Recent Decline in Hepatocellular Carcinoma Rates in the United States

Abstract

In the United States, hepatocellular carcinoma (HCC), a highly lethal cancer with limited treatment options, is primarily caused by chronic infection with hepatitis B virus (HBV) or hepatitis C virus (HCV), nonalcoholic steatohepatitis (NASH), and excessive alcohol consumption.1Islami F. et al.CA Cancer J Clin. 2017; 67: 273-289Crossref PubMed Scopus (161) Google Scholar HCC development generally requires decades-long exposure to 1 or more risk factors. HCC rates in the United States have increased over several decades.1Islami F. et al.CA Cancer J Clin. 2017; 67: 273-289Crossref PubMed Scopus (161) Google Scholar This trend was largely attributed an HCV epidemic during the 1960s–1980s and an increase in obesity-associated NASH. Based on data from 18 Surveillance, Epidemiology, and End Results (SEER) cancer registries during 2000 through 2012, HCC rates were projected to increase approximately 2.8%/y through 2030 if observed trends continued into the future.2Petrick J.L. et al.J Clin Oncol. 2016; 34: 1787-1794Crossref PubMed Scopus (307) Google Scholar However, a recent analysis of 13 SEER registries suggested that HCC rates plateaued during 2010–2015.3Rich N.E. et al.Clin Gastroenterol Hepatol. 2020; 18: 242-248Abstract Full Text Full Text PDF PubMed Scopus (47) Google Scholar Here, we use data from 21 SEER registries to provide updated HCC rates through 2016. HCC incidence data were obtained from SEER-21, population-based cancer registries covering 37% of the US population. Cases were identified using International Classification of Diseases for Oncology, 3rd edition, codes (site: C22.0; histology: 8170-8175). Incidence rates were age-standardized to the 2000 US population and estimated overall and by sex, age group, and race/ethnicity. Annual percent changes (APCs) were estimated using Joinpoint (National Cancer Institute, Bethesda, MD) regression, which identifies statistically significant inflection points in rate trajectories.4Kim H.J. et al.Stat Med. 2000; 19: 335-351Crossref PubMed Scopus (3799) Google Scholar Rate ratios (RRs) comparing 2015 and 2016 were estimated. We also estimated delay-adjusted rates of all liver/intrahepatic bile duct cancers (HCC-specific estimates are unavailable) to assess the potential impact of delayed reporting.5Midthune D.N. et al.J Am Stat Assoc. 2005; 100: 61-70Crossref Scopus (47) Google Scholar During 2000–2016, 119,078 HCC cases occurred in SEER-21 registries (rate, 5.84/100,000). In 2016, rates were higher in men (10.6/100,000) than women (2.83/100,000), increased with age (20–34-year-olds, 0.23/100,000; ≥65-year-olds, 27.6/100,000), and were highest among American Indian/Alaska Native (14.5/100,000) and Hispanic (9.74/100,000) individuals. Overall, HCC rates increased 5.64%/y (P < .001) during 2000–2007, increased 2.68%/y (P < .001) during 2007–2013, and then plateaued starting in 2013 (APC, –1.44%/y; P = .12) (Figure 1 and Supplementary Table 1). Rates plateaued among men and women beginning in 2012 and 2013, respectively. For ≥65-year-olds, rates increased during the entire time period (APC, 2.69%/y; P < .001). In contrast, inflections and significant decreases occurred among 35–49-year-olds starting in 2006 (APC, –4.93%/y; P < .001) and among 50–64-year-olds in 2014 (APC, –6.64%/y; P = .04). After years of strong increases, HCC rates plateaued among white individuals beginning in 2013 (APC, –0.91%/y; P = .45) and among black individuals in 2009 (APC, 0.27%/y; P = .65). Among Hispanic individuals, HCC incidence decreased nonsignificantly (–6.55%/y; P = .08) beginning in 2014. Among Asian/Pacific Islander (API) individuals, starting in 2007, HCC rates decreased –2.72%/y (P < .001). In contrast, for American Indian/Alaska Native individuals, HCC rates increased 4.60%/y (P < .001) across the time period. Comparing rates in 2016 to those in 2015, we observed a significant decrease in HCC incidence overall (RR, 0.96; P = .007) (Supplementary Table 1). Statistically significant (P < .05) declines were present among men (RR, 0.96), 50–64-year-olds (RR, 0.93), Hispanic individuals (RR, 0.92), and API individuals (RR, 0.90). There were 171 fewer HCC cases in 2016 than 2015 (Supplementary Figure 1). The deficit was greater among men than women (–146 vs –25 cases). By age group, the largest decrease in HCC cases was among 50–64-year-olds (–249 cases), whereas an increase occurred among ≥65-year-olds (+124 cases). By race/ethnicity, the largest decreases in HCC cases occurred among Hispanic individuals (–80 cases) and API individuals (–65 cases). Similar declines occurred between 2015 and 2016 in analyses restricted to SEER-18 (RR, 0.94; P = .01) and SEER-13 (RR, 0.94; P = .0005) registries and when liver/intrahepatic bile duct rates were estimated using delay adjustment (RR, 0.97; P < .05). In the United States, HCC rates increased for decades and were projected to continue to rise through 2030.2Petrick J.L. et al.J Clin Oncol. 2016; 34: 1787-1794Crossref PubMed Scopus (307) Google Scholar In that context, it is surprising that HCC rates flattened in 2013 and then declined in 2016. It seems unlikely that this change reflects a decrease in HCC due to NASH or alcoholic liver disease, because the US obesity epidemic has not improved, and alcohol-induced death rates have increased, including deaths due to cirrhosis.6Centers for Disease Control and PreventionMMWR Morb Mortal Wkly Rep. 2017; 66: 491Crossref PubMed Scopus (5) Google Scholar Improved treatment of viral hepatitis is likely to have contributed to the decline in HCC incidence. Treatment of chronic HBV, which can reduce HCC risk, has improved considerably with the introduction of effective nucleotide/nucleoside analogues in the mid-2000s.7Terrault N.A. et al.Hepatology. 2018; 67: 1560-1599Crossref PubMed Scopus (2444) Google Scholar That timeframe is temporally consistent with the HCC decline among API individuals, the demographic group most likely to have HBV-associated HCC.1Islami F. et al.CA Cancer J Clin. 2017; 67: 273-289Crossref PubMed Scopus (161) Google Scholar More recently, the availability of direct-acting antiviral agents (DAAs) against HCV has resulted in high cure rates, even among patients with advanced fibrosis or cirrhosis who responded poorly to interferon alfa–based treatment.8El-Serag H.B. et al.Hepatology. 2016; 64: 130-137Crossref PubMed Scopus (296) Google Scholar A prior study estimated that if 80% of HCV-infected individuals were treated with second-generation DAAs, HCC rates would decline.2Petrick J.L. et al.J Clin Oncol. 2016; 34: 1787-1794Crossref PubMed Scopus (307) Google Scholar These treatments were not available until late in 2014; therefore, DAA regimens cannot account for the plateau in HCC incidence starting in 2013 and are unlikely to fully account for the decrease in 2016. Because HCV prevalence is highest in the baby boomer generation,1Islami F. et al.CA Cancer J Clin. 2017; 67: 273-289Crossref PubMed Scopus (161) Google Scholar an increase in HCC rates among 50–64-year-olds would have been expected with the aging of this cohort (to approximately 50- to 70-years-old in 2016); however, rates plateaued and then declined in this age group, perhaps reflecting better HCV treatments. The Hispanic population contributed disproportionately to the decrease in HCC cases from 2015 to 2016; the drivers behind this decline deserve further investigation. Trends varied across registries. Declines or plateaus observed in recent years occurred in 8 registries (representing 55% of HCCs), likely reflecting etiologic heterogeneity in HCC by region. Because SEER does not capture information about etiologic factors, possible explanations for changing HCC trends are speculative. Increased competing risk of death may also contribute to the decline in HCC risk, particularly given rising US death rates due to cirrhosis and drug use,9Shiels M.S. et al.Lancet. 2017; 389: 1043-1054Abstract Full Text Full Text PDF PubMed Scopus (170) Google Scholar which are both associated with viral hepatitis. Although it is premature to conclude that a long-term decrease in HCC has begun, the sensitivity analysis based on delay-adjusted rates provides reassurance that the decrease in 2016 is not an artifact. Furthermore, the proportion of HCCs that were distant stage decreased from 15.9% (2004) to 13.9% (2016), suggesting that underascertainment is unlikely to have driven these findings. Our analysis suggests that the tide has begun to turn for HCC incidence in the United States. Future studies should monitor this trend and assess the factors that underlie these changing rates. Author contributions: The study concept and design, interpretation of data, and drafting of the manuscript were shared by Meredith S. Shiels and Thomas R. O’Brien. Meredith S. Shiels carried out the statistical analysis. Supplementary Table 1Age-standardized incidence rates, and annual percent changes in hepatocellular carcinoma rates, 2000-2016Segment 1Segment 2Segment 3Segment 4Incidence rates per 100,000YearsAPC (95% CI)YearsAPC (95% CI)YearsAPC (95% CI)YearsAPC (95% CI)20152016Overall2000–20075.64 (4.90 to 6.38)2007–20132.68 (1.72 to 3.66)2013–2016–1.44 (–3.32 to 0.48)6.726.46aIndicates significant declines from 2015 to 2016, P < .05.Sex Men2000–20075.61 (4.55 to 6.67)2007–20122.85 (0.90 to 4.84)2012–2016–0.64 (–2.36 to 1.11)11.110.6aIndicates significant declines from 2015 to 2016, P < .05. Women2000–20133.83 (3.32 to 4.35)2013–2016–2.11 (–5.85 to 1.77)2.952.83Age, y 20–342000–2016–0.21 (–1.66 to 1.27)0.270.23 35–492000–20060.87 (–1.61 to 3.41)2006–2016–4.93 (–6.17 to –3.67)1.591.46 50–642000–20069.95 (8.19 to 11.74)2006–20115.12 (2.87 to 7.43)2011–20140.74 (–5.18 to 7.03)2014–2016–6.64 (–12.3 to –0.65)18.417.1aIndicates significant declines from 2015 to 2016, P < .05. 65+2000–20084.16 (3.36 to 4.97)2008–20162.69 (2.06 to 3.32)27.927.6Race/ethnicity White2000–20085.57 (4.78 to 6.37)2008–20133.31 (1.51 to 5.14)2013–2016–0.91 (–3.48 to 1.73)5.034.93 Black2000–20096.49 (5.26 to 7.73)2009–20160.27 (–0.98 to 1.54)8.878.54 Hispanic2000–20065.46 (3.39 to 7.56)2006–20141.72 (0.56 to 2.89)2014–2016–6.55 (–13.4 to 0.85)10.69.74aIndicates significant declines from 2015 to 2016, P < .05. API2000–20071.38 (–0.08 to 2.86)2007–2016–2.72 (–3.54 to –1.90)10.79.70aIndicates significant declines from 2015 to 2016, P < .05. AIAN2000–20164.60 (2.92 to 6.31)14.714.5NOTE. Rates were standardized to the 2000 US population in 5-year age categories. Rates among AIAN individuals were limited to counties in purchased/referred care delivery areas to limit racial misclassification. Joinpoint regression model requires a minimum of 2 observations between a Joinpoint and the first or last data point, and between 2 Joinpoints.AIAN, American Indian/Alaska Native; CI, confidence interval.a Indicates significant declines from 2015 to 2016, P < .05. Open table in a new tab NOTE. Rates were standardized to the 2000 US population in 5-year age categories. Rates among AIAN individuals were limited to counties in purchased/referred care delivery areas to limit racial misclassification. Joinpoint regression model requires a minimum of 2 observations between a Joinpoint and the first or last data point, and between 2 Joinpoints. AIAN, American Indian/Alaska Native; CI, confidence interval.

Topics & Concepts