Association of Gestational Weight Gain and Maternal Body Mass Index in Early Pregnancy With Risk for Nonaffective Psychosis in Offspring

Key Points

Question  Are gestational weight gain during pregnancy and maternal body mass index in early pregnancy associated with a risk for nonaffective psychosis in offspring?

Findings  In this population-based cohort study of 526 042 individuals born in Sweden from 1982 through 1989, extremely inadequate gestational weight gain was associated with a significantly increased risk for nonaffective psychosis in offspring in adjusted and sibling comparison models. A weak, U-shaped association was found between maternal body mass index at the beginning of pregnancy and risk for nonaffective psychosis in offspring in adjusted models.

Meaning  Insufficient weight gain during pregnancy may increase the risk for nonaffective disorders in offspring, even in an affluent and well-nourished population.

Importance  Prenatal exposure to famine is associated with a 2-fold risk for nonaffective psychoses. Less is known about whether maternal nutrition states during pregnancy modify offspring risk for nonaffective psychoses in offspring in well-fed populations.

Objective  To determine whether gestational weight gain (GWG) during pregnancy and maternal body mass index (BMI) in early pregnancy are associated with risk for nonaffective psychoses in offspring.

Design, Setting and Participants  This population-based cohort study used data from Swedish health and population registers to follow up 526 042 individuals born from January 1, 1982, through December 31, 1989, from 13 years of age until December 31, 2011. Cox proportional hazards regression models adjusted for socioeconomic status and potential risk factors were used to examine the risk for developing nonaffective psychoses. Family-based study designs were used to further test causality. Data were analyzed from February 1 to May 14, 2016.

Exposures  Gestational weight gain during pregnancy, maternal body mass index at the first antenatal visit, and paternal body mass index at the time of conscription into the Swedish military (at 18 years of age).

Main Outcomes and Measures  Hazard ratios (HRs) for the diagnosis of nonaffective psychoses (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10] codes F20 to F29 and International Classification of Diseases, Ninth Revision [ICD-9] codes 295, 297 and 298, except 298A and 298B) and narrowly defined schizophrenia (ICD-9 code 295 and ICD-10 code F20).

Results  The 526 042 individuals in the cohort (48.52% female and 51.47% male; mean [SD] age, 26 [2.3] years) included 2910 persons with nonaffective psychoses at the end of follow-up, of whom 704 had narrowly defined schizophrenia. Among the persons with nonaffective psychosis, 184 (6.32%) had mothers with extremely inadequate GWG (<8 kg for mothers with normal baseline BMI), compared with 23 627 (4.52%) of unaffected individuals. Extremely inadequate GWG was associated with an increased risk for nonaffective psychoses among offspring in adjusted models (HR, 1.32; 95% CI, 1.13-1.54) and in matched-sibling analysis (HR, 1.61; 95% CI, 1.02-2.56). Similar patterns were observed when considering narrowly defined schizophrenia as the outcome. Maternal mild thinness in early pregnancy was weakly associated with an increased risk for nonaffective psychosis in offspring (HR for BMI≥17.0 and <18.5, 1.21; 95% CI, 1.01-1.45), as was paternal severe thinness (HR for BMI<16.0, 2.53; 95% CI, 1.26-5.07) in mutually adjusted models. In matched-sibling analysis, no association was observed between maternal underweight (HR, 1.46; 95% CI, 0.90-2.35), overweight (HR, 1.11; 95% CI, 0.73-1.68), or obesity (HR, 0.56; 95% CI, 0.23-1.38) and risk for nonaffective psychosis in offspring.

Conclusions and Relevance  Inadequate GWG was associated with an increased risk for nonaffective psychosis in offspring, consistent with historical studies on maternal starvation. These findings support the role of maternal undernutrition in nonaffective psychosis pathogenesis.

Nonaffective psychoses, or schizophrenia spectrum disorders, are increasingly considered neurodevelopmental disorders.^1-5 Prenatal exposure to famine during the Dutch Hunger Winter (1944-1945) was associated with a 2-fold increase in the risk for nonaffective psychosis in offspring.^6-8 Similarly, exposure to prenatal famine during the Chinese Great Leap Forward (1959-1961) led to a 2-fold increased relative risk for schizophrenia.^9,10 Results from such disparate settings demonstrate that maternal malnutrition during pregnancy may increase the risk for psychosis among offspring.^11,12

Deficits in maternal nutrition during pregnancy, including micronutrient deficiencies (eg, folate, vitamin D, iron) and protein-caloric malnutrition, have been associated with abnormalities in offspring neurodevelopment.^11,13 Obesity, paradoxically, has been associated with deficiencies in nutrients vital to neurodevelopment, such as vitamin A, folate, vitamin D, and essential fatty acids,^14,15 and offspring risk for neural tube defects.¹⁶ A range of maternal nutritional states during pregnancy may contribute to the risk for psychoses in offspring. Khandaker et al¹⁷ reviewed several studies that used maternal body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared) as a proxy for maternal nutrition, but these produced conflicting results hampered by low numbers of cases.

This study’s aim was to investigate the association among maternal baseline BMI, gestational weight gain (GWG), and offspring risk for nonaffective psychosis in the largest cohort studied to date. We hypothesize that extremes in baseline maternal BMI or GWG, signifying suboptimal prenatal nutrition, would contribute to an increased risk for nonaffective psychosis in offspring. We posit that extremely low GWG is analogous to early gestational exposure to starvation seen in the famine studies. We used 2 family-based study designs—paternal-offspring comparisons and matched-sibling comparisons^18-20—to evaluate the weight of evidence for any observed associations.

This national, population-based cohort study used data from Psychiatry Sweden, a linkage of Swedish health and population registers.²¹ Ethical approval was granted by the Regional Ethical Committee of Stockholm. No informed consent was required for the analysis of anonymized register data.

The study population included all nonadopted individuals born in Sweden from January 1, 1982, through December 31, 1989 (n = 798 934), who were followed up from 13 years of age until December 31, 2011, for diagnoses of nonaffective psychoses.²¹ Children were excluded who died or emigrated before their 13th birthday (2.7%), had incomplete Medical Birth Register data (0.4%), were missing information on their biological father (0.5%), or were part of multiple births (1.9%) (eFigure 1 in the Supplement). In addition, 30.2% of eligible mother-child pairs lacked maternal BMI or GWG data. Those individuals excluded from the final study population were demographically similar to those included (eTable 1 in the Supplement).

Data on psychiatric history were taken from the National Patient Register, which has been collecting diagnoses for inpatient care since 1973 and psychiatric outpatient care since 2001. Nonaffective psychosis status was defined as receipt of 1 of the following diagnoses from International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), or International Classification of Diseases, Ninth Revision (ICD-9): ICD-10 codes F20 to F29 and ICD-9 codes 295, 297, and 298 (except 298A and 298B) (eTable 2 in the Supplement) before December 31, 2011.²¹ Narrowly defined schizophrenia (ICD-9 code 295 and ICD-10 code F20) was also considered as an outcome.

Maternal weight and height at the first antenatal visit were used to approximate baseline maternal BMI. Such data were recorded by midwives in the Medical Birth Register beginning in 1982. The timing of the first antenatal visit was unavailable. However, first trimester weight gain was on average low, and 90% of initial antenatal visits in Sweden occurred before 12 weeks’ gestation.^22,23 Weights of less than 40 kg or greater than 140 kg were censored as unrealistic or indicating an existing medical condition, as were heights of less than 140 cm or greater than 210 cm. Maternal baseline BMI values were categorized according to World Health Organization guidelines²⁴ into standard and extended BMI classifications.

Maternal weight was also recorded before delivery. Gestational weight gain was calculated as the difference in maternal weight between the first antenatal visit and delivery. Based on Institute of Medicine guidelines,²⁵ GWG was categorized as ideal, inadequate, or excessive according to the maternal baseline BMI categories of underweight (12.5-18.0 kg), normal weight (11.5-16.0 kg), overweight (7.0-11.5 kg), and obese (5.0-9.0 kg). The GWG categories of inadequate and excessive were divided at their respective medians (by BMI category) to create the following 5 extended GWG categories (eTable 3 in the Supplement): ideal, extremely inadequate, inadequate, excessive, and extremely excessive.

Paternal BMI was calculated from Swedish conscription register data, collected since 1969. Weight and height were measured objectively at the time of conscription into the Swedish military (at 18 years of age). Measurements were censored and categorized as with maternal BMI. Of the offspring with maternal BMI data, 64.74% also had paternal BMI data and are considered as the paternal BMI subcohort (eFigure 1 in the Supplement).

Covariates were considered as potential confounders based on current literature. The following covariates were included in the study and classified as in Blomström et al²¹: birth year, offspring sex,³ household income at birth (in quintiles, with highest quintile as the reference category),²⁶ highest level of parental education achieved, single-parent household status,²⁶ urban birth (child born in a municipality with ≥200 000 inhabitants in 1980),²⁷ parental immigration status (categorized as 0-2 parents born outside Sweden),²⁸ parent older than 35 years at the time of birth (0-2 parents),²⁹ parental nonaffective psychosis diagnosis (0-2 parents), and parental history of psychiatric care (0-2 parents).³⁰

Data were analyzed from February 1 to May 14, 2016. Statistical analyses were performed using STATA/IC software (version 14.1; StataCorp). We analyzed BMI as a categorical variable, then a continuous variable, using Cox proportional hazards regression to calculate hazard ratios (HRs) and 95% CIs for nonaffective psychosis (or schizophrenia) in offspring with robust SEs to account for clustering of observations with mothers. Offspring were followed up from 13 years of age until the diagnosis of nonaffective psychosis, emigration, death, or December 31, 2011, whichever came first. Basic HRs were adjusted for birth year and sex of offspring. The final model for BMI was adjusted for birth year, offspring sex, family income quintile, parent older than 35 years, parental history of nonaffective psychosis, and parent born outside Sweden. Parental educational level was not included in the model owing to colinearity with income quintile. Normal BMI was the reference category in categorical analyses. Categorical analyses were repeated with extended BMI categories.

For continuous analyses, we fit Cox proportional hazards regression models using restricted cubic splines with 5 knots. Restricted cubic spline models allow for the flexible fitting of nonmonotonic associations between variables.³¹ Postestimation xbrcspline³¹ was used, with the reference category set as BMI of 21.0, denoting minimal risk.²²

Similar to BMI, GWG was also analyzed as a categorical (reference category, ideal GWG) and continuous (reference category, 11 kg) variable. Models were adjusted as described above for BMI. Additional models considered maternal baseline BMI and gestational age as covariates. Continuous analysis was repeated with stratification by maternal BMI category.

Matched-sibling analyses comparing affected individuals with their unaffected full siblings were performed to investigate whether observed associations among maternal BMI and GWG and offspring nonaffective psychosis could be the result of confounding by shared familial factors. Narrowly defined schizophrenia was not considered as an outcome in sibling analyses owing to lack of power. Cox proportional hazards regression, stratified by family identity, was performed for matched full siblings, discordant on outcome, adjusted for birth order and sex. We considered BMI and GWG as categorical and continuous exposures as above.

Analyses of maternal and paternal BMI were repeated among those individuals with a paternal BMI observation (paternal BMI subcohort) and were analyzed individually as above and in a mutually adjusted model. Finally, the main analyses were repeated stratified by sex.

The 526 042 individuals in the study cohort (48.52% female and 51.47% male; mean [SD] age, 26 [2.3] years) included 2910 cases of nonaffective psychoses at the end of follow-up. As expected, offspring who developed nonaffective psychosis or narrowly defined schizophrenia were more likely to be male and to be born in an urban center or to a single parent, an immigrant, or a parent with a history of psychiatric care compared with unaffected offspring (Table 1 and eTable 4 in the Supplement). In the paternal BMI subcohort, offspring were less likely to have a foreign-born parent compared with the full cohort and were less likely to have a parent 35 years or older, because the conscription register began in 1969 (eTable 1 in the Supplement). Covariates by maternal BMI and GWG categories are presented in eTables 5 and 6 in the Supplement.

In the categorical analysis (Table 2), offspring of underweight mothers displayed a somewhat increased risk for nonaffective psychosis (adjusted HR, 1.14; 95% CI, 1.00-1.30). In the analysis of extended BMI categories, offspring of mothers with mild thinness (BMI≥17.0 and <18.5) had an increased risk for nonaffective psychosis (adjusted HR, 1.21; 95% CI, 1.06-1.39), as did offspring of mothers with class 2 obesity (BMI≥35.0 and <40.0; adjusted HR, 1.93; 95% CI, 1.00-3.71).

Similarly, in continuous analysis of maternal BMI (Figure 1), we observed a U-shaped association between maternal BMI and nonaffective psychosis among offspring in crude or adjusted models, although with wide CIs. Maternal underweight was usually associated with schizophrenia; no association was apparent between elevated maternal BMI and schizophrenia risk (Figure 1). In matched-sibling analyses, we found little apparent association between maternal BMI and offspring nonaffective psychosis in categorical (Table 2) or continuous analysis (Figure 1).

Normal-weight and underweight mothers were more likely to gain weight within their recommended ranges (44.59% and 56.52%, respectively), compared with overweight (26.91%) and obese (34.20%) mothers (eTable 7 in the Supplement). Broad GWG categories were not associated with offspring nonaffective psychosis (Table 3). For extended GWG categories, the offspring of mothers with extremely inadequate weight gain had an increased risk for nonaffective psychosis (adjusted HR, 1.36; 95% CI, 1.16-1.58), even after accounting for gestational age and maternal BMI (Table 3).

In continuous analysis, nonaffective psychosis risk in offspring was associated with a low GWG (<11 kg) in unadjusted (Figure 2) and adjusted (eFigure 2 in the Supplement) models. Similar patterns were observed for narrowly defined schizophrenia (Figure 2 and eFigure 3 in the Supplement). When stratified by baseline maternal BMI category, the association of low GWG and the risk for nonaffective psychosis remained for the normal-weight and overweight or obese groups (BMI≥25.0), but not for the underweight group.

In the matched-sibling analysis, the risk for nonaffected psychosis increased for offspring born to mothers with extremely inadequate GWG (HR, 1.61; 95% CI, 1.02-2.56) (Table 3). In continuous analysis, a similar finding was observed, although the CIs were wide and included 1 (Figure 2).

In the analysis of paternal BMI, we observed an increased risk for nonaffective psychosis among the offspring of severely thin fathers (BMI<16.0; adjusted HR, 2.52; 95% CI, 1.26-5.04) (eTable 8 in the Supplement) and a weak, U-shaped association between paternal BMI and nonaffective psychosis in offspring in continuous analysis, a pattern comparable with that of maternal BMI (eFigure 4 in the Supplement). The association between maternal obesity and the risk for nonaffective psychosis in offspring was strengthened by adjusting for paternal BMI, although the CIs remained wide (eFigure 4 and eTable 8 in the Supplement). In mutually adjusted models, maternal mild thinness in early pregnancy was weakly associated with an increased risk for nonaffective psychosis in offspring (HR for BMI≥17.0 and <18.5, 1.21; 95% CI, 1.01-1.45), as was paternal severe thinness (HR for BMI<16.0, 2.53; 95% CI, 1.26-5.07).

After stratification by sex (eTables 9 and 10 in the Supplement), maternal mild thinness was associated with nonaffective psychosis in male (HR, 1.30; 95% CI, 1.10-1.55) but not female (HR, 1.14; 95% CI, 0.91-1.43) offspring. Extremely inadequate GWG was associated with risk for psychosis in male (HR, 1.32; 95% CI, 1.07-1.63) and female (HR, 1.60; 95% CI, 1.27-2.01) offspring but was more pronounced in the latter.

Extremely inadequate GWG was associated with an increased risk for nonaffective psychosis in offspring in categorical and continuous analyses, even after adjustment for potential confounders. The sibling analysis suggests that this result is unlikely to be attributable to unmeasured familial confounding. Together these results indicate, similarly to the Dutch Hunger Winter and Great Leap Forward studies,^7-9 that inadequate maternal nutrition during pregnancy increases the risk for nonaffective psychosis in offspring, even in the context of an affluent and well-nourished population.

This study also demonstrated a weak U-shaped association between maternal BMI at the beginning of pregnancy and the increased risk for nonaffective psychosis in offspring. However, the results of the paternal comparison indicate that the associations between the risk for nonaffective psychosis in offspring and parental BMI may be partly attributable to genetic or other shared familial factors.

Previous studies have reported contradictory associations between maternal BMI and risk for psychosis in offspring.¹⁷ Low late-pregnancy maternal BMI (≤24.0) was associated with a 3-fold risk for schizophrenia in offspring (reference BMI>30.0).³² However, late-pregnancy maternal BMI was used as a combined proxy for prepregnancy BMI and GWG, obscuring the contribution of each.

High maternal prepregnancy^33,34 and late pregnancy³⁵ BMI have also been linked to psychosis in offspring. We observed a 2-fold increased risk associated with maternal obesity (class 2) and an elevated risk for obesity and its subclasses. The results of our family comparison studies indicate that the associations between maternal BMI and the risk for nonaffective psychosis in offspring are at least partly confounded by genetic or other shared familial factors. The only study to date considering a potential correlation between genetic determinants of BMI and the risk for schizophrenia reported an inverse association: evidence of genetic correlation between low BMI and risk for schizophrenia.³⁶

We observed a greater effect for female offspring of mothers with extremely inadequate GWG compared with male offspring, but effect estimates overlapped between the sexes. The initial Dutch famine study reported a similar sex effect,⁸ but this difference was not apparent in a later, more rigorous analysis.⁶ Our result merits cautious interpretation owing to the relative youth of our cohort and the differential age at onset for male and female offspring.³⁷

Although other mechanisms cannot be ruled out based on these observational studies, the association of inadequate GWG with nonaffective psychosis in concert with the findings of the Dutch and Chinese famine studies implicates malnutrition as the effector. Multiple nutrient deficiencies have been demonstrated to affect neurodevelopment and risk for schizophrenia in offspring.^4,11,38 Low GWG during pregnancy may therefore represent an inability to meet the nutrient demands of the placental-fetal unit. Suboptimal nutrient status of mothers with extremely inadequate GWG in our study is evidenced by their lower rates of male births (eTable 10 in the Supplement). Male fetuses place a higher energy demand on mothers during pregnancy,³⁹ such that states of maternal deprivation lead to fewer male births.⁴⁰ Fetal growth restriction, indicated by infants who are small for gestational age, can result from inadequate GWG, although other factors also contribute. Small size for gestational age has also been linked to an increased risk for nonaffective psychosis.^41,42

Severely inadequate GWG may otherwise be indicative of an existing maternal medical condition, such as endrocrinologic disorders, malabsorption, anorexia nervosa, bulimia nervosa, or hyperemesis gravidarum. Further research is necessary to understand the association between conditions that lead to insufficient maternal weight gain and the risk for nonaffective psychosis in offspring. Insufficient weight gain can also occur in otherwise healthy individuals owing to insufficient medical guidance or by a drive to conform to societal (but not medical) standards of appropriate weight gain.⁴³

This study is, to our knowledge, the largest to date to examine the association of maternal BMI and the risk for psychosis in offspring and the first to examine the role of GWG. The large sample size facilitated the use of family-based study designs for more rigorous inference of causation.^20,44 We calculated BMI and GWG from objectively measured, prospectively recorded register data. Swedish registry data on nonaffective diagnoses have high validity.⁴⁵

We were able to adjust GWG for gestational age. Gestational weight gain and gestational age are highly correlated, with inadequate GWG associated with preterm births and low birth weights.^23,46 Accounting for parental psychosis likewise strengthened our results. Nonaffective psychoses are highly heritable,^47,48 and traditional antipsychotics can lead to pronounced weight gain.⁴⁹

One issue is the limited follow-up time: ages of offspring at the end of follow-up varied from 22 to 29 years. Nonaffective psychoses manifest typically from the third decade of life onward.⁵⁰ As such, our sample is considerably right censored. Although we statistically accounted for this, we may have captured more early-onset, possibly phenotypically distinct cases, which limits generalizability.¹⁷ Future studies will allow for reanalysis as the cohort continues to age. Another limitation is the rate of missing BMI and GWG in the eligible study population, although these data seem to be missing at random from the Medical Birth Register.²²

Paternal BMI at conscription was used to examine the independence of any observable effect of maternal BMI on nonaffective psychosis among offspring.^18-20 Paternal BMI was unavailable for any later time points. Using paternal BMI at 18 years of age allowed for exploration of the contribution of paternal factors to nonaffective psychosis in offspring while removing the effects of the shared parental environment at the time of pregnancy. Increasing paternal age is related to increased BMI and increased risk for nonaffective psychosis in offspring. By capturing BMI at the same age for all fathers, we avoided potential confounding owing to these associations.²²

Sibling comparisons were used to test for unmeasured familial confounding. Using sibling analysis in extended categories for a rare outcome reduced power, possibly obscuring true associations. Also, in discordant sibling design, only mothers who varied in BMI or GWG between their 2 index pregnancies contribute to effect estimates; such designs are susceptible to confounding by nonshared factors that might lead to such changes in the same mother.⁴⁴

We improve on the Dutch and Chinese famine studies by using individual measures of parental BMI and GWG as proxies for nutrition in place of population-level measures of starvation. However, BMI and GWG are incomplete representations of metabolic health and nutritional intake and cannot discount other mechanisms of action. We were also limited by the small number of mothers at the extremes of BMI categories.

Last, we recognize the dissonance of applying Institute of Medicine 2009 GWG guidelines to mothers in the 1980s. Using the rationale of Holowko et al,⁵¹ we believe the guidelines represent optimal maternal and child health outcomes, regardless of advice at the time. Our study seemingly validates the 2009 guidelines for mothers with BMI in the normal range, because ideal GWG conferred the lowest risk. However, offspring of overweight and obese women showed an elevated risk for nonaffective psychosis, even at the lower end of the Institute of Medicine ideal GWG range, potentially raising the question of the adequacy of weight gain guidelines for populations outside the normal BMI range, specifically for nonobstetric outcomes.^46,52-54

Our results corroborate evidence from previous research and indicate that inadequate weight gain during pregnancy contributes to the risk for nonaffective psychosis in offspring. Weight gain outside Institute of Medicine guidelines may have deleterious effects on offspring neurodevelopment.

Corresponding Author: Renee M. Gardner, PhD, EoiOMH, Department of Public Health Sciences, Karolinska Institutet, Tomtebodavägen 18A, Stockholm, Sweden (renee.gardner@ki.se).

Accepted for Publication: December 16, 2016.

Published Online: February 22, 2017. doi:10.1001/jamapsychiatry.2016.4257

Author Contributions: Dr Gardner had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Dalman, Gardner.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Mackay, Dalman.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Mackay, Gardner.

Obtained funding: Dalman, Karlsson.

Study supervision: Dalman, Gardner.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by grant 523 2010 1052 from the Swedish Research Council for data linkage and staffing costs; by the Stanley Medical Research Institute for additional staff costs; and by grant 2007008 from the Stockholm County Council, grant 2007-2064 from the Swedish Council for Working Life and Social Research, grant 523-2010-1052 from the Swedish Research Council, and the Swedish Regional agreement on medical training and clinical research for data linkages and staff costs.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.