- Independent risk factors for ICC
1.1. Patient characteristics
Of the 232 patients, 177 (76.3%) underwent LND, and 71 (40.1%) had at least one LNM. A minimum of 6 lymph nodes were dissected in 66 patients (37.3%). There were some differences in baseline between the two groups (Table 1). POD time, operative time, intraoperative blood loss, ALT, cirrhosis rate, intraoperative blood transfusion and T stage were all higher in the LND group.
1.2.Prognostic factor analysis
The median follow-up time was 19 months. The total median OS was 28 months (95 confidence interval [CI] 19.025-36.975 months), and the 5-year OS rate was 32.7%. The total median DFS was 12 months, and the 5-year DFS rate was 28.0%. The 5-year OS of the nLND group and LND group was 50.6% vs. 25.7%, and the median OS was 61 vs. 21 months (95% CI 34.891- 87.109 months vs. 13.765-28.235 months, P < 0.01). The DFS of the nLND group was better than that of the LND group; the 5-year DFS was 44.2% vs. 22.3%, respectively, and the median DFS was 34 vs. 12 months, respectively (95% CI 17.612-50.388 months vs 9.039-14.961 months, P < 0.01).
LND, POD time >9 days, CEA >7 ng/mL, CA 19-9 >75 U/mL, operation time >235 min, blood loss >300 mL, intraoperative blood transfusion, tumor size >6 cm, multi-disease, positive margin, vascular invasion, nerve invasion, T stage >T1, ALT >50U/L, AST >40U/L, TBIL >21μmol/L, and ALB <35g/L were risk factors for OS in univariate analysis. These variables were included in the multivariate analysis; CA 19-9 >75 U/mL, CEA >7 ng/mL, positive margin, and T stage >T1 were statistically significant (Table 2).
LND, POD time >9 d, CEA >7 ng/mL, CA 19-9 >75 U/mL, operation time >235 min, blood loss>300 mL, tumor size >6 cm, multi-disease, positive margin, differentiation, vascular invasion, nerve invasion, and T stage >T1 were the primary factors influencing DFS. CEA and differentiation were independent risk factors in multivariate analysis (P <0.05) (Table 2).
1.3.Effects of LND on prognosis
In univariate analysis, we found significant difference in survival between LND and nLND, with 5-year OS of 50.6% vs. 25.7% in nLND and LND groups. Median OS was 61 vs. 21 months (95% CI 34.891-87.109 vs. 13.765-28.235 months, P <0.01). DFS in nLND group was better than that in LND group. 5-year DFS was 44.2% vs 22.3%, and median DFS was 34 vs 12 months (95%CI 17.612-50.388 vs 9.039-14.961 months, P <0.01). This was clearly at odds with the expected results. However, in multivariate analysis, the survival difference did not exist, suggesting that the difference between nLND and LND may be due to the difference on baseline.
To further verify whether LND affected the prognosis of ICC, we used inverse probability of treatment weighting (IPTW) to reduce confounding factors to the greatest extent. We included factors that differed between groups and were likely to influence prognosis. Finally, T staging, CA199, nerve invasion, vascular invasion, and cirrhosis were included in the adjustment (Figure 1). The standardized mean differences (SMD) showed that IPTW (weighted) effectively balanced the between-group differences. The HR of OS was 1.04(95CI 0.61-1.77, P= 0.89). The HR of DFS was 1.38 (0.84-2.28, P=0.21). There was no statistical significance in prognosis between the two groups. (Figure 2a-d).
- Effect of LNM on prognosis
The current AJCC staging system defines the N stage in N0 (non-LNM) and N1(at least 1 LNM), which is too simplistic. Further, we investigated the effect of LNM number on prognosis. We only conducted the study in the LND group since the situation of LNM was unknown in nLND group. The median OS was 48 months (95%CI 28.209-67.791) with a 5-year survival rate of 38.9% in LNM negative group, while the median OS was 13 months (95%CI 9.613-16.387) with a 5-year survival rate of 4.1% (P=0.00) in LNM-positive group. The median DFS of the LNM-negative group was 19.5 months (95%CI 8.759-30.241), with a 5-year DFS of 29.6%, and the median DFS of the LNM-positive group was 7 months (95%CI 4.121-9.879), with a 5-year DFS of 0%, P=0.00 (Figure 3a-b). Then, the grouping was defined as LNM 1-3 (N=52) and LNM >3 (N=19). The results showed that LNM >3 had worse OS (HR 1.310 [95CI 0.657-2.612], P=0.44). The median DFS of LNM group was 7 months, and the 2-year DFS was 10.8%. The DFS of LNM >3 was worse, but still there was no statistical significance (HR 1.249 [95CI 0.629-2.477], P=0.53). (Figure 3c-d)
Considering of LND number may affect the evaluation of LNM, we conducted a stratified analysis on LND >6 or not. In LND =<6 group, LNM >3 was not a prognostic risk factor (OS: HR 0.372[95CI 0.050-2.739], P=0.33; DFS: HR 1.278[95CI 0.383-4.262], P=0.69). In LND>6 group, there was no difference on DFS (HR 1.249 [95CI 0.629-2.477], P=0.9). The OS of LNM >3 group was worse (HR 1.892 [95CI 0.805-4.447], P=0.14), but still not statistically significant even though the P value was small. Risk table showed that all LNM >3 patients relapsed or died within 2 years (Figure 4a-d).
- The adjuvant therapy for nLND patients
The indications of adjuvant therapy for ICC are controversial, but the idea adjuvant chemotherapy should be performed for patients with LNM has been accepted by most doctors. The effect of adjuvant therapy for nLND patients is not clear. As we had studied above, adjuvant chemotherapy was not a independent prognostic factor in the whole cohort. 87(37.5%) patients in the cohort underwent adjuvant chemotherapy. LND and nLND groups have a similar proportion of adjuvant therapy (37.3% vs. 38.2%). In nLND group, adjuvant chemotherapy showed a protective effect on both OS (HR 0.458[95CI 0.202-1.041, P=0.62]) and DFS (HR 0.553 [95CI 0.250-1.222, P=0.14]), although this was not statistically significant.
Then we difined adjuvant therapy as having adjuvant chemotherapy, radiotherapy and both. 103 (44.4%) patients had adjuvant therapy. The proportion of LND and nLND groups was 44.6% vs 43.6%. In nLND group, the OS of adjuvant therapy was significant better than non-adjuvant therapy (HR 0.293 [95CI 0.128-0.669, P=0.004]). The same conclusion applied to DFS (HR 0.391 [95CI 0.176-0.867, P=0.02]). (Figure 5a-d)