Abstract
The aim of this paper is to examine risk factors for schizophrenia and to speculate on possible population-based interventions that may reduce the incidence of schizophrenia. Universal interventions have strengths and weaknesses that relate to the features of both the exposure (the nongenetic risk-modifying variable) and of the disorder [1]. In particular, this approach alleviates the need to identify a minority of individuals who are ‘high risk’ – the focus of much current research in the prevention of schizophrenia. If we can identify such individuals, and if we can reduce their risk, then this is a highly desirable goal. However, if we cannot identify high-risk individuals, we need to consider alternative strategies. In the present paper, I will argue that we also need to consider population-based, universal interventions for the primary prevention of schizophrenia.
Geoffrey Rose has emphasised that population-based interventions are best suited to risks that are spread diffusely across the entire community, rather than segregated to a small minority [2]. In other words, if a large proportion of the community is exposed to a small risk, then population-based interventions may avert more illness (a greater number of cases prevented) than interventions based on rare, high-risk individuals. Rose also introduced the concept of the ‘prevention paradox’ – a preventive measure that brings large benefits to the community but which offers little to each participating individual [2]. For example, many population-based interventions (e.g. vaccination, wearing a seatbelt) bring little direct benefit to the individual; however, members of the community are willing to accept them because they cause little inconvenience. Inconvenience is weighted against the frequency of the undesirable outcome and the severity of this outcome.
In order to implement this type of prevention, we need to identify risk factors that are weak (i.e. small effect) and prevalent. We need then to consider whether these risk factors can be modified and if they can, whether it is feasible to deliver an intervention that is safe and cheap to a large group in the community.
The first part of the present paper discusses the definitions of the different types of risk factors. Next, the evidence for several risk factors for schizophrenia is reviewed, with a particular emphasis on the potential for population-based interventions. The final part speculates on what we can do now, and what may be possible in the future.
What is a risk factor?
In recent times, there has been an effort to clarify terminology surrounding risk factors [3],[4]. One important point regarding variables that may correlate with outcomes, but do not precede the outcome, is that these factors should not be labelled risk factors but rather sequelae, consequences or concomitants. Another important point is that when considering risk factors, it is important to understand that there are risk indicators that can be epiphenomena or proxy markers of an underlying risk-modifying factor. A major problem with a brain disease like schizophrenia is that the current knowledge base is limited and, therefore, we can not confidently predict if a risk factor is causally related or whether it is a proxy marker.
The term ‘risk-modifying factor’ should be reserved for factors that appear to operate within the causal chain (contribute to the outcome). If this type of risk factor could be reduced, then the incidence of the resultant disorder should also be reduced. While risk-modifying variables can be proximal to disease identification, in neurodevelopmental models of schizophrenia, we are looking for distal or ‘upstream’ risk-modifying variables. Also, these factors may operate directly or indirectly (sometimes referred to as first or second order effects). Risk factors can be fixed (e.g. gender) or variable (alcohol intake), and they can also be protective or adverse.
In order to reduce the incidence of schizophrenia, we need to design an intervention that affects risk-modifying variables rather than risk indicators. For example, imagine if one was trying to prevent lung cancer in the absence of clear biological mechanisms. Case control studies may identify that those with lung cancer (compared to healthy controls) had increased odds of smoking, having ashtrays in the house, having nicotine-stained fingers and matches in their pockets. Interventions designed to reduce smoking would impact on the incidence of the disorder, while interventions designed to reduce the other factors (all risk indicators) would not directly impact on disease incidence. Thus, our challenge in the prevention of schizophrenia is the correct identification of risk-modifying factors rather than risk indicators.
Candidate exposures
Research has identified a number of pre- and perinatal candidate risk factors for schizophrenia. These include family history, season of birth, place of birth, obstetric complications, prenatal exposure to viruses and prenatal famine.
Risk factors for schizophrenia
It should be noted that the PAR is a problematic concept. For example, the total PAR for various risk factors could total more than 100%% and the measure does not take into account interactions between different risk factors. However, PAR does serve to rank the order of risk factors in a manner that is of interest to universal interventions. It captures a central feature of this level of intervention – namely that small risk factors, if widely distributed among the community, may ‘cause’ more cases than rarer but larger risk factors. For example, if one wanted to reduce lung cancer, then one could turn to one of the strongest known risk-modifying factors for lung cancer, namely asbestos exposure. Fortunately, this exposure is now relatively rare, thus total elimination of this exposure would only avert a small fraction of cases. On the other hand, smoking is associated with a much smaller odds ratio of lung cancer compared to asbestos exposure; however, this exposure is very prevalent in the community. Thus, prevention of smoking would avert many more cases of lung cancer than would the elimination of asbestos exposure.
In summary, in order to prevent schizophrenia by universal interventions we need to first identify weak but prevalent risk factors. The next section will review such candidate exposures.
Genetic factors
Of the presently known risk factors for schizophrenia, family history is by far the most robust [6]; however, as yet, no single gene has been identified. Having a mother with schizophrenia (and a healthy father) increases the risk of schizophrenia in the offspring about ten-fold, yet the prevalence of mothers with schizophrenia in the general community is low. Mortensen and colleagues reported that the PAR of having one or both parents affected was only 3.8%% [7]. While the interaction between genetic and nongenetic risk factors is almost certainly more complicated than these figures suggest, the finding reinforces the fact that while risk factors with large odds ratios are attractive targets, their real effect on the population may be small.
It is plausible that several genes, each of small effect, contribute to the risk of schizophrenia and that most individuals with these genes are unaffected. If we could identify these susceptible individuals, then we could develop neuroprotective interventions for their offspring. At present, genetic risk factors do not readily translate into universal interventions. Perhaps in the future, genetic screening could lead to ‘selective’ prevention for the high-risk fetus and infant.
Season of birth
People born in winter and spring tend to have a slightly increased relative risk of developing schizophrenia compared to those born in autumn and summer. While data on the season of birth effect in the northern hemisphere population are quite robust [8], the same cannot be said for data from the southern hemisphere. Recently, a meta-analysis performed on data from a range of southern hemisphere studies (in Australia, South Africa and the Reunion Islands) did not support a season of birth effect [9]. This north–south difference may help clarify the nature of the underlying risk-modifying factors. Mortensen and colleagues reported a very small effect size (1.11) for season of birth and a PAR of 10.5%% [7].
Season of birth is a risk indicator, and thus can only serve to generate candidate risk-modifying variables. Presently, candidate exposures that may be related to the season of birth effect include perinatal viral exposures [8] and low prenatal vitamin D [10].
Place of birth
People born in the city tend to have a greater risk of developing schizophrenia compared to those born in rural regions. The quality of evidence for place of birth as a risk factor is now quite robust following the recent publication of two major population-based studies: one from The Netherlands [11] and one from Denmark [7]. The relative risk of developing schizophrenia when born in the city versus the country is about 2.4. However, as urban-birth is relatively frequent, the Danish study reported the PAR for this variable to be 34.6%%. Place of birth, once again, appears to be a proxy marker for a risk-modifying variable, such as a viral infection, nutrition, low vitamin D or other, as yet unidentified, factors.
Pregnancy and birth complications
Links between pregnancy and birth complications and an increased risk of schizophrenia have been examined over many decades. While not entirely consistent, the weight of the evidence (based on case control studies, birth cohorts and population-based register linkage) supports an association. A recent meta-analysis based on 700 patients with schizophrenia and 835 controls found significantly increased odds ratios for premature rupture of membranes (odds ratio = 3.11), gestational age shorter than 37 weeks (odds ratio = 2.44), and use of resuscitation or an incubator (odds ratio = 2.21) [12]. There were trend level associations between schizophrenia and two other perinatal variables; that is, low birthweight, and a forceps delivery. The authors reported attributable risks for significant variables as between 3 and 7%%. There is an extensive literature on how to improve perinatal outcomes, much of it based on randomised controlled trials. Options for intervention at a population level (all pregnant women) could include periconceptual counselling, optimising general antenatal care, and cessation of smoking. Two attractive features of this type of intervention are that disorders other than schizophrenia could be averted, and pregnant women tend to be more receptive to health promotion than the general community.
Prenatal infection
A range of prenatal infective agents can impact on brain development. The main clue implicating pre- and perinatal infections as candidate risk factors for schizophrenia stems from the excess of winter/spring births for schizophrenia. While influenza has been the focus of much research over the past decade, the strength of the evidence of prenatal exposure to influenza is weak and inconsistent [13]. Other candidate viruses include rubella [14], Coxsackie B [15] and Borna virus [16]. Prenatal exposure to rubella has long been known to impact on brain development. Susser and colleagues have followed up offspring of women who had serologically proven exposure to rubella during pregnancy [14]. When those with congenital deafness, blindness and intellectual handicap were excluded, there was an increased risk of schizophrenia in this group. Several case control studies have reported increased seroprevalence of Borna virus in patients with schizophrenia versus healthy controls; however, many of these studies have been plagued by methodological problems. A large, multicentre study is presently under way in order to explore this issue.
In summary, there are no robust data to either support or reject the theory that pre- or perinatal exposure to infection increases the risk of schizophrenia. Therefore, until there are convincing data from well-designed and adequately powered studies, the viral theory should remain a candidate risk factor. From the perspective of universal interventions, viral illness can be prevented by vaccinations, and there are now examples of public health interventions where mass vaccinations have eliminated viruses completely.
Prenatal nutrition as a risk factor
Prenatal nutritional deprivation is a biologically plausible risk-modifying factor for schizophrenia. While studies of the incidence of schizophrenia in developing nations (where poor nutrition is more prevalent) do not show higher rates, there remains the possibility that deficits in specific micronutrients may play a role. For example, folate has been implicated in the pathogenesis of another disorder of brain development, neural tube defects [17].
The evidence for prenatal famine as a risk factor has been derived from one study by Susser and colleagues, who identified increased risk of schizophrenia in the offspring of women who were pregnant during a famine in The Netherlands during World War II [18]. A range of adverse health outcomes has been associated with exposure to this prenatal famine (e.g. neural tube defect and intellectual handicap). In addition, there is considerable interest in recent years about the impact of prenatal nutrition and various adult-onset disorders, such as diabetes, cardiovascular disease and hyptertension [19].
In summary, while the evidence implicating prenatal nutrition as a risk factor in schizophrenia is scant, it is an attractive candidate for universal intervention. Better maternal nutrition is safe, relatively cheap, and could feasibly impact on a range of distal health outcomes in the offspring.
What can we do about the candidate exposures?
In order to develop universal intervention programs, we need also to consider whether the candidate exposure is able to be modified or averted. With viruses, there is the option of vaccination. For example, congenital rubella has been drastically reduced through vaccinations. Nutrition is another area of potential intervention, with large, controlled trials showing that increasing maternal folate levels (a safe and cheap intervention) is associated with reduced incidence of neural tube defect in the offspring. Another safe and inexpensive public health intervention that could improve obstetric outcomes are stop smoking campaigns. Once again, these interventions are relatively cheap, have proven efficacy, and are associated with a broad range of improvements in health care to both the mother and the offspring.
Current preventive research at the Queensland Centre for Schizophrenia Research
Two streams of prevention research are currently being conducted at the Queensland Centre for Schizophrenia Research (QCSR). One examines ways to optimise antenatal care for high-risk fetuses and the other examines whether low prenatal vitamin D is a risk-modifying factor for schizophrenia.
Antenatal care for women with high-risk fetuses
Women with schizophrenia tend to experience more obstetric complications at delivery [20],[21]. It has also been suggested that the effect of family history together with obstetric complications compounds the subsequent risk of schizophrenia in the offspring [22]. If this is correct, it follows that optimising antenatal care to women who have a ‘high risk’ fetus may be suitable for ‘selective’ prevention [1]. In the absence of antenatal genetic testing, the only current way to identify a fetus who has a genetic vulnerability to schizophrenia is to examine the mental health of the mother and father.
Researchers at QCSR are presently involved in a case control study examining social, psychological and obstetric variables in women with a history of psychotic illness attending an antenatal clinic versus healthy controls. Mothers and their infants are followed up six months after birth. The study aims to generate directions for simple, pragmatic interventions that lead to improved health care for these women. Preliminary analysis suggests that there are many socioeconomic factors that are impacting on the optimal antenatal care. It seems appropriate to provide these women with targeted antenatal care (as you would with a mother who had diabetes, epilepsy or a past history of premature labour or low birthweight). In addition, mothers with a history of psychosis tend to have more life events (and be more distressed by them) compared to healthy controls. While these life events tend to relate to lack of stable accommodation and poverty, those with supportive family members are able to cope with these stressors better. Therefore, future interventions might include streamlined antenatal–mental health coordination with assertive community follow up and more support for these women and their caregivers (home help, parent aid, etc.) [23].
Nevertheless, even if we were able to improve the outcome for mothers with psychosis and their offspring, the impact on the incidence of schizophrenia would be very small. Most women with psychosis are already mothers at the time of their first admission [24], and most people who develop schizophrenia (97%%) do not have mothers with schizophrenia. Regardless of these issues, providing improved antenatal care to mothers with psychosis should result in more immediate outcomes related to maternal and infant health.
Is low prenatal vitamin D a risk-modifying factor for schizophrenia?
The second prevention study being undertaken at QCSR relates to identifying candidate nongenetic exposures and involves examining whether low prenatal vitamin D is a risk-modifying factor for schizophrenia. Vitamin D is a steroid hormone with a diverse range of actions obtained through exposure to sunlight and diet. Vitamin D is a potent inducer of nerve growth factor and there are vitamin D receptors in the developing and adult brain. While there is a range of literature describing the importance of vitamin D in bone and calcium homeostasis, it is clear that vitamin D, like other steroids, has an impact on nearly all tissues [25]. Vitamin D has features that could explain many of the epidemiological landmarks for schizophrenia. Vitamin D fluctuates across seasons. There are city and rural differences (people in the city are more prone to low vitamin D due to factors such as architecture, reduced outdoor activity and smog). Vitamin D also provides a possible explanation for the increased risk of schizophrenia in second-generation, dark-skinned migrants who have moved to live in cooler climates (their skin is less efficient at producing vitamin D). Although hypovitaminosis D is an attractive candidate risk factor for schizophrenia, there is no direct evidence to implicate this steroid in the aetiology of schizophrenia. QCSR has embarked on a research program involving epidemiology and animal experiments in order to explore this hypothesis [10].
This candidate exposure has many desirable features when it comes to selecting potential universal interventions for public health. There is a growing body of evidence linking prenatal nutrition to several adult health disorders, and it is feasible that close attention to maternal nutrition could translate into a reduced burden of chronic disorders in the offspring. Nutritional interventions, on the whole, are safe, cheap and acceptable to women. Finally, the demonstration that folate supplementation can reduce the incidence of another neurodevelopmental disorder (neural tube defect) provides inspiration to those interested in primary prevention of schizophrenia.
Future directions
Searching for candidate risk factors for schizophrenia demands collaboration with developmental neurobiologists and researchers involved in perinatal medicine, public health and nutrition. Unfortunately, the environment cannot be systematically ‘mapped’ in order to generate candidate exposures as the genome can be mapped, linked and sequenced. It is essential that the data used to address this problem is derived from epidemiologically informed sampling frames. Sample sizes need to be representative of the underlying population so the PAR that guides public health planning can be generated. Links between candidate exposures and genetic epidemiology will also be crucial for future research.
In conclusion, the following factors should be considered when selecting candidate nongenetic risk factors for schizophrenia:
Does the exposure gain support from schizophrenia epidemiology? Is there evidence from animal experiments and controlled human studies to support the candidate? What research needs to be done in order to rank order competing exposures? Can interventions be delivered at a universal level which will impact on the prevalence of the exposure? Can these interventions result in the primary prevention of disorders other than schizophrenia? Are these interventions safe, inexpensive and acceptable to the general community?
Sartorius and Henderson proposed three options for the research community interested in the primary prevention of mental illness [26]. First, forget about primary prevention completely and concentrate on better treatments and cures. Second, fund more research in order to discover the causes of serious psychiatric illness and then commence primary prevention. Finally, start primary prevention based on existing imperfect knowledge. Clearly we need a balance of all three approaches. Basic strategic research needs to go hand in hand with attempts at primary prevention. We need to encourage a vision that primary prevention of schizophrenia is achievable, balanced by an appreciation that a huge amount of work needs to be done over the next few decades. The First Australian Schizophrenia Prevention Conference has provided an important early step along this journey.
Footnotes
Acknowledgements
This project was supported by the Stanley Foundation.