During phase 2, a total of 17 confirmed case-patients were interviewed with hypothesis-generating questionnaires about items eaten in the 2 weeks before symptom onset. The instrument included questions about restaurant history with meal details; grocery stores frequented; and yes/no questions about >70 fruits and vegetables, 8 herbs, and 16 mixed foods (e.g., salsa, pesto) previously implicated in outbreaks of foodborne disease. No common restaurants or events were identified.

Frequently reported foods were compared with population controls from Canadian (Waterloo, Ontario) and American (Oregon; US Foodborne Diseases Active Surveillance Network [FoodNet]) published food consumption surveys (3–5). Although such measurements may be limited by the timing of questionnaire administration and the recall period considered, they can be useful comparators during the hypothesis-generating stages of an investigation. By the end of phase 2, strawberries, cilantro, and sweet basil were reported more often than expected by case-patients (Table 2). Garlic and red peppers also were commonly eaten by case-patients; however, population comparisons were unavailable. Eighty-eight percent of case-patients reported having eaten romaine lettuce; 85% of controls in the Waterloo survey (4,5) had eaten lettuce of any type, and romaine lettuce consumption was much less commonly reported in the FoodNet survey (3) (Table 2). Other foods assessed were not eaten more often than expected. A formal case–control study was considered premature in the early stages of phase 2 because no strong hypothesis emerged from early interviews and comparisons to population controls. We further explored the plausibility of various hypotheses through a combination of methods described below that allowed room for additional hypotheses to emerge or existing hypotheses to strengthen as cases accrued.

Detailed questionnaires asked whether foods were eaten in a restaurant or were store-bought and about type of packaging and method of preparation (because C. cayetanensis is heat-sensitive) (6). We reinterviewed early case-patients using the second questionnaire and interviewed later case-patients using both questionnaires. In phase 1, garlic eaten at restaurants by all 4 persons with confirmed infections were traced back to different suppliers; only 1 case-patient ate raw garlic in a restaurant. Three case-patients also reported eating cooked garlic at home; cooking would have inactivated the pathogen.

Early and proactive collaboration with CFIA involved a general assessment of the country of origin and distribution patterns for frequently eaten foods. According to CFIA records, romaine lettuce and red peppers sold during the exposure period were not imported from a known Cyclospora-endemic country and were widely distributed in Canada and the United States. This was not consistent with case distribution (R. Cardinal, CFIA, pers. comm.).

Because interviews, population control comparisons, and product distribution limited suspected foods to strawberries, cilantro, and basil, we began preliminary traceback of all 3 suspected items. Environmental health officers and regional CFIA staff interviewed grocery store owners, restaurant managers, and distributors to trace produce to its supplier. Local strawberries eaten by case-patients from 3 small markets were traced back to 2 local farms in geographically separate regions of British Columbia. Cilantro eaten by case-patients was traced to 2 suppliers; both supplied home-grown rather than imported produce. Of 14 case-patients with confirmed basil exposures, 4 (57%) ate only organic basil supplied by distributor A. Additionally, 4 (29%) reported multiple basil exposures, including exposure to organic basil from distributor A (Figure 2). In British Columbia, organic basil enjoys a smaller market share than the conventional product.

In phase 2, 12 (71%) of 17 case-patients reported shopping at grocery C. Records of any grocery store purchases for the households of 8 consenting case-patients were obtained through grocery C’s savings card program; other case-patients were not cardholders. All purchase histories were requested for 1 month before symptom onset to account for the typical incubation period plus product shelf life. Records from 3 (38%) case-patients showed purchases of the same organic basil supplied by distributor A. Two case-patients had bought organic basil on the same day at the same location. Of the remaining 5 case-patients who recalled purchasing organic basil but whose consumer card records did not confirm it, 2 had not used their cards for large portions of the incubation period.

We collected supplier information for organic basil during a visit to the distribution warehouse and local farm site of distributor A. The remaining 2 (14%) case-patients with basil exposure previously unlinked to distributor A were confirmed through trace-forward from distributor A. The first had eaten organic basil at a smaller market supplied by distributor A under another trade name. The second had eaten conventional basil from a grocery store supplied by distributor A. Distributor A confirmed using organic basil to supplement conventional basil shipments when supply was low. Late summer outbreaks of cyclosporiasis in British Columbia are unusual; distributor A confirmed that imported product was used throughout the summer in 2007 because of a poor local growing season.