Relying on previous spatial analysis of BU incidence in Akonolinga District, we analyzed a series of cases that occurred in the highest BU-risk area of the district, located along the Nyong River upstream of Akonolinga (3). This area includes 24,469 inhabitants of the town of Akonolinga and 24 surrounding villages. We analyzed 562 new cases of BU that originated in this area from January 2002 through May 2012, after aggregation by month of diagnosis. Biological confirmation was obtained from the National Reference Centre for Mycobacteria for 354 (63%) cases. The BU incidence rate remained stable over the 10-year period at 2.2 cases/1,000 person-years.

Median BU incidence peaked in March, and a second peak occurred in September (Technical Appendix Figure 2), but monthly medians did not differ significantly (Kruskal-Wallis test, p = 0.149). Given the specificities (nonstationarity) of the BU case series, wavelet analysis was the appropriate method for analysis (Technical Appendix). A 1-year periodic signal was identified in the BU-case time series from 2005 to 2011, and this periodicity was statistically significant from mid-2005 to the beginning of 2009 (Figure 1).

Next, we analyzed the links between BU and seasonal changes by using wavelet association and phase analyses between BU case incidence and total monthly rainfall (in mm) or mean Nyong River flow (in cubic meters per second). Strong seasonality was found in the series of monthly total rainfall and of monthly mean Nyong River flow; a 1-year period and a weaker 6-month period corresponded to the 2 rainy seasons separated by a period of lesser rainfall (the small dry season, mid-July to mid-August) (Figure 1; Technical Appendix Figure 3). Because of its shape, the wavelet detected yearly rainfall oscillations between a minimum in December (dry season) and a maximum in July (the middle of the rainy period) instead of the maximal rainfall months of October and November (Technical Appendix Figure 3).

We assessed the association of the incident case signal with environmental variables (Technical Appendix). The 1-year periodic signal of the BU case series was associated with Nyong River flow from the end of 2005 to the end of 2009 (Figure 1) and with rainfall from the end of 2005 to the beginning of 2011 (Technical Appendix Figure 3). Under the assumption that changes in the environment preceded changes in BU incidence, phase analysis indicated that cases lagged 6 months behind Nyong River flow oscillations (Figure 1). When the 2 signals were associated, a 9-month lag behind rainfall oscillations was observed (Technical Appendix Figure 3).

In the BU-endemic focus of Akonolinga, Cameroon, significant 1-year seasonal variations in BU incidence occur. The incubation period for BU has been estimated to be ≈4.5 months when data from Australia are used (10) and ≈3 months when data from Uganda are used (7). The median delay between symptom onset and health care seeking was reported to be 5 weeks in Akonolinga (interquartile range 3–12 weeks), yielding a delay between infection and diagnosis of 5–6 months. Given this delay and a finding of BU diagnosis peaks during March–April, the number of infections would therefore be highest from August through October (Figure 2). Such a pattern was observed in the 1970s in Uganda (6,7) and Cameroon (11) and more recently in Côte d’Ivoire (5). In low BU-endemicity French Guiana, an overseas territory located near Brazil at the same latitude as Cameroon, periodic peaks after the 2 rainy seasons have been reported (12).

Variations in BU incidence result from variations in population exposure (13) combined with variations in environmental presence of M. ulcerans (14). We hypothesize that one of the main drivers of these variations is the seasonal flooding of the Nyong River, which rises 3–5 m from April through November, creating temporary bodies of water and swamps on a vast surface, deeply affecting the ecosystem.

Although M. ulcerans was identified year-round in specific environments such as permanent swamps, its presence and abundance were maximal during the rainy months, July–October (14) (Figure 2). Prevalence of M. ulcerans in rivers was high at the beginning of the rainy season and was high in flooded areas during the following small dry season and high rainy season (14).

Human activity patterns follow these seasonal changes, resulting in seasonal variations in exposure to M. ulcerans. In Uganda, the contribution of permanent swamps to BU risk and the increased risk associated with temporary swamps during the rainy season have been documented (8). According to residence, age, and/or sex, the inhabitants of the Akonolinga district face varying exposures to aquatic environments; during the period identified as high risk, populations frequent seasonally flooded environments for water collection, fishing, and harvest of dry season cultures (Figure 2) (15).

During the study period, the intensity of the association between BU incidence and rainfall or Nyong River flow varied. The seasonal signal was detected over 5 consecutive years and was strongest when yearly variations in the Nyong River flow were lower (2005, 2006, 2008), which could indicate transient forcing of BU incidence by seasonal phenomena. Assessment of the effects of lower frequency climatic events, such as El Niño Southern Oscillation, is needed. In French Guiana, where BU endemicity is low, such events were shown to affect BU incidence dynamics (12).

We showed that BU incidence in this region varies significantly by season and linked these variations to the fluctuations of M. ulcerans occurrence in the environment, which are probably driven by the dynamics of freshwater ecosystems of the Nyong River. In Akonolinga, during the high rainy season when risk for M. ulcerans transmission seems to be highest, populations should increase their protective behaviors, and case detection efforts should be intensified in subsequent months to ensure early diagnosis and access to care.

We thank the staff of the Akonolinga district hospital, who provided diagnoses and delivered treatment to BU patients, and the staff of the Service de Mycobactériologie at Centre Pasteur du Cameroun, who provided laboratory confirmation of diagnoses.