- Email: [email protected]
Seasonality of Birth Weight in Chile: Environmental and Socioeconomic Factors
Seasonality of Birth Weight in Chile: Environmental and Socioeconomic Factors FLORENCIA TORCHE, PHD, AND ALEJANDRO CORVALAN, MSC
PURPOSE: Research suggests a relationship between birth weight and season of birth, but findings vary across countries and underlying factors are not well understood. We examine the seasonality of birth weight and explore alternative hypotheses for its etiologydexposure to environmental factors and varying socioeconomic composition of mothersdin Chile. METHODS: Birth weight of approximately 5 million Chilean singleton live births 37 of 41 weeks of gestation between 1987 and 2007 were analyzed for seasonality by using regression models with month dummies and parametric sinusoidal specifications. Multivariate models with socioeconomic covariates and interactions across geographic regions examine potential factors accounting for seasonal variation. RESULTS: Marked 12-month and 6-month periodic cycles were found. The amplitude and phase of the seasonal variation change across geographic regions. In the low-latitude northern region, there is a spring peak and a fall nadir, while in middle-latitude colder regions, a bimodal periodicity emerges with peaks in spring and fall, a pronounced winter nadir, and smaller nadir in the summer. Socioeconomic composition of mothers is found to vary with annual periodicity, but it does not account for the seasonality in birth weight. CONCLUSIONS: Environmental factors rather than the socioeconomic composition of mothers likely account for seasonal variation in birth weight. The change in periodicity of birth weight across latitudes is consistent with a beneficial exposure to sunlight both early and late in the pregnancy, and a detrimental late exposure to cold temperatures only in areas with low winter temperatures. Ann Epidemiol 2010;20:818–826. Ó 2010 Elsevier Inc. All rights reserved. KEY WORDS:
Birth Weight, Seasonality, Time-series Analysis, Socioeconomic Factors, Temperature, Light.
INTRODUCTION A substantial association exists between season of birth and individual outcomes throughout the life cycle. These outcomes include early health measures such as birth weight and gestational age (1–3), life expectancy (4), schizophrenia (5), and educational attainment (6). Birth weight is particularly important because it is the main proximate determinant of infant mortality, and it affects health outcomes in childhood and adulthood (7–11) as well as cognitive, educational, and socioeconomic attainment (12, 13). Given the influence of birth weight on later attainment, the seasonality of adult outcomes may be at least partly mediated by birth weight. Several studies have documented seasonal variation in birth weight, but results are far from homogeneous. Even if studies are restricted to term or gestational age–adjusted births to avoid confounding by length of gestation, some find a marked annual cycle, with most detecting peaks in spring (14, 15) and winter (16, 17). However, autumn and summer peaks have also been found (18–20). Others find bimodal periodicities with peaks consistently falling in fall and spring (21–23). Part of the variation may be due to From the Departments of Sociology (F.T.) and Economics (A.C.), New York University. Address correspondence to: Florencia Torche, PhD, Department of Sociology, New York University, New York. E-mail: [email protected]. Received April 7, 2010; accepted August 18, 2010. Ó 2010 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
different specification in the dependent variable (term births vs. gestation-adjusted weight) or to the fact that some studies do not explicitly check for a semiannual periodicity. Substantive reasons may also account for the differences in findings. Variation in birth weight depends on a complex set of genetic and epigenetic factors affecting the maternalplacental-fetal unit. Season of birth provides a natural experiment for the generation of alternative hypotheses about the effect of environmental factors (15). A useful strategy is to examine changes in seasonal variation across latitudes, under the assumption that latitude provides a proxy for important factors such as temperature and sunlight exposure. Meta-analysis suggests the seasonality of birth weight varies across latitudes, although different study methodologies may hamper conclusive interpretation (3). An alternative hypothesis proposes that the seasonality of birth weight emerges from the varying socioeconomic composition of mothers conceiving in different months of the year (6, 24, 25). To the extent that sociodemographic characteristics are correlated with offspring’s birth weight and that different sociodemographic groups are more likely to give birth in certain seasons, part of the variation attributed to environmental factors may be driven by the changing distribution of risk factors in the pregnancy set. For example, if the months of highest temperature correspond to those in which low–socioeconomic status (SES) women are most likely to give birth, temperature may appear predictive 1047-2797/$ - see front matter doi:10.1016/j.annepidem.2010.08.005
AEP Vol. 20, No. 11 November 2010: 818–826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
819
of low birth weight only because low SES is a risk factor for low birth weight. METHODS We examined seasonality of birth weight in Chile during a 21year period. Populated areas in Chile range from latitude 18 S to 56 S (comparable to the range between Niger and Sweden), with substantial climatic variation. We exploit such variation to examine potential environmental determinants of birth weight seasonality. Data were derived from the Chilean Birth Registry database established by the Chilean Ministry of Health. By Chilean law, all births must be reported to the Ministry of Health. The database is constructed on the basis of live-birth certificates, a form completed by the professional (MD or certified midwife) who attends the delivery (99.8% of deliveries are attended by a professional in Chile) and measured with little error (26). The birth certificate contains information on birth weight, gestational age, maternal residence, and maternal demographic characteristics. We examined birth weight in all live-born singletons with gestations between 37 and 41 weeks from the years 1987 through 2007 (both inclusive) for a total of 4,968,912 births. Records of individual birth weight were summarized into a time series of monthly means. To observe short-run fluctuations, the trend in these means was then extracted using a Hodrick-Prescott filter with monthly smoothing parameter l Z 14,400 (27). Spectral analysis showed that the fundamental frequency components of the series are annual and semiannual (results from spectral analysis not shown, available from authors upon request). Regression models were then implemented to account for the seasonal fluctuation in birth weight. Birth weight was predicted by 11 month dummies (January used as baseline for comparison) and a full set of year dummies to account for trend. Additionally, we model month effects parametrically by using sinusoidal functions. As suggested by the two component frequencies extracted from the spectral analysis, we allowed for a 12-month periodicity [A*sin([month4] * 2p/12)]d‘‘annual model’’d and for a 6-month periodicity [A*sin([month-4 ] * 2p/6)]d ‘‘semi-annual model’’. In both cases, 4 denotes an integer phase shift, which controls the location of the seasonal peak, and A identifies amplitude calculated as peak-to-zero (or nadir-to-zero) distance or half of the maximal difference between highest and lowest weights during the year. Nonlinear least squares was used to estimate the cyclical phase. The phase parameter was then approximated to result in an integer value for the peak month in the annual and semiannual cycle. Then, amplitude was calculated using ordinary least square regression for a fixed value of the phase. Regression models were subsequently extended in two ways. First, we stratify the sample by geographic region and estimate region-specific regression models to evaluate
FIGURE 1. Average monthly temperature in degrees Celsius across geographic regions, Chile. Monthly temperatures are averages over the second half of the 20th century. Regional averages are obtained from arithmetic mean of temperatures in the larger cities comprising most of the population living in each region. Cities included are Arica, Antofagasta, Copiapo´, and La Serena (northern region); Valparaı´so, Concepcio´ n, (central-coastal region), Santiago and Linares (central-interior region), and Temuco and Puerto Montt (southern region).
variation in phase and amplitude. Four regions are distinguished: north (latitudes 18 S–31 S), central-coastal and central-interior (32 S–37 S), and south (38 S–56 S). The low-latitude northern region has milder weather and slight temperature variation throughout the year (Fig. 1). Colder temperatures are observed in the central, and particularly the southern, regions. The distinction between centralcoastal and central-interior is not based on latitude but on the summer-winter temperature variation, which is much narrower in the coastal area due to the moderating influence of the Pacific Ocean. The central-interior region displays maximum summer temperatures as high as those in the northern region and minimum winter temperatures as low as those in the southern region. According to the standard Ko¨ppen climate system, the northern area is characterized by a dry arid and semi-arid climate, whereas the central and southern areas feature a temperate/mesothermal climate type (28). Second, we added measures of maternal sociodemographic characteristics to the models to assess whether the varying socioeconomic composition of births across the year accounts for seasonal variation in birth weight. The following variables were included: proportion of mothers that are married (married) each month of birth, proportion of mothers older than 16 years of age (adult), proportion of mothers with postsecondary education (college), and proportion of mothers living in urban areas (urban). RESULTS Figure 2 presents the original series and extracted trend for the birth-weight data between 1987 and 2007. Some
820
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
AEP Vol. 20, No. 11 November 2010: 818–826
increase in average birth weight occurs between 1987 and 1995dfrom 3,300 to 3,400 grams, a well-documented trend potentially associated with economic growth (29). Increase is followed by stability between 1995 and 2007. Evaluation of the original series with respect to the trend suggests a possible seasonal component with annual periodicity. Regression analysis shows that birth weight has a bimodal peak in October–November (spring) and March (fall) (model 1 in Table 1, ‘‘All births’’ specification). A sinusoidal formulation with 12-month and 6-month periodicity fits the observed data extremely well, accounting for most of the variation across months (Model 2 in Table 1, ‘‘All births’’ specification). The variable ‘‘Annual cycle’’ identifies 12-month periodicity, while ‘‘Semiannual cycle’’ accounts for any additional 6-month periodicity. In the analysis of all births, the bestfitting model includes an annual phase with peak in January and semiannual peak in the same month. Note that because the parameter estimates are of approximately the same magnitude but opposite sign, they should offset each other, resulting in a value close to zero for the first month of the year. Figure 3 depicts the parameter estimates associated with the month dummies and their 95% confidence intervals, superimposing the sum of the annual and semiannual sinusoidal functions. Figure 3 shows peaks in spring and fall, a pronounced winter nadir, and a smaller nadir in the summer. Infants born in spring or fall are, on average, 16 grams heavier than those born in the winter (p ! 0.001). To explore potential determinants of birth weight seasonality, we stratify our sample by geographic region and replicate the regression analysis across groups. Results are displayed in the remaining columns of Table 1. They indicate that the phase and amplitude of seasonal variation substantially changes across region. In the low-latitude northern region, there is marked annual periodicity with
highest birth weight in spring and lowest birth weight in the fall. An annual sinusoidal function with peak in October and nadir in April suffices to account for the data, with no 6-month periodicity found (Fig. 4, panel 1). The phase of the cycle shifts for the other three geographic regions (panels 2–4 in Fig. 4). In these regions the periodicity is markedly bimodal, with twin peaks in spring and fall, and a pronounced winter nadir. This pattern remains unchanged across latitudes 32 S to 56 S (central and southern regions of the country). Furthermore, the amplitude of the cycle varies across regions, with wider amplitude in the centralinterior area, which features the widest seasonal variation in temperature as indicated by Figure 1. In the centralinterior region, infants born in the spring and fall weigh, on average, 21 grams more than those born in the winter (p ! 0.001). This amplitude is, expectedly, greater than the average across the country reported in Figure 3, which combines seasonal patterns with different phases (30). Geographic variation suggests that environmental factors may affect seasonality of birth weight. However, no controls have been introduced for the socioeconomic characteristics of pregnancies, so the findings may be an artifact of the changing sociodemographic composition of births throughout calendar months. As with birth weight, these variables display seasonal variation. The compositional change of maternal marital status and age show a pronounced annual pattern (Fig. 5). Married women and those young women 17 years of age and older are more likely to give birth in the summer and least likely to give birth in the winter, potentially accounting for the winter peak found in central and southern regions. Capturing the seasonality of births to college-educated and urban women requires annual and semiannual components. While the seasonal variation in the proportion of urban women is limited, college-educated women are more likely to give birth in the spring, potentially accounting for the spring peak found throughout the sample. Table 2 adds these sociodemographic factors to the regression models by using sinusoidal specifications. To the extent sociodemographic compositional variation of births accounts for the seasonality of birth weight, we expect that the sinusoidal parameters will decline in magnitude and become statistically insignificant. The findings are not consistent with this hypothesis. The parameters remain virtually unchanged across all regions, indicating that the seasonality of birth weight is not accounted for by the changing distribution of risk factors related to sociodemographic characteristics throughout the year.
FIGURE 2. Time series of mean monthly birth weight. Singleton births at 37–41 weeks of gestation, 1987–2007, Chile. Trend extracted by use of a Hodrick-Prescott filter with monthly smoothing parameter l Z 14,400.
DISCUSSION We find a marked seasonality of birth weight in Chile. Seasonal variation changes across geographic regions. In
AEP Vol. 20, No. 11 November 2010: 818–826
TABLE 1. Linear regression model, dependent variable birth weight, singleton births 37–41 weeks of gestation, 1987–2007, Chile. Models for all births and for births by geographical region* All births Model 1 b (SE) January (omitted) February March April May June July August September October November December Annual cycley Semiannual cycley Constant N R2
6.02{ (1.57) 8.10{ (1.40) 4.47{ (1.27) 2.80x (1.36) 3.07z (1.71) 8.83{ (1.62) 3.68x (1.63) 1.11 (1.42) 6.31{ (1.40) 7.03{ (1.43) 5.00x (1.78)
3314 252 98.26
Region 1dNorth Model 2 b (SE)
4.76{ (0.51) 4.62{ (0.46) 3316 252 98.13
Model 1 b (SE) 3.16 (3.43) 2.76 (3.18) 3.60 (3.10) 2.16 (2.98) 2.43 (3.40) 1.73 (3.49) 7.31x (3.42) 1.80 (3.08) 8.75x (3.59) 9.13{ (3.12) 4.89 (3.50)
3308 252 90.92
Model 2 b (SE)
5.27{ (0.91) 1.03 (0.96) 3310 252 90.46
Region 2dCentral-Coastal
Region 3dCentral-Interior
Model 1 b (SE)
Model 1 b (SE)
9.20x (3.96) 10.70{ (3.65) 10.97{ (3.43) 7.49z (3.87) 0.88 (3.67) 4.70 (3.40) 3.23 (4.05) 3.80 (3.66) 7.97x (3.77) 3.39 (3.83) 4.45 (3.85)
3313 252 93.71
Model 2 b (SE)
4.21{ (0.95) 5.20{ (0.90) 3317 252 93.52
6.69{ (1.88) 9.93{ (1.74) 4.15x (1.72) 2.10 (1.54) 5.92{ (2.05) 11.52{ (2.00) 5.96{ (2.02) 1.40 (2.01) 7.35{ (1.75) 8.52{ (1.55) 5.08x (1.89)
3313 252 97.50
Model 2 b (SE)
6.43{ (0.60) 5.76{ (0.57) 3314 252 97.32
Region 4dSouth Model 1 b (SE) 8.23{ (2.81) 7.26{ (2.75) 7.05{ (2.39) 5.66x (2.73) 0.22 (2.88) 11.23{ (2.80) 4.68 (2.90) 4.26 (2.62) 2.46 (2.32) 1.87 (2.89) 2.58 (3.27)
3328 252 95.78
Model 2 b (SE)
6.34{ (0.87) 3.07{ (0.82) 3329 252 95.45
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
b Z parameter estimate; SE Z standard error. *Robust standard errors in parentheses. All regressions include a full set of year dummies to account for trend. For ‘‘all birth’’ and regional samples, model 1 includes a full set of month dummies and model 2 includes annual and semiannual parameter estimates from the best-fitting sinusoidal specification. y The phase (4) of the annual cycle varies across regions: 4 Z –2 for all births, 1 in north region, –1 in central-coastal region, –2 in central-interior region, and –1 in south region. Consequently, annual peaks are as follows: January for all births, April in north region, February in central-coastal region, January in central-interior region, and February in the south region. In the semiannual cycle, 4 Z –1/2, and peak is in January for the ‘‘all birth’’ and all regional samples. z p ! .05. x p ! .01. { p ! .001.
821
822
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
FIGURE 3. Birth weight (in grams) across month of birth using month dummies (mean and 95% confidence interval) and sinusoidal specification, including annual and semiannual components. Singleton births 37–41 weeks of gestation, 1987–2007, Chile.
low latitude, mild-weather areas with little temperature oscillation throughout the year, there is a manifest spring peak and fall nadir. In the rest of the country, characterized by colder winters and wider seasonal variation in temperature, there is a bimodal distribution with winter and summer nadirs, and peaks in the fall and spring. We also show that these patterns are not accounted for by the socioeconomic composition of mothers, although sociodemographic characteristics do vary seasonally.
AEP Vol. 20, No. 11 November 2010: 818–826
Seasonal variation of birth weight fluctuates between 13 and 21 grams across geographic regions, a figure in line with previous research using comparable large data sets (16, 20, 22, 23). Even if this magnitude is small, it is substantial in the context of an ecological study that captures populationlevel averages. Under the reasonable assumption that the seasonal variation is heterogeneous across individuals and affects only a segment of the population, the effect may be substantially greater among affected pregnancies. For example, if one half of pregnancies were affected, the effect would be twice as large as the reported population-level average, and if only 10% of pregnancies were affected, the effect would be 10 times as large. The potentially large effect of the ‘‘treatment on the treated’’ suggests the importance of further studying heterogeneity in seasonality of birth weight along demographic and socioeconomic lines, in order to detect the population most at risk. In addition, the geographic variation in seasonality of birth weight sheds light on its etiology. Which factors can explain the observed patterns? Factors highlighted in some developing contexts, such as seasonal variation in nutrient deficiency, physical exertion, and exposure to malaria or other infections, have been largely eradicated in Chile. In countries where these exposures have been eradicated, the literature focuses on two environmental factors: exposure to sunlight and to low temperatures. Sunshine exposure in early pregnancy has been found to be related to higher birth weight (18, 31). One potential mechanism suggests that
FIGURE 4. Birth weight deviations from the January mean (in grams) across month of birth using month dummies (mean and 95% confidence interval) and sinusoidal specification. Singleton births 37–41 weeks of gestation, 1987–2007, in Chile, by geographic region.
AEP Vol. 20, No. 11 November 2010: 818–826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
823
FIGURE 5. Maternal socioeconomic characteristics across month of birth using month dummies (mean and 95% confidence interval) and sinusoidal specification. Singleton births 37–41 weeks of gestation, 1987–2007, in Chile. Upper left panel: Percentage of births to women older than 16 years of age. Upper right panel: Percentage of births to married women. Lower left panel: Percentage of births to women with at least some college education. Lower right panel: Percentage of births to women living in urban areas. All values are deviations from the January mean (equal zero in all figures). Seasonality of births to women older than 16 years (adult) and married women (married) modeled by a 12-month sinusoidal parameter with peak in February and nadir in August. Seasonality of percentage of births to women with college education (college) modeled by a 12-month sinusoidal parameter with peak in October and nadir in April, and a 6-month sinusoidal parameter with peaks in April and October and nadirs in January and July. Seasonality of percentage of women living in urban areas (urban) modeled by a 12-month sinusoidal parameter with peak in November and nadir in May, and a 6-month sinusoidal parameter with peaks in May and November and nadirs in February and August.
sunlight exposure during the first trimester of gestation may stimulate growth hormone release by inhibiting production of maternal pineal melatonin (32). Some evidence for an inverse relationship between melatonin and growth hormone (GH) production exists (33). However, a caveat about this hypothesis is that GH receptors are functionally immature prior to approximately 36 weeks of gestation (34). An alternative mechanism refers to insulin-like growth factor (IGF)–1. This growth factor displays seasonal fluctuation (35) and is associated with fetal growth (36). IGF-1 plays an important role in prenatal growth, and animal studies show that it varies with exposure to light (37, 38). Another avenue for the influence of sunlight exposure is the production of vitamin D. Vitamin D3 is unique in that its production depends primarily on the action of sunlight on the skin (39). Its production is strongly and consistently associated with the duration of the photoperiod, which in turn is influenced by latitude and season (40, 41). Vitamin D deficiency is more prevalent during the winter months, even in low-latitude areas (41). Fetal vitamin D requirements increase during pregnancy, related to the increased need for fetal calcium; thus maternal vitamin D
levels tend to fall during the third trimester, especially if it occurs during the winter (42). This alternative mechanism may result in a deficit of sunlight exposure being more consequential for birth weight late rather than early in the pregnancy. Other studies suggest that seasonal variation of birth weight is related to the exposure to cold temperatures in mid and late gestation. Exposure to low temperatures raises fibrinogen levels in plasma (43), increases blood viscosity (44), and might induce vascular constriction in the placenta, thereby diminishing uteroplacental blood flow. This may result in lower birth weight (16, 45). This effect may be more pronounced in mid or late gestation when the fetus experiences rapid growth. Research has documented a potential late-pregnancy effect in areas with cold winters (19). We found a pronounced spring peak throughout the regions analyzed. This peak could be induced by earlypregnancy exposure to sunlight during the summer, and it would consistently affect all latitudes insofar as the seasonal fluctuation of sunlight also affects low-latitude areas. A second fall peak emerges in the central and southern areas. A potential explanation is the late-pregnancy exposure to
3314 252 97.32 3317 252 93.52 3310 252 90.50 3316 252 98.13
AEP Vol. 20, No. 11 November 2010: 818–826
b Z parameter estimate; SE Z standard error. *Robust standard errors in parentheses. All regressions include a full set of year dummies to account for trend. Values indicate percentage of women who are adult, married, college educated, and urban multiplied by 100. y The phase (4) of the annual cycle varies across regions: 4 Z –2 for all births, 1 in north region, –1 in central-coastal region, –2 in central-interior region, and –1 in south region. Consequently, annual peaks are as follows: January for all births, April in north region, February in central-coastal region, January in central-interior region, and February in the south region. In the semiannual cycle, 4 Z –1/2, and peak is in January for the ‘‘all birth’’ sample and all regional samples. z p ! .05. x p ! .01. { p ! .001.
3329 252 95.45
2.31x (1.13) 4.12{ (0.83) 1.65 (1.40) 2.99{ (0.61) 2.66{ (0.84) 1.01y (0.59) 2895 252 96.07
b (SE) b (SE)
6.34{ (0.87) 3.07{ (0.82)
5.83{ (0.77) 5.64{ (0.64) 1.07 (2.02) 0.92 (0.59) 0.66 (0.85) 0.55 (0.80) 3387 252 97.36
b (SE) b (SE)
6.43{ (0.60) 5.76{ (0.57)
4.00{ (1.27) 4.93{ (0.95) 2.14 (1.92) 0.83 (0.64) 0.07 (0.72) 0.96 (1.10) 3541 252 93.62
b (SE) b (SE)
4.21{ (0.95) 5.20{ (0.90)
5.06{ (1.02) 1.16 (0.96) 1.69 (1.79) 0.06 (0.61) 0.72 (0.91) 0.92 (0.98) 3554 252 90.63
b (SE) b (SE)
5.27{ (0.91) 1.03 (0.95)
4.09{ (0.70) 4.63{ (0.52) 1.13 (1.93) 0.58 (0.59) 0.17 (0.95) 0.20 (0.81) 3403 252 98.16
b (SE) b (SE)
All births
4.76{ (0.51) 4.62{ (0.46)
Annual cycley Semiannual cycley Adult Married College Urban Constant Observations R2
Region 1dNorth
Region 2dCentral-Coastal
Region 3dCentral-Interior
Region 4dSouth
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
TABLE 2. Birth weight variation across month of birth using sinusoidal specification. Controls added for socioeconomic characteristics of mothers. Chile singleton births, 37–47 weeks of gestation, 1987–2007, Chile by geographic region*
824
sunshine for these births; the absence of a peak in the northern region is likely because, in this mild climate, individuals spend more time outdoors even in the winter, thus reaching a satisfactory exposure level. A particularity of the central and southern regions is the pronounced winter nadir, which does not exist in the low-latitude northern region. We hypothesize that this nadir may be due to the exposure to cold during late gestation. Again, this detrimental effect would not manifest itself in the northern region because minimum winter temperature is rather high, reaching 14 C (see Fig. 1), which is close to the minimum indoor temperature of 16 C recommended by the World Health Organization (31). Furthermore, the amplitude of the seasonal cycle is most pronounced in the central-interior region of the country, which is consistent with the wider seasonal variation in between its summer and winter temperature and with its low winter temperatures. Populations living in this region may be more exposed to low winter temperatures than those living in the centercoastal area (who are protected by milder temperatures) and may be less prepared for the cold weather than those living in the homogeneously colder southern region. An alternative parsimonious explanation of the bimodal pattern of birth weight in the central and southern regions is the combination of a 12-month seasonal exposure and a nonlinear exposure risk relationship to ambient temperature (23). If both extremely cold and extremely hot temperatures are detrimental for birth weight, but the effects are asymmetric, a bimodal pattern such as the one observed in Chile will emerge. Evidence suggesting that birth weight is associated not only with cold temperatures, but also with heat stress emerging from hot weather (46) renders this hypothesis plausible. Even if seasonal variation in temperature is relatively moderate in the Chilean context, even mild variation may result in substantial effects if a dose-response effect of temperature on birth weight exists. In addition to ambient temperature and exposure to sunshine, the literature suggests other potential environmental exposures with seasonal patterns, such as atmospheric air pollution and drinking water disinfection by-products, as potential determinants of the seasonality of birth weight (47–50). In the case of Chile, atmospheric pollution may be a particularly relevant factor. The capital city of Santiago comprises most of the population in the central-interior region of the country and is one of the most polluted cities in the world, with high levels of particulate material, carbon monoxide, and sulfur monoxide, among others (51). Pollution has a marked seasonal pattern with acute levels in the winter due to the ‘‘thermal inversion’’ phenomenon, whereby a mass of warm air traps cooler air underneath, impeding the dissipation of particulate material and gases. If, as suggested by some studies, the effect of air pollution is stronger in late gestation (52, 53), the large winter decline in birth weight in
AEP Vol. 20, No. 11 November 2010: 818–826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
825
FIGURE 6. Birth weight deviations from the January mean (in grams) across month of birth using month dummies (mean and 95% confidence intervals) and sinusoidal specification. Singleton births 37–41 weeks of gestation, 1987–2007, in central-interior region of Chile, by residence in the Santiago metropolitan area.
the central-interior region may be driven by increase in pollution. In order to examine this possibility, Figure 6 compares the seasonal pattern of birth weight in Santiago vis-a`-vis other areas in the central-interior region, where the pollution levels are substantially lower. Figure 6 shows an identical seasonal pattern in Santiago and other areas of the central-interior region. The amplitude of the cycle is, if anything, slightly smaller in Santiago than in less polluted areas, suggesting that Santiago’s high winter pollution levels are an unlikely confounder for the stronger seasonality of birth weight in the central-interior region. Some of the advantages of this study are the use of a large sample with virtually no selectivity and the availability of gestational age that allows us to select term births, preventing confounding by gestational age. The substantial latitude variation within a single country allows examination of potential etiology of seasonal change in birth weight. Given that we focus on a single country, variation across geographic regions cannot be attributed to differences in samples, analytical techniques, or national demographic or institutional characteristics. Variation in the seasonality of birth weight across regions suggests that it is likely related to environmental rather than sociodemographic composition of births and that exposure to ambient temperature and sunshine may have distinct influences at different gestational ages, specifically at the beginning and end of the pregnancy. However, as in previous studies, latitude and climatic variation are just proxies for a multiplicity of factors that can shape birth weight. These include change in maternal behaviors including smoking, diet and physical activity, and exposure to infectious diseases, among others. The substantial variation found in our study suggests the
relevance of further exploiting natural experiments such as climatic variation within single sites to advance the understanding of the etiology of the seasonal variation of birth weight. The authors thank Danuta Rajs, MD, Head of the Statistics Unit of the Chilean Ministry of Health for providing the birth registry database; and Ghislaine Echevarria MD for helpful advice.
REFERENCES 1. Keller CA, Nugent RP. Seasonal patterns in perinatal mortality and preterm delivery. Am J Epidemiol. 1983;118:689–698. 2. 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. 3. Chodick G, Flash S, Deoitch Y, Shalev V. Seasonality in birth weight: review of global patterns and potential causes. Hum Biol. 2009;81:463– 477. 4. Doblhammer G, Vaupel JW. Lifespan depends on month of birth. Proc Natl Acad Sci U S A. 2001;98:2934–2939. 5. Davies G, Welham J, Chant D, Torrey EF, McGrath J. A systematic review and meta-analysis of Northern Hemisphere season of birth studies in schizophrenia. Schizophr Bull. 2003;29:587–593. 6. Buckles K, Hungerman D. Season on birth and later outcomes: old questions, new answers. NBER Working paper 14673; 2009. 7. Hales CN, Barker DJ, Clark PM, Cox LJ, Fall C, Osmond C, et al. Fetal and infant growth and impaired glucose tolerance at age 64. BMJ. 1991;303(6809):1019–1022. 8. Barker DJ, Gluckman PD, Godfrey KM, Harding JE, Owens JA, Robinson JS. Fetal nutrition and cardiovascular disease in adult life. Lancet. 1993; 341:938–941. 9. Paneth NS. The problem of low birth weight. Future Child. 1995;5(1):19–34. 10. Frankel S, Elwood P, Sweetnam P, Yarnell J, Smith GD. Birthweight, body-mass index in middle age, and incident coronary heart disease. Lancet. 1996;348:1478–1480.
826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
11. Pallotto EK, Kilbride HW. Perinatal outcome and later implications of intrauterine growth restriction. Clin Obstet Gynecol. 2006;49:257–269. 12. Behrman JR, Rosenzweig MR. Returns to birthweight. Rev Econ Stat. 2004;86:586–601. 13. Case A, Fertig A, Paxson C. The lasting impact of childhood health and circumstance. J Health Econ. 2005;24:365–389. 14. Wohlfahrt J, Melbye M, Christens P, Andersen AM, Hjalgrim H. Secular and seasonal variation of length and weight at birth. Lancet. 1998;352:1990. 15. McGrath JJ, Barnett AG, Eyles DW. The association between birth weight, season of birth and latitude. Ann Hum Biol. 2005;32:547–559. 16. Murray LJ, O’Reilly DP, Betts N, Patterson CC, Davey Smith G, Evans AE. Season and outdoor ambient temperature: effects on birth weight. Obstet Gynecol. 2000;96:689–695. 17. Gloria-Bottini F, Magrini A, Bottini E. The effect of genetic and seasonal factors on birth weight. Early Hum Dev. 2009;85:439–441. 18. Waldie KE, Poulton R, Kirk IJ, Silva PA. The effects of pre- and post-natal sunlight exposure on human growth: evidence from the Southern Hemisphere. Early Hum Dev. 2000;60:35–42. 19. Lawlor DA, Leon DA, Davey Smith G. The association of ambient outdoor temperature throughout pregnancy and offspring birthweight: findings from the Aberdeen Children of the 1950s cohort. BJOG. 2005; 112:647–657. 20. Chodick G, Shalev V, Goren I, Inskip PD. Seasonality in birth weight in Israel: new evidence suggests several global patterns and different etiologies. Ann Epidemiol. 2007;17:440–446. 21. Matsuda S, Hiroshige Y, Furuta M, Doi T, Sone T, Kahyo H. Geographic differences in seasonal variation of mean birth weight in Japan. Hum Biol. 1995;67:641–656. 22. McGrath JJ, Keeping D, Saha S, Chant DC, Lieberman DE, O’Callaghan MJ. Seasonal fluctuations in birth weight and neonatal limb length; does prenatal vitamin D influence neonatal size and shape? Early Hum Dev. 2005;81:609–618. 23. McGrath J, Barnett A, Eyles D, Burne T, Pedersen CB, Mortensen PB. The impact of nonlinear exposure-risk relationships on seasonal time-series data: modelling Danish neonatal birth anthropometric data. BMC Med Res Methodol. 2007;7:45. 24. Darrow LA, Strickland MJ, Klein M, Waller LA, Flanders WD, Correa A, et al. Seasonality of birth and implications for temporal studies of preterm birth. Epidemiology. 2009;20:699–706. 25. Lokshin M, Radyakin S. Month of birth and children’s health in India. World Bank Research Working Paper. 2009(October): p. 1–40. 26. Mardones F, Villarroel L, Karzulovic L, Barja S, Arnaiz P, Taibo M, et al. Association of perinatal factors and obesity in 6- to 8-year-old Chilean children. Int J Epidemiol. 2008;37:902–910. 27. Hodrick RJ, Prescott E. Postwar U.S. Business Cycles: An Empirical Investigation. J Money, Credit, and Banking. 1997;29(1):1–16. 28. Ko¨ppen WP. Das geographisca System der Klimate. In: Ko¨ppen WP, Geiger R, eds. Handbuch der klimatologie. Berlin: Gebru¨der Borntraeger; 1936:1–44. 29. Amigo H, Vargas C, Rona RJ. A feeling of well-being accompanied by a period of prosperity and birthweight in Chile: a possible link? Paediatr Perinat Epidemiol. 2005;19:426–434.
AEP Vol. 20, No. 11 November 2010: 818–826
33. Bellastella A, Criscuolo T, Iorio S, Parlato F, Sinisi AA, Sinisi AM, et al. Normal plasma insulin-like growth factor I levels and impaired final stature in adult blind subjects. J Endocrinol Invest. 1994;17:37–40. 34. Gluckman PD, Pinal CS. Regulation of fetal growth by the somatotrophic axis. J Nutr. 2003;133(5 Suppl. 2):1741S–1746S. 35. Gomez JM, Maravall FJ, Gomez N, Navarro MA, Casamitjana R, Soler J. Relationship between 25-(OH) D3, the IGF-I system, leptin, anthropometric and body composition variables in a healthy, randomly selected population. Horm Metab Res. 2004;36:48–53. 36. Clapp JF 3rd, Schmidt S, Paranjape A, Lopez B. Maternal insulin-like growth factor-I levels (IGF-I) reflect placental mass and neonatal fat mass. Am J Obstet Gynecol. 2004;190:730–736. 37. Suttie JM, White RG, Breier BH, Gluckman PD. Photoperiod associated changes in insulin-like growth factor-I in reindeer. Endocrinology. 1991;129:679–682. 38. Webster JR, Corson ID, Littlejohn RP, Stuart SK, Suttie JM. Effects of photoperiod on the cessation of growth during autumn in male red deer and growth hormone and insulin-like growth factor-I secretion. Gen Comp Endocrinol. 1999;113:464–477. 39. DeLuca HF, Zierold C. Mechanisms and functions of vitamin D. Nutr Rev. 1998;56(2 Pt 2):S4–S10; discussion S54–75. 40. Webb AR, Kline L, Holick MF. Influence of season and latitude on the cutaneous synthesis of vitamin D3: exposure to winter sunlight in Boston and Edmonton will not promote vitamin D3 synthesis in human skin. J Clin Endocrinol Metab. 1988;67:373–378. 41. Holick MF. Environmental factors that influence the cutaneous production of vitamin D. Am J Clin Nutr. 1995;61(3 Suppl):638S–645S. 42. MacLennan WJ, Hamilton JC, Darmady JM. The effects of season and stage of pregnancy on plasma 25-hydroxy-vitamin D concentrations in pregnant women. Postgrad Med J. 1980;56:75–79. 43. Neild PJ, Syndercombe-Court D, Keatinge WR, Donaldson GC, Mattock M, Caunce M. Cold-induced increases in erythrocyte count, plasma cholesterol and plasma fibrinogen of elderly people without a comparable rise in protein C or factor X. Clin Sci (Lond). 1994;86:43–48. 44. Keatinge WR, Coleshaw SR, Cotter F, Mattock M, Murphy M, Chelliah R. Increases in platelet and red cell counts, blood viscosity, and arterial pressure during mild surface cooling: factors in mortality from coronary and cerebral thrombosis in winter. Br Med J (Clin Res Ed). 1984;289:1405–1408. 45. Elter K, Ay E, Uyar E, Kavak ZN. Exposure to low outdoor temperature in the midtrimester is associated with low birth weight. Aust N Z J Obstet Gynaecol. 2004;44:553–557. 46. Deschenes O, Greenstone M, Guryan J. Climate change and birth weight. Am Econ Rev. 2009;99:211–217. 47. Bove F, Shim Y, Zeitz P. Drinking water contaminants and adverse pregnancy outcomes: a review. Environ Health Perspect. 2002;110(Suppl. 1):61–74. 48. Maisonet M, Correa A, Misra D, Jaakkola JJ. A review of the literature on the effects of ambient air pollution on fetal growth. Environ Res. 2004;95:106–115. 49. Stillerman KP, Mattison DR, Giudice LC, Woodruff TJ. Environmental exposures and adverse pregnancy outcomes: a review of the science. Reprod Sci. 2008;15:631–650. 50. Wang L, Pinkerton KE. Air pollutant effects on fetal and early postnatal development. Birth Defects Res C Embryo Today. 2007;81:144–154.
30. Barnett AG, Dobson AJ. Analysing seasonal health data. Statistics for biology and health. Berlin London: Springer; 2010; xiii.
51. Hewitt CN, Jackson AV. Handbook of atmospheric science. Malden (MA): Blackwell Publishing; 2003.
31. Tustin K, Gross J, Hayne H. Maternal exposure to first-trimester sunshine is associated with increased birth weight in human infants. Dev Psychobiol. 2004;45:221–230.
52. Maisonet M, Bush TJ, Correa A, Jaakkola JJ. Relation between ambient air pollution and low birth weight in the Northeastern United States. Environ Health Perspect. 2001;109(Suppl. 3):351–356.
32. Valcavi R, Zini M, Maestroni GJ, Conti A, Portioli I. Melatonin stimulates growth hormone secretion through pathways other than the growth hormone-releasing hormone. Clin Endocrinol (Oxf). 1993;39:193–199.
53. Ritz B, Yu F. The effect of ambient carbon monoxide on low birth weight among children born in southern California between 1989 and 1993. Environ Health Perspect. 1999;107:17–25.
PURPOSE: Research suggests a relationship between birth weight and season of birth, but findings vary across countries and underlying factors are not well understood. We examine the seasonality of birth weight and explore alternative hypotheses for its etiologydexposure to environmental factors and varying socioeconomic composition of mothersdin Chile. METHODS: Birth weight of approximately 5 million Chilean singleton live births 37 of 41 weeks of gestation between 1987 and 2007 were analyzed for seasonality by using regression models with month dummies and parametric sinusoidal specifications. Multivariate models with socioeconomic covariates and interactions across geographic regions examine potential factors accounting for seasonal variation. RESULTS: Marked 12-month and 6-month periodic cycles were found. The amplitude and phase of the seasonal variation change across geographic regions. In the low-latitude northern region, there is a spring peak and a fall nadir, while in middle-latitude colder regions, a bimodal periodicity emerges with peaks in spring and fall, a pronounced winter nadir, and smaller nadir in the summer. Socioeconomic composition of mothers is found to vary with annual periodicity, but it does not account for the seasonality in birth weight. CONCLUSIONS: Environmental factors rather than the socioeconomic composition of mothers likely account for seasonal variation in birth weight. The change in periodicity of birth weight across latitudes is consistent with a beneficial exposure to sunlight both early and late in the pregnancy, and a detrimental late exposure to cold temperatures only in areas with low winter temperatures. Ann Epidemiol 2010;20:818–826. Ó 2010 Elsevier Inc. All rights reserved. KEY WORDS:
Birth Weight, Seasonality, Time-series Analysis, Socioeconomic Factors, Temperature, Light.
INTRODUCTION A substantial association exists between season of birth and individual outcomes throughout the life cycle. These outcomes include early health measures such as birth weight and gestational age (1–3), life expectancy (4), schizophrenia (5), and educational attainment (6). Birth weight is particularly important because it is the main proximate determinant of infant mortality, and it affects health outcomes in childhood and adulthood (7–11) as well as cognitive, educational, and socioeconomic attainment (12, 13). Given the influence of birth weight on later attainment, the seasonality of adult outcomes may be at least partly mediated by birth weight. Several studies have documented seasonal variation in birth weight, but results are far from homogeneous. Even if studies are restricted to term or gestational age–adjusted births to avoid confounding by length of gestation, some find a marked annual cycle, with most detecting peaks in spring (14, 15) and winter (16, 17). However, autumn and summer peaks have also been found (18–20). Others find bimodal periodicities with peaks consistently falling in fall and spring (21–23). Part of the variation may be due to From the Departments of Sociology (F.T.) and Economics (A.C.), New York University. Address correspondence to: Florencia Torche, PhD, Department of Sociology, New York University, New York. E-mail: [email protected]. Received April 7, 2010; accepted August 18, 2010. Ó 2010 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
different specification in the dependent variable (term births vs. gestation-adjusted weight) or to the fact that some studies do not explicitly check for a semiannual periodicity. Substantive reasons may also account for the differences in findings. Variation in birth weight depends on a complex set of genetic and epigenetic factors affecting the maternalplacental-fetal unit. Season of birth provides a natural experiment for the generation of alternative hypotheses about the effect of environmental factors (15). A useful strategy is to examine changes in seasonal variation across latitudes, under the assumption that latitude provides a proxy for important factors such as temperature and sunlight exposure. Meta-analysis suggests the seasonality of birth weight varies across latitudes, although different study methodologies may hamper conclusive interpretation (3). An alternative hypothesis proposes that the seasonality of birth weight emerges from the varying socioeconomic composition of mothers conceiving in different months of the year (6, 24, 25). To the extent that sociodemographic characteristics are correlated with offspring’s birth weight and that different sociodemographic groups are more likely to give birth in certain seasons, part of the variation attributed to environmental factors may be driven by the changing distribution of risk factors in the pregnancy set. For example, if the months of highest temperature correspond to those in which low–socioeconomic status (SES) women are most likely to give birth, temperature may appear predictive 1047-2797/$ - see front matter doi:10.1016/j.annepidem.2010.08.005
AEP Vol. 20, No. 11 November 2010: 818–826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
819
of low birth weight only because low SES is a risk factor for low birth weight. METHODS We examined seasonality of birth weight in Chile during a 21year period. Populated areas in Chile range from latitude 18 S to 56 S (comparable to the range between Niger and Sweden), with substantial climatic variation. We exploit such variation to examine potential environmental determinants of birth weight seasonality. Data were derived from the Chilean Birth Registry database established by the Chilean Ministry of Health. By Chilean law, all births must be reported to the Ministry of Health. The database is constructed on the basis of live-birth certificates, a form completed by the professional (MD or certified midwife) who attends the delivery (99.8% of deliveries are attended by a professional in Chile) and measured with little error (26). The birth certificate contains information on birth weight, gestational age, maternal residence, and maternal demographic characteristics. We examined birth weight in all live-born singletons with gestations between 37 and 41 weeks from the years 1987 through 2007 (both inclusive) for a total of 4,968,912 births. Records of individual birth weight were summarized into a time series of monthly means. To observe short-run fluctuations, the trend in these means was then extracted using a Hodrick-Prescott filter with monthly smoothing parameter l Z 14,400 (27). Spectral analysis showed that the fundamental frequency components of the series are annual and semiannual (results from spectral analysis not shown, available from authors upon request). Regression models were then implemented to account for the seasonal fluctuation in birth weight. Birth weight was predicted by 11 month dummies (January used as baseline for comparison) and a full set of year dummies to account for trend. Additionally, we model month effects parametrically by using sinusoidal functions. As suggested by the two component frequencies extracted from the spectral analysis, we allowed for a 12-month periodicity [A*sin([month4] * 2p/12)]d‘‘annual model’’d and for a 6-month periodicity [A*sin([month-4 ] * 2p/6)]d ‘‘semi-annual model’’. In both cases, 4 denotes an integer phase shift, which controls the location of the seasonal peak, and A identifies amplitude calculated as peak-to-zero (or nadir-to-zero) distance or half of the maximal difference between highest and lowest weights during the year. Nonlinear least squares was used to estimate the cyclical phase. The phase parameter was then approximated to result in an integer value for the peak month in the annual and semiannual cycle. Then, amplitude was calculated using ordinary least square regression for a fixed value of the phase. Regression models were subsequently extended in two ways. First, we stratify the sample by geographic region and estimate region-specific regression models to evaluate
FIGURE 1. Average monthly temperature in degrees Celsius across geographic regions, Chile. Monthly temperatures are averages over the second half of the 20th century. Regional averages are obtained from arithmetic mean of temperatures in the larger cities comprising most of the population living in each region. Cities included are Arica, Antofagasta, Copiapo´, and La Serena (northern region); Valparaı´so, Concepcio´ n, (central-coastal region), Santiago and Linares (central-interior region), and Temuco and Puerto Montt (southern region).
variation in phase and amplitude. Four regions are distinguished: north (latitudes 18 S–31 S), central-coastal and central-interior (32 S–37 S), and south (38 S–56 S). The low-latitude northern region has milder weather and slight temperature variation throughout the year (Fig. 1). Colder temperatures are observed in the central, and particularly the southern, regions. The distinction between centralcoastal and central-interior is not based on latitude but on the summer-winter temperature variation, which is much narrower in the coastal area due to the moderating influence of the Pacific Ocean. The central-interior region displays maximum summer temperatures as high as those in the northern region and minimum winter temperatures as low as those in the southern region. According to the standard Ko¨ppen climate system, the northern area is characterized by a dry arid and semi-arid climate, whereas the central and southern areas feature a temperate/mesothermal climate type (28). Second, we added measures of maternal sociodemographic characteristics to the models to assess whether the varying socioeconomic composition of births across the year accounts for seasonal variation in birth weight. The following variables were included: proportion of mothers that are married (married) each month of birth, proportion of mothers older than 16 years of age (adult), proportion of mothers with postsecondary education (college), and proportion of mothers living in urban areas (urban). RESULTS Figure 2 presents the original series and extracted trend for the birth-weight data between 1987 and 2007. Some
820
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
AEP Vol. 20, No. 11 November 2010: 818–826
increase in average birth weight occurs between 1987 and 1995dfrom 3,300 to 3,400 grams, a well-documented trend potentially associated with economic growth (29). Increase is followed by stability between 1995 and 2007. Evaluation of the original series with respect to the trend suggests a possible seasonal component with annual periodicity. Regression analysis shows that birth weight has a bimodal peak in October–November (spring) and March (fall) (model 1 in Table 1, ‘‘All births’’ specification). A sinusoidal formulation with 12-month and 6-month periodicity fits the observed data extremely well, accounting for most of the variation across months (Model 2 in Table 1, ‘‘All births’’ specification). The variable ‘‘Annual cycle’’ identifies 12-month periodicity, while ‘‘Semiannual cycle’’ accounts for any additional 6-month periodicity. In the analysis of all births, the bestfitting model includes an annual phase with peak in January and semiannual peak in the same month. Note that because the parameter estimates are of approximately the same magnitude but opposite sign, they should offset each other, resulting in a value close to zero for the first month of the year. Figure 3 depicts the parameter estimates associated with the month dummies and their 95% confidence intervals, superimposing the sum of the annual and semiannual sinusoidal functions. Figure 3 shows peaks in spring and fall, a pronounced winter nadir, and a smaller nadir in the summer. Infants born in spring or fall are, on average, 16 grams heavier than those born in the winter (p ! 0.001). To explore potential determinants of birth weight seasonality, we stratify our sample by geographic region and replicate the regression analysis across groups. Results are displayed in the remaining columns of Table 1. They indicate that the phase and amplitude of seasonal variation substantially changes across region. In the low-latitude northern region, there is marked annual periodicity with
highest birth weight in spring and lowest birth weight in the fall. An annual sinusoidal function with peak in October and nadir in April suffices to account for the data, with no 6-month periodicity found (Fig. 4, panel 1). The phase of the cycle shifts for the other three geographic regions (panels 2–4 in Fig. 4). In these regions the periodicity is markedly bimodal, with twin peaks in spring and fall, and a pronounced winter nadir. This pattern remains unchanged across latitudes 32 S to 56 S (central and southern regions of the country). Furthermore, the amplitude of the cycle varies across regions, with wider amplitude in the centralinterior area, which features the widest seasonal variation in temperature as indicated by Figure 1. In the centralinterior region, infants born in the spring and fall weigh, on average, 21 grams more than those born in the winter (p ! 0.001). This amplitude is, expectedly, greater than the average across the country reported in Figure 3, which combines seasonal patterns with different phases (30). Geographic variation suggests that environmental factors may affect seasonality of birth weight. However, no controls have been introduced for the socioeconomic characteristics of pregnancies, so the findings may be an artifact of the changing sociodemographic composition of births throughout calendar months. As with birth weight, these variables display seasonal variation. The compositional change of maternal marital status and age show a pronounced annual pattern (Fig. 5). Married women and those young women 17 years of age and older are more likely to give birth in the summer and least likely to give birth in the winter, potentially accounting for the winter peak found in central and southern regions. Capturing the seasonality of births to college-educated and urban women requires annual and semiannual components. While the seasonal variation in the proportion of urban women is limited, college-educated women are more likely to give birth in the spring, potentially accounting for the spring peak found throughout the sample. Table 2 adds these sociodemographic factors to the regression models by using sinusoidal specifications. To the extent sociodemographic compositional variation of births accounts for the seasonality of birth weight, we expect that the sinusoidal parameters will decline in magnitude and become statistically insignificant. The findings are not consistent with this hypothesis. The parameters remain virtually unchanged across all regions, indicating that the seasonality of birth weight is not accounted for by the changing distribution of risk factors related to sociodemographic characteristics throughout the year.
FIGURE 2. Time series of mean monthly birth weight. Singleton births at 37–41 weeks of gestation, 1987–2007, Chile. Trend extracted by use of a Hodrick-Prescott filter with monthly smoothing parameter l Z 14,400.
DISCUSSION We find a marked seasonality of birth weight in Chile. Seasonal variation changes across geographic regions. In
AEP Vol. 20, No. 11 November 2010: 818–826
TABLE 1. Linear regression model, dependent variable birth weight, singleton births 37–41 weeks of gestation, 1987–2007, Chile. Models for all births and for births by geographical region* All births Model 1 b (SE) January (omitted) February March April May June July August September October November December Annual cycley Semiannual cycley Constant N R2
6.02{ (1.57) 8.10{ (1.40) 4.47{ (1.27) 2.80x (1.36) 3.07z (1.71) 8.83{ (1.62) 3.68x (1.63) 1.11 (1.42) 6.31{ (1.40) 7.03{ (1.43) 5.00x (1.78)
3314 252 98.26
Region 1dNorth Model 2 b (SE)
4.76{ (0.51) 4.62{ (0.46) 3316 252 98.13
Model 1 b (SE) 3.16 (3.43) 2.76 (3.18) 3.60 (3.10) 2.16 (2.98) 2.43 (3.40) 1.73 (3.49) 7.31x (3.42) 1.80 (3.08) 8.75x (3.59) 9.13{ (3.12) 4.89 (3.50)
3308 252 90.92
Model 2 b (SE)
5.27{ (0.91) 1.03 (0.96) 3310 252 90.46
Region 2dCentral-Coastal
Region 3dCentral-Interior
Model 1 b (SE)
Model 1 b (SE)
9.20x (3.96) 10.70{ (3.65) 10.97{ (3.43) 7.49z (3.87) 0.88 (3.67) 4.70 (3.40) 3.23 (4.05) 3.80 (3.66) 7.97x (3.77) 3.39 (3.83) 4.45 (3.85)
3313 252 93.71
Model 2 b (SE)
4.21{ (0.95) 5.20{ (0.90) 3317 252 93.52
6.69{ (1.88) 9.93{ (1.74) 4.15x (1.72) 2.10 (1.54) 5.92{ (2.05) 11.52{ (2.00) 5.96{ (2.02) 1.40 (2.01) 7.35{ (1.75) 8.52{ (1.55) 5.08x (1.89)
3313 252 97.50
Model 2 b (SE)
6.43{ (0.60) 5.76{ (0.57) 3314 252 97.32
Region 4dSouth Model 1 b (SE) 8.23{ (2.81) 7.26{ (2.75) 7.05{ (2.39) 5.66x (2.73) 0.22 (2.88) 11.23{ (2.80) 4.68 (2.90) 4.26 (2.62) 2.46 (2.32) 1.87 (2.89) 2.58 (3.27)
3328 252 95.78
Model 2 b (SE)
6.34{ (0.87) 3.07{ (0.82) 3329 252 95.45
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
b Z parameter estimate; SE Z standard error. *Robust standard errors in parentheses. All regressions include a full set of year dummies to account for trend. For ‘‘all birth’’ and regional samples, model 1 includes a full set of month dummies and model 2 includes annual and semiannual parameter estimates from the best-fitting sinusoidal specification. y The phase (4) of the annual cycle varies across regions: 4 Z –2 for all births, 1 in north region, –1 in central-coastal region, –2 in central-interior region, and –1 in south region. Consequently, annual peaks are as follows: January for all births, April in north region, February in central-coastal region, January in central-interior region, and February in the south region. In the semiannual cycle, 4 Z –1/2, and peak is in January for the ‘‘all birth’’ and all regional samples. z p ! .05. x p ! .01. { p ! .001.
821
822
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
FIGURE 3. Birth weight (in grams) across month of birth using month dummies (mean and 95% confidence interval) and sinusoidal specification, including annual and semiannual components. Singleton births 37–41 weeks of gestation, 1987–2007, Chile.
low latitude, mild-weather areas with little temperature oscillation throughout the year, there is a manifest spring peak and fall nadir. In the rest of the country, characterized by colder winters and wider seasonal variation in temperature, there is a bimodal distribution with winter and summer nadirs, and peaks in the fall and spring. We also show that these patterns are not accounted for by the socioeconomic composition of mothers, although sociodemographic characteristics do vary seasonally.
AEP Vol. 20, No. 11 November 2010: 818–826
Seasonal variation of birth weight fluctuates between 13 and 21 grams across geographic regions, a figure in line with previous research using comparable large data sets (16, 20, 22, 23). Even if this magnitude is small, it is substantial in the context of an ecological study that captures populationlevel averages. Under the reasonable assumption that the seasonal variation is heterogeneous across individuals and affects only a segment of the population, the effect may be substantially greater among affected pregnancies. For example, if one half of pregnancies were affected, the effect would be twice as large as the reported population-level average, and if only 10% of pregnancies were affected, the effect would be 10 times as large. The potentially large effect of the ‘‘treatment on the treated’’ suggests the importance of further studying heterogeneity in seasonality of birth weight along demographic and socioeconomic lines, in order to detect the population most at risk. In addition, the geographic variation in seasonality of birth weight sheds light on its etiology. Which factors can explain the observed patterns? Factors highlighted in some developing contexts, such as seasonal variation in nutrient deficiency, physical exertion, and exposure to malaria or other infections, have been largely eradicated in Chile. In countries where these exposures have been eradicated, the literature focuses on two environmental factors: exposure to sunlight and to low temperatures. Sunshine exposure in early pregnancy has been found to be related to higher birth weight (18, 31). One potential mechanism suggests that
FIGURE 4. Birth weight deviations from the January mean (in grams) across month of birth using month dummies (mean and 95% confidence interval) and sinusoidal specification. Singleton births 37–41 weeks of gestation, 1987–2007, in Chile, by geographic region.
AEP Vol. 20, No. 11 November 2010: 818–826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
823
FIGURE 5. Maternal socioeconomic characteristics across month of birth using month dummies (mean and 95% confidence interval) and sinusoidal specification. Singleton births 37–41 weeks of gestation, 1987–2007, in Chile. Upper left panel: Percentage of births to women older than 16 years of age. Upper right panel: Percentage of births to married women. Lower left panel: Percentage of births to women with at least some college education. Lower right panel: Percentage of births to women living in urban areas. All values are deviations from the January mean (equal zero in all figures). Seasonality of births to women older than 16 years (adult) and married women (married) modeled by a 12-month sinusoidal parameter with peak in February and nadir in August. Seasonality of percentage of births to women with college education (college) modeled by a 12-month sinusoidal parameter with peak in October and nadir in April, and a 6-month sinusoidal parameter with peaks in April and October and nadirs in January and July. Seasonality of percentage of women living in urban areas (urban) modeled by a 12-month sinusoidal parameter with peak in November and nadir in May, and a 6-month sinusoidal parameter with peaks in May and November and nadirs in February and August.
sunlight exposure during the first trimester of gestation may stimulate growth hormone release by inhibiting production of maternal pineal melatonin (32). Some evidence for an inverse relationship between melatonin and growth hormone (GH) production exists (33). However, a caveat about this hypothesis is that GH receptors are functionally immature prior to approximately 36 weeks of gestation (34). An alternative mechanism refers to insulin-like growth factor (IGF)–1. This growth factor displays seasonal fluctuation (35) and is associated with fetal growth (36). IGF-1 plays an important role in prenatal growth, and animal studies show that it varies with exposure to light (37, 38). Another avenue for the influence of sunlight exposure is the production of vitamin D. Vitamin D3 is unique in that its production depends primarily on the action of sunlight on the skin (39). Its production is strongly and consistently associated with the duration of the photoperiod, which in turn is influenced by latitude and season (40, 41). Vitamin D deficiency is more prevalent during the winter months, even in low-latitude areas (41). Fetal vitamin D requirements increase during pregnancy, related to the increased need for fetal calcium; thus maternal vitamin D
levels tend to fall during the third trimester, especially if it occurs during the winter (42). This alternative mechanism may result in a deficit of sunlight exposure being more consequential for birth weight late rather than early in the pregnancy. Other studies suggest that seasonal variation of birth weight is related to the exposure to cold temperatures in mid and late gestation. Exposure to low temperatures raises fibrinogen levels in plasma (43), increases blood viscosity (44), and might induce vascular constriction in the placenta, thereby diminishing uteroplacental blood flow. This may result in lower birth weight (16, 45). This effect may be more pronounced in mid or late gestation when the fetus experiences rapid growth. Research has documented a potential late-pregnancy effect in areas with cold winters (19). We found a pronounced spring peak throughout the regions analyzed. This peak could be induced by earlypregnancy exposure to sunlight during the summer, and it would consistently affect all latitudes insofar as the seasonal fluctuation of sunlight also affects low-latitude areas. A second fall peak emerges in the central and southern areas. A potential explanation is the late-pregnancy exposure to
3314 252 97.32 3317 252 93.52 3310 252 90.50 3316 252 98.13
AEP Vol. 20, No. 11 November 2010: 818–826
b Z parameter estimate; SE Z standard error. *Robust standard errors in parentheses. All regressions include a full set of year dummies to account for trend. Values indicate percentage of women who are adult, married, college educated, and urban multiplied by 100. y The phase (4) of the annual cycle varies across regions: 4 Z –2 for all births, 1 in north region, –1 in central-coastal region, –2 in central-interior region, and –1 in south region. Consequently, annual peaks are as follows: January for all births, April in north region, February in central-coastal region, January in central-interior region, and February in the south region. In the semiannual cycle, 4 Z –1/2, and peak is in January for the ‘‘all birth’’ sample and all regional samples. z p ! .05. x p ! .01. { p ! .001.
3329 252 95.45
2.31x (1.13) 4.12{ (0.83) 1.65 (1.40) 2.99{ (0.61) 2.66{ (0.84) 1.01y (0.59) 2895 252 96.07
b (SE) b (SE)
6.34{ (0.87) 3.07{ (0.82)
5.83{ (0.77) 5.64{ (0.64) 1.07 (2.02) 0.92 (0.59) 0.66 (0.85) 0.55 (0.80) 3387 252 97.36
b (SE) b (SE)
6.43{ (0.60) 5.76{ (0.57)
4.00{ (1.27) 4.93{ (0.95) 2.14 (1.92) 0.83 (0.64) 0.07 (0.72) 0.96 (1.10) 3541 252 93.62
b (SE) b (SE)
4.21{ (0.95) 5.20{ (0.90)
5.06{ (1.02) 1.16 (0.96) 1.69 (1.79) 0.06 (0.61) 0.72 (0.91) 0.92 (0.98) 3554 252 90.63
b (SE) b (SE)
5.27{ (0.91) 1.03 (0.95)
4.09{ (0.70) 4.63{ (0.52) 1.13 (1.93) 0.58 (0.59) 0.17 (0.95) 0.20 (0.81) 3403 252 98.16
b (SE) b (SE)
All births
4.76{ (0.51) 4.62{ (0.46)
Annual cycley Semiannual cycley Adult Married College Urban Constant Observations R2
Region 1dNorth
Region 2dCentral-Coastal
Region 3dCentral-Interior
Region 4dSouth
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
TABLE 2. Birth weight variation across month of birth using sinusoidal specification. Controls added for socioeconomic characteristics of mothers. Chile singleton births, 37–47 weeks of gestation, 1987–2007, Chile by geographic region*
824
sunshine for these births; the absence of a peak in the northern region is likely because, in this mild climate, individuals spend more time outdoors even in the winter, thus reaching a satisfactory exposure level. A particularity of the central and southern regions is the pronounced winter nadir, which does not exist in the low-latitude northern region. We hypothesize that this nadir may be due to the exposure to cold during late gestation. Again, this detrimental effect would not manifest itself in the northern region because minimum winter temperature is rather high, reaching 14 C (see Fig. 1), which is close to the minimum indoor temperature of 16 C recommended by the World Health Organization (31). Furthermore, the amplitude of the seasonal cycle is most pronounced in the central-interior region of the country, which is consistent with the wider seasonal variation in between its summer and winter temperature and with its low winter temperatures. Populations living in this region may be more exposed to low winter temperatures than those living in the centercoastal area (who are protected by milder temperatures) and may be less prepared for the cold weather than those living in the homogeneously colder southern region. An alternative parsimonious explanation of the bimodal pattern of birth weight in the central and southern regions is the combination of a 12-month seasonal exposure and a nonlinear exposure risk relationship to ambient temperature (23). If both extremely cold and extremely hot temperatures are detrimental for birth weight, but the effects are asymmetric, a bimodal pattern such as the one observed in Chile will emerge. Evidence suggesting that birth weight is associated not only with cold temperatures, but also with heat stress emerging from hot weather (46) renders this hypothesis plausible. Even if seasonal variation in temperature is relatively moderate in the Chilean context, even mild variation may result in substantial effects if a dose-response effect of temperature on birth weight exists. In addition to ambient temperature and exposure to sunshine, the literature suggests other potential environmental exposures with seasonal patterns, such as atmospheric air pollution and drinking water disinfection by-products, as potential determinants of the seasonality of birth weight (47–50). In the case of Chile, atmospheric pollution may be a particularly relevant factor. The capital city of Santiago comprises most of the population in the central-interior region of the country and is one of the most polluted cities in the world, with high levels of particulate material, carbon monoxide, and sulfur monoxide, among others (51). Pollution has a marked seasonal pattern with acute levels in the winter due to the ‘‘thermal inversion’’ phenomenon, whereby a mass of warm air traps cooler air underneath, impeding the dissipation of particulate material and gases. If, as suggested by some studies, the effect of air pollution is stronger in late gestation (52, 53), the large winter decline in birth weight in
AEP Vol. 20, No. 11 November 2010: 818–826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
825
FIGURE 6. Birth weight deviations from the January mean (in grams) across month of birth using month dummies (mean and 95% confidence intervals) and sinusoidal specification. Singleton births 37–41 weeks of gestation, 1987–2007, in central-interior region of Chile, by residence in the Santiago metropolitan area.
the central-interior region may be driven by increase in pollution. In order to examine this possibility, Figure 6 compares the seasonal pattern of birth weight in Santiago vis-a`-vis other areas in the central-interior region, where the pollution levels are substantially lower. Figure 6 shows an identical seasonal pattern in Santiago and other areas of the central-interior region. The amplitude of the cycle is, if anything, slightly smaller in Santiago than in less polluted areas, suggesting that Santiago’s high winter pollution levels are an unlikely confounder for the stronger seasonality of birth weight in the central-interior region. Some of the advantages of this study are the use of a large sample with virtually no selectivity and the availability of gestational age that allows us to select term births, preventing confounding by gestational age. The substantial latitude variation within a single country allows examination of potential etiology of seasonal change in birth weight. Given that we focus on a single country, variation across geographic regions cannot be attributed to differences in samples, analytical techniques, or national demographic or institutional characteristics. Variation in the seasonality of birth weight across regions suggests that it is likely related to environmental rather than sociodemographic composition of births and that exposure to ambient temperature and sunshine may have distinct influences at different gestational ages, specifically at the beginning and end of the pregnancy. However, as in previous studies, latitude and climatic variation are just proxies for a multiplicity of factors that can shape birth weight. These include change in maternal behaviors including smoking, diet and physical activity, and exposure to infectious diseases, among others. The substantial variation found in our study suggests the
relevance of further exploiting natural experiments such as climatic variation within single sites to advance the understanding of the etiology of the seasonal variation of birth weight. The authors thank Danuta Rajs, MD, Head of the Statistics Unit of the Chilean Ministry of Health for providing the birth registry database; and Ghislaine Echevarria MD for helpful advice.
REFERENCES 1. Keller CA, Nugent RP. Seasonal patterns in perinatal mortality and preterm delivery. Am J Epidemiol. 1983;118:689–698. 2. 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. 3. Chodick G, Flash S, Deoitch Y, Shalev V. Seasonality in birth weight: review of global patterns and potential causes. Hum Biol. 2009;81:463– 477. 4. Doblhammer G, Vaupel JW. Lifespan depends on month of birth. Proc Natl Acad Sci U S A. 2001;98:2934–2939. 5. Davies G, Welham J, Chant D, Torrey EF, McGrath J. A systematic review and meta-analysis of Northern Hemisphere season of birth studies in schizophrenia. Schizophr Bull. 2003;29:587–593. 6. Buckles K, Hungerman D. Season on birth and later outcomes: old questions, new answers. NBER Working paper 14673; 2009. 7. Hales CN, Barker DJ, Clark PM, Cox LJ, Fall C, Osmond C, et al. Fetal and infant growth and impaired glucose tolerance at age 64. BMJ. 1991;303(6809):1019–1022. 8. Barker DJ, Gluckman PD, Godfrey KM, Harding JE, Owens JA, Robinson JS. Fetal nutrition and cardiovascular disease in adult life. Lancet. 1993; 341:938–941. 9. Paneth NS. The problem of low birth weight. Future Child. 1995;5(1):19–34. 10. Frankel S, Elwood P, Sweetnam P, Yarnell J, Smith GD. Birthweight, body-mass index in middle age, and incident coronary heart disease. Lancet. 1996;348:1478–1480.
826
Torche and Corvalan SEASONALITY OF BIRTH WEIGHT
11. Pallotto EK, Kilbride HW. Perinatal outcome and later implications of intrauterine growth restriction. Clin Obstet Gynecol. 2006;49:257–269. 12. Behrman JR, Rosenzweig MR. Returns to birthweight. Rev Econ Stat. 2004;86:586–601. 13. Case A, Fertig A, Paxson C. The lasting impact of childhood health and circumstance. J Health Econ. 2005;24:365–389. 14. Wohlfahrt J, Melbye M, Christens P, Andersen AM, Hjalgrim H. Secular and seasonal variation of length and weight at birth. Lancet. 1998;352:1990. 15. McGrath JJ, Barnett AG, Eyles DW. The association between birth weight, season of birth and latitude. Ann Hum Biol. 2005;32:547–559. 16. Murray LJ, O’Reilly DP, Betts N, Patterson CC, Davey Smith G, Evans AE. Season and outdoor ambient temperature: effects on birth weight. Obstet Gynecol. 2000;96:689–695. 17. Gloria-Bottini F, Magrini A, Bottini E. The effect of genetic and seasonal factors on birth weight. Early Hum Dev. 2009;85:439–441. 18. Waldie KE, Poulton R, Kirk IJ, Silva PA. The effects of pre- and post-natal sunlight exposure on human growth: evidence from the Southern Hemisphere. Early Hum Dev. 2000;60:35–42. 19. Lawlor DA, Leon DA, Davey Smith G. The association of ambient outdoor temperature throughout pregnancy and offspring birthweight: findings from the Aberdeen Children of the 1950s cohort. BJOG. 2005; 112:647–657. 20. Chodick G, Shalev V, Goren I, Inskip PD. Seasonality in birth weight in Israel: new evidence suggests several global patterns and different etiologies. Ann Epidemiol. 2007;17:440–446. 21. Matsuda S, Hiroshige Y, Furuta M, Doi T, Sone T, Kahyo H. Geographic differences in seasonal variation of mean birth weight in Japan. Hum Biol. 1995;67:641–656. 22. McGrath JJ, Keeping D, Saha S, Chant DC, Lieberman DE, O’Callaghan MJ. Seasonal fluctuations in birth weight and neonatal limb length; does prenatal vitamin D influence neonatal size and shape? Early Hum Dev. 2005;81:609–618. 23. McGrath J, Barnett A, Eyles D, Burne T, Pedersen CB, Mortensen PB. The impact of nonlinear exposure-risk relationships on seasonal time-series data: modelling Danish neonatal birth anthropometric data. BMC Med Res Methodol. 2007;7:45. 24. Darrow LA, Strickland MJ, Klein M, Waller LA, Flanders WD, Correa A, et al. Seasonality of birth and implications for temporal studies of preterm birth. Epidemiology. 2009;20:699–706. 25. Lokshin M, Radyakin S. Month of birth and children’s health in India. World Bank Research Working Paper. 2009(October): p. 1–40. 26. Mardones F, Villarroel L, Karzulovic L, Barja S, Arnaiz P, Taibo M, et al. Association of perinatal factors and obesity in 6- to 8-year-old Chilean children. Int J Epidemiol. 2008;37:902–910. 27. Hodrick RJ, Prescott E. Postwar U.S. Business Cycles: An Empirical Investigation. J Money, Credit, and Banking. 1997;29(1):1–16. 28. Ko¨ppen WP. Das geographisca System der Klimate. In: Ko¨ppen WP, Geiger R, eds. Handbuch der klimatologie. Berlin: Gebru¨der Borntraeger; 1936:1–44. 29. Amigo H, Vargas C, Rona RJ. A feeling of well-being accompanied by a period of prosperity and birthweight in Chile: a possible link? Paediatr Perinat Epidemiol. 2005;19:426–434.
AEP Vol. 20, No. 11 November 2010: 818–826
33. Bellastella A, Criscuolo T, Iorio S, Parlato F, Sinisi AA, Sinisi AM, et al. Normal plasma insulin-like growth factor I levels and impaired final stature in adult blind subjects. J Endocrinol Invest. 1994;17:37–40. 34. Gluckman PD, Pinal CS. Regulation of fetal growth by the somatotrophic axis. J Nutr. 2003;133(5 Suppl. 2):1741S–1746S. 35. Gomez JM, Maravall FJ, Gomez N, Navarro MA, Casamitjana R, Soler J. Relationship between 25-(OH) D3, the IGF-I system, leptin, anthropometric and body composition variables in a healthy, randomly selected population. Horm Metab Res. 2004;36:48–53. 36. Clapp JF 3rd, Schmidt S, Paranjape A, Lopez B. Maternal insulin-like growth factor-I levels (IGF-I) reflect placental mass and neonatal fat mass. Am J Obstet Gynecol. 2004;190:730–736. 37. Suttie JM, White RG, Breier BH, Gluckman PD. Photoperiod associated changes in insulin-like growth factor-I in reindeer. Endocrinology. 1991;129:679–682. 38. Webster JR, Corson ID, Littlejohn RP, Stuart SK, Suttie JM. Effects of photoperiod on the cessation of growth during autumn in male red deer and growth hormone and insulin-like growth factor-I secretion. Gen Comp Endocrinol. 1999;113:464–477. 39. DeLuca HF, Zierold C. Mechanisms and functions of vitamin D. Nutr Rev. 1998;56(2 Pt 2):S4–S10; discussion S54–75. 40. Webb AR, Kline L, Holick MF. Influence of season and latitude on the cutaneous synthesis of vitamin D3: exposure to winter sunlight in Boston and Edmonton will not promote vitamin D3 synthesis in human skin. J Clin Endocrinol Metab. 1988;67:373–378. 41. Holick MF. Environmental factors that influence the cutaneous production of vitamin D. Am J Clin Nutr. 1995;61(3 Suppl):638S–645S. 42. MacLennan WJ, Hamilton JC, Darmady JM. The effects of season and stage of pregnancy on plasma 25-hydroxy-vitamin D concentrations in pregnant women. Postgrad Med J. 1980;56:75–79. 43. Neild PJ, Syndercombe-Court D, Keatinge WR, Donaldson GC, Mattock M, Caunce M. Cold-induced increases in erythrocyte count, plasma cholesterol and plasma fibrinogen of elderly people without a comparable rise in protein C or factor X. Clin Sci (Lond). 1994;86:43–48. 44. Keatinge WR, Coleshaw SR, Cotter F, Mattock M, Murphy M, Chelliah R. Increases in platelet and red cell counts, blood viscosity, and arterial pressure during mild surface cooling: factors in mortality from coronary and cerebral thrombosis in winter. Br Med J (Clin Res Ed). 1984;289:1405–1408. 45. Elter K, Ay E, Uyar E, Kavak ZN. Exposure to low outdoor temperature in the midtrimester is associated with low birth weight. Aust N Z J Obstet Gynaecol. 2004;44:553–557. 46. Deschenes O, Greenstone M, Guryan J. Climate change and birth weight. Am Econ Rev. 2009;99:211–217. 47. Bove F, Shim Y, Zeitz P. Drinking water contaminants and adverse pregnancy outcomes: a review. Environ Health Perspect. 2002;110(Suppl. 1):61–74. 48. Maisonet M, Correa A, Misra D, Jaakkola JJ. A review of the literature on the effects of ambient air pollution on fetal growth. Environ Res. 2004;95:106–115. 49. Stillerman KP, Mattison DR, Giudice LC, Woodruff TJ. Environmental exposures and adverse pregnancy outcomes: a review of the science. Reprod Sci. 2008;15:631–650. 50. Wang L, Pinkerton KE. Air pollutant effects on fetal and early postnatal development. Birth Defects Res C Embryo Today. 2007;81:144–154.
30. Barnett AG, Dobson AJ. Analysing seasonal health data. Statistics for biology and health. Berlin London: Springer; 2010; xiii.
51. Hewitt CN, Jackson AV. Handbook of atmospheric science. Malden (MA): Blackwell Publishing; 2003.
31. Tustin K, Gross J, Hayne H. Maternal exposure to first-trimester sunshine is associated with increased birth weight in human infants. Dev Psychobiol. 2004;45:221–230.
52. Maisonet M, Bush TJ, Correa A, Jaakkola JJ. Relation between ambient air pollution and low birth weight in the Northeastern United States. Environ Health Perspect. 2001;109(Suppl. 3):351–356.
32. Valcavi R, Zini M, Maestroni GJ, Conti A, Portioli I. Melatonin stimulates growth hormone secretion through pathways other than the growth hormone-releasing hormone. Clin Endocrinol (Oxf). 1993;39:193–199.
53. Ritz B, Yu F. The effect of ambient carbon monoxide on low birth weight among children born in southern California between 1989 and 1993. Environ Health Perspect. 1999;107:17–25.