Seasonal Average Temperature Differences and CVD Incidence: Results from the US-Based Nurses' Health Study, Nurses' Health Study II, and the Health Professional Follow-Up Study

Abstract

Background: Climate change is one of the greatest health threats facing humanity. Multiple studies have documented the impact of short-term temperature exposure on human health. However, long-term temperature exposures are far less studied.

Objectives: We examined whether exposures to higher or lower summer and winter average temperatures compared to long-term average temperatures were associated with cardiovascular disease (CVD) incidence in three US-based cohorts.

Methods: We followed 276,618 participants from the Nurses' Health Study (NHS) (1991-2018), the Nurses' Health Study II (NHSII) (1994-2017), and the Health Professionals' Follow-Up Study (1991-2015). We used data (1986-2018) from PRISM Spatial Climate Datasets ($800\text{-}\times 800\text{-m}$ spatial resolution) to calculate differences between the summer (June-August) and winter (December-February) average temperatures and the previous 5-year summer and winter average temperatures at residential addresses of each participant. CVD incidence was defined as first nonfatal or fatal myocardial infarction (MI) or nonfatal or fatal stroke. Cox proportional hazard models were used to examine associations with between average temperatures and CVD incidence. Hazard ratios (HRs) and 95% confidence intervals (95% CI) were pooled using random effect meta-analysis. We also examined associations in the populations $<65$ and $65+$ years of age.

Results: After pooling HRs, we found no association of summer average temperatures higher than the previous 5-year average temperature, with CVD incidence. A winter average temperature lower than the previous 5-year average was associated with CVD incidence ($\text{HR}=0.95$ per 2.7°C increase; 95% CI: 0.89, 1.01). Among persons $<65$ years of age, we observed increased CVD risks with higher summer average temperatures (pooled $\text{HR}=1.03$ per 1.3°C increase; 95% CI: 1.00, 1.07) and lower winter average temperatures (pooled $\text{HR}=0.91$ per 2.7°C increase; 95% CI: 0.87, 0.95) compared to the previous 5-year average temperature.

Discussion: Exposure to a winter average temperature lower than the previous 5-year average was suggestively associated with an increased CVD risk. Exposure to a summer average temperature higher than the previous 5-year average was associated with CVD incidence in the population $<65$ years of age but not in the full population. https://doi.org/10.1289/EHP14677.

Figure 1.

Forest plots of associations of summer and winter temperature difference with CVD, MI, and stroke incidence in the NHS ($n=\mathrm{112,650}$ women), NHSII ($n=\mathrm{116,076}$ women), and HPFS ($n=\mathrm{47,892}$ men) cohorts. There were 11,468 CVD, 5,933 MI, and 5,535 stroke cases in 30,093,911 person-months in NHS. There were 2,006 CVD, 1,063 MI, and 943 stroke cases in 31,822,076 person-months in NHSII. There were 2,332 CVD, 894 MI, and 1,438 stroke cases in 11,174,144 person-months in HPFS. Associations are expressed per IQR (averaged over all three cohorts) increase. The average IQR was 1.3°C for the summer temperature difference and 2.7°C for the winter temperature difference. Cox proportional hazard models included summer and winter temperature difference compared to the 5-year average. Models were stratified by age in months and adjusted for race, calendar year, marital status, living alone, retirement status, smoking status, pack years, diet, alcohol consumption, family history of MI, BMI, physical activity, nSES, cancer, and comorbidities at baseline. For the NHS and NHSII, we also adjusted for occupation of both of the nurse’s parents when she was 16, educational status of her spouse/partner, menopausal status, postmenopausal hormone use, and night shift work. Associations were pooled with a random effect model. Note: BMI, body mass index; CI, confidence interval; CVD, cardiovascular disease; FE, fixed effects; HPFS, Health Professional Follow-up Study; HR, hazard ratio; IQR, interquartile range; MI, myocardial infarction; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; nSES, neighborhood socioeconomic status; RE, random effect.

Figure 2.

Forest plots of associations of summer and winter temperature difference with CVD, MI, and stroke incidence in subset populations of NHS, NHSII, and HPFS including participants $<65$ years of age. There were 1,656 CVD, 964 MI, and 692 stroke cases in 10,323,388 person-months in NHS ($n=\mathrm{89,007}$). There were 1,843 CVD, 979 MI, and 864 stroke cases in 30,434,929 person-months in NHSII ($n=\mathrm{116,075}$). There were 369 CVD, 158 MI, and 211 stroke cases in 4,205,374 person-months in HPFS ($n=\mathrm{32,680}$). Associations are expressed per IQR (averaged over all three cohorts) increase. The average IQR was 1.3°C for the summer temperature difference and 2.7°C for the winter temperature difference. Cox proportional hazard models included summer and winter temperature difference compared to the 5-year average. Models were stratified by age in months and adjusted for race, calendar year, marital status, living alone, retirement status, smoking status, pack years, diet, alcohol consumption, family history of MI, BMI, physical activity, nSES, cancer, and comorbidities at baseline. For the NHS and NHSII, we also adjusted for occupation of both of the nurse’s parents when she was 16, educational status of her spouse/partner, menopausal status, postmenopausal hormone use, and night shift work. Associations were pooled with a random effect model. Note: BMI, body mass index; CI, confidence interval; CVD, cardiovascular disease; FE, finite element; HPFS, Health Professional Follow-up Study; HR, hazard ratio; IQR, interquartile range; MI, myocardial infarction; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; nSES, neighborhood socioeconomic status.