In 1993 about a thousand people in Germany fell ill, most of them children. The source turned out to be paprika and paprika-seasoned potato chips, and the snacks carried between 0.04 and 0.45 Salmonella cells per gram. The investigators put the dose people actually swallowed at 4 to 45 cells.
The infectious dose is the number of organisms it takes to make someone ill. Everything else follows from it: whether washing your greens is worth the effort, whether the fridge saves you, whether cutting away the bad bit is enough. Textbooks carried figures in the hundreds of thousands of cells for a long time, and those figures turned out to be too high.
Where the textbook numbers came from
The classic doses were measured on volunteers who swallowed a known culture and were watched to see who got sick. Blaser and Newman put nine such studies and eleven real outbreaks side by side in 1982, because the two sets of numbers did not agree. In six outbreaks out of eleven, the calculated dose was under a thousand organisms.
The authors named the limits of the volunteer work plainly: the strains chosen for it, the same subjects tested more than once, and too few volunteers at the lower doses. The studies measured what it takes for a healthy adult to fall ill from a laboratory strain, which is a different question from what it takes in a kitchen.
A team at RIVM, the Dutch national institute for public health, rebuilt the Salmonella dose-response curve in 2010 using outbreak data instead. The dose that infects half the people who eat it came out at 7 CFU, and the dose that makes half of them ill at 36 CFU. The authors describe this as far higher infectivity and pathogenicity than the feeding studies suggested.
E. coli O157: under ten cells
Nobody has ever fed O157:H7 to volunteers and nobody will: the disease is too severe, and Strachan and colleagues say so directly in their 2005 modeling paper. Every number for this organism is reconstructed from outbreak investigations.
The best known reconstruction is Tuttle and colleagues at the CDC, published in 1999. The outbreak ran through the western United States in the winter of 1992 to 1993 and was traced to ground beef patties at one fast-food chain. The team counted bacteria in 76 samples of recalled patties: a median of 1.5 cells per gram, which works out at 67.5 cells per patty. That is what led them to conclude that the infectious dose for O157:H7 is fewer than 700 organisms.
The Japanese estimate is lower still. Hara-Kudo and Takatori went through 17 outbreaks from 2004 to 2006 and, for one O157 outbreak, arrived at a dose of 2 to fewer than 9 CFU. Everyone who ate was infected.
Salmonella: a range spanning millions
In the same Japanese work, Salmonella produced the widest spread, and the spread shows how dose and the severity of an outbreak travel together.
| Organism in the outbreak | Estimated dose | Share infected among those who ate |
|---|---|---|
| E. coli O157 | 2 to fewer than 9 CFU | 100% |
| Salmonella, one group of outbreaks | 81-1560 CFU or MPN | 10-66.4% |
| Salmonella, three outbreaks | 259,000 to 14,000,000,000 CFU | 89-100% |
| Enterotoxigenic E. coli | 25-1000 CFU | 93% |
| Campylobacter | 360 MPN | 37.5% |
The German paprika outbreak of 1993 sits at the bottom of that scale. Lehmacher and colleagues estimated the dose at 4 to 45 cells, with one person falling ill for every 10,000 who ate the snack. Their conclusion: even extremely low numbers of Salmonella, adapted to a dry product, were able to cause illness.
Norovirus and Cyclospora
Norovirus holds the record. The same RIVM group had estimated two years earlier that the probability of infection from a single virus particle is close to 0.5, higher than for any other virus studied. Infection does not always mean illness: among infected subjects the probability of becoming ill rose from 0.1 at a dose of a thousand genomes to 0.7 at a hundred million.
For Cyclospora the honest answer is different. In a 2019 update, authors from the FDA and the US Department of Agriculture list what is still unknown about the parasite, and the infective dose is the first item on that list. The figures that circulate online have no primary source behind them. What that means for washing is covered in our piece on the Michigan cyclosporiasis outbreak.
What this means for the salad on your plate
Danyluk and Schaffner built a risk model for leafy greens in 2011 and tested it against the 2006 E. coli O157:H7 spinach outbreak. To reproduce an outbreak of that size, the model needed a field level of 0.1 CFU per gram and 0.1 percent of servings contaminated. In a contaminated serving that is about ten cells per 100 grams.
Ten cells in a bag will not show up in sampling and will not come off in the sink. An industrial wash tank takes off 1 to 2 log, a factor of ten to a hundred, and against a dose like that the reduction decides nothing. Why the tank is not there for that job at all is covered in our guide to wash water sanitizers, and why home washing does not close the gap either is in our piece on biofilms.
The other half of the same model is more useful than any washing advice: cut greens held at abusive temperatures support growth of up to 1 log per day. Ten cells become a hundred in a day in a warm kitchen and a thousand in two. That is why the fridge matters more than the faucet, and why a bag of salad forgotten in a warm car is worth a second thought.
An honest correction: dose is a probability, not a verdict
In the German paprika outbreak, one person in ten thousand fell ill. The other 9,999 ate the same chips and noticed nothing. A low infectious dose means the risk stops being zero. It does not make illness certain.
The second factor is your stomach. Smith, at the USDA Agricultural Research Service, made the point in 2003: gastric fluid at pH 1 to 2 is deleterious to many pathogens, while antacids and drugs that suppress acid secretion remove that protection and raise the risk of food- and waterborne illness. E. coli and Salmonella can also adapt to acid and survive the passage.
So the same plate of salad carries a different risk for different people. Pregnancy, age over 65, heartburn medication and a weakened immune system are on a separate line in every set of guidelines for exactly this reason. Unpasteurized juice sits on those lists for the same reason: in the 1996 apple juice outbreak the median age of the patients was five years, and the child who died was 16 months old. That case is covered in our piece on E. coli in apple juice.
Key points
- Textbook doses were measured on volunteers and laboratory strains. Reconstructions from real outbreaks come out orders of magnitude lower.
- For O157:H7 the estimates are fewer than 700 cells in the 1992-93 ground beef outbreak and 2 to fewer than 9 CFU in a Japanese outbreak where everyone who ate was infected.
- Salmonella in a dry product caused an outbreak at a dose of 4 to 45 cells. Modeled on outbreak data, the dose that infects half the people who eat it is 7 CFU.
- For Cyclospora the infective dose is unknown, and the FDA review says so.
- Reproducing the 2006 spinach outbreak took a modeled 0.1 CFU per gram, roughly ten cells in a serving. A wash removing 1 to 2 log does nothing against that.
- Cut greens in a warm room add up to 1 log per day. Cold holds the dose down; rinsing does not.
- A low dose raises the probability rather than guaranteeing illness. Heartburn medication, pregnancy, age and a weakened immune system raise it further.
How those cells reach your greens once they are in the kitchen is covered in our piece on kitchen cross-contamination. Why a few cells on a sprout seed turn into millions is in our comparison of sprouts and microgreens. What the words on a bag of washed salad actually cover is in our look at triple-washed claims. Why a freezer pauses bacteria rather than killing them is covered in our piece on refreezing. How many cells cross onto food dropped on the floor is covered in our look at the five-second rule. When illness comes from toxin already made in the food and cell counts matter less, see our piece on reheating rice. The big picture is in our food safety hub.
