An examination of sex differences and cigarette smoking as predictors of COPD prevalence and incidence in older US adults

Abstract

Introduction:

This study examined the main and interactive effects of sex, cigarette smoking status, cigarette pack-years, and second-hand smoke exposure on COPD prevalence and incidence.

Methods:

COPD prevalence was estimated for U.S. adults aged 40+ years from Wave 1 of the Population Assessment of Tobacco and Health Study (N=12,296). Incidence analyses included adults from the initial sample without a COPD diagnosis (N=6,611). Multivariable Poisson regressions assessed prevalence and incidence based on self-reported sex and cigarette smoking, adjusted for covariates.

Results:

COPD prevalence was 7.4% and 9.4% and incident COPD was 5.0% and 8.7% for males and females, respectively. The adjusted prevalence ratio (aPR) for COPD for females was 1.26 [95% CI=1.11, 1.44] and the adjusted risk ratio (aRR) for incident COPD was 1.73 [1.41, 2.12]. Stratified by smoking status, female (vs male) sex was associated with aPRs of 1.26 [1.10, 1.44] and 1.35 [0.98, 1.84] and aRRs of 1.32 [1.00, 1.75] and 2.58 [1.79, 3.72] for adults who ever or never smoked, respectively. Smoking status (p=.003) and pack-years (p=.006) increased risk of COPD incidence for both males and females, but to a greater extent for males.

Conclusions:

Female sex was associated with significantly higher COPD incidence, which was not explained by cigarette smoking, second-hand smoke exposure, e-cigarette use, or other covariates. Cigarette-related COPD risk factors increased risk of COPD incidence for both males and females, but to a greater extent for males. Future research can include examining alternative risk factors or diagnostic biases contributing to higher incident COPD among females.

INTRODUCTION

Cigarette smoking is the primary driver of chronic obstructive pulmonary disease (COPD).¹ In a global systematic review and meta-analysis published in 2018, men had slightly higher COPD prevalence than women overall, with the highest COPD prevalence among women seen in North America (7.3% among women vs 8.1% among men).² National U.S. survey data suggest that COPD prevalence for women is approaching³ or surpassing^4,5 prevalence for men, and in 2023, the Centers for Disease Control and Prevention published national statistics showing that age-adjusted prevalence of COPD estimates from 2011–2022 were higher among women compared to men across years in the U.S.⁶ Epidemiologic data on incident COPD are important for understanding the upward trend in COPD prevalence among women, though there have been relatively few studies examining sex differences in COPD incidence in the U.S. In these few studies, which were published between 2009 and 2016, COPD incidence was consistently greater for men compared to women,^7–9 leaving it unclear as to why prevalence for women would be increasing. Given the increasing COPD prevalence and associated morbidity and mortality risk among women in the U.S., more up-to-date epidemiological work is needed.

In addition to rising prevalence, women receive COPD diagnosis at younger ages^10,11 and have more severe disease than men,^11–13 including more frequent exacerbations and more severe dyspnea—all despite lower cigarette smoking prevalence and lower average pack-years of smoking compared to men. Additionally, among adults who report never smoking, women have a higher prevalence of COPD than men, suggesting that cigarette smoking alone cannot account for sex differences in COPD prevalence.^14–18 There is growing evidence about how second-hand tobacco smoke exposure contributes to COPD,¹⁹ and new questions about how use of alternative nicotine products, like e-cigarettes, may also play a role.²⁰ Given that cigarette smoking alone does not completely account for sex differences in COPD risk,^14–18 more work is needed to understand whether other tobacco-related risk factors may explain differences seen for men and women. The current study used data from a nationally representative longitudinal cohort survey to assess the relationships between tobacco-related risk factors, including cigarette smoking status, cigarette pack-years, and second-hand smoke exposure; self-reported male or female sex; and COPD prevalence and incidence accounting for sociodemographic characteristics, e-cigarette use, and other important covariates. First, we examined the main and interactive effects of sex, cigarette smoking status, cigarette pack-years, and second-hand smoke exposure in relation to COPD prevalence and incidence in the full sample of adults aged 40+ years. Subsequently, we examined the main effect of sex on COPD prevalence and incidence in models stratified by never/ever cigarette smoking status.

METHODS

Sample

We analyzed restricted-use survey data from Waves 1 (2013–2014), 2 (2014–2015), 3 (2015–2016), 4 (2016–2018), and 5 (2018–2019) of the Population Assessment of Tobacco and Health (PATH) Study.²¹ For Wave 1 prevalence estimates, all adults aged 40+ years who had valid COPD data were included (N=12,296). Incidence analyses included adults aged 40+ years without a COPD diagnosis at Wave 1 who had data available at Waves 1–5 (N=6,611). Supplemental Figure 1 shows a flow diagram of the analytic samples for prevalence and incidence analyses. This study qualified as exempt per guidelines of the Westat Institutional Review Board and the Dartmouth Health Human Research Protection Program.

Measures

COPD prevalence and incidence.

At Wave 1, respondents were asked, “Has a doctor, nurse or other health professional EVER told you that you had any of the following lung or respiratory conditions? Choose all that apply: COPD, chronic bronchitis, emphysema, asthma, some other lung or respiratory condition, none of the above, don’t know, refused.” Any COPD, chronic bronchitis, or emphysema diagnoses were used for COPD prevalence. At Waves 2–5, respondents who had not reported COPD, chronic bronchitis, or emphysema diagnoses at any prior waves were asked about new respiratory disease diagnoses over the past 12 months. Any COPD, chronic bronchitis, or emphysema diagnoses across Waves 2–5 among adults who had reported no diagnosis at Wave 1 were combined to create a single COPD incidence measure at Wave 5.²²

Self-reported sex.

At Wave 1, respondents were asked, “What is your sex?”, with response options “male” or “female”.

Tobacco-related covariates.

Tobacco-related covariates from Wave 1 included cigarette smoking status, age of cigarette onset, pack-years of cigarette smoking (winsorized to the 99^th percentile),^23,24 second-hand smoke exposure, and e-cigarette use status. We used the PATH Study-derived variables for never, current established, and former established cigarette smoking to examine smoking status (nominal categorical variable). Current established cigarette smoking was defined as smoking at least 100 cigarettes lifetime and currently smoking every day or some days. Former established cigarette smoking was defined as smoking at least 100 cigarettes lifetime and not smoking in the past 12 months. Never smoking was defined as smoking fewer than 100 cigarettes in the respondent’s lifetime. In order for respondents to be retained in models that controlled for age of cigarette onset and pack-years, age of cigarette onset was categorized as never smoking, smoking onset <15 years of age, 15–24 years, or 25+ years (nominal categorical variable); and pack-years of smoking for respondents who reported never smoking were entered as “0” (continuous variable). Although ordinal in nature, age of cigarette smoking onset was included in models as a nominal categorical variable because it was not linearly associated with COPD prevalence. Second-hand smoke exposure at Wave 1 was included based on responses to the question, “In the past 7 days, number of hours that you were in close contact with others when they were smoking?”²⁵ (continuous variable). The PATH Study-derived variables for never, ever, and past 30-day (referred to as current) e-cigarette use at Wave 1 were used to examine e-cigarette use status (nominal categorical variable).

Other covariates.

Sociodemographic covariates from Wave 1 were categorized as shown in Table 1, which included urbanicity, age, race, and total household income in the past 12 months. We used PATH Study derived variables for age, sex, and race, which included imputations for missing data²¹; n=19 from the prevalence sample and n=11 from the incidence sample had data on sex imputed. Urban/non-urban classification criteria were based on the 2010 U.S. Census.²¹ All sociodemographic covariates were included in models as nominal categorical variables. Although age and total household income are ordinal in nature, they were not linearly associated with COPD prevalence. Other covariates included ever asthma diagnosis (yes/no) and use of cannabis, categorized as never use, ever but no past 30-day use, or past 30-day use at Wave 1 (both nominal categorical variables).

Table 1.

Sociodemographic characteristics of the overall sample and by self-reported respondent sex at Wave 1 of the Population Assessment of Tobacco and Health Study, 2013–2014.

	Overall (N=12,296)	Male (N=5,968)	Female (N=6,328)
	Weighted Column % or (Weighted Mean) (95% CI)
Age
40–49	27.2 (26.4–28.1)	27.6 (26.3–28.9)	26.9 (25.7–28.1)
50–59	29.6 (28.8–30.4)	29.8 (28.7–30.9)	29.5 (28.3–30.6)
60–69	23.6 (22.6–24.7)	24.1 (22.7–25.6)	23.2 (21.8–24.7)
70–79	12.9 (12.3–13.6)	12.9 (11.9–14.0)	12.9 (12.0–14.0)
80+	6.6 (6.0–7.3)	5.6 (4.8–6.5)	7.4 (6.5–8.4)
Race
White	81.7 (81.0–82.3)	82.2 (80.9–83.3)	81.3 (80.2–82.4)
Black	10.9 (10.4–11.5)	10.6 (9.9–11.4)	11.2 (10.4–12.0)
Other/Multiple	7.4 (7.0–7.8)	7.2 (6.3–8.3)	7.5 (6.7–8.4)
Household income
$100,000+	18.3 (17.3–19.4)	21.1 (19.6–22.6)	16.0 (14.9–17.2)
$50,000-$99,999	23.5 (22.3–24.7)	24.8 (23.2–26.4)	22.4 (20.9–23.9)
$25,000-$49,999	20.3 (19.3–21.4)	21.2 (19.7–22.7)	19.6 (18.3–21.1)
<$25,000	26.0 (24.9–27.0)	23.2 (21.9–24.6)	28.2 (26.8–29.8)
Urbanicity
Urban	75.3 (71.5–78.7)	73.9 (69.7–77.7)	76.4 (72.8–79.7)
Non-urban	24.7 (21.3–28.5)	26.1 (22.3–30.3)	23.6 (20.3–27.2)
Cigarette smoking status
Never	62.8 (61.4–64.1)	58.2 (56.4–60.0)	66.5 (65.0–68.1)
Former	23.0 (21.9–24.2)	26.1 (24.6–27.8)	20.4 (19.2–21.7)
Current	14.2 (13.7–14.8)	15.6 (14.9–16.3)	13.0 (12.3–13.8)
E-cigarette use status
Never	89.5 (89.0–89.9)	89.6 (89.0–90.2)	89.3 (88.7–90.0)
Ever, no current	6.5 (6.2–6.9)	6.7 (6.3–7.2)	6.4 (5.9–6.9)
Current	4.0 (3.7–4.3)	3.6 (3.3–4.0)	4.3 (3.9–4.7)
Cannabis use status
Never	70.1 (68.9–71.3)	64.7 (62.9–66.5)	74.6 (73.3–75.9)
Ever, no current	25.7 (24.6–26.8)	29.8 (28.1–31.6)	22.3 (21.1–23.6)
Current	4.2 (3.8–4.6)	5.5 (4.9–6.1)	3.1 (2.7–3.7)
Age of cigarette onset
<15	5.4 (4.9–5.8)	7.0 (6.2–7.8)	4.0 (3.6–4.5)
15–24	28.6 (27.4–29.8)	32.0 (30.4–33.7)	25.7 (24.4–27.2)
25+	3.3 (2.9–3.7)	2.8 (2.3–3.3)	3.7 (3.2–4.3)
Cigarette pack-years	(10.2) (9.7–10.7)	(12.9) (12.0–13.7)	(8.0) (7.5–8.6)
Second-hand smoke exposure (hours in past 7 days)	(4.6) (4.3–4.9)	(5.2) (4.9–5.6)	(4.1) (3.7–4.5)
COPD Prevalence at Wave 1	8.5 (7.8–9.1)	7.4 (6.6–8.2)	9.4 (8.5–10.3)
Asthma Diagnosis	10.3 (9.7–11.0)	7.9 (7.0–8.9)	12.3 (11.4–13.3)

Statistical analysis

Respondents who were missing data on any of the variables used in analyses were excluded. Supplemental Figure 1 shows a flow diagram of the participant sample for final prevalence and incidence analyses. The weights employed in our analysis accounted for attrition and non-response bias to obtain estimates for the U.S. population. Prevalence analyses were weighted using the Wave 1 full-sample and 100 replicate weights, which account for oversampling of some groups and non-response bias at Wave 1. Incidence analyses were weighted using the Wave 5 all-waves full-sample and 100 replicate survey weights, which account for oversampling at Wave 1 as well as non-response and attrition bias. Variances were computed using the balanced repeated replication method with Fay’s adjustment set to 0.3 to increase estimate stability. All analyses were conducted using Stata/MP 17.0 (www.stata.com/statamp/).

Missing data analysis.

Wave 5 all-waves weights addressed missingness related to respondents who were lost to follow-up across waves. Still, to determine whether large differences may limit generalizability of findings, we examined Wave 1 sample characteristics (COPD prevalence, sex, tobacco-related covariates, and other covariates) based on whether respondents were excluded from incidence analyses because they did not have Wave 5 all-waves survey weights available (i.e., they did not complete all five waves of data collection) using weighted (Wave 1 weights) chi-square or independent-samples t-tests as appropriate (Supplemental Table 1).

Sample characteristics overall and by sex.

Weighted descriptive statistics were used to examine characteristics of the sample overall (N=12,196) and by respondent sex (Table 1).

Associations between respondent sex and COPD among all adults aged 40+ years.

A series of weighted bivariable and multivariable Poisson regressions were used to evaluate main and interactive effects of sex, cigarette smoking status, cigarette pack-years, and second-hand smoke exposure on COPD prevalence or incidence among the overall samples of adults with prevalent COPD at Wave 1 (N=12,196) or adults with incident COPD at Waves 2–5 (N=6,611), respectively. Specific bivariable and multivariable models examined were: 1) sex on Wave 1 COPD prevalence (N=12,196; Table 2); 2) sex on Waves 2–5 COPD incidence (N=6,611; Table 3); 3) sex*ever/never smoking on COPD prevalence (N=12,196; Table 4); 4) sex*pack-years on COPD prevalence (N=12,196; Table 4); 5) sex*second-hand smoke on COPD prevalence (N=12,196; Table 4); 6) sex*ever/never Smoking on COPD incidence (N=6,611; Table 4); 7) sex*pack-years on COPD incidence (N=6,611; Table 4); and 8) sex*second-hand smoke on COPD incidence (N=6,611; Table 4). All multivariable models were adjusted for urbanicity, age, race, household income, cannabis use, e-cigarette use, second-hand smoke exposure, cigarette use status, age of cigarette onset, and cigarette pack-years. All covariates were treated as nominal categorical variables in models except for second-hand smoke exposure and cigarette pack-years, which were treated as continuous.

Table 2.

Bivariable and multivariable associations between self-reported sex and COPD prevalence (N=12,296; Wave 1, 2013–2014) among adults aged 40+ years at Wave 1 of the Population Assessment of Tobacco and Health (PATH) Study.

			Prevalence Ratio (95% CI)
		COPD Prevalence (Weighted Row %)	Bivariable	Multivariable
Sex	Male	7.4	Ref	Ref
	Female	9.4	1.27 (1.11, 1.46)	1.26 (1.11, 1.44)
Cigarette smoking status	Never	4.4	Ref	Ref
	Former	12.8	2.93 (2.46, 3.50)	1.68 (1.22, 2.33)
	Current	19.5	4.46 (3.82, 5.21)	2.05 (1.51, 2.79)
Age of cigarette onset	Never	4.4	Ref	Ref
	<15	22.4	5.13 (4.07, 6.46)	1.31 (0.96, 1.78)
	15–24	14.3	3.27 (2.82, 3.80)	1.07 (0.81, 1.43)
	25+	13.2	3.03 (2.26, 4.07)	1.00 (1.00, 1.00)
E-cigarette use status	Never	7.0	Ref	Ref
	Ever, no current	20.7	2.96 (2.60, 3.38)	1.57 (1.35, 1.83)
	Current	20.9	2.98 (2.58, 3.44)	1.66 (1.42, 1.93)
Cigarette pack-years1			1.23 (1.21, 1.25)	1.09 (1.07, 1.12)
Second-hand smoke exposure2			1.07 (1.06, 1.07)	1.02 (1.01, 1.03)
Race	White only	8.9	Ref	Ref
	Black only	7.1	0.80 (0.65, 0.99)	0.82 (0.69, 0.99)
	Other or multiple	5.3	0.60 (0.46, 0.78)	0.71 (0.55, 9.92)
Household income	$100K+	2.0	Ref	Ref
	$50-$99K	5.4	2.72 (1.81, 4.09)	2.15 (1.43, 3.22)
	$25–50K	8.9	4.49 (3.06, 6.60)	2.75 (1.86, 4.06)
	<$25K	15.1	7.57 (5.13, 11.16)	4.14 (2.76, 6.21)
	Refused/DK	9.2	4.60 (2.97, 7.12)	2.87 (1.84, 4.45)
Age	40–49	3.5	Ref	Ref
	50–59	8.1	2.35 (1.91, 2.89)	2.14 (1.74, 2.63)
	60–69	10.2	2.94 (2.42, 3.58)	2.86 (2.32, 3.51)
	70–79	13.9	4.02 (3.13, 5.16)	3.88 (2.99, 5.05)
	80+	13.7	3.95 (2.87, 5.43)	4.28 (3.15, 5.83)
Urbanicity	Urban	7.7	Ref	Ref
	Non-urban	10.7	1.38 (1.17, 1.63)	1.14 (0.99, 1.32)
Asthma diagnosis	No	6.6	Ref	Ref
	Yes	24.9	3.80 (3.31, 4.36)	3.56 (3.12, 4.05)
Cannabis use	Never	8.0	Ref	Ref
	Ever, no current	8.6	1.08 (0.94, 1.25)	1.01 (0.87, 1.17)
	Current	15.1	1.89 (1.52, 2.36)	1.18 (0.93, 1.50)

Bolded values denote statistical significance, p<0.05;

²Represents increased risk associated with each additional 10 pack-years;

²Represents the increased risk associated with each additional 5 hours of second-hand smoke exposure in the past 7 days.

Table 3.

Bivariable and multivariable associations between self-reported sex and COPD incidence (N=6,611; Waves 2–5, 2015–2019) among adults aged 40+ years at Wave 1 of the Population Assessment of Tobacco and Health (PATH) Study.

			Risk Ratio (95% CI)
		COPD Incidence (Weighted Row %)	Bivariable	Multivariable
Sex	Male	5.0	Ref	Ref
	Female	8.7	1.73 (1.39, 2.16)	1.73 (1.41, 2.12)
Cigarette smoking status	Never	5.1	Ref	Ref
	Former	7.6	1.48 (1.11, 1.96)	0.95 (0.57, 1.58)
	Current	16.5	3.22 (2.71, 3.82)	1.57 (1.00, 2.47)
Age of cigarette onset	Never	5.1	Ref	Ref
	<15	15.0	2.91 (2.14, 3.97)	1.15 (0.69, 1.91)
	15–24	10.2	1.98 (1.58, 2.48)	1.00 (0.63, 1.59)
	25+	10.1	1.96 (1.27, 3.03)	1.00 (1.00, 1.00)
E-cigarette use status	Never	6.2	Ref	Ref
	Ever, no current	16.4	2.67 (2.20, 3.24)	1.40 (1.11, 1.76)
	Current	15.5	2.51 (2.01, 3.15)	1.30 (0.95, 1.76)
Cigarette pack-years 1			1.18 (1.15, 1.21)	1.11 (1.06, 1.17)
Second-hand smoke exposure 2			1.07 (1.06, 1.09)	1.04 (1.01, 1.06)
Race	White only	6.9	Ref	Ref
	Black only	9.0	1.31 (1.02, 1.69)	1.19 (0.91, 1.55)
	Other or multiple	6.2	0.90 (0.55, 1.49)	1.03 (0.63, 1.69)
Household income	$100K+	2.3	Ref	Ref
	$50-$99K	5.6	2.46 (1.59, 3.81)	2.10 (1.36, 3.24)
	$25–50K	7.2	3.16 (2.08, 4.81)	2.34 (1.53, 3.57)
	<$25K	11.5	5.04 (3.38, 7.51)	3.18 (2.08, 4.86)
	Refused/DK	8.9	3.89 (2.46, 6.15)	3.02 (1.91, 4.78)
Age	40–49	5.0	Ref	Ref
	50–59	7.3	1.46 (1.09, 1.95)	1.46 (1.10, 1.95)
	60–69	7.1	1.43 (1.07, 1.90)	1.58 (1.18, 2.11)
	70–79	10.8	2.17 (1.53, 3.09)	2.46 (1.67, 3.62)
	80+	8.3	1.66 (0.74, 3.68)	1.86 (0.82, 4.22)
Urbanicity	Urban	7.0	Ref	Ref
	Non-urban	7.2	1.04 (0.77, 1.40)	0.93 (0.70, 1.24)
Asthma diagnosis	No	6.3	Ref	Ref
	Yes	14.9	2.37 (1.80, 3.12)	2.25 (1.69, 3.01)
Cannabis use	Never	6.6	Ref	Ref
	Ever, no current	7.8	1.18 (0.95, 1.47)	1.23 (0.98, 1.55)
	Current	10.0	1.53 (1.06, 2.20)	1.04 (0.68, 1.59)

Bolded values denote statistical significance, p<0.05;

¹Represents increased risk associated with each additional 10 pack-years;

²Represents the increased risk associated with each additional 5 hours of second-hand smoke exposure in the past 7 days.

Table 4.

Main and interactive effects between sex and cigarette smoking status, cigarette pack-years, or second-hand smoke exposure on COPD prevalence (N=12,196; Wave 1, 2013–2014; Interaction Models 1–3) or COPD incidence (N=6,611; Waves 2–5, 2015–2019; Interaction Models 4–6) among adults aged 40+ years at Wave 1 of the Population Assessment of Tobacco and Health (PATH) Study.

	Unadjusted Prevalence Ratio (95% CI)	p value for unadjusted	Adjusted Prevalence Ratio (95% CI)1	p value for adjusted
	Interaction Model 1: Sex * Cigarette smoking status interaction on COPD prevalence
Female sex (vs male)	1.79 (1.30, 2.45)	<0.001	1.41 (1.04, 1.91)	0.025
Ever cigarette smoking (vs never)	4.48 (3.52, 5.72)	<0.001	2.05 (1.38, 3.06)	<0.001
Female sex * Ever cigarette smoking	0.72 (0.51, 1.02)	0.063	0.85 (0.61, 1.17)	0.316
	Interaction Model 2: Sex * Cigarette pack years interaction on COPD prevalence
Female sex (vs male)	1.50 (1.25, 1.80)	<0.001	1.33 (1.08, 1.64)	0.009
Cigarette pack-years	1.24 (1.21, 1.26)	<0.001	1.10 (1.07, 1.14)	<0.001
Female sex * Cigarette pack-years	1.01 (0.98, 1.05)	0.397	0.98 (0.94, 1.02)	0.372
	Interaction Model 3: Sex * Second-hand smoke exposure interaction on COPD prevalence
Female sex (vs male)	1.28 (1.11, 1.48)	0.001	1.24 (1.07, 1.44)	0.004
Second-hand smoke exposure	1.06 (1.05, 1.08)	<0.001	1.02 (1.00, 1.03)	0.019
Female sex * Second-hand smoke exposure	1.00 (0.99, 1.02)	0.640	1.01 (0.99, 1.02)	0.372
	Unadjusted Risk Ratio (95% CI)	p value for unadjusted	Adjusted Risk Ratio (95% CI)1	p value for adjusted
	Interaction Model 4: Sex * Cigarette smoking status interaction on COPD incidence
Female sex (vs male)	2.95 (2.05, 4.26)	<0.001	2.58 (1.80, 3.69)	<0.001
Ever cigarette smoking (vs never)	3.86 (2.60, 5.73)	<0.001	2.02 (1.13, 3.61)	0.018
Female sex * Ever cigarette smoking	0.44 (0.28, 0.69)	<0.001	0.49 (0.31, 0.78)	0.003
	Interaction Model 5: Sex * Cigarette pack years interaction on COPD incidence
Female sex (vs male)	2.21 (1.78, 2.76)	<0.001	2.11 (1.69, 2.63)	<0.001
Cigarette pack-years	1.24 (1.20, 1.28)	<0.001	1.17 (1.11, 1.22)	<0.001
Female sex * Cigarette pack-years	0.94 (0.89, 0.98)	0.010	0.91 (0.85, 0.96)	0.002
	Interaction Model 6: Sex * Second-hand smoke exposure interaction on COPD incidence
Female sex (vs male)	1.82 (1.43, 2.30)	<0.001	1.75 (1.40, 2.18)	<0.001
Second-hand smoke exposure	1.08 (1.06, 1.11)	<0.001	1.04 (1.00, 1.08)	0.032
Female sex * Second-hand smoke exposure	0.99 (0.97, 1.02)	0.541	0.99 (0.96, 1.03)	0.770

¹All adjusted models included urbanicity, age, race, household income, age of cigarette onset, e-cigarette use status, cannabis use status, and ever asthma diagnosis as covariates.

Models 1 and 4 were also adjusted for cigarette pack-years and second-hand smoke exposure. Models 2 and 5 were also adjusted for cigarette smoking status and second-hand smoke exposure. Models 3 and 6 were also adjusted for cigarette smoking status and cigarette pack-years.

Associations between respondent sex and COPD in models stratified by cigarette smoking status.

Additional bivariable and multivariable Poisson regressions were used to evaluate main effects of sex on COPD prevalence or incidence among subsamples stratified by never/ever cigarette smoking status. Specific bivariable and multivariable models examined were: 1) sex on COPD prevalence among respondents who ever smoked (former + current) (N=6,729; Supplemental Table 2); 2) sex on COPD incidence among respondents who ever smoked (N=3,292; Supplemental Table 3); 3) sex on COPD prevalence among respondents who never smoked (N=5,567; Supplemental Table 4); and 4) sex on COPD incidence among respondents who never smoked (N=3,319; Supplemental Table 5). Multivariable models stratified to only those who never smoked were adjusted for urbanicity, age, race, household income, cannabis use, e-cigarette use, and second-hand smoke exposure. Multivariable models stratified only those who ever smoked were adjusted for these same covariates, plus cigarette pack-years and age of cigarette onset. All covariates were treated as nominal categorical variables in models except for second-hand smoke exposure and cigarette pack-years, which were treated as continuous.

Sensitivity analysis.

Given that age of cigarette onset, cigarette pack years, and ever smoking status were interdependent, we examined additional interaction models for smoking status*sex and pack years*sex that did not include interdependent covariates. Specifically, we removed pack years and age of onset from the smoking status*sex interaction analyses and removed smoking status and age of onset from the pack years*sex interaction analyses for both COPD prevalence (N=12,196) and incidence (N=6,611) in four additional interaction models as a sensitivity analysis (Supplemental Table 6).

RESULTS

Sample characteristics

Overall, respondents were largely from urban areas (75.3%), female (51.5%), and of White race (81.7%). There were more respondents with incomes <$25,000 per year (26.0%) relative to those with incomes $100,000+ per year (18.3%) and there were more respondents aged 40–59 (56.8%) than those aged 60+ (43.1%). Males and females were similar on most sociodemographic characteristics, though males tended to report higher household incomes compared to females (Table 1).

Respondents included in the Wave 1 COPD prevalence analysis but who were missing Wave 5 all-waves survey weights (i.e., they did not complete all five waves of data collection) and excluded from COPD incidence analyses were significantly older, were more likely to be male, to report ever use of cigarettes and e-cigarettes, to begin smoking cigarettes at a younger age, to have a higher pack-years of smoking, and to be diagnosed with COPD at Wave 1. Respondents excluded from incidence analyses based on attrition across waves were also less likely to report ever use of cannabis, an asthma diagnosis, and to have an income ≥$50,000/year (Supplemental Table 1).

Sample characteristics by sex

The overall weighted Wave 1 prevalence of COPD was 8.5% (7.4% males, 9.4% females), and Waves 2–5 incidence was 7.0% (5.0% males, 8.7% females). Females (vs males) had higher prevalence and incidence of COPD despite lower prevalence of ever smoking (33.4% vs 41.7%), lower average cigarette pack-years (8.0 vs 12.9), lower prevalence of childhood cigarette smoking (4.0% vs 7.0% started smoking <15 years of age), lower second-hand smoke exposure (4.1 vs 5.2 hours in the past 7 days), and lower prevalence of ever cannabis use (25.4% vs 35.3%) (Table 1). Females also had higher prevalence of asthma at Wave 1 (12.3% vs 7.9%) (Table 1).

Associations between respondent sex and COPD among all adults aged 40+ years

Female (vs male) sex was associated with a significantly increased prevalence of Wave 1 COPD prevalence in bivariable (prevalence ratio (PR)=1.27; 95% CI=1.11, 1.46) and multivariable regression analyses (aPR=1.26; 95% CI=1.11, 1.44) (N=12,196; Table 2). Female (vs male) sex was also significantly associated with increased Waves 2–5 incidence in bivariable (RR=1.73; 95% CI=1.39, 2.16) and multivariable regressions (aRR=1.73; 95% CI=1.41, 2.12) (N=6,611; Table 3).

In models that examined the potential interactions between respondent sex and never/ever smoking status, cigarette pack-years, or second-hand smoke exposure on COPD prevalence, there were no significant interactions (Table 4, Interaction Models 1–3). In contrast, for COPD incidence, sex*smoking status (Table 4, Interaction Model 4) was statistically significant in bivariable (p<0.001) and multivariable (p=.003) models. Sex*cigarette pack-years (Table 4, Interaction Model 5) was also statistically significant in bivariable (p=.010) and multivariable models (p=.002) for COPD incidence. Sex*second-hand smoke exposure was nonsignificant for COPD incidence (Table 4, Interaction Model 6). For the significant sex*smoking status interaction on COPD incidence, post-hoc tests revealed that the aRRs for ever smoking for males were 3.86 (95% CI=2.60, 5.73) and 2.02 (95% CI=1.13, 3.61) for bivariable and multivariable models, respectively, which were significantly higher than the aRRs for ever smoking for females (bivariable: 1.70; 95% CI=1.34, 2.15; multivariable: 1.00; 95% CI=0.62, 1.60). Similarly, for the significant sex*cigarette pack-years interaction on COPD incidence, post-hoc tests revealed that the aRRs for (per 10) pack-years for males were 1.24 (95% CI=1.20, 1.28) and 1.17 (95% CI=1.11, 1.22) for bivariable and multivariable models, respectively, which were significantly higher than the aRRs for females (bivariable: 1.16; 95% CI=1.11, 1.21; multivariable: 1.06; 95% CI=1.00, 1.11), though both were statistically significant.

Associations between respondent sex and COPD in models stratified by cigarette smoking status

The increased risk for COPD prevalence seen for females (vs males) was somewhat dependent upon smoking status. Among respondents who had ever smoked, female (vs male) sex was significantly associated with increased prevalence of COPD in bivariable (PR=1.29; 95% CI=1.13, 1.47) and multivariable regressions that controlled for covariates (aPR=1.26; 95% CI=1.10, 1.44) (N=6,729; Supplemental Table 2). Female (vs male) sex was not significantly associated with increased risk of COPD incidence in bivariable analysis (RR-1.30; 95% CI=0.98, 1.73), but became significantly associated in multivariable analysis after controlling for covariates (aRR=1.32; 95% CI=1.00, 1.75) (N=3,292; Supplemental Table 3).

Among respondents who had never smoked, female (vs male) sex was significantly associated with increased prevalence of COPD in bivariable analysis (PR=1.79; 95% CI=1.30, 2.45), but became nonsignificant after controlling for covariates (aPR=1.35; 95% CI=0.98, 1.84) (N=5,567; Supplemental Table 4). However, female (vs male) sex was significantly associated with increased risk of COPD incidence in both bivariable (RR=2.95; 95% CI=2.05, 4.26) and multivariable (aRR=2.58; 95% CI=1.79, 3.72) regressions (N=3,319; Supplemental Table 5).

Sensitivity analysis

In multivariable interaction models that removed interdependent covariates, results were consistent with primary interaction models (Supplemental Table 6). There were no significant interactions for COPD prevalence. For COPD incidence, sex*smoking status (p=0.001) and sex*cigarette pack-years (p=0.002) were statistically significant. For the significant sex*smoking status interaction on COPD incidence, post-hoc tests revealed that the aRR for ever smoking for males was 2.64 (95% CI=1.72, 4.07), which was significantly higher than the aRR for ever smoking for females (1.22 (95% CI=0.92, 1.62)). Similarly, for the significant sex*cigarette pack-years interaction on COPD incidence, post-hoc tests revealed that the aRR for (per 10) pack-years for males was 1.19 (95% CI=1.14, 1.23), which was significantly higher than the aRR for females (1.08; 95% CI=1.03, 1.13).

DISCUSSION

In this large nationally representative study of U.S. adults aged 40+, females had significantly higher COPD prevalence and incidence compared to males. This association was independent of lifetime cigarette smoking, second-hand smoke exposure, e-cigarette use, and other covariates. When further evaluating the association between sex and COPD prevalence, female (vs male) sex was significantly associated with increased COPD prevalence among those who had ever smoked cigarettes, but not among those who never smoked in multivariable models stratified by smoking status. None of the interactions examined (i.e., sex*smoking status, sex*pack years, sex*second-hand smoke exposure) were significant for COPD prevalence. In contrast, for COPD incidence, female (vs male) sex was associated with significantly increased risk for both groups (never and ever cigarette smoking) in multivariable models stratified by smoking status. We also observed significant sex*smoking status and sex*cigarette pack-years interactions for COPD incidence. Post-hoc tests for the significant interactions revealed that both ever (vs never) cigarette smoking and cigarette pack-years were associated with significantly increased risk of COPD incidence for males and females, but had significantly stronger associations with COPD incidence among males, when controlling for other smoke-related risk factors (e.g., second-hand smoke exposure) and covariates.

Findings from this study are consistent with those indicating an increased risk of COPD among women despite lower smoking exposures (eg, cigarette pack-years). Our study extends prior findings by showing 1) an increased risk of COPD incidence for women compared to men regardless of ever smoking status, but especially for those who report never smoking, which is not explained by greater second-hand smoke exposure and 2) cigarette smoking COPD risk factors (cigarette smoking status, cigarette pack-years) have stronger associations with COPD incidence for men than women. We did not anticipate finding that current smoking and pack-years would be stronger COPD incidence risk factors for men compared to women in interaction analyses. We searched for statistical anomalies that could explain the finding and found none (i.e., the association for pack years was linear for both prevalence and incidence). We also conducted additional sensitivity analyses which removed other interdependent smoking-related covariates from these interaction models that may have led to underestimation of the effect of smoking status or pack years on COPD for women (e.g., removing pack-years and age of cigarette onset from the sex*smoking status interaction model), but results were consistent with models that included all covariates. We also found that men had more cigarette pack years on average than women, and more women who never smoked (and thus had zero pack years) had prevalent and incident COPD, which may account for these findings. It is also possible that unmeasured sex differences in cigarette smoking or COPD diagnosis in the modern era (for example bias in how smoking is reported, or bias in how COPD is diagnosed in men vs women²⁶) could explain the finding. However, it is critical to first replicate the finding before there is much speculation as to the cause, and certainly before concluding that cigarette smoking is a less important COPD risk factor for women than men.

We found no significant interactions between sex and second-hand smoke exposure on COPD prevalence or incidence, which is in contrast to prior work suggesting that passive smoke exposure is a COPD risk factor for women, but not men.¹⁴ Other prior work has led to the suggestion that women may be more susceptible to the lung-damaging effects of cigarette smoking.^12,27 Indeed, women who smoke cigarettes are at greater risk of airflow obstruction²⁸ and have significantly faster annual decline in FEV1% predicted with increasing age²⁹ compared to males who smoke. In addition, women have more pronounced proteomic alterations in respiratory tract lining fluid due to active smoking compared to men.³⁰ However, results from our analyses for COPD incidence do not support the suggestion that females are more susceptible to the effects of cigarette smoking. In fact, although cigarette smoking was a risk factor for COPD prevalence and incidence for both men and women, we found stronger relationships between cigarette smoking and COPD incidence for men in interaction analyses. We also observed that ever (vs never) cigarette smoking was not significantly associated with COPD incidence for women after controlling for other important factors that put women at increased risk for COPD (e.g., higher prevalence of asthma).³¹ Important, however, is that cigarette pack-years was significantly associated with COPD incidence for females in the pack years*sex interaction model, suggesting that cigarette pack years (vs ever smoking status) may better capture the increased COPD incidence risk associated with lifetime cigarette smoking for females. We also found that a higher adjusted relative risk for COPD incidence for females (vs males) among those reporting ever or never smoking (1.32 and 2.58 respectively), in models stratified by smoking status, which is consistent with results from the sex*smoking status interaction analysis for COPD incidence and suggests that there are explanations outside of cigarette smoking and second-hand smoke exposure to consider. Findings highlight the importance of understanding COPD risk among people who have never smoked for public health.¹⁶

Recent work has reported smaller airway lumen sizes in women (vs men) and thicker airway walls for men (vs women) who have never smoked after controlling for important covariates, including height and lung capacity.³² This same work showed that among those with a history of smoking, sex-related airway differences were associated with lower FEV1/FVC, more dyspnea, poorer respiratory quality of life, and higher respiratory mortality in women compared to men.³² Smaller airway diameter results in greater airway resistance, and may help to explain sex differences in COPD risk regardless of cigarette smoking status. Still, we found that COPD prevalence and incidence among females were higher than their male counterparts, which was not explained by second-hand smoke exposure,^14,19 e-cigarette use,^33,34 sociodemographic characteristics like rural/urban residence and household income,³⁵ or other important covariates that have been independently associated with risk of COPD among adults who have never smoked. It is possible that such differences could be explained by alternative COPD risk factors that we did not account for, but that have been shown to increase COPD risk in women specifically, including respondent weight⁴, hormonal influences³⁶, and exposure to biomass fuel for heating and cooking.^14,37 Findings from the current study should prompt investigation into alternative risk factors contributing to the much higher incident COPD among females who have or have not smoked cigarettes.

This study had several strengths, but also had limitations. Other COPD risk factors not accounted for in the current study due to not being available in the PATH Study include prenatal and early life exposures (eg, genetics, respiratory infections) and occupational exposures, which have been shown to impact lung development, lung function trajectories, risk of COPD, and COPD morbidity.^38–40 In addition, the second-hand smoke exposure variable from the PATH Study asked respondents to report on their hours of exposure in the past 7 days. It is possible that alternative measures of second-hand smoke exposure (e.g., years of living in a household where someone smokes), which were not available in the PATH Study, would have influenced results differently. The use of this large survey database also introduces challenges related to self-reported measures of COPD and sex, which may have led to biased results. Without clinical data such as spirometry to confirm airflow obstruction, it is possible to over and under diagnose COPD.⁴¹ Women may be more likely to go to the doctor than men,⁴² which could contribute to diagnostic biases. In addition, higher incidence of COPD among women might simply represent later onset of the disease in that group given that women tend to have higher life expectancies compared to men. In addition, consistent with prior work,^22,43 we used Wave 1 variables for both prevalence and incidence analyses. Results may differ if instead time-varying variables were used in incidence analyses that capture changes in tobacco product use and other risk factors over time. We also found that e-cigarette use was associated with increased COPD for both men and women, which is consistent with prior work and deserving of attention in future studies.^22,33,34,44 Still, this study did not account for nuances in e-cigarette use, such as differences in devices used or use frequency, and therefore cannot account for the possible true impact of e-cigarettes on COPD or relevant sex differences. Finally, we may be overestimating the impact of sex on COPD incidence given that more respondents who had ever smoked cigarettes, including those with higher pack-years of smoking and earlier age of cigarette onset, and more male respondents were excluded from COPD incidence analyses due to attrition across waves.

Findings from the current analysis using nationally representative data from U.S. adults aged 40+ years showed that female respondents were at significantly increased risk for COPD prevalence and incidence compared to male respondents. Interaction analyses revealed that cigarette smoking-related COPD risk factors (ever smoking status, cigarette pack-years) were related to COPD incidence for both men and women, but were more strongly related to COPD incidence for men. In addition, COPD incidence among females who never smoked was substantially higher than their male counterparts who never smoked, which was not explained by second-hand smoke exposure, e-cigarette use, or other covariates. Findings should prompt a search for alternative risk factors or diagnostic biases contributing to higher incident COPD among females. Future research is also needed to determine how various COPD risk factors interact with one another to confer greater disease risk for women and to aid in the development of preventive and therapeutic clinical innovations that are tailored to those who are at greatest risk for COPD.

IMPLICATIONS.

Prior studies show that COPD prevalence has been increasing for women in the U.S., but the basis for this change remains unclear. This study shows how female (vs male) sex is associated with significantly increased risk for COPD prevalence and incidence among a nationally representative sample of older (aged 40+ years) U.S. adults using data from 2013 to 2019, which was not accounted for by cigarette smoking, second-hand smoke exposure, e-cigarette use, or other covariates. Work is needed on alternative COPD risk factors or diagnostic biases contributing to higher incident COPD among females.

DATA AVAILABILITY

Public-use and restricted-use data from the PATH Study are available through the National Addiction & HIV Data Archive Program (NAHDAP): https://www.icpsr.umich.edu/web/NAHDAP/studies/36231