Incidence of Hepatocellular Carcinoma in Nonalcoholic Fatty Liver Disease

Orci LA, Sanduzzi-Zamparelli M, Caballol B, et al. Incidence of hepatocellular carcinoma in patients with nonalcoholic fatty liver disease: a systematic review, meta-analysis, and meta-regression. Clin Gastroenterol and Hepatol 2021;20:283–292. The epidemiology of hepatocellular carcinoma (HCC) is changing over time, with decreases in viral hepatitis-related cases and increases in the proportion attributed to nonviral etiologies, including nonalcoholic fatty liver disease (NAFLD). Historically, HCC risk and the targeting of surveillance programs were largely restricted to patients with cirrhosis, outside of hepatitis B infection. However, this paradigm was challenged by studies that demonstrated that up to one-third of HCC cases in the setting of NAFLD occur in the absence of cirrhosis, raising the question if surveillance programs should be expanded to noncirrhotic NAFLD. These studies underscored the importance of defining NAFLD subgroups with sufficiently high HCC risk to warrant HCC surveillance. Orci et al conducted a systematic review and meta-analysis of studies characterizing HCC incidence among patients with NAFLD, across stages of fibrosis, published between January 1950 and July 2020 (Clin Gastroenterol Hepatol 2022;20:28–292). The literature review identified 18 eligible studies with a total of 470,404 patients with NAFLD. Criteria for the diagnosis of NAFLD varied across studies, including liver histology in 5 studies, ultrasound examination only in 3 studies, laboratory results alone in 1 study, a combination of criteria in 5 studies, and not reported in 4 studies. Using the National Institutes of Health Quality Assessment Tool, 1 study was classified as high risk of bias, 11 as moderate risk, and 6 as low risk. The authors first examined pooled HCC incidence among studies that included any patients with cirrhosis and found a pooled incidence of 2.39 per 100 person-years (95% confidence interval [CI], 1.40–4.08); however, heterogeneity was very high with the I2 exceeding 90%. When restricting analyses to studies exclusively evaluating patients with cirrhosis, the pooled HCC incidence was higher at 3.78 per 100 person-years (95% CI, 2.47–5.78). In a subgroup analysis, the pooled HCC incidence estimates were 4.62 per 100 person-years (95% CI, 2.77–7.72) for studies with dedicated surveillance programs and lower at 2.35 per 100 person-years (95% CI, 0.99–5.61) in those that included patients not undergoing regular surveillance. Among 3 studies evaluating patients with noncirrhotic NAFLD, the pooled HCC incidence was low at 0.03 per 100 person-years (95% CI, 0.01–0.07), although heterogeneity was high, with the I2 exceeding 90%. Sensitivity analyses failed to remove the high observed between-study heterogeneity. A meta-regression analysis only revealed publication data as statistically significant; however, global heterogeneity remained high even after allowing this factor. NAFLD is an increasingly common etiology for cirrhosis and HCC worldwide, with a recent modeling study suggesting the global prevalence of NAFLD-related HCC will increase by >100% over the next decade (Hepatology 2018;67;123–133). Orci et al provide important data that directly inform clinical practice recommendations for this growing population. First, these study results reinforced that the annual incidence of HCC in patients with NAFLD cirrhosis exceeds the cost-effectiveness threshold for HCC surveillance, supporting professional society guidelines recommendations in this patient population. The authors found a pooled incidence of 3.78 per 100 person-years, similar to that reported in other cirrhosis etiologies, although the exact point estimate is limited by high heterogeneity and should be interpreted with some caution. Recent studies including prospectively enrolled patients with cirrhosis have demonstrated HCC incidence in NAFLD cirrhosis is lower than other etiologies, but still above the cost-effectiveness threshold. Although professional society guidelines include a historical threshold of 1.5% per year, more recent analyses suggest surveillance would be cost effective in patients with cirrhosis if annual HCC incidence exceeds 0.8% per year (Am J Gastroenterol 2020;115:1642–1649). Second, the authors addressed the controversy of HCC risk in noncirrhotic NAFLD. These data are important in light of the high proportion of NAFLD-related HCC diagnosed in the absence of underlying cirrhosis. For example, a large cohort study from the Veterans Affairs health system found that 34.6% of NAFLD-related HCC occurred in absence of cirrhosis, compared with 8.9% and 11.1% of those with hepatitis C-related HCC and alcohol-associated HCC, respectively (Clin Gastroenterol Hepatol 2016;14:124–131). A subsequent systematic review of studies examining the pooled risk of HCC in the absence of cirrhosis found that the prevalence of HCC in noncirrhotic nonalcoholic steatohepatitis was 38.0%, compared with only 14.2% among other etiologies (odds ratio, 2.61; 95% confidence interval, 1.27–5.35) (Aliment Pharmacol Ther 2018;48:696–703). There is a multifactorial etiology for the pathogenesis of NAFLD to HCC, involving diabetes, obesity-mediated mechanisms, and genetic polymorphisms. The overlap of risk factors for the development of NAFLD and HCC may explain why HCC risk in noncirrhotic NAFLD is higher than in some other etiologies. The mechanisms of tumorigenesis in NAFLD are distinct, but may overlap with the mechanisms involved in cirrhosis. Hence, noncirrhotic HCC in NAFLD is argued to be more than just a sporadic event (Liver Int 2016;36:317–324). However, it is unclear if the high proportion of HCC occurring in noncirrhotic NAFLD is related to a high annual risk of HCC in patients or low annual risk among a very large base of at-risk patients. In this systemic review, the authors found NAFLD patients without cirrhosis have a low HCC annual incidence of only 0.03 per 100 person-years, well below established cost-effectiveness thresholds in noncirrhotic patients. As with the analysis among patients with cirrhosis, the estimate for HCC incidence in noncirrhotic NAFLD was limited by high heterogeneity, which could not be resolved in subgroup analyses or meta-regression. This heterogeneity may in part be related to the inclusion of studies including a wide range of fibrosis stages. A recent retrospective study from Sweden demonstrated a dose-dependent increase in HCC incidence among NAFLD patients with worsening histological severity from simple steatosis (0.8 per 1000 person-years) to nonfibrotic nonalcoholic steatohepatitis (1.2 per 1000 person-years), noncirrhotic fibrosis (2.3 per 1000 person-years), and cirrhosis (6.2 per 1000 person-years) (Hepatology 2021;74:2410–2423). Although there is little debate about HCC surveillance among patients with simple steatosis or nonfibrotic nonalcoholic steatohepatitis, there continue to be questions about patients with advanced fibrosis without cirrhosis. Therefore, there continues to be a need for multicenter studies specifically examining HCC incidence in this subgroup of patients. The high observed heterogeneity in pooled HCC incidence estimates underscores a need for risk stratification tools—both for patients with and those without cirrhosis. Several approaches to risk stratification are currently being explored. Simple clinical risk scores, including readily available demographic and clinical data, have moderate accuracy for predicting clinical outcomes, including HCC (J Hepatol 2021;75:786–794). More nuanced risk stratification approaches include use of polygenic risk scores, incorporating single nucleotide polymorphisms such as PNPLA3 and MBOAT7 (J Hepatol 2021;74:775–782) or blood-based biomarkers, such as prognostic liver secretome (PLSec) (Medicine 2021;2:836–850), which have both been shown to be associated with HCC risk in patients with cirrhosis. These risk stratification strategies could identify patients without cirrhosis who are higher risk and would benefit from enrollment in a surveillance program. Similarly, risk stratification tools could identify patients with cirrhosis who have sufficiently low HCC risk who do not benefit from HCC surveillance as well as high risk patients in whom more intensive surveillance strategies could be cost effective. In light of ultrasound-based surveillance being prone to suboptimal visualization and sensitivity in obese patients and those with nonviral liver disease, alternative strategies such as magnetic resonance imaging may be considered in high-risk individuals. Therefore, HCC surveillance recommendations could be personalized to individual risk, rather than the current one-size-fits-all approach of surveillance being recommended in all patients with cirrhosis (Clin Transl Gastroenterol 2017;8:e101). The results of this meta-analysis should be interpreted considering its limitations. In addition to the heterogeneity in pooled estimates as discussed, there was also variation in the NAFLD diagnostic criteria and variation in HCC surveillance strategies across studies. This latter point is particularly noteworthy, considering data showing underuse of HCC surveillance, with lower use among patients with NAFLD than other etiologies, which can lead to the underdiagnosis of HCC cases and underestimation of HCC incidence (Hepatology 2021;73:713–725). This possibility was suggested by the authors finding a higher incidence rate among studies in which patients were under surveillance programs compared with those in whom surveillance status was unknown. Overall, the data from Orci et al provide important insights into HCC risk in cirrhotic and noncirrhotic NAFLD, which inform the value of surveillance in these populations and highlight areas in need of further research. Emerging risk stratification tools for HCC surveillance offer opportunities to improve precise targeting of surveillance in the NAFLD population in the future.