Let the sun shine in: effects of ultraviolet radiation on invasive pneumococcal disease risk in Philadelphia, Pennsylvania

Abstract

Background: Streptococcus pneumoniae is a common cause of community acquired pneumonia and bacteremia. Excess wintertime mortality related to pneumonia has been noted for over a century, but the seasonality of invasive pneumococcal disease (IPD) has been described relatively recently and is poorly understood. Improved understanding of environmental influence on disease seasonality has taken on new urgency due to global climate change.

Methods: We evaluated 602 cases of IPD reported in Philadelphia County, Pennsylvania, from 2002 to 2007. Poisson regression models incorporating seasonal smoothers were used to identify associations between weekly weather patterns and case counts. Associations between acute (day-to-day) environmental fluctuations and IPD occurrence were evaluated using a case-crossover approach. Effect modification across age and sex strata was explored, and meta-regression models were created using stratum-specific estimates for effect.

Results: IPD incidence was greatest in the wintertime, and spectral decomposition revealed a peak at 51.0 weeks, consistent with annual periodicity. After adjustment for seasonality, yearly increases in reporting, and temperature, weekly incidence was found to be associated with clear-sky UV index (IRR per unit increase in index: 0.70 [95% CI 0.54-0.91]). The effect of UV index was highest among young strata and decreased with age. At shorter time scales, only an association with increases in ambient sulphur oxides was linked to disease risk (OR for highest tertile of exposure 0.75, 95% CI 0.60 to 0.93).

Conclusion: We confirmed the wintertime predominance of IPD in a major urban center. The major predictor of IPD in Philadelphia is extended periods of low UV radiation, which may explain observed wintertime seasonality. The mechanism of action of diminished light exposure on disease occurrence may be due to direct effects on pathogen survival or host immune function via altered 1,25-(OH)2-vitamin-D metabolism. These findings may suggest less diminution in future IPD risk with climate change than would be expected if wintertime seasonality was driven by temperature.

Figure 1

Periodogram Constructed from Spectral Decomposition of Weekly Pneumococcal Case Counts. Spectral density is represented on the y-axis, and can be conceptualized as a measure of goodness-of-fit for oscillatory regression models at different frequencies. The large peak at a frequency of 51 weeks suggests that invasive pneumococcal disease is a process that oscillates with annual periodicity (and is, in other words, compatible with wintertime seasonality). The two peaks at lower frequencies are lower harmonics illustrating bi- and tri-annual behaviour.

Figure 2

Autocorrelogram for Weekly Invasive Pneumococcal Case Counts in Philadelphia, 2002 to 2007. Positive autocorrelation is observed at intervals of 12 months, and negative autocorrelation is observed at intervals of 6 months, consistent with steady seasonal oscillation in disease risk.

Figure 3

Trends in Invasive Pneumococcal Disease in Philadelphia County. Bars represent actual case counts, and the curve depicts expected incidence of disease occurrence based on a multivariable Poisson regression model including sine and cosine oscillators, an annual term, and a linear spline term with a knot at January 1, 2004. Reported cases are seasonal (wintertime predominant); incidence increases sharply in 2002 and 2003 with the introduction of mandatory reporting, and more slowly thereafter, as described in the text.

Figure 4

Changing Effect of UV Radiation on Invasive Pneumococcal Disease Risk by Age Group. Figure constructed by performing meta-regression using age-stratum specific estimates of the effect of a single unit increase clear-sky UV index on the incidence of invasive pneumococcal disease. A log-linear increase in the effect of UV index is observed with decreasing age.