Evaluating the seasonality of growth in infants using a mobile phone application

The study investigates whether infant growth velocity varies by season, a phenomenon previously documented in toddlers and school-aged children but not well characterized in infants. Grounded in the infancy–childhood–puberty (ICP) model, infancy is marked by rapid growth influenced primarily by nutrition, with known sex differences in linear growth. Prior reports from low-income settings suggested seasonality in infant growth driven by food availability, but there are no reports from developed countries, potentially due to challenges in collecting dense, prospective growth data. Leveraging mobile app-based data from a widely used childcare application, the authors aim to determine if climatic seasons influence length and weight gain in infants in Japan.

Background theories include the ICP model that delineates distinct growth phases and emphasizes rapid infancy growth. WHO standards show sex differences in length gain in the first year. Studies in low-income countries (Ethiopia, Timor-Leste) reported seasonal growth patterns likely linked to food availability. Among older children, small-scale studies have shown enhanced growth from spring to summer, though mechanisms remain unclear. Potential biological determinants considered in prior literature include growth hormone and vitamin D status, with GH being less influential in infancy and seasonal variation in vitamin D well documented. A gap exists regarding seasonality of infant growth in developed countries due to data collection limitations, motivating this app-based approach.

Design and data source: Secondary analysis of de-identified longitudinal data entered by caregivers using the free Papatto Ikuji mobile application (Android/iOS) in Japan, which tracks infant care and measurements. The app had substantial national uptake (e.g., 73,496 installs in 2016, ~7.5% of newborns). Ethics approval was obtained with opt-out; data were anonymized. Population and inclusion criteria: Children aged 0–400 days with at least two measurements (length and/or weight) recorded 15–100 days apart between January 2014 and October 2017. The lower bound reduces influence of measurement error for closely spaced measures; the upper bound approximates a season length. After applying criteria and z-score range exclusions (LAZ/WAZ outside −3.0 to +3.0 removed), 9,409 infants contributed to analyses (5,870 infants with ≥1 ΔLAZ/day; 9,165 with ≥1 ΔWAZ/day). Across all infants, there were 20,007 ΔLAZ/day and 33,236 ΔWAZ/day observations. Median number of measurements per child was 4 (IQR 2–7) for length and 4 (IQR 2–8) for weight; averages were 4.8 (length) and 5.4 (weight). Variables and processing: For each pair of consecutive measurements, length-for-age and weight-for-age z-scores (LAZ, WAZ) were computed using Japanese age- and sex-specific standards (Japan Society for Pediatric Endocrinology). Daily change in z-score was computed at the reference date (midpoint between two measurement dates): ΔLAZ/day = (LAZ_T2 − LAZ_T1)/(T2 − T1); ΔWAZ/day = (WAZ_T2 − WAZ_T1)/(T2 − T1). Seasons were assigned by Japan Meteorological Agency definitions based on the reference date: winter (Dec–Feb), spring (Mar–May), summer (Jun–Aug), autumn (Sep–Nov). Nutritional exposure was categorized via a breastfeeding index = breastfeeding events/(breastfeeding + formula events): breastfeeding (≥0.90), mixed (0.50–0.89), formula-dominant (<0.50). Statistical analysis: Descriptive statistics summarized cohort characteristics. ΔLAZ/day and ΔWAZ/day were aggregated by year-season and by month (pooled across years) with means and 95% CIs. To assess seasonal effects on linear growth controlling for repeated measures, a multilevel linear regression was performed allowing only one ΔLAZ/day per person per season (averaged if multiple within a season). Covariates included age at reference date, sex, nutritional group, season of reference date, season of birth, and a time-sequence variable (Period 1–4) representing sequential seasons from birth to measurement; the time sequence was modeled as a random effect. Models were fitted via maximum likelihood using SAS 9.4.

• Sample: 9,409 infants aged 0–400 days; 20,007 ΔLAZ/day observations (mean −0.0022) and 33,236 ΔWAZ/day (mean 0.0005). Means and SDs reported in the cohort were ΔLAZ/day −0.002 ± 0.022 and ΔWAZ/day 0.001 ± 0.014.
• Seasonality: Visual aggregation showed ΔLAZ/day consistently increased in summer and decreased in winter across years; ΔWAZ/day showed no seasonal pattern.
• Multilevel model (adjusted for age, sex, nutrition, season of birth, period): Summer ΔLAZ/day was 0.0026 higher than winter (95% CI 0.0015–0.0036; P < 0.001). Spring and autumn were not significantly different from winter.
• Clinical magnitude: The summer–winter difference corresponds to ~13% of average linear growth at 6 months (0.18 cm difference given ~1.3 cm/month growth).
• Nutrition: Formula-dominant feeding associated with higher ΔLAZ/day than breastfeeding by 0.0015 (95% CI 0.0006–0.0024; P = 0.001). Mixed feeding was not significantly different from breastfeeding.
• Period effects: The period (time sequence) variable showed associations, potentially reflecting differences between the app user population and the 2000 reference used for growth standards.
• Weight gain: No evidence of seasonality in ΔWAZ/day.

The findings demonstrate that infant linear growth exhibits significant seasonal variation in a developed country context, with higher velocity in summer and lower in winter, addressing the primary question about seasonality in infancy. This parallels observations in older children and suggests a consistent biological or environmental influence across pediatric ages. Potential mechanisms were considered: food availability is unlikely given the absence of seasonal variation in weight gain; growth hormone-related mechanisms are less plausible because infancy growth is largely GH-independent and seasonality is observed even in GH-deficient children; variation in vitamin D status is a plausible contributor due to known seasonal fluctuations and links to growth, although vitamin D data were not available for confirmation. While the absolute magnitude of seasonal variation may be modest and not directly practice-changing, identifying these patterns has public health relevance for monitoring potential seasonal deficiencies (e.g., vitamin D) at the population level. The large, app-based dataset enhances detection of subtle effects that might be missed in routine clinical measurements, and the representativeness of sex and breastfeeding distributions supports generalizability within Japan.

Using nationwide mobile app-based data, the study reveals that infant linear growth in Japan is seasonally patterned, increasing in summer and decreasing in winter, while weight gain does not show seasonality. The effect size is modest but statistically robust and corresponds to a meaningful fraction of typical monthly growth at 6 months. These results highlight a previously unrecognized seasonal influence on infant growth in a developed country and motivate further research into underlying mechanisms, particularly the role of vitamin D and other environmental factors, and evaluation across diverse geographic regions.

• No biochemical data (e.g., growth hormone, IGF-1, vitamin D) to investigate mechanisms underlying seasonality.
• Reliance on caregiver-entered measurements without standardized protocols across users, introducing potential non-differential measurement error and attenuation bias.
• Lack of residential location data prevented assessment of latitude effects; however, most of the Japanese population resides between latitudes 33–37°.
• Weight-for-height (ponderal growth) was not analyzed due to lack of appropriate age-specific standardized references; WAZ is a composite of length and weight.
• Minor discrepancies in counts across stages reflect exclusion of extreme z-scores (|z| > 3) and data cleaning.