Introduction

Insulin resistance is a growing metabolic disorder worldwide and is associated with some of the most common diseases affecting the modern society, including diabetes, high blood pressure, obesity, and coronary heart disease1. Direct methods of assessing insulin resistance include euglycemic-hyperinsulinemia clamp and insulin suppression tests and simple indirect indicators are estimated by the homeostasis model assessment of insulin resistance (HOMA-IR)2,3,4. However, these tests are invasive, complex, and expensive, making their application difficult in large-scale population studies and clinical practice5. Recently, the triglyceride and glucose (TyG) index, a simple and accurate marker of insulin resistance, has been proposed, which uses fasting triglyceride and blood glucose levels for calculation6. The TyG index can help screen people at high risk of diabetes mellitus with a simple blood test. Furthermore, studies using the TyG index in adults in the Republic of Korea showed that an increase in the TyG index was associated with an increase in the prevalence of coronary artery calcification or arterial stiffness7,8 and suggested that it is a useful tool for evaluating insulin resistance6.

Smoking is a lifestyle factor that may directly or indirectly affect insulin resistance9. Several prospective studies on the relationship between smoking and insulin resistance have shown that smoking is a risk factor for insulin resistance10,11,12,13. However, these studies have mostly used self-reporting as a method of measuring exposure to smoking, and this may have led to incorrect measurement, as self-reported and biomarker results show a consistency of only 46–53%; in addition, self-reports tend to be unreliable for quantitative assessments of smoking volume14,15. These findings suggest that an objective method of measuring smoking volume is required to account for the inherent bias in self-reported data.

Cotinine is the main metabolite of nicotine present in the blood, urine, hair, and saliva and is considered an indicator of exposure to nicotine smoke or current smoking16. While nicotine has a half-life of around 2 h in the blood, cotinine has a half-life of 18–24 h and reflects the accumulated exposure to environmental tobacco smoke17. In particular, urine cotinine levels may help determine the contribution of smoke in the air during the sampling process to the total smoking exposure18. 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNAL) has been used extensively to assess the accuracy of self-reported smoking status19,20. NNAL is widely known as a biomarker of nicotine-derived nitrosamine ketone, a tobacco-specific lung carcinogen21,22. Furthermore, NNAL, due to its half-life of approximately 40 days, is useful for its tobacco specificity, association with carcinogen intake, facilitation of consistent detection of people exposed to tobacco, and evaluation of long-term exposure to harmful substances23.

Identifying an association between cotinine and NNAL, objective biomarkers of tobacco exposure, and insulin resistance may help assess the effect of smoking on the risk of insulin resistance. Although several studies based on self-reported data have been published regarding the influence of smoking on the risk of insulin resistance, to the best of our knowledge, no studies have examined this effect using objective smoking-related biomarkers. Therefore, this study investigated the relationship between smoking patterns and insulin resistance using cotinine and NNAL as biomarkers of tobacco exposure.

Results

Demographic characteristics

Of 4,043 participants, 2,067 (51.1%) were males (Table 1). Of the 2,067 male participants, 839 (40.6%), 454 (22.0%), 12 (0.6%), and 762 (36.9%) were continuous-smokers, past-smokers, current-smokers, and non-smokers, respectively. Of the 1976 (48.9%) female participants, 201 (10.2%), 452 (22.9%), 22 (1.1%), and 1301 (65.8%) were continuous-smokers, past-smokers, current-smokers, and non-smokers, respectively. The insulin resistance groups differed with respect to all factors except educational levels, household income, region, occupational categories, energy intake levels, secondhand smoking exposure, and the survey year.

Table 1 General characteristics of the study population.
Full size table

Association between smoking patterns and insulin resistance

Table 2 presents the associations between smoking patterns and insulin resistance for male and female participants after adjusting for all control variables. Compared to non-smokers, men who were continuous-smokers (odds ratio [OR] = 1.74, 95% confidence interval [CI] = 1.27–2.38) and past-smokers (OR = 1.47, 95% CI = 1.03–2.09) were at an increased risk of insulin resistance. Similarly, compared to non-smokers, women who were continuous-smokers (OR = 2.01, 95% CI = 1.33–3.03) and past-smokers (OR = 1.37, 95% CI = 1.00–1.87) were at an increased risk of insulin resistance.

Table 2 Association between short-term smoking patterns and the triglyceride and glucose index.
Full size table

Table 3 presents the results of subgroup analyses stratified by the independent variable. Compared to non-smokers, male participants in the drinking group had an increased risk of insulin resistance in both the continuous-smoker and past-smoker groups (OR = 2.08, 95% CI = 1.53–2.64 and OR = 1.80, 95% CI = 1.23–2.64, respectively); female participants in the drinking group had an increased insulin resistance risk in the continuous-smoker group (OR = 1.98, 95% CI = 1.25–3.13).

Table 3 Subgroup analysis stratified by independent variables.
Full size table

Obesity affected the risk of insulin resistance in male continuous-smokers and past-smokers (OR = 2.23, 95% CI = 1.33–3.76 and OR = 2.08, 95% CI = 1.15–3.76, respectively). Among women, relative to participants with normal weight, underweight participants who were continuous-smokers had the highest risk of insulin resistance (OR = 1.70, 95% CI = 1.02–2.82). Additionally, the risk of insulin resistance was the highest in the continuous-smoker group among men who rarely walked (OR = 2.42, 95% CI = 1.55–3.79) and among women who walked sufficiently (OR = 2.26, 95% CI = 1.13–4.51). Relative to non-smokers, male continuous-smokers and past-smokers with adequate energy intakes had an increased risk of insulin resistance (OR = 2.10, 95% CI = 1.33–3.32 and OR = 1.79, 95% CI = 1.12–2.87, respectively); for women, this association was observed in the past-smoker group (OR = 1.77, 95% CI = 1.05–2.98).

Discussion

Several previous studies have examined the relationship between smoking and insulin resistance using self-reported data; however, studies on this relationship using biomarkers remain rare. Therefore, this study is one of the few studies to investigate the relationship between smoking patterns and insulin resistance using biomarkers.

Our study found that NNAL and cotinine concentrations in short-term smoking patterns were associated with insulin resistance risk in continuous- and past-smokers who met the smoking criteria. We also found that continuous smoking was significantly associated with the highest risk of insulin resistance in both men and women. However, no association was found between current smokers and insulin resistance; in this group, the smoking criteria were based only on cotinine levels. This suggests that groups with short smoking durations of around 16–20 h could be protected from complications, such as insulin resistance, by applying smoking cessation guidelines and practices.

The findings of the present study are consistent with those of previous studies, despite the use of different data sources13. Furthermore, our findings indirectly support those of prior studies regarding a dose–response relationship between smoking and insulin resistance12,24. Previous studies have shown that the amount and duration of smoking increase the risk of insulin resistance in a dose-dependent manner24. This finding may be due to hormonal changes associated with smoking. Moreover, smoking may induce insulin resistance directly, owing to its effect on abdominal obesity, which may partly occur due to nicotine absorption during smoking9. Another possible mechanism involves the smoking-triggered secretion of hormones such as cortisol, catecholamines, and growth hormones, which oppose the effects of insulin. These hormones increase lipolysis, subsequently increasing free fatty acid release and impairing endothelial function, which may contribute to insulin resistance12. Finally, smoking is negatively associated with adiponectin levels in a dose–response manner25. Therefore, these mechanisms indirectly support the association between continuous and past smoking (representing continuous smoking for > 40 days in our sample) and the risk of insulin resistance.

The stratified subgroup analysis we conducted revealed that continuous-smokers and past-smokers were at an increased risk of insulin resistance; specifically, men with a high body mass index (BMI) had an OR that was more than two-fold higher than that of non-smokers. Previous studies have suggested that smoking may cause insulin resistance by triggering processes associated with fat accumulation in the abdomen and increasing the waist-to-hip circumference ratio9. In addition, an increased body fat percentage has been shown to increase blood levels of non-esterified fatty acids, glycerol, hormones, pro-inflammatory cytokines, and other factors involved in the development of insulin resistance26, suggesting that a high BMI may increase the insulin resistance risk. Additionally, in both male and female participants, continuous-smokers with unhealthy behaviors, such as alcohol intake and lack of exercise, had a more than two-fold higher risk of insulin resistance than their counterparts. This finding supports those from previous studies on the association between unhealthy behaviors, including alcohol consumption and smoking, and serious metabolic abnormalities27, including insulin resistance. Moreover, continuous-smokers and past-smokers with adequate energy intakes had a two-fold higher risk of insulin resistance than their counterparts. In this study, energy intake was stratified into categories defined by the Korean nutrient intake standards28. However, given that smokers consume fewer essential nutrients such as vitamins, calcium, and potassium than non-smokers, it is likely that smokers meet their energy requirements by eating foods that adversely affect insulin resistance29. These findings may account for the increase in the insulin resistance risk that we observed in continuous-smokers and past-smokers. Further studies on the relationships between nutrient intake, smoking, and insulin resistance are required.

This study had several limitations. First, the cross-sectional study design precludes any meaningful conclusions about causality. Second, although we estimated smoking exposure and insulin resistance using urine and blood samples, respectively, data on the remaining variables were obtained from the Korea National Health and Nutrition Examination Survey (KNHANES VII) data, which were based on self-reported information; consequently, some of the estimates used may have been subject to recall bias. Third, participants with type 2 diabetes mellitus were excluded to help control for confounding factors that could affect insulin resistance; nevertheless, this restriction may have obscured or reduced the association between exposure to smoking and the risk of insulin resistance. Fourth, the study sample size was relatively small, specifically the size of the current-smoker group; this limitation was associated with the data source, whereby only half of the total sample was randomly investigated for NNAL and cotinine levels30. However, the KNHANES survey provides data that are nationally representative and inclusive of biomarker information, whereas previous studies did not examine these parameters. Therefore, to support our findings, future studies using larger sample sizes are required.

This study had various strengths. First, this study was based on biomarkers, in contrast to previous studies based on self-reported data. Second, this study utilized nationally representative data from the Republic of Korea, allowing us to evaluate the association between smoking patterns and insulin resistance using high-quality information; the influence of both recall bias and measurement bias on the findings is likely to be small. Finally, some previous studies have used cotinine levels to estimate smoking exposure. Herein, we also included NNAL concentrations; NNAL has a long half-life, contributing to the analysis of smoking patterns.

In conclusion, this study showed that long-term smokers (≥ 40 days) were at an increased risk of insulin resistance in short-term smoking patterns. Our findings regarding short-term smokers (16–20 h) suggest that smoking cessation may protect against complications such as insulin resistance. Therefore, there is a need to educate first-time smokers about the health benefits of quitting smoking.

Methods

This study was based on data collected by the 2016–2018 KNHANES VII. The KNHANES comprises three parts: health surveys, health check-ups including laboratory tests, and nutrition surveys. The KNHANES is a nationwide population-based cross-sectional survey that has been conducted annually since 1998, under the direction of the Korea Centers for Disease Control and Prevention (KCDC) of the Ministry of Health and Welfare, to accurately assess the population’s health and nutritional status31. The KNHANES was approved by the Institutional Review Board of the KCDC, and written informed consent was obtained from all survey participants. This study adhered to the doctrine of the Declaration of Helsinki for Biomedical Research.

The total number of respondents during 2016–2018 was 24,269. As the KNHANES does not evaluate smoking behavior in participants younger than 19 years, data on this age group were excluded (N = 4880). In addition, we excluded participants with diabetes mellitus and those who were either menstruating or pregnant at the time of data collection (N = 4886). Finally, participants with missing values for NNAL, cotinine, and other independent variables were excluded (N = 10,460). Thus, a total of 4043 participants (2067 men and 1976 women) were evaluated (Fig. 1).

Figure 1
figure 1

Schematic diagram of the study eligibility.

Full size image

Variables

The dependent variable in this study was the TyG index, a product of fasting triglyceride and glucose blood levels, which helps assess insulin resistance32. In the KNHANES data, fasting (starting after 7 p.m. the day before the survey) blood samples were provided for testing. The TyG index was calculated using the formula, ln(triglyceride [mg/dL] × fasting blood glucose [mg/dL]/2), and expressed on a logarithmic scale5.

We defined short-term smoking patterns by measuring the concentrations of NNAL and cotinine. Spot urinary samples were collected for urinary NNAL and cotinine at the time of health checkup. Urine cotinine was examined in all subjects aged 6 years and above at the health checkup, and NNAL was randomized in half of the health checkup subjects30. Fresh urine samples were collected and immediately underwent routine urinalysis, and the remaining aliquots were stored at –20 °C until the analysis of cotinine and NNAL33. Urine concentrations of cotinine and total NNAL (free NNAL plus NNAL-glucuronide) were analyzed by liquid chromatography-tandem mass spectrometry (LC–MS/MS) using Agilent 1100 Series API 4000 (AB Sciex, Foster City, CA, USA) and Agilent 1200 Series Triple Quadrupole 5500 (AB Sciex, Foster City, CA, USA), respectively34,35. The limit of detection was 0.27399 ng/mL for cotinine and 0.1006 pg/mL for NNAL30,36.

In this study, NNAL and cotinine concentrations were used to classify the participants into smoking and non-smoking groups using the smoking concentration standards of the KCDC (2.13 pg/mL and 50 ng/mL for NNAL and cotinine, respectively)37. We defined “short-term smoking pattern” based on half-life values (i.e. 18–24 h, 40 days for cotinine and NNAL)23,38. A "Continuous smoker" was defined as a participant who met both smoking criteria for 18–24 h and 40 days. Thus, both NNAL and cotinine concentrations are participants who meet smoking criteria. "Current-smokers" were defined as those who did not meet the 40 days smoking criteria but met the 18–24 h smoking criteria. Therefore, participants who did not meet the NNAL smoking concentrations criteria but did meet the Cotinine smoking concentrations criteria. "Past-smokers" were defined as those who met the 40 days smoking criteria but not the 18–24 h smoking criteria. Therefore, participants who meet the NNAL smoking concentrations criteria but not the cotinine smoking concentrations criteria. "Non-smoker" was defined as a person who did not meet both smoking criteria for 18–24 h and 40 days. Therefore, both NNAL and cotinine concentrations do not meet the smoking criteria (Fig. S1).

Potential confounding variables included sociodemographic and health-related characteristics and the study year. Sociodemographic characteristics included age, marital status, educational level, household income, region, and occupation. Health-related characteristics included BMI, drinking status, walking frequency, energy intake level, chronic disease diagnosis, secondhand smoking exposure, family history of diabetes, and pack-year estimates.

Statistical analyses

Before the analysis, we excluded cases where there was no response to the variables required for the study (Fig. 1). Therefore, in this study, all estimates were calculated using sample weights assigned to the study participants. The sample weights were constructed by the KNHANES to represent the population in the Republic of Korea while accounting for the complex survey design and survey non-response31. Additionally, we performed a pre-analysis to classify participants into “low” and “high” insulin resistance groups. We analyzed the TyG index using receiver operating characteristic curves to estimate valid cut-off values for impaired fasting glucose levels, and the effective cut-off values for the TyG index were 8.3878 and 8.60248 for men and women, respectively; these were similar to previously reported values39 and were used in this study. A univariate linear regression analysis was conducted to investigate the general characteristics of the study population. Multiple regression analyses were performed and adjusted for covariates to analyze the association between smoking patterns and insulin resistance. Further, subgroup analyses were performed with multiple linear regression models stratified by sex to investigate the associations of BMI, drinking status, walking frequency, and energy intake levels with insulin resistance. ORs and 95% CIs were calculated to compare non-smokers to continuous-smokers, current-smokers, and past-smokers. All statistical analyses were performed using SAS software version 9.4 (SAS Institute, Inc., Cary, NC, USA). Findings were considered significant at P values  < 0.05.