This historical cohort research was conducted at Shanghai First Maternity and Infant Hospital in China. We analyzed singleton pregnancies in women registered between January 2013 and September 2022. Exclusion criteria included: 1) Pre-existing hypertension or other cardiovascular diseases; 2) Implausible gestational age or incomplete clinical information and UA data records; 3) duplicate pregnancy records. Finally, 84,298 eligible cases were enrolled (Figure 1). Demographic and clinical data were prospectively collected through standardized clinical documentation, including age at conception, residential origin (divided into Shanghai/non-Shanghai), education level of the mother, pre-pregnancy anthropometrics (height, weight), and the calculated body mass index (BMI, weight[kg]/height[m]²). Gestational age was calculated from the last menstrual period. The diagnoses of “gestational hypertension”, “preeclampsia”, and “eclampsia”, as well as the mode of delivery, were retrieved from the electronic records. This study has received approval from the Institutional Review Board (IRB) of the Shanghai First Maternity and Infant Hospital, School of Medicine, Tongji University (Ethics Approval Number: KS1998). Abbreviations mentioned in this work were listed in Supplementary Table S1.

Serum UA levels were measured during every prenatal examination before the 24^th week of gestation. Fasting venous blood samples were collected at the clinic, and immediately centrifuged at 4,500 rpm for serum extraction. Levels of serum UA were quantified using enzymatic methods on a Hitachi 7600 chemical analyzer (Hitachi Co., Tokyo, Japan) at the clinical laboratory of our hospital. The intra-assay coefficient of variation (CV) in the clinical laboratory was below 4.25% and the inter-assay CV was below 5.67%. All measurements were performed in the same hospital laboratory using a consistent analytic platform with daily internal quality-control procedures and annual external proficiency testing throughout the study period. Though temporal assay drift or batch effects across the 10-year period cannot be entirely ruled out, such bias would likely be non-differential.

All statistical analyses were performed using R software (version 4.3.0). Continuous variables were presented as mean ± standard deviation or median (interquartile range) as appropriate, while categorical variables were expressed as counts (percentages). For the longitudinal analysis, we used Kendall’s rank correlation test to evaluate trends in UA levels over time within each group. The median and interquartile ranges of UA levels were calculated for each gestational window (the 4^th-8^th, 9^th-12^th, 13^th-16^th, 17^th-20^th, and 21^st-24^th weeks).

Potential nonlinear relationships were explored by generalized additive models (GAMs) with logit link functions with a cubic regression spline and a basis dimension of k = 10 for gestational weeks, and internal validation was performed by 10-fold cross-validation, to accommodate potential nonlinear relationships and interaction effects between UA and other linear confounders. In this way, the nonlinear associations between serum UA levels and the risk of each HDP outcome were revealed, respectively; separate models were fitted with smooth terms and interaction terms. Adjusted predictions were generated to visualize the probability of HDP outcomes across UA levels at representative gestational weeks. The non-faceted plots were derived via locally weighted scatterplot smoothing (LOESS); the faceted plots were estimated using GAMs. The above analyses were conducted in R using the mgcv package for GAMs.

Multivariable logistic regression models were also developed to assess the independent association between UA levels and HDP risk after adjusting for potential confounders. As an exposure variable, UA levels were categorized into quartiles (Q1-Q4). The models adjusted for: 1) maternal age (per 5-year increase); 2) pre-pregnancy BMI; 3) glucose metabolism disorder (GMD); 4) maternal education level. Adjusted odds ratios (aORs) with 95% confidence intervals were calculated for each variable. Tests for linear trend across UA quartiles were performed by modeling quartiles as an ordinal variable. Model assumptions were verified using variance inflation factors (all <2.0) and Hosmer-Lemeshow goodness-of-fit tests (all p>0.05).

All statistical tests were two-sided, with p-values <0.05 considered statistically significant. Data visualization was performed using the ggplot2 package, including smoothed curves for GAM predictions and forest plots for odds ratios. About missing data handling, for each analysis, pregnancies with missing exposure or outcome information would be excluded. Given the small amount of missing data and lack of systematic patterns in missingness, such exclusion hardly introduced meaningful bias in the estimated associations.

The study included a total of 84,298 pregnancies (Figure 1), with their demographic and clinical information exhibited based on the diagnosis of hypertensive diseases (Table 1). Maternal age distribution across grouped pregnancy cases seems to be similar, while subjects with HDP diagnosis had significantly higher pre-pregnancy BMI (GH: 23.2 ± 3.7; PE: 22.9 ± 3.5; HDP: 23.1 ± 3.7) compared to normotensive pregnancies (21.2 ± 2.7), Moreover, the prevalence of pre-pregnancy obesity was significantly higher in GH (9.9%), PE (7.5%), and HDP (8.6%) compared to normotensive women (2.2%), meanwhile women with PE (17.1%) and HDP (16.2%) had higher rates of glucose metabolism disorders than normotensive women (11.2%), including pre-gestational diabetes mellitus (PGDM) and impaired glucose tolerance (IGT). Majority of the enrolled pregnant women held bachelor’s degree, while lower proportions of bachelor (Normotensive 61.0% vs PE 56.6%, HDP 58.8%) and advanced education (Normotensive 14.2% vs GH 9.4%, PE 12.1%, HDP 10.8%), and higher proportions of low education (Normotensive 8.0% vs GH 9.5%, PE 9.9%, HDP 9.8%) were observed in cases with hypertensive diseases.

Remarkably, the average level of UA across each case’s all gestational age was elevated in GH (190–255μmol/L), PE (186–251μmol/L), and combined HDP cohorts (188–253μmol/L) compared to normotensive women (174–229μmol/L) (Table 1).

Across the 4^th-8^th week and the afterward every 4-week gestational window until the 24^th week, UA levels were consistently higher in women who later developed HDP (including GH and PE) than in normotensive ones (Table 2), and such consistent elevation was also illustrated in Figure 2. From the gestation of 4^th–24^th weeks, pregnancies later diagnosed with HDP exhibited persistently higher UA levels than normotensive pregnancies, with no discernible temporal trend across the whole gestational age, which means serum UA changed significantly with advancing gestational age, yet no consistent or interpretable pattern emerged across these changes.

In our GAMs models using log-transformed maternal uric acid [log(UA)] (Supplementary Figure S1) as a continuous predictor, the associations between UA levels and the risk of HDP diseases varied across gestational weeks, with distinct temporal patterns to different subtypes of HDP.

Overall, higher maternal UA concentrations were consistently associated with an increased risk of HDP, at any gestational week between 4^th to 24^th weeks and for all HDP subtypes (Figures 3–6). For GH (Figure 3A), the estimated association of log(UA) with adverse outcome was approximately linear, indicating a stable positive relationship throughout 4^th to 24^th gestational weeks [effective degrees of freedom (edf) = 2.00, χ² = 357.7, p < 0.0001]. As for PE (Figure 3B), the relationship displayed a non-linear N-shaped curve, with two peaks of association strength occurring around gestational weeks 10 and 24 (edf = 5.60, χ² = 333.1, p < 0.0001). In the GAM analysis simultaneously including GH and PE cases (Figure 3C), the smooth term of gestational week interacting with maternal log(UA) was highly significant (edf = 5.51, χ² = 662.3, p < 0.0001). It indicated a markedly nonlinear association between log(UA) and HDP risk over gestation. The fitted smooth function showed a pattern resembling the GAM model of PE, characterized by an overall N-shaped relationship with more pronounced upward trends in later gestational weeks.

Furthermore, LOESS curves representing the predicted risks of the adverse outcomes based on the established GAM models were shown. It’s obvious that GH showed the most stable enhanced relation between maternal UA and hypertensive outcome from 4^th to 24^th weeks’ gestation (Figure 4) among all the models. Though for both PE and overall HDP models, associations between maternal UA levels and outcomes varied non-monotonically across gestational weeks with great significance, higher UA concentrations were still associated with an increased risk of HDP diseases (Figures 4, 5).

Supported by previous analyses, elevated UA within 4^th-24^th weeks may signify a sustained pro-hypertensive state. The research further conducted multivariable regression analyses using UA as the exposure variable and adjusted maternal covariates. It revealed significant associations between elevated UA and increased risks of GH, PE, and any HDP after confounder adjustments (Table 3). Compared to the lowest UA quartile (Q1), women with the highest quartile (Q4) had substantially higher odds of GH (aOR=1.82, 95%CI: 1.59–2.08), PE (aOR=1.67, 95%CI: 1.48–1.89), and the combined risk of HDP (aOR=1.77, 95%CI: 1.61–1.94)(all p<0.001), with intermediate risk for Q3 (GH aOR=1.41, 95%CI: 1.23-1.63; PE aOR=1.17, 95%CI: 1.02-1.33; HDP aOR=1.28, 95%CI: 1.16–1.41) and no significant association for Q2, indicating evident dose-response relationship.

Moreover, pre-pregnancy BMI showed a strong effect, with each unit increase associated with 16–19% higher odds of GH, 13–16% higher odds of PE, and totaling 16–18% higher odds of HDP (all p<0.001); and each 5-year more advanced conception age linked to higher 11% GH risk (p<0.001). GMDs were associated with PE risk (aOR=1.21, p<0.001). Risks of the three hypertensive outcomes were also differentially adjusted by maternal education level: advanced education above bachelor’s degree served as protective factor against GH (aOR=0.71, p<0.001) and any HDP development (aOR=0.87, p<0.01), whereas associate education or lower associated with increased PE risk (aOR=1.28 for associate degree, p<0.001; aOR=1.26 for high school education or less, p<0.01), as well as the whole HDP risk (aOR=1.20 for associate degree, p<0.001; aOR=1.15 for high school education or less, p<0.05).

In all, the Q3 (in GH and the HDP perspective) and Q4 (in all three types of hypertensive outcomes) quantiles outperformed other maternal covariates, including pre-pregnancy BMI as an effective indicator (Figures 7–9), which suggests that maternal UA above the average during pregnancy serves as an important feature of HDP development.

The stratified analyses depend on demographic and clinical factors (including education levels, pre-pregnancy BMI, conception age, and GMD status) are presented, all used population with UA in Q1 as the reference group. Overall, except in the underpowered subgroup (the underweight divided by pre-pregnancy BMI) due to the limited sample sizes, GH risk (Table 4, Figure 10) increased consistently across ascending serum UA quartiles (Q2 to Q4), with the highest quartile (Q4) commonly bearing the greatest burden of disease risk. Among pre-pregnancy obese cases, GH risk was markedly accentuated, with the aOR increasing progressively from 2.48 in the Q2 cohort to 3.96 in the Q4 cohort. No substantial difference in GH risk was observed when divided by conception age of 35 years. In the GMD subgroup, GH risk was significantly elevated from Q3 onward (Q3 aOR = 1.45, Q4 aOR = 1.94), whereas in non-GMD participants, significance was reached only in Q4 (aOR = 1.79).

Likewise, PE risk (Table 5, Figure 11) rose progressively across UA level in Q2 to Q4, with the most pronounced elevation observed in Q4. When stratified by pre-pregnancy BMI, PE risk rose significantly in the Q4 cohort among women with normal weight (aOR = 1.78) and those who were overweight (aOR = 1.73); in obese women, no statistically significant association was observed across any quartile. Distinct from GH, the association between serum UA and PE differed by maternal age. Among younger women, elevated UA conferred a significant risk in Q3, Q4, whereas in the advanced conception age group (≥35 years), the relationship was stronger, with Q4 aOR of 2.21 versus 1.70 in the younger group. UA–PE association was more consistent among women with GMD, showing significant elevations from Q3 onward (aOR = 1.25 and 1.76, respectively), whereas in non-GMD women, significance was reached only in Q4 (aOR = 1.70). Combinedly, the effect of maternal UA level differed most significantly when divided by GMD status and pre-pregnancy BMI (Table 6, Figure 12).