Taken together, our findings demonstrate that physiological indicators of heat strain in pregnant women are significantly influenced by environmental conditions while working in the heat, particularly in humid conditions and in the third trimester of pregnancy compared to the second. This study provides information on the pathophysiological effects of heat stress on pregnant women and not on the respective health outcomes of mothers or babies. We applied a series of methods, namely the Pearson product moment correlation method, mixed effect linear models with random intercepts, and confirmatory composite analysis to assess the association between each environmental and physiological indicator. This approach enabled us to disentangle the independent effects of each of the environmental variables on a set of different physiological variables and assess the potential compound effects between climate variables, while taking into consideration variance inflation factors. Overall, our findings provide an input to quantify the risk of pregnant women experiencing pathophysiological effects from working in the heat and can thus support further research on adaptive measures to alleviate heat strain.
Our methodology allowed us to demonstrate that the heat strain of pregnant women was more severe under conditions above a relative humidity of 50%, whereby the impact of rising temperature on skin temperature increased at a greater rate. This is due to the limited capacity of the human body to thermoregulate in hot and humid environments, because the proportion of evaporated sweat, and thus the sweating efficiency, decreases with rising relative humidity above the 35°C threshold 15. The existence of the 35°C temperature threshold as a human adaptability limit to heat stress 22 was found to be even lower by the recent study of Vecciello et al 23 in non-pregnant individuals.
Furthermore, we found robust evidence for the increase in susceptibility to heat strain throughout pregnancy at rising temperatures across all physiological indicators, namely skin temperature, heart rate, tympanic temperature and the thereby estimated core temperature. More precisely, women in their third trimester of pregnancy experienced higher heat strain than women in their second trimester of pregnancy did. It would be interesting for future research to compare these findings with data from early pregnancy, given that a previous study has shown that the risk of heat-related stillbirth might be particularly high in early pregnancy 24, whilst a global analysis and meta-analysis detected an increased risk of preterm birth at exposure to extreme heat during the last seven days of gestation 12, 25.
Our study demonstrated the convergent validity of heat stress indices through correlations between each other as well as their construct validity through their correlations with skin temperature and tympanic temperature. The Heat Index, UTCI, Apparent Temperature and Wet Bulb Globe Temperature were strongly associated with each other, even though they incorporate different input variables such as relative humidity, solar radiation, black globe temperature and air velocity, in addition to air temperature. The underlying reason for the associations found is the strong weight that each of the heat stress indices assigns to air temperature as an input variable. Furthermore, the selected heat stress indices showed similar associations with both skin temperature and tympanic temperature, which supports their applicability to our study setting. Even though heat stress indices performed similarly in our assessment, we recommend including indices such as the UTCI or WBGT in heat-health warning systems, given that these indices incorporate air temperature, relative humidity and solar radiation, which showed associations to physiological variables in our study (Section 2.2). These indices could potentially be included in heat-health warning systems as a preventive measure for heat strain. Heat stress indices would need to be communicated in a targeted way, such as through simply understandable risk levels, as to allow for the public to respond appropriately.
There are five key limitations to our study. First, we tested the applicability of four out of more than 100 existing heat stress indices 26. However, the Heat Index, Apparent Temperature, WBGT, and UTCI are indices that are often referred to in the literature 13 and require input variables covered with our datasets, together with additional solar radiation modelling. We also tried to overcome this limitation by assessing the separate effects of environmental variables on maternal physiology, which increases the generalizability of the results. Second, given data availability constraints, our study design omitted variables such as clothing thermal insulation, systolic blood pressure, and diastolic blood pressure, which have been included in other studies 16, 21. Likewise, water intake or cloud cover could have been added to the models as confounders. Third, we acknowledge the potential imprecision of results sourced from the matching of datasets with different temporal resolutions, especially for heart rate measurements, which fluctuate over time. We reached no clear conclusions about the effects of heat stress on heart rate and could not determine construct validity of heat stress indices on heart rate, core temperature estimates or the physiological strain index, possibly due to behavioural adaptations to heat not captured in estimated metabolic rate. Even while applying a moving average merging technique (Supplementary Table 5) and confirmatory composite analysis (Supplementary Table 9), we found no robust estimates for the compound effects of air temperature and relative humidity on heart rate, possibly also due to non-accounted behavioural responses. This limitation might be resolved in future studies while measuring metabolic rate more effectively to disentangle the effects of activity and heat stress. Fourth, we did not account for individual differences in our confirmatory composite analysis; therefore, our results do not allow us to make inferences about repeated measurements of the same study participants. Such individual-level differences were only included in our mixed effect models with random intercepts. Fifth, we considered a linear association between variables and did not account for non-linearity, based on previous assessments with this data 1.
Despite these limitations, our understanding of the interplay between environmental variables and the physiology of pregnant women in The Gambia first and foremost underlines the necessity of climate change mitigation and adaptation to protect one of the most vulnerable population subgroups from heat strain. We suggest developing adequate and targeted adaptation strategies that increase the resilience of pregnant women to climatic changes, given the evidence of high levels of physiological impacts while working in hot environments. To protect the maternal health of agricultural workers, particularly during the third trimester of pregnancy, when vulnerability to heat strain increases, measures that reduce exposure to heat stress, including work schedules adapted to climatic conditions, or measures that reduce the impacts of heat stress, such as protective clothing or microclimate cooling equipment, could be tested in future research 27. The establishment of heat-health warning systems with tested heat stress indices will be crucial for the implementation of adaptive measures 4. Cherisch et al. 3 suggest interventions to protect maternal and newborn health in Africa at the individual level through behavioral change, at the infrastructural level through changes in health systems and services, at the structural level through policy and financing options, or at the environmental level through nature-based solutions. For optimal policy design and implementation, adaptive measures should be developed in close collaboration with the local community, and investigations of the respective measures should be undertaken through a participatory research approach 28.