- Email: [email protected]
Cyclic time patterns of death from suicide in northern Finland
Journal of Affective Disorders 78 (2004) 11 – 19 www.elsevier.com/locate/jad
Research report
Cyclic time patterns of death from suicide in northern Finland Timo Partonen a,*, Jari Haukka a, Kaisa Viilo b, Helina¨ Hakko b, Sami Pirkola a, Erkki Isometsa¨ a, Jouko Lo¨nnqvist a, Terttu Sa¨rkioja c, Erkki Va¨isa¨nen b, Pirkko Ra¨sa¨nen b a
Department of Mental Health and Alcohol Research, National Public Health Institute, Mannerheimintie 166, FIN-00300 Helsinki, Finland b Department of Psychiatry, University of Oulu, Oulu, Finland c County Administrative Board, Oulu, Finland Received 22 March 2002; received in revised form 28 June 2002; accepted 5 July 2002
Abstract Background: Time patterns of suicide have been attributed not only to social and psychological factors but also to direct geophysical effects. Seasonal variations in day length and temperature seem likely to contribute to the timing of the suicide process. Methods: We analysed all suicides (n = 1658) committed in a northern province of Finland during a period of 153 months. Daily data on the number of suicides, local weather conditions and geomagnetic storms were compiled and modelled with Poisson regression using the province population as the denominator, and with the means of harmonic series for seasonal variation. Time series analysis of monthly numbers of suicides was carried out using the seasonal-trend decomposition procedure based on loess. Results: Marked fluctuations in the number of suicides occurred during the study period ( P = 0.01). There was significant seasonal variation in death from suicide ( P = 0.01), but analysis of the meteorological data showed no evidence of effect on the risk of suicide. Limitations: Assessment of mental disorder or alcohol consumption was missing, since only data derived from death certificate was available for each case. Conclusions: The seasonal effect was significant, but remained modest compared to sex and age as risk factors for suicide. Preventive measures need to be tailored according to time of the year. D 2002 Elsevier B.V. All rights reserved. Keywords: Season; Suicide; Time
1. Introduction Suicides are not evenly distributed over particular periods of time, since social or psychological factors have immediate or delayed effect on the individual’s decision to commit suicide (Mann, 1998). Most stud-
* Corresponding author. Tel.: +358-9-4744-8660; fax: +358-94744-8563. E-mail address: [email protected] (T. Partonen). 0165-0327/$ - see front matter D 2002 Elsevier B.V. All rights reserved. doi:10.1016/S0165-0327(02)00236-7
ies have found that suicide rates tend to peak during spring and summer (Yip et al., 1998; Altamura et al., 1999). Moreover, it appears that climate and weather conditions can somewhat influence the occurrence of death from suicide, possibly by interacting with social factors (Preti, 1997; Salib and Gray, 1997; Preti and Miotto, 1998). Physical time-givers, as well as social and endogenous ones, can modify complex behaviours. Seasonal transition may contribute to the pathogenesis of the suicide process due to the substantial changes in
12
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
day length and temperature that occur in spring and autumn. The asynchrony between changes in temperature and light – dark cycles challenges the stability of endogenous clocks (Czeisler et al., 1999). It may acutely result in disordered periods of activity (Taillard et al., 2001), or disorganised behaviour, possibly leading to suicide in adverse circumstances. 1.1. Objectives Since geophysical factors may markedly influence daily behaviour patterns, it is important in terms of prevention to study whether they can also affect the occurrence of death from suicide. We therefore set out to study the time series of death from suicide in the population of a far northern region where there is extensive variation in levels of ambient temperature and daylight over the year. We hypothesised that increasing light exposure on still cold days during spring in specific may drive the individual into a later activity phase and the failure to adjust to time-giving signals from the natural habitat during this period of seasonal mismatch may lead to death from suicide more frequently than by chance alone.
2. Methods The study material consisted of all suicides (n = 1658) committed from 1 April 1987 to 31 December 1999 in the province of Oulu, which is located 64– 66jN in northern Finland, a northern European country with approximately 5 million inhabitants. Date, sex and age were recorded for each case. Population data on the number of inhabitants, sex ratio and age groups in the province during the study period were derived from statistical yearbooks published by Statistics Finland (1988 – 2000). Cause of death information for this study was obtained from the official death certificates that were collected as part of the National Suicide Prevention Project in Finland from 1 April 1987 to March 1988, and provided by the county administrative board thereafter. Since 1987, death certificate forms in Finland have complied with the recommendations of the World Health Organization’s International Classification of Diseases. The detailed content of the Finnish death certificate provides excellent accuracy and reli-
ability of information on the cause of death. Finnish law requires a medicolegal investigation of the cause and the manner of death whenever it has been unnatural, or is suspected to be so, or whenever it has been sudden or unexpected. The proportion of medicolegal investigations for suicide cases has been stable for many decades (Na¨yha¨, 1981), and the overall autopsy rate has remained relatively high, both of which allow for reliable conclusions on suicide trends over time. 2.1. Meteorological data The following daily data on local weather conditions in the Oulu borough of Kajaani were obtained from the meteorological observatory of the Finnish Meteorological Institute: mean temperature, mean cloud coverage (by eighths: 1 = clear, 4 = half-cloudy, and 8 = cloudy), mean air pressure, and wind direction and velocity. The day-to-day fluctuation in mean temperature was used for the analysis. The direction of wind was coded into four indicator variables for north, east, south and west. As for wind strength, a value of 1 was assigned if the wind velocity from a direction was over 10 m/s as measured at midday, indicating strong wind; otherwise the value was zero. Magnetic field variations routinely arise from current systems that are propagated by regular changes in solar radiation. However, interactions of the solar wind with the magnetosphere, and between the magnetosphere and ionosphere, also induce irregular current systems and magnetic field changes. Magnetic activity indices have been designed to describe variation in the geomagnetic field caused by these irregular current systems. The effect of geomagnetic storms was analysed using the geomagnetic activity indices measured in the borough of Nurmija¨rvi, Finland, at the geophysical observatory of the Finnish Meteorological Institute (for entry, see http://sumppu. fmi.fi/MAGN/K-index/NURMIJARVI/). 2.2. Astronomic data The year was divided into the astronomic seasons: spring (spring equinox to summer solstice), summer (summer solstice to autumn equinox), autumn (autumn equinox to winter solstice), and winter (winter solstice to spring equinox). The lunar cycles and
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
calendar weeks were also recorded in order to assess the effect of cyclic time patterns independent of day length and temperature. Both are ancient systems of reckoning time, the former a natural one and the latter purely man-made. The Almanac Office of the Department of Astronomy, University of Helsinki, provided these data. 2.3. Seasonal mismatch The year was also divided into the thermal seasons: spring (daily mean temperature above 0 jC), summer (above 10 jC), autumn (below 10 jC), and winter (below 0 jC). Because of the mismatch between the two definitions of seasons (see Fig. 1), we assessed the periods of conflicting dates, or the seasonal mismatches, as follows: spring gap (spring equinox to thermal spring), summer gap (thermal summer to summer solstice), autumn gap (thermal autumn to autumn equinox), and winter gap (thermal winter to winter solstice). The mismatch data were calculated on a day-by-day basis over the study period. The primary event in the spring gap is a change in the length of day, whereas the other three seasonal mismatches begin with a change in daily mean temperature. The spring gap may thus impart a unique time-giving stimulus to endogenous clocks. Specifically, we hypothesised that increasing light exposure on still cold days during spring may drive the individual into a later activity phase (Majercak et al., 1999) that would compromise well-being. This may end with subsequent delays (Beersma et al., 1999) on top of the delayed phase position and finally stop the clock running (Jewett et al., 1991). Exposure to colder
13
days during autumn may in turn reverse the delayed activity phase and induce accelerating phase advances that would result in attempts to slow down the clockwork. 2.4. Statistics Time series analysis of monthly numbers of suicides was carried out using locally weighted regression, the seasonal-trend decomposition procedure based on loess (STL; Cleveland and Devlin, 1988; Cleveland et al., 1990). The STL decomposes time series in three components (trend, seasonal, and the remainder) using a sequence of smoothing operations. This method is robust in detecting both trends and seasonal variation. Separate analyses were applied for men and women. Daily data containing the number of suicides were compiled. Each day was represented by six observations because the data included all combinations of sex and age categories (under 30, 30– 60 and over 60 years). The daily number of suicides (range 0 –3) was modelled with Poisson regression using population in the province of Oulu as the denominator. Potential explanatory variables were as follows: the year as a categorical variable (1987 –1999); sex; age; seasonal components with a total of four cosine and sine oscillations in a year; the lunar phase in four categories (equal intervals of 7.4 days around the phases, new moons as the reference); weekdays in two categories (weekends, and the remainder as the reference); changes in daily mean temperature (jC) as a categorical variable using first and ninth deciles ( 3.7 and + 3.7) as the cut-off points; daily mean air pressure
Fig. 1. Schematic representation of the concept of seasonal mismatch. Abbreviations: T, thermal seasons, the lengths of which vary depending on local weather conditions each year; A, astronomic seasons. Numbers from 1 to 4 represent seasonal gaps from spring towards winter.
14
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
Table 1 Number of suicides (n) and incidence per 1000 person-years
3. Results
Variable
Of the 1413 male and 313 female suicides there were complete data available for analysis on 1658 (96%) subjects (Table 1). Male sex (v2 = 735.3, df = 1, P < 0.0000001) and two older age groups (30 – 60, plus over 60 years, v2 = 344.7, df = 2, P < 0.0000001) were associated with the risk of suicide. Our analyses showed that marked changes in the number of suicides occurred during the study period (v2 = 28.0, df = 12, P = 0.01). The risk of suicide decreased significantly over the three consecutive years from 1994 to 1996 (Table 2), and the weekend was linked to increased risk of death from suicide ( + 13% on average). The time patterns of suicide occurrence are
Year: 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Sex: Male Female Age: < 30 years 30 – 60 years >60 years
n
Personyears
Incidence
109 150 133 135 148 146 127 98 113 112 118 136 133
327.7 435.5 436.1 439.9 441.4 446.9 446.8 449.7 451.8 454.1 452.9 452.8 454.8
0.33 0.34 0.31 0.31 0.34 0.33 0.28 0.22 0.25 0.25 0.26 0.30 0.29
1359 299
2843.3 2847.2
0.48 0.11
Table 2 Risk ratios (RR) and 95% confidence intervals (CI) from the Poisson regression model
389 1017 252
2469.5 2279.5 941.5
0.16 0.45 0.27
Variable
RR
95% CI
Year: 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
1.00 1.06 0.93 0.93 1.02 0.99 0.86 0.65 0.75 0.74 0.78 0.89 0.87
0.83 – 1.35 0.72 – 1.20 0.73 – 1.20 0.80 – 1.31 0.77 – 1.27 0.66 – 1.11 0.50 – 0.86 0.58 – 0.97 0.57 – 0.96 0.60 – 1.01 0.69 – 1.15 0.67 – 1.12
Lunar cycle: New moon First quarter Full moon Third quarter
447 393 436 382
1404.7 1430.1 1429.0 1426.7
0.32 0.27 0.31 0.27
Week: Weekdays Weekend
896 762
3251.4 2439.1
0.28 0.31
(kPa) as a categorical variable using first and ninth deciles (994.8 and 1026.0) as the cut-off points; four daily indicator variables of wind; major geomagnetic storms using the value of 50 aK as the cut-off point; and mismatch of the astronomic and thermal seasons, reflecting the conflicting signals of light and temperature from the habitat. The seasonal components were modelled with the means of harmonic series (Jones et al., 1988), and year and the harmonics of seasonal components were included in all models (McCullagh and Nelder, 1994). Other potential explanatory variables were added sequentially (as above) into a model one by one, the significance was tested using the likelihood ratio test, and a given variable was left in the model if significant.
Sex: Male Female Age: < 30 years 30 – 60 years >60 years
1.00 0.22
0.19 – 0.25
1.00 2.84 1.96
2.53 – 3.19 1.67 – 2.30
Lunar cycle: New moon First quarter Full moon Third quarter
1.00 0.86 0.96 0.84
0.75 – 0.99 0.84 – 1.09 0.73 – 0.96
Weeks: Weekdays Weekend
1.00 1.13
1.03 – 1.25
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19 Fig. 2. A seasonal-trend decomposition of suicide occurrence among men over the study period. Panels from the top: the original data, the smoothed data for trend using a time window size of 47 months, the seasonal component, and the remainder from the model. The scale bars to the right represent the corresponding unit in each figure.
15
16 T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19 Fig. 3. Seasonal-trend decomposition of suicide occurrence among women over the study period. Panels from the top: the original data, the smoothed data for trend using a time window size of 47 months, the seasonal component, and the remainder from the model. The scale bars to the right represent the corresponding unit in each figure.
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
Fig. 4. Annual variation in the relative risk of suicide in the study population. The line represents the fitted curve of the monthly data for risk ratios using January as the reference, and the bars represent the 95% confidence limits.
presented for men and women in Figs. 2 and 3, respectively. We also found that the seasonal variation in suicide occurrence was of significance among the study population (v2 = 14.6, df = 4, P = 0.01), with peaks during long days over the year (Fig. 4). Men and women did not differ in the extent of this variation. Analysis of the meteorological data provided no evidence of an effect on the risk of suicide (data not shown). Neither did the seasonal mismatch contribute to death from suicide during spring (v2 = 3.1, df = 1, P = 0.08) nor autumn (v2 = 1.3, df = 1, P = 0.2), thus falsifying our hypothesis. In addition, the risk of suicide was associated with the lunar cycle (v2 = 8.4, df = 3, P = 0.04), and the calendar week (v2 = 6.5, df = 1, P = 0.01). The risk of suicide was highest around the phase of a new moon and lower for the next phase ( 14% on average), around the phase of a full moon ( 4% on average), and again during the final phase ( 16% on average) of the lunar cycle. There was no interaction between the seasonal and lunar effects concerning the risk of suicide.
4. Discussion Our main finding was that contrary to recently published findings (Yip et al., 2000; Parker et al., 2001) there was a significant seasonal variation in death from suicide. The highest incidence was ob-
17
served at the turn of summer to autumn, and a second peak occurred during spring. Our findings also suggest that despite a marked difference in time patterns for suicide between men and women in the long term, there is none in the seasonal variation. The seasonal effect was modest though, as compared with sex and age as the known risk factors for suicide. We also showed, using versatile and minute data, that local weather conditions, geomagnetic storms, or the periods of seasonal mismatch had no effect on the risk of suicide among a population living in northerly regions. The principal strength of our study was that we had an unselected suicide population with a relatively high number of suicide cases in a remote northern region which experiences a relatively extreme change of the seasons. Both the lunar cycle and the calendar week showed significant associations, but these were secondary to the seasonal effect on the risk of suicide. Suicide occurrence has not previously been linked to phases of the lunar cycle (MacMahon, 1983; Maldonado and Kraus, 1991; Gutierrez-Garcia and Tusell, 1997), except in one study which found fewer suicides in the phase of a full moon than in the first quarter of the lunar cycle (Yvonneau, 1996). Our results showed that the phase of a new moon was associated with increased risk of suicide. Although alcohol consumption appears to increase on average by 26% at the new moon relative to the full moon (de Castro and Pearcey, 1995), there is no immediately apparent explanation for our finding. However, it is of note here that the association, although a significant one, was weak. One reason for the existence of seasonal and circadian variations in suicide may be that the function of intrinsic time-keeping mechanisms (Ashkenazi et al., 1993) is compromised prior to death. The body clocks are driven by signals such as changes in luminance and ambient temperature from the individual’s natural habitat to match the solar day, in such a way that the internal phase is a reliable predictor of solar time. Rhythmic variations in bodily functions and overt behaviour allow individuals to anticipate environmental changes, and to seek protection. The accuracy of this minute synchronisation depends on the individual response characteristics of body clocks (Beersma et al., 1999) on which the knowledge of time, physical activity, and state of health have a vital influence. Depressive disorders (Soueˆtre et al., 1991)
18
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
and alcohol misuse (Wasielewski and Holloway, 2001) easily affect these characteristics and may thereby routinely compromise the intrinsic clockwork, especially when there is a shortage of the stimuli that give time from the habitat. However, the present findings of the effect of seasonal mismatch, as calculated on the basis of actual meteorological data and tested against our hypothesis, gave no evidence of disordered intrinsic clockwork in suicide. 4.1. Limitations One weakness of our trial was the absence of data on mental health for each case. Mental disorders are frequent among individuals completing suicide (Henriksson et al., 1993), and carry high risk for death from suicide (Harris and Barraclough, 1997). From the epidemiological perspective, the conditions most relevant for prevention are affective disorders and alcohol dependence (Mortensen et al., 2000). Seasonal changes in depressive disorder, its treatment indices, and suicide seem to coincide, underlying the importance of depressive episodes as a cause of death from suicide (Eastwood and Peacocke, 1976; Maes et al., 1993). Our finding of seasonal variation in suicide occurrence links to a previous report from Finland that feelings of depression were most prevalent in northerly regions during the spring (Na¨yha¨ et al., 1994). Recently, another study attributed the declining suicide rate to the improved care of those with depressive disorder (Rihmer et al., 1998). We did not have individual data on alcohol consumption either. The suicide rate of Finnish men aged under 50 years appears to be associated with per capita alcohol consumption (Ma¨kela¨, 1996). Typical habits of alcohol consumption also seem to be linked to weekly variations in the number of suicides (Pirkola et al., 1997). Our data agreed with this, since the risk of suicide was increased at weekends. In addition, there was a significant decrease in suicide occurrence from 1994 to 1996. Interestingly, there was also a marked decrease in the consumption of alcohol beverages, from 8.7 to 6.0 l per capita calculated as 100% alcohol, from 1990 to 1996 in the province of Oulu (OY Alko AB, 1988 – 1995; STAKES, 1996 –2000). This fall in alcohol consumption apparently coincided with reduced suicide mortality in the province.
Another study weakness was that the meteorological data were derived from only one point in the study area. However, this type of measurement error usually tends to attenuate results, which indicates a stronger real effect compared to the observed one.
5. Conclusion Marked changes in the number of deaths from suicide were observed during the study period. There was a substantial seasonal variation in suicide occurrence among the study population. Compared to sex and age, however, the seasonal effect on the risk of suicide remained modest. Periods of seasonal mismatch, local weather conditions, and geomagnetic storms had no effect on the risk.
Acknowledgements We thank Heikki Nevanlinna, PhD, from the Finnish Meteorological Institute, for help with the interpretation of geomagnetic data, and Markus Henriksson, MD, from the National Public Health Institute, for intellectual support.
References Altamura, C., VanGastel, A., Pioli, R., Mannu, P., Maes, M., 1999. Seasonal and circadian rhythms in suicide in Cagliari, Italy. J. Affect. Disord. 53, 77 – 85. Ashkenazi, I.E., Reinberg, A., Bicakova-Rocher, A., Ticher, A., 1993. The genetic background of individual variations of circadian-rhythm periods in healthy human adults. Am. J. Hum. Genet. 52, 1250 – 1259. Beersma, D.G.M., Daan, S., Hut, R.A., 1999. Accuracy of circadian entrainment under fluctuating light conditions: contributions of phase and period responses. J. Biol. Rhythms 14, 320 – 329. Cleveland, W.S., Devlin, S.J., 1988. Locally-weighted regression: an approach to regression analysis by local fitting. J. Am. Stat. Assoc. 83, 596 – 610. Cleveland, R.B., Cleveland, W.S., McRae, J.E., Terpenning, I., 1990. STL: a seasonal-trend decomposition procedure based in loess. J. Official Statistics 6, 3 – 73. Czeisler, C.A., Duffy, J.F., Shanahan, T.L., Brown, E.N., Mitchell, J.F., Rimmer, D.W., Ronda, J.M., Silva, E.J., Allan, J.S., Emens, J.S., Dijk, D.-J., Kronauer, R.E., 1999. Stability, precision, and near-24-h period of the human circadian pacemaker. Science 284, 2177 – 2181.
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19 de Castro, J.M., Pearcey, S.M., 1995. Lunar rhythms of the meal and alcohol intake of humans. Physiol. Behav. 57, 439 – 444. Eastwood, M.R., Peacocke, J., 1976. Seasonal patterns of suicide, depression and electroconvulsive therapy. Br. J. Psychiatry 129, 472 – 475. Gutierrez-Garcia, J.M., Tusell, F., 1997. Suicides and the lunar cycle. Psychol. Rep. 80, 243 – 250. Harris, E.C., Barraclough, B., 1997. Suicide as an outcome for mental disorders: a meta-analysis. Br. J. Psychiatry 170, 205 – 228. Henriksson, M.M., Aro, H.M., Marttunen, M.J., Heikkinen, M.E., Isometsa¨, E.T., Kuoppasalmi, K.I., Lo¨nnqvist, J.K., 1993. Mental disorders and comorbidity in suicide. Am. J. Psychiatry 150, 935 – 940. Jewett, M.E., Kronauer, R.E., Czeisler, C.A., 1991. Light-induced suppression of endogenous circadian amplitude in humans. Nature 350, 59 – 62. Jones, R.H., Ford, P.M., Hamman, R.F., 1988. Seasonality comparison among groups using incidence data. Biometrics 44, 1131 – 1144. MacMahon, K., 1983. Short-term temporal cycles in the frequency of suicide, United States, 1972 – 1978. Am. J. Epidemiol. 117, 744 – 750. Maes, M., Meltzer, H.Y., Suy, E., De Meyer, F., 1993. Seasonality in severity of depression: relationships to suicide and homicide occurrence. Acta Psychiatr. Scand. 88, 156 – 161. Majercak, J., Sidote, D., Hardin, P.E., Edery, I., 1999. How a circadian clock adapts to seasonal decreases in temperature and day length. Neuron 24, 219 – 230. Ma¨kela¨, P., 1996. Alcohol consumption and suicide mortality by age among Finnish men, 1950 – 1991. Addiction 91, 101 – 112. Maldonado, G., Kraus, J.F., 1991. Variation in suicide occurrence by time of day, day of the week, month, and lunar phase. Suicide Life Threat. Behav. 21, 174 – 187. Mann, J.J., 1998. The neurobiology of suicide. Nat. Med. 4, 25 – 30. McCullagh, P., Nelder, J.A., 1994. In: Generalized linear models, 2nd ed. Chapman and Hall, London. Mortensen, P.B., Agerbo, E., Erikson, T., Qin, P., Westergaard-Nielsen, N., 2000. Psychiatric illness and risk factors for suicide in Denmark. Lancet 355, 9 – 12. Na¨yha¨, S., 1981. Short and medium-term variations in mortality in Finland: a study on cyclic variations, annual and weekly periods and certain irregular changes in mortality in Finland during period 1868 – 1972. Scand. J. Soc. Med. 8 (Suppl. 21), 1 – 101.
19
Na¨yha¨, S., Va¨isa¨nen, E., Hassi, J., 1994. Season and mental illness in an arctic area of northern Finland. Acta Psychiatr. Scand. 89 (Suppl. 377), 46 – 49. OY Alko AB, 1988 – 1995. Alcohol Statistical Yearbooks, 1987 – 1994. OY Alko AB, Helsinki. Parker, G., Gao, F., Machin, D., 2001. Seasonality of suicide in Singapore: data from the equator. Psychol. Med. 31, 549 – 553. Pirkola, S., Isometsa¨, E., Heikkinen, M., Lo¨nnqvist, J., 1997. Employment status influences the weekly patterns of suicide among alcohol misusers. Alcohol. Clin. Exp. Res. 21, 1704 – 1706. Preti, A., 1997. The influence of seasonal change on suicidal behaviour in Italy. J. Affect. Disord. 44, 123 – 130. Preti, A., Miotto, P., 1998. Seasonality in suicides: the influence of suicide method, gender and age on suicide distribution in Italy. Psychiatry Res. 81, 219 – 231. Rihmer, Z., Rutz, W., Pihlgren, H., Pestality, P., 1998. Decreasing tendency of seasonality in suicide may indicate lowering rate of depressive suicides in the population. Psychiatry Res. 81, 233 – 240. Salib, E., Gray, N., 1997. Weather conditions and fatal self-harm in North Cheshire 1989 – 1993. Br. J. Psychiatry 170, 473 – 477. Soueˆtre, E., Salvati, E., Candito, M., Darcourt, G., 1991. Biological clocks in depression: phase-shift experiments revisited. Eur. Psychiatry 6, 21 – 29. STAKES, 1996 – 2000. Intoxicants Statistical Yearbooks, 1995 – 1999. STAKES, Helsinki. Statistics Finland, 1988 – 2000. Statistical Yearbooks, 1987 – 1999. Statistics Finland, Helsinki. Taillard, J., Philip, P., Chastang, J.-F., Diefenbach, K., Bioulac, B., 2001. Is self-reported morbidity related to the circadian clock? J. Biol. Rhythms 16, 183 – 190. Wasielewski, J.A., Holloway, F.A., 2001. Alcohol’s interactions with circadian rhythms: a focus on body temperature. Alcohol Res. Health 25, 94 – 100. Yip, P.S.F., Chao, A., Ho, T.P., 1998. A re-examination of seasonal variation in suicides in Australia and New Zealand. J. Affect. Disord. 47, 141 – 150. Yip, P.S.F., Chao, A., Chiu, C.W.F., 2000. Seasonal variation in suicides: diminished or vanished. Br. J. Psychiatry 177, 366 – 369. Yvonneau, M., 1996. Apercß us de Dordogne, et de la lune, sur le suicide [Views from Dordogne, and the moon, on suicide]. Encephale 22 (Suppl. 4), 52 – 57.
Research report
Cyclic time patterns of death from suicide in northern Finland Timo Partonen a,*, Jari Haukka a, Kaisa Viilo b, Helina¨ Hakko b, Sami Pirkola a, Erkki Isometsa¨ a, Jouko Lo¨nnqvist a, Terttu Sa¨rkioja c, Erkki Va¨isa¨nen b, Pirkko Ra¨sa¨nen b a
Department of Mental Health and Alcohol Research, National Public Health Institute, Mannerheimintie 166, FIN-00300 Helsinki, Finland b Department of Psychiatry, University of Oulu, Oulu, Finland c County Administrative Board, Oulu, Finland Received 22 March 2002; received in revised form 28 June 2002; accepted 5 July 2002
Abstract Background: Time patterns of suicide have been attributed not only to social and psychological factors but also to direct geophysical effects. Seasonal variations in day length and temperature seem likely to contribute to the timing of the suicide process. Methods: We analysed all suicides (n = 1658) committed in a northern province of Finland during a period of 153 months. Daily data on the number of suicides, local weather conditions and geomagnetic storms were compiled and modelled with Poisson regression using the province population as the denominator, and with the means of harmonic series for seasonal variation. Time series analysis of monthly numbers of suicides was carried out using the seasonal-trend decomposition procedure based on loess. Results: Marked fluctuations in the number of suicides occurred during the study period ( P = 0.01). There was significant seasonal variation in death from suicide ( P = 0.01), but analysis of the meteorological data showed no evidence of effect on the risk of suicide. Limitations: Assessment of mental disorder or alcohol consumption was missing, since only data derived from death certificate was available for each case. Conclusions: The seasonal effect was significant, but remained modest compared to sex and age as risk factors for suicide. Preventive measures need to be tailored according to time of the year. D 2002 Elsevier B.V. All rights reserved. Keywords: Season; Suicide; Time
1. Introduction Suicides are not evenly distributed over particular periods of time, since social or psychological factors have immediate or delayed effect on the individual’s decision to commit suicide (Mann, 1998). Most stud-
* Corresponding author. Tel.: +358-9-4744-8660; fax: +358-94744-8563. E-mail address: [email protected] (T. Partonen). 0165-0327/$ - see front matter D 2002 Elsevier B.V. All rights reserved. doi:10.1016/S0165-0327(02)00236-7
ies have found that suicide rates tend to peak during spring and summer (Yip et al., 1998; Altamura et al., 1999). Moreover, it appears that climate and weather conditions can somewhat influence the occurrence of death from suicide, possibly by interacting with social factors (Preti, 1997; Salib and Gray, 1997; Preti and Miotto, 1998). Physical time-givers, as well as social and endogenous ones, can modify complex behaviours. Seasonal transition may contribute to the pathogenesis of the suicide process due to the substantial changes in
12
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
day length and temperature that occur in spring and autumn. The asynchrony between changes in temperature and light – dark cycles challenges the stability of endogenous clocks (Czeisler et al., 1999). It may acutely result in disordered periods of activity (Taillard et al., 2001), or disorganised behaviour, possibly leading to suicide in adverse circumstances. 1.1. Objectives Since geophysical factors may markedly influence daily behaviour patterns, it is important in terms of prevention to study whether they can also affect the occurrence of death from suicide. We therefore set out to study the time series of death from suicide in the population of a far northern region where there is extensive variation in levels of ambient temperature and daylight over the year. We hypothesised that increasing light exposure on still cold days during spring in specific may drive the individual into a later activity phase and the failure to adjust to time-giving signals from the natural habitat during this period of seasonal mismatch may lead to death from suicide more frequently than by chance alone.
2. Methods The study material consisted of all suicides (n = 1658) committed from 1 April 1987 to 31 December 1999 in the province of Oulu, which is located 64– 66jN in northern Finland, a northern European country with approximately 5 million inhabitants. Date, sex and age were recorded for each case. Population data on the number of inhabitants, sex ratio and age groups in the province during the study period were derived from statistical yearbooks published by Statistics Finland (1988 – 2000). Cause of death information for this study was obtained from the official death certificates that were collected as part of the National Suicide Prevention Project in Finland from 1 April 1987 to March 1988, and provided by the county administrative board thereafter. Since 1987, death certificate forms in Finland have complied with the recommendations of the World Health Organization’s International Classification of Diseases. The detailed content of the Finnish death certificate provides excellent accuracy and reli-
ability of information on the cause of death. Finnish law requires a medicolegal investigation of the cause and the manner of death whenever it has been unnatural, or is suspected to be so, or whenever it has been sudden or unexpected. The proportion of medicolegal investigations for suicide cases has been stable for many decades (Na¨yha¨, 1981), and the overall autopsy rate has remained relatively high, both of which allow for reliable conclusions on suicide trends over time. 2.1. Meteorological data The following daily data on local weather conditions in the Oulu borough of Kajaani were obtained from the meteorological observatory of the Finnish Meteorological Institute: mean temperature, mean cloud coverage (by eighths: 1 = clear, 4 = half-cloudy, and 8 = cloudy), mean air pressure, and wind direction and velocity. The day-to-day fluctuation in mean temperature was used for the analysis. The direction of wind was coded into four indicator variables for north, east, south and west. As for wind strength, a value of 1 was assigned if the wind velocity from a direction was over 10 m/s as measured at midday, indicating strong wind; otherwise the value was zero. Magnetic field variations routinely arise from current systems that are propagated by regular changes in solar radiation. However, interactions of the solar wind with the magnetosphere, and between the magnetosphere and ionosphere, also induce irregular current systems and magnetic field changes. Magnetic activity indices have been designed to describe variation in the geomagnetic field caused by these irregular current systems. The effect of geomagnetic storms was analysed using the geomagnetic activity indices measured in the borough of Nurmija¨rvi, Finland, at the geophysical observatory of the Finnish Meteorological Institute (for entry, see http://sumppu. fmi.fi/MAGN/K-index/NURMIJARVI/). 2.2. Astronomic data The year was divided into the astronomic seasons: spring (spring equinox to summer solstice), summer (summer solstice to autumn equinox), autumn (autumn equinox to winter solstice), and winter (winter solstice to spring equinox). The lunar cycles and
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
calendar weeks were also recorded in order to assess the effect of cyclic time patterns independent of day length and temperature. Both are ancient systems of reckoning time, the former a natural one and the latter purely man-made. The Almanac Office of the Department of Astronomy, University of Helsinki, provided these data. 2.3. Seasonal mismatch The year was also divided into the thermal seasons: spring (daily mean temperature above 0 jC), summer (above 10 jC), autumn (below 10 jC), and winter (below 0 jC). Because of the mismatch between the two definitions of seasons (see Fig. 1), we assessed the periods of conflicting dates, or the seasonal mismatches, as follows: spring gap (spring equinox to thermal spring), summer gap (thermal summer to summer solstice), autumn gap (thermal autumn to autumn equinox), and winter gap (thermal winter to winter solstice). The mismatch data were calculated on a day-by-day basis over the study period. The primary event in the spring gap is a change in the length of day, whereas the other three seasonal mismatches begin with a change in daily mean temperature. The spring gap may thus impart a unique time-giving stimulus to endogenous clocks. Specifically, we hypothesised that increasing light exposure on still cold days during spring may drive the individual into a later activity phase (Majercak et al., 1999) that would compromise well-being. This may end with subsequent delays (Beersma et al., 1999) on top of the delayed phase position and finally stop the clock running (Jewett et al., 1991). Exposure to colder
13
days during autumn may in turn reverse the delayed activity phase and induce accelerating phase advances that would result in attempts to slow down the clockwork. 2.4. Statistics Time series analysis of monthly numbers of suicides was carried out using locally weighted regression, the seasonal-trend decomposition procedure based on loess (STL; Cleveland and Devlin, 1988; Cleveland et al., 1990). The STL decomposes time series in three components (trend, seasonal, and the remainder) using a sequence of smoothing operations. This method is robust in detecting both trends and seasonal variation. Separate analyses were applied for men and women. Daily data containing the number of suicides were compiled. Each day was represented by six observations because the data included all combinations of sex and age categories (under 30, 30– 60 and over 60 years). The daily number of suicides (range 0 –3) was modelled with Poisson regression using population in the province of Oulu as the denominator. Potential explanatory variables were as follows: the year as a categorical variable (1987 –1999); sex; age; seasonal components with a total of four cosine and sine oscillations in a year; the lunar phase in four categories (equal intervals of 7.4 days around the phases, new moons as the reference); weekdays in two categories (weekends, and the remainder as the reference); changes in daily mean temperature (jC) as a categorical variable using first and ninth deciles ( 3.7 and + 3.7) as the cut-off points; daily mean air pressure
Fig. 1. Schematic representation of the concept of seasonal mismatch. Abbreviations: T, thermal seasons, the lengths of which vary depending on local weather conditions each year; A, astronomic seasons. Numbers from 1 to 4 represent seasonal gaps from spring towards winter.
14
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
Table 1 Number of suicides (n) and incidence per 1000 person-years
3. Results
Variable
Of the 1413 male and 313 female suicides there were complete data available for analysis on 1658 (96%) subjects (Table 1). Male sex (v2 = 735.3, df = 1, P < 0.0000001) and two older age groups (30 – 60, plus over 60 years, v2 = 344.7, df = 2, P < 0.0000001) were associated with the risk of suicide. Our analyses showed that marked changes in the number of suicides occurred during the study period (v2 = 28.0, df = 12, P = 0.01). The risk of suicide decreased significantly over the three consecutive years from 1994 to 1996 (Table 2), and the weekend was linked to increased risk of death from suicide ( + 13% on average). The time patterns of suicide occurrence are
Year: 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Sex: Male Female Age: < 30 years 30 – 60 years >60 years
n
Personyears
Incidence
109 150 133 135 148 146 127 98 113 112 118 136 133
327.7 435.5 436.1 439.9 441.4 446.9 446.8 449.7 451.8 454.1 452.9 452.8 454.8
0.33 0.34 0.31 0.31 0.34 0.33 0.28 0.22 0.25 0.25 0.26 0.30 0.29
1359 299
2843.3 2847.2
0.48 0.11
Table 2 Risk ratios (RR) and 95% confidence intervals (CI) from the Poisson regression model
389 1017 252
2469.5 2279.5 941.5
0.16 0.45 0.27
Variable
RR
95% CI
Year: 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
1.00 1.06 0.93 0.93 1.02 0.99 0.86 0.65 0.75 0.74 0.78 0.89 0.87
0.83 – 1.35 0.72 – 1.20 0.73 – 1.20 0.80 – 1.31 0.77 – 1.27 0.66 – 1.11 0.50 – 0.86 0.58 – 0.97 0.57 – 0.96 0.60 – 1.01 0.69 – 1.15 0.67 – 1.12
Lunar cycle: New moon First quarter Full moon Third quarter
447 393 436 382
1404.7 1430.1 1429.0 1426.7
0.32 0.27 0.31 0.27
Week: Weekdays Weekend
896 762
3251.4 2439.1
0.28 0.31
(kPa) as a categorical variable using first and ninth deciles (994.8 and 1026.0) as the cut-off points; four daily indicator variables of wind; major geomagnetic storms using the value of 50 aK as the cut-off point; and mismatch of the astronomic and thermal seasons, reflecting the conflicting signals of light and temperature from the habitat. The seasonal components were modelled with the means of harmonic series (Jones et al., 1988), and year and the harmonics of seasonal components were included in all models (McCullagh and Nelder, 1994). Other potential explanatory variables were added sequentially (as above) into a model one by one, the significance was tested using the likelihood ratio test, and a given variable was left in the model if significant.
Sex: Male Female Age: < 30 years 30 – 60 years >60 years
1.00 0.22
0.19 – 0.25
1.00 2.84 1.96
2.53 – 3.19 1.67 – 2.30
Lunar cycle: New moon First quarter Full moon Third quarter
1.00 0.86 0.96 0.84
0.75 – 0.99 0.84 – 1.09 0.73 – 0.96
Weeks: Weekdays Weekend
1.00 1.13
1.03 – 1.25
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19 Fig. 2. A seasonal-trend decomposition of suicide occurrence among men over the study period. Panels from the top: the original data, the smoothed data for trend using a time window size of 47 months, the seasonal component, and the remainder from the model. The scale bars to the right represent the corresponding unit in each figure.
15
16 T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19 Fig. 3. Seasonal-trend decomposition of suicide occurrence among women over the study period. Panels from the top: the original data, the smoothed data for trend using a time window size of 47 months, the seasonal component, and the remainder from the model. The scale bars to the right represent the corresponding unit in each figure.
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
Fig. 4. Annual variation in the relative risk of suicide in the study population. The line represents the fitted curve of the monthly data for risk ratios using January as the reference, and the bars represent the 95% confidence limits.
presented for men and women in Figs. 2 and 3, respectively. We also found that the seasonal variation in suicide occurrence was of significance among the study population (v2 = 14.6, df = 4, P = 0.01), with peaks during long days over the year (Fig. 4). Men and women did not differ in the extent of this variation. Analysis of the meteorological data provided no evidence of an effect on the risk of suicide (data not shown). Neither did the seasonal mismatch contribute to death from suicide during spring (v2 = 3.1, df = 1, P = 0.08) nor autumn (v2 = 1.3, df = 1, P = 0.2), thus falsifying our hypothesis. In addition, the risk of suicide was associated with the lunar cycle (v2 = 8.4, df = 3, P = 0.04), and the calendar week (v2 = 6.5, df = 1, P = 0.01). The risk of suicide was highest around the phase of a new moon and lower for the next phase ( 14% on average), around the phase of a full moon ( 4% on average), and again during the final phase ( 16% on average) of the lunar cycle. There was no interaction between the seasonal and lunar effects concerning the risk of suicide.
4. Discussion Our main finding was that contrary to recently published findings (Yip et al., 2000; Parker et al., 2001) there was a significant seasonal variation in death from suicide. The highest incidence was ob-
17
served at the turn of summer to autumn, and a second peak occurred during spring. Our findings also suggest that despite a marked difference in time patterns for suicide between men and women in the long term, there is none in the seasonal variation. The seasonal effect was modest though, as compared with sex and age as the known risk factors for suicide. We also showed, using versatile and minute data, that local weather conditions, geomagnetic storms, or the periods of seasonal mismatch had no effect on the risk of suicide among a population living in northerly regions. The principal strength of our study was that we had an unselected suicide population with a relatively high number of suicide cases in a remote northern region which experiences a relatively extreme change of the seasons. Both the lunar cycle and the calendar week showed significant associations, but these were secondary to the seasonal effect on the risk of suicide. Suicide occurrence has not previously been linked to phases of the lunar cycle (MacMahon, 1983; Maldonado and Kraus, 1991; Gutierrez-Garcia and Tusell, 1997), except in one study which found fewer suicides in the phase of a full moon than in the first quarter of the lunar cycle (Yvonneau, 1996). Our results showed that the phase of a new moon was associated with increased risk of suicide. Although alcohol consumption appears to increase on average by 26% at the new moon relative to the full moon (de Castro and Pearcey, 1995), there is no immediately apparent explanation for our finding. However, it is of note here that the association, although a significant one, was weak. One reason for the existence of seasonal and circadian variations in suicide may be that the function of intrinsic time-keeping mechanisms (Ashkenazi et al., 1993) is compromised prior to death. The body clocks are driven by signals such as changes in luminance and ambient temperature from the individual’s natural habitat to match the solar day, in such a way that the internal phase is a reliable predictor of solar time. Rhythmic variations in bodily functions and overt behaviour allow individuals to anticipate environmental changes, and to seek protection. The accuracy of this minute synchronisation depends on the individual response characteristics of body clocks (Beersma et al., 1999) on which the knowledge of time, physical activity, and state of health have a vital influence. Depressive disorders (Soueˆtre et al., 1991)
18
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19
and alcohol misuse (Wasielewski and Holloway, 2001) easily affect these characteristics and may thereby routinely compromise the intrinsic clockwork, especially when there is a shortage of the stimuli that give time from the habitat. However, the present findings of the effect of seasonal mismatch, as calculated on the basis of actual meteorological data and tested against our hypothesis, gave no evidence of disordered intrinsic clockwork in suicide. 4.1. Limitations One weakness of our trial was the absence of data on mental health for each case. Mental disorders are frequent among individuals completing suicide (Henriksson et al., 1993), and carry high risk for death from suicide (Harris and Barraclough, 1997). From the epidemiological perspective, the conditions most relevant for prevention are affective disorders and alcohol dependence (Mortensen et al., 2000). Seasonal changes in depressive disorder, its treatment indices, and suicide seem to coincide, underlying the importance of depressive episodes as a cause of death from suicide (Eastwood and Peacocke, 1976; Maes et al., 1993). Our finding of seasonal variation in suicide occurrence links to a previous report from Finland that feelings of depression were most prevalent in northerly regions during the spring (Na¨yha¨ et al., 1994). Recently, another study attributed the declining suicide rate to the improved care of those with depressive disorder (Rihmer et al., 1998). We did not have individual data on alcohol consumption either. The suicide rate of Finnish men aged under 50 years appears to be associated with per capita alcohol consumption (Ma¨kela¨, 1996). Typical habits of alcohol consumption also seem to be linked to weekly variations in the number of suicides (Pirkola et al., 1997). Our data agreed with this, since the risk of suicide was increased at weekends. In addition, there was a significant decrease in suicide occurrence from 1994 to 1996. Interestingly, there was also a marked decrease in the consumption of alcohol beverages, from 8.7 to 6.0 l per capita calculated as 100% alcohol, from 1990 to 1996 in the province of Oulu (OY Alko AB, 1988 – 1995; STAKES, 1996 –2000). This fall in alcohol consumption apparently coincided with reduced suicide mortality in the province.
Another study weakness was that the meteorological data were derived from only one point in the study area. However, this type of measurement error usually tends to attenuate results, which indicates a stronger real effect compared to the observed one.
5. Conclusion Marked changes in the number of deaths from suicide were observed during the study period. There was a substantial seasonal variation in suicide occurrence among the study population. Compared to sex and age, however, the seasonal effect on the risk of suicide remained modest. Periods of seasonal mismatch, local weather conditions, and geomagnetic storms had no effect on the risk.
Acknowledgements We thank Heikki Nevanlinna, PhD, from the Finnish Meteorological Institute, for help with the interpretation of geomagnetic data, and Markus Henriksson, MD, from the National Public Health Institute, for intellectual support.
References Altamura, C., VanGastel, A., Pioli, R., Mannu, P., Maes, M., 1999. Seasonal and circadian rhythms in suicide in Cagliari, Italy. J. Affect. Disord. 53, 77 – 85. Ashkenazi, I.E., Reinberg, A., Bicakova-Rocher, A., Ticher, A., 1993. The genetic background of individual variations of circadian-rhythm periods in healthy human adults. Am. J. Hum. Genet. 52, 1250 – 1259. Beersma, D.G.M., Daan, S., Hut, R.A., 1999. Accuracy of circadian entrainment under fluctuating light conditions: contributions of phase and period responses. J. Biol. Rhythms 14, 320 – 329. Cleveland, W.S., Devlin, S.J., 1988. Locally-weighted regression: an approach to regression analysis by local fitting. J. Am. Stat. Assoc. 83, 596 – 610. Cleveland, R.B., Cleveland, W.S., McRae, J.E., Terpenning, I., 1990. STL: a seasonal-trend decomposition procedure based in loess. J. Official Statistics 6, 3 – 73. Czeisler, C.A., Duffy, J.F., Shanahan, T.L., Brown, E.N., Mitchell, J.F., Rimmer, D.W., Ronda, J.M., Silva, E.J., Allan, J.S., Emens, J.S., Dijk, D.-J., Kronauer, R.E., 1999. Stability, precision, and near-24-h period of the human circadian pacemaker. Science 284, 2177 – 2181.
T. Partonen et al. / Journal of Affective Disorders 78 (2004) 11–19 de Castro, J.M., Pearcey, S.M., 1995. Lunar rhythms of the meal and alcohol intake of humans. Physiol. Behav. 57, 439 – 444. Eastwood, M.R., Peacocke, J., 1976. Seasonal patterns of suicide, depression and electroconvulsive therapy. Br. J. Psychiatry 129, 472 – 475. Gutierrez-Garcia, J.M., Tusell, F., 1997. Suicides and the lunar cycle. Psychol. Rep. 80, 243 – 250. Harris, E.C., Barraclough, B., 1997. Suicide as an outcome for mental disorders: a meta-analysis. Br. J. Psychiatry 170, 205 – 228. Henriksson, M.M., Aro, H.M., Marttunen, M.J., Heikkinen, M.E., Isometsa¨, E.T., Kuoppasalmi, K.I., Lo¨nnqvist, J.K., 1993. Mental disorders and comorbidity in suicide. Am. J. Psychiatry 150, 935 – 940. Jewett, M.E., Kronauer, R.E., Czeisler, C.A., 1991. Light-induced suppression of endogenous circadian amplitude in humans. Nature 350, 59 – 62. Jones, R.H., Ford, P.M., Hamman, R.F., 1988. Seasonality comparison among groups using incidence data. Biometrics 44, 1131 – 1144. MacMahon, K., 1983. Short-term temporal cycles in the frequency of suicide, United States, 1972 – 1978. Am. J. Epidemiol. 117, 744 – 750. Maes, M., Meltzer, H.Y., Suy, E., De Meyer, F., 1993. Seasonality in severity of depression: relationships to suicide and homicide occurrence. Acta Psychiatr. Scand. 88, 156 – 161. Majercak, J., Sidote, D., Hardin, P.E., Edery, I., 1999. How a circadian clock adapts to seasonal decreases in temperature and day length. Neuron 24, 219 – 230. Ma¨kela¨, P., 1996. Alcohol consumption and suicide mortality by age among Finnish men, 1950 – 1991. Addiction 91, 101 – 112. Maldonado, G., Kraus, J.F., 1991. Variation in suicide occurrence by time of day, day of the week, month, and lunar phase. Suicide Life Threat. Behav. 21, 174 – 187. Mann, J.J., 1998. The neurobiology of suicide. Nat. Med. 4, 25 – 30. McCullagh, P., Nelder, J.A., 1994. In: Generalized linear models, 2nd ed. Chapman and Hall, London. Mortensen, P.B., Agerbo, E., Erikson, T., Qin, P., Westergaard-Nielsen, N., 2000. Psychiatric illness and risk factors for suicide in Denmark. Lancet 355, 9 – 12. Na¨yha¨, S., 1981. Short and medium-term variations in mortality in Finland: a study on cyclic variations, annual and weekly periods and certain irregular changes in mortality in Finland during period 1868 – 1972. Scand. J. Soc. Med. 8 (Suppl. 21), 1 – 101.
19
Na¨yha¨, S., Va¨isa¨nen, E., Hassi, J., 1994. Season and mental illness in an arctic area of northern Finland. Acta Psychiatr. Scand. 89 (Suppl. 377), 46 – 49. OY Alko AB, 1988 – 1995. Alcohol Statistical Yearbooks, 1987 – 1994. OY Alko AB, Helsinki. Parker, G., Gao, F., Machin, D., 2001. Seasonality of suicide in Singapore: data from the equator. Psychol. Med. 31, 549 – 553. Pirkola, S., Isometsa¨, E., Heikkinen, M., Lo¨nnqvist, J., 1997. Employment status influences the weekly patterns of suicide among alcohol misusers. Alcohol. Clin. Exp. Res. 21, 1704 – 1706. Preti, A., 1997. The influence of seasonal change on suicidal behaviour in Italy. J. Affect. Disord. 44, 123 – 130. Preti, A., Miotto, P., 1998. Seasonality in suicides: the influence of suicide method, gender and age on suicide distribution in Italy. Psychiatry Res. 81, 219 – 231. Rihmer, Z., Rutz, W., Pihlgren, H., Pestality, P., 1998. Decreasing tendency of seasonality in suicide may indicate lowering rate of depressive suicides in the population. Psychiatry Res. 81, 233 – 240. Salib, E., Gray, N., 1997. Weather conditions and fatal self-harm in North Cheshire 1989 – 1993. Br. J. Psychiatry 170, 473 – 477. Soueˆtre, E., Salvati, E., Candito, M., Darcourt, G., 1991. Biological clocks in depression: phase-shift experiments revisited. Eur. Psychiatry 6, 21 – 29. STAKES, 1996 – 2000. Intoxicants Statistical Yearbooks, 1995 – 1999. STAKES, Helsinki. Statistics Finland, 1988 – 2000. Statistical Yearbooks, 1987 – 1999. Statistics Finland, Helsinki. Taillard, J., Philip, P., Chastang, J.-F., Diefenbach, K., Bioulac, B., 2001. Is self-reported morbidity related to the circadian clock? J. Biol. Rhythms 16, 183 – 190. Wasielewski, J.A., Holloway, F.A., 2001. Alcohol’s interactions with circadian rhythms: a focus on body temperature. Alcohol Res. Health 25, 94 – 100. Yip, P.S.F., Chao, A., Ho, T.P., 1998. A re-examination of seasonal variation in suicides in Australia and New Zealand. J. Affect. Disord. 47, 141 – 150. Yip, P.S.F., Chao, A., Chiu, C.W.F., 2000. Seasonal variation in suicides: diminished or vanished. Br. J. Psychiatry 177, 366 – 369. Yvonneau, M., 1996. Apercß us de Dordogne, et de la lune, sur le suicide [Views from Dordogne, and the moon, on suicide]. Encephale 22 (Suppl. 4), 52 – 57.