Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease with an estimated worldwide prevalence of 32.4%.1 The multisystem condition is related to an increased risk of liver-related and cardiovascular extrahepatic diseases.2 Smoking is the leading preventable risk factor for premature disability and mortality. As NAFLD and smoking are both associated with the development of metabolic features, there has been increasing interest in testing the relationship between smoking and NAFLD. The causal relevance of smoking to NAFLD incidence have been implicated in cross-sectional and prospective cohort studies.3,4 However, previous studies mainly focused on the effect of active smoking, discussion on the influence of passive smoking on NAFLD will facilitate illustrating the association between smoking and NAFLD. We conducted a national two-center cross-section study of active, passive smoking and NAFLD risk in Chinese and European population. This design allowed us to (1) test for associations between active, passive smoking and NAFLD risk by sex and consolidate evidence for causality by estimating dose-response relationship, (2) identify mediated factors, and (3) resolve their possible interaction with smoking.
In this study, we reported results from the UK Biobank study (application number: 85248), a large-scale, prospective study which recruited more than 500,000 participants from 40 to 70 years of age,5 and independently validated the association in a Chinese population, the Nanjing Health Examination Cohort (NJHE Cohort). All participants provided written informed consent. The UK Biobank obtained ethical approval from the NHS National Research Ethics Service. After excluded those with missing data on NAFLD diagnosis and smoking status, high alcohol consumption or baseline liver diseases (Supplementary Table 1), 14,348 and 11,211 participants were included in further analysis, respectively (Supplementary Fig. 1). Smoking status was classified as non-, current, former, and passive smokers. NAFLD diagnosis was based on the presence of three findings, (1) evidence of hepatic steatosis by either histology or imaging, (2) without heavy alcohol consumption, (3) without history of specific diseases that could lead to steatosis.
Log-binomial logistic regression was used to examine the association of active, passive smoking with NAFLD risk by sex, in which age, ethnicity, education level, physical activity, and drinking status were adjusted. To assess the influence of potential mediating factors, we further adjusted for body mass index (BMI), triglycerides, fasting blood glucose, high-density lipoprotein cholesterol (HDL-C) and waist-hip ratio (UK Biobank only) and performed mediation analysis using the above variables as potential mediators. Stratified analyses were conducted to assess interaction effects of aforementioned variants. Generalized additive models (GAMs) were used to evaluate nonlinear relations between exposure to secondhand smoke and the value of liver proton density fat fraction (PDFF) among nonsmokers in UK Biobank. Mendelian randomization (MR) analyses between smoking initiation and NAFLD were performed to consolidate the association. We then calculated category-specific population attributable fraction (PAF), a fraction of total NAFLD risk in the population that would be eliminated if persons in the specific exposure category shifted to the low-risk group.6 We performed a series of sensitivity analyses to test the stability of the results. First, we restricted the analysis to participants with complete covariates. Second, in order to consist with the definition in UK Biobank, we redefined passive smoker in the NJHE Cohort. Third, we restricted the analysis to those 40 years of age or older in the NJHE Cohort to coincide with the age distribution in UK Biobank. Fourth, we adopted a more rigorous exclusion criteria to ensure the specificity of smoking with NAFLD. Detailed information on methods and statistical analyses are provided in the Supplementary File 1.
The baseline characteristics of study participants from the UK Biobank and NJHE Cohort are shown in Table 1. We observed a higher prevalence of NAFLD in Chinese men (40.03%), but a lower rate in Chinese women (13.40%). Compared with nonsmokers, current smokers had a significantly higher NAFLD risk in both sexes in the UK Biobank (adjusted OR, 1.39 [1.06–1.81] and 1.75 [1.37–2.24] for men and women, respectively; Fig. 1). An equivalent effect for men was observed in the NJHE Cohort, with an adjusted OR of 1.36 [1.19–1.57]. Female passive smoking in women was found prominently associated with an increased risk of NAFLD (adjusted OR, 1.60 [1.37–1.89] and 1.26 [1.07–1.48] in the UK Biobank and NJHE Cohort, respectively), not in men. Strength of the aforementioned association was weakened after adjusted for potential mediators. The results were substantially unchanged in sensitivity analyses (Supplementary Tables 2–5). What is more, greater cumulative cigarettes consumptions, shorter durations of smoking cessation and longer duration of secondhand smoke exposure showed stronger associations with NAFLD risk (Supplementary Table 6). The dose-response relationship was also found through the GAM (Supplementary Fig. 2), where substantially rising curves were observed when we assessed the association between absolute exposure of secondhand smoke and value of liver PDFF among female nonsmokers in the UK Biobank (p=0.001 for exposure at home and p<0.001 for exposure outside home). Ensured the absence of pleiotropy (MR Egger’s intercept p-value=0.086) and homogeneity (the p-value for heterogeneity was 0.218) of instrumental variables, we used the results from inverse-variance weighted (IVW) in MR analysis (Supplementary Table 7), and found that genetic liability to smoking initiation was positively associated with NAFLD (OR, 1.76 [1.29–2.41]; p=3.65×10−4).
Table 1Baseline characteristics of participants according to NAFLD by sex†
Characteristic | UK Biobank, n=14,348
| NJHE Cohort, n=11,211
|
---|
Men
| Women
| Men
| Women
|
---|
NAFLD, n=1,738 | No NAFLD, n=4,332 | p-value | NAFLD, n=1,595 | No NAFLD, n=6,683 | p-value | NAFLD, n=2,029 | No NAFLD, n=3,040 | p-value | NAFLD, n=823 | No NAFLD, n=5,319 | p-value |
---|
Age, year | 55.05 (7.42) | 55.25 (7.80) | 0.363 | 54.41 (7.03) | 53.66 (7.36) | <0.001 | 39.09 (10.91) | 36.55 (10.79) | <0.001 | 44.89 (11.84) | 37.46 (9.40) | <0.001 |
Ethnicity | | | | | | | | | | | | |
White ethnicity/Han nationality | 1,685 (96.95) | 4,200 (96.95) | 0.657 | 1,551 (97.24) | 6,536 (97.80) | 0.220 | 1,984 (97.78) | 2,982 (98.09) | 0.443 | 811 (98.54) | 5,226 (98.25) | 0.550 |
Other | 50 (2.88) | 119 (2.75) | | 42 (2.63) | 133 (1.99) | | 45 (2.22) | 58 (1.91) | | 12 (1.46) | 93 (1.75) | |
Unknown | 3 (0.17) | 13 (0.30) | | 2 (0.13) | 14 (0.21) | | 0 | 0 | | 0 | 0 | |
Education level | | | | | | | | | | | | |
College or University degree | 702 (40.39) | 2,192 (50.60) | <0.001 | 576 (36.11) | 3,083 (46.13) | <0.001 | 1,899 (93.59) | 2,907 (95.63) | 0.001 | 712 (86.51) | 5,023 (94.44) | <0.001 |
Other level | 875 (50.35) | 1,810 (41.78) | | 836 (52.41) | 3,109 (46.52) | | 130 (6.41) | 133 (4.38) | | 110 (13.37) | 295 (5.55) | |
Unknown | 161 (9.26) | 330 (7.62) | | 183 (11.47) | 491 (7.35) | | 0 | 0 | | 1 (0.12) | 1 (0.02) | |
Body mass index, kg/m2 | 29.16 (4.10) | 26.19 (3.34) | <0.001 | 29.56 (5.09) | 25.25 (4.04) | <0.001 | 26.71 (3.02) | 23.44 (2.50) | <0.001 | 25.62 (3.33) | 21.49 (2.33) | <0.001 |
Physical activity | | | | | | | | | | | | |
Yes | 744 (42.81) | 2,181 (50.35) | <0.001 | 600 (37.62) | 2,965 (44.37) | <0.001 | 416 (20.50) | 489 (16.09) | <0.001 | 242 (29.40) | 1,559 (29.31) | 0.393 |
No | 773 (44.48) | 1,632 (37.67) | | 706 (44.26) | 2,554 (38.22) | | 1,613 (79.50) | 2,547 (83.78) | | 577 (70.11) | 3,748 (70.46) | |
Unknown | 221 (12.72) | 519 (11.98) | | 289 (18.12) | 1,164 (17.42) | | 0 | 4 (0.13) | | 4 (0.49) | 12 (0.23) | |
Drinking status | | | | | | | | | | | | |
Never | 44 (2.53) | 104 (2.40) | 0.985 | 64 (4.01) | 191 (2.86) | 0.107 | 1,520 (74.91) | 2,312 (76.05) | 0.488 | 791 (96.11) | 5,155 (96.92) | 0.003 |
Former | 47 (2.70) | 114 (2.63) | | 33 (2.07) | 149 (2.23) | | 9 (0.44) | 11 (0.36) | | 0 | 1 (0.02) | |
Current | 1,646 (94.71) | 4,112 (94.92) | | 1,498 (93.92) | 6,342 (94.90) | | 499 (24.59) | 717 (23.59) | | 30 (3.65) | 163 (3.06) | |
Unknown | 1 (0.06) | 2 (0.05) | | 0 | 1 (0.01) | | 1 (0.05) | 0 | | 2 (0.24) | 0 | |
Smoking status | | | | | | | | | | | | |
Nonsmoker | 914 (52.59) | 2,628 (60.66) | <0.001 | 900 (56.43) | 4,413 (66.03) | <0.001 | 1,003 (49.43) | 1,742 (57.30) | <0.001 | 502 (61.00) | 3,491 (65.63) | 0.051 |
Passive smoker | 258 (14.84) | 626 (14.45) | | 252 (15.80) | 755 (11.30) | | 290 (14.29) | 450 (14.80) | | 315 (38.27) | 1,785 (33.56) | |
Former smoker | 472 (27.16) | 902 (20.82) | | 346 (21.69) | 1,258 (18.82) | | 160 (7.89) | 165 (5.43) | | 1 (0.12) | 15 (0.28) | |
Current smoker | 94 (5.41) | 176 (4.06) | | 97 (6.08) | 257 (3.85) | | 576 (28.39) | 683 (22.47) | | 5 (0.61) | 28 (0.53) | |
Pack years of smoking | 24.32 (18.70) | 19.37 (15.80) | <0.001 | 20.26 (15.54) | 15.66 (12.14) | <0.001 | 12.11(12.97) | 10.24(12.79) | 0.004 | 5.95(6.78) | 2.40(3.31) | 0.042 |
Duration of smoking cessation, year | 18.51 (11.49) | 22.04 (11.57) | <0.001 | 17.62 (11.26) | 19.41 (11.12) | 0.005 | 6.91(8.30) | 5.96(7.24) | 0.366 | 1.13(1.24) | 4.58(3.62) | 0.012 |
Exposure of SHS at home, hour per week‡ | 0.50 (4.16) | 0.21 (2.59) | 0.007 | 0.41 (3.56) | 0.27 (2.88) | 0.168 | 19.07 (13.06) | 16.01 (12.38) | 0.007 | 21.17 (11.03) | 16.72 (10.13) | <0.001 |
Exposure of SHS outside home, hour per week§ | 0.48 (2.58) | 0.32 (1.62) | 0.013 | 0.38 (1.89) | 0.21 (0.86) | <0.001 | 2.01 (2.65) | 1.86 (2.58) | 0.216 | 1.58 (2.54) | 1.58 (2.39) | 0.999 |
Fasting blood glucose, mmol/L | 5.16 (1.17) | 4.95 (0.83) | <0.001 | 5.07 (1.05) | 4.91 (0.71) | <0.001 | 5.34 (1.18) | 4.99 (0.65) | <0.001 | 5.47 (1.35) | 4.87 (0.50) | <0.001 |
Triglycerides, mmol/L | 2.27 (1.17) | 1.76 (0.91) | <0.001 | 1.84 (0.92) | 1.31 (0.65) | <0.001 | 2.18 (1.53) | 1.33 (0.82) | <0.001 | 1.84 (1.10) | 1.01 (0.49) | <0.001 |
HDL-C, mmol/L | 1.19 (0.24) | 1.31 (0.27) | <0.001 | 1.48 (0.31) | 1.66 (0.34) | <0.001 | 1.14 (0.20) | 1.29 (0.25) | <0.001 | 1.29 (0.25) | 1.54 (0.29) | <0.001 |
Waist-hip Ratio | 0.95 (0.06) | 0.91 (0.06) | <0.001 | 0.85 (0.06) | 0.79 (0.06) | <0.001 | – | – | – | – | – | – |
In stratified analysis (Supplementary Figs. 3–4), the associations between smoking status and NAFLD risk were stronger among those with abnormal metabolic condition. On this basis, we found that the percentages of the effect mediated by BMI, triglycerides, HDL-C, and waist-hip ratio were estimated as 25.00%, 10.71%, 2.38% and 14.29% in the association between female passive smoking and NAFLD risk in the UK Biobank (Fig. 2), and the relations between active smoking and NAFLD risks were significantly mediated by aforementioned factors in both sexes (Supplementary Fig. 5). PAFs for population counterfactuals were reported in Supplementary Table 8. In the UK Biobank, if passive smokers avoided exposure to secondhand smoke, 5.70% [3.40–7.95%] of observed NAFLD cases could have been averted in the whole population, while an absolute higher PAF (8.25% [5.19–11.22%]) was calculated for women. More detailed description of results is provided in Supplementary File 2.
In our study, patterns of association between smoking status and NAFLD varied among the multi-ethnic populations. Both male and female active smokers were related to increased risk of NAFLD in the UK Biobank, while the positive association was only observed among men in the NJHE Cohort. Similarly, a cross-sectional analysis also reported a positive association between current smoking and NAFLD risk among Korean men but not among women.7 Low exposure rate of smoking in Asian women may account for the sex heterogeneity.8 Additionally, dose-response relationship results further reinforced the causality correlation. In the present study, accumulated pack-year was strongly associated with the severity of NAFLD. Among former smokers who have quit smoking for more than 15 years, the association with the risk of NAFLD was weakened. This variation trend might be regulated by a greater decrease in insulin resistance, which was verified in a Korean study.9 Previous studies on a potential association between passive smoking and the NAFLD risk have shown conflicting results. A Finnish longitudinal cohort revealed that passive smoking in both child and adult lives were associated with increased risk of adult fatty liver.3 Data from the National Health and Nutrition Examination Survey failed finding the positive association between NAFLD and serum cotinine level.10 Our study demonstrated that a longer duration of secondhand smoke exposure was significantly associated with the risk of NAFLD in both sexes.
Smoking is proved to be associated with an increase in low-density lipoprotein cholesterol (LDL-C), plasma triglycerides, and insulin resistance as well as a decrease in plasma HDL-C levels,11,12 which are also relevant to the occurrence of NAFLD.13 Emerging evidence now suggests that nicotine in the blood exacerbates hepatic steatosis through increased oxidative stress, hepatocellular apoptosis, and decreased phosphorylation (inactivation) of adenosine-5-monophosphate-activated protein kinase, leading to increased hepatic lipogenesis.14 Given these findings, we explored the potential pathways and affirmed that the associations of smoking with NAFLD were mediated through above factors, consistent with previous findings that cigarette smoking is a cofactor of lipid profiles in hepatic steatosis,15 highlighting the necessity of smoking cessation, especially among those of abnormal metabolic markers who were more vulnerable to fatty liver. To the best of our knowledge, this is the first study to reveal the public health implications of cigarette control on the incidence of NAFLD using the calculation of PAFs. Among women about 7% of NAFLD cases could be attributed to secondhand smoke exposure, suggesting that effective strategies should be implemented on preventing secondhand smoke exposure.
The present study has several limitations. The cross-sectional study limited our ability to establish a temporal relationship between smoking and NAFLD, and internal exposure such as serum cotinine level requires evidence of the association of smoking exposure and NAFLD. In summary, our study extends the range of adverse health outcomes positively associated with cigarette exposure, lending robust support to smoking intervention on the reduction of NAFLD in multi-ethnic populations.
Supporting information
Supplementary Table 1
ICD-10 and ICD-9 codes used to define baseline liver disease in UK Biobank.
(DOCX)
Supplementary Table 2
Association between risk of NAFLD and smoking status restricted to participants with complete covariates.
(DOCX)
Supplementary Table 3
Association between risk of NAFLD and smoking status after redefining passive smoker in NJHE Cohort.
(DOCX)
Supplementary Table 4
Association between the risk of NAFLD and smoking status restricted to participants aged 40 years or older in NJHE Cohort.
(DOCX)
Supplementary Table 5
Association between risk of NAFLD and smoking status after adopting a more rigorous exclusion criteria.
(DOCX)
Supplementary Table 6
Associations of NAFLD and different smoking statuses by sex.
(DOCX)
Supplementary Table 7
Association of genetically-predicted smoke initiation with NAFLD in MR.
(DOCX)
Supplementary Table 8
PAFs% (95% CIs) for incident nonalcoholic fatty liver disease by smoking status.
(DOCX)
Supplementary Fig. 1
Flowchart for selection of study participants.
(DOCX)
Supplementary Fig. 2
Nonlinear relations between exposure to secondhand smoke and value of liver PDFF among never-smokers in UK Biobank.
PDFF, proton density fat fraction; SHS, secondhand smoke. Generalized additive models were used to estimate degree of freedom and p-values.
(DOCX)
Supplementary Fig. 3
Association between smoking status and risk of nonalcoholic fatty liver disease stratified by BMI, TG, HDL-C, FBG, and WHR in UK Biobank.
BMI, body mass index; FBG, fasting blood glucose; HDL-C, high-density lipoprotein cholesterol; NAFLD, nonalcoholic fatty liver disease; TG, triglycerides; WHR, waist-hip ratio. Risk estimates were adjusted for age, ethnicity, physical activity, education level, and drinking status.
(DOCX)
Supplementary Fig. 4
Association between smoking status and risk of nonalcoholic fatty liver disease stratified by BMI, TG, HDL-C, and FBG in NJHE Cohort.
BMI, body mass index; FBG, fasting blood glucose; HDL-C, high-density lipoprotein cholesterol; NAFLD, nonalcoholic fatty liver disease; TG, triglycerides. Risk estimates were adjusted for age, ethnicity, physical activity, education level, and drinking status.
(DOCX)
Supplementary Fig. 5
Mediation effects of BMI, TG, HDL-C, FBG and WHR on the association between different smoking status and risk of nonalcoholic fatty liver disease by sex in UK Biobank.
*p<0.05 for coefficients different from 0. Data are regression coefficients with adjustment for age, ethnicity, physical activity, education level, and drinking status. BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; NAFLD, nonalcoholic fatty liver disease; TG, triglycerides.
(DOCX)
Supplementary File 1
Supplementary Methods.
(DOCX)
Supplementary File 2
Supplementary Results.
(DOCX)
Abbreviations
- BMI:
body mass index
- CI:
confidence interval
- GAM:
generalized additive model
- HDL-C:
high-density lipoprotein cholesterol
- IVW:
inverse-variance weighted
- LDL-C:
low-density lipoprotein cholesterol
- MR:
mendelian randomization
- MRI:
magnetic resonance imaging
- NAFLD:
nonalcoholic fatty liver disease
- NJHE Cohort:
Nanjing Health Examination Cohort
- OR:
odds ratio
- PAF:
population attributable fraction
- PDFF:
proton density fat fraction
Declarations
Acknowledgement
This research was conducted using the UK Biobank Resource (Application Number: 85248). We thank the investigators and participants in the UK Biobank and Nanjing Health Examination Cohort for their contributions to this study.
Ethical statement
UK Biobank has full ethical approval from the NHS National Research Ethics Service (21/NW/0157).
Data sharing statement
The Nanjing Health Examination Cohort data used to support the findings of this study have not been made available; The UK Biobank data are available from https://www.ukbiobank.ac.uk/. Restrictions apply to the availability of these data, which were used under license for the current study (Project ID: 85248). Data are available for bona fide researchers upon application to the UK Biobank.
Funding
This work was supported by the National Natural Science Foundation of China (No. 81903382); Natural Science Foundation of Jiangsu Province (Nos. BK20190652, BK20220320); Science and Technology Young Scientific and Technological Talents Project of Jiangsu Province (No. 2021-50); China Postdoctoral Science Foundation (Nos. General Program, 2019M651900, 2021M701757); CAMS Innovation Fund for Medical Sciences (No. 2019RU038).
Conflict of interest
The authors have no conflict of interests related to this publication.
Authors’ contributions
Conceived the study design and supervised the entire project (HS, CS), data interpretation, data analysis, and writing of the draft (XG, CS), and study design and data interpretation in this analysis (XG, CY, JL). All authors reviewed or revised the draft, and approved the submitted draft.