Research report
Seasonal affective disorder and latitude: a review of the literature

https://doi.org/10.1016/S0165-0327(98)00097-4Get rights and content

Abstract

Background: The aim of the study is to investigate the relationship between the prevalence of SAD and latitude. Methods: An overview of the epidemiological literature on the prevalence of SAD is given and studies relevant for the latitudinal dependency of prevalence will be analyzed and discussed. Results: The mean prevalence of SAD is two times higher in North America compared to Europe. Over all prevalence studies, the correlation between prevalence and latitude was not significant. A significant positive correlation was found between prevalence and latitude in North America. For Europe there was a trend in the same direction. Conclusions: The influence of latitude on prevalence seems to be small and other factors like climate, genetic vulnerability and social–cultural context can be expected to play a more important role. Additional controlled studies taking these factors into account are necessary to identify their influence.

Introduction

By definition, seasonal affective disorder (SAD) is associated with (the changing of) the seasons (Rosenthal et al., 1984). A major hypothesis is that SAD is triggered by photoperiod variation. Since photoperiod variation over the seasons is larger closer to the poles, it is hypothesized that with an increase in latitude there is an increase in the prevalence of SAD. This association has been suggested by the outcome of two studies in the USA and Norway (Lingjaerde et al., 1986Potkin et al., 1986). Although both studies have their methodological limitations, they have stimulated more research into the relationship between prevalence and latitude. Since then, the prevalence of SAD has been explored in a number of studies, mainly in North America and Europe.

In the present report the importance of the results of these studies for the latitude–prevalence issue will be discussed. First, studies will be discussed that either investigated both prevalence and latitude directly or studies that did not investigate prevalence per se, but within specific sub-populations the relationship of SAD or depressive symptoms with latitude. Second, analysis will be performed on prevalence studies at different latitudes. Since the legitimacy of comparing figures from different studies depends on the comparability of the applied methodology and assessment instruments, this aspect will be given some consideration. For the sake of conciseness the characteristics of the studies are not extensively discussed but are summarized in tables for the different continents.

In the Potkin et al. (1986)and the Lingjaerde et al. (1986)studies a questionnaire consisting of 15 symptoms of SAD was published in nationwide newspapers in the USA and Norway with the request to return the questionnaire if eight or more symptoms were present (caseness SAD). Both studies showed a positive correlation between latitude and prevalence. This result is merely indicative. The characteristics of the readers of the newspaper are unknown and may have been different in different parts of the country. Also, it is unknown how many of the subjects who met the criteria of SAD (eight symptoms or more) returned their questionnaire; these percentages may have differed across the country due to cultural differences. Finally, since the questionnaire used in these studies was not validated, the lack of direct clinical assessment is more important in these studies than in those using a validated instrument.

Rosen et al. (1990)studied the prevalence at four locations in the USA (see Table 1, Table 2, Table 3 for a description of the different studies), using the same methods.

Positive correlations were found between the prevalence of winter-SAD, and winter-SAD and subsyndromal-SAD (S-SAD; see methodological issues) combined, and latitude. Recently, Levitt and Boyle (1997)studied the prevalence of SAD in the Province of Ontario, Canada, across eight strata of 1° latitude (from 42°N to 50°N). They found no association between latitude and prevalence. In a study in Italy (Muscettola et al., 1995) the prevalence rates at three of the five locations at which the study was performed were compared: Catanzaro (39°N), Napoli (41°N), and Trieste (46°N). The correlation was in the unexpected direction (r=−0.50). The results in the latter study may be biased, however, by the low response rates (overall 13.6%). Nevertheless, as the authors claim, comparison between the three locations in the study may have some value since the response rates were approximately the same.

In Japan (Sakamoto et al., 1993) the prevalence of SAD was assessed among patients with a mood disorder who contacted 53 outpatient university psychiatric clinics in Japan for the first time between 1 September 1990 and 31 March 1991. The clinics were located between 26°N and 44°N. The results show a nonsignificant correlation between prevalence and latitude (Spearman's r=0.33, P<0.10). Okawa et al. (1996)studied seasonal variation in six cities in Japan at latitudes ranging from 32°N to 43°N. The global seasonality score showed a significant correlation with latitude. Partonen et al. (1993)assessed the frequency of depressive symptoms among 1000 subjects (801 women, 199 men), all employees of a nationwide bank in Finland. The SIGH-SAD self-rating scale was returned by 486 subjects, living between 60°N and 70°N. The results showed that depressive symptoms were not more common at higher latitudes than at lower latitudes. In a study among the winter-over personnel of three antarctic stations (from 64°S to 90°S), Palinkas et al. (1996)did not find an association between seasonality and latitude. The sample was too small (n=87) and the stay on Antarctica was too short to draw conclusions on prevalence. Finally, the prevalence of seasonal symptoms in relation to latitude was studied in children (age range, 9–12 years) by Carskadon and Acebo (1993). Questionnaires were sent to teachers of 78 schools across the USA, who asked the parents of the children to complete them. Six questions concerning seasonal symptoms were taken from the SPAQ and adapted for this occasion. The schools were located in three geographic zones: a northern zone (>42°N), a central zone (between 36°N and 42°N) and a southern zone (<36°N). The results showed a significant higher incidence of seasonal symptoms in the winter in the northern and central zones versus the southern zone.

The conclusion from this overview is that, although confirmed by some studies, the evidence in favor of a latitude–prevalence hypothesis is not conclusive.

Another approach is to compare the results of the different prevalence studies in relation to the latitudes at which they were performed. For a valid comparison it is necessary that the studies share a comparable and sound methodology. Therefore, some methodological issues concerning the prevalence studies will be discussed first.

In all studies presented in Table 1, Table 2, Table 3 (except the Potkin et al. (1986), the Lingjaerde et al. (1986)and the Partonen et al. (1993)studies) the same assessment instrument was used: the Seasonal Pattern Assessment Questionnaire (SPAQ; Rosenthal et al., 1987). This facilitates comparison of the results of the different studies. The criteria for SAD on the SPAQ have been formulated in the Kasper et al. (1989a)study and are based on data from 168 SAD patients (Kasper et al., 1989a, Hardin et al., 1991). The SPAQ applies three criteria for SAD, which are presented in Table 1 in the study by Kasper et al., 1989a). The first is based on the Global Seasonality (GS) scale, providing a composite measure for change of mood, social activities, appetite, sleep, weight and energy across the seasons. Item scales range from (0) `no change' to (4) `extremely marked change'. Thus, the total scale ranges from 0 to 24. The suggested cut-off score for caseness on this criterion is 10 for (telephone) interviews and 11 for the paper and pencil method.

A second criterion for SAD is based on one question, i.e. whether seasonal changes are considered a problem. The response possibilities are 0=no problem, 1=a mild problem, 2=a moderate problem, 3=a marked problem, 4=a severe problem, and 5=a disabling problem. A score of at least 2 is necessary to reach the SAD threshold.

The final criterion is the `window', i.e. the time interval within which the problems should recur. The timing of the problems is determined by asking what months subjects feel worst. The width of the window varies across studies (see Table 1, Table 2, Table 3), and may thus be a confounding factor.

Subsyndromal-SAD (S-SAD) (Kasper et al., 1989b) is defined as a cluster of seasonal complaints, which are not severe enough to allow for a diagnosis of SAD. The prevalence of S-SAD will not be discussed since there is confusion about the second set of the criteria: a GS score of 8 or 9 (9 or 10 for the self-report method) “and seasonal changes are either a problem or not” (Kasper et al., 1989a; p. 829). In some studies this criterion was not followed (Terman, 1988, Rosen et al., 1990, Booker and Hellekson, 1992, Hagfors et al., 1992, Muscettola et al., 1995) and changed into at least mild problems with the changes of season. In view of the way the criteria are formulated, this latter definition would indeed make more sense. It is clear that prevalence rates for S-SAD are influenced by the use of different criteria.

Different sampling methods are employed. The most reliable method is the drawing of a random sample of the general population from community registers (Magnusson and Stefansson, 1993, Mersch et al., 1995, Mersch et al., 1998). In this case systematic sampling error is avoided. A second method is the random selection of subjects from the telephone directory (Terman, 1988Rosen et al., 1990Muscettola et al., 1995). In this case there is a risk of systematic error, because people without a telephone are excluded from the sample. Moreover, some people with a telephone are not listed in the directory. Kasper et al. (1989a)employed an elegant method to reduce the latter chance of error by random number dialling, a method also used by Levitt and Boyle (1997).

A method particularly sensitive to sampling bias is the study of subgroups of the population (Ito et al., 1992, Magnusson and Axelsson, 1993, Partonen et al., 1993, Ozaki et al., 1995, Eagles et al., 1996, Hedge and Woodson, 1996, Madden et al., 1996). Especially the selection of subjects from companies is questionable, since it is not likely that organisations select their employees randomly. Furthermore, bias as a consequence of the illness may influence the results. In a study in The Netherlands, respondents who met the criteria of SAD were significantly more often unemployed or on sick leave (Mersch et al., 1998). Moreover, concern with the protection of their privacy towards the employers may bias the ratings of company employees. For these reasons the seven above-mentioned studies are left out of the comparison. Since the study on children by Swedo et al. (1995)is obviously not a representative sample of the population this study is also left out of the analysis. Finally, the earlier discussed Italian study (Muscettola et al., 1995) is left out of the comparisons because of the extremely low response rate.

Also the survey methods differ. Most often (Terman, 1988, Rosen et al., 1990, Magnusson and Stefansson, 1993, Hagfors et al., 1995, Mersch et al., 1995, Muscettola et al., 1995) a questionnaire was mailed to the research population, while in some cases subjects were interviewed by telephone (Kasper et al., 1989a, Hagfors et al., 1992, Wirz-Justice et al., 1992, Levitt and Boyle, 1997). In one study (Booker and Hellekson, 1992) it is unclear whether subjects were interviewed at home or by telephone. Kasper et al. (1989a)preferred telephone interviews to mail out procedures because of the higher response rate. According to these authors the literature on survey methodology provides evidence that both methods are equally valid. Therefore, studies using either one of these survey methods will be included in the analysis.

Response rates may influence prevalence figures. For instants, it is possible that the probability of responding to the questionnaire may partly be a function of the presence of SAD symptoms. To test this an analysis of variance was performed on the selected studies showing no significant interaction (F(1,10)=1.44, P=0.58). Also, there was no influence of latitude on response rate (F(1,10)=1.43, P=0.48).

Correlations between prevalence and latitude (Spearman rs; significancies are one-tailed) were calculated on North American (Terman, 1988, Kasper et al., 1989a, Rosen et al., 1990, Booker and Hellekson, 1992, Levitt and Boyle, 1997) and European studies (Hagfors et al., 1992, Hagfors et al., 1995, Wirz-Justice et al., 1992, Magnusson and Stefansson, 1993, Mersch et al., 1995). Some studies were performed over an area that included a range of latitudes (Hagfors et al., 1992, Hagfors et al., 1995, Wirz-Justice et al., 1992, Magnusson and Stefansson, 1993, Levitt and Boyle, 1997). The prevalence rates for these countries are averaged and plotted at the mean latitude. If available, figures for winter SAD were used. In studies that did not report separate values for winter and summer SAD (Booker and Hellekson, 1992, Hagfors et al., 1992, Hagfors et al., 1995, Wirz-Justice et al., 1992, Levitt and Boyle, 1997), the prevalence rates include both patterns. Because the rates for summer SAD are extremely low in the studies that did report these figures (see Table 1, Table 2, Table 3), it is not likely that this procedure has influenced the results substantially. In Fig. 1 the prevalence rates are plotted as a function of latitude. The relationship is shown by linear curve-fits.

The correlation shows a very weak insignificant positive relationship between prevalence and latitude: r(n=13)=0.07, P=0.415. Visual inspection shows, however, that this low correlation can be explained for a large part by the difference between the North American and the European prevalence figures (MNA=6.24, S.D.=3.06 and MEur=3.90, S.D.=1.69). This difference is significant as shown by an analysis of variance with latitude as covariate (F(1,10)=20.33, P=0.001). Correlations between prevalence and latitude for the North American and European data, separately, are r(n=7)=0.90, P=0.003 and r(n=6)=0.70, P=0.061, respectively.

Section snippets

Discussion

The results of the correlation between latitude and prevalence are puzzling. The small overall correlation coefficient does not support the hypothesis of an existing relationship between latitude and prevalence. The correlation coefficient for the North American studies, however, is highly significant, while the coefficient for the Europe studies shows a trend in the same direction. The conclusion is that if latitude influences prevalence, this influence is only weak. Apparently, other factors

Conclusions

The conclusion from the overview must be that the evidence for a positive correlation between prevalence and latitude is still unclear. It seems safe to conclude that if such a relationship exists, its impact on prevalence is smaller than (the combined effect of) a number of other factors of which climatological, social and cultural influences and genetic factors are most prominent. Or, as Rosen et al. (1990)concluded in the most supportive study for the latitude hypothesis: “…only a small

References (55)

  • L.N. Rosen et al.

    Prevalence of seasonal affective disorder at four latitudes

    Psychiatry Res.

    (1990)
  • D.R. Schuller et al.

    A comparison of personality characteristics of seasonal and nonseasonal major depression

    Comp. Psychiatry

    (1993)
  • J.M. Allen et al.

    Depressive symptoms and family history in seasonal and nonseasonal mood disorders

    Am. J. Psychiatry

    (1993)
  • American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, 4th ed. American...
  • D.G. Blazer et al.

    The prevalence and distribution of major depression in a national community sample: The National Comorbidity Survey

    Am. J. Psychiatry

    (1994)
  • J.M. Booker et al.

    Prevalence of seasonal affective disorder in Alaska

    Am. J. Psychiatry

    (1992)
  • Bouhuys, A.L., Meesters, Y., Jansen, J.H.C., Bloem, G.M., 1997. Personality in remitted seasonal affective disorder...
  • M.J. Garvey et al.

    Does cold winter weather produce depressive symptoms?

    Int. J. Biometeorol.

    (1988)
  • C. Hagfors

    `Light sensitivity': real or expressiveness subscale on the SPAQ?

    Light Treat. Biol. Rhythms

    (1993)
  • C. Hagfors et al.

    Seasonal affective disorder (SAD) in Finland: an epidemiological study

    Soc. Light Treat. Biol. Rhythms

    (1992)
  • C. Hagfors et al.

    Seasonality in Finland and Sweden, an epidemiologic study, preliminary results

    Soc. Light Treat. Biol. Rhythms

    (1995)
  • A.L. Hedge et al.

    Prevalence of seasonal changes in mood and behaviour during the winter months in central Texas

    Psychiatry Res.

    (1996)
  • A. Ito et al.

    Prevalence of seasonal mood changes in low latitude area: seasonal pattern assessment questionnaire score of Quezon city workers

    Jpn. J. Psychiatry Neurol.

    (1992)
  • K.L. Jang et al.

    The relationship between seasonal mood change and personality: more apparent then real?

    Acta Psychiatr. Scand.

    (1997)
  • S. Kasper et al.

    Phototherapy in individuals with and without subsyndromal seasonal affective disorder

    Arch. Gen. Psychiatry

    (1989)
  • S. Kasper et al.

    Epidemiological changes in mood and behavior. A telephone survey of Montgomery County

    Md Arch. Gen. Psychiatry

    (1989)
  • A.J. Levitt et al.

    Latitude and the variation in seasonal depression and Seasonality of depressive symptoms

    Soc. Light Treat. Biol. Rhythms

    (1997)
  • Cited by (174)

    • Tryptophan and melatonin as treatments

      2023, Encyclopedia of Sleep and Circadian Rhythms: Volume 1-6, Second Edition
    • Who is more prone to depression at higher latitudes? Islanders or mainlanders?

      2021, Current Research in Ecological and Social Psychology
    • Light therapy for mood disorders

      2021, Handbook of Clinical Neurology
    • Music sentiment and stock returns

      2020, Economics Letters
    View all citing articles on Scopus
    View full text