Two Blood Signals Diverge Early and Refine Liver Cancer Therapy Response

For patients with hepatocellular carcinoma, the first weeks of treatment are a diagnostic gray zone. The tumor is too large to remove surgically, so oncologists deploy combinations of immunotherapy and targeted drugs, then wait for the first scan. That scan, performed roughly two months in, frequently shows stable disease — a category so broad it can conceal durable control, emerging progression, or treatment-related change. A new real-world study from Nantong University in China suggests that two routinely measured blood tests, read together rather than separately, can cut through much of this ambiguity within seventy days of starting therapy.

The research, published in Frontiers in Oncology, tracked 166 patients with unresectable hepatocellular carcinoma who received first-line immune-combination therapy. The team, led by Haitao Wu, Pengju Tang, and Zhong Chen of the Affiliated Hospital of Nantong University, paired early changes in alpha-fetoprotein (AFP), the most widely available circulating marker of tumor activity, with early changes in the modified albumin-bilirubin (ALBI) grade, an objective measure of liver reserve. What they found was striking: when these two signals moved in opposite directions, the discordance itself carried prognostic weight that neither signal alone could provide.

AFP is informative only in AFP-producing tumors, but early declines in this protein have repeatedly been linked to radiographic response and survival during immune checkpoint inhibitor treatment. A meta-analysis identified a decline of more than twenty percent within the first several weeks as a practical threshold. The ALBI score, calculated from total bilirubin and albumin, quantifies hepatic function without the subjective elements of the Child–Pugh classification. Modified ALBI categories provide finer discrimination within the intermediate range, where relatively small biochemical shifts can determine whether a patient remains eligible for continued therapy.

The study used a fixed day-70 landmark, a design choice that prevents follow-up from beginning before phenotype status can be determined. Patients with baseline AFP of at least 20 ng/mL were classified by whether AFP declined by at least twenty percent and whether modified ALBI grade worsened by at least one category. Four phenotypes emerged: concordant favorable (AFP response without liver deterioration), response-liver vulnerability (AFP response with deterioration), nonresponse-liver preserved (no AFP response, liver intact), and concordant failure (no AFP response with deterioration). Fifty-eight patients fell into the first category, thirty-six into the second, forty-three into the third, and twenty-nine into the fourth.

After adjustment for performance status, macrovascular invasion, extrahepatic metastasis, tumor burden score, baseline continuous ALBI score, and systemic regimen class, concordant failure showed the strongest association with shorter progression-free survival, with a hazard ratio of 2.85 (95% CI 1.63–4.99; P <0.001). Nonresponse-liver preserved also remained significantly associated with shorter progression-free survival (HR 1.69, 95% CI 1.02–2.79; P = 0.040), indicating that maintained hepatic reserve should not be equated with effective tumor control. The estimate for response-liver vulnerability was smaller and imprecise (HR 1.49, 95% CI 0.89–2.49; P = 0.130). When the three non-favorable phenotypes were collapsed into a single group, the adjusted hazard ratio was 1.79 (95% CI 1.18–2.72; P = 0.006).

Imaging response reinforced the biochemical picture. Progressive disease was most frequent in concordant failure by both RECIST 1.1 (44.8%) and modified RECIST (41.4%), while partial response was most common in response-liver vulnerability by modified RECIST (61.1%). The first post-treatment scan was performed a median of 66 days after treatment initiation, and 79 patients had RECIST stable disease while 59 had modified RECIST stable disease — precisely the equivocal categories where a serum marker could add the most value.

Overall survival estimates were less precise. Only 64 deaths were recorded, and although the largest point estimate again occurred in concordant failure (HR 2.09, 95% CI 1.00–4.38; P = 0.050), the confidence interval reached the null boundary. The authors treated the survival analysis as exploratory rather than confirmatory. Secondary post-landmark outcomes, including hepatic decompensation within three months (OR 2.42, 95% CI 0.90–6.50; P = 0.081), progression or death within 90 days (OR 2.31, 95% CI 0.94–5.73; P = 0.069), and treatment discontinuation within 90 days (OR 1.45, 95% CI 0.50–4.17; P = 0.492), were all imprecise after parsimonious adjustment.

Sensitivity analyses strengthened the main finding. The progression-free survival association for the combined non-favorable group was consistent in direction across a day-42 restricted analysis (HR 2.63, 95% CI 1.32–5.21), the actual laboratory-date analysis (HR 1.78, 95% CI 1.19–2.65), exclusion of Child–Pugh class B (HR 1.92, 95% CI 1.03–3.59), exclusion of early local therapy (HR 1.68, 95% CI 1.06–2.65), alternative AFP-response thresholds, and deterioration in standard ALBI grade. In a continuous analysis, each 0.5-unit increase in ALBI score was associated with shorter progression-free survival (HR 1.33, 95% CI 1.04–1.68; P = 0.021), while AFP response remained associated with longer survival (HR 0.57, 95% CI 0.39–0.83; P = 0.003). The interaction between AFP response and continuous ALBI change was not evident (P = 0.828), suggesting the two signals capture largely independent dimensions of early treatment assessment.

The clinical implication is nuanced. The authors emphasize that a non-favorable biochemical pattern should not be used as a stand-alone rule for changing therapy. Instead, in AFP-evaluable patients whose first imaging assessment is ambiguous, a non-favorable AFP-ALBI phenotype may support earlier multimodal reassessment of tumor status, liver function, and treatment feasibility. The framework does not describe AFP-nonproducing hepatocellular carcinoma, which accounted for 23.8% of the source cohort, nor does it capture the earliest treatment failures. The retrospective, single-center design leaves scope for residual confounding, and the phenotype may partly encode baseline liver reserve and tumor burden despite direct adjustment.

Prospective studies should standardize the timing of AFP, albumin, bilirubin, treatment-exposure, decompensation, and imaging assessments. Incorporating des-gamma-carboxy prothrombin, also known as PIVKA-II, may extend the approach to patients with low AFP. External cohorts with different dominant first-line regimens are needed to test transportability, while longitudinal models could determine whether recovery from, or persistence of, liver deterioration adds prognostic information. The most useful monitoring strategy, the authors suggest, should ultimately answer three separate questions: whether the tumor is responding, whether the liver can tolerate continued therapy, and whether the patient remains eligible for subsequent treatment.

Subject of Research: Early discordance between alpha-fetoprotein response and albumin-bilirubin deterioration refines response assessment in unresectable hepatocellular carcinoma receiving first-line immune-combination therapy: a real-world landmark cohort study

Article Title: Early discordance between alpha-fetoprotein response and albumin-bilirubin deterioration refines response assessment in unresectable hepatocellular carcinoma receiving first-line immune-combination therapy: a real-world landmark cohort study

Article References: Wu, H., Tang, P., Gong, J., Shi, X., Zhao, Y., & Chen, Z. (2026). Early discordance between alpha-fetoprotein response and albumin-bilirubin deterioration refines response assessment in unresectable hepatocellular carcinoma receiving first-line immune-combination therapy: a real-world landmark cohort study. Frontiers in Oncology, 16, Article 1954746. https://doi.org/10.3389/fonc.2026.1954746

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1954746

Keywords: Early, discordance, between, alpha-fetoprotein, response, albumin-bilirubin, deterioration, refines, assessment, unresectable, hepatocellular, carcinoma