Analysis of clinical data
A total of 604 patients who underwent DIXON surgery for rectal cancer were retrospectively collected. The incidence of AL was 7.9% (48/604), of which the incidence of AL in the model group and AL in the validation group were 7.5% (31/413) and 8.9% (17/191), respectively.
Analysis Of Patient Related Factors In The Model Group
Compared with the NAL group, there were more male patients (P = 0.008) and PCT (P = 0.034) in the AL group. Patients with HbA1c (> 6.3%) also had significantly higher AL after surgery (P = 0.036). There was no significant difference in age, smoking, Alcohol excess, BMI, history of malignant tumor, Diabetes mellitus, hypertension and ASA between AL group and NAL group.
Analysis Of Surgical Related Factors In Model Group
There were more patients with LCA non preservation (P = 0.044) and with more than 2 staples (P = 0.008) in the AL group compared with the NAL group. There was no significant difference in operation time, intraoperative blood loss, surgical approach and DS between the AL group and the NAL group.
Analysis Of Tumor Related Factors In The Model Group
Compared with the NAL group, the AL group had more patients with TD (< 7cm) (P = 0.000). There was no significant difference in pathological T stage, pathological N stage, TNM stage and tumor histological differentiation.
The clinical information and results of the univariate analysis of the patients in the model group are shown in Table 1
Information For Validation Group
The incidence of AL in 191 patients undergoing DIXON surgery for rectal cancer was 8.9% (17/191). Patients with TD (< 7cm) were 50.8% (97/191) which was slightly higher than 48.9% (202/413) of the model group. Patients with DS were 21.5% (41/191) which was slightly lower than 21.8% (90/413) of the model group. Among them, there are 174 cases of NAL and 17 cases of AL. The information of the patients in the validation group was shown in the Table 2.
Construction Of Rareal Model
Transformations that convert continuous variables to categorical variables
For the convenience of constructing the RAREAL model, the glycated hemoglobin (HbA1c) variable, which is a continuous variable, was converted into a dichotomous variable (> 6.3%, ≤ 6.3%)[> 45mmol/mol,≤45mmol/mol] according to the clinical test reference range of HbA1c (3.9%-6.3%)[30 mmol/mol − 45mmol/mol] .
Model Building
AL after DIXON operation for rectal cancer was used as the dependent variable, and age, gender, smoking, alcohol excess, BMI, history of malignant tumor, hypertension, PCT, diabetes mellitus, HbA1c, ASA, operation time, intraoperative blood loss, LCA preservation, number of staples, surgical approach, DS, TD, pathological T stage, pathological N stage, TNM stage, and tumor histological differentiation were used as independent variables. Univariate analysis showed that male (P = 0.008), HbA1c (> 6.3%) (P = 0.036), PCT (P = 0.034), LCA non preservation (P = 0.044), TD (< 7cm) (P = 0.044) and more than 2 staples fired during surgery (P = 0.008) were risk factors for AL. Further logistic regression analysis found that HbA1c (> 6.3%) (odds ratio (OR) 7.831; 95% confidence interval (CI) 1.943–31.557; P = 0.004), LCA non preservation (OR 7.035; 95% CI1.968-25.142; P = 0.003), TD (< 7cm) (OR 14.246; 95% CI 3.504–57.913; P = 0.000) were independent of AL risk factors (Table 3). Multivariate equation: Y=-7.620 + 2.058X1 + 1.951X2 + 2.656X3 (Y: AL, X1: HbA1c (> 6.3%), X2: LCA non preservation, X3: TD (< 7cm)).
The partial regression coefficients of the equation were 2.058, 1.951, and 2.656, respectively. The partial regression coefficients were rounded to the corresponding scores of the indicators to establish the RAREAL model for AL. The score for HbA1c (> 6.3%) was 2, for LCA non preservation was 2, and for TD (< 7cm) was 3.
Diagnostic effect of RAREAL model analyzed by ROC curve and comparison of ROC curve of two groups
When the maximum Youden index of the model group was 0.415, the corresponding best cut-off point of RAREAL was 4.5, the AUC (95%CI) was 0.764 (0.660–0.868), and its diagnostic sensitivity and specificity were 0.419, 0.995. When the RAREAL score was 4.5 as the diagnostic cutoff in the validation group, the AUC (95%CI) was 0.757 (0.610–0.904), and its corresponding sensitivity and specificity were 0.471 and 0.989, respectively. There was no significant difference in the AUC of the ROC between the two groups (P = 0.939) (Table 4)
Factors
|
AL (n = 31)
|
NAL (n = 382)
|
P
|
Table 1
characteristics of patients in the model group
Patient related factors
|
Age(years, \(\stackrel{-}{x}\pm s\))
|
60.6 ± 8.4
|
59.8 ± 11.9
|
0.624
|
Gender, n = No (%)
|
|
|
0.008
|
Female
|
4(12.9%)
|
139(36.4%)
|
|
Male
|
27(87.1%)
|
243(63.6%)
|
|
Smoking, n = No (%)
|
|
|
0.556
|
Yes
|
12(38.7%)
|
128(33.5%)
|
|
No
|
19(61.3%)
|
254(66.5%)
|
|
Alcohol excess, n = No (%)
|
|
|
0.796
|
Yes
|
10(32.3%)
|
132(34.6%)
|
|
No
|
21(67.7%)
|
250(65.4%)
|
|
History of malignant tumor, n = No (%)
|
|
|
0.821
|
Yes
|
28(90.3%)
|
340(89.0%)
|
|
No
|
3(9.7%)
|
42(11.0%)
|
|
BMI(kg/m2, \(\stackrel{-}{x}\pm s\))
|
22.3 ± 2.1
|
22.6 ± 2.7
|
0.503
|
HbA1c, n = No (%)
|
|
|
0.036
|
≤ 6.3%
|
26(83.9%)
|
361(94.5%)
|
|
> 6.3%
|
5(16.1%)
|
21(5.5%)
|
|
Diabetes mellitus, n = No (%)
|
|
|
0.861
|
Yes
|
6(19.4%)
|
79(20.7%)
|
|
No
|
25(80.6%)
|
303(79.3%)
|
|
Hypertension, n = No (%)
|
|
|
0.404
|
Yes
|
20(64.5%)
|
217(56.8%)
|
|
No
|
11(35.5%)
|
165(43.2%)
|
|
PCT, n = No (%)
|
|
|
0.034
|
Yes
|
9(29.0%)
|
53(13.9%)
|
|
No
|
22(71.0%)
|
329(86.1%)
|
|
ASA grade, n = No (%)
|
|
|
0.224
|
3
|
4(12.9%)
|
45(11.8%)
|
|
2
|
8(25.8%)
|
146(38.2%)
|
|
1
|
19(61.3%)
|
191(50.0%)
|
|
Surgical related factors
|
|
|
|
Operation time(min)
|
|
|
0.734
|
Median(IQR)
|
180(160, 210)
|
180(165, 211)
|
|
Intraoperative blood loss (ml)
|
|
|
0.739
|
Median(IQR)
|
40 (30,70)
|
40 (40,50)
|
|
LCA preservation, n = No (%)
|
|
0.044
|
Yes
|
23(74.2%)
|
337(88.2%)
|
|
No
|
8(25.8%)
|
45(11.8%)
|
|
Number of staples fired, n = No (%)
|
|
|
0.008
|
1 or 2
|
20(64.5%)
|
319(83.5%)
|
|
༞2
|
11(35.5%)
|
63(16.5%)
|
|
Surgical approach, n = No (%)
|
|
|
0.199
|
Laparoscopic
|
24(77.4%)
|
326(85.7%)
|
|
Open
|
7(22.6%)
|
56(14.3%)
|
|
Diverting stoma, n = No (%)
|
|
|
0.427
|
Yes
|
5(16.1%)
|
85(22.3%)
|
|
No
|
26(83.9%)
|
297(77.7%)
|
|
Tumor related factors
|
|
|
|
TD, n = No (%)
|
|
|
0.000
|
≥ 7cm
|
5(16.1%)
|
206(53.9%)
|
|
< 7cm
|
26(83.9%)
|
176(46.1%)
|
|
Pathological T stage, n = No (%)
|
|
|
0.145
|
Tis, T1, T2,
|
8(25.8%)
|
149(39.0%)
|
|
T3 and T4
|
23(74.2%)
|
233(61.0%)
|
|
Pathological N stage, n = No (%)
|
|
|
0.312
|
N0
|
6(19.4%)
|
106(27.7%)
|
|
N1 and N2
|
25(80.6%)
|
276(72.3%)
|
|
TNM stage, n = No (%)
|
|
|
0.303
|
Ⅰ
|
4(12.9%)
|
43(11.3%)
|
|
Ⅱ
|
8(25.8%)
|
152(39.8%)
|
|
Ⅲ
|
19(61.3%)
|
187(49.0%)
|
|
Tumor histological differentiation, n = No (%)
|
|
0.473
|
High
|
14(45.2%)
|
173(45.3%)
|
|
Moderate
|
11(35.5%)
|
103(27.0%)
|
|
Low
|
6(19.4%)
|
106(27.7%)
|
|
AL anastomotic leakage, NAL non anastomotic leakage, BMI body mass index, HbA1c glycated hemoglobin, PCT preoperative chemoradiotherapy, ASA American Society of Anesthesiologists physical status classification, IQR interquartile range, LCA left colic artery, TD tumor distance from the anal margin.
Factors
|
AL(n = 17)
|
NAL(n = 174)
|
P
|
Table 2
characteristics of patients in the validation group
Patient related factors
|
Age(years, \(\stackrel{-}{x}\pm s\))
|
59.0 ± 9.2
|
57.4 ± 12.2
|
0.594
|
Gender, n = No (%)
|
|
|
0.097
|
Female
|
3(17.6%)
|
66(37.9%)
|
|
Male
|
14(82.4%)
|
108(62.1%)
|
|
Smoking, n = No (%)
|
|
|
0.867
|
Yes
|
6(35.3%)
|
65(37.4%)
|
|
No
|
11(64.7%)
|
109(62.6%)
|
|
Alcohol excess, n = No (%)
|
|
|
0.488
|
Yes
|
5(29.4%)
|
66(37.9%)
|
|
No
|
12(70.6%)
|
108(62.1%)
|
|
History of malignant tumor, n = No (%)
|
|
|
0.446
|
Yes
|
1(5.9%)
|
21(12.1%)
|
|
No
|
16(94.1%)
|
153(87.9%)
|
|
BMI(kg/m2, \(\stackrel{-}{x}\pm s\))
|
23.2 ± 1.8
|
23.9 ± 2.2
|
0.253
|
HbA1c, n = No (%)
|
|
|
0.096
|
≤ 6.3%
|
14(82.4%)
|
164(94.3%)
|
|
> 6.3%
|
3(17.6%)
|
10(5.7%)
|
|
Diabetes mellitus, n = No (%)
|
|
|
0.916
|
Yes
|
4(23.5%)
|
39(22.4%)
|
|
No
|
13(76.5%)
|
135(77.6%)
|
|
Hypertension, n = No (%)
|
|
|
0.793
|
Yes
|
7(41.2%)
|
66(37.9%)
|
|
No
|
10(58.8%)
|
108(62.1%)
|
|
PCT, n = No (%)
|
|
|
0.170
|
Yes
|
5(29.4%)
|
27(15.5%)
|
|
No
|
12(70.6%)
|
147(84.5%)
|
|
ASA grade, n = No (%)
|
|
|
0.958
|
3
|
2(11.8%)
|
23(13.2%)
|
|
2
|
5(29.4%)
|
55(31.6%)
|
|
1
|
10(58.8%)
|
96(55.2%)
|
|
Surgical related factors
|
Operation time (min)
|
|
|
0.682
|
Median(IQR)
|
180(172.50,210)
|
180(165,215)
|
|
Intraoperative blood loss (ml)
|
|
|
0.810
|
Median(IQR)
|
40(25,90)
|
40(40,50)
|
|
LCA preservation, n = No (%)
|
|
|
0.071
|
Yes
|
12(70.6%)
|
152(87.4%)
|
|
No
|
5(29.4%)
|
22(12.6%)
|
|
Number of staples fired, n = No (%)
|
|
|
0.872
|
1 or 2
|
13(76.5%)
|
136(78.2%)
|
|
༞2
|
4(23.5%)
|
38(21.8%)
|
|
Surgical approach, n = No (%)
|
|
|
0.454
|
Laparoscopic
|
14(82.4%)
|
153(87.9%)
|
|
Open
|
3(17.6%)
|
21(12.1%)
|
|
Diverting stoma, n = No (%)
|
|
|
0.535
|
Yes
|
2(11.8%)
|
39(22.4%)
|
|
No
|
15(88.2%)
|
135(77.6%)
|
|
Tumor related factors
|
TD, n = No (%)
|
|
|
0.006
|
≥ 7cm
|
3(17.6%)
|
91(52.3%)
|
|
< 7cm
|
14(82.4%)
|
83(47.7%)
|
|
Pathological T stage, n = No (%)
|
|
|
0.595
|
Tis, T1, T2,
|
6(35.3%)
|
73(42.0%)
|
|
T3 and T4
|
11(64.7%)
|
101(58.0%)
|
|
Pathological N stage, n = No (%)
|
|
|
0.403
|
N0
|
3(17.6%)
|
53(30.5%)
|
|
N1 and N2
|
14(82.4%)
|
121(69.5%)
|
|
TNM stage, n = No (%)
|
|
|
0.979
|
Ⅰ
|
2(11.8%)
|
19(10.9%)
|
|
Ⅱ
|
8(47.1%)
|
79(45.4%)
|
|
Ⅲ
|
7(41.2%)
|
76(43.7%)
|
|
Tumor histological differentiation, n = No (%)
|
|
0.814
|
High
|
9(52.9%)
|
81(46.6%)
|
|
Moderate
|
5(29.4%)
|
51(29.3%)
|
|
Low
|
3(17.6%)
|
42(24.1%)
|
|
AL anastomotic leakage, NAL non anastomotic leakage, BMI body mass index, HbA1c glycated hemoglobin, PCT preoperative chemoradiotherapy, ASA American Society of Anesthesiologists physical status classification, IQR interquartile range, LCA left colic artery, TDtumor distance from the anal margin.
Factor
|
B
|
P值
|
OR
|
95% CI
|
Table 3
Multivariate logistic regression analysis.
Male
|
0.624
|
0.292
|
1.866
|
0.584–5.955
|
HbA1c (> 6.3%)
|
2.058
|
0.004
|
7.831
|
1.943–31.557
|
PCT
|
0.506
|
0.363
|
1.659
|
0.557–4.939
|
LCA non preservation
|
1.951
|
0.003
|
7.035
|
1.968–25.142
|
TD
|
2.656
|
0.000
|
14.246
|
3.504–57.913
|
Number of staples fired
|
-0.784
|
0.118
|
0.456
|
0.171–1.220
|
HbA1c glycated hemoglobin, PCT preoperative chemoradiotherapy, LCA left colic artery, TDtumor distance from the anal margin.
Group
|
Youden index
|
Best cut-off point
|
AUC (95%CI)
|
Sensitivity
|
Specificity
|
SE
|
Z
|
P
|
Table 4
Evaluation of the consistency of ROC curve between the model group and the validation group
Model group
|
0.415
|
4.5
|
0.764(0.660–0.868)
|
0.419
|
0.995
|
0.053
|
|
|
Validation group
|
|
|
0.757(0.610–0.904)
|
0.471
|
0.989
|
0.075
|
|
|
|
|
|
|
|
|
|
0.076
|
0.939
|
SE standard error