BOSTON — A 2009 outbreak of illness from exposure to E. coli was believed to have been caused by flour in cookie dough, but no definitive evidence emerged. A 2016 E. coli outbreak was conclusively attributed to flour but only after investigators overcame numerous challenges.

A research paper describing the 2016 outbreak in great detail was published Nov. 23 in the New England Journal of Medicine. “Shiga Toxin-Producing E. coli Infections Associated with Flour” was written by a team of more than 20 scientists led by Samuel J. Crowe from the Centers for Disease Control and Prevention. The paper represents a significant expansion of a C.D.C. September 2016 statement about the episode.

The outbreak of E. coli O121 sickened 55 and prompted numerous product recalls (the C.D.C. originally connected the contaminated flour to 63 illnesses). The findings underscore the importance of consumers following instructions not to eat raw cookie dough or batter.

The researchers acknowledge that flour is not a “usual suspect” as an outbreak vehicle since its low-water content is not conducive to bacterial growth.

“Nevertheless, pathogenic microorganisms on the wheat or other ingredients in flour can survive the drying process and remain viable in flour for months,” the researchers said.

Shiga toxin-producing E. coli (STEC), including types O121 and O26, which causes more than a quarter of a million illnesses annually in the United States, has been identified as a pathogen able to survive in flour.

“Symptoms appear three to four days after infections and include mild fever, abdominal pain, vomiting and diarrhea, which is often bloody,” the researchers wrote. “The hemolytic–uremic syndrome, a form of kidney failure, also develops in some patients with STEC infection.”

In February 2016, the C.D.C. and state health departments began investigating a cluster of patients who were infected with STEC O121.

“All the patients were infected with a strain of STEC O121 that had the same uncommon pulsed-field gel electrophoresis pattern combination, which suggested a common source of illness,” the researchers said.

A total of 56 cases were identified in 24 states of which 55 were infections with STEC O121. Sixteen patients were hospitalized. Among these one patient, an adolescent girl, experienced hemolytic–uremic syndrome but recovered. No deaths were reported.

The C.D.C., together with state health departments, in February 2016 began investigating the cluster of patients infected with STEC O121. Consistent with normal investigative protocol, a seven-day food-consumption survey was collected from patients, but responses did not lead investigators to a strong hypothesis regarding the outbreak source, the researchers said. Several of the patients then participated in open-ended interviews by a C.D.C. epidemiologist, and responses prompted the development of a second questionnaire.

“Consumption of raw or undercooked flour is not included on most routine state and national foodborne disease questionnaires, so epidemiologists were not initially able to assess whether case patients had consumed raw flour,” the researchers wrote.

Leafy green vegetables initially were suspected, having been reported by numerous early case patients. Ultimately, though, those sickened had not eaten leafy green vegetables any more than the overall population.

All 10 patients participating in the open-ended interviews said that they baked frequently or regularly consumed home-baked foods. Five of the patients recalled baking the week before the onset of illness, and three others thought they may have baked during that period.

“Of the five case patients who remembered baking, four reported eating or tasting homemade batter or dough, three of whom used brand A flour,” the researchers said. “The fourth used either brand A or another brand.”

When interviewed, two patients still had the bags of flour they had used in the week before the illness onset.

Shortly afterward state investigators recognized that three sickened children had been exposed to raw flour at restaurants in Maryland, Virginia and Texas.

“Restaurant staff had given them raw dough to play with while they waited for their food to be served,” the researchers said.

The second questionnaire included questions about baking, flour and raw dough exposures together with other possible food exposures.

Responses included:

22 of 26 ate baked or made homemade cookies, muffins, pancakes, cakes or other foods containing flour,

19 of 30 used brand A flour to make something homemade or from scratch,

17 of 30 ate, tasted or licked uncooked or unbaked homemade dough or batter,

15 of 24 ate chocolate chips or chunks by themselves or in homemade foods and

23 of 29 ate peanut butter.

“Some case patients did not report exposure to flour in the week before illness onset, but this is not uncommon in outbreak investigations,” the researchers said. “Interviews often occur weeks to months after the illness, which makes it difficult for the patient to recall exposures accurately.”

Meanwhile, trace-back investigations of the bags of brand A flour collected from patients included a bag of unbleached all-purpose flour milled Nov. 14, 2015, and a bag of bleached all-purpose flour milled Nov. 15, 2015.

“The two bags were produced in the same facility,” the researchers said. “The flour that was used in the raw dough given to the children exposed in the Maryland, Virginia and Texas restaurants also was from this facility, as was flour from three additional bags collected from case patients residing in Arizona, California and Oklahoma.”

Still, initial testing of the flour collected from the homes of case patients did not identify STEC O121. With additional screening, STEC O121 was identified, and the laboratory testing protocol was modified. Additional recovered bags also showed STEC O121.

No source of contamination in the mill was identified by the Food and Drug Administration, the researchers said, suggesting “the ingredients might have been contaminated further back in the production chain.” The milling company involved in the episode isolated STEC from flour produced at that facility and shared the isolates with the F.D.A., the researchers noted.

In addition to the three recalls by company A, additional recalls were issued by companies that had used recalled flour in their products.

“In total, nearly 250 products containing flour were recalled,” the researchers said.

Throughout the paper, no specific companies or brands are identified. General Mills, Inc. conducted recalls of its family flour during this period in connection with the E. coli outbreaks.  By the end of its third recall, the company had recalled 45 million lbs of flour.

While the flour came from a single mill, the trace-back investigation was unable to definitively determine whether the contaminated flour “shared a common source of wheat,” the researchers said.

“Wheat from several states was used to produce the flour, and grains from different fields are frequently commingled shortly after harvest and further mixed during transport and milling,” the researchers said.

While the source of contamination was not identified, the researchers have theories.

“On the basis of what is known about the ingredients of flour, wheat is the ingredient most likely to be contaminated, perhaps in the field before harvest,” the researchers said. “Some farmers use manure from cattle, a reservoir of STEC, to fertilize their wheat fields, which could lead to contamination of the wheat if the cattle are colonized. Another source might be white-tailed deer, which are ubiquitous in the United States and are also reservoirs for STEC. Given that a specific wheat field was not implicated in this investigation, we could not evaluate whether animal intrusion was a source of contamination.”

While recognizing that the source of the illness was commercially distributed contaminated flour, the researchers said other factors should be recognized as well in connection with the outbreak.

“The behaviors of both consumers and retailers increased the risk of illnesses resulting from the contaminated flour,” the researchers noted. “These behaviors included the consumption of raw or undercooked homemade dough or batter, which has long been discouraged because of the known risk of salmonellosis from consuming raw eggs, as well as allowing children to play with raw dough in restaurants and using flour to make play-dough for children at home. Our data show that although it is a low-moisture food, raw flour can be a vehicle for foodborne pathogens.”