We read with interest the article1 and subsequent correspondence2 by Khalili et al. regarding the use of biopsy for diagnosing small (1-2 cm) liver nodules that remain indeterminate after imaging studies performed during hepatocellular carcinoma (HCC) surveillance. The authors found a low (23%) prevalence of malignancy in these nodules, along with low rates of biopsy positivity, and they concluded that biopsy should be reserved for lesions displaying arterial hypervascularization or associated with synchronous HCC. In our opinion, this study has several obvious major weaknesses that need to be highlighted.
First, the reference standard for classifying nodules as malignant were growth on follow-up imaging, which was defined as “a 30% change in lesional diameter,” and features of HCC on biopsy, whereas nodules were considered benign “only if they remained stable on imaging for ≥18 months, irrespective of biopsy findings.” This means, for instance, that a nodule with histological features of HCC whose diameter increased by 25% (e.g., from 15 to 18.75 mm, which corresponds to a 2-fold increase in volume, i.e., from 1.77 to 3.45 mL) was considered benign! This obvious underestimation goes a long way toward explaining the low prevalence of malignancy reported in this series.
Second, the authors' aim was to evaluate the impact of biopsy, and yet only 30 of the 85 nodules they analyzed were biopsied (the decision to biopsy or not being left to the hepatologist in charge of the patient). The result is an obvious bias due to the retrospective design of the study. In addition, one of the reasons given for not performing biopsy was that “nodules were not visible on grayscale US.” This is puzzling, since nodule detection on “routine US surveillance” was the criteria used for case inclusion.
Third, 12 (40%) out of the 30 biopsies that were performed yielded “nonlesional parenchyma,” which suggests that they were sampling errors. Indeed, at least two of these nodules displayed growth on follow-up imaging and were ultimately classified as malignant.
In light of the lesions apparent stability over time and the low rate of malignancy diagnosed in this series—both results that may well derive from methodological errors—the authors warn that HCC may be “overdiagnosed” and wonder whether histologically malignant nodules not progressing to advanced HCC can justifiably be diagnosed and treated as HCC. They later assert2 that “long-term stability (at imaging) represents a stronger reference standard than biopsy.”
If HCC has recently become a potentially curable disease, this is due mainly to its early diagnosis. There is a massive body of evidence showing that successful treatment is most likely when the tumor is diagnosed in the very early stages. Given the obvious methodological weaknesses of their study, the findings reported by Khalili et al.1 have to be regarded as questionable and the conclusions based on these findings dangerously misleading.