Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
It has been reported that the incidence of suicide reaches a peak during the early summer. 1,2 This finding has been very difficult to explain and, in fact, seems to be counterintuitive, given the impression of most people that their mood deteriorates during fall and winter. It has been suggested that sunshine exposure may affect suicide risk through regulation of serotonin 3,4 or melatonin 5 levels, but supportive empirical evidence is lacking. We have evaluated the hypothesis that sunshine may trigger suicide by collecting data on the monthly distribution of suicide deaths in countries with reliable statistics and by estimating sunshine exposure during the corresponding months.
In order to insure reliable data, we have focused on countries-members belonging to the Organisation for Economic Co-operation and Development (OECD). A thorough Internet search was performed for each of the 29 OECD countries. No Internet site was found for Portugal and Poland. For the 27 remaining countries, e-mails were sent and on-line forms were submitted, some times more than once, to officials or webmasters responsible for the data. For four countries (Belgium, Ireland, Luxembourg, and the United Kingdom), an answer was received that the requested data were not available, whereas no answer was received from three countries (Italy, Korea, and Turkey). Thus, distribution of suicide deaths by month was sent from 20 OECD countries for a period covering 4 to 24 of the most recent years.
For the calculation of sunshine during the months with the highest density of suicide deaths, we have retrieved data from two Internet sites. Daylight duration data for every country were retrieved from the Internet web site http://www.sunrisesunset.com. Total sunshine per month was calculated by multiplying total daylight duration during the particular month by the average fraction of time during the month that the sun was not obstructed by clouds. This fraction was derived using “WINDISP4,” a Map and Image Display and Analysis System which was downloaded from the Internet web site of the Food and Agriculture Organisation of the United Nations. All months were evaluated on an equal duration basis.
Parameters of seasonality within countries were estimated using the Poisson regression variant of the circular normal distribution 6 (appendix) which is an analog of the classic Edward’s procedure. This procedure efficiently estimates the month of peak suicide incidence, the relative risk of committing suicide during the month of peak incidence (compared with the minimum incidence during the whole year), and confidence intervals for those measures. The advantage of this analysis within countries is that it eliminates confounding factors that differ across countries but remain reasonably constant within a country over the study time. For correlation analyses across countries, Spearman’ s correlation coefficient was used.
We studied three different issues that are relevant to our hypothesis of the triggering role of sunshine on suicide.
Of the 20 OECD countries in the study, 18 are in the northern hemisphere (Austria, Canada, the Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Japan, Mexico, Netherlands, Norway, Spain, Sweden, Switzerland, and the United States of America) and 2 are in the southern hemisphere (Australia and New Zealand). For all 20 countries, there was evidence of seasonality of suicide rates, with relative risk estimates varying from 8% to 50%, and with peak incidence in May or June in the countries of the northern hemisphere and November or December in the two countries of the southern hemisphere (Table 1).
Figure 1 shows the positive association between the relative risk of suicide during the month of peak incidence in the 20 examined OECD countries and the average sunshine in hours during this month in the corresponding country. Spearman’s correlation coefficient for this association is +0.7. Correlation taking into account as weights the corresponding population sizes generates a Spearman’s correlation coefficient of +0.6.
Exclusion of Summer Months
It is possible that the excess suicide risk during the summer months could be associated with behavioural changes of the persons not attributed directly to sunshine. The most important such factor is that people tend to take their vacation from work in the summer months, so that suicide risk could be affected by factors associated with more free time rather than more sunshine. To address this possibility, we also analyzed the data within each country by removing the three summer months (June-August for northern hemisphere, December-February for southern hemisphere). Seasonality was assessed with our procedure for the remaining 9 months. All 18 countries in the northern hemisphere showed a peak suicide rate in May or April, whereas Australia and New Zealand showed peak rates in October and November respectively. Exclusion of four consecutive spring-summer months had no effect either.
This is the first time that the early summer peak incidence of suicide has been documented across many countries, through a demonstrably unbiased process (inclusion of all OECD countries), although many earlier studies have indicated the existence of this pattern in individual populations and at least one important investigation included several countries. 7 Further, our data indicate that the degree of seasonality, that is, the amplitude of the seasonal variation in Figure 1, can be largely explained in terms of monthly sunshine, both within and across countries. Last, our analyses indicate that the relation between sunshine and suicide risk is not a by-product of the covariable of vacation time.
There are concerns when an ecological approach has to be used for the investigation of a phenomenon, and these concerns have been reviewed by Morgenstern. 8 The applicability of climatological/meteorological data becomes more questionable as the geographical size of a country increases, and there is no way to examine whether sunshine or a climatological correlate of this variable is the one more relevant. Nevertheless, the consistency of an early summer excess incidence of suicide around the world, and the further association of suicide with hours of sunshine, strongly suggest that a physical environmental factor plays an important role in the triggering of suicide.
The evidence presented concerns triggering of suicide and just one dominant pattern in the distribution of suicides over time. Other time patterns, eg, within week 9 place patterns (latitude, 10 urban/rural 11,12) and personal characteristic patterns (age, 13 gender, 13,14 social class, 15 family status, 13,16,17 etc.) may have different explanatory factors, as it is common in diseases with multifactorial etiology. It should be pointed out that the overall rate of suicide in the northern hemisphere increases with latitude and this is likely due to social factors or to more complete reporting of suicide in these countries. The indicated factors, however, are unlikely to have a seasonal pattern and thus, are unlikely to confound the seasonal pattern of suicides across the world.
It is not obvious how sunshine or correlated aspects of it, including short-term variability, could affect the decision to commit suicide. Melatonin is affected by sunshine 18–22 and plays a role in mood regulation, 6,23–24 but no attempt has been made to investigate a possible sequence of events. Other hormones, such as cortisol 25 and serotonin, 26 as well as L-tryptophan, 27 are also known to show sunshine dependence, and singling out melatonin at this time is probably premature. There should be no doubt however, that the seasonality of suicide is a striking epidemiological characteristic of this phenomenon that could have substantial preventive implications, eg, by administering melatonin, the secretion of which is suppressed by sunshine. Further progress could be made by switching from ecological to individual based investigations 28 or by focusing on daily variations of sunshine and number of suicides, with allowance for possible lag time between exposure and outcome.
The authors would like to thank all employees and organizations that provided the data used in this study, and in particular Ms. Jeanne MacDonald (Canada), Ms. Shari Mason (New Zealand), Ms. Mary Heanue (Ireland), Ms. Kari Kristiansen (Norway), Mr. Manuel Berzal Burgos (Spain), Ms. Johanna Janecek (Austria), Mr. Erwin Wüest (Switzerland), Mr. Thomas Norup (Denmark), Mr. Mauno Huohvanainen (Finland), Mr. Stan Bordeaux (Australia), Ms. Eva Lundquist (Sweden), Ms. Kristin Krah (Germany), Ms. Francoise Pequignot (France), Mr. Ed Koderitsch (Netherlands), Ms. Ivana Valterova (Czech Republic), Mr. Peter Jozan (Hungary), George Kotsyfakis (Greece), National Center for Health Statistics (USA), Mr. Jorge Domínguez Romero (Mexico), and Mr. Hirako (Japan).
Because the number of days in each month varies slightly, we used the circular distribution procedure, which is analogous to but more flexible than Edwards’ procedure. The frequency of the circular normal distribution at a certain date is proportional to the periodic factor ek cos 2π (date –maxdate)/totaldate, where totaldate is the total length of time under consideration, maxdate is the date of maximum frequency, ek is the ratio of maximum over expected frequency, and e2k is the ratio of maximum over minimum frequency. 6 When k = 0 there is no seasonality. For the 12 months in our study, we estimated the risk and date of peak by taking the number of suicides for each month as a Poisson count with frequency proportional to the above factor, adjusted in the exponent for the different number of days in each month.
1. Bazas T, Jemos J, Stefanis K, Trichopoulos D. Incidence and seasonal variation of suicide
mortality in Greece. Compr Psychiatry 1979; 20: 15–20.
2. Kevan SM. Perspectives on season of suicide
: a review. Soc Sci Med 1980; 14: 369–378.
3. Turecki G, Briere R, Dewar K, Antonetti T, Lesage AD, Seguin M, Chawky N, Vanier C, Alda M, Joober R, Benkelfat C, Rouleau GA. Prediction of level of serotonin
2A receptor binding by serotonin
receptor 2A genetic variation in post-mortem brain samples from subjects who did or did not commit suicide
. Am J Psychiatry 1999; 156: 1456–1458.
4. Stockmeier CA. Neurobiology of serotonin
in depression and suicide
. Ann N Y Acad Sci 1997; 836: 220–232.
5. Stanley M, Brown GM. Melatonin
levels are reduced in the pineal glands of suicide
victims. Psychopharmacol Bull 1988; 24: 484–488.
6. Gumbel EJ, Greenwood JA, Durand D. The circular normal distribution: theory and tables. J Am Stat Assoc 1953; 48: 131–152.
7. Chew KS, McCleary R. The spring peak in suicides: a cross-national analysis. Soc Sci Med 1995; 40: 223–230.
8. Morgenstern H. Uses of ecologic analysis in epidemiologic research. Am J Public Health 1982; 72: 1336–1344.
9. MacMahon K. Short-term temporal cycles in the frequency of suicide
. United States 1972–1978. Am J Epidemiol 1983; 117: 744–750.
10. Lester D, Shephard R. Variation of suicide
and homicide rates by longitude and latitude
. Percept Mot Skills 1998; 87: 186.
11. Yip PS, Callanan C, Yuen HP. Urban/rural and gender differentials in suicide
rates: East and West. J Affect Disord 2000; 57: 99–106.
12. Saunderson T, Haynes R, Langford JH. Urban-rural variations in suicides and undetermined deaths in England and Wales. J Public Health Med 1998; 20: 261–267.
13. Yip PS. Age, sex, marital status and suicide
: an empirical study of east and west. Psychol Rep 1982; 82: 311–322.
14. Zhao S, Qu G, Peng Z, Peng T. The sex ratio of suicide
rates in China. Crisis 1994; 15: 44–48.
15. Kaplan MS, Geling O. Sociodemographic and geographic patterns of firearm suicide
in the United States, 1989–1993. Health Place 1999; 5: 179–185.
16. Lester D, Yang B. The relationship between divorce, unemployment and female participation in the labour force and suicide
rates in Australia and America. Aust N Z J Psychiatry 1991; 25: 519–523.
17. Lester D. Effect of divorce on the suicide
rates of single and married men. Psychol Rep 1989; 65: 658.
18. Hatonen T, Alila-Johansson A, Mustanoja S, Laakso ML. Suppression of melatonin
by 2000-lux light in humans with closed eyelids. Biol Psychiatry 1999; 46: 827–831.
19. Kennaway DJ, Rowe SA. Effect of stimulation of endogenous melatonin
secretion during constant light exposure on 6-sulphatoxymelatonin rhythmicity in rats. J Pineal Res 2000; 28: 16–25.
20. Bunnell DE, Treiber SP, Phillips NH, Berger RJ. Effects of evening bright light exposure on melatonin
, body temperature and sleep. J Sleep Res 1992; 1: 17–23.
21. Reiter RJ. Action spectra, dose-response relationships, and temporal aspects of light’s effects on the pineal gland. Ann N Y Acad Sci 1985; 453: 215–230.
22. Olcese J. Melatonin
after Four Decades: An Assessment of Its Potential. New York: Kluwer Academic/Plenum, 1999.
23. Dollins AB, Zhdanova IV, Wurtman RJ, Lynch HJ, Deng MH. Effect of inducing nocturnal serum melatonin
concentrations in daytime on sleep, mood, body temperature, and performance. Proc Natl Acad Sci USA 1994; 91: 1824–1828.
24. Lam RW, Berkowitz AL, Berga SL, Clark CM, Kripke DF, Gillin JC. Melatonin
suppression in bipolar and unipolar mood disorders. Psychiatry Res 1990; 33: 129–134.
25. Posener JA, Schildkraut JJ, Samson JA, Schatzberg AF. Diurnal variation of plasma cortisol and homovanillic acid in healthy subjects. Psychoneuroendocrinology 1996; 21: 33–38.
26. Caga Mpang FR, Inouye ST. Diurnal and circadian changes of serotonin
in the suprachiasmatic nuclei: regulation by light and an endogenous pacemaker. Brain Res 1994; 639: 175–179.
27. Maes M, Scharpe S, Vekerk R, D’Hondt P, Peeters D, Cosyns P, Thompson P, De Meyer F, Wauters A, Neels H. Seasonal variation in plasma L-tryptophan availability in healthy volunteers. Relationships to violent suicide
occurrence. Arch Gen Psychiatry 1995; 52: 937–946.
28. Van Wijngaarden E, Savitz DA. Occupational sunlight exposure in relation to suicide
among electric utility workers. Am J Ind Med 2000; 38: 149–154.