Cumulative Meta-analysis of Job Strain and CHD

To the Editor: In a cumulative meta-analysis, the results from individual studies are added chronologically and the aggregate effect estimate recomputed with the addition of each study.1 The purpose of this technique is to determine the point at which further studies carried out in a similar manner are unlikely to change the conclusion. Here, we apply this approach to evaluate whether further observational studies are needed to examine the association between job strain (the most widely studied marker of psychosocial stress at work)2 and incident coronary heart disease. We used effect estimates from 26 prospective cohort studies included in the most recent meta-analysis of this association (full references for the constituent studies are available in the eAppendix; https://links.lww.com/EDE/A772).3 Statistical analysis was conducted using Stata-MP (version 11.1) software. With coronary heart disease as the outcome, the summary hazard ratio for job strain compared with no job strain was 1.4 (95% confidence interval = 1.0–1.8) in 2003, when the first six cohort studies had been published (Figure). Although an additional 20 cohort studies on the same topic have been published, the summary estimate remained essentially unchanged at 1.3. The precision of the estimate has improved, with confidence intervals of 1.2–1.5 when all studies are included.FIGURE: Cumulative meta-analysis of cohort studies on job strain and incident coronary heart disease for studies identified in the most recent systematic review.3 Full references for the constituent studies are available in the eAppendix (https://links.lww.com/EDE/A772).Should further observational studies be conducted in the presence of such abundant evidence? One justification has been limitations in exposure measurement in past studies.4 At least 11 sets of questions have been used to measure the components of job strain. Even when the item content is identical, numerous alternative ways of defining job strain have been used.5 The proliferation of instruments, together with a lack of agreement on a single gold standard assessment, has been claimed to underestimate the real effects of job strain by biasing the results toward the null.4 Our cumulative meta-analysis included prospective cohort studies with a variety of measurement approaches.3 Despite this, variation in results among the studies was small: the I2 index summarizing between-study heterogeneity in effect estimates was 21% (P = 0.20). In addition, a recent analysis of 200,000 men and women produced similar effect estimates in the various strata of sex, age, socioeconomic status, and European region (Scandinavia, Continental Europe, and the United Kingdom).6 Single studies using repeat measures of job strain over time have shown larger effects in a smaller group of chronically strained participants, but the conclusions remain unchanged.7 Thus, on the basis that the assessment of job strain should be refined, it seems difficult to justify further observational studies in Europe. Consistency across studies does not necessarily mean we have the final answer, but it does mean that we will not learn anything new from more studies conducted in the same manner. It is possible that improved designs or alternative approaches will yield new information about the association between job strain and coronary heart disease. These could include quasi-experiments where the effects of planned structural changes within organizations are explored (eg, conversion of the traditional factory assembly line mode of production into an open shop floor plan based on a teamwork approach). Cluster-randomized trials could help determine the extent to which interventions might reduce job strain and prevent heart disease above and beyond existing preventive strategies. Current strategies that emphasize smoking cessation, healthy diet, physical activity, antihypertensive treatment, and statin therapy would represent the “standard care” control group in such trials.8 Mika Kivimäki Department of Epidemiology and Public Health University College London London, United Kingdom email protected G. David Batty Department of Epidemiology and Public Health University College London London, United Kingdom Jane E. Ferrie School of Social and Community Medicine University of Bristol Bristol, United Kingdom Ichiro Kawachi Department of Social and Behavioral Sciences Harvard School of Public Health Boston, MA