Data
This study used pooled files from the Integrated Public Use Microdata Series (IPUMS) for NHIS surveys with information on cigarette and cigar smoking (1987, 1991, 1992, 1998, 2000, 2005 and 2010) and linked to the National Death Index (NDI) (2) to obtain follow-up from year of interview to year of death or end of follow-up, December 31, 2015. Details of the data linkage can be found in our previous study (9).
Study population
The total number of participants for all NHIS survey years was 752,153. Our analyses were restricted a priori to men age 40–79 years (n = 52,710) since cigars are rarely used by women (6), and at younger ages tobacco use may be less stable and deaths are rare (10).
About 97% of these men (n = 51,062) were eligible for mortality linkage. Men who died the same year as their survey enrollment (n=366) accrued no person-years, so they were not eligible for analysis. We also excluded men with incomplete information on cigarette or cigar use or demographic characteristics (i.e. race/ethnicity, education and marital status), so the final sample for our analyses was 43,202 men (age 40-59 years, n = 27,229 and age 50-79 years, n = 15,973) with 588,761 person-years and 14,657 deaths.
Measures
Tobacco Status
The main predictor of mortality outcomes was cigarette and cigar smoking status at the survey enrollment. We used standard definitions for cigarette smokers. Never cigarette smokers had never smoked 100 cigarettes in their lifetime. Current cigarette smokers had smoked at least 100 cigarettes and smoked every day or some days at the time of the survey. Former cigarette smokers had smoked 100 cigarettes but did not smoke at the time of the survey.
During the time frame of this study, NHIS surveys collected little or no information about cigar types (2). Prior to 2000, NHIS surveys asked respondents “Have you ever smoked cigars?” The NHIS 2000 survey asked “Have you ever smoked a cigar?” and added an instruction to “Include small, thin, cigars called 'cigarillos', 'puritos' or 'chicos', that are wrapped in tobacco leaf rather than paper, and are made by machine or handrolled.” In 2005 and 2010, the NHIS asked “Have you ever smoked a cigar EVEN ONE TIME?” In all years participants were then asked “Have you smoked at least 50 cigars in your entire life?” (2)
We defined never cigar smokers as those who had never smoked at least 50 cigars in their lifetime. Current cigar smokers had smoked at least 50 cigars and smoked every day or some days at the time of the survey. Former cigar smokers had smoked at least 50 cigars but did not smoke at the time of the survey.
Next, we constructed 9 categories using cigarette and cigar status: 1- never smokers (never cigarette or cigar), 2- never cigarette and current cigar (i.e. exclusive cigar), 3- never cigarette and former cigar, 4- current cigarette and never cigar (i.e. exclusive cigarette), 5- current cigarette and cigar (i.e. dual users), 6- current cigarette and former cigar, 7- former cigarette and never cigar, 8- former cigarette and current cigar, and 9- former for both products.
Individual characteristics
We included the following characteristics as confounders: age, race/ethnicity (non-Hispanic white, non-Hispanic black, other), marital status (never married, married, divorced/separated, widowed), educational attainment (< high school, high school, some college, college and higher), family income ($0-$34,999, $35,000-$74,999, ≥$75,000), region of residence (Northeast, South, Midwest, West) and survey year.
Age, race/ethnicity, socioeconomic and marital status are well established factors in adult mortality studies (11). For example, Hispanics have lower adult mortality rates than non-Hispanic whites even though they have more disadvantaged conditions (11, 12). There is substantial evidence that marriage is associated with lower mortality (13-15). For example, non-married white men have elevated risks for the mortality from all-causes, cardiovascular diseases, cancers and other diseases than married white men, especially at ages 45-64 years (15).
A recent study using the NHIS surveys linked to the National Death Index suggested that education has an inverse effect on mortality, but only after middle age, around 55 years (16). Similarly, there is evidence in the literature suggesting that mortality differs by level of income (11, 17). Specifically, people with lower income have higher mortality than those with higher income, controlling for health status (17). Mortality rates vary by geographic locations (18,19), and the inclusion of survey years was an attempt to capture any variations due to unobserved characteristics.
Mortality outcomes
We examined all-cause and cause-specific mortality from heart diseases (I00-I09, I11, I13, and I20-I51), malignant neoplasms (C00-C97), chronic lower respiratory diseases (J40-J47), and cerebrovascular disease (I60-I69). In addition, we combined those diseases with diabetes mellitus (E10–E14) and influenza/pneumonia (J09–J18) to make a category called smoking-related diseases similar to, but somewhat broader than those recognized by the Surgeon General (20). This category was mutually exclusive and exhaustive with respect to all other causes, which consisted of accidents (V01-X59,Y85-Y86), Alzheimer's disease (G30), nephritis, nephrotic syndrome and nephrosis (N00-N07, N17-N19, N25-N27) and all other residual causes (9).
Statistical analysis
Cox proportional hazards models were used to examine the associations between cigar status and mortality outcomes based on underlying causes of death from the 10th revision of International Statistical Classification of Diseases and Related Health Problems (ICD-10)(21), reported as hazard ratios (HRs, with 95% confidence intervals, CI) with never smokers as the referent group. Follow-up, in years between survey enrollment and death or survival until December 31, 2015, ranged from 1 to 28 years (mean= 13.6 years; median = 14 years, standard deviation = 7.3 years).
We estimated HRs adjusted for age, race/ethnicity, marital status, education, income, region, survey year, BMI categories and self-reported health status We included an indicator for missing family income and BMI categories, and we applied sample weights adjusted for NDI linkage eligibility in all regression models. Results are reported separately for younger (ages 40-59 years), older (age 60-79 years) and pooled age groups.