Abstract
Bipolar disorder is a multifactorial illness with uncertain aetiology. Knowledge of potential risk factors enables clinicians to identify patients who are more likely to develop bipolar disorder, which directs further investigation, follow up and caution when prescribing. Ideally, identifying directly causative factors for bipolar disorder would enable intervention on an individual or population level to prevent the development of the illness, and improve outcomes through earlier treatment. This article reviews the epidemiology of bipolar disorder, along with putative demographic, genetic and environmental risk factors, while assessing the strength of these associations and to what extent they might be said to be ‘causative’. While numerous genetic and environmental risk factors have been identified, the attributable risk of individual factors is often small, and most are not specific to bipolar disorder but are associated with several mental illnesses. Therefore, while some genetic and environmental factors have strong evidence supporting their association with bipolar disorder, fewer have sufficient evidence to establish causality. There is increasing interest in the role of specific gene–environment interactions, as well as the mechanisms by which risk factors interact to lead to bipolar disorder.
Introduction
Bipolar affective disorder (bipolar) is a multicomponent illness involving episodes of severe mood disturbance, neuropsychological deficits, immunological and physiological changes, and disturbances in functioning. 1 It is one of the leading causes of disability worldwide 2 and is associated with high rates of premature mortality from both suicide and medical comorbidities.3,4
The aetiology of bipolar is not well understood and research into the disorder lags behind disorders such as psychosis. However, the last decade has seen an expanding evidence into the genetics of the disorder, underlying developmental pathways, risks and vulnerability factors, gene–environment interactions and the putative features of the bipolar prodrome.
This article summarizes the research into demographic, genetic and environmental risk factors for the development of bipolar, with a focus on recent updates and the role of environmental triggers. To identify relevant literature, searches were conducted in PubMed and PsycINFO using the terms ‘Bipolar Disorder’, combined with ‘risk factors’ or ‘epidemiology’. Results were reviewed with a focus on the most recent evidence and systematic reviews or large prospective studies, and further individual searches were then expanded for each risk factor category identified. A summary of the included studies relating to specific risk factors for bipolar are included in Table 1.
Studies investigating specific risk factors for bipolar disorder.
ADHD, attention deficit hyperactivity disorder; BDNF, brain-derived neurotrophic factor; bipolar, bipolar disorder; CI, confidence interval; COMT, Catechol-O-methyltransferase; 5-HTTL, serotonin system gene; 5-HTTLPR, serotonin-transporter-linked polymorphic region; GWAS, genome-wide association study; HR, hazard ratio; MDD, major depressive disorder; mRNA, messenger ribonucleic acid; OR, odds ratio; SNP, single nucleotide polymorphism; TLR2, toll-like receptor 2.
Epidemiology of bipolar disorder
Epidemiological studies have suggested a lifetime prevalence of around 1% for bipolar type I in the general population.54,55 A large cross-sectional survey of 11 countries found the overall lifetime prevalence of bipolar spectrum disorders was 2.4%, with a prevalence of 0.6% for bipolar type I and 0.4% for bipolar type II. 56 Although findings varied across different countries, this suggested a lower prevalence of bipolar type I and II than previous studies,55,57 while the prevalence of bipolar type I in USA was found to be 1%, slightly higher than the other countries. It is unclear whether differences were due to more stringent diagnostic criteria used in this study, or true differences in rates of bipolar across countries and ethnic groups. In one of the very few epidemiological investigations in England, the recent Adult Psychiatric Morbidity Survey 2014 found lifetime prevalence of likely bipolar was 2%. The measurement method suggests that this was an underestimate, but the study did not distinguish bipolar subtypes. 58 A recent meta-analysis of 25 studies found a pooled lifetime prevalence of 1.06% and 1.57% for bipolar type I and II, respectively, although the majority of the included studies were from North or South America. 59 Nevertheless, a similar prevalence has been found in the UK, Germany and Italy, 60 and a lifetime prevalence between 0.1–1.83% was found in a systematic review of studies from African countries. 61
The reason for international variations in the prevalence of bipolar is not entirely clear, and ethnicity, 49 cultural factors 62 and variations in diagnostic criteria and study methodology 59 may each have an impact. The evidence for differing rates of bipolar in different ethnicities is conflicting, with some studies showing higher rates in Caucasians63,64 and others in nonwhite populations. 65 A systematic review found no clear evidence for differences across ethnic groups, and suggested individual study differences may be related to cultural factors, migration and higher rates of misdiagnosis of black ethnic groups as having schizophrenia rather than bipolar. 49 With regards to sex, several studies report equal distribution in bipolar, 49 while others have identified a higher prevalence of manic episodes and bipolar type I in males and higher rates of bipolar type II in females. 56 Overall, the evidence is not sufficiently strong to deviate from the view that bipolar appears to have a roughly equal distribution across sex and ethnicity.
The mean age of onset for bipolar appears to be in the early twenties, 56 although findings vary between 20–30 years. 55 A bimodal distribution of the incidence of bipolar has been suggested, 66 supported by a large population-based cohort study, which found two peaks in age of onset at 15–24 years and at 45–54 years. 67 However, age of onset estimates are very difficult to define accurately for bipolar, given the long periods of untreated illness, when symptoms can be nascent or apparent without individuals accessing services, which is often used as the measure of onset in many studies. 68 Moreover, there appear to be differences in the presentation and clinical course of bipolar depending on age of onset, 69 with higher rates of psychiatric and medical comorbidities such as suicidality and vascular disease in later-onset mania. 70
A number of studies have investigated rates of bipolar according to sociodemographic variables, with generally inconsistent findings. 49 There is some evidence of higher rates in low income, unemployed and unmarried groups, 49 although the social disruption caused by severe mental illness giving rise to such associations cannot be ruled out. 54 Conversely, an interesting finding among some studies is that higher socioeconomic status and higher occupational level, as well as creativity,54,71 are associated with increased risk of bipolar,72,73 which is opposite to that of unipolar depression and schizophrenia. 54 However, these studies are limited by small sample sizes and a lack of replication. 74 Explanations for this association include the possibility of referral bias for those with higher socioeconomic status, while some have suggested that those with high-functioning creative traits may confer a genetic risk of bipolar. 54
There is also emerging evidence for an association between urban environments and increased rates of bipolar. 49 While the evidence is stronger for schizophrenia, where there have been multiple suggested explanations, 75 the reason for the association between urbanization and bipolar is less clear. However, a cohort study found that there was a strong association between urban residence and the incidence of psychotic bipolar, but no association for bipolar without psychosis. 76 This may suggest that urban residence is a transdiagnostic risk factor for psychotic illness rather than bipolar per se.
Genetics and gene environment interactions
The contribution of genetic factors to bipolar has long been identified, with evidence from twin studies suggesting monozygotic concordance of between 40–70%, and lifetime risk in first-degree relatives is 5–10%; around seven times higher than the general population risk. 5 However, relatives of patients with bipolar are more likely to develop unipolar depression than bipolar themselves, suggesting the genetic risk transcends diagnostic categories. 5 There is also evidence of shared genetic risk between bipolar, schizophrenia and autism.77,78 Nonetheless, bipolar clearly does not follow a Mendelian pattern of inheritance, and linkage studies have not identified individual genes with a strong association with the disorder. 79 The genetic risk for bipolar in part is likely due to multiple single nucleotide polymorphisms, which are highly prevalent in the general population and confer a very small increased risk individually. 80 Technological advances have allowed for genome-wide association studies that have pooled data and identified multiple genetic loci associated with bipolar patients, suggesting aggregated polygenic risk. 6
Whilst many important genetic loci have been identified, how these translate to risk of illness is a second frontier of discovery. Studies have identified polymorphisms in genes coding for brain-derived neurotrophic factor (BDNF) to be associated with bipolar. 7 BDNF is suspected to be involved in the pathogenesis of bipolar as well as being a potential biomarker of disease activity. 81 Associations with catechol-O-methyl transferase (COMT) and monoamine transporters have also been observed.8,82 Genes for voltage-gated calcium channel subunits such as CACNA1C are located near to single nucleotide polymorphisms that have an association with bipolar, as well as proteins involved in cell signalling such as ODZ4, 6 and genes encoding for gamma-aminobutyric acid (GABA) receptor subunits. 83 The fact that many of the medications used as prophylactic agents in bipolar act on calcium channels or GABA receptors 84 suggests these proteins may be involved in the neurobiology of the disorder, and this evidence is guiding the search for new therapeutic targets. 85
However, it is clear that the effect size of each single nucleotide polymorphism is very small. For example, the odds of having bipolar in those with the polymorphism around CACNA1C is 1.14, and the majority of those with this polymorphism do not go on to develop the disorder.6,80 There has therefore been increasing interest in the role of how gene–environment interactions contribute to the onset of bipolar, although this remains an under-researched area, compared with schizophrenia.86,87 Nevertheless, interaction between childhood abuse and BDNF gene polymorphisms have been shown in several studies,9,86 while toll-like receptor 2 polymorphisms may interact with stressful life events and Toxoplasma gondii infection to increase the risk of bipolar.10,11 A COMT polymorphism has been found to interact with stressful life events for bipolar depressive episodes, 12 while serotonin transporter genes have interactions with cannabis use on the presence of psychotic symptoms in bipolar. 13 With the increasing ability of genome-wide association studies to identify polymorphisms conferring a very small increased risk, further study of how these genes interact with environmental factors to trigger bipolar is required.
Environmental risk factors
Prenatal and perinatal factors
Prenatal viral infections have been implicated in a number of mental illnesses, including bipolar.88 –90 A recent review by Barichello and colleagues 14 investigated associations between bipolar and 10 infectious agents. Findings between studies were generally inconsistent, and no association was found for Epstein-Barr virus, human herpesvirus 6 or varicella zoster virus. Five of the eleven studies investigating cytomegalovirus found an association between antibody levels and bipolar, while two studies found an association between maternal influenza infection and bipolar with psychosis,91,92 although other studies found no association.93 –95 None of these studies were prospective or longitudinal and it is uncertain whether these infections occurred during pregnancy or subsequently. Therefore, the evidence for maternal viral infection as a risk factor for bipolar remains weak, overall.
However, there is stronger evidence for an association between bipolar and seropositivity for T. gondii infection, demonstrated in two recent meta-analyses.15,16 The first included 11 studies and demonstrated overall increased odds of having bipolar in those with immunoglobulin G (IgG) to T. gondii, with an odds ratio of 1.52 (95% confidence interval 1.06–2.18). 15 A second meta-analysis of eight studies also found a significant association between bipolar and T. gondii seropositivity, with an odds ratio of 1.26 (95% confidence interval 1.08–1.47). 16 However, the included studies were not prospective and it remains uncertain when T. gondii exposure occurred. Notwithstanding, there is preclinical evidence suggestive of a relationship between T. gondii and development of mental illness, with studies showing behavioural changes in mice 96 and humans.97,98 Moreover, there is evidence that infection with T. gondii causes changes in dopamine metabolism leading to increased dopamine production, 99 similar to that suggested as a potential mechanism for manic episodes in bipolar. 100 Furthermore, there is evidence that following T. gondii infection, the local inflammatory response leads to alteration in cytokines, 101 such as IL-6, 102 which have been implicated in mental illness and bipolar specifically, 103 and may be related to cognitive deterioration in this patient group. 102
Evidence regarding other prenatal exposures such as maternal smoking and severe psychological stressors are inconsistent, with only a small number of studies investigating these factors. 50 Obstetric complications have generated interest as a risk factor for later development of bipolar, 104 but a meta-analysis found no significant evidence for this association, 17 and bipolar patients were less likely to have experienced obstetric complications than those with schizophrenia. A systematic review by Marangoni and colleagues 50 identified prospective studies which suggested extreme prematurity (less than 32 weeks’ gestation) conferred a significant risk of developing bipolar.
In general, the evidence for prenatal and perinatal factors as an independent risk factor for developing bipolar is relatively weak and inconsistent, and such factors appear to confer greater risk for developing other mental disorders, such as schizophrenia. 17 The evidence for T. gondii infection is more substantial, while maternal CMV and influenza infection warrant further investigation as to their associations with bipolar.
Postnatal factors
Childhood maltreatment
Childhood maltreatment is a well-studied environmental risk factor with high-quality evidence that it confers a risk for later development of bipolar, 51 although it is also associated with behavioural problems and other mental illnesses.105,106 When investigating specific subtypes of abuse, several studies have identified a link between emotional abuse or emotional neglect and the later the development of bipolar,18,19 while emotional abuse appears to be the most frequent subtype of abuse experienced in bipolar patients. 20 A recent high quality meta-analysis of childhood adversity in bipolar patients compared with healthy controls found significant associations between development of bipolar and prior physical, sexual and emotional abuse, and physical and emotional neglect. 21 The largest association was for emotional abuse which was four times more likely to have occurred in bipolar patients than in controls. 21 Moreover, higher rates of childhood adversity were found in patients with bipolar compared with unipolar depression, although rates were similar to schizophrenia. 21 Gilman and colleagues 52 also found that a history of childhood abuse increased the risk of transitioning to bipolar following a depressive episode. This suggests that abuse and neglect during childhood confer some specific risk to more severe forms of mental illness.
As well a risk factor, childhood maltreatment appears to be associated with poorer clinical outcomes in bipolar, with more severe and more frequent mood episodes, 22 earlier onset, increased risk of suicide and comorbid substance misuse. 23 The relationship between childhood abuse and the severity of bipolar adds further weight to its position as a potential causative factor for the disorder. Notwithstanding, childhood maltreatment does not appear to be specifically related to psychotic symptoms or a diagnosis of bipolar type I over type II.21,24
Whilst it seems likely that childhood traumatic events increase the risk of bipolar, why or how they do this remains unclear but is the focus of ongoing research. Traumatic events are linked to increased levels of affective instability or emotional dysregulation more generally in people with bipolar and this represents one possible mechanism of action. 107 Other dimensions of psychopathology such as hostility and impulsivity, along with affective instability have been shown to mediate the association between childhood maltreatment and outcomes in bipolar, 108 while alterations in the hypothalamic–pituitary–adrenal (HPA) axis, 109 increased levels of BDNF and inflammatory cytokines 110 and reduced limbic grey matter volume 111 represent possible neurobiological underpinnings of the effect of childhood trauma and how this may lead to later psychopathology and bipolar, in particular.
It should be noted that there is difficulty in determining to what extent childhood maltreatment is a cause or consequence of the predisposition to develop bipolar, as parental psychopathology may confer a genetic risk of the disorder, as well as increased risk of childhood maltreatment. 112 The retrospective nature of these studies introduces the possibility of recall bias with regard to childhood adversity, and at present, there are few prospective studies investigating the association between childhood maltreatment and bipolar.
Psychological stressors
Recent stressful life events are known to affect the course of bipolar, 113 although their relationship with the onset of the disorder has been less extensively investigated compared with unipolar depression. 49 A systematic review by Tsuchiya and colleagues 49 identified four studies investigating stressful life events prior to the onset of bipolar, the three largest of which found an increased risk of onset within 6 months of such events. A meta-analysis found that patients experience more life events prior to relapses into either manic or depressive episodes than during euthymic periods, although the rate of significant life events prior to the onset of bipolar was similar to unipolar depression. 25 Other studies have supported the association between life events and the onset of bipolar, including a large case-control study which found that stressful life events were associated with a first hospitalization for a manic episode, particularly suicide of a first-degree relative, but also recent marriage, divorce, disability or unemployment. 26 There are a number of confounders to these associations, particularly with regard to suicide of a first-degree relative, where genetic factors play a significant role, as death due to other causes was not associated with hospitalization. 26 A bidirectional relationship has also been suggested for stressful life events in bipolar, as there is evidence that these events occur both prior to and following mood episodes. 27
There is also evidence for specific life events conferring a risk for bipolar, such as early parental loss and childbirth. The systematic review by Tsuchiya and colleagues 49 found that only 3 of the 10 studies investigating parental loss identified an association with bipolar, although it is noteworthy that one of these was a very large cohort study which adjusted for a number of confounders, including family history of mental illness. 53 A meta-analysis found that childbirth specifically increased the risk of mood episodes in patients with bipolar, more so than relapses in unipolar depression or schizophrenia. 25 Tsuchiya and colleagues 49 identified only three studies investigating onset of bipolar following childbirth, but each found an association with subsequent bipolar diagnosis within 12 months. This is perhaps unsurprising, considering the association between puerperal psychosis and bipolar, 114 but it is unclear whether the reason for the association is genetic, hormonal or related to childbirth as a life event.
However, life events are relatively nonspecific in relation to mental and physical illness, and appear to be associated not only with the onset of bipolar disorder and unipolar depression, but also psychosis, 115 anxiety disorders, 116 ischaemic stroke 117 and circulatory disorders. 118 While gene–environment interactions have been identified between life events and the onset of specific disorders, 12 the use of checklists to identify life events in such studies has been criticized as lacking sufficient detail with regard to the severity and context of such events. 119 These methodological issues make it difficult to establish causation between life events and development of bipolar.
Substance misuse
Bipolar is frequently comorbid with misuse of substances, including cannabis, opioids, cocaine, sedatives and alcohol,50,52 and causality has been suggested in both directions. 120 While the high level of comorbidity is undeniable, causality it much harder to ascertain as there is often difficulty in establishing the temporal relationship between substance misuse and the onset of mental illness. This is compounded by the relative lack of prospective, longitudinal studies examining the relationship between substance misuse and bipolar. 121
There is increasing evidence that cannabis use can act as a risk factor for the development of bipolar as well as psychotic disorders. A recent systematic review by Gibbs and colleagues 28 identified several studies supporting a link between cannabis use and subsequent relapse of manic symptoms. This review also included a meta-analysis of two large prospective cohort studies29,30 which found that cannabis use almost trebled the risk of new-onset subthreshold manic symptoms after adjusting for potential confounding factors. A further large prospective cohort study found cannabis use increased the risk of first episode bipolar by a factor of 5 after adjusting for confounders, and demonstrated evidence of a dose–response relationship. 31 Other studies were more equivocal, finding increased risk of bipolar only in those with weekly to daily cannabis use and no dose–response relationship, 32 or increased risk only in those with a past year episode of depression. 52 Recently, a prospective analysis has demonstrated cannabis use at age 17 is associated with hypomania in young adulthood independent of psychotic symptoms and other important confounders. Further path analysis indicated cannabis use is one mechanism by which childhood abuse translates to increased risk of bipolar symptoms. 33
Other substances of abuse are also important in the risk of bipolar. Prospective studies have linked opioid use to an increased risk of developing bipolar, which is greater than other mood disorders.34,35 A further study found that alcohol and drug abuse or dependence before the age of 25 increased the odds of developing subsequent bipolar, although differences between specific drugs were not examined. 36 Cocaine use has also been implicated, although is less well studied, 37 and as stimulant use can precipitate mania or similar symptoms, 120 this may lead to inappropriate diagnosis of bipolar, 122 rather than act as a causative factor.
There are significant confounding factors to associations between bipolar and substance misuse, which remain despite attempts at adjustment within the studies. It has been suggested that cannabis may serve as self medication for bipolar illness, 123 and therefore may be used by those with subthreshold symptoms prior to the onset of bipolar. Furthermore, there is evidence that shared genetic factors confer risk for developing both substance misuse disorders and bipolar,124,125 while childhood maltreatment is also associated with both disorders.20,22,113
Medical comorbidity
Bipolar is known to be comorbid with a number of medical and psychiatric conditions.51,38,39 There are multiple reasons for this, including shared genetic and environmental vulnerabilities, consequences of treatment, recognition bias on the part of clinicians as well as the potential for a direct causal relationship in either direction.
There is strong evidence for the association between bipolar and irritable bowel syndrome (IBS) 51 highlighted in a recent large meta-analysis of retrospective cohort studies. 40 However, potentially important confounders, such as antidepressant use, were not adjusted for. There is also evidence that both disorders may share inflammatory51,126,127 and stress-related aetiologies,25,128 which could give rise to this association.
Similarly, recent meta-analyses have shown asthma, 41 obesity, 42 migraine 43 and head injury 44 are associated with bipolar. The evidence for these associations is mediated by the relatively small number of studies included, most of which were cross sectional and lacked data to adjust for confounding factors. However, for asthma, a retrospective cohort 45 and large prospective study 46 also support the association, which may be mediated by shared inflammatory pathways126,127 or the use of corticosteroids during early childhood.38,45 Medication and lifestyle factors significantly confound the association with obesity, for which there are few prospective studies and weak evidence for a directly causal relationship, while the association with traumatic brain injury is potentially confounded by ‘accident proneness’ or physical abuse. 129 There is evidence of increased prevalence of bipolar in patients with multiple sclerosis (MS)47,130 which cannot be completely accounted for by steroid-induced mania, and in some instances, psychiatric symptoms may predate the diagnosis of MS. 131 However, other studies have not supported this association. 38
A meta-analysis reported high lifetime prevalence of anxiety disorders in bipolar patients, 48 while ADHD, conduct disorders, aggression and impulsivity also appeared to increase risk of developing bipolar. 39
Prodromal features and bipolar at-risk criteria
It is becoming increasingly recognised that bipolar, like schizophrenia, has a prodromal phase which can be identified prior to development of the full illness.132,133 However, one issue with research into this area is the potential conflation of the concepts of a prodrome for bipolar, referring to symptoms that can be retrospectively identified as preceding the onset of the disorder, and a ‘risk syndrome’ consisting of clinical features, comorbidities and risk factors which increase the risk of later developing bipolar. 134 At present, neither prodrome nor risk syndrome has been fully defined, although the bipolar at-risk (BAR) assessment tool has demonstrated predictive validity and reliability for identifying those at risk of bipolar, with around 23% of those identified transitioning to mania or hypomania. 135 A study using the BAR assessment tool criteria found that cyclothymia had the best overall clinical utility for case finding and screening when focusing on depressed youths with an early transition to bipolar. The clinical utility profile of subthreshold mania, antidepressant emergent elation, family history of bipolar and atypical depression suggested they were better for screening out noncases. 136 However, other studies have questioned the associations between clinical characteristics of depression and transition to bipolar. 52
The low positive predictive value of these precursors reduces their usefulness, and of the significant proportion of those ‘at risk’ who do not go on to develop bipolar there is limited understanding of what factors are protective against this transition, or how this group differs from those who do develop bipolar. 134 Future research should focus on identifying differences in this group, while continuing to refine screening tools for prodromal identification and risk syndromes in prospective studies. Focusing on transition to first-episode mania may have greater reliability in identifying cases. 134
First-episode bipolar mania has an annual incidence of around 5 per 100,000 of population, 137 and peak incidence occurs between 21–25 years. 138 Although the incidence of first-episode mania is equal between males and females, 137 studies have found that age of onset is around 5 years earlier for men. 139 A meta-analysis of longitudinal studies of first-episode mania found that 87.5% of patients achieve syndromal recovery within the first year, meaning they no longer meet criteria for diagnosis. However, the symptomatic recovery rates (essentially defined as being symptom free) were 62.1% within the first year, while 41% experience a recurrence of a manic, mixed or depressed episode over the same period. 140 Considering the relatively poor outcome in such patients, the potential to identify a risk syndrome or prodromal phase of bipolar in those presenting with a depressive illness offers the opportunity to intervene at an earlier stage, leading to improved outcomes. 68
Conclusion
Risk factors for bipolar are numerous, both genetic and environmental, but low attributable risk, inconsistency of results, inability to identify the temporality of the relationship, lack of a clear biological mechanism and the nonspecific nature of many risk factors means that causation is difficult to assign in an individual patient. Studies of environmental risk were also unable to completely adjust for confounding. However, there is evidence that severity of bipolar is related to childhood emotional abuse and the degree of cannabis misuse, suggesting a dose–response relationship. The association with T. gondii is also strong, with some evidence of biological plausibility, although concerns remain about temporality. Bipolar is associated with medical comorbidities such as IBS and asthma, which may point towards shared inflammatory pathophysiology of the disorders, while other psychiatric disorders and clinical features that predate the onset of bipolar may point towards an identifiable ‘risk syndrome’. Future research into these risk factors should focus on establishing temporality, whether the severity of bipolar is linked to the risk factor, and identifying potential neurobiological and environmental mechanisms to explain the associations. Finally, research into gene–environment interactions is required to link existing evidence on genetic and environmental risks.
Footnotes
Funding
TR is partly funded through the National Institute of Health Research as an academic clinical fellow. The authors received no financial support for the research, authorship, or publication of this article.
Conflict of interest statement
The authors declare that there is no conflict of interest.
