CK7-positive hepatocellular carcinoma represents a distinct subtype: a retrospective cohort study

Abstract

Background

Hepatocellular carcinoma (HCC) remains a common and highly lethal malignancy worldwide, with a substantial and persistent disease burden. Although systemic therapies have advanced in recent years, overall prognosis is still far from satisfactory, reflecting the pronounced heterogeneity of HCC and the incomplete understanding of its key molecular drivers. Cytokeratin 7 (CK7), a marker related to biliary epithelium and hepatic progenitor–cell lineage, is routinely applied in diagnostic pathology and contributes to phenotypic classification of liver tumors. In particular, CK19-positive HCC has been recognized as a distinct subtype characterized by more aggressive behavior, higher risks of early recurrence and metastasis, and unfavorable survival outcomes. Ki-67 (Ki67), a widely used indicator of proliferative activity, has likewise been associated with tumor progression and poor prognosis in HCC. Against this background, we sought to delineate the clinicopathological features and outcomes of CK7-positive HCC using readily available clinical data, and to examine these patterns in relation to CK19 and Ki67 expression, thereby generating clinically grounded clues for mechanistic inference and future precision-oriented research.

Methods

This single-center retrospective cohort study included patients who underwent hepatectomy at Guangxi Medical University Cancer Hospital between 2014 and 2021 and had postoperative pathological confirmation of HCC. Among 1,992 eligible cases, 1,668 patients with complete CK7/CK19/Ki-67 immunohistochemical data and key clinicopathological variables constituted the complete-case analysis cohort. Data were extracted on demographics; exposure-related factors (e.g., alcohol use and consumption of raw fish dishes- used here as a proxy for potential biliary/parasitic exposure); liver function and cirrhosis-associated variables (Child–Pugh class, phenotypes suggestive of portal hypertension, splenomegaly, etc.); tumor burden and stage (maximum tumor diameter, tumor number, and BCLC stage); pathological features of invasiveness (including microvascular invasion and the presence of portal vein and/or bile duct tumor thrombus); and follow-up outcomes, including overall survival (OS), recurrence-free survival (RFS), and distant metastasis-free survival (DMFS). Immunohistochemical staining for CK7, CK19, and Ki67 was dichotomized as negative/positive according to prespecified cutoffs. Continuous variables were compared using Student’s t test or Welch’s t test for approximately normally distributed data, and the Wilcoxon rank-sum test otherwise. Categorical variables were analyzed with the chi-square test or Fisher’s exact test, as appropriate. Odds ratios (ORs) with 95% confidence intervals (CIs) were reported for binary outcomes. Survival was assessed using Kaplan–Meier estimates and log-rank tests; Cox proportional hazards models were fitted for univariable and multivariable analyses, with adjustment for clinically relevant confounders (including age, Child–Pugh class, and portal hypertension), and hazard ratios (HRs) with 95% CIs were reported.

Results

Among the 1,668 patients included in the complete-case analysis, positivity rates for CK7, CK19, and Ki-67 were 19.8%, 23.6%, and 81.8%, respectively. CK7 positivity was associated with a history of alcohol use (OR 1.73, 95% CI 1.35–2.21; P < 0.001) and consumption of raw fish dishes (OR 1.52, 95% CI 1.18–1.97; P = 0.001), and was inversely associated with HBsAg positivity (OR 0.70, 95% CI 0.50–0.98; P = 0.044). Clinically, CK7-positive tumors were characterized by lower AFP levels (P < 0.001), modestly higher triglyceride and total bilirubin levels (P = 0.033 and P = 0.029), and a tendency toward smaller maximum tumor diameter (P = 0.001; P < 0.001 in the single-tumor subset), whereas the proportion of multifocal disease did not differ (P = 0.762). In outcome analyses, CK7 positivity was not associated with recurrence-free survival (RFS), distant metastasis–free survival (DMFS), or overall survival (OS) (RFS HR 1.04; DMFS HR 0.97; OS HR 1.01; all P > 0.05).

In contrast, CK19 positivity was enriched among patients with anti-HCV positivity (OR 2.93, 95% CI 1.24–6.96; P = 0.018) and was linked to more advanced stage and invasive features, including portal vein tumor thrombus (PVTT), microvascular invasion (MVI), and metastasis, and consistently predicted poorer RFS/DMFS/OS. Similarly, Ki-67 positivity was significantly associated with adverse tumor biology and inferior survival outcomes.

Conclusion

Using routinely available clinical data, we provide a panoramic characterization of HCC spanning putative exposures, phenotypic stratification, tumor aggressiveness, and clinical outcomes. CK7-positive HCC displayed a clinicopathological and prognostic profile distinct from the CK19-positive subtype, suggesting that CK7 expression may denote a biologically discrete subset with a different etiologic trajectory. These findings offer a clinically grounded rationale for mechanistic studies and the development of subtype-tailored intervention strategies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-16001-6.

Introduction

Hepatocellular carcinoma (HCC) is among the most prevalent primary malignancies of the liver worldwide and remains a leading cause of cancer-related death, imposing a substantial public health burden [1, 2]. Although systemic treatment has evolved rapidly over the past decade—particularly with the advent of targeted agents and immune-based therapies—overall outcomes for HCC remain disappointing, and long-term survival is still limited. This persistent gap highlights the profound clinical, pathological, and molecular heterogeneity of HCC, and suggests that key mechanisms underpinning tumor initiation and progression have yet to be fully elucidated [1]. Accordingly, defining biologically meaningful HCC subsets is essential for more precise risk stratification, individualized therapeutic decision-making, and ultimately improved prognosis [2].

Cytokeratins (CKs) are intermediate filament proteins that constitute the cytoskeletal framework of epithelial cells and are central to routine histopathological evaluation of liver tumors [3]. However, the role of CK7 in tumor biology has not been comprehensively characterized. In diagnostic practice, CK7 immunoreactivity is often used as part of an immunophenotypic panel to aid inference regarding tumor lineage and origin; in some settings, CK7 positivity supports particular epithelial lineages, whereas in others, the absence of CK7 expression has been linked to more aggressive behavior [4]. Beyond oncology, CK7 has also been reported to have diagnostic utility in autoimmune hepatitis [5]. Notably, CK7 and related lineage markers have been found to be upregulated in an alcohol-associated liver disease (ALD) mouse model [6], further complicating the interpretation of CK7 expression in hepatocarcinogenesis. In a study focusing on hepatic immune responses, CK7 was used as an indicator to assess the stage of primary biliary cirrhosis progression [7].

Accumulating evidence indicates that a subset of HCCs aberrantly express biliary epithelial markers. These tumors are often hypothesized to arise from hepatic progenitor cells (HPCs) and may exhibit a more aggressive clinical course [3, 8]. Among these, CK19-positive HCC is the best studied and is consistently associated with enhanced invasiveness, a higher propensity for nodal metastasis, early postoperative recurrence, and inferior overall survival, supporting its recognition as a biologically distinct HCC subtype [1, 3]. While HCC is typically negative for both CK7 and CK19, the minority of cases that express these markers—including progenitor-phenotype HCC—may carry unfavorable prognostic implications, underscoring the growing relevance of immunohistochemical profiling for both differential diagnosis and prognostic stratification [9]. In parallel, Ki-67 (KI67), a widely used marker of proliferative activity, has been repeatedly linked to accelerated tumor progression and adverse outcomes in HCC [10]. Accordingly, CK19 and Ki-67 were selected as biologically grounded comparator markers because they represent, respectively, progenitor/biliary-lineage differentiation and proliferative activity—two complementary dimensions of aggressive HCC biology.

In this study, we conducted a single-center retrospective cohort analysis to delineate the clinicopathological phenotype and outcome profile of CK7⁺ HCC. By benchmarking CK7⁺ tumors against CK19-positive and Ki-67–high HCC, we aimed to clarify whether CK7 expression marks a clinically meaningful subtype with distinct biological behavior and prognostic implications. We further sought to provide clinically anchored evidence to support mechanistic hypotheses and to inform the development of subtype-tailored precision strategies for HCC.

Methods

Study design and population

This was a single-center retrospective cohort study. We consecutively enrolled patients who underwent hepatectomy at Guangxi Medical University Cancer Hospital between January 2014 and December 2021 and had postoperative pathological confirmation of hepatocellular carcinoma (HCC). A total of 1,992 eligible cases were initially identified. Patients were excluded from the primary analyses if substantial missing immunohistochemical, clinicopathological, or follow-up data precluded the predefined comparisons; therefore, the final analytical population represented a complete-case cohort.

Data collection and variable definitions

Clinical data were retrospectively extracted from the hospital information system and curated using a standardized data dictionary. Variables of interest included the following.

Demographic and virologic characteristics

Age, sex, and ethnicity were recorded. Hepatitis B surface antigen (HBsAg) status and hepatitis C virus antibody (HCV-Ab) status were collected; where available, HBV DNA levels were also retrieved.

Etiology-related factors and exposures. Alcohol use history was coded as a binary variable (0/1). Consumption of raw fish dishes was also coded as 0/1 and was considered a surrogate for possible liver fluke related or broader biliary/parasitic exposure, rather than a direct diagnostic substitute for confirmed liver fluke infection.

Liver function and cirrhosis-related features

Child–Pugh class was documented. Phenotypes related to portal hypertension were captured using routinely available clinical and imaging-derived proxies (e.g., integrated splenomegaly indicators and composite portal hypertension metrics, as applicable). Standard liver biochemistry, including total bilirubin, alanine aminotransferase (ALT), and aspartate aminotransferase (AST), was collected.

Tumor burden, stage, and invasive features

Maximum tumor diameter (cm) and tumor number were recorded; tumor multiplicity was defined radiologically (solitary vs. multifocal). Tumor stage was classified according to the Barcelona Clinic Liver Cancer (BCLC) system (0/A/B/C). Pathological and imaging-based invasive features included microvascular invasion (MVI; 0/1), portal vein tumor thrombus (PVTT; 0/1), and bile duct tumor thrombus (BDTT; 0/1).

Tumor markers and metabolic indices

Laboratory measures included alpha-fetoprotein (AFP), carbohydrate antigen 19 − 9 (CA19-9), triglycerides (TG), and total bilirubin (TBil). Laboratory values reported with inequality signs (e.g., “<1”, “>1000”, “>1200”) were harmonized using prespecified rules to yield analyzable numeric values; sensitivity analyses additionally applied a log transformation (log10[x + 1]) to attenuate the influence of extreme values.

Immunohistochemistry evaluation and grouping

Immunohistochemistry (IHC) results for CK7, CK19, and Ki-67 were obtained from postoperative specimens as documented in the pathology reports and extracted from the medical record system. Each marker was dichotomized a priori as negative versus positive. For CK19, any unequivocal tumor-cell staining (> 0%) was considered positive, consistent with recent evidence suggesting that a 0% threshold may represent an optimal prognostic cutoff in HCC [11]. For CK7, because no universally accepted cutoff exists in HCC and expression is often focal and heterogeneous, any unequivocal tumor-cell staining (> 0%) was considered positive for exploratory phenotypic stratification. Ki-67 was categorized as low (≤ 10%) or high (> 10%) according to prior HCC-oriented studies [12, 13].

Outcomes and follow-up

Primary outcomes were overall survival (OS) and recurrence-free survival (RFS), while secondary outcomes included distant metastasis-free survival (DMFS) and clinicopathologic associations. OS was defined as the time from surgery to death from any cause or last follow-up. RFS was defined as the time from surgery to first recurrence/progression or last follow-up. DMFS was defined as the time from surgery to the occurrence of distant metastasis or last follow-up. Postoperative recurrence and metastasis events, as well as time to metastasis, were additionally captured. Follow-up was censored on May 2024 or at the last available follow-up visit, whichever occurred first. Patients lost to follow-up were censored at the date of last confirmed contact.

Statistical analysis

All analyses were performed using R (R Foundation for Statistical Computing).

Continuous variables were assessed for distributional characteristics and summarized as mean ± standard deviation (SD) when approximately normal, or as median (interquartile range [IQR]) otherwise. Categorical variables were summarized as counts (percentages). Between-group comparisons for continuous variables used Welch’s t test when normality was acceptable but variances were unequal, or the Wilcoxon rank-sum test for non-normal distributions. Categorical variables were compared using Pearson’s chi-square test; Fisher’s exact test was applied when expected cell counts were < 5.

For binary outcomes—including PVTT, BDTT, MVI, metastasis, tumor multiplicity (solitary vs. multifocal), and histologic differentiation (dichotomized as grades 1–2 vs. 2.5–4)—odds ratios (ORs) with 95% confidence intervals (CIs) were estimated and displayed using forest plots. The direction of effect was consistently defined as marker-positive (+) relative to marker-negative (−).

Survival distributions for OS, RFS, and DMFS were estimated using the Kaplan–Meier method and compared with log-rank tests. Hazard ratios (HRs) with 95% CIs were derived using Cox proportional hazards models.

All tests were two-sided, and P < 0.05 was considered statistically significant. For exploratory analyses involving multiple endpoints or comparisons, results were interpreted in conjunction with effect sizes and internal consistency rather than statistical significance alone. Because postoperative adjuvant therapies (e.g., TACE, targeted agents, or immunotherapy) were heterogeneous, incompletely captured for some patients, and often time-dependent, they were not incorporated into the primary Cox models; this issue is acknowledged as a limitation.

Ethical considerations

The study protocol was approved by the Ethics Committee of Guangxi Medical University Cancer Hospital and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients prior to surgery. Data were de-identified for research purposes, and confidentiality was maintained throughout the study.

Results

Baseline distribution of CK7, CK19, and Ki-67

Baseline characteristics of the study cohort stratified by CK7, CK19, and Ki-67 status are summarized in Tables 1, 2, 3. Overall, patients were predominantly male and of Han ethnicity, with most presenting with BCLC stage 0–B disease. Across the three stratifications, demographic variables (age, sex, and ethnicity), etiological factors (drinking history, raw fish consumption, HBsAg positivity, and HCV antibody positivity), metabolic comorbidities (diabetes and hypertension), tumor burden (tumor number and maximum tumor diameter), invasive features (PVTT, BDTT, and MVI), and laboratory indices (AFP, CA19-9, triglycerides, and total bilirubin) were compared between marker-negative and marker-positive groups. Significant between-group differences were observed for selected variables (P values shown in Tables 1, 2, 3), indicating that CK7/CK19/Ki-67 expression is associated with distinct clinical–pathological profiles in resected HCC.

Table 1.

Baseline characteristics stratified by CK7 status

Variable	Level	Overall	CK7-	CK7+	P value
Age, years		55.14 ± 11.17	55.04 ± 11.18	55.53 ± 11.10	0.473
Sex					0.964
	Female	223 (13.4%)	179 (13.4%)	44 (13.3%)
	Male	1445 (86.6%)	1158 (86.6%)	287 (86.7%)
Ethnicity					0.915
	Han	971 (59.9%)	775 (59.8%)	196 (60.1%)
	Zhuang	651 (40.1%)	521 (40.2%)	130 (39.9%)
Drinking history					< 0.001
	No	1159 (69.5%)	964 (72.1%)	195 (58.9%)
	Yes	509 (30.5%)	373 (27.9%)	136 (41.1%)
Raw fish consumption					0.001
	No	1209 (72.5%)	993 (74.3%)	216 (65.3%)
	Yes	459 (27.5%)	344 (25.7%)	115 (34.7%)
HBsAg positive					0.027
	No	207 (12.4%)	154 (11.5%)	53 (16.0%)
	Yes	1460 (87.6%)	1182 (88.5%)	278 (84.0%)
HCV Ab positive					0.399
	No	1648 (98.8%)	1319 (98.7%)	329 (99.4%)
	Yes	20 (1.2%)	18 (1.3%)	2 (0.6%)
Diabetes					0.101
	No	1560 (93.5%)	1257 (94.0%)	303 (91.5%)
	Yes	108 (6.5%)	80 (6.0%)	28 (8.5%)
Hypertension					0.352
	No	1490 (89.3%)	1199 (89.7%)	291 (87.9%)
	Yes	178 (10.7%)	138 (10.3%)	40 (12.1%)
AFP, µg/L		144.70 (8.36, 1000.00)	178.40 (9.30, 1000.00)	63.62 (6.04, 1000.00)	0.001
CA19-9, U/mL		13.80 (5.93, 25.80)	13.90 (5.80, 25.88)	13.30 (6.14, 25.55)	0.995
Triglycerides		0.96 (0.74, 1.30)	0.94 (0.73, 1.28)	1.02 (0.77, 1.36)	0.033
Total bilirubin		13.70 (10.20, 18.20)	13.50 (10.00, 18.10)	14.50 (10.50, 19.15)	0.029
BCLC stage					0.016
	0	94 (5.6%)	70 (5.2%)	24 (7.3%)
	A	1053 (63.1%)	859 (64.2%)	194 (58.6%)
	B	208 (12.5%)	152 (11.4%)	56 (16.9%)
	C	313 (18.8%)	256 (19.1%)	57 (17.2%)
Metastasis present					0.767
	No	1582 (94.8%)	1267 (94.8%)	315 (95.2%)
	Yes	86 (5.2%)	70 (5.2%)	16 (4.8%)
Tumor number					0.762
	Single	1509 (90.5%)	1211 (90.6%)	298 (90.0%)
	Multiple	159 (9.5%)	126 (9.4%)	33 (10.0%)
Tumor size, cm		6.00 (3.80, 10.00)	6.00 (4.00, 10.00)	5.50 (3.50, 8.00)	0.001
Portal vein tumor thrombus (PVTT)					0.387
	No	1458 (87.4%)	1164 (87.1%)	294 (88.8%)
	Yes	210 (12.6%)	173 (12.9%)	37 (11.2%)
Bile duct tumor thrombus (BDTT)					0.071
	No	1639 (98.5%)	1317 (98.8%)	322 (97.3%)
	Yes	25 (1.5%)	16 (1.2%)	9 (2.7%)
Microvascular invasion (MVI)					0.201
	No	749 (44.9%)	590 (44.1%)	159 (48.0%)
	Yes	919 (55.1%)	747 (55.9%)	172 (52.0%)
Splenomegaly					0.075
	No	834 (50.0%)	654 (48.9%)	180 (54.4%)
	Yes	834 (50.0%)	683 (51.1%)	151 (45.6%)
CK19					< 0.001
	CK19-	1275 (76.4%)	1093 (81.8%)	182 (55.0%)
	CK19+	393 (23.6%)	244 (18.2%)	149 (45.0%)
Ki-67					0.168
	Ki-67-	304 (18.2%)	235 (17.6%)	69 (20.8%)
	Ki-67+	1364 (81.8%)	1102 (82.4%)	262 (79.2%)

Table 2.

Baseline characteristics stratified by CK19 status

Variable	Level	Overall	CK19-	CK19+	P value
Age, years		55.00 (47.00, 64.00)	56.00 (48.00, 64.00)	52.00 (44.00, 59.00)	< 0.001
Sex					0.938
	Female	223 (13.4%)	170 (13.3%)	53 (13.5%)
	Male	1445 (86.6%)	1105 (86.7%)	340 (86.5%)
Ethnicity					0.803
	Han	971 (59.9%)	745 (60.0%)	226 (59.3%)
	Zhuang	651 (40.1%)	496 (40.0%)	155 (40.7%)
Drinking history					0.059
	No	1159 (69.5%)	901 (70.7%)	258 (65.6%)
	Yes	509 (30.5%)	374 (29.3%)	135 (34.4%)
Raw fish consumption					0.449
	No	1209 (72.5%)	930 (72.9%)	279 (71.0%)
	Yes	459 (27.5%)	345 (27.1%)	114 (29.0%)
HBsAg positive					0.972
	No	207 (12.4%)	158 (12.4%)	49 (12.5%)
	Yes	1460 (87.6%)	1116 (87.6%)	344 (87.5%)
HCV Ab positive					0.013
	No	1648 (98.8%)	1265 (99.2%)	383 (97.5%)
	Yes	20 (1.2%)	10 (0.8%)	10 (2.5%)
Diabetes					0.566
	No	1560 (93.5%)	1190 (93.3%)	370 (94.1%)
	Yes	108 (6.5%)	85 (6.7%)	23 (5.9%)
Hypertension					0.016
	No	1490 (89.3%)	1126 (88.3%)	364 (92.6%)
	Yes	178 (10.7%)	149 (11.7%)	29 (7.4%)
AFP, µg/L		144.70 (8.36, 1000.00)	63.47 (6.08, 1000.00)	1000.00 (123.40, 1000.00)	< 0.001
CA19-9, U/mL		13.80 (5.93, 25.80)	13.67 (5.80, 25.39)	14.80 (6.90, 28.00)	0.124
Triglycerides		0.96 (0.74, 1.30)	0.96 (0.74, 1.30)	0.94 (0.73, 1.30)	0.788
Total bilirubin		13.70 (10.20, 18.20)	13.80 (10.30, 18.30)	13.30 (9.90, 18.20)	0.309
BCLC stage					< 0.001
	0	94 (5.6%)	75 (5.9%)	19 (4.8%)
	A	1053 (63.1%)	835 (65.5%)	218 (55.5%)
	B	208 (12.5%)	152 (11.9%)	56 (14.2%)
	C	313 (18.8%)	213 (16.7%)	100 (25.4%)
Metastasis present					< 0.001
	No	1582 (94.8%)	1224 (96.0%)	358 (91.1%)
	Yes	86 (5.2%)	51 (4.0%)	35 (8.9%)
Tumor number					0.916
	Single	1509 (90.5%)	1154 (90.5%)	355 (90.3%)
	Multiple	159 (9.5%)	121 (9.5%)	38 (9.7%)
Tumor size, cm		6.00 (3.80, 10.00)	6.00 (3.80, 9.75)	6.50 (3.50, 10.00)	0.115
Portal vein tumor thrombus (PVTT)					0.002
	No	1458 (87.4%)	1132 (88.8%)	326 (83.0%)
	Yes	210 (12.6%)	143 (11.2%)	67 (17.0%)
Bile duct tumor thrombus (BDTT)					0.142
	No	1639 (98.5%)	1255 (98.7%)	384 (97.7%)
	Yes	25 (1.5%)	16 (1.3%)	9 (2.3%)
Microvascular invasion (MVI)					< 0.001
	No	749 (44.9%)	606 (47.5%)	143 (36.4%)
	Yes	919 (55.1%)	669 (52.5%)	250 (63.6%)
Splenomegaly					0.018
	No	834 (50.0%)	658 (51.6%)	176 (44.8%)
	Yes	834 (50.0%)	617 (48.4%)	217 (55.2%)
CK7					< 0.001
	CK7-	1337 (80.2%)	1093 (85.7%)	244 (62.1%)
	CK7+	331 (19.8%)	182 (14.3%)	149 (37.9%)
Ki-67					< 0.001
	Ki-67-	304 (18.2%)	265 (20.8%)	39 (9.9%)
	Ki-67+	1364 (81.8%)	1010 (79.2%)	354 (90.1%)

Table 3.

Baseline characteristics stratified by Ki-67 status

Variable	Level	Overall	Ki-67-	Ki-67+	P value
Age, years		55.00 (47.00, 64.00)	60.00 (51.00, 66.00)	54.00 (47.00, 62.00)	< 0.001
Sex					0.216
	Female	223 (13.4%)	34 (11.2%)	189 (13.9%)
	Male	1445 (86.6%)	270 (88.8%)	1175 (86.1%)
Ethnicity					0.151
	Han	971 (59.9%)	190 (63.5%)	781 (59.0%)
	Zhuang	651 (40.1%)	109 (36.5%)	542 (41.0%)
Drinking history					0.138
	No	1159 (69.5%)	222 (73.0%)	937 (68.7%)
	Yes	509 (30.5%)	82 (27.0%)	427 (31.3%)
Raw fish consumption					0.219
	No	1209 (72.5%)	229 (75.3%)	980 (71.8%)
	Yes	459 (27.5%)	75 (24.7%)	384 (28.2%)
HBsAg positive					0.113
	No	207 (12.4%)	46 (15.1%)	161 (11.8%)
	Yes	1460 (87.6%)	258 (84.9%)	1202 (88.2%)
HCV Ab positive					1
	No	1648 (98.8%)	301 (99.0%)	1347 (98.8%)
	Yes	20 (1.2%)	3 (1.0%)	17 (1.2%)
Diabetes					0.266
	No	1560 (93.5%)	280 (92.1%)	1280 (93.8%)
	Yes	108 (6.5%)	24 (7.9%)	84 (6.2%)
Hypertension					< 0.001
	No	1490 (89.3%)	253 (83.2%)	1237 (90.7%)
	Yes	178 (10.7%)	51 (16.8%)	127 (9.3%)
AFP, µg/L		144.70 (8.36, 1000.00)	13.00 (3.84, 376.10)	229.50 (11.37, 1000.00)	< 0.001
CA19-9, U/mL		13.80 (5.93, 25.80)	13.20 (6.90, 27.00)	14.00 (5.80, 25.72)	0.559
Triglycerides		0.96 (0.74, 1.30)	1.01 (0.77, 1.40)	0.95 (0.73, 1.27)	0.098
Total bilirubin		13.70 (10.20, 18.20)	13.50 (10.20, 17.90)	13.70 (10.20, 18.30)	0.484
BCLC stage					< 0.001
	0	94 (5.6%)	17 (5.6%)	77 (5.6%)
	A	1053 (63.1%)	232 (76.3%)	821 (60.2%)
	B	208 (12.5%)	37 (12.2%)	171 (12.5%)
	C	313 (18.8%)	18 (5.9%)	295 (21.6%)
Metastasis present					0.006
	No	1582 (94.8%)	298 (98.0%)	1284 (94.1%)
	Yes	86 (5.2%)	6 (2.0%)	80 (5.9%)
Tumor number					0.669
	Single	1509 (90.5%)	277 (91.1%)	1232 (90.3%)
	Multiple	159 (9.5%)	27 (8.9%)	132 (9.7%)
Tumor size, cm		6.00 (3.80, 10.00)	6.00 (3.50, 8.50)	6.00 (4.00, 10.00)	0.094
Portal vein tumor thrombus (PVTT)					< 0.001
	No	1458 (87.4%)	294 (96.7%)	1164 (85.3%)
	Yes	210 (12.6%)	10 (3.3%)	200 (14.7%)
Bile duct tumor thrombus (BDTT)					0.293
	No	1639 (98.5%)	302 (99.3%)	1337 (98.3%)
	Yes	25 (1.5%)	2 (0.7%)	23 (1.7%)
Microvascular invasion (MVI)					< 0.001
	No	749 (44.9%)	211 (69.4%)	538 (39.4%)
	Yes	919 (55.1%)	93 (30.6%)	826 (60.6%)
Splenomegaly					0.205
	No	834 (50.0%)	162 (53.3%)	672 (49.3%)
	Yes	834 (50.0%)	142 (46.7%)	692 (50.7%)
CK7					0.168
	CK7-	1337 (80.2%)	235 (77.3%)	1102 (80.8%)
	CK7+	331 (19.8%)	69 (22.7%)	262 (19.2%)
CK19					< 0.001
	CK19-	1275 (76.4%)	265 (87.2%)	1010 (74.0%)
	CK19+	393 (23.6%)	39 (12.8%)	354 (26.0%)

Continuous variables are presented as mean ± SD when approximately normally distributed; otherwise as median (IQR). P values were computed using Welch’s t-test or Wilcoxon rank-sum test for continuous variables, and Pearson’s chi-square test or Fisher’s exact test for categorical variables, as appropriate. The “Overall” column in Tables 1, 2, 3 reflects the complete-case analysis cohort (n = 1668)

Stratification by CK7, CK19, and Ki-67 expression

Among the 1,668 patients in the complete-case analysis cohort, CK7 positivity (CK7 + vs. CK7−) was significantly associated with etiological exposures: CK7 + patients more frequently had a drinking history (41.1% vs. 27.9%, P < 0.001) and raw fish consumption (34.7% vs. 25.7%, P = 0.001), while HBsAg positivity was slightly less common in CK7 + patients (84.0% vs. 88.5%, P = 0.0268). Consistent with a distinct clinicopathologic phenotype, CK7 + cases showed lower AFP (median 63.62 vs. 178.40 µg/L, P = 0.00142), smaller tumor size (median 5.5 vs. 6.0 cm, P = 0.00106), and a different BCLC distribution (P = 0.016).

For CK19 stratification, CK19 + patients were younger (median 52 vs. 56 years, P < 0.001) and demonstrated a higher prevalence of HCV antibody positivity (2.5% vs. 0.8%, P = 0.0131), highlighting a specific viral association with CK19 expression. CK19 + status also aligned with more aggressive disease features, including higher AFP (median 1000.0 vs. 63.47 µg/L, P < 0.001), more advanced BCLC stage (P < 0.001), more metastasis at presentation (8.9% vs. 4.0%, P < 0.001), increased PVTT (17.0% vs. 11.2%, P = 0.00231), and higher MVI (63.6% vs. 52.5%, P < 0.001).

For Ki-67 stratification, Ki-67 + cases showed patterns consistent with a proliferative/aggressive phenotype: younger age (median 54 vs. 60 years, P < 0.001), markedly higher AFP (median 229.5 vs. 13.0 µg/L, P < 0.001), more advanced BCLC stage (P < 0.001), higher metastasis prevalence (5.9% vs. 2.0%, P = 0.00553), higher PVTT (14.7% vs. 3.3%, P < 0.001), and substantially higher MVI (60.6% vs. 30.6%, P < 0.001).

Associations between CK7/CK19/Ki-67 and etiological factors

As shown in Fig. 1, we examined the associations of CK7, CK19, and Ki-67 expression with etiological factors in the complete-case cohort (n = 1,668).

Fig. 1.

Associations of CK7/CK19/Ki-67 with etiological factors. Note: (Figs. 1A, C and E) The proportions of etiological factors (drinking history, raw fish consumption, HBsAg positivity, and HCV antibody positivity) stratified by immunohistochemical marker status (negative vs. positive) are shown as 100% stacked bar charts. (Figs. 1B, D and F) Forest plots summarize odds ratios (ORs) with 95% confidence intervals (CIs) for etiological factor positivity comparing marker-positive versus marker-negative patients (reference group: marker-negative). ORs and P values were derived from 2 × 2 contingency analyses (Fisher’s exact test when appropriate). CK7 positivity was associated with higher odds of drinking history and raw fish consumption, whereas CK7 positivity was associated with lower odds of HBsAg positivity. CK19 positivity was associated with higher odds of HCV antibody positivity. Ki-67 status showed no significant associations with the etiological factors evaluated

In the etiological factor analysis, CK7 positivity was significantly associated with both alcohol exposure and raw fish consumption. Specifically, CK7-positive tumors were more likely to occur in patients with a drinking history compared with CK7-negative tumors (OR 1.73, 95% CI 1.35–2.21; P < 0.001). A similar association was observed for raw fish consumption (OR 1.52, 95% CI 1.18–1.97; P = 0.001). In contrast, CK7 positivity was inversely associated with HBsAg positivity (OR 0.70, 95% CI 0.50–0.98; P = 0.043), and CK7 showed no significant association with HCV antibody positivity (OR 0.43, 95% CI 0.10–1.85; P = 0.404).

For CK19, no significant associations were found with drinking history, raw fish consumption, or HBsAg positivity (all P > 0.05). Notably, CK19 positivity was significantly enriched among patients with positive HCV antibody status (OR 2.93, 95% CI 1.24–6.96; P = 0.018), suggesting a potential link between HCV-related liver disease and the CK19-positive HCC phenotype. Ki-67 status was not significantly associated with any of the etiological factors assessed (all P > 0.05).

Further validation of CK7-positive–associated laboratory profiles

As shown in Fig. 2, we further compared continuous laboratory variables across CK7, CK19, and Ki-67 expression groups to define their distinct biochemical profiles.

Fig. 2.

Continuous variables by CK7, CK19, and Ki-67 status. Note: Panel A shows the distribution of age, triglycerides, AFP, CA19-9, and total bilirubin stratified by CK7/CK19/Ki-67 (Negative vs. Positive) using violin + box + jitter plots. AFP and CA19-9 are displayed on a log10(value + 1) scale to improve visualization of skewed distributions, while P values were computed on the raw (untransformed) values using the Wilcoxon rank-sum test. Panel B summarizes group differences as the Hodges–Lehmann median shift (Positive − Negative) with 95% confidence intervals, where values > 0 indicate higher levels in the positive group and values < 0 indicate lower levels in the positive group

When stratified by immunohistochemical marker status (Fig. 2), CK7 positivity was associated with lower AFP levels (P < 0.001) and modest but statistically significant higher levels of triglycerides (P = 0.033) and total bilirubin (P = 0.029), whereas age (P = 0.587) and CA19-9 (P = 0.995) did not differ between CK7 − and CK7 + groups.In contrast, CK19 positivity was linked to a younger age distribution (P < 0.001) and higher AFP (P < 0.001), with no significant differences in triglycerides (P = 0.788), CA19-9 (P = 0.124), or total bilirubin (P = 0.309). Similarly, Ki-67 positivity was associated with younger age (P < 0.001) and higher AFP (P < 0.001), while CA19-9 (P = 0.559), triglycerides (P = 0.098), and total bilirubin (P = 0.484) showed no statistically significant differences. Across markers, Panel B confirmed these patterns by demonstrating the direction and magnitude of the median shifts (Positive − Negative) with corresponding 95% confidence intervals.

Beyond laboratory correlates: CK7/CK19/Ki-67 and tumor invasive phenotypes

As shown in Fig. 3, we examined the associations of CK7, CK19, and Ki-67 expression with tumor invasive phenotypes, including differentiation, metastasis, MVI, PVTT, BDTT, and tumor multiplicity.

Fig. 3.

Associations of CK7, CK19, and Ki-67 with tumor biological behaviors. Note: Stacked bar plots show the proportions of each binary outcome (Yes/No) stratified by marker status (positive vs. negative) for CK7, CK19, and Ki-67. Outcomes included bile duct tumor thrombus (BDTT), differentiation grade (III–IV vs. I–II), metastasis, microvascular invasion (MVI), portal vein tumor thrombus (PVTT), and tumor number (multiple vs. single). Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated using 2 × 2 contingency tables, comparing marker-positive versus marker-negative groups; P values were obtained using Pearson’s chi-square test or Fisher’s exact test, as appropriate. Forest plots summarize ORs (log scale) for each marker across all outcomes

In analyses of tumor biological behaviors, CK19 positivity was consistently associated with more aggressive features, including poor differentiation (III–IV) (OR 1.51, P < 0.001), metastasis (OR 2.35, P < 0.001), MVI (OR 1.58, P < 0.001), and PVTT (OR 1.63, P = 0.002), whereas no significant association was observed with tumor number (multiple) (OR 1.02, P = 0.916) or BDTT (OR 1.84, P = 0.142). Ki-67 positivity showed strong associations with adverse pathological characteristics, including poor differentiation (OR 3.85, P < 0.001), metastasis (OR 3.09, P = 0.006), MVI (OR 3.48, P < 0.001), and PVTT (OR 5.05, P < 0.001), with no significant association for tumor number (OR 1.10, P = 0.669) or BDTT (OR 2.60, P = 0.293). In contrast, CK7 positivity showed no statistically significant associations with differentiation, metastasis, MVI, PVTT, or tumor number (all P > 0.05), while a borderline association was observed for BDTT (OR 2.30, P = 0.071). Overall, the forest plots highlighted a gradient of aggressiveness, with Ki-67 demonstrating the strongest effect sizes across outcomes, followed by CK19, whereas CK7 showed limited associations.

As shown in Fig. 4, we further analyzed tumor diameter in the single-tumor subgroup to clarify whether CK7, CK19, or Ki-67 expression was associated with differences in solitary tumor burden. In the single-tumor subgroup, CK7 positivity was associated with a smaller tumor diameter, as reflected by a negative Hodges–Lehmann median shift (Positive − Negative) and a significant Wilcoxon test (P < 0.001). In contrast, tumor size did not differ significantly by CK19 status (P = 0.177) and showed only a non-significant trend with Ki-67 positivity (P = 0.080), suggesting that the inverse relationship with tumor size was most pronounced for CK7 rather than CK19 or Ki-67 in patients with solitary tumors.

Fig. 4.

Tumor size in the single-tumor cohort stratified by CK7, CK19, and Ki-67 status. Note: Panel A shows violin/box/jitter plots of tumor diameter (cm) by marker status (Negative vs. Positive). Distributions are visually clipped at the 1st–99th percentiles for readability, while statistical tests were performed on raw values. Panel B presents Hodges–Lehmann median shifts (Positive − Negative) with 95% confidence intervals; P values were obtained using two-sided Wilcoxon rank-sum tests

BCLC stage distribution and survival outcomes by CK7/CK19/Ki-67 status

As shown in Fig. 5, we assessed the prognostic impact of CK7, CK19, and Ki-67 expression on recurrence-free survival, distant metastasis-free survival, and overall survival.

Fig. 5.

BCLC stage distribution and survival outcomes stratified by CK7, CK19, and Ki-67 status. Note: Top row: stacked bar charts showing the proportional distribution of BCLC stage (0/A/B/C) in marker-negative versus marker-positive groups.Rows 2–4: Kaplan–Meier curves for recurrence-free survival (RFS), metastasis-free survival, and overall survival (OS) stratified by CK7, CK19, and Ki-67 status, respectively. Hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated using Cox proportional hazards models (positive vs. negative), and log-rank tests were used for unadjusted comparisons. Positive and negative groups are indicated on each survival panel

BCLC stage distribution and survival outcomes by CK7/CK19/Ki-67 status.

In survival analyses, CK7 expression was not associated with prognosis, with no significant differences in RFS, metastasis-free survival, or OS between CK7-positive and CK7-negative patients (Fig. 5). In contrast, CK19 positivity and high Ki-67 expression showed poorer survival across endpoints, consistent with prior reports describing their links to aggressive tumor biology and adverse outcomes. Given that the prognostic relevance of CK19 and Ki-67 has been well established in the literature, our analyses primarily serve as an internal validation, whereas CK7 did not appear to confer additional prognostic stratification in this cohort.

Overall, these findings indicate that CK7-positive HCC was characterized by distinct exposure-related and clinicopathological features, including associations with alcohol use, raw-fish consumption, lower AFP, and smaller tumor size, but without significant differences in survival outcomes. In contrast, CK19 and Ki-67 positivity were more consistently associated with aggressive tumor behavior and poorer prognosis. An integrated schematic summary of these findings is shown in Fig. 6.

Fig. 6.

Graphical summary of biomarker-associated tumor biology and prognosis. Conceptual framework summarizing hypothesized pathways and cohort-supported associations linking etiologic exposures, CK7/CK19/Ki-67 phenotypes, tumor biology, and outcomes in hepatocellular carcinoma. Solid arrows denote hypothesized causal links based on biological plausibility together with the observed pattern in this cohort; dashed arrows indicate statistical associations without causal inference. Orange pathways highlight features associated with a more aggressive course (later BCLC stage and shorter OS/RFS/DMFS). In this cohort, alcohol drinking and raw-fish consumption were associated with CK7 positivity, which also showed mild associations with alcohol-related liver-injury markers (slightly higher triglycerides and total bilirubin), alongside lower AFP and smaller tumor size in the single-tumor subset. Despite these phenotypic differences, CK7 + vs. CK7 − showed no significant differences in BCLC stage or OS/RFS/DMFS. Additional associations included HBsAg positivity inversely related to CK7, and HCV Ab positivity positively related to CK19

Discussion

Principal novel findings and clinical meaning

This study provides an integrated view of how CK7, CK19, and Ki-67 phenotypes relate to exposures, laboratory profiles, tumor biology, and outcomes in HCC. The most notable new finding is that CK7 positivity delineated a distinct clinical–biochemical phenotype—characterized by a history suggestive of alcohol use and raw-fish exposure and mild associations with alcohol-related liver injury markers (e.g., triglycerides and total bilirubin), together with lower AFP and smaller tumor size in the single-tumor subgroup. These results suggest that CK7 expression may capture a background etiologic/liver-injury context or a specific differentiation lineage, rather than a straightforward marker of aggressive behavior. Clinically, this implies that CK7 positivity alone may be more useful for phenotyping and etiologic inference than for risk stratification of long-term outcomes.

Etiologic signals: alcohol/raw-fish exposure and viral hepatitis

Alcohol consumption and the habitual intake of raw fish dishes—a lifestyle pattern closely linked to liver fluke exposure—were positively associated with CK7 expression. From a causal standpoint, alcohol-related liver injury and fluke-associated hepatobiliary inflammation are more plausibly upstream drivers, whereas CK7 expression in tumor cells is likely a downstream phenotypic consequence. This interpretation is supported by experimental evidence showing upregulation of CK7 and related markers in an animal model of alcohol-associated liver disease [6]. Collectively, these observations raise the possibility that CK7 captures an alcohol-/fluke-associated molecular program along the trajectory from chronic liver injury to hepatocarcinogenesis, or alternatively that CK7-positive HCC constitutes a biologically distinct subgroup within the broader HCC spectrum.

In addition, we also noted that HBsAg positivity was inversely associated with CK7 positivity, supporting the notion that CK7 + defined tumors may represent a subgroup less driven by HBV-related carcinogenesis.

Distinct tumor biology of CK7-positive HCC

AFP is widely regarded as a marker reflecting hepatoblast-like features in liver tumors [14]. CK7 and CK19, by contrast, are commonly used to indicate biliary-lineage (cholangiocytic) differentiation within HCC cells [15]. Although CK7 and CK19 showed a clear statistical correlation with each other in our cohort (Figure S1), their relationships with AFP moved in opposite directions. Notably, among patients with solitary tumors, CK7-positive HCC was associated with a smaller maximum tumor diameter (Fig. 4). Taken together, these findings suggest that CK7-positive HCC may represent a phenotypically distinct subgroup and, at least in part, may be linked to a less aggressive growth pattern with the potential for more favorable outcomes.

CK19/Ki-67 axis and aggressive biology: replication plus extension

Consistent with prior literature [16–18], our cohort corroborated that CK19 positivity and Ki-67 positivity cluster with more aggressive tumor behavior and poorer prognosis. In addition to validating these established patterns, we observed a positive association between HCV antibody positivity and CK19 positivity, which may suggest a potential etiologic relationship between HCV-related inflammatory signaling and progenitor/biliary-like differentiation programs in a subset of HCC. An animal experiment conducted years ago on HCV infection provides supporting evidence as well [19]. While causality cannot be established here, this signal warrants targeted mechanistic and external-cohort validation—particularly in settings where HCV prevalence remains clinically relevant. Clinically, CK19 and Ki-67 may complement conventional prognostic factors such as AFP, MVI, and BCLC stage by helping to identify patients at higher risk of recurrence and invasive disease who may merit closer postoperative surveillance.

Bile duct tumor thrombus and CA19-9

Among the invasive phenotypes evaluated, bile duct tumor thrombus (BDTT) emerged as a particularly distinctive entity (Fig. 3). Owing to the relatively small number of BDTT cases in our cohort, most comparisons did not reach conventional statistical significance. Nevertheless, CK7 showed the largest estimated effect size for BDTT, with a borderline association. When we further profiled the clinical correlates of BDTT, CA19-9 stood out as the most prominent marker (Figure S2), suggesting that BDTT may be accompanied by a characteristic biomarker pattern.

Diabetes and Ki-67 expression

At an early stage of the analysis, we observed that patients with diabetes and hypertension appeared to have more favorable outcomes. Upon closer examination, this pattern was accompanied by a significant shift in Ki-67 status among patients with diabetes, with a higher proportion of Ki-67–negative tumors (OR = 0.55, P = 0.037). We subsequently considered age as a potential explanation for this association (Figure S3). However, a direct biological link between diabetes and Ki-67 cannot be excluded. Prior work has reported reduced Ki-67 expression in pancreatic β cells in patients with diabetes [20], providing a plausible context in which altered proliferative signaling may extend beyond endocrine tissue and be reflected in tumor proliferation indices.

Limitations and clinical implications

Several limitations merit consideration. First, this was a retrospective single-center study, which may limit the generalizability of the findings and underscores the need for multicenter or external validation. Second, because the final analytical cohort was based on complete-case inclusion, selection bias related to missing data cannot be fully excluded. To address this concern, we compared the included analytical cohort with cases excluded due to missing data and found that the two groups were generally comparable in terms of age, sex, diabetes, hypertension, tumor size, PVTT, BDTT, and OS events, although differences were observed in HBsAg positivity, tumor multiplicity, BCLC stage distribution, and MVI status (Supplementary Table X). These findings suggest that while the final cohort remained broadly representative of the overall surgical population, some degree of selection bias may still be present and should be considered when interpreting the results. Third, residual confounding remains possible despite multivariable adjustment, particularly because postoperative adjuvant therapies were heterogeneous, partially incomplete, and often time-dependent. Fourth, raw-fish consumption was used only as a surrogate for possible biliary/parasitic exposure and should not be interpreted as direct evidence of liver fluke infection. Finally, although CK19 and Ki-67 may have value for identifying higher-risk patients who could potentially benefit from intensified surveillance or future biomarker-guided adjuvant strategies, including immunotherapy-oriented approaches, these implications remain hypothesis-generating and require prospective validation.

Conclusion

CK7 positivity appears to delineate a distinct clinicopathological pattern of hepatocellular carcinoma. Rather than uniformly indicating an aggressive subtype, CK7 expression may reflect a specific exposure-related and differentiation-related background, thereby warranting further mechanistic study and external validation.

Abbreviations

AFP

Alpha-fetoprotein

ALD

Alcohol-associated liver disease

BDTT

Bile duct tumor thrombus

BCLC

Barcelona Clinic Liver Cancer

CI

Confidence interval

CK

Cytokeratin

DMFS

Distant metastasis–free survival

HBsAg

Hepatitis B surface antigen

HCC

Hepatocellular carcinoma

HCV-Ab

Hepatitis C virus antibody

HR

Hazard ratio

IHC

Immunohistochemistry

MVI

Microvascular invasion

OR

Odds ratio

OS

Overall survival

PVTT

Portal vein tumor thrombus

RFS

Recurrence-free survival

TBil

Total bilirubin

TG

Triglycerides

Authors’ contributions

XL conceived the study, curated the data, performed statistical analyses, and drafted the manuscript. JL, CY, and ZS contributed to data collection and interpretation. BX supervised the study and critically revised the manuscript. All authors read and approved the final manuscript.

Funding

This research was supported by the Guangxi Natural Science Foundation (Grant No. 2025GXNSFAA069076) and the Guangxi Science and Technology Program (Grant No. AD25069077).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was approved by the Ethics Committee of Guangxi Medical University Cancer Hospital and was conducted in accordance with the Declaration of Helsinki(protocol code: KY2026006;). Written informed consent was obtained from all patients prior to surgery.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.