We retrospectively established a PD cohort between July 2013 and February 2024 at the Department of Nephrology, Tianjin First Central Hospital, China. Patients aged 18–80 years who initiated PD treatment for at least 3 months were enrolled in this study. All patients were treated according to the high-quality goal-directed peritoneal dialysis strategy of the International Society for Peritoneal Dialysis Practice Recommendations at our hospital before enrollment and during the observation period. Patients were excluded if they had malignant tumors, were receiving chemotherapy, or had insufficient clinical information. Patients who were transferred to another hospital, switched to hemodialysis, or received a kidney transplant during the study period were treated as censored and included in the survival analysis. This study has been granted an exemption from requiring written informed consent because it is a retrospective review of medical records. Prior to analysis, the patient records and information were anonymized and de-identified. This study was conducted at the PD Center of the Nephrology Department at our hospital, approved by the institutional review board (IRB) of the hospital (2015009S), and was conducted in accordance with the principles of the Helsinki Declaration.

Data were obtained from medical records. Information was collected on sociodemographic characteristics, medical history, anthropometric measurements, and laboratory tests. PD patients in our center underwent monthly hematological and biochemistry examinations during their outpatient follow-up.

Baseline demographic and clinical information, including age, sex, original disease, height, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), diabetes mellitus (DM), hypertension history, and body mass index (BMI), was collected at the initiation of PD. SBP and DBP were measured in the morning of the outpatient visits. Diabetes was defined as a self-reported history of diabetes or FBG ≥ 7.0 mmol/L. Hypertension was defined as a self-reported history of hypertension, the use of antihypertensive medication, SBP ≥ 140 mmHg, or DBP ≥ 90 mmHg. Hypercholesterolemia was defined as a self-reported history of dyslipidemia, or TC ≥ 5.17 mmol/L. Body mass index (BMI) was calculated as weight (kg)/height (m)².

Biochemical and medication data were collected 1 month after PD onset as a baseline. Blood parameters included hemoglobin (Hb), albumin (Alb), blood urea nitrogen (BUN), creatinine (Cr), uric acid (UA), sodium (Na), potassium (K), chlorine (Cl), venous carbon dioxide (CO₂), calcium (Ca), inorganic phosphorus (P), fasting blood glucose (FBG), triglyceride, total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), high-sensitivity C-reactive protein (hsCRP), and intact parathyroid hormone (iPTH). All blood samples were measured and analyzed via commercially available kits and an automated analyzer (Cobas 8000 or e411 for iPTH, Roche Diagnostics Ltd. for others, Germany). The serum albumin concentration was measured via the bromocresol green method. The average 24-h urinary urea and creatinine clearances were calculated to assess residual renal function (RRF). The total PD treatment volume, total ultrafiltration volume, total effluent dialysate volume, and daily urine volume were recorded daily. The total protein content in the effluent dialysate was measured. The protein equivalent nitrogen appearance (nPNA) was calculated and normalized to the actual weight. In accordance with previous studies (23–27), the TyG index was calculated as Ln [TG (mg/dL) × FBG (mg/dL)/2]. TyG–BMI = TyG × BMI (22). Treatment with angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers (both combined as renin-angiotensin-aldosterone system (RAAS) inhibitors, RAASi) was also recorded. PD types and associated parameters were recorded. PD patients received standard prescriptions for continuous ambulatory peritoneal dialysis (CAPD using Baxter’s Low Calcium Peritoneal Dialysis Solution [Lactate-G1.5/2.5%]) or automated peritoneal dialysis (APD using Baxter’s Peritoneal Dialysis Solution [Lactate-G1.5/2.5%]), accordingly. Twenty-four-hour urine and dialysate samples were collected 1 month after PD commencement and every month thereafter to measure dialysis adequacy, which was defined as the urea clearance index (Kt/V) and creatinine clearance (CCr), via standard methods (28). Weekly CCr (WCCr) was calculated by multiplying the 24-h Ccr by 7 days.

The primary and secondary endpoints were all-cause mortality and CVD mortality, respectively. The criteria for CVD death were death attributed to congestive heart failure, acute myocardial infarction, atherosclerotic heart disease, cardiac arrest, cardiac arrhythmia, cardiomyopathy, cerebrovascular disorders, anoxic encephalopathy, ischemic brain injury, or peripheral arterial disease (29). The cause of death was identified by a comprehensive management team consisting of junior and senior professors at our PD Center.

All patients were followed up until death, transferred to hemodialysis therapy, received a kidney transplant, transferred to another center, lost to follow-up, or until 29 February 2024.

Normally distributed continuous variables are presented as means±standard deviations (SDs), while non-normally distributed continuous variables are presented as medians and interquartile ranges. Categorical variables are presented as frequencies and percentages. All patients were divided into three groups according to TyG–BMI values using the tri-quantile method: the high group (TyG–BMI>227.15), middle group (TyG–BMI 189.57–227.15), and low group (TyG–BMI < 189.57). Comparisons of continuous variables among the three groups were performed using a one-way ANOVA or the Kruskal–Wallis H test. The chi-squared test was used to compare categorical variables. Survival analysis was performed using Cox regression to compare the effects on all-cause and cardiovascular mortality among the high, middle, and low groups. A restricted cubic spline plot with five knots (at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles) was used to explore the relationship between TyG–BMI, as a continuous variable, and the outcomes. Model 1 was univariate. Model 2 was adjusted for sex, age, original disease, DM, hypertension, and SBP. Model 3 was additionally adjusted for Hb, Alb, K, CO₂, P, total Kt/V urea, nPNA, RRF, iPTH, and RAASi. Survival curves were plotted using the Kaplan–Meier method, and comparisons between groups were performed using the log-rank test. All statistical analyses were conducted using SPSS version 22 (IBM, Japan) or the statistical package R (version 4.5.1). A p-value of less than 0.05 was considered statistically significant.

In this study, we excluded patients with missing data (n = 25), patients with malignant disease (n = 2), and patients who received chemotherapy during the follow-up period (n = 1). Therefore, 865 eligible patients were included in the current study (Figure 1). The mean age of the 865 participants was 57.28 ± 15.27 years, and 480 (55.5%) of them were male. The mean TyG–BMI of the entire study population was 212.27 ± 46.64. The clinical characteristics and laboratory data of the three groups categorized by the TyG–BMI are shown in Table 1. Patients in the high TyG–BMI group had the lowest prevalence of glomerulonephritis as the primary disease, the highest prevalence of diabetic nephropathy and hypertensive nephropathy, and the greatest number of individuals with diabetes. They also had the highest mean SBP, triglyceride, BMI, Hb, and hs-CRP levels, as well as the lowest mean total cholesterol level. In the middle group, patients were more likely to be male and had the lowest prevalence of hypertensive nephropathy as the original disease. These patients also exhibited the lowest mean triglyceride levels and the highest HDL-C levels. Patients in the low group exhibited the highest prevalence of glomerulonephritis as the original disease and the lowest prevalence of diabetic nephropathy. This group also had the fewest individuals diagnosed with diabetes. Furthermore, they presented with the lowest mean levels of SBP, BMI, Hb, HDL-C, and hs-CRP, along with the highest mean total cholesterol level.

As shown in Table 2, the majority of PD patients in this cohort used the daytime automated peritoneal dialysis (DAPD) modality (80.6%), and the median RRF was 2.8 mL/min/1.73m², the urine volume was 800 mL/day, and the nPNA was 0.93 g/kg/d. The mean total Kt/V was 1.96 ± 0.84, and the total WCCr was 70.81 ± 45.20 L/w/1.73m². Patients in the High TyG–BMI group had the highest RRF, urine volume, total WCCr, and renal WCCr, and the lowest total Kt/V, renal Kt/V, dialysate Kt/V, and nPNA. In the middle group, patients had the lowest RRF. Patients in the low group exhibited the lowest urine volume, total WCCr, and renal WCCr, while demonstrating the highest values for total Kt/V, renal Kt/V, dialysate Kt/V, and nPNA.

During the 46.6 (22.4–78.0) months of follow-up, 266 (30.75%) deaths occurred, of which 110 (41.35%) were due to CVD. When analyzed as a continuous variable for Cox regression, TyG–BMI did not affect all-cause mortality. Categorically, patients in the high and low groups were significantly associated with an increased risk of all-cause mortality compared to those in the middle group according to the univariate and multivariate analyses (Table 3). TyG–BMI was significantly associated with CVD mortality, which was analyzed as a continuous variable in all three models. Only patients in the high TyG–BMI group had an increased risk of CVD mortality compared to those in the middle group. There was no significant difference between the low group and the middle group (Table 4).

The Kaplan–Meier analysis and the log-rank test were performed among the three groups, with survival curves illustrated in Figure 2. Regarding all-cause mortality, a statistically significant difference was observed among the three groups (p = 0.0043) (Figure 2A). In the context of multiple comparisons, patients in both the high and low groups demonstrated a significantly increased risk of mortality when compared to those in the middle group. There was also a statistically significant difference in CVD mortality among the three groups (p = 0.0098) (Figure 2B). In the context of multiple comparisons, patients classified in the high group exhibited a significantly elevated risk of CVD mortality when compared to those in the middle group (p = 0.008).

Restricted cubic spline regression analysis revealed a significant U-shaped relationship between TyG–BMI and all-cause mortality. The risk of death was found to be lowest at a TyG–BMI value of 209.73, with an increasing trend observed both before and after this point (p for non-linearity = 0.002). For TyG–BMI < 209.73, the HR increases by 0.66 times for every one standard deviation decrease (95% CI [0.55–0.80]), whereas for TyG–BMI > 209.73, the HR increases by 1.36 times for every one standard deviation increase (95% CI [1.24–1.49]) (Figure 3A). No similar relationship was evident between TyG–BMI and cardiovascular mortality; instead, a J-shaped pattern was observed (non-linear p = 0.024). The risk of CVD mortality remains relatively stable until TyG–BMI reaches 206.64, after which it begins to increase rapidly (non-linear p = 0.024). When TyG–BMI is < 206.64, there is no statistically significant difference in HR for every one standard deviation reduction (95% CI [0.69–1.38]); when TyG–BMI is > 206.64, the risk increases 1.52 times for every one standard deviation increase (95% CI [1.36–1.69]) (Figure 3B).

In the sensitivity analysis, our research findings remained robust, with the exception of sex as a factor. We observed that higher TyG–BMI values were associated with increased all-cause and cardiovascular mortality in men, whereas in women, lower TyG–BMI values were linked to elevated all-cause mortality but showed no difference in cardiovascular mortality risk between the groups (p = 0.016, p = 0.046). Therefore, adjustments for comorbidities such as hypertension and diabetes were not necessary (Figure 4).

In this study, we identified an independent association between the TyG–BMI and all-cause as well as CVD mortality in a single PD center. Furthermore, we observed for the first time that the relationship between the TyG–BMI and all-cause mortality exhibited a U-shaped pattern in the population undergoing PD.

However, our study has several limitations. First, the levels of triglycerides and glucose may have been influenced by prescribed medications, which were not reported in our study, potentially introducing bias to the results. Second, data on triglycerides, glucose, and BMI were collected only once at baseline, and it remains unclear whether changes in TyG–BMI over time could impact its association with mortality. Therefore, longitudinal cohort studies are necessary to explore the persistence of the association between TyG–BMI and mortality over time. Third, we did not assess the homeostasis model assessment of insulin resistance or compare it with the TyG–BMI because of insufficient data on insulin levels during follow-up. Finally, potential residual confounding factors should be acknowledged, given that this is an observational study.