Study participants included 786 febrile patients at the healthcare center for the villages of Dielmo and Ndiop; 78% of the patients were <15 years of age, and the sex ratio was 1:1. For all patients with fever (defined as axillary temperature of >37.5°C), we conducted a medical examination, completed a questionnaire, and collected a whole-blood finger-prick sample (200-?L [4 drops]) (8). In parallel, we collected blood samples from a control group of 385 healthy, afebrile villagers; 62.5% of these study participants were <15 years of age, and the sex ratio was 1:1.

We performed genotyping of T. whipplei as previously described (32). We attempted to amplify and sequence each of 4 multispacer sequences (TW133, ProS, SecA, and Pro184) from positive specimens. When sequences were obtained, we compared them with those available in the GenBank database and our internal laboratory database to determine their corresponding genotype.

We performed statistical analyses by using Epi Info 6 software (http://www.cdc.gov/epiinfo/index.html); results with p<0.05 were considered statistically significant. The corrected ?2 test or the Fisher exact test was used where indicated.

A total of 786 febrile patients and 385 healthy controls were included in the study, among whom 36 (4.6%) and 1 (0.25%), respectively, were positive for T. whipplei DNA (p<0.00007). The positive control participant was a 13-year-old boy who had low concentrations of T. whipplei DNA (Ct of 36.85 and 37.99). The Ct for febrile patients ranged from 26.10 to 36.41 (mean ISD 33.40 ± 2.53).

The prevalence of T. whipplei bacteremia was 4% (3/75) for febrile patients <12 months of age, 4.8% (12/250) for those 1–3 years of age, 4.2% (5/119) for those 4–6 years of age, 5.4% (9/167) for those 7–15 years of age, 2.7% (2/75) for those 16–29 years of age, and 5.2% (5/97) for those >30 years of age. Age data were not available for 3 patients. No significant differences in age distribution were observed.

Clinical data were available for 786 febrile patients (Table 1). The main symptoms in the 36 T. whipplei–positive febrile patients were headache (23 [68.9%]), cough (13 [36.1%]), rhinorrhea (8 [22.2%]), nausea (5 [13.9%]), vomiting (4 [11.1%]), and diarrhea (3 [8.3%]). No significant clinical differences were observed by Ct level.

All 36 T. whipplei cases detected among the 786 febrile patients were in the 466 patients tested during the June–October rainy season; no cases were detected among the 320 febrile patients sampled during the November–May dry season (p = 0.0000001). Moreover, 33 (92%) of these 36 cases were diagnosed during the 2010 rainy season, and the other 3 were diagnosed during August 2011 (2 cases) and October 2011 (1 case) (Figure). The highest prevalence of T. whipplei bacteremia cases was detected during August, when 28 (30%) of 93 febrile patients were found to be positive (19 [28%] of 73 patients in Dielmo and 9 [45%] of 20 patients in Ndiop). In fact, the data were affected by the high prevalence of cases observed in August 2010, which seemed to be indicative of an outbreak.

In July 2010, T. whipplei infection was detected in 2 febrile patients, an 18-year-old boy in Dielmo (case detected July 24) and a 15-year-old girl in Ndiop (case detected July 27). In August 2010, a total of 29 febrile patients from Dielmo were tested; 17 (58.5%) of the 29 patients had samples (18 total samples) positive for T. whippleibacteremia. During the same month in Ndiop, 9 (69%) of 13 febrile patients had positive samples. In September 2010, 2 patients were positive in Dielmo and 1 in Ndiop, and in October, 2 patients were positive in Dielmo and none in Ndiop. For almost 1 year, all specimens from febrile patients were negative for T. whipplei. Then, in August 2011, only 2 patients were positive in Dielmo, and in October 2011, only 1 patient was positive in Ndiop.

Data about antimicrobial drug therapy was available for 33 patients, 23 of whom benefited from treatment with amoxicillin (18 patients), metronidazole (3 patients), or cotrimoxazole (2 patients). In Dielmo, 24 specimens from 23 patients were positive for T. whipplei; 1 patient was sampled twice 15 days apart, and both specimens were positive. For 17 patients, blood specimens were also sampled during other febrile episodes. Nine specimens from 5 patients were sampled from 15 days to 13 months before the positive sample was detected, and 43 specimens from 17 patients were sampled from 3 weeks to 16 months after the positive sample was detected; all of these samples were negative. Moreover, our previously published data (8) included test results for a 4-year-old boy who was diagnosed with T. whipplei bacteremia in January 2009 (19 months before August 2010). Four other blood specimens from this patient were tested 1 month before (1 sample) or 4, 11, and 15 months after (3 samples) the positive specimen was detected, and all were negative for T. whipplei.

In Ndiop, 12 specimens from 12 patients were positive. For 8 of these patients, blood specimens were sampled during other febrile episodes. The specimen for 1 patient was sampled 1 month before the positive sample, and 9 specimens from 6 patients were sampled from 7 weeks to 18 months after the positive samples; all of these specimens were negative. No data were available for these patients about antibody response against T. whipplei.

Because of the lack of specimens available for genotyping and the low sensitivity of genotyping, we could obtain multispacer sequences for only 8 patients at the time of the 2010 peak in T. whipplei bacteremia cases (Table 2). The T. whipplei genotype corresponds to the concatenation of the 4 spacers (TW133-ProS-SecA-Pro184); however, TW133 sequencing was not successful, so the corresponding spacer was not available (NA) for any of the patients. ProS sequence was obtained for 5 patients, SecA for 6 patients, and Pro184 for all patients. For 4 patients, 3 spacers were available, enabling the detection of the same multispacer sequence combination (NA-7-2-1) for the 4 patients. For another 4 patients, 2 spacers were available, enabling the detection of the NA-7-NA-1 combination for 2 of the patients and the NA-NA-2-1 combination for the other 2 patients. None of the potential combinations has previously been sequenced in Senegal. Moreover, the NA-7-2-1 combination has also not previously been detected in any other area of the world and is thus a new genotype. Overall, our data suggest that the same genotype was detected in Dielmo and Ndiop during the summer of 2010. However, T. whipplei genotyping was performed (sometimes only partially) for only 8 of 36 patients, so we can only suspect, but not confirm, that an epidemic clone was present and that an outbreak was ongoing at that time.

In Dielmo during the peak of the August 2010 outbreak, multiple persons in several households were positive for T. whipplei bacteremia: 4 of 6 persons in household no. 19, 3 of 4 persons in household no. 39, 2 of 2 persons in household no. 9, and 2 of 3 persons in household no. 14. In Ndiop, 2 of 2 persons in household no. 3 and 2 of 3 persons in household no. 8 were positive for T. whipplei bacteremia. Of note, during this time, the family in household no. 39 had a furnace in which they baked bread that they marketed locally. In December 2010, most of the family left the village and the furnace was shut down; no additional T. whipplei bacteremia cases were subsequently observed.