Page 94 - 《南京医科大学学报》自然科学版2026年第2期
P. 94
第46卷第2期
·250 · 南 京 医 科 大 学 学 报 2026年2月
(续表1)
2
Indicator Survival group(n=339) Deceased group(n=252) t/χ /Z P
>10-15 26(7.67) 028(11.11)
>15 03(0.88) 02(0.79)
BMI[kg/m ,n(%)] 12.908 0.005
2
<18.5 15(4.42) 027(10.71)
18.5- < 24 262(77.29) 189(75.00)
24.0- < 28 050(14.75) 034(13.49)
≥28 12(3.54) 02(0.79)
CCI score[M(P25,P75 )] 5.00(4.00,6.00) 6.00(5.00,7.00) 9.773 <0.001
Hemoglobin[g/L,M(P25,P75 )] 112.00(100.00,122.00) 105.50(97.00,114.75) 4.358 <0.001
Albumin[g/L,M(P25,P75 )] 39.60(37.50,41.80) 36.20(33.40,37.98) 11.801 <0.001
Prealbumin[g/L,M(P25,P75 )] 0.29(0.25,0.34) 0.22(0.16,0.29) 9.507 <0.001
Bun/creatinine ratio[M(P25,P75 )] 6.79(5.69,8.33) 7.34(5.74,9.78) 2.681 0.007
Ca×P product[M(P25,P75 )] 48.57(40.07,59.97) 49.07(41.54,61.97) 0.607 0.544
Serum sodium[mmol/L,M(P25,P75 )] 137.80(136.10,140.00) 137.70(135.60,140.68) 0.243 0.808
iPTH[pg/mL,n(%)] 21.579 <0.001
<300 204(60.18) 103(40.87)
300-600 096(28.32) 106(42.06)
≥600 039(11.50) 043(17.06)
C⁃reactive protein[mg/L,M(P25,P75 )] 2.40(1.20,6.60) 4.50(2.03,12.30) 6.092 <0.001
KT/V[n(%)] 0.181 0.914
<1.2 071(20.94) 050(19.84)
1.2-<1.4 091(26.84) 071(28.17)
≥1.4 177(52.21) 131(51.98)
AVF:arteriovenous fistula;TCC:tunneled cuffed catheter;AVG:arteriovenous graft;CCI:Charlson comorbidity index;iPTH:intact parathyroid hormone total.
2.2 预测因子筛选 专及以上学历、独自居住、带涤纶套中心静脉导管
将全部预设变量纳入 LASSO 回归进行变量筛 (tunneled cuffed catheter,TCC)、糖尿病肾病、CCI 评
选,当λ.1se=0.027 时模型表现最佳,此时共得到 12 分、白蛋白、前白蛋白、尿素氮肌酐比值及 iPTH<
个预测因子,包括年龄、性别、初中及以下学历、大 300 pg/mL(图1)。
A 25 21 20 15 8 5 B 26 25 25 25 25 22 21 21 20 20 20 20 17 15 13 10 9 8 8 8 7 5 5 5 2 1
1.4
6
2 1.3
8 2 1.2
18
17
Coefficients 23 Binomial deviance 1.1
22
23
0
21
20
19
28
16
1.0
-2
0.9
5
-4
0.8
24
-7 -6 -5 -4 -3 -2 -7 -6 -5 -4 -3 -2
log(λ) log(λ)
A:Distribution plot of the LASSO regression coefficients of the variables(n=26). B:Use of 10⁃fold cross⁃validation to determine the regular coeffi⁃
cient(lambda,λ)of the Lasso model. The left dashed line corresponded to the minimum λ(λ.min)and the right dashed line corresponded to the λ.1se.
图1 基于LASSO回归的预测因子选择
Figure 1 Predictor selection based on LASSO regression

