Birth month associations with height, head circumference, and limb lengths among peruvian children

Authors


  • The copyright line for this article was changed on 9 December 2015 after original online publication.

ABSTRACT

Associations between season of birth and body size, morbidity, and mortality have been widely documented, but it is unclear whether different parts of the body are differentially sensitive, and if such effects persist through childhood. This may be relevant to understanding the relationship between early life environment and body size and proportions. We investigated associations between birth month and anthropometry among rural highland (n = 162) and urban lowland (n = 184) Peruvian children aged 6 months to 8 years. Stature; head-trunk height; total limb, ulna, tibia, hand, and foot lengths; head circumference; and limb measurements relative to head-trunk height were converted to internal age-sex-specific z scores. Lowland and highland datasets were then analyzed separately for birth month trends using cosinor analysis, as urban conditions likely provide a more consistent environment compared with anticipated seasonal variation in the rural highlands. Among highland children birth month associations were significant most strongly for tibia length, followed by total lower limb length and stature, with a peak among November births. Results were not significant for other measurements or among lowland children. The results suggest a prenatal or early postnatal environmental effect on growth that is more marked in limb lengths than trunk length or head size, and persists across the age range studied. We suggest that the results may reflect seasonal variation in maternal nutrition in the rural highlands, but other hypotheses such as variation in maternal vitamin D levels cannot be excluded. Am J Phys Anthropol 154:115–124, 2014. © 2014 The Authors. American Journal of Physical Anthropology Published by Wiley Periodicals, Inc.

Associations have been widely reported between season of birth and a range of characteristics, including birth weight and length (Wohlfahrt et al., 1998; McGrath et al., 2005b, 2007; Torche and Corvalan, 2010; Krenz-Niedbała et al., 2011) and body size in childhood (Kościński et al., 2004; Puch et al., 2008) and adulthood (Weber et al., 1998; Waldie et al., 2000). Season of birth associations with phenotype and health in developing countries are thought to reflect seasonal variation in diet, activity, and disease (Roberts et al., 1982; Adair and Pollitt, 1983; Moore et al., 1997; Rayco-Solon et al., 2005; Chodick et al., 2009). For example, the incidence of low birth weight is double for babies born during the hungry season (July–October) in rural Gambia, which coincides with seasonal rains, increased malaria and diarrheal disease, a greater agricultural workload and poorer nutrition (Moore et al., 1997; Rayco-Solon et al., 2005). Similar patterns are documented in Bangladesh (Shaheen et al., 2006), Taiwan (Adair and Pollitt, 1983), Tanzania (Kinabo, 1993), and Zaire (Fallis and Hilditch, 1989).

While determinants of season of birth associations with phenotype in developed countries may differ from those in developing countries, they may also offer insight into processes which may influence these associations in addition to disease load and/or nutrition. Previous studies have demonstrated that the stature of Austrian 18-year-old males (Weber et al., 1998), birth weight and neonatal limb lengths in Australia (McGrath et al., 2005a, 2005b), Danish birth weight and length (Wohlfahrt et al., 1998), and birth weight in Northern Ireland (Murray et al., 2000), and north Chile (Torche and Corvalan, 2010) are highest among those born in late winter–spring, and lowest in autumn-early winter births.

Some variation exists in the seasonal pattern of growth in developed countries (Banegas et al., 2001; Chodick et al., 2007). For example, an October–April peak and May–September nadir was reported for child height in north Poland (Kościński et al., 2004; Krenz-Niedbała et al., 2011), while in Dunedin, New Zealand, birth weight and length peaked in October (spring) births and were lowest in January (summer) births (Waldie et al., 2000). Furthermore, dual peaks and nadirs have been identified in Japan (Matsuda et al., 1995, 1993) and southern Chile (Torche and Corvalan, 2010), although this may reflect in part concomitant seasonal variation in gestation length (Matsuda et al., 1995, 1993). Proposed explanations for seasonal variation in birth weight and associations between body size and birth month in developed countries include ambient temperature in the first or second trimester (Murray et al., 2000; Lawlor et al., 2005; Chodick et al., 2007), vitamin D levels during the first trimester in relation to UV exposure (McGrath et al., 2005b; Torche and Corvalan, 2010; Krenz-Niedbała et al., 2011), or day length effects on maternal hormone profiles (Weber et al., 1998; Waldie et al., 2000), but the causes are unclear and multiple factors that vary geographically may be involved (Matsuda et al., 1995; Chodick et al., 2009).

Associations between birth month and body size persist into adulthood, suggesting that observed differences have their origin in prenatal or early postnatal life when growth trajectories are particularly sensitive to the environment (Smith et al., 1976; Lucas, 1991; Martorell et al., 1994; Schroeder et al., 1995; Mei et al., 2004; Dewey and Adu-Afarwuah, 2008; Stein et al., 2010). While patterns may be expected to persist through childhood, studies investigating child height or body mass in relation to birth season provide mixed results in terms whether any patterns are observed, at what ages they are found, and when in the year the peaks and nadirs fall (Shephard et al., 1979; Henneberg and Louw, 1990, 1993; Waldie et al., 2000; Kościński et al., 2004; McGrath et al., 2005b; Tanaka et al., 2007; Puch et al., 2008). This may in part relate to the fact that environmental factors like sunlight exposure or day length have separate effects on growth in childhood which may mask or modify patterns arising form birth season (Waldie et al., 2000). Hence it remains unclear whether early seasonal influences on growth are carried through childhood and into adulthood and how this varies between populations.

Whether birth month is associated with the size of different parts of the body, such as the limbs and trunk, has not been well investigated. The lengths of the limbs and especially their distal segments (ulna and radius, tibia and fibula) appear particularly susceptible to environmental stresses in early life compared with trunk and especially head size. Tibia length (or knee height), in particular, has been proposed to be a sensitive indicator of early life conditions (Leitch, 1951; Meadows Jantz and Jantz, 1999; Bogin and Varela-Silva, 2010; Pomeroy et al., 2012). For example, exposure and/or susceptibility to high-altitude hypoxia is associated with reduced relative lower limb, and especially tibia length (Bailey and Hu, 2002; Bailey et al., 2007; Pomeroy et al., 2013). This may be a result of thrifty phenotype-style trade-offs among different parts of the body (Pomeroy et al., 2012) or the fact that the limbs grow relatively faster than the trunk or head postnatally, especially in infancy (Dangour et al., 2002; Wadsworth et al., 2002; Fredriks et al., 2005). To our knowledge McGrath et al. (2005b) is the only study to consider the question of whether body proportions are influenced by birth month, and reported a peak in limb lengths among those born in October and a nadir in April in an Australian sample, as for birth weight, while trunk length and head circumference appear less affected.

Relative limb lengths correlate with both early life environment and adult chronic disease risk (Gunnell et al., 1998a; Davey Smith et al., 2001; Wadsworth et al., 2002; Samaras, 2007; Whitley et al., 2008; Bogin and Varela-Silva, 2010; Pomeroy et al., 2012). Therefore if season of birth effects vary among different parts of the body in the same manner as relative limb proportions correlate with other environmental stressors, this could indicate that birth month might be a useful model for investigating the basis of the Developmental Origins of Health and Disease (DOHaD) hypothesis, since it is a stress that is essentially random in its effects on individuals.

The purpose of our study was to investigate whether birth month associations with anthropometry exist among Peruvian lowland urban and highland rural children, and whether these associations vary between different body measurements. We analyzed height; limb, limb segment and trunk lengths; and head circumference for associations with birth month, among rural highland and urban lowland Peruvian children aged 6 months to 8 years. As children from developing countries may experience much greater seasonal stress in early life than children from wealthier countries, it may be easier to identify the impacts of early life environmental stress exposure on anthropometry in the former. We also investigated whether the relative strength of associations demonstrate the hierarchical ordering of environmental sensitivity previously documented, whereby head circumference is least associated with environmental stress while trunk length, autopod (hand or foot) lengths, stature, total limb lengths and zeugopod (ulna or tibia) lengths, show a gradient of increasing environmental sensitivity (Pomeroy et al., 2012).

MATERIALS AND METHODS

The study focuses on children aged 6 months to 8 years from rural agropastoral communities in Vinchos and Santillana Districts, Ayacucho Region, Peru, located at between 3100 and 4400 m altitude (Supporting Information Fig. S1), and urban lowland children living in the unplanned but well-established peri-urban community (shanty-town) of Pampas de San Juan de Miraflores in Lima (Miranda et al., 2009, 2011).

The highland population lives principally from subsistence agriculture, and is expected to experience seasonal fluctuations in dietary quality and adequacy related to the agricultural cycle. While data are unavailable for the specific populations we studied, detailed analyses of diet and activity in an agropastoral population at 4000 m altitude in Nuñoa, southern Peru (Supporting Information Fig. S1) conducted from 1982 to 1983 demonstrated seasonal food shortages prior to the May-July harvest (i.e. in January–May: Supporting Information Fig. S2), when adults of both sexes experienced a negative energy balance, estimated at 75% of their energy requirement. Once the harvest began more food was available and the average adult had positive energy balance (June–August, mean 106% of estimated requirements) (Leonard and Thomas, 1989; Leonard, 1991). In Quinua (3300 m altitude, Supporting Information Fig. S1), another agropastoral community in Ayacucho Region, food availability followed a similar seasonal pattern to that in Nuñoa. Food was least available in March–April just before the harvest in late April–June, followed by a post-harvest peak (Mitchell, 1978, 1991). It is important to note that these data from Nuñoa and Quinua were collected several decades ago, and so may not directly parallel the current situation where commercial foods are more widely available in rural areas. Nonetheless, agricultural production is still a major source of subsistence for populations like the highland one studied here so seasonal food shortages may still affect food availability.

In contrast, while of generally low socioeconomic status, the lowland population is not expected to be subject to seasonal fluctuations in food availability since supplies are likely to be much more readily accessed year-round, thanks to the urban food supply chain of Peru's capital city, Lima. Seasonality in disease exposure has been documented, with diarrhea more common in the summer months among infants in this community (Checkley et al., 2002), and respiratory infections expected to be a greater problem in the very damp, cool conditions of the winter months. However, given the easier year-round access to food supplies and healthcare among the lowland children, it is expected that overall seasonal variation in environmental stress exposure will be less than in the highlands.

The study was approved by the Institutional Ethics Committee at the Universidad Peruana Cayetano Heredia, Lima, and the Health Directorate for Ayacucho Region (Dirección Régional de Salud Ayacucho, DIRESA). The study adhered to international ethical standards (World Medical Association, 2008), and participation was voluntary. After explaining the study to the participant (where age-appropriate) and to their parents/guardians, written informed consent was obtained from a parent/legal guardian and participants aged 6 years or over gave verbal or written assent.

While children aged 6 months to 14 years participated in the study, only those aged up to 8 years were included in the analyses of birth month associations with anthropometry to avoid the potentially confounding effects of highly variable growth around puberty. Peruvian girls at living at Cerro de Pasco start menarche around 14.4 years (Gonzales et al., 1996), and virtually all children below 8.5 years were pre-pubertal in a similar low SES population in the Americas (Wilson et al., 2011) so a cut-off of 8 years should exclude the vast majority of pubertal children. Data on children <6 months were not available, and for this reason alone younger children were not included.

Height, head-trunk height, total upper and lower limb lengths, zeugopod (ulna and tibia) lengths, and autopod (hand and foot) lengths, and head circumference were measured by a single trained observer using standard methods as previously described (Pomeroy et al., 2012). To investigate season of birth associations with relative limb proportions, indices of total upper and lower limb length and ulna or tibia length relative to head-trunk height were also calculated by dividing the relevant limb measurement by head-trunk height. This was preferred to computing indices relative to stature, since lower limb and tibia lengths are components of stature, and so would influence variation in both the numerator and denominator of the ratio, while head-trunk height is independent of limb lengths. Although postnatal growth is seasonally patterned (e.g. Bogin, 1978; Pollitt and Arthur, 1989), all children were measured between July and September 2010, so any effect of season in which they were measured is unlikely.

Anthropometric data and body proportion indices were converted to age- and sex-specific z scores by fitting LMS curves to pooled highland and lowland data on children aged 6 months to 14 years using decimal age accurate to the day, based on birth and measurement dates. Date of birth was confirmed using government identity documents, birth records or school lists. To test for associations between birth month and anthropometry, we performed cosinor analysis on the anthropometry z scores. Cosinor analysis involves fitting a linear regression model of the z scores on sine and cosine terms of birth month. A significant model indicates significant seasonal variation in anthropometry z score. The MESOR is the midline value around which the cosine curve oscillates, while the amplitude is that of the fitted cosine curve (i.e. difference between MESOR and peak of the fitted curve) and the acrophase is the timing of the peak in the fitted cosine curve across the time period of interest, in this case the month of the peak (Fernández et al., 2009: see Fig. 1). Highland and lowland datasets were analyzed separately as we hypothesized greater seasonal variation in environmental stress exposure in the former population, while the sexes were pooled. Age and sex were initially included in the regression models to allow for potential variation by age and sex in the relationship between anthropometry and birth month, but were removed as their influence was not significant (P > 0.10). Similarly, interaction terms for the cosinor model and sex were not significant, so were also excluded from subsequent analyses. Analyses were performed in SPSS for Windows v. 21.0.

RESULTS

Tables 1 and 2 summarize the sample size and anthropometry z scores by birth month. The total sample included in the birth month analyses is 162 highland children (89 male, 54%) and 184 lowland children (92 male, 50%), but varies between measurements due to missing data (lowest n = 123 for highland hand length). There was no difference in the sex distribution between birth months, age groups, or age group and birth month (χ2 tests, P > 0.05). Highland children had lower mean z scores for all measurements than lowland children, as previously reported (Pomeroy et al., 2012: readers are referred to this publication for a further breakdown of anthropometric data by age for the two samples). Supporting Information Figure S3 shows that median height among highland children was near the 3rd centile of the WHO standard (WHO Multicentre Growth Reference Study Group, 2006) and reference (de Onis et al., 2007) while lowland children were close to the WHO median.

Table 1. Summary statistics for anthropometry z scores by birth month for highland Peruvian children age 6 months to 8 years
Z scoreMean z score by birth monthAverage standard deviationa
JanFebMarAprMayJunJulAugSepOctNovDec
  1. a

    Average standard deviation = mean of monthly standard deviations

Stature−1.0−0.9−0.8−0.9−0.9−1.0−1.0−0.4−0.6−0.7−0.5−0.40.69
Head-trunk height−1.0−0.4−0.6−0.8−0.7−0.7−0.90.0−0.5−0.6−0.7−0.20.74
Total upper limb length−1.0−1.0−0.9−1.0−1.0−1.1−1.0−0.6−0.8−0.7−0.5−0.70.64
Ulna length−0.9−1.1−0.9−1.1−0.8−1.0−1.0−0.5−0.9−0.8−0.5−0.70.62
Hand length−1.0−0.8−0.7−0.8−1.0−1.1−0.8−0.6−0.9−0.8−0.2−0.60.72
Total lower limb length−0.9−1.1−0.9−0.9−0.8−1.0−1.1−0.6−0.5−0.6−0.3−0.40.78
Tibia length−1.0−1.2−1.0−1.1−1.0−1.1−1.1−0.7−0.8−0.6−0.5−0.60.66
Foot length−0.9−0.9−0.7−0.9−0.8−1.0−1.2−0.5−0.7−0.5−0.4−0.60.72
Head circumference−0.4−0.9−0.5−0.3−0.7−0.4−0.5−0.3−0.5−0.2−0.9−0.20.81
n6–97–1019–2210–1312–189–1110–1511–149–1111–156–812–16 
Table 2. Summary statistics for anthropometry z scores by birth month for lowland Peruvian children aged 6 months to 8 years
Z scoreMean z score by birth monthAverage standard deviationa
JanFebMarAprMayJunJulAugSepOctNovDec
  1. a

    Average standard deviation = mean of monthly standard deviations.

Stature0.30.80.80.40.70.70.90.50.80.41.00.90.63
Head-trunk height0.10.60.60.10.50.70.70.30.50.21.30.80.83
Total upper limb length0.30.90.90.50.60.60.80.50.90.21.00.70.65
Ulna length0.41.01.00.40.60.70.70.50.60.21.00.70.67
Hand length0.30.70.70.50.50.60.60.20.70.30.80.80.77
Total lower limb length0.40.80.90.50.70.60.90.50.90.60.60.90.65
Tibia length0.41.00.90.50.80.70.90.50.80.30.80.90.60
Foot length0.30.80.80.20.50.50.90.40.60.21.00.90.74
Head circumference−0.10.50.3−0.10.30.50.30.50.30.31.40.40.92
n17–1818–1916–1814–1815–1616–171114–1615–197–1011–1210–11 

Cosinor analysis of the highland dataset (Table 3) indicates a significant birth month association with tibia length, relative tibia length, total lower limb length, stature and relative upper limb length in order of decreasing strength, but not for other anthropometric measures. The adjusted R2 values are low, indicating that birth month explains a only small proportion of the variance (highest for tibia length: R2 = 0.06). Figure 1 shows the cosinor model fitted to the highland tibia length data as an example, and illustrates the MESOR, acrophase and amplitude. For all the significant models, the acrophase (peak) occurs in late October or early November, with 95% confidence intervals of mid-late September to early-mid December (Table 3). Amplitudes of the acrophase of these models were approximately 0.25 z scores, indicating a difference of ∼0.5 z scores between the peak and nadir months. Monthly means for the different measurements among the highland sample are shown in Figures 2 and 3, and demonstrate that children born between January and June have shorter stature, total upper and lower limb lengths and tibia length. The cosinor results were not significant for any variables in the lowland sample (model P > 0.3: results not shown).

Table 3. Results of cosinor analysis of anthropometric variables by birth month among highland Peruvian children aged 6 months to 8 years
Z scoreModel PAdjusted R2 (z scores)MESORAmplitudeAcrophase (Peak month)
Value (z scores)95% CIDate95% CI
  1. Only variables for which the cosinor results are significant are given here, a summary of results for other measurements are given in Supporting Information Table 1. Variables are ordered by P value.

  2. Bold text indicates statistically significant results. Relative lengths are z scores of length/head-trunk height.

  3. CI, confidence interval.

Tibia length0.0050.06−0.880.260.11–0.411st November26th September–7th December
Relative tibia length0.0070.06−0.740.280.11–0.452nd November27th September–9th December
Total lower limb length0.020.04−0.760.250.07–0.421st November20th September–13th December
Stature0.020.03−0.730.220.061–0.3729th October16th September–10th December
Relative total upper limb length0.0450.03−0.690.230.05–0.4130th October13th September–15th December
Figure 1.

Mean values of age- and sex-specific tibia length z score by month of birth in the highland sample among highland Peruvian children aged 6 months to 8 years, showing fitted cosinor model. The MESOR, amplitude and acrophase are labeled for reference.

Figure 2.

Variation in mean age- and sex-specific anthropometry z score by month of birth among highland Peruvian children aged 6 months to 8 years. The same data points are repeated for December and January on the left and right sides of the graph (separated by dotted lines) to better illustrate the trend from December to January.

Figure 3.

Variation in age- and sex-specific z scores for limb and limb segment length relative to head-trunk height by month of birth among highland Peruvian children aged 6 months to 8 years. The same data are repeated for December and January on the left and right sides of the graph (separated by dotted lines) to better illustrate the trend from December to January.

DISCUSSION

Our results demonstrate seasonal associations between birth month and tibia length (absolute and relative to head-trunk height), total lower limb length, relative total upper limb length and stature among rural highland Peruvian children aged 6 months to 8 years, with the peak in October–November. This contrasts with the urban lowland dataset, where no anthropometry varied seasonally. Stronger birth month effects in the highland than lowland sample are consistent with our prediction that the rural highland population experience greater seasonal fluctuation in environmental stress exposure than the urban lowland population.

In the highland dataset, the strongest seasonal signal is in tibia length (absolute and relative) followed by (absolute) total lower limb length and stature, while there is no significant month of birth association with head-trunk height, autopod length or particularly head circumference. This pattern could be consistent with brain-sparing (Hales and Barker, 1992) and the hierarchy with which different parts of the body appear to be buffered from environmental stress (e.g. Lampl et al., 2003; Bailey et al., 2007; Frisancho, 2007; Bogin and Varela-Silva, 2010; Pomeroy et al., 2012), although it may reflect a prenatal cephalo-caudal growth pattern resulting in greater relative growth in limb lengths than in trunk or head size postnatally (Dangour et al., 2002; Wadsworth et al., 2002; Fredriks et al., 2005; Bogin and Varela-Silva, 2010).

While the results for the upper limb were not significant, except for relative total upper limb length, they follow a similar pattern as those for the lower limb. Comparisons between the highland and lowland children previously demonstrated a similar pattern of environmental sensitivity in the lengths of the upper and lower limbs, and their component elements (Pomeroy et al., 2012), suggesting that similar patterns in the upper and lower limb may be expected. Given the lack of a significant result for the ulna, it is unsurprising that total upper limb length is not significant since it is the sum of ulna and humerus lengths.

Our results support previous work in showing that linear body size, especially total lower limb and particularly tibia length (both absolute and relative to trunk length) are sensitive to early life environment. Several previous studies have demonstrated associations between body size and proportions and early life conditions (Leitch, 1951; Gunnell et al., 1998b; Bogin et al., 2002; Wadsworth et al., 2002; Frisancho, 2007; Li et al., 2007; Whitley et al., 2008), including differences between highland and lowland Andeans (Haas et al., 1977; Mueller et al., 1978; Palomino et al., 1978; Stinson and Frisancho, 1978; Stinson, 1990, 2009; Pomeroy et al., 2012). Tibia length is thought to be a particularly sensitive marker of the early environment (Meadows Jantz and Jantz, 1999; Lampl et al., 2003; Bailey et al., 2007; Bogin and Varela-Silva, 2010; Pomeroy et al., 2012). The relatively low R2 values in our analysis indicate that while some seasonal associations are significant, other factors also impact on anthropometry. This is, however, is to be expected as genetics, parental body size and nutrition, gestational age, birth weight and postnatal environment all contribute to this variation (e.g. Bogin, 1999; Haeffner et al., 2002; Gigante et al., 2006; Li et al., 2007; Whitley et al., 2008; Howe et al., 2012). We were unable to control for these variables in our analysis, but their influence warrants investigation in future studies.

The results are also interesting since they suggest that in rural highland Peru, environmental factors associated with season of birth may be sufficiently strong in early life to influence the development of both body size and proportions even though offspring are buffered from the impacts of the external environment by maternal physiology during gestation and breastfeeding (Wells, 2003). Moreover, our results suggest that season of birth may be associated with altered patterns of body proportions, so similar analyses could be helpful in elucidating the basis of associations between birth weight, relative limb/trunk proportions, and chronic disease risk in adulthood (see Introduction). Since birth weight and body proportions share similar associations with chronic disease risk in later life (Bogin and Baker, 2012), birth weight, body proportions and metabolic function might share critical windows for environmental sensitivity in early life, although a lack of association between birth weight and body proportions in later life may indicate that critical windows for different physical characteristics are somewhat distinct (Gunnell et al., 1999; Gunnell, 2002; Wadsworth et al., 2002; Bogin and Baker, 2012).

Understanding which environmental factors are associated with different months of the year and when they operate is challenging for several reasons, and an issue that we could not address directly. Season of birth correlates with the season at conception, the trimesters of pregnancy, and early infancy: short- or long-term seasonal growth effects could act during any of these times. Furthermore, seasonal variations in a range of environmental parameters are correlated, such as food availability, activity, morbidity, temperature, day length, and UV intensity, and can elicit complex and correlated behavioral responses (McGrath et al., 2006; Chodick et al., 2009). This makes it difficult to confidently identify causal relationships between seasonal environmental variation and phenotype.

Considering the potential influences of various seasonal environmental factors that have been proposed to explain associations between birth season and phenotype (see Introduction), temperature and UV exposure seem less likely explanations of the results found among our highland sample. Temperature is relatively stable through the year in this region due to its altitude and latitude (Thomas and Winterhalder, 1976; Jablonski and Chaplin, 2010: Supporting Information Fig. S4). There are no direct data on seasonal variation in Vitamin D status in the Andean highlands, but given that UV exposure and skin pigmentation are major determinants of vitamin D status (Jablonski and Chaplin, 2010; Holick, 2011) and as most rural people work outside, travel extensively by foot, have moderately pigmented skin, and regional UV levels are high (Jablonski and Chaplin, 2010), it seems unlikely that variation in Vitamin D levels in relation to seasonal fluctuations in UV exposure would explain the results.

Seasonal variation in diet and/or morbidity appear to be more plausible explanations for our results. Although we lack dietary data, studies of similar rural highland communities already outlined indicate seasonality in nutrition is likely. It is well established that the effects of disease on nutrient intake, absorption and utilization in the immune response can have major impacts on the resources available for somatic growth (Scrimshaw and SanGiovanni, 1997; Lunn, 2000). Furthermore, various parasitic and infectious diseases show seasonal variation in their prevalence in other regions, including malaria, gastrointestinal and respiratory infections (Rowland et al., 1988; Lunn, 2000; Hadi et al., 2004; Rayco-Solon et al., 2005) and could act through maternal exposure during pregnancy (Rayco-Solon et al., 2005), or postnatal exposure in the offspring (Lunn, 2000). Unfortunately there are insufficient data to fully evaluate the potential role of seasonal variation in disease in our sample. Malaria is not present at this altitude so can be excluded. Respiratory infections appear to be more common than gastrointestinal infections in the Andean highlands (Way, 1976; Niermeyer et al., 2009) and are exacerbated by hypoxia and air pollution from indoor cooking (Niermeyer et al., 2009). However, seasonal variation in morbidity in highland Andean communities is not well documented so we are unable to evaluate fully whether these factors might explain our results.

The differences between populations we observed might be due to a number of factors, including altitude, temperature or rural/urban status, and we are not able to disentangle these possible effects. Previous studies suggest that rural populations grow more slowly and/or achieve shorter adult height regardless of altitude (Frisancho et al., 1975; Leonard et al., 1990). This may reflect the more reliable food supply in urban compared with rural environments, but based on current evidence we cannot speculate as to whether lowland rural children may show similar seasonal patterns in body and limb sizes as our rural highland sample.

It is not currently possible to discern whether this seasonal influence is prenatal (and if so at which stage of pregnancy) or postnatal. As childhood and adulthood lower limb and trunk lengths show similar associations with birth weight (Gunnell et al., 1999; Wadsworth et al., 2002) and birth weight appears unrelated to relative lower limb length in childhood (Bogin and Baker, 2012), it has been argued that relative lower limb length indexes postnatal environment. As these data derive from high-income countries, the extent to which they represent patterns in low- and middle-income countries is unknown. Growth in the first 2 years of life may be particularly influential in setting subsequent growth trajectories (Martorell et al., 1994; Schroeder et al., 1995; Mei et al., 2004; Dewey and Adu-Afarwuah, 2008; Stein et al., 2010) and so it is plausible that early postnatal influences may have long term consequences for body size and proportions. However, while limb proportions certainly appear sensitive to postnatal conditions, some evidence suggests that they also are sensitive to prenatal growth stress. Some studies have shown an association between neonatal limb size and proportions and exposure to maternal smoking or diabetes (Lindsay et al., 1997; Catalano et al., 2003; Lampl et al., 2003; Lampl and Jeanty, 2004). Thus relative limb lengths may reflect both pre- and postnatal environment, and the contributions of pre- and postnatal stressors in determining relative limb proportions is currently unclear.

Supporting a degree of prenatal sensitivity of limb lengths to environmental conditions, McGrath et al. (2005b) demonstrated a seasonal pattern in limb and limb segment lengths among Australian newborns that appears to be independent of the pattern in trunk length. Interestingly, they also found a peak in southern hemisphere spring (October), similar the present study and potentially indicative of a similar underlying mechanism. They argued that maternal vitamin D status in early pregnancy could be responsible for this pattern, but a clear link between early pregnancy vitamin D status and limb proportions at birth remains to be demonstrated. Furthermore, without data on annual variation in maternal vitamin D status in our highland and lowland populations, this explanation cannot be excluded at present.

A limitation of our study is that we lacked data on birth weight and length of gestation. Rates of prematurity and birth weight both demonstrate seasonality (Rayco-Solon et al., 2005; Lee et al., 2006), and may influence subsequent growth. Furthermore, sex differences in season of birth associations with phenotype have been reported in some studies (e.g. Murray et al., 2000; Krenz-Niedbała et al., 2011), but our sample size was inadequate for separate analyses by sex.

To conclude, we have demonstrated associations between birth month and tibia length, total lower limb length, stature and relative upper limb length among children aged 6 months to 8 years from rural highland Peru, but not in an urban lowland sample. The results suggest that limb lengths, especially tibia length, are sensitive to season of birth. The seasonally patterned stress exposure responsible for this pattern among highland children could reflect the results in brain-sparing growth and growth trade-offs among different parts of the body, or the relatively faster growth rate of the limbs compared with the trunk and head in infancy and childhood. The results suggest that season of birth associations with relative limb proportions could be a useful route for investigating the basis of the DOHaD hypothesis and factors influencing variation in body proportions.

ACKNOWLEDGMENTS

Thank you to all the participants in the study, and their parents, for generously giving their time to take part. Thanks to the field staff and especially to Angela Huamán Gomez and Lilia Cabrera of PRISMA for their help with data collection. Climate data in Supporting Information Figure S4 were obtained from the NASA Langley Research Center Atmospheric Science Data Center Surface meteorological and Solar Energy (SSE) web portal supported by the NASA LaRC POWER Project (https://eosweb.larc.nasa.gov/sse/, accessed 25/07/2013). The authors have no conflicts of interest to declare. We thank the associate editor and two anonymous reviewers for their insightful and constructive comments that helped improve the article.

Ancillary