Re: "Personality and All-Cause Mortality: Individual-Participant Meta-Analysis of 3,947 Deaths in 76,150 Adults"

We join Dr. Benjamin P. Chapman (1) in acknowledging the importance of Jokela et al.'s recently published paper (2). The study of personality-trait predictors of mortality is consequential, and with this study it is clear that low conscientiousness deserves a seat at the table with major risk factors for early death. Jokela et al. accomplished this by performing an individual participant meta-analysis of data from 7 cohorts (2) to follow up on Kern and Friedman's work (3) revealing the importance of low conscientiousness as a risk factor. We expect that this paper will be oft cited and that this statistical approach will be used in similar future studies. However, we depart from the authors' conclusion that “the present results [of their study] suggest that low conscientiousness is the only higher-order personality trait of the five-factor model that predicts higher mortality risk across populations” (2, p. 673). This generalization is overly broad, premature, and not tenable, given 2 limitations of their study. The first limitation is that while the combined sample size of Jokela et al.'s study was large (n = 76,150), in proportional hazards analysis, statistical power is determined not by the number of participants but by the number of events (4, 5). In light of this, compared with some previous studies (6, 7), the British Household Panel Survey (8), the Household, Income and Labour Dynamics in Australia Survey (9), and the Survey of Midlife Development in the United States (10) all had low power. This problem with power is further illustrated by the fact that, despite the authors' assertion, conscientiousness effects were not significant in the Household, Income and Labour Dynamics in Australia Survey (9) or in the Survey of Midlife Development in the United States (10), just identified as lower in power than other studies, or in the Wisconsin Longitudinal Study sibling sample (11), which had the next-lowest power (see Jokela et al.'s Figure 1 (2, p. 670)). Clearly, the total number of deaths is a key determinant of a study's findings or lack thereof. This limitation is not ameliorated by Jokela et al.'s data analytical approach or by any other approach that we are aware of. As Munafò et al. pointed out, “Meta-analysis, while valuable and offering the potential to generate new hypotheses, is not a replacement for adequately powered primary studies” (12, p. 278). The second limitation is the inadequate measurement of personality, which could have seriously compromised the scales' predictive validities (13). The 7 studies used several different very brief scales to measure the 5 higher-order factors, which is likely to have produced content and construct deficiency of their putative personality factors. Thus, the authors may not have found consistent results for factors other than conscientiousness because these factors were not adequately measured. Unfortunately, the authors present no evidence regarding construct validity, that is, high positive correlations between the different brief scales. Thus, as in other studies (12–15), the inconsistent findings displayed in Jokela et al.'s Figure 1 (2, p. 670) could reflect this limitation of their study. In short, very little confidence can be placed in their conclusion that conscientiousness is the only important personality predictor of mortality. None of this is to say that we do not admire the researchers' efforts or appreciate the importance of their findings concerning conscientiousness. However, well-powered primary studies with valid measures of all 5 factors might identify other important predictors of mortality, and thus we think it is too soon to “close the door” on other personality factors as important predictors of mortality. Dr. Paul T. Costa, Jr., receives royalties for use of the NEO Personality Inventory–Revised.