Abstract
Background
To assess the nonlinear association between city-scale storm events and incident metabolic syndrome (MetS) in the China Health and Retirement Longitudinal Study (CHARLS) cohort of middle-aged and older adults, and to examine the effect modification and mediating effects of PM₂.₅, as well as spatial heterogeneity.
Methods
Based on CHARLS data from the 2011 baseline survey followed up to 2015, 4,085 participants without pre-existing MetS at baseline were included, with 512 incident MetS cases identified. Cox proportional hazards models were used to estimate risks; natural cubic splines (RCS) were applied to characterize the exposure–response relationship. RCS×RCS interactions was “storm peak rainfall × PM₂.₅,” with simple slope curves estimated at the 10th, 50th, and 90th percentiles (P10, P50, P90) of pollution levels. Parallel multiple mediation analysis assessed the indirect effects of PM₂.₅, and psychosocial factors. Global/local spatial autocorrelation and multiscale geographically weighted regression (MGWR) were employed to evaluate spatial clustering and non-stationarity.
Results
In quantile-based Cox models, compared with Q1 (lowest rainfall), participants in Q2–Q4 showed lower MetS risks (Model 2: HR_Q2 = 0.67, HR_Q3 = 0.61, HR_Q4 = 0.58; all significant). After full adjustment, associations attenuated but remained directionally consistent (Model 3: HR_Q2 = 0.76, HR_Q3 = 0.74, HR_Q4 = 0.81). PM₂.₅ was positively associated with MetS (HR ≈ 1.01 per unit), while annual mean temperature was protective (HR ≈ 0.96 per unit). The RCS revealed a U-shaped curve: in low storm peak rainfall areas (≈ 0–25 units), HR was slightly above 1; minimal risk occurred at moderate storm peak rainfall (≈ 50–150 units, HR ≈ 0.5–0.7); beyond 175 units, risk increased sharply (HR > 8 at ≈ 200 units). Significant effect modifications were observed in RCS×RCS analyses: PM₂.₅ stratification: At low pollution (P10 ≈ 37.1 µg/m³), the association increased with storm peak rainfall intensity; neutral at median (P50 ≈ 55.2 µg/m³); protective at high pollution (P90 ≈ 79.6 µg/m³). Mediation analysis indicated a significant indirect effect through PM₂.₅, and nonsignificant psychosocial pathways. Spatial analysis revealed clustering of MetS incidence and non-stationary storm effects—predominantly negative in northern/northeastern China, but neutral or positive in parts of central/southern coastal regions.
Conclusion
City-scale storm peak rainfall exhibited a nonlinear association with incident MetS, where moderate intensity may be protective but extreme intensity harmful. PM2.5 served as the primary modifying and mediating pathway and marked regional variability. Health risk assessments of storm peak rainfall must consider both exposure ranges and geographic heterogeneity.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12942-026-00457-7.
Keywords: Storm peak rainfall, Metabolic syndrome, Natural cubic splines, Interaction, Mediation analysis, MGWR, Spatial heterogeneity
Background
Metabolic syndrome (MetS) is a cluster of interconnected conditions—including central obesity, hypertension, elevated blood glucose, and dyslipidemia—that significantly increase the risk of cardiovascular disease and type 2 diabetes [1]. In recent years, the prevalnce of MetS has been rising sharply in China, particularly among older adults, due to lifestyle changes, dietary shifts, and environmental factors associated with rapid urbanization [2].
At the same time, urban areas in China are increasingly affected by extreme storm events, which have become more frequent and intense under climate change [3]. Storm event not only cause physical infrastructure damage and economic losses but also disrupt daily activities and health services. Prolonged exposure to such extreme weather conditions can limit outdoor physical activity, induce psychological stress, and interfere with chronic disease management—all of which are potential contributors to the development and progression of MetS [4, 5].
Current research has primarily revealed the direct impacts and short-term effects of storm events on human health, while studies focusing on the indirect, long-term effects on chronic diseases such as metabolic syndrome remain relatively limited. Significant short-term effects on the cardiovascular system have been observed: studies have confirmed that storm and extreme precipitation can have notable short-term impacts on cardiovascular health. For example, a study covering 30 major cities in East Asia found that storm events (such as 5-year and 10-year return-period rainstorms) were significantly associated with a short-term increase in the risk of respiratory disease mortality [6].
Another study conducted in Nanchang City also demonstrated that short-term exposure to higher storm levels was linked to an increased risk of hospital admissions for stroke, with the effect being most pronounced at a lag of 1 to 2 days.These acute cardiovascular events are often related to the physiological stress, sudden changes in temperature and humidity, and increased levels of air pollutants and bioaerosols (such as pollen and mold spores) caused by heavy rainfall [7]. These factors may exacerbate preexisting endothelial dysfunction and inflammatory conditions in individuals with metabolic syndrome.
Despite the growing body of research linking climate variability to health outcomes, there is limited empirical evidence on the indirect, long-term impact of storm events on chronic metabolic conditions. Most existing studies focus on acute outcomes such as infectious diseases or injuries. The interplay between environmental hazards like storm events and chronic non-communicable diseases remains understudied, especially in aging populations.Previous studies have suggested that meteorological conditions, air pollution, and behavioral–psychological factors jointly contribute to the development of metabolic syndrome (MetS). However, prospective evidence on the exposure–response shape, potential mediation pathways, and regional heterogeneity of the “storm–metabolic outcome” relationship remains limited.
Using nationally representative data from the China Health and Retirement Longitudinal Study (CHARLS), we aim to:1) To characterize the nonlinear exposure–response relationship between peak rainfall and the incidence of metabolic syndrome (MetS);2) To quantify the mediating effects of temperature, PM₂.₅, and psychosocial factors;3) To examine the spatial clustering and non-stationarity of MetS incidence; and4) To validate effect modification across temperature and pollution quantiles using RCS×RCS interactions, thereby linkingthese findings with the spatial mechanisms identified by MGWR to form an integrated analytical framework.
The China Health and Retirement Longitudinal Study (CHARLS), a nationally representative panel survey of middle-aged and older Chinese adults, provides a unique opportunity to investigate this relationship. By integrating individual health data with geolocated meteorological information, this study aims to evaluate how exposure to storm events is associated with the risk of MetS, thereby contributing to a more nuanced understanding of climate-related chronic disease risks in rapidly urbanizing societies.
Methods
Study design and population
This study utilized data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative longitudinal cohort survey conducted between 2011 and 2020 to assess the social, economic, and health status of Chinese residents [8]. The CHARLS baseline survey, carried out from June 2011 to March 2012, included 10,257 households and 17,708 participants [8]. All participants were subsequently followed up every two years via face-to-face interviews using standardized questionnaires.
The data and related information used in this study are publicly available for download from the CHARLS project website (http://charls.pku.edu.cn/).
This study used publicly available data of the China Health and Retirement Longitudinal Study (CHARLS) which is a nationwide survey of residents in China. Although surveys of the CHARLS have been conducted four times, the blood biomarkers such as cholesterol levels were measured only at the first survey (conducted from May 2011 to March 2012) and the third survey (conducted from July to September 2015) of the CHARLS. In the present study, participants from the first and the third surveys of the CHARLS were included. Briefly, individuals aged 45 years or older, underwent standardized questionnaires about basic information, health status and function, medical insurance, and provided fasting blood samples to measure the biomarkers of MetS [including fasting glucose (FG), triglycerides (TG) and high-density lipoprotein cholesterol (HDL-C)], were considered in the study.Eventually, 4085 participants were included in the current analysis, a flowchart of the selection of participants is attached in the material (As shown in Fig. 1 of the uploaded file). Each participant had the blood pressure (BP) measured on the left arms by an electronic sphygmomanometer (Omron HEM-7200 Monitor), and then the average value of three consecutive measurements was taken. A height meter was used to measure height and it’s accurate to 0.1 cm. A weight scale was employed to measure weight and it’s accurate to 0.1 kg. Waist was measured to the nearest 0.1 cm. We calculated BMI by dividing weight by height squared (kg/m2). The fasting venous blood samples were collected and then using an enzymatic colorimetric test to measure the concentrations of FG, TG and HDL-C.
Fig. 1.
Flowchart of the selection of participants
Exposure: storm Count, storm Duration, storm Level, storm peak rainfall, storm volume
Storm count refers to the total number of rainstorm events within a given period, reflecting the frequency of extreme precipitation occurrences in a region [9]. Storm Duration(h) represents the length of time that each rainstorm event persists, which can influence the extent of flooding and infrastructure impact [10]. Storm Level(mm/h) denotes the classification or intensity grade of rainstorm events, commonly based on standardized thresholds of rainfall amount [11]. Storm Peak Rainfall(mm/d) is the highest rainfall intensity observed during a single rainstorm event, serving as an indicator of short-term precipitation extremes [12]. Storm Volume(mm) indicates the total amount of precipitation produced during a storm event, reflecting the potential for surface runoff and flood risk [13]. These five indicators are widely used in meteorological studies to evaluate the characteristics, intensity, and potential impacts of rainstorm events [14]. In recent years, global climate change has led to frequent occurrences of extreme weather events. As one of these events, rainstorms have had a profound impact on both the ecological environment and public health in China. Nationwide rainstorm datasets provide an important foundation for studying the spatiotemporal distribution characteristics, intensity changes, and underlying mechanisms of rainstorms.All rainfall data were obtained from: Bo He, Ming Yisen, Liu Qihang, et al. China Rainstorm Dataset, 2001–2019. https://cstr.cn/31253.41.sciencedb.j00001.00290.005E952E.
Storm data from 2011 to 2015 were utilized in the present analysis.
Outcome: metabolic syndrome
Metabolic syndrome (MetS) is a cluster of interrelated metabolic abnormalities, including central obesity, insulin resistance, hypertension, hyperglycemia, and dyslipidemia. Individuals with metabolic syndrome are at a significantly increased risk for developing type 2 diabetes, cardiovascular diseases, and other related chronic conditions [15]. The diagnosis of MetS is typically based on the presence of at least three out of five criteria: increased waist circumference, elevated fasting glucose, elevated triglycerides, reduced high-density lipoprotein cholesterol, and elevated blood pressure [16]. In recent years, the prevalence of metabolic syndrome has risen globally, becoming a major public health concern due to shifts in lifestyle and dietary patterns [17].
The specific diagnostic criteria for metabolic syndrome are as follows:
(1) Waist circumference
For Chinese individuals, ≥ 90 cm for men and ≥ 80 cm for women.
(2) Hypertriglyceridemia
Triglycerides ≥ 1.7 mmol/L, or currently receiving treatment.
(3) Reduced high-density lipoprotein cholesterol (HDL-C)
<1.03 mmol/L for men, < 1.29 mmol/L for women, or currently receiving treatment.
(4) Elevated blood pressure
Systolic blood pressure ≥ 130 mmHg and/or diastolic blood pressure ≥ 85 mmHg, or currently receiving antihypertensive treatment.
(5) Elevated fasting blood glucose
≥5.6 mmol/L, or previously diagnosed with type 2 diabetes.Diagnosed as metabolic syndrome if any three of the above criteria are met [16].
Confounding variables
Based on a directed acyclic graph (DAG) (Fig. 2), the following covariates were included: age (45–60 years, ≥ 60 years); gender (male, female); education level (no formal education, elementary school or below, middle school or above); marital status (married/unmarried); body mass index (BMI); smoking (yes: having the habit of smoking cigarettes, smoking a pipe, or chewing tobacco; no); drinking (1: having the habit of drinking; 0: no); social activity frequency; disability (0: without disabilities; 1: having one type of disability, such as physical disability, brain injury, intellectual disability, visual impairment, hearing loss, or speech disorder); health insurance (yes/no); income status (1: with wage income; 0: without wage income); CESD-10 psychological scale score; PM2.5 concentration; and annual mean temperature [18].
Fig. 2.
DAG
Statistical analysis
A total of 4,085 participants free of metabolic syndrome (MetS) at baseline in 2011 were followed up through 2015, during which 512 incident MetS cases were identified (Table 1).
Table 1.
Characteristics of the study population
| Characteristic | Overall, N = 4,085 | 0, N = 3,573 | 1, N = 512 |
|---|---|---|---|
| Age(years) | 60.00 (54.00, 67.00) | 60.00 (54.00, 67.00) | 60.00 (54.00, 67.00) |
| Educational level, n (%) | |||
| Illiterate | 2,031.0 (49.7%) | 1,777.0 (49.7%) | 254.0 (49.6%) |
| Primary school and below | 966.0 (23.6%) | 831.0 (23.3%) | 135.0 (26.4%) |
| Middle school and high school | 773.0 (18.9%) | 685.0 (19.2%) | 88.0 (17.2%) |
| Technical school and above | 315.0 (7.7%) | 280.0 (7.8%) | 35.0 (6.8%) |
| Sex, n (%) | |||
| Female | 1,809.0 (44.3%) | 1,487.0 (41.6%) | 322.0 (62.9%) |
| Male | 2,276.0 (55.7%) | 2,086.0 (58.4%) | 190.0 (37.1%) |
| Marriage, n (%) | |||
| Unmarried | 557.0 (13.6%) | 494.0 (13.8%) | 63.0 (12.3%) |
| Married | 3,528.0 (86.4%) | 3,079.0 (86.2%) | 449.0 (87.7%) |
| Medical insurance, n (%) | |||
| No | 230.0 (5.6%) | 194.0 (5.4%) | 36.0 (7.0%) |
| Yes | 3,855.0 (94.4%) | 3,379.0 (94.6%) | 476.0 (93.0%) |
| BMI, Median (IQR) | 21.78 (19.93, 24.02) | 21.50 (19.75, 23.56) | 24.31 (22.23, 26.60) |
| Alcohol consumption, n (%) | |||
| Never drinking | 2,580.0 (63.2%) | 2,220.0 (62.1%) | 360.0 (70.3%) |
| Former drinking | 1,505.0 (36.8%) | 1,353.0 (37.9%) | 152.0 (29.7%) |
| Smoke, n (%) | |||
| Never smoking | 2,527.0 (61.9%) | 2,138.0 (59.8%) | 389.0 (76.0%) |
| Former smoking | 1,558.0 (38.1%) | 1,435.0 (40.2%) | 123.0 (24.0%) |
| Cesd10, Median (IQR) | 8.00 (4.00, 13.00) | 8.00 (4.00, 13.00) | 9.00 (4.00, 14.00) |
| Income status, n (%) | |||
| No | 257.0 (6.3%) | 228.0 (6.4%) | 29.0 (5.7%) |
| Yes | 3,828.0 (93.7%) | 3,345.0 (93.6%) | 483.0 (94.3%) |
| Social activity frequency, n (%) | |||
| 0 | 2,111.0 (51.7%) | 1,861.0 (52.1%) | 250.0 (48.8%) |
| 1 | 1,398.0 (34.2%) | 1,205.0 (33.7%) | 193.0 (37.7%) |
| 2 | 471.0 (11.5%) | 414.0 (11.6%) | 57.0 (11.1%) |
| 3 | 88.0 (2.2%) | 77.0 (2.2%) | 11.0 (2.1%) |
| 4 | 14.0 (0.3%) | 13.0 (0.4%) | 1.0 (0.2%) |
| 5 | 3.0 (0.1%) | 3.0 (0.1%) | 0.0 (0.0%) |
| Disability, n (%) | |||
| No | 3,332.0 (81.6%) | 2,901.0 (81.2%) | 431.0 (84.2%) |
| Yes | 753.0 (18.4%) | 672.0 (18.8%) | 81.0 (15.8%) |
1Wilcoxon rank sum test; Pearson’s Chi-squared test; Fisher’s exact test
Time to incident MetS was modeled using Cox proportional hazards regression to evaluate the association between city-level storm peak rainfall and MetS onset. Confounders were progressively adjusted in three models: Model 1 included storm peak rainfall exposure only; Model 2 additionally controlled for demographic characteristics; and Model 3 further adjusted for health behaviors and environmental factors(Table 2) [19]. And conducted within–between decomposition using Cox regression.Cluster-robust standard errors were estimated at the city level (or alternatively using province-level fixed effects or stratification, as described in the Supplementary Materials). To explore potential nonlinear associations, storm peak rainfall was modeled with restricted cubic splines (RCS) using 3–5 knots (degrees of freedom = 3 in the main analysis) [19]. Nonlinearity was assessed by joint Wald and likelihood ratio tests, and the minimum-risk point (MRP) and effective exposure range (P5–P95) were reported. Interaction analyses examined storm peak rainfall × PM₂.₅ interactions (RCS×RCS), with simple slope curves plotted at selected quantiles of temperature and PM₂.₅. A parallel multiple mediation framework was applied to estimate total effects (TE), natural direct effects (NDE), and natural indirect effects (NIE).The addition of bootstrap sensitivity analysis and PMR computation has been implemented, aiming to mitigate reliance on proportional interpretations. Bootstrap methods (≥ 1000 iterations) were used to generate confidence intervals, and mediation proportions as well as sensitivity analyses (using alternative temperature metrics and different spline knot settings) were reported. Spatial analyses were conducted using global Moran’s I and Local Indicators of Spatial Association (LISA) to detect clustering of MetS incidence, and multiscale geographically weighted regression (MGWR) to assess spatial non-stationarity of storm peak rainfall effects, yielding coefficient surfaces and corresponding uncertainty estimates. Robustness analyses using an alternative diagnostic definition of MetS confirmed the consistency of storm peak rainfall–MetS associations in both direction and magnitude [20]. All statistical analyses and figure generation were performed using R software (version 4.3.2), and p < 0.05 was considered statistically significant.
Table 2.
Cox proportional hazards regression
| Group | Characteristic | HR1 | 95% CI1 | p-value | |
|---|---|---|---|---|---|
| Model 1 | Storm peak rainfall(mm/day) | ||||
| Q1 | Reference | Reference | |||
| Q2 | 0.69 | 0.55, 0.87 | 0.002 | ||
| Q3 | 0.64 | 0.50, 0.81 | < 0.001 | ||
| Q4 | 0.59 | 0.47, 0.76 | < 0.001 | ||
| Model 2 | Storm peak rainfall(mm/day) | ||||
| Q1 | Reference | Reference | |||
| Q2 | 0.67 | 0.53, 0.84 | < 0.001 | ||
| Q3 | 0.61 | 0.48, 0.78 | < 0.001 | ||
| Q4 | 0.58 | 0.45, 0.74 | < 0.001 | ||
| Model 3 | Storm peak rainfall(mm/day) | ||||
| Q1 | Reference | Reference | |||
| Q2 | 0.76 | 0.59, 0.99 | 0.045 | ||
| Q3 | 0.74 | 0.55, 0.99 | 0.044 | ||
| Q4 | 0.81 | 0.58, 1.12 | 0.197 | ||
1HR = Hazard Ratio, CI = Confidence Interval
Results
A total of 4,085 participants free of metabolic syndrome (MetS) at baseline in 2011 were followed up through 2015, during which 512 incident MetS cases were identified.
Quantile-based Cox models showed that.Other storm-related events showed no significant associations, and the corresponding results are presented in the Supplementary Materials. However, storm peak rainfall exhibited distinct findings; therefore, our analyses focused on this indicator. Compared with the lowest quartile of storm peak rainfall (Q1), participants in Q2–Q4 had significantly lower risks of developing MetS (Model 2: HR_Q2 = 0.67, HR_Q3 = 0.61, HR_Q4 = 0.58). After full adjustment, the direction remained consistent but the magnitude was attenuated. PM₂.₅ was positively associated with MetS (≈ 1.01 per unit), while annual mean temperature showed a protective effect (≈ 0.96 per unit).
RCS analysis revealed a clear U-shaped association between storm peak rainfall and MetS risk. Slightly elevated risks were observed under low storm peak rainfall, with the lowest risk at moderate storm peak rainfall levels (≈ 50–150 units), followed by a sharp increase beyond 175 units. The test for nonlinearity was statistically significant (Fig. 3). In the within-between decomposition of two waves of fixed follow-up cohorts, the regional within-effect (within) remains insignificant, while the regional between-effect (between) persists as protective under various adjusted control schemes. (Supplementary material).
Fig. 3.
RCS analysis (between rainfall and MetS risk)
RCS×RCS interaction effects
Storm peak rainfall × PM₂.₅ (Fig. 4): In low-pollution settings (P10 ≈ 37 µg/m³), increasing storm peak rainfall corresponded to higher risk; at median pollution (P50 ≈ 55 µg/m³), the effect was neutral; and in high-pollution areas (P90 ≈ 80 µg/m³), risk markedly declined with stronger storm peak rainfall. This suggests that under higher pollution, storm peak rainfall-induced wet scavenging provided greater net benefit, while under cleaner conditions, disaster- and behavior-related risks predominated. The interaction was statistically significant (Table 3).
Fig. 4.
Rainfall × PM₂.₅
Table 3.
Results of joint tests and interactions
| Factor | Chi-Square | P |
|---|---|---|
| TOTAL NONLINEAR + INTERACTION | 22.27 | 0.0023 |
| Total Factor | 151.77 | <0.0001 |
| Storm (Factor+Higher Order Factors) | 20.83 | 0.002 |
| PM2.5 (Factor+Higher Order Factors) | 20.49 | 0.0023 |
| Temp | 19.77 | 0.0001 |
| Nonlinear | 4.59 | 0.0321 |
Sensitivity analyses showed that the widening confidence intervals at the upper tail of the low-temperature curve reflected sparse extreme exposures; restricting analyses to the common support region yielded consistent conclusions.
Mediation and Spatial effects
In the present mediation analysis, only PM2.5 demonstrated a statistically significant mediating effect, albeit with an extremely small effect size (ACME ≈ -0.00007). Social activity, self-rated health, and depressive symptoms (CES-D10) were not significant mediators. (Fig. 5)
Fig. 5.
Mediation analysis
Spatial analysis demonstrated spatial clustering of MetS incidence. According to the results of spatial autocorrelation analysis, the incidence rate of metabolic syndrome (MetS) in the study area exhibits a statistically significant spatial clustering pattern. Specifically, the Moran’s I value is 0.057, which, although small, indicates a weak but definite positive spatial correlation. This suggests that areas with high and low incidence rates tend to cluster spatially. The corresponding z-score is 4.361, which is much higher than the critical reference value of 2.58, and the p-value is less than 0.001. These results indicate that the likelihood of such a clustering pattern arising from random processes is extremely low. Therefore, the finding is highly significant at the 99% confidence level, confirming the existence of a clear spatial dependence in MetS incidence.MGWR results revealed spatial non-stationarity, with predominantly negative associations in northern and northeastern China (β ≈ -0.92 to -0.12), and neutral to positive associations in central-southern and coastal regions (β ≈ -0.12 to + 0.27), suggesting that climatic and pollution contexts jointly shaped the regional heterogeneity of MetS risk. (Fig. 6).
Fig. 6.
Spatial Effects
Discussion
In this nationally representative longitudinal study, we found a nonlinear, U-shaped association between storm peak rainfall and the risk of incident metabolic syndrome (MetS) in Chinese adults. Moderate rainfall appeared to be protective, particularly more polluted conditions, while both low and extreme storm peak rainfall were associated with elevated risks. Spatially, the associations varied markedly across regions, reflecting the influence of climatic zones, pollution levels, and local adaptive capacity [21].
The observed U-shaped pattern may be explained by distinct underlying mechanisms. Mechanistically, moderate rainfall may reduce ambient particulate matter exposure via wet scavenging, thereby transiently improving local air quality and environmental conditions. In contrast, extreme storm events likely elevate metabolic risk through multiple concurrent stress-related pathways. These include reduced healthcare accessibility due to flooding and transportation disruptions, curtailed daily activities and physical exercise, interruptions in the management of chronic conditions, and heightened psychological stress. Consequently, the overall influence of storm events on metabolic health is more likely to reflect a composite of environmental, behavioral, and service-accessibility factors rather than a single, isolated physiological pathway.
Within our parallel mediation framework, only PM₂.₅ exhibited a statistically detectable indirect effect. However, the average causal mediation effect (ACME) for PM₂.₅ was numerically close to zero (negative but extremely small in magnitude), and the proportion mediated was also minimal (on the order of ~ 0.02). This indicates that, even with statistical significance achieved in our large sample, the absolute change in MetS risk attributable to this specific mediation pathway is very limited and does not support a clinically meaningful or perceptible mediating mechanism. We therefore emphasize the effect size over statistical significance in this context: in terms of magnitude, PM₂.₅ contributes negligibly as a mediator and may at most hint at a potentially involved pathway, not a key mechanism explaining the primary exposure–outcome association. Other hypothesized mediators—social activity, self-rated health, and depressive symptoms (CES-D10)—were not supported, as their ACMEs were approximately zero with confidence intervals crossing zero. This further indicates limited evidence for strong mediation mechanisms in the present study.
Spatially, MetS incidence exhibited statistically significant yet weak clustering (small but significant Moran’s I). Multiscale geographically weighted regression (MGWR) revealed significant spatial non-stationarity in the storm peak rainfall–MetS association. The association was predominantly negative in northern and northeastern regions but tended to be neutral or positive in central-southern and coastal areas. This pattern suggests that climatic zones, regional pollution context, and local adaptive capacity jointly shape this regional heterogeneity. Coupled with the within–between decomposition results—which showed a non-significant within-region effect but more prominent between-region differences—we advocate for a cautious interpretation of our findings. This pattern highlights potential ecological bias and underlying structural differences between regions. Therefore, the main conclusions of this study should be grounded in the converging evidence regarding the exposure–response shape, interaction context, and spatial heterogeneity, rather than equating the observed between-region associations with strong individual-level causal effects.
Key strengths of this study include its nationally representative cohort, prospective design, and the integration of complementary analytical evidence from restricted cubic spline (RCS) modeling, interaction analyses, and spatial methods. Nonetheless, several limitations warrant consideration. First, sparse observations at the extreme tails of the storm distribution increase statistical uncertainty in those ranges. Second, the use of city-level exposure assignment may introduce ecological bias. Third, unmeasured or incompletely measured region-level differences in lifestyle, diet, and socioeconomic factors may result in residual confounding. Finally, as illustrated by our mediation results, statistically significant findings can occur with trivial effect sizes, which limits their clinical or practical relevance.
In conclusion, our findings illustrate a complex, context-dependent relationship between storm events and metabolic health, shaped by environmental, spatial, and potentially behavioral factors. Future research should aim to incorporate more granular, individual-level behavioral and physiological data collected during extreme weather events (e.g., blood pressure, heart rate, and energy metabolism), utilize dynamic real-world indicators of healthcare accessibility, and integrate biomarker measurements. Such approaches are needed to validate the potential pathways suggested here and to inform actionable, precision public health strategies for climate-related metabolic risk prevention.
Supplementary Information
Acknowledgements
Not applicable.
Author contributions
ZiJie Cai: Writing – original draft, Supervision, Project administration, Funding acquisition.ChunZhi Tang: Writing – review, Supervision, Project administration.LiXiang Gan: Writing – review & editing, Supervision, Project administration. GuangPeng Zhang: Writing – review & editing, Writing – Software, Methodology, Investigation, Formal analysis, Data curation. YaLong Qiu: Writing – review & editing, Methodology, Investigation. HongYing Tian: Writing – review & editing, Visualization, Validation, Software, Resources, Formal analysis.
Funding
Not applicable.
Data availability
Data is provided within the manuscript or supplementary information files.
Declarations
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Alberti KGMM, Zimmet P, Shaw J. Metabolic syndrome–a new world-wide definition. A consensus statement from the international diabetes federation. Diabet Med Engl. 2006;23(5):469–80. PMID:16681555IF: 3.4 Q2 B3. [DOI] [PubMed] [Google Scholar]
- 2.Pan Y, Wu X, Liu Y, Li Z, Yang Y, Luo Y. Urbanization and cognitive function among Middle-Aged and old adults in China. J Gerontol B Psychol Sci Soc Sci United States. 2022;77(12):2338–47. PMID:35908238IF: 3.2 Q1 B2. [DOI] [PubMed] [Google Scholar]
- 3.Deshpande A, Chang HH, Levy K. Heavy rainfall events and diarrheal diseases: the role of Urban-Rural geography. Am J Trop Med Hyg United States. 2020;103(3):1043–9. PMID:32700663IF: 1.6 Q3 B4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Romanello M, di Napoli C, Green C, Kennard H, Lampard P, Scamman D, Walawender M, Ali Z, Ameli N, Ayeb-Karlsson S, Beggs PJ, Belesova K, Berrang Ford L, Bowen K, Cai W, Callaghan M, Campbell-Lendrum D, Chambers J, Cross TJ, van Daalen KR, Dalin C, Dasandi N, Dasgupta S, Davies M, Dominguez-Salas P, Dubrow R, Ebi KL, Eckelman M, Ekins P, Freyberg C, Gasparyan O, Gordon-Strachan G, Graham H, Gunther SH, Hamilton I, Hang Y, Hänninen R, Hartinger S, He K, Heidecke J, Hess JJ, Hsu S-C, Jamart L, Jankin S, Jay O, Kelman I, Kiesewetter G, Kinney P, Kniveton D, Kouznetsov R, Larosa F, Lee JKW, Lemke B, Liu Y, Liu Z, Lott M, Lotto Batista M, Lowe R, Odhiambo Sewe M, Martinez-Urtaza J, Maslin M, McAllister L, McMichael C, Mi Z, Milner J, Minor K, Minx JC, Mohajeri N, Momen NC, Moradi-Lakeh M, Morrissey K, Munzert S, Murray KA, Neville T, Nilsson M, Obradovich N, O’Hare MB, Oliveira C, Oreszczyn T, Otto M, Owfi F, Pearman O, Pega F, Pershing A, Rabbaniha M, Rickman J, Robinson EJZ, Rocklöv J, Salas RN, Semenza JC, Sherman JD, Shumake-Guillemot J, Silbert G, Sofiev M, Springmann M, Stowell JD, Tabatabaei M, Taylor J, Thompson R, Tonne C, Treskova M, Trinanes JA, Wagner F, Warnecke L, Whitcombe H, Winning M, Wyns A, Yglesias-González M, Zhang S, Zhang Y, Zhu Q, Gong P, Montgomery H, Costello A. The 2023 report of the lancet countdown on health and climate change: the imperative for a health-centred response in a world facing irreversible harms. Lancet Engl. 2023;402(10419):2346–94. PMID:37977174IF: 88.5 Q1 B1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Khraishah H, Alahmad B, Ostergard RLJ, AlAshqar A, Albaghdadi M, Vellanki N, Chowdhury MM, Al-Kindi SG, Zanobetti A, Gasparrini A, Rajagopalan S. Climate change and cardiovascular disease: implications for global health. Nat Rev Cardiol Engl. 2022;19(12):798–812. PMID:35672485IF: 44.2 Q1 B1. [DOI] [PubMed] [Google Scholar]
- 6.He C, Kim H, Hashizume M, Lee W, Honda Y, Kim SE, Guo YL, Schneider A, Zhu Y, Zhou L, Chen R, Kan H. The overlooked health impacts of extreme rainfall exposure in 30 East Asian cities. Nat Sustain. 2024;7(4):423–31. 10.1038/s41893-024-01294-x. [Google Scholar]
- 7.ZENG F. Association between rainfall and stroke admissions:based on distributional lag nonlinear modeling. 2024;27(20):2458–65. 10.12114/j.issn.1007-9572.2024.0010
- 8.Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China health and retirement longitudinal study (CHARLS). Int J Epidemiol Engl. 2014;43(1):61–8. PMID:23243115IF: 5.9 Q1 B2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhou L, Liu L. Enhancing dynamic flood risk assessment and zoning using a coupled hydrological-hydrodynamic model and Spatiotemporal information weighting method. J Environ Manage Engl. 2024;366:121831. PMID:39018862IF: 8.4 Q1 B2. [DOI] [PubMed] [Google Scholar]
- 10.Song Y, Park M. A study on setting disaster-Prevention rainfall by rainfall duration in urban areas considering natural disaster damage: focusing on South Korea. Water. 2020;12(3). 10.3390/w12030642.
- 11.Wan W, Lei X, Zhao J, Wang M, Khu S, Wang C. A Forecast-Skill-Based dynamic Pre-Storm level control for reservoir Flood-Control operation. Water. 2021;13(4). 10.3390/w13040556.
- 12.Feld G, Randell D, Wu Y, Ewans K, Jonathan P. Estimation of storm peak and intrastorm Directional-Seasonal design conditions in the North sea. J OFFSHORE Mech Arct ENGINEERING-TRANSACTIONS ASME. 2015;137(2). 10.1115/1.4029639.
- 13.Chen L, Yan Z, Li Q, Xu Y. Flash flood risk assessment and driving factors: A case study of the Yantanxi river Basin, southeastern China. Int J DISASTER RISK Sci. 2022;13(2):291–304. 10.1007/s13753-022-00408-3. [Google Scholar]
- 14.Zhou K. Analysis on the Jul.20 extreme rainstorm and flood control countermeasures in Zhengzhou, China. J WATER Clim CHANGE. 2024;15(8):3549–65. 10.2166/wcc.2024.647. [Google Scholar]
- 15.Grundy SM, Cleeman JI, Daniels SR, Donato KA, Eckel RH, Franklin BA, Gordon DJ, Krauss RM, Savage PJ, Smith SCJ, Spertus JA, Costa F. Diagnosis and management of the metabolic syndrome: an American heart Association/National heart, Lung, and blood Institute scientific statement. Circulation United States. 2005;112(17):2735–52. PMID:16157765IF: 38.6 Q1 B1. [DOI] [PubMed] [Google Scholar]
- 16.Alberti KGMM, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA, Fruchart J-C, James WPT, Loria CM, Smith SCJ, the International Diabetes Federation Task Force on Epidemiology and Prevention. Harmonizing the metabolic syndrome: a joint interim statement of ; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation United States; 2009;120(16):1640–1645. PMID:19805654IF: 38.6 Q1 B1. [DOI] [PubMed]
- 17.Saklayen MG. The global epidemic of the metabolic syndrome. Curr Hypertens Rep United States. 2018;20(2):12. PMID:29480368IF: 5.1 Q1 B2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wei S, Yu S, Qian C, Xu Z, Jia Y, Chen B. Exploring the U-shaped nonlinear relationship of CHG index with metabolic syndrome and mortality risks in metabolic syndrome patients. Lipids Health Dis Engl. 2025;24(1):294. PMID:41023729IF: 4.2 Q1 B2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Tan M-Y, Zhang Y-J, Zhu S-X, Wu S, Zhang P, Gao M. The prognostic significance of stress hyperglycemia ratio in evaluating all-cause and cardiovascular mortality risk among individuals across stages 0–3 of cardiovascular-kidney-metabolic syndrome: evidence from two cohort studies. Cardiovasc Diabetol Engl. 2025;24(1):137. PMID:40128747IF: 10.6 Q1 B1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Zhang Q, Lou S, Meng Z, Ren X. Gender and age impacts on the correlations between hyperuricemia and metabolic syndrome in Chinese. Clin Rheumatol Ger. 2011;30(6):777–87. PMID:21181218IF: 2.8 Q2 B3. [DOI] [PubMed] [Google Scholar]
- 21.He S, Hao C. Analysis on spatial-temporal variation characteristics of climate in Qinling-Huaihe demarcation zone since 1961. Ecol Ind. 2024;158:111345. 10.1016/j.ecolind.2023.111345. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data is provided within the manuscript or supplementary information files.