The Figure shows the close association between total monthly rainfall, as recorded at Darwin Airport, and the number of cases of melioidosis in the Top End for each month during the 12 years. The correlation between monthly cases of melioidosis and rainfall at Darwin Airport in the preceding calendar month (r = 0.617; p < 0.0001) was slightly tighter than the correlation with rainfall in the concurrent month (r = 0.574).

Table 1 shows the median 14-day rainfall correlation with various risk factors, clinical signs and symptoms, and outcomes. The correlation with median 14-day rainfall was significantly higher for patients with pneumonia, those with bacteremia and septic shock, and those who died. For those patients with pneumonia, correlation with median 14-day rainfall was significantly higher if they were bacteremic. Patients with diabetes, alcohol excess, and chronic renal disease were all associated with significantly higher median 14-day rainfall; correlation with 14-day rainfall did not significantly differ for age, sex, ethnicity, chronic lung disease, and smoking (data not shown).

Table 2 shows that, on univariate analysis, 14-day rainfall ≥125 mm correlated significantly with primary symptoms of pneumonia, bacteremia, and septic shock, and with death. Table 3 shows independent predictors of clinical signs and symptoms and outcome when multivariate analysis with 14-day rainfall as a categorical variable was used. When 14-day rainfall was used as a continuous variable, it was an independent risk factor for a admission with pneumonia (p = 0.023), bacteremic pneumonia (p = 0.001), septic shock (p = 0.005), and of death (p < 0.0001). Notably, while absence of any risk factors was a predictor of primary signs and symptoms other than pneumonia and of less severe disease, neither diabetes, alcohol excess, nor chronic renal disease was an independent predictor of signs and symptoms, disease severity, or death.

Our data confirm our observations that patients admitted with melioidosis 1–2 weeks after heavy monsoonal rainfall are more ill and more likely to die. Median rainfall in the 14 days before admission was highest in those who died with melioidosis (211 mm). For those admitted with bacteremic pneumonia, prior 14-day median rainfall was 188 mm, compared with 89 mm in patients who were nonbacteremic and did not have pneumonia. Multivariate analysis showed that rainfall in the 14 days before admission was an independent predictor of septic shock and death. Patients were 2.5 times more likely to die from melioidosis if the rainfall in the 14 days before admission was ≥125 mm. Overall, 68% of deaths occurred in this high rainfall group. Furthermore, prior heavy rainfall was an independent predictor of admission with pneumonia rather than with no pneumonia. Patients were almost twice as likely to have bacteremic pneumonia if the rainfall in the 14 days before admission was ≥125 mm.

Earlier literature, including that involving soldiers from the Vietnam War, suggests that inhalation is a common mode of infection with B. pseudomallei (13,14). This scenario was proposed for those exposed to dust raised by helicopter rotor blades in Vietnam (15). However, recent reviews have supported the predominant role of percutaneous inoculation of B. pseudomallei after exposure to muddy soils or surface water in endemic locations (10,11,16). Admissions with melioidosis pneumonia after presumptive inoculating skin injuries have been documented in patients with soil-contaminated burns (17) and are also common in our hospital (3,10). This finding suggests hematogenous spread to the lung rather than inhalation or spread from the upper respiratory tract. This finding is analogous to postprimary tuberculosis, with disease from hematogenous spread localizing in the upper lung zones, where highest alveolar oxygen tension exists (18). Moreover, septicemic melioidosis pneumonia patients are often more systemically ill than is suggested by initial chest x-ray, supporting the concept of spread to, rather than from, the lung.

Even if percutaneous inoculation is more common overall, the association of prior heavy rainfall with both pneumonia and more severe disease may well reflect a shift towards inhalation as the mode of acquiring B. pseudomallei. The periods of intense monsoonal rainfall are usually also associated with heavy winds and melioidosis cases, and outbreaks are documented after cyclonic winds and rain (19,20). Aerosolization of bacteria from surface soil and water under such conditions is probable, resulting in the potential for inhalation of B. pseudomallei. Melioidosis following near-drowning is well recognized, with aspiration considered the likely infecting event, followed by pneumonia after an incubation period as short as 2 days (21–23).

That melioidosis can potentially be more severe after inhalation than after percutaneous inoculation is not surprising. This finding is well recognized for anthrax, plague, and tularemia and has implications for biological warfare considerations (24–26). However, as with melioidosis, septicemia with pulmonary involvement after percutaneous inoculation is well recognized with anthrax, plague, and tularemia. The lack of clarity of correlation between mode of infection, site of disease, and clinical course in the melioidosis literature is also evident in descriptions of the closely related disease, glanders (infection with B. mallei) (15).

Additional possible explanations for more severe disease after heavy rainfall include a larger bacterial inoculating dose and infection with more virulent bacteria. The association of melioidosis with the wet season has also been postulated to be due to movement of B. pseudomallei from deeper soil layers to the surface with the rising water table (27). Early studies also speculated that the increased isolation of B. pseudomallei from surface water after heavy rains resulted from increased growth of the bacteria (28). More recently, the possibility has been raised that B. pseudomallei may persist in the environment in a viable nonculturable state during times of stress, such as in prolonged dry seasons (20,29). Differential gene activation likely allows such environmental bacteria to respond and adapt to different environmental conditions (30). Recently, viable but nonculturable cells of Francisella tularensis have been shown to be avirulent in mice (31). Thus, both increased environmental bacterial load and increased virulence of environmental B. pseudomallei may possibly result from periods of heavy rainfall. A possible confounder to analyzing associations of rainfall with disease severity is that, whatever the mechanisms of more severe disease, such cases will tend to have shorter incubation periods. Therefore, the prior 14-day rainfall is likely to more closely reflect the rainfall associated with infection in these cases than in less severe cases, where incubation periods >14 days might occur.

In patients with melioidosis in this study, diabetes, alcohol excess, and chronic renal disease were associated with higher prior rainfall. We previously suggested that the predisposition to melioidosis in persons with these three conditions may relate primarily to impaired polymorphonuclear leukocyte (PMNL) functions (3). This hypothesis is supported by data from an observational, uncontrolled study showing improved survival with use of granulocyte colony-stimulating factor (G-CSF) in melioidosis septic shock (32). Recent animal data suggest an important role for lung-derived G-CSF in controlling intrapulmonary infection (33). Therefore, in diabetes, alcoholism, or chronic renal disease, both impaired phagocytic activity of alveolar macrophages and impaired recruitment of PMNL into the lungs as a result of acquired dysfunction of alveolar macrophages may be critical, in addition to impaired PMNL function, in determining the predisposition to melioidosis pneumonia. Such a predisposition is likely to be especially important in influencing whether infection becomes established after inhalation of B. pseudomallei. The association of diabetes, alcohol excess, and chronic renal disease with higher prior rainfall may therefore reflect a particular susceptibility to inhalation as a mode of infection in patients with these risk factors. Alternatively, this finding may reflect a greater influence of bacterial load or organism virulence in these risk groups. Either explanation is consistent with the observation from Thailand that risk factors and level of environmental exposure to B. pseudomallei have a compound interaction, as is evident in the especially high rates of melioidosis in diabetic rice farmers (34).

We have shown that the intensity of rainfall in the 14 days before a person is admitted to hospital with melioidosis is an independent predictor of the patient’s having pneumonia, septic shock developing, and death. We postulate that this may reflect a shift towards inhalation of B. pseudomallei as the mode of transmission after heavy monsoonal rains and winds.