In the current study, survival analysis was performed using different cut-off values of the CK19 expression levels, and ROC curves were drawn to compare its predictive performance. The results suggested that 0% might be the best cut-off value for defining CK19-positive HCC, and poor prognosis was associated with high expression levels of CK19. The univariate and multivariate regression analyses showed positive CK19, macrovascular invasion, increased CA125, decreased BMI, poorly differentiated Edmondson grade, more than one tumor, and increased BCLC stage were the independent risk factors, affecting postoperative OS of HCC patients. The univariate and multivariate COX and LASSO regression analyses identified the five clinicopathological factors, including CK19, BMI, CA125, Edmondson grade and tumor number, as predictive markers, and an OS prediction model was established based on these factors, which exhibited higher prediction potential. The predictive ability provided a reference to predict the individual survival outcome of CK19-positive HCC patients.
The positive CK19 expression in HCC and its clinical significance have received greater attention. However, the definition of CK19-positive HCC remains controversial. Different studies have used different cut-off values and defined positive CK19 expression in patients based on different cut-off values. In 1996, Wu et al.[21] from Hong Kong, China used 1% as the cutoff value (anti-CK19 antibody, 1:100; Dako), and the positive rate of CK19 expression was 29.3%. Uenishi et al.[22] from Japan (anti-CK19 antibody, 1:100, Dako, 2003), Durnez et al.[23] from Belgium (anti-CK19 antibody, 1:20, Dako, 2006), Yang team from China[24] (Anti-CK19 antibody, 1:100, Dako, 2008), Zhang’s group[27] (anti-CK19 antibody, 1:200, Kit-0030, Maixin, Fuzhou, China, 2021), Lee’s group from Korea[25] (anti-CK19 Antibody, 1:50, Dako, 2012), and Choi's team[26] (2018) and Kmeid et al.[28] (2021) from the United States adopted a cut-off value of 5% and showed that the corresponding CK19-positive rates were 40.85%, 16.51%, 18.67%, 36.84%, 15.70%, 25.7%, and 13.4% respectively. The Yamamoto team from Japan[32] (anti-human CK19 antibody, 1:100, Dako, 2005), Kim’s group from Korea[29] (2011), and Wei team[30] (2022) from China used 10% as the cut-off value and showed that the corresponding CK19-positive rates 11.1%, 18.2%, and 33.1%, respectively. However, Lu’s team from China[31] used an anti-CK19 antibody (1:50, Dake) (2011) and adopted 15% as the cutoff value and showed that the positive rate of CK19 was 10.1%. These results showed that the cut-off values used by different research teams were not uniform. The main reason for these differences might be that the researchers empirically defined the cut-off values, which were based on the source of the anti-CK19 antibody, the concentration of the antibody, the concentration of the experiment, conditions, etc. In general, previous studies have used 1%, 10%, and 15% as the cut-off value; however, the studies using CK19-positive (anti-CK19 antibody, concentration 1:100, Dako Company) based on 5% as the cut-off value are more. In the current study, the authoritative Abcam anti-CK19 antibody (Anti-Cytokeratin 19 antibody [KRT19/800], 1:500, Cat#ab220193, USA)[33] was used for the identification of CK19 expression. The previously used threshold as well as the optimal cut-off value of CK19 expression were selected for the survival analysis of patients, and the ROC curve was drawn to determine the optimal cut-off value. It was observed that the 1% and 2% CK19-positive rates might not be well-distinguished when judged by pathologists, and the predictive potentials of the two were almost similar in the current study. Therefore, it was suggested that if even a small number of CK19-positive cells were found on a pathological section during clinical practice, the patient might be considered a CK19-positive HCC patient. This phenomenon might also be explained by the existence of high heterogeneity of HCC tumor cells. In this study, our results showed that there was no significant difference in AUC values between the 0% and 2% cutoff values, and considering that using 0% as the cutoff value is more convenient for researchers, as long as staining is found, it can be defined as CK19 positive, which can greatly reduce the missed diagnosis rate and improve diagnostic efficiency. Therefore, the current study indicated that 0% might be the best cut-off value for defining CK19-positive HCC.
Several studies have shown that CK19-positive HCC patients have unique clinical and pathological features. Zhang et al.[27] reported that among the CK19-positive patients, the number of patients with age < 45 years and AFP ≥ 400 ng/mL was more as compared to those among CK19-negative patients, which was consistent with the findings in the current study. However, the correlation between BMI and HCC is controversial. In a large prospective cohort study, the research group of Professor Peter from Epidemiology of the American Cancer Society found that high BMI, large waist circumference, and type 2 diabetes were correlated with an increased risk of liver cancer[34]. The meta-analysis of the prospective cohort studies also showed a certain positive correlation between BMI and liver cancer[35, 36]. However, the current study showed that HCC patients with low BMI had a higher risk of CK19-positive expression, and low BMI was an independent risk factor for HCC. Haeryoung Kim et al[29] showed that CK19-positive HCC cells showed poor differentiation as compared to CK19-negative HCC cells, and the proteins related to epithelial-mesenchymal transition (EMT) and invasiveness were more frequently expressed in CK19-positive HCC. Another study showed that the CK19 expression in HCC was positively correlated with tumor enlargement, decreased tumor differentiation, increased metastasis, and MVI[18]. These findings were also consistent with the findings of the current study.
In the current study, the univariate and multivariate Cox and LASSO-cox regression analyses were performed, and CK19, CA125, BMI, Edmondson grade and tumor number were screened to successfully construct an HCC prognosis prediction model with good predictive ability. Similar to the current study, numerous previous studies have shown that CK19 is an adverse factor for the prognosis of HCC patients[22–24, 30, 37]; however, its individual predictive ability is low. The studies on the prognosis model based on CK19 as a predictor are limited. Kawai et al[38] showed that HCC preoperative serum CK19 fragment (CK19-fragment, CYFRA21-1) was a sensitive predictor of CK19 expression in tumors. Xiang et al. showed that the combined detection of CYFRA21-1, HSP90β, CEA, and CA125 levels in the peripheral blood of lung cancer patients might provide certain reference values for the early diagnosis and prognosis prediction of lung cancer[39]. The combined CEA, CYFRA21-1, and NSE could effectively predict the diagnosis and prognosis of lung cancer[40]. A study showed that the preoperative mRNA expression level of CK19 in peripheral blood was an independent risk factor for the prognosis of patients with gastric cardia cancer (GCC), and the elevated preoperative CK19 and CEA mRNA levels might be good biomarkers to predict lymph node metastasis and poor prognosis[41].
Multiple studies have shown that Edmondson Steiner grading may become an important prognostic indicator for HCC. Chapel et al.[42]showed that Edmondson grade evaluated by percutaneous biopsy is closely related to survival in patients with unresectable HCC. Zhou's study[43]showed that the Edmondson grade has the strongest predictive power for recurrence and survival of HCC. CA125, a high-molecular-weight glycoprotein, has been used for the diagnosis and prognosis of various cancers, including ovarian cancer, cholangiocarcinoma, and lung cancer[44–46]. As compared to healthy subjects and benign liver diseases, the HCC patients exhibited increased expression levels of serum CA125, showing a sensitivity of 92% for the diagnosis of HCC[47]. Zhou et al.[48] reported that elevated preoperative serum CA125 levels could predict the increase in tumor diameter and poor prognosis in R0 resected HCC patients with AFP ≤ 200 ng/mL. Huang et al[49] found that elevated preoperative serum CA125 was an independent prognostic factor for OS and RFS in HCC patients; moreover, HCC patients with higher CA125 levels had shorter OS and higher recurrence rates. Tumor number is also a major factor, affecting the prognosis of liver cancer patients; multiple tumors are associated with poor prognosis[50–52]. A retrospective study in China[53] showed that HCC patients with lower BMI had poorer prognoses; this was consistent with the results in the current study.
The above-mentioned studies showed that the five predictors were associated with prognosis in HCC and confirmed the feasibility and practicability of the nomogram model constructed in the current study. This might provide a certain reference value for clinicians to judge and treat the prognosis of CK19-positive HCC patients. However, there are some shortcomings of the current study, which included lacking multi-center data of HCC patients, single sample size, and lack of external or multi-center validation cohort. The predictive performance of the model requires further validation through multi center and large sample studies.