Introduction

Climate change is one of the most significant threats to human health [1], characterized by an increase in global temperatures amongst other environmental changes. Global temperatures have increased by approximately 1·2 °C, and are projected to increase beyond a critical threshold of 1·5 °C in the next 5–10 years [2]. Increasingly, heat exposure is being linked with a multitude of short- and long-term health effects in vulnerable populations, including children [3], the elderly, and pregnant women [4]. The effect on pregnant women extends to the health of the foetus, with significant detrimental effects associated with heat exposure including preterm birth, stillbirth, and decreased birth weight [5]. Impacts of heat exposure are increasingly important in populations in resource-constrained settings, where heat adaptation measures such as active (air-conditioning) and passive cooling (water, green and blue spaces) are limited, and often inaccessible [6]. These populations are often found in some of the hottest climates and in areas whose contribution to global warming is negligible, thus compounding inequities [7]. In addition, research in this field is biased towards Europe, North America and Asia and is profoundly underrepresented in Africa and South America [8]. Understanding the scope and distribution of research conducted is key to guiding future research, including biological studies to explore possible mechanisms, and interventional studies to alleviate any observed negative effects. Multiple previous systematic reviews have explored the short-term impacts of heat on the foetus [3, 5, 9] but only one has explored the long-term impacts of heat exposure on mental health [10]. The in-utero environment has long been considered important in the long-term health and wellbeing of individuals [11, 12], although it has been challenging to delineate specific causal pathways. This study aims to systematically review the literature on the long-term effects of heat exposure in-utero on the foetus, and explore possible casual pathways.

Materials and methods

This review forms part of a larger systematic mapping survey of the effect of heat exposure, and adaptation interventions on health (PROSPERO CRD 42019140136) [13]. The initial literature search was conducted in September 2018, where the authors searched MEDLINE (PubMed), Science Citation Index Expanded, Social Sciences Citation Index, and Arts and Humanities Citation Index using a validated search strategy (Supplementary Text 1). This search was updated in April 2023 through a MEDLINE search, as all previous articles were located in this database. Screening of titles and abstracts was done independently in duplicate, with any differences reconciled by MFC, with subsequent updates conducted by NB and DL. The authors only included studies on humans, published in Chinese, English, German, or Italian. Studies on heat exposure from artificial and endogenous sources were excluded, and only exogenous, weather-related heat exposure during pregnancy was included. All study designs were eligible except modelling studies and systematic reviews. No date restrictions were applied. EPPI-Reviewer software [14] provided a platform for screening, reviewing of full text articles, and for data extraction. No additional information was requested or provided by the authors. Long-term effects were defined as any outcomes that were not apparent at birth.

Articles meeting the eligibility criteria were extracted in duplicate after the initial search and then by a single reviewer in the subsequent update (NB/DL). Data were extracted to include characteristics outlined in Supplementary file 1.

This systematic review was conducted according to the Systematic Review without Meta-Analysis (SWiM) guidelines, broadly based on PRISMA [15], as the outcomes, statistical techniques, and heat exposure measurements were heterogenous, rendering a meta-analysis untenable. Outcomes were grouped clinically, reviewed for the magnitude and direction of effect, and their statistical significance, and included negative or null findings when reported on. A text-based summary of these findings was made. ‘Vote-counting’ was utilized to summarise direction of effect findings. Analysis was conducted on the geographical areas, climate zones [16], mean annual temperature and socioeconomic classification of the country where the studies were conducted. Furthermore, an attempt was made to identify at-risk population sub-groups.

The principal investigator assessed each study for a risk of bias using the tools developed by the Joanna-Briggs Institute (JBI) [17] (Supplementary file 1). Each study was classified as high or low risk of bias. Studies that did not score ‘yes’ on two or more applicable parameters were classified as high risk of bias [5]. Due to the limited research in this field, no studies were excluded based on risk of bias. The certainty of the evidence was assessed using the GRADE approach, with the body of evidence assessed on a scale of certainty: very low, low, moderate and high [18]. Due to the heterogeneity of outcomes, and the reporting thereof, assessment of publication bias was not possible.

The funder of the study had no role in study design, data collection, analysis, interpretation, or writing of the report.

Results

The updated search identified 18 621 non-duplicate records, and after screening 229 full-text articles were reviewed for inclusion, with a total of 29 studies included in the final analysis (Fig. 1: flow chart). The included studies were conducted in 25 countries across six continents, including six Low-Income Countries (LIC), two Lower-Middle Income Countries (LMIC), one Upper-Middle Income Country (UMIC) and 16 High Income Countries (HIC) [19]. They included 25 Köppen-Geiger climate zones [16], and mean annual temperatures ranging from 2.1 °C in Norway to 30.0 °C in Burkina Faso [20] (Figs. 2 and 3). All studies were observational, with 17 cohort, five case-control and eight cross-sectional studies. The timeline of the data is from 1913 to 2019, and individuals included ranged in age from neonates to adults, and the elderly. The studies were grouped by outcomes as follows: behavioural, educational and socioeconomic (n = 6), cardiovascular disease (n = 6), respiratory disease (n = 3), growth and anthropometry (n = 2), mental health (n = 12) and longevity and mortality (n = 3). The measures of heat exposure were variable, with minimum, mean, maximum, and apparent temperature being utilized, as well as temperature variability, heat wave days and discreet shocks (number of times exposure exceeded a specific threshold). The majority of studies measured heat using mean temperature (n = 27/29). In addition, the statistical comparison was diverse, with some studies making a continuous linear comparison by degree Celsius, while others compared heat exposure by quartiles, amongst other categorical comparisons. Furthermore, heat exposure by any definition was not reported over the same timeframes, with some studies including variable periods before birth, during pregnancy and at birth in their analysis. Levels of temporal resolution of heat exposure were also diverse, ranging from monthly effects to effects observed over the entire gestational period, or year of birth. In addition, differing use of heat adaptation mechanisms was not uniformly described and adjusted for. Various confounders were adjusted for, and although not uniform, these were generally inadequate. The effect on female infants was greater than on male infants in eight of nine studies differentiating by sex, with increased effects on marginalised groups (African-Americans) in one further study. Overall, the quality of the evidence, as assessed by the GRADE approach, was low in respiratory and longevity outcome groups to very low in all other groups, primarily as a result of their observational nature and high risk of bias, due to insufficient consideration of confounders, and inadequate measures of heat exposure.

Fig. 1
figure 1

PRISMA flow diagram

Full size image
Fig. 2
figure 2

Map showing countries where studies were conducted relative to mean annual temperature [21]

Full size image
Fig. 3
figure 3

Map showing countries where studies were conducted relative to climate zones [16]

Full size image

A total of six studies reported on behaviour, educational and socioeconomic outcomes, which were detrimentally affected by increases in heat exposure (Fig. 4; Table 1), although the quality of the evidence was very low. End-points were not uniform, but included earnings, completion of secondary school or higher education, number of years of schooling, and gamified cooperation-rates in a public-goods game (where test scores represent achieving maximal public benefit in hypothetical situations).

Table 1 Summary table-behaviour, educational and socioeconomic (Grade of evidence: Very low)
Full size table

Two large studies reported a detrimental effect of heat exposure on adult income, with the greatest effect noted in first trimester exposure. These studies noted a reduction in earnings of up to 1·2% per 1 °C increase in temperature, with greater effects in females [22], and a decrease of $55.735 (standard error(SE): 15·425, P < 0·01) annual earnings at 29–31 years old, per day exposure > 32 °C [26]. Two studies reported worse educational outcomes, with the greatest effect noted in the second trimester [23]. Rates of completing secondary education were found to be reduced by 0·2% per 1 °C increase in temperature (P = 0·05) [22], illiteracy was increased by 0·18% (SE=(0·0009); P < 0·05) and mean years of schooling was lowered by 0.02 (SE=(0·009) P = 0·07) [23]. Two studies reported a beneficial effect of heat exposure on educational outcomes, although both studies suffered from significant methodological flaws, and effects were < 0·01% when effect estimates were noted [24, 27]. One small study reported lower cooperation rates by 20% (P < 0·01) in a public-goods game, with lower predicted public wellbeing [25].

The studies generally exhibited a dose-response effect with evidence for a critical threshold of effect of 28 °C in one study [22]. All studies were at a high risk of bias.

Fig. 4
figure 4

Figure showing vote counting across all outcome groups. No Effect = No direction of effect noted in study

Full size image

Six studies reported on cardiovascular pathology and risk factors thereof, which were detrimentally affected by increased exposure to heat (Fig. 4; Table 2), although measures and surrogates of this outcome were heterogenous. The quality of the evidence was very low, and the sample sizes were small. Outcomes included blood pressure, a composite cardiovascular disease indicator, and specific cardiovascular disease risk factors such as diabetes mellitus (type I), insulin resistance, waste circumference, and triglyceride levels.

Table 2 Summary of findings cardiovascular pathology (Grade of evidence: Very low)
Full size table

Three studies found a detrimental effect of heat exposure on hypertension rates, and increased blood pressure [31], with a maximum of 1·6 mm Hg increase noted per interquartile range (IQR) increase (95% Confidence interval (CI) = 0·2, 2·9, P = 0·024) in children [30], with increased effects on women in the largest study (N = 11,237) [32]. Another study found increasing heat exposure at conception was detrimentally associated with an increase in coronary heart disease (P = 0·08) [32], although one of the smaller studies (N = 4286) found a beneficial effect of heat exposure at birth on diverse cardiovascular outcomes, including coronary heart disease (P = 0·03 for trend), triglyceride levels (P = 0·06 for trend) and insulin resistance (P = 0·04 for trend) [27]. One study found lower odds of type I diabetes mellitus with increasing heat exposure, with odds ratio (OR) = 0·73 (95%CI = 0·48, 1·09, P-value not stated) [28]. Another study did not detect statistically significant relationships between heat exposure and hypertension or a composite cardiovascular disease indicator, but did not provide effect estimates [29]. Five studies were at a high risk of bias [27, 29,30,31,32], with only one case-control study at a low risk of bias [28].

Respiratory pathology was reported by three studies, assessing different outcomes. Outcomes were detrimentally associated with increasing heat (Fig. 4; Table 3), however the quality of the evidence was low. The outcomes were primarily measured in infants and children, with no studies on adult outcomes. The largest study (N = 1681) found that increasing heat exposure increased the odds of having childhood asthma [33], and another small study (N = 343) noted worsened lung function with increasing heat exposure [34].

Table 3 Summary of findings of respiratory pathology (Grade of evidence: low)
Full size table

An additional study noted increased odds of childhood pneumonia with increasing diurnal temperature variation (DTV) in pregnancy, with a maximum OR = 1·85 (95%CI = 1·24, 2·76) in the third trimester [35].

Exposure in the third trimester had the greatest effect across all three studies [33,34,35]. Females showed an increased susceptibility to heat exposure’s effects on lung function, but males were more susceptible to heat’s effect on childhood pneumonia. There was a critical threshold noted in the asthma study of 24·6 °C, with a dose-response effect. The asthma study was assessed as low risk of bias, however the other studies were at high risk.

Growth and anthropometry was reported on by two studies, with differing outcomes, although in both, heat exposure was associated with detrimental, although heterogenous, outcomes (Fig. 4; Table 4). The overall quality of the evidence was very low. One study found a positive association with heat exposure and increased body mass index (BMI), r = 0·22 (P < 0·05) in the third trimester with greater effects noted in females and in African-Americans [36]. Another large study (N = 23 026) found increased odds of stunting (OR = 1·28, 95%CI = not stated, p < 0·001) with a negative correlation with height noted (r=-0·083 P < 0·01) [37]. Effects were greatest in the first and third trimester. Both studies were at a high risk of bias.

Table 4 Summary of findings growth and anthropometry (Grade of evidence: Very low)
Full size table

Mental health was reported on by 12 studies. Increasing heat exposure generally had a detrimental association with mental health outcomes (Fig. 4; Table 5), although these were heterogenous. The overall quality of the evidence was very low. Five studies reported on schizophrenia rates, with only one study showing a strongly positive association of heat exposure at conception with schizophrenia rates (r = 0·50, p < 0·025) [38]. Another study noted the same effect with increasing heat in the summer before birth, however this was not statistically significant [39]. The third study reported no association of this outcome [40], with another small study (N = 2985) showing a negative correlation with temperatures at birth, without reporting on heat exposure during other periods of gestation [41]. The fifth study failed to report direction of effect, but noted non-significant findings [42]. Six studies reported on eating disorders, with all six showing a detrimental effect with increasing heat exposure. Of the three studies on clinical anorexia nervosa, one reported increasing rates of anorexia nervosa compared to other eating disorders (χ²= 4·48, P = 0·017) [43], another reported increasing rates of a restrictive-subtype (χ²= 3·18, P = 0·04) as well as reporting worse assessments of restrictive behaviours [44], which was supported by a third study in a different setting [45]. Three studies examined non-clinical settings, with some inconsistent effects. The first study showed a weak positive association with heat exposure, and drive for thinness (Spearman’s ⍴ = 0·46, P < 0·05) and bulimia scores (Spearman’s ⍴ = 0·25, P < 0·05) [46], which was supported by a replication study [47], and one other study [48]. The most significant and consistent effects noted in the third trimester, at birth, and in females [47, 48]. One study reported a beneficial effect of increased temperatures in the first trimester on rates of depression, however no other directions of effect were noted for other periods of exposure [49]. These studies were at a high risk of bias.

Table 5 Summary of findings mental health (Grade of evidence: Very low)
Full size table

Increasing heat exposure had a detrimental effect on longevity and mortality across various outcomes (Fig. 4; Table 6), although despite large sample sizes, the quality of the evidence was low. One study found a negative correlation of heat exposure with longevity (r=-0·667, P < 0·001), with a greater effect on females [50]. A second study showed a detrimental effect on telomere length, as a predicter of longevity, with the greatest effect towards the end of gestation (3·29% shorter TL, 95%CI = − 4·67, − 1·88, per 1 C increase above 95th centile) [51]. Conversely, a third study noted no correlation with mortality [24]. All but the study on telomere length [51] were at a high risk of bias.

Table 6 Summary of findings longevity and mortality (Grade of evidence: low)
Full size table

Discussion

This study establishes significant patterns of effects amongst the outcomes reviewed, with increasing heat exposure being associated with an overall detrimental effect on multiple, diverse, long-term outcomes. These effects are likely to increase with rising temperatures, however modelling this is beyond the scope of this review.

The most notable detrimental outcomes are related to neurodevelopmental pathways, with behavioural, educational, socioeconomic and mental health outcomes consistently associated with increasing heat exposure, in addition to having the greatest body of literature to support this. Importantly, other systems such as the respiratory and cardiovascular systems also suggest harmful effects of heat exposure, culminating in detrimental associations with longevity and mortality. Some studies illustrated a possible beneficial effect in some disease-processes, such as coronary heart disease and depression showing the potential for shifting disease profiles with rising temperatures.

The detrimental effects of heat exposure became more significant with increasing temperatures, with many studies describing increasing effects beyond critical thresholds which, although varied across studies, suggest that there is a limit of heat adaption strategies, both biological and behavioural [52, 53].

In addition, the effect of increasing heat exposure was associated with worse outcomes in already marginalised communities, such as women [22, 32, 34, 36, 44, 47, 48, 50] and certain ethnic groups (African-Americans) [46]. The reasons for sub-population vulnerabilities are unclear and likely complex. In the case of female foetuses being more susceptible to changes in the in-utero environment, it is possible that there is a ‘survivorship bias’. This would occur if women with harmful exposure lose male infants during pregnancy at a higher rate, and thus the surviving female infants appear more at risk. However, despite an increased risk of early pregnancy loss, there are no studies that have assessed this differential vulnerability. This still has the effect of potentially increasing the burden of disease on an already marginalised group.

In the case of certain population groups being more at risk, it is likely that both physiological differences in vulnerability as well as socio-economic effect-modifiers exist to explain these differences, however, the included literature lacks sufficient evidence to assess this. The vulnerabilities of different populations to the long-term effects of heat exposure in-utero likely contributes to the unequal impacts of climate change that have already been established [54], and will be an important contributor to inequality with future increases in temperature. Further research in this area is critical to inform targeted redistributive interventions.

Although the associations may be clear, establishing causality is fraught with difficulty, with no consensus on an infallible approach [55,56,57]. However, it is prudent to highlight supporting evidence in this review.

The hypothesis that the in-utero environment had significant long-term impacts on the foetus was first suggested by Barker, in the context of maternal nutrition and cardiovascular disease [11]. Further studies supported this hypothesis, and expanded on the effects the in-utero environment has on the foetus and its long-term wellbeing [58]. Long-term heat exposure may also be associated with changes in nutritional availability [11], and is likely one of many complex but important environmental exposures in-utero.

Maternal comorbidities, associated with increasing heat exposure such as hypertensive disorders of pregnancy and gestational diabetes mellitus, are known to negatively affect the foetus in the long-term [59, 60]. These comorbidities may be part of the long-term pathogenicity of heat exposure, through short-term exposure-outcome pathways. Placental dysfunction is central to the pathology of pre-eclampsia, and is a significant cause for foetal pathology [61, 62]. The placenta is not auto-regulated and is therefore acutely affected by changes to blood volume, heart rate and blood pressure, culminating in cardiac output as it is delivered to the placenta as an end-organ with resultant negative effects on the foetus [63]. Heat-acclimatisation mechanisms are hypothesized to affect this delicate balance [52, 64], with observational studies supporting this [64]. It has been suggested that heat exposure’s increase in inflammation is a possible causative mechanism for pre-term birth [5, 52], but inflammation has numerous additional effects on the immune system and could prove an insult to the mother and developing foetus [62, 65]. These effects may only manifest in the long-term.

Heat was one of the earliest described teratogens [66], with significant effects on neurodevelopment noted in animal models in keeping with the observed associations of this review [67]. Biological organisms are extremely dependent on heat as a trigger for various processes. Plants and animals undergo significant change in response to the seasons, which are often guided by fluctuations in temperature. These changes are often mediated by epigenetic mechanisms, allowing the modification and modulation of gene expression [68, 69].

Thus, from an evolutionary perspective, DNA, is sensitive to changes in temperature. The mechanism of this sensitivity has been shown to be primarily epigenetic in nature [69]. Increasing heat results in modifications to histone deacetylation and DNA methylation [69]. This is required to provide fast-acting adaptions to acute stressors, but can have long-term effects too [70]. Thus, it is likely that humans are sensitive to changes in temperature, which can alter epigenetic modifications, and thus our exposome. This sensitivity, may have provided a survival benefit in times of increasing heat, or it may simply be a vestigial function which provides no survival benefit, and may in fact have detrimental effects [71]. Epigenetic changes have been shown to have significant effects on metabolic diseases and risk profiles, and an in-depth review is provided by Wu et al. [72]. The exact processes and genes involved would be an area requiring further research, where similar research exists on the effects of nutrition on exact epigenetic pathways [73]. An important pattern requiring further research involves the effect heat may have on neurodevelopment [67, 74]. The above pathways provide additional mechanisms for the long-term lag between exposure-outcome pathways. In addition, acute heat exposure at the time of birth has been associated with various possibly pathogenic mechanisms such as preterm birth [5], low APGAR scores [75] and foetal distress [76], as well as a possible effect on the maternal microbiome and the seeding thereof to the neonate [10, 64, 77, 78]. These effects, can all provide plausible causes for the long-term outcomes observed through short-term insults. The interplay of these, and additional factors is highlighted in Fig. 5 [79]. Importantly, the periods of vulnerability are likely different for these various pathways, but specific outcomes may have multiple periods of vulnerability through different pathways.

Fig. 5
figure 5

Causal pathways

Full size image

The outcomes associated with increasing heat exposure highlight the health, social, and economic cost of global warming, establishing current estimates and future predictions for this are beyond the scope of this research but would provide a valuable area for future research. This would entail estimating disease-burden due to climate change through attribution studies. Traditional health impact studies conflate adverse outcomes from natural variations in climate (‘noise’) with adverse outcomes from anthropogenic climate change. However, not every climate-related adverse outcome is the result of anthropogenic climate change, and these effects are likely different in vulnerable populations. This highlights the benefit of studying and implementing effective heat adaptation strategies in areas where the greatest effect is likely to be observed, and where the greatest impacts in lessening the economic and human impact of global warming are possible [80, 81].

Limitations

The difficulty in assessing the data is compounded by the heterogenous measures of heat exposure. No studies used widely accepted heat exposure indices that consider important environmental modifying factors like humidity and windspeed [82, 83]. In addition, effect modifiers, heat acclimatisation and adaptation strategies were seldom considered [84,85,86]. It may be prudent for future studies to consider the measure of ionizing radiation exposure as an analogous environmental exposure, where different measures exist for the intensity, total quantity (a function of duration of exposure) and biologically-adjusted quantity absorbed [87]. Differing time-periods of exposure made it difficult to evaluate specific periods of sensitivity, which are likely different for various outcomes, depending on critical periods of development.

Despite consistency across different contexts in this review, the analysis of the distribution of the included studies highlights the unequal weight of studies towards relatively cooler climates, in regions with higher socioeconomic levels and likely greater heat adaptation uptake, and must therefore be interpreted in this context. It is possible that myriad factors that differ geographically, including physiological and socio-economic differences, will influence the effects of heat, and thus there is likely no underlying universal truth to associations and effect estimates.

Quantifying, describing and comparing the effect size across studies was rendered more difficult due to heterogenous statistical analyses.

Although some studies adjusted for possible confounding variables, not all reported on this, with the effects of seasonal, foetal, and maternal biological factors that may not lie on the causal pathway seldom considered [3, 5, 9, 88,89,90,91,92].

Data extraction and assessment of risk of bias was not uniformly undertaken in duplicate due to resource constraints, which may predispose to extraction errors or bias. The high risk of bias of included studies, limits the utility of the overall assessment of effects and suggestions for further action. In addition, publication bias is likely skewing the results towards statistically significant detrimental results, with studies with smaller sample sizes not necessarily showing wider distribution of findings as would be expected.

Conclusions

Climate change, and in particular, global warming, is a significant emerging global public health threat, with far reaching, and disproportionate effects on the most vulnerable populations. The effects of increasing heat exposure in utero are associated with, and possibly causal in, wide-ranging long-term impacts on socioeconomic and health outcomes with a significant cost associated with increasing global temperatures. This association is as a result of a complex interplay of factors, including through direct and indirect effects on the mother and foetus. Further research is urgently required to elicit biological pathways, and targets for intervention as well as predicting future disease-burden and economic impacts through attribution studies.