Seasonal patterns of invasive pneumococcal disease

Pneumococcal infections increase each winter, a phenomenon that has not been well explained. We conducted population-based active surveillance for all cases of invasive pneumococcal disease in seven states; plotted annualized weekly rates by geographic location, age, and latitude; and assessed correlations by time-series analysis. In all geographic areas, invasive pneumococcal disease exhibited a distinct winter seasonality, including an increase among children in the fall preceding that for adults and a sharp spike in incidence among adults each year between December 24 and January 7. Pneumococcal disease correlated inversely with temperature (r -0.82 with a 1-week lag; p<0.0001), but paradoxically the coldest states had the lowest rates, and no threshold temperature could be identified. The pattern of disease correlated directly with the sinusoidal variations in photoperiod (r +0.85 with a 5-week lag; p<0.0001). Seemingly unrelated seasonal phenomena were also somewhat correlated. The reproducible seasonal patterns in varied geographic locations are consistent with the hypothesis that nationwide seasonal changes such as photoperiod-dependent variation in host susceptibility may underlie pneumococcal seasonality, but caution is indicated in assigning causality as a result of such correlations.