Background: Summer predominance of tuberculosis (TB) was reported previously in temperate regions. No consistent data were available for lower latitudes.
Method: The monthly TB notification data in Hong Kong from 1991 to 2002 were examined for seasonal fluctuation. A seasonal model was then developed after standardization by period, sex, age, history of TB, form of disease, and bacteriological status.
Results: The raw monthly counts showed remarkably consistent seasonal fluctuation across different periods, sexes, and age groups. A sine model was fitted for 82 104 notifications (adjusted R(2) = 0.373, P < 0.001). A summer peak was observed with seasonal fluctuation of 18.4% (P < 0.001), which was substantially higher than that reported previously for temperate regions. The amplitudes of fluctuation were 35.0, 15.0, 19.0, and 20.2% for those aged < or =14, 15-34, 35-64, and >/=65 years, respectively (all P < 0.001). No gender difference was noted (18.2% vs 19.0%, P = 0.790). Seasonal pattern was detected among new cases (18.6%, P < 0.001), but not retreatment cases (5.2%, P = 0.333). Culture-positive cases showed greater fluctuation than culture-negative cases (29.4% vs 6.4%, P < 0.001). No significant difference was found between pulmonary and extrapulmonary cases (16.8% vs 21.6%, P = 0.356). TB cases notified in summer were more likely to be smear-positive [odds ratio (OR) 1.100, 95% confidence interval (CI) 1.045-1.158, P < 0.001] and culture-positive (OR 1.175, 95% CI 1.121-1.232, P < 0.001) than those notified in winter, even after stratification by other key variables.
Conclusion: A consistent seasonal pattern was found, with variable amplitudes of fluctuation in different subgroups and differing disease characteristics in different seasons. These observations are suggestive of the presence of a seasonal disease-modifying factor.