Baseline characteristics
3,312 individuals with CKD were registered in a single-center SMP-CKD cohort between July 1, 2015 and December 31, 2022. 2,518 of them were excluded for not having CKD stage 3–5, 1,609 were missing baseline lipid parameters (1,255 were missing lipid profiles, 187 were missing body composition measurements, and 131 were missing anthrometric measurements, of which 36 were missing 2 indicators) and 61 survived less than 3 months.
A total of 794 individuals (429 males and 365 females; median age 58.35 [48.50, 66.69] years with moderate to severe CKD (median eGFR 33.68 [17.55, 47.81] ml/min/1.73 m2 and median urea 10.87 [7.92, 15.75] mmol/L and median UPCR 0.87 [0.27, 2.03] mg/g) were included in the analysis. 34.01% were primary glomerulonephritides. 5.54% had hypertensive renal disease and 13.85% had diabetic nephropathy. 14.86% had other secondary nephrosis, and at least 30% are estimated to have had unknown protopathy.
82.49% of patients had a history of hypertension, and 73.68% had a history of hyperuricemia, and 42.06% were diabetic. 20.28% had a history of CVDs. Treatment with ACEI/ARB was reported in 42.32% of them. 34.36% reported the use of lipid-lowering therapy. Baseline characteristics of the study population are described in Table 1.
Table 1
Baseline Cohort Characteristics
Variables | N(%)/Mean(SD)/Median[25th,75th] | Variables | N(%)/Mean(SD)/Median[25th,75th] |
Age | | 58.35 [48.48, 66.78] | Medication | |
Gender | Male | 429(54.03) | ACEI/ARB | 336(42.32) |
Female | 365(45.97) | Other antihypertensive drugs | 432(54.41) |
Marriage | Unmarried | 40(5.04) | Hypoglycemic agents | 171(21.54) |
Married | 737(92.82) | Urate-lowering drugs | 404(50.89) |
Alcohol Consumption | No | 779(98.11) | Lipid-lowering drugs | 272(34.36) |
Yes | 15(1.89) | Folic acid tablets | 86(10.83) |
Current Smoker | No | 725(91.31) | Polysaccharide iron | 122(15.37) |
Yes | 69(8.69) | EPO | 133(16.75) |
Education | Elementary school | 118(14.90) | Hb, g/L | 119.64(21.75) |
| Junior high school | 229(28.91) | TCO2, mmol/L | 22.90 [20.78, 25.00] |
| High school | 257(32.45) | UA, µmmol/L | 427.00 [370.75, 501.75] |
| College degree or above | 188(23.74) | Urea, mmol/L | 10.87 [7.92, 15.75] |
Self-care Ability | Yes | 52(6.55) | ALB, g/L | 43.60 [40.28, 46.10] |
No | 742(93.45) | eGFR, ml/min/1.73 m2 | 33.68 [17.55, 47.81] |
Protopathy | Primary Glomerulonephritides | 270(34.01) | K+, mmol/L | 4.48 [4.18, 4.83] |
Hypertensive Renal Disease | 44(5.54) | Na+, mmol/L | 141.00 [139.00, 142.00] |
Diabetic nephropathy | 110(13.85) | Ca, mmol/L | 2.33 [2.25, 2.42] |
Others | 118(14.86) | P, mmol/L | 1.28 [1.12, 1.44] |
Unknown | 252(31.74) | UPCR, mg/g | 0.87[0.27,2.03] |
Comorbidity | | TG, mmol/L | 1.49 [1.10, 2.18] |
Hypertension management | | TC, mmol/L | 4.79 [4.00, 5.57] |
Hypertension with normal BP | 164(20.66) | HDL-C, mmol/L | 1.26 [1.00, 1.52] |
Hypertension with abnormal BP | 396(49.87) | LDL-C, mmol/L | 3.03 [2.30, 3.73] |
Hypertension with without regular assessment | 95(11.97) | TSKF, cm | 1.40 [1.00, 1.90] |
Non-hypertension | 139(17.51) | MUAC, cm | 27.00 [25.40, 29.00] |
Glu management | | BMI, kg/m2 | 22.80 [20.60, 25.03] |
Diabetes with normal Glu | 142(17.89) | BFM, kg | 15.45 [11.35, 19.90] |
Diabetes with abnormal Glu | 117(14.74) | FMI, % | 5.90 [4.30, 7.50] |
Diabetes without regular assessment | 75(9.45) | PBF, % | 26.19(8.38) |
Non-diabetes | 460(57.94) | VFA, cm2 | 68.70 [51.87, 91.63] |
UA management | | TBW, kg/L | 32.65 [27.80, 37.51] |
Hyperuricemia with normal uric acid | 198(24.94) | ICW, kg/L | 20.00 [16.90, 22.90] |
Hyperuricemia with anormal uric acid | 387(48.74) | ECW, kg/L | 12.65 [10.90, 14.60] |
Non-hyperuricemia | 206(25.95) | FFM, kg | 44.20 [37.90, 51.00] |
With history of cardiovascular disease | 161(20.28) | FFMI, % | 16.83(2.05) |
Note:Primary Glomerulonephritides included chronic nephritis, nephropathy syndrome and IgA nephropathy.Other secondary nephrosis included systemic lupus erythematosus nephritis, Henoch-Schonlein purpura,Hepatitis B virus-associated nephritis and obstructive nephropathy, etc. Blood Pressure, BP; Angiotensin converting enzyme, ACE; Angiotensin receptor blocker, ARB; Uric acid, UA; Erythropoietin, EPO; Hemoglobin, Hb; Triglyceride, TG; total cholesterol, TC; high-density lipoprotein cholesterol,HDL-C; low-density lipoprotein cholesterol, LDL-C; Total carbon dioxide, TCO2; albumin, ALB; Urine protein-to-creatinine, UPCR; Body Mass Index, BMI; triceps skinfold thickness, TSKF; mid-arm circumference, MUAC; Body Fat Mass, BFM; Fat Mass Index, FMI; Percent Body Fat, PBF; Visceral Fat Area, VFA; Total Body Water, TBW; Intracellular Water, ICW; Extracellular Water, ECW; Fat Free Mass, FFM; Fat Free Mass Index, FFMI. |
Poisson regression on time-split data
During a median follow-up of 2.03 [1.06, 3.19] years, 255 subjects reached the composite outcome. We split the follow-up duration into seven timescales. Table 2 described the survival rate of each timescale’s composite endpoint. Taking the censored samples as reference, 42.86% and 42.79% of events occurred with crude survival rates of 0.108 [0.086, 0.135] and 0.186 [0.150, 0.228], respectively, in the first and second timescale. The events (crude survival rate [95% CI]) in the 3 to 6 timescale were 25.29% (0.144 [0.104, 0.193]), 29.09% (0.197 [0.135, 0.278]), 11.67% (0.088 [0.035, 0.181]) and 3.92% (0.069 [0.008, 0.249]), respectively. No events occurred in the last timescale.
Table 2
Survial Rate of Composite Endpoint in Timescales
Follow-up Timescales | Person-years | Events n(%) | Crude Survival Rate[95%CI] |
Timeband = 1 | 182 | 78(42.86) | 0.108[0.086,0.135] |
Timeband = 2 | 215 | 92(42.79) | 0.186[0.150,0.228] |
Timeband = 3 | 174 | 44(25.29) | 0.144[0.104,0.193] |
Timeband = 4 | 110 | 32(29.09) | 0.197[0.135,0.278] |
Timeband = 5 | 60 | 7(11.67) | 0.088[0.035,0.181] |
Timeband = 6 | 51 | 2(3.92) | 0.069[0.008,0.249] |
Timeband = 7 | 2 | 0(0) | 0.000[0.000,0.835] |
Multivariable-adjusted Poisson Models |
After adjusting the offset and timescales (as factors), the univariate Poisson models in five datasets generated variables with P < 0.10 including covariates (age, marital status, Hb, TCO2, Urea, ALB, UPCR, eGFR, K+, Na+, Ca, P and ECW), circulating lipid profile (TC), manual measurements (TSKF, MUAC and BMI) and body composition (BFM, PBF, VFA and FMI). Model 1 was covariates and lipid fractions; Model 2 was covariates and manual measurements; Model 3 was covariates and body composition; Model 4 contained all variables with P < 0.10. Supplement 2 described the results of univariate Poisson regression for each dataset.
The category-based based multivariable-adjusted Poisson models were constructed with a “backward” stepwise regression. The AIC values gradually decreased across all datasets and models. This indicated that Model 4 of Dataset 5 had the best goodness of fit. Hence, the results of Datasets 1–4 are detailed in Supplement 3 and Supplemental Figs. 1–4. We displayed results from Dataset 5 in Table 3.
Model 1 showed that TC was a significant risk factor of composite outcome (1.153 [1.050, 1.266], P = 0.003; AIC = 1,656.4). No manual measurements had statistical significance in Model 2 (BMI:0.966 [0.925,1.009], P = 0.117; AIC = 1,662.2). PBF was a significant protective factor of composite outcome (0.978 [0.963, 0.994], P = 0.005; AIC = 1,657.1) in Model 3. In Model 4, both TC and PBF were independent factors with more obvious effects and better goodness of fit than Models 1 and 3 (TC: (1.158 [1.056, 1.271], P = 0.002); PBF: 0.948 [0.906, 0.992], P = 0.021; AIC = 1,651.1). The AUCs for Models 1, 2 and 3 were 0.712 [0.678, 0.746], 0.710 [0.676, 0.745] and 0.711 [0.677, 0.745], respectively. Model 4 had the largest AUC at 0.717 [0.684, 0.751] with P for ROC < 0.05 (Table 4 and Fig. 2).
Table 4
| Model 1 | Model 2 | Model 3 | Model 4 |
AIC | 1656.4 | 1662.2 | 1657.1 | 1651.1 |
AUC | 0.712[0.678,0.746] | 0.710[0.676,0.745] | 0.711[0.677,0.745] | 0.717[0.684,0.751] |
P for ROC | 0.013 | 0.027 | 0.047 | Ref |
Notes: Akaike information criterion, AIC; the area under the receiver operating curve, AUC; receiver operating curve, ROC. |
Table 5
Multivariable-adjusted Poisson Models of Sensitivity Analysis
| Model 1 | | Model 2 | |
Variables | RRs[95%CI] | P | RRs[95%CI] | P |
Age | 0.984[0.973,0.995] | 0.005 | 0.986[0.974,0.997] | 0.017 |
Urea | 1.041[1.011,1.071] | 0.006 | 1.047[1.016,1.078] | 0.003 |
UPCR | 1.239[1.178,1.304] | 0.000 | 1.243[1.178,1.313] | 0.000 |
eGFR | 0.960[0.946,0.973] | 0.000 | 0.959[0.945,0.973] | 0.000 |
Na+ | 0.923[0.870,0.980] | 0.009 | 0.929[0.874,0.987] | 0.018 |
P | 1.756[1.147,2.688] | 0.010 | 1.672[1.088,2.570] | 0.019 |
ECW | -- | -- | 0.827[0.708,0.967] | 0.017 |
BFM | -- | -- | 1.081[0.991,1.180] | 0.079 |
PBF | -- | -- | 0.928[0.862,1.000] | 0.049 |
FFMI | -- | -- | 1.187[1.022,1.379] | 0.025 |
Notes: Poisson regression with a “backward” stepwise. Model1: Age, Hb, TCO2, Urea, ALB, UPCR, eGFR, K+, Na+, Ca, P, ECW + TSKF; Model2: Age, Hb, TCO2, Urea, ALB, UPCR, eGFR, K+, Na+, Ca, P, ECW + BFM, PBF, VFA, FMI, FFMI; Model3:Age, Hb, TCO2, Urea, ALB, UPCR, eGFR, K+, Na+, Ca, P, ECW + TSKF, BFM,PBF,VFA,FMI. The model 2 and model 3 had same results. |
Subgroup analysis
Subgroup analysis with strata of sex, age (group by median), CKD stages, with or without hypertension, and with or without diabetes, were constructed on the basis of Model 4. In Fig. 3, TC was a relative risk factor for the group with age less than 58.35 years (1.170 [1.025, 1.335], P = 0.020), females (1.248 [1.082, 1.439], P = 0.002), the group with CKD stage 5 (1.185 [1.030, 1.363], P = 0.017), the group with hypertension (1.180 [1.065, 1.300], P = 0.001) and both with and without diabetes (with diabetes: 1.173 [1.023, 1.346], P = 0.022; no diabetes: 1.183 [1.036, 1.350], P = 0.013). There was an interaction effect between TC and sex (P for interaction = 0.037).
In Fig. 4, PBF was a relative protective factor of the group with age over 58.35 years (0.884 [0.822, 0.952], P = 0.001), females (0.936 [0.877,0.999], P = 0.048), the CKD stage 5 group (0.920 [0.850, 0.996], P = 0.039), the hypertension group (0.944 [0.899, 0.991], P = 0.021) and the no diabetes group (0.933 [0.878, 0.992], P = 0.027). There was an interaction effect between PBF and age (P for interaction = 0.012). No other interactions were found in subgroup analysis.
Sensitivity analysis
For sensitivity analysis, incomplete cases were dropped before repeating the Poisson regression. In the univariate Poisson regression, variables with P < 0.10 were covariates (age, Hb, TCO2, Urea, ALB, UPCR, eGFR, K+, Ca, P, ECW), manual measurement (TSKF) and body composition (BFM, PBF, VFA, FMI, FFMI). Model 1 was covariates and manual measurements; Model 2 was covariates and body composition; Model 3 contained all variables with P < 0.10. After adjusting covariates with “backward” stepwise regression, Model 2 was the same as Model 3. The details of Models 1 and 2 are displayed in Table 6 and Fig. 5. PBF was still a protective factor of composite outcome (0.928 [0.862, 1.000], P = 0.049) in Model 2. It had a smaller AIC of 1,183.5 and a larger AUC (0.696 [0.654,0.738]) than Model 1, while the P for ROC was not significant.
Table 6
Models Comparison of Sensitivity Analysis
| Model 1 | Model 2 |
AIC | 1183.5 | 1185.2 |
AUC | 0.696[0.654,0.738] | 0.698[0.664,0.733] |
P for ROC | 0.075 | Ref |
Notes: Akaike information criterion, AIC; the area under the receiver operating curve, AUC; receiver operating curve, ROC. |