The NHANES is a study project that aims to assess the health and nutritional status of American adults and children (18). There are demographic, socioeconomic, nutritional, and health-related questions in the NHANES interviews. The National Center for Health Statistics (NCHS) Research Ethics Review Board approved the NHANES research protocols, and all participants provided written informed consent.

We looked at NHANES descriptive data from 2005–2010, 2013–2014, and 2017–2018. Participants aged ≥50 years were enrolled, and those with missing serum cotinine and bone mineral density (BMD) data were excluded.

During the household interview, adults aged 20 and up self-reported their smoking status. Participants who claimed to have smoked fewer than 100 cigarettes in their lives were labeled ‘never smokers’. Former smokers were individuals who had smoked more than 100 cigarettes in their lives but had quit, while current smokers were those who were currently smoking.

Cotinine is a primary nicotine metabolite used as a marker of active smoking and as an indicator of exposure to secondhand smoke (19). Serum cotinine was determined by isotope dilution-high-performance liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometry (20). As in previous investigations (21), those below the lower detection limit were considered unexposed. We generated cotinine categories representing smoking exposure and utilized the newly recommended cut-off point of 3 ng/ml by Benowitz et al. (22) to separate smokers from nonsmokers. Cotinine levels were ranked as follows: cotinine <0.05 ng/ml, cotinine 0.05–2.99 ng/ml, and cotinine ≥3 ng/ml.

BMD (measured in grams/cm²) was evaluated using a dual X-ray absorptiometry technique (QDR 4500A fan-beam densitometers [Hologic Inc]) while the participants visited mobile examination centers. The left hip (or right hip, in case of left hip replacement or metal object injection) was routinely scanned to report total BMD of the femur, femoral neck, and trochanter. The exclusion criteria for assessing participants’ dual X-ray absorptiometry followed those of the NHANES recommendations.

The WHO criteria for osteopenia and osteoporosis (23) identify low bone mineral density for male and female individuals aged 50 years and older. This method uses the BMD data from young male and female individuals as threshold values. Male and female individuals aged between 20 and 29 years were selected as the reference group in the current study since prospective data demonstrated femur bone loss in female participants in their thirties (24). Osteopenia was defined as a BMD value that was between 1 and 2.5 standard deviations (SDs) below the mean of male and female participants aged between 20 and 29 years; osteoporosis was defined as a BMD value of more than 2.5 SD below the young reference mean. Both of these conditions were considered to have low bone density. This criteria has been applied to each region of interest.

In the NHANES, data were collected using a standard participant questionnaire administered throughout a household interview, along with a medical assessment for each participant. The covariates considered in this study included age, sex, race, education level, poverty, drinking status, physical activity, energy intake level, total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), serum calcium, history of prednisone or cortisone, self-reported diabetes, self-reported hypertension, self-reported cardiovascular disease (CVD), and self-reported cancer. Poverty was assessed using the poverty income ratio (PIR) and was defined as a PIR of 1 for a particular family. For drinking status, participants were categorized as being nondrinkers, low-to-moderate drinkers (<2 drinks/day in men and <1 drink/day in women), or heavy drinkers (≥2 drinks/day in men and ≥1 drink/day in women). The intake of energy was calculated by averaging the two values for the two 24-hour recall interviews. In line with their physical activity levels, the participants were classified as active, insufficiently active, and inactive (25). Descriptions of each variable are presented in https://www.cdc.gov/Nchs/Nhanes/continuousnhanes/.

Continuous variables were reported as the means (standard deviations) or medians (interquartile ranges) and compared by adopting Student’s t test (normal distribution) or the Mann−Whitney U test (nonnormal distribution). By adopting the chi-square test, categorical variables were represented as absolute values (percentages) and compared. As continuous variables, cotinine levels were log2-transformed to achieve a normal distribution. The “mice” package utilized the random forest algorithm for multiple interpolation of the missing data. All statistical analyses were conducted by utilizing R Statistical Software, version 4.2.0, and results with p values < 0.05 (two-sided) were considered statistically significant.

The connections between tobacco exposure and the prevalence of osteoporosis as well as osteopenia were investigated by adopting three consecutive multivariate logistic regression models. Model 1 was adjusted for age (continuous), sex (male or female), and race/ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, or other race). Model 2 was adjusted for Model 1 by adding education level (below high school, high school, or above high school), drinking status (nondrinker, low-to-moderate drinker, or heavy drinker), family income-poverty ratio (<1.0, or ≥1.0), physical activity (inactive, insufficiently active, active), and total energy intake (log2-transformed). Model 3 was based on Model 2 and included additional adjustments for TC (continuous), HDL-C (continuous), serum calcium (continuous), history of prednisone or cortisone (yes or no), self-reported hypertension (yes or no), self-reported diabetes (yes or no), self-reported cardiovascular disease (yes or no), and self-reported cancer (yes or no). Furthermore, restricted cubic spline (RCS) with three knots (10th, 50th, and 90th percentiles) was adopted to study dose−response associations. Nonlinearity was examined by analysis of variance (ANOVA). By adopting segmented regression, the threshold inflection of linearity was computed to fit the piecewise-linear relationship between tobacco exposure and the prevalence of osteoporosis and osteopenia.

A total of 50463 participants from NHANES 2005–2010, 2013–2014 and 2017–2018 were included. Of these, those aged < 50 (n=36297) and those with missing data on serum cotinine (n=1309) and BMD (n=2293) were excluded. In total, 10564 eligible participants were enrolled ( Figure 1 ). Table S1 presents the baseline characteristics of participants according to NHANES 2005-2010, 2013-2014, and 2017-2018.

In this study, the thresholds for osteoporosis in men were 0.73 gm/cm² or less for the total femur, 0.61 gm/cm² or less for the femoral neck, 0.50 gm/cm² or less for the trochanter, and 0.86 gm/cm² or less for the intertrochanter; among female smokers, the thresholds were 0.65 gm/cm² or less for the total femur, 0.56 gm/cm² or less for the femoral neck, 0.44 gm/cm² or less for the trochanter, and 0.77 gm/cm² or less for the intertrochanter. The cutoff values for osteopenia in men were 0.73 to 0.94 gm/cm², 0.61 to 0.83 gm/cm², 0.50 to 0.69 gm/cm² and 0.86 to 1.11 gm/cm² for the total femur, femoral neck, femoral rotor and intertrochanter, respectively; for women, they were 0.65 to 0.84 gm/cm², 0.56 to 0.76 gm/cm², 0.44 to 0.61 gm/cm² and 0.77 to 0.99 gm/cm² for the total femur, femoral neck, femoral rotor and intertrochanter, respectively ( Table 1 ). Moreover, the overall mean age of the reference group (men or women aged 20 to 29 years) was 24.45 ± 2.91 years ( Table S2 ). A total of 797 (36.0%) participants had cotinine levels ≥3 ng/ml, and 663 (30.0%) participants in the reference group were current smokers. The proportion of cotinine levels was significantly higher in men than in women.

The characteristics of the enrolled individuals are presented in Table 2 . The mean age of the study population was 64.85 ± 9.54 years, and the patients were predominantly male (51.9%). By applying cotinine levels, 2228 (21.1%) participants had cotinine levels ≥3 ng/ml, 2179 (20.6%) participants had cotinine levels between 0.05 and 2.99 ng/ml, and 6157 (58.3%) participants had cotinine levels <0.05 ng/ml. The self-reported smoking status revealed that there were 1747 (16.5%) current smokers, 3612 (34.2%) former smokers, and 5205 (49.3%) nonsmokers among the participants. Participants with cotinine level ≥3 ng/ml and self-reported smokers were more likely to be younger, male, and non-Hispanic Black, to have lower education, to have poverty, to be a drinker, to have a low level of physical activity, to have a history of prednisone or cortisone use, and to have more comorbidity than those with a cotinine level <0.05 ng/ml and those who were self-reported nonsmokers.

Table 3 displays the associations between serum cotinine levels and the prevalence of osteoporosis in both continuous and categorical analyses. Regardless of adjustment for covariates, the continuous analysis revealed that log2-transformed cotinine levels showed a noticeable positive association with the prevalence of osteoporosis. The categorical analysis indicated that the association between higher serum cotinine levels (≥3 ng/ml) and the prevalence of osteoporosis (Model 1: OR=2.27 [1.91-2.69]; Model 2: OR=2.03 [1.70-2.43]; Model 3: OR=2.04 [1.70-2.45]; all p for trend <0.001) remained significant after adjustment for covariates using the lowest serum cotinine levels (<0.05 ng/ml) as a reference. Similar results were observed in that current smokers were highly associated with the prevalence of osteoporosis compared with nonsmokers (Model 1: OR=2.30 [1.90-2.79]; Model 2: OR=2.16 [1.77-2.64]; Model 3: OR=2.16 [1.77-2.65]; all p for trend <0.001).

Further examination of the associations between serum cotinine levels, smoking status, and the prevalence of osteopenia is presented in Table 4 . As illustrated by the continuous analysis, after adjustment for covariates, there was a markedly positive relationship between log2-transformed serum cotinine levels and the prevalence of osteopenia. In Model 3, the categorical analysis revealed that the multivariate odds ratios (95% confidence intervals [CI]) for osteopenia increased monotonically to 0.95 (0.85-1.06) and 1.60 (1.42-1.80) (p for trend < 0.001) with higher serum cotinine levels (≥3 ng/ml). Similar relationships were observed between current smokers and the prevalence of osteopenia compared with nonsmokers (OR=1.70 [1.49-1.94]; p for trend < 0.001).

RCS regression with multivariable-adjusted associations was adopted to demonstrate dose−response associations between log2-transformed serum cotinine levels and the prevalence of osteoporosis as well as osteopenia ( Figure 2 ). Serum cotinine levels were nonlinearly and positively correlated with the prevalence of osteoporosis (p for nonlinearity = 0.001) and osteopenia (p for nonlinearity < 0.001), with inflection points of 5.82 ng/ml and 3.26 ng/ml, respectively.