Air Temperature and Cardiovascular Mortality in Bavaria
Air Temperature and Cardiovascular Mortality in Bavaria
During the study period from 1990 to 2006 (6209 days), there were 90 879 cardiovascular deaths in Munich, 47 573 in Nuremberg and 49 491 in the Augsburg region (city of Augsburg and administrative districts of Augsburg and Aichach-Friedberg). Table 1 presents descriptive statistics for the cause-specific cardiovascular mortality categories. Slightly more deaths occurred during winter periods. Table 2 shows very similar average mean temperatures for Munich and Nuremberg, but lower average mean temperatures for the Augsburg region; relative humidity was much higher in Augsburg (77.6%) than in Munich (68.6%). Descriptive statistics for the meteorological variables and air pollution for warm periods (June to August) and for cold periods (December to February) are shown in the online supplementary table S1 http://heart.bmj.com/content/100/16/1272/suppl/DC1.
Figure 1 shows 3D plots of cardiovascular mortality by lag and temperature in Munich, Nuremberg and the Augsburg region. The estimated associations for temperature and mortality were non-linear, with associations at high temperatures that persisted up to a lag of 2–3 days; also, longer lagged effects were observed in the cold temperature range.
(Enlarge Image)
Figure 1.
Relative risk of daily cardiovascular mortality in Munich, Nuremberg and the Augsburg region and mean air temperature by lag period, from distributed lag non-linear models (dlnm), 1990–2006. The dlnm comprised four degrees of freedom (df) for the effect of temperature and four df for the lag space. The z axis represents the relative increase in daily counts of mortality with respect to 10°C. Footnote: All models were adjusted long-term trend/seasonality, weekday variations, influenza epidemics, relative humidity and barometric pressure.
Figure 2 presents city/region-specific cumulative effects of 2-day (lag0–1) and 15-day (lag0–14) average air temperatures on cardiovascular mortality, comparing the RR of mortality over the temperature range with the reference temperature 10°C. We found a pronounced effect at higher temperatures for 2-day average temperatures, whereas cold effects only became predominant for 15-day average temperatures in Munich and the Augsburg region.
(Enlarge Image)
Figure 2.
Relative risk in daily cardiovascular mortality by mean air temperature at lag 0–1 (left panel) and lag 0–14 (right panel), using a distributed lag non-linear model, 1990–2006. The y axes represent the relative risk for mortality with a risk of 1.0 at 10°C; the grey-shaded areas are 95% CIs. Footnote: All models were adjusted long-term trend/seasonality, weekday variations, influenza epidemics, relative humidity and barometric pressure.
Table 3 summarises the city/region-specific results by tabulating the RR of daily (cause-specific) cardiovascular mortality in association with increases in 2-day and decreases in 15-day average air temperatures. The first column compares the 99th centiles with the 90th centiles of temperature, reflecting essentially heat effects. For all, Munich, Nuremberg and the Augsburg region, we found consistent associations between heat and increases in (cause-specific) cardiovascular mortality. The second column of Table 3 shows the effects of low temperatures on mortality, comparing 1st centiles to 10th centiles. Decreases in the 15-day average temperatures were associated with increases in overall cardiovascular mortality for Munich and the Augsburg region, but not for Nuremberg.
Combined estimates of the exposure-response relationship between air temperature and cause-specific cardiovascular mortality are given in figure 3 and Table 4, showing significant increases in all subcategories for high 2-day average temperatures, except for mortality due to myocardial infarctions. An increase from the 90th centile (20.0°C) to the 99th centile (24.8°C) of 2-day average temperature led to an increase in cardiovascular mortality by 9.5% (95% CI 4.6% to 14.5%) (Table 4). Increases in mortality due to heart failure and cerebrovascular diseases were found for low 15-day average temperatures (figure 3 and Table 4). There was nearly no or only moderate heterogeneity in the city-specific exposure-response relationships (figure 3). We found stronger associations for the elderly (≥75 years) (Table 4).
(Enlarge Image)
Figure 3.
Combined relative risks in cause-specific cardiovascular mortality by air temperature at lag 0–1 (left panel) and lag 0–14 (right panel), 1990–2006. The continuous bold black line represents the population-average curve, whereas the dashed grey lines are the city-specific estimates. The y axes represent the relative risk for mortality with a risk of 1.0 at 10°C; the grey-shaded areas are 95% CIs.
In addition, we explored the effects of air temperature on mortality within warm periods (June to August) and cold periods (December to February) separately (see online supplementary table S2 http://heart.bmj.com/content/100/16/1272/suppl/DC1). Overall, the results were comparable or stronger for the summer period compared with the whole-year analysis, whereas for the winter period no effects could be seen anymore.
We performed sensitivity analyses to check our main findings. For example, the effect estimates were stable or slightly increased when using three df or seven df per year for the spline of trend (see online supplementary table S3 http://heart.bmj.com/content/100/16/1272/suppl/DC1). When we changed df for the temperature space in the dlnm, the estimated RRs were slightly smaller than in the main analysis; however, this did not change the statistical significance of the risk estimates. Overall, results of the analysis using a time-stratified case cross-over approach were consistent and comparable with the main analysis results (see online supplementary table S4 http://heart.bmj.com/content/100/16/1272/suppl/DC1).
Results
During the study period from 1990 to 2006 (6209 days), there were 90 879 cardiovascular deaths in Munich, 47 573 in Nuremberg and 49 491 in the Augsburg region (city of Augsburg and administrative districts of Augsburg and Aichach-Friedberg). Table 1 presents descriptive statistics for the cause-specific cardiovascular mortality categories. Slightly more deaths occurred during winter periods. Table 2 shows very similar average mean temperatures for Munich and Nuremberg, but lower average mean temperatures for the Augsburg region; relative humidity was much higher in Augsburg (77.6%) than in Munich (68.6%). Descriptive statistics for the meteorological variables and air pollution for warm periods (June to August) and for cold periods (December to February) are shown in the online supplementary table S1 http://heart.bmj.com/content/100/16/1272/suppl/DC1.
Figure 1 shows 3D plots of cardiovascular mortality by lag and temperature in Munich, Nuremberg and the Augsburg region. The estimated associations for temperature and mortality were non-linear, with associations at high temperatures that persisted up to a lag of 2–3 days; also, longer lagged effects were observed in the cold temperature range.
(Enlarge Image)
Figure 1.
Relative risk of daily cardiovascular mortality in Munich, Nuremberg and the Augsburg region and mean air temperature by lag period, from distributed lag non-linear models (dlnm), 1990–2006. The dlnm comprised four degrees of freedom (df) for the effect of temperature and four df for the lag space. The z axis represents the relative increase in daily counts of mortality with respect to 10°C. Footnote: All models were adjusted long-term trend/seasonality, weekday variations, influenza epidemics, relative humidity and barometric pressure.
Figure 2 presents city/region-specific cumulative effects of 2-day (lag0–1) and 15-day (lag0–14) average air temperatures on cardiovascular mortality, comparing the RR of mortality over the temperature range with the reference temperature 10°C. We found a pronounced effect at higher temperatures for 2-day average temperatures, whereas cold effects only became predominant for 15-day average temperatures in Munich and the Augsburg region.
(Enlarge Image)
Figure 2.
Relative risk in daily cardiovascular mortality by mean air temperature at lag 0–1 (left panel) and lag 0–14 (right panel), using a distributed lag non-linear model, 1990–2006. The y axes represent the relative risk for mortality with a risk of 1.0 at 10°C; the grey-shaded areas are 95% CIs. Footnote: All models were adjusted long-term trend/seasonality, weekday variations, influenza epidemics, relative humidity and barometric pressure.
Table 3 summarises the city/region-specific results by tabulating the RR of daily (cause-specific) cardiovascular mortality in association with increases in 2-day and decreases in 15-day average air temperatures. The first column compares the 99th centiles with the 90th centiles of temperature, reflecting essentially heat effects. For all, Munich, Nuremberg and the Augsburg region, we found consistent associations between heat and increases in (cause-specific) cardiovascular mortality. The second column of Table 3 shows the effects of low temperatures on mortality, comparing 1st centiles to 10th centiles. Decreases in the 15-day average temperatures were associated with increases in overall cardiovascular mortality for Munich and the Augsburg region, but not for Nuremberg.
Combined estimates of the exposure-response relationship between air temperature and cause-specific cardiovascular mortality are given in figure 3 and Table 4, showing significant increases in all subcategories for high 2-day average temperatures, except for mortality due to myocardial infarctions. An increase from the 90th centile (20.0°C) to the 99th centile (24.8°C) of 2-day average temperature led to an increase in cardiovascular mortality by 9.5% (95% CI 4.6% to 14.5%) (Table 4). Increases in mortality due to heart failure and cerebrovascular diseases were found for low 15-day average temperatures (figure 3 and Table 4). There was nearly no or only moderate heterogeneity in the city-specific exposure-response relationships (figure 3). We found stronger associations for the elderly (≥75 years) (Table 4).
(Enlarge Image)
Figure 3.
Combined relative risks in cause-specific cardiovascular mortality by air temperature at lag 0–1 (left panel) and lag 0–14 (right panel), 1990–2006. The continuous bold black line represents the population-average curve, whereas the dashed grey lines are the city-specific estimates. The y axes represent the relative risk for mortality with a risk of 1.0 at 10°C; the grey-shaded areas are 95% CIs.
In addition, we explored the effects of air temperature on mortality within warm periods (June to August) and cold periods (December to February) separately (see online supplementary table S2 http://heart.bmj.com/content/100/16/1272/suppl/DC1). Overall, the results were comparable or stronger for the summer period compared with the whole-year analysis, whereas for the winter period no effects could be seen anymore.
We performed sensitivity analyses to check our main findings. For example, the effect estimates were stable or slightly increased when using three df or seven df per year for the spline of trend (see online supplementary table S3 http://heart.bmj.com/content/100/16/1272/suppl/DC1). When we changed df for the temperature space in the dlnm, the estimated RRs were slightly smaller than in the main analysis; however, this did not change the statistical significance of the risk estimates. Overall, results of the analysis using a time-stratified case cross-over approach were consistent and comparable with the main analysis results (see online supplementary table S4 http://heart.bmj.com/content/100/16/1272/suppl/DC1).