Long-term dynamic effect of body mass index on adverse cardiovascular outcomes with targeted maximum likelihood estimation method: result from the KNOW-CKD study

Abstract

The relationship between body mass index (BMI) and cardiovascular outcomes in chronic kidney disease (CKD) remains controversial. While the obesity paradox has been observed in this population, the impact of longitudinal BMI patterns over time is not well established. In this study, we investigated the association between BMI trajectories and the risk of cardiovascular events (CVEs) using data from 1061 patients enrolled in the Korean Cohort Study for Outcome in Patients With Chronic Kidney Disease (KNOW-CKD). BMI was categorized as high (≥ 23 kg/m²) or low (< 23 kg/m²), and assessed at three time points over a 7-year period. We applied targeted maximum likelihood estimation (TMLE) and marginal structural models (MSMs) to adjust for time-varying covariates, including estimated glomerular filtration rate, blood pressure, hemoglobin, albumin, and C-reactive protein, as well as baseline demographic and clinical characteristics. Patients with persistently high BMI had a significantly reduced risk of CVEs (relative risk 0.279; 95% CI 0.143–0.546; P < 0.001) compared to those with consistently low BMI. Transient increases or decreases in BMI did not show the same benefit. Our findings suggest that sustained high BMI may be protective against cardiovascular events in patients with CKD, challenging current weight management recommendations and supporting individualized approaches based on long-term risk profiles.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-45135-7.

Introduction

Over the past few decades, extensive research has explored the relationship between obesity and adverse clinical outcomes. Obesity is widely recognized as a major risk factor for cardiovascular disease, metabolic disorders, and the development of chronic kidney disease (CKD) in the general population^1–5. Paradoxically, however, numerous studies have indicated that obesity and overweight may be linked to improved survival in patients with chronic diseases, such as congestive heart failure, chronic obstructive pulmonary disease, and rheumatoid arthritis^6–8. This phenomenon, known as the “obesity paradox,” has also been consistently observed in patients with end-stage kidney disease (ESKD) undergoing hemodialysis or peritoneal dialysis^9–12, as well as in CKD patients who do not yet require dialysis^13,14. However, some studies conducted in CKD patients have shown inconsistent results, so there remains ongoing controversy regarding the obesity paradox^15–17.

Body mass index (BMI) is the most commonly used measure for assessing obesity in both clinical practice and population studies. In addition to indicating body size, BMI also reflects nutritional status, which is often associated with conditions like protein-energy wasting or inflammation^18–20. BMI can fluctuate over time, and the cardiovascular risk or survival outcomes may differ among patients depending on their weight patterns. Those who consistently maintain a high or low body weight, those who steadily gain or lose weight, and those who experience weight fluctuations may have different health outcomes. In fact, high variability in BMI has been linked to increased cardiovascular risk and mortality in the general population²¹. In patients with coronary artery disease, body weight fluctuations have been linked to a higher risk of cardiovascular events and mortality²². Moreover, significant changes in body weight, such as extreme weight gain or loss, have also been associated with an elevated risk of cardiovascular mortality in patients with diabetes and those at high cardiovascular risk²³. In patients with CKD, relying on a single baseline BMI measurement may be insufficient to capture the dynamic nature of body weight change, given that change in BMI over time may reflect distinct pathophysiologic states, such as inflammation, sarcopenia, or progressive metabolic derangement. Therefore, using BMI from a single time point may not adequately capture its association with long-term outcomes, especially in CKD patients who experience more dynamic BMI changes due to chronic inflammation and malnutrition. To accurately assess BMI’s impact on clinical outcomes, changes over time should be considered. Most previous studies only used baseline BMI, leading to potential bias as BMI evolves and is influenced by time-dependent factors. Evaluating long-term BMI patterns enables a more accurate assessment of cumulative exposure and more faithfully reflects the prognostic impact of weight change on long-term outcomes. Incorporating dynamic BMI effects into cardiovascular risk assessment is therefore important for informing clinical management in patients with CKD. This study aims to evaluate the relationship between BMI changes and cardiovascular outcomes in CKD patients, analyzing whether consistent high or low BMI, gradual changes, or transitions between BMI categories differently affect cardiovascular outcomes using baseline and time-varying covariates.

Methods

Study population

The KoreaN Cohort Study for Outcome in Patients with Chronic Kidney Disease (KNOW-CKD) is a nationwide, prospective observational cohort study that enrolled adults with CKD stages G1–G5 (non-dialysis) from diverse etiologies. The study’s design, methods, and protocols have been thoroughly detailed in previous publications²⁴. A total of 2,238 patients aged 20 to 75 were enrolled in the KNOW-CKD study between 2011 and 2016 (NCT01630486). The study adhered to the Declaration of Helsinki, and ethical approval was obtained from each institution. All participants provided written informed consent.

Data collection and measurements

At enrollment, baseline demographic data including sex, age, smoking status, medical history, and medication use were collected. Body weight and height were measured to calculate BMI (kg/m²). Blood and urine samples were collected after overnight fasting, with serum creatinine and proteinuria levels sent to the central laboratory of KNOW-CKD for analysis. Other biochemical tests were conducted at each participating institution. Serum creatinine was measured using the isotope dilution mass spectrometry method, and estimated glomerular filtrationrate (eGFR) was calculated using the CKD Epidemiology Collaboration Creatinine Equation²⁵. Proteinuria was quantified by measuring the urine protein-to-creatinine ratio (UPCR) from random midstream urine samples. Participants were followed up regularly according to the study protocol and closely monitored for the occurrence of outcome events as defined by the study. Those lost to follow-up were considered censored at their last recorded visit date.

Variables

The primary outcome of the study was extended major adverse cardiovascular events (eMACEs), which included a range of cardiovascular events such as acute myocardial infarction, hospitalization for unstable angina or heart failure, percutaneous coronary interventions, coronary artery bypass grafting, ischemic or hemorrhagic stroke, symptomatic arrhythmias requiring hospitalization, peripheral arterial disease, and other cardiovascular events necessitating hospitalization or intervention. The participants were followed up until March 31, 2022. The eMACEs outcome was not arbitrarily defined for the present study, but rather represents a pre-specified and standardized cardiovascular endpoint within the KNOW-CKD cohort. In the KNOW-CKD cohort, individual cardiovascular event rates are relatively low. To address this, eMACEs was developed as extended composite cardiovascular events. This outcome definition and ascertainment process has been consistently applied across multiple studies using the KNOW-CKD cohort and has supported several peer-reviewed publications^26–28.

The time-varying exposure variable was the individual’s BMI at enrollment, 3 years, and 7 years post-enrollment. Patterns of BMI change were classified into four scenarios for the short-term period (baseline to 3 years)—high–high, high–low, low–high, low–low—and six scenarios for the long-term period (baseline, 3 years, and 7 years)—high–high–high, low–high–high, low–low–high, high–low–high, low–high–low, low–low–low.

Baseline covariates included age, sex, smoking status, diabetes, cardiovascular disease, and UPCR levels. Time-varying covariates were systolic blood pressure (SBP), eGFR, hemoglobin, serum albumin, and high-sensitivity C-reactive protein (hsCRP) levels, measured at the 3- and 7-year follow-ups.

Statistical analyses

Continuous variables were presented as mean ± standard deviation, and categorical variables were reported as frequencies and percentages. A BMI cutoff value of 23.0 kg/m² was used to categorize participants into low and high BMI groups, following the World Health Organization (WHO) BMI classification for Asian adults²⁹. We assessed the independent association of fixed BMI levels at three time points (baseline, 3 years, and 7 years after enrollment) with the risk of eMACEs using Cox proportional hazard models. The association between baseline BMI and eMACE risk was evaluated using Cox Model 1, adjusted for age, sex, smoking status, diabetes, cardiovascular disease, urine proteinuria, baseline SBP, eGFR, hemoglobin, serum albumin, and hsCRP levels. Cox Model 2 assessed the relationship between BMI at 3 years and eMACE risk in participants who remained event-free for up to 3 years. Cox Model 3 evaluated BMI at 7 years in participants without events for up to 7 years.

We used Targeted Maximum Likelihood Estimation (TMLE) to assess the dynamic impact of BMI on eMACEs and infer causality in longitudinal data. Causality is inferred when the observed outcome (e.g., eMACEs under high BMI) differs from the counterfactual outcome (e.g., eMACEs under low BMI) for the same individual. Since counterfactual outcomes are unobserved, TMLE estimates them using predicted values, enabling more accurate causal inference³⁰. The outcome model included time-invariant covariates, prior outcomes, exposures (high BMI), and time-varying covariates, generating conditional expectations of eMACEs. Two additional models accounted for BMI (propensity score) and loss to follow-up. This iterative modeling mitigates bias from time-varying confounders^31,32. Using the same covariates across models (doubly robust property) further reduces bias³². Final comparisons yielded a causal risk ratio, with inferences drawn from the efficient influence curve equation.

TMLE was applied to five distinct BMI status scenarios (Scenario #1: High–High–High; Scenario #2: Low–High–High; Scenario #3: Low–Low–High; Scenario #4: High–Low–High; Scenario #5: Low–High–Low) compared to a counterfactual pattern (Low–Low–Low)^33,34. Estimation utilized a machine learning approach via the Superlearner algorithm, which combined various estimators to determine the best weighted combination through cross-validation, thus addressing bias from model misspecification³⁵.

To evaluate the effects of time-varying BMI on eMACEs while adjusting for time-dependent confounders, a marginal structural model (MSM) was applied using stabilized inverse probability of treatment weights (IPTW) for BMI exposure categories and inverse probability of censoring weights (IPCW)^36–38. Multinomial logistic regression models were fitted at each time point from baseline to the end of follow-up to obtain IPTW and IPCW based on previous BMI levels, time-invariant, and time-varying covariates. Stabilized weights were calculated by multiplying the unstabilized weights by the probability of treatment (for IPTW) and censoring (for IPCW), estimated using logistic regression models. All statistical analyses were conducted using R software with the ltmle package, with statistical significance set at a two-sided p-value of < 0.05.

Results

Baseline characteristics of the study population

Of the 2238 participants enrolled in the cohort, 1061 were included in the baseline analysis after excluding participants without baseline BMI data or with incomplete covariate information. For the 7-year follow-up analysis, after further excluding participants with missing longitudinal BMI or time-varying covariate data, a total of 456 participants remained in the final analytic sample (Fig. 1). The mean age of the cohort was 53.9 ± 12.0 years, with 59.8% of participants being male. The mean baseline BMI was 24.6 ± 3.4 kg/m², and the mean eGFR was 58.4 ± 28.8 mL/min/1.73 m². At the time of enrollment, 306 participants (28.8%) had diabetes, and 150 (14.1%) had a history of cardiovascular disease.

Fig. 1.

Flow diagram of study cohort: Among 2238 enrolled participants, 1061 were included in the primary analysis after excluding individuals with missing baseline BMI or covariate data within the first 3 years of follow-up. For the 7-year follow-up analysis, 456 participants with complete longitudinal BMI and time-varying covariate data were included.

Baseline characteristics categorized by BMI are detailed in Table 1. Patients with high BMI were older and more likely to have diabetes, a history of cardiovascular disease, higher SBP, and UPCR, while exhibiting lower eGFR levels compared to those with low BMI.

Table 1.

Baseline characteristics by BMI category.

	Total	BMI ≥ 23 kg/m2	BMI < 23 kg/m2
	(N = 1061)	(N = 732)	(N = 329)
Age, year	53.9 ± 12.0	55.2 ± 11.4	50.8 ± 12.8
Male gender, n (%)	634 (59.8%)	473 (64.6%)	161 (48.9%)
BMI, kg/m2	24.6 ± 3.4	26.2 ± 2.7	21.0 ± 1.5
Smoking, n (%)	490 (46.2%)	369 (50.4%)	121 (36.8%)
Diabetes, n (%)	306 (28.8%)	242 (33.1%)	64 (19.5%)
Cardiovascular disease, n (%)	150 (14.1%)	116 (15.8%)	34 (10.3%)
SBP, mmHg	126.2 ± 14.8	127.2 ± 14.7	124.2 ± 14.7
eGFR, ml/min/1.73m2	58.4 ± 28.8	56.6 ± 27.1	62.4 ± 32.0
UPCR, g/g	0.4 (0.1, 1.)	0.4 (0.1, 1.0)	0.3 (0.1, 0.8)
RAS blocker, n (%)	911 (85.9%)	647 (88.4%)	264 (80.2%)
Statin, n (%)	553 (52.1%)	427 (58.3%)	126 (38.3%)
Hemoblobin, g/dl	13.2 ± 1.8	13.3 ± 1.9	12.8 ± 1.6
Albumin, g/dl	4.3 ± 0.3	4.3 ± 0.3	4.3 ± 0.3
Total cholesterol, mg/dl	174.2 ± 36.4	173.8 ± 37.	175.1 ± 35.1
hs-CRP, mg/dl	0.6 (0.2, 1.6)	0.8 (0.3, 1.9)	0.4 (0.1, 1.1)
Calcium, mg/dl	9.2 ± 0.4	9.2 ± 0.5	9.2 ± 0.4
Phosphorus, mg/dl	3.6 ± 0.6	3.6 ± 0.5	3.6 ± 0.6

BMI, body mass index; SBP, systolic blood pressure; eGFR, estimated glomerular filtration rate; UPCR, urine protein-to-creatinine ratio; RAS, renin-angiotensin-aldosterone system; hs-CRP, high-sensitivity C-reactive protein.

BMI and risk of eMACEs by conventional cox regression analyses

During a median follow-up of 8.8 years, the primary outcome occurred in 73 patients. Of the 1061 participants, 456 had BMI data measured at three time points (baseline, 3 years, and 7 years after enrollment). We conducted Cox regression analyses to assess the association between fixed BMI levels at specific time points and the risk of extended major adverse cardiovascular events (eMACEs), using time-invariant baseline covariates.

Table 2 presents the adjusted hazard ratios (HRs) for eMACEs corresponding to high BMI categories at different time points. In model 1, only baseline BMI was included, while model 2 incorporated both baseline and three-year BMI categories. Model 3 utilized data from patients with BMI measured at all three time points.

Table 2.

Association between BMI and eMACEs from conventional Cox regression analyses.

	Model 1 (n = 1,061)		Model 2 (n = 1,033)		Model 3 (n = 352)
	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
BMI ≥ 23 kg/m2 at baseline	0.786 (0.467−1.323)	0.365	1.348 (0.572–3.175)	0.495	0.147 (0.005–4.313)	0.266
BMI ≥ 23 kg/m2 at 3 year			0.386 (0.169–0.884)	0.024	0.447 (0.016–12.684)	0.584
BMI ≥ 23 kg/m2 at 7 year					1.618 (0.219–11.966)	0.637

Model 1: adjusted for age, sex, smoking, diabetes, cardiovascular disease, UPCR(g/g), and baseline SBP, eGFR, Hb, Alb, hsCPR.

Model 2: model 1 plus BMI at 3 year.

Model 3: model 2 plus BMI at 7 year.

BMI, body mass index; HR, hazard ratio; CI, confidence interval; UPCR, urine protein-to-creatinine ratio; SBP, systolic blood pressure; eGFR, estimated glomerular filtration rate; Hb, hemoglobin; Alb, albumin; hsCRP, high-sensitivity C-reactive protein.

Conventional Cox regression analyses revealed no significant associations between fixed BMI levels at specific times and the risk of eMACEs, except for the three-year BMI in model 2, where high BMI at three years was linked to a decreased risk of eMACEs (HR 0.386; 95% CI, 0.169–0.884; P = 0.024).

BMI and risk of eMACEs by TMLE and MSM analyses

Table 3 presents the results from the TMLE analyses. Patients with sustained high BMI over the seven-year follow-up had a significantly reduced risk of eMACEs compared to those with sustained low BMI (relative risk [RR] 0.279; 95% confidence interval [CI] 0.143–0.546; P < 0.001). However, no significant risk reduction was observed in patients who transitioned from low to high BMI, either early or late, compared to those with sustained low BMI (RR 0.319; 95% CI 0.001–136.890; P = 0.712 and RR 0.988; 95% CI 0.296–3.298; P = 0.985, respectively).

Table 3.

Association between BMI and eMACEs from TMLE.

BMI	RR	95% CI	P
High–High–High vs. Low–Low–Low	0.279	0.143–0.546	< 0.001
Low–High–High vs. Low–Low–Low	0.319	0.001–136.890	0.712
Low–Low–High vs. Low–Low–Low	0.988	0.296–3.298	0.985
High–Low–High vs. Low–Low–Low	0.610	0.067–5.585	0.662
High–High–Low vs. Low–Low–Low	0.076	0.0002–28.438	0.395

Time-varying exposure variable was change in BMI: High BMI, BMI ≥ 23 kg/m²; Low BMI, BMI < 23 kg/m².

Time points: baseline, 3-year, 7-year.

Baseline covariates: age, sex, smoking, diabetes, cardiovascular disease and UPCR (g/g).

Time-varying covariates: SBP, eGFR, and Hb/Albumin/CRP.

BMI, body mass index; RR, relative risk; CI, confidence interval; UPCR, urine protein-to-creatinine ratio; SBP, systolic blood pressure; eGFR, estimated glomerular filtration rate; Hb, hemoglobin; Alb, albumin; hsCRP, high-sensitivity C-reactive protein.

In the MSM analyses, the HR for cumulative high BMI frequency regarding eMACEs was 0.925 (95% CI 1.185–0.663; P = 0.645) for up to three years and 0.562 (95% CI 0.406–0.779; P = 0.001) for up to 7 years (Fig. 2). These findings suggest that a significant reduction in the risk of eMACEs is observed in patients with sustained high BMI over a long duration, but not in those with short-term changes.

Fig. 2.

Adjusted HRs with 95% CIs for cardiovascular outcomes according to cumulative exposure to high BMI, based on conventional Cox regression and marginal structural model analyses. (A) Conventional Cox proportional hazards regression analyses showing the association between high BMI and cardiovascular outcomes based on BMI status measured at a single time point. The x-axis indicates the time point at which BMI was assessed (baseline, 3 years, and 7 years of follow-up). (B) Marginal structural model (MSM) analyses incorporating longitudinal BMI measurements to account for cumulative exposure to high BMI over time. “0–3Y” represents MSM results using BMI data measured at baseline (0 years) and at 3 years, whereas “0–3–7Y” represents MSM results using BMI data measured at baseline, 3 years, and 7 years. All models were adjusted for relevant baseline and time-varying covariates.

Sensitivity analyses

To evaluate the robustness of the observed associations, we performed conventional cox regression and group-based trajectory modeling. To address concerns regarding potential information loss form dichotomizing BMI, we performed additional analyses modeling baseline BMI using quartile-based categorization. In the conventional Cox regression analyses with the first quartile as the reference, higher BMI quartiles were generally associated with lower risk of eMACEs. Although these associations did not reach statistical significance in the 3-year follow-up analysis (Supplementary Table S1), they became statistically significant over the 7 years follow-up period (HR 0.221, 95% CI 0.080–0.611; P = 0.004; Supplementary Table S2).

Additionally, we performed group-based trajectory modeling using repeated BMI measurements at multiple time points as a sensitivity analysis. Using BMI measurements obtained at baseline, 6 months, and annually up to 3 years, three distinct BMI trajectories were identified and classified as sustained low, intermediate, and sustained high BMI trajectories. When the sustained low trajectory group was used as the reference, the sustained high BMI trajectory group exhibited a lower risk of eMACEs, although this association did not reach conventional statistical significance (HR 0.549, 95% CI 0.297–1.013; P = 0.055; Supplementary Table S3). In an extended trajectory analysis incorporating BMI measurements from baseline to 7 years, three similar trajectory groups were identified. In this longer-term model, both the intermediate (HR 0.180, 95% CI 0.071–0.460; P < 0.001; Supplementary Table S2) and sustained high BMI trajectory groups (HR 0.177, 95% CI 0.068–0.461; P < 0.001; Supplementary Table S4) were associated with a significantly lower risk of eMACEs compared with the sustained low trajectory group. These findings were consistent with the primary TMLE and MSM analyses, supporting the robustness of the observed association between sustained higher BMI and reduced cardiovascular risk over long-term follow-up.

Discussion

In this prospective cohort study of patients with chronic kidney disease (CKD), we examined the longitudinal association between dynamic body mass index (BMI) patterns and the risk of extended major adverse cardiovascular events (eMACEs). We found that patients who persistently maintained a high BMI over time had a significantly lower risk of eMACEs compared with those who consistently remained at a low BMI. In contrast, patients who transitioned from low to high BMI did not experience a corresponding reduction in cardiovascular risk, regardless of the timing of weight gain. Marginal structural model (MSM) analyses further demonstrated that an inverse association between BMI and cardiovascular outcomes was not evident during the initial three years of follow-up but became apparent when the exposure window was extended to seven years. These findings indicate that cardiovascular benefit in CKD is associated not simply with attaining a higher BMI at a single time point, but with maintaining a higher BMI consistently over the long term, underscoring the potential importance of avoiding sustained underweight status.

Our findings align with the obesity paradox reported in patients with end-stage kidney disease and non-dialysis CKD, in which higher BMI has been associated with improved survival and cardiovascular outcomes^{9,10,13,14,39–42}. However, most prior studies relied on a single baseline BMI measurement, which does not capture longitudinal weight changes that may be clinically relevant. By applying MSM and targeted maximum likelihood estimation (TMLE) to account for time-varying BMI and confounders, our study extends previous work by explicitly modeling BMI as a longitudinal exposure and demonstrates that sustained higher BMI, rather than short-term BMI change, is associated with improved cardiovascular outcomes⁴³. Notably, transitions from low to high BMI did not reduce cardiovascular risk compared with remaining at low BMI, highlighting the importance of long-term BMI stability.

Emerging evidence supports the adverse prognostic implications of weight variability in CKD. A recent study in predialysis CKD patients reported that greater BMI variability was associated with worse outcomes, with both weight gain and weight loss conferring higher risk than weight stability⁴⁴. These observations are concordant with our findings and may explain why BMI transitions did not translate into cardiovascular benefit, whereas sustained BMI status—particularly sustained high BMI—was protective. Differences across studies may reflect variations in study populations, analytic approaches, and follow-up duration, but collectively suggest that stability of body weight plays a critical role in cardiovascular prognosis.

Evidence from non-nephrology populations further reinforces this interpretation. In patients with type 2 diabetes and established cardiovascular disease, weight loss has been associated with increased cardiovascular risk, whereas weight gain predicted improved survival⁴⁵. Similarly, extreme weight gain or loss has been linked to higher cardiovascular mortality and heart failure hospitalization²³. A large population-based study examining BMI changes and sudden cardiac arrest reported that sustained underweight was associated with the highest risk, and recovery from underweight did not fully mitigate this excess risk⁴⁶. Together, these findings across diverse populations support the concept that sustained BMI status, rather than transient or reactive weight change, is a key determinant of long-term cardiovascular outcomes.

Several biological mechanisms may underlie the protective association between sustained higher BMI and cardiovascular outcomes in CKD. Higher BMI may reflect greater metabolic reserve, improved nutritional status, and resistance to catabolic stress, all of which may enhance tolerance to chronic inflammation and metabolic burden. Within the mechanistic framework proposed by Manta et al.⁴⁷, obesity-related hemodynamic, inflammatory, and neurohormonal pathways may coexist with protective factors against wasting and frailty. In CKD, low BMI often reflects protein–energy wasting, sarcopenia, and systemic inflammation, conditions that are strongly associated with cardiovascular events. Thus, the lower cardiovascular risk observed among patients with persistently high BMI may reflect a balance between obesity-related cardiometabolic stress and protection against wasting-related vulnerability, rather than a simple linear relationship between adiposity and cardiovascular risk. Additional contributing mechanisms may include favorable adipokine profiles^48,49, endotoxin-neutralizing effects of higher lipoprotein levels⁵⁰, and a phenotype of metabolically healthy obesity characterized by preserved cardiorespiratory fitness^51,52.

Several limitations of this study should be acknowledged. BMI is a crude surrogate for body composition and does not distinguish lean mass from adiposity, although both may differentially influence outcomes. Body weight in CKD may also fluctuate due to fluid retention associated with declining renal function or heart failure, which may not reflect true changes in nutritional status. Although we adjusted for relevant time-varying covariates, residual confounding related to unmeasured volume status cannot be excluded. The high BMI category encompassed both overweight and obese individuals, potentially masking heterogeneity in cardiovascular risk across finer BMI strata, and intentional versus unintentional weight change could not be distinguished. As an observational study, residual confounding remains possible despite the use of advanced causal inference methods. In addition, the distinction between pre-existing low BMI and CKD-related weight loss could not be assessed due to the lack of longitudinal BMI data prior to CKD onset. However, to reduce bias from disease-related weight loss and reverse causation, we used TMLE/MSM and adjusted for tim-varying indicators of CKD severity and catabolic or inflammatory status that are associated with CKD-related weight loss. Another potential consideration is the influence of sodium-glucose cotransporter 2 (SGLT2) inhibitors, which are known to induce modest weight loss while reducing the risk of heart failure. However, given the enrollment era, reimbursement restrictions of South Korea, and the low proportion of diabetic patients (28.8%), confounding or effect modification by SGLT2 inhibitor is unlikely to explain the association observed in this study. Beyond these limitations, sample size reduction over long-term follow-up and stratification into multiple BMI change patterns likely limited statistical power to detect modest associations for less prevalent transitions. Finally, the exclusive inclusion of a Korean CKD cohort may limit generalizability to other racial and ethnic populations.

In conclusion, sustained high BMI was associated with a lower risk of eMACEs among patients with CKD, whereas BMI transitions from low to high did not confer cardiovascular benefit. These associations were evident only with long-term exposure, emphasizing the importance of longitudinal weight stability rather than short-term weight change. Our findings suggest that regular monitoring and long-term maintenance of body weight may be relevant considerations in cardiovascular risk assessment and management in CKD.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceived and designed the experiments: J.Y.J.; data acquisition: Y.J.O., J.K., S.S., S.W.K., Y.H.K., K.O., W.C., Y.Y.H., and J.Y.J.; data analysis/interpretation: Y.J.O. and J.Y.J.; manuscript writing: Y.J.O. and J.Y.J.; counsel and advice: Y.Y.H. and J.K,; reviewing of the draft: Y.J.O, J.K., S.S., S.W.K., Y.H.K., K.O., W.C., Y.Y.H., and J.Y.J. Each author provided important intellectual content by presenting and solving questions about the accuracy or integrity of all parts of the work. All authors approved the final version of the manuscript.

Funding

This work was supported by the National Institutes of Health (NIH) research projects (2025E110100) and the Research Program funded by the Korea Disease Control and Prevention Agency (2011E3300300, 2012E3301100, 2013E3301600, 2013E3301601, 2013E3301602, 2016E3300200, 2016E3300201, 2016E3300202, 2019E320100, 2019E320101, 2019E320102, and 2022-11-007).

Data availability

The data underlying this article cannot be publicly shared due to the privacy of the study participants. However, the data can be made available upon reasonable request to the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.