Sunshine and Suicide Incidence

To the Editor: In an ecologic study across 20 countries, Petridou et al. 1 found a positive relation between relative risk of suicide during the peak month of suicide incidence and same-month average sunshine duration (+0.7; Spearman correlation). They concluded that sunshine exposure, via sunshine-regulated hormones like melatonin, may have a role in the triggering of suicide. Although there is no harm in this sort of speculation, acceptance of this effect should clearly await more direct evidence. However, we suspect that the finding itself rests on misleading methods and data, and we marshal evidence for this contention as follows. First, the authors did not show that relative risk measures for suicide peak months are more closely related to seasonal variation in sunshine than to other environmental variables, did not mention findings opposite to their own (reviewed elsewhere), 2,3 and did not address the fact that suicide peak months generally are not the ones with the most intense sun exposure. Second, mere size of correlational findings is not evidence for actual relations. Using the relative risk estimates from Petridou et al. 1 (Table 1), we obtained sizable cross-national correlations with other variables 4 as well, with physician density (+0.62), tuberculosis rate (+0.81), and computer ownership density (−0.57). Does this mean there is a role for doctors or tuberculosis cases in increasing countries’ amplitude of suicide seasonality, whereas computer ownership reduces the amplitude? Third, the authors make no mention of the perhaps most startling finding in suicide seasonality research—over the past few decades, suicide seasonality has notably diminished almost everywhere. The main agent of this secular trend remains unresolved. 3,5,6 If indeed seasonal variation in sunshine triggers within-country suicide peaks, and differences in sunshine account for cross-national differences in suicide seasonality, we look forward to hearing that seasons within countries, as well as climate differences across countries, have recently decreased. Fourth, we doubt the accuracy of countries’ suicide peak months as determined by Petridou et al.1 Other peak months have been identified for Australia, 7 Finland, 8 Ireland, 9 Japan, 10 New Zealand, 7 Sweden, 11 and Austria (1970–1999 data: May, not June). Some of these findings stem from time series considerably longer (Sweden: 1911–1993) 11 than that of Petridou et al.; others indicate either gender differences in peak months 7,8 (including Austrian data) or biseasonality in suicide incidence. 7,8,10 The Petridou et al. 1 time-series data vary greatly in length (4 to 24 years), which obviously led to misidentification of suicide peak months, because there is evidence for them shifting from spring to summer with increasing latitude 8 (ie, a positive relation). Conversely, in the Petridou et al. 1 data, this relation is negative (−0.28; correlation between peak month number, recoded for southern hemisphere, and capitals’ latitude). Fifth, we question the accuracy of the Petridou et al. 1 relative risk estimates for countries’ peak suicide months. Monthly variation in suicide is still strong in the United States, 12 although, in the Petriodou et al. table, the smallest estimate is for the United States. The relative risk estimate is exceptionally large in Japan, 10 although not presented as such in the table; rather, in the table, the relative risk estimate for Japan is identical to that for Australia, where seasonality is weak. 7 Again, high cross-country variation in time-series’ length, in concert with the statistical method used, obviously led to erroneous estimation of seasonality effects. The circular normal distribution method used by Petridou et al.1 tests for one-cycle seasonality only, thus missing seasonality increments attributable to within-year cycles, and it is sensitive to outliers in the data that gain influence in short time series. 3 The accuracy of suicide seasonality estimates can be tested using their positive relation to latitude, as has been found both within the United States 13 and internationally 3,5 (ie, seasonality increases with increasing equatorial distance). Conversely, in the Petridou et al. 1 data, the correlation is negative (−0.35). It is more parsimonious to assume misidentification of peak months and mistaken estimation of seasonality effects in the Petridou et al. 1 study, attributable to factors unique to their database and the statistical method used, than to suggest that a great many established findings on suicide seasonality are incorrect. Their finding of a relation between sunshine duration and suicide incidence rests heavily on correctly identified suicide peak months and correctly estimated suicide seasonality effects. Because the data are demonstrably odd, the authors’ conjecture might be unwarranted. On a final note, we express our irritation regarding the claim by Petridou and colleagues 1 for scientific priority regarding the cross-national documentation of seasonal suicide peaks. Actually, the Chew and McCleary 5 study deserves such priority—it was not merely about “several” countries, but was, rather, a large-scale (28-country) investigation covering 16 of the 20 countries sampled by Petridou et al.1 Martin Voracek Maryanne L. Fisher

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.