Associations between meteorological factors and pregnancy complications during different pregnancy trimesters: a multicenter retrospective study in eastern China

Study design and participants

This study is based on a multicenter retrospective cross-sectional study conducted in Ningbo, China (Zhu et al., 2024a; Zhu et al., 2024b). Ningbo, located on the eastern coast of China, is one of the most economically developed cities, with a resident population of 9.697 million in 2023. Ningbo is situated in a subtropical monsoon climate zone characterized by distinct seasons. The average annual temperature is approximately 16.4 °C, the average annual sunshine duration is 1,850 h, and the average annual precipitation is 1,480 millimeters.

This study utilized clinical data from obstetric inpatients at four cooperative medical centers in Ningbo between January 1, 2013, and December 31, 2023. Under a collaborative framework approved by institutional review boards (ethical approval no. SX201916), standardized electronic medical records were consolidated through structured data abstraction protocols to establish a retrospective multicenter database for systematic analysis.

The inclusion criteria for all the subjects were as follows: aged 20–49 years, singleton pregnancies, natural conception, gestational week of at least 28 weeks, availability of diagnostic information for mothers, continuous residence during pregnancy, and Han Chinese ethnicity. Participants with pre-existing diagnoses of diabetes, cardiovascular disease, renal disorders, severe hepatic conditions, or cancer prior to pregnancy were excluded from the study. Pregnant individuals who experienced spontaneous or induced abortions during pregnancy or lacked documented last menstrual period (LMP) data before conception were also excluded. A total of 92,332 pregnant women were ultimately included in the analysis.

In our study, the minimum required sample size was calculated using the formula as following: ${\displaystyle n=\left({\left({Z_{\alpha }}_{\mathrm{/2}}\right)}^2\cdot p\cdot \left(1-p\right)\right)/E^2}$ where: Z_α/2 = 1.96: critical value for a two-tailed 95% confidence level (α = 0.05); p = 0.1976: anticipated prevalence of GDM which derived from prior epidemiological study (Qian et al., 2023); E = 0.02: allowable margin of error (±2%).

This yielded a minimum requirement of 1,526 participants to ensure the estimated prevalence would fall within ± 2% of the true population value. Our study ultimately enrolled 92,332 participants, exceeding the minimum requirement by over 60-fold. This substantial sample size could minimize sampling error and enhance generalizability of findings to the target population.

Data collection encompassed demographic characteristics, medical records, and delivery details, systematically abstracted from hospital electronic medical records systems with de-identification of sensitive fields. Gestational weeks were determined by LMP date and ultrasound examination. Conception seasons (spring: March–May, summer: June–August, autumn: September–November and winter: December–February) were calculated based on LMP.

In accordance with the information security technology guide for health data security of China, sensitive personal information of all participants would undergo anonymization and structural modification through standardized encryption protocols to ensure irreversible privacy protection prior to any analytical processing occurs. This study was approved by the Ningbo University Medical Science Research Ethics Committee, as well as ethical consent from all collaborative medical centers (ethical application ref: SX201916). Furthermore, written informed consent for participation was not required due to the retrospective nature of this study, in accordance with Chinese national legislation and institutional requirements.

Criteria for pregnancy complications

In this study, four common pregnancy complications were included in the analysis as outcomes: GDM, GH, PE, and hypothyroidism.

The diagnosis of GDM was performed using the standardized “one-step” 75-g oral glucose tolerance test (OGTT) according to the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) criteria (McIntyre et al., 2016), with reference to the American Diabetes Association (ADA) diagnostic guidelines (American Diabetes Association, 2012). As previously described, the diagnostic procedure required: (1) administration of a 75-g glucose load after ≥ 8-hour overnight fasting at 24–28 gestational weeks; (2) glucose measurement at fasting, 1-hour and 2-hour intervals using standardized laboratory procedures. Diagnosis was confirmed when meeting any IADPSG/ADA threshold: fasting ≥ 5.1 mmol/L (92 mg/dL), 1-hour ≥ 10.0 mmol/L (180 mg/dL), or 2-hour ≥ 8.5 mmol/L (153 mg/dL).

GH and PE were diagnosed based on the 2021 guidelines issued by the International Society for the Study of Hypertension in Pregnancy (ISSHP) (Magee et al., 2022). GH is new-onset hypertension at ≥20 weeks of gestation in the absence of proteinuria or other findings suggestive of preeclampsia. PE is gestational hypertension with ≥1 new onset of organ or uteroplacental dysfunction: proteinuria, other maternal end-organ dysfunction, or uteroplacental dysfunction.

The diagnostic criteria for hypothyroidism during pregnancy were established through adherence to the 2017 American Thyroid Association (ATA) Guidelines for Thyroid Disease Management in Pregnancy and Postpartum (Alexander et al., 2017). The diagnosis of hypothyroidism during pregnancy is based on the serum levels of thyroid-stimulating hormone (TSH) and free thyroxine (FT4). Clinical hypothyroidism (CH) is defined as a TSH level higher than the upper limit of the specific reference range (or 4.0 mU/L in the first trimester) and an FT4 level less than the lower limit of the specific reference range during pregnancy. Subclinical hypothyroidism (SCH) is defined as a TSH level higher than the upper limit of the specific reference range (or 4.0 mU/L in the first trimester) and an FT4 level in the normal range during pregnancy.

Meteorological factor data

We collected the daily mean temperature (T_mean), relative humidity (RH), surface pressure, wind speed, precipitation, sunshine duration, maximum temperature (T_max), minimum temperature (T_min) and DTR in Ningbo from January 1, 2013, to December 31, 2023, to evaluate the relationships between meteorological factors and these four pregnancy complications. T_mean, RH, surface pressure, wind speed, precipitation, sunshine duration, T_max, and T_min records were obtained from the National Aeronautics and Space Administration (NASA) Goddard Earth Sciences (GES) Data and Information Services Center (DISC) Global Land Data Assimilation System (GLDAS) (Rodell et al., 2004; Beaudoing &.Rodell, 2020). The DTR is calculated as the difference between T_max and T_min.

We averaged these variables for the first day of each pregnancy week and the subsequent six days to assess the average individual exposure levels of meteorological factors per week of pregnancy for each participant. We calculated the average individual exposure levels of meteorological factors for the following three periods: the first trimester (1–13 weeks), the second trimester (14–27 weeks), and the first two trimesters combined (1–27 weeks).

Definition of the MFS

In this study, we calculated the MFS according to these nine meteorological factors (Wang et al., 2021). Initially, individual exposure values for the nine meteorological factors across all participants were normalized using min–max standardization. The equation: ${\displaystyle X_{new}=\left(X-X_{min}\right)/\left(X_{max}-X_{min}\right).}$ We subsequently summed the exposure values of the nine meteorological factors and weighted them according to adjusted multivariate estimates of high-frequency risk (β coefficients) to derive the weighted MFS. The equation was as follows: ${\displaystyle MFS=\left({{\beta }_{\mathrm{T}}}_{\mathrm{mean}}\times T_{mean}+{\mathrm{\beta }}_{\mathrm{RH}}\times RH+{\mathrm{\beta }}_{\mathrm{surface pressure}}\times surfacepressure+{\mathrm{\beta }}_{\mathrm{wind speed}}\times windspeed+{\mathrm{\beta }}_{\mathrm{precipitation}}\times precipitation+{\mathrm{\beta }}_{\mathrm{sunshine duration}}\times sunshineduration+{{\mathrm{\beta }}_{\mathrm{T}}}_{\max }\times T_{max}+{{\mathrm{\beta }}_{\mathrm{T}}}_{\min }\times T_{min}+{\mathrm{\beta }}_{\mathrm{DTR}}\times DTR\right)\times \left(9/\mathrm{sum of the}\mathrm{\beta }\mathrm{coefficients}\right).}$ A higher MFS indicates a greater degree of exposure to environmental meteorological factors. In addition, the participants were divided into five groups according to the quintiles of the MFS.

Statistical analysis

We used descriptive statistics to summarize the demographic characteristics of the study population and the exposure to meteorological factors during the study period. Continuous variables are presented as the means and standard deviations (means ± SDs), and categorical variables are presented as frequencies and percentages. The Pearson linear correlation test was employed to examine the relationships between the various meteorological factors. We adjusted all the models for the same confounding factors, including maternal age, parity (1, 2, ≥3), gravidity (1, ≥2), season of conception (spring, summer, autumn, or winter), and year of conception. The results for all the models are presented as odds ratios (ORs) with 95% confidence intervals (CIs).

We assessed the impact of each individual meteorological factor and MFS on pregnancy complications. Because GDM, GH and PE were mainly diagnosed in the second trimester in the study population, we used a logistic regression model to estimate the associations between each meteorological factor and MFS on the risk of GDM, GH and PE during the first trimester, the second trimester and the first two trimesters. As hypothyroidism is diagnosed mainly in the first trimester, we used a logistic regression model to estimate the associations between each meteorological factor and MFS with hypothyroidism during the first trimester.

We then assessed the effects of sensitive windows of exposure to extreme meteorological factors on the risk of pregnancy complications. Distributed lag nonlinear models (DLNMs) incorporating logistic regression were applied to estimate the exposure−lag−response effects of exposure to extreme meteorological factors on the risk of pregnancy complications (Gasparrini, 2014). The 95th, 97th and 99th percentiles of each meteorological variable were defined as extremely high values, whereas the 5th, 3rd and 1st percentiles were defined as extremely low values, with the median of each meteorological variable serving as the reference value (Zhang et al., 2021). Considering that GDM is diagnosed between weeks 24 and 28 of pregnancy and that GH and PE are diagnosed after week 20 of pregnancy (McIntyre et al., 2016; Magee et al., 2022), the lag time was set from 1–24 gestational weeks. Hypothyroidism is diagnosed early in pregnancy (Alexander et al., 2017), so the lag time range was set from 1–13 gestational weeks. Given the potential nonlinear effects of meteorological factors, we employed natural cubic spline functions to model the delayed response relationship between exposure at each gestational week and the outcome (Wang et al., 2020). The optimal degrees of freedom were determined based on the minimum Akaike information criterion (AIC) (Gasparrini, 2014).

Finally, we evaluated the effects of extreme meteorological factors during different stages of pregnancy on the risk of pregnancy complications. Meteorological factors were categorized into three groups. The 95th, 97th, and 99th percentiles of each meteorological variable were defined as extremely high values, whereas the 5th, 3rd, and 1st percentiles were defined as extremely low values. The ranges from the 5th–95th percentiles, 3rd–97th percentiles, and 1st–99th percentiles were designated as reference values. Logistic regression models were used to assess whether extreme meteorological factors, compared with reference values, influenced the occurrence of pregnancy complications. For the interaction analysis, we used the relative risk owing to interaction (RERI), proportion attributable (AP) and synergy index (SI) to estimate the additive interactive effects of extreme meteorological factors on the risk of pregnancy complications in different trimesters (Lou et al., 2018; Zhang et al., 2021). Each statistically significant extreme meteorological factor was explored for its two-by-two additive interaction. If there is an additive interaction between the two factors, the RERI with 95% CI and AP with 95% CI should not include 0, whereas the SI with 95% CI should not include 1. RERI and AP > 0 and SI > 1 indicate positive interactions, whereas RERI and AP < 0 and SI > 1 indicate negative interactions.

All the statistical analyses were conducted using R software (version 4.2.3). A two-sided test with P < 0.05 was considered statistically significant. Each reported odds ratio in this study was the adjusted OR (aOR).

Basic characteristics of the participants and exposure to meteorological factors during pregnancy

A total of 92,332 participants were included in the study, with diagnoses as follows: gestational diabetes mellitus (GDM) in 17,814 participants (19.29%), gestational hypertension (GH) in 3,860 participants (4.18%), preeclampsia (PE) in 3,101 participants (3.36%), and hypothyroidism in 17,418 participants (18.86%) (Table 1). Overall, the mean (SD) maternal age of the participants was 30.02 (4.57) years, and 49,248 (53.34%) were primiparous. In general, advanced maternal age, increased gravidity and spring pregnancy increased the risk of GDM, GH and PE (all P < 0.001). Primiparous individuals had increased risks of GH, PE and hypothyroidism (all P < 0.01). Residents had increased risks of GDM, whereas immigrants had increased risks of GH and PE (all P < 0.001) (Tables S1–S4).

Table S5 presents the average exposure levels of meteorological factors during different trimesters in patients with GDM, GH, PE and hypothyroidism and in healthy pregnant women. During the first trimester, patients with GDM, GH, PE and hypothyroidism had higher exposure levels of T_mean, RH, precipitation, T_max and T_min and lower exposure levels of surface pressure (all P < 0.001). During the second trimester, patients with GDM, GH and PE exhibited significantly elevated exposure levels for T_mean, RH, T_max, and T_min (all P < 0.02).

The correlations among the meteorological factors are shown in Fig. S1. Table S6 summarizes the trimester-specific distributions of meteorological factors across all participants. Weekly distributions of meteorological factors during the first 24 and 13 gestational weeks are detailed in Tables S7 and S8, respectively.

Associations between meteorological factors and pregnancy complications

Table 2 summarizes trimester-specific associations between meteorological factors and pregnancy complications after adjusting for confounders. During the first trimester, higher temperatures (T_mean, T_max, T_min) consistently increased risks of GDM, GH, PE, and hypothyroidism. Increased precipitation was linked to higher GH and hypothyroidism risks, while prolonged sunshine duration raised the odds of GDM and GH. Conversely, rising surface pressure reduced risks of all complications. Higher RH increased GDM and PE risks but decreased hypothyroidism likelihood, whereas wind speed reduced PE risk but heightened hypothyroidism risk. In the second trimester, elevated temperatures continued to amplify GDM and GH risks. Higher RH remained positively associated with GDM and PE, and increased precipitation was linked to higher GDM risk, while increased surface pressure and DTR reduced GH and PE risks, respectively. All reported associations were statistically significant (P < 0.05), with detailed aORs and 95% CIs shown in Table 3.

Analysis between MFS and pregnancy complications revealed dose–response relationships. Participants in the highest MFS quintile during the first trimester had greater odds of GDM (aOR = 1.186, 95% CI [1.079–1.304]), GH (aOR = 1.596, 95% CI [1.323–1.925]), PE (aOR = 1.347, 95% CI [1.094–1.658]) and hypothyroidism (aOR = 1.257, 95% CI [1.141–1.385]) compared to the lowest quintile. In the second trimester, high MFS quintile exposure increased GDM and GH risks but decreased PE likelihood. Cumulative exposure across the first two trimesters further amplified GDM, GH, and PE risks (Table 3). All trend tests reached statistical significance (P < 0.05), with consistent patterns observed per 10-point MFS increment (Table S9). These findings indicated that cumulative exposure to adverse meteorological conditions during the first and second trimesters elevated complication risks in a dose-dependent manner.

Effects of extreme meteorological factors on risks of pregnancy complications

Tables S12–S15 demonstrate trimester-specific associations between extreme meteorological factors and pregnancy complications risks. Figures 1 and 2 identifies susceptibility windows for extremely high meteorological exposures on risks of GDM, GH and PE during the first 24 gestational weeks and hypothyroidism during the first 13 gestational weeks. Compared with the median meteorological exposure, extremely high precipitation and sunshine duration elevated GDM risks during 13th–20th and 1st–8th gestational weeks, respectively. For GH, elevated risks clustered between during 15th–23rd weeks with exposure to extremely high T_mean, RH, wind speed, precipitation, sunshine, and T_max. PE risks increased during early (1st–10th weeks) and late (18th–24th weeks) gestation under extremely high RH, surface pressure, precipitation, sunshine, and T_max. Hypothyroidism showed heightened sensitivity to multiple extremely high factors (T_mean, RH, surface pressure, wind speed, precipitation, T_max, T_min, DTR) predominantly in the first trimester (1st–13th weeks).

Figures 3 and 4 reveals susceptibility windows for extremely low meteorological exposures. Compared with the median meteorological factors, extremely low DTR increased GDM risk during the 1st gestational week. GH risks rose during 9th–19th weeks with extremely low surface pressure, wind speed, and T_min. PE risks were elevated during 10th–21st weeks under low precipitation, T_max, and DTR. Hypothyroidism exhibited broad sensitivity to extremely low meteorological values (T_mean, RH, surface pressure, wind speed, precipitation, T_max, T_min, DTR), primarily within the first trimester (1st–13th weeks).

Non-significant associations are detailed in Tables S10–S11. In summary, extreme meteorological exposures influenced complications within distinct windows: risks of GDM, GH, and PE concentrated in mid-pregnancy (3rd–5th months), while hypothyroidism showed first-trimester vulnerability.

Effects of interactions between extreme meteorological factors on risks of pregnancy complications

Analysis of potential additive interactions between extreme meteorological factors revealed no statistically significant effects on risks of GDM, GH and PE (Tables S16–S18). In contrast, Table S19 demonstrates significant interactions between these factors on the risk of hypothyroidism during the first trimester.

Significant interactions between extreme meteorological factors on hypothyroidism risk were identified under varying percentile thresholds (Table 4). When extremes were defined as the 5th/95th percentiles, positive interactions occurred between extremely low T_min (3.97 °C) and low precipitation (2.36 mm), high wind speed (4.03 m/s), or high DTR (8.93 °C), as well as between high wind speed and high DTR (all RERI > 0.60). For the 3rd/97th percentile thresholds, positive interactions emerged between extremely low T_mean (7.74 °C) and high wind speed (4.17 m/s; RERI = 3.02) or high DTR (9.07 °C; RERI = 2.53), while a negative interaction was observed between low precipitation (2.13 mm) and high wind speed (RERI = −0.73). Under the 1st/99th percentile definition, a positive interaction linked low T_mean (7.44 °C) and high surface pressure (1,023.13 hPa; RERI = 2.02), whereas negative interactions involved low RH (65.32%) with low T_min (2.76 °C; RERI = −1.66) or high surface pressure (RERI = −1.09). All interaction metrics (RERI, AP, SI) with 95% CIs are detailed in Table 4. No additional significant interactions were detected (Table S19).