The predictive ability of serum alpha-fetoprotein for hepatocellular carcinoma is linked with the characteristics of the target population at surveillance.

BACKGROUND

Serum alpha-fetoprotein (AFP) is the most important tumor marker for hepatocellular carcinoma (HCC). The reported predictive accuracy of AFP for HCC widely varied. This study investigated the factors contributory to the heterogeneity of the ability of AFP to detect HCC.

METHODS

A total of 1,135 patients were categorized into four groups: HCC undergoing surgical resection (n = 248), chronic hepatitis B (CHB, n = 413), chronic hepatitis C (CHC, n = 207), and liver cirrhosis (LC, n = 267). The area under the receiver operating characteristic curve (AUC) was estimated in different combinations.

RESULTS

The AUC was the highest when HCC patients were co-analyzed with patients with LC (0.805), followed by co-analyzing patients with CHB (0.797) and CHC (0.740). The optimal cutoffs for AFP were between 26 and 32 ng/ml. Patients with tumor size </=3 cm had a lower positive predictive value (PPV, 30%) compared to patients with tumor size >3 cm (46%), and HBsAg-negative and anti-HCV-positive patients had the lowest PPV (38%) compared to other groups (60%, 100%, and 82%) at a cutoff at 20 ng/ml. The AUC was lower for hepatitis B-negative subjects (0.684 and 0.509), compared to hepatitis B-positive subjects (0.826 and 0.806) stratified by the status of HCV. Elevated ALT >80 U/L and HCC independently predicted increased (>20 ng/ml) AFP levels.

CONCLUSIONS

Serum AFP had a fairly stable predictive accuracy for HCC, with an optimal cutoff around 30 ng/ml. ALT level, viral status, and tumor size may significantly confound its ability to detect HCC.