Incidence of bipolar affective disorder per 100 000 person-years at risk according to age and sex on a Danish population-based cohort of 2.1 million people, where 2299 developed bipolar affective disorder during the 31.8 million person-years of follow-up. Incidence indicates the number of new cases that occurred per time period.
Mortensen PB, Pedersen CB, Melbye M, Mors O, Ewald H. Individual and Familial Risk Factors for Bipolar Affective Disorders in Denmark. Arch Gen Psychiatry. 2003;60(12):1209–1215. doi:10.1001/archpsyc.60.12.1209
Few population-based studies have addressed risk factors for bipolar affective disorder.
To study the possible association between bipolar affective disorder and history of mental illness in a parent or sibling; urbanicity of birth place; season of birth; sibship characteristics, including birth order; influenza epidemics during pregnancy; and early parental loss.
We used a population-based cohort of 2.1 million individuals based on data from the Danish Civil Registration System linked with the Danish Psychiatric Central Register.
Nationwide population-based sample of all individuals hospitalized or in outpatient clinic contact for the first time with bipolar affective disorder.
Overall, 2299 individuals were first diagnosed with bipolar affective disorder during the 31.8 million person-years of follow-up.
Risk of bipolar affective disorder was associated with a history of bipolar affective disorder as well as other psychiatric disorders, including schizophrenia and schizoaffective disorder, in parents or siblings. People with a first-degree relative with bipolar affective disorder had a 13.63-fold(95% confidence interval, 11.81-15.71) increased risk of bipolar affective disorder. No other consistent associations were found with the exception of an association between early parental loss, in particular maternal, and bipolar affective disorder. Children who experienced maternal loss before their fifth birthday had a 4.05 (95% confidence interval, 1.68-9.77) increased risk of bipolar affective disorder.
Early parental loss may represent both environmental and genetic risk factors for bipolar affective disorder. Most of the risk factors included in our study that previously have been associated with schizophrenia were not associated with bipolar affective disorder, supporting that the 2 disorders may be at least partially separate etiological entities.
IT IS well established that genetic factors are important in the etiology of bipolar affective disorder, and molecular genetic studies have identified a number of chromosomal regions of interest. However, there have been relatively few studies investigating possible nongenetic risk factors for bipolar affective disorder. It has been suggested that known or suspected risk factors for schizophrenia also may be of importance for other psychoses, but the empirical evidence regarding this is limited.1,2 The urbanicity of birth place has been shown to be related to schizophrenia risk, but few studies of this regarding bipolar affective disorder or other affective disorders exist. Winter or early spring birth has been associated with increased schizophrenia risk.3 This has also been found in some studies on bipolar affective disorder,4,5 although the findings are less uniform for bipolar affective disorder than for schizophrenia. One study suggested a possible link between influenza exposure in utero and subsequent risk of bipolar affective disorder,6 but another study found an inverse relationship.7 A number of studies have also found this for schizophrenia,8,9 but others have not.10 Some sibship or family characteristics, for example the birth order, age distance to older siblings, being from a large sibship, or a short age distance to other siblings, have been associated with schizophrenia risk,10 but to our knowledge, they have not been studied in relation to bipolar affective disorder. Finally, Agid et al11 found that early parental loss could be associated with risk for psychoses. Generally, there have been very few large-scale population-based studies of risk factors for bipolar affective disorder. The objective of this study was to determine if the findings regarding schizophrenia mentioned here would also apply in a study of bipolar affective disorder using a large Danish population-register–based sample.
Using data from the Danish Civil Registration system,12 we identified all individuals with known maternal identity who were born in Denmark between January 1, 1950, and December 31, 1983. The identity of the father was available for 96.8%. This study population was identical to the one used in a previous study of schizophrenia,13 except that in this study individuals were required to be alive at their 15th birthday(n = 2.1 million). Dates of death are available for all deaths in Denmark during the period from April 1, 1968, to December 31, 1998.
The study population and their mothers, fathers, and siblings were linked with the Danish Psychiatric Central Register.14 The Danish Psychiatric Central Register has been computerized since April 1, 1969. It contains data on all admissions to Danish psychiatric inpatient facilities and at present includes data on approximately 450 000 individuals and 1.6 million admissions. Information on outpatient visits to psychiatric departments has been included in the register since 1995. There are no private psychiatric departments in Denmark. From April 1969 to December 1993, diseases were classified according to the International Classification of Diseases, Eighth Revision (ICD-8),15 and from January 1994, International Classification of Diseases, 10th Revision (ICD-10).16
Overall, 2.1 million individuals were followed from their 15th birthday or April 1, 1970, whichever came later, until the date of onset of bipolar affective disorder, the date of death, the date of emigration, or December 31, 1998, whichever came first. Individuals diagnosed with schizoaffective disorder or schizophrenia were censored at the time of diagnosis. April 1, 1970, was chosen as the entry date, as we assumed the register to be nearly complete after the first year of operation. Cohort members were recorded as having bipolar affective disorder if they had been admitted to a psychiatric hospital or had been in outpatient care with a diagnosis of bipolar affective disorder. The date of onset was defined as the first day of the first contact leading to a diagnosis of bipolar affective disorder. In our previous study of schizophrenia based on the same study population,13 we started follow-up at the fifth birthday. However, owing to the very small number of cases registered with bipolar affective disorder prior to the 15th birthday, we chose to start follow-up at this age. Although this reduces the number of cohort members to approximately 600 000, contributing approximately 6 million person-years of observation, this reduces the statistical power of our study very little, because only 0.8% of all cases of bipolar affective disorder were excluded (19 cases excluded, 2299 cases included).
Parents and siblings were recorded hierarchically with a history of schizoaffective disorder, schizophrenia, bipolar affective disorder, recurrent depression, other affective illness, or other mental disorders if they had been admitted or had been in outpatient care with one of these diagnoses. For mothers, fathers, and siblings we only used the diagnosis with the highest hierarchy (Table 1). However, due to lack of power, the hierarchical diagnoses for siblings do not include recurrent depression. The hierarchy was not based on assumptions about genetic relatedness between the disorders, because this is not well documented at present. Rather it was based on our wish to create mutually exclusive categories, as well as the assumption that the use of the diagnoses of schizophrenia and schizoaffective disorder have been more conservative in Denmark than the use of the diagnoses of mania or bipolar affective disorder. We therefore placed schizophrenia and schizoaffective disorder highest in the hierarchy.
Degree of urbanization of birth place was grouped into 5 categories: capital (Copenhagen), capital suburb, provincial city with more than 100 000 inhabitants, provincial town with more than 10 000 inhabitants, and rural areas. This categorization is identical to the categorization used in previous studies showing urban-rural differences on schizophrenia risk.13,17
Using the complete person-identifiable information for all cohort members, we calculated their sibship size; number of older and younger siblings; and interval between nearest, older, and younger sibling at their 15th birthday. To ensure that this information was complete, these variables were only calculated for those having a mother born in Denmark later than April 1, 1935.18
History of maternal and paternal loss (defined as death of a parent) was classified according to the child's age at the loss, from birth to 5th birthday, 5th birthday to 10th birthday, 10th birthday to 15th birthday, or not before the 15th birthday. History of sibling loss was classified as before the 15th birthday or later than the 15th birthday. Data regarding place of birth, sibship characteristics, and parental loss were all available from the Civil Registration System.12
Monthly registrations of the number of mandatory reported cases of influenza in Denmark were obtained from the National Board of Health for the period from 1950 through 1979 and from the State Serum Institute from 1980 through 1983. The size of the Danish population per year was obtained from Statistics Denmark, Copenhagen. To investigate the possible association between bipolar affective disorder and the number of reported influenza cases per population during the month of birth and for each of the 9 months preceding the month of birth, exposure to influenza during a month was categorized as low, intermediate, or high, respectively, if the number of influenza notifications was less than 5, 5 to less than 10, or 10 or more per 1000 individuals in the population. This categorization was identical to that of a previous study showing no association between influenza prevalence during fetal life and the risk of schizophrenia.10
The relative risk of bipolar affective disorder was estimated by log-linear Poisson regression19 with the GENMOD procedure in SAS version 6.12 (SAS Institute Inc, Cary, NC).20 All relative risks were adjusted for age, sex, interaction between age and sex, calendar year of diagnosis, and age of the mother and father at the time of the person's birth. Age, calendar year of diagnosis, and history of mental illness in siblings were treated as time-dependent variables,21 whereas all other variables were treated as time-fixed variables. P values were based on likelihood ratio tests, and 95% confidence limits were calculated by the Wald test.21
Our study included 2299 incident cases of bipolar affective disorder in a population of 2.1 million, representing 31.8 million person-years of follow up. The term person-years means the total sum of the number of years that each individual in the study population has been under observation (ie, the time from their 15th birthday or April 1, 1970, whichever came later) until the date of a first diagnosis of bipolar affective disorder, the date of death, the date of emigration, or December 31, 1998, whichever came first. If a diagnosis of schizophrenia or schizoaffective disorder had been made prior to these dates, they were censored (ie, the observation of new cases, as well as the cumulation of person-years, was terminated at that time).
Figure 1 shows the crude incidence of bipolar affective disorder according to age and sex. In our study, the incidence of first admissions or outpatient clinic contacts with a diagnosis of bipolar affective disorder peaked in our oldest age group of 45 years or older, with rates of 13.3 per 100 000 person-years for men and 19.8 per 100 000 person-years for women, respectively. It should be noted, however, that our study sample was restricted to persons born in 1950 or later, so we were unable to compare our results with total population incidence rates from other studies that included a broader range of age groups. Table 2 presents the distribution of bipolar affective disorder cases and the crude incidence of bipolar affective disorder according to risk factors and confounders. Table 3 presents our main results. The first column shows only the univariate associations adjusted for age and its interaction with sex, parental age, and variables in the same category (history of mental illness, history of loss, or urbanicity), whereas the second column represents the full model adjusting further for history of mental illness in a parent or sibling and their interactions and history of loss of a parent or sibling.
The main finding was a positive association between a parental or sibling history of bipolar or other affective disorder and the risk of bipolar affective disorder. There was some difference between the different diagnostic categories of affective disorders with respect to the risk increase for bipolar affective disorder in the offspring. Bipolar affective disorder in parents or siblings was associated with higher risk than recurrent depression, which in turn increased risk more than other affective disorders or other mental disorders. Schizoaffective disorder in parents or siblings increased risk as much as bipolar affective disorder, whereas schizophrenia, excluding schizoaffective disorder, in parents or siblings increased the risk for bipolar affective disorder by a factor ranging from 3.7 to 5.7. However, if the hierarchy had not been applied, the association with schizoaffective disorders could be ascribed to the fact that 11 of 13 parents diagnosed with schizoaffective disorder had also been diagnosed with bipolar affective disorder.
Risk increased further when more than one parent or sibling was affected with mental illness. When interpreting our results, we found statistical evidence for interaction between mental illnesses in parents and siblings (Table 3). Therefore under our model, a person with a bipolar mother and a bipolar father has a 112.8-fold (12.20 × 14.91 × 0.62) increased risk of bipolar affective disorder compared with an individual with no history of parental mental illness. The term interaction in this context only refers to a characteristic of the statistical model and is not necessarily an indication of actual interaction between genes or other risk factors.
Neglecting the defined hierarchy of mental illness in a parent or sibling, we performed additional analyses and found that people with a history of bipolar affective disorder and not schizoaffective disorder or schizophrenia in a parent or sibling had a risk of 13.63 (95% confidence interval, 11.81-15.71) as compared with people without such a history. Furthermore, people with a history of schizophrenia or schizoaffective disorder and no history of bipolar disorder in a parent or sibling had a 4.17-fold (95% confidence interval, 3.24-5.36) increased risk of bipolar affective disorder as compared with people without such a history.
Conversely, there was no effect of season of birth, number of siblings, birth order, distance in age to other siblings, or rates of influenza during the in utero period. Data regarding these variables were analyzed in many different ways, including those used in our previous studies of schizophrenia10,17 (data not shown), but no trends or significant associations were found. Urbanicity of birth place was significantly associated with the risk of bipolar affective disorder, but only through a slightly increased rate among those born in provincial cities. One variable added to the analysis was the death of a parent or a sibling. This had a significant association with the risk for bipolar affective disorder, also after adjustment for history of mental illness in a parent or sibling. The effect was most marked if the parent died before the child was 5 years old and declined after that but was significant for all deaths before the child was 15 years old. It was not possible to subdivide by causes of death in parents and siblings.
As mentioned earlier, it is difficult to compare our results regarding incidence to findings from other studies because our sample only included persons younger than 49 years. However, the pattern of increasing first hospitalization rates with increasing age within this age span is similar to older findings from England,22 as well as more recent findings from Finland.23 We found higher rates among women. Community-based surveys have generally not found this difference (eg, Weissman et al24 and Kringlen et al25), and only a few studies of treated incidence bipolar affective disorder have reported higher rates in women (Helgason,26 Brewin et al,27 and Tsuchiya et al28). We do not know if our finding is due to sex differences in treatment-seeking, the age composition of the sample, or other factors.
The increased occurrence of bipolar affective disorder, schizoaffective disorder, and recurrent depression among parents and siblings is in accordance with previous family studies of probands with bipolar affective disorder.29- 32 The relative risks associated with a parental history of bipolar disorder, recurrent depression, and other affective disorders, as well as schizophrenia, did not differ significantly between mothers and fathers, respectively.
We applied a diagnostic hierarchy, which maximized the number of diagnoses with schizoaffective disorders among parents. However, considering the hierarchy and the negative results from most previous studies, it is somewhat surprising that we found an increased occurrence of schizophrenia among parents and siblings to cases. This may be due to misclassification of a severe manic episode as schizophrenia resulting in this diagnosis in our study. However, it is possible that our findings could also reflect some genetic overlap between schizophrenia and bipolar affective disorder,33- 35 because there was an association with a parental, especially maternal, history of schizophrenia even after excluding parents ever diagnosed with bipolar affective disorder.
The other general finding in our study was the lack of association between the risk of bipolar affective disorder and a number of risk factors that have been suggested for schizophrenia.
Although many studies have found schizophrenia to be linked to winter-spring birth,4,17 the absence of seasonality of birth effects for bipolar affective disorder is in line with our own recent negative findings for schizophrenia.13 Furthermore, the evidence regarding season of birth and bipolar affective disorder from the literature is conflicting,4 and further studies are needed.
The lack of effect of birth order is in line with the findings of Petterson.36 We find no effects at all, which probably exclude strong effects of birth order or sibship characteristics due to our large sample size. Although one study found a significant association between bipolar affective disorder and exposure to influenza epidemic during the second trimester,6 another study did not find this,37 and our study adds no positive support to a link between population levels of influenza during the gestational period and bipolar affective disorder.
Urbanicity of birth place was associated with the risk of bipolar affective disorder, but only through a slightly increased rate among those born in 1 of the 3 Danish provincial cities with more than 100 000 inhabitants, ie, Aarhus, Odense, and Aalborg. There was no dose-response relationship between risk and degree of urbanicity. It is in strong contrast with our findings regarding urbanicity and schizophrenia risk, however,13,17,38 and thereby adds some further credibility to those findings because the present study would be subject to many of the same possible biases that could be proposed to explain the findings for schizophrenia. In particular, our findings regarding bipolar affective disorder would not support the possibility that, for example, a lower nosocomial threshold in cities, perhaps due to better access to services, is the explanation because the same mechanism would apply to both diseases. We do not know the reason for the increased rates in provincial cities but speculate that slight differences in diagnostic traditions may have contributed to this pattern. Marcelis et al39 found a relatively weak but significant association between urban birth and the risk for a broader category of affective psychoses. Our failure to replicate this could be hypothesized to be due to the fact that we were able to adjust for the effect of history of mental illness in a parent or sibling. However, in our study there are no urban-rural differences even without this adjustment (Table 3). The explanation may simply be our more narrowly defined outcome (ie, bipolar affective disorder is less associated with urbanicity than manic-depressive illness, including unipolar depression, in general).
Regarding early parental loss, there are few studies specifically related to bipolar affective disorder. Agid et al11 found similar effects but could not separate the effects of maternal and paternal loss in bipolar affective disorder. However, in both major depression and to some extent schizophrenia, they had findings similar to ours (ie, a stronger effect of maternal loss, and a stronger effect if the parent was lost during early childhood). Others have not excluded the possible confounding effect of mental illness in the parents. In our study, where the estimates for early parental loss could be adjusted for history of mental illness in a parent or sibling, we still found a significant effect. Still, we cannot exclude that the effect may reflect mental disorders in the parents (eg, the cause of death could be suicide, which quite possibly would reflect an underlying affective disorder). Independent analyses in Denmark (L. V. Kessing, DrMedSc, unpublished data, 2003, and K. Tsuchiya, MD, unpublished data, 2003) may indeed suggest that this may be at least part of the explanation, but the finding of a greater effect of maternal loss, as well as the age-dependent effect, may suggest that the association cannot solely be ascribed to risk genes for mental disorders that are shared by the deceased parents and their children.
Our main finding was a strong association between the risk of bipolar affective disorder and a history of bipolar affective disorder, as well as other psychiatric disorders, including schizophrenia and schizoaffective disorder, in parents or siblings. The association between death of a parent, in particular during early childhood, and bipolar affective disorder may be an indicator of nongenetic as well as genetic risk factors. Our study suggests that some factors that have been associated with schizophrenia, including urbanicity of birth place, season of birth, influenza during pregnancy, and sibship characteristics, either do not contribute to the risk for bipolar affective disorder or show a distinctly different pattern of association. Of course, our study does not exclude that some risk factors may be shared between schizophrenia and bipolar affective disorder. For example, head injury has been associated with the risk of onset of both disorders,40- 42 and also some obstetric factors have been associated with both schizophrenia and affective disorders (Hultman et al43). Also our data would be compatible with some genetic overlap between the two disorders. However, our findings do provide some epidemiological evidence that schizophrenia and bipolar affective disorder are at least partially separate etiological entities.
Corresponding author: Preben Bo Mortensen, DrMedSc, National Centre for Register-Based Research, Aarhus University, Taasingegade 1, DK-8000 Aarhus C, Denmark (e-mail: firstname.lastname@example.org).
Submitted for publication October 14, 2002; final revision received April 8, 2003; accepted April 11, 2003.
This study was supported by the Stanley Medical Research Institute, Bethesda, Md. The National Centre for Register-Based Research is financially supported by the Danish National Research Foundation, Copenhagen, Denmark.