Adverse effects of temperature on perinatal and pregnancy outcomes: methodological challenges and knowledge gaps

Evidence linking temperature with adverse perinatal and pregnancy outcomes is emerging. We searched for literature published until 30 January 2023 in PubMed, Web of Science, and reference lists of articles focusing on the outcomes that were most studied like preterm birth, low birth weight, stillbirth, and hypertensive disorders of pregnancy. A review of the literature reveals important gaps in knowledge and several methodological challenges. One important gap is the lack of knowledge of how core body temperature modulates under extreme ambient temperature exposure during pregnancy. We do not know the magnitude of non-modulation of body temperature during pregnancy that is clinically significant, i.e., when the body starts triggering physiologic counterbalances. Furthermore, few studies are conducted in places where extreme temperature conditions are more frequently encountered, such as in South Asia and sub-Saharan Africa. Little is also known about specific cost-effective interventions that can be implemented in vulnerable communities to reduce adverse outcomes. As the threat of global warming looms large, effective interventions are critically necessary to mitigate its effects.


. Introduction
Global surface temperature has been on the rise since the Industrial Revolution due to the rapid increase in greenhouse gas emissions.The rising trend is likely to continue, increasingly exposing the world population to extreme climatic conditions, more frequently for longer duration and with greater intensity.The 2019 Global Burden of Disease study estimated the burden of extreme temperature, not including the effect of adverse pregnancy or perinatal outcomes because of the lack of causal evidence (1).This manuscript begins with a narrative presentation of the association between temperature and specific pregnancy or perinatal outcomes.In the following section, we discuss the methodological challenges in these types of studies.The final section highlights the gaps in the existing literature.Addressing these gaps should shed light on the nature of the relationship between temperature and adverse outcomes.This manuscript focused on highlighting the gaps and methodology issues in the current literature, rather than conducting a systematic or scoping review, because several of those already exist (2)(3)(4).
We focused on the outcomes that were most studied and searched for relevant literature using PubMed, Web of Science, published reviews, and reference lists.We used the following Medical Subject Headings (MeSH): "Temperature, " "Heat, " "Perinatal Outcomes, " "Pregnancy Outcomes, " "Preterm Birth, " "Low Birth Weight, " "Stillbirth, " "Hypertensive Disorders of Pregnancy."The MeSH terms were combined with "and, " "or, " and "not" operators, and the search was restricted to articles on humans and published before 30 January 2023.

. Preterm birth
We found 24 studies that examined short-term exposure to temperature and preterm birth (PTB) (5-28).The associations reported in these studies are quantitatively different [i.e., hazard ratio (HR), relative risk (RR), or odds ratio (OR)], vary in magnitude and precision, comparison groups are heterogeneous, and exposure ranges from a few individual days before delivery to cumulative average over a period.Eight studies were from the US, seven were conducted on different European populations, three each were from China and Australia, two were from Iran, and one was from Canada (Table 1).The studies vary in size ranging from a few thousand to over a million births, with percentages of PTB between 3% and 17% (Table 1).While many studies included births throughout the year, some included from one season.Four studies from California (9,(16)(17)(18) one from Minnesota (14), North Carolina (15), Italy (13), and Spain (6) examined the association separately by season.Most of the studies used birth record data employing retrospective cohort, case-crossover, or time series designs.Individual lag days (e.g., lag 1 representing the day before birth) or cumulative lags (e.g., lag 0-6 representing an average of 7 days, starting from the day of birth and including 6 preceding days), in the weeks and months prior to the day of birth, were used as exposure windows at risk.Additionally, various exposure contrasts were examined, e.g., for heat, 75th percentile or higher, and for cold, 10th percentile or lower temperatures were compared with the percentiles in between (e.g., median).Studies that examined heat waves defined in various ways, e.g., number of days above a threshold, and consecutive number of days above a threshold.The reported associations across studies broadly ranged from 1.01 (1.01, 1.02) to 2.5 (1.02, 6.15) (15, 24); the smallest association was reported by a US study for a 1-day lag temperature of 73-74 • C compared with 71-72 • C, and the largest was reported by a Swedish study for a 30-day cumulative lag at the 75th percentile (17.5 • C) compared with the median (7.7 • C).The study-specific associations are presented in Table 1.Notably, most of the studies reported that exposure-outcome associations were below 20%, i.e., the increased risk of PTB was below 20%.While heterogeneity between studies contributed to the diversity of evidence, summarizing results across studies becomes challenging without making additional assumptions.Many exposure windows were often combined with several exposure contrasts, and numerous tests of significance were conducted.
Four studies examined the association between long-term exposure to temperature and PTB (Table 1) (5, [29][30][31].Associations ranging from 1.05 (1.01, 1.19) to 1.31 (1.11, 1.56) were reported in studies conducted among Chinese, Korean, Israeli, and US populations (5, [29][30][31]. Effect modification by infant sex, race, ethnicity, maternal characteristics, and air pollution exposure has also been reported (8,9,26,31).Three studies reported the associations with temperature were higher for non-Hispanic African Americans and Hispanics compared with non-Hispanic white people (8,9,31).Maternal smoking, alcohol consumption, hypertension, and diabetes have also been suggested to increase susceptibility to temperature (9).Preliminary evidence of synergism between heat waves and PM 2.5 has been reported (10,29).However, in another study, neither PM 10 nor ozone modified the association between apparent temperature (AT) and PTB, although maternal age, education, and pre-existing chronic diseases modified the association (13).
Two studies examined short-term exposure (Table 2).An Australian study reported the association between a daily maximum temperature of 20-25 • C in the last week of pregnancy and 11 g (8, 18) higher estimated mean birth weight compared with > 30 • C (37); another study from Spain reported an RR of 1.003 (1.002, 1.004) for LBW per 10 • C increase in maximum temperature on the 14th week (39).

. Spontaneous abortion and stillbirth
We found 13 studies that examined the association of temperature with spontaneous abortions and stillbirths (12,39,46,(48)(49)(50)(51)(52)(53)(54)(55)(56)(57)) (Table 3).A unique study from Hungary reported that the likelihood of unobserved pregnancy loss is 0.22 (0.12, 0.33) per 100, 000 women aged 16-44 years if there is 1 day in the first 2 weeks of pregnancy with a mean temperature of >25 • C compared with 15-20 • C (48).Compared with the other outcomes, relatively more studies on stillbirth are from low-and middle-income countries, likely because of the higher incidence of  stillbirth in these regions.These studies employed cross-sectional, case-crossover, retrospective, or time-series designs (Table 3).The associations for stillbirth ranged from 1.03 (1.00, 1.06), when the temperature was above 97.5thpercentile for 4 consecutive days in the week prior to birth compared to 0 days above that level, to 3.7 (3.1, 4.5), per 1 • C increase in the mean temperature in the week prior to birth in the warm season (46,55).The association between temperature and stillbirth has been reported to differ by race, ethnicity, or socioeconomic status.Higher associations were reported for non-Hispanic African Americans and Hispanics compared with no association for non-Hispanic white (52).Higher association was also reported only in most socioeconomically deprived population but not in the overall study population (56).

. Hypertensive disorders of pregnancy
Temperature has been associated with hypertensive disorders of pregnancy.Two-to three-fold increase in the risk of preeclampsia per 9 • C has been found for 1st, 2nd, and 3rd trimester averages of daily mean temperatures (58).Before 12 weeks of conception, colder temperature has been reported to increase the odds of preeclampsia or eclampsia (59).However, in the first 20 weeks of pregnancy, colder temperature reduced the odds, while hotter temperature increased the odds of preeclampsia or eclampsia (59).No association with preeclampsia was found for heat stress in the 3 weeks prior to delivery (12).

. Methodological considerations
Choice of metric is an important consideration because each represents qualitatively different exposure.For example, AT incorporates humidity and reflects how hot it feels outdoors.High humidity in moderately warm conditions can cause discomfort leading to physiological stress (60).High minimum temperature over consecutive days reflects the lack of overnight cooling, which could cause sleep perturbations (61).High maximum outdoor temperature increases physical discomfort, but wind velocity may increase or decrease physiological stress.When the body is sweating, wind can considerably influence regulatory physiological mechanisms.The other less commonly used metrics to measure heat stress are excessive heat factor, physiological equivalent temperature (PET), and Universal Thermal Climate Index (UTCI) (12,62,63).Regardless of the primary choice of metric, at least a few (e.g., maximum and minimum AT) should be uniformly reported to allow comparisons of association across studies.Furthermore, if temperature thresholds or non-linear relationships are observed, additional results should be reported in a manner that will allow between study comparisons and pooling of estimates across studies.
The importance of the comparison group can be highlighted using the example of the tri-country Latin American study (34).In the study, the means of the monthly averages for Brazil, Mexico, and Chile were 22, 19, and 14 • C, respectively.Because of the annual weather patterns in each country, for Brazil, the comparison between the 50th percentile (22) and the reference group (19 • C, average across all study sites) reflects the effect of heat, while for Chile, the same comparison (14) reflects the effect of cold.Thus, based on the same exposure contrasts, the results are essentially different effects of temperature.
Measuring exposure accurately is an enduring challenge in environmental epidemiology.Most studies use fixed site or satellite measurements as proxies for true personal exposure (30,33,58).Small deviations (∼0.5 • C) in true exposure may have significant repercussions on physiological processes during critical windows of pregnancy, which, if missed, can produce misleading results.Residential mobility during pregnancy, which has been studied more in the context of air pollution, may also affect exposure to extreme temperature (64)(65)(66).The presence and extent of exposure misclassification depend on the proportion-changing residences during pregnancy, which vary by region (67,68).The amount of time spent in temperature-controlled environments could also misclassify exposure, which requires prospective studies designed to measure the time activity patterns of the participants.This information is seldom available from routine sources such as birth registries or hospital records.
Studies should use standardized outcome definitions, especially when countries use different definitions, e.g., stillbirths are recorded at 20 (US), 24 (UK), or 28 (WHO) weeks of gestation (46,49,(51)(52)(53)(54)(55)(56).Another example is late fetal death defined in a study as "death within the first 24 h of life" (39), which is inconsistent with the definition by WHO that states the first 24 h after birth as the early neonatal period.The availability of reliable data on spontaneous abortion and stillbirth is a challenge across regions.Accuracy of gestational age determines the classification of PTB, especially when the last menstrual period-based assessment is not verified by first trimester ultrasound.Studies examining temperature and birth weight should conceptually consider gestational age because it has been suggested to be in the causal pathway and adjustment could produce biased estimates (69).
As temperature, air pollution, and season are inter-related, to understand the independent associations, studies should develop frameworks including all known and potential risk factors in the context of the temperature-outcome relationship (e.g., Directed Acyclic Graphs).Temperature and air pollution vary by season.Perinatal outcomes have also been observed to vary by season, as do births (70).Several studies adjusted for air pollution, while others examined effect modification (10,13,28,35,39,54).Buckley et al. argued that temperature and some pollutants (e.g., ozone) likely share predecessors (71).The nature of the pollutant warrants careful consideration before inclusion with temperature in the same model.The same argument also applies to seasonality and temperature.Restricting births to a season is an option, though it limits generalizability and sample size.
Fixed cohort bias is a threat to validity, but it has generally been considered (19,72).It occurs when births, only within a fixed period, is included.This results in shorter pregnancies being missed at study initiation, while longer pregnancies are missed at the end of the study (19).To avoid fixed cohort bias, studies have used conception dates within a fixed period or excluded pregnancies conceived before a fixed time prior to the starting date of the study (5, 8, 55).

. Gaps in current knowledge
From the perspective of exposure to extreme temperatures, several questions remain unanswered.Several studies have reported associations with multiple exposures windows, separately.It remains unclear, which exposure(s) is(are) relatively more harmful -a single extreme, several consecutive extremes, several interspersed extremes or moderately high prolonged exposures?Examining the overlap of these extremes with stages of fetal development will advance the understanding of the pathophysiological pathways.Fetal growth is programmed around the first month, when the placental bed is formed, which could be sensitive to exogenous exposures and merits investigation.Conversely, lean body mass growth occurs in the third trimester, a period that is also vulnerable to environmental exposures.It will be analytically complex to examine more than one exposure simultaneously (e.g., short-and long-term exposures), especially when they are biologically relevant and statistically correlated.Models have been developed that can examine multiple correlated exposures (e.g., Bayesian Kernel Machine Regression and distributed linear and non-linear lag models) (73)(74)(75).
Studies have provided some insights into potential biological mechanisms.For example, heat has been shown to affect thermoregulation and blood pressure during pregnancy (46,(76)(77)(78).Around 37 weeks when thermoregulation becomes less efficient because of increased body mass and decreased ability to dissipate heat by sweating, heat-related uterine contraction has been observed (28,79).Extreme temperature exposure during pregnancy has been linked with systemic inflammation, cell impermeability, upregulation of heat shock proteins, release of endotoxins (46,80), and gene sequencing (81,82).Nonetheless, we do not know how core body temperature modulates under extreme ambient temperature, which could be an intermediate link between exposure and biological mechanisms.The fetus is usually ∼0.5 • C hotter than the maternal core and can lose heat only through the umbilical artery and amniotic fluid (83).In the event of maternal core temperature increase, heat likely gets transferred to the fetus (84,85).Any physiologic or systemic effects interfering with maternal-fetal exchange will likely affect fetal development, growth, and potential survival.We do not know the magnitude of non-modulation of body temperature that is clinically important, i.e., when the body starts triggering physiological counterbalances.Does non-modulation differ in different populations and what factors (e.g., lifestyle characteristics) influence it?None of these questions have been addressed to date.Bonell et al. (86) are taking an important stride in that direction.More studies are needed in different populations to understand the modulation of core body temperature during pregnancy, transfer of excess heat from the mother to the fetus, and its consequences.Heat sensing and pregnancy monitoring wearables have been feasibly implemented in various settings including in low-and middleincome countries (87-92).These wearable devices are small.Prices range from fifty to several hundred US Dollars and come in various forms such as watches, rings, bracelets, and small buttons.The devices can store a reasonable amount of data, and the frequency of measurement determines the duration of a cycle.Technology involved in transferring data from these devices for further use in research varies in complexity.In addition to body temperature, some of these devices can measure ambient temperature, sleep, physical activity, etc. Technology has opened up avenues for accurate measurement of environmental exposures and helped advance research, though it is necessary to test these devices for accuracy before extensive use.
Populations acclimatize to their natural surroundings.A maximum temperature that is heat wave in temperate zones will be perceived differently in tropical countries.Localized behavioral adaptations during extreme heat or cold, such as modifying outdoor activities and consumption of specific food types, help mitigate adverse effects, predisposing populations to varying levels of risks.Most of the existing studies were conducted in the Western Hemisphere.New studies should be conducted on populations exposed to extreme temperatures and in low-and middle-income countries to understand the global risk pattern.
Little is known about the cost-effective interventions that can be implemented in vulnerable communities to reduce adverse outcomes.Vulnerable communities are likely to have fewer mitigation measures (e.g., air conditioning and lack of access to health care), particularly in rural and socioeconomically deprived areas.These communities are also likely to have a higher prevalence of pre-existing conditions and high-risk lifestyle (e.g., substance use and drinking), disposing pregnant women to higher risk, regardless of heat exposure.
For a comprehensive understanding of the nature of the effect, studies should be conducted in places where populations are frequently exposed to extreme temperature.Several other gaps in the available literature exist, including no knowledge about modulation of core body temperature during pregnancy, when exposed to extreme ambient temperature, which can help understand the biological mechanisms.Even if the epidemiologic associations of temperature with PTB, LBW, SGA, and stillbirths are small, these are clinically relevant indicators of newborn health and are observed across populations in varying magnitude.Despite a small association, continuing global warming will potentially increase exposure prevalence, thereby increasing the death and disability burden attributable to temperature, as well as the associated healthcare cost of the affected infants.

TABLE Summary
TABLE Summary of studies conducted between and examining the association between temperature and birth weight, low birth weight or small-for-gestational age.