Prenatal ambient temperature and risk for schizophrenia

Abstract

Objective:

We conducted a systematic review of the published literature to test the hypothesis that maternal exposure to extremes of ambient temperatures during pregnancy is associated with the risk for psychiatric disorders or congenital malformations in offspring, both of which are indicative of perturbations of fetal neurodevelopment.

Method:

This study was conducted in accordance with the recommendations outlined in the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting proposal. Electronic databases (Ovid MEDLINE, Ovid Embase, Ovid PsycINFO, Ovid Global Health, Web of Science, and Cochrane Library) were searched. Four independent reviewers selected studies with the following criteria: (1) prenatal maternal ambient temperature exposure; (2) outcome of offspring psychiatric disorder or congenital defects; (3) empirical study; (4) full-length article, no conference presentations or abstracts.

Results:

Twenty-two studies met criteria and one was added from a reference list (n=23). Of these, schizophrenia (n=5), anorexia nervosa (n=3) and congenital cardiovascular malformations (n=6) studies were the most common. Each of these categories showed some evidence of association with an early pregnancy maternal ambient heat exposure effect, with other evidence for a late pregnancy cold effect.

Conclusion:

Some evidence supports a role for early pregnancy maternal exposure to extreme ambient heat in the development of psychiatric disorders, but large-scale, prospective cohort data on individual births is essential. Optimal studies will be conducted in seasonally variable climates, accounting for actual maternal residence over the pregnancy and at parturition, local environmental temperature records, and appropriate covariates, similar to studies identified by this systematic review for congenital malformations.

1. Introduction

Nearly a century ago, (Tramer, 1929) first observed that the birth pattern for individuals with schizophrenia was not distributed over the seasons as it was for the general population. His seminal paper and many subsequent articles pointed to a season of birth effect whereby individuals with schizophrenia were more likely to be born in the winter and early spring. Several underlying mechanisms of this phenomenon are proposed including seasonal reproductive proclivities, differences in light and vitamin D availability, infectious cycles, external toxins, and even meteorological influences (Cheng et al., 2013; Torrey et al., 1997).

One relatively understudied but highly plausible mechanism is ambient temperature. Although hyperthermia was the first teratogen identified in humans (Edwards, 1986), the full scope of adverse pregnancy outcomes associated with environmental temperature extremes in humans is only recently being illuminated, as most of the extant research employed animal models. This data is crucially needed as the global average temperature has risen by approximately 0.5 – 1°C since the 1970s and may increase by another 1.4–5.8°C by 2100 (Environmental Panel on Climate Change) (Hayhoe et al., 2018).

For most winter and spring births, conception and early gestation occur over the prior year’s spring or summer. All brain developmental processes are temperature dependent so they can be perturbed by elevated or depressed maternal-fetal temperatures. While fetal and maternal temperatures are highly correlated, high fetal metabolic activity causes the intrauterine temperature to be higher than the maternal temperature (Mann, 1968). Fetal growth during the first trimester is more dependent on the rate of blood flow rather than on nutrient concentrations in the intervillous blood (James et al., 2017). With a rise in ambient heat, more blood is shunted to the maternal skin surface to aid in heat dissipation, diminishing circulation to the fetus and vital maternal organs, with dehydration from heat stress further compromising uterine blood flow. Fetal development may additionally be sensitive to ambient heat because the structure of proteins or enzymes can be influenced, altering the kinetics of neuronal migration, with temperature dependence of RNA expression also playing a role (Lin et al., 2014). A circumscribed exposure to extreme heat in a pig model produced reduced placental efficiency and expression of glucose transporter and cationic exchange molecules, with lasting effects on muscle development (Zhao et al., 2020). Prior to conception, elevated temperatures can impact sperm quality, particularly among smokers (Harlev et al., 2015) and for men of older ages with reduced DNA repair enzymes in seminal fluid (Malaspina et al., 2015), in turn affecting offspring health. The consequences of disrupted development include congenital malformations and minor perturbations evident at parturition as birth defects, or psychiatric disorders presenting in later life.

Not only is climate change producing elevations in the overall global temperature, but is also producing extreme heat waves and extreme cold periods. Cold temperatures in late gestation or soon after birth could also be relevant, and one or both factors could be relevant. Understanding the effects of prenatal ambient temperature extremes thus emerges as an urgent priority for public health, obstetrics and indeed, psychiatry (Malaspina et al., 2020).

Here we conducted a systematic review to examine studies that documented temperature extremes during pregnancy and offspring outcomes. Specifically, we tested the hypothesis that early pregnancy ambient heat associates with greater psychiatric disorders including schizophrenia and congenital malformations.

2. Methods

This study was conducted in accordance with the recommendations outlined in the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting proposal (Moher et al., 2009).

2.1. Literature Search

A comprehensive search strategy using both keywords and index terms was created by a Medical Librarian (SW) and executed in the Ovid MEDLINE, Ovid Embase, Ovid PsycINFO, Ovid Global Health, Web of Science, and Cochrane Library databases on March 31, 2020. The search was designed to yield all observational studies on congenital defects and psychiatric disorders or neurodevelopment associated with ambient temperature or extreme ambient temperature fluctuations during pregnancy in humans. No limits were placed on date or language. The complete Medline search strategy is available in Appendix A. Articles were included if they reported an observational study that recorded ambient temperature during gestation as an exposure and congenital malformation or psychiatric disorder in a patient population as an outcome. Studies in healthy populations that examined variability in subthreshold symptoms, cognitive ability, temperament, traits, or behaviors (e.g. restrictive eating behavior) were excluded. Literature reviews, editorials, case reports, and conference proceedings also were excluded.

2.2. Study Selection, Risk of Bias, and Data Extraction

Study selection was conducted independently by a group of reviewers (JP, AA, DM, JS) in two successive rounds using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). Results were first screened on the basis of their titles and abstracts. All results identified as potentially eligible on this basis were advanced to have the full text screen. Each stage required two reviewers to screen an article and conflicts were resolved by a third reviewer. A data extraction form was generated and the following data were abstracted from each study: title of paper, type of paper, country in which the study was conducted, corresponding author contact details, study funding source, possible conflicts of interest, aim of study, study design, start date, end date/duration, ambient temperature exposure, population description and comparison group, source/setting of the population, method of recruitment of participants or name type of records used, total number of participants, maternal age group, baseline characteristics, gestation period examined, ambient temperature category, how temperature was measured, whether air pollution was assessed, outcome names, time points measured, time points reported, outcome definitions, type of measurement (e.g. percentage, odds ratio, risk ratio), whether the outcome too is validated or standardized, relevant results and findings, subgroups (if any), loss to follow up (if any), other results reported, unit of analysis, statistical methods used and appropriateness of these methods, whether all systematic and random error was adjusted, confidence intervals, stengths, limitations, strategies to overcome limitations, key conclusions, and additional/relevant bibliography items. An investigator also independently assessed the quality of each study using the Newcastle-Ottawa Scale (Wells, 2013).

3. Results

The systematic review returned 2,604 records for screening (Figure 1). This included 359 studies on season of birth and psychiatric disorder, but only eight included environmental temperatures and met our criteria. Of these, five focused on schizophrenia, three on anorexia nervosa, and one on multiple outcomes including depression. The systematic review returned 13 studies on congenital malformations and environmental temperature, and one was added based on a reference list for a total of 14 studies. Six of these focused on cardiovascular defects; two each on neural tube, hypothyroidism, and overall birth defects, and one each on hypospadias and orofacial cleft. Information from the 23 selected articles is presented in Tables 1 (Schizophrenia), 2 (Anorexia Nervosa) and 3 (Congenital Malformations). Main findings pertinent to this review are discussed below.

Figure 1.

PRISMA Flow Chart

Table 1:

Survey of Studies Concerned with Ambient Temperature Exposure and Schizophrenia

First Author, Year Study design: size, location, years	Exposures and method of assessment	Outcomes	Results	Key findings	Newcastle-Ottawa Quality Assessment Scale Score
Boland, 2018 PMID: 29036387 Case control study, electronic medical record data for 10.5 million births across six sites in three countries (USA, South Korea, Taiwan)	Ecological study of 6 climate variables (mean sunshine hours, minimum temperature, maximum temperature, rainfall in inches, relative humidity, days of precipitation and 6 other environmental exposures by month, assembled from National Weather and Meteorological Agencies in the three countries.	12 prenatal exposures were considered for 133 disease outcomes; multiple conditions (e.g. ADHD, depression, diabetes)	-Depressive disorder significantly anti-correlated with sunlight (R=0.625, 95% CI, 0.452 to 0.753) and temperature (high temperature, R=0.645, 95% CI, 0.462 to 0.779; low temperature, R=0.651, 95% CI, 0.446 to 0.790).	Lack of sunlight during the first trimester was related to depressive disorder development as elevated sunlight exposure and higher temperature saw lower risk for depression. Exposure to lower low temperature, lower high temperature and higher carbon monoxide-levels predicted high relative risk for depression as well.	2/9
Hare and Moran, 1981 PMID: 7315486 Case control study, cases were those who were admitted to a psychiatric hospital in England and Wales between 1970–1977; controls were a random 1 % sample from the 1971 Census, of persons born in England and Wales (193,759) Case conditions and numbers: Schizophrenia, 17,217; affective psychosis 14,004; neurosis 43,657; and personality disorder (16,265).	Ecological study using mean monthly temperatures at sea level for England and Wales (as a single geographical unit) provided by the Meteorological Office. Authors calculated mean quarterly temperatures as the average of 3 months (ex: Jan-March, etc.)	Schizophrenia, affective psychosis, neurosis, and personality disorder	-Association tested between schizophrenia birth rate from first and second quarters of year using mean temperature from each quarter and the first half of the year over 30-year period -Second quarter birth rate shows a significant negative correlation with the temperature of the first quarter and the first half of the year -Birth rates for the first half of the year are similarly correlated with the temperatures of the first quarter and first half of the year	Spring birth rate correlated with winter mean temperature indicating influence of cold temperatures around time of birth instead of warm temperature in early pregnancy. Comparison of years with the coldest and with the warmest seasons showed the schizophrenia birth rate was consistently higher in the coldest years. Correlation between winter birth rate and previous summer temperature was notably positive despite failure to reach significance. No comparable relations were identified for affective psychosis, neurosis or personality disorder.	1/9
McNeil, 1975 PMID: 1217973 Retrospective study: a local sample of schizophrenic or paranoid cases (N=301) admitted in 1876–1946 and a nationwide sample of Swedish schizophrenics (N = 13, 440)	The monthly means of the daily temperature at 14:00 (2 p.m.) in Lurid (Hamberg, 1887–1918; KSVA, 1876–1948; SMHA, 1919–1949) were used to calculate the mean temperature of June-August. The 71 birth years were rank ordered by mean summer temperature at representative geographical locations.	Schizophrenia or paranoia	-A positive relationship between summer temperature and increased patient birthrates was not seen in the Malmo sample (all male, primary diagnosis of SZ or paranoia) – the warmest quartile did not have the highest patient birthrate For total Swedish sample, no positive linear relationship was found between temperature quartile and patient birthrates.	The hypothesis of a positive relationship between summer temperature and subsequent patient birthrates was not supported. No linear relationship was identified between the rank ordered yearly data by mean summer temperature and births of schizophrenia, in either the local Malmo sample or total Swedish sample. Neither was a positive linear relationship found between temperature quartile and patient birthrates.	0/9
Tatsumi, 2002 PMID: 119505452985 patients born between 1920 and 1965 in hospitals in Japan	Observed/expected ratio of births; spearman correlation analysis between the ratio and average monthly ambient temperature	Observed incidences and expected numbers of schizophrenia births within each month	-Distributions of winter births across the twelve months (January–December) were not significantly different between the observed and expected numbers (chi-square=7.54, df=11, p=0.75). -Incidence of schizophrenia births was significantly decreased in the warmest three months (=July-September) (–7%, chi-square = 4.47, df=1, p = 0.04), compared with expected number -In males, a statistical trend for a difference between the observed and expected number of SZ births was found across the twelve months of a year (chi-square = 17.31, df = 11, p = 0.09) -Decrease of summer births was more marked in male vs all other subjects with rates of −8, −10, and −14% in the warmest five, four and three months, respectively (chi-square = 6.53, 7.62, 10.63, df = 1, p = 0.01, 0.006, 0.001, respectively) -A trend towards increased winter births was also found in male subjects during the coldest five and three months (+6 and +8%, chi-square = 6.23 and 4.21, df=1, p = 0.01, 0.04)	No differences were found across the twelve months for the observed and expected numbers of cases. A decrease in summer births of those with schizophrenia were found especially among males. Evidence was found that cold temperatures could increase risk in males. Seasonality of births appeared more marked in males than in females.	0/9
Templer, 1980 Re-analysis of two studies (Templer et al. 1978 and Torrey et al. 1977)	Average monthly temperature-Data obtained from Torrey et al. who found the ratio of schizophrenic births/general population births for each of the 12 months for 19 states across 3 regions: New England, Midwest, and South-Ratios of schizophrenia/general population births across states were ranked and correlated with temperature rank of month; temperature rank of month was also correlated with rank of conception ratios-Rank of birthtemperature correlations were correlated with January temperature and rank of birth-conception correlations with July temperatures	Schizophrenia	-Significant positive correlation coefficients were demonstrated between mean monthly temperature and conception rate for those developing schizophrenia in each state (ranging from .43-.83 for nine states) - rank order correlation coefficient between rank of birth correlations and rank of number of significant birth ratios for each state was .48 (p<.025, 1-tailed); comparable conception correlation was .50 (p<.025, 1-tailed)	Higher temperature at conception were associated with more conceptions of offspring who developed schizophrenia. A negative relationship was found between number of schizophrenic births and temperature of month.	Re-analysis – N/A
Watson, 1984 PMID: 6691787 Retrospective cohort study, 3246 individuals with schizophrenia; 392 to 1599 in analyses from 1915–1959 and 1973–1978 in the USA	Average St. Cloud monthly temperature 1915–1959 from National Oceanic and Atmospheric Administration Environmental Service	Schizophrenia	-Chi square used to compare the number of winter and nonwinter schizophrenia births in sets of years with high and low temperature winters and summers. -One of the 26 analyses on married schizophrenics was significant at .05 (summer disease total x2 [1] for year of birth, 6.86; winter proportions for high- and low-disease years, .328 and .444, respectively); no additional tests were significant even at .10. -Five of the 26 single-schizophrenic analyses were significant at .05 and nine at .10	A greater winter birth seasonality effect was seen in the years directly following those marked by high levels of infectious disorders than in years directly following those with low incidences of these diseases only for unmarried SZ. Analyses comparing the strength of December to March seasonality trends immediately preceding and following particularly cold/warm summers and during particularly cold/warm winters did not result in significant findings.	0/9

Table 2:

Survey of Studies Concerned with Ambient Temperature Exposure and Anorexia

First Author, Year Study design: size, location, years	Exposures and method of assessment	Outcomes	Results	Key findings	Newcastle-Ottawa Quality Assessment Scale Score
Watkins, 2002 PMID: 12183938 Retrospective Study, 408 patients (259 diagnoses of anorexia nervosa, 149 other eating disorders) in the UK	Average monthly temperature at assumed time of conception was taken from meteorological records	Anorexia nervosa (early onset)	-Quarterly analysis identified April-June as the peak period in the birth pattern of patients with anorexia nervosa, but association failed to reach significance (x2=5.41; df=3; p=.144) -Mean temperature during the assumed month of conception for anorexia nervosa group was 101.1 C (SD= 4.82), versus 9.1 C (SD= 4/82) for the other eating disorders – this difference was significant (z=2.26; one-tailed p = 0.024)	Patients with anorexia nervosa were more likely to be born between April-June and more likely to be conceived during warmer months (July-September).	1/9
Willoughby, 2002 PMID: 12183942 Retrospective Study, 458 patients born in the UK and Australia between 1971 and 1990	Ambient temperature exposure was mean temperature during the month of assumed conception	Anorexia nervosa (including subtypes - restrictive, binge/purge)	-The chi-squared goodness of fit did not approach significance (x2 = 13.1; df = 11; p = .287) when assessing the monthly birth rates of Australians with anorexia nervosa as well as the seasonality effects in Australia based on quarterly birth patterns. - No significant associations between diagnosis and month of birth (x2 = 7.68; df-11; two-tailed p−.741) or quarter of birth (x2 = 4.45;df-3; two-tailedp-.217) -A proportionately lower number of restrictive anorexics (40/134; 30%) were conceived in cooler months (mean temperature ≤ 16.95 degrees C) relative to the number of binge/purge subtype anorexics born in warm months (23/54; 43%). This association was significant [x2(Yate's continuity correction)-2.81;df-1; one-tailed p−.047].	A significant link was found between temperature at the assumed time of conception and restrictive anorexia birth months for restrictive but not binge/purge anorexic subtypes in Australia were less likely to be conceived in relatively cool weather; i.e. more likely to be conceived in warm weather, supporting a temperature at conception hypothesis rather than a simple seasonal pattern of birth. There was little change across the year in the birth patterns of young people with anorexia nervosa in the southern hemisphere.	0/9
Waller, 2002 PMID: 12436015 Retrospective Study, 195 adult female anorexics, 117 being of the restrictive subtype and 78 being binge-purging subtype from three specialist eating disorders clinics in southeast England	Utilized Meteorological Office records of mean monthly temperatures for the central area of England.	Anorexia nervosa diagnosis, restrictive subtype and binge-purging subtype	-A significantly higher proportion of restrictive anorexics were born in the high-risk period (April through June) (chi-squared = 2.91, df = 1, one tailed p<0.5) and were more likely to be conceived in warm months (mean temperature > 14.1 degrees C) than cold or mild months (mean temperature <14.1 degrees C) -Among the 117 restrictive anorexics, 32 (27.4%) were born during the relatively warm months with only 12 (15.4%) of the 78 binge-purging anorexics born during the warm months -Restrictive anorexics born during the high-risk period (April – June) had slightly lower BMI (mean = 15.4, SD = 4.30) than the restrictive anorexics born during other months (mean = 15.9, SD = 3.23); this difference was not significant -No significant difference in BMI (Mann-Whitney z = .41, NS) between binge-purge anorexics born in April through June (mean = 17.3, SD = 2.12) and those born in other months (mean = 16.9, SD = 3.38) -Mean temperature at conception of the restrictive anorexics was 10.1 degrees C (SD = 4.80) while the mean for the binge-purge anorexics was slightly lower at 9.34 degrees C (SD = 4.21); difference was not statistically significant (Mann-Whitney z = 1.20, one-tailed p .12). -Restrictive eating attitudes (EAT-26 dieting and oral control scores) were significantly correlated with temperature at assumed conception but only among the restrictive anorexics.	Higher proportion of restrictive anorexics were born in the high-risk period - April through June (conceived July through September). Higher environmental temperature at assumed conception was more likely to be found in restrictive anorexics vs binge-purge subtype. Higher environmental temperature at assume conception was associated with more restrictive eating attitudes among restrictive anorexics only.	0/9

Table 3:

Survey of Studies Concerned with Ambient Temperature Exposure and Congenital Malformations

First Author, Year Study design: size, location, years	Exposures and method of assessment	Outcomes	Results	Key findings	Newcastle-Ottawa Quality Assessment Scale Score
Agay-Shay, 2013 PMID: 23739216 Registry-based retrospective cohort study, 135, 527 live and stillbirths in the Tel-Aviv region of Israel in 2000–2006	Average ambient temperature and extreme heat events (>90th percentile); data from automatic monitoring stations in Tel-Aviv region (half-hourly measurements), Israeli Ministry of Environmental Protection and the Israel Electric Corporation	Congenital heart defects	-After stratifying by season of conception, continuous exposure to average ambient temperature and maximum peak temperature (1°C increase) during the cold season increased the risk for multiple clinical heart defects [odds ratio (OR) 1.05, 95% confidence interval (CI): 1.00, 1.10 and OR 1.03, 95% CI: 1.01, 1.05, respectively]. A 1-day increase in extreme heat events increased the risk for multiple clinical heart defects (OR 1.13, 95% CI: 1.06, 1.21) and also for isolated atrial septal defects (OR 1.10, 95% CI: 1.02, 1.19).	Maternal exposure to extreme heat days during the cold season (weeks 3–8 postconception) increased the risk for multiple CHDs and isolated ASD cases	5/9
Aminzadeh, 2010 PMID: 20956721 Case control study, 47,075 births in Iraq from 2006–2008	Ambient temperature exposure (average monthly from meteorological station)	Suspected or confirmed congenital hypothyroidism (CH)	- The lowest proportion of births for those suspicious for CH and those with confirmed CH in summer were 19.9% and 13.4%; their highest frequencies in winter were 31.3% and 35.2%, respectively. -32.4% of cases born in warm period and 67.6% of cases were born in cold period -The traditional warmest (July/August) and coldest (January/February) months were the exact time of nadir (2.8%) and peak (15.5%) for CH. -The fifth month (July/August) was the warmest month at 28.3 – 50C (AMT=38.98C) and correlated with the lowest incidence of CH. -The tenth month (January/February) was the coldest month at 20.2 – 23.9C (AMT=11.15C) and correlated with the highest incidence of CH -Relationship between the occurrence of CH and average temperature at the time of birth (Spearman’s rho correlation method. r= 0.87; P< 0.001) -For each 1 degrees C increase in temperature at the time of birth, infants were 4% less likely to be a case of CH (OR 0.96, 95% CI 0.94 – 0.98; P,0.001).	The prevalence of CH was negatively correlated with temperature – the risk of delivering an infant with CH was significantly higher during the cold season and lower during the warm season regardless of sex. Geographical region is of interest due to extremely hot summers (mean temperature > 50 degrees C) and cold winters (mean temperature < 0 degrees C)	5/9
Auger, 2017 PMID: 27494594 Retrospective cohort study, 704,209 fetuses between 2–8 weeks post conception from April-September in Quebec, Canada, 1988– 2012	Exposure to maximum temperature ≥ 30°C; data from 18 meteorological stations representative of each health region in Quebec from Environment Canada	Congenital heart defects (both critical and noncritical)	-Fetuses exposed to 15 days of temperature ≥ 30°C between 2 and 8 weeks post conception had 1.06 times the risk of critical defects (0.67, 1.67) and 1.12 times the risk of noncritical defects (0.98, 1.29) relative to 0 days exposure. Associations strengthened progressively as more days of exposure to temperatures ≥ 30C.	First trimester extreme heat exposure may be associated with noncritical heart defects, particularly atrial septal defects.	6/9
Auger, 2017 PMID: 27881468 Retrospective cohort study, 3–4 weeks postconception for 887,710 fetuses in utero in Quebec between 1988–2012	Maximum daily/weekly temperature (historical weather data from meteorological station)	Neural tube defects: spina bifida & anencephaly/encephalocele	Compared to 20°C, a maximum daily temperature of 30°C was associated with 1.56 times the risk of any neural tube defect on day 5 (95% CI 1.04 to 2.35) and 1.49 times the risk of day 6 (95% CI 1.00 to 2.21) of the fourth week postconception which fell between April and September. The prevalence of spina bifida was slightly higher for maximum weekly temperatures of ≥30°C during the fourth week post-conception compared with 20–24.9°C [29.5 per 100 000 (95% CI 21.3 to 37.8) vs 25.0 per 100 000 (95% CI 18.2 to 31.7].	The risk of neural tube defects was weakly associated with elevated ambient temperatures between April and September – the prevalence of spina bifida was slightly higher for maximum weekly temperatures of ≥30°C during the fourth week postconception.	6/9
Gu, 2007 PMID: 17956160 Retrospective Study, 1,586 participants identified as having CH by neonatal screening in Japan from 1994–2003	Monthly average temperatures at 11 main observatories in Japan were obtained from the electronic database of the Japan Meteorological Agency	Congenital hyperthyroidism (at birth— neonatal screening)	-A statistically significant negative correlation was observed between monthly incidence and monthly average temperature among males (Spearman r = −0.84, n = 703, p < 0.001), females (Spearman r = −0.82, n = 876, p < 0.001), and total cases (Spearman r =− 0.89, n = 1586, p < 0.001).	Higher incidence of CH was associated with lower temperature in both males and females. The highest incidence of cases occurred when the average temperature was 5.4C (January). Warm temperature at birth were unrelated to CH, conception period was not addressed.	4/9
Judge, 2004 PMID: 15367322 Case control study, Cases and controls (845 cases, 1983 controls) were drawn from the population of all liveborn infants born in New York state (all 14 counties) from 1988–1991	Participants reported number of hours/week spent in ‘extreme cold (<0 degrees F)’, in ‘extreme heat (>100 degrees F)’, and number of times/week they used a ‘hot bath, hot tub or sauna’ during the ‘before’ and ‘early’ stages of pregnancy; data collected 3–8 years after infant’s birth	CCM (congenital cardiovascular malformations)	-No significant increased risk of CCM was found in mothers who reported exposure to extreme heat (> 100F) early in pregnancy (OR = 1.13,95% CI 0.59, 2.19) or exposure to extreme cold (< 0F) (OR = 1.19, 95% CI 0.66, 2.15). Reported use of a hot tub, hot bath, or sauna during early pregnancy had no increased risk of CCM in offspring (OR = 0.88, 95% CI 0.65, 1.18).	In early pregnancy, no significantly increased risk of CCM was found in mothers who reported exposure to extreme heat, exposure to extreme cold, or use of hot tub/hot bath/sauna.	9/9
Kilinc, 2016 PMID: 27567375 Retrospective case-control study, children whose gestational periods (8–14 weeks) occurred in the Ankara and Istanbul regions and who had hypospadias repair or other urological treatments between 2000 and 2015	The monthly maximum and average temperatures measured in Ankara and Istanbul were obtained from the records of the Turkish General Directorate of Meteorology	Hypospadias repair or other urological treatments (circumcision, urinary tract infection, pyeloplasty, nephrolithotomy, etc.)	-For patients with hypospadias, gestational weeks 8–14 occurred in in July (207 patients, 12.1%) and August (191 patients, 11.1%); the average and maximum monthly temperatures during this period were significantly higher (p = 0.01) -Compared to other seasons, the average and maximum monthly ambient temperatures during the summer increased the risk for hypospadias (OR, 1.32; 95% CI, 1.08–1.52; and OR, 1.22; 95% CI, 0.99–1.54, respectively).	High ambient temperature exposure during the period of hypospadias development (weeks 8–14 of gestation) may contribute to its increased incidence. The magnitude of mean monthly temperature in summer was significantly related to increased hypospadias risk.	4/9
Lin, 2018 PMID: 29886237 Case control study, 5,848 CHD cases and 5,742 controls in Arkansas, Texas, North Carolina, Georgia, New York, Utah, California, and Iowa from 1997–2007	Extreme heat events (EHEs) were defined by using the 95th (EHE95) or 90th (EHE90) percentile of daily maximum temperature and its frequency and duration at 3–8 weeks GA	Congenital heart defect	-Findings of 2.17–3.24 fold increased risk for VSD was found after cumulative exposure to EHE (3–11 days) during summer -A long duration of unseasonably warm weather in spring was associated with CHDs in multiple regions with ORs ranging from 1.23 to 9.62 (10 days of cumulative EHE90 for VSD).	Cumulative exposure to EHE in pregnant women may be associated with an increased risk for CHDs in spring.	9/9
Soim, 2017 PMID: 28766872 Case control study, Cases - 326 NTDs (210 spina bifida, 81 anencephaly, 35 encephalocele) and Controls – 1781 in the USA (10 states) from 1997 to 2007	Data obtained on daily maximum temperature (Tmax) in degrees F, dew point (F), wind speed (knots), and atmospheric pressure (millibars) for each site from the National Center for Atmospheric Research. Daily universal apparent maximum temperature (UATmax) was exposure	Neural tube defects: anencephaly, craniorachischisis, spina bifida, encephalocele	-No significant association between EHEs and NTDs was observed. Authors did not observe any statistically significant associations between NTDs and EHE95 or EHE90. Consistently elevated estimates observed for EHE95 in New York (Northeast), North Carolina and Georgia (Southeast), and Iowa (Upper Midwest) for two consecutive days of exposure to extreme heat events (temperature > 95th percentile UATmax) though findings were not significant.	No significant findings found	9/9
Soim, 2018 PMID: 30338937 Case control study, 907 OFC cases (294 CP cases and 613 CL +/−P cases) and 2,206 controls in the USA (8 states) from 1997 to 2007	Daily maximum temperature in degrees F, dew point, wind speed, and atmospheric pressure data was obtained from the National Centers for Environmental Information for each weather monitoring station. Maternal residential addresses from 3 months before conception through the end of pregnancy were geocoded centrally and linked with the closest weather monitoring station.	Cleft palate, cleft lip +/− palate	-Assessment of heat events (temperature > 90th percentile) and extreme heat events (temperature > 95th percentile) in first 8 weeks of gestation after conception -Significantly increased aORs [1.89 CI (1.11, 3.23)] were found in NC and GA (Southeast) among mothers who experienced 3-day long EHE95 but not 2- or 4-day long EHE95 -Compared to no exposure to EHE90, significantly increased aORs [1.70 CI (1.02, 2.81)] were observed among mothers who experienced 4-day long EHE90, but not 3 or 5-day long EHE90	Prolonged duration of EHEs may increase the risk of OFCs in some geographic locations. No statistically significant general association was observed between maternal exposure to EHEs and OFCs.	9/9
Stingone, 2019 PMID: 32091506 Case control study, 4,0333 controls and 2,632 cases with dates of delivery between 1999–2007 in the USA (6 states)	Meteorological data (daily temperature, dew point, wind speed, atmospheric pressure) for each participating NBDPS study center from the National Center for Atmospheric Research was garnered. Geocoded maternal residencies were linked with the closest weather station and assigned meteorological data from that station.	Congenital heart defects - conotruncal heart defects, left/right ventricular outflow tract obstruction defects (LVOTO and RVOTO, respectively), and septal defects, as well as further subgrouping for total perimembranous ventricular septal defects and atrial septal defects	-Compared with low PM2.5 exposure and no EHE exposure, the odds of a ventricular septal defect in offspring associated with high PM2.5 exposure was elevated only among women who experienced an EHE (OR 2.14 95% CI 1.19, 3.38 vs. OR 0.97 95% CI 0.49, 1.95; RERI 0.82 95% CI −0.39, 2.17) [The elevated odds observed for VSDpm (OR: 2.14; 95% CI: 1.19, 3.83) occurred for mothers with at least 1 day of early pregnancy in the summer season who experienced both high PM2.5 exposure and an EHE95 during early pregnancy] -Among those exposed to an EHE, women exposed to high levels of PM2.5 had 1.59 (95% CI 0.94, 2.71) times the odds of having a child with a VSDpm than women with low PM2.5 exposure -Among women unexposed to an EHE, the association between PM2.5 and VSDpm was estimated at 0.97 (95% CI 0.49, 1.95)	EHEs may contribute to the association between prenatal exposure to PM2.5 and CHD occurrence - the odds of VSD development among those with high PM2.5 exposure was elevated only among women who experienced an EHE	9/9
VanZutphen, 2014 PMID: 24407473 Case control study, 13,044 selected birth defect cases and 59,884 controls from New York State resident live births in 19922006	Authors divided New York state into 14 weather regions - each weather region was assigned a daily average value of temperature (minimum, mean, and maximum in 1C), barometric pressure, dew point, and wind speed by using all available airport data for that region. Universal apparent temperature was calculated using temperature, vapor pressure, and wind speed.	Birth defect cases were classified into the 45 birth defects categories; of these the following were excluded: chromosomal anomalies (Trisomies 13, 18, and 21), fetal alcohol syndrome, and case groups with fewer than 50 cases (anencephalus, encephalocele, anophthalmia/mi crophthalmia, aniridia, tricuspid atresia/stenosis, Ebstein's anomaly, biliary atresia, bladder exstrophy, and amniotic bands)	-For every 1 degrees C decrease in the average daily mean universal apparent temperature during the critical period (3–8 weeks post-conception), there was a significantly associated increase in the odds of coarctation of the aorta (OR 1.06, 95% CI 1.02–1.11) and reduced odds of hypoplastic left heart syndrome (OR 0.92, 95% CI 0.86–0.98) -Increased odds of coarctation of the aorta was associated with cold spell (OR 1.61, 95% CI 1.11–2.34) and number of extreme cold days (OR 1.04, 95% CI 1.01–1.08)	Positive and consistent associations were found between cold exposure indices during the critical period (3–8 weeks post-conception) and development of coarctation of the aorta.	7/9
Van Zutphen, 2012 PMID: 23031822 23031822 Population-based case-control study, 6,422 cases from the New York State Congenital Malformations Registry and 59,328 controls (10% random sample of live births) that shared at least 1 week of the critical period of embryogenesis in summer from 1992–2006 in New York, US	Universal apparent temperature, heat wave prevalence, number of heat waves, and number of days above the 90th percentile; data obtained from the National Center for Atmospheric Research, 14 regions of relatively homogeneous weather and ozone exposures	Congenital structural malformations (CNS, eye, cardiovascular, craniofacial, genitourinary, and musculoskeletal) rate	- A 5° F increase in mean daily minimum apparent temperature was significantly associated with higher rates of congenital cataracts (adjusted odds ratio (aOR) = 1.51: 1.14, 1.99) and lower rates of anophthalmia/microphthalmia (aOR = 0.71: 0.54,0.94). -Congenital cataracts were associated with heat waves (aOR = 1.97: 1.17–3.32), number of heat waves (aOR = 1.45” 1.04–2.02), and number of days above the 90th percentile (aOR = 1.09: 1.02,1.17).	Positive and consistent associations were found between multiple heat indices during the critical period of embryogenesis (weeks 3–8 post-conception) and congenital cataracts.	7/9
Zhang, 2019 PMID: 30696385 Case control study, 5742 Controls, 5848 Cases in the USA (10 states) from 1997–2007	Daily maximum temperatures were assessed using hourly temperature estimates within grid cell (accounted for 36×36-km regions). Exposure period average heat exposure during 3–8 weeks post conception for pregnant women with at least 1 day of this post conception period overlapping with spring or summer. For each pregnancy, the count of excessively hot days (EHD) was the number of days with Tmax exceeding either the 90th (EHD90) or 95th (EHD95) percentile for the same season of the baseline period. Frequency and duration of extreme heat events were evaluated as well. The frequency of extreme heat events (EHE) were described as the number of occurrences of at least 3 consecutive EHD90 days (EHE90) or 2 consecutive EHD95 days (EHE95). Duration of EHE was described as the number of days for the longest EHE within the 42-day period.	Congenital heart defects	-Midwest found to have the highest potential increase in summer maternal exposure to excessively hot days (3.42; 95% CI, 2.99–3.88 per pregnancy), heat event frequency (0.52; 95% CI, 0.44–0.60) and heat even duration (1.73; 95% CI, 1.49–1.97) -In the South, potential increases in frequency of exposure to spring EHE95 may be related to a 12.3% (95%CI, 5.9%–18.9%) increase in the number of total CHDs, a 19.7% (95% CI, 7.4%– 33.5%) increase in conotruncal heart defects, and an 18.9% (95% CI, 6.7%–32.6%) increase in VSD. -Large increases in specific CHD subtypes during spring were found, including a 34.0% (95% CI, 4.9%–70.8%) increase in conotruncal CHD in the South and a 35.3% (95% CI, 6.2%–72.1%) increase in VSDs-Duration of similar exposures in Northeast may result in a 38.6% (95% CI, 9.9%–75.1%) increase in ASDs, a 17.9% (95%CI, 7.2%–30.1%) increase in septal heart defects, and a 23.4% (95% CI, 9.9%–39.0%) increase in VSDs. -In the Northeast, septal defects may increase by 33.9% (95% CI,14.9%–56.8%) due to maternal exposure to increased number of excessively hot days; similar changes were projected with EHE90.	Maternal heat exposure is projected to increase across the US (projection period: 2025–2035) especially in the Midwest. The South and Northeast regions were found to have the largest number of projected CHD increases related to frequency of exposure to elevated temperatures especially during the spring	8/9

3.1. Schizophrenia

Of the five studies focusing on schizophrenia, only one reported a link to prior summer mean temperature (Templer, 1980). This U.S.-based study spanned 19 states across three geographic regions (New England, Midwest, South) and 26 years. Significant positive correlation coefficients were demonstrated between mean monthly temperature and conception rate for those developing schizophrenia in each state (ranging from .43 – .83 for nine states), suggesting that higher temperatures at conception, which in the U.S. occur in summer, were associated with more conceptions of offspring who developed schizophrenia.

Four other studies did not find significant effects of ambient temperatures. First, a Swedish study performed a quartile split across 71 years of mean summer temperatures (June-August) (McNeil et al., 1975). If a fetus spent any amount of time in utero during a given summer (e.g. one day to three months), regardless of trimester, and based on a study-wide assignment of 38 weeks gestational age at birth, that summer’s mean temperature was assigned to that fetus. Chi-square analysis of schizophrenia birth incidence across the four quartiles was not significant.

A second study from England and Wales examined mean temperatures for each quarter and the first half of the year over a 30 year period finding a significant negative correlation between the second-quarter birthrate of persons with schizophrenia and the mean temperature of the first quarter of the year. Second quarter (April – June) birth rate for schizophrenia was increased by colder first quarter (Jan – March) temperatures, with no effect of warmer temperatures in early pregnancy (Hare and Moran, 1981). Although it did not reach significance, the correlation between the first quarter birth rate and previous third quarter temperature was notably positive.

Third, a U.S. study in Minnesota used 2 × 2 chi-square test to compare schizophrenia birth rate in winter (December to March) to a ‘control period’ defined as the flanking four months combined and the 10 warmest vs. 10 coolest summers (June to August) across a 44 year period (Watson et al., 1984). The same test was run for the 10 warmest vs. 10 coolest winters (December to February). These tests yielded no significant results. Fourth, a Japanese study correlated the monthly birth rate for schizophrenia verses the monthly mean temperature across 45 years and only identified an inverse relationship in males (Tatsumi et al., 2002). These results also demonstrate an influence of cold temperatures around the time of birth. Unfortunately, this study did not analyze winter/spring birth rates in association with prior summer months so information on an influence of warm temperature in early pregnancy was not available.

3.2. Anorexia Nervosa

Three articles focusing on anorexia-nervosa and ambient elevated temperatures supported early pregnancy heat as a factor for the season of birth effect. First, a retrospective chart review from the United Kingdom defined month of conception as nine months prior to birth, finding that those with restrictive anorexia were significantly more likely to be conceived in warm months (mean temperature > 14.1°C) than in cold- or mild months (mean temperature ≤ 14.1°C), relative to those with a binge-purge subtype of anorexia (Waller et al., 2002). The same pattern held in a comparison between early-onset anorexia and other eating disorders (e.g. bulimia nervosa, food avoidance emotional disorder, pervasive refusal, or selective eating) (Watkins et al., 2002). A comparable pattern was found in an Australian sample showing that those with restrictive anorexia were less likely to be conceived in the cooler months (mean temperature ≤ 16.95°C) relative to the hot months (mean temperature > 16.95°C) (Willoughby et al., 2002). Additionally, conception month mean temperatures correlated positively with self-report ratings of dieting and oral control behaviors in restrictive anorexia patients (Waller et al., 2002).

3.3. Congenital Malformations

Congenital malformation research frequently targets the very early pregnancy period (2 – 8 weeks postconception) as this is the critical period for organogenesis. With respect to congenital cardiovascular defects, our systematic review indicates a relationship between very early pregnancy exposures and elevated ambient heat. First a retrospective cohort study in Tel-Aviv, Israel, examined the cold season, a time with relatively mild average temperatures in the Mediterranean region (September 23^rd to March 30^th), and found a small continuous increase in the risk for multiple congenital heart defects associated with early pregnancy exposure (3 – 8 weeks post conception) to maximum peak temperature (per 1°C increase) (Agay-Shay et al., 2013). In the same season, slightly stronger risks for multiple congenital heart defects and isolated atrial septal defects were observed in the context of so-called extreme heat event exposure (i.e. number of days of daily average temperature > 24.4°C). A second retrospective cohort study in Quebec, Canada, found an association between 15 days of extreme heat event exposure (i.e. daily maximum temperature ≥ 30°C) during the 2^nd through 8^th weeks of gestation in spring and summer (March through September) and atrial septal defects and other noncritical defects, compared to offspring with 0 days of exposure (Auger et al., 2017b). Analyses by week revealed an emergence of the association in postconception week 3.

The U.S. National Birth Defects Prevention Study used a case-control design across multiple geographically diverse states to test the effect of 3 to 11 days of extreme heat event exposure (at least two consecutive days of maximum daily temperature above 95^th percentile for the season/year or at least three consecutive days of maximum daily temperature above the 90^th percentile for the season/year) between the 3^rd and 8^th weeks postconception in spring and summer on congenital heart defects (Lin et al., 2018). Five to 11 cumulative summer days of 90^th and 95^th percentile extreme heat exposures and 10 cumulative spring days of 90^th percentile extreme heat exposure were associated with ventricular septal defects in a study-wide assessment incorporating all states. Further, spring time 95^th percentile extreme heat event exposures were associated with ventricular septal defects in the South (Arkansas, Texas) and Northeast (New York State) and atrial septal defects in the Northeast, consistent with the studies above. From the same cohort, this particular risk in the Northeast in spring is predicted to increase by 38.6% in 2025–2035, among other risk increases across spring and summer in other regions (Zhang et al., 2019). Also from the same cohort, high air pollution (PM_2.5) interacted with an extreme heat event exposure on ventricular septal defects (Stingone et al., 2019).

Finally, an earlier case-control study utilizing the New York Congenital Malformations Registry across 14 regions in New York that did not show associations between maternal ambient heat exposure and congenital heart defects only used maternal self-report data on number of hours per week spent in > 100°F in the first trimester collected 3–8 years after the infant’s birth (Judge et al., 2004). This study also did not include septal defects as a congenital heart defects outcome.

Other birth defect study results returned by the systematic review had fewer studies. For example, a prospective cohort study in southwest Iran demonstrated an association between the risk for congenital hypothyroidism and ambient temperature (Aminzadeh et al., 2010), as only 32.4% of cases were born in the warm period whereas 67.6% were born in the cold period. The odds of congenital hypothyroidism increased by 4% for every 1°C decrease in monthly mean temperature. This geographical region is of interest due to its extremely hot summers (mean temperature > 50°C) and cold winters (mean temperature < 0 °C). A Japanese retrospective study also found a higher incidence of congenital hypothyroidism in association with lower temperature at birth, with highest incidence occurring during January (mean temperature = 5.48°C) (Gu et al., 2007). Although warm temperatures at birth were unrelated to congenital hypothyroidism, temperatures in the conception period were not addressed in this study.

Another retrospective case-control study identified by the systematic review considered the risk for hypospadias in Ankara and Istanbul, Turkey, from ambient temperature exposure between 8 and 14 weeks gestation when this defect is thought to occur. Offspring with hypospadias were more likely to have their critical gestational window occur in July and August than the comparison patients having other urological treatments (Kilinc et al., 2016). Moreover, the mean monthly temperature in summer was significantly related to increasing hypospadias risk.

Early pregnancy ambient heat exposure was also associated with oro-facial cleft malformations in the data of the U.S. National Birth Defects Prevention Study (Soim et al., 2018). This investigation assessed extreme heat events (at least two consecutive days of daily universal apparent temperature (calculated using temperature, dewpoint, wind speed and atmospheric pressure) maximum above the 95^th percentile for the summer/year. Alternatively, effects of at least three consecutive days of daily universal apparent temperature maximum above the 90^th percentile for the summer/year in the first 8 weeks postconception were assessed. Although, no overall effects were found, increased risk for oro-facial abnormalities in the Southeast (North Carolina, Georgia) was identified for 3 versus 0 days of extreme heat (daily universal apparent temperature maximum > 95^th percentile) and for 4 versus 0 days of extreme heat (daily universal apparent temperature maximum > 90^th percentile).

The two studies selected by the systematic review on neural tube malformations returned mixed findings. First, relative to 20° C, exposure to 30° C daily temperature maximum in day 5 of the 4^th gestational week postconception, between April and September in the Quebec, Canada cohort was significantly associated with the risk for neural tube malformations (Auger et al., 2017a). However, in the U.S. National Birth Defects Prevention Study, elevated but not significant risks for neural tube defects were related to extreme heat exposure (at least two consecutive days of daily universal apparent temperature maximum > 95^th percentile for the summer/year) in the 3^rd and 4^th week postconception in the Northeast (New York), Southeast (North Carolina, Georgia) and Upper Midwest (Iowa) (Soim et al., 2017).

Finally, a study encompassing 14 regions in New York State linked data from the State Congenital Malformations Registry with temperature records, found evidence that extreme heat events (at least three consecutive days of daily universal apparent temperature mean > 90^th percentile over summer) in early pregnancy (3–8 weeks of gestation) predicted congenital cataracts, with no effects observed for cardiovascular, oro-facial cleft and neural tube (Van Zutphen et al., 2012). In the same cohort, exposure to extreme cold (i.e. at least three consecutive days of daily universal apparent temperature mean < 10^th percentile over winter) was associated with increased risk for aortic coarctation, but not for other malformations (Van Zutphen et al., 2014).

3.4. Other

One article returned by the systematic review did not fit into a clear offspring outcomes category. This case-control study utilized 10.5 million electronic medical records across six sites and three countries (U.S., South Korea, Taiwan) to examine the association between cumulative exposure to multiple climate factors, pollutants and influenza-like illness in individual pregnancy trimesters and the relative risk by birth month for over 133 disease outcomes (Boland et al., 2018). The disease outcomes were selected such that each site must have at least 1,000 cases of the disease. Patient median age was 35–53 years old, and all patients were assumed to have 38.5–39.17 gestational age at birth depending on the mean for that country. First trimester exposures to lower low temperature, lower high temperature, fewer sunshine hours and higher carbon monoxide-levels predicted higher relative risk for depression by birth month. The study did not examine schizophrenia and anorexia nervosa as outcomes.

4. Discussion

This systematic review assessed the available literature linking maternal environmental extreme temperature exposures in pregnancy to offspring psychiatric outcomes and congenital malformations. The literature has supported strong season of birth effects, but has overall lacked the methodological rigor to evaluate if maternal exposures to extreme environmental temperatures during critical gestational periods could account for the effect.

Nonetheless, a vulnerable period in the first trimester (gestational age 3 – 8 weeks) was identified for high ambient heat exposure in association with congenital malformations, which are detectable at birth, and for some psychiatric outcomes that emerge later in life. Elevated ambient heat exposure, most likely above a temperature threshold that a particular pregnant woman can physiologically manage, may increase the intrauterine temperature, perturbing protein conformations, enzymatic activity, cellular proliferation and neuronal migration. This information is critically important as climate change is warming the earth and also producing periods of excessive heat and excessive cold.

4.1. Schizophrenia

Data demonstrating excess schizophrenia births in the late winter and early spring date back nearly a century (Tramer, 1929), with many dozens of replications since (Torrey et al., 1997), and with 10% of all cases attributable to the season of birth effect (Mortensen et al., 1999). These findings predominate in northern hemisphere studies, whereas southern hemisphere studies are more variable, including no effect (McGrath and Welham, 1999) or demonstrating excess births between May and October, which is the winter and spring in the southern hemisphere (Parker and Neilson, 1976).

Only five studies identified in our search probed if maternal ambient temperature was the critical factor for the season of birth, and these were published over three decades ago. All are characterized by substantial limitations that feasibly affected their success in finding a signal for elevated ambient temperature. Interestingly, the seminal study to consider ambient pregnancy temperature as the seasonally varying exposure was conducted in central Ohio, U.S. (Pasamanick et al., 1960). This study was not identified in our review as it was published in a book chapter, but it is widely cited, and it briefly inspired this line of investigation. These authors had already identified a role for early gestational temperature elevation in producing “mental deficiency” in children (Knobloch and Pasamanick, 1958), particularly in the 2^nd and 3^rd month, which they identified as a key time of neurodevelopment, defining a “critical window”. In their subsequent schizophrenia study, they found that the excess in winter births only occurred after warmer summers, proposing the relevance of hot temperatures in early gestation with disrupted neurodevelopment. This first study had more accurate data on temperature exposures than all subsequent studies as it was confined to a specific geographical area, the U.S. state of Ohio. Of the five subsequent schizophrenia studies testing if ambient temperatures explained a season of birth effect, only a study that employed regional U.S. data (New England, Midwest, and Southern states) found this relationship (Templer, 1980). Strengths of these two studies included the seasonally variably U.S. climate, which has more extreme periods of summer heat and winter cold, than the other locations studied. While all locations likely endured more recent extreme heat periods, effects of these can be obscured by using mean monthly or quartile temperatures from just a “representative” location in an ecological design.

Notably, two studies found evidence for the influence of cold temperatures around the time of delivery on the risk for schizophrenia (Hare and Moran, 1981; Tatsumi et al., 2002). Perinatal cold temperatures do not preclude the plausibility of detrimental heat effects in early pregnancy, but may reflect a different mechanism, perhaps one propelling early birth. Two studies examined within season temperatures, comparing schizophrenia birth rates across gradations from the coolest to warmest summers (McNeil et al., 1975; Watson et al., 1984) and the coolest to warmest winters (Watson et al., 1984). These studies did not find significant results, perhaps because the temperature measures were so imprecise for the actual gestational exposures. Studies attempting to include very large numbers of participants had less accurate environmental temperature data for the included local regions, and none of these had information on the mother’s residence in early pregnancy. Also, most studies just sought correlations between mean monthly temperatures and proportions of schizophrenia births, without considering critical cut off temperatures or the availability of cooling or heating for the pregnant mother or newborn.

4.2. Anorexia Nervosa

Season of birth effects were found in all three anorexia studies identified in this review but only for restrictive eating. Anorexia was linked to slightly later births than schizophrenia, peaking in April through June (Morgan and Lacey, 2000; Rezaul et al., 1996). Waller and colleagues (2002) found this pattern was specific for restrictive anorexics and less so for binge-purge anorexics, proposing that higher environmental temperature in the month of assumed conception produced the risk for restrictive anorexia. Although the mean monthly conception temperatures were similar for the two anorexia subtypes (10.1 and 9.34° C), only the severity of restrictive eating behaviors (i.e. dieting and oral control) in the restrictive anorexia cases was significantly correlated with elevated ambient temperature at conception. Watkins et al. (2002) nonetheless discounted a deleterious effect of extreme ambient temperature as the pathogenic factor, and instead proposed that the pregnancies of women with anorexia or subthreshold restrictive eating behavior just “survived” better if conceived in the warm months, assuming the mothers of these offspring also had restricted eating.

Willoughby et al. (2002) considered this pattern in the southern hemisphere. Although no season of birth effect was identified, they again demonstrated a significant association between the temperature at conception and diagnostic subtype, showing that restrictive anorexics from the southern hemisphere are less likely to be conceived in relatively cool weather, comparable to more likely to be conceived in warmer times.

4.3. Overlapping Schizophrenia and Anorexia Nervosa Risk Patterns

The positive findings of warm ambient temperatures in the conception months for schizophrenia and anorexia nervosa are of interest, despite the slightly later birth excess in anorexia. As described over 20 years ago, persons with eating disorders hold incorrect beliefs about their body habitus that are resistant to counter example, similar or identical to delusions, and share other features (Morgan and Lacey, 2000; Watkins et al., 2002). Indeed, recent evidence confirms the comorbidity of these disorders and furthermore implicates a specific phenotype for the comorbid cases and clinically high risk subjects (Malaspina et al., 2019; Sarac et al., 2021). Premorbid eating disorders were recently identified in 9.4% of adult cases (Malaspina et al., 2019) and in 27% of clinically high risk cases (Sarac et al., 2021), affecting significantly more females in both samples. Very similar prevalences of eating disorders were identified in a recent study of first episode cases (13%) and ultra-high risk subjects (31%) (Rasmussen et al., 2020). Somewhat older data from the NAPLS (North American Prodrome Longitudinal Study) study found 9% of 744 clinically high risk subjects had current disordered eating, although the age of onset of the eating disorders was not specified (Addington et al., 2017). As to phenotypes, the Malaspina et al. (2019) study reported that established cases who had premorbid eating disorders had higher IQ scores than other cases, but more severe psychotic symptoms, uniquely reporting gustatory hallucinations and delusions. The groups with and without premorbid eating disorders were similar in negative symptoms and functional impairments. Comparably, prodromal cases with eating disorders in the Sarac et al. (2021) study demonstrated significantly faster processing speed, better cognition and larger brain volumes than other prodromal subjects. Together, these data suggest that a distinct causal pathway may determine psychosis in persons with comorbid schizophrenia and eating disorders. Further support for this notion is the finding that there is a greater proportion of females among the excess winter births (Dassa et al., 1996). Prenatal temperature exposure can be tested for these cases in larger replication studies, as they may benefit from specific interventions.

4.4. Congenital Malformations

Overall, evidence was found for early pregnancy maternal ambient extreme heat event exposures and septal defects, cardiovascular defects considered to be noncritical, but those that nonetheless can carry lifelong health risks (Agay-Shay et al., 2013; Auger et al., 2017b; Lin et al., 2018; Stingone et al., 2019; Zhang et al., 2019). This extreme heat event exposure also associated with increased risks for neural tube defects (Auger et al., 2017a), oro-facial cleft (Soim et al., 2018), and cataracts (Van Zutphen et al., 2012). Consistent with these findings, higher hypospadias risk was associated with higher summer temperatures in early pregnancy (Kilinc et al., 2016). Notably, these studies included early pregnancy extreme heat events exposures not only in summer periods when peak heat may be expected, but also in cooler periods such as spring months and September (Auger et al., 2017b; Lin et al., 2018; Zhang et al., 2019) or the cold season of a Mediterranean climate (Agay-Shay et al., 2013). In fact, some extreme heat events findings were isolated to these cooler periods only (Agay-Shay et al., 2013; Lin et al., 2018). Together these results raise the possibility that ambient heat effects are specific, perhaps to days of unusual heat.

Evidence for the role of cold temperature exposure in association with congenital malformations was also identified. In two studies, an increase in hypothyroidism risk linked with birth month cold temperatures (Aminzadeh et al., 2010; Gu et al., 2007). Also, Van Zutphen et al. (2014) found increased aortic coarctation risk linked to extreme maternal cold exposure in early pregnancy. Although these results stem from studies examining different timepoints in pregnancy, they do provide evidence that cold temperatures also may influence neurodevelopmental outcomes.

4.5. Limitations

Although most studies covered by this systematic review offer strengths in their very large sample sizes and meteorological record utilization for objective ambient temperature data, many suffer from substantive limitations. Apart from studies that pinpointed unordinary temperatures for the time of year, e.g. extreme heat events (temperature > 90^th percentile or temperature > 95^th percentile), several studies only utilized mean monthly or mean quarterly (or seasonal) temperatures, that could obscure extreme temperature events. Indeed, it may not be that ‘general warmness’ or ‘coldness’ are teratogenic, but exposures in the extreme. Moreover, some studies that selected only summer months, when the greatest heat was expected, would have completely missed periods when heat occurred at times when women may not have been acclimated to warmer temperatures, such as spring, fall and the cold season. In fact, there is some evidence for significant exposure in these periods. As elevated maternal and fetal temperatures might impact fetal development over very short periods, the use of mean trimester and mean monthly temperatures could easily dilute a high temperature effect of a heat wave.

Additionally, unless studies utilized the birth record, where place of birth and gestational age at birth were noted, these variables otherwise were assumed. The assumptions were broadly reasonable (e.g. Stockholm for mean temperatures in a Swedish nationwide sample; 38 week gestational periods (McNeil et al., 1975)), but nonetheless were imprecise, obfuscating potentially important individual differences, and in the case of gestational age at birth, not accounting for the well-documented and potentially mediating or moderating influence of preterm birth in association with psychiatric outcomes. Moreover, early pregnancy maternal ambient heat exposure is known to be associated with preterm birth in some studies (Chersich et al., 2020), so preterm birth may be another cascading risk factor in this pathway.

An additional cascading risk factor may be maternal stress. For example, Project Ice Storm considers the effects of freezing winter rain storms in Quebec, Canada, in 1998, causing widespread power outages. In this project, authors focus on maternal self-reports of objective vs. subjective stress rather than ambient temperature exposure. One study showed that for intellectual abilities, children whose mothers experienced moderate or high objective stress during their first or second trimester had significantly lower mental development scores at 2 years of age (King and Laplante, 2005). As part of the objective stress assessment, participants reported loss of electricity for an average of 14.9 days. Moreover, lower cognitive and linguistic scores found in the study extended to 5.5 years of age (Laplante et al., 2008). Our systematic review raises the possibility that some of these effects may have been driven in part by sheer ambient cold temperature exposure, but the overall body of studies from Project Ice Storm demonstrates that maternal stress from temperature shocks may be one important intermediate variable between maternal ambient temperature exposure and offspring psychiatric outcomes.

There are additional potential sources of noise, particularly related to the studies on schizophrenia, and also more broadly relevant to the evaluation of all these studies. For example, diagnostic reliability should be considered as it was typically based only on administrative registries and not on clinical criteria. Case selection furthermore was limited to institutionalized/hospitalized patients in each study, introducing a “Berkson’s bias” by omitting ill persons not in treatment or institutionalized (Westreich, 2012), thus diminishing the magnitude of the season of birth effect for schizophrenia spectrum conditions. Indeed, a review of 30 studies found a greater season of birth effect for cases with an early onset of psychosis who carried less genetic risk (Boyd et al., 1986), and more recent studies confirm a season of birth effect for schizotypy (Konrath et al., 2016).

Unless maternal prenatal residence was documented, as in the National Birth Defects Prevention Study which asked women to self-report their residence over the pregnancy and three months prior, studies assumed that the maternal residence throughout the whole pregnancy was near the place of birth. This assumption does not account for women who traveled to a specific location to give birth or lived elsewhere, but delivered unexpectedly in another place. Finally, it was not known whether women were actually exposed to ambient heat or cold in the period in question as they may have been out of town or may have utilized adequate air conditioning or heating.

Representative locations chosen for estimating temperature exposures is another consideration. For example, Hare and Moran (1981) found consistently higher births in the coldest years for schizophrenia, related to low temperature around the time of birth, with no such effect for affective psychosis, neurosis or personality disorder. Here, no early pregnancy warmth effect was identified, but the use of mean monthly temperatures for all of England and Wales surely obfuscated locally fluctuating temperatures which are influenced by geography; typically cooler for proximity to water and warmer in city locations. Of course, with global warming, many locations are experiencing more extremes of summer heat compared to decades ago, but the critical periods for the effect are not clearly demonstrated. If specific temperature thresholds are needed to unmask a relationship to early pregnancy, then the relative temperature differences across different years used by all these studies will be insufficient to reveal the relationship. It is interesting that this methodology did reveal consistent associations between anorexia nervosa and temperature at conception, perhaps suggesting better diagnostic classification of the disorder, reducing the overall noise in the study. The recent studies on congenital malformations that pinpoint the temperature of the self-report maternal residence to using the most proximal meteorological station to assess specific extreme heat events have overcome these limitations of previous work.

Another consideration is the inclusion of males and females in samples. For example, a local Malmo, Sweden, sample of schizophrenia cases (more accurate temperature) and the entire national Swedish sample of cases (larger numbers, but temperatures based on “representative locations”) were examined by McNeil et al. (1975), with no influence of summer warmth over the pregnancy identified. However, the local sample was only men, whereas sex specific effects are important in this season of birth research area on schizophrenia. For example, five of seven studies found significant effects for females, with four showing far greater effects in females (Dalen, 1975; Jones and Frei, 1979; Parker and Balza, 1977; Parker and Neilson, 1976; Pulver et al., 1981; Pulver et al., 1983; Roche, 1974). Of note, these female specific effects may owe to the greater vulnerability of the male fetus to undergo spontaneous abortions in response to early pregnancy trauma.

Torrey et al. (1977) speculated that summer temperature extremes could cause genetic damage or that the genetic predisposition for schizophrenia made these offspring more vulnerable to seasonal environmental factors. Low socioeconomic status is proposed, as a season of birth effect was identified from public but not private hospitals (Barry and Barry, 1964; Barry and Bary, 1961). This would be consistent with heat effects if these mothers had more outdoor work with heat exposures. Vitamin D deficiency is also proposed (McGrath, 1999), but this pathway would not be expected to thusly vary with social class, as persons with lower social class spent more time outdoors. With the advent of air conditioning many more pregnant persons will be protected from extreme heat in early pregnancy. Another possibility is a sociocultural explanation. These models concern the youth of those born latest in their school class relative to the rest of their class peers, a relative age effect (i.e. those born in December for a calendar year cut off or those born in August for a September cut off deciding grade). Notably, Boland et al. (2018) applied a sweeping bioinformatics approach testing 12 prenatal exposures for 133 disease outcomes and only found this effect for attention deficit hyperactivity disorder, a diagnosis linked significantly to age at commenced schooling.

Several of the studies spanned decades, and many other factors could influence birth rates and illness risk over a decades-long study period, including war, famine, infection, pandemic, and other events. It is notable that findings for associations between cold winter temperatures in late pregnancy and increased winter and spring birth rates, may be due to measurement, where the temperature data would have been recorded closer in time to the birth compared to summer temperatures. Women in late pregnancy may have been less likely to move or travel in the late pregnancy period vs. early, thus their location in the assumed place of birth may have been more accurate.

Finally, we note other methodological issues. Unless the study utilized a prospective cohort design, with no a priori knowledge of the exposure or outcome, there is the potential for selection bias. In the retrospective cohort design, those who were lost to follow up would not have been identified, for example the offspring who did not survive. Although this concern could apply to studies on congential malformations, it may be lessened because the exposure and outcome occurred closer in time (i.e. prenatal temperature exposures and congential birth outcomes) compared to the outcome of psychiatric disorders emerging later in life. Additionally, several studies returned by this systematic review included a large number of statistical tests without description of multiple comparisons correction.

5. Microbiome Research as a New Horizon

Ambient temperature extremes that perturb internal maternal temperatures could influence the maternal microbiome, as demonstrated in animal models (Khakisahneh et al., 2020). Moreover, the microbiome can be impacted by seasonality due to differential availability of food sources especially in populations following traditional lifestyles (Smits et al., 2017). While Western populations generally have access to most food throughout the year, the seasonality of food cost (Bai et al., 2020) still can impact the diet of pregnant women, and dietary quality varies by season (van der Toorn et al., 2020). Altogether, seasonal changes in diet can significantly impact the microbiome of both women and their offspring and subsequently alter behavior (Gacias et al., 2016) and health outcomes, including psychiatric disorders (Szeligowski et al., 2020). Notably, it is the maternal vaginal microbiome that seeds the neonatal gut during delivery, organizing the development of the gut and immune system (Hoffman et al., 2020; Tamburini et al., 2016). The impact of seasonality in diet and nutrient intake and its subsequent effect on the microbiome is generally not described in the studies returned by this systematic review. This is an important limitation that needs to be accounted for in future work, though this research with fetal brain development as an outcome may be better suited to animal models for initial hypothesis testing given that large prospective cohort studies will entail a span of several decades.

6. Implications

Remarkably, there is more than a 2,500-year history of attending to the birth month as a means to predict a person’s characteristics. The Babylonians developed astrology, around 2,400 years ago which then spread to the eastern Mediterranean and Hippocrates later acknowledged seasonal effects on human behavior. While astrology is considered a pseudoscience, Hippocrates’ observations and the consistent empirical findings over the last century associating birth months to an array of behavioral conditions, temperamental styles, functioning and disorders are compelling. What remains unclear are the factors that could account for these differences.

There is an urgency to examine if ambient temperatures contribute to this effect. Needed are large-scale, prospective designs in a seasonally changeable climate with information on maternal residence over the pregnancy and at parturition, with local environmental temperature records, and occupational and other exposures to extreme ambient temperatures. The interpretation of these findings will also entail additional information on such factors as prenatal stressors, infection, nutrition, parental ages, obstetric history, socioeconomic status, and family history.

This possibility is highly consequential in light of global climate change resulting from warming of the earth. According to the U.S. National Center for Environmental Information, the number of days with mean temperature above 32°C in the average U.S. county is forecasted to increase from approximately one per year to 43 per year by 2070–2099. (https://www.ncdc.noaa.gov/sotc/national/2019). As such, the impact of climate extremes on pregnancy outcomes and offspring health is only expected to increase.

Appendix A

Medline Search

1. exp Temperature/ or exp Seasons/

2. (exposure or temperature or climate or index).tw,kf.

3. 1 and 2

4. exp Climate Change/

5. exp greenhouse effect/

6. exp Humidity/

7. exp air conditioning/

8. exp Seasons/

9. 3 or 4 or 5 or 6 or 7

10. ((Heat or hot or cold) adj3 (exposure or temperature or climate or index)).tw,kf.

11. Heatwave.tw,kf.

12. climate change.tw,kf.

13. Global warming.tw,kf.

14. elevated ambient temperature.tw,kf.

15. (temperature adj2 extreme*).tw,kf.

16. (season adj2 birth).tw,kf.

17. air conditioning.tw,kf.

18. 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17

19. 9 or 18

20. exp Pregnancy/

21. exp Pregnancy complications/

22. exp pregnant women/

23. exp maternal exposure/

24. 20 or 21 or 22 or 23

25. Pregnant.tw,kf.

26. pregnancy.tw,kf.

27. gestation.tw,kf.

28. mother.tw,kf.

29. maternal.tw,kf.

30. birth.tw,kf.

31. prenatal.tw,kf.

32. 25 or 26 or 27 or 28 or 29 or 30 or 31

33. 24 or 32

34. 19 and 33

35. exp Mental Disorders/

36. exp Congenital Abnormalities/

37. exp Brain/gd [Growth & Development]

38. ((Mental or psych* or neuro*) adj3 (illness* or disorder* or disease* or outcome* or health)).tw,kf.

39. Congenital.tw,kf.

40. birth defect*.tw,kf.

41. Neurodevelopment.tw,kf.

42. ((neuro* or brain) adj2 (development or architecture)).tw,kf.

43. (behavior or temperament).tw,kf.

44. (educational attainment or social function).tw,kf.

45. ADHD.tw,kf.

46. Attention deficit.tw,kf.

47. (psychosis or schizophrenia or depression or bipolar or anxiety).tw,kf.

48. Autism.tw,kf.

49. Spina bifida.tw,kf.

50. Cleft.tw,kf.

51. (Clubfoot or club foot).tw,kf.

52. Down syndrome.tw,kf.

53. 35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43 or 44 or 45 or 46 or 47 or 48 or 49 or 50 or 51 or 52

54. 34 and 53

55. 54 not (Animals/ not (Animals/ and Humans/))