Figures
Abstract
Backgrounds/Objective
The prevalence of diabetes is increasing rapidly all over the world. However, studies on passive smoking and type 2 diabetes have not been systematically assessed. Therefore, we conducted a meta-analysis to explore whether an association exists between passive smoking and risk of type 2 diabetes.
Methods
We searched PubMed, EMBASE, Cochrane library and Web of Science up to April 9th, 2013, to identify prospective cohort studies that assessed passive smoking and risk of type 2 diabetes. The fixed-effect model was used to calculate the overall relative risk (RR).
Result
4 prospective cohort studies were included for analysis, with a total of 112,351 participants involved. The pooled RR was 1.28 (95% confidence interval (CI) 1.14 to 1.44) comparing those who were exposed to passive smoking with those who were not. Subgroup, sensitivity analysis and publication bias test suggested the overall result of this analysis was robust.
Citation: Wang Y, Ji J, Liu Y-j, Deng X, He Q-q (2013) Passive Smoking and Risk of Type 2 Diabetes: A Meta-Analysis of Prospective Cohort Studies. PLoS ONE 8(7): e69915. https://doi.org/10.1371/journal.pone.0069915
Editor: Adrian V. Hernandez, Universidad Peruana de Ciencias Aplicadas (UPC), Peru
Received: February 1, 2013; Accepted: June 13, 2013; Published: July 26, 2013
Copyright: © 2013 Wang et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The prevalence of diabetes is increasing rapidly all over the world, and it is estimated that 439 million adults will be affected by diabetes by 2030 [1]. Type 2 diabetes(T2DM), which is characterized by reduced insulin sensitivity and relative insulin deficiency [2], consists of over 95% of diabetes worldwide [3]. Therefore, identification of risk factors of T2DM is of significant importance to the primary prevention of this disease.
A recent study reported that 40% of children, 33% of male non-smokers and 35% of female non-smokers were exposed to passive smoking worldwide [4]. It has been shown that passive smoking can cause disease, disability, and death [5]. However, the association between passive smoking and T2DM risk has not been summarized. Therefore, we performed this meta-analysis to systematically assess the association between passive smoking and risk of T2DM based on prospective cohort studies.
Methods
Data sources and searches
We conducted this meta-analysis according to the Meta-Analysis of Observational Studies in Epidemiology guidelines [6]. We performed a systematic search of PubMed, EMBASE, the Cochrane library and Web of Science up to April 9th, 2013 to identify relevant prospective cohort studies regarding the association between passive smoking and risk of T2DM. We also searched the reference lists of all retrieved articles to identify any additional literatures. However, we did not search the gray literature. There was no language restriction.
The search terms were (Diabetes mellitus, type 2 OR Diabetes mellitus OR Prediabetic state OR impaired fasting glucose OR impaired glucose tolerance OR Metabolic syndrome OR Glucose intolerance OR Hyperglycemia OR Glucose metabolism disorders OR Insulin resistance OR Glucose) AND (Tobacco smoke pollution OR Passive smoking OR Air pollution, tobacco smoke OR Second-hand smoking OR Involuntary smoking) (Table 1).
Study selection
We first screened the titles and abstracts of all the articles to identify the possible eligible studies, and then read the full articles to include eligible studies. Studies were included if they met the following criteria: had a prospective cohort design, the exposure was passive smoking, the outcome was T2DM, reported estimates of the odds ratio (OR) or relative risk (RR) or hazard ratio (HR) and its 95% confidence interval (CI) or reported data to calculate them. Only the latest study was included if there were duplicates or data were originated from the same study population. Review or studies that did not report available information were excluded.
Data extraction and quality assessment
Data extraction was conducted by two independent reviewers; disagreements would be resolved by consensus. The reference groups were never smokers who were not exposed to passive smoking, except that in one study the reference group were“those currently exposed to passive smoke but did not actively smoke, irrespective of past smoke” [7], The corresponding risk estimates (including RRs, ORs and HRs) adjusted for the maximum number of confounding variables with corresponding 95%CIs were extracted. We also extracted the following data: name of the first author, publication year, study location, age of the participants, total number of patients and participants involved, percentage of female, information of exposed groups, confounding factors that were adjusted for in the analysis.
A 9-star system based on the Newcastle-Ottawa Scale (NOS) [8] was employed for quality assessment. 4, 2, 3 scores were respectively assigned for selection of study groups, comparability of study groups, assessment of outcomes and adequacy of follow-up. Studies with scores of 0–3, 4–6, 7–9 were considered as low, moderate and high quality, respectively.
Data analysis
The RR was used as the common measure of association across studies. As HR was broadly equivalent to relative risk (RR) [9], [10], HRs were directly considered as RRs. ORs were transformed into RRs according to the formula RR = OR/[(1−P0)+(P0×OR)] where P0 stands for the incidence of T2DM in nonexposed group [11]. In addition, the Miettinen test-based approach was used to calculate the variance of lnRR (variance lnRR = variance lnOR×[lnRR/lnOR]) [12]. Heterogeneity across studies was assessed using the Cochrane Q statistic (significance level at P<0.10) and the I2 statistic [13], [14]. The heterogeneity was considered statistically insignificant if P>0.10 and I2≤50%, then the Mantel-Haenszel fixed-effect model was used to calculate pooled RR among studies. Otherwise, the DerSimonian and Laird [15] random-effect model was used to combine the results. Sensitivity analysis was performed to detect the effects of individual study on the pooled result by omitting one study in each turn. We conducted the subgroup analyses according to study location, percentage of female, study quality, and confounding factors being adjusted for to examine the cause of potential heterogeneity.
Potential publication bias was detected by both Begg's test [16] and Egger's test [17]. In order to further assess the possible effect of publication bias on the pooled RR, the nonparametric trim and fill method [18] was used. This method considers the possibility of hypothetical “missing” studies that might exist, then imputes their RRs, and recalculates a pooled RR which incorporates the hypothetical missing studies as though they actually exist. STATA version 11.0 (Stata Corporation, College Station, TX) was employed to conduct all data analysis.
Results
Literature search
Figure 1 shows the study selection process. A total of 481 articles were identified by the search strategy. 173 articles were removed as they were duplicates, left 308 articles for screening. By screening of titles or abstracts, 303 articles were excluded as they were not cohort studies or not clearly relevant. After reading the full text of the remained 5 articles, we excluded 1 articles [19] because the outcome of interest was not T2DM.. Finally, 4 studies [7], [20]–[22] were included for meta-analysis.
Study characteristics
Table 2 summaries the main characteristics of the selected studies for analysis. The 4 studies were conducted in United States, Japan, Germany and South Korea, respectively. The population size per study ranged from 885 to 100,526, with a total of 112,351 participants involved. 3 studies consisted of males and females while 1 studies involved only females. Adjusted RRs (ORs/HRs) were reported in all studies. Adjustment for potential confounding factors differed across studies, and the main adjusted factors were age, sex, body mass index (BMI), alcohol drinking, family or parental history of diabetes, physical activity and hypertension. The maximum follow-up years were 4, 6, 7 and 24 years, respectively. Based on the quality assessment of NOS, 1 study was in high quality (scored 7) while the other 3 were in moderate quality (2 studies scored 6 and 1 study scored 5).
Main analysis
Figure 2 shows a forest plot presenting the association between passive smoking and type 2 diabetes risk. No statistically significant heterogeneity across studies was found (Pheterogeneity = 0.13, I2 = 47.1%). Meta-analysis of the 4 included studies using fixed-effect model suggested an increased risk of T2DM in those who were exposed to passive smoke compared to those who were not (Overall RR = 1.28, 95% CI: 1.14–1.44).
Subgroup and sensitivity analysis
Table 3 presents the results of subgroup analysis with both models according to study location, percentage of female, quality scores and adjustment for important confounding factors including BMI, family/parental history of diabetes and physical activity. For the results using fixed-model effect model, the associations between passive smoking and T2DM were similar to the overall result in subgroups. For the results by random-effect model, the pooled RRs were generally similar to the overall RR.
Sensitivity analysis was conducted by omitting one study each time and re-calculating the pooled results. The overall risk estimates did not vary materially with a range from 1.21(95% CI: 1.05–1.40) to 1.47(95% CI: 1.24–1.76), indicating that the pooled RR was not substantially influenced by any of the individual study.
Publication bias
Publication bias was not found in the included studies, as suggested by the Begg's and Egger's test (P = 0.17 by Egger's test and P = 0.16 by Begg's test). In addition, considering the limited number of included studies, we used the “fill and trim” method to assess the possible effects of potential publication bias on the pooled RR. Although The “fill and trim” method identified hypothetical 3 “missing” studies, the recalculated overall result with random-effect model continued to show a positive association between passive smoking and T2DM (RR = 1.27, 95% CI: 1.04–1.55).
Discussion
To our knowledge, this is the first meta-analysis of prospective cohort studies on passive smoking and risk of T2DM. Result of the present analysis suggested that passive smoking was associated with a significantly increased risk of T2DM. Subgroup analysis, sensitivity analysis and publication bias test suggested the overall result of this analysis was robust.
A recent meta-analysis [23] based on 25 cohort studies found that active smoking was associated with a 44% increase in RR (RR = 1.44, 95%CI: 1.31–1.58) of T2DM. Nevertheless, in our analysis, passive smoking was associated with a 28% increase in RR of T2DM (RR = 1.28, 95%CI: 1.14–1.44), a smaller increase in RR than active smoking. Similar results were found in the study by Zhang L et al [22]. They found that active smoking can increase 39% (≤14 cigarettes per day, RR = 1.39, 95%CI: 1.17–1.64) or 98% (≥25 cigarettes per day, RR = 1.98, 95%CI: 1.57–2.36) risk of T2DM among nonsmokers, while passive smoking was only associated with 10% (occasionally exposure, RR = 1.10, 95%CI: 0.94–1.23) or 16% (regularly exposure, RR = 1.16, 95%CI: 1.00–1.35) increase in risk of T2DM for nonsmokers who were not exposed to passive smoking. Similar results were also reported in the studies on coronary heart disease and stroke [24], [25]. These might be explained by the dilution of passive smoke by room air, which makes a non-smoker be exposed to less tobacco smoke than an active smoker [26].
Although passive smoking was associated with a smaller increase in RR of diabetes compared with active smoking, the burden of diabetes caused by the high prevalence of passive smoking in workplace and at home should not be neglected. One smoker might result in the exposure of passive smoke to several workmates or families, especially in those countries with insufficient anti-smoking legislation. In places where people are not allowed to smoke, the concentration of nicotine in the air is lower than in places where smoking is allowed [27]. This implies the importance of anti-smoking legislation to the reduction of T2DM risk by population level. Moreover, to achieve a significant reduction of T2DM in the burden to society, prohibitions on both passive and active smoking should be targeted.
Several mechanisms might be involved in the effect of passive smoking on diabetes. The environmental tobacco smoke consists of nearly 5,000 chemical compounds, including 43 known carcinogens, carbon monoxide, nicotine and other toxic ingredients [28]. Nicotine is an important ingredient in cigarette smoke that can cause insulin resistance by affecting insulin action [29]. Animal studies also suggested that prenatal or neonatal exposure to nicotine will lead to loss of pancreatic β-cells [30]. In addition, epidemiological studies have found that exposure to environmental tobacco smoke in the childhood was associated with increased risk of pancreatic cancer [31], [32]; this suggests that tobacco smoke might have a chronic toxic effect to the pancreas. Third, like active smoking, passive smoking has been related to oxidative stress, systemic inflammation and endothelial dysfunction [33], which were strongly involved in insulin resistance and diabetes risk [22].
There are several potential limitations in this study that warrant consideration. First, the results of this analysis were based on prospective cohort studies. However, observational studies cannot prove causality.
Second, although all the included studies controlled several known risk factors for T2DM, including age, sex, BMI, alcohol, and physical activity etc, residual or unmeasured confounding may still affect the observed association. Furthermore, there was heterogeneity between studies in regard to adjusting for confounding factors, which may lead to misleading overall results.
Third, as the exposure status of passive smoking was generally self-reported, there was a possibility that participants tend to narrow their exposure status; and this may lead to underestimate or exaggerate of the risk estimates if the lessened extent in two groups were different. However, Sasaki et al. in their study found that self-reported passive smoking might lead to underestimate the true exposure status and thus underestimate the strength of association between passive smoking and T2DM risk [34].
Fourth, although no publication bias was found by the Begg's and Egger's test, the power to detect bias of this two tests was low with small numbers of studies [16], [17]. However, although the risk estimates were slightly reduced, the fill and trim analysis showed no significant change of the general result. Nevertheless, the possibility of publication bias cannot be fully excluded by this analysis.
Finally, our analysis failed to assess a dose-response relationship between passive smoking and T2DM, because there was only one included study that explored this relationship, although a dose-response relationship was found both in workplace and home in this study [20].
In conclusion, the present meta-analysis suggests that passive smoking is associated with an increased risk of T2DM. However, considering the limited number and moderate quality of included studies, further well-designed studies are warranted to confirm this observed association. The dose-response relationship should be well explored by studies in the future.
Author Contributions
Conceived and designed the experiments: QQH. Performed the experiments: QQH YW JJ. Analyzed the data: YW JJ. Contributed reagents/materials/analysis tools: QQH YW JJ YJL XD. Wrote the paper: QQH YW JJ.
References
- 1. Shaw JE, Sicree RA, Zimmet PZ (2010) Global estimates of the prevalence of diabetes for 2010 and 2030. Diabetes Res Clin Pract 87: 4–14.
- 2. McCrimmon RJ, Ryan CM, Frier BM (2012) Diabetes and cognitive dysfunction. Lancet 379: 2291–2299.
- 3. American Diabetes Association (2010) Diagnosis and Classification of Diabetes Mellitus. Diabetes Care 33: S62–S69.
- 4. berg M, Jaakkola MS, Woodward A, Peruga A, Prüss-Ustün A (2011) Worldwide burden of disease from exposure to second-hand smoke: a retrospective analysis of data from 192 countries. The Lancet 377: 139–146.
- 5. He J, Vupputuri S, Allen K, Prerost MR, Hughes J, et al. (1999) Passive Smoking and the Risk of Coronary Heart Disease — A Meta-Analysis of Epidemiologic Studies. New England Journal of Medicine 340: 920–926.
- 6. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, et al. (2000) Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 283: 2008–2012.
- 7. Hayashino Y, Fukuhara S, Okamura T, Yamato H, Tanaka H, et al. (2008) A prospective study of passive smoking and risk of diabetes in a cohort of workers: the High-Risk and Population Strategy for Occupational Health Promotion (HIPOP-OHP) study. Diabetes Care 31: 732–734.
- 8. Stang A (2010) Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 25: 603–605.
- 9. de Lemos ML (2002) How to survive the survival plots. Lancet 360: 954.
- 10. Spruance SL, Reid JE, Grace M, Samore M (2004) Hazard ratio in clinical trials. Antimicrob Agents Chemother 48: 2787–2792.
- 11. Zhang J, Yu KF (1998) What's the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 280: 1690–1691.
- 12. Miettinen O (1976) Estimability and estimation in case-referent studies. Am J Epidemiol 103: 226–235.
- 13. Higgins JP, Thompson SG (2002) Quantifying heterogeneity in a meta-analysis. Stat Med 21: 1539–1558.
- 14. Higgins JP, Thompson SG, Deeks JJ, Altman DG (2003) Measuring inconsistency in meta-analyses. BMJ 327: 557–560.
- 15. DerSimonian R, Laird N (1986) Meta-analysis in clinical trials. Control Clin Trials 7: 177–188.
- 16. Begg CB, Mazumdar M (1994) Operating characteristics of a rank correlation test for publication bias. Biometrics 50: 1088–1101.
- 17. Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias in meta-analysis detected by a simple, graphical test. BMJ 315: 629–634.
- 18. Duval S, Tweedie R (2000) Trim and fill: A simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 56: 455–463.
- 19. Houston TK, Person SD, Pletcher MJ, Liu K, Iribarren C, et al. (2006) Active and passive smoking and development of glucose intolerance among young adults in a prospective cohort: CARDIA study. BMJ 332: 1064–1069.
- 20. Ko KP, Min H, Ahn Y, Park SJ, Kim CS, et al. (2011) A prospective study investigating the association between environmental tobacco smoke exposure and the incidence of type 2 diabetes in never smokers. Ann Epidemiol 21: 42–47.
- 21. Kowall B, Rathmann W, Strassburger K, Heier M, Holle R, et al. (2010) Association of passive and active smoking with incident type 2 diabetes mellitus in the elderly population: the KORA S4/F4 cohort study. Eur J Epidemiol 25: 393–402.
- 22. Zhang L, Curhan GC, Hu FB, Rimm EB, Forman JP (2011) Association between passive and active smoking and incident type 2 diabetes in women. Diabetes Care 34: 892–897.
- 23. Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz J (2007) Active smoking and the risk of type 2 diabetes: a systematic review and meta-analysis. JAMA 298: 2654–2664.
- 24. Oono IP, Mackay DF, Pell JP (2011) Meta-analysis of the association between secondhand smoke exposure and stroke. J Public Health (Oxf) 33: 496–502.
- 25. Shinton R, Beevers G (1989) Meta-analysis of relation between cigarette smoking and stroke. BMJ 298: 789–794.
- 26. Eriksen MP, LeMaistre CA, Newell GR (1988) Health hazards of passive smoking. Annu Rev Public Health 9: 47–70.
- 27. Sikorska-Jaroszynska MH, Mielnik-Blaszczak M, Krawczyk D, Nasilowska-Barud A, Blaszczak J (2012) Passive smoking as an environmental health risk factor. Ann Agric Environ Med 19: 547–550.
- 28. Brownson RC, Figgs LW, Caisley LE (2002) Epidemiology of environmental tobacco smoke exposure. Oncogene 21: 7341–7348.
- 29. Xie XT, Liu Q, Wu J, Wakui M (2009) Impact of cigarette smoking in type 2 diabetes development. Acta Pharmacol Sin 30: 784–787.
- 30. Bruin JE, Petre MA, Raha S, Morrison KM, Gerstein HC, et al. (2008) Fetal and neonatal nicotine exposure in Wistar rats causes progressive pancreatic mitochondrial damage and beta cell dysfunction. PLoS One 3: e3371.
- 31. Chuang SC, Gallo V, Michaud D, Overvad K, Tjonneland A, et al. (2011) Exposure to environmental tobacco smoke in childhood and incidence of cancer in adulthood in never smokers in the European Prospective Investigation into Cancer and Nutrition. Cancer Causes Control 22: 487–494.
- 32. Vrieling A, Bueno-de-Mesquita HB, Boshuizen HC, Michaud DS, Severinsen MT, et al. (2010) Cigarette smoking, environmental tobacco smoke exposure and pancreatic cancer risk in the European Prospective Investigation into Cancer and Nutrition. Int J Cancer 126: 2394–2403.
- 33. Yanbaeva DG, Dentener MA, Creutzberg EC, Wesseling G, Wouters EF (2007) Systemic effects of smoking. Chest 131: 1557–1566.
- 34. Sasaki S, Braimoh TS, Yila TA, Yoshioka E, Kishi R (2011) Self-reported tobacco smoke exposure and plasma cotinine levels during pregnancy–a validation study in Northern Japan. Sci Total Environ 412–413: 114–118.