Association between secondhand smoke exposure and serum sex hormone concentrations among US female adults: a cross-sectional analysis using data from the National Health and Nutrition Examination Survey, 2013-2016


Secondhand smoke (SHS) refers to the mixture of the smoke released at the end of the burning of tobacco or tobacco products and the smoke released by smokers exhaling.1 SHS exposure is a common public health problem that can cause many adverse health outcomes, including lung cancer, acute cardiovascular effects and respiratory outcomes, due to it containing many actually and probably toxic and radiative substances, such as cotinine, nicotine, carbon monoxide, thiocyanate and 4-aminobiphenyl-haemoglobin adduct.2 In 2004, the proportions of children, non-smoking males and non-smoking females exposed to SHS worldwide were 40%, 30%, and 35%, respectively.3 The loss of disability-adjusted life years due to SHS exposure (10.9 million) accounted for approximately 0.7% of the total global burden of disease in disability-adjusted life years in 2004.3 SHS exposure is estimated to cause additional 600 000 deaths each year among non-smokers.4 The Global Tobacco Surveillance System has led efforts to fight against tobacco use, but more than 90% of the world population still lives in countries that are not fully smoke-free under public health regulations.3 5 SHS therefore remains a global challenge.

Cotinine is a key metabolite of nicotine, and its level in the blood is proportional to the amount of tobacco smoke exposure, which can be used as an indicator of environmental tobacco smoke or SHS exposure.6 Cotinine may influence the probability of a successful pregnancy by dysregulating and changing the reproductive and hormonal systems, exerting adverse effects on mechanisms such as gametogenesis, tubal transport, ovulation, fertilisation, implantation of fertilised oocytes and placental development.7 Serum sex hormones play important roles in regulating many critical developmental differentiation events that greatly affect the fetus, and subsequently reproductive health.8 Serum sex hormones were also found to be involved in several regulatory systems and processes of diseases with sex-related differences.9 10 Total testosterone (TT) is an essential precursor for oestradiol (E2) synthesis in females and can act directly as an androgen.11 12 Testosterone has favourable cardiovascular effects that can be measured by surrogate outcomes.12 E2 secretion is responsible for ovarian quality and reproductive ageing.13 Serum sex hormone-binding globulin (SHBG) is the blood transport protein for sex hormones, and is involved in sex hormone metabolism, blood circulation and endocrine balance.14

There is some epidemiological evidence that cigarette smoke causes changes in circulating hormone concentrations by interfering with the transport, storage, metabolism and clearance of steroid hormones.10 15 16 Several previous studies have focused on the association between smoking and serum sex hormones,17–19 but some of their conclusions have been contradictory. Moreover, some of the few studies that specifically focused on SHS mostly evaluated SHS using questionnaires, which are highly subjective. The characteristics of the menstrual cycle have also been found to differ between non-smokers and smokers.20 Considering the above situation, we selected female adults (age ≥20 years) from the 2013–2016 National Health and Nutrition Examination Survey (NHANES, a large-scale national representative cross-sectional population study) to estimate the statistical association between SHS exposure and serum sex hormone levels, adjusting for potential confounders, with the understanding that causality cannot be inferred from our cross-sectional data in never smokers and former smokers. Our analysis seeks to lay the groundwork for future longitudinal or experimental studies that can more directly address the causal impact of SHS on hormone levels. The obtained results may provide a new insight into the association between SHS and serum sex hormones in females.


Study population

We used the 2013–2016 NHANES database, which is a cross-sectional, nationally representative and unique data source that was designed by the National Center for Health Statistics of the US Centers for Disease Control and Prevention to evaluate the health and nutritional status of the US non-institutionalised civilian population. A detailed description of the NHANES database can be found elsewhere.21

We acquired the demographic, examination, laboratory and questionnaire data from the 2013–2016 NHANES. Demographic information was obtained from 5983 females aged ≥20 years. Different types of data, including examination, laboratory and questionnaire data, were combined with demographic data using unique survey markers (respondent sequence number). This original sample study excluded pregnant females and some participants with incomplete data on sex hormones (TT, E2 and SHBG), serum cotinine and the covariates of body mass index (BMI), race, education, marital status, family poverty-to-income ratio (PIR), menstrual status, time of blood sampling and drinking status.

We also focused on people exposed to SHS and excluded active smokers. Data processing resulted in the final analysis of data from 622 participants. The specific data processing flow is shown in figure 1.

Figure 1
Figure 1

Flow chart for inclusion of study participants. NHANES, National Health and Nutrition Examination Survey.

Sex steroid hormones

Isotope dilution–liquid chromatography tandem mass spectrometry was used to measure serum TT and E2. The SHBG concentration is based on the reaction of SHBG to immunoantibodies and chemoluminescence measurements of the reaction products that occur after two incubation periods and exposure to a magnetic field. The ratio of TT to E2 (TT/E2) was calculated to estimate aromatase activity. The formula [(TT/SHBG)×100] was used to calculate the free androgen index (FAI).

SHS exposure

Smokers were classified as active smokers (smoked ≥100 cigarettes in lifetime and smoking now), former smokers (smoked ≥100 cigarettes in lifetime and do not smoke at all now) and never smokers (smoked <100 cigarettes in lifetime) based on two questions: (1) ‘Have you/has your study participant smoked at least 100 cigarettes in your/their entire life?’ and (2) ‘Do you/does your study participant smoke cigarettes now?’ Active smokers were excluded from the study. The serum cotinine concentration was considered the biomarker of tobacco exposure. Isotope dilution-high-performance liquid chromatography/atmospheric-pressure chemical-ionisation tandem mass spectrometry was used to measure serum cotinine. SHS exposure was defined as a serum cotinine concentration of 0.05–10 ng/mL.22


Covariates considered in the models included age, race, education, marital status, BMI, drinking status, family PIR, menstrual status and time of blood sampling. Age was classified into three groups (20–40, 40–60 and >60 years). Race was divided into five groups (non-Hispanic black, Mexican American, Hispanic, non-Hispanic white and other). Education was classified as less than high school, high school or equivalent, and college or above. Marital status was categorised as married or living with a partner, single, divorced or widowed. BMI was calculated by dividing weight in kilograms by height in metres squared, and was grouped into three categories (normal, <25 kg/m2; overweight, ≥25 and <30 kg/m2; and obese, ≥30 kg/m2). The drinking status was divided into drinkers and non-drinkers, with those who answered ‘yes’ to the question ‘In 1 year, have you/has your study participant had at least 12 drinks of any type of alcoholic beverage?’ were considered to be drinkers, and those who answered ‘no’ were non-drinkers. Female menstruation (premenopausal and postmenopausal) was defined as reproductive health in the questionnaire. If participants answered ‘yes’ to the first question ‘Have you/has your study participant had at least one menstrual period in the past 12 months (not including bleeding caused by medical conditions, hormone therapy or surgeries)?’, they were considered as premenopausal. If participants answered ‘no’, they were asked the second question ‘What is the reason that (you have/your study participant has) not had a period in the past 12 months?’ The females who answered ‘pregnant’ or ‘breast feeding’ to the second question were excluded. Moreover, those who had their ovaries removed or had ever used female hormone supplements were excluded. Postmenopausal females were therefore defined as those who had not experienced at least one menstrual period in the past 12 months, not due to breast feeding and pregnancy, and those who had their ovaries removed and used female hormone supplements, and they were excluded. The time of blood sampling was categorised into morning, afternoon and evening.

Statistical analysis

Continuous variables are expressed as weighted median with IQR, while categorical variables are expressed as unweighted numbers and weighted percentages. Serum cotinine and sex hormone levels were natural logarithm transformed based on their skewed distribution. A linear association pattern was used based on previous literature similar to our research.23–25 Weighted multiple linear regression was used to estimate the associations between SHS and sex hormone concentrations. We evaluated the percentage change in serum sex hormone concentrations when the serum cotinine level was doubled. The formula (exp(ln2×β)−1)×100% was used to calculate the percentage change, and exp(ln2×(β±1.96×SE)−1)×100% was calculated as the 95% CI.26 SHS was also distributed into quartiles (quartile 1 (Q1), <0.077; quartile 2 (Q2), 0.077–0.151; quartile 3 (Q3), 0.151–0.441; and quartile 4 (Q4), >0.441). For serum cotinine quartiles, Q1 was deemed to have the least SHS exposure and was used as the reference group, and the percentage change was evaluated by comparing each quartile with this reference group. Probability values were obtained by regarding the SHS quartiles as an ordinal variable. The p value for trend was calculated using serum cotinine quartiles as a continuous variable, which this assumes that the distance between the quintiles is the same, in addition. All models were analysed by adjusting the age, race, education, marital status, BMI, drinking status, family PIR, menstrual status and time of blood sampling based on prior literature and their potential relevance to both SHS exposure and serum hormone levels.7 10 27

To ensure the robustness of the results, we assessed the associations between SHS and serum hormones in never smokers and former smokers separately, and subgroup analysis was conducted with menstruation type stratified since it plays an important role in the changes of serum sex hormones. In linear regression, the interaction effect can be calculated by including an interaction term between SHS (logarithm-transformed serum cotinine) and menstrual status in the regression model. This can be done by multiplying the two variables together and including this new interaction term as an additional covariate in the model.

All statistical analyses were performed using R software (V.4.0.2), in which the ‘survey’ packages were also used. All values with two-tailed p<0.05 were considered significant.

Patient and public involvement



Participant characteristics

The study included 622 participants aged ≥20 years from the 2013–2016 NHANES. The demographic characteristics of the population are listed in table 1. Overall, approximately half of the females were aged 20–40 years, and over 70% were never smokers. More than half of the participants had less than high school education (58.76%) and were drinkers (57.07%). The proportion with obesity (BMI ≥30 kg/m2) was 43.97%. For marital status, the proportions of married or living with a partner, single, divorced or widowed people were roughly equal. Natural menopause presented in 63.47%, and blood was mostly sampled in the morning (47.83%). In serum sex hormones, the median (IQR) TT, E2 and SHBG concentrations, and FAI and TT/E2 were 22.3 (21.20–23.85) ng/dL, 36.74 (30.11–45.44) pg/mL, 59.01 (55.73–63.16) nmol/L, 35.64 (32.36–39.47) and 0.61 (0.47–0.80), respectively. The median (IQR) serum cotinine concentration was 0.15 (0.13–0.19) ng/mL.

Table 1

Population characteristics in NHANES

Associations between SHS exposure and serum sex hormones in the study population

Table 2 lists the estimated percentage changes with 95% CIs in serum sex hormones associated with SHS exposure in study participants. After adjusting for potential confounding variables, a doubled serum cotinine concentration was significantly associated with a 5.79% (95% CI=1.06% to 10.74%) higher E2 and 2.75% (95% CI=0.0% to 5.55%) higher FAI. When modelling serum cotinine as quartiles, participants with serum cotinine in Q4 had 31.07% (95% CI=10.63% to 55.28%, p for trend=0.010) higher E2 and 8.35% (95% CI=−1.92% to 19.69%, p for trend=0.043) higher FAI than those in Q1. However, the TT/E2 in Q4 was 19.96% (95% CI=−32.05% to −5.73%, p for trend=0.041) lower than that in Q1.

Table 2

Estimated per cent change (95% CI) in serum sex hormone concentrations by SHS exposure (serum cotinine levels) in NHANES Study participants

Figure 2 illustrates the potential correction effect of menopausal status on the associations between SHS exposure and serum sex hormone concentrations. Exposure to SHS, which was significantly associated with higher FAI, was observed (5.93%, 95% CI=1.09% to 11.01%, p<0.05) among postmenopausal women. However, no significant interaction was found between SHS exposure (ie, doubled serum cotinine concentration) and menopausal status (p for interaction>0.05).

Figure 2
Figure 2

Estimated percent change (95% CI) in serum sex hormone concentration associated with doubling of serum cotinine stratified by menopausal status. E2, oestradiol; FAI, free androgen index; SHBG, sex hormone-binding globulin; TT, total testosterone; TT/E2, ratio of total testosterone and oestradiol.

Associations between SHS exposure and serum sex hormones in never smokers and former smokers

Table 3 presents the associations between SHS and serum sex hormones in never smokers and former smokers. For never smokers, a doubled serum cotinine concentration was significantly associated with 2.85% higher levels of TT (95% CI=0.29% to 5.47%) and 6.29% (95% CI=0.68% to 12.23%) higher levels of E2 in fully adjusted models, respectively. After modelling serum cotinine levels as quartiles, the never smokers in Q4 exhibited 10.30% (95% CI=0.78% to 20.72%) and 27.75% (95% CI=5.17% to 55.17%) higher levels of TT and E2, respectively, compared with those in Q1. SHBG was lower by 4.36% (95% CI=−8.47% to −0.07%, p for trend=0.049) in former smokers when the serum cotinine concentration was doubled. When serum cotinine was categorised into quartiles, the SHBG in Q4 was lower by 17.58% (95% CI=−31.33% to −1.07%, p for trend=0.018) compared with that in Q1.

Table 3

Estimated per cent change (95% CI) in serum sex hormone concentrations by SHS exposure (serum cotinine levels) in non-smokers and former smokers among NHANES


In this large cross-sectional representative sample of US women aged 20 years and older, we found that SHS was significantly associated with increased level of E2 and FAI. Among never smokers, SHS exposure was associated with higher levels of TT and E2. SHS exposure may be associated with lower levels of SHBG among individuals who were former smokers, and there is also a similar association between SHS and TT/E2 among all participants except for never smokers and former smokers. Most research has focused on the associations between smoking and sex hormones,10 18 27–29 with few epidemiological studies having explored the associations between SHS and serum sex hormones.10 17 19 30

We observed that SHS had a significant association with increased levels of TT in never smokers. Similar conclusions have been drawn from some smoking-related studies. The latest meta-analysis of this association included 19 published peer-reviewed articles and demonstrated that current smokers had higher TT levels than non-smokers among premenopausal women.31 Zhao et al analysed six studies involving 6089 women (mean age=28–62 years), and in fixed-effects models found that exposure to smoking was associated with higher TT.29 Some studies of females who received in vitro fertilisation found that cigarette smoking was associated with increased serum TT or FAI.28 32 Although the SHS exposure status was not available for never smokers or former smokers in these studies and smoking status was measured using self-reported questionnaires, these studies have yielded some data that support the present findings.

Different conclusions have also been drawn about smoking and serum TT in postmenopausal women.33 34 Few studies have assessed the association between SHS and serum TT in females. Li et al recently conducted a secondary analysis of a large randomised controlled trial, the Polycystic Ovary Syndrome Acupuncture and Clomiphene Trial, which involved 27 hospitals during 2012–2015 in China, and found that females exposed to SHS (n=500) had higher serum TT levels (1.7 vs 1.5 nmol/L, p=0.01) and FAI (5.7 vs 4.0, p=0.001) than females who were not exposed (n=271)30; this is somewhat consistent with our findings. Possible explanations for the SHS being associated with higher levels of serum TT are as follows: (1) cotinine increases adrenal androgen production by blocking the adrenal 21-hydroxylase enzyme28 35 and (2) cotinine inhibits TT decomposition.29 FAI, a proxy for circulating free androgen, increases with SHS exposure. This was consistent with the result for SHBG since decreased binding of free androgen by SHBG could cause the level of free androgen to increase.

An increase in SHS was found to be associated with a corresponding increase in E2 levels, both within the general population and never smokers. Smoking/cotinine has been viewed as potentially anti-oestrogenic due to its associations with numerous disorders that have hormonally linked symptoms; however, only a few studies have yielded metabolic data to support this, and they were constrained by a lack of biosampling points or the small number and scope of participants.27 Flouris et al
19 performed a randomised single-blind crossover study that involved 28 non-smoking female adults (n=14) undergoing a 1-hour exposure to moderate SHS and a 1-hour control trial, and found that E2 was significantly reduced after SHS in females.

There are plausible reasons for why the results of our study were inconsistent. On the one hand, the level of serum cotinine may affect the results of serum sex hormones. The mean serum cotinine concentration reached 21.86 ng/mL after SHS in females, which far exceeded the definition of SHS in our study (serum cotinine concentration of 0.05–10 ng/mL). On the other hand, in the control group of females, the serum cotinine level slightly increased after the 1-hour control trial but remained below 10 ng/mL, while the serum E2 concentration also increased slightly (205.21±33.39 vs 209.62±33.41 pg/mL (mean±SD), p<0.05). This also supported the results of our study. Another study concluded that SHS was associated with lower-than-normal median steroid hormone (including E2) concentrations.10 The different results may be attributable to the passive smoking status in our study being determined by a combination of serum cotinine concentrations and a self-reported questionnaire, and only 23.3% of self-reported passive smoking statuses in the study were determined by serum cotinine concentrations, which may have resulted in a large proportion of passive smokers with serum cotinine concentrations below 0.05.

Similarly, there have been some findings supporting our study. A toxicology study demonstrated that exposing pregnant B6C3F1 mice to mainstream cigarette smoke significantly increased their serum oestrogen levels.36 Some researchers also found that in participants exposed to tobacco smoking, the mean E2 level was higher during the entire cycle compared with those who were not exposed to tobacco smoking.17 Similar conclusions have been drawn in some related studies.23 37 SHS being significantly associated with increased E2 may be related to abnormal functions of the ovarian or adrenal glands due to interference from cigarette smoke.

SHS was significantly associated with decreased levels of SHBG in former smokers. Some researchers have found similar results that females exposed to SHS had a lower SHBG than non-exposed females (30.1 vs 35.6 nmol/L, p=0.03). However, our study did not distinguish the specific population exposed to SHS, so we could not determine the SHBG status among former smokers. Several possible mechanisms could explain the association between SHS and SHBG in former smokers. The liver is a major organ for SHBG synthesis and secretion. SHS is considered to be able to affect liver functions,38 39 and some researchers have observed that cotinine can inhibit liver function,40 and so decreased SHBG levels in serum may have resulted from cotinine mediating the synthesis and secretion of SHBG from the liver. The liver function of former smokers may be further impaired compared with that of non-smokers who were also exposed to SHS due to the sustained effect of cotinine in cigarette smoke when they are active smokers, which may have caused the lower SHBG concentration in former smokers. However, this phenomenon may also have been coincidental due to the small number of former smokers. The specific mechanism underlying this phenomenon should therefore be further explored using more appropriate standardised methods and epidemiology studies of larger samples.

We found that SHS exposure was associated with increased levels of FAI in females, and the association remained significant in postmenopausal women. Compared with premenopausal women, we found a significant increase in FAI levels in postmenopausal women after SHS exposure. The difference between postmenopausal and premenopausal women may be due to changes in the internal hormonal environment during menopause, which in turn changes the sensitivity to environmental chemicals (eg, those in SHS).41 42 The synthesis and metabolism of certain hormones after menopause may be affected, thus affecting their levels in the body.

In addition, we did not take into account the cotinine cut-off points established by race based on existing NHANES Study.43 There are two major reasons. On the one hand, our study delves into the association between SHS exposure and hormone levels. The definition of SHS exposure also relies on cotinine concentrations. We have concerns that applying race-specific cotinine thresholds might conflict with the cotinine levels associated with SHS exposure, making it challenging to differentiate between the two exposures accurately. On the other hand, we agree with Eneanya et al
44 that cut-off points or categorisations based on race generally have been deemed inappropriate. Overemphasising racial differences is not conducive to promoting the concept of health equity.

Several limitations of our study should be considered. First, while our multivariable model addresses potential confounding to elucidate the association between SHS and serum sex hormones, we recognise that our cross-sectional design precludes definitive causal conclusions. Our approach assumes that included covariates are related to both exposure and outcome but does not imply causality. Future studies, particularly those employing longitudinal designs, are needed to explore these relationships within an explicit causal framework. Second, we recognise the potential for non-differential misclassification bias due to the reliance on self-reported data for SHS. However, our study’s outcome—serum sex hormone levels—is measured through objective laboratory assays, negating the risk of recall bias affecting our results. This objective measurement strengthens the reliability of our findings, although we acknowledge the inherent limitations of self-reported exposure data. Third, some unknown potential confounders were not adjusted for, which may have introduced bias into the results. Besides, among former smokers, we were unable to determine the effect of length of abstinence on the association between SHS and sex hormones through stratified analyses, due to the smaller sample size of former smokers. Finally, while this study analysed a representative US database, the results may be limited by extrapolation.

This post was originally published on