Source: Saturday Morning Breakfast Cereal, Artist: Zach Weinersmith


A reckoning is coming.

Unless you’ve been living under a food safety rock, everyone’s talking about the Romaine lettuce outbreak that, if we’re defining outbreaks by WGS signature, kinda never actually ended from last year.

“Genetic analysis of the E. coli O157:H7 strains tested to date from patients in this current outbreak are similar to strains of E. coli O157:H7 associated with a previous outbreak from the Fall of 2017 that also affected consumers in both Canada and the U.S. The 2017 outbreak of E. coliO157:H7 was associated with leafy greens in the U.S. and romaine in Canada. This year, romaine lettuce is the suspected vehicle for both the U.S. and Canadian outbreaks. “

The reactions have been varied. There’s been some industry defensiveness, some public outcry, vague politician outcry, and a lot of food safety chatter filled with good ideas, intentions, and blame. Bill Marler put out the question, and I thought it was worth spending a little more time to think about the answer.

“How does this “Romaine” acceptable?  What will growers, processors, shippers, grocery stores, restaurants, consumers, regulators, and politicians do?”

Putting those illnesses in the perspective of the victims, what Bill does, is how we should think about food safety in industry as the distributors of the food supply. But, when we want to actually discuss; not posture, not politicize, but seriously discuss solving a public health problem, we need to re-frame those illnesses.

Consider some data:

Romaine harvested in 2017: 3,075,000,000 lbs (harvested, not necessarily sold).

Total confirmed sick from romaine in 2017-2018: 339.

Romaine lettuce harvested before a single (detectable) illness was recorded based on those numbers: 9,070,796 lbs (somewhere in the neighborhood of 250-300 harvested acres).

So, based on the alarming outbreak/illness numbers above, we detected one illness for every 9 billion lbs of romaine harvested.

If there’s been a recurring theme on this blog, it’s that food safety is a public health risk, not a personal risk. The danger only becomes visible when you extrapolate that risk to millions of meals or people, or you’re one of the people who won the diarrhea lottery. This concept is playing out in the detection and publication of multi-state outbreaks, detected via whole genome sequencing, increasing with the number of Genomes available to compare.




This isn’t a perfect cause and effect, things are more complicated than that. But this theory doesn’t just come from me, CDC agrees:

“because many foods are distributed widely and rapidly, and because detection methods have improved, outbreaks that occur in multiple states and that even span the entire country are being recognized with increasing frequency.”

The rate of foodborne illness hasn’t been increasing. WGS is revealing what hazard analysts have known for a while. Raw products without a valid intervention such as cooking, pasteurization, irradiation, or some other significant bacterial reduction will have some potential presence of environmental pathogens. This means that there is always going to be a minuscule potential for illness in these products, and that potential is so low, it may only be visible with the huge amount of data now available through WGS.

Consider a 6-log or 99.9999% reduction. A million bacteria aren’t hard to come by. Grab 1/4 teaspoon of dirt or surface water. Boom, reduction overcome. Now, in 3 billion lbs of romaine, how much dirt do you think is on those leaves? How many lbs of lettuce has enough that bacteria survive the 6-log reduction in real numbers, even if the percentages are tiny? We talked about this before when we considered how statistical detection rates make end item testing useless.

This stuff is more complicated than I’m presenting it here, but do you see my point? David Hand calls it “The Improbability Principle“.  Basically, if you have a large enough data set, even if something is rare or unlikely, it will still happen frequently if there are enough opportunities.

You will not win the lottery this month, but someone is very likely to win.

By all means, if there is a specific farm or region who’s negligent practices led to dangerous foods, let’s sick the government and MarlerClark on them. That’s partially the approach we’re taking by using traceability pushes to find smoking guns. But when it’s no longer traced to a disgusting production facility that still hasn’t figured out how to clean things…EVER, we’ll arrive at a point where we need to determine how many illnesses are “acceptable” for these types of products, and communicate the risk to the public.

Huitres Cancale
This is the approach we’ve taken with Oysters. Despite the US eating way fewer raw oysters than heads of lettuce, about 80,000 people get vibriosis – and 100 people die from it – in the United States every year.  Oysters are considered a food where there is an inherent and natural risk of Vibrio when consumed raw. I’ve spoken with representatives from WSDA about some of the oyster and raw milk outbreaks they dealt with in 2017, and I wanted to know what the preventive actions for those types of products were after an outbreak.

Their answer was basically the manufacturer/restaurant has to clean, throw away traceback products, then it’s back to business as usual. The root cause wasn’t in the factories or kitchens (in general), but in an entire state selling these products some of them will be dangerous. Rather than the servers fault, it’s the product.

WSDA agrees, root cause is the product and it’s “raw” status.

While the incidence varies depending on food categories, we may find that for grains, flour, lettuce, or fruit the risk of illness may be no different than communicating the risk of oysters, non-irradiated sprouts, or other “raw” products. There will inevitably be safer versions available, either those that are cooked or receive a microbial reduction step (irradiation, some sort of validated wash, “processing”), or grown in a more sterile environment (e.g. hydroponics). Provide a lower risk option for high-risk populations, share data to help people understand the risks of consuming raw products, and set a realistic public health standard of “same or better”.

We live on a bacterial planet. Visibility of these outbreaks does not change the risk we have always faced, but it will give us opportunity to reduce those risks. We’re not anywhere near a stop point yet, we have a lot of opportunties to address with contaminated irrigation systems, unsanitary distribution networks, food animal vectors, or production facilities that still haven’t figured out how to clean things (or that mice are bad). But I’m going to bet now that this is going to dominate our food safety policy discussions for the foreseeable future.

Eventually, there’s going to be a diminishing return on the resources we invest to get just one more log reduction, or it’s going to become prohibitively expensive and someone is going to ask why we aren’t working just as hard on what’s killing most of us.  I think that turning point will be when we start revealing all the illnesses that can be traced back to small establishments (restaurants and cottage foods) that are hiding statistically from the outbreak data. Once we can demonstrate that ma and pop have similar contamination rates when measured on a population scale, that’s when we’ll set an “acceptable” illness rate.


Besser, J. 2017.Whole Genome Sequencing (WGS) at CDC (PulseNet).

Crowe, Samuel J., et al. “Vital signs: multistate foodborne outbreaks-United States, 2010-2014.” MMWR Morb Mortal Wkly Rep 64.43 (2015): 1221-5.

National Agricultural Statistics Service .2018.Vegetables Annual Summary, 02.13.2018

Nguyen, Von D., et al. “Increase in multistate foodborne disease outbreaks—United States, 1973–2010.” Foodborne pathogens and disease 12.11 (2015): 867-872.