Study Design and Participants

Figure 1

Figure 1. Locations of Buruli ulcer–endemic areas included in comprehensive case-control study of protective and risk factors for Buruli ulcer, Victoria, Australia. Colors indicate risk classifications at beginning of the study period,...

Figure 2

Figure 2. Flow diagrams of study recruitment, participation, and exclusion criteria in comprehensive case–control study of protective and risk factors for Buruli ulcer, southeastern Australia. A) Case-patient recruitment; B) control recruitment.

We performed a postcode matched, case–control study in BU-endemic areas surrounding Port Phillip Bay, Victoria, Australia (Figure 1; Appendix Table 1). Ethics approval was granted by the Victoria Department of Health Human Research Ethics Committee (project 10–18). We invited adults (>18 years of age) to participate in the study who resided in Victoria and were notified to the Department of Health in Victoria as having laboratory-confirmed BU during June 2018–June 2020. We extracted case data from the Victoria Department of Health Public Health Events Surveillance System. We recruited case-patients via regular mail after receiving permission for contact from the patient’s general practitioner or treating medical team. We restricted analysis to residents or holiday homeowners in the study areas (Figure 2).

We matched control participants (residents of Victoria >18 years of age) to patients according to residential postal codes within the study area. We selected controls from both the Victorian Population Health Survey (participants who had provided consent to be contacted for other research studies) and the electoral roll of Australia (when additional matched controls were required for a particular postcode). We excluded controls if they or a household member had been previously diagnosed with BU (Figure 2).

Participation for both patients and controls involved the return of a completed study questionnaire. In addition, a subsample of patients and controls were enrolled in an environmental survey of residential properties that investigated the presence of M. ulcerans (14).

Data Collection and Measurements

We used a self-administered questionnaire to examine the amount of time participants spent in the study areas, outdoor and lifestyle behaviors, insect exposure, medical history, and environmental characteristics of the participants’ properties. We evaluated those details and formulated response and collapsed categories for analysis (Appendix Table 2). Participant-reported medications and conditions that might affect the immune system were reviewed by a physician specializing in infectious diseases (D.P.O.) to ascertain those likely to cause immunosuppression. We devised an occupational classification related to potential environmental exposure to M. ulcerans through employment by using participant responses to 2 questions: what proportion of your time do you spend outside as part of your occupation and are you in contact with the soil during your work? We examined the effects of working outdoors and having soil contact among participants whose employment was based in the study (disease-endemic) areas only.

Statistical Analysis

We evaluated host, environmental, and behavioral factors according to BU case status. We examined relationships between those factors and the likelihood of developing BU by using multivariable conditional logistic regression; cases and controls were matched by postcode. We calculated odds ratios adjusted for age and sex (aORs) and 95% CIs for the total participant sample (residents and holiday homeowners) and separately for residents only (Appendix Tables 3–11). Percentages of missing data were generally low (<3% for most factors); if missing data were >10%, we included a separate category for those participants with missing exposure data in the model unless otherwise stated. Given the expectation that participants might have multiple potentially protective health behaviors, we examined patterns and clustering of those behaviors by using polychoric correlations and exploratory factor analysis (Appendix Figures 2, 3).

We conducted a post-hoc sensitivity analysis to explore the robustness of the observed relationship between BCG vaccination and BU case status; we restricted analysis to participants 47–70 years of age who were within the age-range eligible for BCG vaccination as part of the routine vaccination schedule for schoolchildren in Victoria from the 1950s to 1985 (24). We analyzed those reporting receipt of BCG vaccination and those unsure of vaccination status as a single category (under the assumption of likely vaccination through routine vaccination) and compared them with age-matched participants reporting no BCG vaccination. We performed analyses by using Stata 15 (StataCorp LLC, https://www.stata.com) except for factor analysis, which we performed by using Stata 16.

Demographic and Clinical Characteristics of Participants

We examined data from 245 (57% participation rate) BU case-patients and 481 (18%) postcode-matched control participants from across the BU-endemic areas; 171 (70%) patients and 469 (97.5%) controls were permanent residents in the study areas, and most (71%) were homeowners in high BU-endemic areas of Mornington Peninsula (Figure 1). Half (123/245) of case-patients were 60–79 years of age, signifying an overrepresentation when compared with all notified cases in the study areas (204/550 [37%] 60–79 years of age). In contrast, patients 18–39 years of age were underrepresented in our participant sample (35/245 [14%] compared with 134/550 [24%] among notified cases) (Appendix Table 12). We also observed an overrepresentation of controls 60–79 years of age and a large underrepresentation of controls 18–39 years of age when compared with population proportion estimates (Appendix Table 12). Male sex was associated with BU case status (57.6% of BU cases vs. 44.7% of controls; aOR 1.52 [95% CI 1.06–2.19]).

BU cases were reported predominantly during winter (44%) and spring (38%) (Table; Appendix Figure 1). The median time between symptom onset and diagnosis was 5 (interquartile range [IQR] 3–12) weeks; duration was longer for patients who were holiday homeowners (8 [IQR 4–13]) weeks than for those who were residents (4 [IQR 3–10] weeks; p<0.0001 by rank-sum test). An insect bite, wound, or injury to the affected area was reported in 36% of BU cases before ulcers appeared.

Host Factors

Figure 3

Figure 3. Odds of developing Buruli ulcer according to different host factors in comprehensive case–control study of protective and risk factors for Buruli ulcer, southeastern Australia. Host characteristics are shown for case-patients...

We evaluated associations between host factors and BU case status (Figure 3). Persons with a history of diabetes mellitus had a higher probability of developing BU than those without diabetes (aOR 2.26 [95% CI 1.13–4.49]). An association was observed with prednisolone therapy (aOR 2.56 [95% CI 1.28–5.13]); however, this result could be confounded by persons commencing prednisolone therapy during their BU treatment.

Receipt of BCG vaccination was associated with lower odds of BU (aOR 0.59 [95% CI 0.39–0.90]) than for participants reporting no BCG vaccination. No relationship between BU and vaccination timing (<20 or >20 years ago) was observed. Of note, 41% of patients and 31% of controls reported that they were unsure whether they had received the vaccination. In the sensitivity analysis that restricted participant age to 47–70 years (those unsure were assumed vaccinated), the observed association between BU and BCG vaccination persisted but was attenuated; aOR was 0.71 (95% CI 0.41–1.22) for the entire age-restricted participant sample (Appendix Table 11).

Environmental Factors

Figure 4

Figure 4. Odds of developing Buruli ulcer according to different environmental factors in comprehensive case–control study of protective and risk factors for Buruli ulcer, southeastern Australia. Environmental factors are shown for case-patients...

The presence of possums around the property was strongly associated with BU in residents (aOR 5.30 [95% CI 1.82–15.49]) and, to a lesser extent, in the entire participant sample (aOR 2.33 [95% CI 1.15–4.71]). The likelihood of developing BU increased with the number of possums reported around the residential property (Figure 4; Appendix Table 5); large amounts of possum feces (compared with none) (aOR 1.88 [95% CI 1.05–3.36]); and with the presence of tea trees (Leptospermum sp.), a common habitat for possums, on the property (aOR 1.72 [95% CI 1.10–2.69]).

Most (98%) properties used piped (town) water for drinking, bathing, and garden watering. Participants drinking filtered town water (274/721, 38% of total participants) had lower odds of developing BU than those not drinking filtered town water (aOR 0.64 [95% CI 0.46–0.90]). Of those not drinking filtered town water, 433/447 (97%) drank unfiltered town water, and 14 (3%) drank water from other sources only, such as tank or bottled water. Use of bore water by residents for bathing or garden watering was associated with BU (aOR 1.56 [95% CI 0.98–2.50]). Water sources around the property were not associated with BU case status, except for the presence of ponds (aOR 1.69 [95% CI 0.99–2.89]) for residents (Figure 4). We observed no associations between case status and the presence of other nonpossum wildlife or biting insects; use of garden products (mulch or potting mix) among residents; or with earthworks, major renovations, or sewerage works near the property (Appendix Table 7).

Exposures

Figure 5

Figure 5. Odds of developing Buruli ulcer according to potential outdoor exposures in comprehensive case–control study of protective and risk factors for Buruli ulcer, southeastern Australia. Potential outdoor exposures are shown for...

Working outdoors was associated with higher odds of BU than working indoors in BU-endemic areas (Figure 5); highest odds were associated with occupations involving soil contact (aOR 2.89 [95% CI 1.01–8.25]). Outdoor occupations that involved soil contact were commonly gardeners, carpenters, and other construction-related roles.

We found no association between gardening frequency and BU case status among residents (Figure 5); however, the entire participant sample comprising more holiday homeowner cases had lower odds for BU (aOR 0.50 [95% CI 0.34–0.74]). Participants partaking in outdoor activities (>95% of participants) had a lower likelihood of developing BU than those not undertaking outdoor activities (aOR 0.34 [95% CI 0.17–0.68]). However, we observed no strong associations between participants undertaking individual activities (beach walks/jogging, wetland walks/jogging, bushwalking, golf, sports on an oval, swimming in local lakes/rivers, sailing, outdoor barbeques, or other activities) and those not undertaking the activity (Appendix Table 9).

Protective Behavioral Factors

Figure 6

Figure 6. Odds of developing Buruli ulcer according to protective behavioral factors in comprehensive case-control study of protective and risk factors for Buruli ulcer, southeastern Australia. Potential protective behavioral factors are shown...

We analyzed associations between protective health behaviors and BU case status (Figure 6). Several protective behaviors were associated with lower odds of developing BU: tending immediately to cuts and scratches received during outdoor activity by washing the area and then applying antiseptic or dressings (aOR 0.56 [95% CI 0.36–0.87]), wearing insect repellant during warmer months (aOR 0.62 [95% CI 0.43–0.89]), and covering arms and legs with clothing (aOR 0.59 [95% CI 0.36–0.90]). Participants who combined protective behaviors had the strongest correlations between tending to new wounds, covering preexisting wounds, washing hands after outdoor activity, and using gloves for gardening (Appendix Figures 2, 3). Combining protective behaviors was associated with lower odds of BU; we observed a gradient of decreasing odds for BU in those undertaking higher numbers of protective behaviors (Figure 6).