Abstract
Background
It is unclear whether the risk of hepatocellular carcinoma (HCC) decreases over time following hepatitis C virus (HCV) eradication.
Aims
We aimed to determine if patients who have accrued longer time since SVR have a lower risk of HCC than patients who have accrued less time since SVR.
Methods
We conducted a retrospective cohort study of all HCV-infected Veterans Affairs patients who achieved SVR before 1/1/2018 and remained alive without a diagnosis of HCC as of 1/1/2019 (n=75,965). We ascertained their baseline characteristics as of 1/1/2019 (“time zero”), including time accrued since SVR, and followed them for the subsequent 12 months until 1/1/2020 for incident HCC. Multivariable Cox proportional hazards regression was used to determine the association between time since SVR and HCC risk after adjusting for baseline characteristics: age, race/ethnicity, sex, diabetes, hypertension, BMI, alcohol use, Charlson Comorbidity Index, Fibrosis-4 score, HCV genotype, hepatitis B virus co-infection, and HIV co-infection.
Results
The majority (96.0%) were male and mean age was 64.6±7 years. Among those with cirrhosis (n=19,678, 25.9%), compared to patients who had accrued only ≥1 to 2 years since SVR (HCC incidence 2.71/100 person-years), those who had accrued >2 to 4 years (2.11/100 person-years, aHR 0.80, 95% CI 0.63–1.01) and >4 to 6 years (1.65/100 person-years, aHR 0.61, 95% CI 0.41–0.90) had progressively lower HCC risk, but HCC risk appeared to plateau for those with >6 years since SVR (1.68/100 person-years, aHR 0.70, 95% 0.46–1.07). Among those without cirrhosis, HCC risk was low (0.23–0.27/100 person-years) without a significant association between time since SVR and HCC risk.
Conclusions
Among patients with cirrhosis and cured HCV, HCC risk declines progressively up to 6 years post-SVR – although it remained well above thresholds that warrant screening. This suggests time since SVR can inform HCC surveillance strategies in patients with cured HCV and can be incorporated in HCC risk prediction models.
Keywords: HCC, SVR, HCV, liver cancer, incidence
Graphical Abstract
Introduction
Chronic infection with hepatitis C virus (HCV) is a leading cause of hepatocellular carcinoma (HCC); however, most patients with HCV have already achieved sustained virologic response (SVR) through treatment with direct acting antivirals (DAAs), or are expected to be treated in the near future.1,2 In turn, SVR has been shown to significantly reduce the risk of HCC when compared to having persistent, active infection.3–5
Whether HCC risk continues to decline over time following SVR remains unclear. Multiple mechanisms may be at play. First, SVR has been clearly shown to cause regression of liver fibrosis and liver remodeling.6,7 Second, HCV directly induces carcinogenic effects via its proteins and transcripts and HCV eradication can reduce HCC risk by eliminating these direct carcinogenic effects.8,9 It is tempting to speculate that these mechanisms may subsequently lead to decreased risk of HCC risk over time. Indeed, prior observational studies have suggested that HCC incidence continues to decline years after SVR.4,5,10 However, other non-HCV related factors may be confounding this observation. As time accrues after SVR, patients become older or may acquire additional comorbidities that impact HCC risk (e.g., diabetes, obesity, alcohol use); prior observational studies have not adjusted for these confounders.
Understanding HCC risk over time is particularly important as it will help inform HCC screening strategies for patients with cured HCV. In this retrospective cohort study, we aimed to use Veterans Affairs (VA) Health Care System data to assess whether patients who have accrued longer time since SVR have a lower risk of HCC than those who have accrued less time since SVR.
Patients and methods
Data source
The VA Health Care system is the largest integrated healthcare system in the United States, providing care at 168 VA Medical Centers and 1,053 outpatient clinics, and serving more than 6 million Veterans each year. We derived electronic patient-level data from the VA Corporate Data Warehouse, a national repository of all VA electronic health records. This study was approved by the Institutional Review Board of the VA Puget Sound Health Care System.
Study design and rationale
We created a cohort of all HCV-infected VA patients who achieved SVR before 1/1/2018 and were still alive and had not undergone liver transplantation or developed HCC as of 1/1/2019 (“time zero”). We required SVR to be achieved before 1/1/2018 so that all cohort participants would have accrued at least 1 year of time since SVR by 1/1/2019 when follow-up for HCC incidence began. We ascertained patients’ baseline characteristics as of 1/1/2019, including number of years since SVR categorized as ≥1 to 2, >2 to 4, >4 to 6, and >6 years. We followed these patients for incident HCC for 12 months from 1/1/2019 (time zero) until 1/1/2020 (end of follow-up) (Figure 1). This study design allowed us to have the same time zero (i.e. 1/1/2019) for all study participants in all 4 comparison groups of time since SVR, to establish eligibility and ascertain all baseline characteristics as of this time zero, and to begin follow-up for incident HCC beginning at time zero and extending for the following 12 months. This avoided biases that occur due to time zero inconsistencies between comparison groups and also mirrored clinical practice whereby clinicians attempt to assess a person’s HCC risk at a given time point (e.g. a clinic visit) based on the person’s baseline characteristics as of that time. We chose 1/1/2019 as the most recent time zero that we could have employed because 12 months of follow-up after that reaches 1/1/2020, just before the COVID-19 pandemic, which spuriously affected observed HCC incidence in the VA and elsewhere due to disruptions in healthcare delivery.11
Figure 1. Study design and incidence rates.
We aimed to determine HCC incidence in a cohort of patients with cured HCV who were alive and at risk of HCC as of 1/1/2019 (time zero) and had achieved SVR at least one year prior to 1/1/2019. Follow-up for the development of HCC began on 1/1/2019 and continued for 12 months. Patients were categorized according to how much time they had accrued since SVR as of the beginning of follow-up on 1/1/2019 (≥1–2, >2–4, >4–6, and >6 years). We also ascertained other baseline characteristics as of 1/1/2019 to enable multivariable adjustment for baseline characteristics to determine the association between time accrued since SVR and HCC risk.
Study population
We identified 121,350 patients who achieved SVR in the VA system between 10/1/1999 and 1/1/2018 (Figure 2). We excluded 13,755 patients who died, 3,344 patients who were previously diagnosed with HCC, and 1,110 patients who underwent liver transplantation before 1/1/2019 (time zero, the date on which follow-up for HCC incidence began). We additionally excluded 28,176 patients who did not have a serum aspartate aminotransferase [AST], alanine aminotransferase [ALT] or platelet count performed in the 12 months before 1/1/2019, that we could use for baseline laboratory characteristics (Supplemental Table 1). This resulted in an analytic cohort of 75,965 patients, who achieved SVR before 1/1/2018, who were still alive and without a diagnosis of HCC as of 1/1/2019 and had received care and laboratory testing in the VA Health Care system during the 12-month period before the initiation of follow-up in 1/1/2019.
Figure 2. Patient inclusion and exclusion flow diagram.
We identified 121,350 patients who achieved SVR prior to 1/1/2018 (i.e. at least one year prior to the initiation of follow-up for HCC on 1/1/2019). We excluded patients who died, were diagnosed with HCC, or underwent liver transplantation prior to 1/1/2019 (time zero). Finally, we excluded patients who did not have available laboratory studies (AST, ALT, platelet count) in the 12 months prior to 1/1/2019. This left a final analytic cohort of 75,965 patients.
Baseline patient characteristics
We determined the presence of medical comorbidities including type 2 diabetes mellitus, hypertension, alcohol use disorder, and human immunodeficiency virus (HIV) infection based on appropriate International Classification of Diseases (ICD)-9 revision and ICD-10 codes recorded at least twice in any inpatient or outpatient encounter prior to 1/1/2019 (Supplemental Table 2). We ascertained age, body mass index (BMI), and Charlson Comorbidity Index as of 1/1/2019. We determined baseline laboratory tests (platelet count, AST, ALT, international normalized ratio [INR], albumin, total bilirubin, creatinine, haemoglobin) performed in the 12 months before 1/1/2019, and used the one closest to 1/1/2019 as the “baseline” level.
The diagnosis of cirrhosis was based on the presence of specific ICD-9 and ICD-10 codes for cirrhosis or complications of cirrhosis (Supplemental Table 2) recorded at least twice in any inpatient or outpatient encounter prior to 1/1/2019, an approach that has been validated and widely used in VA-based studies by us and others. The diagnosis of cirrhosis using two ICD-9 codes in VA data was shown to have a 97% positive predictive value compared to chart extraction.12 The presence of cirrhosis was not defined using the fibrosis-4 score.13 In addition, as has previously been done in other VA studies14,15, hepatic encephalopathy was defined by the prescription of lactulose, rifaximin, or neomycin for more than 90 days to maximize sensitivity.
Exposure ascertainment: time since SVR
We defined SVR as an undetectable serum HCV RNA viral load performed at least 12 weeks after the end of HCV treatment.16,17 The date of SVR was defined as the date at which the HCV viral load first became undetectable without any subsequent detectable or positive viral loads. Time since SVR was calculated as the difference between 1/1/2019 (i.e. the beginning of follow-up) and the date of SVR. As the follow-up time began in 1/1/2019 and all participants achieved SVR before 1/1/2018, everyone included in our analytic cohort had already accrued at least one year of time since SVR the start of follow-up (Figure 1). This approach is similar to that of a landmark analysis18 whereby only those alive and without prior HCC by 1/1/2019 will contribute follow-up time. We included this 1-year landmark, or “washout”, period in order to avoid spurious changes in HCC incidence that may occur during the same year as SVR when patients are undergoing close evaluation and surveillance and many prevalent HCCs may be diagnosed. We also wanted to make sure that baseline laboratory tests reflected those established after SVR, even for those who had only accrued one year of follow-up.
Follow-up period and ascertainment of incident HCC
All patients were followed for the same 12 calendar months from 1/1/2019 to 1/1/2020 for the primary outcome of incident HCC. The diagnosis of HCC was based on the ICD-9 code 155.0 or ICD-10 code C22.0 (the VA switched to ICD-10 codes on 10/1/2015). Using this ICD-9 code to define HCC with VA records has been shown to have a positive predictive value of 84–94% when compared to chart review.19,20 Incident HCCs were defined when the relevant ICD-10 code was recorded for the first time between 1/1/2019 and 1/1/2020, without any prior documentation of C22.0 or ICD-9 code 155.0 at any time point before 1/1/2019 (Supplemental Table 2). We did not collect data on imaging studies or receipt of HCC screening.
Statistical Analysis
We used univariable and multivariable Cox proportional hazards regression to determine the association between time since SVR and the risk of developing HCC. Patients who died or underwent liver transplantation during follow-up (1/1/2019–1/1/2020) were censored at the date of death or transplant. We also performed a competing risks proportional hazards analysis with death as a competing risk to the development of HCC.21
We decided a priori to include in our multivariable models the following baseline characteristics and potential confounders that may be associated with both time since SVR and the development of HCC: HIV co-infection, age, diabetes, hypertension, alcohol use disorder, Charlson Comorbidity Index, race/ethnicity, sex, BMI, HCV genotype, fibrosis-4 (FIB-4) score, hepatitis B virus (HBV) exposure, and HBV co-infection. HBV co-infection is based on a positive HBV antigen study or a detectable HBV viral load. FIB-4 was calculated as (Age × AST)/(Platelet count × √ALT) based on tests performed in 2018. It is a non-invasive marker of hepatic fibrosis that is also very strongly associated with HCC.16,17 We also performed a sensitivity analysis in which we did not simultaneously adjust for the FIB-4 score, alcohol use, and BMI,22,23 because longer time since SVR may result in reduction or improvement in these factors such that they may be on a direct, causal pathway between length of time since SVR and HCC risk. In that case, adjusting for FIB-4, alcohol use, and BMI may potentially attenuate or mask an association between time since SVR and HCC risk.
Results
Characteristics of the study population
Among the 75,965 patients who achieved SVR, the majority (96.0%) were male and either non-Hispanic White (54.6%) or non-Hispanic Black (34.2%) (Table 1). The mean age was 64.6 ± 7.0 years and a quarter of patients (25.9%) had cirrhosis at baseline. Alcohol use disorder (73.5%), hypertension (66.72%), and diabetes (33.1%) were common comorbidities. Approximately one-third (34.9%) of patients were previously exposed to HBV as defined by a positive anti-hepatitis B core immunoglobulin G test; however, only 2.8% had an active HBV co-infection at baseline. The mean time accrued since the date of SVR until the start of follow-up date (i.e. 1/1/2019) was 3.98 ± 3.44 years (range 1 to 19.2 years).
Table 1.
Baseline characteristics of patients who achieved SVR prior to 1/1/2018 and had an available FIB-4 score in the calendar year 2018
| All patients (n = 75,965) |
1–2 years since SVR (n=16,759) |
>2–4 years since SVR (n=42,259) | >4–6 years since SVR (n=5,208) |
>6 years since SVR (n=11,739) |
|
|---|---|---|---|---|---|
| Age, yrs (mean±SD) | 64.7 ± 7.0 | 63.9 ± 8.0 | 65.0 ± 6.7 | 64.6 ± 6.8 | 64.9 ± 6.6 |
| Male | 72,928 (96.0%) | 16,153 (96.4%) | 40,694 (96.3%) | 4,955 (95.1%) | 11,126 (94.8%) |
| Race/Ethnicity | |||||
| Non-Hispanic White | 41,455 (54.6%) | 8,624 (51.5%) | 21,416 (50.7%) | 3,269 (62.8%) | 8,146 (69.4%) |
| Non-Hispanic Black | 25,988 (34.2%) | 6,260 (37.4%) | 16,472 (39.0%) | 1,300 (25.0%) | 1,956 (16.7%) |
| Hispanic | 4,432 (5.8%) | 923 (5.5%) | 2,264 (5.4%) | 359 (6.9%) | 886 (7.6%) |
| Other/ Refused | 2,813 (3.7%) | 630 (3.8%) | 1,454 (3.4%) | 204 (3.9%) | 525 (4.5%) |
| Missing | 1,154 (1.5%) | 322 (1.9%) | 653 (1.6%) | 76 (1.5%) | 226 (1.9%) |
| BMI | |||||
| <18.5 | 2,975 (3.9%) | 870 (5.2%) | 1,537 (3.7%) | 157 (3.0%) | 411 (3.5%) |
| 18.5–24.9 | 19,780 (26.0%) | 5,042 (30.1%) | 11,044 (26.1%) | 1,128 (21.7%) | 2,566 (21.9%) |
| 25–30 | 27,520 (36.2%) | 5,833 (34.8%) | 15,466 (36.6%) | 1,942 (37.3%) | 4,279 (36.5%) |
| >30 | 25,690 (33.8%) | 5,014 (29.9%) | 14,212 (33.6%) | 1,981 (38.0%) | 4,483 (38.2%) |
| FIB-4 score (mean±SD) | |||||
| ≤1.4 | 31,599 (41.6%) | 7,066 (42.2%) | 10,7052 (40.4%) | 1,900 (36.5%) | 5,581 (47.5%) |
| >1.4 – 2.1 | 24,884 (32.8%) | 5,470 (32.6%) | 13,891 (32.9%) | 1,653 (31.7%) | 3,870 (33.0%) |
| >2.1 – 3.25 | 13,064 (17.2%) | 2,847 (17.0%) | 7,501 (17.8%) | 1,009 (19.4%) | 1,707 (14.5%) |
| >3.25–4.5 | 3,414 (4.5%) | 749 (4.5%) | 2,015 (4.9%) | 320 (6.1%) | 330 (2.8%) |
| >4.5 | 3,004 (4.0%) | 627 (3.7%) | 1,800 (4.3%) | 326 (6.3%) | 251 (2.1%) |
| AST (mean±SD) | 25.7 ± 15.6 | 25.5 ± 17.0 | 25.5 ± 15.0 | 27.0 ± 15.2 | 26.2 ± 15.5 |
| ALT (mean±SD) | 24.4 ± 15.9 | 22.7 ± 15.3 | 23.9 ± 15.2 | 26.7 ± 18.3 | 27.1 ± 17.4 |
| Platelet count(k/microL) (mean±SD) | 215.8 ± 72.6 | 219.0 ± 73.1 | 213.8 ± 72.9 | 201.9 ± 74.8 | 224.9 ± 68.8 |
| HCV Genotype | |||||
| Genotype 1 | 41,902 (59.1%) | 10,126 (60.4%) | 27,560 (65.2%) | 2,651 (50.9%) | 4,565 (38.9%) |
| Genotype 2 | 6,872 (9.1%) | 1,365 (8.1%) | 2,738 (6.5%) | 702 (13.5%) | 2,067 (17.6%) |
| Genotype 3 | 3,952 (5.2%) | 945 (5.6%) | 1,603 (3.8%) | 299 (5.7%) | 1,105 (9.4%) |
| Other | 562 (0.7%) | 142 (0.9%) | 335 (0.8%) | 24 (0.5%) | 61 (0.5%) |
| Missing | 19,677 (25.9%) | 4,181 (25.0%) | 10,023 (23.7%) | 1,532 (29.4%) | 3,941 (33.6%) |
| HBV co-infection † | 1,972 (2.6%) | 260 (1.6%) | 1,310 (3.1%) | 119 (2.3%) | 283 (2.4%) |
| Hepatitis B core antibody | |||||
| Positive | 26,476 (34.9%) | 6,506 (38.8%) | 14,866 (35.2%) | 1,782 (34.2%) | 3,322 (28.3%) |
| Negative | 36,574 (48.2%) | 8,606 (51.4%) | 19,987 (47.3%) | 2,537 (48.7%) | 5,444 (46.4%) |
| Never tested | 12,915 (17.0%) | 1,647 (9.8%) | 7,406 (17.5%) | 889 (17.1%) | 2,973 (25.3%) |
| HIV co-infection | 2,619 (3.5%) | 383 (2.3%) | 1,655 (3.9%) | 219 (4.2%) | 362 (3.1%) |
| Cirrhosis | 19,678 (25.9%) | 3,670 (21.9%) | 12,047 (28.5%) | 2,125 (40.8%) | 1,836 (15.6%) |
| Decompensated cirrhosis ‡ | 5,213 (6.9%) | 970 (5.8%) | 3,046 (7.2%) | 550 (10.6%) | 647 (5.5%) |
| Hypertension | 50,686 (66.7%) | 10,242 (61.1%) | 28,772 (68.1%) | 3,503 (67.3%) | 8,169 (69.6%) |
| Diabetes | 25,150 (33.1%) | 5,167 (30.8%) | 14,378 (34.0%) | 1,806 (34.7%) | 3,799 (32.4%) |
| History of alcohol use disorder | 55,840 (73.5%) | 12,444 (74.3%) | 30,999 (73.4%) | 3,747 (72.0%) | 8,650 (73.7%) |
| Charlson comorbidity index (%) | |||||
| 0–1 | 29,448 (38.8%) | 5,623 (33.6%) | 15,802 (37.4%) | 2,176 (41.38%) | 5,847 (49.8%) |
| 2–4 | 30,731 (40.5%) | 7,462 (44.5%) | 17,144 (40.6%) | 1,920 (36.9%) | 4,205 (35.8%) |
| ≥5 | 15,786 (20.8%) | 3,674 (21.9%) | 9,313 (22.0%) | 1,112 (21.4%) | 1,687 (14.4%) |
MELD: Model for End Stage Liver Disease, HCV: Hepatitis C virus, BMI: Body Mass Index
HBV co-infection is based on a positive HBV antigen study or a detectable HBV viral load
Decompensated cirrhosis is defined by the diagnosis of ascites, spontaneous bacterial peritonitis, encephalopathy, bleeding gastroesophageal varices, hepatorenal syndrome, or hepatopulmonary syndrome
When grouped by time accrued since SVR, those who had achieved SVR >6 years prior were more likely to have had HCV genotype 2 or 3 infection (27% for >6 years vs. 10.3% for >2 to 4 years, 13.7% for ≥1 to 2 years, and 19.2% for >4 to 6 years). Patients who had accrued >4 years since SVR had the highest proportion with cirrhosis (40.8% for >4–6 years vs. 21.9% for ≥1 to 2 years, 28.5% for >2 to 4 years, and 15.6% for >6 years).
Association between time accrued since SVR and risk of HCC among those with baseline cirrhosis
A total of 547 patients (0.72%) were diagnosed with HCC, which included 404 patients who had baseline cirrhosis and 143 without baseline cirrhosis. Among the 19,678 patients with cirrhosis, those with more time accrued since SVR had lower HCC incidence rates: the highest incidence was among those who accrued ≥1–2 years since SVR (2.71 per 100 person years [P-Ys]), and declined for those who accrued >2 to 4 years (2.11 per 100 P-Ys) and even more for those who accrued >4 to 6 years (1.65 per 100 P-Ys), but appeared to remain stable thereafter at >6 years (1.68 per 100 P-Ys) (Table 2, Figure 3a).
Table 2.
Association between number of years since SVR and risk of HCC among patients 75,965 patients who achieved SVR in the national Veterans Affairs healthcare system.
| Number of years† since SVR | Number of patients | Number of HCC | Number of person-years | HCC Incidence per 100 person-years | Hazard Ratio | Adjusted Hazard Ratio‡ | Adjusted Hazard Ratio without adjusting for FIB-4, alcohol use, or BMI§ | Sub-hazard ratio | Adjusted sub-hazard ratio‡ |
|---|---|---|---|---|---|---|---|---|---|
| All patients (n = 75,965) | |||||||||
| ≥1 to 2 | 16,759 | 125 | 16,341 | 0.76 | 1 | 1 | 1 | 1 | 1 |
| >2 to 4 | 42,259 | 326 | 41,332 | 0.79 | 1.03 (0.84–1.27) | 0.89 (0.72–1.10) | 0.89 (0.72–1.10) | 1.03 (0.84–1.27) | 0.90 (0.73–1.11) |
| >4 to 6 | 5,208 | 42 | 5,113 | 0.82 | 1.07 (0.76–1.52) | 0.72 (0.51–1.03) | 0.73 (0.52–1.04) | 1.08 (0.76–1.53) | 0.73 (0.51–1.03) |
| >6 | 11,739 | 54 | 11,525 | 0.47 | 0.61 (0.45–0.84) | 0.84 (0.60–1.17) | 0.78 (0.56–1.08) | 0.62 (0.45–0.85) | 0.84 (0.61–1.16) |
| With cirrhosis (n = 19,678) | |||||||||
| ≥1 to 2 | 3,670 | 95 | 3,506 | 2.71 | 1 | 1 | 1 | 1 | 1 |
| >2 to 4 | 12,047 | 245 | 11,628 | 2.11 | 0.78 (0.61–0.99) | 0.80 (0.63–1.01) | 0.79 (0.62–1.01) | 0.78 (0.62–0.99) | 0.80 (0.63–1.02) |
| >4 to 6 | 2,125 | 34 | 2,059 | 1.65 | 0.61 (0.41–0.90) | 0.61 (0.41–0.90) | 0.62 (0.42–0.92) | 0.61 (0.42–0.91) | 0.61 (0.41–0.91) |
| >6 | 1,836 | 30 | 1,783 | 1.68 | 0.62 (0.41–0.94) | 0.70 (0.46–1.07) | 0.63 (0.41–0.95) | 0.63 (0.42–0.95) | 0.70 (0.46–1.07) |
| Without cirrhosis (n = 56,287) | |||||||||
| ≥1 to 2 | 13,089 | 30 | 12,835 | 0.23 | 1 | 1 | 1 | 1 | 1 |
| >2 to 4 | 30,212 | 81 | 29,704 | 0.27 | 1.17 (0.77–1.77) | 1.26 (0.83–1.92) | 1.24 (0.81–1.89) | 1.17 (0.77–1.78) | 1.26 (0.83–1.93) |
| >4 to 6 | 3,083 | 8 | 3,054 | 0.26 | 1.12 (0.51–2.45) | 1.36 (0.62–2.99) | 1.34 (0.61–2.94) | 1.13 (0.52–2.47) | 1.39 (0.63–3.04) |
| >6 | 9,903 | 24 | 9,742 | 0.25 | 1.05 (0.62–1.80) | 1.28 (0.73–2.25) | 1.25 (0.71–2.18) | 1.06 (0.62–1.81) | 1.30 (0.75–2.24) |
Number of years counts from 1/1/2019 to the date of SVR. Since the inclusion criteria for this cohort was achievement of SVR prior to 1/1/2018, there was a minimum of 1 year from SVR date to start of follow-up time in 1/1/2019 for all participants.
Adjusted for Time since SVR, age, HIV status, diabetes, alcohol use disorder, hypertension, BMI, Charlson Comorbidity Index, race/ethnicity, sex, HCV genotype, FIB-4 score, HBV co-infection, and prior HBV exposure
Adjusted for Time since SVR, age, HIV status, diabetes, hypertension, Charlson Comorbidity Index, race/ethnicity, sex, HCV genotype, HBV co-infection, and prior HBV exposure
Sub-hazard ratio was obtained using Fine-Gray Hazards regression for the competing risk of death
Figure 3. Cumulative HCC incidence curves for the 12-month time period from 01/01/2019 to 01/01/2020 for patients with cirrhosis (a) and without cirrhosis (b), categorized according to length of time since SVR.
Among those with cirrhosis (n = 19,678), there was a significant difference in incident HCC rates when stratified by time since SVR (logrank test p = 0.02). Among those without cirrhosis (n = 56,287), there was not a significant difference in incident HCC rates when stratified by time since SVR (logrank test p = 0.90).
In the unadjusted analysis, compared to those who had accrued only ≥1 to 2 years of time since SVR, those who accrued >2 to 4 years (HR 0.78, 95% CI 0.61 – 0.99) and >4 to 6 years (HR 0.61, 95% CI 0.41–0.90) had progressively lower HCC risk, which appeared to plateau in those with >6 years since SVR (HR 0.62, 95% CI 0.41–0.94). There was minimal change in the degree and no change in the direction of these associations after adjusting for routinely available clinical and laboratory characteristics or when accounting for the competing risk of death (Table 2). In a separate multivariate analysis excluding FIB-4, alcohol use, and BMI from the set of adjusted characteristics, the accrual of >4–6 and >6 years was significantly associated with a lower risk of HCC (aHR 0.62, 95% CI 0.42–0.92, and aHR 0.63, 95% CI 0.41–0.95, respectively) (Table 2). On average, for each year of SVR accrued beyond 1 year, HCC risk decreased by a factor of 0.95 (95% CI 0.91–0.99, p=0.03).
In the adjusted model, increasing FIB-4 values, a higher Charlson Comorbidity Index, and HCV genotype 3 (vs. genotype 1) were independently associated with increased risk of HCC (Table 3). Older age and HBV co-infection or prior exposure were not associated with risk of HCC.
Table 3.
Adjusted hazard ratios for the associations between baseline characteristics and risk of HCC among patients who achieved SVR and stratified by cirrhosis status
| Adjusted‡ hazard ratio (95% confidence interval) | |||
|---|---|---|---|
| All patients (n = 75,965) |
With cirrhosis (n = 19,678) |
Without cirrhosis (n = 56,287) |
|
| Time since SVR (years) | |||
| ≥1 to 2 | 1 | 1 | 1 |
| >2 to 4 | 0.89 (0.72–1.10) | 0.80 (0.63–1.01) | 1.26 (0.83–1.92) |
| >4 to 6 | 0.72 (0.51–1.03) | 0.61 (0.41–0.90) | 1.36 (0.62–2.99) |
| >6 | 0.84 (0.60–1.17) | 0.70 (0.46–1.07) | 1.28 (0.73–2.25) |
| Age (years) | |||
| ≤60 | 1 | 1 | 1 |
| >61–64 | 1.50 (1.06–2.14) | 1.33 (0.89–1.98) | 2.21 (1.04–4.69) |
| >64–67 | 1.91 (1.35–2.71) | 1.75 (1.18–2.59) | 2.52 (1.18–5.40) |
| >67–70 | 1.80 (1.26–2.56) | 1.49 (0.99–2.22) | 3.23 (1.52–6.85) |
| >70 | 1.60 (1.11–2.30) | 1.31 (0.87–1.99) | 2.92 (1.35–6.31) |
| Diabetes | 0.99 (0.82–1.21) | 0.95 (0.76–1.20) | 1.14 (0.77–1.69) |
| Alcohol use disorder | 1.27 (1.03–1.56) | 1.19 (0.93–1.52) | 1.50 (0.99–2.27) |
| Hypertension | 1.08 (0.89–1.33) | 1.03 (0.82–1.30) | 1.24 (0.81–1.88) |
| BMI | |||
| <18.5 | 1 | 1 | 1 |
| 18.5–24.9 | 1.44 (0.82–2.55) | 1.30 (0.66–2.59) | 1.80 (0.65–5.00) |
| 25–30 | 1.55 (0.89–2.72) | 1.51 (0.77–2.98) | 1.65 (0.59–4.56) |
| >30 | 1.29 (0.73–2.27) | 1.30 (0.66–2.55) | 1.18 (0.42–3.35) |
| Charlson comorbidity index | |||
| 0–1 | 1 | 1 | 1 |
| 2–4 | 1.40 (1.11–1.76) | 1.23 (0.94–1.63) | 1.75 (1.15–2.67) |
| ≥5 | 1.65 (1.26–2.16) | 1.46 (1.07–1.99) | 2.24 (1.31–3.84) |
| Race/ethnicity | |||
| Non-Hispanic White | 1 | 1 | 1 |
| Non-Hispanic Black | 0.99 (0.81–1.21) | 0.90 (0.71–1.15) | 1.25 (0.86–1.82) |
| Hispanic | 1.11 (0.80–1.53) | 1.10 (0.77–1.57) | 1.12 (0.51–2.44) |
| Other/ Refused | 1.21 (0.80–1.84) | 1.42 (0.91–2.22) | 0.42 (0.13–1.33) |
| Missing | 0.85 (0.40–1.80) | 1.20 (0.57–2.55) | N/A† |
| Male sex | 2.97 (1.23–7.18) | 2.10 (0.87–5.09) | N/A† |
| HCV Genotype | |||
| Genotype 1 | 1 | 1 | 1 |
| Genotype 2 | 0.78 (0.55–1.12) | 0.74 (0.48–1.15) | 0.93 (0.49–1.79) |
| Genotype 3 | 1.65 (1.21–2.26) | 1.50 (1.05–2.13) | 2.31 (1.20–4.46) |
| Other | 0.76 (0.24–2.38) | 1.00 (0.32–3.14) | 1.27 (0.87–1.86) |
| Missing | 0.91 (0.74–1.12) | 0.78 (0.61–1.01) | N/A* |
| FIB-4 score | |||
| ≤1.4 | 1 | 1 | 1 |
| >1.4 – 2.1 | 1.40 (1.07–1.82) | 1.41 (0.98–2.00) | 1.35 (0.91–2.01) |
| >2.1 – 3.25 | 1.93 (1.47–2.53) | 2.05 (1.45–2.89) | 1.44 (0.88–2.35) |
| >3.25–4.5 | 2.49 (1.79–3.47) | 2.36 (1.59–3.50) | 3.39 (1.69–6.78) |
| >4.5 | 3.79 (2.81–5.12) | 3.75 (2.62–5.36) | 4.40 (1.87–10.38) |
| HBV co-infection | 0.85 (0.49–1.48) | 1.02 (0.57–1.82) | 0.29 (0.04–2.07) |
| Hepatitis B core antibody | |||
| Negative | 1 | 1 | 1 |
| Positive | 1.02 (0.84–1.22) | 1.12 (0.90–1.40) | 0.81 (0.56–1.16) |
| Not tested | 0.99 (0.77–1.29) | 1.27 (0.95–1.69) | 0.50 (0.28–0.89) |
| Cirrhosis | 5.24 (4.25–6.48) | - | - |
Among those without cirrhosis, no females, missing race/ethnicity, and “other” HCV genotype had developed HCC
Adjusted for diabetes, alcohol use disorder, HCV genotype, hypertension, FIB-4 score, age, BMI, Charlson Comorbidity Index, HIV, race and ethnicity, sex, and cirrhosis (model with all patients, only)
Association between time accrued since SVR and risk of HCC among those without baseline cirrhosis
Among 56,287 patients without cirrhosis at baseline, 143 (0.25%) developed HCC during the 1-year follow-up. The 1-year HCC incidence was similar across all groups of time accrued since SVR (0.23 to 0.27 per 100 P-Ys). There was no significant association between time accrued since SVR and risk of HCC among those without cirrhosis on either unadjusted or adjusted analysis, regardless of whether FIB-4 was included in the model (Table 2, Figure 3b). On the adjusted analysis, older age, a higher Charlson Comorbidity Index, HCV genotype 3 (vs. genotype 1), and increasing FIB-4 values were associated with HCC risk (Table 3). There was not a significant association between the accrual of one year after SVR and HCC risk (HR 1.00, 95% CI 0.99–1.01, p=1).
Discussion
The widespread use of direct-acting antiviral medications has dramatically increased the number of patients who have achieved SVR. Understanding how the risk of HCC changes over time following successful treatment is important to guide HCC screening practices, particularly as this population ages and more time accrues since SVR. In this retrospective cohort study, we found that for patients with cirrhosis, those who have accrued more time since SVR have a significantly lower risk of developing HCC compared to those who have accrued less time since SVR. HCC incidence was highest for those who had accrued 1 to 2 years since SVR (2.71 per 100 P-Ys), declined for those who had accrued >2 to 4 (2.11 per 100 P-Ys), and further declined for those who accrued 4 to 6 yrs (1.65 per 100 P-Ys), without any apparent further decline for those who accrued >6 yrs (1.68 per 100 P-Ys). However, HCC risk even for patients who accrued >6 years since SVR exceeded the ~1% per year threshold considered to be cost-effective with current modalities24. Our findings support continued screening indefinitely for those with cirrhosis and even for patients who have accrued >6 years after SVR. In contrast, for those without cirrhosis, the risk of HCC remained below the cost-effective threshold regardless of time accrued since SVR (0.23 to 0.27 per 100 P-Ys).
The decline in HCC risk as time accrues since SVR carries significant implications for HCC risk estimation. Several multivariable models have been developed to predict risk of HCC for those with HCV-cured patients, including one derived from a VA population.17,25–30 These models estimate HCC risk using pre-treatment or SVR-derived predictors which leads to major limitations. As demonstrated in this study, these risk models may over-estimate HCC risk for those patients with cirrhosis who achieved SVR several years prior. Future HCC risk models that consider the time accrued since SVR are needed to evaluate HCC risk more accurately and reliably for those with cured HCV and cirrhosis.
One possible mechanism for this observed association between time since SVR and HCC risk is that HCV eradication reduces liver fibrosis, a strong risk factor for HCC. The effects of liver remodelling and regression of fibrosis post-SVR has been demonstrated in other studies. Facciorusso and colleagues used transient elastography to measure the liver stiffness of 36 patients with cirrhosis annually for up to five years following SVR.6 They found that liver stiffness continued to decline even several years after SVR: liver stiffness was highest at SVR (mean 14.7 kPa), lower at two years (mean 12.6 kPa), and lowest at five years (mean 10.8 kPa). In another prospective cohort study, Mauro and colleagues found that among 37 patients with biopsy-proven cirrhosis (i.e. Metavir stage 4), 16 (43.2%) had regression of at least one Metavir stage on a subsequent biopsy post-SVR.31 Another plausible mechanism is that HCV treatment eliminates the hepatocarcinogenic effects of this virus. HCV induces epigenetic changes that promote HCC, such as the downregulation of tumour suppressor genes, activation of oncogenes, and dysregulation of hepatocellular apoptosis.32 These HCV-induced carcinogenic changes can persist even upon viral eradication,33 but may resolve in the years following SVR as the liver regenerates in the absence of active HCV replication.
This study has several limitations. First, the data was acquired retrospectively and from a predominantly male, VA patient population. We were unable to ascertain the presence of advanced fibrosis (and consequently elevated HCC risk) among those without baseline cirrhosis. Second, we specifically chose a follow-up period that ended on 1/1/2020 prior to the start of the COVID-pandemic. The pandemic has had a detrimental impact on HCC screening and diagnosis.11 In this real-world study, patients did not necessarily undergo regular HCC screening, especially those without cirrhosis in whom HCC screening is not recommended. HCC screening rates have been previously described to decline gradually following SVR35. However, those with HCC who are not diagnosed via screening are expected to eventually be diagnosed with HCC as the tumour grows. If the initial decline seen in HCC incidence was due to “missed” HCC from a decline in HCC surveillance after SVR, then one would expect to see a subsequent rise in HCC incidence in the groups who accrued the most years since SVR as these missed lesions are later diagnosed. Instead, we see that among those with cirrhosis, HCC risk remains highest for those who accrued 1 to 2 (2.71 per 100 P-Ys) and >2 to 4 years (2.11 per 100 P-Ys), and lowest for those who accrued >4 to 6 (1.65 per 100 P-Ys) and >6 years (1.68 per 100 P-Ys) since SVR. It is also possible that in some patients, HCC was present but undiagnosed at the time of HCV treatment, only to be diagnosed shortly after treatment. These prevalent tumours may inflate HCC incidence particularly for those with fewer years of time accrued since SVR. We sought to minimize this possibility by excluding all HCCs that were diagnosed within one year after SVR. Finally, due to the abrupt and complete transition from interferon to DAAs in 2014, it is impossible to disentangle time since SVR from the receipt of IFN versus DAA therapies in our study. That is, in our study, almost everyone with >5 years since SVR was treated with IFN-based therapy and almost everyone with <5 years since SVR was treated with DAAs. However, to date, there has not been any convincing evidence or physiological mechanism that would indicate an association between the type of antiviral therapy used and HCC risk. We acknowledge that there could be a type of selection bias in that some patients with advanced liver disease would not be eligible for interferon-based therapies, but would be eligible for DAA based therapies (i.e. patients treated >5 years ago with IFN may have had less severe disease and hence lower risk of HCC, than those treated <5 years ago with DAAs). However, despite this potential bias, we did not find that HCC risk continued to decline beyond 6 years after SVR (for those who were exclusively treated with IFN-based therapies), even after adjusting for available baseline characteristics.
In conclusion, in this national study of 75,965 patients successfully treated for HCV in the VA healthcare system, we found that HCC risk declined progressively for those with cirrhosis up to 6 years post-SVR and appeared to plateau thereafter, although the absolute risk remained well above thresholds to warrant ongoing screening. The decline in HCC risk following SVR carries important implications for refining HCC surveillance strategies in patients with cured HCV and holds potential for integration into future HCC risk prediction models.
Supplementary Material
Acknowledgements
Other than the authors listed, there are no additional contributors to this manuscript.
Financial Support
- Authors’ declaration of personal interests:
- AASLD Clinical, Translational, and Outcomes Research Award to Dr. Philip Vutien and Andrew Moon.
- Andrew Moon serves as a consultant for TARGET RWE
- All other authors disclose no funding interests in relation to this manuscript.
Declaration of funding interests: This study was funded in part by NIH/NCI grant number R01CA196692, VA CSR&D grant number I01CX001156 and VA HSR&D grant number I01HX003062 to Dr. George N. Ioannou.
List of Abbreviations
- ALT
Alanine aminotransferase
- AST
Aspartate aminotransferase
- BMI
Body mass index
- DAA
Direct acting antiviral
- FIB-4
Fibrosis-4 score
- HCC
Hepatocellular carcinoma
- HIV
Human immunodeficiency virus
- ICD
International Classification of Diseases
- INR
International normalized ratio
- SVR
Sustained virologic response
- VA
Veterans Affairs
References
- 1.Kim D, Cholankeril G, Dennis BB, et al. Trends in the Prevalence of Hepatitis C Virus Infection based on the Insurance Status in the United States from 2013 to 2018. Liver Int. 2022;42(2):340–349. [DOI] [PubMed] [Google Scholar]
- 2.Chhatwal J, Wang X, Ayer T, et al. Hepatitis C Disease Burden in the United States in the era of oral direct-acting antivirals. Hepatology. 2016;64(5):1442–1450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ioannou GN, Green PK, Berry K. HCV eradication induced by direct-acting antiviral agents reduces the risk of hepatocellular carcinoma. J Hepatol. 2017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kanwal F, Kramer JR, Asch SM, Cao Y, Li L, El-Serag HB. Long-Term Risk of Hepatocellular Carcinoma in HCV Patients Treated With Direct Acting Antiviral Agents. Hepatology. 2020;71(1):44–55. [DOI] [PubMed] [Google Scholar]
- 5.Kim NJ, Vutien P, Berry K, Ioannou GN. Hepatocellular carcinoma risk declines but remains high enough for screening in the first 7 years after HCV cure with DAAs in patients with cirrhosis or high FIB-4. Gastroenterology. 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Facciorusso A, Del Prete V, Turco A, Buccino RV, Nacchiero MC, Muscatiello N. Long-term liver stiffness assessment in hepatitis C virus patients undergoing antiviral therapy: Results from a 5-year cohort study. J Gastroenterol Hepatol. 2018;33(4):942–949. [DOI] [PubMed] [Google Scholar]
- 7.Singh S, Facciorusso A, Loomba R, Falck-Ytter YT. Magnitude and Kinetics of Decrease in Liver Stiffness After Antiviral Therapy in Patients With Chronic Hepatitis C: A Systematic Review and Meta-analysis. Clin Gastroenterol Hepatol. 2018;16(1):27–38 e24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bandiera S, Billie Bian C, Hoshida Y, Baumert TF, Zeisel MB. Chronic hepatitis C virus infection and pathogenesis of hepatocellular carcinoma. Curr Opin Virol. 2016;20:99–105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Jeong SW, Jang JY, Chung RT. Hepatitis C virus and hepatocarcinogenesis. Clin Mol Hepatol. 2012;18(4):347–356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kim NJ, Vutien P, Cleveland E, Cravero A, Ioannou GN. Fibrosis Stage-specific Incidence of Hepatocellular Cancer After Hepatitis C Cure With Direct-acting Antivirals: A Systematic Review and Meta-analysis. Clin Gastroenterol Hepatol. 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kim NJ, Rozenberg-Ben-Dror K, Jacob DA, Berry K, Ioannou GN. The COVID-19 Pandemic Highlights Opportunities to Improve Hepatocellular Carcinoma Screening and Diagnosis in a National Health System. Am J Gastroenterol. 2022;117(4):678–684. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ioannou GN, Green P, Lowy E, Mun EJ, Berry K. Differences in hepatocellular carcinoma risk, predictors and trends over time according to etiology of cirrhosis. PLoS One. 2018;13(9):e0204412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Sterling RK, Lissen E, Clumeck N, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology. 2006;43(6):1317–1325. [DOI] [PubMed] [Google Scholar]
- 14.Tapper EB, Parikh ND, Green PK, et al. Reduced Incidence of Hepatic Encephalopathy and Higher Odds of Resolution Associated With Eradication of HCV Infection. Clin Gastroenterol Hepatol. 2020;18(5):1197–1206 e1197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Vutien P, Kim NJ, Moon AM, et al. Fibroscan liver stiffness after anti-viral treatment for hepatitis C is independently associated with adverse outcomes. Aliment Pharmacol Ther. 2020;52(11–12):1717–1727. [DOI] [PubMed] [Google Scholar]
- 16.Ioannou GN, Beste LA, Green PK, et al. Increased Risk for Hepatocellular Carcinoma Persists Up to 10 Years After HCV Eradication in Patients With Baseline Cirrhosis or High FIB-4 Scores. Gastroenterology. 2019;157(5):1264–1278 e1264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ioannou GN, Green PK, Beste LA, Mun EJ, Kerr KF, Berry K. Development of models estimating the risk of hepatocellular carcinoma after antiviral treatment for hepatitis C. J Hepatol. 2018;69(5):1088–1098. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Dafni U. Landmark analysis at the 25-year landmark point. Circ Cardiovasc Qual Outcomes. 2011;4(3):363–371. [DOI] [PubMed] [Google Scholar]
- 19.Kanwal F, Hoang T, Kramer JR, et al. Increasing prevalence of HCC and cirrhosis in patients with chronic hepatitis C virus infection. Gastroenterology. 2011;140(4):1182–1188 e1181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Davila JA, Weston A, Smalley W, El-Serag HB. Utilization of screening for hepatocellular carcinoma in the United States. J Clin Gastroenterol. 2007;41(8):777–782. [DOI] [PubMed] [Google Scholar]
- 21.Fine JP, Gray RJ. A Proportional Hazards Model for the Subdistribution of a Competing Risk. Journal of the American Statistical Association. 1999;94(446):496–509. [Google Scholar]
- 22.Kim NJ, Pearson M, Vutien P, et al. Alcohol Use and Long-Term Outcomes Among U.S. Veterans Who Received Direct-Acting Antivirals for Hepatitis C Treatment. Hepatol Commun. 2020;4(2):314–324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Pearson MM, Kim NJ, Berry K, et al. Associations Between Alcohol Use and Liver-Related Outcomes in a Large National Cohort of Patients With Cirrhosis. Hepatol Commun. 2021;5(12):2080–2095. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Farhang Zangneh H, Wong WWL, Sander B, et al. Cost Effectiveness of Hepatocellular Carcinoma Surveillance After a Sustained Virologic Response to Therapy in Patients With Hepatitis C Virus Infection and Advanced Fibrosis. Clin Gastroenterol Hepatol. 2019;17(9):1840–1849 e1816. [DOI] [PubMed] [Google Scholar]
- 25.Yamashita Y, Joshita S, Sugiura A, et al. aMAP score prediction of hepatocellular carcinoma occurrence and incidence-free rate after a sustained virologic response in chronic hepatitis C. Hepatol Res. 2021;51(9):933–942. [DOI] [PubMed] [Google Scholar]
- 26.Azzi J, Dorival C, Cagnot C, et al. Prediction of hepatocellular carcinoma in Hepatitis C patients with advanced fibrosis after sustained virologic response. Clin Res Hepatol Gastroenterol. 2022;46(6):101923. [DOI] [PubMed] [Google Scholar]
- 27.Tamaki N, Kurosaki M, Yasui Y, et al. Hepatocellular Carcinoma Risk Assessment for Patients With Advanced Fibrosis After Eradication of Hepatitis C Virus. Hepatol Commun. 2022;6(3):461–472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Semmler G, Meyer EL, Kozbial K, et al. HCC risk stratification after cure of hepatitis C in patients with compensated advanced chronic liver disease. J Hepatol. 2022;76(4):812–821. [DOI] [PubMed] [Google Scholar]
- 29.Alonso Lopez S, Manzano ML, Gea F, et al. A Model Based on Noninvasive Markers Predicts Very Low Hepatocellular Carcinoma Risk After Viral Response in Hepatitis C Virus-Advanced Fibrosis. Hepatology. 2020;72(6):1924–1934. [DOI] [PubMed] [Google Scholar]
- 30.Fan R, Papatheodoridis G, Sun J, et al. aMAP risk score predicts hepatocellular carcinoma development in patients with chronic hepatitis. J Hepatol. 2020;73(6):1368–1378. [DOI] [PubMed] [Google Scholar]
- 31.Mauro E, Crespo G, Montironi C, et al. Portal pressure and liver stiffness measurements in the prediction of fibrosis regression after sustained virological response in recurrent hepatitis C. Hepatology. 2018;67(5):1683–1694. [DOI] [PubMed] [Google Scholar]
- 32.Zhao P, Malik S, Xing S. Epigenetic Mechanisms Involved in HCV-Induced Hepatocellular Carcinoma (HCC). Front Oncol. 2021;11:677926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Hamdane N, Juhling F, Crouchet E, et al. HCV-Induced Epigenetic Changes Associated With Liver Cancer Risk Persist After Sustained Virologic Response. Gastroenterology. 2019;156(8):2313–2329 e2317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Kanwal F, Kramer J, Asch SM, Chayanupatkul M, Cao Y, El-Serag HB. Risk of Hepatocellular Cancer in HCV Patients Treated With Direct-Acting Antiviral Agents. Gastroenterology. 2017;153(4):996–1005 e1001. [DOI] [PubMed] [Google Scholar]
- 35.Hamill V, Gelson W, MacDonald D, et al. Delivery of biannual ultrasound surveillance for individuals with cirrhosis and cured hepatitis C in the UK. Liver Int. 2023;43(4):917–927. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.