- Email: [email protected]
Relationship between photoperiod and hospital admissions for mania in New South Wales, Australia
Author’s Accepted Manuscript Relationship between photoperiod and hospital admissions for mania in New South Wales, Australia Gordon Parker, Dusan Hadzi-Pavlovic, Adam Bayes, Rebecca Graham www.elsevier.com/locate/jad
PII: DOI: Reference:
S0165-0327(17)31062-5 http://dx.doi.org/10.1016/j.jad.2017.09.014 JAD9222
To appear in: Journal of Affective Disorders Received date: 28 May 2017 Revised date: 11 September 2017 Accepted date: 13 September 2017 Cite this article as: Gordon Parker, Dusan Hadzi-Pavlovic, Adam Bayes and Rebecca Graham, Relationship between photoperiod and hospital admissions for mania in New South Wales, Australia, Journal of Affective Disorders, http://dx.doi.org/10.1016/j.jad.2017.09.014 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Relationship between photoperiod and hospital admissions for mania in New South Wales, Australia Gordon Parkera,b,*, Dusan Hadzi-Pavlovica,b, Adam Bayesa, Rebecca Grahama,b a
School of Psychiatry, UNSW, Sydney, NSW, Australia
b
Black Dog Institute, Hospital Rd, Randwick, NSW, 2031, Australia
*
Corresponding author: Hospital Rd, Randwick, NSW, 2031, Australia. [email protected]
Abstract: Background: Causes for a seasonal impact on admissions for mania remain to be clarified. We examined the impact of photoperiod, rate of change of photoperiod and hours of sunshine on admissions over an extended period. Methods: Monthly admission data to NSW psychiatric hospitals for more than twenty thousand patients admitted for mania over a fifteen-year period were correlated with photoperiod and sunshine changes. Results: While the peak in admissions occurred in spring, the shift in admissions being underrepresented to being precipitously over-represented corresponded with the photoperiod commencing to increase in winter (i.e. July). Analyses identified rate of change in photoperiod as somewhat more influential than change in photoperiod and with hours of sunshine not making a distinctive contribution. Immediate and delayed impacts of rate of change as well as change in photoperiod across the whole year accounted for a distinctive 20% of the variance in hospital admissions. Limitations: Validity of mania diagnoses cannot be established from the data set, admission data were obtained from across the state while meteorological data were obtained from the capital city, lag periods between onset of a mania and hospitalization (while identified) would impact on associations, social factors were not included and study associations do not imply causality. Conclusions: The lack of a strong year-long correlation may reflect photoperiod changes being only a weak causal factor or that its influence may be through a strong impact phase after the winter solstice and with the spring peaking of admissions reflecting secondary photoperiod or other influences. Keywords: mania, bipolar disorder, photoperiod
1. Introduction In non-equatorial regions hospital admissions for mania demonstrate a seasonal pattern. As reviewed previously (Geoffroy et al., 2014; Parker and Graham, 2016, Wang and Chen, 2013), admissions in the northern hemisphere show a bimodal peak in spring and summer, and either show a similar pattern in southern hemisphere countries or a single peak in spring. While socio-cultural factors such as holiday periods have been nominated as possible causes, most studies have focused on climatic variables. Numerous candidates have been proposed, including photoperiod and hours of bright sunshine (Parker and Walter, 1982), relative humidity, air ionization and barometric pressure (Mawson and Smith, 1981), solar/ultraviolet radiation, temperature (Myers and Davies, 1978; Silverstone and Romans-Clarkson, 1989), snow cover and rain (Geoffroy et al., 2014). While any impact of photoperiod has not been demonstrated as necessarily a primary causal factor, we pursue its potential impact in this report. The ‘photoperiod’ is the interval in a 24-hour period when the individual is exposed to light and, as measured in this study, is day length. Its length varies across the year, having maximal range at the highest latitudes, and minimal at the lowest latitudes (where day length and night length are always approximately equal). Photoperiod changes are linked with a number of biological systems in nature, such as signaling plants to flower as well as hibernation and sexual behaviours in certain animals (Taiz and Zeiger, 2006). The photoperiod has also been linked to bipolar disorder, with increase in its length reported as associated with admissions for mania, either in the month of – or month prior to – admission. The latter could reflect a time lag between actual onset of a manic episode and any necessitated hospitalization, with Winokur (1976) reporting that approximately two-thirds of admissions for mania occurred within a month of onset, and Francis and Gasparo (1994) reporting a mean duration of 3.3 weeks before admission. Hours of bright sunshine (and their variations) might be expected to correlate with photoperiod – but do show some variation as a consequence of state weather variables – and have been previously examined (Parker and Walter,1982; Volpe and Del Porto, 2006) in relation to their impact on admissions for mania. Our rationale for pursuing the impact of photoperiod principally (as well as any additional impact of hours of bright sunshine) reflects varying results from several previous studies. In a New Zealand study, Sayer et al. (1991) reported that the previous month’s day length and mean daily temperature had the greatest effect on admissions for mania, while in a UK study, Suhail and Cochrane (1998) quantified that 68% of monthly fluctuations in admissions for mania was accounted for by hours of daylight and hours of sunshine. In a North American study, Cassidy and Carroll (2002) reported that the greatest increases in hospital admissions for mania occurred in February and March – corresponding with the largest lengthenings of photoperiod (suggesting that the rate of change in photoperiod should be studied in addition to the change in photoperiod). However, several studies have failed to demonstrate any such associations (Lee et al., 2002; Volpe et al., 2010), albeit with the latter study undertaken in Brazil. However, in an earlier Brazilian study (Volpe et al. 2006) undertaken in the same subtropical mesothermic climate (and latitude some 20 degrees), admissions for mania correlated positively with (i) the average index and previous month’s hours of sunshine and (ii) differenced mean temperature.
2. Methods 2.1. Database Our data were obtained from the New South Wales (NSW) Department of Health and comprised 21,882 individuals hospitalised in NSW mental health units from 2000 to 2014 for ICD-10 diagnosed mania (i.e. the Department of Health records only ICD-10 diagnoses made on discharge), and which would therefore not have included ‘mixed states’. Admissions for each month were expressed as a rate per 100,000, based on the NSW population for the year, and adjusted to a 30-day month. Ethical approval was not required as the study analysed publicly available aggregated data. Photoperiod data were calculated using the R package geosphere (based on year, day of year, and latitude). These provided two measures of photoperiod: daily photoperiod, and the daily rate of change in photoperiod, each of which was converted to a mean for each month. For the state of NSW (the source of our admission data) the maximum difference in photoperiod between the northern (28S) and southern (37S) ends of the state is 50 minutes at the two solstices. The capital Sydney - where the bulk of the NSW population resides and contributes to the majority of psychiatric admissions has a minimum day length of 9 hrs 45 min and a maximum of 14 hrs 25 min. Mean monthly hours of sunshine for 2000–2014 were obtained from the Australian Bureau of Meteorology. 2.2. Analysis The primary statistical analysis was a set of vector autoregressive (VAR) models (Milhoj, 2016) fitted using the R package vars. Essentially, these model a time series (e.g. rate of admission) as a function of previous values (lags) of the time series as well as previous values of other time series in the model. If a time series lacks certain statistical properties it is said to be non-stationary, and it is often recommended in the literature that before fitting VAR models such series should be made stationary. This can be achieved by not using the observed values but rather using the difference between each value and the previous value (differencing). We used the Dickey-Fuller test (Milhoj, 2016) to examine if a series was stationary or not. We examined the relationship between rate of admissions and mean monthly values of (a) photoperiod, (b) the rate of change in photoperiod, and (c) hours of sunshine. For the 180-month time series (15 years x 12 months), we fitted sets of VAR models (see Table 1). These equations modelled the rate using (i) previous values of the rate of admissions (up to a lag of four months), (ii) previous values of the photoperiod variables up to a lag of four months, and (iii) a constant and linear trend. The rate of change in photoperiod was always superior to photoperiod (in terms of R2). As both rate and hours of sunshine were identified as non-stationary (and stationary after first differences) we also report results for differenced series.
3. Results In Figure 1, number of monthly admissions - expressed as standardised differences (residuals) between the observed and the expected (the latter being the average of all monthly admissions adjusted for population and to a 30-day month) numbers of admissions - are charted (in grey) for each month. The red (solid) line represents the photoperiod or day length (averaged across the 15 years), and the blue (dashed) line
represents the rate of change of the photoperiod across the year in Sydney. The photoperiod is at its maximum at the summer solstice (December 21 - 22). The days then shorten at an increasingly rapid rate until the autumn equinox (March 20 - 21) after which the photoperiod continues to decrease, but it does so more slowly. The photoperiod achieves its briefest duration at the winter solstice (June 20 - 21) when the rate of decrease in the photoperiod reaches its minimum. Thereafter the days start to become longer, lengthening at an increasingly rapid rate until the spring equinox (September 22 - 23), after which the photoperiod keeps lengthening (but at a decreasing rate) until the summer solstice - when the rate of increase in the photoperiod reaches its minimum and the days change from lengthening to shortening again. Turning to admissions, for all but January and February in the first half of the year, as the days become shorter, there were fewer than expected admissions for mania, and for all months in the second half of the year, as the days lengthen, there was a greater than expected number of admissions for mania. The switch from fewer than expected to greater than expected occurred abruptly in July (during winter) after the period with the least apparent shortening of the days, and just as the days started to lengthen. More specifically, the fewest admissions for mania occurred in April (following maximal shortening) and June (corresponding with the shortest length). However, while the photoperiod is still increasing in December, the peak in admissions occurred in November, rather than in December, arguing against a consistent association between oscillations in photoperiod and admission rates. In terms of seasonal pattern, the figure shows that the peak in admissions occurred in spring (September to November in NSW) and with the peak month for the year being the last month of spring. 3.1. Comparison of models Table 1 reports comparisons of the relative contributions of the variables based on the variance accounted for (R2) and, while the tabled results differ according to whether the series are differenced or not, the conclusions are consistent. In both cases, the addition of hours of sun (Model II) to a model with only the lagged rate (Model I which predicts current admission solely from previous admissions) produced just a small increase in R2 (3% and 5%), whereas the addition of rate of change in photoperiod (Model III) had a much greater increase in R2 (13% and 14%). The final model (Model IV) showed a small improvement over Model III (p = 0.07 in the undifferenced data, and p = 0.006 in the differenced data), indicating that both variables can be included usefully. We also report (in Table 2) the individual variable effects (via regression coefficients) in the final models, and now consider those data under relevant headings. 3.2. Rate of admission In the undifferenced series the current admission rate was positively associated only with the previous (lag 1) rate (p = 0.02), while in the differenced series all four lags were negatively associated and significant (p 0.01). 3.3. Rate of change in photoperiod (length of day) In both series, the rate of change was non-significant at lag 1 and significant at lags 2 - 4. Photoperiod has a cyclic nature which results in a particular pattern of
correlations between any given month and previous months. For example, values six months apart are near mirror images but negatively correlated close to 1, while values 3 months apart are correlated near to zero. Consequently, the coefficients for the multivariate models reported in Table 2 are not readily interpretable, with negative coefficients the opposite of what they would be if included singly. They do, however, locate change at a lag of 3 months as more prominent than other lags. 3.4 Hours of sunshine In the undifferenced series none of the lags were significant, though lags 1 and 4 were near significance. In the differenced series only lag 4 was significant (p = 0.016). In summary, admissions in the previous month, photoperiod length and magnitude of change in length accounted for about 20% of the variance (an R = 0.45) in rate of admissions. Length of photoperiod appeared relevant at up to four months previously, with rate of change and absolute change in photoperiod length most relevant three to four months previously. Hours of sunshine appeared a less salient variable.
4. Discussion 4.1 Principal findings Our study confirmed the common finding of a peak number of hospital admissions for mania occurring in spring. Perhaps the most distinctive finding was the abrupt change in admissions in winter (June to August in NSW), from fewer than expected admissions in June to a substantively greater than expected number in July and coinciding with the photoperiod changing from shortening to lengthening. Such a winter inflection point in admission rates for those with mania has been reported in several other photoperiod studies (Carney et al., 1989; Cassidy and Carroll, 2002; Myers and Davies, 1978), and this phenomenon may be as important as the spring excess emphasized in the literature. Changes quantified over the study period (2000-2014) are similar to those we documented in our earlier study involving 1,876 patients with a diagnosis of ‘manicdepressive psychosis, manic phase’ again admitted to a psychiatric facility in NSW over the 1971-1976 period, but did show some differences. For example, the earlier study quantified a peak in admissions in September and October (as against October and November in this study). Possible explanations may reflect change in diagnostic practices over the two study periods and the much larger sample in the current study, but also the impact of variations in total light exposure and, in particular, from nonsolar light sources. Bauer et al. (2017) have considered how changes in light exposure (such as LED lighting) may have influenced adaptability to a springtime circadian challenge. Thus, we need to concede that temporal changes in light exposure may explain differences across published studies. In focusing on any photoperiod effects, we examined (as detailed in Table 2) both the impact of variations in the photoperiod and the rate of change in the photoperiod, as well as examining for both more immediate and for lagged effects. We also examined for any impact of hours of sunshine but established a minimal contribution from that variable. Overall, photoperiod and rate of change in
photoperiod correlated similarly with admissions and with their overall contribution, along with the previous month’s admissions rate, accounting for some 20% of the variance in the level of admissions, indicative of a distinctive seasonal contribution. In the Introduction we noted that several studies had identified a lag between changes in photoperiod and admission patterns. Our study identified a number of lag impacts but with the most distinct effects associated with 3-4 month lags. Our findings suggest that, while there is an overall association between admissions for mania and the month of the year, the causal impact of photoperiod on admission rates may arise only at certain critical periods. The sudden and distinctive change from underrepresentation to over-representation of admissions corresponds with the photoperiod commencing to increase and suggests a ‘kick start’ or initial activation effect. The following attenuation in admissions might reflect initial photoperiod changes becoming less salient after the abrupt increase in July. The subsequent spring/summer peaking may arise from additional photoperiod mechanisms (e.g. when day length reaches a critical threshold), though it must be conceded that factors other than photoperiod might be in operation. The overall modest association between admission rates and both photoperiod rate of change and photoperiod change may reflect a diffusion of the initial ‘kick start’ effect, homeostatic mechanisms coming into play or multiple photoperiod mechanisms operating on differing groups of bipolar individuals at different times of the year and thus obscuring pristine delineation. Thus, to the extent that photoperiod changes impact on mania, rather than there being any single mechanism, there may be multiple photoperiod effects linked with differing bipolar genotypes. We now consider some possible mechanisms whereby environmental light is transduced by the nervous system into downstream effects on sleep and mood. Circadian rhythms in humans are primarily entrained by the ‘zeitgeber’ light. Entering through the eyes, light is detected by intrinsically photosensitive retinal ganglion cells (ipRGCs) which express the photopigment melanopsin (which has the greatest spectral sensitivity for blue light wavelengths). Via the retinohypothalamic tract, the ipRGCs signal daytime to the body’s ‘master clock’, the suprachiasmatic nucleus (SCN), which in turn inhibits pineal gland production of melatonin (Henriksen et al., 2016). Melatonin, which is produced from its precursor serotonin (5HT), is involved in numerous physiological processes including synchronizing the biological rhythms to the earth’s 24-hour day and has been shown to affect sleep patterns (Arendt, 2006; Wehr, 1991), and which Wehr et al. (1987) described as the ‘final common pathway’ in causing mania. In an earlier study undertaken in NSW we reported associations between photoperiod changes, as well as the rate of increase in bright sunshine (luminance), and increased hospital admissions for mania in spring (Parker and Walter, 1982). We hypothesized that the rapid increase in luminance in July and August stimulates the pineal gland and reduces melatonin secretion, with downstream effects on sleep and mood control so contributing to the seasonal pattern. There are, however, a number of other potential mechanisms linking photoperiod to the onset of manic episodes including the impact of altering circadian rhythms in individuals with a bipolar disorder. It has been postulated that bipolar individuals are vulnerable to a range of seasonal and climatic factors (Volpe et al., 2008), with supersensitivity to light a trait marker for bipolar disorder (Lewy et al., 1985; Nathan et al., 1999; Wright and Lack, 2001). In bipolar I individuals, a heritable excessive sensitivity of melatonin to light was reported, independent of illness state (Lewy et al, 1981; Hallam et al, 2006). Lee
et al. (2013) reported differing variants of the NF1A gene region between bipolar individuals which might play a role in seasonal mania – and they postulated the existence of a NF1A related sub-phenotype of bipolar disorder, with affected individuals experiencing seasonally-related manic episodes while those unaffected have manic episodes without a seasonal component. Sleep deprivation related to the increase in daylight hours as winter ends might also be linked to the induction of mania (Wehr et al., 1987). Additionally, Dulcis et al (2013) demonstrated that adult rats exposed to long active summer-like period lengths exhibited risk taking behaviours and elevated dopamine levels, with both mechanisms associated with mania. Levels of neurotransmitters in the brain, in particular serotonin (5HT), have also been reported as showing associations with variations in sunlight. In another Australian study, Lambert et al. (2002) reported that turnover of 5HT is lowest in winter and that the rate of its production in the human brain was directly related to the prevailing duration of bright sunlight, rapidly rising with increasing luminosity, while dopamine turnover did not show a seasonal pattern. Similarly, low extracellular 5HT has been found in autumn and winter, and high extracellular 5HT around the summer solstice (Pail et al., 2011). Thus it is possible that photoperiod (and sunlight) variations alter neurotransmitter concentrations that might play a role in triggering manic episodes. The effects of light treatment also implicate the possible association of photoperiod with manic episodes. The use of light therapy for bipolar depression and seasonal affective disorder is well described (Pail et al., 2011), while an intervention for mania, so-called ‘dark therapy’ may be salient. The therapy involves a reduction in light exposure, with a recent randomized control trial by Henriksen et al. (2016) showing that the use of blue light blocking glasses compared to placebo (in addition to treatment as usual) for bipolar mania was an effective treatment. Here the hypothesized mechanism is that the ipRGCs are blocked from being activated, creating a state of ‘virtual darkness’ in the brain and preserving melatonin production and with subsiding of the mania accruing from sleep and circadian rhythms being restored. 4.2. Study limitations A number of study limitations are noted. First, the data on admissions for mania were derived from a data set of state-wide health statistics and the validity of diagnoses of mania cannot be established. Second, we did not analyse data for any other – or all other – psychiatric conditions, so that we cannot reject the possibility of an unspecified effect of photoperiod on admissions of other diagnostic sub-sets or on admissions in general. Third, we examined data for all state hospitals and, while the great majority of admissions occurred in Sydney hospitals, a minority would have occurred in rural regions and where weather variables would have differed to some degree. Fourth, we cannot discount an impact from the move to daylight saving (when the clocks are moved forward an hour) which occurs in NSW on the first of October, although our Figure 1 data indicate that the swing to over-representation occurred earlier. Fifth, and as noted earlier, there is likely to be lag period between onset of a manic episode and being hospitalized as a consequence (and this may explain our finding of lagged effects). Sixth, data were only acquired for those admitted to hospital and where any seasonality impact may differ (e.g. be more severe) from the pattern for those whose episodes of mania do not result in hospital
admission. Seventh, we have described associations which do not allow causality to be claimed, especially when there may be a number of mediating factors. Eighth, while photoperiod changes over the year, it is largely constant from year to year, and so it can only really explain “constant” changes across a year and not fluctuations from year to year that might be influenced by other meteorological factors (e.g. temperature, ambient light) which change across the years and even social (e.g. the impact of holidays) and service (e.g. admission procedures) factors.
4.2. Implications In this study examining the impact of photoperiod changes on hospitalizations for mania, we quantified a modest level of variance accounted for across the whole year by our two study weather variables (rate of change in photoperiod and change in photoperiod), and which could argue for photoperiod having less impact than other meteorological variables. However, if photoperiod changes are causal, our findings allow an explanatory hypothesis. The greatest shift from fewer than expected to greater than expected admissions for mania occurred in July, when the days change from shortening to lengthening. We therefore hypothesize that there is a ‘kick start’ impact in July and with the subsequent peaking of admissions reflecting the ‘impact’ phenomenon attenuating or a second threshold photoperiod mechanism operating – if photoperiod changes are the key determining factor. We overview studies suggesting that changes in photoperiod potentially operate on certain bipolar individuals by altering circadian rhythms and inducing mania via such mechanisms as sleep disruption or changes in neurotransmitter levels. Author statement Role of funding The study was supported by NHMRC Program Grant (1037196).
Contributors GP designed the study, DHP undertook the statistical analyses, RG and AB undertook the literature review, GP and AB wrote first draft. All authors contributed to and have approved the final manuscript.
Conflicts of interest none
Acknowledgments This study was supported by an NHMRC Program Grant (1037196).
References Arendt, J., 2006. Melatonin and human rhythms. Chronobiol. Int. 23, 21–37.
Bauer, M., Glenn, T., Alda, M., Aleksandrovich, M.A., Andreassen, O.A., Angelopoulos, E., Ardau, R., Ayhan, Y., Baethge, C., Bharathram, S.R., Bauer, R., Bauna, B.T., Becerra-Palars, C., Bellivier, F., Belmaker, R. H., Berk, M., Bersudsky, Y., Bicakci, Ş., Birabwa-Oketcho, H., Bjella, T.D., Bossini, L., Cabrera, J., Cheung, E.Y.W., Del Zompo, M., Dodd, S., Donix, M., Etain, B., Fagiolini, A., Fountoulakis, K.N., Frye, M.A., Gonzalez-Pinto, A., Gottlieb, J.F., Grof, P., Harima, H., Henry, C., Isometsä, E.T., Janno, S., Kapczinski, F., Kardell, M., Khaldi, S., Kliwicki, S., König, B., Kot, T.L., Krogh, R., Kunz, M., Lafer, B., Landén, M., Larsen, E.R., Lewitzka, U., Licht, R.W., LopezJaremillo, C., MacQueen, G., Manchia, M., Marsh, W., MartinezCengotitabengoa, M., Melle, I., Meza-Urzúa, F., Yee Ming, M., Monteith, S., Morken, G., Mosca, E., Munoz, R., Mythri, S.V., Nacef, F., Nadella, R.K., Nery, F.G., Nielsen, R.E., O’Donovan, C., Omrani, A., Osher, Y., Østermark Sørensen, H., Quali, U., Pica Ruiz, Y., Pilhatsch, M., Pinna, M., da Ponte, F.D.R., Quiroz, D., Ramesar, R., Rasgon, N., Reddy, M.S., Reif, A., Ritter, P., Rybakowski, J.K., Sagduyu, K., Scippa, Â.M., Severus, E., Simhandl, C., Stein, D.J., Strejilevich, S., Subramaniam, M., Sulaiman, A.H., Suominen, K., Tagata, H., Tatebayashi, Y., Tondo, L., Torrent, C., Vaaler, A.E., Veeh, J., Vieta, E., Viswanath, B., Yoldi-Negrete, M., Zetin, M., Zgueb, Y., Whybrow, P.C., 2017. Solar insolation in springtime influences age of onset of bipolar I disorder. Acta Psychiatr. Scand. 1-12 Carney, P.A., Fitzgerald, C.T., Monaghan, C., 1989. Seasonal variation mania. In Thompson, C., Silverstone, T. (Eds). Seasonal Affective Disorder. CNS (Clinical Neuroscience) Publishers, London, UK. Cassidy, F., Carroll, B.J., 2002. Seasonal variation of mixed and pure episodes of bipolar disorder. J. Affect. Disord. 68, 25-31. Dulcis, D., Jamshidi, P., Leutgeb, S., Spitzer, N.C. 2013. Neurotransmitter switching in the adult brain regulates behavior. Science. 340, 449-453. Francis, A., Gasparo, P., 1994. Interval between symptom onset and hospitalization in mania. J. Affect. Disord. 31, 179-185. Geoffroy, P.A., Bellivier, F., Scott. J., Etain, B., 2014. Seasonality and bipolar disorder: a systematic review, from admission rates to seasonality of symptoms. J. Affect. Disord. 168, 210-223. Hallam, K.T., Olver, J.S., Chambers, V., Begg, D.P., McGrath, C., Norman, T.R, 2006. The hereditability and sensitivity to bright nocturnal light in twins. Psychoendocrinology, 31, 867-875. Henriksen, T.E.G., Skrede, S., Fasmer, O.B., Schoeyen, H., Leskauskaite, I., Bjørke‐ Bertheussen, J., et al., 2016. Blue‐blocking glasses as additive treatment for mania: a randomized placebo‐controlled trial. Bipolar Disord. 18, 221–232. Lambert, G.W., Reid, C., Kaye, D.M., Jennings, G.L., Esler, M.D., 2002. Effect of sunlight and season on serotonin turnover in the brain. Lancet 360, 1840–1842. Lee, H.J., Kim, L., Joe, S-H., Suh, K-Y., 2002. Effects of season and climate on the first manic episode of bipolar affective disorder in Korea. Psychiatry Res. 113, 151-159. Lee, H.J., Woo, H.G., Greenwood, T.A., Kripke, D.F., Kelsoe, J.R., 2013. A genomewide association study of seasonal pattern mania identifies NF1A as a possible susceptibility gene for bipolar disorder. J. Affect. Disord. 145, 200-207. Lewy, A.J., Nurnberger, J.I., Wehr, T.A., Pack, D., Becker, L.E., Powell, R., Newsome, D.A., 1985. Supersensitivity to light: possible trait marker for manicdepressive illness. Am. J. Psychiatry 142, 725–727.
Mawson, D.A., Smith, A.L., 1981. Relative humidity and manic admissions in the London area. Br. J. Psychiatry 138, 134-138. Milhoj, A. 2016. Multiple time series modelling using the SAS VARMAX procedure. Cary, NC: SAS Institute Inc. Myers, D.H., Davies, P., 1978. The seasonal incidence of mania and its relationship to climatic variables. Psychol. Med. 8, 433-440. Nathan, P.J., Burrows, G.D., Norman, T.R., 1999. Melatonin sensitivity to dim white light in affective disorders. Neuropsychopharmacology 21, 408–413. Pail, G., Huf, W., Pjrek, E., Winkler, D., Willeit, M., Praschak-Rieder, N., et al., 2011. Bright-light therapy in the treatment of mood disorders. Neuropsychobiology 64, 152–162. Parker, G., Walter, S., 1982. Seasonal variation in depressive disorders and suicidal deaths in New South Wales. Br. J. Psychiatry 140, 626-632. Parker, G.B., Graham, R.K., 2016. Seasonal variations in rates of hospitalisation for mania and hypomania in psychiatric hospitals in NSW. J. Affect. Disord. 191, 289-291. Sayer, H.K., Marshall, S., Mellsop, G.W., 1991. Mania and seasonality in the southern hemisphere. J. Affect. Disord. 23, 151-156. Silverstone, T.R., Romans-Clarkson, S., 1989. Bipolar affective disorder: causes and prevention of relapse. Br. J. Psychiatry 154, 321-335. Suhail, K., Cochrane, R., 1998. Seasonal variations in hospital admissions for affective disorders by gender and ethnicity. Soc. Psychiatry Psychiatr. Epidemiol. 33, 211-217. Taiz, L., Zeiger, E., 2006. Plant Physiology (4th Edition). Sinauer Associates, Inc., Sunderland MA. Volpe, F.M, Del Porto, J.A., 2006. Seasonality of admission for mania in a psychiatric hospital of Belo Horizonte, Brazil. J. Affect Disord. 94, 243-248. Volpe, F.M., Tavares, A., Del Porto, J.A., 2008. Seasonality of three dimensions of mania: psychosis, aggression and suicidality. J. Affect. Disord. 108, 95–100. Volpe, F.M., da Silva, E.M., dos Santos, T.N., de Freitas, D.E., 2010. Further evidence of seasonality of mania in the tropics. J. Affect. Disord. 124, 178-182. Wang, B., Chen, D. 2013. Evidence of seasonal mania: a review. J Psychiat Practice. 19, 301-308. Wehr, T.A., Sack, D.A., Rosenthal, N.E., 1987. Sleep reduction as a final common pathway in the genesis of mania. Am. J. Psychiatry 144, 201-204. Wehr, T.A., 1991. The durations of human melatonin secretion and sleep respond to changes in day length (photoperiod). J. Clin. Endocrinol. Metab. 73, 1276-1280. Winokur, G., 1976. Duration of illness prior to hospitalization (onset) in the affective disorders. Neuropsychobiology 2, 87-93. Wright, H.R., Lack, L.C., 2001. Effect of light wavelength on suppression and phase delay of the melatonin rhythm. Chronobiol. Int. 18, 801–808.
Sep
Oct
Nov
Dec
Jul
Aug
Sep
Oct
Nov
Dec
2.1
2.1
Aug
-1.6 -2.1
-1.6 -2.1
-1.1
-0.5
0
0.5
1.1
1.6
Admissions above/below expected Photoperiod Daily Change in Photoperiod (minutes/day) Solstice (Jun/Dec) & Equinox (Mar/Sep)
1.6
13.8
Jul
1.1
13.3
Jun
0.5
12.7
May
0
12.1
Apr
-0.5
11.5
Mar
-1.1
10.4 9.8
11
Feb
Admissions above/below expected (standardized residuals)
Photoperiod (hours) / Day-to-day Change in Photoperiod (minutes/day)
14.4
Jan
Jan
Feb
Mar
Apr
May
Jun
Time of Year
Figure 1. Change and rate of change in photoperiod for Sydney and admissions for mania in NSW. Table 1.Relative contribution of the variables based on the variance accounted for (Rsquared) in the series of VAR models fitted to undifferenced and differenced data. Model Variables in model Undifferenced Differenced1 I
Rate of admissions
R2 = 0.09
R2 = 0.26
II
Admissions + hours of sunshine
R2 = 0.12
R2 = 0.31
III
Admissions + change in photoperiod
R2 = 0.22
R2 = 0.39
IV
Admissions + hours of sunshine + change in photoperiod
R2 = 0.26
R2 = 0.45
1.Rate of change in photoperiod not differenced as it was stationary. Table 2. Final VAR models examining individual variable effects via regression coefficients for rate of admissions in undifferenced and differenced series Undifferenced series Differenced series Variable1
Lag
Est.
Rate of admissions
1
0.183 0.08
2.29 0.023
–0.08 0.08
– 1.02 0.309
2 3 4
SE
t
p
Est.
SE
t
P
–0.65 0.08
– 8.53
< 0.001
–0.61 0.09
– 7.00
< 0.001 < 0.001 0.012
0.10 0.08
1.29 0.198
–0.35 0.09
– 4.05
0.09 0.08
1.11 0.271
–0.19 0.08
– 2.54
Rate of change in photoperiod2
Hours of sunshine
1
6.30 3.50
1.80 0.074
5.56 3.71
1.50
0.136
2
– 13.46 6.50
– 2.07 0.040
– 16.47 6.77
– 2.43
0.016
3
16.82 6.29
2.68 0.008
23.35 6.78
3.45
0.007
4
–7.57 3.33
– 2.27 0.024
– 13.40 3.67
– 3.65
< 0.001
1
–0.03 0.01
– 1.88 0.062
–0.02 0.01
– 1.31
0.193
0.00 0.01
– 0.08 0.933
–0.01 0.02
– 0.82
0.413
0.00 0.01
– 0.02 0.984
0.00 0.02
– 0.20
0.845
0.03 0.01
1.91 0.058
0.03 0.01
2.43
0.016
2 3 4 2
R = 0.26 p = < 0.001
2
R = 0.45 p = < 0.001
1. Estimates for the constant and linear trend not reported. 2. Rate of change in photoperiod was stationary and not differenced. 3. Values in bold indicate statistically significant estimates. Highlights
Admissions for mania switched from being under- to over-represented when the photoperiod started to increase. Change and rate of change in photoperiod accounted for a distinctive 20% of the variance in admissions. The lack of a strong year-long correlation may reflect photoperiod changes being only a weak causal factor. Photoperiod may operate through a strong impact phase after the winter solstice and with the spring peaking of admissions reflecting secondary photoperiod or other influences.
PII: DOI: Reference:
S0165-0327(17)31062-5 http://dx.doi.org/10.1016/j.jad.2017.09.014 JAD9222
To appear in: Journal of Affective Disorders Received date: 28 May 2017 Revised date: 11 September 2017 Accepted date: 13 September 2017 Cite this article as: Gordon Parker, Dusan Hadzi-Pavlovic, Adam Bayes and Rebecca Graham, Relationship between photoperiod and hospital admissions for mania in New South Wales, Australia, Journal of Affective Disorders, http://dx.doi.org/10.1016/j.jad.2017.09.014 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Relationship between photoperiod and hospital admissions for mania in New South Wales, Australia Gordon Parkera,b,*, Dusan Hadzi-Pavlovica,b, Adam Bayesa, Rebecca Grahama,b a
School of Psychiatry, UNSW, Sydney, NSW, Australia
b
Black Dog Institute, Hospital Rd, Randwick, NSW, 2031, Australia
*
Corresponding author: Hospital Rd, Randwick, NSW, 2031, Australia. [email protected]
Abstract: Background: Causes for a seasonal impact on admissions for mania remain to be clarified. We examined the impact of photoperiod, rate of change of photoperiod and hours of sunshine on admissions over an extended period. Methods: Monthly admission data to NSW psychiatric hospitals for more than twenty thousand patients admitted for mania over a fifteen-year period were correlated with photoperiod and sunshine changes. Results: While the peak in admissions occurred in spring, the shift in admissions being underrepresented to being precipitously over-represented corresponded with the photoperiod commencing to increase in winter (i.e. July). Analyses identified rate of change in photoperiod as somewhat more influential than change in photoperiod and with hours of sunshine not making a distinctive contribution. Immediate and delayed impacts of rate of change as well as change in photoperiod across the whole year accounted for a distinctive 20% of the variance in hospital admissions. Limitations: Validity of mania diagnoses cannot be established from the data set, admission data were obtained from across the state while meteorological data were obtained from the capital city, lag periods between onset of a mania and hospitalization (while identified) would impact on associations, social factors were not included and study associations do not imply causality. Conclusions: The lack of a strong year-long correlation may reflect photoperiod changes being only a weak causal factor or that its influence may be through a strong impact phase after the winter solstice and with the spring peaking of admissions reflecting secondary photoperiod or other influences. Keywords: mania, bipolar disorder, photoperiod
1. Introduction In non-equatorial regions hospital admissions for mania demonstrate a seasonal pattern. As reviewed previously (Geoffroy et al., 2014; Parker and Graham, 2016, Wang and Chen, 2013), admissions in the northern hemisphere show a bimodal peak in spring and summer, and either show a similar pattern in southern hemisphere countries or a single peak in spring. While socio-cultural factors such as holiday periods have been nominated as possible causes, most studies have focused on climatic variables. Numerous candidates have been proposed, including photoperiod and hours of bright sunshine (Parker and Walter, 1982), relative humidity, air ionization and barometric pressure (Mawson and Smith, 1981), solar/ultraviolet radiation, temperature (Myers and Davies, 1978; Silverstone and Romans-Clarkson, 1989), snow cover and rain (Geoffroy et al., 2014). While any impact of photoperiod has not been demonstrated as necessarily a primary causal factor, we pursue its potential impact in this report. The ‘photoperiod’ is the interval in a 24-hour period when the individual is exposed to light and, as measured in this study, is day length. Its length varies across the year, having maximal range at the highest latitudes, and minimal at the lowest latitudes (where day length and night length are always approximately equal). Photoperiod changes are linked with a number of biological systems in nature, such as signaling plants to flower as well as hibernation and sexual behaviours in certain animals (Taiz and Zeiger, 2006). The photoperiod has also been linked to bipolar disorder, with increase in its length reported as associated with admissions for mania, either in the month of – or month prior to – admission. The latter could reflect a time lag between actual onset of a manic episode and any necessitated hospitalization, with Winokur (1976) reporting that approximately two-thirds of admissions for mania occurred within a month of onset, and Francis and Gasparo (1994) reporting a mean duration of 3.3 weeks before admission. Hours of bright sunshine (and their variations) might be expected to correlate with photoperiod – but do show some variation as a consequence of state weather variables – and have been previously examined (Parker and Walter,1982; Volpe and Del Porto, 2006) in relation to their impact on admissions for mania. Our rationale for pursuing the impact of photoperiod principally (as well as any additional impact of hours of bright sunshine) reflects varying results from several previous studies. In a New Zealand study, Sayer et al. (1991) reported that the previous month’s day length and mean daily temperature had the greatest effect on admissions for mania, while in a UK study, Suhail and Cochrane (1998) quantified that 68% of monthly fluctuations in admissions for mania was accounted for by hours of daylight and hours of sunshine. In a North American study, Cassidy and Carroll (2002) reported that the greatest increases in hospital admissions for mania occurred in February and March – corresponding with the largest lengthenings of photoperiod (suggesting that the rate of change in photoperiod should be studied in addition to the change in photoperiod). However, several studies have failed to demonstrate any such associations (Lee et al., 2002; Volpe et al., 2010), albeit with the latter study undertaken in Brazil. However, in an earlier Brazilian study (Volpe et al. 2006) undertaken in the same subtropical mesothermic climate (and latitude some 20 degrees), admissions for mania correlated positively with (i) the average index and previous month’s hours of sunshine and (ii) differenced mean temperature.
2. Methods 2.1. Database Our data were obtained from the New South Wales (NSW) Department of Health and comprised 21,882 individuals hospitalised in NSW mental health units from 2000 to 2014 for ICD-10 diagnosed mania (i.e. the Department of Health records only ICD-10 diagnoses made on discharge), and which would therefore not have included ‘mixed states’. Admissions for each month were expressed as a rate per 100,000, based on the NSW population for the year, and adjusted to a 30-day month. Ethical approval was not required as the study analysed publicly available aggregated data. Photoperiod data were calculated using the R package geosphere (based on year, day of year, and latitude). These provided two measures of photoperiod: daily photoperiod, and the daily rate of change in photoperiod, each of which was converted to a mean for each month. For the state of NSW (the source of our admission data) the maximum difference in photoperiod between the northern (28S) and southern (37S) ends of the state is 50 minutes at the two solstices. The capital Sydney - where the bulk of the NSW population resides and contributes to the majority of psychiatric admissions has a minimum day length of 9 hrs 45 min and a maximum of 14 hrs 25 min. Mean monthly hours of sunshine for 2000–2014 were obtained from the Australian Bureau of Meteorology. 2.2. Analysis The primary statistical analysis was a set of vector autoregressive (VAR) models (Milhoj, 2016) fitted using the R package vars. Essentially, these model a time series (e.g. rate of admission) as a function of previous values (lags) of the time series as well as previous values of other time series in the model. If a time series lacks certain statistical properties it is said to be non-stationary, and it is often recommended in the literature that before fitting VAR models such series should be made stationary. This can be achieved by not using the observed values but rather using the difference between each value and the previous value (differencing). We used the Dickey-Fuller test (Milhoj, 2016) to examine if a series was stationary or not. We examined the relationship between rate of admissions and mean monthly values of (a) photoperiod, (b) the rate of change in photoperiod, and (c) hours of sunshine. For the 180-month time series (15 years x 12 months), we fitted sets of VAR models (see Table 1). These equations modelled the rate using (i) previous values of the rate of admissions (up to a lag of four months), (ii) previous values of the photoperiod variables up to a lag of four months, and (iii) a constant and linear trend. The rate of change in photoperiod was always superior to photoperiod (in terms of R2). As both rate and hours of sunshine were identified as non-stationary (and stationary after first differences) we also report results for differenced series.
3. Results In Figure 1, number of monthly admissions - expressed as standardised differences (residuals) between the observed and the expected (the latter being the average of all monthly admissions adjusted for population and to a 30-day month) numbers of admissions - are charted (in grey) for each month. The red (solid) line represents the photoperiod or day length (averaged across the 15 years), and the blue (dashed) line
represents the rate of change of the photoperiod across the year in Sydney. The photoperiod is at its maximum at the summer solstice (December 21 - 22). The days then shorten at an increasingly rapid rate until the autumn equinox (March 20 - 21) after which the photoperiod continues to decrease, but it does so more slowly. The photoperiod achieves its briefest duration at the winter solstice (June 20 - 21) when the rate of decrease in the photoperiod reaches its minimum. Thereafter the days start to become longer, lengthening at an increasingly rapid rate until the spring equinox (September 22 - 23), after which the photoperiod keeps lengthening (but at a decreasing rate) until the summer solstice - when the rate of increase in the photoperiod reaches its minimum and the days change from lengthening to shortening again. Turning to admissions, for all but January and February in the first half of the year, as the days become shorter, there were fewer than expected admissions for mania, and for all months in the second half of the year, as the days lengthen, there was a greater than expected number of admissions for mania. The switch from fewer than expected to greater than expected occurred abruptly in July (during winter) after the period with the least apparent shortening of the days, and just as the days started to lengthen. More specifically, the fewest admissions for mania occurred in April (following maximal shortening) and June (corresponding with the shortest length). However, while the photoperiod is still increasing in December, the peak in admissions occurred in November, rather than in December, arguing against a consistent association between oscillations in photoperiod and admission rates. In terms of seasonal pattern, the figure shows that the peak in admissions occurred in spring (September to November in NSW) and with the peak month for the year being the last month of spring. 3.1. Comparison of models Table 1 reports comparisons of the relative contributions of the variables based on the variance accounted for (R2) and, while the tabled results differ according to whether the series are differenced or not, the conclusions are consistent. In both cases, the addition of hours of sun (Model II) to a model with only the lagged rate (Model I which predicts current admission solely from previous admissions) produced just a small increase in R2 (3% and 5%), whereas the addition of rate of change in photoperiod (Model III) had a much greater increase in R2 (13% and 14%). The final model (Model IV) showed a small improvement over Model III (p = 0.07 in the undifferenced data, and p = 0.006 in the differenced data), indicating that both variables can be included usefully. We also report (in Table 2) the individual variable effects (via regression coefficients) in the final models, and now consider those data under relevant headings. 3.2. Rate of admission In the undifferenced series the current admission rate was positively associated only with the previous (lag 1) rate (p = 0.02), while in the differenced series all four lags were negatively associated and significant (p 0.01). 3.3. Rate of change in photoperiod (length of day) In both series, the rate of change was non-significant at lag 1 and significant at lags 2 - 4. Photoperiod has a cyclic nature which results in a particular pattern of
correlations between any given month and previous months. For example, values six months apart are near mirror images but negatively correlated close to 1, while values 3 months apart are correlated near to zero. Consequently, the coefficients for the multivariate models reported in Table 2 are not readily interpretable, with negative coefficients the opposite of what they would be if included singly. They do, however, locate change at a lag of 3 months as more prominent than other lags. 3.4 Hours of sunshine In the undifferenced series none of the lags were significant, though lags 1 and 4 were near significance. In the differenced series only lag 4 was significant (p = 0.016). In summary, admissions in the previous month, photoperiod length and magnitude of change in length accounted for about 20% of the variance (an R = 0.45) in rate of admissions. Length of photoperiod appeared relevant at up to four months previously, with rate of change and absolute change in photoperiod length most relevant three to four months previously. Hours of sunshine appeared a less salient variable.
4. Discussion 4.1 Principal findings Our study confirmed the common finding of a peak number of hospital admissions for mania occurring in spring. Perhaps the most distinctive finding was the abrupt change in admissions in winter (June to August in NSW), from fewer than expected admissions in June to a substantively greater than expected number in July and coinciding with the photoperiod changing from shortening to lengthening. Such a winter inflection point in admission rates for those with mania has been reported in several other photoperiod studies (Carney et al., 1989; Cassidy and Carroll, 2002; Myers and Davies, 1978), and this phenomenon may be as important as the spring excess emphasized in the literature. Changes quantified over the study period (2000-2014) are similar to those we documented in our earlier study involving 1,876 patients with a diagnosis of ‘manicdepressive psychosis, manic phase’ again admitted to a psychiatric facility in NSW over the 1971-1976 period, but did show some differences. For example, the earlier study quantified a peak in admissions in September and October (as against October and November in this study). Possible explanations may reflect change in diagnostic practices over the two study periods and the much larger sample in the current study, but also the impact of variations in total light exposure and, in particular, from nonsolar light sources. Bauer et al. (2017) have considered how changes in light exposure (such as LED lighting) may have influenced adaptability to a springtime circadian challenge. Thus, we need to concede that temporal changes in light exposure may explain differences across published studies. In focusing on any photoperiod effects, we examined (as detailed in Table 2) both the impact of variations in the photoperiod and the rate of change in the photoperiod, as well as examining for both more immediate and for lagged effects. We also examined for any impact of hours of sunshine but established a minimal contribution from that variable. Overall, photoperiod and rate of change in
photoperiod correlated similarly with admissions and with their overall contribution, along with the previous month’s admissions rate, accounting for some 20% of the variance in the level of admissions, indicative of a distinctive seasonal contribution. In the Introduction we noted that several studies had identified a lag between changes in photoperiod and admission patterns. Our study identified a number of lag impacts but with the most distinct effects associated with 3-4 month lags. Our findings suggest that, while there is an overall association between admissions for mania and the month of the year, the causal impact of photoperiod on admission rates may arise only at certain critical periods. The sudden and distinctive change from underrepresentation to over-representation of admissions corresponds with the photoperiod commencing to increase and suggests a ‘kick start’ or initial activation effect. The following attenuation in admissions might reflect initial photoperiod changes becoming less salient after the abrupt increase in July. The subsequent spring/summer peaking may arise from additional photoperiod mechanisms (e.g. when day length reaches a critical threshold), though it must be conceded that factors other than photoperiod might be in operation. The overall modest association between admission rates and both photoperiod rate of change and photoperiod change may reflect a diffusion of the initial ‘kick start’ effect, homeostatic mechanisms coming into play or multiple photoperiod mechanisms operating on differing groups of bipolar individuals at different times of the year and thus obscuring pristine delineation. Thus, to the extent that photoperiod changes impact on mania, rather than there being any single mechanism, there may be multiple photoperiod effects linked with differing bipolar genotypes. We now consider some possible mechanisms whereby environmental light is transduced by the nervous system into downstream effects on sleep and mood. Circadian rhythms in humans are primarily entrained by the ‘zeitgeber’ light. Entering through the eyes, light is detected by intrinsically photosensitive retinal ganglion cells (ipRGCs) which express the photopigment melanopsin (which has the greatest spectral sensitivity for blue light wavelengths). Via the retinohypothalamic tract, the ipRGCs signal daytime to the body’s ‘master clock’, the suprachiasmatic nucleus (SCN), which in turn inhibits pineal gland production of melatonin (Henriksen et al., 2016). Melatonin, which is produced from its precursor serotonin (5HT), is involved in numerous physiological processes including synchronizing the biological rhythms to the earth’s 24-hour day and has been shown to affect sleep patterns (Arendt, 2006; Wehr, 1991), and which Wehr et al. (1987) described as the ‘final common pathway’ in causing mania. In an earlier study undertaken in NSW we reported associations between photoperiod changes, as well as the rate of increase in bright sunshine (luminance), and increased hospital admissions for mania in spring (Parker and Walter, 1982). We hypothesized that the rapid increase in luminance in July and August stimulates the pineal gland and reduces melatonin secretion, with downstream effects on sleep and mood control so contributing to the seasonal pattern. There are, however, a number of other potential mechanisms linking photoperiod to the onset of manic episodes including the impact of altering circadian rhythms in individuals with a bipolar disorder. It has been postulated that bipolar individuals are vulnerable to a range of seasonal and climatic factors (Volpe et al., 2008), with supersensitivity to light a trait marker for bipolar disorder (Lewy et al., 1985; Nathan et al., 1999; Wright and Lack, 2001). In bipolar I individuals, a heritable excessive sensitivity of melatonin to light was reported, independent of illness state (Lewy et al, 1981; Hallam et al, 2006). Lee
et al. (2013) reported differing variants of the NF1A gene region between bipolar individuals which might play a role in seasonal mania – and they postulated the existence of a NF1A related sub-phenotype of bipolar disorder, with affected individuals experiencing seasonally-related manic episodes while those unaffected have manic episodes without a seasonal component. Sleep deprivation related to the increase in daylight hours as winter ends might also be linked to the induction of mania (Wehr et al., 1987). Additionally, Dulcis et al (2013) demonstrated that adult rats exposed to long active summer-like period lengths exhibited risk taking behaviours and elevated dopamine levels, with both mechanisms associated with mania. Levels of neurotransmitters in the brain, in particular serotonin (5HT), have also been reported as showing associations with variations in sunlight. In another Australian study, Lambert et al. (2002) reported that turnover of 5HT is lowest in winter and that the rate of its production in the human brain was directly related to the prevailing duration of bright sunlight, rapidly rising with increasing luminosity, while dopamine turnover did not show a seasonal pattern. Similarly, low extracellular 5HT has been found in autumn and winter, and high extracellular 5HT around the summer solstice (Pail et al., 2011). Thus it is possible that photoperiod (and sunlight) variations alter neurotransmitter concentrations that might play a role in triggering manic episodes. The effects of light treatment also implicate the possible association of photoperiod with manic episodes. The use of light therapy for bipolar depression and seasonal affective disorder is well described (Pail et al., 2011), while an intervention for mania, so-called ‘dark therapy’ may be salient. The therapy involves a reduction in light exposure, with a recent randomized control trial by Henriksen et al. (2016) showing that the use of blue light blocking glasses compared to placebo (in addition to treatment as usual) for bipolar mania was an effective treatment. Here the hypothesized mechanism is that the ipRGCs are blocked from being activated, creating a state of ‘virtual darkness’ in the brain and preserving melatonin production and with subsiding of the mania accruing from sleep and circadian rhythms being restored. 4.2. Study limitations A number of study limitations are noted. First, the data on admissions for mania were derived from a data set of state-wide health statistics and the validity of diagnoses of mania cannot be established. Second, we did not analyse data for any other – or all other – psychiatric conditions, so that we cannot reject the possibility of an unspecified effect of photoperiod on admissions of other diagnostic sub-sets or on admissions in general. Third, we examined data for all state hospitals and, while the great majority of admissions occurred in Sydney hospitals, a minority would have occurred in rural regions and where weather variables would have differed to some degree. Fourth, we cannot discount an impact from the move to daylight saving (when the clocks are moved forward an hour) which occurs in NSW on the first of October, although our Figure 1 data indicate that the swing to over-representation occurred earlier. Fifth, and as noted earlier, there is likely to be lag period between onset of a manic episode and being hospitalized as a consequence (and this may explain our finding of lagged effects). Sixth, data were only acquired for those admitted to hospital and where any seasonality impact may differ (e.g. be more severe) from the pattern for those whose episodes of mania do not result in hospital
admission. Seventh, we have described associations which do not allow causality to be claimed, especially when there may be a number of mediating factors. Eighth, while photoperiod changes over the year, it is largely constant from year to year, and so it can only really explain “constant” changes across a year and not fluctuations from year to year that might be influenced by other meteorological factors (e.g. temperature, ambient light) which change across the years and even social (e.g. the impact of holidays) and service (e.g. admission procedures) factors.
4.2. Implications In this study examining the impact of photoperiod changes on hospitalizations for mania, we quantified a modest level of variance accounted for across the whole year by our two study weather variables (rate of change in photoperiod and change in photoperiod), and which could argue for photoperiod having less impact than other meteorological variables. However, if photoperiod changes are causal, our findings allow an explanatory hypothesis. The greatest shift from fewer than expected to greater than expected admissions for mania occurred in July, when the days change from shortening to lengthening. We therefore hypothesize that there is a ‘kick start’ impact in July and with the subsequent peaking of admissions reflecting the ‘impact’ phenomenon attenuating or a second threshold photoperiod mechanism operating – if photoperiod changes are the key determining factor. We overview studies suggesting that changes in photoperiod potentially operate on certain bipolar individuals by altering circadian rhythms and inducing mania via such mechanisms as sleep disruption or changes in neurotransmitter levels. Author statement Role of funding The study was supported by NHMRC Program Grant (1037196).
Contributors GP designed the study, DHP undertook the statistical analyses, RG and AB undertook the literature review, GP and AB wrote first draft. All authors contributed to and have approved the final manuscript.
Conflicts of interest none
Acknowledgments This study was supported by an NHMRC Program Grant (1037196).
References Arendt, J., 2006. Melatonin and human rhythms. Chronobiol. Int. 23, 21–37.
Bauer, M., Glenn, T., Alda, M., Aleksandrovich, M.A., Andreassen, O.A., Angelopoulos, E., Ardau, R., Ayhan, Y., Baethge, C., Bharathram, S.R., Bauer, R., Bauna, B.T., Becerra-Palars, C., Bellivier, F., Belmaker, R. H., Berk, M., Bersudsky, Y., Bicakci, Ş., Birabwa-Oketcho, H., Bjella, T.D., Bossini, L., Cabrera, J., Cheung, E.Y.W., Del Zompo, M., Dodd, S., Donix, M., Etain, B., Fagiolini, A., Fountoulakis, K.N., Frye, M.A., Gonzalez-Pinto, A., Gottlieb, J.F., Grof, P., Harima, H., Henry, C., Isometsä, E.T., Janno, S., Kapczinski, F., Kardell, M., Khaldi, S., Kliwicki, S., König, B., Kot, T.L., Krogh, R., Kunz, M., Lafer, B., Landén, M., Larsen, E.R., Lewitzka, U., Licht, R.W., LopezJaremillo, C., MacQueen, G., Manchia, M., Marsh, W., MartinezCengotitabengoa, M., Melle, I., Meza-Urzúa, F., Yee Ming, M., Monteith, S., Morken, G., Mosca, E., Munoz, R., Mythri, S.V., Nacef, F., Nadella, R.K., Nery, F.G., Nielsen, R.E., O’Donovan, C., Omrani, A., Osher, Y., Østermark Sørensen, H., Quali, U., Pica Ruiz, Y., Pilhatsch, M., Pinna, M., da Ponte, F.D.R., Quiroz, D., Ramesar, R., Rasgon, N., Reddy, M.S., Reif, A., Ritter, P., Rybakowski, J.K., Sagduyu, K., Scippa, Â.M., Severus, E., Simhandl, C., Stein, D.J., Strejilevich, S., Subramaniam, M., Sulaiman, A.H., Suominen, K., Tagata, H., Tatebayashi, Y., Tondo, L., Torrent, C., Vaaler, A.E., Veeh, J., Vieta, E., Viswanath, B., Yoldi-Negrete, M., Zetin, M., Zgueb, Y., Whybrow, P.C., 2017. Solar insolation in springtime influences age of onset of bipolar I disorder. Acta Psychiatr. Scand. 1-12 Carney, P.A., Fitzgerald, C.T., Monaghan, C., 1989. Seasonal variation mania. In Thompson, C., Silverstone, T. (Eds). Seasonal Affective Disorder. CNS (Clinical Neuroscience) Publishers, London, UK. Cassidy, F., Carroll, B.J., 2002. Seasonal variation of mixed and pure episodes of bipolar disorder. J. Affect. Disord. 68, 25-31. Dulcis, D., Jamshidi, P., Leutgeb, S., Spitzer, N.C. 2013. Neurotransmitter switching in the adult brain regulates behavior. Science. 340, 449-453. Francis, A., Gasparo, P., 1994. Interval between symptom onset and hospitalization in mania. J. Affect. Disord. 31, 179-185. Geoffroy, P.A., Bellivier, F., Scott. J., Etain, B., 2014. Seasonality and bipolar disorder: a systematic review, from admission rates to seasonality of symptoms. J. Affect. Disord. 168, 210-223. Hallam, K.T., Olver, J.S., Chambers, V., Begg, D.P., McGrath, C., Norman, T.R, 2006. The hereditability and sensitivity to bright nocturnal light in twins. Psychoendocrinology, 31, 867-875. Henriksen, T.E.G., Skrede, S., Fasmer, O.B., Schoeyen, H., Leskauskaite, I., Bjørke‐ Bertheussen, J., et al., 2016. Blue‐blocking glasses as additive treatment for mania: a randomized placebo‐controlled trial. Bipolar Disord. 18, 221–232. Lambert, G.W., Reid, C., Kaye, D.M., Jennings, G.L., Esler, M.D., 2002. Effect of sunlight and season on serotonin turnover in the brain. Lancet 360, 1840–1842. Lee, H.J., Kim, L., Joe, S-H., Suh, K-Y., 2002. Effects of season and climate on the first manic episode of bipolar affective disorder in Korea. Psychiatry Res. 113, 151-159. Lee, H.J., Woo, H.G., Greenwood, T.A., Kripke, D.F., Kelsoe, J.R., 2013. A genomewide association study of seasonal pattern mania identifies NF1A as a possible susceptibility gene for bipolar disorder. J. Affect. Disord. 145, 200-207. Lewy, A.J., Nurnberger, J.I., Wehr, T.A., Pack, D., Becker, L.E., Powell, R., Newsome, D.A., 1985. Supersensitivity to light: possible trait marker for manicdepressive illness. Am. J. Psychiatry 142, 725–727.
Mawson, D.A., Smith, A.L., 1981. Relative humidity and manic admissions in the London area. Br. J. Psychiatry 138, 134-138. Milhoj, A. 2016. Multiple time series modelling using the SAS VARMAX procedure. Cary, NC: SAS Institute Inc. Myers, D.H., Davies, P., 1978. The seasonal incidence of mania and its relationship to climatic variables. Psychol. Med. 8, 433-440. Nathan, P.J., Burrows, G.D., Norman, T.R., 1999. Melatonin sensitivity to dim white light in affective disorders. Neuropsychopharmacology 21, 408–413. Pail, G., Huf, W., Pjrek, E., Winkler, D., Willeit, M., Praschak-Rieder, N., et al., 2011. Bright-light therapy in the treatment of mood disorders. Neuropsychobiology 64, 152–162. Parker, G., Walter, S., 1982. Seasonal variation in depressive disorders and suicidal deaths in New South Wales. Br. J. Psychiatry 140, 626-632. Parker, G.B., Graham, R.K., 2016. Seasonal variations in rates of hospitalisation for mania and hypomania in psychiatric hospitals in NSW. J. Affect. Disord. 191, 289-291. Sayer, H.K., Marshall, S., Mellsop, G.W., 1991. Mania and seasonality in the southern hemisphere. J. Affect. Disord. 23, 151-156. Silverstone, T.R., Romans-Clarkson, S., 1989. Bipolar affective disorder: causes and prevention of relapse. Br. J. Psychiatry 154, 321-335. Suhail, K., Cochrane, R., 1998. Seasonal variations in hospital admissions for affective disorders by gender and ethnicity. Soc. Psychiatry Psychiatr. Epidemiol. 33, 211-217. Taiz, L., Zeiger, E., 2006. Plant Physiology (4th Edition). Sinauer Associates, Inc., Sunderland MA. Volpe, F.M, Del Porto, J.A., 2006. Seasonality of admission for mania in a psychiatric hospital of Belo Horizonte, Brazil. J. Affect Disord. 94, 243-248. Volpe, F.M., Tavares, A., Del Porto, J.A., 2008. Seasonality of three dimensions of mania: psychosis, aggression and suicidality. J. Affect. Disord. 108, 95–100. Volpe, F.M., da Silva, E.M., dos Santos, T.N., de Freitas, D.E., 2010. Further evidence of seasonality of mania in the tropics. J. Affect. Disord. 124, 178-182. Wang, B., Chen, D. 2013. Evidence of seasonal mania: a review. J Psychiat Practice. 19, 301-308. Wehr, T.A., Sack, D.A., Rosenthal, N.E., 1987. Sleep reduction as a final common pathway in the genesis of mania. Am. J. Psychiatry 144, 201-204. Wehr, T.A., 1991. The durations of human melatonin secretion and sleep respond to changes in day length (photoperiod). J. Clin. Endocrinol. Metab. 73, 1276-1280. Winokur, G., 1976. Duration of illness prior to hospitalization (onset) in the affective disorders. Neuropsychobiology 2, 87-93. Wright, H.R., Lack, L.C., 2001. Effect of light wavelength on suppression and phase delay of the melatonin rhythm. Chronobiol. Int. 18, 801–808.
Sep
Oct
Nov
Dec
Jul
Aug
Sep
Oct
Nov
Dec
2.1
2.1
Aug
-1.6 -2.1
-1.6 -2.1
-1.1
-0.5
0
0.5
1.1
1.6
Admissions above/below expected Photoperiod Daily Change in Photoperiod (minutes/day) Solstice (Jun/Dec) & Equinox (Mar/Sep)
1.6
13.8
Jul
1.1
13.3
Jun
0.5
12.7
May
0
12.1
Apr
-0.5
11.5
Mar
-1.1
10.4 9.8
11
Feb
Admissions above/below expected (standardized residuals)
Photoperiod (hours) / Day-to-day Change in Photoperiod (minutes/day)
14.4
Jan
Jan
Feb
Mar
Apr
May
Jun
Time of Year
Figure 1. Change and rate of change in photoperiod for Sydney and admissions for mania in NSW. Table 1.Relative contribution of the variables based on the variance accounted for (Rsquared) in the series of VAR models fitted to undifferenced and differenced data. Model Variables in model Undifferenced Differenced1 I
Rate of admissions
R2 = 0.09
R2 = 0.26
II
Admissions + hours of sunshine
R2 = 0.12
R2 = 0.31
III
Admissions + change in photoperiod
R2 = 0.22
R2 = 0.39
IV
Admissions + hours of sunshine + change in photoperiod
R2 = 0.26
R2 = 0.45
1.Rate of change in photoperiod not differenced as it was stationary. Table 2. Final VAR models examining individual variable effects via regression coefficients for rate of admissions in undifferenced and differenced series Undifferenced series Differenced series Variable1
Lag
Est.
Rate of admissions
1
0.183 0.08
2.29 0.023
–0.08 0.08
– 1.02 0.309
2 3 4
SE
t
p
Est.
SE
t
P
–0.65 0.08
– 8.53
< 0.001
–0.61 0.09
– 7.00
< 0.001 < 0.001 0.012
0.10 0.08
1.29 0.198
–0.35 0.09
– 4.05
0.09 0.08
1.11 0.271
–0.19 0.08
– 2.54
Rate of change in photoperiod2
Hours of sunshine
1
6.30 3.50
1.80 0.074
5.56 3.71
1.50
0.136
2
– 13.46 6.50
– 2.07 0.040
– 16.47 6.77
– 2.43
0.016
3
16.82 6.29
2.68 0.008
23.35 6.78
3.45
0.007
4
–7.57 3.33
– 2.27 0.024
– 13.40 3.67
– 3.65
< 0.001
1
–0.03 0.01
– 1.88 0.062
–0.02 0.01
– 1.31
0.193
0.00 0.01
– 0.08 0.933
–0.01 0.02
– 0.82
0.413
0.00 0.01
– 0.02 0.984
0.00 0.02
– 0.20
0.845
0.03 0.01
1.91 0.058
0.03 0.01
2.43
0.016
2 3 4 2
R = 0.26 p = < 0.001
2
R = 0.45 p = < 0.001
1. Estimates for the constant and linear trend not reported. 2. Rate of change in photoperiod was stationary and not differenced. 3. Values in bold indicate statistically significant estimates. Highlights
Admissions for mania switched from being under- to over-represented when the photoperiod started to increase. Change and rate of change in photoperiod accounted for a distinctive 20% of the variance in admissions. The lack of a strong year-long correlation may reflect photoperiod changes being only a weak causal factor. Photoperiod may operate through a strong impact phase after the winter solstice and with the spring peaking of admissions reflecting secondary photoperiod or other influences.