Abstract Background The incidence rate and mortality rate of liver cancer were 11.6 (ranked 8th among cancers) and 10.7 (4th) per 100,000 individuals respectively globally in 2020, whereas in Korea, they were higher at 29.5 (7th) and 20.6 (2nd) per 100,000 individuals. The National Cancer Screening Program (NCSP) in Korea conducts biannual concurrent liver ultrasound and serum alpha-fetoprotein (AFP) measurement for hepatocellular carcinoma (HCC) surveillance in high-risk groups, including those with chronic hepatitis B and C, and liver cirrhosis. The effectiveness of surveillance is impacted by varying AFP cutoffs; however, within the NCSP, different cutoffs are being used, and it is not clear which cutoffs are being employed. Harmonization of AFP test results by major reagent manufacturers have been achieved, enabling the adoption of a uniform threshold. This study aims to assess the variation of AFP cutoffs among institutions within the NCSP and suggest a unified and optimal AFP threshold. Methods This study examined HCC screening results from the NCSP between 2018 and 2020, investigating unit and cutoff usage for AFP test across institutions each year (number of institutions were 4452 for 2018, 4754 for 2019, and 4847 for 2020). To determine the optimal AFP cutoff for HCC screening, datasets comprised unique patient results from 2018 to 2020 (number of patients = 819,644). Receiver operating characteristic (ROC) curve analyses were conducted for determine best AFP cutoffs. Cancer diagnosis was defined by billing records indicating HCC within one year post-screening. Results More than 96% of institutions used ng/mL as a unit of AFP test. Among institutions using ng/mL, the most frequently used AFP cutoff was 7 ng/mL, with frequency percentages of 75.7%, 72.2%, and 62.2% in 2018, 2019, 2020, respectively. The percentiles (1st, 10th, 90th, 99th) of AFP cutoff usage for the years 2018, 2019, and 2020 were (7, 7, 8.8, 15), (7, 7, 8.9, 15), and (7, 7, 9, 15) respectively. The AUC of ROC was 0.824 and some cut-off points with their sensitivity and specificity values were as follows: 5 ng/mL (0.648, 0.874); 10 (0.457, 0.976); 20 (0.327, 0.993); 40 (0.246, 0.997); 100 (0.169, 0.998); 200 (0.123, 0.999); 400 (0.084, 1). Conclusions Different cutoffs observed across institutions in NCSP. Our study cautiously suggests that the optimal AFP value for screening HCC in NCSP may range from 10 to 20 ng/mL, considering sensitivity and specificity when AFP is combined with ultrasound in regions with a high prevalence of HCC. For a more suitable AFP cutoff recommendation, additional analyses based on ultrasound findings and underlying diseases should be considered.
Read full abstract