Female vulnerability to the effects of smoking on health outcomes in older people.

Cigarette smoking is among the leading risk factors for mortality and morbidity. While men have a higher smoking prevalence, mechanistic experiments suggest that women are at higher risk for health problems due to smoking. Moreover, the comparison of smoking effects on multiple conditions and mortality for men and women has not yet been done in a population-based group with race/ethnic diversity. We used proportional hazards models and restricted mean survival time to assess differences in smoking effects by sex for multiple health outcomes using data from the U.S. Health and Retirement Study (HRS), a population-representative cohort of individuals aged 50+ (n = 22,708, 1992-2014). Men had experienced more smoking pack-years than women (22.0 vs 15.6 average pack-years). Age of death, onset of lung disorders, heart disease, stroke, and cancer showed dose-dependent effects of smoking for both sexes. Among heavy smokers (>28 pack-years) women had higher risk of earlier age of death (HR = 1.3, 95%CI:1.03-1.65) and stroke (HR = 1.37, 95%CI:1.02-1.83). Risk of cancer and heart disease did not differ by sex for smokers. Women had earlier age of onset for lung disorders (HR = 2.83, 95%CI:1.74-4.6), but men risk due to smoking were higher (Smoking-Sex interaction P<0.02) than women. Passive smoke exposure increased risk of earlier heart disease (HR = 1.33, 95%CI:1.07-1.65) and stroke (HR:1.54, 95%CI:1.07-2.22) for non-smokers, mainly in men. Smoking cessation after 15 years partially attenuated the deleterious smoking effects for all health outcomes. In sum, our results suggest that women are more vulnerable to ever smoking for earlier death and risk of stroke, but less vulnerable for lung disorders. From an epidemiological perspective, sex differences in smoking effects are important considerations that could underlie sex differences in health outcomes. These findings also encourage future mechanistic experiments to resolve potential mechanisms of sex-specific cigarette smoke toxicity.

Implications

This study identifies new sex differences in health outcomes due to smoking exposures. Sex differences in smoking hazards are often missed in epidemiological studies. Specifically, interactions between sex and smoking dosage have not been fully examined with respect to differences in health outcomes in aging. Several rodent studies have evaluated sex differences in responses to cigarette smoke to support further examination of these interactions in humans. Critical gaps for future studies include sex differences in smoking responses at different life stages (development vs adulthood vs aged); and the molecular mechanisms of sex interaction with cigarette smoke.

Introduction

Smoking is a leading cause of global mortality (6.5 million excess deaths) [1]. While men exhibit higher prevalence of smoking compared to women, studies have shown that women smokers have worse outcomes. For instance, women smokers with lung cancer show higher DNA adducts and mutation in P53 gene [2]. Sex-specific biological effects of cigarette smoke are supported by experimental studies in mice. Chronically exposed female mice showed significantly greater deficits than males in lung airway remodeling, increases in biomarkers for oxidative stress, inflammation, [3] and allergic reactivity [4]. In epidemiological studies, female smokers have shown higher risk for coronary heart disease [5], stroke [6], lung cancer [2, 7], bladder cancer [8], and chronic obstructive pulmonary disease [9]. However, epidemiological findings are mixed [10, 11]. Such discrepancies may be attributable to differences in the method of quantifying smoking exposure, use of cross-sectional data, small sample sizes, and use of younger cohorts when studying aging-related diseases.

Smoking hazards are dose-dependent [12], but we lack clear information on sex differences in the dose-dependence [8]. Some studies have found excess risk for men in all-cause mortality when cigarette dosage was quantified as the number of current cigarettes smoked per day [13]. In contrast, other reports on both sexes did not show a clear sex difference [14, 15].

Gender-specific health outcomes are even more obscure for individuals exposed to second-hand smoke, which is estimated to cause 650,000 deaths globally [12]. The few studies examining sex differences in passive smoking effects are inconsistent. For example, men were more vulnerable than women to passive smoking effects on risk of stroke in one study [16], whereas the opposite was reported in others [17, 18].

The current study examines the sex-specific smoking effects on aging-related health outcomes in the U.S. Health and Retirement Study (HRS), a large nationally representative U.S. aging study that has surveyed participants for more than 22 years. The comparison for smoking effects on males and females across multiple conditions and mortality has not yet been done in a population-based group with race/ethnic diversity. To more precisely estimate smoking dosage (vs. number of current cigarettes smoked per day), we calculated a pack-years index to represent the lifetime smoking exposure for each individual. We also examined passive smoking impacts on non-smokers with ever smoking spouses. We focus on how sex alters smoking hazards on the age of death and the onset of lung disorders, heart disease, stroke, and cancer. We further discussed the benefits of smoking cessation for different health outcomes. Evaluating these multiple outcomes allows us to examine the disease-specificity of smoking and the sex interaction. Potential biological mechanisms are discussed for interpretation of findings on sex and disease specificity.

Methods

Study population

Participants were a part of the 1992–2014 waves of the HRS, which is a nationally representative, longitudinal study of health and aging in the United States including adults (50+) and their spouses [19]. HRS is a publicly available data and no new data was collected for our analysis. All HRS participants gave their consent to enter the study. The current analysis used 12 waves of data, collected every two years from 1992 through 2014. Respondent information was obtained from the RAND 2014 HRS datafile, in addition to HRS Core data files for each wave. The data were comprised of five HRS cohorts (the original HRS cohort born 1931–41, Children Of the Depression born 1924–30, War Babies born 1942–47, Early Baby Boomers born 1948–53, and Mid Baby Boomers born 1954–59), which are recruited in the study in years 1992, 1998, 2004, or 2010 at ages 51–61 (S1–S3 Figs). The older AHEAD cohort was excluded from the analysis due to its entry into HRS after age 70. Only 122 individuals from AHEAD cohort who entered between ages 51–61 were included in the study. The original HRS cohort has up to 22 years of follow-up and the Mid Baby Boomers had up to 4 years of follow-up, which are the longest and shortest average years of follow-up in the cohorts (S1 Fig).

Outcome variables

Age of death was computed from the year of death variable in the RAND file (radyear), which is based on the National Death Index and exit interviews with proxy respondents. Two variables were constructed for each of the health conditions: prevalence, which is a binary variable for having been diagnosed with the condition; and age of onset, which is the earliest reported age of the health problem. The incidence of specific diseases was based on the question “whether or not a doctor has told the respondent that s/he had these conditions”. The age of onset was extracted from the responses to “In what year (when) did you have or were diagnosed with the condition”. For individuals with no prior history of the condition, the age at the wave of incidence was considered as the first age of onset for the condition. The health conditions include: 1) Lung disorders including chronic bronchitis and emphysema but not asthma; 2) Heart disease including heart attack, coronary heart disease, angina, congestive heart failure or other heart problems; 3) Stroke or transient ischemic attacks (TIA); and 4) Cancer which included any kind of cancer or malignant tumor, except for skin cancer.

Predictor variables

Lifetime exposure to smoking is indexed as lifetime pack-years smoked. The pack-year variable is calculated as the multiplicand of reported average number of cigarette packs smoked daily by lifetime years of smoking. Briefly, the earliest age of smoking was extracted from responses to “How many years ago”, “what year”, or “what age did you start smoking?” The age of smoking cessation was extracted from questions on “How many years ago”, “what year”, or “what age did you stop smoking?” The earliest age reported for starting and the latest age for cessation were used for each individual to calculate total years of smoking (S4 Fig). Cigarettes smoked per day were calculated from both the average of the reported number of cigarettes per day at each wave for each individual, and the maximum number of cigarettes smoked during the time in which the individual reported smoking the most (S4 Fig). Around 10,123 ever smokers (44%) had at least one missing value for calculating pack-years. These individuals included ever smokers with no reported age of start, former smokers with unknown age of cessation, and ever smokers who did not report the number of cigarettes per day. Since this was a large portion of the population, missing values were imputed using the average age of starting, age of cessation, and number of cigarettes per day calculated from the 12,585 ever smokers who had complete data. S5 Fig shows the number of individuals with or without data imputation in the analysis. The continuous pack-year variable (multiplicand of average daily packs and years of smoking) was then classified into dosage quartiles for analysis. The reported results are based on the whole data (imputed and non-imputed), however, sensitivity analysis confirmed the same pattern of findings in the sub-population with no imputation (S2 Table).

Passive smokers are defined as never smokers who lived with at least one smoker spouse. The difference between the age of smoking cessation and the latest age in the study was defined as years since smoking cessation in former smokers. This variable was converted to a categorical variable as follows: <5, 5–15, >15 years of cessation.

The demographic characteristics for sex, race (White/Caucasian, Black/African American, Other), and ethnicity (Hispanic/non-Hispanic) were extracted from the HRS RAND file. The ethnicity variable was constructed from the self-reported race and ethnicity as follows: White (non-Hispanic White), African American (non-Hispanic African American), Hispanic, and Other (non-Hispanic others).

Statistical analysis

Hazard ratios (HRs) for sex, smoking pack-years, passive smoking were calculated using Cox proportional hazard modeling for age of death, and onset of health conditions. The models estimated time after age 50 to event, and included an interaction term for sex and smoking to evaluate differences between men and women. All models were adjusted for ethnicity. The HRs were also calculated in sex-stratified models for ever and passive smoking effects by sex. The restricted mean survival time (RMST) of each group was calculated from the Cox proportional hazard model. Survival curves were fitted for the sex-stratified data on the Cox models that included a sex-smoking interaction term to estimate RMST of each group. The RMST can be interpreted as the average of event-free survival time from 50 to 85 years old age that is adjusted for ethnicity [20-22]. The analysis was done in R (version 3.5.3), using the survival package. The Cox-proportional hazard formula is: where h(t) represents expected hazard at age t; the h0(t) is the baseline hazard when all of the predictors are 0; β, coefficients; X, the predictors which included sex, ethnicity, different categories of pack-years (or passive smoke), and sex interaction with each pack-year categories (or passive smoke).

S3 and S4 Tables summarize the results of cox proportional hazard models with additional controls for other confounders including years of education, and cohort. Adjusting for these confounders did not affect the results on the smoking hazards.

Results

Demographics of the HRS sample with pack-year categories are in S1 Table. The 22 years of the study included 22,708 age-eligible individuals, ages 50–85 years. Men and women had similar age (mean baseline age, 66) and were balanced for most variables, with exceptions of a female excess for passive smokers (10.7% of women vs. 6.3% of men), non-smokers (26.8% of women vs. 19.7% of men), and medium smokers with 15–20 pack-years history (20.0% of women vs. 15.0% of men). Men had a greater percentage of very high smokers with >28 pack-years (25.7%) than women (13.3%). The health conditions with the highest and lowest prevalence were heart disease (26.0% in men, 22.0% in women) and lung disorders (5.9% in men, 7.0% in women).

Dose-dependent smoking hazard ratios (HR)

Ever smokers had consistent dose-dependent HR for earlier death, and earlier onset of lung disorders, heart disease, and stroke for both men and women. Smoking-related HR for risk of death and lung disorders were elevated even at the lowest smoking levels (0.03–15 pack-years) (Table 1). The highest HR from ever smoking was observed for lung disorders, which ranged from 3.0 (95%CI 1.85–4.96) to 7.0 (95%CI 4.42–10.92), with HRs progressively increasing with higher smoking dosage. Increase in smoking dosage from 15 to >28 pack-years caused a 20% increase in the HR of earlier onset for heart disease and stroke: an HR = 1.2 (95%CI 1.06–1.41) to 1.5 (95%CI 1.34–1.73). While smoking increased the risk of earlier onset of cancer (1.2, 95%CI 1.07–1.45), there was no clear pattern of dose-dependence.

Table 1

Hazard ratios for age of death, and age of onset of lung disorders, heart disease, stroke and cancer according to lifetime smoking level and the interaction with sex.

variable	level HR (95%CI)	Age of death	Lung disorders	Heart disease	Stroke	Cancer
Sex	Men (ref)
	Women	0.65 (0.52,0.8)***	2.83 (1.74,4.6)***	0.82 (0.71,0.95)**	0.81 (0.64,1.02)	1.14 (0.97,1.34)
ethnicity	White/Caucasian (ref)
	African American	1.64 (1.53,1.77)***	0.93 (0.8,1.07)	1.04 (0.97,1.12)	2.06 (1.87,2.27)***	0.86 (0.79,0.94)**
	Hispanic	1.04 (0.93,1.15)	0.79 (0.65,0.97)*	0.8 (0.73,0.88)***	1.34 (1.16,1.54)***	0.73 (0.66,0.82)***
	Other	1.05 (0.86,1.28)	1.34 (0.99,1.81)	1.03 (0.87,1.21)	1.17 (0.89,1.54)	0.73 (0.59,0.91)**
Pack years	Non-smokers (ref)
	Low	1.38 (1.14,1.67)**	3.03 (1.85,4.96)***	1.1 (0.95,1.27)	1.21 (0.96,1.53)	1.17 (0.99,1.39)
	Medium	1.45 (1.2,1.75)***	3.07 (1.88,5)***	1.22 (1.06,1.41)**	1.24 (0.98,1.56)	1.22 (1.03,1.44)*
	High	1.43 (1.2,1.71)***	3.34 (2.09,5.35)***	1.33 (1.16,1.51)***	1.24 (1,1.54)*	1.12 (0.96,1.31)
	Very high	2.24 (1.89,2.66)***	6.95 (4.42,10.92)***	1.52 (1.34,1.73)***	1.52 (1.23,1.87)***	1.25 (1.07,1.45)**
Women x Smoking interaction	Low	1.1 (0.85,1.43)	0.44 (0.25,0.78)**	1.19 (0.98,1.45)	0.96 (0.7,1.31)	0.87 (0.69,1.08)
	Medium	1.02 (0.79,1.31)	0.42 (0.24,0.73)**	0.95 (0.79,1.15)	1.01 (0.75,1.38)	0.79 (0.64,0.98)*
	High	1.17 (0.92,1.49)	0.45 (0.26,0.76)**	0.9 (0.75,1.08)	1.1 (0.83,1.47)	0.86 (0.7,1.06)
	Very high	1.3 (1.03,1.65)*	0.56 (0.34,0.94)*	1.07 (0.89,1.28)	1.37 (1.02,1.83)*	0.98 (0.79,1.2)
Total N		22708	21486	22708	22695	22689

* p < 0.05,

** p < 0.01,

*** p < 0.001

Sex-specific ever smoking hazards

Overall, women died at older ages than men and had later onset of specific health conditions (Table 1). Women had a lower HR for earlier death (HR = 0.65, 95%CI 0.52–0.8) and earlier diagnosis with heart disease (HR = 0.82, 95%CI 0.71–0.95). In contrast, women had a higher HR for earlier onset of lung disorders than men (HR = 2.83, 95%CI 1.74–4.6).

Gender interactions with ever smoking varied by outcome and smoking dosage. For very heavy smoking (> 28 pack-years) women had a higher HR for earlier death (HR = 1.3, 95%CI 1.03–1.65) and earlier stroke (HR = 1.37, 95%CI 1.02–1.83) than men (Fig 1A). In contrast, women smokers showed a lower risk of earlier lung disorders diagnosed than men smokers, particularly in low, medium and high smokers (Fig 1B). As noted above, women had a higher main effect of lung disorders (HR = 2.8, 95%CI 1.74–4.6) than men. Level of smoking and sex did not show a strong interaction on the outcomes of earlier onset for heart disease or cancer. The sensitivity analysis in non-imputed smoking data showed similar but stronger sex interactions with smoking hazards (S2 Table). Thus, in the main text, we describe the observed hazards in the full sample.

Fig 1

Smoking hazards are modified by sex for earlier age of death, and earlier onset of lung disorders, and stroke.

A) The smoking pack-years hazard and Kaplan-Meier survival curves for all-cause mortality in men and women between 50–85 years old age in HRS. The hazard ratio (HR) of smoking pack-years (B) and passive smoke (C) on health outcomes. HRs are calculated using Cox proportional hazard models separately by sex. *p-value <0.05 based on Wald test in the models. The significant sex differences are based on the interaction terms in the full model and are indicated in each figure. The stratified models were adjusted for ethnicity. The baseline effects of sex are reported in Tables 1 and 2.

Table 2

Hazard ratios of age of death, and age of onset of lung disorders, heart disease, stroke and cancer according to passive smoke and the interaction with sex.

	HR (95%CI)	Age of death	Lung disorders	Heart disease	Stroke	Cancer
Sex	Men (ref)
	Women	0.73 (0.54,0.98)	2.92 (1.51,5.65)**	0.89 (0.73,1.08)	1.02 (0.74,1.41)	1.33 (1.07,1.65)*
Ethnicity	White/Caucasian (ref)
	African American	1.37 (1.05,1.8)*	1.5 (0.95,2.39)	1.2 (1,1.43)*	2.93 (2.24,3.82)***	0.8 (0.64,0.99)*
	Hispanic	1.03 (0.75,1.42)	1.22 (0.72,2.05)	0.72 (0.57,0.9)**	1.48 (1.05,2.09)*	0.76 (0.6,0.96)*
	other	0.73 (0.37,1.42)	1.03 (0.37,2.83)	0.61 (0.39,0.96)*	1.16 (0.59,2.28)	0.5 (0.3,0.84)**
Smoking	Never smokers (ref)
	Passive Smokers	1 (0.72,1.38)	1.22 (0.5,2.95)	1.33 (1.07,1.65)*	1.54 (1.07,2.22)*	1.26 (0.97,1.64)
Women x passive smoking interaction		0.71 (0.46,1.1)	0.86 (0.32,2.29)	0.78 (0.58,1.05)	0.52 (0.32,0.84)**	0.69 (0.5,0.96)*
Total N		5309	5201	5309	5305	5298

* p < 0.05,

** p < 0.01,

*** p < 0.001

The order of conditions based on earlier RMST age (an indicator of average event-free age), is as follows: heart disease (men, RMST age 74.2; women, 75.3), cancer (men, 77.8; women, 77.1), death (men, 77.7; women, 79.3), stroke (men, 80.5; women, 82), and lung disorders (men, 81.1; women, 79.2) (Fig 2A). RMST analysis showed that very high smoking (>28 pack-years) has greater impact on women than men for earlier age of death (4.7 years earlier for women, 4.1 years earlier for men), and stroke (2 years earlier for women, 1.1 years earlier for men). Men showed greater vulnerability to smoking for the earlier age of onset for lung disorders, particularly in smokers with less than 28 pack-years (women, -0.8 years; men, -1.3 years). Very heavy smoking had similar effects on the earlier onset of heart diseases and cancer in both sexes (Heart diseases: women, -3.3 years, men, -3.1 years; Cancer: women -1.6 years, men -1.5 years).

Fig 2

Restricted mean survival time (RMST) of death and disease onset in men and women smokers.

The calculated RMST is based on Cox proportional hazard model of (A) smoking pack-years, and (B) passive smoke. The results are reported as RMST±SEM. The lines are showing the number of years that was reduced by smoking from the baseline RMST age in each sex.

Passive smoking hazards

Passive smoking increased the risk of earlier onset of heart disease (HR = 1.3, 95%CI 1.07–1.65) and stroke (HR = 1.5, 95%CI 1.07–2.22) (Table 2). Age of death and age of onset of lung disorders, and cancer were not affected by passive smoking. In HRS, hazards of passive smoke were close to those associated with very high smoking exposure for earlier onset for heart disease (HR Passive/Ever, 1.33/1.52; RMST Passive/Ever 77/75) and stroke (HR Passive/Ever, 1.54/1.52; RMST Passive/Ever, 82/80).

Based on Cox proportional hazard models, passive smoking effects were modified by sex only for earlier age of stroke. Only passive smoking males showed an increase in the risk of earlier age of stroke compared to females (Fig 1C). The RMST analysis showed 2 years earlier onset for heart diseases (at around age 75), and 1.1 years earlier onset for stroke (at age 81) only in male passive smokers (Fig 2B). Heart disease and stroke onset did not show any baseline differences between male and female never smokers with non-smoker spouses.

Smoking cessation and health outcomes

Because ever smokers include both active and former smokers, we conducted sensitivity analysis based on smoking status to examine the hazards of smoking and potential sex differences. Former smokers were categorized by the number of years since cessation. Former smokers with >15 years of cessation recovered from ever smoking hazards for earlier onset of stroke (HR = 0.89, 95%CI 0.79–1.10) and cancer (HR = 1.08, 95%CI 0.93–1.25) (S5 Table). This group of former smokers (>15 years cessation) were still at higher risk for lung disorders (HR = 2.67, 95%CI 1.68–4.23) and heart disease (HR = 1.22, 95%CI 1.08–1.39) than the non-smokers. Nonetheless, they were at lower risk compared to current smokers. Smoking cessation of >15 years also showed protective effects against mortality compared to non-smokers (HR = 0.48, 95%CI 0.4–0.57) (S5 Table, Fig 3A).

Fig 3

Smoking cessation >15 years reduced the hazards of smoking on age of death in both men and women.

A) Kaplan-Meier survival curves for all-cause mortality in men and women between 50–85 years for smoking cessation status. B) Histogram and average of age of smoking cessation in different groups of former smokers in HRS.

Former smokers with fewer than 15 years of cessation showed increased risks for all health outcomes compared to current smokers: earlier death (HR for current smokers vs. cessation <5 years: 1.29 vs 6.26); lung disorders (4.3 vs 8.24); heart disease (1.29 vs 1.5); stroke (1.32 vs 2); cancer (1.14 vs 1.55). Former smokers with more than 15 years since cessation, on average stopped smoking around age 55 (Fig 3B), which was at least 5 years earlier than the former smokers who reported 10–15 years of cessation.

This sensitivity analysis did not enable us to detect sex differences in the health hazards of current smokers. The effects of smoking cessation itself were modest in that they differed by sex for only some health outcomes: earlier death and risk of lung disorders (S5 Table).

Discussion

Our findings show dose-dependent sex differences in smoking for premature mortality and morbidity among a nationally-representative sample of U.S. adults. Women were more vulnerable to ever smoking for premature death and stroke incidence. In contrast, for <28 pack-years, men smokers had earlier lung disorders than women. The onset age of cancer and heart disease in smokers did not differ by sex, even for very heavy smokers. Passive smoke exposure indicated excess risk for men for earlier heart disease and stroke.

These findings for women are consistent with experimental studies. Mice chronically exposed to cigarette smoke for 6 months had greater responses among females than males for lung small airway remodeling and increased distal airway resistance. Corresponding biomarkers included 2-fold increase in 3-nitrotyrosine (oxidative stress) and 1.5-fold increase in transforming growth factor β (inflammation) [3]. In contrast, male mice of this study had higher induction of genes mediating oxidative stress responses (Nrf2, Nqo1), and detoxification (Cyp1a1, Cyp1b1). The role of sex steroids is shown by ovariectomy, which ameliorated lung airway remodeling [4]. In other studies replacement of 17βc-estradiol (E2) decreased autophagy and increased Nrf2 responses of hippocampus to cerebral ischemia [23]. Thus, steroid hormones could underlie some of these sex differences to cigarette smoke.

Analysis of older birth cohorts from 1800–1935 in US and European counties showed increasing excess adult mortality for men born after 1900 [24]. Around 30% of this excess in men’s mortality was attributed to smoking; however, this analysis was restricted to using smoking status (smokers vs non-smokers), as pack-years were not known. From this analysis of HRS, we can infer that more of the excess male mortality observed in earlier historical cohorts reflects the smoking behaviors of men. As we show that men smoked 30% more pack-years than women (S1 Table, an average of 22.0 vs 15.6). Of key relevance, the current findings now provide a characterization of individual smoking exposure and mortality for those born after 1934.

In other studies, women smokers had 2-fold more DNA adducts or frameshift mutations in the P53 gene than men smokers [2]. This higher DNA damage among women smokers was also associated with a more accelerated risk of cancer and other age-related chronic diseases [25, 26]. While the current study showed that smoking increased risk for earlier age of cancer, it did not show a clear sex interaction. In sex stratified analysis, the risk of an earlier age of cancer onset was significant only for women smokers with very high pack-years, which is consistent with the women’s excess of mutations in lung cancer cited above [2, 7]. Future studies should examine sex-smoking interactions for specific cancer types.

Lung disorders in the current study included all chronic lung conditions other than asthma, such as chronic bronchitis and emphysema. Thus, the assessment of the outcome includes chronic obstructive pulmonary disease (COPD). North American regional studies have shown men excess of COPD from cigarette smoking [27-29], which parallels our finding of greater risk for earlier lung disorders onset in men smokers. These findings contradict those in mouse models, that found only female responses in airway remodeling due to chronic exposure [3]. However, these mouse models may not represent older human ages. One concern in interpreting these findings on lung conditions is the potential under-diagnosis of COPD in North America, particularly for women [30]. Further experimental studies may resolve the sex-specific cigarette smoke effects for the onset of lung disorders at later ages.

Hazards of passive smoke were comparable to very high pack-years for earlier onset for heart disease and stroke. Importantly, the chemical composition of side-stream (passive) smoke differs from main-stream (inhaled) smoke, with much lower density of particles and gases, by 10 to-100 fold [31]. However, per cigarette, side-stream smoke has 2- to 30-fold higher concentrations of nicotine and organic toxins (benz[a]pyrene, and other polycyclic aromatic hydrocarbons); volatile hydrocarbons (ethene, propene); and N-nitrosamines; and gases (carbon monoxide, nitric oxide) [31, 32]. Our recent studies on ambient air pollution particulate matter showed that chemical and physical characteristics of the particles can largely affect the toxicity of air pollution samples with the same mass concentration [33, 34]. Thus, it is not surprising that main-stream and second-hand smoke would diverge in toxicity and have different gender-specific effects. In HRS, men exposed to passive smoke had a higher risk of heart disease and stroke than women. In a mouse model of prenatal exposure to passive smoke, males had greater alteration in adult lung tidal volume [35].

Despite the well-documented magnitude of cigarette smoking hazards and decades of research on potential carcinogens and other toxins, we have still a surprisingly limited understanding of sex interactions. This analysis revealed that sex-specific smoking effects depend on the aging condition. The parallel experimental findings for sex differences in mice highlight the possibility of broadly shared biological mechanisms. Further population-level analyses are needed on sex differences in cigarette toxicity that may be shared with air pollution, including diseases of arteries, lungs, and brain [36]. For example, lung cancer risk scales with pack-years and air pollution levels of PM2.5 independently, while the combination has multiplicative synergies [37]. Further experimental studies of developmental and adult exposure to cigarette smoke could include mice with transgenes for detoxification gene variants associated with vulnerability to air pollution, e.g. alleles of the glutathione S-transferase gene GSTP1 [38] and MET receptor of tyrosine kinase [39].

In the last part of our analysis, we examined smoking hazards in former smokers by duration since cessation. These results suggested that smoking cessation before age 60 could attenuate the smoking hazards for all our target health outcomes. However, quitting smoking after age 60 was associated with additional stress and increased the risk of earlier death and other outcomes compared to current smokers. This aging effect could represent the declining regenerative capacity of many tissues after middle-age [40, 41]. Little is known of how smoking interacts with basic aging processes of cell senescence and systemic inflammation.

We must also consider whether spontaneous smoking cessation by older heavy smokers was due to their recognition of underlying diseases. Former smokers with less than 20 years of cessation had higher incidence of lung or prostate cancer than current smokers did [42-44]. In the Canadian National Breast Screening Study, which includes 49165 women aged 40–59, those who quit smoking at older ages had a higher relative risk for lung cancer mortality than current smokers with high number of cigarettes per day or years of smoking [44].

Chronic damage from heavy smoking cannot be recovered from simply by cessation. An experimental study of male hypertensive rats modeled the long-term sequela of smoking (up to at age 10 months) and identified persistent damage effects that did not reverse 10 months after cessation [45]. Changes included higher mortality at age 21 months, and persistent inflammation in several lung regions. Examining the smoking cessation in HRS enabled us to characterize the complexity of smoking effects on health in the older population. However, we are not able to draw strong conclusions about sex differences in the benefits of smoking cessation. Understanding this relationship necessitates a larger cohort that allows the inclusion of both smoking status and smoking pack-years in one model. The benefits of smoking cessation could depend on several factors such as age, sex, smoking dosage, underlying disease, and other environmental risk factors.

Further studies of sex differences in response to cigarette smoke should consider physiologically distinct stages of the lifespan: development (0–18), young adulthood (18–35), middle-age including post-menopause (36–60), and older ages when chronic diseases increase exponentially (60+). For example, a mouse model of air pollution toxicity had attenuated responses of lung and brain by middle-age (18 months) [46, 47]. We should also consider potential middle-aged survivor bias in older age cohorts such as HRS, which combines individuals in late-middle age and older ages.

The current study examined the association of smoking and several aging-related conditions in a nationally representative dataset with follow-up over two decades and a robust measure for exposure to different smoking doses over a lifetime. Cigarette smoke effects may interact with diverse socioeconomic and environmental factors including birth cohort, diet and lifestyle, and exposure to air and noise pollution. In HRS, we showed that additional controls for years of education and birth cohort do not alter the smoking hazards or sex-specificity of the findings (S3 and S4 Tables). However, this conclusion is obscured by the complexity of the interaction between socio-economic status, cultural habits, genetic diversity, and cigarette smoke toxicity. Moreover, our results can be affected by the quality of self-reported smoking data. For example, former smokers reported a slightly higher history for the number of cigarettes per day than current smokers (0.5 higher average cigarettes per day, S4 Fig). It is unclear if such reporting captures true differences, or is due to reporting bias from former or current smokers. A series of controlled experiments can disentangle the contribution of these individual confounders on the smoking hazards and also contribute to the development of a biomarker that can estimate true smoking dosage, such as DNA methylation levels [48]. Future analyses of smoking effects on aging-related conditions should also consider interactions with other biological (e.g. age) and environmental (e.g. outdoor and indoor ambient air pollution) factors.

Despite the partially successful decrease of cigarette smoking in most developed countries, Asian and African markets for tobacco are still growing, which anticipates the need for new therapeutic and preventive measures with sex-specificity by age and life stage.

Histogram of number of years of in different cohorts enrolled in HRS.

(DOCX)

Click here for additional data file.

Histogram of year of birth of different cohorts enrolled in HRS.

(DOCX)

Click here for additional data file.

Histogram of year of recruitment for each cohort in HRS.

(DOCX)

Click here for additional data file.

Box plot of average number of cigarettes per day, smoking years, and pack years of 17,399 men and women current or former smokers in HRS.

The significance of difference between current and former smokers was assessed by t-test. * p<0.05.

(DOCX)

Click here for additional data file.

Histogram of the number of individuals with imputations used in creating the pack year variable.

In total, 12585 respondents had complete data, and 10123 respondents had at least one missing variable that had to be imputed (e.g. age of start smoking, age of quitting smoking, average daily cigarettes smoked).

(DOCX)

Click here for additional data file.

Demographic characteristics of the HRS sample, 1992–2014.

(DOCX)

Click here for additional data file.

Hazard ratios of age of death, and age of onset of lung disorders, heart disease, and stroke according to ever smoking and the interaction with gender in data with no imputation.

This subpopulation showed a similar pattern but stronger sex-smoking interaction for the age of death, heart disease, and cancer. Thus, the main text reported the results of the whole population.

(DOCX)

Click here for additional data file.

Confounders: Years of education, Ethnicity.

(DOCX)

Click here for additional data file.

Confounders: Years of education, Cohort, Ethnicity.

(DOCX)

Click here for additional data file.

Hazard ratios of age of death, and age of onset of lung disorders, heart disease, stroke and cancer according to years since quitting smoking and the interaction with sex.

(DOCX)

Click here for additional data file.

18 Mar 2020

PONE-D-20-03538

Female vulnerability to the effects of smoking on health outcomes in older people

PLOS ONE

Dear Dr. Haghani,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

I agree with the reviewers that there are a number of strengths to the manuscript, but also some concerns that would need to be satisfactorily addressed before it could be accepted for publication. In particular, both reviewers' comments about providing more detail about the data and the analyses, and the recommendations to account for potential sources of bias, are important to address.

We would appreciate receiving your revised manuscript by May 02 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Neal Doran

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at http://www.plosone.org/attachments/PLOSOne_formatting_sample_main_body.pdf and http://www.plosone.org/attachments/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please ensure you have thoroughly discussed any potential limitations of this study within the Discussion section, for example the impact of confounding factors.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: No

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Thank you for giving me the chance to review this manuscript.

The study investigated the association between smoking and different health outcomes in a large population-based sample. Furthermore, sex-specific and dose-dependent smoking effects were evaluated. The paper is well written and contributes with interesting results to the previous literature. However, some points need to be clarified or could be discussed in more detail. Please see below for my specific comments.

Specific comments

• It could be mentioned in the abstract which statistical methods have been applied.

• In the abstract, HR 0.44-0.56 for the risk of lung disorders in women smokers is stated. This needs to be checked.

• What was the reason to exclude asthma? Only lung disorders such as chronic bronchitis and emphysema were considered.

• It is mentioned in the methods section that 12 waves of the Health and Retirement Study (1992-2014) were used for the current analyses. Does this imply that 12 follow-ups were available? Maybe some more information on the different waves and times of follow-up could be given.

• Would it be possible to specify the cause of death? Besides smoking there might be other risk factors which could lead to an early death. Have further potential confounding factors such as socio-economic status, comorbidities and other lifestyle-related factors been tested in the regression models?

• It is stated in the methods section that missing smoking values were imputed using average values. What was the proportion of imputed values? If the proportion was quite high it would be good to perform sensitivity analyses only including those with self-reported smoking information (i.e. exclusion of subjects with imputed values).

• The exact definition of the smoking categories is not entirely clear (presented in Table S1). The percentages for never, passive and active smokers sum up to 100% for men and women, respectively, meaning that these three groups are disjunctive. However, passive smokers are per definition also never smokers. Does this imply that the never smoker group comprises only never smokers without passive smoke exposure? It seems that former smokers are included in the active smoking category. Have it been investigated whether there are different health effects when differentiating former (e.g. quit smoking a long time ago) and current smokers?

• The resolution of the Kaplan-Meier curve is not sufficient as the legend is not readable.

• Potential biological mechanisms behind the sex-specific smoking effects could be discussed in more detail.

Reviewer #2: Comments to the Author:

This study examines sex-specific vulnerability to health outcomes from smoking at older ages in the HRS. They find that older women are more susceptible to earlier death and risk of stroke but are less likely to report a doctor-diagnosed lung disorder. Sex-specific incidence and comorbidity of smoking and other substance use disorders is critically under-examined in population studies, and the paper makes a novel and important contribution to this literature. However, I have some major and minor concerns that need to be addressed before I can recommend it for publication. These concerns are highlighted below.

Major Concerns:

My main concern is whether the study is well powered, particularly for the sex-specific findings for females in the “very high pack years” group who comprise only 13.3% of female smokers. These findings form the basis of the authors’ claim that older female smokers are more susceptible to earlier death and risk of stroke so additional evidence and discussion of power are needed.

The authors conduct several sex-specific tests across multiple outcomes and should adjust p-value significance thresholds for multiple hypothesis testing.

There are huge gradients in smoking by education and birth cohort. These variables are also highly correlated with life expectancy. The authors either need to control for education and birth cohort in their models or provide a strong justification for why these controls are not included.

The authors do not address how mortality selection affects their results. For example, it’s possible that only the healthiest male smokers survived past age 50 and this explains why we see an earlier onset of death for female smokers. This is a major limitation of the HRS data that at a minimum needs to be discussed if not corrected for somehow using a weighting strategy (e.g., see Domingue et al., 2017). Because of this, I think it would also strengthen the paper immensely if the authors were able to replicate their findings in another dataset that is slightly younger (e.g., UK Biobank).

[Domingue, B. W., D. W. Belsky, A. Harrati, D. Conley, D. R. Weir, and J. D. Boardman. 2017. "Mortality selection in a genetic sample and implications for association studies." International Journal of Epidemiology 46 (4):1285-1294.]

Along these lines, I would like to know which HRS cohorts were used in the study. For example, respondents in the AHEAD cohort were over the age of 70 when they entered the HRS. Including these cohorts in the analysis may exacerbate any bias from mortality selection.

It was unclear to me why active female smokers were more likely to have a stroke whereas passive male smokers were more likely to have a stroke, and I did not understand the authors’ rational behind these findings on p. 10 (“Thus it is not surprising the main-stream and second-hand smoke would diverge in toxicity, and have different gender interactions”). Why would we expect level of toxicity to affect sex differences in an outcome? Clarification and more discussion are needed here.

I like how the authors cite findings from animal models to provide some biological basis for their findings. However, the findings from mouse models that are reported do not seem to match the findings of the study. The authors claim that this contradiction is a result of the mouse models not representing “older human ages.” Please elaborate or provide more evidence for why the biological process of aging may change the sex-specific interactions we observe.

I like how the authors present findings for active and passive smoking exposure, and I thought the way they used the HRS data to calculate pack years was an excellent utilization of the data that’s available. However, if I remember correctly, there is a lot of missingness for the age of onset and cessation questions in the HRS. How many respondents in the study have imputed data for pack years? How might this affect the results?

In Table S1, the authors report that 80.3% of males and 73.2% of females are active smokers in the HRS. These percentages seem way too high and I would double-check these numbers (other publications report closer to 25% for males and 24% for females in the HRS).

In the HRS, former smokers are asked their max CPD over their entire smoking history and current smokers are asked their current CPD. Since individuals tend to ramp down their average daily number of cigarettes as they age, I would guess that active smokers on average have a lower CPD than former smokers. The authors should discuss how this reporting difference in the HRS may affect their overall results.

Minor Concerns:

On page 3, I found the sentence “this study highlights the biological basis of gender differences in smoking effects” to be misleading. This study does not explore the actual biological underpinnings of sex differences in smoking comorbidities.

Throughout, the authors use the word gender. If the hypothesis is that observed population-level differences have a biological basis I would use the word sex instead of gender.

I may have missed it, but what methods were used for imputation of the pack years variable?

Could the passive smoking variable also be calculated in pack years? If you know how long an individual has been married to their spouse it seems like you could multiply their spouse’s average packs per year times the number of years they’ve been married. Not sure if this would work in the HRS, but it seems like these findings may also be dose dependent.

The authors should provide a discussion of the limitations of their study.

The figures were fuzzy and I had difficulty reading them. Please provide higher quality files in future revisions.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

7 Apr 2020

Reviewer #1:

Thank you for giving me the chance to review this manuscript.

The study investigated the association between smoking and different health outcomes in a large population-based sample. Furthermore, sex-specific and dose-dependent smoking effects were evaluated. The paper is well written and contributes with interesting results to the previous literature. However, some points need to be clarified or could be discussed in more detail. Please see below for my specific comments.

Specific comments

1. It could be mentioned in the abstract which statistical methods have been applied.

Answer: Information on statistical methods was added to the abstract.

2. In the abstract, HR 0.44-0.56 for the risk of lung disorders in women smokers is stated. This needs to be checked.

Answer: We apologize for the confusion. The sentence was revised for clarification, and in order to refer to the main finding of sex differences found. “Women had earlier age of onset for lung disorders (HR=2.83, 95%CI:1.74-4.6), but men risk due to smoking were higher (Smoking-Sex interaction P<0.02) than women".

3. What was the reason to exclude asthma? Only lung disorders such as chronic bronchitis and emphysema were considered.

Answer: Data for asthma was not available in HRS. HRS asked about a limited number of diseases or conditions regarded as having age-related onset.

4. It is mentioned in the methods section that 12 waves of the Health and Retirement Study (1992-2014) were used for the current analyses. Does this imply that 12 follow-ups were available? Maybe some more information on the different waves and times of follow-up could be given.

Answer: New figures that summarize the number of individuals in each cohort, year of the entry, and number of years of follow-up were added to the supplementary data (Fig. S1-3). The HRS began in 1992 and 1993 with a cohort born from 1931 to 1941 (original HRS); subsequently 5 different cohorts have been added ending with the mid baby boomers. The HRS cohort can have up to 12 follow-up interviews and those who were the youngest, added in 2010, could only have 2 follow-ups. Since, most of the AHEAD cohort was entered at ages older than 70, they were predominantly excluded from the analysis. But a small number of AHEAD respondents who were age eligible were included. We have clarified this in the text.

5. Would it be possible to specify the cause of death? Besides smoking there might be other risk factors which could lead to an early death. Have further potential confounding factors such as socio-economic status, comorbidities and other lifestyle-related factors been tested in the regression models?

Answer: Unfortunately, we do not have access to cause of death data to answer the reviewer question. However, analysis of multiple health outcomes allows us to conclude that cigarette smoke can increase the risk of death from multiple chronic diseases.

Because it is possible that other variables confound the analysis, we have added additional analysis including years of education and cohort as potential confounders of the association between smoking and health outcomes (Table S2-3). Controlling for these variables did not alter our original results or conclusions. We have added this to the 2nd to last paragraph in the discussion.

6. It is stated in the methods section that missing smoking values were imputed using average values. What was the proportion of imputed values? If the proportion was quite high it would be good to perform sensitivity analyses only including those with self-reported smoking information (i.e. exclusion of subjects with imputed values).

Answer: We appreciate the reviewer’s comment and add information about data imputation in the supplementary data. The sensitivity analysis (Table S2) shows the same smoking-sex interaction as discussed in the paper. The analysis based on non-imputed data showed stronger p-values for sex-smoking interaction but did not change any of the conclusions or main findings. Thus, we report and refer to those results in the supplement.

7. The exact definition of the smoking categories is not entirely clear (presented in Table S1). The percentages for never, passive and active smokers sum up to 100% for men and women, respectively, meaning that these three groups are disjunctive. However, passive smokers are per definition also never smokers. Does this imply that the never smoker group comprises only never smokers without passive smoke exposure? It seems that former smokers are included in the active smoking category. Have it been investigated whether there are different health effects when differentiating former (e.g. quit smoking a long time ago) and current smokers?

Answer: The demographic table was updated for clarification. “Active smokers” was replaced by “Ever smokers” in Table S1.

We took advantage of the opportunity and added additional analysis about effects of smoking cessation effects as suggested by the reviewer to the results section, Table 3, figure 3, and additional discussion points. Only former smokers who quit smoking for more than 15 years partially had decrease of hazards rate for health outcomes than active smokers.

8. The resolution of the Kaplan-Meier curve is not sufficient as the legend is not readable.

Answer: We apologize for this. It appears that the PlosOne website reduced the quality of the figures during conversion to PDF. We replaced the figures as suggested. Hopefully, they will have better quality in the revision. Please download the image separately from the link on the corner of the PDF file to see the image in original resolution.

9. Potential biological mechanisms behind the sex-specific smoking effects could be discussed in more detail.

Answer: We have added to our discussion (2nd paragraph) the potential role of sex hormones in environmental responses.

Reviewer #2:

This study examines sex-specific vulnerability to health outcomes from smoking at older ages in the HRS. They find that older women are more susceptible to earlier death and risk of stroke but are less likely to report a doctor-diagnosed lung disorder. Sex-specific incidence and comorbidity of smoking and other substance use disorders is critically under-examined in population studies, and the paper makes a novel and important contribution to this literature. However, I have some major and minor concerns that need to be addressed before I can recommend it for publication. These concerns are highlighted below.

Major Concerns:

1. My main concern is whether the study is well powered, particularly for the sex-specific findings for females in the “very high pack years” group who comprise only 13.3% of female smokers. These findings form the basis of the authors’ claim that older female smokers are more susceptible to earlier death and risk of stroke so additional evidence and discussion of power are needed.

Answer: We understand the reviewer’s concern. As shown in Table S1 of the supplement, there was sufficient sample size throughout the range of levels of smoking. The 13.3% of females who are classified as very high smokers consist of 1560 women. This is a large number of individuals with up to 22 years of follow up after entry into the study of this population-representative sample. We added additional discussion points on the interaction of age and cigarettes, and the role of sex hormones in response to cigarette smoke for better clarification and interpretation on sex-specific findings.

2. The authors conduct several sex-specific tests across multiple outcomes and should adjust p-value significance thresholds for multiple hypothesis testing.

Answer: Since, these are independent outcomes examining a pattern of morbidity and mortality with aging, we did not adjust p-values for multiple test corrections. However, most of the findings are quite strong and persist with multiple test correction for 5 interaction tests. Thus, we emphasize the pattern of effects and report the p-values, so that readers can extract the most important differences and extrapolate the findings to those relevant to other research.

3. There are huge gradients in smoking by education and birth cohort. These variables are also highly correlated with life expectancy. The authors either need to control for education and birth cohort in their models or provide a strong justification for why these controls are not included.

Answer: Additional analysis was added to Table S3-4 to examine whether the inclusion of years of education and birth cohort would change the association between smoking, sex and health outcomes. The analysis revealed that these confounders do not alter the smoking hazards or sex-interactions. As mentioned in our response to reviewer 1, comment 5, we added additional discussion points to the limitations about disentangling the effects of confounders on smoking hazards.

4. The authors do not address how mortality selection affects their results. For example, it’s possible that only the healthiest male smokers survived past age 50 and this explains why we see an earlier onset of death for female smokers. This is a major limitation of the HRS data that at a minimum needs to be discussed if not corrected for somehow using a weighting strategy (e.g., see Domingue et al., 2017). Because of this, I think it would also strengthen the paper immensely if the authors were able to replicate their findings in another dataset that is slightly younger (e.g., UK Biobank).

Answer: We thank the reviewer for this suggestion. A discussion paragraph was added to the paper about the potential selection effects resulting from lack of survival in HRS. Our sample includes primarily people who joined the HRS in their 50s, rather than those who joined at older ages (see supplemental Figures S1-S3). Our sample is different from the HRS genetic data that was used in the Domingue et al., 2017 paper. One of the major issues in that paper was that people who joined in 1992 had to survive to 2006 or 2008 to be included in the genetic data.

Mortality before age 50 from smoking is fairly unusual and studying mortality among people in this age would be a very selected group. Studying smoking effects in younger ages is outside the scope of our paper.

5. Along these lines, I would like to know which HRS cohorts were used in the study. For example, respondents in the AHEAD cohort were over the age of 70 when they entered the HRS. Including these cohorts in the analysis may exacerbate any bias from mortality selection.

Answer: The reviewer is correct in indicating we should have made this clear in the earlier draft. We have added material in the text and supplementary figures to clarify the sample used (Figures S1-S3). We did not include the AHEAD in this analysis.

6. It was unclear to me why active female smokers were more likely to have a stroke whereas passive male smokers were more likely to have a stroke, and I did not understand the authors’ rational behind these findings on p. 10 (“Thus it is not surprising the main-stream and second-hand smoke would diverge in toxicity, and have different gender interactions”). Why would we expect level of toxicity to affect sex differences in an outcome? Clarification and more discussion are needed here.

Answer: We expanded this discussion on the chemical differences between main- and side-stream smoke (page 12) to answer this question.

7. I like how the authors cite findings from animal models to provide some biological basis for their findings. However, the findings from mouse models that are reported do not seem to match the findings of the study. The authors claim that this contradiction is a result of the mouse models not representing “older human ages.” Please elaborate or provide more evidence for why the biological process of aging may change the sex-specific interactions we observe.

Answer: We added additional discussion points and now describe how age can alter the sex differences in cigarette toxicity (beginning on page 13). We highlighted how sexual hormones can affect the biological responses to the environmental stressors. Thus, sex-cigarette smoke interaction can differ at different life stages (e.g. pre and post menopause).

8. I like how the authors present findings for active and passive smoking exposure, and I thought the way they used the HRS data to calculate pack years was an excellent utilization of the data that’s available. However, if I remember correctly, there is a lot of missingness for the age of onset and cessation questions in the HRS. How many respondents in the study have imputed data for pack years? How might this affect the results?

Answer: We added supplementary information on the imputation approach used and provide information on how much data were imputed. Comparative analysis using all the data and the data with no imputation was added to the supplement, Figure S5. The analysis based on non-imputed data actually showed stronger p-values for sex-smoking interaction, but did not change any of the conclusions or main findings. Thus, we report the unimputed results in the supplement and explain this in the main paper.

9. In Table S1, the authors report that 80.3% of males and 73.2% of females are active smokers in the HRS. These percentages seem way too high and I would double-check these numbers (other publications report closer to 25% for males and 24% for females in the HRS).

Answer: We apologize for using the word “active”. This was a mistake as this number includes both current and former smokers so these would be “ever” or current smokers in conventional terminology. The table was revised for clarification.

10. In the HRS, former smokers are asked their max CPD over their entire smoking history and current smokers are asked their current CPD. Since individuals tend to ramp down their average daily number of cigarettes as they age, I would guess that active smokers on average have a lower CPD than former smokers. The authors should discuss how this reporting difference in the HRS may affect their overall results.

Answer: As the reviewer suggested, we added a Figure S4 to compare reports of the number of cigarettes per day between current and former smokers. It is true that there is a modest excess CPD in former smokers. We include this as a limitation in the discussion.

Minor Concerns:

11. On page 3, I found the sentence “this study highlights the biological basis of gender differences in smoking effects” to be misleading. This study does not explore the actual biological underpinnings of sex differences in smoking comorbidities.

Answer: The sentence was modified.

12. Throughout, the authors use the word gender. If the hypothesis is that observed population-level differences have a biological basis I would use the word sex instead of gender.

Answer: Gender was replaced by sex throughout the text.

13. I may have missed it, but what methods were used for imputation of the pack years variable?

Answer: We apologize if the text was not clear. We have expanded the methods section to describe the imputation method. In addition, Figure S5 and Table S2 were added to the supplement providing details on the imputation and sensitivity analysis using only the non-imputed sample.

14. Could the passive smoking variable also be calculated in pack years? If you know how long an individual has been married to their spouse it seems like you could multiply their spouse’s average packs per year times the number of years they’ve been married. Not sure if this would work in the HRS, but it seems like these findings may also be dose dependent.

Answer: This was our original intent. However, it was not possible to calculate the dosage of secondhand smoke reliably because of the large number of people with multiple spouses and varying lengths of marriage.

15. The authors should provide a discussion of the limitations of their study.

Answer: We added further discussion of our limitations.

16. The figures were fuzzy and I had difficulty reading them. Please provide higher quality files in future revisions.

Answer: We apologize for this. It appears that the PlosOne website reduced the quality of the figures during conversion to PDF. We replaced the figures as suggested. Hopefully, they will have better quality in the revision. Please download the image separately from the link on the corner of the PDF file to see the image in original resolution.

Submitted filename: Respond to reviewers.docx

Click here for additional data file.

14 May 2020

PONE-D-20-03538R1

Female vulnerability to the effects of smoking on health outcomes in older people

PLOS ONE

Dear Dr. Haghani,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

More specifically, we ask that you carefully consider Reviewer 1's comments about definitions and analytic suggestions.

We would appreciate receiving your revised manuscript by Jun 28 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Neal Doran

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I thank the authors for the thorough revision of the manuscript and conducting several additional analyses. I just have one further comment regarding the analyses on smoking cessation. Investigating sex-specific effects of smoking cessation on different health outcomes might be beyond the scope of this manuscript and its aims. As in the first version of the manuscript the exact definition of “active smokers” (now indicated as “ever smokers”) was not entirely clear and the assessment of lifetime pack-years smoked differs by smoking status (average number of cigarettes per day vs. maximum number of cigarettes), sensitivity analyses investigating whether the effects of smoking on health outcomes differ by smoking status might have been reasonable. Therefore, ever smokers could have been separated into former and never smokers (taking lifetime pack-years into account) independent of years since smoking cessation. Table 3 in the manuscript shows some quite high hazard ratios with broad confidence intervals. This might be due to the small sample sizes, esp. for women who quit smoking <5 or 5-15 years ago and who have lung disorders. Thus, these results should be interpreted with caution. I would suggest to move these results to the supplement and not put too much emphasis on this kind of analyses. Rather, I would just add this as kind of sensitivity analyses (as was done e.g. for the comparison of imputed vs. non-imputed pack-years data) to see whether the effects change and just mention this briefly. Nevertheless, I would like to thank the authors for all their efforts.

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

14 May 2020

Reviewer #1: I thank the authors for the thorough revision of the manuscript and conducting several additional analyses. I just have one further comment regarding the analyses on smoking cessation. Investigating sex-specific effects of smoking cessation on different health outcomes might be beyond the scope of this manuscript and its aims. As in the first version of the manuscript the exact definition of “active smokers” (now indicated as “ever smokers”) was not entirely clear and the assessment of lifetime pack-years smoked differs by smoking status (average number of cigarettes per day vs. maximum number of cigarettes), sensitivity analyses investigating whether the effects of smoking on health outcomes differ by smoking status might have been reasonable. Therefore, ever smokers could have been separated into former and never smokers (taking lifetime pack-years into account) independent of years since smoking cessation. Table 3 in the manuscript shows some quite high hazard ratios with broad confidence intervals. This might be due to the small sample sizes, esp. for women who quit smoking <5 or 5-15 years ago and who have lung disorders. Thus, these results should be interpreted with caution. I would suggest to move these results to the supplement and not put too much emphasis on this kind of analyses. Rather, I would just add this as kind of sensitivity analyses (as was done e.g. for the comparison of imputed vs. non-imputed pack-years data) to see whether the effects change and just mention this briefly. Nevertheless, I would like to thank the authors for all their efforts.

Answer: Thank you for this comment. As suggested, table 3 was moved to the supplement. We also softened the language about the conclusions on sex differences in the benefits of smoking cessation.

Submitted filename: Respond to reviewers, May 14.docx

Click here for additional data file.

18 May 2020

Female vulnerability to the effects of smoking on health outcomes in older people

PONE-D-20-03538R2

Dear Dr. Haghani,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Neal Doran

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

22 May 2020

PONE-D-20-03538R2

Female vulnerability to the effects of smoking on health outcomes in older people

Dear Dr. Haghani:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Neal Doran

Academic Editor

PLOS ONE