Recurrence after surgery for hepatocellular carcinoma: “The seed and the soil” - a multicentre retrospective cohort study

Abstract

BACKGROUND

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related mortality. Liver resection provides a curative option, though recurrence following resection remains a major clinical challenge, as adjuvant therapies evolve, and preservation of liver function becomes increasingly important.

AIM

To evaluate factors associated with HCC recurrence in a real-world setting.

METHODS

We retrospectively reviewed records of patients with HCC who underwent primary liver resection at three tertiary referral hospitals between January 2008 and May 2022. Baseline patient and tumour characteristics were assessed. Risk factors for recurrence were analysed using multivariate Cox regression.

RESULTS

In this cohort, 235 patients underwent surgical resection for HCC. Median survival was 80.9 months. HCC recurred in 94 (40%) patients with median time to recurrence 18.8 months (interquartile range: 8.6-35). An albumin-bilirubin (ALBI) grade of ≥ 2 [adjusted hazard ratio (aHR): 2.02 (1.27-3.21), P = 0.003], the presence of cirrhosis [aHR: 2.03 (1.27-3.25), P = 0.003], and Barcelona Clinic Liver Cancer (BCLC) stage B/C [aHR: 2.23 (1.23-4.03) P = 0.008] were independently associated with increased risk of recurrence. When combining these risk factors, the adjusted HR for recurrence was 2.73 (1.57-4.75), 4.32 (2.29-8.12) and 7.52 (2.47-22.93) in those with 1, 2, or 3 factors compared to those with 0 (P < 0.001).

CONCLUSION

HCC recurrence following primary liver resection was independently associated with cirrhosis, ALBI grade ≥ 2 and BCLC stage B/C. The seed (the liver cancer as reflected by BCLC stage), and the soil (the liver function and presence or absence of cirrhosis) are relevant considerations for recurrence prediction.

Key Words: Hepatocellular carcinoma; Prognostic prediction; Surgery; Recurrence; Albumin-bilirubin grade

Core Tip: Recurrence following curative liver resection for hepatocellular carcinoma (HCC) remains a major clinical challenge. In this multicentre retrospective cohort study, cirrhosis, albumin-bilirubin grade ≥ 2, and Barcelona Clinic Liver Cancer stage B/C were independently associated with recurrence. These findings highlight the importance of both tumour burden and underlying liver function when estimating postoperative recurrence risk. Improved preoperative risk stratification may support patient counselling and multidisciplinary decision-making in the management of HCC.

INTRODUCTION

Hepatocellular carcinoma (HCC) is the leading cause of primary liver cancer globally, responsible for 80% of cases and ranking as the third leading cause of cancer-related mortality, with over 800000 deaths annually[1-4]. In Australia, HCC incidence and mortality continue to rise, with liver cancer mortality increasing from 4 to 7 per 100000 over the past two decades[5]. Chronic hepatitis C virus (HCV) infection is a leading cause of HCC in Australia, although other aetiologies such as hepatitis B virus (HBV), alcohol-related liver disease, and metabolic dysfunction-associated steatotic liver disease (MASLD) are increasingly significant contributors[6].

HCC typically develops in the setting of chronic liver disease and cirrhosis, where repeated cycles of hepatocyte injury, inflammation, and regeneration promote hepatocarcinogenesis[6-8]. The most common global risk factors include chronic HBV and HCV infection, alcohol-related liver disease, and MASLD, although their relative contributions vary geographically[8]. In addition to underlying liver disease, tumour-related characteristics such as tumour size, vascular invasion, multiplicity of lesions, and elevated serum alpha-fetoprotein have been associated with tumour progression and recurrence following curative treatment[8]. Understanding the interaction between tumour biology and the underlying hepatic environment remains essential when considering treatment strategies and predicting outcomes following liver resection.

HCC is frequently diagnosed at advanced stages, limiting curative options and underscoring gaps in surveillance and early intervention strategies[9]. Current treatment pathways are developed in a multidisciplinary team (MDT), guided by a staging system, most commonly the Barcelona Clinic Liver Cancer (BCLC) system, which utilises both tumour characteristics and underlying liver function to guide clinical decision-making. For BCLC stage 0/A (early-stage) HCC with single tumours or ≤ 3 nodules < 3 cm in size, surgical resection is considered potentially curative[10]. Additionally, suitability for surgery requires an anticipation that resection will be complete, that adequate functional liver remnant remains and that there are no major comorbidities or significant portal hypertension that would create a high risk of surgical morbidity or mortality[11,12]. Improved surgical techniques and perioperative management have expanded the proportion of patients eligible for curative-intent resection such that a significant number of patients with HCC outside BCLC stage 0/A may be offered surgery, influenced by local practice and surgical expertise[11]. The enthusiasm for surgery, even beyond BCLC 0/A, is supported by several meta-analyses suggesting surgical resection has been associated with improved survival in selected patients with intermediate- and advanced-stage HCC compared with non-surgical therapies[11-13]. Although HCC recurrence is an acknowledged problem, the determination of recurrence risk is not a formal part of the surgical decision making in any current guidelines[14-16].

Recurrence after surgery is related to tumour factors, including multi-nodularity, tumour size ≥ 5 cm, lymphovascular invasion, elevated alpha-fetoprotein (AFP) (> 400 ng/mL), poor tumour differentiation, liver capsule involvement, and R1 resection[17,18]. In transplantation, in an effort to avoid wasting grafts, scores such as Milan and University of California San Francisco (UCSF) tumour criteria have been designed to determine recurrence risk and accordingly transplant eligibility[19,20]. It makes sense that tumour factors alone predict recurrence after the entire diseased liver is removed with liver transplantation. Outside of the transplant setting, the diseased liver remains and therefore factors such as metachronous microscopic disease, intra-hepatic seeding during surgery despite clear histological margins or subsequent de novo tumour formation in patients with underlying field defects such as cirrhosis or HBV become relevant[17,18].

The interplay between tumour biology and the underlying hepatic environment reflects a consideration in cancer recurrence unique to the liver, where elements about the tumour itself (the seed) and the underlying liver disease (the soil) are relevant.

This retrospective study examined a multicentre HCC cohort undergoing primary liver resection with curative intent aiming particularly to characterise pre-operative tumour and liver factors that may be associated with recurrence.

MATERIALS AND METHODS

Study design and patient population

This was a retrospective cohort study of patients with incident HCC managed at three tertiary referral hospitals in New South Wales, Australia, between January 2008 and May 2022. Patients were identified through an established multicentre database populated using the International Classification of Diseases, 11^th Revision code for malignant neoplasm of the liver (2C12.0) within the electronic medical records of the three participating hospitals. The cohort was complete with all patients with confirmed HCC, who had a date of HCC diagnosis within the study period and received their initial/primary HCC resection at one of the three hospitals involved. Patients with combined HCC-cholangiocarcinoma (n = 9) or fibrolamellar-HCC (n = 4) were excluded. Prior publications have been reported from this cohort whose focus initially was on HCV treatment related outcomes in patients with HCC. Data collection involved reviewing clinical notes, multidisciplinary team reports, medication lists, and laboratory, transient elastography (FibroScan^®), histopathology, and imaging results. Missing data were addressed through cross-referencing hospital databases, and efforts were made to minimise selection bias by including all patients meeting the inclusion criteria across three centres. The study protocol was approved by St Vincent’s Hospital Human Research Ethics Committee (2025/ETH01121), as well as local ethics committees at all study sites. The initial ethics approval was titled in consideration of the primary endpoint, but all elements collected for that were analysed here within the ethics approval of secondary endpoint analysis. The study met the ethical guidelines of the Declarations of Helsinki and the Declaration of Istanbul. As this was a retrospective study using de-identified patient data, the requirement for informed consent was waived by the ethics committees. This study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.

Study definitions

Patients were categorised as HCV-HCC, HBV-HCC or non-viral-HCC, based on their primary aetiology. All patients with a history of HCV, confirmed by a positive HCV-RNA or a history of HCV therapy, were classified as HCV-HCC. Patients with HBV, confirmed by a positive surface antigen, were classified as HBV-HCC. Patients without HCV or HBV were classified as non-viral-HCC. Alcohol related liver disease and MASLD were ascertained from clinical documentation. The presence of coexisting alcohol-related liver disease was recorded for patients in the HCV-HCC and HBV-HCC groups.

HCC diagnosis was confirmed according to the American Association for the Study of Liver Diseases diagnosis guidelines[21]. The date of HCC diagnosis was recorded as the date of multiphase imaging, or the date of biopsy if multiphase imaging was not performed. Cirrhosis was determined by histopathology, a history of decompensation (ascites, hepatic encephalopathy, or portal hypertensive bleeding), a nodular liver surface on imaging, or the impression documented in clinical notes.

Patients were considered to have HCC detected through surveillance if they were asymptomatic and the indication for imaging was to screen for HCC. HCC cure was determined by clear histological margins and complete radiological response. HCC recurrence was defined as local recurrence, or a new intrahepatic lesion determined through imaging with or without a MDT discussion, measured from time of HCC cure.

Statistical analysis

Clinical characteristics of patients with HCC diagnosed in June 2008 to May 2022 were compared (overall and by HCC aetiology). A χ² test for categorical variables and a Mann-Whitney U test for continuous variables were used to identify variables associated with recurrence. Variables with P < 0.10 were entered into Cox proportional hazards regression models to identify independent predictors while adjusting for potential confounders. Results were expressed as hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between baseline factors and HCC recurrence. For factors with potential collinearity (determined by a variance inflation factor equal to or greater than 4), a separate multivariate analysis was conducted. Model for end-stage liver disease (MELD) and Child-Pugh (CP) scores were analysed separately and restricted to cirrhotic patients due to overlapping components and resultant collinearity. Non-cirrhotic patients were excluded from these analyses. ALBI score, developed in liver cancer cohorts and applicable irrespective of cirrhosis status, was analysed across the full cohort together with cirrhosis presence. All results with a P value of < 0.05 were considered significant. All analyses were performed using Stata (v17.0; Stata Corp, College Station, TX, United States). Retrospective study design limitations, including potential confounding variables, were mitigated through multivariate analyses.

RESULTS

Baseline characteristics

A total of 308 patients with HCC underwent surgical resection. Seventy-three were excluded as they either had metastatic disease (n = 13) or surgical resection was not the primary HCC treatment (n = 60). Thus, 235 patients were included in this analysis. A single patient (0.4%) died within 90-day of surgery. Median survival was 80.9 months and 80% of deaths were liver related. Median age at time of diagnosis was 64 years (57-70), and the majority were male (78%). Aetiology of liver disease was HCV in 71 (30%), HBV in 94 (40%) and non-viral aetiologies in 70 (30%). Ninety-two (39%) were non-cirrhotic. Of 143 (61%) with cirrhosis, 112 (48%) were CP-A5 and 31 (13%) CP-A6 or greater. MELD 3.0 score in those with cirrhosis was 6, 7, 8 and 9 or greater in 27 (11%), 35 (15%), 31 (13%) and 50 (21%) of patients respectively. ALBI grade was 1 in 157 (67%) and ≥ 2 in 78 (33%) of patients.

Tumour characteristics were as follows: AFP was < 400 ng/mL in 188 (80%) and ≥ 400 ng/mL in 22 (9%) (unknown in 25). HCC tumour burden was single nodule in 192 (82%), two nodules in 29 (12%), and 3 or more nodules in 14 (6%) patients. The largest nodule was less than 3.0 cm in size in 100 (43%) patients, 3.0-4.9 cm in 68 (29%), 5.0-10.0 cm in 50 (22%), and greater than 10.0 cm in 17 (7%). Macrovascular invasion was present in 14 (6%) of patients. Most patients, 195 (83%), were within UCSF criteria. BCLC stage was 0, A, B and C in 42 (18%), 159 (68%), 20 (8%) and 14 (6%) respectively. Histological assessment of the resected lesion post-procedure identified incomplete resection in view of positive margins in 16 (7%) patients.

Recurrence of HCC following resection with curative intent

Recurrence occurred in 94 patients (39%) with median time to recurrence of 18.8 months (interquartile range: 8.6-35). Recurrence incidence was 10.5 per 100 person-years (95%CI: 8.6-12.9). Recurrence was associated with a trend toward worse survival [HR: 1.51 (0.98-2.32), P = 0.059].

We examined individual factors and the associations with post-operative recurrence. Recurrence was similar in HCV-HCC (51%), compared to those with HBV-HCC (35%) and non-viral-HCC (36%) (Table 1). Recurrence occurred 35/92 (38%) in those without and 59/143 (41%) in those with cirrhosis. Recurrence was lower in those with MELD 3.0 score of 6 (22%) compared to higher scores: 7 (49%), 8 (42%), and ≥ 9 (46%). Recurrence was lower in those with CP-A5 (38%) and higher in those with more advanced CP ≥ 6 (52%). Recurrence occurred in 39% of those with ALBI grade 1 and 42% of those with ALBI grade ≥ 2. Recurrence occurred in the same proportion (40%) of those inside and outside UCSF criteria, and similar in those with AFP < 400 ng/mL (40%) and AFP ≥ 400 ng/mL (41%). Recurrence occurred in 42%, 31% and 36% of patients with 1 ng/mL, 2 ng/mL, and ≥ 3 tumour nodules, respectively. Based on the size of the largest nodule/s, recurrence occurred in 36%, 40%, 52%, and 29% of those with largest nodule/s < 3.0 cm, 3.0-4.9 cm, 5.0-10.0 cm, and > 10.0 cm, respectively. Recurrence occurred in 36% and 40% of those with and without evidence of macrovascular invasion, respectively.

Table 1 Baseline characteristics of patients with hepatocellular carcinoma undergoing primary liver resection.

	Baseline characteristics of cohort (n = 235)	Recurrence, n = 94 (40)	No recurrence, n = 141 (60)
Patient factors
Recurrence by aetiology
HCV	71 (30)	36 (51)	35 (49)
HBV	94 (40)	33 (35)	61 (65)
MASLD/alcohol	70 (30)	25 (36)	45 (34)
Age, years [median (q1-q3)]	64 (57-70)	64 (56-71)	65 (57-70)
Male sex	183 (78)	74 (40)	109 (60)
Diabetes (T2DM)	79 (34)	28 (35)	51 (65)
Liver factors
Cirrhosis
No	92 (39)	35 (38)	57 (62)
Yes	143 (61)	59 (41)	84 (59)
Child Pugh
No cirrhosis	92 (39)	35 (38)	57 (62)
A5	112 (48)	43 (38)	69 (62)
≥ A6	31 (13)	16 (52)	15 (48)
MELD 3.0 score
No cirrhosis	92 (39)	35 (38)	57 (62)
6	27 (11)	6 (22)	21 (78)
7	35 (15)	17 (49)	18 (51)
8	31 (13)	13 (42)	18 (58)
≥ 9	50 (21)	23 (46)	27 (54)
ALBI grade
1	157 (67)	61 (39)	96 (61)
≥ 2	78 (33)	33 (42)	45 (58)
Platelet count
≤ 150	62 (26)	23 (37)	39 (63)
> 150	173 (74)	71 (41)	102 (59)
Tumour factors
AFP (ng/mL)
< 400	188 (80)	75 (40)	113 (60)
≥ 400	22 (9)	9 (41)	13 (59)
Unknown AFP at diagnosis	25 (11)	10 (40)	15 (60)
BCLC
0	42 (18)	14 (33)	28 (67)
A	159 (68)	66 (42)	93 (58)
B	20 (8)	9 (45)	11 (55)
C	14 (6)	5 (36)	9 (64)
UCSF criteria
In	195 (83)	78 (40)	117 (60)
Out	40 (17)	16 (40)	24 (60)
Number of nodules
1	192 (82)	80 (42)	112 (58)
2	29 (12)	9 (31)	20 (69)
≥ 3	14 (6)	5 (36)	9 (64)
Size of largest nodule/s
< 3 cm	100 (43)	36 (36)	64 (64)
3-4.9 cm	68 (29)	27 (40)	41 (60)
5-10 cm	50 (22)	26 (52)	24 (48)
> 10 cm	17 (7)	5 (29)	12 (71)
Presence of macroscopic vascular invasion (or tumour rupture/bleed)
Yes	14 (6)	5 (36)	9 (74)
No	221 (94)	89 (40)	132 (60)

Clinical, liver disease-related, and tumour-related characteristics of the study cohort stratified by recurrence status. Values are presented as n (%) or median (interquartile range). We performed univariate (unadjusted) analysis and identified factors associated with hepatocellular carcinoma recurrence including age, the presence of cirrhosis, Child-Pugh score, model for end-stage liver disease 3.0 score, albumin-bilirubin grade grade, aetiology, and University of California San Francisco criteria, Barcelona Clinic Liver Cancer staging system (BCLC) stage, nodule size and macrovascular invasion (Table 2). Multivariate analysis found the albumin-bilirubin score, if ≥ 2; adjusted hazard ratio (aHR): 2.24 (1.43-3.52) P < 0.001, the presence of cirrhosis; aHR: 2.21 (1.42-3.43) P < 0.001 and those with Barcelona Clinic Liver Cancer stage B/C; aHR: 2.51 (1.41-4.47) P = 0.002 as independent predictors of recurrence. AFP: Alpha-fetoprotein; ALBI: Albumin-bilirubin grade; BCLC: Barcelona Clinic Liver Cancer; HBV: Hepatitis B virus; HCV: Hepatitis C virus; MASLD: Metabolic dysfunction-associated steatotic liver disease; MELD: Model for end-stage liver disease; T2DM: Type 2 diabetes mellitus; UCSF: University of California San Francisco criteria.

Table 2 Factors associated with hepatocellular carcinoma recurrence following primary liver resection, Cox proportional hazard model assessing hepatocellular carcinoma recurrence following primary liver resection (n = 235).

		Unadjusted model		Adjusted model 1		Adjusted model 2		Adjusted model 3		Adjusted model 4
Characteristics	Rate of recurrence (95%CI), per 100 person-years	HR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)	P value
Patient factors
Age, year	N/A	1.02 (1.01-1.04)	0.019	1.03 (1.00-1.05)	0.019	1.02 (1.00-1.05)	0.043	1.02 (1.00-1.04)	0.108	1.03 (1.01-1.05)	0.017
Sex
Male	10.5 (8.4-13.2)	1.00
Female	11.3 (7.3-17.6)	1.14 (0.70-1.88)	0.598
Diabetes (T2DM)
Nil diabetes	10.8 (8.5-13.8)	1.00
T2DM	10.3 (7.1-15.0)	1.09 (0.70-1.71)	0.704
Liver factors
Cirrhosis
No	7.8 (5.6-10.9)	1.00								1.00
Yes	13.5 (10.5-17.3)	2.21 (1.42-3.43)	< 0.001							2.03 (1.27-3.25)	0.003
Child Pugh
No cirrhosis	8.0 (5.8-11.2)	1.00		1.00
A5	11.8 (8.8-15.9)	1.87 (1.17-2.97)	0.008	1.73 (1.06-2.83)	0.029
> A6	19.7 (12.1-32.2)	3.24 (1.76-5.98)	< 0.001	2.85 (1.50-5.42)	0.001
MELD 3.0 score
No cirrhosis	7.8 (5.6-10.9)	1.00				1.00
6	6.6 (2.9-14.6)	1.18 (0.48-2.86)	0.718			1.13 (0.46-2.80)	0.790
7	12.1 (7.5-19.5)	1.68 (0.93-3.04)	0.083			1.51 (0.81-2.82)	0.191
8	16.2 (9.4-27.8)	3.06 (1.57-5.94)	0.001			2.77 (1.35-5.67)	0.005
≥ 9	18.0 (12.1-26.9)	3.29 (1.89-5.71)	< 0.001			2.97 (1.68-5.27)	< 0.001
ALBI grade
1	9.2 (7.1-11.8)	1.00						1.00
≥ 2	15.2 (10.8-21.4)	2.24 (1.43-3.52)	< 0.001					2.02 (1.27-3.21)	0.003
Tumour factors
Aetiology of HCC
HBV	8.8 (6.3-12.4)	1.00		1.00		1.00		1.00		1.00
HCV	14.6 (10.5-20.3)	1.75 (1.08-2.81)	0.022	1.22 (0.73-2.03)	0.448	1.18 (0.70-1.99)	0.532	1.45 (0.89-2.35)	0.132	1.20 (0.72-1.99)	0.478
MASLD/ETOH	9.6 (6.5-14.1)	1.27 (0.74-2.17)	0.378	0.91 (0.51-1.62)	0.758	0.85 (0.48-1.53)	0.595	0.99 (0.56-1.73)	0.966	0.87 (0.49-1.55)	0.644
AFP (ng/mL)
< 400 (n = 190)	10.2 (8.1-12.8)	1.00
≥ 400 (n = 27)	15.7 (8.2-30.2)	1.53 (0.76-3.10)	0.234
BCLC
0-A	9.9 (7.9-12.3)	1.00		1.00		1.00		1.00		1.00
B-C	19.9 (11.8-33.7)	2.51 (1.41-4.47)	0.002	2.01 (1.12-3.60)	0.019	2.23 (1.23-4.03)	0.008	2.38 (1.33-4.25)	0.003	2.05 (1.15-3.67)	0.015
Number of nodules
1	10.3 (8.2-12.8)	1.00
2	13.9 (7.3-26.8)	1.77 (0.88-3.59)	0.112
≥ 3	13.5 (5.6-32.3)	1.45 (0.59-3.60)	0.418
Size of largest nodule/s
< 3 cm	8.1 (5.9-11.3)	1.00
3-4.9 cm	11.0 (7.5-16.1)	1.50 (0.90-2.49)	0.118
5-10 cm	17.2 (11.7-25.3)	2.67 (1.58-4.52)	< 0.001
> 10 cm	11.9 (5.0-28.6)	1.87 (0.73-4.82)	0.195
Presence of macroscopic vascular invasion (or tumour rupture/bleed)
No	10.3 (8.4-12.7)	1.00
Yes	26.8 (11.2-64.4)	2.96 (1.19-7.39)	0.020

Univariate and multivariate Cox proportional hazards regression analyses evaluating factors associated with recurrence following curative-intent liver resection. Results are presented as hazard ratios with 95% confidence intervals. AFP: Alpha-fetoprotein; ALBI: Albumin-bilirubin grade; BCLC: Barcelona Clinic Liver Cancer; CP: Child-Pugh score; MELD: Model for end-stage liver disease; HR: Hazard ratio; HCV: Hepatitis C virus; MASLD: Metabolic dysfunction-associated steatotic liver disease; ETOH: Chronic ethanol; HBV: Hepatitis B virus; HCC: Hepatocellular carcinoma; MELD: Model for end-stage liver disease; T2DM: Type 2 diabetes mellitus.

We performed univariate (unadjusted) analysis and identified factors associated with HCC recurrence including age, the presence of cirrhosis, CP score, MELD 3.0 score, ALBI grade, aetiology, and UCSF criteria, BCLC stage, nodule size and macrovascular invasion (Table 2). Multivariate analysis found the ALBI score, if ≥ 2; adjusted hazard ratio (aHR): 2.24 (1.43-3.52) P < 0.001, the presence of cirrhosis; aHR: 2.21 (1.42-3.43) P < 0.001 and those with BCLC stage B/C; aHR: 2.51 (1.41-4.47) P = 0.002 as independent predictors of recurrence.

We notice that 25.9%, 43.4%, 26.3% and 4.3% of the cohort had a 0, 1, 2 or 3 of these negative factors respectively (Figure 1). If we were to add the risks, the adjusted HR of recurrence risk was 2.74 (1.57-4.76), 4.32 (2.29-8.12) and 7.53 (2.47-22.93) for either 1, 2 or 3 of these risk factors respectively compared with the group with no risk factors (P < 0.001) (Figure 1, Table 3).

Figure 1 Kaplan-Meier analysis of hepatocellular carcinoma recurrence following curative-intent primary liver resection. The Kaplan-Meier curve demonstrates recurrence-free survival stratified by the number of pre-operative risk factors present: Cirrhosis, albumin-bilirubin grade ≥ 2, and Barcelona Clinic Liver Cancer stage B or C. Increasing numbers of these risk factors were associated with progressively lower recurrence-free survival following curative-intent resection. HCC: Hepatocellular carcinoma.

Table 3 Recurrence risk according to the number of pre-operative risk factors (cirrhosis, albumin-bilirubin grade ≥ 2, Barcelona Clinic Liver Cancer staging system stage B/C).

Number of risk factors	Criteria present	aHR (95%CI)	P value	n (%)
0	None	Reference	Reference	60 (26)
1	Any one of: Cirrhosis, ALBI ≥ 2, or BCLC stage B/C	2.74 (1.57-4.76)	< 0.001	102 (43)
2	Any two	4.32 (2.29-8.12)	< 0.001	62 (26)
3	All three	7.53 (2.47-22.93)	< 0.001	10 (4)

aHR: Adjusted hazard ratio; CI: Confidence interval; ALBI: Albumin-bilirubin; BCLC: Barcelona Clinic Liver Cancer.

DISCUSSION

HCC continues to pose a significant global health challenge, with high recurrence even after curative-intent resections. Our multicentre cohort study identified three factors that were independently associated with an increased risk of recurrence, namely the presence of cirrhosis, ALBI grade ≥ 2, and BCLC stage B/C. The greater the number of these factors present, the higher the associated recurrence rate. Recurrence in this setting likely reflects the influence of tumour biology and the underlying hepatic microenvironment. In this framework, tumour stage represents the “seed”, while the condition of the liver parenchyma, reflected by cirrhosis and liver function, represents the “soil”. Incorporating these considerations into preoperative planning in this time of evolving systemic and locoregional therapies may tip the balance to more conservative strategies when the risk of recurrence is very high.

In our cohort, HCC recurrence occurred in 39% of patients following liver resection, with a median time to recurrence of 18.8 months. This aligns with previous studies reporting similarly high recurrence after curative resection. For example, Hu et al[22] observed recurrence in 45% of patients within BCLC criteria, paralleling the rates seen in our cohort. The high rate of recurrence reflects not only residual microscopic disease but also the field defect inherent in chronic diseased livers where de novo carcinogenesis is the culprit, even after successful surgical resection of a tumour[23]. Recurrence tended to confer a worse survival in our cohort. We did not analyse predictors of survival due to the many confounders including changing treatment options for recurrence over the study period.

Poor liver function, indicated by a MELD 3.0 score ≥ 7, 8 and > 9, was associated with higher risk of recurrence. In our study, measures of liver dysfunction, namely the ALBI grade and CP score were also found to be independently associated. Previous studies, such as Xu et al[18], have reported that advanced liver disease, as measured by CP scores, is a predictor of late recurrence. The CP is subjective and a blunt instrument for detecting subtle liver dysfunction. In our study, only 13% of patients were CP-A6 or more. Whilst a MELD score is a more granular measure of liver dysfunction, allowing greater separation of the cohort than CP, MELD was developed and validated to predict short-term survival in patients with advanced cirrhosis, not for recurrence risk stratification in patients undergoing resection. For example, creatinine and prothrombin time in transplant populations may be disturbed due to liver dysfunction but may indicate other things in those with HCC under consideration for surgery. We found the ALBI grade, increasingly used in other HCC management settings, contained a simple objective score more relevant to resection consideration[24-26]. In a retrospective series Lee et al[25], ALBI grade has been shown to be an independent risk factor for early recurrence, and when combined with 4 other variables (ALBI grade ≥ 2, multiple tumours, tumour size equal to or greater than 2 cm, serum AFP equal or greater than 20 ng/mL and total tumour volume equal or larger than 227 cm³) into a nomogram was also found to be associated with recurrence after surgery with curative intent[26].

The BCLC staging system demonstrated significant prognostic value in predicting recurrence in our cohort. In this cohort who already had sufficient CP score and performance status to be considered for surgery, the BCLC criteria thus reflects tumour characteristics which we know is a predictor of recurrence[9,17,18]. Nevertheless, in our and other groups, the willingness to offer surgery outside of BCLC algorithm in an effort to offer cure is not uncommon. In our cohort, 34 of 235 patients (14%) underwent resection despite being classified as BCLC stage B/C[16,26]. The selection of such patients usually reflects consideration of surgical risk. Our data suggests that BCLC may provide useful information regarding recurrence risk even when surgical feasibility and risk is considered acceptable.

In contrast to some prior studies, neither HCC aetiology (whether HBV, HCV, or non-viral), positive margins, or AFP levels emerged as independent predictors of HCC recurrence in our cohort[17-19]. Earlier research has suggested that viral aetiologies, particularly HBV, correlate with a higher risk of post-operative recurrence[9,20,21]. Regional and chronological variations in aetiological characteristics may influence the impact of aetiology on recurrence for example by the reduced recurrence of HCC described after HCV treatment[27]. A type 2 error due to limitations in subgroup size may be responsible for lack of demonstrated risk with these factors.

Overall, when considering surgery for HCC, an appreciation of recurrence risk using the presence of cirrhosis, the BCLC stage and the ALBI score might be useful. If the liver remnant is borderline or risks of surgery high or likelihood of complete resection low, then the risk of recurrence could be a tipping point at which alternate therapies may be considered. Multidisciplinary teams should consider the potential benefits of resection against the risks of recurrence. Clinical trials will determine the role of neoadjuvant therapies. Systemic immune or targeted, radiation or locoregional therapies are continuously evolving yet can only be made available to those with sufficient liver reserve.

This study has limitations in view of its retrospective nature. We could not discern whether the recurrence was true recurrence or de novo carcinogenesis in a diseased liver. There is no cut off to distinguish these two events, and our study reflects this, with recurrences emerging continually rather than a clear bi-modal time distribution which could allow interpretation of different mechanisms. There is always potential for selection bias and unmeasured confounders in a retrospective study design, despite efforts to mitigate these through multivariate analyses. Our cohort was derived from three tertiary hospitals in Australia, which may limit the generalisability of these findings. We do not have data on the nature of the surgery (lobar hepatectomy or more limited resection), or the nature of the recurrence in each case. All surgery was performed by qualified hepatobiliary surgeons in academic centres. Nevertheless, this is a limitation in the retrospective data set extracted from available electronic medical record.

Prospective validation of this model would be valuable. Such studies may help determine whether preoperative risk stratification can meaningfully influence surgical decision-making or guide the use of adjunctive and neo-adjuvant therapies.

Future directions and clinical relevance

The study examines preoperative predictors of HCC recurrence after surgery. Current surgical decision-making primarily focuses on tumour resectability, functional liver remnant, and perioperative risk. However, recurrence remains common even after technically successful resections. Incorporating a philosophy of considering the seed and the soil particularly the cirrhosis status, the ALBI grade, and BCLC stage may assist clinicians in better identifying patients at higher risk of recurrence.

Future prospective studies are required to validate these findings in independent cohorts and to determine whether recurrence risk prediction can meaningfully guide clinical management. Patients at high risk may be the subgroup who benefit from emerging neoadjuvant or adjuvant therapies, closer postoperative surveillance, or alternative non-surgical locoregional or systemic liver preserving treatment strategies. As systemic therapies and immunotherapies for HCC continue to evolve, integrating recurrence risk assessment into multidisciplinary decision-making may improve long-term outcomes.

CONCLUSION

In conclusion, our study identified three independent predictors of HCC recurrence following primary resection: Cirrhosis, ALBI ≥ 2 and BCLC stage B/C. To avoid futile surgery, further evidence is required to refine the approach for patients at highest risk of recurrence. Guideline recommendations about surgery, which consider the operative risk, the estimation of functional liver remnant and the BCLC criteria should additionally encourage discussion about recurrence risk until a more certain approach is evidenced by prospective studies.