Skip to main content


Post-Treatment Alpha-Fetoprotein Reduction Prognostic for Survival Outcomes in HCC

January 26, 2021


Ghassan Abou‑Alfa, MD, Memorial Sloan Kettering Cancer Center (New York, NY), discusses results from a study evaluating the association between post-treatment α‑fetoprotein reduction and real-world outcomes in US patients with hepatocellular carcinoma (HCC).

These results were presented at the virtual 2021 ASCO Gastrointestinal Cancers Symposium.


Hello, this is Ghassan Abou‑Alfa from Memorial Sloan Kettering Cancer Center. On behalf of all co‑authors, I'm honored to discuss with you the association between post‑treatment α‑fetoprotein reduction and outcome in real‑world US patients with advanced hepatocellular carcinoma.

No doubt we all know that meta‑analyses suggesting post‑treatment α‑fetoprotein reduction could be associated with a better outcome in hepatocellular carcinoma, but, of course, the cutoff was varied.

In the CELESTIAL study of cabozantinib versus placebo, if you recall, the 20% cutoff was brought in. Patients with more than or equal to 20% decrease from baseline α‑fetoprotein was associated with longer survival and progression‑free survival, regardless of the treatment.

However, the association between α‑fetoprotein reduction after initiation of first‑line therapy with tyrosine kinase inhibitor and outcomes is remaining unclear with advanced hepatocellular carcinoma patients.

Thus, what we did, we went ahead with the nationwide Flatiron Health electronic health record system and looked at deidentified data for patients with hepatocellular carcinoma and α‑fetoprotein who received first‑line tyrosine kinase inhibitor therapy, and looked at the closest to the initiation of the first‑line therapy being from about 30 days before up to 7 days plus for α‑fetoprotein, whatever value we had.

Afterwards, we looked at the post‑treatment α‑fetoprotein defined on the AFP level, the closest, again, to the 8 weeks after first‑line therapy and within plus/minus 2 weeks. We cut it off the same like the cabozantinib study, +/- 20% as a reduction from the baseline.

We looked at the real‑world overall survival data and progression‑free survival data. Of course, all this was done within a statistical analysis that looked at reduction versus no reduction.

We collected data on 436 patients. The etiologies were similar to what we have in our practice. This is real‑world data after all. 8% had documented hepatitis B, 50% hepatitis C, 47% obesity and diabetes, and 42% with heavy alcohol use. Sorafenib, understandably, was the most common used TKI, and 91% of the patients received sorafenib in first‑line TKI.

The median baseline α‑fetoprotein was 210 for the patients with more than or equal to 20% AFP reduction and 145 for the patients with less than 20 or no AFP reduction. There was a 35% decrease in the hazard ratio for the real‑world progression‑free survival observed in patients more than or equal to 20% AFP reduction versus those with less than 20% or no AFP reduction, with a hazard ratio of 0.65.

In addition, there was a 35% decrease in the hazard of death for patients with more than/equal to 20% AFP reduction versus less than 20% or no AFP reduction, again with a hazard ratio of 0.65.

For overall survival, more than/equal to 20% reduction in the AFP versus less than 20% or no reduction in AFP has shown the improvement in overall survival. If anything, it was clearly among patient with obesity or heavy alcohol versus those without the respective risk factor that I mentioned. Interestingly though, there was no statistically significant interaction observed.

If anything, this exercise can let us know that post‑treatment α‑fetoprotein reduction may be prognostic for real‑world progression‑free survival and real‑world overall survival among patients with advanced hepatocellular carcinoma on first‑line tyrosine kinase inhibitor.

By all means, this is great data that is dependent on first and largest study assessing the association between AFP post‑treatment and survival outcome. The data, as curated with abstraction, confirmed hepatocellular carcinoma diagnosis, etiologic risk factor, and longitudinal treatment assessments.

Understandably though, the data is limited by the potential survival bias, as patients had to survive long enough to get the post‑treatment AFP level, single cutoff and single time window to post‑treatment AFP reduction, and potential for missing certain AFP measurements and the complete clinical information.

Nonetheless, this is a great suggestion for clarifying further if prognostic value differs by HCC risk factor profile and, of course, explore alternative thresholds to define α‑fetoprotein reduction.

Back to Top