Cigarette Smoking and Competing Risks for Fatal and Nonfatal Cardiovascular Disease Subtypes Across the Life Course

All data and materials are made publicly available by the National Heart, Lung, and Blood Institute bioLINCC (Biologic Specimen and Data Repository Information Coordinating Center). The LRPP (Lifetime Risk Pooling Project) is an individual‐level pooled and harmonized data set from 20 community‐based or population‐based cardiovascular disease (CVD) cohorts in the United States available in BioLINCC spanning the past 70 years with detailed methods of cohort development and harmonization previously published.⁹ There is nearly 100% complete follow‐up for vital status. For the current analysis, we included data from participants free of clinical CVD at baseline from the following 9 prospective cohorts in the LRPP: the Coronary Artery Risk Development in Young Adults, the Atherosclerosis Risk in Communities Study, the Cardiovascular Health Study, the Multi‐Ethnic Study of Atherosclerosis, the Framingham Heart Study, the Framingham Offspring Study, the Kaiser Permanente Study of the Oldest Old, the Chicago Heart Association Detection Project in Industry Study, and the National Health and Nutrition Examination Survey I Epidemiologic Follow‐Up Study. All data were deidentified, and all study protocols were approved by the institutional review board at Northwestern University. Requests to access the data set from qualified researchers may be submitted at https://biolincc.nhlbi.nih.gov/studies/.

We included data from only 1 examination for each cohort from participants who had available responses for self‐reported smoking status, available data on covariates, and at least 10 years of follow‐up. Smoking status was defined as yes or no based on the participant response in each cohort to current smoking (asked in each cohort at the baseline examination for current smoking or smoking within the past 30 days). We excluded 3192 participants who did not report White or Black race, 3961 participants with prevalent CVD at baseline, and 7773 participants without exposure or covariate information for the final analytic sample of 106 165 (which represented a complete case analysis). Baseline demographics of age, sex, and race and ethnicity were self‐reported. In each cohort, trained clinical staff measured blood pressure, weight, and height. Fasting glucose and lipid profile were directly measured. Diagnosis of diabetes and use of diabetes/hypertensive medications were self‐reported. Fatal and nonfatal MI were adjudicated based on defined clinical criteria that included at least 2 of the following: electrocardiographic changes consistent with MI, biomarker elevations consistent with myocardial injury, and typical chest pain or International Classification of Diseases, Eighth Revision (ICD‐8)/International Classification of Diseases, Ninth Revision (ICD‐9) codes. Stroke was adjudicated using the clinical criteria of new neurological deficit persisting for >24 hours consistent with neurological injury and/or ICD‐8/ICD‐9 codes. HF was determined by clinical criteria and/or ICD‐8/ICD‐9 codes. Vital status was obtained through linkage with the National Death Index, and cause of death was adjudicated by review of medical records and autopsy records, when available.

The study included 1 949 658 person‐years of follow‐up, including observations from 1948 to 2017. The participants were stratified by sex, index age groups (20–39 years for young adults, 40–59 years for middle‐aged adults, and 60–79 years for older adults), and smoking status at the index examination. Long‐term risk was calculated using a modified Kaplan–Meier analysis, which accounts for non‐CVD death as a competing risk. This model accounts for fatal non‐CVD events as a separate end point rather than a censoring event. Long‐term risk of total CVD was estimated for each index age group by calculating the rates of CVD death, MI, stroke, and HF. We calculated cumulative risk for CVD events and non‐CVD death by smoking status in age‐stratified and sex‐stratified analyses. We determined the proportion of incident first events, in which the occurrence of 1 type of event (CVD event or non‐CVD death) precluded consideration of another event. The Lunn and McNeil data augmentation method was used to fit Cox proportional hazard models to estimate HRs for all CVD events combined compared with non‐CVD death as the first event in those who currently smoke, stratified by index age group and sex.^{10, 11} We also used the method described by Fine and Gray to estimate the joint and simultaneous competing hazards for each CVD event subtype and non‐CVD death as the first event in those who smoke, stratified by index age group and sex. Similarly, we used the Fine and Gray method to estimate the competing hazards for fatal and nonfatal CVD and non‐CVD death as the first event in those who smoke, stratified by index age group and sex.^{11, 12} The analyses were adjusted for age, race and ethnicity, body mass index, hypertension, hyperlipidemia (defined as total cholesterol ≥240 mg/dL or lipid‐lowering therapy), and diabetes. Mean overall survival time, mean CVD‐free survival time, and mean survival time with CVD were estimated using the Irwin restricted mean to compare compression of morbidity by smoking status. The restricted mean is mathematically equivalent to the area under the survival curve.^{11, 13} For each index age group, the restriction time point was set as 95 years of age. The survival times were then compared between individuals who do and do not currently smoke, stratified by the age groups and sex. We performed a sensitivity analysis for middle‐aged participants with a baseline exam before and after 1985 to mitigate bias related to changing secular trends in smoking status. We used a P value of <0.05 for a 2‐sided significance test. All statistical analyses were performed with SAS version 9.2 (SAS Institute) and R version 3.1.2 (The R Foundation).

A total of 106 165 people were included in the pooled cohort, including 17 205 (16.2%) Black participants and 53 527 (50.4%) women. Baseline characteristics stratified by smoking status and sex in middle‐aged, young, and older adults are described in Table 1.

In middle‐aged adults (index age of 40–59 years), there were 7002 incident fatal or nonfatal coronary heart disease events, 3527 incident fatal or nonfatal strokes, and 4559 incident HF events during the duration of the study. In addition, there were a total of 6224 deaths from all CVD causes (coronary heart disease, stroke, HF, and other) and 11 578 non‐CVD deaths in middle‐aged adults. The median (interquartile range) of follow‐up was 23 years (15–32 years). The unadjusted event rates for CVD and non‐CVD death were higher in those who reported smoking compared with those who did not in all adults (men and women across all 3 age groups; Table 2). The competing cumulative risks for both non‐CVD death and total CVD events were higher in those who reported smoking compared with those who did not in all adults (across all sex and age groups; Figure 1 and Figures S1 and S2). Similarly, the long‐term risk of non‐CVD death and CVD were higher in those individuals who reported smoking compared with those who did not (Table 3). Middle‐aged men had the largest absolute difference in long‐term risk for CVD between those who reported smoking (46%) and those who did not (36%). MI was the most common CVD subtype, with those who reported smoking having a higher long‐term risk than those who did not. Young men had the most pronounced absolute difference in long‐term risk for HF between those who reported smoking (12%) and those who did not (5%).

Total survival in follow‐up is composed of years lived free of CVD and years lived with CVD and is shown in Figure 2 and Figures S3 and S4 for middle‐aged, young, and older men and women, respectively. Across all sex and age groups, overall survival was significantly shorter in those participants who reported baseline smoking compared with those who reported not smoking, which was attributed to the significantly lower number of years lived free of CVD or health span. For example, among middle‐aged men, overall survival was significantly higher in those who reported not smoking (32.6 years) compared with those who reported smoking (27.1 years) at baseline, leading to a significantly higher number of years lived free of CVD (29.1 versus 24.0 years), indicating a delay in CVD onset by 5.1 years.

The proportions of middle‐aged men and women who had incident CVD or non‐CVD death as the first event are shown in Table 4, stratified by sex and smoking status. Incident CVD event as the first event occurred in a larger proportion of middle‐aged men and women who reported smoking (52.1% and 44.1%, respectively) compared with those who reported not smoking (49.5% and 41.2%, respectively). Similarly, non‐CVD death as the first event was more common in middle‐aged men and women who reported smoking (36.5% and 33.9%, respectively) compared with those who reported not smoking (28.6% and 27.9%, respectively) at baseline. Similar findings were observed in young men and women (Table 4). However, among older men and women, incident CVD event as the first event occurred in a larger proportion of those individuals who reported smoking (59.2% and 57.5%, respectively) than those who reported not smoking (54.7% and 52.1%, respectively). Non‐CVD death continued to occur in a greater proportion of older men and women who smoked compared with those who did not smoke. Among all age groups, a majority of the first CVD events were fatal or nonfatal MI. Although fatal and nonfatal stroke was the next leading cause of first CVD event in the young age group, HF became the second‐leading cause in the middle‐aged and older age groups. Adjusted competing HRs as determined by the Lunn and McNeil method was higher for CVD compared with non‐CVD death in both middle‐aged and older adults who reported smoking and those who reported not smoking.

Adjusted competing HRs as determined by the Fine and Gray method for first CVD subtype or non‐CVD death in men and women across all 3 age groups are shown in Table 5 and Table S1. Individuals who reported smoking had an increased risk of non‐CVD death and overall CVD as the first event across all age and sex groups compared with those who did not report smoking at baseline. Those who reported smoking had an increased risk of both fatal and nonfatal CVD events compared with those who did not smoke across all age and sex groups. The risk of fatal CVD event was higher than the risk of nonfatal CVD event across all age and sex groups except in older women. Sensitivity analysis stratified by index year before and after 1985 demonstrated similar findings (Table S2).

Key limitations of our study include the lack of data on smoking intensity given the known graded association between number of cigarettes per day and CVD. However, there is a sharp increase in CVD risk even with low levels of cigarette exposure that we were able to quantify for baseline smoking status. There is the potential to bias our results toward the null by classifying those who smoked but quit >1 year ago as not smoking at baseline, which may be underestimating the true risk of CVD morbidity and mortality related to prior cigarette smoking exposure. Another limitation is the use of baseline risk factor levels, which does not incorporate change over time. Lastly, the long‐term data required for lifetime risk estimates lead to the inclusion of relatively older data and the potential for results to be biased by changes in secular trends, such as declines in smoking prevalence over time. However, our sensitivity analysis in recent years were consistent, and prior analyses in the LRPP have also examined differences across birth cohorts with similar findings for the relationships of each risk factor with CVD.²⁰ However, our study has many strengths including a large sample size of adults with a representative sample of women and Black participants who were free of CVD at baseline with adjudicated CVD events by type and non‐CVD death in follow‐up. Our inclusion of young adults and extended follow‐up allows us to quantify the effects of smoking at the beginning and end of adulthood.