We looked for atypical weather patterns that could confound, and explain large inconsistencies in, conventional estimates of mortality due to SO(2), CO, and smoke. Using Greater London data for 1976-1995 in the linear temperature/mortality range 0-15 degrees C we determined weather patterns associated with pollutants (all deseasonalized) by single regressions of daily temperature, wind, rain, humidity, and sunshine at successive days advance and delay. Polluted days were colder (P<0.01 for SO(2), CO, and smoke) and less windy and rainy than usual, and this cold weather was more prolonged than usual with 50% maximum temperature depression 5.9 days (95% interval 4.0-7.7) before high SO(2), compared to 2.0 (1.6-2.3) days before average cold days. We also used multiple regression of mortality at 50+ years of age on all these weather factors and pollutants at 0-, 1-, 2- to 4-, 5- to 13-, and 14- to 24-day delays to allow for the atypical weather patterns. This showed cold weather associated with 2.77 excess deaths per million during 24 days following a 1 degrees C fall for 1 day, but no net excess deaths with SO(2) (mean 28.0 ppb) or CO (1.26 ppm). It suggested (P>0.05) some increase with smoke, perhaps acting as surrogate for PM(10), for which data were too scanty to analyze.
Copyright 2001 Academic Press.