Established risk prediction models for the incidence of a low lean tissue index in patients with peritoneal dialysis

Figures & data

Table 1. Comparison of general clinical data between the two groups.

Characteristics	Total (n = 104)	Low LTI (n = 49)	Normal LTI (n = 55)	p Value
Age, years ($\overline{x}$±s)	64.93 ± 10.78	63.16 ± 9.13	66.50 ± 11.93	0.12
Male, n (%)	61 (58.65)	27 (55.10)	34 (61.82)	0.49
PD length (month)	31.12 (17.19–53.83)	39.53 (23.15–57.97)	29.37 (13.27–52.07)	0.15
BMI (kg/m^2)	24.70 (21.10–26.75)	22.80 (20.90–25.50)	25.60 (22.90–27.80)	0.01
BSA (m^2, $\overline{x}$±s)	1.71 ± 0.18	1.68 ± 0.18	1.73 ± 0.19	0.17
Complications, n (%)
DM	51 (49.04)	25 (51.02)	26 (47.27)	0.70
CVD	28 (26.92)	17 (34.69)	11 (20.00)	0.09
HBP	81 (77.88)	37 (75.51)	44 (80.00)	0.58
Blood pressure (mmHg)
SBP ($\overline{x}$±s)	148.15 ± 22.66	148.16 ± 24.83	148.15 ± 20.76	0.99
DBP	81.00 (76.00–85.75)	81.00 (78.50–89.00)	81.00 (75.00–84.00)	0.10
SGA rating
Grade A	85 (81.73)	35 (71.43)	50 (90.91)	0.01
Grade B	19 (18.27)	14 (28.57)	5 (9.09)
CCI	3.00 (2.00–4.00)	4.00 (2.00–4.00	3.00 (2.00–4.00)	0.21
Hb (g/L, $\overline{x}$±s)	104.37 ± 16.21	104.55 ± 18.15	104.20 ± 14.42	0.91
RDW (%)	13.60 (12.73–14.40)	13.90 (13.05–14.90)	13.00 (12.50–13.90)	0.01
sAlb (g/L)	33.00 (30.00–36.00)	33.00 (29.00–36.00)	33.50 (30.75–37.25)	0.61
Pre-alb (mg/L, $\overline{x}$±s)	307.83 ± 75.98	293.42 ± 66.61	320.66 ± 81.93	0.07
Glu (mmol/L)	6.80 (5.70–9.20)	7.20 (5.65–9.95)	6.50 (5.70–8.60)	0.55
sCr (μmol/L, $\overline{x}$±s)	879.25 ± 282.33	880.79 ± 281.23	877.88 ± 285.89	0.96
TG (mmol/L)	1.72 (1.08–2.10)	1.49 (1.05–2.09)	1.79 (1.15–2.12)	0.30
TC (mmol/L, $\overline{x}$±s)	4.54 ± 1.27	4.50 ± 1.33	4.59 ± 1.23	0.72
HDL-C (mmol/L)	0.94 (0.82–1.15)	0.93 (0.80–1.13)	0.97 (0.82–1.16)	0.55
LDL-C (mmol/L)	2.55 (1.79–3.40)	2.62 (1.70–3.28)	2.51 (1.96–3.44)	0.94
hs-CRP (mg/L)	1.17 (0.50–4.13)	1.03 (0.50–5.80)	1.19 (0.50–3.79)	0.96
IL-6 (ng/L, $\overline{x}$±s)	31.50 ± 7.71	33.93 ± 7.21	29.30 ± 7.55	0.01
Adjusted Ca (mmol/L)	2.45 (2.38–2.54)	2.48 (2.42–2.60)	2.44 (2.34–2.54)	0.07
P (mmol/L, $\overline{x}$±s)	1.44 ± 0.37	1.44 ± 0.39	1.44 ± 0.35	0.97
iPTH (pg/ml)	240.70 (128.38–396.00)	212.45 (111.60–452.25)	249.30 (141.28–378.53)	0.93

BMI: body mass index; BSA: body surface area; DM: diabetes mellitus; CVD: cardiovascular disease; HBP: high blood pressure; SGA: subject global assessment; CCI: Charlson comorbidity index; Hb: hemoglobin; RDW: red cell distribution width; sAlb: serum albumin; Glu: glucose; sCr: creatinine; TG: triglyceride; TC: total cholesterol; HDL-C: high density lipoprotein cholesterol; LDL-C: low density lipoprotein cholesterol; hs-CRP: high-sensitivity C-reactive protein; IL-6: interleukin 6; Ca: calcium; P: phosphor; iPTH: intact parathyroid.

Data are presented as the Md (P25, P75), unless indicated otherwise. The boldface fronts mean p vaule considered statistically significant.

Table 2. Comparison of PD-related indicators in patients with different LTIs.

	Total (n = 104)	Low LTI (n = 49)	Normal LTI (n = 55)	p Value
Glucose exposure (g/day)	130.00 (110.00–140.00)	140.00 (110.00–155.00)	120.00 (90.00–140.00)	0.17
Total Kt/V	1.65 (1.38–1.97)	1.64 (1.35–2.00)	1.68 (1.47–1.97)	0.56
Total CCr (mL/min)	46.09 (38.39–65.96)	44.66 (36.92–56.91)	50.76 (38.49–75.09)	0.18
RRF	0.72 (0–3.45)	0 (0–2.32)	1.06 (0–4.02)	0.11
nPCR [g/ (kg·day)]	0.67 (0.60–0.78)	0.66 (0.61–0.75)	0.68 (0.57–0.80)	0.60
4 h D/Pcr	0.58 (0.51–0.67)	0.58 (0.49–0.67)	0.61 (0.53–0.68)	0.19
24 h urine(ml)	300.00 (0–675.00)	0 (0–500.00)	400.00 (0–1000.00)	0.08
Peritoneal transport function, n (%)
H	5 (4.81)	3 (6.12)	2 (3.64)	0.28
HA	23 (22.12)	10 (20.41)	13 (23.64)
LA	20 (19.23)	13 (26.53)	7 (12.73)
L	56 (53.84)	23 (46.94)	33 (60.00)
OH (L)	3.00 (1.93–4.68)	3.80 (2.10–4.80)	2.80 (1.80–3.70)	0.04

Kt/V: urea clearance index; CCR: creatinine clearance rate; RRF: residual renal function; PCR: protein catabolic rate; H: high transport; HA: high average transport; LA: low average transport; L: low transport; OH: overhydration.

Data are represented as Md (P25, P75), unless indicated otherwise.

Table 3. Logistic regression model analysis results of risk factors associated with a low LTI in PD patients.

Variables	Univariate analysis			Multivariate analysis
	OR	95% CI	p Value	OR	95% CI	p Value
Age (years)	0.97	0.94–1.01	0.12	–	–	–
Male	0.79	0.35–1.69	0.49	–	–	–
PD length	1.01	0.99–1.03	0.12	–	–	–
BMI (kg/m^2)	0.83	0.73–0.94	0.01	0.79	0.68–0.91	0.01
DM	1.16	0.54–2.51	0.70	–	–	–
SGA rating	4.00	1.32–12.12	0.01	6.33	1.59–25.30	0.01
DBP	1.03	0.99–1.07	0.12	–	–	–
Hb (g/L)	1.00	0.98–1.03	0.91	–	–	–
sAlb (g/L)	0.98	0.90–1.06	0.59	–	–	–
Pre-alb (mg/L)	0.99	0.99–1.00	0.07	–	–	–
hs-CRP (mg/L)	1.01	0.99–1.04	0.32	–	–	–
IL-6 (ng/L)	1.09	1.03–1.16	0.01	1.10	1.02–1.18	0.01
RDW (%)	1.90	1.29–2.81	0.01	1.87	1.18–2.97	0.01
RRF	0.88	0.75–1.03	0.10	–	–	–
Adjusted Ca (mmol/L)	5.75	0.79–41.86	0.08
OH(L)	1.24	1.01–1.53	0.04	–	–	–
Kt/V (L/wk/1.73 m^2)	0.82	0.36–1.85	0.63

BMI: body mass index; BSA: body surface area; DM: diabetes mellitus; SGA: subject global assessment; DBP: diastolic blood pressure; Hb: hemoglobin; sAlb: serum albumin; hs-CRP: high-sensitivity C-reactive protein; IL-6: interleukin 6; RDW: red cell distribution width; RRF: residual renal function; OH: overhydration; Kt/V: urea clearance index.

A multifactor logistic regression analysis model was adjusted for gender, age, pro-alb, calcium, and OH.

Figure 1. The ROC curves of each machine learning model.

Table 4. The metrics of different machine learning models.

Model	Accuracy	Precision	Recall	F1 score	AUC
DT	1.000000	1.000000	1.000000	1.000000	1.000000
SVM	0.562500	0.777778	0.368421	0.500000	0.712551
RF	0.812500	0.882353	0.789474	0.833333	0.896761
GBDT	1.000000	1.000000	1.000000	1.000000	1.000000
GBM	0.968750	0.950000	1.000000	0.974359	1.000000

DT: decision tree; SVM: support vector machine; RF: random forest; GBDT: gradient boosted decision tree; GBM: gradient boosted machine; AUC: area under the curve. The boldface fonts mean that DT, GBDT and GBM show the best performance in terms of AUC values.

Figure 2. The feature importance of machine learning models and visual DT flow diagram. (A) The feature ranking analysis produced by the decision tree model. (B) A pruned decision tree flowchart for predicting low LTI incidence in PD patients. (C) A pruned decision tree flow chart based on nutrition-related indicators for predicting the risk stratification of low LTI incidence in PD patients. DT: decision tree; LTI: lean tissue index; PD: peritoneal dialysis.

Figure 3. Kaplan–Meier survival curves of patients in the low LTI and normal LTI groups. LTI: lean tissue index.

Table 5. Cox regression analysis of all-cause death in PD patients with low LTI.

Variables	Before correction			After correction
	HR values	95% CI	p Values	HR values	95% CI	p Values
Age (years)	1.03	0.99–1.08	0.12
Male	0.66	0.25–1.76	0.41	–	–	–
PD length (m)	1.00	0.99–1.02	0.59
CVD history	2.14	0.83–5.52	0.12
BMI (kg/m^2)	0.90	0.79–1.03	0.12
SGA rating	0.86	0.49–1.49	0.59
IL-6 (ng/L)	1.01	0.95–1.07	0.84
sAlb (g/L)	0.97	0.88–1.06	0.47
RDW (%)	0.99	0.68–1.42	0.94
Low LTI	3.17	1.13–8.90	0.03	3.14	1.12–8.80	0.03
OH (L)	1.10	0.90–1.34	0.36
Kt/V(L/wk/1.73 m^2)	0.75	0.27–2.11	0.59

PD: peritoneal dialysis; CVD: cardiovascular disease; BMI: body mass index; SGA: subject global assessment; sAlb: serum albumin; IL-6: interleukin 6; RDW: red cell distribution width; LTI: lean tissue in; OH: overhydration; Kt/V: urea clearance index.