Adverse impacts of high temperature on adults and the elderly have been documented.1–3 Few studies have evaluated pregnant women, another potentially vulnerable population.4,5 Although the mechanisms by which temperature could adversely affect pregnancy are unclear, one possible route is for short-term heat stress to trigger labor earlier than expected.6 Researchers have investigated the relationship between ambient heat and preterm birth.6–12 In the 1990s, ambient heat in the United States was correlated with aggregated preterm labor rates, but not preterm delivery8 or mean gestational age.10 Time-series analysis has shown that high ambient temperature was associated with preterm birth and premature rupture of membranes in Israel,12 but not in the United Kingdom.9 Thus, findings have been mixed or limited by underpowered ecologic approaches that poorly account for bias arising from aggregation of individual-level data.
More recently, support for an association between increased temperature and preterm birth was found in California using case-crossover analysis,7 a study design that inherently accounts for individual characteristics (each birth is in its own control, and date of birth is matched to days nearby).13 Case-crossover designs have, however, been criticized for their potential to inadequately account for short-term variation in conception patterns,14 a source of bias that arises because birth timing is a function of underlying conception rates from months before that later may generate more births on warm (or cool) days.15 This concern is potentially justified by the association between season and preterm birth described in other research.16,17 Moreover, preterm births arise from an underlying population of fetuses-at-risk, the majority resulting in term births, including some whose gestational age may have been shortened by exposure to high temperature but not enough to cause preterm birth. Thus, if high ambient temperature is a trigger for labor, the overall risk of delivery (ie, both term and preterm births) would be increased after hot days. All these issues imply that analyses of preterm births alone may not be fully informative and may even conclude that heat is associated with preterm birth when seasonal differences in conception were responsible or an expected number of preterm births occurred relative to term births.
An alternate approach to these problems is to use the entire population of births in individual-level analyses aimed at determining whether the overall number of births is greater for hot days compared with cooler ones, while accounting for potential bias from seasonal conception patterns using a time-to-event study design (because the conception date cannot affect length of gestation).18,19 A recent study using such an approach found that high outdoor temperatures in a subtropical urban area were associated with preterm birth.11 A study on hospital births in Spain suggested a slight reduction in gestational age with temperatures of 30°C or higher,6 but sample size limitations prevented the researchers from determining the association with preterm birth, and bias from seasonal conception patterns could not be excluded (as a time-to-event design accounting for fetuses-at-risk was not used). Because preterm birth is associated with significant morbidity and mortality,20 and rates are increasing in many countries,21 a better understanding of the association with high outdoor temperature is warranted.
In view of the increasing frequency of heat waves due to climate change,4,22,23 we aimed to determine the association between maximum outdoor temperature and birth timing in a time-to-event analysis of individual-level data for an urban population of North America.
Data and Variables
We used a time-to-event study design that accounted for all fetuses-at-risk, rather than births only.19 Data were obtained from the Quebec, Canada, birth file that provides complete coverage of live births (ie, newborns showing any sign of life at the moment of delivery) in the province through registration certificates. We extracted singleton live births in the city of Montreal for the 30-year period spanning 1981 to 2010. For each year, the risk period started June 1 and ended September 30, covering the 4-month span when extremely high temperatures occur in Montreal. Births from October through May, when temperatures were much cooler or sub-zero, were not considered events but were included as fetuses-at-risk if they were at risk of preterm or term birth from June to September. The final sample consisted of 219,319 births (not including 2,220 with missing data on gestational or maternal age) plus an additional 193,207 fetuses born after September but at risk during June to September.
The outcomes were preterm and term birth. Gestational age <37 weeks was defined as preterm. Term births were further categorized as early term (37–38 weeks) or full term (≥39 weeks).24 Gestational age was available in completed weeks based on ultrasound examinations, providing relatively accurate estimates.25 Menstrual dating may have been used in the 1980s, potentially misclassifying gestational age for a proportion of births (though with no reason to suspect that misclassification differed by temperature).
Exposure was defined as the maximum outdoor temperature reached on the day of birth or any of the 6 preceding days (ie, the week before birth), following previous research which used similar time-fixed measures.6,7,12 Maximum temperature can capture thermal stress during summer, but not during cold months when minimum temperature is a more relevant measure of exposure in Montreal. We restricted the exposure window to 1 week because it was not clear that earlier exposures could directly trigger labor via thermoregulatory mechanisms. Hourly temperature data were obtained from the Environment Canada meteorological center situated 20 km from the city core.26 Maximum temperature was evaluated as a continuous exposure, using 20°C as referent. A temperature of 20°C is unlikely to be associated with thermal stress and has been used as a referent in previous Montreal-based research.26 In addition, we examined the maximum temperature of each preceding day separately, in the event that lagged associations were masked by one maximum temperature over the week. We also considered mean (rather than maximum) temperature.
Because labor might be triggered by persistent heat-related stress rather than brief heat exposures of short duration, we evaluated two indicators of extreme heat episodes—the first defined as 3 or more consecutive days with temperatures of 32°C or above in the week before birth (dichotomous)27 and the second defined as cumulative exposure to 0, 1, 2, 3, and 4 to 7 days with maximum temperature of at least 32°C (categorical). We used both indicators because there is no standard definition of extreme heat episodes.6
Humidity and air pollution were considered potential confounders of the association between temperature and risk of delivery. Humidity was measured using mean percent relative humidity during the week before birth (continuous). Air pollution was captured using mean daily concentration of ozone (O3) in parts per billion and particulate matter smaller than 2.5 μm (PM2.5) in microgram per cubic meter across all Environment Canada monitoring stations in Montreal during the preceding week, expressed continuously.28 Unlike humidity, air pollution data were available only after year 1999. The following variables were included as covariates: maternal age (<20, 20–34, or ≥35 years), education (no high school diploma, high school diploma, some postsecondary, some university or more, or unknown), marital status (legally married or not legally married), immigration (Canadian-born, foreign-born, or unknown), language spoken at home (French, English, other language, or unknown), parity (0, 1, or ≥2 previous deliveries), month (categorical), and period (1981–1990, 1991–2000, or 2001–2010).
We used Cox proportional hazards regression with preterm or term birth as the outcome and gestational week in the time axis.18,19 We estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for maximum temperature (or extreme heat episodes) and birth in unadjusted models, and in models adjusted for relative humidity, maternal age, education, marital status, immigration, language, parity, month, and period. Adjustment for O3 and PM2.5 was undertaken in subset analyses for the year 2000 onward. We ran separate analyses for preterm (censoring at ≥37 gestational weeks), early-term (censoring at ≥39 weeks and excluding births <37 weeks), and full-term births (excluding births <39 weeks). For each outcome, we also censored pregnancies that delivered during October or later that were part of the pool of fetuses-at-risk from June through September. We tested the proportional hazards assumption using temperature-by-gestational age (or extreme heat episode-by-gestational age) interaction terms.29 Nonlinear associations between continuous temperature and the hazard of birth were assessed with restricted cubic splines.30 Knots for splines were placed at the 10th, 50th, and 90th temperature percentiles (additional knots did not influence results). A period-by-temperature interaction term was tested to evaluate potential modifying effects of calendar time on the relation between temperature and birth risk.
Statistical analyses were undertaken using the RCS macro in SAS 9.2 (SAS Institute Inc., Cary, NC).32 Data were de-identified, and the institutional review board of the University of Montreal Hospital Centre waived the requirement for ethics review.
Maximum daily temperatures in the week preceding birth ranged from 10.1 to 35.4°C (median 27.6°C; interquartile range 4.5°C) (Table 1). There was little difference in maximum temperature between preterm and term births. Maximum temperatures reached 32°C or more for 19,829 births (9.0%; Table 2), of which 1121 were preterm and 4322 early-term (Table 2). The mean gestational age of births did not vary with temperature, and was not lower during extreme heat episodes. Preterm birth rates did increase slightly, from 5.4% at maximum temperatures less than 20°C to 5.8% at temperatures 28.0°C or higher. There was no increase, however, in preterm birth rates during extreme heat episodes.
Results of the Cox regression analysis suggested a linear association between increased maximum temperature and risk of delivery among term pregnancies (Figure 1), with little difference between models adjusted and unadjusted for maternal characteristics. There was a steady rise in the hazard of delivery at term with increasing temperature up to 35°C, at which point the hazard was almost 4% higher relative to 20°C. Additional adjustment for air pollution in the subset of births after 1999 did not weaken the associations, although it did reduce statistical power. There was, however, no evidence that increased temperatures were associated with risk of preterm delivery before 37 weeks. In addition, the association among term births was more pronounced before 39 weeks, reaching a 10% greater hazard of delivery at 35°C relative to 20°C, compared with a 3% greater hazard of delivery at 39 weeks and later (Figure 2). Temperature-by-gestational age interaction terms suggested that hazards were proportional over the preterm, early-term, and full-term periods. In models using lagged temperature, the hazard of early-term (but not preterm) birth tended to increase with greater time since exposure, such that associations with maximum temperature were strongest for the fourth-day preceding birth (Figure 3).
Extreme heat episodes were also associated with a higher risk of delivery among early-term pregnancies, despite the low number of exposed pregnancies. Early-term pregnancies exposed to 3 consecutive days of temperatures of 32°C or higher had a 17% higher hazard of delivery relative to unexposed pregnancies (Table 3). Associations tended to be stronger for more prolonged heat exposures, such that early-term pregnancies exposed to 4 to 7 days of 32°C or higher had a 27% higher hazard of delivery relative to no days of exposure (HR = 1.27 [95% CI = 1.07–1.52]). Exposure to extreme heat episodes was not associated with a higher risk of delivery among preterm or full-term pregnancies.
There was little evidence that month of delivery was associated with risk of delivery among early-term or full-term pregnancies, but risk of preterm birth was 10% higher for June–August relative to September. Finally, interaction terms for period-by-temperature suggested no change in associations over calendar time, and mean temperature over the preceding week was not associated with risk of delivery.
Exposure to high ambient summer temperature and extreme heat episodes was weakly associated with increased risk of delivery among term pregnancies (but not among preterm pregnancies) in a continental North American city. Associations were more pronounced during the early-term (37–38 weeks) than full-term (≥39 weeks) period. There seemed to be a linear trend with increasing maximum temperature, as well as a lagged association, such that high ambient temperatures of ≥30°C approximately 4 days before delivery were most strongly associated with delivery among early-term pregnancies. These findings suggest that exposure to very hot ambient temperature may shorten gestational length of pregnancies that reach term during summer, but not enough to discernibly affect preterm birth rates. Although the impact on health of infants born early term is unclear, evidence is emerging that morbidity of early-term births is greater than later births.24,33,34
Our findings contrast somewhat with research carried out in Brisbane, a subtropical Australian city. Average temperatures of 25°C during the week before birth in Brisbane were associated with 10% higher hazards of preterm delivery relative to that at 21°C, with a less increased risk of delivery among term pregnancies (approximate HR = 1.02). However, CIs were wide, and mean temperature is not a marker of extreme heat exposure. Indeed, our analysis suggested no association with mean temperature; thus, maximum temperature may be more appropriate to capture extreme heat exposures. Lagged associations during the week before birth and extreme heat episodes were not considered in the Brisbane study.11 Generalizability of the findings to areas with continental climates is also unclear. In Barcelona, heat indices exceeding the 95th percentile (>30°C) during the day preceding birth were associated with a 1 to 2 day reduction in gestational age.6 A linear trend with increasing temperature was also found in the Barcelona study, although lagged analyses suggested that the critical exposure window was the day before birth rather than earlier. The sample included only 7585 pregnancies, and so the association with preterm birth could not be evaluated. Similarly, there were not enough births to determine the association with prolonged extreme heat episodes. The number of preterm births in Montreal was higher, but the absence of an association with ambient temperature or extreme heat episodes could simply reflect the possibility that heat triggers labor at term primarily, with little impact before term.
Unfortunately, little attention has been paid to clinical mechanisms by which heat may affect pregnancy. Studies of small samples of pregnant women (<30) suggest that heat stress increases uterine contractility,35,36 and that sensitivity to heat is greater late in gestation when thermoregulation may be less efficient.7 Dehydration due to heat could reduce uterine blood flow, potentially increasing excretion of pituitary hormones which induce labor.37 Pregnant women with comorbidities may be more sensitive to thermal stress and to endocrine hormones that trigger parturition.12 A hospital-based study in Paris linked a heat wave with oligohydramnios, a condition characterized by insufficient amniotic fluid,38 which aligns with a follow-up study of 42 pregnant women suggesting that amniotic fluid volume fluctuates with ambient temperature.39 Although the pathophysiology of oligohydramnios is complex, heat-related mechanisms are plausible and could be part of the pathway leading to labor onset during extreme heat. Mechanisms linking thermal stress and labor are clearly an avenue for future research.
In contrast to clinical research, limitations of studies addressing seasonal exposures have received greater attention in the literature, for several reasons. First, temperature and season are highly correlated, and conception rates from months before that were seasonally dependent may in part account for later associations between temperature and birth.14,15 A related issue is that associations between seasonal exposures and birth outcomes may be affected by “fixed cohort bias” from noninclusion of shorter (or longer) pregnancies at the start (or end) of data collection.11,40 However, we censored fetuses that were born after September, ruling out bias from noninclusion of longer pregnancies. Furthermore, recent evidence suggests that bias from left truncation is low when truncation varies by gestational age and is nondifferential over the temperature range.41 This was likely the case in our study, because the study start was set to June 1 regardless of gestational age or temperature in preceding months.
Time-stratified case-crossover analysis is an emerging tool to assess relations between temperature and perinatal outcomes.7 Case-crossover designs inherently adjust for individual characteristics including season and are useful for transient exposures such as temperature fluctuation with acute events such as birth.13 However, case-crossover analyses do not account for fetuses-at-risk, and are inappropriate for common outcomes such as term birth. Further, results may be biased by selection of controls from days after delivery when cases are no longer at risk.42,43 Methods for analyzing cohort data, such as time-to-event regression, are advantageous, particularly when time to birth is the outcome. Bias due to seasonal conception rates can be minimized by restricting data to months when high temperatures are reached. It is difficult to see how seasonal exposures at conception could influence the timing of delivery more than immediate exposures such as temperature, and thus time-to-event analysis is a natural way to account for exposures at conception. Future research may benefit from addressing ways to account for possible feedback relations between time-varying temperature and confounders,44 particularly in geographic regions where wide variation in exposures occur over the course of pregnancy with changes in season.
This study had limitations. We evaluated deliveries only, and could not determine if temperature was associated with labor not resulting in birth. We could not distinguish spontaneous preterm birth from other subtypes, which may have masked associations at low gestational ages. We did not have data on cesarean delivery or miscarriage. Stillbirths were not included as a competing outcome, but fetal deaths are rare in Montreal and the causal pathways potentially differ—a topic for future research. We investigated heat in the week preceding birth but not longer exposure windows. Given that associations were strongest 4 days before birth in Montreal, and the day before birth in Spain,6 immediate exposures seem more relevant and biologically plausible. We did not have information on individual-level covariates not recorded on birth registration certificates such as maternal comorbidity, income, or smoking. Similarly, we did not have information on use of air conditioning, which may mitigate heat exposure. We suspect that misclassification of temperature exposure was nondifferential, and thus our associations likely represent conservative estimates. An important aspect of our study was the use of individual-level birth data rather than aggregated daily number of births, a characteristic of ecologic studies. Although common in environmental research,45 ecologic studies constrain inference at the individual level.
This study found that increased ambient temperature during summer, as well as extreme heat episodes with prolonged exposure to very high temperatures, was weakly associated with risk of delivery among early-term pregnancies, but not among preterm pregnancies. Thus, extreme heat in continental areas may trigger labor late in gestation. Research is needed to verify the generalizability of these study findings and to explore the relationship with other birth outcomes.
We thank Allan Brand, Constance Brossier and Alison Park for research assistance.
1. Baccini M, Biggeri A, Accetta G, et al. Heat effects on mortality in 15 European cities. Epidemiology. 2008;19:711–719
2. Hajat S, O’Connor M, Kosatsky T. Health effects of hot weather: from awareness of risk factors to effective health protection. Lancet. 2010;375:856–863
3. Lin S, Luo M, Walker RJ, Liu X, Hwang SA, Chinery R. Extreme high temperatures and hospital admissions for respiratory and cardiovascular diseases. Epidemiology. 2009;20:738–746
4. Sheffield PE, Landrigan PJ. Global climate change and children’s health: threats and strategies for prevention. Environ Health Perspect. 2011;119:291–298
5. Balbus JM, Malina C. Identifying vulnerable subpopulations for climate change health effects in the United States. J Occup Environ Med. 2009;51:33–37
6. Dadvand P, Basagaña X, Sartini C, et al. Climate extremes and the length of gestation. Environ Health Perspect. 2011;119:1449–1453
7. Basu R, Malig B, Ostro B. High ambient temperature and the risk of preterm delivery. Am J Epidemiol. 2010;172:1108–1117
8. Lajinian S, Hudson S, Applewhite L, Feldman J, Minkoff HL. An association between the heat-humidity index and preterm labor and delivery: a preliminary analysis. Am J Public Health. 1997;87:1205–1207
9. Lee SJ, Hajat S, Steer PJ, Filippi V. A time-series analysis of any short-term effects of meteorological and air pollution factors on preterm births in London, UK. Environ Res. 2008;106:185–194
10. Porter KR, Thomas SD, Whitman S. The relation of gestation length to short-term heat stress. Am J Public Health. 1999;89:1090–1092
11. Strand LB, Barnett AG, Tong S. Maternal exposure to ambient temperature and the risks of preterm birth and stillbirth in Brisbane, Australia. Am J Epidemiol. 2012;175:99–107
12. Yackerson N, Piura B, Sheiner E. The influence of meteorological factors on the emergence of preterm delivery and preterm premature rupture of membrane. J Perinatol. 2008;28:707–711
13. Maclure M, Mittleman MA. Should we use a case-crossover design? Annu Rev Public Health. 2000;21:193–221
14. Darrow LA, Strickland MJ, Klein M, et al. Seasonality of birth and implications for temporal studies of preterm birth. Epidemiology. 2009;20:699–706
15. Basso O, Olsen J, Bisanti L, Juul S, Boldsen J. Are seasonal preferences in pregnancy planning a source of bias in studies of seasonal variation in reproductive outcomes? The European Study Group on Infertility and Subfecundity. Epidemiology. 1995;6:520–524
16. Lee SJ, Steer PJ, Filippi V. Seasonal patterns and preterm birth: a systematic review of the literature and an analysis in a London-based cohort. BJOG. 2006;113:1280–1288
17. Strand LB, Barnett AG, Tong S. The influence of season and ambient temperature on birth outcomes: a review of the epidemiological literature. Environ Res. 2011;111:451–462
18. Auger N, Gamache P, Adam-Smith J, Harper S. Relative and absolute disparities in preterm birth related to neighborhood education. Ann Epidemiol. 2011;21:481–488
19. Platt RW, Joseph KS, Ananth CV, Grondines J, Abrahamowicz M, Kramer MS. A proportional hazards model with time-dependent covariates and time-varying effects for analysis of fetal and infant death. Am J Epidemiol. 2004;160:199–206
20. Moster D, Lie RT, Markestad T. Long-term medical and social consequences of preterm birth. N Engl J Med. 2008;359:262–273
21. Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. Lancet. 2008;371:75–84
22. Patz JA, Campbell-Lendrum D, Holloway T, Foley JA. Impact of regional climate change on human health. Nature. 2005;438:310–317
23. Ebi KL, Mills DM, Smith JB, Grambsch A. Climate change and human health impacts in the United States: an update on the results of the U.S. national assessment. Environ Health Perspect. 2006;114:1318–1324
24. Spong CY. Defining “term” pregnancy: recommendations from the Defining “Term” Pregnancy Workgroup. JAMA. 2013;309:2445–2446
25. Taipale P, Hiilesmaa V. Predicting delivery date by ultrasound and last menstrual period in early gestation. Obstet Gynecol. 2001;97:189–194
26. Smargiassi A, Goldberg MS, Plante C, Fournier M, Baudouin Y, Kosatsky T. Variation of daily warm season mortality as a function of micro-urban heat islands. J Epidemiol Community Health. 2009;63:659–664
27. Smoyer-Tomic KE, Kuhn R, Hudson A. Heat wave hazards: an overview of heat wave impacts in Canada. Nat Hazards. 2003;28:463–485
28. Jacob DJ, Winner DA. Effect of climate change on air quality. Atmos Environ. 2009;43:51–63
29. Hess KR. Assessing time-by-covariate interactions in proportional hazards regression models using cubic spline functions. Stat Med. 1994;13:1045–1062
30. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med. 1989;8:551–561
32. Heinzl H, Kaider A. Gaining more flexibility in Cox proportional hazards regression models with cubic spline functions. Comput Methods Programs Biomed. 1997;54:201–208
33. Boyle EM, Poulsen G, Field DJ, et al. Effects of gestational age at birth on health outcomes at 3 and 5 years of age: population based cohort study. BMJ. 2012;344:e896
34. Crump C, Sundquist K, Winkleby MA, Sundquist J. Early-term birth (37-38 weeks) and mortality in young adulthood. Epidemiology. 2013;24:270–276
35. Khamis Y, Shaala S, Damarawy H, Romia A, Toppozada M. Effect of heat on uterine contractions during normal labor. Int J Gynaecol Obstet. 1983;21:491–493
36. Vähä-Eskeli K, Erkkola R. The effect of short-term heat stress on uterine contractility, fetal heart rate and fetal movements at late pregnancy. Eur J Obstet Gynecol Reprod Biol. 1991;38:9–14
37. Stan C, Boulvain M, Hirsbrunner-Amagbaly P, Pfister R. Hydration for treatment of preterm labor. Cochrane Database Syst Rev. 2002:CD003096
38. Luton D, Alran S, Fourchotte V, Sibony O, Oury JF. Paris heat wave and oligohydramnios. Am J Obstet Gynecol. 2004;191:2103–2105
39. Sciscione AC, Costigan KA, Johnson TR. Increase in ambient temperature may explain decrease in amniotic fluid index. Am J Perinatol. 1997;14:249–251
40. Strand LB, Barnett AG, Tong S. Methodological challenges when estimating the effects of season and seasonal exposures on birth outcomes. BMC Med Res Methodol. 2011;11:49
41. Schisterman EF, Cole SR, Ye A, Platt RW. Accuracy loss due to selection bias in cohort studies with left truncation. Paediatr Perinat Epidemiol. 2013;27:491–502
42. Levy D, Lumley T, Sheppard L, Kaufman J, Checkoway H. Referent selection in case-crossover analyses of acute health effects of air pollution. Epidemiology. 2001;12:186–192
43. Darrow LA. Invited commentary: application of case-crossover methods to investigate triggers of preterm birth. Am J Epidemiol. 2010;172:1118–1120
44. Buckley JP, Samet JM, Richardson DB. Commentary: does air pollution confound studies of temperature? Epidemiol. 2014;25:242–245
45. Herring A, Naimi AI. The ecological design. BJOG. 2012;119:1638–1639