Abstract
Background
In patients with allergic rhinitis (AR), the dynamic longitudinal association between transition from normal body mass index (BMI) to overweight/obesity and metabolic inflammation remains unclear.
Methods
We conducted a retrospective longitudinal cohort study involving 1715 AR patients with normal baseline BMI between 2014 and 2025 at the Affiliated Hospital of Guangdong Medical University in Zhanjiang, Guangdong, China. Using targeted maximum likelihood estimation with time-window analysis, we estimated the average treatment effects of BMI increase on inflammatory and metabolic markers over 1–3 years and quantified risks of metabolic diseases.
Results
BMI transition was associated with progressively increased systemic inflammation and persistent metabolic dysregulation. Inflammatory markers showed a time-dependent increase, whereas metabolic abnormalities, characterized by elevated triglycerides and reduced high-density lipoprotein cholesterol, appeared earlier and remained stable. BMI gain was also associated with increased risks of several metabolic diseases, particularly fatty liver disease.
Conclusion
In AR patients, transitioning from normal BMI to overweight or obesity was associated with a gradual increase in systemic inflammation and persistent metabolic dysfunction that was associated with increased risks of various metabolic diseases. When implementing weight management for patients with AR, it is recommended to monitor changes in body weight over time and intervene early in the weight gain process.
Keywords: Allergic rhinitis, Body mass index, Inflammation, Metabolic diseases, Cohort studies
Introduction
Allergic rhinitis (AR) is the most common immune-mediated respiratory disorder, affecting approximately 10%–40% of the global population and 8%–20% of the Chinese population, with an estimated nearly 250 million patients in China.1,2 Although AR is not life-threatening, inadequate disease control can result in a range of adverse consequences, including sleep disturbances, headaches, memory impairment, worsening sinusitis, and asthma attacks, substantially impairing patients quality of life.3 Currently, no curative treatment is available for AR. Existing medications such as antihistamines and intranasal corticosteroids only alleviate symptoms.4 Therefore, identifying adjunctive therapeutic strategies that reduce the inflammatory burden of AR is of great clinical importance.
Recent studies have found that systemic low-grade chronic inflammation is an important aggravating factor in the pathogenesis of AR.5 Obesity, a metabolic disease characterized by adipose tissue dysfunction, can induce or exacerbate systemic inflammation through the secretion of various adipokines and inflammatory cytokines (eg,TNF-α, IL-6). 6, 7, 8, 9 In theory, obesity may influence the onset and severity of AR by exacerbating inflammation throughout the body and in the nasal mucosa.10
Obesity itself has become a global public health crisis. According to data from the 2021–2023 National Health and Nutrition Examination Survey (NHANES), approximately 40.3% of U.S. adults are obese (body mass index (BMI)≥30 kg/m2).11 The prevalence of obesity continues to increase globally, including in low- and middle-income countries. Its associations with various chronic diseases, such as type 2 diabetes and cardiovascular disease, have been well established.8,12 However, evidence regarding the impact of weight change on inflammatory status in AR patients is primarily derived from cross-sectional studies, making it difficult to establish causality. Moreover, there is a lack of longitudinal observations on the dynamic changes in inflammatory and metabolic markers in AR patients following weight gain.
This study employed a retrospective cohort design using longitudinal health examination data from AR patients. Targeted maximum likelihood estimation (TMLE) combined with time-window analysis was applied to evaluate the dynamic longitudinal associations of weight gain on inflammatory and metabolic markers. We further quantified the risk of long-term metabolic diseases associated with weight gain. This study aims to provide longitudinal evidence regarding the potential impact of weight management in AR patients and to offer a scientific basis for the early prevention of metabolic inflammation and the selection of optimal intervention timing.
Materials and methods
Data sources and study population
Data sources
This study employed a retrospective observational design, utilizing data from the health examination database of the Affiliated Hospital of Guangdong Medical University spanning from 2014 to 2025. The Affiliated Hospital of Guangdong Medical University was selected as the study setting because it provides a large-scale longitudinal health examination database with standardized examination procedures, uniform laboratory measurements, and continuous follow-up records, enabling reliable assessment of temporal changes in BMI, inflammatory markers, and metabolic outcomes. The Affiliated Hospital of Guangdong Medical University is a major regional medical center serving Western Guangdong, and its health examination center receives individuals from multiple areas across the region. Thus, although the study was conducted at a single institution, the underlying health examination population had a relatively broad geographic catchment. The use of a single institutional database also ensured uniform examination procedures, laboratory measurements, and longitudinal data collection. Patients with AR who underwent multiple health examinations at this center were enrolled. The inclusion criteria required a clear record of a previous AR diagnosis in the health examination database, with AR patients identified based on explicit documentation of “allergic rhinitis” in their examination records. All health examination data were collected using standardized procedures, and laboratory tests were uniformly performed by the hospital's clinical laboratory using regularly calibrated equipment and methods.
The data extraction process in this study strictly adhered to the principles of the Declaration of Helsinki and the ethical requirements of the hospital. All personally identifiable information (eg, names, ID numbers, and contact numbers) was de-identified by the hospital's information technology department prior to data export. Researchers only accessed anonymized study codes and health examination data fields. The study protocol was approved by the Ethics Committee of the Affiliated Hospital of Guangdong Medical University (ethics approval number: YJKT2026-103-01), and the requirement for informed consent from participants was waived.
Study cohort construction
We constructed the study cohort by screening the database for patients with AR who met the following criteria: (1) at least 1 health examination record with a confirmed diagnosis of AR; (2) age greater than 18 years and less than 100 years;13 (3) body mass index (BMI) within the normal adult range (18.5–23.9 kg/m2) at baseline; and (4) at least 2 physical examination records.
Participants aged >100 years were excluded because centenarians represent a biologically distinct population with unique aging trajectories, metabolic characteristics, and disease patterns that may differ substantially from the general adult population. Including these individuals could introduce substantial age-related heterogeneity and affect the evaluation of longitudinal associations between BMI transition and inflammatory/metabolic outcomes.
Exclusion criteria included: (1) Abnormal or missing key inflammatory and metabolic outcomes;14 (2) Missing weight data during follow-up, making it impossible to determine the timing of weight gain. A total of 1715 eligible patients with AR were ultimately included. Fig. 1 illustrates the study population screening process. In addition, extreme/outlier values of key variables were reviewed and excluded when considered clinically implausible or potentially attributable to measurement or recording errors, in order to minimize the influence of abnormal observations on subsequent analyses.
Fig. 1.
Flowchart of study population selection
Outcome assessment
Outcome measures are categorized into 3 types: inflammatory markers, metabolic markers, and clinical outcomes. Inflammatory markers included neutrophil-to-lymphocyte ratio (NLR), white blood cell count (WBC), absolute neutrophil count (ANC), and absolute lymphocyte count (ALC). All these markers were obtained through routine blood tests. Metabolic markers included triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). These markers are obtained through biochemical testing. Clinical outcomes included fatty liver disease, diabetes mellitus, hyperuricemia, kidney disease, hyperlipidemia, coronary heart disease, hypertension, hepatitis, liver cirrhosis, and hyperthyroidism. These diseases were selected based on their established associations with metabolic inflammation: fatty liver disease, diabetes mellitus, hyperuricemia, hyperlipidemia, coronary heart disease, and hypertension are core components of metabolic syndrome; hepatitis (particularly non-alcoholic steatohepatitis) and liver cirrhosis represent hepatic manifestations of fatty liver progression; kidney disease is a common target organ damage of metabolic syndrome; and hyperthyroidism is closely related to dysregulated energy metabolism. Disease information was extracted from the text of health examination reports using natural language processing in R (version 4.2.1) and was independently verified by 2 researchers.
Definition of exposure and construction of the time window
The exposure variable was “weight gain” defined as progression from a normal weight at baseline (18.5 ≤ BMI <24 kg/m2) to overweight or obesity (BMI ≥24 kg/m2). The date of the first health examination at which the participant met the criteria for overweight was recorded as the exposure time point. The normal-weight maintenance group consisted of individuals who maintained a normal weight throughout the study; a virtual exposure time was assigned to them via random matching to ensure temporal comparability. To assess the dynamic changes in post-exposure associations, we established fixed-time-window outcomes at 1, 2, and 3 years post-exposure and calculated the mean change (Δ) in inflammatory and metabolic markers relative to baseline within each time window.
Definition of confounders
To establish a rigorous causal inference framework, this study defined age, gender, baseline BMI as core confounders because they are established determinants of body weight trajectories and are associated with inflammatory and metabolic outcomes.15,16 Baseline inflammatory markers (baseline WBC, baseline ANC, baseline ALC, etc.), baseline metabolic indicators (baseline TG, baseline HDL-C, baseline LDL-C), baseline visceral fat rating, baseline blood pressure indices, baseline renal function indicators, and baseline pulmonary function indicators were additionally considered as potential confounders. A directed acyclic graph (DAG) was employed for topological modeling of the logical relationships among variables, as shown in Fig. 2.
Fig. 2.
Association network of metabolic etiology in allergic rhinitis population
Statistical analysis methods
TMLE was used to estimate average treatment effects under the assumptions of consistency, positivity, and conditional exchangeability given the measured baseline covariates. Fixed 1-year, 2-year, and 3-year post-exposure time windows were applied to characterize temporal changes in inflammatory and metabolic markers following BMI increase. For each time window, TMLE estimated the average treatment effect (ATE) of exposure on the change (Δ) of each inflammatory and metabolic marker from baseline, adjusting for pre-exposure confounders. To avoid model misspecification, we used the SuperLearner algorithm with 10-fold cross-validation to select the optimal prediction models for the outcome (Q-model) and the propensity score (g-model). The SuperLearner library included 5 algorithms: SL. glm, SL. glm.interaction, SL. glmnet, SL. ranger, and SL. mean. Cross-validation selected the convex combination of these algorithms that minimized the cross-validated risk. TMLE then updated the initial Q-model using the estimated propensity score, yielding a doubly robust and asymptotically efficient ATE estimate. Sensitivity analyses were performed to verify the robustness of our primary results. We used a stepwise hierarchical covariate adjustment strategy for inflammatory, metabolic, and clinical outcomes. A total of 5 gradually adjusted models were constructed to assess potential confounding bias. The minimal model included adjustments for age, gender, and baseline BMI. The inflammatory model was based on the minimal model and further adjusted for baseline inflammatory markers. These markers included NLR, WBC, ANC, and ALC. The metabolic model was also built upon the minimal model. It additionally controlled for baseline metabolic markers, including TG, HDL, LDL, and visceral fat level. The comprehensive model combined all inflammatory and metabolic covariates on the basis of the minimal model. The full model added extra clinical indicators to the comprehensive model. These indicators covered blood pressure and renal function, and creatinine was used to evaluate renal function status. All analyses were performed in R version 4.2.1 using the tmle, SuperLearner, glmnet, ranger, and lme 4 packages. A two-sided p < 0.05 was considered statistically significant.
Results
Baseline characteristics and distribution of exposure duration in the study population
This study included a total of 1715 A R patients, 477 (27.8%) experienced weight gain during follow-up and 1238 (72.2%) maintained a normal weight during the follow-up. Table 1 presents the baseline characteristics of the study population. Compared with the normal-weight maintenance group, the BMI gain group had a higher proportion of male participants (60.6% vs. 51.8%, p < 0.01) and a significantly higher baseline BMI (22.7 ± 1.1 kg/m2 vs. 21.2 ± 1.4 kg/m2, p < 0.001). There was no significant difference in age between the 2 groups. These findings suggest that male and higher BMI at baseline may be potential risk factors for weight transition.
Table 1.
Baseline characteristics of the study population
| Characteristics | Overall (N = 1715) | BMI gain group (N = 477) | BMI-stable group (N = 1238) | P value |
|---|---|---|---|---|
| Age (years), mean ± SD | 43.3 ± 13.8 | 42.9 ± 12.8 | 43.4 ± 14.1 | 0.491 |
| Male, n (%) | 930 (54.2%) | 289 (60.6%) | 641 (51.8%) | <0.01 |
| Baseline BMI (kg/m2), mean ± SD | 21.6 ± 1.5 | 22.7 ± 1.1 | 21.2 ± 1.4 | <0.001 |
Among the 477 patients with increased BMI, the median follow-up time was 2.29 years (interquartile range (IQR, 1.49–4.05 years), with a mean time of 2.95 ± 1.90 years. As illustrated in Fig. 3, BMI increase events were predominantly concentrated within the first 4 years of follow-up.
Fig. 3.
Distribution of time to weight transition
Dynamic changes in inflammatory and metabolic markers after weight gain
To investigate the temporal evolution of systemic inflammatory and metabolic status following weight gain, this study applied the TMLE method to evaluate the ATE of each indicator at 1 year, 2 years, and 3 years post-exposure.
Inflammatory markers: sustained and progressively enhanced pro-inflammatory changes
Data analysis showed that inflammatory markers exhibited sustained and progressively increasing pro-inflammatory changes following weight transition (Table 2). At 1 year post-exposure, ANC and NLR began to increase, but the effects did not reach statistical significance (NLR: ATE = +0.109, 95% CI: −0.195–0.414, p = 0.481; ANC: ATE = +0.449 × 109/L, 95% CI: −0.097–0.995, p = 0.107). By the second year after exposure, the effects were further strengthened, and ALC increased significantly (ALC: ATE = +0.118 × 109/L, 95% CI: 0.092–0.144, p < 0.001). By the third year post-exposure, significant elevations were observed in NLR(ATE = +0.134, 95% CI: 0.042–0.227, p = 0.005), WBC (ATE = +0.453 × 109/L, 95% CI: 0.194–0.712, p < 0.001), ANC (ATE = +0.369 × 109/L, 95% CI: 0.077–0.661, p = 0.013), and ALC (ATE = +0.107 × 109/L, 95% CI: 0.021–0.192, p = 0.014). Fig. 4 presents a composite view of these dynamic changes, showing the time-specific ATE estimates, confidence intervals, and significance levels for each inflammatory marker across the 1-, 2-, and 3-year post-exposure windows, as well as the overall average effects. This pattern indicates that the systemic low-grade inflammation observed following weight gain is not transient, but a persistent pathological state that intensifies over time. The overall pooled analysis demonstrated significant long-term increases in WBC (ATE = +0.279 × 109/L, 95% CI: 0.135–0.423, p < 0.001) and ANC (ATE = +0.198 × 109/L, 95% CI: 0.081–0.315, p < 0.001), consistent with the time-window findings at year 3, further supporting the long-term association between weight gain on inflammation.
Table 2.
Dynamic changes in inflammatory markers after weight transition (time-window TMLE analysis and overall average effect).
| Markers | Analysis method | N | BMI gain group | BMI-stable group | ATE (95% CI) | P value |
|---|---|---|---|---|---|---|
| NLR | 1 year | 833 | 196 | 637 | 0.109 (−0.195, 0.414) | 0.481 |
| 2 year | 1092 | 328 | 764 | 0.136 (−0.102, 0.375) | 0.262 | |
| 3 year | 1142 | 355 | 787 | 0.134 (0.042, 0.227) | 0.005 | |
| Overall analysis | 1705 | 476 | 1229 | 0.009 (−0.067, 0.085) | 0.817 | |
| WBC ( × 109/L) | 1 year | 833 | 196 | 637 | 0.400 (−0.317, 1.117) | 0.274 |
| 2 year | 1092 | 328 | 764 | 0.391 (−0.226, 1.008) | 0.214 | |
| 3 year | 1142 | 355 | 787 | 0.453 (0.194, 0.712) | <0.001 | |
| Overall analysis | 1705 | 476 | 1229 | 0.279 (0.135, 0.423) | <0.001 | |
| ANC ( × 109/L) | 1 year | 833 | 196 | 637 | 0.449 (−0.097, 0.995) | 0.107 |
| 2 year | 1092 | 328 | 764 | 0.309 (−0.168, 0.786) | 0.205 | |
| 3 year | 1142 | 355 | 787 | 0.369 (0.077, 0.661) | 0.013 | |
| Overall analysis | 1705 | 476 | 1229 | 0.198 (0.081, 0.315) | <0.001 | |
| ALC ( × 109/L) | 1 year | 833 | 196 | 637 | 0.063 (−0.201, 0.327) | 0.64 |
| 2 year | 1092 | 328 | 764 | 0.118 (0.092, 0.144) | <0.001 | |
| 1 year | 1142 | 355 | 787 | 0.107 (0.021, 0.192) | 0.014 | |
| Overall analysis | 1705 | 476 | 1229 | 0.055 (−0.001, 0.110) | 0.054 |
Note: the overall analysis used the traditional TMLE method, incorporating all follow-up data to estimate the average effect of weight transition throughout the entire follow-up period.
Fig. 4.
Forest plot of dynamic changes in inflammatory markers following weight change
Metabolic parameters: characteristics of persistent dysregulation
Unlike the persistent worsening of inflammatory markers, metabolic markers exhibited a sustained and stable pattern of dysregulation following weight change (Table 3). Time-window analysis revealed that TG were significantly elevated in all time windows following exposure: by 0.208 mmol/L at 1 year (95% CI: 0.085–0.331, p < 0.001), 0.202 mmol/L at 2 years (95% CI: 0.100–0.304, p < 0.001), and 0.203 mmol/L at 3 years (95% CI: 0.107–0.298, p < 0.001). HDL-C decreased significantly at all time points: by 0.094 mmol/L at 1 year (95% CI: −0.140 to −0.049, p < 0.001), decreased by 0.084 mmol/L at 2 years (95% CI: −0.126 to −0.043, p < 0.001), and decreased by 0.093 mmol/L at 3 years (95% CI: −0.134 to −0.052, p < 0.001). LDL-C did not show significant changes at any time point (all p > 0.05). Fig. 5 presents the time-window and overall estimates for TG, HDL-C, and LDL-C, illustrating that TG increases and HDL-C decreases persist across the 3 years while LDL-C remains unchanged. This “high TG–low HDL-C″ lipid profile is a hallmark of obesity-related metabolic dysfunction, and this pattern was observed within 1 year of weight gain and persisted thereafter. The results of the overall analysis were highly consistent with those of the time-window analysis, further confirming the persistent nature of metabolic dysfunction.
Table 3.
Dynamic changes in metabolic markers after weight transition (time-window TMLE analysis and overall average effect).
| Markers | Analysis method | N | BMI gain group | BMI-stable group | ATE (95% CI) | P value |
|---|---|---|---|---|---|---|
| TG (mmol/L) | 1 year | 1206 | 184 | 1022 | 0.208 (0.085, 0.331) | <0.001 |
| 2 year | 1476 | 307 | 1169 | 0.202 (0.100, 0.304) | <0.001 | |
| 3 year | 1533 | 335 | 1198 | 0.203 (0.107, 0.298) | <0.001 | |
| Overall analysis | 1657 | 447 | 1210 | 0.205 (0.125, 0.285) | <0.001 | |
| HDL-C (mmol/L) | 1 year | 1204 | 184 | 1020 | −0.094 (−0.140, −0.049) | <0.001 |
| 2 year | 1474 | 307 | 1167 | −0.084 (−0.126, −0.043) | <0.001 | |
| 3 year | 1531 | 335 | 1196 | −0.093 (−0.134, −0.052) | <0.001 | |
| Overall analysis | 1654 | 447 | 1207 | −0.063 (−0.096, −0.029) | <0.001 | |
| LDL-C (mmol/L) | 1 year | 1105 | 155 | 950 | 0.048 (−0.086, 0.181) | 0.485 |
| 2 year | 1425 | 286 | 1139 | −0.039 (−0.138, 0.060) | 0.441 | |
| 3 year | 1499 | 322 | 1177 | −0.016 (−0.110, 0.078) | 0.737 | |
| Overall analysis | 1620 | 433 | 1187 | 0.023 (−0.075, 0.121) | 0.65 |
Note: the overall analysis used the traditional TMLE method, incorporating all follow-up data to estimate the average effect of weight transition throughout the entire follow-up period.
Fig. 5.
Dynamic changes of metabolic indexes after BMI changes
Dose-effect relationship between continuous BMI changes and inflammatory and metabolic markers
To validate these findings from a dose-response perspective, we used a mixed-effects model to analyze the effect of continuous changes in BMI on each indicator. As shown in Table 4, for every 1 kg/m2 increase in BMI, multiple inflammatory and metabolic markers increased significantly: WBC (β = 0.116 × 109/L, 95% CI: 0.081–0.150, p < 0.001), ANC (β = 0.098 × 109/L, 95% CI: 0.069–0.127, p < 0.001), ALC (β = 0.015 × 109/L, 95% CI: 0.003–0.027, p = 0.013), NLR (β = 0.034, 95% CI: 0.014–0.053, p < 0.001), TG (β = 0.136 mmol/L, 95% CI: 0.116–0.155, p < 0.001), and LDL-C (β = 0.053 mmol/L, 95% CI: 0.032–0.074, p < 0.001). In contrast, HDL-C decreased significantly with rising BMI (β = −0.027 mmol/L, 95% CI: −0.034 to −0.020, p < 0.001). Fig. 6 displays the nonlinear dose-response relationships between continuous BMI changes and these 7 inflammatory and metabolic markers. These results demonstrate a clear dose-response association between increased BMI and inflammation and metabolic disorders, consistent with the direction and significance of the time-window TMLE findings.
Table 4.
Effect of each 1 kg/m2 increase in BMI on inflammatory and metabolic markers (mixed-effects model).
| Markers | Number of observations | BMI effect (95% CI) | P value |
|---|---|---|---|
| WBC ( × 109/L) | 2646 | 0.116 (0.081, 0.150) | <0.001 |
| ANC ( × 109/L) | 2647 | 0.098 (0.069, 0.127) | <0.001 |
| ALC ( × 109/L) | 2647 | 0.015 (0.003, 0.027) | 0.013 |
| NLR | 2647 | 0.034 (0.014, 0.053) | <0.001 |
| TG (mmol/L) | 2499 | 0.136 (0.116, 0.155) | <0.001 |
| HDL-C (mmol/L) | 2499 | −0.027 (−0.034, −0.020) | <0.001 |
| LDL-C (mmol/L) | 2286 | 0.053 (0.032, 0.074) | <0.001 |
Fig. 6.
Nonlinear dose-response relationship between BMI change and indicators
The impact of increased BMI on the risk of metabolic diseases and patterns of comorbidity
To assess the impact of BMI increase from the normal range (18.5–23.9 kg/m2) to overweight/obesity (≥24 kg/m2) on the risk of long-term metabolic diseases, this study compared the post-exposure incidence rates of 10 metabolic-related diseases between the BMI gain group and the BMI-stable group, after excluding individuals who had the corresponding disease at baseline. The results are presented in Table 5. Risk Difference (RD) represents the absolute difference in incidence rates between the 2 groups; Number Needed to Treat (NNT) indicates the number of individuals who would need to be intervened upon to prevent 1 additional case of the disease; Odds Ratio (OR) reflects the strength of the association between exposure and disease.
Table 5.
Incidence of metabolic diseases after BMI increase from normal to overweight/obesity.
| Disease | BMI gain group | BMI-stable group | RD (95% CI) | Adjusted OR (95%CI) | P Value | NNT |
|---|---|---|---|---|---|---|
| Fatty liver disease | 49.0% (169/345) | 23.4% (253/1079) | 0.255 (0.197, 0.314) | 2.64 (1.96, 3.55) | <0.001 | 3.9 |
| Diabetes mellitus | 62.7% (153/244) | 45.2% (408/902) | 0.175 (0.106, 0.244) | 1.99 (1.43, 2.78) | <0.001 | 5.7 |
| Hyperuricemia | 35.6% (108/303) | 26.7% (269/1007) | 0.089 (0.029, 0.150) | 1.51 (1.10, 2.07) | 0.003 | 11.2 |
| Hyperlipidemia | 44.1% (116/263) | 36.7% (323/881) | 0.074 (0.007, 0.142) | 1.55 (1.12, 2.15) | 0.03 | 13.5 |
| Kidney disease | 26.5% (113/426) | 19.8% (232/1172) | 0.067 (0.020, 0.115) | 1.50 (1.10, 2.05) | 0.005 | 14.9 |
| Liver cirrhosis | 14.2% (63/443) | 9.7% (116/1192) | 0.045 (0.008, 0.082) | 1.43 (0.98, 2.09) | 0.012 | 22.3 |
| Coronary heart disease | 62.9% (129/205) | 56.2% (449/799) | 0.067 (−0.007, 0.142) | 1.53 (1.06, 2.20) | 0.096 | – |
| Hepatitis | 20.2% (71/352) | 16.7% (171/1022) | 0.034 (−0.013, 0.082) | 1.41 (0.99, 2.02) | 0.145 | – |
| Hypertension | 20.9% (74/354) | 21.6% (218/1009) | −0.007 (−0.056, 0.042) | 1.05 (0.73, 1.50) | 0.822 | – |
| Hyperthyroidism | 13.3% (56/420) | 16.1% (189/1176) | −0.027 (−0.066, 0.011) | 1.03 (0.71, 1.49) | 0.207 | – |
Note:Analysis restricted to individuals free of the disease at baseline. Adjusted for age, gender, and baseline BMI. NNT is presented only for diseases with positive and statistically significant risk differences (P < 0.05).
The post-exposure incidence rates of 6 diseases—fatty liver disease, diabetes mellitus, hyperuricemia, hyperlipidemia, kidney disease, and liver cirrhosis—were significantly higher in the BMI gain group than in the BMI-stable group. Fatty liver disease exhibited the strongest clinical impact, with a significant RD of 25.5% (95% CI: 19.7%–31.4%, p < 0.001), an adjusted OR of 2.64 (95% CI: 1.96–3.55, p < 0.001), and the smallest NNT of 3.9. The other 5 significant diseases had adjusted OR ranging from 1.43 to 1.99. No significant RD differences were observed for coronary heart disease, hepatitis, hypertension, or hyperthyroidism (all p > 0.05).
Fig. 7 illustrates the impact of BMI increase on the risk of metabolic diseases. The left panel presents a forest plot showing the risk difference with 95% confidence intervals for each disease, while the right panel displays a horizontal bar chart directly comparing the incidence rates between the 2 groups. These findings demonstrate that BMI increase is an important risk factor for multiple metabolic diseases in patients with AR, with fatty liver disease showing the most pronounced risk elevation, highlighting its clinical preventive significance.
Fig. 7.
Incidence and risk difference of metabolic diseases in AR patients after BMI gain
To further elucidate the intrinsic associations among metabolic diseases in individuals with increased BMI, this study calculated the Jaccard comorbidity similarity among 6 significant metabolic diseases (fatty liver, diabetes, hyperuricemia, hyperlipidemia, kidney disease, and cirrhosis) based on the BMI gain group (N = 477) and constructed a comorbidity network diagram (Fig. 8).
Fig. 8.
Circular comorbidity network of metabolic diseases in the BMI gain group
Comorbidity network analysis revealed varying degrees of comorbidity associations among the 6 conditions, with the strongest comorbidity observed between fatty liver and diabetes, diabetes and hyperlipidemia, and fatty liver and hyperlipidemia, with Jaccard similarities of 76.6%, 68.1%, and 60.2%, respectively. This indicates a close pathophysiological link among fatty liver disease, diabetes, and hyperlipidemia, suggesting that dyslipidemia and impaired glucose metabolism are highly co-occurring in individuals with elevated BMI. Hyperuricemia also showed moderate comorbidity associations with hyperlipidemia (Jaccard similarity 51.5%), diabetes (56.4%), and fatty liver disease (51.0%), suggesting that abnormal uric acid metabolism is closely related to disorders of glucose and lipid metabolism. The comorbidity rates of kidney disease with fatty liver, diabetes, hyperlipidemia, and hyperuricemia were 40.9%, 37.5%, 37.2%, and 33.1%, respectively, suggesting that the kidneys are a key target organ for metabolic syndrome. The comorbidity rates of cirrhosis with various metabolic diseases were relatively low (18.6%–22.1%), which may be related to the long course of cirrhosis and the fact that its full manifestation had not yet been observed during the follow-up period.
In Fig. 8, node size represents disease prevalence (with the highest prevalence for fatty liver disease and the lowest for cirrhosis), node color distinguishes different diseases, and arc thickness represents Jaccard comorbidity similarity (thicker arcs indicate stronger comorbidity associations). The network structure reveals that fatty liver disease, diabetes, and hyperlipidemia occupy central positions in the network; these 3 conditions are closely interconnected. Hyperuricemia, kidney disease, and cirrhosis are located on the periphery, exhibiting varying degrees of connectivity with the core diseases.
Sensitivity analysis
Sensitivity analyses confirmed the robustness of the core study findings regarding the associations between BMI increase and metabolic-inflammatory outcomes. As shown in Table 6, the direction of effect, magnitude of effect, and statistical significance of the core outcomes remained consistent across all adjusted models. This indicates that the main findings of this study are robust and reliable and are not significantly influenced by unmeasured confounders.
Table 6.
Sensitivity analysis: Effect estimates across different covariate adjustment models
| Outcome | Minimal | Basic | Metabolic | Comprehensive | Full |
|---|---|---|---|---|---|
| WBC ( × 109/L) (ATE) | 0.306 | 0.312 | 0.33 | 0.325 | 0.293 |
| ANC ( × 109/L) (ATE) | 0.202 | 0.219 | 0.246 | 0.216 | 0.217 |
| ALC ( × 109/L) (ATE) | 0.073 | 0.072 | 0.046 | 0.048 | 0.049 |
| NLR(ATE) | 0.019 | 0.012 | 0.05 | 0.023 | 0.027 |
| TG (mmol/L) (ATE) | 0.207 | 0.208 | 0.209 | 0.207 | 0.208 |
| HDL-C (mmol/L) (ATE) | −0.045 | −0.045 | −0.045 | −0.046 | −0.045 |
| LDL-C (mmol/L) (ATE) | 0.011 | 0.01 | 0.011 | 0.011 | 0.011 |
The sensitivity analysis results for the primary clinical outcomes are shown in Fig. 9. Across different covariate adjustment models, the direction of risk differences remained consistent for fatty liver disease, diabetes mellitus, hyperuricemia, hyperlipidemia, kidney disease, and liver cirrhosis. The effect magnitudes showed only limited variation. These findings indicate a relatively robust association between BMI increase and an elevated risk of these diseases. Notably, fatty liver disease consistently exhibited the largest risk difference across all models, followed by diabetes mellitus and hyperuricemia, further supporting the stability of the main findings of this study.
Fig. 9.
Sensitivity analysis of risk differences for primary clinical outcomes across different covariate adjustment models
Discussion
This longitudinal cohort study applied the TMLE method with a time-window analysis to assess the temporal associations between increases in BMI from normal weight to overweight/obesity and subsequent inflammatory and metabolic changes from normal weight to overweight/obese in patients with AR. BMI transition was associated with progressively higher inflammatory markers over 3 years and with early and persistent metabolic dysregulation characterized by higher TG and lower HDL-C. Specifically, inflammatory markers progressively worsened over the three-year follow-up period. In contrast, metabolic dysregulation, characterized by high TG and low HDL-C, emerged early and remained stable. BMI increase was also associated with higher risks of several metabolic diseases, with the strongest association observed for non-alcoholic fatty liver disease.
Previous studies on the association between obesity and AR have yielded inconsistent results. Some meta-analyses found no significant association between abnormal BMI and AR in either children or adults, while others reported a positive association, particularly in Western pediatric populations 17, 18, 19. Age and gender differences further contributed to the heterogeneity of the findings: 1 large cross-sectional study observed that obesity was associated with an increased risk of non-allergic rhinitis but not AR in adults, and even a decreased risk of AR in children;20,21 some studies reported a positive correlation exclusively among women.22,23 These discrepancies may partly reflect methodological heterogeneity, including differences in AR diagnostic criteria, obesity measures, and confounder adjustment. More importantly, most previous studies used cross-sectional designs, which cannot establish causality or capture temporal dynamics. In contrast, our study provides longitudinal evidence characterizing temporal changes following BMI transition by leveraging repeated measurements and a causal inference framework. Consistent with general population studies, we confirm that weight gain was associated with systemic low-grade inflammation and an atherogenic lipid profile.24 However, we extend these findings by revealing temporal heterogeneity specific to AR patients: metabolic abnormalities manifest within 1 year of weight gain, whereas inflammatory escalation requires longer accumulation, particularly for composite markers such as NLR. Moreover, the marked clustering of fatty liver disease, diabetes, and hyperlipidemia observed in AR patients has not been emphasized before, suggesting that AR may share common immuno-metabolic pathways with these conditions 25, 26, 27.
Our study extends the understanding of obesity-related changes from a static association to a dynamic longitudinal perspective and provides clinically relevant evidence for early weight management in patients with AR. Importantly, our findings suggest that metabolic deterioration may begin before the onset of overt obesity, as the transition from normal BMI to overweight was already associated with measurable metabolic abnormalities. The early emergence of elevated TG and reduced HDL-C within 1 year after weight gain indicates that metabolic screening should not be limited to patients with established obesity but may also be considered for AR patients experiencing progressive weight increase. Furthermore, the gradual worsening of inflammatory markers over subsequent years suggests that preventing weight gain may help reduce the long-term inflammatory burden associated with AR. The low NNT (3.9) observed for fatty liver disease indicates a potentially meaningful clinical benefit of preventing BMI progression, supporting the integration of routine weight monitoring, metabolic assessment, and early lifestyle intervention into comprehensive AR management.
Beyond the statistical associations observed in this study, our findings have potential implications for the clinical management of patients with allergic rhinitis. Importantly, the transition from normal BMI to overweight was already associated with measurable metabolic abnormalities, suggesting that metabolic risk assessment should not be restricted to patients with established obesity. In the clinical setting, monitoring BMI trajectories rather than relying solely on a single BMI measurement may help identify AR patients at an earlier stage of metabolic deterioration. For patients with progressive weight gain, periodic assessment of metabolic parameters, including lipid profiles, and early lifestyle interventions focusing on weight control, dietary optimization, and physical activity may represent potential strategies to reduce future metabolic risks.
A key strength of this study lies in the application of the TMLE method combined with SuperLearner, a robust causal inference framework that provides doubly robust, data-adaptive estimates. The fixed-time-window design further enabled the evaluation of temporal changes following BMI transition. Nevertheless, several limitations should be acknowledged. First, the retrospective single-center health examination cohort limited the availability of detailed biological and anthropometric information, including specific inflammatory cytokines (eg, hs-CRP, IL-6, and TNF-α) and fine-grained body weight change phenotypes (eg, rate of weight gain, and weight fluctuations). Although this was a single-center study, the hospital serves a broad population across Western Guangdong, providing greater geographic coverage than a narrowly local single-center cohort. Nevertheless, broad regional catchment does not necessarily ensure population representativeness. Participants undergoing repeated health examinations may differ from other patients with AR in healthcare utilization, socioeconomic characteristics, health awareness, and lifestyle behaviors. Therefore, the findings may be most directly applicable to adults with AR undergoing periodic health examinations in Western Guangdong and to populations with similar characteristics. Further validation in multicenter cohorts from other regions and healthcare settings is warranted. Second, the binary definition of BMI transition from normal weight to overweight/obesity may have overlooked the potential impact of the magnitude and velocity of BMI increase. Third, although AR diagnosis was based on documented diagnostic records in the health examination database, detailed clinical phenotypes of AR were unavailable. Information regarding symptom severity, allergen sensitization, disease control status, and AR-specific treatments could not be evaluated, which may have introduced heterogeneity among AR patients. Future prospective studies incorporating standardized AR assessments are needed to further characterize these variations. Fourth, metabolic disease outcomes were primarily identified from routine health examination reports using natural language processing followed by manual verification. Although 2 independent researchers reviewed the extracted disease information to improve accuracy, misclassification of disease status remains possible because detailed clinical diagnostic criteria, medication records, and longitudinal medical histories were not fully available. Some conditions may have been underdiagnosed or inconsistently documented during routine examinations, which could have influenced the estimated associations between BMI transition and metabolic disease risks. Future studies integrating electronic medical records, standardized diagnostic criteria, laboratory confirmation, and physician-adjudicated outcomes are needed to further validate these findings. Finally, although we adjusted for a comprehensive set of potential confounders, including demographic characteristics, baseline inflammatory markers, metabolic parameters, visceral fat level, renal function, and blood pressure indicators, residual confounding cannot be completely excluded. Several potentially important factors, such as smoking status, physical activity, dietary patterns, socioeconomic status, and medication use, were unavailable in the health examination database. These unmeasured or incompletely measured factors may influence both BMI transition and metabolic-inflammatory outcomes. Although sensitivity analyses demonstrated the robustness of the main findings across different adjustment models, the observed associations should still be interpreted within the framework of measured covariates. Future studies incorporating more comprehensive lifestyle information and biological measurements are needed to further clarify these associations.
In conclusion, this study provides longitudinal evidence that BMI transition from normal to overweight/obesity in AR patients is associated with both progressive systemic inflammation and persistent metabolic dysregulation, with distinct temporal patterns—metabolic changes appearing early and inflammatory markers worsening gradually. Clinically, these findings suggest that weight management in AR should not be delayed until obesity is established; rather, routine monitoring of BMI trajectories and periodic lipid profiling could enable earlier detection of metabolic risk. The particularly pronounced risk elevation for fatty liver disease, reflected by a low number needed to treat of 3.9, underscores a potentially high-yield target for preventive intervention. Nevertheless, these findings should be interpreted in light of the single-center design and the possibility of residual confounding. Future prospective studies incorporating comprehensive lifestyle data and standardized AR phenotyping are warranted to confirm these associations. If validated, integrating weight monitoring, metabolic assessment, and early lifestyle counseling into routine AR care could represent a pragmatic strategy to mitigate long-term cardiometabolic comorbidities in this population.
Authors’ consent for publication
All authors have reviewed and approved the final manuscript and consent to its publication.
Availability of data and materials
The data that support the findings of this study are available from the health examination database of the Affiliated Hospital of Guangdong Medical University but restrictions apply to the availability of these data, which were used under ethical approval for the current study, and so are not publicly available.
Ethics approval
This study was conducted in accordance with the principles of the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of the Affiliated Hospital of Guangdong Medical University (ethics approval number: YJKT2026-103-01). The requirement for informed consent was waived by the ethics committee due to the retrospective nature of the study and the use of de-identified data.
Author contributions
Conceptualization: Siying Pan.
Data curation: Jiayuan Wu, Junxian Wang, Zuyi Zhao.
Formal analysis: Siying Pan, Linxi He, Xiaohan Xu;
Funding acquisition: Bingsong Zhang;
Investigation: Jiayuan Wu, Junxian Wang;
Methodology: Siying Pan, Linxi He;
Software: Siying Pan, Linxi He, Xiaohan Xu;
Supervision: Xiaohan Xu, Bingsong Zhang;
Validation: Zuyi Zhao; Visualization: Siying Pan, Linxi He;
Writing: Siying Pan, Linxi He; Review and editing: Xiaohan Xu, Bingsong Zhang.
Disclosure of the use of generative AI and AI-assisted technologies
During the preparation of this work, the authors used ChatGPT and DeepSeek to improve the readability and linguistic quality of the manuscript. The tool was applied solely for language polishing, including grammar correction, sentence structure refinement, and wording optimization. It was not used to generate scientific content, draw clinical conclusions, provide medical recommendations, or produce any scientific, pedagogical, or medical insights. All authors have carefully reviewed and revised the AI-assisted suggestions and take full responsibility for the accuracy, integrity, and originality of the final manuscript.
Funding
This work is supported by the Big Data Platform of Affiliated Hospital of Guangdong Medical University; the National Natural Science Foundation of China Youth Program (82304253); the Special Project for Clinical and Basic Sci&Tech Innovation of Guangdong Medical University (GDMULCJC2025053, GDMULCJC2024006).
Competing interests
The authors declare that they have no competing interests.
Footnotes
Full list of author information is available at the end of the article
Contributor Information
Juhong Yang, Email: megii0315@126.com.
Bingsong Zhang, Email: zhangbingsong@gdmu.edu.cn.
References
- 1.Kogias C., Drylli A., Panagiotakos D., Douros K., Antonogeorgos G. Allergic rhinitis systematic review shows the trends in prevalence in children and adolescents in Greece since 1990. Allergie. 2023;3:220–228. [Google Scholar]
- 2.Wang N., Yao Y., Liu Y.H., Zeng M., Liu Z. Allergic rhinitis in China: trends, challenges and implications over the past two decades. Clin Exp Allergy. 2025;55:648–658. doi: 10.1111/cea.70118. [DOI] [PubMed] [Google Scholar]
- 3.Conway A.E., Verdi M., Anagnostou A., et al. Sleep disorders in the allergy-immunology clinic. J Allergy Clin Immunol Pract. 2025;13:2940–2950. doi: 10.1016/j.jaip.2025.07.052. [DOI] [PubMed] [Google Scholar]
- 4.Yamani I., Bu Saeed K., Alsulami A., Sait S., Althumali A.H. Efficacy of biologic therapies in the management of allergic rhinitis: a systematic review. Cureus. 2024;16 doi: 10.7759/cureus.71408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hui D., Shi C., Fan Y., et al. Systemic immune dysregulation in allergic rhinitis: mechanisms, comorbidities, and implications for targeted therapy. J Asthma Allergy. 2026;19:1–11. doi: 10.2147/JAA.S589195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lamabadusuriya D.A., Jayasena H., Bopitiya A.K., De Silva A.D., Jayasekera P. Obesity-driven inflammation and cancer risk: a comprehensive review. Semin Cancer Biol. 2025;114:256–266. doi: 10.1016/j.semcancer.2025.07.007. [DOI] [PubMed] [Google Scholar]
- 7.Renovato-Martins M., Moreira-Nunes C., Atella G.C., Barja-Fidalgo C., de Moraes J.A. Obese adipose tissue secretion induces inflammation in preadipocytes: role of toll-like receptor-4. Nutrients. 2020;12:2828. doi: 10.3390/nu12092828. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Guh D.P., Zhang W., Bansback N., Amarsi Z., Birmingham C.L., Anis A.H. The incidence of co-morbidities related to obesity and overweight: a systematic review and meta-analysis. BMC Public Health. 2009;9:88. doi: 10.1186/1471-2458-9-88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kolb H. Obese visceral fat tissue inflammation: from protective to detrimental? BMC Med. 2022;20:494. doi: 10.1186/s12916-022-02672-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Engin A. Reappraisal of adipose tissue inflammation in obesity. Adv Exp Med Biol. 2024;1460:297–327. doi: 10.1007/978-3-031-63657-8_10. [DOI] [PubMed] [Google Scholar]
- 11.Emmerich S.D., Fryar C.D., Stierman B., Ogden C.L. Obesity and severe obesity prevalence in adults: united States, August 2021–August 2023. NCHS Data Brief. 2024;508:1–8. doi: 10.15620/cdc/159281. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Xiao N., Ding Y., Cui B., et al. Navigating obesity: a comprehensive review of epidemiology, pathophysiology, complications and management strategies. Innov Med. 2024;2 [Google Scholar]
- 13.Holstege H., Beker N., Dijkstra T., et al. The 100-plus study of cognitively healthy centenarians: rationale, design and cohort description. Eur J Epidemiol. 2018;33:1229–1249. doi: 10.1007/s10654-018-0451-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ludbrook J. Outlying observations and missing values: how should they be handled? Clin Exp Pharmacol Physiol. 2008;35:670–678. doi: 10.1111/j.1440-1681.2007.04860.x. [DOI] [PubMed] [Google Scholar]
- 15.Mäkinen V.P., Ala-Korpela M. Influence of age and sex on longitudinal metabolic profiles and body weight trajectories in the UK biobank. Int J Epidemiol. 2024;53 doi: 10.1093/ije/dyae055. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Fransson E.I., Batty G.D., Tabák A.G., et al. Association between change in body composition and change in inflammatory markers: an 11-year follow-up in the whitehall II study. J Clin Endocrinol Metab. 2010;95:5370–5374. doi: 10.1210/jc.2010-0730. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Xing X., Zhu S., Zhou G., Ma Y., Wang H. Childhood obesity and allergic rhinitis: a meta-analysis. Biomol Biomed. 2026;26:774–787. doi: 10.17305/bb.2025.12982. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Yeo B.S.Y., Guan E.J., Ng K., et al. Association of abnormal body weight and allergic Rhinitis—A systematic review and meta-analysis. Clin Exp Allergy. 2025;55:142–165. doi: 10.1111/cea.14604. [DOI] [PubMed] [Google Scholar]
- 19.Sybilski A.J., Raciborski F., Lipiec A., et al. Obesity – a risk factor for asthma, but not for atopic dermatitis, allergic rhinitis and sensitization. Public Health Nutr. 2015;18:530–536. doi: 10.1017/S1368980014000676. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Han Y.Y., Forno E., Gogna M., Celedón J.C. Obesity and rhinitis in a nationwide study of children and adults in the United States. J Allergy Clin Immunol. 2016;137:1460–1465. doi: 10.1016/j.jaci.2015.12.1307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Akyuz M., Ozdemir H., Ceylan A., Guler T., Artac H. The effect of central obesity in children with allergic and non-allergic rhinitis. Asthma Allergy Immunol. 2023:183–191. [Google Scholar]
- 22.Cazzoletti L., Ferrari M., Olivieri M., et al. The gender, age and risk factor distribution differs in self-reported allergic and non-allergic rhinitis: a cross-sectional population-based study. Allergy Asthma Clin Immunol. 2015;11:36. doi: 10.1186/s13223-015-0101-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Kilpeläinen M., Terho E.O., Helenius H., Koskenvuo M. Body mass index and physical activity in relation to asthma and atopic diseases in young adults. Respir Med. 2006;100:1518–1525. doi: 10.1016/j.rmed.2006.01.011. [DOI] [PubMed] [Google Scholar]
- 24.Bays H.E., Kirkpatrick C.F., Maki K.C., et al. Obesity, dyslipidemia, and cardiovascular disease: a joint expert review from the obesity medicine association and the national lipid association 2024. J Clin Lipidol. 2024;18:e320–e350. doi: 10.1016/j.jacl.2024.04.001. [DOI] [PubMed] [Google Scholar]
- 25.Kuniyoshi Y., Tsujimoto Y., Banno M., Taito S., Ariie T., Kimoto T. Association of obesity or metabolic syndrome with various allergic diseases: an overview of reviews. Obes Rev. 2025;26 doi: 10.1111/obr.13862. [DOI] [PubMed] [Google Scholar]
- 26.Rodriguez-Coira J., Villaseñor A., Izquierdo E., et al. The importance of metabolism for immune homeostasis in allergic diseases. Front Immunol. 2021;12 doi: 10.3389/fimmu.2021.692004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Kwon M.J., An J., Yu J.W., et al. Associations between metabolic syndrome and allergic diseases a nationwide study in Korea and literature review. Sci Rep. 2026;16 doi: 10.1038/s41598-026-41559-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data that support the findings of this study are available from the health examination database of the Affiliated Hospital of Guangdong Medical University but restrictions apply to the availability of these data, which were used under ethical approval for the current study, and so are not publicly available.