Relationship between daylength and suicide in Finland

Background Many previous studies have documented seasonal variation in suicides globally. We re-assessed the seasonal variation of suicides in Finland and tried to relate it to the seasonal variation in daylength and ambient temperature and in the discrepancy between local time and solar time. Methods The daily data of all suicides from 1969 to 2003 in Finland (N = 43,393) were available. The calendar year was divided into twelve periods according to the length of daylight and the routinely changing time difference between sun time and official time. The daily mean of suicide mortality was calculated for each of these periods and the 95% confidence intervals of the daily means were used to evaluate the statistical significance of the means. In addition, daily changes in sunshine hours and mean temperature were compared to the daily means of suicide mortality in two locations during these afore mentioned periods. Results A significant peak of the daily mean value of suicide mortality occurred in Finland between May 15th and July 25th, a period that lies symmetrically around the solstice. Concerning the suicide mortality among men in the northern location (Oulu), the peak was postponed as compared with the southern location (Helsinki). The daily variation in temperature or in sunshine did not have significant association with suicide mortality in these two locations. Conclusions The period with the longest length of the day associated with the increased suicide mortality. Furthermore, since the peak of suicide mortality seems to manifest later during the year in the north, some other physical or biological signals, besides the variation in daylight, may be involved. In order to have novel means for suicide prevention, the assessment of susceptibility to the circadian misalignment might help.


Background
Current data on the routinely occurring peaks of deaths from suicide are conflicting [1,2]. However, for the past four decades in Finland, the seasonal pattern has been stronger the lower the suicide mortality has been [3]. There is a clear peak of suicide occurrence around May or June [4][5][6][7] and a preceding peak in suicide attempts around April [8]. Furthermore, another smaller peak of suicide occurrence exists around October [7,9]. These two mortality peaks, being similar and more robust the further away the country locates from the equator, have been explained by socio-demographic and socio-economic factors [10], but since this seasonal pattern has existed for decades [11], if not centuries [12], biological factors are likely.
Major depressive episodes are known to contribute to suicide substantially [13,14], and a history of mood disorders and psychiatric hospitalization associates clearly with the seasonal occurrence of suicides [15,16]. Desynchronization of physiological rhythms, e.g. desynchronization of the circadian rhythm of core body temperature with the sleep-wake cycle [17][18][19] and some clock gene variants [20,21], can be associated with mood disorders. Based on our earlier psychological autopsy studies of death from suicide [22] and the data from the nationwide suicide program in Finland [23], we hypothesized that the circadian misalignment among the depressed may increase during spring, and thereby predispose to suicidal behaviors [24].
Rest-activity cycles during the day [25] and sleep stages at night [26] are controlled by circadian clocks, but they are frequently disturbed among the depressed. Furthermore, the principal circadian clock entrains to the sun light [27][28][29], by tracking the daily changes in rise and set times of the sun and the variation in the length of the day [30][31][32]. Thus, the timing of light exposure is relevant to entrainment and influences the course of mood disorders [33,34]. Therefore, we hypothesized that it is the key to the suicide mortality peaks whether the light-dark transitions give the principal circadian pacemaker a signal to accelerate or decelerate, especially among the depressed. In addition, since sunshine and ambient temperature are potential timegivers, modulate the function of biological clocks [35], and associate with deaths from suicide [3,36], we aimed to test their effect, as well.  (Tables 1 and 2). Two phenomena, which affect the timing and the speed of the light-dark transitions regularly each year, were selected a priori as the potential factors that might challenge the biological clocks and produce circadian misalignment. First, we focused on the length of the photoperiod, because at high to temperate latitudes around spring and fall equinoxes the transitions between day and night are most rapid and the durations of twilight short, as a consequence of the rotation of the earth. Second, we focused on the constant mismatch between the sun time (hereafter ST) and the coordinated universal time (hereafter UCT), arising from the earth's tilt and elliptical orbit around the sun.

Methods
The nominal calendar year was split into twelve periods according to these two phenomena; first into four astronomical seasons, which are determined by spring and fall equinoxes and summer and winter solstices (for the definition, see http://asa.nao.rl.ac.uk/), and second into eight periods, by the equation of time (for the definition, see http://www.nmm.ac.uk/explore/astronomyand-time/time-facts/the-equation-of-time), as follows (see also Figure 1). From February 11 th to May 14 th (hereafter marked as X1) and from July 26 th to November 3 rd (X2) ST goes fast compared with UCT and in between those periods, that is, from May 15 th to July 25 th (Y1) and from November 4 th to February 10 th (Y2) it goes slow. Furthermore, another categorization was made based on the equation of time separating periods when ST is either ahead or behind the UCT. In other words, ST is constantly ahead of the UCT, from April 15 th to June 13 th (A1) and from September 1 st to December 25 th (A2), and constantly behind the UCT, from June 14 th to August 31 st (D1), and from December 26 th to April 14 th (D2). Hence, ST deviates from UCT constantly and is maximally behind at February 11 th (approximately 14 minutes) and vice versa maximally ahead at November 3 rd (approximately 16 minutes). The Almanac Office at the University of Helsinki http:// almanakka.helsinki.fi/ both provided the dates for the astronomical seasons and calculated the dates for the To evaluate the effect of daily sunshine hours and temperature on suicide mortality, we focused on two cities on a similar longitude but with dissimilar photoperiod: first, Helsinki (60°9.7'N, 24°57.3'E), which is the capital of Finland in the south, and second, Oulu (65°1 .0'N, 25°30.0'E), which is a central city of the northern part of the country, 600 km north from Helsinki. In Helsinki 5062 suicides were committed by men, and 2160 by women, whereas 903 by men and 278 by women in Oulu. The Finnish Meteorological Institute http://www.fmi.fi/ provided us with the daily data on sunshine hours and temperature, measured within the 25-km radius from these cities throughout the study period. For the day to day analysis, the daily sunshine, measured in minutes per day, (hereafter S) and the daily temperature, measured in degrees in Celsius and averaged as the daily mean value, (hereafter T) were compared with those on the previous day and changes were marked as (+) indicating an increase, and (-) indicating a decrease from the previous day. Thus, we ended with four types of days according to weather changes, coded as T+S+, T+S-, T-S+, and T-S-, concerning the data from Helsinki and Oulu regions.
In order to take into account the differences in the yearly population sizes within and between Helsinki and Oulu, the daily means of suicides were calculated into daily means of suicide mortality rates (suicides per 100,000), for men and women, per each year, and for both cities (Tables 1 and 2). Furthermore, in order to control for the different lengths of each period studied, and to avoid the bias of having dominance of certain type of weather changes within any period of the year, the daily mean of suicide mortalities (number of suicides per day, with 95% confidence intervals [CIs]) was calculated for each period in separate (Tables 3, 4, and 5).
Finally, to rule out a potential confounder, we analyzed whether daylight saving time (hereafter DST) had any effect on the suicide mortality. DST was introduced in Finland 1981. From 1981 to 1994 DST lasted from the end of March until the end of September (hereafter DST1), and since 1995 DST has been in use from the end of March until the end of October, as in most parts of Europe (hereafter DST2). We calculated suicide mortality rates during one month period before, and after the transitions into and out of DST, separately for the years 1981 to 1994 (DST1) and years 1995 to 2003 (DST2), for which the suicide mortality rates of the corresponding periods during the years 1969 to 1980 were used as controls (Tables 6,7,8,9,10,and 11).
The 95% CIs of the daily mean values, controlled for the length of a period of study and the male and female population sizes in a region of study, were used to evaluate the statistical significance, so that if they did not overlap with each other, it was judged to indicate a marked statistical significance.

Results
In Finland, during the years 1969 to 2003, the daily mean of suicide mortality was at the highest, with a statistical significance, for both men (mean = .124, CI = .118-.129) and women (mean = .032, CI = .030-.033),     (Table 4).

Local photoperiod
The highest daily mean of suicide mortality seem to have emerged later in Oulu compared with Helsinki, but only for men. Therefore, the results of men are reported here in more detail. The daily mean of suicide mortality was at the highest during the period Y1 in Helsinki (mean = .142, CI = .130-.154, Table 4), but during the period X2 i.e. from July 26 th to November 3 rd in Oulu (mean = .150, CI = .131-.169, Table 4). The same postponed pattern was found also when the time pattern of suicide mortality was evaluated by seasons. The daily mean of suicide mortality was highest in Helsinki during spring (mean = .138, CI = .128-.148), but during summer in Oulu (mean = .159, CI = .138-.180). Furthermore, a similar postponed pattern was seen from A1 (Helsinki) to D1 (Oulu) periods (Table 5). However, these results did not reach statistical significance.

Local daily weather changes
For men, the days with T+S+ seem to have had the highest daily mean of suicide mortality both in Helsinki, during the period Y1 (mean = .149, CI = .131-.167), and in Oulu, during the period X2 (mean = .206, CI = .122-.290), which were the most "dangerous" periods in these cities. However, when estimated by the 95% confidence intervals, there was no statistical difference in the variation of means of suicide mortality between the four types of weather changes. The daily mean of suicide mortality in Helsinki and Oulu, however, do exceed the nationwide daily means of suicide mortality (mean = .124 for Y1 in Finland, and mean = .112 for X2 in Finland), as do all the underlined values for different types of weather changes in Helsinki and Oulu compared with each period at issue in Tables 3, 4, and 5.

Daylight saving time
The use or timing of daylight saving time did not have a significant effect on the suicide mortality (Tables 6, 7, 8,  9, 10, and 11).

Nationwide results
Our key finding of statistical significance demonstrates the increased suicide mortality on nationwide level in Finland during the period from May 14 th to July 25 th . This 76-day period covers symmetrically both sides of summer solstice (Figure 1). During this period there is only 1 to 4 hours of darkness during the night in Helsinki but no darkness at all in Oulu. For the photoperiod dynamics in these locations, see http://www. gaisma.com/en/location/helsinki.html and http://www. gaisma.com/en/location/oulu.html, whose sunrise and sunset calculations are based on the algorithms displayed on National Oceanic and Atmospheric Administration Surface Radiation Research Branch web site at http://esrl.noaa.gov/gmd/grad/solcalc/calcdetails.html, and e.g. for latitudes less than 72°north and south, accuracy is approximately one minute. It is of note here that the photoperiod in Finland due to its time zone is asymmetrical throughout the year, the period of daylight being always shorter for the a.m. hours than it is for the p.m. hours. This phenomenon influences the mechanisms that decode the duration of the melatonin signal in the melatonin-target tissues. From the circadian-clock point of view, this period (May 14 th to July 25 th ) is a challenge to alignment of the circadian rhythms with the sleep-wake cycle, and it resembles "the critical spring photoperiodic window" on intermediate to long days that has been characterized in sheep [37]. Some possible biological mechanisms for our current finding are briefly discussed in the following. The very long day (20 to 24 hours of daylight) might challenge the network within the circadian pacemaker that is comprised of the so-called evening and morning active cells, and that takes part in the seasonal adaptation in diurnal animals such as fruit flies [38,39] and   sheep [37,[40][41][42]. If this holds for humans as well, it is not known at the moment. If it does, it could mean that, when day lengths approximate fall and winter, the morning active cells dominate the circadian output, e.g. the sleep-wake behavior. This dominance of hierarchy is gradually transferred to the evening active cells as the days get longer in spring [38,39], the coincidence effect of the morning and evening active cells disappearing when the melatonin signal duration becomes insufficient to sensitize adenylate cyclase and to support a peak expression of the morning-active cells [37]. Interestingly, the speeding up of the evening active cells (e.g. by sunshine) makes the morning active cells run faster in long (summer) but not in short (winter) days [38]. In Finland, which is located at high to temperate latitudes with the light-dark transitions being most rapid around equinoxes, the asymmetrical photoperiod possibly favors the evening-active cells, and produces pronounced melatonin-dependent effects on gene expression during spring and fall. Whether such "locked morning active cells" contribute to the peak in deaths from suicide in spring in particular is not known. However, CRY2 and PER2 genetic variants, which might influence the evening and morning signals from the circadian pacemaker system, associate with depression vulnerability [43,44] in humans. Therefore, depressed individuals in particular Daily mean of suicide mortality during one month period before (-1 m), and after (+1 m) the switch away daylight saving time. * = DST was not in use in Finland m = daily mean of suicide mortality CI = confidence interval of the mean  might suffer from entrainment errors during periods that challenge the circadian pacemaker and predispose to circadian misalignment.

Local daily weather changes
The complexity of the circadian pacemaker system suggests that signals other than the seasonal changes in photoperiod, such as temporary variations in local weather conditions, are likely to play a role in the entrainment process [35,45]. Our finding of the later suicide peak in the northern area of study, Oulu-region, supports this. However, the daily mean of suicide mortality was almost as high also during the Y1 period in Oulu, as in Helsinki.
Hereafter we discuss the potential influence of daily weather changes for the suicide mortality in Helsinki and Oulu during the peak periods.
During the most dangerous periods, Y1 in Helsinki and X2 in Oulu, days with T+S+ seemed to be the worst for suicide mortality. From the circadian point of view the long daylight combined with the daily increase in ambient temperature and sunshine hours (T+S+) may have further phase advanced the circadian rhythm of the male suicide victims. An increase in sunshine hours and exposure to light may accelerate and advance the phase of the principal circadian clock, but an increase in ambient temperature and exposure to heat may have a similar effect [46]. The peaks of suicides have associated Daily mean of suicide mortality during one month period before (-1 m), and after (+1 m) the switch away daylight saving time. * = DST was not in use in Finland m = daily mean of suicide mortality CI = confidence interval of the mean  with ambient temperature in earlier studies [47][48][49], but so far, to our knowledge, the role of the circadian clocks has not been addressed. Many lines of evidence suggest that abnormalities in the thermoregulatory processes are common among the depressed and therefore may cause or maintain the circadian misalignment. Patients with a major depressive episode tend to have elevated body temperature throughout the night, not during the day, and a phase advance of the circadian rhythm of core body temperature [18]. As hot nights might advance the phase of the circadian clock [50], and nocturnal body temperature during rapid-eye-movement sleep is influenced by hot, not cold, ambience [51], the dynamics of nocturnal temperatures might contribute to the advanced and rather fixed phase positions of circadian rhythms in major depressive episodes. In addition, sleep abnormalities, characteristically excessive rapid-eye-movement sleep at the cost of slow-wave sleep [17], are likely to give an abnormal (accelerating) feedback to the principal circadian pacemaker [26]. Further, during winter the duration of rapid-eye-movement sleep per night tends to increase [52], giving no support to deceleration and thereby favoring the desynchronization that may result in lowered mood and the subsequent increase in risk of suicide.
Daily fluctuations in temperature may play a part in the timing of suicides, either in combination with the long day length, or possibly also as a separate stressor. Studies concerning the over-activity in the functions of brown adipose tissue among the depressed [53] are most interesting in this respect, since the over-activity of brown adipose tissue may lead to reduced adaptation to rapid changes in ambient temperature that are typical during spring and fall. Once being activated, brown adipose tissue does not become quiescent easily [54], and if having been over-activated, it may through the thermoregulatory defect lead to disruption of the sleep-wake cycle and appetite control, and lead to early morning awakenings and loss of weight of the affected individual. Whether this kind of "vernalization failure" characterizes a suicide process and contributes to a mortality peak year after year is not known, but needs experimental data for analysis. However, in line with this background, for the Y1 period the daily mean of suicide mortality of men was at its lowest during the days of T-S-in Helsinki (mean = .124, CI = .106-.142, Table 4) and during the days of T-S+ in Oulu (mean = .140, CI = .090-.190, Table 4), suggesting that T-is a common nominator for the "safer" weather changes in both locations. T-S+ days were the "safest" also during the X2 period in Oulu. The daily decrease in temperature could therefore serve as a protective change during otherwise warm season. However, as the daily means did not differ significantly between the four types of weather changes, this is somewhat speculative thus far.

Limitations
Our limitation here is that we did not have the diagnostic information of the suicide victims, and that we demonstrate associations only, which do not necessarily tell anything about causality. Another limitation is that we did not have access to a suitable method, e.g. molecular-timetable methods [55] to be applied to a range of tissues, such as the brain and brown adipose tissue, from autopsy studies, to be able to analyze a mechanism of action and thereby to demonstrate a potential link between abnormalities in the circadian pacemaker system and death from suicide. On the other hand, our strengths include the nationwide sample of suicides for a long period of time, from a country with a high suicide mortality rate.

Conclusions
Our main findings here are that suicide mortality is higher during summer months and that daily changes in sunshine and ambient temperature are likely to modify the suicide mortality. Our findings presented herein now wait for tests by others in independent materials and is thus open to replication and the subsequent verification or falsification of the hypothesis. Some experimental data would be urgently needed for explanation of the mechanisms of action that take place in the brain of depressed patients and predispose them to suicide within those particular periods of time that we identified here. Suicide is a long process, whereas the timing of death from suicide appears far from random. In Finland from 1969 to2003 suicide mortality was elevated from May 15 th to July 25 th . This phenomenon should be considered also in clinical practice, since it bears implications for suicide prevention. managed the literature searches and analyses and author LH wrote the first draft of the manuscript. All authors contributed to and approved the final manuscript.