Fibrosis-4 Index Predicts Long-Term Outcomes After Sustained Virologic Response in Hepatitis C Virus Patients with Cirrhosis

Abstract

Background: Direct-acting antivirals (DAAs) achieve high sustained virologic response (SVR) in hepatitis C virus (HCV) patients with cirrhosis, yet the risk of hepatic decompensation or hepatocellular carcinoma (HCC) persists. Fibrosis-4 (FIB-4), a pre-treatment index that predicts advanced fibrosis, is linked to HCC risk post SVR. We compared post-SVR outcomes and care engagement, and determined the optimal pre-treatment FIB-4 index to predict the risk of HCC or decompensation in HCV patients with cirrhosis treated at the Department of Corrections (DOC) and non-DOC clinics. Methods: HCV patients with cirrhosis treated with DAAs since 2014 in the HCV Treatment Registry were included. Cirrhosis was defined by elastography, imaging, or clinical criteria. Patients with prior decompensation or HCC were excluded. Outcomes (HCC, decompensation) were collected from records. The FIB-4 index was compared between DOC and non-DOC groups. Results: Among 2104 cirrhotic patients (mean age 54; 76% male), 53% were treated in DOC via telemedicine and 47% in non-DOC clinics. HCC developed in 4.8% and decompensation in 8.1%. DOC patients had lower FIB-4 scores and SVR, partly from higher loss to follow-up (9% vs. 1%). Of 1581 with follow-up, surveillance was more common in non-DOC, which also had higher HCC and decompensation. A higher baseline FIB-4 index independently predicted HCC and decompensation (cutoffs: 3.24, 3.7; AUROC 0.79, 0.75, respectively). Conclusions: Despite SVR, cirrhotic patients—especially with a high FIB-4 index—remain at risk for HCC and decompensation. Outcomes differ by care setting, highlighting the need for continued AASLD-recommended surveillance post-SVR.

1. Introduction

Chronic hepatitis C virus (HCV) is among the most common causes of hepatic cirrhosis in the developed world. HCV accounts for an estimated 41% of the 2.2 million adult cirrhosis cases in the United States [1]. The advent of direct-acting antivirals (DAAs) has transformed HCV treatment, with sustained virologic response (SVR) rates exceeding 95% with shorter treatment courses (8–12 weeks) and improved safety profiles [2]. The ease of an all-oral regimen has also allowed HCV treatment through telemedicine in difficult-to-access patient populations, such as the Department of Corrections (DOC) [3,4]. However, even after achieving SVR, the risk of liver decompensation and hepatocellular carcinoma (HCC) is not entirely eliminated. It is estimated that there is up to an 86% reduction in the risks of hepatic decompensation [5], with the annual risk of hepatic decompensation averaging about 0.5–1% for those with compensated cirrhosis [6,7]. Additionally, systematic reviews and meta-analyses found a 71% relative risk reduction in the development of HCC after achieving SVR [6]. Therefore, ongoing monitoring after SVR is crucial in individuals with advanced fibrosis or cirrhosis as they remain at elevated risk for HCC and hepatic decompensation despite viral clearance. Surveillance and follow-up facilitate early detection and treatment of conditions such as HCC and portal hypertension, which may arise years after SVR, and ultimately improve patient outcomes through timely intervention [7].

The fibrosis-4 (FIB-4) index—which includes age, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and platelet count—was developed to identify advanced fibrosis [8]. Current guidelines recommend ongoing imaging surveillance for HCC after SVR in those with a baseline or post-SVR FIB-4 index of 3.25 or greater [9]. However, there are limited data on the use of pre-treatment FIB-4 in predicting hepatic decompensation after SVR. Furthermore, the engagement in care in those treated by telemedicine is poorly defined—particularly as compared to those followed in in-person clinics. Current surveillance guidelines need to be evaluated in different settings. We hypothesized that remaining engaged in care would be higher in the non-DOC population and that those with a higher pre-treatment FIB-4 would be at higher risk of developing HCC or decompensation after SVR. To fill these gaps in knowledge, we investigated the degree of engagement in care for HCC surveillance and the development of HCC or hepatic decompensation, and compared those treated in dedicated in-person HCV clinics compared to those treated through telemedicine. We also used this retrospective data to define the optimal pre-treatment FIB-4 index threshold that predicts both HCC and decompensation after HCV treatment.

2. Methods

2.1. Study Population

This was a retrospective study of all HCV patients with cirrhosis treated with DAAs between 2014–2025 at Virginia Commonwealth University (VCU) by experienced hepatology providers. All subjects were prospectively enrolled in our institutional review board (IRB)-approved HCV Treatment Registry. After obtaining written informed consent, comprehensive demographic and clinical data were systematically collected and entered into a secure REDCap (VCU, Richmond, VA) database for longitudinal follow-up. The study was conducted to comply with the principles of the Declaration of Helsinki for medical research involving human subjects.

2.2. Clinical and Laboratory Data

Baseline data included demographics (age, sex, race), laboratory tests (hepatic panel, heme 8, HCV RNA and genotype, and HIV and hepatitis B virus [HBV] serologies), controlled attenuated parameter (CAP), FIB-4 index, and HCV treatment clinic (in-person dedicated non-DOC HCV clinic or the telemedicine clinic for DOC subjects) as previously described [4]. The target population was those treated with compensated cirrhosis. Cirrhosis at baseline was determined using a combination of diagnostic modalities, including vibration-controlled transient elastography (VCTE, FibroScan^®, Paris, France) [10] of at least 12.5 kPa, cross-sectional imaging (nodular liver), and established clinical criteria including the presence of Terry nails, gynecomastia, caput medusa, facial telangiectasia, palmar erythema, decreased body hair, and testicular atrophy [1]. Those with untreated HBV, total bilirubin ≥ 2.0 mg/dL, history of HCC or hepatic decompensation before DAA treatment, or pregnancy were excluded from this analysis. We include those patients with non-hepatic cancers in our HCV treatment practice as long as their life expectancy is long enough to benefit from HCV treatment.

2.3. Follow-Up Data

A detailed review of the electronic health records was performed for each patient and data were collected from the date they completed DAA until the last date they had an office visit or laboratory tests, to determine the degree of engagement in care after they completed DAA and whether they were still alive. The degree of engagement was measured by the percentage of patients who remained in care with office visits and/or laboratory studies after the SVR-12 visit. Hepatic decompensation was defined as the development of one or more of the following complications: ascites, hepatic encephalopathy (HE), portal hypertensive bleeding (e.g., variceal hemorrhage), or clinically apparent jaundice or total bilirubin ≥ 2.5 mg/dL. HCC was determined by imaging according to the LIRADS criteria [11] or biopsy.

2.4. Statistical Analysis

Continuous variables were expressed as mean and standard deviation (SD) or as median and interquartile range (IQR) for normally distributed or skewed data, respectively, and categorical variables as percentages. The primary clinical outcomes of interest were the development of hepatic decompensation and/or HCC after SVR-12 in those that remained engaged in care. Those treated through the non-DOC in-person clinics were compared to DOC subjects treated via HCV telemedicine. Those variables with a p-value < 0.2 on univariate analysis were included in the multivariate models to predict HCC or hepatic decompensation (JMP Pro 18 statistical software). The area under the receiver operating curve (AUROC), in conjunction with a Youden index calculation, was used to identify the FIB-4 threshold for developing hepatic decompensation or HCC after DAA treatment. Once we established said thresholds for either liver-related outcome, bootstrapping with 1000 resamples was performed to estimate optimism-corrected AUROC and derive stable Youden cut points in order to reduce bias from the data-driven thresholds in the subsequent survival analysis. Kaplan–Meier analyses were done to evaluate for any difference in the development of a liver-related outcome. Furthermore, a Cox proportional hazard analysis (for hazard ratio [HR] and 95% CI) was completed for both hepatic decompensation and HCC. We included variables that had significant univariate association as covariates in the Cox models, while excluding variables used as components in FIB-4 to reduce multicollinearity. A sensitivity analysis was performed on DOC and non-DOC patients separately, conducting the same AUROC, followed by bootstrap corrected AUROC and Youden indexes used for Kaplan–Meier and adjusted Cox Proportional Hazards analysis (using R Version 4.5.1).

The author(s) used Chat Generative Pre-Trained Transformer (ChatGPT, GPT-5.3, OpenAI, San Francisco, CA) to assist in preparing this work, which included generating portions of the graphical abstract and proofreading the manuscript. All content was subsequently reviewed and edited by the author(s), who take full responsibility for the final publication.

3. Results

3.1. Patient Characteristics

A total of 2104 patients with chronic HCV-related cirrhosis without HCC or prior decompensation were included in this analysis (Figure 1,

Table 1. Baseline characteristics of the study cohort.

Characteristic	Total	Non-DOC Clinic	DOC Clinic	p-Value
Number of patients	2104	991 (47%)	1113 (53%)
Age (years) 1	54 (11)	58 (8)	50 (11)	<0.0001
Sex (% male)	76	62	89	<0.0001
Race (% White/African-American/Other)	58/40/2	47/51/2	69/30/1	<0.0001
HIV positive (%)	3.7	5.4	2.4	0.0003
HBV sAg positive (%)	1.7	1.9	1.5	ns
AST (U/L) 2	69 (46–105)	76 (50–115)	63 (42–96)	<0.0001
ALT (U/L) 2	73 (45–121)	70 (44–114)	76 (45–125)	0.007
Total bilirubin (mg/dL) 1	0.75 (0.43)	0.87 (0.47)	0.71 (0.38)	<0.0001
Albumin (g/L) 1	4.0 (0.87)	3.8 (0.52)	4.1 (0.54)	<0.0001
Platelet count (×1000) 1	172 (73)	159 (74)	184 (70)	<0.0001
Pre-DAA FIB-4 score	2.6 (1.5–4.8)	3.7 (2.1–6.0)	2.0 (1.2–3.5)	<0.0001
Genotype 1 (%)	82	87	77	<0.0001
CAP (dB/m) 1	238 (82)	250 (58)	233 (90)	0.002
In HCC surveillance imaging after SVR-12 (%)	71	81	63	<0.0001
Lost to care before SVR-12 visit (%)	4	1	9	<0.0001
SVR-12 (%)	91	94	89	<0.0001
Remained engaged in care (%) after SVR-12 3	80	70	89	<0.0001

¹ Mean ± standard deviation (SD). ² Median (interquartile range, IQR). ³ Engaged in care regardless of in HCC surveillance. HIV, human immunodeficiency virus; HBV, hepatitis B virus; AST, aspartate aminotransferase; ALT, alanine aminotransferase; DAA, direct acting antiviral; CAP, controlled attenuated parameter; SVR-12, sustained virologic response at 12 weeks after completing DAA treatment; ns, not significant.

). At baseline, significant differences were observed between patients treated through telemedicine in DOC setting versus those managed in-person in the non-DOC environments. Individuals treated in the DOC had lower baseline FIB-4 scores (2.0 vs. 3.7; p < 0.0001) and also had lower SVR rates (89% vs. 94%; p < 0.0001). The reduced SVR in the DOC cohort was likely attributable to a lower rate of engagement during treatment and higher loss to follow-up (9% vs. 1%; p < 0.0001) before the SVR-12 assessment.

3.2. Liver-Related Outcomes

Longitudinal follow-up data were available for 1518 patients (72% of the total cohort). Following SVR-12, more patients treated in the DOC did not remain engaged in care compared to those treated in the non-DOC (in-person) clinics (30% vs. 15%; p < 0.0001) (Figure 1). Among these, HCC surveillance imaging was more frequently performed in non-DOC patients compared to those in the DOC group (81% vs. 63%; p < 0.0001) (

Table 2. Liver-related outcomes in those engaged in care by treatment clinic after SVR-12.

	Total	Non-DOC Clinic	DOC Clinic	p-Values
N	1518	683	835
Yes (%)	12.9	19.3	6.8	<0.0001
Decompensation (%)	8.1	11.4	3.0	<0.0001
Ascites (%)	7.0	11.5	2.8	<0.0001
Hepatic encephalopathy (%)	5.0	7.5	2.8	<0.0001
Portal hypertensive bleeding (%)	2.5	3.5	1.6	0.009
Jaundice (%)	1.8	1.8	1.7	0.86
HCC (%)	4.8	8.3	1.7	<0.0001
Alive (%)	92	89	96	<0.0001
Not engaged in care after SVR (%)	29	23	36	p < 0.0001

HCC, hepatocellular carcinoma; SVR, sustained virologic response.

). Of those who remained in follow-up, 12.9% had a liver-related clinical outcome. Of these patients, HCC developed in 4.8% of them, while 8.1% experienced hepatic decompensation. Of those who decompensated, 7% developed ascites, 5% experienced HE, 2.5% suffered from portal hypertensive bleeding, and 1.8% presented with jaundice, defined as a serum bilirubin level exceeding 1.5 mg/dL. Clinical outcomes varied significantly between treatment settings (Table 2). Patients followed in non-DOC clinics had a markedly higher incidence of both HCC (8.3% vs. 1.7%; p < 0.0001) and hepatic decompensation (11.4% vs. 3%; p < 0.0001) compared to those followed in the DOC.

3.3. Predictors of Liver-Related Outcomes

A univariate analysis found that baseline age, male sex, AST, ALT, total bilirubin, albumin, platelet count, FIB-4, CAP, and non-DOC clinic were associated with developing hepatic decompensation, while age, AST, total bilirubin, albumin, platelet counts, FIB-4, SVR-12, and non-DOC clinic were associated with developing HCC (

Table 3. Factors associated with liver-related outcomes (n = 1518).

	Hepatic Decompensation	HCC	No Liver Related Outcome	Univariate for Decompensation	Multivariate for Decompensation 3	Univariate for HCC	Multivariate for HCC 4
N	158	95	1265
Age (years) 1	56 (10)	60 (5)	53 (11)	0.01		<0.0001
Sex (% male)	64	76	77	0.001		0.66
Race (% White/Black/other)	67/32/1	52/44/4	58/41/1	0.25		0.28
HIV positive (%)	3.2	3.2	3.9	0.86		0.51
HBV sAg positive (%)	2.5	2.1	1.6	0.68		0.99
AST (U/L) 2	75 (48–106)	82 (51–113)	67 (45–103)	0.003		<0.0001
ALT (U/L) 2	66 (41–116)	72 (43–124)	75 (45–124)	0.01		0.255
Total bilirubin (mg/dL) 1	1.06 (0.47)	1.38 (0.81)	0.72 (0.36)	<0.0001		<0.001	<0.0001
Albumin (g/L) 1	35 (0.62)	3.6 (0.55)	4.07 (0.52)	<0.0001	0.0001	<0.0001
Platelet count (×1000) 1	122 (68)	118 (66)	181 (71)	<0.0001		<0.0001
Pre-DAA FIB-4 score 2	4.0 (2.1–6.8)	3.9 (2.0–6.8)	2.38 (1.4–4.1)	<0.0001	0.0002	<0.0001	0.033
Genotype 1 (%)	72	84	81	0.004		0.68
CAP (dB/m) 2	197 (99)	238 (95)	241 (80)	0.0018		0.96
SVR-12 (%)	97	88	95	0.08		0.001
Non-DOC clinic (%)	34	19	56	<0.0001		<0.0001	0.0003

¹ Mean (SD). ² Median (IQR). Variables that had significant univariate association were included as covariates in the multivariable models; however, the individual components of FIB-4 were not included as covariates regardless of significance in order to avoid multilinearity, as they had elevated variance inflation factors (VIF) relative to other covariates, and including them would introduce multicollinearity and compromise the interpretability of the FIB-4 composite score. ³ Model for decompensation includes sex, total bilirubin, albumin, CAP, SVR-12, DOC clinic, and FIB-4 (age, AST, ALT, and platelet count not included since they are part of FIB-4). ⁴ Model for HCC includes total bilirubin, albumin, FIB-4, SVR-12, and DOC clinic (age, AST, ALT, and platelet count not included since they are part of FIB-4). HIV, human immunodeficiency virus; HBV, hepatitis B virus; HBsAg, hepatitis B surface antigen; AST, aspartate aminotransferase; ALT, alanine aminotransferase; DAA, direct acting antiviral; CAP, controlled attenuated parameter; SVR-12, sustained virologic response at 12 weeks after completing DAA treatment; HCC, hepatocellular carcinoma; DOC, Department of Corrections; ns, not significant.

). Multivariate analysis adjusting for demographic, laboratory, and treatment clinic demonstrated that baseline FIB-4 was highly predictive of both HCC (p = 0.033) and decompensation (p = 0.0002). The optimal pre-treatment (DAA) FIB-4 threshold that best predicted post-SVR development of HCC or decompensation was identified as 3.17 and 3.7, respectively, with an unadjusted AUROC of 0.79 for HCC development and 0.75 for decompensation, indicating good predictive accuracy (Figure 2a,b). When performing bootstrap-corrected AUROC, AUROCs remained nearly identical: 0.79 (0.001 optimism) for HCC and 0.75 (−0.001 optimism) for decompensation. Bootstrap resampling yielded a slight increase in the Youden threshold for HCC, with a median optimal threshold of 3.24 (IQR 0.48), while the threshold for decompensation remained identical at 3.70 (IQR 0.34). When testing for multicollinearity in the hazard models for AST, ALT, and platelets for both decompensation and HCC, the results did not appreciably change further justifying the decision to exclude these variables in the model in order to preserve the interpretability of the FIB-4 index.

Kaplan–Meier analyses using the bootstrap derived thresholds determined that there was a highly significant difference (p < 0.0001) between patients who had a FIB-4 score less than the above thresholds: 3.24 for HCC (Figure 3a) and 3.70 for decompensation (Figure 3b). Cox proportional hazards analysis on HCC outcome showed FIB-4 and patients who remained in follow up after SVR-12 visit were significant predictors (p = 0.001 and p < 0.001, respectively) (

Table 4. (a) Cox proportional hazard for developing HCC. (b) Cox proportional hazard for developing hepatic decompensation.

(a)
	Cox Proportional Hazard on HCC Outcome Using Youden Index
Predictors	Estimates	CI	p
FIB 4 [≥3.24]	3.96	1.76–8.88	0.001
TB	1.62	0.84–3.14	0.150
Alb	0.94	0.61–1.46	0.790
Non-DOC Clinic	0.77	0.39–1.51	0.449
(b)
Cox Proportional Hazard on Decomposition Outcome Using Youden Index
Predictors	Estimates	CI	p
FIB 4 [≥3.7]	2.66	1.25–5.64	0.011
Sex [Male]	0.74	0.34–1.61	0.450
TB	2.60	1.19–5.69	0.017
Alb	0.54	0.37–0.78	0.001
CAP	1.00	0.99–1.00	0.379
Non-DOC Clinic	0.61	0.27–1.35	0.218

a). Conversely, a Cox analysis found FIB-4 and serum albumin to be significant predictors of hepatic decompensation (p = 0.011 and p = 0.001, respectively) (Table 4b).

In a sensitivity analysis for only DOC patients, FIB-4 yielded an unadjusted AUC of 0.807 for HCC and 0.744 for decompensation, with Youden indexes of 3.43 and 4.77, respectively. Optimism-corrected bootstrapping yielded an AUROC of 0.804 (0.003 optimism) and a median Youden index of 3.43 (IQR 0.685) for HCC, and an AUROC of 0.744 (0.001 optimism) with a median Youden index of 4.03 (IQR 1.074) for decompensation. Using the optimism corrected Youden thresholds, FIB-4 > 3.43 yielded significantly worse survival in Kaplan–Meier analysis (p = 0.04) for HCC and had an adjusted hazard ratio of 4.02 (p = 0.057). For decompensation, FIB-4 > 4.03 yielded significantly worse survival (p < 0.001) in Kaplan–Meier analysis, and had an adjusted hazard ratio of 5.16 (p = 0.001).

In a sensitivity analysis for only non-DOC patients, FIB-4 yielded an unadjusted AUC of 0.727 for HCC and 0.706 for decompensation, with Youden indexes of 2.92 and 3.7, respectively. Optimism-corrected bootstrapping yielded an AUROC of 0.728 (−0.001 optimism) and a median Youden index of 4.082 (IQR 2.604) for HCC, and an AUROC of 0.708 (−0.002 optimism) with a median Youden index of 4.06 (IQR 0.66) for decompensation. Using the optimism corrected Youden thresholds, FIB-4 > 4.08 yielded significantly worse survival in Kaplan–Meier analysis (p < 0.001) for HCC and had an adjusted hazard ratio of 2.58 (p = 0.016). For decompensation, FIB-4 > 4.06 yielded significantly worse survival (p < 0.001) in Kaplan–Meier analysis, but had a non-significant adjusted hazard ratio of 0.46 (p = 0.31).

4. Discussion

The development of DAAs has dramatically changed the management of chronic HCV. However, those with cirrhosis who achieve SVR are still at risk for developing HCC, albeit at a lower risk [5]. In this study, we compared post-SVR outcomes and care engagement, and determined the optimal pre-treatment FIB-4 index to predict the risk of HCC or decompensation in HCV patients with cirrhosis treated at Department of Corrections (DOC) and non-DOC clinics. While prior studies used the FIB-4 threshold to identify advanced fibrosis to estimate the risk of developing HCC after SVR [9], our data identified unique, outcome-specific FIB-4 thresholds (3.24 for HCC and 3.70 for hepatic decompensation) after SVR. We also show the differences among healthcare settings in influencing long-term follow-up and outcomes, revealing marked disparities between DOC, a unique and underserved group, and non-DOC treatment populations—while at the same time supporting the importance of surveillance irrespective of the healthcare setting.

Our study revealed significant differences in baseline characteristics between patients treated in DOC settings and those treated in non-DOC clinics. DOC patients were more likely to have lower median baseline FIB-4 scores, suggesting less advanced liver disease at treatment initiation. This may have been due to referral bias. Despite these differences, SVR rates were high across both groups, though SVR documentation was more often incomplete in the DOC cohort given its higher loss to follow-up rate. These findings are consistent with existing literature demonstrating that DAA therapies are highly effective across diverse populations, but that care continuity and documentation can vary substantially in correctional or resource-limited settings [12,13].

4.1. Engagement in Care by Treatment Site

Engagement in post-SVR longitudinal liver care was notably higher in non-DOC settings (81%) compared to DOC clinics (63%), with a 9% loss to follow-up prior to the SVR-12 visit in the DOC group versus 1% in non-DOC. These disparities underscore systemic barriers to long-term care among incarcerated or recently released individuals, including fragmented care transitions and limited access to specialty services. This pattern mirrors findings from Tapper et al., which emphasized the role of care infrastructure in maintaining surveillance adherence and optimizing long-term outcomes post-SVR [14].

4.2. Hepatocellular Carcinoma and Decompensation Outcomes in Those Engaged in Care (By Treatment Site)

Among those who remained engaged in care, non-DOC patients experienced a higher burden of liver-related complications, including HCC and hepatic decompensation. The incidence of HCC was 8.3% and decompensation was 11.4% in non-DOC patients, compared to 1.7% and 3%, respectively, in the DOC group—further reinforcing the fact that achievement of SVR decreases the risk of hepatic decompensation and/or development of HCC but does not eliminate it [6]. However, these differences may be partially attributable to lower baseline FIB-4 scores and the aforementioned under-documentation of outcomes in the DOC group. Moreover, lower rates of HCC surveillance imaging in DOC patients (63% vs. 81%) raise concern for missed diagnoses and underscore the urgent need for standardized, guideline-concordant follow-up protocols across treatment settings.

4.3. Use of FIB-4 to Predict Liver-Related Outcomes

Several studies have shown that increased baseline FIB-4 predicts both hepatic decompensation and HCC after SVR [9,15], signifying its utility and versatility as a clinical risk stratification tool. Our findings reinforce the utility of the FIB-4 score as a practical and accessible tool for post-SVR risk stratification. Previous studies used a FIB-4 threshold of 3.25 (developed to identify advanced fibrosis prior to HCV treatment) [9,15] to predict post-SVR HCC. We identified a similar baseline optimal threshold (3.24) for identifying those who might develop HCC after SVR. Given that better outcomes are associated with earlier HCC detection, the threshold should identify more patients eligible for locoregional therapy and transplantation.

Although increased baseline FIB-4 predicts hepatic decompensation after SVR, there is no established pre-treatment threshold to stratify patients’ risk of developing hepatic decompensation. Rather than use the advanced fibrosis FIB-4 threshold of 3.25, we identified a pre-treatment FIB-4 threshold of 3.70 as the optimal cutoff to predict the future development of decompensation following treatment. This aligns with prior studies that validate FIB-4 scoring as a prognostic value [8]. Our data suggest that a slightly higher cutoff of 3.70 may be more appropriate for identifying patients needing intensified decompensation monitoring, and that different cutoffs, specifically that of 3.24 for identifying risk of HCC development, may be more applicable and more specific to different populations. While age, AST, and platelet count were also significant on univariate analysis, these variables are included in FIB-4 and, therefore, are not included in the final models to reduce multicollinearity.

4.4. Strengths and Limitations of the Study

Key strengths of this study include its real-world, large cohort, multicenter design, and the inclusion of a racially diverse population. This allows for more generalizable insights into treatment disparities and long-term outcomes in both correctional and academic settings. The use of routinely collected laboratory data to assess FIB-4 enhances the study’s translational value.

However, several limitations must be acknowledged. First, the patient population studied was mostly composed of male patients. Second, while telemedicine is a viable option to access healthcare, we observed a higher loss to follow-up—particularly in the DOC population—after SVR-12 with lower clinical and HCC surveillance intensity. The DOC operates under a unique, structured system of health care that is not replicable outside of the DOC. This may have led to an underestimation of adverse outcomes and biased our results. Given the higher rate of non-engagement in the DOC cohort, the reported lower incidence rates of decompensation and HCC are minimum estimates, and the true population risk is likely higher. Therefore, the attrition bias might differentially affect the DOC vs. non-DOC groups, and these outcomes should not be compared between the groups. The lower FIB-4 among DOC treated patients may have resulted in bias in our data. Third, about half of the patient population studied were DOC patients, which may not be representative of all telemedicine populations including those treated by telemedicine outside of the DOC. Fourth, we did not analyze dynamic changes in fibrosis markers over time, which could provide deeper prognostic insight. Furthermore, the retrospective nature of the study and the possible residual confounding effects from factors such as alcohol use, metabolic disease, or incomplete imaging adherence may have influenced clinical trajectories despite statistical adjustment—especially in that of the non-DOC cohort. Therefore, the hazard estimates and AUROCs should be interpreted cautiously. Additionally, our use of Youden’s index on unadjusted ROC curves and then in our Kaplan–Meir curves and Cox models may have resulted in optimistic estimates of discrimination. Our FIB-4 thresholds should be used as a pragmatic risk stratification tool that needs external validation and limits the generalizability of a single threshold in all clinic settings. Additionally, because VCTE are not recommended after SVR [10], we were not able to assess its use after SVR to predict liver-related outcomes.

The individual components of the FIB-4 index were not included in the multivariate models; our aim was to assess the prognostic utility of the clinically established composite score itself. As such, their effects are inherently captured within the composite score, though this approach prevents evaluation of their individual contributions to outcomes. However, when tested for multicollinearity, the results did not change. We also did not adjust for prior decompensation because our inclusion/exclusion criteria eliminated those patients. Additionally, all patients in our cohort had compensated cirrhosis at baseline, resulting in uniformly low Model for End Stage Liver Disease (MELD) scores with little variability. The MELD score was not included because it is primarily used to predict short-term mortality and guide transplant prioritization in patients with decompensated cirrhosis. In this setting, MELD would not provide additional discriminatory value beyond what was already captured by our study variables. Another limitation is the potential for follow-up bias, as only 72% of the initial 2104 subjects remained in follow-up after SVR with differences in engagement in care between DOC and non-DOC subjects. Patients who remained engaged in care were more likely to undergo laboratory testing and have FIB-4 values available, while those lost to follow-up may have experienced worse outcomes that went undocumented, which reflects a real-world experience in treating HCV. These differences may underestimate the development of HCC and decompensation in the DOC population and inflate the apparent differences between the two ambulatory settings.

5. Conclusions

Elevated pre-treatment FIB-4 scores are a predictor of HCC after SVR, serving as a low-cost and easily accessible surrogate marker for more severe hepatic injury and fibrotic burden at the time of antiviral treatment initiation [8,15,16]. Our data suggest that pre-treatment FIB-4 thresholds of 3.70 for decompensation and 3.24 for HCC development can identify those at highest risk for these liver-related outcomes in chronic HCV after SVR. However, these FIB-4 thresholds require validation in purely non-carceral, community-based populations. There are differences in engagement in care after SVR among DOC patients treated through telemedicine and those treated in-person, reinforcing previous research on the mortality impact of such disparities in post-HCV treatment. In defining this accessible FIB-4 marker, our research takes a step towards ameliorating the burden of liver outcomes following HCV infection.