Only recently has improving the lives of the 20% of children estimated to have a parent with a mental disorderReference Bassani, Padoin C, Philipp and Veldhuizen1–Reference Abel, Hope, Swift, Parisi, Ashcroft and Kosidou3 become a public health priority.Reference Puras, Kolaitis and Tsiantis4–Reference Abel, Hope, Faulds and Pierce6 Thus, despite these children being likely to experience multiple deprivations and challenges, notably little is known about them or their health needs. Most of the information available about these children relates to their higher risk of developing psychiatricReference Rasic, Hajek, Alda and Uher7 and neurodevelopmental problems.Reference Fairthorne, De and Leonard8–Reference Berg, Back, Vinnerljung and Hjern10 Prior research has shown that children of parents with mental disorder have a considerably increased risk of congenital anomalies,Reference Webb, Pickles, King-Hele, Appleby, Mortensen and Abel11 of having been born with obstetric complications, including premature birth and low birthweight,Reference Stein, Pearson, Goodman, Rapa, Rahman and McCallum12 of being stillborn and of premature death (the last persists into adulthood).Reference Webb, Abel, Pickles and Appleby13,Reference Webb, Sc, Abel, Appleby, King-Hele and Mortensen14 However, little attention has been paid to a link between parental mental disorder and poorer offspring physical health, defined here as diseases such as asthma or diabetes affecting somatic rather than mental health. This is important not least because poor physical health has a detrimental effect on a child's development, with chronic ill health affecting social functioningReference Meijer, Sinnema, Bijstra, Mellenbergh and Wolters15 and academic progress.Reference Thies16,Reference Currie17 Also, health disparities in childhood often persist into adulthood, leading to lower life expectancy.Reference Rough, Goldblatt, Marmot, Nathanson, Mansfield, Foyle and Nathanson18 Moreover, such a readily identifiable high-risk group could be a suitable target for early interventions.
This systematic review examined whether having a parent with a mental illness increases the risk of physical illness throughout childhood. Prior reviews have focused on maternal depression or anxiety and specific child health outcomes of obesityReference Milgrom, Skouteris, Worotniuk, Henwood and Bruce19–Reference Benton, Skouteris and Hayden21 and asthma.Reference Easter, Sharpe and Hunt22–Reference Andersson, Hansen, Larsen, Hougaard, Kolstad and Schlünssen24 Most of these included self-reported outcomes, cross-sectional designs and studies in which the childhood illness precedes exposure to the parental illness. Our review and meta-analysis included a broad range of parental mental disorders, specified a priori, and any clinically diagnosed child physical health outcome. We limited the scope of the review by using strict definitions of exposure and outcome and by pre-specifying the type of study design. This enabled us to summarise the literature on physical health of children whose parents have mental illness, focusing on studies of higher quality and with clearly defined measures.
Method
This review was developed according to the Centre for Reviews and Dissemination's ‘Guidance for undertaking reviews in health care’.Reference Tacconelli25 The protocol was registered on the PROSPERO database (reference: CRD42017072620).
Selection criteria
Cohort or case–control studies that quantify associations between parental mental disorders and physical health in the offspring were included. This review focused on childhood disease; therefore, we excluded studies reporting outcomes occurring only in the neonatal period (0–28 days) or beyond age 18. Studies were also excluded if: the outcome was measured prior to exposure; the sample size was <10; or the cohort sample was drawn from diseased children. If more than one study had overlapping study populations and definitions of exposure and outcome, the study with the largest sample was chosen.
We specified a priori parental mental disorders that were of interest: substance use disorders (ICD-10 category: F10–19); schizophrenia, schizotypal and delusional disorders (F20–29); mood disorders (F30–39); anxiety disorders (F40–41); obsessive–compulsive disorders (F42); post-traumatic stress disorder (F43.1); and eating disorders (F50). We included studies where: mental illness was defined by a clinical diagnosis, using ICD or DSM criteria; mental illness was measured using a peer-reviewed instrument; or where the parent received treatment for a mental disorder.
We included studies reporting any physical disease in the offspring, clinically diagnosed and in the World Health Organizations’ ICD-10 framework. This excluded psychological or neurobehavioural disorders (i.e. chapter V of ICD-10) and any disorder categorised in ICD-10 chapter VIII: ‘Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified’.
Search strategy
We searched the following databases for published papers, reports, conference abstracts and theses: MEDLINE; PsycINFO; PsycARTICLES; Embase; Web of Science; ProQuest ASSIA. We included studies published within the dates 1 January 1970 to the search date (21 September 2017), and did not initially include language restrictions.
We searched using the following general terms: (children) AND (parent) AND (mental illness) AND (physical disease) AND (case–control OR cohort study). We included terms for the specific mental illnesses and, for the purposes of conducting the search, an a priori list of common diseases of childhood that was developed by clinical collaborators (R.P. and K.M.A.). We included MESH terms and included variants and synonyms using truncations and wildcards where helpful. The full search strategy can be found in supplementary Appendix A available at https://doi.org/10.1192/bjp.2019.216. Following screening, we sought to identify further studies by hand-searching review papers arising from our searches. We also conducted hand-searches of reference lists of identified papers and we conducted a cited-reference search using Web of Science.
Screening
Duplicates were removed and then two reviewers (M.P. and A.K.) screened titles and abstracts, initially piloting 250 papers to ensure that consistent features were selected. All studies categorised as ‘yes’ or ‘maybe’ for inclusion were extracted for full text screening. A full-text screening tool was piloted by three reviewers (M.P., A.K. and J.G.) using 30 papers and then the same three reviewers divided all full-text papers between them so that each paper was screened twice. Disagreements and ambiguities were resolved in a group discussion, calling on a fourth reviewer (K.M.A.) as necessary. Lack of resources meant that we excluded papers that were not in English.
Data extraction and analysis
Data extraction was carried out by two reviewers (M.P. and A.K.), extracting data on: outcome type, exposure type; assessment instrument used; timing of exposure and outcome; sample size; study setting; statistical model; variables used in adjustment; and effect size. Study quality was assessed using the National Institutes of Health (NIH) Quality Assessment Tools using checklists developed for cohort and case–control studies.26
When studies used multiple adjusted models, we report from the model with the most covariates, unless we judged that any of the extra variables were on the causal pathway between exposure and outcome. When studies presented multiple effect sizes from multiple exposures we report only those exposures considered by the studies' authors to be the most severe and most chronic. All extracted effect sizes are presented in supplementary Appendix B. For meta-analysis, when a study reported multiple outcomes we selected the most frequent and when studies reported the effect of maternal and paternal exposure we selected estimates associated with maternal exposure.
For evidence synthesis, studies were grouped according to outcome type. For each grouping, estimates were converted into odds ratios (where possible) and pooled odds ratios were estimated using random-effects meta-analyses. Between-study heterogeneity was estimated using the I 2 statistic. Two sensitivity analyses were conducted. First, we determined the robustness of the meta-analyses to removing studies that were ranked poor quality. Second, where appropriate we compared estimates by type of study (case–control versus cohort design). Analyses were done using Stata 14 for Windows using the metan command.
Results
The database searches yielded 15 945 non-duplicate studies (Fig. 1). Of these, 251 were considered for full-text screening and 221 were excluded (94 failed the outcome criteria; 37 failed the exposure criteria; 36 were non-relevant study designs; 12 were reviews; 12 selected a diseased-only cohort; 7 were in a language other than English; 7 did not report sufficient data; 6 were outside the age range; 4 had exposure subsequent to disease; 2 were nested within a larger study; and we were unable to locate the full text for 4). This resulted in 30 studies for inclusion. After additional searches, a further 11 studies were included (5 from other reviews, 5 from reference lists of included papers and 1 from the cited reference search), giving a final total of 41 studies.
Fig. 1 Flow diagram for studies included in systematic review.
Overview of included studies
The vast majority of included studies (31/41) investigated exposure to parental depression; 7 investigated anxiety, 5 substance misuse and 1 psychotic disorder (Table 1). One study defined exposure as ‘psychiatric morbidity’, another as ‘common mental disorder’ and another considered the effect of post-traumatic stress disorder. None investigated the effect of parental eating, bipolar or obsessive–compulsive disorders. All studies examined maternal mental disorder; 17% (7/41) also examined exposure to paternal mental disorder. Forty-two percent of studies (17/41) measured exposure only during the perinatal period, defined here as from the start of pregnancy to 1 year after the birth. The median age at the last follow-up was 5 years (interquartile range IQR = 2–7.5) and the median sample size was 1696 (IQR = 294–12 618).
Table 1 Descriptive summary of the 41 studies included in the systematic review
a. These categories not mutually exclusive.
The majority (26) were prospective cohort studies; 8 were retrospective cohort studies and 7 were case–control studies. Thirty-nine percent (16) came from Europe (half of which were from Scandinavia) and 29% (12) were from North America. Sixty-eight percent (28) were carried out after 2010 and the earliest study was from 1981.
Studies were grouped into the categories according to their outcomes: accidents and injuries (n = 10); asthma (n = 8); other atopic diseases (n = 3); overweight and obesity (n = 10). The remaining studies were split between those from a low- and middle-income country (LMIC) setting (n = 8) (mainly consisting of studies examining diarrhoea or malnutrition) and those from a high-income country setting (n = 2).
Accidents or injuries (ten studies)
Of the ten studies that reported accidents and injuries,Reference Phelan, Khoury, Atherton and Kahn27–Reference Howard, Goss, Leese and Thornicroft36 nineReference McKinlay, Kyonka, Grace, Horwood, Fergusson and MacFarlane28–Reference Howard, Goss, Leese and Thornicroft36 and a combined sample of 314 132 children were included in the pooled analysis for accidents and injuries (Fig. 2). This revealed a 15% increase in the likelihood of a child having an accident or injury if they were exposed to parental mental illness (ORpool = 1.15, 95% CI 1.04–1.26, I 2 = 76.4%).
Fig. 2 Forest plot displaying estimates and pooled estimates of the effect of childhood exposure to parental mental illness on childhood disease.
Five of the six studies that considered the risk of offspring accidents or injuries associated with maternal depression found a positive associationReference Phelan, Khoury, Atherton and Kahn27–Reference Baker, Kendrick, Tata and Orton31 and one study, which was the smallest and the only one to examine paternal as well as maternal depression, did not find an effectReference Schwebel and Brezausek32 (Table 2). Most studies examined outcomes in the first 5 years; however, one study of a birth cohort of 1265 found an effect of maternal depression on traumatic brain injuries up to age 15.Reference McKinlay, Kyonka, Grace, Horwood, Fergusson and MacFarlane28
Table 2 Summary of analyses from studies included in the systematic review
Prosp., prospective; retro., retrospective; gest., gestation; a, adjusted analyses; HR, hazard ratio; IRR, incidence rate ratio; OR, odds ratio; RR, risk ratio; m, month; y, year; wks, weeks; GI, gastrointestinal infection; LRTI, lower respiratory tract infection.
a. From the most severe and chronic exposures reported in each study. Other relevant analyses are reported in supplementary Appendix B.
Three studies examined parental substance use disorder and injuries.Reference Winqvist, Jokelainen, Luukinen and Hillbom33–Reference Raitasalo and Holmila35 One reported a doubling in the risk of traumatic brain injury if either parent had misused alcoholReference Winqvist, Jokelainen, Luukinen and Hillbom33 and another small study (n = 125) estimated a similar effect of maternal substance misuse,Reference Wilson, Desmond and Wait34 although the lower confidence interval included a null effect. A large retrospective cohort study from Finland (n = 113 813) did not find an effect of maternal or paternal substance misuse on risk of injuries in the first 6 years.Reference Raitasalo and Holmila35
One study examined accidents in the first year of life for 199 children with maternal psychotic disorder matched to 787 children without.Reference Howard, Goss, Leese and Thornicroft36 They did not find an increased risk of accidents associated with maternal exposure.
Asthma (eight studies)
From eight studiesReference Cookson, Granell, Joinson, Ben-Shlomo and Henderson37–Reference Kozyrskyj, Mai, McGrath, HayGlass, Becker and MacNeil44 and a combined sample of 450 202 children, we estimated a 19% increase in the odds of childhood asthma for children exposed to parental mental illness (ORpool = 1.19, 95% CI 1.08–1.32, I 2 = 77.0%).
Six of the eight studies found a positive association between maternal depression or anxiety and childhood asthma.Reference Cookson, Granell, Joinson, Ben-Shlomo and Henderson37–Reference Magnus, Wright, Røysamb, Parr, Karlstad and Page41 The remaining two studies also had effect sizes indicating a positive relationship, albeit with confidence intervals that include a null effect.Reference Klinnert, Nelson, Price, Adinoff, Leung and Mrazek42,Reference Kozyrskyj, Letourneau, Kang and Salmani43 Three of these studies also reported on exposure to paternal mental disorderReference Lange, Bunyavanich, Silberg, Canino, Rosner and Celedón39–Reference Magnus, Wright, Røysamb, Parr, Karlstad and Page41 but only one detected an effect.Reference Brew, Lundholm, Viktorin, Lichtenstein, Larsson and Almqvist40 Those that investigated short-term versus chronic depression found a greater effect of the latter.Reference Giallo, Bahreinian, Brown, Cooklin, Kingston and Kozyrskyj38,Reference Kozyrskyj, Mai, McGrath, HayGlass, Becker and MacNeil44 One study that looked at depression exposure during pregnancy versus exposure postnatally did not find a difference.Reference Brew, Lundholm, Viktorin, Lichtenstein, Larsson and Almqvist40
Other atopic diseases (three studies)
In the pooled analysis, using three studiesReference Wang, Wen, Chiang, Lin and Guo45–Reference Elbert, Duijts, den Dekker, de Jong, Nijsten and Jaddoe47 and 23 471 children, there was inconclusive evidence for an association between parental mental illness and childhood atopy (ORpool = 1.36, 95% CI 0.91–2.03, I 2 = 92.9%).
However, a large cohort study from Taiwan reported a positive effect of maternal depression on risk of infant eczema in the first 6 months of life.Reference Wang, Wen, Chiang, Lin and Guo45 One study reported an association between atopic dermatitis and maternal anxiety, but not maternal depressionReference Letourneau, Kozyrskyj, Cosic, Ntanda, Anis and Hart46 and another reported an association with maternal (but not paternal) depression for inhalant (but not food) allergies.Reference Elbert, Duijts, den Dekker, de Jong, Nijsten and Jaddoe47
Overweight or obesity (ten studies)
The pooled analysis for the effect of parental mental illness and being overweight or obese in childhood included seven studiesReference Ajslev, Andersen, Ingstrup, Nohr and Sørensen50–Reference Audelo, Kogut, Harley, Rosas, Stein and Eskenazi54, Reference Wojcicki, Holbrook, Lustig, Epel, Caughey and Muñoz56–Reference Figueiredo, Roos, Eriksson, Simola-Ström and Weiderpass57 and 36 309 children. To facilitate pooling, overweight was selected in the meta-analysis. The pooled estimate showed borderline evidence to conclude a positive association (ORpool = 1.16, 95% CI 0.97–1.39, I 2 = 55.0%). The three studies unsuitable for pooling reported equivocal results.Reference De Sousa48,Reference Bronte-Tinkew, Zaslow, Capps, Horowitz and McNamara49,Reference Blanco, Sepulveda, Lacruz, Parks, Real and Martin-Peinador55
Of the nine studies that examined childhood obesity and its association with maternal depression or anxiety, one small cohort study (n = 160) estimated a positive association,Reference De Sousa48 seven were equivocalReference Bronte-Tinkew, Zaslow, Capps, Horowitz and McNamara49–Reference Blanco, Sepulveda, Lacruz, Parks, Real and Martin-Peinador55 and one reported a negative effect.Reference Wojcicki, Holbrook, Lustig, Epel, Caughey and Muñoz56 Of the three studies that examined duration of maternal depressive disorder, two found an increased risk associated with cumulative exposure to depression, but not of exposure to periodic depression.Reference Santos, Matijasevich, Domingues, Barros and Barros51,Reference Audelo, Kogut, Harley, Rosas, Stein and Eskenazi54 Two large prospective cohort studies, based in The Netherlands and the USA,Reference Guxens, Tiemeier, Jansen, Raat, Hofman and Sunyer52,Reference Wang, Anderson, Dalton, Wu, Liu and Zheng53 indicated an effect of maternal depression on childhood overweight in their unadjusted analyses, but when adjusted for potential confounders (including maternal/paternal body mass index) this effect disappeared.
One Finnish study of 4525 teenagers that examined the association between parental harmful drinking and offspring obesity did not find an effect.Reference Figueiredo, Roos, Eriksson, Simola-Ström and Weiderpass57
Other studies from low- and middle-income countries (eight studies)
Five studies from LMICs with a combined sample of 851 children investigated the effect of maternal mental illness on childhood malnutrition.Reference de Miranda, Turecki, de Jesus Mari, Andreoli, Marcolim and Goihman58–Reference Ashaba, Rukundo, Beinempaka, Ntaro and LeBlanc62 The pooled risk of malnutrition was more than double for exposed compared with unexposed children (ORpool = 2.55, 95% CI 1.74–3.73, I 2 = 0.0%).
Three studies and 13 430 children were pooled to investigate the link between perinatal maternal depression and diarrhoea or gastrointestinal infection and, similarly, the odds of diarrhoea doubled for exposed children (ORpool = 2.16, 95% CI 1.65–2.84, I 2 = 52.0%).Reference Rahman, Iqbal, Bunn, Lovel and Harrington63–Reference Weobong, ten Asbroek, Soremekun, Gram, Amenga-Etego and Danso65
Other studies from high-income countries (two studies)
One retrospective cohort study of 107 587 children in the UK reported an association between maternal perinatal depression and offspring gastrointestinal infection and respiratory tract infections.Reference Ban, Gibson, West and Tata66 Another retrospective cohort study of 2552 children of mothers with alcohol or substance use disorder in Finland found a marginal effect on diseases of the eye, ear and mastoid process.Reference Sarkola, Gissler, Kahila, Autti-Rämö and Halmesmäki67
Study quality
Overall, 10 studies were graded ‘good’, 19 studies were ‘fair’ and 12 studies were graded ‘poor’ (supplementary Appendix C). Particular methodological problems were the lack of clarity regarding sample selection and the measurements used, and overlapping timing of exposure and outcome. Removing studies that were graded ‘poor’ (supplementary Appendix D) made little difference to the pooled estimates. Also, for the two meta-analyses that included case–control studies, the results were consistent by type of study design (supplementary Appendix E).
Discussion
Summary of findings
For the first time, this systematic review summarises current evidence on risk for poor physical health in offspring of parents with mental disorder. Overall, this detailed evidence synthesis paints a picture of relatively poor physical health in the children of parents with mental disorder, with pooled effect estimates revealing an increased risk of injuries, asthma, malnutrition and diarrhoea. However, we highlight striking gaps in the evidence: over three-quarters of the studies focused exclusively on maternal depression and, of those, half on postnatal depression; there is little information about children of parents with mental disorders other than depression or anxiety; and few studies investigated the impact of paternal mental disorder.
In total, 63% of studies (26/41) reported an effect of parental mental disorder on physical health outcomes in the offspring. A further 17% of studies (7/41) estimated a positive association, albeit with 95% confidence intervals that include a null effect. Most of the identified studies examined the relationship between exposure to parental mental illness (predominantly maternal postnatal depression) and risk of childhood injuries (n = 10), obesity (n = 10) or asthma (n = 8). Of note, no eligible studies assessed risk of childhood cancers, diabetes, epilepsy or migraine and only one assessed the impact of serious mental disorders such as schizophrenia or bipolar disorder on childhood physical health. Similarly, no studies assessed effects of dual diagnosis, maternal or paternal personality disorder or eating disorders.
Research in context
We identified evidence of an increased risk of childhood accidents or injuries associated with parental mental illness (ORpool = 1.15, 95% CI 1.04–1.26).Reference Phelan, Khoury, Atherton and Kahn27–Reference Raitasalo and Holmila35 Self-reported data have shown that periods of depression affect a mother's ability to supervise her child.Reference Phelan, Morrongiello, Khoury, Xu, Liddy and Lanphear68 Substance misuse, particularly alcohol dependence, is associated with violent behaviour,Reference Cleaver, Nicholson, Tarr and Cleaver69 which may confer additional risk of injury to the child. The one study that examined the effect of psychosis on risk of childhood accidents did not find an effect, although the confidence intervals indicate that it was underpowered to detect less than a doubling in the rate of accidents.Reference Howard, Goss, Leese and Thornicroft36 All but two of these studies on accident and injury followed children in the first 6 years of life and therefore we do not know whether the risk is ameliorated by school entrance.
We highlight evidence to of an increased risk of childhood asthma associated with parental mental illness (ORpool = 1.26, 95% CI 1.12–1.26) and some evidence to suggest an increased risk of other atopic disorders (ORpool = 1.36, 95% CI 0.91–2.03). The association was observed most strongly in studies in which the exposure was categorised as severeReference Cookson, Granell, Joinson, Ben-Shlomo and Henderson37 or chronic.Reference Brew, Lundholm, Viktorin, Lichtenstein, Larsson and Almqvist40,Reference Wang, Wen, Chiang, Lin and Guo45 Prior research has reported psychosis and atopic disorders clustering in individualsReference Chen, Lee and Lin70,Reference Pedersen, Benros, Agerbo, Børglum and Mortensen71 and families.Reference Pedersen, Benros, Agerbo, Børglum and Mortensen71 The familial link between mental and atopic disorders could arise as a result of shared aetiological factors but also as a result of the effects of adversity. Parental mental disorder increases risks for a range of adversities during childhood, such as povertyReference Fitzsimons, Goodman, Kelly and Smith72,Reference Wickham, Whitehead, Taylor-Robinson and Barr73 and trauma.Reference Björkenstam, Kosidou and Björkenstam74 Exposure to adversity and stress can alter a child's immune response,Reference Fagundes and Kiecolt-glaser75 which in turn increases the risk for atopy.Reference Sternthal, Enlow, Cohen, Canner, Staudenmayer and Tsang76 Yet, the link between mental disorders and atopy is likely complex and might include combinations of direct and indirect effects of parental disorder, as well as shared environmental factors (such as smoking).
The evidence for a link between parental mental illness and childhood overweight or obesity is inconclusive. Prior reviews that included cross-sectional studies did report a correlation between maternal depression or anxiety and childhood obesity;Reference Milgrom, Skouteris, Worotniuk, Henwood and Bruce19,Reference Benton, Skouteris and Hayden21 another that included only prospective designs found a link with chronic, but not with episodic (mostly postnatal), maternal depression.Reference Lampard, Franckle and Davison20 All of the studies we identified that investigated this relationship used prospective cohort or case–control designs, which generally had smaller samples than studies that used registry-based cohorts. Therefore, the lack of definitive findings may be due to lack of power of individual studies.
All the studies from LMICs reported a positive association between maternal depression and offspring malnutrition or diarrhoea. Half of these exclusively examined the effect of postnatal depression in the first year after birth. Recent reviews have reported rates of maternal postnatal depression in LMICs of around 20%,Reference Fisher, Cabral de Mello, Patel, Rahman, Tran and Holton77 considerably higher than the 6.5–12.9% seen in affluent Western populations.Reference Gavin, Gaynes, Lohr, Meltzer-Brody, Gartlehner and Swinson78 Evidence also suggests that maternal mental ill health is associated with both poor fetal and child growth.Reference Gelaye, Rondon, Araya and Williams79,Reference Surkan, Kennedy, Hurley and Black80 Therefore, maternal mental health has been highlighted as a priority target for screening mothers in these settings.Reference Walker, Wachs, Meeks Gardner, Lozoff, Wasserman and Pollitt81
Strengths and limitations
The review highlights how children of parents with mental illness have multiple physical health challenges, on top of previously identified mortalityReference Webb, Abel, Pickles and Appleby13,Reference King-Hele, Webb, Mortensen, Appleby, Pickles and Abel85 and neurodevelopmental risks.Reference Rasic, Hajek, Alda and Uher7,Reference Stein, Pearson, Goodman, Rapa, Rahman and McCallum86 Several strengths of our review reinforce the findings. First, although the different outcomes under consideration preclude direct comparison of estimates, there is general consistency in the effect sizes estimated, lending weight to the evidence. Also, we use strict definitions of study design; this means that the estimates are better approximations of effect sizes than if weaker study designs were used: i.e. cross-sectional studies. Therefore, we indicate where potential mechanistic explanations should be explored, so that modifiable factors might be identified to help this vulnerable group. The included studies come from heterogeneous samples, countries, settings, measures and designs. This strengthens the conclusions of the review because, as the results generally show consistency, we can assume that the findings are independent of these factors.
To date, this is the most comprehensive review and meta-analysis of the associations between parental mental illness and offspring physical illness during childhood, however there are a number of limitations. Few studies investigated the effect of both maternal and paternal mental illness and, generally, they noted weaker effects for paternal exposure.Reference Howard, Goss, Leese and Thornicroft36,Reference Magnus, Wright, Røysamb, Parr, Karlstad and Page41,Reference Klinnert, Nelson, Price, Adinoff, Leung and Mrazek42,Reference Wang, Anderson, Dalton, Wu, Liu and Zheng53 This indicates that maternal condition plays a more important role in the risk of offspring physical ill health, either as a result of intrauterine exposures, or through early childhood experiences or both; it also provides some evidence against a purely genetic cause for these relationships.
All the identified studies come from observational settings and are therefore subject to confounding bias. Most attempted to account for this bias, primarily using regression adjustment. Notably, most of the adjusted analyses were closer to a null effect size. This indicates that, overall, factors that increase the risk of parental mental illness are also likely to increase the risk of poor physical health in offspring (and vice versa). One obvious candidate for a variable of this kind is socioeconomic or multiple deprivation, inextricably linked with parental mental ill healthReference Elliott82 and poor child physical health.Reference Rough, Goldblatt, Marmot, Nathanson, Mansfield, Foyle and Nathanson18,Reference Bradley and Corwyn83
Despite adjustment for confounders, we must be cautious before attributing the causes of poor child physical health to parental mental disorders. First, not all the potential confounders are likely to be identified and measured in data available to researchers. Two of the more recent studies tried to account for residual confounding from familial factors by investigating whether associations still persist using sibling or cousin analysisReference Brew, Lundholm, Viktorin, Lichtenstein, Larsson and Almqvist40,Reference Magnus, Wright, Røysamb, Parr, Karlstad and Page41 or when investigating paternal exposure.Reference Brew, Lundholm, Viktorin, Lichtenstein, Larsson and Almqvist40 Second, some studies may be subject to overadjustment, where analyses adjust for variables on the pathway between the exposure and the outcome. For example, poor fetal growth is associated with both prenatal maternal mental disorderReference Lewis, Austin and Galbally84 as well as many health outcomes. Therefore, this might be the mechanism by which prenatal maternal mental disorder influences child health and studies that adjust for this might be underestimating the effect of maternal mental illness on child health. Third, although we excluded studies in which it was clear that the outcome was measured before the exposure occurred, for many studies this was unclear. Therefore, we cannot rule out that at least some portion of the results were because poor child health affects parental mental illness and not vice versa. Fourth, although we actively tried to include unpublished research, all the identified studies were from the published literature. Therefore, it might well be that some positive findings are the result of publication bias.
Implications for future research and policy
This systematic review shines a stark light on the gaps in our knowledge about the physical health of children whose parents have mental illness, and highlights a need to shift the focus of research towards parental mental disorders other than maternal postnatal depression. Future studies should interrogate the extent to which antenatal, perinatal and postnatal exposures have differential effects on offspring's risk of physical illness. Also, maternal mental disorder may pose more risk to child physical health than paternal disorder but there is a strong need for future studies to include paternal exposure where possible if we are to understand the mechanisms behind these effects in particular outcomes. This chimes well with recent calls for research and policy to place a greater emphasis on the role of fathers in children's lives.Reference Clapton87
Future research should interrogate the associations highlighted here to explore behavioural, environmental and genetic causes. To do this we highlight a need to develop fresh approaches to understanding the links between child physical health and parental mental health. These include the necessity of accounting for key confounders and to consider the use of alternative design strategies, including negative controls, sibling or quasi-experimental designs.Reference Lawlor, Tilling and Smith88 If future research is to be able to deepen our understanding of when and how these vulnerable children are at risk of preventable illnesses, large high-quality cohorts must be identified. For example, to investigate mechanisms of childhood asthma (still the most common childhood illness) we need to look at effects of different parental illnesses and effects of maternal versus paternal mental disorder. Combining data across such cohorts from different countries may offer such an opportunity.
Finally, from a policy perspective, such approaches can offer the detail needed to plan resource allocation and develop new service provision. Studies describing patterns of healthcare utilisation by these children and parents may be particularly valuable for this.
Funding
This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement no. GA682741) and the National Institute for Health Research Grant (grant reference 111905).
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2019.216.
eLetters
No eLetters have been published for this article.