In 1854, as a cholera epidemic killed hundreds in London, an English physician named John Snow was determined to find out how the disease was transmitted.


Snow doubted the prevailing belief that disease was spread by breathing “bad air.” He noted that the disease was centered near a public water pump on Broad Street. The water in that area, he argued, was polluted by sewage, while people in other neighborhoods who drank from unpolluted wells, were unaffected.

When the handle was removed from the suspect pump, the epidemic ebbed. And Snow had demonstrated that cholera is spread not by air, but by food or water.

Snow’s work came as Louis Pasteur and other pioneers were beginning to probe the microbial world of bacteria. Together, they helped establish the new science of epidemiology, the study of disease and how it is transmitted.

Today, epidemiology consists of two fundamental processes. One is in the lab, where scientists chase microbes using microscopes and, more recently, high-tech genetic analysis.

The other, perhaps more prosaic process is fieldwork, variations on the case-control study pioneered by John Snow. This consists of observing or interviewing sick people, noting what they have eaten or experienced, and comparing those results with similar people who are not sick.  If sick people have consumed food or water that healthy people have not, one can establish a statistical probability for the cause of the outbreak.  

Epidemiologists have used case-control studies for more than a century, tracing the sources of countless epidemics of food poisoning and other illnesses. Perhaps the best-known examples were studies in Britain and the U.S. in the 1950s that showed that tobacco smoking was by far the most common cause of lung cancer.

Yet, after 150 years of experience, not everyone is impressed with the validity of the case-control study.  Witness California advocates of unpasteurized milk who contested epidemiology that pointed to Organic Pastures raw milk as the probable source of a recent outbreak of E.coli O157:H7. While interviews with all five young victims pointed to raw milk, investigators didn’t find contaminated milk — not surprising given the weeks-long time lag between when someone becomes infected, and when tests are conducted.

Sometimes a contaminated product isn’t available to be tested following an outbreak of illness.

Until environmental tests of the dairy came back positive for the outbreak strain of E. coli O157:H7, raw milk advocates had argued that health officials had not made their case, despite the overwhelming epidemiologic evidence.

Last year, executives at Del Monte Fresh Produce, a major distributor of fruits and vegetables, took that inconclusive-proof argument to great lengths, challenging health authorities who contend that at least 20 people in 10 states were sickened with Salmonella poisoning attributed to cantaloupe imported by Del Monte and sold at Costco warehouses.

They believe they were falsely accused of responsibility, and that the resulting recall damaged the company unjustly. “It’s got to be a comprehensive and reliable investigation, and in our opinion this was neither” Del Monte executive Dennis Christou told the New York Times.

This claim is based on lab tests that showed no Salmonella – no smoking gun.

Del Monte’s aggressive response drew cheers from some in the food industry, frustrated by federal regulators who insist on recalls. Food recalls, justified or not, cost companies millions of dollars – not to mention the associated bad publicity. 

The dispute goes back nearly a year ago, when state and federal health officials detected a spike in cases of Salmonella Panama, a rather rare bacterium that causes severe diarrhea.

Dr. Bill Keene, a nationally known and respected epidemiologist in Oregon’s health department, was one of the scientists who worked on the case. Detailed interviews with victims showed that all or most had eaten cantaloupe.  Investigators then tracked the suspect melons to a Guatemala farm that supplies Del Monte Fresh Produce.

Cantaloupes from the same farm were taken to a lab for pulsed field gel electrophoresis (PFGE), a process that identifies a genotype, or genetic fingerprint. If stool samples from sick people show the same genetic fingerprint, scientists can establish with near certainty that they were sickened by the same food.

The cantaloupe, however, tested negative for Salmonella. And that is the basis for Del Monte’s “not guilty” plea.

Not so fast, say the epidemiologists.The case-control studies still pointed to Del Monte. PFGE analysis is a powerful scientific tool, but so is the more traditional case-control investigation pioneered by John Snow.

“It would be great if we could just buy the product, take it to the lab and find Salmonella,” Keene explained in an interview last spring. “That’s something anybody can understand.  But when you offer up P values and probabilities, people want to say: ‘That’s statistical mumbo jumbo.’ “

It often takes weeks for outbreaks of food poisoning to be detected by health authorities. By that time, the offending food — especially if it is a perishable product — is likely to have been consumed, discarded or no longer on store shelves. So, as often as not, the tainted food is simply no longer available to be tested. And that was the case with the Del Monte outbreak.

However, the lab work can show that these sick people were stricken by the same genetic strain of Salmonella, so it can be assumed they were sickened by the same food product.

Sick people, or “case patients,” are carefully questioned about what they have eaten over previous days and weeks. If all or most of your cases ate cantaloupe, epidemiologists are onto something.  Even if one or two of your victims doesn’t remember eating cantaloupe, they may have eaten something that was prepared on the same surface and contaminated by tainted melon.

The next step is to establish the background rate. What is the likelihood that people will have eaten cantaloupe?  If, for example, the background rate is 30 percent, epidemiologists can calculate the statistical probability that the outbreak was caused by tainted cantaloupe.  That doesn’t make it a certainty, but the probability is in the same statistical range as a positive lab test.

“Everybody loves lab results,” Keene said last spring. “Me too. Nobody sends more stuff to the lab than I do.”

But, from a scientific and legal standpoint, a case-control study is just as powerful, he says.

Keene’s work on the cantaloupe outbreak was “a high quality study,” says Dr. Mike Osterholm, nationally known epidemiologist and director of the Center for Infectious Disease Research and Policy (CIDRAP).

“Case-control is a time-tested approach, an extremely effective method for identifying a product.”

While he understands Del Monte’s frustration, the company “clearly lacks the sophistication to understand the epidemiology here,” Osterholm says. “For them to question this science is like saying you can’t convict a murderer unless you have video of the crime being committed.”

Epidemiologists are not infallible. Osterholm painfully recalls the 2008 Salmonella epidemic that was initially blamed on tomatoes. Investigators later shifted the focus to jalapeno peppers — perhaps peppers used in tomato salsa — but by that time tomato farmers were economically clobbered.

“The FDA can be wrong, and the shattered tomato industry had no place to go to get back its reputation, or its financial losses,” food industry blogger Jim Prevor wrote in recalling that mistake.


But high-tech lab work can lead officials astray as well.  Osterholm points out that Germany’s frightening outbreak of E. coli O104 last spring was initially traced to cucumbers that tested positive for  pathogenic E. coli.  But further tests showed the E. coli was not the outbreak strain; and it was the case-control studies that eventually led the investigators to the real culprit – sprouts grown from contaminated seeds.

The science that led investigators to Del Monte cantaloupe is solid, Osterholm insists.  “Epidemiology is NOT on trial in this case.  To say that would give too much weight to what’s going on here.”

Those who don’t want to believe epidemiology, however, remain deeply doubtful of case-control studies.

So did English decision makers 150 years ago when, after London’s cholera epidemic subsided, the handle was replaced on the Broad Street pump. Nobody will ever know how many innocent Londoners were sickened because people chose to ignore John Snow’s case-control study.