Sunshine and suicide at the tropic of Capricorn , São Paulo , Brazil , 1996 – 2004

Several studies have confi rmed seasonal variation in suicide rates according to hours of sunshine. The suicide pattern was assessed in São Paulo, southeastern Brazil, at the tropic of Capricorn from 1996 to 2004. Poisson regression was employed to estimate parameters of seasonality, as well as to verify associations for each day between daylight duration and suicide. During the nine-year study period, there were 3,984 suicides (76.9% in men; median age=38.7 years old). Seasonal averages of suicides were similar, as were monthly averages. Poisson regression did not reveal any association between suicide rates and hours of sunshine (p=0.45) for both sexes. In conclusion, no seasonal pattern was observed for suicides.

In his masterpiece, "Suicide," Emile Durkheim noted that "when the days grow longer quickly, suicides increase greatly," i.e., suicide has a peak in the spring or summer and a trough in winter, which has been confi rmed in studies conducted in the Northern Hemisphere. 4,5 Although few studies have addressed this issue in the Southern Hemisphere, their fi ndings were similar. 2 They have shown that higher suicides rates occur during months with more daily sunlight hours. In contrast, in Singapore (at the Equator line), a country with little daily sunlight variation throughout the year, there are no seasonal fl uctuations of suicide rates. 3 These studies suggest that sunshine-regulated hormones, such as melatonin, may have a role in triggering suicide. If true, one would expect to see greater seasonal variation in suicide risk with increasing distance from the Equator.
To test this hypothesis, suicide pattern in a city located at the tropic of Capricorn was investigated.

METHODS
An ecological study was carried out correlating daily data from suicide and solar irradiance and temperature, in the city of São Paulo, Southeastern Brazil, from 1996 to 2004. The city is crossed by the tropic of Capricorn There are two meteorological stations located in the northern and southern area of the city. No differences were found for sunshine and temperature measurements between the two stations. In the northern one, data were collected on bright sunlight hours, defi ned as the period of time during which sunlight intensity was suffi cient to burn a trace on a standard chart when focused with a glass sphere (Campbell-Stokes recorder). In the southern station, data were collected using a pyranometer (Eppley Lab Inc., model PBW) to determine the global solar irradiation (in megajoule/m 2 ).
First, temporal changes of suicide during the study period were tested to avoid that increased or decreased suicide rates could affect seasonal distribution of deaths and, most importantly, the association with sunshine and temperature. Then, the analysis of the series was carried out to detect trends. Since temporal analyses showed signifi cant decrease of suicide rates (p=0.0002), data had to be adjusted using a statistical approach as means of the transformed series to analyze seasonality by months and seasons. Poisson regression variant of the circular normal distribution was applied to estimate parameters of seasonality, as well as to verify associations for each day among meteorological factors and suicide. The statistical package Stata v.7 was used in the analysis.

RESULTS
During the nine-year study period, there were 3,984 suicides, of which 76.9% in men. Median age was 38.7 years old. Seasonal suicide averages had similar distribution according to seasons (fall=107; winter=110; spring=110; summer=112) as were monthly averages (the highest suicide average occurred in August and November and the lowest in September and April). The Figure displays the mean number of suicides as a function of mean direct solar irradiance by month. Poisson regression showed no association between number of suicides by day and direct solar irradiance (p=0.45). The same results were obtained when hours of daylight were considered. In addition, the results were also not significant for average temperature (p=0.34). Data analyses according to gender and age group did not show any differences on the seasonal distribution of suicides and the association between sunshine and suicide.

DISCUSSION
Data studied from the largest city in the Southern Hemisphere showed no seasonal fl uctuation of suicide rates. There was also no association between the number of suicides assessed on a daily basis with sunshine and average temperature measurements. The analyses of suicide behavior were consistent when, instead of day, months and seasons were considered as the unit of observation.
The infl uence of sunshine on suicide risk was proved to be signifi cant in countries with wider fl uctuations of solar irradiation throughout the year as in European Nordic countries. 4,5 A less pronounced association was found in temperate zones, 2 and no association was seen in Singapore, located at the Equator line. 3 The present study data support the fi ndings of Chew & McCleary, 2 who compared suicidal behavior among 28 countries and concluded that seasonal variation in suicide is high in temperate zones (between 30° and 60° north and south of the Equator) and low in the tropics. To the INTRODUCTION authors' knowledge, there is no data about seasonal changes of suicide in Hawaii, a place located at the same latitude as São Paulo but in the Northern Hemisphere at the tropic of Cancer.
There is a hypothetical role of melatonin metabolism explaining how sunshine or correlated factors are related to suicide. One putative factor emphasized by various authors, including Petridou et al 5 would be related to one of several mood regulators, the melatonin variation according to season and differences of solar irradiation. Melatonin secretion has been shown to be abnormal in some severely depressed patients but the majority of these studies were conducted in the Northern Hemisphere. Recently, in São Paulo, a comparison of drug-free depressive patients with healthy volunteers matched for age and gender did not reveal any difference in major urinary metabolite of melatonin, and further matching for body mass index and season did not change these results. 1 The present study revealed (data not shown) that age-adjusted suicide rates are declining for both sexes. Although, they were adjusted using means of the transformed series to analyze seasonality according to months and seasons it is possible that this decline could affect the seasonality of suicide.
The strengths of the present study included the quality of solar irradiation measurements in the two stations using different methods, and the good quality ascertainment of death in São Paulo by autopsy and further epidemiological surveillance. The study's limitation was the lack of statistical power to identify differences according to gender and/or age groups.
In conclusion, in a large urban area at the tropic of Capricorn, no seasonal fl uctuations of suicide rates as well as no association between solar irradiance and suicide rates were seen.