Long working hours and alcohol use: systematic review and meta-analysis of published studies and unpublished individual participant data
BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.g7772 (Published 13 January 2015) Cite this as: BMJ 2015;350:g7772All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
We all know that alcohol misuse causes health risk behaviors such as the finding in Vietnam that alcohol is a contributing factor to risky sexual activity and to risky driving behavior (Parker, L 2010). The result of pooled estimates of this meta-analysis show that the prevalence of alcohol use in women with long working hours is higher than that in men (OR: 1.36 vs. 1.11), and people with long working hours usually have a higher education level. It seems like education is a risk factor, and one research study in Vietnamese women reported that those with a higher education had significantly higher rates of alcohol consumption-related problems than those with lower education (Giang, Allebeck, Spak, Van Minh, & Dzung, 2008). The authors also reported that separated, divorced, and widowed women are at higher risk of alcohol consumption than married ones. This finding suggests that the further monitoring of alcohol use and problems among women is important to follow-up changes in consumption pattern (Giang et al., 2008).
Future research should focus on the association between working hour and the frequency and amount of alcohol use to classify the positive versus negative drinking behavior among employees because alcohol use is not always bad. An acceptable amount of alcohol consumption may have a positive effect on health such as in reducing the risk of heart attacks, ischemic strokes and death from all cardiovascular causes. In addition, in certain events alcohol use (with some Champagne and wine) could cheer up and motivate employees after working hard for a long time and help in improving efficiency at work.
This study gives evidence for policy makers and employers to create appropriate working regulations and environment to reduce health risks by reducing alcohol misuse among employees (especially women) and thus improve production efficiency.
References
Parker, L 2010, 'Alcohol consumption behaviours and attitudes in Vietnam: An exploratory analysis', in Rebekah Russell-Bennett; Sharyn Rundle-Thiele (ed.) Proceedings of the International Non-Profit and Social Marketing Conference (INSM), Brisbane, Australia, 15 - 16 July 2010, pp. 143-147.
Giang, K. B., Allebeck, P., Spak, F., Van Minh, H., & Dzung, T. V. (2008). Alcohol Use and Alcohol Consumption–Related Problems in Rural Vietnam: An Epidemiological Survey Using AUDIT. Substance Use & Misuse, 43(3-4), 481-495. doi: doi:10.1080/10826080701208111
Competing interests: No competing interests
Recently, with people working so hard, they get more and more pressure at the office or work environment. One way that people may choose to cope with stress is by turning to alcohol. Drinking may lead to positive feelings and relaxation, at least in the short term. Problems arise, however, when stress is ongoing and people continue to try to deal with its effects by drinking alcohol (Alcohol Alert, 2012).
In some parts of Indonesia theere is also an alcohol drinking problem. The National Household Health Survey (NHHS) 1995 and 2001 only showed prevalence at national level, but in 2007 NHHS with a new name called “Riskesdas” elaborate health indicators up to district levels. Suhardi found the presence of provincial clusters and preferences of alcohol drinkers (Suhardi, 2011).
So, basically I agreed with the link between long working hours and increased alcohol use by Virtanen et al. According to this finding, we still need to explore more how cultural and geographical factors might affect groups. If people have good health literacy associated with risk of alcohol use, this might culturally play a big role in their reason for drinking alcohol.
References:
Suhardi. 2011. Alcohol Drinker preferences in Indonesia according to Riskesdas 2007. Buletin Penelitian Kesehatan. Vol. 39. No.4. 2011:154-164
Alcohol Alert, 2012. Alcohol Research: The Link between Stress and Alcohol. Number 85. Vol. 34. Number 4. National Institute on Alcohol Abuse and Alcoholism.
Competing interests: No competing interests
The link between long working hours and increased alcohol use found in the systemic review and meta-analysis by Virtanen et al. has been widely reported by media worldwide (Virtanen et al., 2015). According to the authors, the observed associations could be explained by stress-alleviating response to prolonged working hours and heavy workloads. We would like to point out that alcohol drinking patterns vary greatly by socioeconomic positions and by culture-specific contexts.
Of the 34 cross-sectional studies and 2 prospective studies included in the meta-analysis, 13 studies were from Japan and 1 study was from Taiwan, where long working hours and karoshi have been major issues of workplace safety and health (Iwasaki, Takahashi, & Nakata, 2006). In a national survey of Taiwanese manual workers, long work hours was found to be associated with alcoholic energy drinks consumption, and most importantly, 60% of them drank at work (Cheng, Cheng, Huang, & Chen, 2012). The mechanism explaining the association between work stress and alcohol use can be different according to drinking contexts (Frone, 2008). The combination of alcohol and caffeine in alcoholic energy drinks rationalized their use in the workplace as to boost energy and increase work efficiency. The specific culture of consuming alcoholic energy drinks at work posed double risks to manual workers, i.e. problem drinking and workplace injury.
In Taiwan and many other Asian countries, prolonged working hours are commonplace and stress-related diseases have been the center of occupational health controversy (Cheng et al., 2014; Cheng, Park, Kim, & Kawakami, 2012; Iwasaki, Takahashi, & Nakata, 2006). While we agree with Virtanen et al. that working hours should be considered in the design of preventive interventions against problematic alcohol use, we further advocate that the use of substances to cope with stress associated with long working hours being considered a major occupational health issue in many Asian countries.
References:
Cheng WJ, Cheng Y, Huang MC, Chen CJ. Alcohol dependence, consumption of alcoholic energy drinks and associated work characteristics in the Taiwan working population. Alcohol and Alcoholism 2012; 47: 372-379.
Cheng Y, Du CL, Hwang JJ, Chen, IS, Chen MF, Su TC. Working hours, sleep duration and the risk of acute coronary heart disease: A case–control study of middle-aged men in Taiwan. International Journal of Cardiology 2014; 171: 419-422.
Cheng Y, Park, J, Kim, Y, Kawakami, N. The recognition of occupational diseases attributed to heavy workloads: experiences in Japan, Korea, and Taiwan. International Archives of Occupational and Environmental Health 2012; 85: 791-799.
Frone MR. Are work stressors related to employee substance use? The importance of temporal context assessments of alcohol and illicit drug use. J Appl Psychol 2008; 93: 199-206.
Iwasaki K, Takahashi, M, Nakata, A. Health problems due to long working hours in Japan: working hours, workers' compensation (Karoshi), and preventive measures. Industrial Health 2006; 44: 537-540.
Virtanen M, Jokela, M, Nyberg, ST, Madsen, IEH, Lallukka, T, Ahola, K, et al. Long working hours and alcohol use: systematic review and meta-analysis of published studies and unpublished individual participant data. British Medical Journal 2015; 350.
Competing interests: No competing interests
Virtanen et al. (2015) reported a positive association between long working hours and alcohol use in their systematic review and meta-analysis conducted from more than 300 000 participants from 14 countries. The authors came to the conclusion that individuals whose working hours exceed standard recommendations are more likely to increase their alcohol use to levels that pose a health risk.
Massively relayed by a large panel of media world-wide (e.g. Time, Guardian, Telegraph, Independent, New Scientist), the sensational news had the effect of a ‘media bomb’. Although the study by Virtanen et al. (2015) certainly helps to highlight the dangers of alcohol consumption in the main-stream media, authors should not forget that spreading information about effects of drug consumption or medical treatments can be dangerous and it should be done with caution. Unfortunately, the published meta-analytic study is such a case where some serious problems associated with model interpretation exist and we want to highlight them.
The situation that the authors discovered a heterogeneity (discrepancy) among the reported results of the studies is not unusual or indeed surprising, but certain statistical procedures must be performed then to avoid misleading results (Viechtbauer 2007). How big a problem the discovered heterogeneity represents could be expressed by computing an I2 statistics (Huedo-Medina et al. 2006) which is very easy to interpret (it ranges from 0 to 100%). If some larger amount of heterogeneity in the model is confirmed (>75%), researchers have two options. The first, to incorporate additional variables (predictors) into the model which may explain the differences. The second, to conduct stratification of studies (i.e. dividing the dataset of studies into smaller parts based on some factor and analysing each sample individually). If either first or second option does not reduce the heterogeneity in a meta-analytic model, then the meta-analytic approach is not appropriate to answer the given question and the researcher cannot go beyond a systematic review.
Although the authors tried to explain this heterogeneity by some meta-regression and stratification, neither predictors (sex, region, population) nor stratified models (all with non-significant p-values) resolved the problem with high proportion of heterogeneity. Therefore they demonstrated no true evidence about existence of association between long working hours and alcohol use. In the mentioned case, the amount of heterogeneity in all models yielding significant results could be thus interpreted as fatal (p-values are thus not relevant here). It means that studies do not represent random sampling variation around a single estimate, and that there exists some underlying difference between the studies that makes it non-suitable to pool them all together. In spite of the problem described, the authors seem to have ignored it and made an unambiguous conclusion ‘more working = more drinking’. Rather we would conclude, that there exist various factors influencing alcohol consumption but working hours seems not to be one of them.
Ladislav Naďo & Peter Kaňuch
References:
Viechtbauer W. Accounting for heterogeneity via random-effects models and moderator analyses in meta-analysis. J Psychol 2007;215:104–121.
Virtanen M, Jokela M, Nyberg ST, Madsen IEH, Lallukka T, Ahola K et al. Long working hours and alcohol use: systematic review and meta-analysis of published studies and unpublished individual participant data. BMJ 2015;350:g7772.
Huedo-Medina TB, Sánchez-Meca J, Marín-Martínez F, Botella J. Assessing heterogeneity in meta-analysis: Q statistic or I2 index? Psychol Methods 2006;11:193–206.
Competing interests: No competing interests
Do long working hours really increase alcohol use?
High quality systematic reviews are regarded as the highest quality evidence (Harbour and Miller, 2001), which means that we have enough confidence about the validity of results. This systematic review indicated that individuals were likely to increase alcohol use for long working hours (OR=1.12, 95%CI: 1.04-1.20, P=0.96, I2=0%). Although, Virtanen et al tried to use systematic review, the most rigorous method, to answer the PICO question, there was still some bias. For example, they didn’t assesse the risk of bias of the included studies. Therefore, how much could we trust the results? In the GRADE approach, if relevant evidence is supported by the studies with high risk of bias, the quality of evidence could be rated down (Guyatt et al., 2011a). Hence, we used GRADE to rate the quality of evidence of this systematic review to verify the certainty of the results.
Firstly, we used the AHRQ Cross-Sectional/Prevalence Study Quality Checklists (Rostom et al., 2004) and Newcastle-Ottawa Scale (NOS) for Cohort Studies (Wells et al., 2010) to assess the risk of bias of cross-sectional studies and cohort studies respectively. However, when reviewing the 63 studies included (except for 4 studies without full-texts), we found that 18 studies didn’t report the data on working hours and alcohol consumption simultaneously. Moreover, Virtanen et al reported that there were 15 studies of which the main topics were not the association between long working hours and alcohol consumption. Meanwhile, Virtanen et al didn’t report whether they contacted the authors to collect the data of primary studies if it could not be abstracted from full-texts. Furthermore, different definitions about the exposure and outcome were found in included studies. Owing to the potential deviation between the total pooled data and the true value, Virtanen et al conducted a subgroup analyses for the six studies (Liu et al., 2002; Nash et al., 2010; Au et al., 2013; Gibb et al., 2012; Holtermann et al., 2010; Cheng et al., 2012) which strictly defined the exposure and outcome. Similarly, it showed that long working hours could affect the alcohol consumption (OR=1.14, 95%CI: 0.89-1.47, P=0.007, I2=68.6%). Therefore, in order to avoid other bias, we only rated the quality of evidence from these six studies.
The downgrading factors
Risk of bias. All of the six studies were cross-sectional studies. Thus, the AHRQ Cross-Sectional/Prevalence Study Quality Checklists was applied to assess the risk of bias (see appendix). Most studies had high risk of bias in each item except for item 1, 3 and 10. And it should be noted that all the six studies were found with high risk of bias in item 5, 8 and 9, which indicated that the research didn’t control the results and other confounding well in primary studies, and would influence the confidence of results. Thus, we considered that there was serious risk of bias.
Indirectness. We did not downgrade the evidence for indirectness, because the exposure and outcome in the six studies and this systematic review were similar.
Inconsistency. In figure 3 of the systematic review, we found differences in direction, little overlap and large I2 among the six studies. Thus, we considered that there was serious bias in inconsistency.
Imprecision. The 95% confidence intervals include the odds ratio of 1. Thus, we considered that there was serious bias in imprecision.
Publication bias. Virtanen et al conducted a systematic search and the estimates in Egger’s test indicated no evidence for publication bias. Thus, we decided not to downgrade for publication bias.
The upgrading factors
Large magnitude of effect. The effect size was not large (OR=1.14) enough to rate up the evidence.
Dose-response gradient. In figure 7 of this systematic review, the alcohol consumption would increase with prolonging working hours. Thus, we considered rating up the evidence for dose-response gradient.
Plausible confounding can increase confidence in estimated effects. The strict definitions of the exposure and outcome in the six studies was a plausible confounding that might decrease the confidence of estimated effect. However, the effect size of the six studies was larger than that of total effect size. Thus, we considered to rate up the evidence for this factor.
Cross-sectional studies belong to the observational study group. And in GRADE, observational studies (except for diagnostic accuracy tests studies) are regarded as low quality evidence at the beginning (Guyatt et al., 2011a). When observational studies are methodologically rigorous and the evidence body meets the criteria, we will consider rating up the quality of evidence (Guyatt et al., 2011b).
In conclusion, although the quality of evidence might be rated up for dose-response gradient and plausible confounding, there was a high risk of bias in the method of the six studies. Therefore, we could not consider rating up the quality of evidence. Finally, we downgraded the quality of evidence from low to very low.
In summary, although the systematic review indicated that individuals whose working hours exceeded standard recommendations were more likely to increase their alcohol consumption, we had little confidence about this conclusion.
References:
Harbour R, Miller J. A new system for grading recommendations in evidence based guidelines. BMJ. 2001 Aug 11;323(7308):334-6.
Guyatt GH, Oxman AD, Vist G, Kunz R, Brozek J, Alonso-Coello P, et al. GRADE guidelines: 4. Rating the quality of evidenced study limitations (risk of bias). J Clin Epidemiol 2011 Apr;64(4):407-15.
Rostom A, Dubé C, Cranney A, et al. Celiac Disease. Rockville (MD): Agency for Healthcare Research and Quality (US); 2004 Sep. (Evidence Reports/Technology Assessments, No. 104.) Appendix D. Quality Assessment Forms. Available from: http://www.ncbi.nlm.nih.gov/books/NBK35156/
GA Wells, B Shea, D O'Connell, J Peterson, V Welch, M Losos, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2010. Available from: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp
Liu Y, Tanaka H. Overtime work, insufficient sleep, and risk of non-fatal acute myocardial infarction in Japanese men. Occup Environ Med 2002;59:447-51.
Nash LM, Daly MG, Kelly PJ, van Ekert EH, Walter G, Walton M, et al. Factors associated with psychiatric morbidity and hazardous alcohol use in Australian doctors. Med J Aust 2010;193:161-6.
Au N, Hauck K, Hollingsworth B. Employment, work hours and weight gain among middle-aged women. Int J Obes (Lond) 2013;37:718-24.
Gibb SJ, Fergusson DM, Horwood LJ. Working hours and alcohol problems in early adulthood. Addiction 2012;107:81-8.
Holtermann A, Mortensen OS, Burr H, Sogaard K, Gyntelberg F, Suadicani P. Long work hours and physical fitness: 30-year risk of ischaemic heart disease and all-cause mortality among middle-aged Caucasian men. Heart 2010;96:1638-44.
Cheng WJ, Cheng Y, Huang MC, Chen CJ. Alcohol dependence, consumption of alcoholic energy drinks and associated work characteristics in the Taiwan working population. Alcohol Alcohol 2012;47:372-9.
Guyatt GH, Oxman AD, Sultan S, Glasziou P, Akl EA, Alonso-Coello P, et al. GRADE guidelines: 9. Rating up the quality of evidence. J Clin Epidemiol. 2011 Dec;64(12):1311-6.
Competing interests: No competing interests.