CHeCS includes patients aged 18 years and above who received health care services on or after January 1, 2006 at one of 4 health care systems—Henry Ford Health System (HFHS), Detroit, MI; Geisinger Health System (GHS), Danville, PA; Kaiser Permanente Northwest (KPNW), Portland, OR; and Kaiser Permanente Hawai’i (KPHI), Honolulu, HI. The study follows all guidelines of the US Department of Health and Human Services (HHS) regarding the protection of human subjects; the CHeCS protocol was approved and is renewed annually by the institutional review board at each of the 4 sites. Due to the deidentified nature of this observational study, requirements for written informed consent were waived.

CHeCS methods have been described elsewhere.^8,9 Briefly, CHeCS has a comprehensive data collection system based on electronic health records. Patients are identified electronically using a combination of laboratory and International Classification of Disease (ICD) 9 and 10–based criteria; CHB is subsequently confirmed through chart abstraction.⁸ CHeCS uses a “dynamic” sampling design; for each data collection cycle, a random sample of new patients is added to the cohort, while existing patients continue to be followed, this method allows us to generalize our findings to other populations. Retrospective patient data are captured through the early 1990s (when available), and data have been collected prospectively through December 31, 2016.

Across the entire 2006–2016 study period, we identified a total of 5528 confirmed CHB patients, with 952 cases of any cirrhosis [17%; including 401 cases of decompensated cirrhosis (7%)] and 628 deaths (11%). Cohort size increased across time—from 2564 in 2006 to 3411 in 2016. The yearly distribution of select patient demographics are presented in Table 1; a summary for the entire study period is presented in Supplementary Table 1 (Supplemental Digital Content 1, http://links.lww.com/JCG/A674). The proportion of female patients remained roughly the same across the study period (45% to 48%). Consistent with our previous report,¹⁷ the rate of antiviral treatment for CHB rose from 20% in 2006 to 35% in 2016.

In a diverse “real world” cohort of over 5000 CHB patients, we observed that overall prevalence of cirrhosis doubled from 2006 to 2016, while incidence of all-cause mortality remained flat across the study period. The rate of treatment increased from 20.5% in 2006 to 35.4% in 2016; receipt of antiviral medication was more common among patients with cirrhosis, a finding that reflects current guidelines to initiate treatment for patients with advanced fibrosis. However, the relative increase in cirrhosis prevalence among treated patients was significantly lower than among untreated patients (aAPC = +2.4% vs. +6.2%, respectively). Furthermore, there was a significant treatment status-by-time interaction for all-cause mortality. In 2006, mortality was significantly higher among treated patients than untreated patients; by 2016, that difference had narrowed and became nonsignificant. This is reflected in aAPCs for mortality, which decreased roughly 4% per year among treated patients, but increased at a similar rate among untreated patients. We also observed treatment status-by-time interactions across the study period, with a slower increase in cirrhosis and a decline in mortality among treated patients, but the opposite among untreated patients. Given that cirrhosis is an indication for initiation of antiviral treatment, receipt of treatment is in some cases a rough proxy for disease severity. Our findings—slower increases in cirrhosis prevalence and reductions in mortality among treated patients—demonstrate the benefit of antiviral treatment, even among CHB patients with more severe liver disease.^18,19 A sensitivity analysis also demonstrated that antiviral treatment reduced mortality among patients who were cirrhotic at baseline.

The implications of our findings relative to untreated patients are rather complicated and less clear. Several studies have suggested the benefits of treatment even among patients with low-level viremia who do not necessarily meet the criteria for treatment initiation.⁶ However, guidance regarding criteria for treatment initiation reflects the complex physiological interaction between viral load and immune response in CHB. Current recommendations prioritize patients with more severe liver disease or higher levels of viral replication. Importantly, there are many patients for whom treatment would be indicated, but who do not receive treatment due to lack of access to appropriate care.¹⁴ As shown in Table 1, a large but decreasing proportion of our sample had never received antiviral treatment—roughly 80% in 2006, which fell to 65% in 2016. It is beyond the scope of the current analysis to differentiate between those patients for whom treatment was and was not indicated under guidelines current during each study year. Nonetheless, the significant increases in cirrhosis and mortality among untreated patients over the study period is of concern, and contrasts with the trends observed among treated patients; given suggestions that patients even with low-level viremia derive benefits from antiviral treatment, studies of additional indicators for treatment (eg, aminotransferase levels or new biomarkers such as quantitative hepatitis B surface antigen) may be warranted. In a previous study, we found that antiviral treatment can reduce risk for development of hepatocellular carcinoma by 83% among patients with baseline hepatitis B virus (HBV) DNA levels > 20,000.⁷ A planned analysis will focus on the impact of treatment indication status and access to care as risk factors for these outcomes among patients with CHB.

In addition, while mortality did not increase for AAPI/other or white patients, rates increased significantly for African American patients over time. This is despite lower rates of cirrhosis observed among African Americans versus whites. It is possible this is attributable to higher overall all-cause mortality among African Americans in general²⁰ or possible racial disparities in access to quality CHB care.² However, it is a limitation of our analysis that we were not able to assess this directly. Likewise, cirrhosis prevalence was low among AAPI patients, despite high rates of CHB in this community. It is possible that there are differences in social determinants of health or natural history of CHB that influence these observations, but because our study was not designed to investigate risk factors for cirrhosis development, we are limited in the interpretation of these findings.

Furthermore, independent of treatment status, we found significant differences in APCs for cirrhosis across age categories. CHB patients aged 60 years and above had the highest prevalence of cirrhosis and all-cause mortality; similar findings have been reported among US veterans.²¹ Notably, although the oldest patients in our cohort (60 y and above) had the highest cirrhosis prevalence, the highest aAPC (indicating the largest proportional increases) occurred among patients aged 50 to less than 60 years, at least 2 times faster than other age groups; this suggests that CHB patients in this age group may benefit from close monitoring. Rates of complications were consistently higher in male patients than female. The 2:1 male: female ratio remained across time, even as rate of cirrhosis increased over the study period. Mortality was also higher among male patients. This is consistent with a number of studies of death among CHB patients.^22–24

One of the strengths of this analysis was our use of validated methods for classifying patients with cirrhosis to ensure consistency across time. These methods, which encompass a variety of observational data sources (liver biopsy, transient elastography, FIB4 index, and ICD-9/10 diagnosis codes) address many of the limitations of relying solely on diagnosis code-based data in an observational cohort. Another strength of this analysis is the ongoing “dynamic” cohort accrual over 5000 CHB patients from 4 large health systems with a combined population of over 2 million adult patients. We are aware of at least one cross-sectional study that sought to estimate prevalence of cirrhosis among chronic HBV patients seen in a US health care system during a similar time period (2014–2016). They found an overall prevalence of 27.7%; rates differed by sex (lower among female than male patients) but not by race. The authors of this study noted a few limitations in their analysis, specifically that their sample included largely uninsured and underinsured patients drawn from a safety net health care system that were referred for specialty gastroenterology care. These differences likely underlie the higher rate of cirrhosis observed among their population compared with our own.⁴ We note that the dynamic sampling design employed to generate our cohort is intended to increase the generalizability of our results by enrolling a random subset of CHB patients seen at our study sites.

Our study has several limitations. We used all-cause mortality instead of liver-related mortality because a previous CHeCS analysis demonstrated that liver-related mortality data is incomplete in routine-care settings.^13,25 Although our analysis is based on ongoing “dynamic” cohort accrual from health system populations, the cohort has aged significantly over time; patients 60 years and above made up only 7% of our 2006 cohort but 11% of the cohort in 2016. However, we did adjust for age in our analysis and found no indication of a qualitative interaction (change in direction of the effect) between age and time. Likewise, we were unable to confirm whether patients who did not receive antiviral treatment were eligible for such treatment, given the complexity of determining if and when treatment may be indicated. In a previous analysis of CHeCS data through 2013, approximately one-third of patients were “treatment eligible” (based on presence of immune-active disease and cirrhosis), of whom 60% had been treated.²⁶ Nevertheless, this analysis was designed to describe overall temporal changes in rates of CHB-related complications, rather than the effect of treatment on individual patients, a topic that has been previously covered. Finally, because our cohort consists of individuals with at least some contact with the health system, we are unable to estimate the prevalence of outcomes in CHB-positive individuals who remain undiagnosed or those without ongoing contact with a health care system. Such individuals are perhaps most at risk for poor outcomes. In addition, our study is based primarily upon health records data and may fail to capture factors that potentially affect clinical outcomes, such as disease duration or undiagnosed substance or alcohol misuse.

In conclusion, although the overall prevalence of cirrhosis doubled from 2006 through 2016 among CHB patients in the US-based CHeCS cohort, rates of all-cause mortality remained consistently low overall. Considering that cirrhosis is an indication for antiviral treatment, rates were higher but increased more slowly among patients who received antiviral treatment compared with untreated patients. Notably, all-cause mortality increased among untreated patients, but decreased among treated patients. These results suggest that antiviral treatment attenuates progression of cirrhosis and risk of death among patients with CHB.