Dr D. A. Lawlor, Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol, BS7 8PR UK.
Background We assessed the effect of mean ambient outdoor temperature during gestation on birthweight.
Objective To assess the effect of mean ambient outdoor temperature during gestation on birth weight.
Design Birth cohort study with record linkage to climate databases.
Setting Aberdeen, Scotland.
Sample A total of 12,150 individuals born in Aberdeen, Scotland between 1950 and 1956.
Methodology/Principal findings Perinatal data from a cohort of 12,150 individuals born in Aberdeen, Scotland between 1950 and 1956 were linked to daily outdoor temperature data. Birthweight was seasonally patterned, with lowest birthweights among those born in the winter months (December–February) and highest birthweights among those born in the autumn months (September–November); P= 0.01 for joint sine–cosine functions. Mean ambient outdoor temperature during the first trimester of pregnancy was inversely associated with birthweight and mean ambient outdoor temperature during the third trimester of pregnancy was positively associated with birthweight. In fully adjusted (for sex, maternal age, birth year, birth order and social class) models a 1°C increase in mean ambient outdoor temperature in the mid 10-day period of the first trimester was associated with a 5.4-g (95% confidence interval [CI] 2.9, 7.9 g) decrease in birthweight, whereas a 1°C increase in the mid 10-day period of the third trimester was associated with a 1.3-g (95% CI 0.50, 2.1 g) increase in birthweight. Ambient outdoor temperature in the first trimester of pregnancy explained the seasonal patterning of birthweight.
Main outcome measure Birthweight.
Results Birthweight was seasonally patterned, with lowest birthweights among those born in the winter months (December–February) and highest birthweights among those born in the autumn months (September–November); P= 0.01 for joint sine–cosine functions. Mean ambient outdoor temperature during the first trimester of pregnancy was inversely associated with birthweight and mean ambient outdoor temperature during the third trimester of pregnancy was positively associated with birthweight. In fully adjusted (for sex, maternal age, birth year, birth order and social class) models a 1°C increase in mean ambient outdoor temperature in the mid 10-day period of the first trimester was associated with a 5.4 g (95% confidence interval 2.9, 7.9 g) decrease in birthweight, whereas a 1°C increase in the mid 10-day period of the third trimester was associated with a 1.3 g (95% confidence interval 0.50, 2.1 g) increase in birthweight. Ambient outdoor temperature in the first trimester of pregnancy explained the seasonal patterning of birthweight.
Conclusion Higher ambient outdoor temperature in the first trimester of pregnancy and/or lower ambient outdoor temperature in the third trimester are associated with reduced offspring birthweight. With the increasing occurrence of temperature extremes, in particular, heat waves, these findings, if replicated in other studies, have important public health implications.
Low birthweight is associated with infant mortality and morbidity, and birthweight is inversely related to cardiovascular disease and diabetes in later life.1–4 Low socio-economic position and smoking during pregnancy are modifiable risk factors for low birthweight,5 but little is known about other potentially modifiable exposures. Several lines of evidence suggest that ambient outdoor temperature during pregnancy also affects offspring birthweight.
Humans at all stages in the life course need to maintain a relatively constant core body temperature in order to maintain normal physiological functioning, and they suffer morbidity and mortality at extremes of both ends of the temperature scale.6–8 Animals subjected to heat stress during pregnancy tend to produce lower birthweight offspring and those subjected to cold stress to produce higher birthweight offspring.9–11 However, these results may have limited relevance to human populations as the animals most frequently studied have markedly different mechanisms to humans for dealing with temperature extremes and have differing physiological responses to ambient outdoor temperature.12 International comparative studies in humans also suggest that heat stress during pregnancy results in low birthweight,13 but these studies are liable to suffer from the ecological fallacy and are unable to assess the effects of temperature variation at different stages of gestation.
Birthweight shows seasonal variability in both developed and developing countries. While the seasonal patterning in developed countries is closely linked to seasonal patterns in food availability and maternal weight loss in pregnancy,14–17 in developed countries, where overall nutrition is generally good, any seasonal patterning of birthweight suggests more subtle or different processes. Three studies in developed countries have found that babies born in the winter (coldest) months have the lowest birthweights.18–20 Two other studies have found that those born in the spring have the lowest birthweight.21,22 Finally, one study in a developed country found no seasonal patterning of birthweight.23 Most of these studies have been unable to adjust for potential confounding factors.18,22,23 One used self-reported birthweight among a cohort of older women and had birthweight data on only one-third of participants.19
Any seasonal patterning may be related to patterns of ambient outdoor temperature exposure during pregnancy. Conflicting results of studies of this phenomenon may be related to geographical, temporal and socio-economic differences between the study populations that would result in differences in their ability to avoid any direct consequences of extremes of outdoor ambient temperature. Birthweight, if influenced by temperature, will reflect differing seasonal patterns over the nine-month course of the pregnancy and therefore investigation of any potential effect of temperature in explaining seasonal patterning in birthweight requires examination of temperature effects in each trimester of gestation. Only one previous study has been able to assess the effect of temperature during pregnancy.21 It concluded that lower birthweight among individuals born in the spring was largely attributable to cold outdoor temperatures during the mothers' second trimester of pregnancy. However, that study did not report on temperature effects in the first and third trimesters.
The aims of this study are to assess the effect of mean ambient outdoor temperature during gestation on birthweight, and to assess whether any associations with temperature during gestation explain seasonal patterns in birthweight.
Data were used from the Aberdeen Children of the 1950s cohort study. This study has been described in detail previously.24 The study cohort is based on participants in the Aberdeen Child Development Survey.25 This survey collected data on the parental and childhood characteristics of 14,938 children who were in Aberdeen primary schools in 1962.25 For 12,150 of these children who were born in Aberdeen, comprehensive information was abstracted from the Aberdeen Maternal and Neonatal Database (AMND) about the course of their mother's pregnancy and the children's physical characteristics at birth.25 These 12,150 individuals born in Aberdeen between 1950 and 1956 form the index members of the Aberdeen Children of the 1950s cohort.24
Birthweight and gestational age were abstracted from obstetric records (AMND) at the time of the 1962 survey.24 The participants' intrauterine growth rate was estimated by calculating sex and gestational age standardised z (standard deviation) scores. Means and standard deviations of birthweight were calculated for participants by sex and single weeks of gestational age. The z scores were then calculated as a participant's birthweight minus the mean for their gestational age and sex category and divided by the standard deviation for their gestational age and sex category.
Temperature data were obtained from the archives of the Meteorological Office via the British Atmospheric Data Centre (http://badc.nerc.ac.uk/). Data from the local Craibstone weather station, which is situated close to Aberdeen Airport and approximately two miles from the city centre, were used. Data from this station were available for the period 1949–1956, which included the years in which the participants were born and their full gestational periods.
For the period of gestation of each subject, we calculated the mean outdoor temperature in the middle 10 days of each trimester of pregnancy. For the first trimester, the mid 10 days were defined as days 42–51 of gestation, for the second trimester, these were defined as days 135–144 and for the third trimester they were defined as days 228–237. The participants' gestational age (based on date of last menstrual period and converted from weeks to days) and date of birth were used to define day zero (the day of the last menstrual period). In addition, we estimated mean temperature around the time of conception as 10 days around ‘day zero’ (−4 to + 5) and mean temperature around the time of birth as 10 days around this date (four days before, the day of birth and five days after). Data on the minimum and maximum air temperatures (taken from 24 hour readings) for most of these days were available. For the maximum temperature data set, there was a recording for all days included in the analysis for 9815 (81%) of the cohort participants. For the minimum temperature data set, there were recordings for all but four of the participants. There was a high level of correlation between the two temperature measures (Pearson's correlation coefficient 0.85) and the results using either the minimum or the maximum temperature were similar. Therefore, results based on the daily minimum temperatures are presented here.
Social class at birth was based on the child's father's occupation as originally stated in the AMND records [seven categories: I—professional; II—managerial; III—non-manual–skilled non-manual; III—manual: skilled manual; IV—semi-skilled; V—unskilled manual and unemployed, which are collapsed in some analyses into two categories non-manual (I–IIInm) and manual (IIIm to V and unemployed)]. Maternal age at the birth of the child was also as originally stated in the AMND records and was recorded in seven 5-year age categories from 15–19 to 45 years or older. Birth order of the participant was obtained at the time of the 1962 survey.
Means of birthweight for each month of birth were estimated and examined graphically. Linear regression models were used to explore seasonal trends and temperature effects on birthweight and birthweight for gestational age z score. In these models, birthweight and birthweight for gestational age were the dependent variables. Month of birth, entered as an indicator variable, was the main exposure variable of interest in the season of birth models, and social class at birth, year of birth, birth order, maternal age at birth of the participant (all entered as indicator variables) and sex were entered as covariates. Very few of the participant's mothers were aged 45 or older when the participant was born (Table 1) and therefore participants in this category were combined with the 40–44 year old category in the regression models. Similarly, the 11 participants who were born in 1956 were combined with those born in 1955 in the regression models. Sex was excluded from the model with sex and gestational age standardised birthweight as the outcome.
Table 1. Study participant characteristics (N= 12,150). Some percentages do not add up to exactly 100 because of rounding. Values are presented as mean [SD] or n (%).
Gestational age (week)
Year of birth
Mother's age at birth of participant (6 participants' mothers could not be classified)
Birth order (data were missing on 90 participants)
7th or greater
Social class at birth (based on father's occupation)
I and II
In order to formally assess seasonal patterns of birthweight, graphs of mean birthweight by month of birth were examined and a sinusoid curve was fitted to the data by fitting sine and cosine functions into a regression model with birthweight as the dependent variable and assuming a six-month periodicity. An F test was used to test the joint effect of the sine and cosine functions.
Linear regression models were used to assess the association of the mean temperature measures with birthweight and birthweight for gestational age. In these models, the mean temperature variables (i.e. mean temperature for each of the 10 day periods around conception, in the middle of the first, second and third trimesters and around birth) were entered as categorical variables (fifths of the distribution) or as continuous variables. Other covariates were entered as described above in the models for the association of month of birth and birthweight. F tests were used to assess departure from a linear association between each temperature variable and birthweight. In order to determine whether any effects of temperature during gestation were independent of seasonal effects, separate analyses were performed with the association between temperature and birthweight for each season of birth: Winter (December–February); Spring (March–May); Summer (June–August) and Autumn (September–November).
F tests were used to assess statistical evidence of interactions with sex. There was no evidence of any statistical interactions between sex and month of birth or any of the temperature measures (all P values > 0.5), therefore results of the regression analyses are presented for both sexes combined. The possibility of social class modifying any of the associations was assessed by stratified analyses and F tests for statistical interactions.
Table 1 shows the distributions of characteristics of study participants.Table 2 shows the results of the regression models assessing the associations of month of birth and other covariates with birthweight. Our results for these associations with sex, maternal age, birth order and social class are in the same direction and of similar magnitudes to what would be expected from the existing literature. Female participants had lower birthweights than male participants and birthweight tended to be lower in the youngest and oldest maternal age groups. Birth order was strongly positively associated with birthweight, such that children who were the seventh or more in their family were 339 g heavier on average than those who were first born. Social class was strongly and linearly related to birthweight, with those from poorer social classes at birth having lower birthweight.
Table 2. Results of linear regression model assessing associations of month of birth with birthweight measures with adjustment for potential confounding factors.
In simple graphical analyses, birthweight (in both sexes) appeared lowest among those born in the winter months (December–February) and highest in those born in the autumn months (September–November) and showed statistical evidence of a seasonal sinusoidal pattern (for joint sine–cosine functions P= 0.04 in girls and P= 0.001 in boys). In crude analyses, birthweight was 30.3 g (95% confidence interval [CI] 0.24, 60.4 g) lower among those born in winter compared with all other months among girls and was 45.0 g (15.3, 74.7 g) lower among boys.
Table 2 shows that in adjusted analyses birthweight and birthweight for gestational age remained seasonally patterned (P= 0.01 for joint sine–cosine functions), with lower weights among those born in the winter months and higher birthweights for those born in the autumn months. With adjustment for gestational age, sex, maternal age, birth order, birth year and childhood social class, those who were born in the winter (December to February) had birthweights that were on average 33.5 g (13.5, 53.4) lower than those born in any other season. In stratified analysis, the seasonal effects on birthweight were similar in both those from manual and non-manual social classes in childhood (Fig. 1; P for interaction with social class = 0.79 for birthweight and 0.87 for z score of birthweight for gestational age and sex).
Figure 2 shows the variations in temperature across the year for each year 1950–1955. This shows similar seasonal patterns, with highest temperatures in summer and lowest in winter for each year.Table 3 shows the correlations between average ambient outdoor temperature around conception, during each trimester of gestation and around the time of birth. Given the average length of a human gestation and the temporal relationship of the five time periods to each other, these correlations are as one would expect. For example, temperature in the middle of the first trimester was strongly inversely correlated with temperature in the middle of the third trimester, which is to be expected as these two time-points will on average be six months a part.
Table 3. Pearson's correlation coefficients between temperature measures during pregnancy. T1, T2, T3 = first, second and third trimesters, respectively.
Table 4 shows the adjusted regression coefficients for the associations of average temperature in the middle of the first, second and third trimesters as well as around the time of conception and around the time of birth with birthweight. There was an inverse linear association between average outdoor temperature in the middle of the first trimester and birthweight, such that women who were in the first trimester of pregnancy when the outdoor temperature was hotter tended to have lower birthweight babies. The opposite effect was seen for temperature in the third trimester with women who were in the third trimester of pregnancy when the outdoor temperature was colder tending to have lower birthweight babies. Temperature during the second trimester of pregnancy positively associated with offspring birthweight, but this association did not reach conventional levels of statistical significance. Temperature around the time of conception tended to show an inverse association (consistent with the first trimester association) and temperature around the time of birth tended to show a direct association (consistent with the third trimester association). Associations for birthweight adjusted for gestational age and sex z scores were similar to those for birthweight. The associations of these temperature measures with birthweight were similar among those from manual and non-manual social classes (all P values for interactions >0.4).
Table 4. Associations of average outdoor temperature in the middle of the first, second and third trimesters of pregnancy, around the time of conception and around the time of birth with birthweight.
Adjusted*difference (95% CI) from those born in the lowest fifth of the distribution of temperatures
z score birthweight for sex and gestational age
All estimates have been adjusted for gestational age (not z score), mother's age, year of birth, birth order, social class and sex (not z score).
Temperature in middle of first trimester
Coldest fifth (range: −6.0 to 0.8)
2nd fifth (range: 0.9 to 3.0)
−12.6 (−42.8, 17.6)
−0.04 (−0.10, 0.02)
3rd fifth (range: 3.1 to 6.0)
−40.4 (−70.5, −10.3)
−0.08 (−0.14, −0.02)
4th fifth (range: 6.1 to 8.9)
−49.6 (−79.5, −19.7)
−0.08 (−0.14, −0.02)
Warmest fifth (range: 9.0 to12.0)
−55.2 (−85.5, −25.0)
−0.10 (−0.16, −0.04)
Change per 1°C
−5.4 (−7.9, −2.9)
−0.01 (−0.02, −0.01)
P linear trend
Temperature in middle of second trimester
Coldest fifth (range: −6.0 to 0.6)
2nd fifth (range: 0.7 to 2.8)
−0.84 (−30.9, 29.2)
0.00 (−0.06, 0.06)
3rd fifth (range: 2.9 to 6.0)
31.1 (1.01, 61.1)
0.04 (−0.02, 0.10)
4th fifth (range: 6.1 to 8.7)
8.4 (−21.7, 38.5)
0.00 (−0.06, 0.06)
Warmest fifth (range: 8.7 to 12.0)
15.7 (−14.9, 46.4)
0.00 (−0.06, 0.06)
Change per 1°C
1.8 (−0.7, 4.3)
0.00 (−0.01, 0.01)
P linear trend
Temperature in middle of third trimester
Coldest fifth (range: −6.8 to 0.7)
2nd fifth (range: 0.8 to 2.9)
6.9 (−21.9, 35.7)
−0.01 (−0.07, 0.05)
3rd fifth (range: 3.0 to 5.7)
18.3 (−10.5, 47.0)
0.03 (−0.03, 0.09)
4th fifth (range: 5.8 to 8.8)
34.7 (5.8, 63.7)
0.06 (0.00, 0.12)
Warmest fifth (range: 8.9 to 17.8)
38.3 (9.6, 66.9)
0.08 (0.02, 0.14)
Change per 1°C
1.3 (0.5, 2.1)
0.01 (0.00, 0.02)
P linear trend
Temperature around time of conception
Coldest fifth (range: −10.6 to 0.6)
2nd fifth (range: 0.7 to 3.3)
−20.7 (−49.6, 8.2)
−0.05 (−0.11, 0.00)
3rd fifth (range: 3.4 to 6.1)
−36.9 (−66.8, −6.9)
−0.07 (−0.13, −0.01)
4th fifth (range: 6.2 to 8.9)
−31.1 (−60.2, −2.0)
−0.06 (−0.12, 0.00)
Warmest fifth (range: 9.0 to 17.6)
−28.2 (−59.2, 2.8)
−0.03 (−0.09, 0.03)
Change per 1°C
−2.4 (−4.6, −0.2)
0.00 (−0.01, 0.00)
P linear trend
Temperature around time of birth
Coldest fifth (range: −10.6 to 0.6)
2nd fifth (range: 0.7 to 2.8)
22.7 (−5.8, 51.1)
0.04 (−0.01, 0.10)
3rd fifth (range: 2.9 to 5.6)
42.6 (13.2, 71.9)
0.06 (0.01, 0.12)
4th fifth (range: 5.6 to 8.9)
51.5 (23.9, 79.1)
0.07 (0.02, 0.13)
Warmest fifth (range: 9.0 to 16.7)
46.6 (20.4, 72.7)
0.07 (0.01, 0.13)
Change per 1°C
3.6 (1.6, 5.7)
0.01 (0.00, 0.01)
P linear trend
When the temperature variables were entered pairwise into the same regression models, the effect of temperature around the time of conception attenuated with adjustment for temperature in the middle of the first or third trimester, but the effect of temperature in the middle of these trimesters was not affected by adjustment for temperature around the time of conception. Similarly, temperature in the middle of the third trimester was independent of temperature around the time of birth, but temperature around the time of birth attenuated with adjustment for both temperature in the middle of the third trimester and temperature in the middle of the first trimester. The effects of temperature in either the middle of the first or third trimester were not affected by adjustment for temperature in the middle of the second trimester. Although temperature variables are correlated (see Table 3) in these regression models, there was no evidence of problems due to collinearity; the standard errors for the regression coefficients did not alter with two temperature variables in the model compared with just one. The only exception to this was when temperature in the middle of the first trimester was included in the same model as temperature in the middle of the third trimester. In this model, the effect of temperature in the middle of the first trimester remained similar to that presented in Table 4 [change in birthweight for a 1°C increase temperature −4.01 (−8.23, −0.01)], but the association with temperature in the middle of the third trimester attenuated to 0.30 (−1.05, 1.66). However, there was evidence of problems with collinearity. The standard error for the regression coefficient with temperature in the first trimester increased from 0.78 when it was the only temperature variable in the model to 2.03 when temperature in the third trimester was included and the standard error for the third trimester regression coefficient increased from 0.69 to 0.91.
Figure 3 shows the effect on the seasonal patterning of birthweight of adjustment for temperature in each of the three trimesters of gestation. With adjustment for temperature in the middle of either the first or third trimester of pregnancy, the winter dip in birthweight is decreased. Adjustment for temperature in the middle of the first trimester of pregnancy also decreased the autumnal high in birthweight. The seasonal pattern of birthweight with adjustment for temperature in the middle of the second trimester was similar to that with no adjustment for temperature. The P value for the joint sine–cosine functions in a regression model with all other covariates and temperature in the first trimester was P= 0.9, with all other covariates and temperature in the second trimester in the model statistical evidence for the sinusoidal pattern remained (P= 0.02) and with temperature in the third trimester there was weak evidence of a sinusoidal effect (P= 0.23).
Table 5 shows the effects of temperature during the first, second and third trimesters and birthweight stratified by season. The associations between ambient outdoor temperature in each of the trimesters of gestation and birthweight are similar for each season of birth and similar to those for the whole cohort, but estimates are imprecise in these stratified analyses.
Table 5. Associations of average outdoor temperature in the middle of the first, second and third trimesters of pregnancy with birthweight, stratified by season of birth. T1, T2, T3 = first, second and third trimesters, respectively.
Adjusted*change (95% CI) in birthweight for a 1°C increase in temperature
Winter (December to February)
Spring (March to May)
Summer (June to August)
Autumn (September to November)
All estimates have been adjusted for gestational age, mother's age, year of birth, birth order, social class and sex.
Temperature in middle of T1
−7.6 (−14.4, −0.3)
−5.2 (−14.8, 1.9)
−5.9 (−15.0, 2.1)
−7.1 (−15.2, 0.3)
Temperature in middle of T2
1.3 (−8.3, 9.9)
1.5 (−8.7, 10.3)
0.9 (−8.4, 8.5)
1.6 (−7.9, 11.1)
Temperature in middle of T3
2.0 (−7.2, 7.1)
1.6 (−7.0, 10.2)
1.3 (−5.9, 8.4)
1.9 (−9.5, 5.4)
In this cohort of individuals born in Aberdeen (Scotland) in the early 1950s, we have shown birthweight to be lower among those born in the winter and birthweight to be inversely associated with temperature in the middle of the first trimester of birth and positively associated with temperature in the middle of the third trimester of birth. Temperature measures in the first and third trimesters of pregnancy explained much of the seasonal patterning of birthweight and these temperature effects remained when associations were stratified by season of birth. These findings suggest that relatively higher ambient outdoor temperatures in the first trimester of gestation and/or lower ambient outdoor temperatures in the third trimester of gestation are associated with low birthweight.
The seasonal patterning of birthweight that we found is similar to that described in three other studies.18–20 The temperature associations we have found are consistent with animal studies and ecological studies which suggest that relative heat stress during pregnancy, particularly in early pregnancy, may result in poor placental growth and subsequent intrauterine growth retardation.12 A study from Northern Ireland of over 400,000 births, which assessed temperature effects on birthweight, found that relatively colder ambient outdoor temperatures in the mother's second trimester of pregnancy were associated with lower birthweight and that temperature in the second trimester explained much of the seasonal patterning of birthweight in boys, although not in girls.21 We also found that colder temperatures in the second trimester were associated with lower birthweight, but this association was non-significant at the conventional 5% level, and with adjustment for temperature in either the first or the third trimester, the association attenuated. However, the study from Northern Ireland did not provide any information on temperature effects at other stages of the pregnancy and whether their second trimester effect was independent of temperatures in these other trimesters.
Strengths of this study are its large size, the availability of extensive details on the participant's intrauterine and birth characteristics and the ability to link these data to daily temperature data from one weather station located in the city where all births occurred. For birthweight and other perinatal data, AMND data were used that were collected at the time of pregnancy and birth using standard procedures with the explicit intention that they should be used in research.24,25 We have no data on placental weight or quality and were therefore unable to assess whether being in early pregnancy during periods of higher outdoor temperature were associated with poor placental growth as seen in animal studies.12
The associations presented may be due to chance, may reflect an effect of outdoor temperature on maternal vascular function, and hence, nutritional supply to the fetus or may reflect other exposures that are more closely linked to temperature than they are to month of birth. The small P values for the first and third gestation temperature effects, the consistency of the seasonal patterning, which has been found in three previous studies,19,20 as well as ours, together with temperature studies in animals and international comparative studies in humans,12,13 make chance an unlikely explanation.
It is possible that temperature during gestation is more strongly linked to maternal behaviours including smoking, diet and physical activity, than is seasonality, and that these provide the mechanism linking temperature during gestation to birthweight. We do not have data on these variables in our study to enable us to directly assess this possibility. Adults tend to increase their total energy and carbohydrate intake in autumn,26 and one study found high carbohydrate intake in the first and mid-trimester to be associated with low birthweight.27 Thus, low birthweight in winter babies may be associated with autumnal increased carbohydrate consumption by their mothers in their second trimester. However, other studies have not found maternal carbohydrate intake to be related to offspring birthweight28 and an autumnal increase in carbohydrate consumption (unless driven specifically by temperature effects) could not explain the independent associations of temperature that we found in stratified analyses. In an Indian population, maternal intake of green vegetables, fruit and milk during pregnancy, as well as maternal circulating levels of folate and vitamin C during pregnancy, were associated with offspring birthweight.29 During the 1950s in Aberdeen there will have been seasonal patterns in fruit and vegetable availability and there may have been temperature effects on the availability of fruit and vegetables. If the results seen in a contemporary Indian population are generalisable to babies born in the North of Scotland in the 1950s, then our results may reflect the seasonal availability of different food sources. Poor nutrition around the time of conception is associated with preterm birth in ewes,30 and in humans energy intake is lower in the spring,26 when winter born babies will have been conceived. However, animal studies may not be directly transferable to humans and in our study birthweight for gestational age was lower in winter suggesting an effect on intrauterine growth rather than prematurity.
A direct effect of temperature during pregnancy is supported by the fact that the seasonal patterning of birthweight was attenuated with adjustment for temperature in the first and third trimesters (Fig. 3) and by our stratified analyses which demonstrate independent (of seasons) effects of temperature. Low birthweight in winter-born babies may be related to exposures to relatively higher outdoor temperatures in the first trimester of gestation and/or colder outdoor temperatures in the third trimester. Our results tend to suggest that exposures to higher outdoor temperatures in the first trimester of gestation are more important than exposures to cold outdoor temperatures in the third trimester. The magnitude of the effect for the first trimester was greater than that for the third trimester, and in models containing both measures simultaneously, the first trimester effect remained whereas the third trimester effect was attenuated to the null. However, the two are dependent upon each other and the results from the model including both may be unreliable due to collinearity. Thus, an effect of cold exposure in the third trimester cannot be ruled out by the results of these statistical models.
In general, humans have greater means of protecting themselves from the cold (by the use of high quality housing, indoor heating, warm clothing and blankets, for example) than they have from protecting themselves from the heat.12 Housing quality in Aberdeen in the 1950s, particularly for those from lower socio-economic groups, was very poor24 and therefore if cold exposure in the third trimester was the important determinant of the effects we have found, one might expect a greater effect among those from the lower social classes who are less able to protect themselves from cold exposure. We found no evidence that socio-economic position modified the seasonal or temperature effects on birthweight, providing some evidence that factors other than cold exposure may be more important.
In animal studies, heat stress in early pregnancy is associated with reduced placental weight, decreased uterine and umbilical blood flow and consequent reduction in offspring birthweight despite no change in maternal appetite or weight loss.12 However, the relevance of these studies to humans is unclear. Relative extremes of temperature (both hot and cold) are known to affect human blood flow with excess cardiovascular deaths occurring in heat waves and during cold winter months.6–8 It is therefore plausible that maternal blood flow and hence fetal nutrition will be affected by temperature extremes at different stages of gestation. Further research in humans is required to determine whether temperature exposures during different stages of gestation importantly influence birthweight. With the increasing occurrence of periods of extreme temperature and a growing epidemiological interest in the acute effects of heat waves,8,31 one area of potential future research would be to examine the effects of women being in early pregnancy during a heat wave on the birthweight of their offspring.
Birthweight is seasonally patterned and this patterning may reflect exposures to relatively higher ambient outdoor temperatures in early pregnancy and/or lower temperatures in later pregnancy. Our results suggest that pregnant women should protect themselves from relative temperature extremes, in particular, high temperatures in early pregnancy. With the increasing occurrence of temperature extremes, in particular, heat waves, these findings, if replicated in other studies, have important public health implications.
The authors would like to thank Raymond Illsley for providing us with the data from the Aberdeen Child Development Survey and for his advice about the study. Graeme Ford played a crucial role in identifying individual cohort members and in helping us initiate the process of revitalisation. Sally Macintyre, Doris Campbell, George Davey Smith, Marion Hall, Bianca de Stavola, Susan Morton, David Batty, David Godden, Di Kuh, Debbie A Lawlor, Glyn Lewis and Viveca Östberg collaborated with David A. Leon to revitalise the cohort. Heather Clark managed the study at the Dugald Baird Centre, Aberdeen with the assistance of Margaret Beveridge. The authors would also like to thank the staff at the ISD (Edinburgh), GRO (Scotland) and NHSCR (Southport) for their substantial contributions and John Lemon who undertook the linkage to the Aberdeen Maternity and Neonatal Databank. Finally, the authors thank the study participants who responded to a mailed questionnaire 40 years after the original survey was completed.
For the analyses presented in this paper, temperature data were obtained from the archives of the Meteorological Office via the British Atmospheric Data Centre (http://badc.nerc. ac.uk/) and Richard Mitchell (University of Edinburgh) helped to download these data. Mark Jitlal (London School of Hygiene and Tropical Medicine) provided valuable advice on how to fit the sine–cosine model to assess seasonality and Yuji Nisihiwaki (London School of Hygiene and Tropical Medicine) made a helpful suggestion regarding the analyses.
The Aberdeen Children of the 1950s study was funded as a component project (G0828205) of a Medical Research Council Co-operative Group Life-course and trans-generational influences on disease risk (G9819083). A project on cognition and adult health in the cohort has been funded by the Chief Scientists Office, Scottish Executive Health Department, which currently funds Heather Clark. Debbie A. Lawlor is funded by a Department of Health (UK) Career Scientist Award. The views expressed in this publication are those of the authors and not necessarily those of any funding body.
All authors developed the idea for this study. D. A. Lawlor undertook the analysis and wrote the first draft of the paper. All authors contributed to the final version of the paper.