiological chain of exposure and no microbiological serotyping to connect Spanish cucumbers to German outbreak victims.

The Hamburg Institute had simply procured and tested lots of vegetables from the huge Hamburg central market. A few Spanish cucumbers tested positive for EHEC (the Institute said) – not O104:H4 or any other particular EHEC strain, just EHEC. (Because O104:H4 was a rare strain of E. coli, most labs weren’t equipped to test for it quickly yet. Even so, some experts have been critical of the delay before the Hamburg Institute cucumber samples were serotyped to see if the strain was O104:H4 or a different strain of EHEC.)

Though contaminated with a still-unnamed and (we later learned) still-unidentified type of EHEC at the time of the May 26 news release, the Spanish cucumbers were a potentially important lead to pursue. But they were an extremely tentative lead, much more tentative than the Hamburg Institute’s news release makes them sound.

If you read carefully enough, the news release does acknowledge uncertainty. The confident assertion in the first paragraph isn’t that Spanish cucumbers “clearly” caused the German O104:H4 outbreak, only that Spanish cucumbers had been found that were “clearly” contaminated with some EHEC strain, leading to “suspicion [not certainty] regarding the trigger of the disease.”

But it would be easy for a casual reader to miss the uncertainty, linking the Hamburg Institute’s confidence that it had found EHEC in Spanish cucumbers to its much less confident suspicion that it had identified the source of Germany’s O104:H4 outbreak. As we will document below, even some not-so-casual readers made the same mistake.

After addressing several other aspects of the outbreak, the news release does return to the chain-of-exposure issue, pointing out that food samples from victims’ homes were still being analyzed and other possible sources of the contamination could not be excluded. This somewhat buried acknowledgment of uncertainty didn’t overcome the confident tone of the news release lede.

If the Hamburg Institute had wanted to proclaim its uncertainty, not merely acknowledge it, there are several ways it could have done so:

  • It could have included in the news release (and the lede) the crucial fact that there hadn’t been time yet to compare the specific EHEC responsible for the German outbreak with the specific EHEC found in three Spanish cucumbers.
  • It could have said that “we think we’ll know the cucumber serotype in a couple of days” or, perhaps, that “we’re asking the National Reference Laboratory for E. coli for assistance in further identifying the serotype of the EHEC we found in the Spanish cucumbers.”
  • Most important, it could have stressed that “until we match the EHEC in our cucumbers with the EHEC in patients, we won’t know whether the contaminated cucumbers we found have anything at all to do with this disastrous outbreak.”
  • It could have added that it had absolutely no epidemiological evidence so far specifically indicating that sick people had eaten Spanish cucumbers.
  • As part of its case for announcing early despite the weakness of the evidence, it could have pointed out that “even if the EHEC strain on the Spanish cucumbers is different from the current outbreak strain, all EHEC strains can cause severe disease, which is why surveillance for EHEC is so important.”

Instead, the release has a distinctly triumphalist tone. It uses the word “Erfolg” – success – three times.

A technical background briefing that accompanied the news release carefully explains how important serotyping is and how labor-intensive and time-consuming it can be. It rightly notes that “only when food samples are identified with the same EHEC serotype as in stool samples of patients has the source of infection been found.” [“Erst wenn man in Lebensmittelproben den gleichen EHEC-Serotyp identifizieren kann wie in den Stuhlproben der Erkrankten, hat man die Infektionsquelle gefunden.”] But even the background briefing never explicitly makes the key point: that the Spanish cucumbers had not yet been serotyped as O104:H4 … or not O104:H4.

Most of the media missed the uncertainty buried in the Hamburg Institute’s announcement. The May 26 Deutsche Welle article, for example, carries this unequivocal lede: “Cucumbers from Spain are the source of the recent E. coli outbreak in northern Germany, the Hamburg Institute for Hygiene announced on Thursday.”

Even the bacteriology expert moderator from ProMED mail, SUNY-Downstate Medical School Professor Larry Lutwick, got the misimpression that the Hamburg Institute had shown that the Spanish cucumbers were contaminated with the outbreak strain. Citing the Deutsche Welle article plus another from a Swedish media source the same day, Lutwick concluded: “Additionally, it appears that German scientists have found the ‘smoking’ cucumber, that is, isolation of the epidemic strain from the presumed vehicle.”

A same-day Reuters story is a bit more tentative, though still not tentative enough. The story ran under the headline “Spanish cucumbers may be E. coli source, Germans say”:

HAMBURG/LONDON (Reuters) – Cucumbers imported from Spain may be the source of an E. coli outbreak that Germany says has killed four people and affected at least 200 more, European health officials said Thursday….
Britain’s Health Protection Agency (HPA) said German authorities believed the source of the infection was likely to have been food. “Early studies implicate raw tomatoes, cucumbers and lettuce,” it said.
Three of four contaminated cucumbers analyzed by the Hamburg Institute for Hygiene and the Environment (HU) came from Spain, according to state health authorities in Hamburg. The country of origin of the other cucumber was not yet known.
“The HU has clearly identified a cucumber from Spain as a carrier of E. coli,” the Health Ministry said in a statement.
“The suspicion concerning the cause of the illnesses now points to cucumbers,” said Hamburg Health Minister Cornelia Pruefer-Storcks.
The European Comission’s Rapid Alert System for Food and Feed said two companies from Andalucia may be implicated….

This story repeats the Hamburg Institute’s misleading conflation of the fact that “a cucumber from Spain” has been “clearly identified” as “a carrier of E. coli” with the implication that the German outbreak is just-as-clearly attributable to Spanish cucumbers. It leaves out the same key fact the Hamburg Institute announcement left out – that nobody had yet analyzed whether the two serotypes of E. coli were the same.

Still, the story creates a somewhat less overconfident impression than the Hamburg Institute announcement and a lot of the other coverage of that announcement, thanks to the tentativeness of its headline and its interjection of a second paragraph from a different source, attributing the outbreak more broadly to “raw tomatoes, cucumbers and lettuce.”

In our judgment, it’s not the media’s fault that they missed how extremely uncertain the Hamburg Institute’s evidence was linking Spanish cucumbers to the outbreak. It’s the Hamburg Institute’s fault for adopting a triumphalist tone and failing to proclaim its uncertainty.

And after the Hamburg Institute issued its May 26 news release, it became the fault of other health and food protection agencies in Germany and throughout Europe (particularly Germany’s National Reference Laboratory for E. Coli in Berlin) for failing to proclaim their uncertainty about the Hamburg Institute’s finding – not uncertainty that Hamburg had found some contaminated Spanish cucumbers, but uncertainty that those cucumbers would turn out to be related to the O104:H4 outbreak.

A variety of political factors may have contributed to these agencies’ reluctance to appear critical of the smaller, less renowned Hamburg Institute. Among other factors, officials at all levels had probably bought into the “speak with one voice” mantra, an extremely widespread crisis communication dogma from which we have long dissented.

[Added December 5, 2011: On November 24, 2011, Germany’s Federal Institute for Risk Assessment (BfR) emailed us a response to this column, pointing out that it did post an “opinion” on its website on May 26, the same day as the Hamburg Institute’s announcement, mentioning that there was not yet any evidence of a link between the Spanish cucumbers and the German outbreak.
We had scrutinized the BfR’s news releases when we were researching the column, but we missed the May 26 BfR opinion, which was entitled “EHEC: Consumers to continue to refrain from eating tomatoes, cucumbers and green salads raw.” The major media in Germany and elsewhere also missed it. Neither the BfR nor any other agency (except in Spain) issued a news release or media statement indicating any dissent from the Hamburg Institute’s claim, which the media were depicting as the universally shared view of the German authorities. See also our reply to the BfR’s response.]

Though they passed up the opportunity – and very likely the temptation – to disagree publicly with the Hamburg Institute, the government agencies involved in the O104:H4 investigation ultimately failed to present a united front anyway. At least that is how they are perceived. As the outbreak waned, they were widely criticized for their divergences.

As Shaun Kennedy, Director of the National Center for Food Protection and Defense at the University of Minnesota, commented in the Center’s July 2011 newsletter: “[D]uring the early stages of the outbreak, simultaneous press conferences across Germany from different agencies and institutes were announcing different recommendations. When public officials don’t agree, it leads to decreased consumer confidence that they are being protected.” And Marc Sprenger, the head of the European Centre for Disease Prevention and Control, urged participants at a July 5–6 “lessons learned” conference to “remember that ‘one voice’ information to policy makers and citizens is key.”

We have not content-analyzed the purported “different recommendations” of different agencies that provoked criticism. But we can document this unfortunate one-voice “success”: During the five-day gap from May 26 (when the Hamburg Institute announced it had found EHEC on Spanish cucumbers) to May 31 (when the Institute announced it was not the outbreak strain), German and European Union officials kept strangely quiet about the uncertainty of the Hamburg Institute’s Spanish cucumber finding.

The “one voice” that dominated during these five days was the Hamburg Institute’s voice. It was mostly a solo voice. Other German and European officials did not really sing along, but they certainly didn’t sing counterpoint. That is, they didn’t explicitly contradict the impression of certainty coming from Hamburg. Especially during the first few days of this five-day gap, EU countries (except for Spain!) spoke and acted as if “Germany” had announced that it had found the outbreak strain on Spanish cucumbers.

Ideally, of course, the Hamburg Institute would have prevented this misimpression by issuing a far more tentative initial announcement.

Failing that, Institute officials could have issued a second announcement on May 27, when they noticed the misimpression their May 26 news release had created. It wouldn’t have had to be a “correction”; it could simply have been a clarification – that the Institute had not identified the outbreak strain on the Spanish cucumbers, at least not yet; and that until the cucumber strain was matched to the outbreak strain no conclusions should be drawn about the role of the Spanish cucumbers in the German outbreak. For extra credit, the second statement could have apologized for creating a misimpression in the first statement.

The Hamburg Institute did eventually correct the misimpression on May 31, once it had firm evidence that the Spanish cucumbers were not connected to the outbreak after all. Even then it never acknowledged that it had created a misimpression, much less apologized for doing so. (Government agencies rarely take responsibility for creating misimpressions; if they say anything at all, they typically blame the media and the public for getting it wrong.) But even an exemplary statement on May 31 would have been far too late. The misimpression of confidence about what the final answer would be needed to be corrected before the final answer was in. And of course the Hamburg Institute’s failure to proclaim uncertainty before it was conclusively proved wrong unduly worsened the already unavoidable woes of the Spanish produce industry.

Given that the Hamburg Institute wasn’t correcting the misimpression of confidence, EU and German federal officials should have done so for them. They wouldn’t have had to blame or even diss the Hamburg Institute. They could simply have proclaimed the uncertainty that the Hamburg Institute had already acknowledged (but not loudly enough). Here’s what they could – and should – have said on May 27:

We would like to call everyone’s attention to the fact that the Hamburg Institute is not saying that it has found the outbreak strain on the Spanish cucumbers. This point has been missed in coverage of the Institute’s announcement yesterday. The Hamburg Institute and our national and EU agencies have not yet had a chance to serotype the Hamburg samples. We hope to have the results in a day or two.
Of course we are all hoping that the Hamburg Institute has found the source of the outbreak. But so far the evidence is extremely weak. While it is entirely appropriate to avoid Spanish cucumbers until we know the serotype of the contaminated cucumbers, it is a huge mistake to believe that Spanish cucumbers are definitely related to the outbreak. Since May 25, we have been recommending avoidance of all cucumbers, tomatoes, and lettuce, and that advice continues.

We can only speculate about why all the national and EU agencies involved in the outbreak left the Hamburg Institute’s misimpression of certainty uncorrected. One major reason, we believe, is probably a deeply felt but poorly reasoned belief in “speaking with one voice.”

Secondarily, we suspect that German federal officials may have avoided saying anything that might irritate the Hamburg Institute because Germany’s National Laboratory wanted to get its hands on the Institute’s cucumber samples as quickly as possible. It takes about three hours to drive from Hamburg to Berlin. But the Hamburg Institute did not send any samples to the national lab in Berlin until May 30 or 31, Hamburg Board of Health press spokesman Rico Schmidt told us in an email responding to our query about this. Despite numerous follow-up queries, we still have no explanation for why it took so long.

Should the Hamburg Institute have gone public about its Spanish cucumber discovery, prior to getting back analytic results showing whether the same serotype of EHEC was in the Spanish cucumbers and the German victims? We think the answer is yes. Despite the likely damage to the Spanish cucumber industry, an early warning about cucumbers from Spain made public health sense – if it was suitably tentative in two important ways:

  • The warning needed to stress that the Spanish cucumbers might still prove innocent – not innocent of containing some strain of EHEC, but innocent of responsibility for the O104:H4 outbreak.
  • The warning needed to stress that other raw vegetables (including cucumbers from other countries) might still prove guilty – that is, responsible for the outbreak.

Many government agencies in Germany and elsewhere were pretty clear about the possible guilt of other raw veggies, and the wisdom of avoiding all raw cucumbers, tomatoes and lettuce until the outbreak was over or its source was more firmly established. But the Hamburg Institute, the other agencies, and therefore the media weren’t nearly clear enough about the possible innocence of the Spanish cucumbers. The experts should have been proclaiming their uncertainty, not just (barely) acknowledging it. They should have been warning consumers about Spanish cucumbers and warning consumers that everything might flip-flop the instant they found evidence that the Spanish cucumbers weren’t responsible for the outbreak. A good uncertainty proclamation might have sounded something like this:

We have found a handful of cucumbers imported from Spain that were contaminated with an unidentified but potentially dangerous strain of E. coli. We don’t know yet where or how the cucumbers got contaminated – in Spain, in Hamburg, or somewhere in between. We don’t know yet whether other Spanish cucumbers are also contaminated, or whether other vegetables from other sources are contaminated as well. Most important, we don’t know yet whether the strain of E. coli we found in the cucumbers is the same strain that is responsible for the deadly ongoing outbreak in Germany and some other European countries (but as far as we know, not Spain). We don’t have a “smoking vegetable” yet – nothing like an uneaten Spanish cucumber that’s contaminated with O104:H4 found in a victim’s refrigerator, for instance.
Until we find out more, it is important for consumers to avoid cucumbers from Spain, and it is also important for consumers to avoid all raw cucumbers, tomatoes, and lettuce. When we learn more, we are likely to owe someone an apology. We just don’t know yet which warning will end up saving lives and which will end up unfairly damaging the sales of safe vegetables.

What did the Spanish government say about the Hamburg Institute’s announcement? Back to the May 26 Reuters story:

[A]ny product that may have been in contact with the infected batch had been removed, Spain’s Health Ministry said.
“While the cucumbers were produced in Spain, an investigation must find out at what stage of the production chain the contamination occurred,” the Ministry said.
Spanish consumers should not adjust eating habits as no case of the infection has been reported in Spain, it said.

To its credit, the Spanish government is not claiming here that Spanish cucumbers are innocent until proven guilty. In this particular story, in fact, the Spanish government doesn’t even mention that Spanish cucumbers haven’t been proven guilty yet. A May 26 Spanish Health Ministry news release does make the point that’s missing from the Hamburg Institute announcement and the Reuters story: “Dado que se trata de un serotipo poco frecuente de la bacteria, la analítica resulta compleja y por tanto quedan todavía extremos por confirmar.” [“Since this is a rare serotype of the bacteria, the analysis is complex and therefore final results are still to be confirmed.”]

But the Spanish government goes a lot further than we would have gone in implying that Spanish consumers are safe. We think the claim that “any product that may have been in contact with the infected batch had been removed” is grossly overconfident, especially for a Day One story. Even assuming that there’s only one infected batch and that it’s feasible to determine what other products might have been in contact with that batch, a 100% effective recall is vanishingly rare even for canned products with batch labels, and virtually impossible for raw vegetables. And we think the claim that Spanish consumers don’t need to worry because no cases have been reported in Spain (so far) is grossly over-reassuring. We have no quarrel with the appropriately uncertain claim in the middle, that no one knows yet whether the Spanish cucumbers were contaminated before or after they left Spain.

A May 27 World Health Organization “Global Alert and Response” was more circumspect about the cucumber connection – maybe too circumspect, in that it discreetly leaves out the cucumbers’ country of origin:

The epidemiological investigation into the source of the outbreak is under way. Although the source has not yet been determined, cucumbers are under suspicion, and the Robert Koch Institute in Germany is advising people, as a precautionary measure, to avoid eating tomatoes, cucumbers and lettuces, in addition to taking the usual hygiene measures in handling fruit and vegetables.

If you want to, you can read skepticism about the Hamburg Institute announcement into this statement – but we see no evidence that the media or the public picked up on such a subtle cue.

Some May 27 news stories were anything but circumspect, discreet, subtle … or uncertain. Here’s the lede of a May 27 AFP story out of Berlin:

More than 270 people in Germany have fallen seriously ill due to potentially deadly bacteria detected in imported Spanish cucumbers, but Madrid said Friday there was “no proof” it is to blame.

There’s no uncertainty whatever in this lede; though the evidence is still weak, AFP unambiguously attributes the German outbreak to Spanish cucumbers. The Spanish defense here is solid; there was, literally, still “no proof” linking the EHEC in those three Spanish cucumbers to the German O104:H4 outbreak. But AFP makes it sound like the Spanish government is defensively resisting an obvious conclusion.

Later in the AFP article, the Spanish government again grounds its rebuttal not in hypothesizing that the Spanish cucumbers might not be responsible for the outbreak, but rather in hypothesizing that the cucumbers might have been contaminated after leaving Spain:

Spain’s agriculture minister, Rosa Aguilar, said it was too early to blame her country and complained the accusations had caused “irreparable damage” to the sector.
“We do not know where the contamination may have taken place and the European Commission has made clear that it could have happened outside the country of origin,” the minister said.
“Until now nothing has been proven and it has not been demonstrated that it happened in the country of origin,” she said, adding: “Our level of safety and quality is extraordinarily high.”

As other stories (including other AFP stories) made clear, the Spanish government and Spanish cucumber growers were acting responsibly, recalling and discarding potentially contaminated cucumbers. They didn’t need evidence linking the cucumbers to the German outbreak to do that. It was enough that they had evidence showing that some Spanish cucumbers were contaminated with some EHEC strain, whether or not it turned out to be the same EHEC strain.

What’s missing from Spanish risk communications in this and other news stories? Something like this:

It may turn out to be us. We just don’t know yet. If it does – if Spanish produce turns out to be responsible for this horrific outbreak of foodborne illness – we will feel terrible, and we will need to reexamine our food safety programs to see what we can do to prevent a future tragedy like this one.

Still, Spanish authorities are less to blame for focusing on “it might not be our fault” and forgetting to say “it might be our fault” than the Hamburg Institute is to blame for sounding far too certain that “it’s their fault” … and the other agencies are to blame for leaving that impression unchallenged until May 31. Again, we acknowledge that the May 26 Hamburg Institute news release did acknowledge uncertainty. But it didn’t proclaim uncertainty – leaving most reporters, many officials, and some experts with a far more confident impression than the evidence justified.

The AFP story ends by reporting that the German state of Saarland had banned the sale of all cucumbers from Spain, and that supermarket chains had withdrawn all Spanish cucumbers from their shelves. But the broader warnings about “cucumbers, tomatoes, and lettuces” were also taking a toll:

Officials meanwhile defended themselves against charges, mainly from farmers in northern Germany, that they had acted rashly in their warnings to the public.
Initial warnings had spoken of possible contamination in tomatoes, lettuce, and cucumbers grown in northern Germany, where most cases of food-poisoning have been reported.
“The protection of the consumer must always take precedence over economic interests,” the consumer ministry spokesman said.
German vegetable growers have suffered losses of some two million euros ($2.8 million) per day since the middle of the week, a spokesman for the Farmers’ Association said Friday.

As non-experts closely monitoring official and expert statements about the outbreak, we think it obviously made sense to advise consumers to avoid all raw tomatoes, lettuce, and cucumbers until the source was more definitely identified or the outbreak was over. And we think it also made sense (for the five days from May 26 to May 31) to advise consumers that Spanish cucumbers were suspected as the possible source – while still urging everyone to continue avoiding all raw tomatoes, lettuce, and cucumbers. In other words, we have no quarrel with the advice that German authorities gave to consumers. Our only risk communication criticism is their failure to proclaim their uncertainty.

That’s the focus of this long column – the difference between merely acknowledging uncertainty and emphatically, dramatically, and repeatedly proclaiming uncertainty.

Would proclaiming uncertainty have helped reduce the cost to German and Spanish farmers and food distributors? Not in the short term. At least we hope not. The only way to protect short-term sales would have been to issue ineffective warnings or no warnings at all, thereby increasing the risk to consumers. The goal is to issue warnings that proclaim uncertainty but still function as warnings … that is, warnings that still keep people from eating raw tomatoes, lettuce, and cucumbers in general (until June 10, when those suspects were exonerated) and Spanish cucumbers in particular (until June 1, when they were exonerated).

We hope and believe that this can be done. People are almost as determined to avoid a cucumber they’ve been told might be deadly as a cucumber they’ve been told is sure to be deadly. Realistically, any food contamination warning is probabilistic. Even if it were certain that a batch of Spanish cucumbers was responsible for the outbreak, most Spanish cucumbers would still be uncontaminated and safe. But unable to tell which ones, sensible consumers would simply avoid Spanish cucumbers … or all cucumbers. For most consumers (and most restaurateurs and retailers), a warning that the outbreak has been tentatively linked to cucumbers from Spain but other cucumbers, tomatoes, and lettuces might also be implicated would be warning enough.

But if proclaiming uncertainty doesn’t protect sales by undermining the warning, what does it accomplish? Two things:

1. Preserving your credibility

First, proclaiming uncertainty protects the credibility of the authorities, and thus the effectiveness of future warnings. If you warn people that X is going to happen, then when X doesn’t happen you were wrong, so we have reason to think you might be wrong next time too. But if you warn people that X might happen, then when X doesn’t happen you weren’t wrong. It was true at the time that X might happen (even though it didn’t); we were lucky this time, but we have reason to worry that we might not be so lucky next time.

That’s especially so if you proclaim your uncertainty, not just acknowledge it; if you stress from the outset that if X doesn’t happen you’ll wish in hindsight that you hadn’t scared people unnecessarily and hurt an innocent industry – but you’ll take that risk rather than the unacceptable risk of leaving people unwarned, unprepared, and unprotected if X materializes.

Proclaiming uncertainty is at least a partial answer, we think, to the confusion over warning fatigue, the so-called “cry-wolf syndrome.” The literature on warning fatigue is all over the map. Some authorities claim warning fatigue is a myth; others say it’s a big problem. Both sides can point to anecdotal evidence (and a small amount of quantitative research) on their side. A false warning sometimes makes people cynical and dismissive; other times it leaves them just as responsive as ever to future warnings, or even more responsive, more alert to the dangerous possibilities. The disagreement over the impact of false warnings is at least partly the result of unclear definitions of what constitutes a false warning.

The distinction between warnings and predictions is crucial here. If you predict that X is going to happen or that X is going to be catastrophic, then when it doesn’t happen on schedule (e.g. a bird flu pandemic) or it isn’t catastrophic (e.g. the swine flu pandemic) you lose credibility, and so do your future predictions. But if you warn that X might happen or might be bad, the credibility of your future warnings should remain comparatively intact no matter how things turn out. People are wise to lose faith in false predictions. But properly constructed warnings aren’t falsified when X doesn’t materialize, any more than the failure of your house to catch fire proves that you were foolish to buy fire insurance.

The paradox is that officials are reluctant to sound uncertain largely because they fear that doing so will call their competence into question. They’re not entirely wrong. People really do prefer confident officials – but only if they turn out right. If you might be wrong, sounding uncertain protects your reputation for competence far better than sounding certain.

Some outside experts have criticized the German authorities for doing a slow, substandard job of tracing the O104:H4 outbreak to its source. We’re not qualified to judge whether they were actually incompetent or not. But they certainly fostered an appearance of incompetence with an overconfident-sounding early claim – it’s the Spanish cucumbers! – that turned out wrong.

Part of the problem is that even appropriately tentative warnings are likely to be perceived as confident predictions. The media play a big part in conveying this misperception. Reporters don’t like hedging; they like to tell their stories in black and white, not gray. So when you announce that X might happen and might be bad, some reporters are likely to report that you said we’re all gonna die. And then if X doesn’t happen (bird flu) or isn’t bad (swine flu), those same reporters look back and ask, “Remember when the experts said we’re all gonna die?”

The only solution we know – and it’s only a partial solution – is to insist aggressively that you’re not saying we’re all gonna die. That’s what we mean by proclaiming uncertainty.

  • It helps to say what you’re not saying: “We’re not saying the culprit is Spanish cucumbers. We’re saying it might be Spanish cucumbers. It’s too soon to be anywhere close to sure how this horrible outbreak began.”
  • It helps to explain why that’s not what you’re saying: “Tracing the source of a foodborne outbreak is like any other detective work. Sometimes the first suspect turns out to be the perp. But as any reader of detective stories knows, sometimes the first suspect is completely innocent. And sometimes, in the real world, you never find out whodunit. We will learn more in the coming weeks. We hope we’ll learn enough to be pretty confident that this outbreak is or isn’t attributable to Spanish cucumbers. We’re certainly not confident yet.”
  • It helps to correct the record: “Yesterday we said we suspected Spanish cucumbers might be responsible for the outbreak. But we apparently didn’t stress the ‘might’ enough, and some of the media ran stories that sounded more confident than we meant to sound. We haven’t even established yet that the strain of E. coli we found in the cucumbers is the same strain we found in the victims. We certainly haven’t tracked the bacteria from cucumbers to victims. Spanish cucumbers are our best guess so far, but we’re a long way from being sure. We’re sorry we gave that impression.”
  • It helps to ask everybody to imagine you turning out wrong. “If we learn that it isn’t the Spanish cucumbers after all, we will sound a little foolish in hindsight. And we’ll feel very apologetic, both to the farmers whose sales we will have damaged and to the consumers to whom we will have given the wrong advice. But in public health you can’t wait till you’re sure before you warn people. For now, don’t eat Spanish cucumbers or any cucumbers. But remember that tomorrow we may learn we were wrong and ask you to avoid eating something else instead.”
  • It helps to empathize with people’s wish that you were certain: “Even though uncertain information is better guidance than no information, everybody finds it painful to act on uncertain information. Everybody wishes we knew for sure, one way or the other. We wish that too. One of the hardest parts of our job is to avoid giving in to the temptation to sound more certain than we actually are.”
  • It helps to remind people that your recommendations, uncertain though they are, are nonetheless serious. “We’re dead serious about advising people not to eat Spanish cucumbers – because Spanish cucumbers are our top suspect so far. And we’re serious about advising people to avoid all raw cucumbers, tomatoes, and lettuce – because we’re casting a wide net and those are also suspects. And we’re serious about advising people to keep paying attention as this story evolves – because the actual source of the outbreak may turn out to be something else entirely.”

2. Speeding the path back from stigma

The second benefit of proclaiming uncertainty, not just acknowledging it, is that proclaiming uncertainty probably diminishes the long-term economic cost of warnings that turn out mistaken.

We say “probably” because we haven’t seen any data on this point, one way or the other. (We’re proclaiming our uncertainty about this effect of proclaiming uncertainty.) But there are lots of data on the generic phenomenon: People who know beforehand that a conclusion is tentative are better prepared to notice and reverse course later if that conclusion is disproved. People who think it’s firm, on the other hand, tend to shrug off the new evidence and stay the course too long. And when they finally absorb the new evidence, they may feel misled and betrayed.

We very much doubt that warnings are an exception to this rule. We would bet that sales of a stigmatized product will rebound faster from a warning that proclaimed its uncertainty than from a warning that sounded certain. But we haven’t seen proof that this is so.

Those are the two main reasons why we think uncertain warnings should proclaim their uncertainty. What about uncertain reassurances?

There, we think, the benefits of proclaiming uncertainty are obvious. It is a truism of risk communication that the alarming side in a risk controversy can get away with more extreme, dramatic, and even exaggerated claims than the reassuring side. (See Peter’s video entitled “First Outrage Management Strategy: Stake out the Middle.”) Overconfident warnings have higher collateral costs than warnings that proclaim their uncertainty, but even overconfident warnings are a public service, because they alert people to possible dangers. Overconfident reassurances are a disservice, because they encourage people to ignore those dangers. This asymmetry in risk communication – the alarming side gets more leeway to dramatize and exaggerate than the reassuring side – is the exact equivalent of the risk assessment principle of conservativeness.

If you’re the Spanish cucumber industry, or even the Spanish government, your ability to survive a foodborne outbreak that is your fault depends largely on your having acknowledged from the outset that it might be. So if you want to tell people it might not be your fault, you damn well better proclaim your uncertainty.

Spanish government officials did not follow this advice. In fact, their denials got more and more confident; far from proclaiming uncertainty as they should have done, they proclaimed certainty that the outbreak could not possibly be the fault of cucumbers contaminated in Spain. It turns out they were right, but there was no way they could have known they were right. They were taking a huge unnecessary risk. Consider these quotations from coverage of a June 1 news conference given by Spanish Deputy Prime Minister Alfredo Perez Rubalcaba:

Spanish Deputy Prime Minister Alfredo Perez Rubalcaba said Spain will not rule out “taking action against the authorities (in Germany) who questioned the quality of our products.”
“The proof is that there are no cases in Spain. If the bacteria came from Spain, we would see infections. Our products are absolutely clean, our cucumbers in this case. It’s not a problem of the origin of the products,” Rubalcaba said.
“If it was from the cucumbers, there would be cases (of illness) in Spain,” Rubalcaba said, adding that there haven’t been. “The Hamburg (Germany) authorities don’t know where it comes from. I understand they have a problem. We have said, ‘You need to say it wasn’t us.’”
Rubalcaba told the news conference there’ll be time to talk about demands for proper compensation but added, “It’s evident they have endangered the jobs of many people without proof. We know (the infection) is a serious problem, but the Spanish products have nothing to do with that crisis.”

At almost the same moment Rubalcaba was insisting without evidence that Spanish cucumbers couldn’t be responsible for the outbreak because nobody in Spain was sick – a non-sequitur – German scientists were proving that Spanish cucumbers were, in fact, not responsible for the outbreak.

Of course they weren’t “absolutely clean” as Rubalcaba claimed; the cucumbers the Hamburg Institute had discovered were genuinely contaminated with EHEC, just a different EHEC serotype than the one causing all the devastation. Neither Spain’s outrage about being “blamed” for the outbreak nor our criticism of the Hamburg Institute’s failure to proclaim uncertainty should be allowed to obscure an important fact: EHEC-contaminated Spanish cucumbers, regardless of the EHEC strain, could have caused severe disease. The coverage between May 26 and May 31 assumed Spanish cucumbers were responsible for Germany’s devastating outbreak; the coverage after May 31 assumed German officials had mistakenly blamed “innocent” Spanish cucumbers. During both periods there was only scant mention of the fact that finding EHEC on Spanish cucumbers was potentially important whether or not it was related to the O104:H4 outbreak.

How and where did those cucumbers become contaminated with EHEC? What happened to other cucumbers from the same batch? Did they make anyone sick? What is the official protocol when EHEC contamination is found during routine tests of produce? Was this protocol followed in the case of the Spanish cucumbers? It’s unclear from the coverage whether anyone (agencies or reporters) tried very hard to answer these questions, once Germany’s top lab had said a different EHEC strain was involved.

The Spanish cucumbers were found “not guilty” of causing the O104:H4 outbreak. They were not found “innocent”; somewhere along the way, they became contaminated with EHEC.

The news that the cucumbers weren’t responsible for the outbreak started to emerge on May 31, when the Hamburg Institute itself announced that the bacteria found on two of the four cucumber samples did not match up with the type of EHEC in the stool samples of patients. The announcement begins this way:

The first EHEC pathogens that have been discovered by the Institute for Hygiene and Environment (HU) of the Board of Health and Consumer Protection (BGV) in the investigation of lettuce and cucumber have now been analyzed. It was found that two of the four samples do not match with the agents of type O104 that were isolated from the stool samples of patients.
“We had hoped to discover the source of these serious cases with HUS syndrome, in these initial results. Unfortunately, this hope was not fulfilled,” said state Health Inspector-Senator Cornelia Prüfer-Storcks.
“Regardless of the outcome of the two pending samples, it was right to publish our findings, because the contaminants may very well trigger EHEC. It would be irresponsible to hold back such a reasonable suspicion. The protection of human life must be more important than economic interests.”
The further search for the source of infection continues. In parallel, the cause of the contamination of the produce will be investigated, since EHEC bacteria are – regardless of serotype – always a looming health problem.

Note the passive voice in the first paragraph “It was found….” We can’t tell if the Hamburg Institute finally analyzed its own samples for O104:H4 or if – as we suspect – it belatedly passed the samples on to a federal agency for analysis.

Note also the unapologetic, defensive tone of the announcement. We fervently agree with the Hamburg Institute that pursuing – and announcing – the cucumber finding was the right thing to do – not just because the cucumbers might have turned out to be the source of the outbreak, but also (as the announcement says) because any EHEC contamination is important. Still, it would have been empathic, courteous, and good risk communication for the May 31 announcement to acknowledge that the Institute had sounded pretty sure, had turned out wrong, and had done real damage to the Spanish cucumber industry along the way.

Finally, note that the words “Spain” and “Spanish” do not appear in the Hamburg Institute’s May 31 announcement. The cucumbers were actively linked to Spain while they were being targeted; now that they had been exonerated, they were referred to simply as “cucumbers” or “samples” (in German, of course).

The Institute’s May 31 exoneration of the Spanish cucumbers took some reporters by surprise. SpiegelOnline reported:

On Tuesday of this week, the plot thickened when Hamburg’s health minister, Cornelia Prüfer-Storcks, made a surprise announcement that the Spanish cucumbers were probably not the source of the wave of EHEC infections. [Note the “probably.” Some media had a hard time letting go of the Spanish cucumbers.]

We can speculate with some confidence about why the May 31 announcement was perceived by some as a “surprise.” Between May 26 and May 31, the media and the public had thoroughly absorbed the previous overconfident-sounding statements from Minister Prüfer-Storcks and the Hamburg Institute – statements to which EU and German federal officials silently assented during those five days.

On May 31, the same day the Hamburg Institute announced that some of the Spanish cucumbers did not test positive for the outbreak serotype, and the other samples’ serotypes were pending, Germany’s Federal Institute for Risk Assessment (BfR) released a statement headlined: “EHEC pathogen not yet typed: tomatoes, cucumbers and salads should nonetheless continue not to be consumed raw.” The thrust of this statement is that until all the Spanish cucumbers have been serotyped, consumers shouldn’t jump the gun and assume that other salad ingredients are safe.

We don’t know if the statement was prepared before the BfR saw the Hamburg Institute’s May 31 announcement, or if the BfR preferred to wait for federal serotype results for all the samples rather than citing Hamburg’s partial results. It’s the sort of uncertainty proclamation that would have been a lot more useful on May 26 or 27.

Also on May 31, the Sydney Morning Herald reported that “Andreas Hensel, president of Germany’s Federal Institute for Risk Assessment, told ZDF television that ‘we have found the so-called EHEC pathogens on cucumbers, but that does not mean that they are responsible for the whole outbreak.’” Given the time difference between Australia and Germany, and the time gap between the ZDF interview and the Morning Herald story, we don’t want to make too much of the fact that Hensel apparently didn’t know when he spoke that at least some of Hamburg’s Spanish cucumbers were completely unconnected to the outbreak. But assuming the newspaper is reporting the interview accurately, it is stunning that four to five days after the initial Hamburg Institute news release a top BfR official was giving interviews conflating “EHEC” with the specific O104:H4 EHEC serotype, and implying that Spanish cucumbers were responsible for at least part of the outbreak.

A day later the serotyping test results were back from all four samples. The BfR made it official in a June 1 news release with a headline as clear as clear can be: “EHEC germs on Spanish cucumbers do not correspond to the pathogen type of the patients concerned.” The first paragraph is equally clear:

A virulent EHEC strain O104:H4 is responsible for the current outbreak of EHEC infections. At the Hamburg-based Institute for Hygiene and Environment cucumbers from Spain were identified as EHEC carriers. These samples have been more thoroughly examined by the National Reference Laboratory for E. coli of the Federal Institute for Risk Assessment (BfR). None of the four samples showed the serotype O104:H4 of the pathogens which had been isolated from the feces samples of the patients. “The source of the on-going infections has not yet been determined,” says BfR President Professor Dr. Andreas Hensel. “Furthermore, it has to be clarified at what point in the food chain the contamination with germs occurred.”

We would have liked this announcement to be a bit more emphatic about what was now certain and what remained uncertain: “We now know for sure that the Spanish cucumbers had the wrong strain of EHEC and could not have caused the human EHEC outbreak of the past several weeks. As for figuring out the real cause of the outbreak, our evidence continues to point to raw salad ingredients. Unfortunately, we still don’t know which ingredient or ingredients.”

We’d also have liked this announcement to specify what EHEC serotype had turned up in the Hamburg Institute’s Spanish cucumbers – in two of them, anyway; the other two apparently had no EHEC at all. To the best of our knowledge, this information was not publicly released until August 2, at a breakout session during the International Association for Food Protection annual meeting. Lothar Beutin from Germany’s National Reference Laboratory for Escherichia coli reported that the EHEC serotype on two of the Spanish cucumbers was E. coli O8:H19.

And of course we wish the June 1 BfR announcement had been much more empathic and apologetic. This would have been a good time and place to express regret – regret that Spain had paid such a high price for EHEC-contaminated cucumbers that turned out unrelated to the outbreak, and regret that consumers were once again left with only the vague warning that some salad ingredient was the probable culprit.

The discovery that Spanish cucumbers might be responsible for the outbreak had been headline news, but the discovery that they weren’t responsible was a significantly smaller story. AP reported it matter-of-factly down deep in its May 31 update, which focuses instead on the increasing scope and severity of the outbreak, with 16 dead and more than a thousand sick so far.

European Union officials said Germany identified cucumbers from the Spanish regions of Almeria and Malaga as possible sources of contamination, and that a third suspect batch, originating either in the Netherlands or in Denmark and sold in Germany, was also under investigation.
On Tuesday [May 31], however, officials said they had found a slightly different type of EHEC on the cucumbers than the strain detected in the feces of sick people in Germany. That means those cucumbers did not cause the outbreak, but posed a health risk nonetheless, the German officials said.
Spain’s agriculture minister, Rosa Aguilar, seized on the find as evidence that “our cucumbers are not responsible for the situation.”

Which, of course, is exactly what it was.

But not all the media thought so, at least not right away. The June 1 Sky News story, for example, is headlined “Killer Cucumbers: Germany Backtracks.” The story never mentions Germany’s announcement the previous day that scientists now knew the two strains of E. coli – the outbreak strain and the Spanish cucumber strain – didn’t match. It begins this way:

German authorities say they do not know whether Spanish cucumbers have caused the outbreak of E. coli which has killed at least 16 people and left hundreds ill.
Health officials say the number of deaths is expected to increase in the coming days across Europe.
The apparent backtrack by Germany has come after Spain and the Netherlands demanded compensation for vegetable producers who have lost money as a result of concerns over their produce.

If the Hamburg Institute and federal German authorities had done a better job of proclaiming their uncertainty, perhaps the Sky News reporter might have realized that the new story was about evolving science, not backtracking in response to compensation demands.

Nor did the June 1 coverage routinely point out that the German authorities had never expressed anything like certainty that the Spanish cucumbers were responsible. A Xinhua story, for example, puts it clearly but wrongly:

German health officials claimed May 26 that cucumbers imported from Spain was one source of the recent deadly E. coli outbreak in its northern states.
But they admitted Tuesday that the latest laboratory tests have shown that the Spanish exports were not behind the outbreak.

An accurate story would have read instead:

German health officials warned [not “claimed”] May 26 that cucumbers imported from Spain might be [not “was”] one source of the recent deadly E. coli outbreak in its northern states.
But they announced [not “admitted”] Tuesday that the latest laboratory tests have shown that the Spanish exports were not behind the outbreak.

The difference between these two is one reason why proclaiming uncertainty is so important.

While officials were having a tough time proclaiming uncertainty, many industry sources found it shockingly easy to proclaim certainty – that their produce, at least, was safe. A June 3 story in Food Quality News, for example, was headlined “Stop dithering and identify E. coli source, say fresh produce players.” An executive at U.K. Salads proclaimed: “We are sure that our product is 100% safe, but something needs to be done as the problems in Germany are affecting our business.”

The impact on cucumber sales was real, but short-lived – even in Germany, even in the face of the Hamburg Institute’s overconfidence. In mid-June, Germany’s Bild am Sonntag newspaper reported a survey showing that 85% of Germans were eating cucumbers, tomatoes, and lettuce again.

The focus of suspicion had turned to sprouts.

Act Two: German Sprouts

The first news stories focusing on sprouts as the possible culprit appeared on June 5. Here’s how the Sprouts Day One AP story starts:

Initial tests have confirmed that bean sprouts grown in northern Germany are the likely cause of an E. coli outbreak that has killed at least 22 people and sickened over 2,200, an agriculture official said Sunday.
Different kinds of sprouts from one organic farm in the greater Uelzen area, between the northern cities of Hamburg and Hannover, could be traced to infected persons in five different German states, Lower Saxony Agriculture Minister Gert Lindemann told reporters.
“There were more and more indications in the last few hours that put the focus on this farm,” Lindemann said at a press conference in Hannover.

The reference to “tests” in this lede (“Initial tests have confirmed….”) is typical of a lot of media coverage of foodborne outbreaks. Epidemiology is more about “tracing” than “testing” – and frequent allusions to testing can only reinforce the misperception that only microbiological evidence from laboratory tests is truly conclusive.

But in fact, so far the evidence was epidemiological, not microbiological. The German authorities hadn’t found any contaminated sprouts on the farm they fingered; in fact, they hadn’t found any contaminated sprouts anywhere, with any EHEC strain. But epidemiological detective work had revealed that many victims in different cities had eaten sprouts traceable to the same farm. It took a while to zero in on sprouts, the authorities later explained, because people didn’t always remember the sprouts that had garnished a salad or sandwich they had eaten weeks earlier.

Perhaps having learned from their cucumber experience, German officials did a somewhat better job of emphasizing their uncertainty about the sprouts. From the AP story again:

Lindemann said 18 different sprout mixtures were under suspicion – including sprouts of beans, broccoli, peas, chickpeas, garlic lentils, mungo beans and radish. The sprouts are often used in mixed salads.
Lindemann urged Germans to not eat sprouts until further notice and said definitive test results would be available Monday. He said authorities could not yet rule out other possible sources for the outbreak and urged Germans to continue avoiding tomatoes, cucumbers and lettuce until further notice.

There is nothing from German officials in this story – or in any story we have found – that acknowledges and shares the public’s frustration about the lack of certainty. How might an official who wanted to come across as human have framed the story:

Until a few days ago we were focusing mostly on cucumbers grown in Spain. Now we’re focusing mostly on sprouts grown on an organic farm in northern Germany. We don’t know what kind of sprouts, and we’re not even sure yet that it’s sprouts – so we’re still saying don’t eat raw cucumbers, tomatoes, and lettuce. Of course people are frustrated! When are we going to know something for sure? Why can’t epidemiology be more like those CSI television shows? It’s driving us a little crazy too.

Can officials actually say this kind of thing? Yes, they can, and sometimes they do. On a U.S. CDC E. coli FAQ page we use, one answer ends with this great line: “It does get a bit confusing – even to microbiologists.”

To our surprise, a much more human-sounding statement came from the company whose sprouts were implicated, Gärtnerhof Bienenbüttel. The statement the company posted on its website homepage on June 8 (Day Four of “it might be sprouts”) begins this way:

We, the Gärtnerhof Bienenbüttel, are shocked and concerned by the June 5 news that some of our goods may be contaminated with EHEC pathogens. Following the announcement by the authorities, we immediately informed all customers and recalled the product. [Wir, der Gärtnerhof Bienenbüttel, sind erschüttert und besorgt über die Nachricht vom 05.06., dass ein Teil unserer Ware durch EHEC-Erreger verunreinigt sein soll. Nach Bekanntgabe durch die Behörden haben wir sofort alle Kunden informiert und die Ware zurückgerufen.]

We can’t give you the original link to that text, because it’s no longer on the Gärtnerhof Bienenbüttel homepage. But you can find it here.

Here’s a translation of the much less empathic, much more typically-outraged statement that has been on the farm’s homepage since July 19:

The Gärtnerhof reopened 14 July 2011 because … all samples taken in connection with the EHEC events showed negative results.
The operation had been closed since 6 June 2011 because of suspected EHEC, at the instigation of the Lower Saxony Ministry of Agriculture. For us, these six weeks were in every respect a state of emergency. It remains incomprehensible to us why Lower Saxony’s Minister of Agriculture Lindemann immediately made the name of our company public on 05/06/2011 because of mere suspicion. The effects were devastating: Continuing to the present day German and international media describe our Gärtnerhof as a source of the EHEC epidemic, although at no point has there been any evidence that this is so….
The operational closure has brought us to the brink of economic existence. Add to this the high psychological stress: We have been identified as the perpetrator responsible for the death and illness of many people, based on nothing but conjecture….
To restore our reputation as a trouble-free organic farm and ensure our economic survival, we will seek legal and expert help to review the conduct of the authorities. This will certainly take some time to complete. In the meantime we are in an open dialogue with the authorities concerned, aiming for an amicable solution.

As we shall see in Act Three, the company is almost certainly right in insisting that the contamination didn’t originate on company premises, though it is almost certainly wrong in implying that the contamination didn’t come from the company’s sprouts. Gärtnerhof Bienenbüttel passed the contamination along, is all – a fact the media got right as soon as the epidemiologists figured it out. Has the company been unfairly blamed? Not that we can see. Has it been unfairly damaged? Perhaps so. More about that in Act Three.

Should it have been publicly named? In our judgment, absolutely – though perhaps officials could have done more in the early days to stress that while they were pretty sure the contaminated sprouts had been sold by Gärtnerhof Bienenbüttel, they didn’t know yet whether it was Gärtnerhof Bienenbüttel that contaminated them. As soon as sprouts were fingered, there were lots of experts accurately telling lots of reporters how easily sprouts can be contaminated during germination, and perhaps particularly so under organic growing conditions. There was virtually no mention of the possibility that dried sprout seeds could be contaminated long before arrival at even the most pristine organic farm – an omission officials could have done more to correct.

Back on June 5 and in the days that followed, officials had to decide how much uncertainty to express about the bare fact that the outbreak had been traced, but still tentatively, to sprouts from Gärtnerhof Bienenbüttel.

On June 5 officials said (as the AP story put it) that they “could not yet rule out other possible sources for the outbreak.” That’s an acknowledgment of uncertainty, but it’s also a signal of confidence: “We can’t rule out Y yet” suggests that you’re pretty sure but not absolutely sure it’s X – X in this case being sprouts from the Gärtnerhof Bienenbüttel organic farm. A day later, testing of sprout samples from the farm had failed to turn up any samples positive for O104:H4. With no microbiological evidence to back up the epidemiological evidence, the June 6 Der Spiegel story reported what felt like a significant decline in official confidence:

Over the weekend, fresh sprouts became the latest suspected culprit. On Sunday, Gert Lindemann, agriculture minister for the western state of Lower Saxony, said officials were conducting lab tests on samples from a farm in the Uelzen district near Hamburg. Indications were apparently so strong that sprouts were the cause of the outbreak that the state’s Consumer Protection and Food Safety Ministry recommended consumers avoid eating them. By Monday afternoon, however, the ministry revealed that 23 of 40 samples taken from various locations on the sprout farm had tested negative for the bacteria. The remaining 17 samples must still be examined, the ministry said. On late Monday afternoon, ministry officials said suspicions remained that the sprouts may have caused the outbreak.
German Health Minister Ilse Aigner also recommended Monday that consumers not eat raw sprouts, “as long as the suspicion hasn’t been eliminated entirely.”
[Added December 5, 2011: On November 24, 2011, Germany’s Federal Institute for Risk Assessment (BfR) emailed us a response to this column, pointing out that Ilse Aigner is not Germany’s Health Minister, but rather its Minister of Food, Agriculture, and Consumer Protection. We failed to fact-check the title in the Der Spiegel article, and we mistakenly referred to Minister Aigner as Health Minister twice later in the column.]

“The suspicion [that it’s sprouts] hasn’t been eliminated entirely” is a lot more tentative than “could not yet rule out other possible sources [than sprouts].”

As far as we can tell, Health Minister Aigner’s June 6 statement that “the suspicion hasn’t been eliminated entirely” is more tentative than it should have been. Finding O104:H4 in a sprouts sample from the Gärtnerhof Bienenbüttel farm would have been pretty convincing evidence that the farm was almost certainly part of the chain of contamination (though not necessarily the source). But failing to find a smoking sprout at the farm did not eliminate sprouts as a highly likely cause of the outbreak, or the farm as a highly likely source or way-station for the contaminated sprouts. A June 6 MSNBC story gets it right:

“All the culture [looking for contamination in a food sample] tells you is that in the particular batch that you happen to have in your hand, it’s not there,” said Dr. Timothy Jones, the Tennessee state epidemiologist who serves on a federal food safety committee. “It doesn’t mean the batch last week was clean or that the box next door was clean.”

The story also paraphrases epidemiologist Michael Osterholm, director of the University of Minnesota Center for Infectious Disease Research & Policy:

European food safety officials appear to rely far more on bacterial cultures than on tracing back what people involved in the outbreak actually ate – and where it came from. But a microbiological approach has repeatedly been shown to fall short of a detailed study of the epidemiology, or health patterns, that characterize foodborne illness outbreaks, Osterholm said.

We can’t assess whether or not German scientists were leaning too heavily on microbiological evidence at the expense of the epidemiological investigation. But it’s very clear that some German and EU politicians preferred a smoking vegetable to the combination of shoe-leather and statistics that is today’s epidemiology of trace-backs and trace-forwards. So did the media. German scientists did far too little to explain why epidemiology’s circumstantial evidence is often the best evidence.

About this time a new leitmotif was emerging in official statements, expert commentary, and media coverage: We may never have microbiological evidence proving where the outbreak came from. Emphasizing that microbiological evidence might never emerge may have been partly a response to the experts’ fear of being wrong again. But it was also a response to the passage of time, to the experts’ belief that the offending produce had probably been discarded by now and thus could never be found, cultured, tested, and identified.

The June 6 Der Spiegel article included this observation:

“As the first people took ill, the contaminated batch had probably long since been sold,” said Lothar Wieler, director of the Institute of Microbiology and Epizootics at Berlin’s Free University. As the first HUS cases emerged, the vegetables would have already been eaten or thrown out, ultimately leaving the source a mystery.

The question is whether the absence of microbiological evidence would necessarily leave the source “a mystery.” And the answer is no. A persuasive epidemiological case without microbiological confirmation does fall short of certainty. But it’s not a mystery either. The experts who monitor infectious disease outbreaks at ProMED commented on June 6:

Negative tests at this point certainly do not exclude previous shipments of sprouts as the vehicle for this large outbreak. It is possible that most if not all of contaminated vehicle may have been eaten or passed the time when it would be eaten and newer lots might not have been exposed to the contamination source. It may be the case that science may not produce a “smoking” sprout here (where the outbreak strain is isolated from the suspected vehicle) and we would be left with epidemiological data alone.

Like many others, Der Spiegel attributed the failure to find microbiological evidence to bureaucratic delay in Germany’s official response. “Eighteen valuable days were lost,” the June 6 article points out, “between when the first patient came down with diarrhea around May 1, and when the RKI was finally alerted.” (The Robert Koch Institute, RKI, is Germany’s official disease control agency, its CDC.)

But for our dissection of uncertainty communication, we are interested in the abrupt switch from overconfidence about the source of the outbreak (it’s probably Spanish cucumbers; no, it’s probably German sprouts) to overconfidence that the source will probably never be established: too much certainty about uncertainty.

Thus a June 7 Associated Press story quotes World Health Organization communicable disease director Guénaël Rodier that “If we don’t know the likely culprit in a week’s time, we may never know the cause.”

As of June 7, then, sprouts from an organic farm in Germany were in the crosshairs, thanks to good epidemiological evidence but no microbiological evidence. And the authorities were telling the public:

  • It’s probably sprouts from that farm.
  • But just in case we’re wrong again, don’t eat raw cucumbers, tomatoes, or lettuce either.
  • We may never know the source (or at least we may never have microbiological evidence of the source) because all the contaminated food has probably been thrown out by now.

Epidemiologists and other experts all understood that the absence of microbiological evidence – a “smoking sprout” – made the epidemiological case pointing to sprouts less-than-certain, but still quite strong – much stronger than the initial wider net of suspicion about raw tomatoes, cucumbers, and lettuce. This kind of nuanced calibration of uncertainty (as opposed to a dichotomy: either you’re sure or it’s a mystery) is difficult to communicate to journalists and the public, even if sources explicitly try to do so. We see little evidence that German authorities tried. We also see little evidence – but we are less certain about this – that they tried very hard to explain to reporters the difference between circumstantial epidemiological evidence and confirmatory microbiological evidence.

For a few days, therefore, many reporters saw the negative microbiological tests of Gärtnerhof Bienenbüttel sprouts as reason to think the experts must have got it wrong again. They were primed to see it that way – primed by the serotyping tests that had exonerated Hamburg’s Spanish cucumbers; primed by most journalists’ automatic preference for microbiological evidence over epidemiological evidence; and primed by all those official statements (from Lower Saxony Agriculture Minister Lindemann and German Health Minister Aigner, among others) that seemed to put a lot of stock in “tests.”

Thus the June 7 New York Times reported:

Suspicion shifted Sunday to bean sprouts from a north German farm, only for scientists to report on Monday that tests on 23 of 40 samples from the facility tested so far had proved negative. A package of bean sprouts found past its sell-by date in the back of an E. coli patient’s refrigerator also showed no traces of contamination, the Hamburg health authorities said Tuesday.

And on the same day CBS posted a story online headlined “If bean sprouts didn’t cause deadly E. coli outbreak, what did?” The lede:

If bean sprouts aren’t the source of the E. coli outbreak that has killed 22 people and sickened thousands in Europe, what is? Officials in Germany, the hardest hit nation, still don’t know – and a World Health Organization expert says time is running out.

A sentence several paragraphs later in this story reads: “In the last week, German officials wrongly accused Spanish cucumbers and then German sprouts of sparking the crisis, which began on May 2.” This sentence explicitly (and incorrectly) claims that the sprouts hypothesis was as mistaken as the cucumbers hypothesis, and implies (with the word “accused”) that both were made confidently. Neither the high degree of uncertainty of the original cucumber hypothesis nor the continuing high likelihood of the new sprouts hypothesis registered with the reporter.

A day later, on June 8, the epidemiological case got stronger connecting the outbreak to sprouts from the Gärtnerhof Bienenbüttel farm, though it still wasn’t ironclad. Der Spiegel summarized the evidence this way:

Earlier on Wednesday, German authorities had said there were growing indications that bean sprouts delivered by an organic farm in Bienenbüttel first identified as a possible source on Sunday may indeed be a cause of the outbreak.
Health officials found two fresh clues that point to the farm near the town of Uelzen in northern Germany, the regional Consumer Protection Ministry of Lower Saxony said.
Ministry spokesman Gert Hahne said a total of 18 people infected with EHEC, the deadly strain of the bacteria, around the northern port city of Cuxhaven ate sprouts from the farm in a company cafeteria.
In addition, three female workers at the farm suffered from diarrhea in the first half of May, and one of them is known to have been contaminated with EHEC. Their work included packaging the sprouts.
The ministry spokesman said it was possible that one of the women accidentally “fed the pathogen into the operations of the company.” Or they might have been infected with EHEC at the farm.…
The Lower Saxony ministry spokesman said that despite the additional clues pointing to the sprout farm, it was still possible that the nationwide epidemic stemmed from several sources.
Four company canteens and three restaurants where people caught EHEC are now known to have been supplied by the Bienenbüttel farm – infecting about 100 of the more than 2,600 EHEC patients in Germany. So far, no EHEC bacteria have been found at the farm.

Expert reliance on epidemiological (circumstantial, less-than-certain) evidence is understandably frustrating to the public. It is even more frustrating to targeted companies and industries, who sometimes feel that they should be considered innocent until proven guilty beyond a reasonable doubt – a standard that would elevate agricultural economics above public health. It’s disappointing but not surprising when a government agriculture official argues the case for waiting for microbiological proof. But when a public health official does it, that’s a bit shocking.

A June 8 BBC story quotes and paraphrases European Union Health and Consumer Affairs Commissioner John Dalli as follows:

Mr Dalli earlier warned Germany against issuing any more premature – and inaccurate – conclusions about the source of contaminated food. Information had to be scientifically sound and foolproof before it was made public, he said.
“It is crucial that national authorities do not rush to give information on the source of infection that is not proven by bacteriological analysis,” he told the European Parliament.
“This spreads unjustified fears [among] the population all over Europe and creates problems for our food producers.”

By contrast, a June 9 BfR update matter-of-factly explains why uncertain evidence justifies a public warning about the suspected food, though not an outright ban:

Consumption recommendations are state measures. They have to weigh the proportionality principle. The health protection of consumers has, as a matter of principle, priority over economic interests…. In the current outbreak event a ban on goods would not have been proportional since the pathogen had not yet been detected. If the pathogen is, however, detected in or on a vegetable, this food is no longer safe and hence no longer marketable. This case has, however, not yet occurred in respect of the current event.

(Flash-forward to July 5, after epidemiological evidence, not microbiological proof, identified dried fenugreek seeds from Egypt as the most likely culprit – seeds that were sold to the Gärtnerhof Bienenbüttel farm among many other customers. Despite the lack of “proven bacteriologic analysis,” Commissioner Dalli changed his tune, announcing in a European Commission news release that the EU would temporarily ban the import of “some Egyptian seeds.” Egyptian authorities, of course, complained at the unfairness of the ban, since there was no microbiological evidence pointing to Egyptian seeds, only circumstantial epidemiological evidence.)

As the epidemiological evidence pointing to sprouts got clearer, the microbiological evidence remained cloudy – underlining the problem with Dalli’s insistence on microbiological proof. In fact, the microbiological evidence got briefly cloudier, as cucumbers, though this time not from Spain, made their second appearance as a possible culprit.

The same June 8 Der Spiegel story that updated the case pointing to sprouts was headlined: “E. coli Mystery Deepens as Deadly Strain Found on Cucumber.” The story begins:

Investigators in Germany on Wednesday discovered the deadly EHEC strain of E. coli on food for the first time since the outbreak – on a piece of cucumber retrieved from the garbage of a family infected with the bacterium.
A spokesman for the Health Ministry of the eastern state of Saxony-Anhalt said the pathogen was the same as the one that has caused 26 deaths and infected more than 2,600 people, most of them in northern Germany, since early May.
However, as the cucumber had been in the garbage bin for two weeks, it was impossible to determine conclusively where the bacteria came from and how it got into the trash, the spokesman said. He added that a member of the family of three had mentioned eating cucumbers before falling ill. The family had not visited northern Germany.

So assuming the Saxony-Anhalt lab got it right – it’s O104:H4 on a victim’s cucumber – what’s a consumer to conclude? Might the cucumber have been contaminated by sharing a market basket, refrigerator, salad, or garbage can with some contaminated sprouts? Might there be two separate contaminations? Or might all that epidemiological evidence against sprouts be mistaken – could the problem be cucumbers after all?

The spokesman for the Saxony-Anhalt health ministry went unnamed in the Reuters story. But at least two other news media, AFP and German Central Broadcasting, named Holger Paech from the state Ministry of Social Affairs (not health) as their Saxony-Anhalt source. All three stories reported basically the same facts: that a piece of cucumber found a week earlier in the compost of a family with EHEC had been identified by local Saxony-Anhalt authorities as contaminated with E. coli O104:H4, raising the possibility that the family had gotten EHEC from the cucumber. (Other possibilities: that the cucumber had gotten EHEC from the family, or from contaminated sprouts in the family’s salad.)

We can find only one media reference to a federal source talking about the Saxony-Anhalt cucumber, an unnamed spokeswoman from Germany’s Federal Office for Consumer Protection and Food Safety quoted in a June 9 story in Stern. Here’s the quote, in toto: “‘This agency can not draw conclusions,’ said a spokeswoman for the Federal Office for Consumer Protection and Food Safety.” [“‘Aus diesem Fund können keine Rückschlüsse gezogen werden,’ sagte eine Sprecherin des Bundesamts für Verbraucherschutz und Lebensmittelsicherheit.”]

The not-quite-smoking cucumber lasted a day – and then disappeared without a trace. (Many news outlets had missed it entirely.) A June 10 AFP story reports its disappearance this way:

Authorities in the eastern German state of Saxony-Anhalt played down the discovery of cucumber pieces with traces of the killer bacteria in the two-week-old rubbish of a Magdeburg family who fell ill.
“According to the information we have now, this is not a decisive lead,” a spokesman for the state social affairs ministry told AFP.
Tests were continuing meanwhile at an organic sprout farm on which suspicion had fallen at the weekend.

We have been unable to determine what happened here.

  • Did the Saxony-Anhalt lab make a technical error? If so, how did it happen, and who discovered it?
  • Did Holger Paech misspeak the day before?
  • Did Saxony-Anhalt refuse to share its sample with federal authorities?
  • Did the federal lab retest the sample and fail to find O104:H4?
  • Was the cucumber contaminated with O104:H4 as Saxony-Anhalt claimed, but the absence of epidemiological evidence pointing to cucumbers led the experts to decide that it was probably cross-contamination from sprouts?

We haven’t a clue. We have found neither an official announcement of the “discovery” of the O104:H4-contaminated cucumber nor an official announcement that the discovery was an error. In fact, we have found nothing on any official website (local, state, or federal) confirming, denying, or clarifying the finding. And we have found no media coverage of the Saxony-Anhalt cucumber after June 10.

Notice that the June 10 AFP story doesn’t deny or retract the earlier story. The reporter says authorities “played down” the cucumber connection. The unnamed official quoted doesn’t say that the cucumber isn’t contaminated with O104:H4 after all, only that the cucumber isn’t “a decisive lead.”

This column urges risk communicators to proclaim their uncertainty – and their degree of uncertainty – about information that is possibly or probably but not definitely true. In this AFP story, a Saxony-Anhalt official is using uncertainty exactly backwards: to back away from previous information, to imply that because it’s uncertain it isn’t true.

On June 10, the Saxony-Anhalt cucumber having come and gone without a trace, German officials finally made a confident claim: It’s sprouts.

A June 10 “Joint Declaration” of the Federal Institute for Risk Assessment (BfR), the Federal Office of Consumer Protection and Food Safety (BVL), and the Robert Koch Institute (RKI) reviews the various epidemiological studies that pointed the finger at the Gärtnerhof Bienenbüttel sprouts. It goes into some detail on the detective work that was needed, making up for victims’ faulty memories about what they ate by studying restaurant menus and group photos taken around the dinner table.

The Joint Declaration concludes that “sprouts should not be consumed raw”; that “all foods originating from the farm business in Lower Saxony” should be taken out of circulation; and that “the existing general recommendation not to consume cucumbers, tomatoes and leaf lettuce in northern Germany no longer has to be complied with.”

The June 10 Reuters story is headlined “Germany pins down E. coli: ‘It’s the bean sprouts.’” It begins:

Germany said on Friday that deadly E. coli bacteria that have killed 31 people and hit farmers across Europe almost certainly came from contaminated bean sprouts grown at an organic farm in northern Germany.
“It’s the bean sprouts,” said Reinhard Burger, head of the German centre for disease control, confirming that the salad vegetable was the common denominator among the thousands who had fallen sick.
Government scientists said traces of the deadly strain were detected in a packet of bean sprouts from the farm found in a family’s rubbish bin in the western state of North Rhine-Westphalia. Two of the family fell ill after eating them.
Their announcement came just hours after Burger’s team in Berlin said it had identified bean sprouts or similar shoots as the most likely source of the outbreak, citing a study of patients and the food they ate….
The state consumer protection agency in North Rhine-Westphalia said it had found the first direct E. coli link to the organic farm in the neighbouring state of Lower Saxony, near the small town of Bienenbuettel….
“All the registered cases in this study had consumed these bean sprouts,” said Burger. “The test method made it possible in an epidemiological way to isolate the source of the outbreak, with a high probability, to the consumption of bean sprouts.” …
Gert Lindemann, the minister, had earlier said that alfalfa, mung bean, radish and arugula sprouts from the farm near Bienenbuettel might all be linked to the outbreak.
“The chain of evidence pointing to bean sprouts is flawless. For us, the source of the outbreak is definitely the farm in Bienenbuettel,” Lindemann said on Friday. The farm has been shut down and is no longer delivering vegetables to market.
Authorities said on Friday it was now safe to eat tomatoes, cucumbers, and leafy salads, food originally suspected as the source, but bean sprouts should be avoided while studies continued.
Germany had also came under fire for hastily blaming the outbreak on Spanish cucumbers, comments it later withdrew, and a failure to produce conclusive evidence of the source.

Early in the course of the outbreak, armed with weak evidence pointing to Spanish cucumbers, German authorities had trouble proclaiming uncertainty, and couldn’t bring themselves to correct the Hamburg Institute’s overly certain-sounding news release – leading most reporters to think they had unequivocally “blamed” the cucumbers. A bit later, armed with much stronger but still not ironclad evidence pointing to sprouts, German authorities bent over backwards to sound uncertain – leading some reporters to think they were far less confident than they actually were. But by the time of the June 10 Joint Declaration, the evidence against sprouts was extremely strong, and the same authorities had no trouble at all proclaiming their near-certainty.

The evidence responsible for the experts’ confidence was chiefly epidemiological, not microbiological. The June 10 Joint Declaration of Germany’s three federal food protection agencies makes no mention of any actual finding of O104:H4 in any sample of sprouts.

But the third paragraph of the Reuters story briefly mentions that O104:H4 had been found in sprouts in a victim’s rubbish bin. A different June 10 news story makes a bigger deal of this discovery. Headlined “Deadly E. coli found on bean sprouts,” the story begins:

German officials on Friday said they had found the first direct evidence of deadly E. coli bacteria on vegetable sprouts thought to have killed 33 people and left over 3,000 ill.
The discovery was made in a packet of sprouts in the garbage can of two sick people living in North Rhine-Westphalia, the state’s Consumer Affairs Ministry announced.
This is “the first time an unbroken chain of evidence” has been found linking the enterohaemorrhagic E. coli (EHEC) bacteria to a northern German sprouts farm, the ministry said….
The tests carried out by health institutes confirmed the presence of the virulent bacteria strain EHEC-O104 responsible for the current outbreak, the ministry added….
North Rhine-Westphalia Consumer Affairs Minister Johannes Remmel welcomed the discovery, but said there remained “a small element of uncertainty” as the bacteria was found in an open packet of sprouts rather than an unopened one which would then have been uncontaminated by the garbage can.

Unlike the two earlier microbiological leads that turned out misleading (the Spanish cucumbers with a strain of EHEC that wasn’t O104:H4, and the Saxony-Anhalt cucumber with – maybe – O104:H4), the North Rhine-Westphalia sprouts sample was confirmed as actually contaminated with the outbreak strain. A June 11 news release from the BfR (Germany’s Federal Institute for Risk Assessment) says:

Scientists of the National Reference Laboratory for Escherichia coli at the Federal Institute for Risk Assessment (BfR) have confirmed the results of the Land authorities in North Rhine-Westphalia: Raw sprouts contaminated with EHEC, which originated from the household of EHEC patients in North Rhine-Westphalia, were contaminated with the EHEC strain O104:H4. “This laboratory diagnostics result is another link in the chain of evidence suggesting that raw sprouts have to be considered as an essential source for the EHEC infections of the last weeks and confirms the epidemiological findings,” said BfR President Professor Dr. Dr. Andreas Hensel.

This piece of microbiological confirmation was useful, but not pivotal. The epidemiological case was already strong before an open package of sprouts from a victim’s household garbage tested positive for O104:H4. And the microbiological evidence had a loophole. It was probable that contaminated sprouts from that package had made the family sick. But it was possible that a sick family member had contaminated the package of sprouts! Or the EHEC could have been in some other food (a batch of radishes, say), and the sprouts in the open package were contaminated by a deadly radish that was later eaten and unavailable for testing. Since the epidemiological evidence also pointed toward sprouts, not some other food, the experts were inclined to dismiss the cross-contamination alternative hypothesis and blame the sprouts. But they couldn’t be sure.

As of August 12, the BfR website still has the June 11 news release confirming the O104:H4 serotyping of the open package of sprouts. But there is some question whether the Federal Office of Consumer Protection and Food Safety (BVL) agrees. On July 12, towards the end of the outbreak, an unnamed person from the BVL gave a presentation entitled “EHEC O104:H4 Outbreak in Germany.” It includes these bullet points:

  • More than 10,000 samples were taken and analyzed by the responsible German authorities
  • E.Coli O104:H4 could not be detected and confirmed in any food of plant origin

Did the BVL official decide to ignore the opened package of sprouts because of the alternative hypothesis of cross-contamination? Did the official decide not to mention this piece of microbiological evidence because the epidemiological evidence was so much stronger? Did he or she decide not to mention it because of the two previous microbiological red herrings (the Hamburg Institute’s Spanish cucumbers that turned out to have the wrong strain of E. coli; and the trash bin cucumber said by a local Saxony Anhalt official to be contaminated with the outbreak strain, but neither confirmed nor disavowed by national agencies)? Or was this instance of O104:H4 “detected and confirmed” in a “food of plant origin” just left out of the BVL presentation by accident?

Another unanswered question about the confirmed sprouts sample is the relative lack of media attention it received, compared with the intense coverage of the suspected (and later ruled out) Spanish cucumbers. We have noted that reporters and politicians generally prefer microbiological evidence to epidemiological evidence – even when the epi evidence is technically stronger. Why then didn’t the media make more of an officially-confirmed instance of O104:H4-contaminated sprouts, after having made so much of an unknown strain of E. coli on Spanish cucumbers?

Had reporters become as skittish about microbiological evidence by this time as they were about epidemiological evidence earlier on? Having been “fooled” about the Hamburg Institute’s Spanish cucumbers, were they reluctant to focus too much on the BfR’s sprouts package? Or had reporters successfully climbed the steep learning curve by then, coming to understand that the epidemiological evidence very strongly pointed to sprouts, with or without a smoking sprout – and that an open package that might have been cross-contaminated wasn’t much of a “smoking sprout” anyway?

We are extremely uncertain about the answers to these questions. We have, in fact, the ultimate degree of uncertainty: We simply do not know.

In general, the media and the public (and officials who are not public health experts) strongly prefer microbiological evidence over epidemiological evidence – while at the same time responding well to warnings based on epidemiological evidence. But there seems to have come a point in this outbreak when nearly everyone “got it” that the epidemiological evidence was good enough.

We strongly believe that this point could have come sooner if officials had proclaimed their uncertainty – and their degree of uncertainty – every step of the way.

Meanwhile, the key remaining question about the source of the outbreak wasn’t whether contaminated sprouts from Gärtnerhof Bienenbüttel were responsible. The key question was how the sprouts got contaminated.

As German authorities stressed often (though not often enough to satisfy the management of Gärtnerhof Bienenbüttel), tracking the outbreak to a particular farm didn’t necessarily mean it originated on that farm. It could have originated earlier or later – in which case produce from other farms could be contaminated as well. The June 10 Joint Statement (the “It’s the sprouts” statement) emphasized this point. It began to focus on a possible contamination source that would soon come to dominate the story of this outbreak: seeds.

According to the latest information available at the moment, it cannot be excluded that the responsible pathogen was introduced to the horticulture company by humans. Introduction via water, pre-suppliers or seed is also conceivable. This is currently being investigated by checking supply relations and conducting lab tests….
The national and regional governments will continue to investigate and analyze distribution channels to find out whether contaminated seed is being used for the production of sprouts in other companies, or otherwise marketed.
Because the bean sprout seed used can also be a possible source of entrance, other bean sprout producing businesses must also be regarded as potential distributors of EHEC O104:H4.

The possibility of seed contamination was raised again a day later. The very short June 12 BfR news release is headlined “EHEC outbreak: BfR also advises against the consumption of home-grown raw sprouts and germ buds”:

Based on epidemiological information provided by the public authorities in Lower Saxony, the suspicion that sprout seeds could have been an underlying cause of contamination of the sprouts is corroborated. Lower Saxony reports about a current case in which home-grown sprouts have possibly been the cause for an EHEC infection in a family. However, it has not yet been possible to detect the pathogen in the seeds. It is standard practice to grow sprouts at home. The cultivation receptacles are available in many stores. “If the seeds are already contaminated with germs, even observance of kitchen hygiene rules does not protect from an EHEC infection,” says BfR President Professor Dr. Andreas Hensel. “For reasons of precaution, BfR, therefore, also recommends to refrain from consuming home-grown raw sprouts for the time being.” …
The federal and regional authorities continue to work at full speed on the tracking of the input pathway for the contamination of sprouts and possibly seeds with EHEC.

This release doesn’t seem to have received much media play. Its focus is on adding home-grown sprouts to the previous official warning against store-bought and restaurant sprouts. The first sentence does mention “the suspicion that sprout seeds could have been an underlying cause.” But it doesn’t stress that suspicion, and it doesn’t draw out its implications. With the benefit of hindsight (see Act Three), here’s what it might have said:

We’re beginning to think the contamination may have originated with sprout seeds, not with fresh sprouts – although it is fresh sprouts that have made people sick. We’re not sure yet. But if that turns out to be true, it has some important implications. It would mean that the organic farm where the contaminated sprouts were grown was not at fault, that they were the unwitting recipient of contaminated seeds. It would mean that other outbreaks are possible, originating with the same seed source but grown on different farms. It would mean that tracking down the source and getting the seeds off the market is still important even though this outbreak seems to be waning, since sprout seeds have a long shelf life and E. coli can survive for a long time in seeds. And it would mean that home-grown sprouts should also be considered potentially dangerous.

The possibility that contaminated sprout seeds might be behind the outbreak was a radical departure from the widespread assumption that the culprit had to be raw fresh vegetables, be it lettuce, tomatoes, cucumbers, or sprouts. Farmers (including the Gärtnerhof Bienenbüttel farm) contributed to this assumption by saying that their produce was highly unlikely to be contaminated because their farming practices were clean – they don’t use manure as fertilizer, their water source doesn’t pass through any cattle farms, etc. Officials (including BfR and WHO) contributed to this assumption by saying that the outbreak would probably tail off soon and its source might never be known because fresh sprouts don’t last long and new batches aren’t coming to market now that everybody has been warned.

From the outset, there had been periodic attention to the possibility that produce could be contaminated after it left the farm – a claim featured heavily in the early Spanish cucumber denials. But very few commentators had noted the possibility that produce could also be contaminated before it got to the farm … via contaminated seeds.

Even after the June 10 Joint Statement and the June 12 BfR home-grown sprouts warning, most people assumed the source of the O104:H4 outbreak was Gärtnerhof Bienenbüttel, that organic farm in northern Germany. Then O104:H4 turned up in a cluster of food poisoning victims in France.

Act Three: French Sprouts … English Sprout Seeds … Egyptian Sprout Seeds

There wasn’t much E. coli news between June 12 and June 24. New cases continued to surface, though with diminishing frequency, most of them still in northern Germany. Perhaps the most interesting new development during this period was the June 17 announcement of the outbreak’s first known case of human-to-human transmission. Twenty people came down sick with O104:H4 after contact with a kitchen employee at a catering company in Hesse who had handled contaminated sprouts but didn’t get sick herself.

Then on June 24 came an entirely new development: French authorities announced that a cluster of seven E. coli patients in the Bordeaux area had been linked to both sprouts and O104:H4. The patients had no known connection to the German outbreak. But they all had bloody diarrhea, most of them had HUS, and several had already tested positive for O104:H4. Most of the victims had attended a school event in a nearby suburb where they ate soup and other dishes garnished with locally produced sprouts.

Given the rarity of O104:H4, it seemed unlikely that the two outbreaks could be completely unrelated – especially since they had sprouts in common. Since the French patients hadn’t been to Germany or eaten sprouts from Germany or had contact with someone who got sick in Germany, was it possible that there could be a shared link earlier in the French and German sprouts supply chains?

The likeliest common denominator: seeds.

The French sprouts were almost immediately traced to a seed company in the United Kingdom, Thompson & Morgan. A June 25 statement from the U.K. Food Standards Agency does a pretty decent job of proclaiming its uncertainty. It acknowledges a “possible link” to the seed supplier (though it doesn’t name the company) and warns consumers not to eat uncooked sprouted seeds:

The investigations into the outbreak of E. coli in France have suggested a possible link to sprouting seeds from a company based in the UK. To date, no cases of food poisoning have been reported in the UK linked to the outbreak in France. We are in close contact with the Health Protection Agency which is actively monitoring the situation.

Reporters also used appropriately nuanced uncertainty language in their June 25 coverage. The BBC’s story, for example, says that there is a “possible link,” and that “it is thought” a number of the French outbreak victims ate sprouts that are “believed to have been grown from seeds sold by Thompson & Morgan.” It quotes France’s secretary of state for consumer affairs, Frederic Lefebvre, as saying: “The link between the symptoms and eating of the sprouts so far has not been definitively established.”

The only overconfident language in the BBC story comes from Thompson & Morgan itself. An unnamed spokeswoman is quoted as saying: “We are very confident the problem is not with our seeds. People can still grow these seeds and use these seeds with absolute confidence.”

And Thompson & Morgan managing director Paul Hansord is quoted as saying: “We make sure that everything we do is to a high standard.” But Hansord also told the BBC that the company bought its seeds in bulk from suppliers around the world – which certainly left open the possibility that it could have unwittingly bought and resold seeds that were already contaminated.

Thompson & Morgan sounds similarly overconfident in the June 25 issue of The Telegraph:

Thompson & Morgan said in statement it was co-operating fully with the investigation but said it was an “unsubstantiated link” between the outbreak and the consumption of sprouting seeds.
The firm said it has sold many hundreds of thousands of packets of seeds throughout the UK and Europe including more than 100,000 packets in France from more than 500 outlets….
“To date, there have been no reported incidents of any problems either in France, the UK or anywhere else they are sold,” it stressed.
“We note that the French outbreak seems to be localised to a specific event, which would indicate to us that something local in the Bordeaux area, or the way the product has been handled and grown, is responsible for the incident rather than our seeds.”

On June 27, Hansord told The Telegraph that Thompson & Morgan had no intention of recalling the seeds it had already sold, though it had promptly stopped selling rocket, white mustard, and fenugreek seeds – the varieties implicated in the French outbreak. And Hansord told Farmers Guardian that “There is only a very slim chance that this type of illness can be carried in seeds and we believe it is highly unlikely that they are to blame.” This was a very over-confident statement, considering that seed researchers have long known that E. coli can remain viable for at least three years in dried sprouting seeds such as alfalfa.

We looked at the Thompson & Morgan website on June 27, and found absolutely nothing about E. coli – not even a denial. So we went onto the company’s Twitter feed and tweeted: “What’s the story about possible E. coli contamination from your seeds?”

The company didn’t post our tweet, or others it was obviously receiving, but it did post its one response to them all: “Although no link has been established we’re keen to offer guidance to customers. See our latest news here –” That link took readers to a “Sprouting Seeds Update” that remained unchanged until some time in late July or early August, when it was replaced with links to two Food Standards Agency updates. It read:

You may be aware of recent news coverage that three varieties of Thompson & Morgan sprouting seeds have been linked to an outbreak of E. coli at a localised event in France, according to claims by French Minister Frederic Lefebvre. This remains unsubstantiated.
Although there has been no established link, as the health and safety of the public is always of paramount concern to Thompson & Morgan, we have temporarily withdrawn from sale the following varieties as a precautionary measure….

Overconfident defensiveness is such a conventional industry response to foodborne outbreaks that we almost forget to notice how offensive and self-defeating it is. Thompson & Morgan could easily have pointed out the absence of microbiological evidence without painting itself into a corner. Even early on, before the Egyptian fenugreek seed link materialized, the company could have put something like this onto its website:

You may be aware of recent news coverage that three varieties of Thompson & Morgan sprouting seeds have been linked to an outbreak of E. coli at an event in France. There is no microbiological evidence linking Thompson & Morgan seeds with this outbreak.
But it is pretty well established that we sold the seeds for the sprouts that most of the victims ate, though we did not sell the seeds for the sprouts that are implicated in the much larger German outbreak of the same rare strain of E. coli. So logically there are four main possibilities.
1. The seeds for the French sprouts could have been contaminated before we got them – and perhaps the seeds for the German sprouts (which never passed through our hands) could also have been contaminated at the same source prior to distribution.
2. The seeds for the French sprouts could have been contaminated here at Thompson & Morgan. That would mean the German contamination happened elsewhere – two separate instances of contamination with the same rare strain.
3. The seeds for the French sprouts could have been contaminated after we shipped them. That would also mean two separate instances of contamination.
4. The seeds could have been uncontaminated. That would mean the French sprouts must have been contaminated during or after the growing process – which would once again mean separate contaminations responsible for the French and German outbreaks.
The first possibility seems the likeliest to us, since it explains both outbreaks with one source of contamination. We think the other three are less likely, but all are possible.
The one that most horrifies us, of course, is #2, the possibility that we could have contaminated some seeds that later led to severe illnesses. We know how careful we are, so we doubt that this is what happened. But it might be. We are certainly taking that possibility seriously, reviewing our procedures to see if we can think of any safety improvements.
We’re also thinking about #1, trying to figure out what more we can do to protect against the possibility of receiving contaminated seeds from our suppliers.
Although there has been no established link, as the health and safety of the public is always of paramount concern to Thompson & Morgan. We have temporarily withdrawn from sale the following varieties as a precautionary measure….

Of course Hansord was absolutely right that there was no microbiological evidence of O104:H4 contamination in any Thompson & Morgan seeds. The only reason food safety authorities were thinking “seeds” at all was the fact of two simultaneous, apparently unconnected outbreaks, both from sprouts contaminated with a rare serotype of E. coli.

But Thompson & Morgan’s seeds were linked only to the Bordeaux sprouts, not to the Gärtnerhof Bienenbüttel sprouts. Thompson & Morgan wasn’t the common denominator connecting the two outbreaks.

The common denominator began to materialize two days later, on June 29, when the European Food Safety Authority and the European Centre for Disease Prevention and Control released a “Joint Rapid Risk Assessment” on the Bordeaux outbreak. By then the number of cases had risen to fifteen. Three of the victims’ illnesses had been identified as O104:H4; at least eleven of them had eaten fenugreek, mustard and rucola sprouts at the same event. The report continued:

These suspected sprouts were locally produced, and were not imported from the farm implicated in the outbreak in Germany. Initial investigations suggested that the seeds used for the sprouts in France were distributed by a UK-based company. Further investigations are now being carried out to determine the origin of the suspected sprout seeds from this French cluster and to establish if there is any link between that cluster and the large outbreak reported from Germany….
The currently available information on the French cluster is in favour of a link between these two health events for several reasons. The clinical picture of the French HUS cases in Bordeaux is similar to that of the cases reported from Germany: the cases are adults, the majority are women and the majority are presenting with HUS. The microbiological characteristics of the isolated strain of E. coli O104:H4 from three of the French HUS patients seem similar to the isolated strain in the German outbreak….
[T]he occurrence of the French cluster, considering the elements outlined above, might support one of the initial hypotheses on the cause of the German outbreak, i.e. that the seeds used for sprouting and distributed to local producers or retail outlets contained a level of E. coli O104:H4 contamination, ultimately leading to contaminated sprouts destined for human consumption.
Following this hypothesis, contamination of the seeds could have occurred at any stage in the long and complex supply chain between seed production, transport, packaging and distribution. This would also mean that other batches of potentially contaminated seeds are still available within the EU and perhaps outside….
The tracing back is progressing and has thus far shown that fenugreek seeds imported from Egypt either in 2009 and/or 2010 by the company AGA SAAT GMBH are implicated in both outbreaks. There is still much uncertainty about whether this is truly the common cause of all the infections as there are currently no positive bacteriological results. In particular, the 2009 lot appears to be implicated in the outbreak in France and the 2010 has been considered to be implicated in the German outbreak.

So the tentative conclusion of the Joint Rapid Risk Assessment was that the French outbreak came from contaminated sprout seeds imported from Egypt by a German company (AGA SAAT), then sold to a British company (Thompson & Morgan), then sold to a French company that sold the seeds to an individual who grew the sprouts that made people sick at a school event near Bordeaux. The German outbreak would presumably have started the same way, with Egyptian seeds imported by AGA SAAT and then sold to various growers, including Gärtnerhof Bienenbüttel.

By the time this June 29 report was issued, the epidemiological evidence linking the German and French outbreaks to Egyptian fenugreek seeds was already strong. It was much stronger than the evidence the Hamburg Institute had had linking the German outbreak to Spanish cucumbers. It was probably stronger than the very early, very general evidence linking the German outbreak to some kind of salad greens. It was at least as strong as the evidence linking the German outbreak to Gärtnerhof Bienenbüttel sprouts when that link was first announced. It wasn’t certain, but it was getting close – although there was still no evidence bearing on where, when, or how the Egyptian fenugreek seeds became contaminated.

The June 29 report bends over backwards to proclaim the remaining uncertainty about the cause of the two outbreaks:

There is still much uncertainty about whether this is truly the common cause of all the infections as there are currently no positive bacteriological results.

It also emphasizes that microbiological evidence may never be forthcoming, and epidemiology may have to do:

However, E. coli contamination is difficult to identify in food items, including seeds, and it is possible that microbiological evidence cannot be established.

For the most part, media coverage of the Egyptian fenugreek seed connection faithfully replicated the uncertainty in the report. We’re not sure why. Were the authorities doing a better job of proclaiming their uncertainty? Or were reporters doing a better job of noticing the uncertainty, especially after the previous false lead (Spanish cucumbers)? It’s probably some of each, we think.

The June 30 Reuters story, for example, carries this lede: “Imported fenugreek seeds from Egypt may be the source of highly toxic E. coli outbreaks in Germany and France that have killed at least 48 people, according to initial investigations by European scientists.” The lede’s tentativeness is amplified later in the story:

The ECDC and EFSA said a batch of fenugreek seeds imported from Egypt in 2009 appeared to be implicated in the outbreak in France, and a 2010 batch was “considered to be implicated in the German outbreak.”
But they said there was still “much uncertainty” about whether these seeds from Egypt were “truly the common cause of all the infections” as there were currently no positive bacteriological results.

If anything, this is a bit too tentative. As readers of this column surely realize by now, the absence of microbiological evidence does not disprove the epidemiological evidence linking Egyptian seeds imported by AGA SAAT to the French and German outbreaks – a point the Reuters article never clearly makes.

We have seen this pattern before in our outbreak chronology, with regard to sprouts from Gärtnerhof Bienenbüttel. Mostly because of their failure to correct the Hamburg Institute’s overly certain-sounding Spanish cucumber announcement, German and EU authorities had acquired a bit of a reputation for guessing prematurely and guessing wrong about the source of the outbreak. Now that they had finally closed in on the source, they were hedging. Official uncertainty communication had moved from one extreme to the other, from sounding too certain to sounding too uncertain.

By the time this Reuters story was distributed on June 30, something important had changed in the Joint Rapid Risk Assessment report. The original version, quoted above, had named the German seed import/export company believed to have imported the contaminated seeds from Egypt and sold them to Thompson & Morgan in the U.K.: AGA SAAT. Here again is the paragraph that names AGA SAAT:

The tracing back is progressing and has thus far shown that fenugreek seeds imported from Egypt either in 2009 and/or 2010 by the company AGA SAAT GMBH are implicated in both outbreaks. There is still much uncertainty about whether this is truly the common cause of all the infections as there are currently no positive bacteriological results. In particular, the 2009 lot appears to be implicated in the outbreak in France and the 2010 has been considered to be implicated in the German outbreak.

That’s the original version. It was a PDF, and many people had downloaded it, including a reporter who sent it to us. We reposted it here for safekeeping. But the version on the ECDC website today reads differently:

The tracing back is progressing and has thus far shown that fenugreek seeds imported from Egypt either in 2009 and/or 2010 are implicated in both outbreaks. There is still much uncertainty about whether this is truly the common cause of all the infections as there are currently no positive bacteriological results. In particular, the 2009 lot appears to be implicated in the outbreak in France and the 2010 has been considered to be implicated in the German outbreak.

AGA SAAT is mentioned twice more in a later paragraph in the original: “In particular, an investigation on the distribution of seeds from these lots throughout Germany and Europe by AGA SAAT has been urgently requested. The export by AGA SAAT of some of its imported seeds to another company in the UK (from where seeds were exported to France) demonstrates the necessity of this information.” These references to AGA SAAT were also removed.

Why? CIDRAP News explains in a July 1 story headlined “ECDC erases name of importer from report on E coli cases”:

[T]he name of the German firm was removed from the online ECDC report a few hours after it was first published, CIDRAP News has learned. The current version of the report omits the name without giving any explanation for the change….
In response to a query today about the reason for removing the company’s name, ECDC spokeswoman Caroline Daamen told CIDRAP News by e-mail, “In the initial risk assessment posted on the website, EFSA and ECDC reported information that had been made available to support the ongoing outbreak investigation. However, some key partners involved felt that it may unnecessarily harm the company to publish its name while the investigations are still ongoing. So it was thought more appropriate to remove the name of the company from the final report….”
Also today, CIDRAP News received an e-mail message from a German attorney who said that he represented AGA SAAT GMBH and demanded that the firm’s name be removed from our Jun 29 news story based on the ECDC report. The message threatened legal action if the name is not removed.

The June 30 Reuters story names AGA SAAT, but can no longer attribute the name to the Joint Rapid Risk Assessment report. It says:

EFSA spokeswoman Lucia de Luca would not confirm or deny media reports that the seeds had come from Egypt via a single German seed importer. “The investigations are still ongoing,” she told Reuters.

But AGA SAAT’s legal threats, aimed at getting its name out of the coverage, did not prevent it from making its own aggressive, overconfident safety claims – leaning on the absence of microbiological evidence to imply that the accumulating epidemiological evidence was not worth considering … and to imply that health authorities agreed. From the same June 30 Reuters story:

German organic seed trader agaSAAT told Reuters it had distributed seeds to Thomson & Morgan, a British seed trader cited as a possible source for the outbreak in France, but had been cleared by health authorities.
“We put our seeds under microbiological testing and there have been no positive tests for E. coli,” agaSAAT’s chief executive Werner Arts said. “This has also been confirmed by German health authorities.”

A different version of the Reuters story has an even more categorically overconfident version of these two paragraphs:

German organic seed trader agaSAAT told Reuters it had distributed seeds to Thomson & Morgan, a British seed trader cited as a possible source for the outbreak in France, but had been cleared by health authorities. “We’ve sent seeds to Thomson & Morgan in Britain, but the seeds were not contaminated with E. coli,” agaSAAT’s chief executive Werner Arts said.
“We put our seeds under microbiological testing and there have been no positive tests for E. coli,” he added. “This has also been confirmed by German health authorities.” Arts said his firm buys seeds from Egypt because “that’s where a lot come from. Everyone does it.”

Curiously, AGA SAAT injected itself into the O104:H4 controversy a week before the Joint Rapid Risk Assessment report (momentarily) identified the company as a possible source of the outbreak. A June 21 AGA SAAT news release begins (in German) this way:

The EHEC pathogens in Germany have killed at least 39 people and Germans are scared. This dangerous intestinal disease was introduced into the food supply by sprouts from an organic farm in Lower Saxony. The German population is uncertain because of the frequent food scares in recent years; people ask themselves if any food is still safe. At agaSAAT GmbH we have a long record of intensive monitoring to make sure our spices and sprouting seeds come from a safe source. Additional de-germination of seeds guarantees safe and healthy enjoyment.
Many consumers today feel strongly about a healthy diet. And sprouts are part of this diet. They are used, for example, as a salad ingredient and are very rich in vitamins. But these very same healthy sprouts have led to the dangerous EHEC infections that have cost the lives of 39 people.
In addition, agaSAAT GmbH in Neukirchen-Vluyn sells sprouting seeds for the production of sprouts. “Our sprouting seeds are purchased exclusively from audited farms after intensive checks,” says Werner Arts, Managing Director of agaSAAT GmbH. In addition to visual controls, examinations in both internal and external laboratories are part of our procedures. “In this manner we can be certain that we purchase only perfectly safe premium goods,” said Arts.

Why did AGA SAAT start defending itself a week before anybody publicly linked it to the outbreak? Well before AGA SAAT issued its June 21 news release, German and European investigators had tentatively tracked the German outbreak to sprouts from the Gärtnerhof Bienenbüttel farm. The authorities were aggressively exploring how the sprouts might have become contaminated, and one of their investigative paths would surely have been to trace back the sources of sprout seeds used by the farm.

We speculate that this ongoing traceback effort may be the reason why AGA SAAT issued its June 21 news release, extolling its safety precautions. Once sprouts from the Gärtnerhof Bienenbüttel farm were implicated, it seems likely that AGA SAAT checked and discovered that it had imported some fenugreek seeds purchased by the farm – and decided to launch a preemptive PR effort. Note we are saying “speculate” and “seems likely.” We are trying to signal our uncertainty that this is why AGA SAAT issued a news release about its confidence in its seeds several days before the French outbreak story broke and public attention began to focus on sprout seeds.

Later, AGA SAAT’s website became highly combative on the question of its possible importation of contaminated fenugreek seeds. Its German-language news page has an angry July 20 entry lambasting “some institutions and the media in recent weeks” for “spreading half-truths” about the O104:H4 outbreaks.

A July 1 Bloomberg story doesn’t name AGA SAAT, but nicely nails the increasing confidence of European health authorities, based entirely on the mounting epidemiological evidence, with no smoking fenugreek seed in sight:

After tracing common food sources, epidemiologists found fenugreek seeds from Egypt could be implicated in both outbreaks, authorities said.
Scientists are checking bacterial specimens to confirm the link and solve a mystery hanging over the outbreak, which caused illness in 4,125 people and killed 49 in 13 European countries.
“The one common source here that keeps coming up over and over again is Egyptian seeds,” Michael Osterholm, director of the Center for Infectious Disease Research and Policy at the University of Minnesota, said in an interview yesterday. “Fenugreek is showing up clearly in the French outbreak and showing up clearly in the German outbreak.” …
Leftover mustard and rocket seeds, gazpacho and tap water samples from the community center have been sent for microbiological analysis, as have samples of rocket, mustard, fenugreek and other seeds from the French gardening retailer. Preliminary results are being analyzed, the researchers said.
Microbiological testing won’t always find the pathogen, especially if the contamination is sporadic and at low levels, Osterholm at the University of Minnesota said.
“Absence of this in a product doesn’t mean anything about what product’s involved,” he said. “Even if 1-to-2 percent of the seeds are contaminated, it still means that potentially many thousands, if not millions, of people have been exposed given the wide distribution of those seeds and how much of this is consumed.”

But other coverage missed or mistrusted the strong scientific basis for the Egyptian fenugreek hypothesis – in large measure (we strongly believe) because of the authorities’ failure to proclaim their uncertainty about earlier, shakier hypotheses.

Here’s the start of a July 1 International Business Times story headlined “Could E. coli Finger-pointing Cripple Egyptian Economy?”:

German officials originally pointed their fingers at Spanish cucumbers after the outbreak of the recent E. coli epidemic, in a costly accusation that still has the Spanish agricultural sector up-in-arms.
Now European investigators from the European Food Safety Authority are looking to Egyptian fenugreek seeds as a potential culprit, in what could be another costly allegation.
“There is still much uncertainty about whether this is truly the common cause of all the infections as there are currently no positive bacteriological results,” said the Authority’s recently published Joint Rapid Risk Assessment.
Unresolved allegations could cripple Egypt’s moves to rebuild its economy on agricultural exports at a time when the outcome of the revolution and democratization depends heavily on economic stability.
Last time European scientists pointed fingers, Spanish farming group Agricultural Cooperatives reported that the international community’s resultant import bans on Spanish cucumbers cost up to $287 million, endangering 70,000 jobs, largely located in economically disadvantaged Andalusia, in Spain’s South.

The first time officials (the Hamburg Institute) “pointed their fingers” at Spanish cucumbers, they had very weak microbiological evidence, but they sounded like they had a strong case. The remaining instances of finger-pointing – at Gärtnerhof Bienenbüttel sprouts, then at Thompson & Morgan seeds, and finally at Egyptian fenugreek seeds imported by AGA SAAT – were based on increasingly strong epidemiological evidence. By the time officials (the European Food Safety Authority and the European Centre for Disease Prevention and Control) “pointed fingers” at Egyptian fenugreek seeds, they had very strong epidemiological evidence, but were reluctant to sound confident without backup microbiological evidence.

In the beginning German and EU officials sounded more confident than they had any right to be (or to be more precise, they did not quickly and emphatically rebut Hamburg state officials, who sounded more confident than they had any right to be). In the end, German and EU officials sounded more uncertain than they had any reason to be.

The International Business Times reporter completely missed the difference between the two occasions. And nothing in the EFSA/ECDC Joint Rapid Risk Assessment report explains the difference – leaving the reporter to interpret the report’s uncertainty claims as an admission that their evidence was once again weak, rather than as an effort to bend over backwards to avoid another overconfident charge.

The lesson here: If your evidence is weak and you’re very unsure you’ve got the goods, proclaim that. If your evidence is strong though not quite ironclad, proclaim that. In both cases, proclaim the distinction: “This isn’t like that other time….”

Predictably, Egyptian denials were overly certain that the contaminated seeds did not originate in Egypt, leaning heavily on the absence of a microbiological smoking seed and on the acknowledgment of uncertainty in the Joint Rapid Risk Assessment report. On June 30, Al-Masry Al-Youm reported:

Minister of Agriculture Ayman Farid Abu Hadid denied on Thursday that Egyptian fenugreek seeds were responsible for the outbreak of E. coli in Europe, pointing to laboratory tests conducted by his ministry proving that Egyptian seeds for export were not infected.” …
The Egyptian Agriculture Export Council called on the European authorities not to announce suspected cases without being sure of the source of the bacteria.

And on July 2, Al Ahram reported:

Egyptian agricultural quarantine chief Ali Suleiman rejected European conclusions that deadly seeds originated there: “The bacteria’s presence in Egypt has never been proven or recorded until now,” Suleiman said, according to state news agency MENA. The European Centre for Disease Prevention and Control (ECDC) and the European Food Safety Authority (EFSA) said on Wednesday that initial investigations implicated the seeds, imported from Egypt, in the German and French outbreaks. But European food and health officials said there was still “much uncertainty” about whether the seeds from Egypt were “truly the common cause of all the infections” as there were currently no positive bacteriological results.

The Egyptian response is yet another example of industry and government officials proclaiming premature certainty about their “innocence,” while demanding that those with grounds for suspecting their probable “guilt” remain silent.

While AGA SAAT and the Egyptian government were trumpeting the authorities’ uncertainty about the source of the O104:H4 contamination, the authorities were becoming progressively less uncertain. On July 5 the European Food Safety Authority released a 23-page technical report. The news release accompanying the report stresses a key point that had repeatedly been missed by journalists and ignored by AGA SAAT and Egypt: “The report highlights that negative results from microbiological tests carried out on seeds cannot be interpreted as proof that a lot is not contaminated with STEC.”

The first paragraph of the release summarizes where the evidence was leading. It was probably a single lot of Egyptian fenugreek seeds that was responsible for the two outbreaks, but it might be several lots. It was pretty definitely Egyptian fenugreek seeds:

The EFSA Task Force established to coordinate investigations to track down the possible source of the French and German outbreaks of E. coli O104:H4 has concluded that one lot of fenugreek seeds imported from Egypt and used to produce sprouts is the most likely common link between the two outbreaks. However, it cannot be excluded that other lots of fenugreek imported from Egypt during the period 2009–2011 may be implicated. Based on these findings, EFSA recommends to the European Commission that all efforts be made to prevent any further consumer exposure to the suspect seeds and that forward tracing be carried out in all countries which may have received seeds from the concerned lots.

The technical report says the 15,000-kilogram fenugreek seed lot #48088 was probably contaminated before leaving Egypt, very likely as a result of farming practices that allowed exposure to animal or human fecal material. The lot left Egypt by boat for Rotterdam, in sealed containers, on November 24, 2009. On December 14 it was trucked to the importer in Germany (AGA SAAT, though the report omits the name). The German farm linked to the German infections (Gärtnerhof Bienenbüttel, also unnamed) received seeds from that lot on the same day it received seeds from another lot from the same Egyptian and German distributors, leading to some remaining uncertainty about whether more than one lot might have been contaminated. A U.K. seed packager (Thompson & Morgan, also unnamed) received seeds from lot #48088 from AGA SAAT (unnamed) on January 13, 2010; these seeds were eventually sold to the French garden supply store from which they were purchased by the individual who germinated the sprouts and served them at a school event near Bordeaux. All told, directly or indirectly, AGA SAAT sold fenugreek seeds from the potentially contaminated lot to 70 different companies, 54 of them in Germany and 16 of them in 11 other EU countries. Although many of the seeds may have already been used, some could still be present in the supply chain.

On the same day, July 5, the European Union banned the importation of Egyptian fenugreek seeds until the end of October.

Egyptian authorities weren’t convinced (of course). Their July 6 response to the EFSA report focuses on the repeated failure to actually find O104:H4 on any fenugreek seeds imported from Egypt. (It predictably fails to note that E. coli is hard to detect on seeds, especially if only a few seeds in a batch are actually contaminated.) The response also claims that there’s no way E. coli could have survived so long on dried seeds:

Lot no. 48088 had been exported on November 2009 as a dried seeds. Scientifically, the bacteria can not remain on this dry surface from 2009 till June 2011.

This latter claim isn’t just overconfident; it is simply false. Numerous studies have shown that E. coli can survive for years in dried seeds.

The last sentence of the July 6 Egyptian response is particularly interesting: “Any audit mission for the Egyptian situation is welcomed.” Actually, European authorities have been trying unsuccessfully since early July to negotiate conditions under which they will be allowed to pursue their investigation within Egypt. As India’s MSN News reported on July 16:

The European Union has delayed sending a delegation to Egypt to probe the presence of highly toxic E. coli bacteria in Egyptian fenugreek seeds, amid failure to agree on the framework of the team and the specific issues to be investigated.
The postponement is until next September as Cairo and Brussels have not agreed on the framework of the delegation and the specific issues to be researched, according to an EU spokesperson.

On August 9, an EU official implied that EU food safety investigators might be permitted to come to Egypt sooner than September. The EU special representative for the southern Mediterranean region, Bernardino León, told AhramOnline: “We will discuss the problem again when the EU’s delegation arrives to Egypt in the upcoming days.” As far as we can tell on August 12, this has not happened yet.

That’s pretty much the end of the story. The first U.S. O104:H4 death was reported on July 8, an Arizona resident who had traveled to Germany. A July 12 European Food Safety Authority report worried about how widely lot #48088 had been distributed, warning of a risk of more clusters and continued sporadic cases until the seeds reach their five-year shelf life in 2014. On July 13, the discovery of some asymptomatic cases in a German school raised new fears of possible new outbreaks to come, perhaps from person-to-person transmission or from food handlers who had been exposed to the contaminated sprouts months earlier.

And on July 26, the Robert Koch Institute (Germany’s federal disease control agency) released a statement declaring a semi-official end to the outbreak. It said the incubation period for EHEC was roughly three weeks, and three weeks had passed since July 4, the most recent illness onset date of a laboratory-confirmed case of O104:H4 with known epidemiologic links to the outbreak. It had tallied a total of 4,321 cases and 50 deaths in Germany. (As of July 21, the World Health Organization counted 4,075 cases and 50 deaths in Europe and North America, the vast majority in Germany.) Of the 4000-plus people sickened by E. coli O104:H4, over 900 had severe kidney damage or kidney failure caused by the infection.

Especially since some of the contaminated seeds could still be on the market or in people’s homes, nobody is ruling out the possibility of new clusters.

Perhaps the best way to end this long column is by quoting from “Crisis Point,” the single worst article we have found about risk communication aspects of this outbreak. It was posted on July 15 on “FreshInfo,” an online service of the Fresh Produce Journal, a U.K. trade publication. You have to register with the site to get access to the article, but it’s worth it for this stunning example of risk communication wrongheadedness.

The boldface intro to the article reads as follows:

Amid a series of allegations and recriminations, the European E. coli outbreak has been a catastrophe for the fresh produce industry. Anna Sbuttoni asks how best to manage a food crisis.

One of the main experts she asked was a man named Gordon Beattie. Here’s the passage that got our adrenalin flowing:

Gordon Beattie, chairman of crisis PR specialist Beattie Communications, insists that the key to managing a high-profile incident is to ban speculation and take control of the public message. His motto is “if in doubt, say nowt” [nothing], especially if the situation is ongoing.
He counts Marks & Spencer, Edam, Disney and GlaxoSmithKline among his clients and has worked on an E. coli outbreak in Scotland that killed 21 people and made many more seriously ill following contamination at a butcher.
“The mistakes that were made in the most recent E. coli outbreak were the same mistakes that were made in Lanarkshire 10 years ago and in many other food scares,” says Beattie. “People started speaking out, saying what the cause might be without knowing what actually caused it. Everyone was getting into a panic and there was all sorts of misinformation.
“We stepped in and everything we put out was official; if it couldn’t be proved, it didn’t go out.
“It takes time to identify the cause of an E. coli outbreak and this means there can be an information gap, while the press repeatedly asks what the cause is. Public health officials feel pressured to make statements that they can’t back up with science. The message to the food industry is don’t say anything until you know the facts. It’s not a sin to say you don’t know yet.”

Beattie isn’t completely wrong. As we have repeatedly stressed in this column, early information is likely to be uncertain and will often turn out wrong, and risk communicators need to resist the temptation to sound like they know. He’s right about that. Prematurely confident – or confident-sounding – statements can mislead consumers, damage credibility, and devastate entire industries. Sounding too certain is especially destructive for those on the reassuring side of the issue (as Beattie’s clients are likely to be). But over-confident warnings are also a serious mistake.

But “say nowt”? Can Beattie really mean that nobody should tell the public anything about a risk until the information is certain? Would he endorse that standard for hurricane warnings or infectious disease pandemics? Is he okay with people eating potentially contaminated sprouts because the authorities aren’t sure yet that sprouts are the source of the outbreak?

Can Beattie possibly want us to think that a company like pharmaceutical giant GlaxoSmithKline – identified in the story as one of his clients – has lots of not-quite-certain information about dangers its drugs might pose but is keeping that information secret on Beattie’s recommendation?

That’s the dilemma we started this column with: what to say when you have uncertain risk information to impart. People hate it when you express the uncertainty. They hate it so much that they tend to miss it. Unless you aggressively, unambiguously proclaim your uncertainty, people are likely to imagine that you’re a lot more certain than you are.

Sounding certain isn’t an acceptable solution. And Beattie notwithstanding, neither is saying nowt. Proclaiming uncertainty is the only option.

Perhaps in one sense, Beattie may be right. Throughout the O104:H4 outbreak, produce companies, trade associations, and sometimes governments issued endless knee-jerk denials and assertions of the absolute safety of their products. They might have been better off saying nowt than sounding prematurely certain that they were not to blame. But premature overconfident denials and saying nowt are both bad choices.

The Hamburg Institute started O104:H4 risk communication down a bad track by sounding far too certain that Spanish cucumbers were responsible for the German outbreak. For whatever reasons, federal officials in Germany and authorities elsewhere in Europe did nothing to correct the record for five critical days, allowing the media and the public (and even infectious disease experts) to get the impression that “officials” (not one local agency) had firmly concluded (not tentatively hypothesized) that the cucumbers were the culprit.

Thereafter the epidemiological detective story evolved fairly straightforwardly, if perhaps a bit too slowly. (There was one other red herring – a quickly abandoned but not explicitly retracted claim that an O104:H4-contaminated cucumber had been found in the household of a German victim.) The German outbreak was traced to sprouts from a single German farm. A much smaller French outbreak was traced to a seed company in the U.K. When it emerged that both the German farm and the U.K. seed company had bought fenugreek seeds from the same lot imported from Egypt by a German company, the mystery was mostly solved – though to this date no one knows how the Egyptian seeds became contaminated with E. coli O104:H4 in the first place.

So far, not a single fenugreek seed tested by officials in Europe has been positive for E. coli O104:H4. The epidemiological investigation found its culprit with minimal help from microbiology.

And yet many people who followed the news with casual interest ended up with an indelible impression of endless false leads.

The Hamburg Institute’s overconfident-sounding announcement about Spanish cucumbers turned out mistaken – and the silence of EU and German federal agencies made them complicit in the mistake. This set up those agencies to sound overly tentative as sprouts and then fenugreek seeds moved into the crosshairs. And for a while, perhaps, it also set up the media and the public to interpret the agencies’ tentativeness not as caution but as evidence that they still didn’t know the source of the outbreak. Overconfident denials from the Spanish government and the Spanish cucumber industry turned out correct, while later overconfident denials from seed companies, industry spokespeople, and government officials in various countries didn’t hold up against increasing epidemiological certainty.

And the European public’s confidence in its food safety and public health officials? We’ll have to wait for the next Eurobarometer or other survey.

Here are our four key conclusions from this uncertainty communication case study:

  1. When talking about uncertain risks, you have to do more than acknowledge the uncertainty. You have to proclaim it. This is new knowledge for us, after decades of simply advising clients to acknowledge uncertainty.
  2. You have to do more than proclaim that you are uncertain; you have to proclaim how uncertain you are – where you are on a scale from “I’m taking a shot in the dark here” to “I’m almost certain but there are still a few remaining doubts to clear up.”
  3. And you have to distinguish your level of uncertainty now from your (or other people’s) level of uncertainty earlier. “This isn’t like when the Hamburg Institute thought it might be Spanish cucumbers.” “We have a lot more evidence linking the outbreak to sprouts today than we had just a few days ago.”
  4. Finally, you have to do it all in a way that comes across as human. “We wish we could nail the source of the outbreak with certainty. That’s what consumers want. It’s what the food industry wants. It’s what we want. We may get close – closer than we are today. But total certainty, sadly, is beyond what epidemiology can accomplish.”

On November 24, 2011, Germany’s Federal Institute for Risk Assessment (BfR) emailed us a response to this column.

Copyright © 2011 by Peter M. Sandman and Jody Lanard