Patients Baseline Characteristics
A total of 812 eligible patients with a diagnosis of ICC in the SEER database from 2011 to 2015 were randomly enrolled in the training cohort(n = 648) and internal validaiton cohort(n = 164), in addition, 136 cases of ICC patients from the West China Hospital of Sichuan University from 2013 to 2015 were enrolled as a external validation cohort. The clinical characteristics of patients in these three cohorts were summarized in Table 1. In the training cohort, 333(51.4%)patients were aged over 65 years and 321(49.5%) were male, 208(32.1%) patients underwent resection; in the internal and external validation cohort, 86(52.4%) and 63(46.3%) patients were aged over 65 years, 83(50.6%) and 88(64.7%) were male, 61(37.2%)and 53(39.0%)patients underwent resection, respecitvely.
Table 1
Baseline clinical characteristics of training cohort, internal validation cohort and external validation cohort.
Characteristic | | N (total) | Patients | p value | | p value |
Training cohort n(%) | Internal Validation cohort n(%) | | External Validation cohort n(%) | |
Total | | 948 | 648 | 164 | | 136 | |
Age (years) | ≤ 65 | 466 | 315(48.6%) | 78(47.6%) | 0.752 | 73(53.7%) | 0.457 |
| > 65 | 482 | 333(51.4%) | 86(52.4%) | | 63(46.3%) | |
Gender | Male | 492 | 321(49.5%) | 83(50.6%) | 0.273 | 88(64.7%) | 0.001* |
| Female | 456 | 327(50.5%) | 81(49.4%) | | 48(35.3%) | |
Maritalstatus | Married | 572 | 379(58.5%) | 106(64.6%) | 0.634 | 87(64.0%) | 0.478 |
| Divorce | 121 | 88(13.6%) | 18(11.0%) | | 15(11.0%) | |
| Other | 255 | 181(27.9%) | 40(24.4%) | | 34(25.0%) | |
AJCC T stage | 1–2 | 564 | 384(59.3%) | 95(57.9%) | 0.504 | 85(62.5%) | 0.483 |
| 3–4 | 384 | 264(40.7%) | 69(42.1%) | | 51(37.5%) | |
Tumor size (cm) | < 5 | 327 | 213(32.9%) | 59(36.0%) | 0.99 | 55(40.4%) | 0.054 |
| 5–10 | 324 | 227(35.0%) | 46(28.0%) | | 51(37.5%) | |
| > 10 | 297 | 208(32.1%) | 59(36.0%) | | 30(22.1%) | |
Tumor grade | I | 63 | 38(5.9%) | 11(6.7%) | 0.506 | 14(10.3%) | 0.181 |
| II | 235 | 152(23.4%) | 49(29.9%) | | 34(25.0%) | |
| III-IV | 194 | 128(19.8%) | 37(22.5%) | | 29(21.3%) | |
| Unknown | 456 | 330(50.9%) | 67(40.9%) | | 59(43.4%) | |
Tumor number | 1 | 896 | 608(93.8%) | 160(97.6%) | 0.99 | 128(94.1%) | 0.896 |
| > 1 | 52 | 40(6.2%) | 4(2.4%) | | 8(5.9%) | |
Surgery | Resection | 322 | 208(32.1%) | 61(37.2%) | 0.653 | 53(39.0%) | 0.128 |
| Partial destruction | 40 | 25(3.9%) | 7(4.3%) | | 8(5.9%) | |
| None | 586 | 415(64.0%) | 96(58.5%) | | 75(55.1%) | |
Chemotherapy | yes | 437 | 294(45.4%) | 76(46.3%) | 0.390 | 67(49.3%) | 0.408 |
| no/unknown | 511 | 354(54.6%) | 88(53.7%) | | 69(50.7%) | |
Radiation | yes | 147 | 111(17.1%) | 19(11.6%) | 0.414 | 17(12.5%) | 0.184 |
| no/unknown | 801 | 537(82.9%) | 145(88.4%) | | 119(87.5%) | |
Independent Risk Factors Of The Training Cohort
Univariate analysis showed that age, AJCC T stage, tumor size, tumor grade, tumor number, surgery ,maritalstatus, were independent factors affecting OS(p < 0.05), gender and radiation had no statistical significance on OS(p > 0.05). Then, multivariable cox regression analysis was carried out and the final results showed 5 factors with statistical significance:T stage(T3-T4:HR = 1.569,95%CI:1.302–1.892,p < 0.001), tumorsize(5-10cm: HR = 1.18, 95%CI:0.938–1.485, p = 0.065; >10cm: HR = 1.229,95%CI:0.972–1.554,p = 0.043), tumor grade(grade II: HR = 1.052, 95%CI: 0.662–1.672, p = 0.830; grade III-IV: HR = 1.595, 95%CI:1.005–2.531;p = 0.048, grade unknown: HR = 1.520, 95%CI:0.980–2.356; p = 0.062), surgery (partial destruction:HR = 0.254,95%CI:0.148–0.434,p < 0.001,resection:HR = 0.224,95%CI:0.172–0.293,p < 0.001),chemotherapy(HR = 0.629,95%CI:0.516–0.767,p < 0.001), the results of univariate and multivariate regression analysis are shown in Table 2.
Table 2: Results of univariate and multivariate Cox regression analysis of OS on training cohort.
Characteristic
|
Univariate analysis
|
Multivariate analysis
|
HR (95%CI)
|
p value
|
HR (95%CI)
|
p value
|
Age (years)
|
|
|
|
≤65
|
Reference
|
Reference
|
>65
|
1.401(1.175-1.670)
|
<0.001**
|
1.121(0.930-1.354)
|
0.231
|
Gender
|
|
|
|
Male
|
Reference
|
|
Female
|
0.931(0.782-1.109)
|
0.422
|
|
|
Maritalstatus
|
|
|
|
Other
|
Reference
|
Reference
|
Married
|
0.792(0.650-0.965)
|
0.020*
|
0.978(0.793-1.207)
|
0.839
|
Divorce
|
0.860(0.644-1.149)
|
0.309
|
0.899(0.667-1.213)
|
0.487
|
T stage
|
|
|
|
T1-T2
|
Reference
|
Reference
|
T3-T4
|
1.705(1.428-2.035)
|
<0.001**
|
1.569(1.302-1.892)
|
<0.001**
|
Tumor size (cm)
|
|
|
|
<5
|
Reference
|
Reference
|
5-10
|
1.363(1.094-1.697)
|
0.006*
|
1.18(0.938-1.485)
|
0.065
|
>10
|
1.992(1.598-2.483)
|
<0.001**
|
1.229(0.972-1.554)
|
0.043
|
Tumor grade
|
|
|
|
I
|
Reference
|
Reference
|
II
|
0.955(0.605-1.508)
|
0.844
|
1.052(0.662-1.672)
|
0.830
|
III-IV
|
1.624(1.031-2.558)
|
0.036*
|
1.595(1.005-2.531)
|
0.048*
|
Unknown
|
2.547(1.663-3.899)
|
<0.001**
|
1.520(0.980-2.356)
|
0.062
|
Tumor number
|
|
|
|
1
|
Reference
|
Reference
|
>1
|
1.797(1.202-2.687)
|
0.004*
|
1.384(0.918-2.086)
|
0.120
|
Surgery
|
|
|
|
None
|
Reference
|
Reference
|
Partial destruction
|
0.272(0.162-0.456)
|
<0.001**
|
0.254(0.148-0.434)
|
<0.001**
|
Resection
|
0.210(0.168-0.263)
|
<0.001**
|
0.224(0.172-0.293)
|
<0.001**
|
Chemotherapy
|
|
|
|
no/unknown
|
Reference
|
Reference
|
yes
|
0.888(0.745-1.059)
|
0.094
|
0.629(0.516-0.767)
|
<0.001**
|
Radiation
|
|
|
|
yes
|
Reference
|
|
no/unknown
|
0.827(0.654-1.044)
|
0.11
|
|
|
Constructing Prognostic Nomograms For Os
Based on the significant independent risk factors of multivariate analysis, a nomogram was constructed to predict the 1-,3-,5-year OS of N0M0 stage ICC patients(Fig. 2). The nomogram shows that the most affected factor for the OS is the surgery, followed by chemotherapy, AJCC T stage, tumor grade and tumor size. The total points is calculated by adding up the points corresponding to each factor, then locating it on the bottom scale to help clinicians predict 1-,3-,5-year OS for each independent N0M0 stage ICC patient. A higher total points reflects a poorer prognosis.
Discrimination And Calibration Of The Nomogram
In the training cohort ,internal and external validation cohort, the calibration plots of the OS predicting nomogram all show good agreement between the predictions and the actual observations of the 1-, 3-, and 5-year OS.(Fig. 3A-I)
The C-index values for OS prediction were 0.750(95% CI:0.731–0.769) in the training cohort, 0.803(95% CI:0.783–0.823) in the internal validation cohort, and 0.681(95% CI:0.524–0.838) in the external validation cohort, which revealed a good prognostic accuracy of the nomogram. We plotted the ROC curves for all cohorts, the AUC values of 1-, 3- and 5-year OS in the training cohort were 0.772,0.809,0.798, respectively; 0.896,0.868,0.896 in the internal validation cohort, respectively, and 0.673,0.786,0.886 in the external validation cohort, respectively, which reveals a nice discriminated ability of the normogram(Fig. 4A-I).
Stratification Of Risk Groups
From the maximum values of the Youden indexes of the ROC curves, we obtained the cutoff values of 110 based on the nomogram. ICC patients in the training cohort, the internal and external validation cohort were then divided into high-risk(total points ≥ 110) and low-risk(total points < 110) groups respectively. The Kaplan-Meier survival curves of the three cohorts were drawn, and the curves all showed that patients in the low-risk group obtained significantly better OS than patients in the high-risk group. (Fig. 5A-C).