Systematic review: Identifying patients with chronic hepatitis C in need of early treatment and intensive monitoring predictors and predictive models of disease progression

Summary

Background

Advances in hepatitis C therapies have led to increasing numbers of patients seeking treatment. As a result, logistical and financial concerns regarding how treatment can be provided to all patients with chronic hepatitis C (CHC) have emerged.

Aim

The aim of this review was to evaluate predictors and predictive models of histologic progression and clinical outcomes for patients with CHC.

Methods

MEDLINE via PubMed, EMBASE, Web of Science and Scopus were searched for studies published between January 2003 and June 2014.Two authors independently reviewed articles to select eligible studies and performed data abstraction.

Results

Twenty-nine studies representing 5817 patients from 20 unique cohorts were included. The outcome incidence rates were widely variable: 16-61% during median follow-up of 2.5-10 years for fibrosis progression; 13-40% over 2.3-14.4 years for hepatic decompensation; and 8-47% over 3.9-14.4 years for overall mortality. Multivariate analyses showed that baseline steatosis and baseline fibrosis score were the most consistent predictors of fibrosis progression (significant in 6/21 and 5/21, studies, respectively) while baseline platelet count (significant in 6/13 studies), aspartate and alanine aminotransferase (AST/ALT) ratio, albumin, bilirubin, and age (each significant in 4/13 studies) were the most consistent predictors of clinical outcomes. Five studies developed predictive models but none were externally validated.

Conclusions

Our review identified the variables that most consistently predict outcomes of patients with CHC allowing the application of risk based approaches to identify patients in need of early treatment and intensive monitoring. This approach maximizes effective use of resources and costly new direct-acting antiviral agents.

Introduction

With the introduction of more efficacious and less toxic drugs, treatment of chronic hepatitis C (CHC) is evolving at a rapid pace. The two new direct-acting antiviral agents (DAA), simeprevir and sofosbuvir, increase rates of sustained virologic response (SVR) with shorter treatment durations compared to prior therapies.^{1, 2} Along with advances in therapy, there has been a focus on the public health impact of CHC. The Centers for Disease Control and Prevention, the Institute of Medicine, and the United States Preventative Services Task Force, have prioritized hepatitis C awareness, screening and diagnosis.^3-5 Treatment is also being advocated as a means to prevent hepatitis C virus (HCV) infection. As a result of these processes, the pool of potential treatment candidates is expected to balloon. This has caused the conundrum in HCV treatment to shift from “Can we improve the efficacy and tolerability of HCV treatment?” to “Can we afford to treat all patients with CHC?”

At the core of the dilemma is the high cost of these new drugs. The estimated wholesale price of a 12-week course of sofosbuvir in the United States (US) is $84,000 and of simeprevir $66,000.^{6, 7} These staggering costs exclude retail markup, and associated cost of pegylated interferon (IFN), ribavirin, physician visits, and laboratory tests. While these new treatment regimens have SVR rates of 80-90%, and SVR has been shown to decrease cirrhosis complications, hepatocellular carcinoma (HCC) and liver-related mortality, even resource-replete countries like the US cannot afford to treat all those who are infected.^{1, 2, 8} The logistical and financial barriers are much higher in resource-limited countries, many of which have higher prevalence of HCV infection than western countries. Clinicians and health policy makers will need to determine an optimal yet practical approach to provide these highly efficacious but extremely costly therapies to this burgeoning patient population.

One solution is to adopt a risk-stratified approach that targets therapy to those at the greatest risk of disease progression. There have been many studies investigating risk factors for disease progression in patients with CHC, but few have employed a longitudinal study design in generalizable patient populations using data that are routinely available in clinical practice. Results of the existing studies have also not been systematically summarized in a single document. Therefore, we performed a systematic review of the literature to (a) identify factors predictive of disease progression (fibrosis progression and clinical outcomes) in patients with CHC and (b) assess existing predictive models.

Methods

Data Sources and Search Strategy

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations in conducting this systematic review.⁹ With the assistance of a medical research librarian, we performed serial literature searches for English and non-English articles. MEDLINE (via PubMed), EMBASE, Web of Science and Scopus were searched using the following keywords: “cirrhosis” or “liver cirrhosis” or “fibrosis”, “hepatitis C” or “hepatitis C, chronic” or “chronic hepatitis C”, “disease progression” or “progression” or “decompensation”. Boolean operators and medical subject heading terms as well as other controlled vocabulary were used to enhance electronic searches. An example of specific search strategy details is shown in Supplement Table 1.

All human subject studies published in full-text or abstract were eligible for inclusion. The search was limited to publications from 2003-2014 as this 10-year period contained the most contemporary and relevant data with respect to treatment and current practice. Additional studies of interest were identified by hand searches of bibliographies and cited reference tracking and consultation with clinical experts on the topic. The initial search was performed in October 2013 and the search was last updated on June 2, 2014.

Study Eligibility and Selection Criteria

Two authors (M.A.K. and A.S.L.) sequentially determined study eligibility. Studies were initially screened by the first author; decisions about study inclusion were made independently by both authors (M.A.K and A.S.L). Differences in opinion regarding study inclusion were resolved through consensus. Studies were included if they: (1) included human studies with participants 18 years of age or older; (2) systematically evaluated predictors of fibrosis progression and/or clinical outcomes for patients with CHC; and (3) used a longitudinal cohort study design. We focused on studies of untreated patients but also included studies with a mix of treated and untreated patients provided that <20% of the study population achieved SVR and results were stratified by treatment outcomes. For studies evaluating predictors of fibrosis progression, we selected studies only when paired biopsy was used to assess progression.

We excluded studies that enrolled (1) patients co-infected with hepatitis B (HBV) or human immunodeficiency virus (HIV); (2) patients with additional causes of chronic liver disease; (3) patients with prior liver transplantation; and (4) specific groups of patients e.g. thalassemia patients only;.These patient populations were excluded because they likely have different rates and risk factors for disease progression compared to the general population of patients with CHC. In addition, studies that evaluated HCC as the only outcome of interest were excluded as we were interested in broad clinical outcomes for patients with CHC, and predictors of HCC development alone may not be the same as predictors of disease progression in CHC in general. Lastly, studies that focused on predictors that are not readily available clinically (e.g. genetic or other serum markers for which commercial assays are not available, and experimental imaging techniques) were excluded given that they would not be relevant to current clinical practice.

Definition of Variables and Outcomes

Patients with CHC were defined as those with detectable HCV ribonucleic acid (RNA). We were interested in two outcomes: histologic progression and clinical progression. The definition of histologic progression was an increase of ≥1 METAVIR (range 0-4) or Ishak (range 0-6) fibrosis stage on follow-up liver biopsy. The definition of clinical progression encompassed the progression from compensated to decompensated cirrhosis, and liver-related or overall mortality. The definition of compensated cirrhosis was based on histology when available (Ishak fibrosis score ≥5 or METAVIR 4) or on the combined results of other non-invasive testing including laboratory tests and imaging. Decompensated cirrhosis was defined by the presence of any of the following: ascites, spontaneous bacterial peritonitis (SBP), variceal bleeding, or hepatic encephalopathy (HE). The presence of HCC as defined by histology or American Association for Study of Liver Diseases radiologic criteria was variably included as a clinical outcome.¹⁰

Data Abstraction and Validity Assessment

Data from eligible studies were abstracted by two authors (M.A.K. and S.Y.) using a standardized template adapted from the Cochrane Collaboration.¹¹ For all studies, we recorded: study design, sample size, patient population characteristics, duration of follow-up, predictor variables studied, outcomes measured, criteria used to define these outcomes, and measures of association/predictiveness of risk for these outcomes. We accepted the outcome definitions as stated by each study without independently validating or reviewing their data. Study authors were directly contacted for additional, unpublished data.

Assessment of Risk of Bias and Study Quality

Two authors (M.A.K and S.Y.) independently assessed the risk of study bias and study quality. Since all the included studies were non-randomized cohort studies, the Newcastle-Ottawa scale was used to judge study quality as recommended by the Cochrane Collaboration.¹² This scale uses a star system to assess the quality of a study based on three domains: selection of the study population, comparability of the study groups, and method of outcomes assessment. For our review, given that no study had a comparison group, we excluded comparability components of the scale across all studies. Studies which received stars in every domain were assessed as being of high quality.

Data Synthesis and Analysis

Given the substantial variation in the design, methods and inclusion/exclusion criteria within our included studies, meta-analysis was not performed. Two authors (M.A.K. and S.Y.) qualitatively synthesized the results of the included studies, focusing on the risk factors evaluated and their independent predictiveness in terms of the outcomes measured and patient populations studied. Studies were categorized according to the outcome of interest: predictors of histologic progression, predictors of clinical outcomes, or studies investigating both clinical and histologic outcomes. All authors had access to the study data and had reviewed and approved the final manuscript.

Results

Studies Included in the Systematic Review

After removal of duplicate entries, 2257 unique articles were identified by our systematic literature search (Figure 1). On the basis of abstract review, 69 were selected for full-text review. Two study authors classified 29 articles as meeting the predefined criteria for analysis. In total, these 29 studies included 5817 unique patients from 20 separate patient cohorts. Sixteen of these studies investigated predictors of histologic progression, eight studies evaluated predictors of clinical outcomes, and the remaining five studies investigated both histologic and clinical outcomes.^13-41 Fourteen studies included treatment-naïve patients only, 5 included both treatment-naïve and treatment-experienced patients, 8 included treatment-experienced patients only, and 2 studies did not describe the treatment status of the patients. We contacted four authors to obtain additional unpublished data.

Figure 1. Flow diagram of studies included in the systematic review.

^aMany studies met multiple exclusion criteria. Each study was coded under a single criterion only.

^bIncludes animal models, pediatric populations, patients who had previously undergone liver transplant, patients with chronic liver disease other than HCV monoinfection, evaluation of only specific subsets of populations with CHC.

^cIncludes studies that were descriptive papers only, studies that did not specifically evaluate for predictors of histologic or clinical progression, and studies that evaluated predictors that are not readily clinically available.

^dIncludes studies that focused on risk factors for the development of HCC only, and studies where some patients achieved SVR and the results were not stratified based on response to treatment.

Characteristics of Studies on Histologic Progression

A total of 21 studies evaluated predictors of histologic progression. The studies included populations from Europe (n=10), Asia (n=2), and North (n=8) and South America (n=1). Only one study was prospective with the remaining 20 being retrospective analyses of previously collected data. The sample size for included studies varied (range 36 to 622 patients) with the majority having <200 patients (n=14). A number of studies had overlapping cohorts. Four studies were derived from the Hepatitis C Antiviral Long-term Treatment Against Cirrhosis (HALT-C) cohort, a US multi-center randomized controlled trial to evaluate the safety and efficacy of low dose pegylated IFN in CHC patients with advanced fibrosis who failed to respond to prior IFN therapy. Four other pairs of studies drew from the same cohort of patients.^{17, 21, 25, 29, 33, 35, 38, 41} These studies were included in the review despite overlapping cohorts given differences in predictors examined, outcomes evaluated and criteria for selection of subsets of patients analyzed within the overall larger cohort. The average duration of follow-up ranged from a median of 2.5 years to 10 years.

The studies had varied inclusion and exclusion criteria as detailed in Table 1. Among the non-HALT-C studies, 11 studies had explicit requirements for baseline Ishak/METAVIR fibrosis stage. Five studies required minimal or no fibrosis at baseline and the remaining 6 studies required lack of cirrhosis on initial biopsy. Only 14 studies described criteria used to determine adequacy of biopsy specimens. The majority of the studies had a single pathologist blinded to clinical data score the biopsies while the HALT-C study had a panel of pathologists review the biopsies and consensus staging was recorded. Exclusionary alcohol intake was described in 9 studies though the cutoff amounts and methods for ascertaining alcohol intake varied across the studies. The studies were predominately comprised of male patients in their late 30's to early 50's.

Table 1.

General Characteristics of Included Studies^*

Study and Country	Sample Size (n)	% Genotype 1	Age	% Male	Study Population
					Inclusion criteria/Patient characteristics	Exclusion criteria
Predictors of Histologic Progression
Baran 2014 Turkey	125	95	Mean 45	38	Ishak <4 on initial biopsy >9 portal tracts on liver biopsy Treatment naïve or non-SVR with prior treatment	HIV co-infected Other chronic liver disease HCC History of immunosuppressive therapy
Boccato 2006 Italy	106	62	Mean 41.6	56	METAVIR F0 or F1 on initial biopsy Biopsy length >15mm and ≥7 portal tracts Minimum 4 yr follow-up Treatment naïve
Castera 2003 France	96	62	Mean 41	61	No cirrhosis on initial biopsy Treatment naïve	HBV or HIV co-infected
Colletta 2005 Italy	40	30	Median 43.5	55	Ishak ≤2 on initial biopsy Serial ALT values < 1.2 times ULN Treatment naïve
Cross 2009 United Kingdom	112	58	Median 44	66	Biopsy length >10mm Treatment naïve	HBV or HIV co-infected Other chronic liver disease Prior liver transplant ETOH intake ≥ 80g/d (male), ≥ 60g/d (female)
Fabris 2012 Italy	93	52	Median 38	46	Ishak ≤1 on initial biopsy Persistently normal or near normal ALT Treatment naïve
Fartoux 2005 France	135	60	Mean 38.5	59	METAVIR ≤1 on initial biopsy Biopsy length >10mm Only one known risk factor for HCV infection Treatment naïve	HBV or HIV co-infected Other chronic liver disease Prothrombin time > 80% Platelets >150,000/mL Hyaluronic acid < 85ug/L
Ghany 2003 United States	123	70	Mean 41	63	Treatment naïve
Khouri 2003 Brazil	55	NR	Mean 38	58	Biopsy length >15mm Minimum of 1 year interval between biopsies 18-75 years old Treatment naïve	HBV or HIV co-infected Immunosuppressed patients Chronic renal failure Using “potentially hepatotoxic drugs”
Kurosaki 2008 Japan	97	88	Median 52	51	No cirrhosis on initial biopsy Treatment with IFN between biopsies without SVR	HBV or HIV co-infected Other chronic liver disease ETOH consumption >20g/d
Levine 2006 Ireland	167	100	Mean 53	0	Women infected from contaminated immunoglobulin Biopsy length >15mm and ≥5 portal tracts Treatment naïve
Mummadi 2010 United States	36	NR	Median 47	75	No cirrhosis on initial biopsy Minimum of 1 year interval between biopsies	HBV or HIV co-infected Other chronic liver disease Prior organ transplant ETOH intake >30g/d HCC
Perumalswami 2006 United States	136	76	Mean 44	58	>10 portal tracts on liver biopsy Treatment naïve	Decompensated cirrhosis HBV or HIV co-infected Other chronic liver disease ETOH ≥60g/d (male), ≥40g/d (female) Malignancy Steroid therapy
Ryder 2004 United Kingdom	214	34	Median 36	59	No cirrhosis on initial biopsy >5 portal tracts on liver biopsy Treatment naïve	HIV co-infected Coagulation disorder Hemodialysis
Tamaki 2013 Japan	314	NR	Mean 53.7	47	Minimum of 1.5 year interval between biopsies Biopsy length >15mm IFN between biopsies, without SVR	HBV or HIV co-infected ETOH ≥40g/d HCC NASH
Williams 2011 United Kingdom	282	44	Mean 37	61	Ishak 0 or 1 on initial biopsy >5 portal tracts on liver biopsy Minimum of 2 year interval between biopsies No treatment during study	HIV co-infected Coagulation disorder Hemodialysis
Predictors of Clinical Outcomes
Bruno 2009 Italy	324	63	Median 59	51.1	Compensated cirrhosis (Child A) ≤ 70 years old IFN based treatment (55%) without SVR	HBV or HIV co-infected Other chronic liver disease HCC “unable to attend regular follow-up visits”
Ghany 2011 United States	470	94	Mean 49.8	71.3	HALT-C cohort: Ishak ≥3 on initial biopsy Prior treatment with IFN based therapy without SVR Evaluated control patients without further treatment	HIV co-infected Other chronic liver disease ETOH abuse within past year CTP score ≥7 History of hepatic decompensation Platelets <75,000 Neutrophil count <1500 Hematocrit <33% HCC or AFP>300 ng/ml Bilirubin >2.5 mg/dl Creatinine >1.5 mg/dl “Serious medical disorder” Use of illicit drugs within past 2 years
Giannini 2003 Italy	63	NR	Mean 52	73		HBV or HIV co-infected Other chronic liver disease ETOH >40g/d
Rincon 2013 Spain	145	NR	Median 51	77	Compensated cirrhosis Treatment naïve or non-SVR	Other chronic liver disease Prior liver transplant HCC >3cm or multilobular or vascular invasion
Sinn 2008 South Korea	647	71	Mean 58.2	49	Compensated cirrhosis Minimum of 1 year follow-up Treatment naïve	HBV or HIV co-infected CTP score >5 HCC
Sinn 2013 South Korea	232	62	Mean 57.2	38	Compensated cirrhosis Minimum of 1 year follow-up ALT< 40 IU/l at baseline Treatment naïve	HBV or HIV co-infected CTP score >5 HCC
VanDerMeer 2012 Europe and Canada	405	76	Median 48	68	Ishak ≥4 at baseline Prior treatment with IFN based therapy without SVR	HBV or HIV co-infected
Vergniol 2011 France	1457	58	Mean 51.2	53.4	52% patients with prior treatment; 38% without SVR 14% SVR with results adjusted for treatment response	HBV co-infected Other chronic liver disease
Predictors of Histologic Progression and Clinical Outcomes
Dienstag 2011 United States	1050 clinical 622 histologic	94	Mean 51	71	HALT-C cohort (See Ghany 2011 above) 517 patients in IFN arm and 533 control arm	HALT-C cohort (See Ghany 2011 above)
Everhart 2009 United States	985 clinical 557 histologic	94	Mean 50.2	71	HALT-C cohort (See Ghany 2011 above) 488 patients from IFN arm and 497 control arm	HALT-C cohort (See Ghany 2011 above)
Fontana 2010 United States	462 clinical 209 histologic	94	Mean 49.5	70.3	HALT-C cohort (See Ghany 2011 above) 49.4% patients in IFN arm	HALT-C cohort (See Ghany 2011 above)
Ghany 2010 United States	1050 clinical 547 histologic	94	Mean 50	71	HALT-C cohort (See Ghany 2011 above) 517 in IFN arm and 533 in control arm	HALT-C cohort (See Ghany 2011 above)
Livingston 2010 United States	52	67	Median 41	51	Alaska Native and American Indian persons Ishak ≤4 on initial biopsy Treatment naïve	HBV or HIV co-infected

AFP= alpha-fetoprotein; ALT= alanine aminotransferase; CTP= Child-Turcotte-Pugh; ETOH= alcohol; HALT-C= hepatitis c antiviral long-term treatment against cirrhosis; HBV=hepatitis B virus; HCC= hepatocellular carcinoma; HCV= hepatitis C virus; HIV= human immunodeficiency virus; IFN= interferon; NASH= non-alcoholic steatohepatitis; PC= analysis of prospectively designed cohort study; RC= retrospective analysis of cohort; SVR= sustained virologic response; ULN= upper limit of normal;

^*For select studies, reported data here reflects only a subset of the total study population based on the patient population and outcome of interest for this systematic review.

^aAll studies of histologic progression required patients to undergo at least 2 liver biopsies as the method to evaluate fibrosis progression.

Characteristics of Studies of Clinical Outcomes

A total of 13 studies evaluated predictors of clinical outcomes. Six studies were conducted in the US (including 5 HALT-C studies), five in Europe and two in Asia. Only two studies were prospective with the remaining 11 being retrospective analyses. Sample size in each study varied from 52 to 1457 patients. Aside from the HALT-C studies, there was only one additional overlapping cohort.^{36, 37} The average duration of follow-up ranged from a median of 2.3 to a maximum of 14.4 years. Compared to studies on histologic progression, the on clinical outcomes consisted of patients who were older, had more advanced fibrosis at baseline, and were more likely to be treatment experienced.

Incidence of Histologic Progression

A summary of the specific outcomes evaluated and incidence of these outcomes in each study is displayed in Tables 2-4. For studies where the outcome was defined as ≥1 fibrosis stage increase on follow-up biopsy (n=13), the incidence of that outcome ranged from 21-61% over a range of follow-up of 2.5-10 years. ^{14, 16, 18, 21, 25, 28-33, 35, 41} Studies applying a stricter definition of fibrosis progression (≥2 stage increase on follow-up biopsy, n=3) had less variability in range of incidence of outcome, reporting 22-34% over a range of follow-up of 3.5-5.8 years.^{13, 23, 26} Studies with higher rates of fibrosis progression tended to have longer follow-up durations (>6 years), though there were several studies with follow-up of ≥6 years that had low rates of fibrosis progression. No identifiable differences in patient characteristics between studies with high vs. low incidence of fibrosis progression were noted.

Table 2.

Outcomes and Predictors Evaluated and Summary of Results: Histologic Progression

Study	Outcomes Evaluated	% with Outcome	Years Follow-up (SD; range)	Predictors Significant on Multivariate Analysis	OR (95% CI)
Only Patients with Minimal Fibrosis at Baseline
Boccato 2006	≥1 METAVIR stage increase	60	Mean 7.8 (1.51; 5-10)	ETOH intake (>40g/d) Baseline steatosis	NR
Colletta 2005	METAVIR ≥2 on follow-up biopsy	35	Median 6.5 (NR;2.25-5.5)	HCV RNA >8.0 ×106 copies/ml ETOH intake >20g/d	NR
Fabris 2012	≥1 Ishak stage increase	61	Median 10 (NR;5.1-10)	HCV RNA >400,000 IU/ml ETOH intake >30g/d IL28B T/*x chol ≤175 mg/dl Follow-up >8yr	4.3 (1.4-13) 100 (8-1300) 4.1(1.5-11) 4.9 (1.8-13)
Fartoux 2005	METAVIR 3 or 4 on follow-up biopsy	16	Mean 5.2 (2.3;1.5-13.1)	Baseline steatosis	4.8 (1.3-18.3)
Williams 2011	≥1 Ishak stage increase	42	Median 4.4 (NR;2-16)	Age (older) Median ALT per 10 IU/L	1.34 (1.03-1.74) 1.07 (1.01-1.13)
Includes Patients with more Advanced Fibrosis at Baseline
Baran 2014	≥2 Ishak stage increase	22	Mean 5.8 (NR; 1.25-18)	Baseline GGT Follow-up ALT (<40 IU/L) Treatment experience (failed)	1.03 (1.01-1.5) 0.16 (0.03-0.93) 5.97 (1.81-19.7)
Castera 2003	≥1 METAVIR stage increase	31	Mean 4 (2.6; 0.8-14.6)	Worsening steatosis	4.7 (1.3-10.8)
Cross 2009	≥1 Ishak stage increase	21	Median 4.2 (NR; 2.8-6.1)	Baseline steatosis	14.3 (2.1-111.1)
Ghany 2003	≥1 Ishak stage increase	39	Mean 3.7 (NR; 0.25-17.6)	Baseline Ishak (low) Baseline HAI Baseline ALT (elevated)	NR
Khouri 2003	≥1 Ludwig stage increase	27	Mean 3.25 (1.1;1-6.8)	None	NR
Kurosaki 2008	≥1 METAVIR stage increase	23	Mean 5.9 (NR;1.2-11.6)	Baseline steatosis Average ALT ≥100 IU/l	5.14 (1.6-15.7) 5.21 (1.4-18.2)
Levine 2006	≥1 Ishak stage increase	27	Mean 5 (NR;NR)	Baseline Ishak Baseline ALT (elevated)	NR
Mummadi 2010	≥1 stage increase on 0-4 scale	53	Median 4 (NR; 2-9)	ΔAPRI ΔFIB-4	NR
Perumalswami 2006	≥1 Ishak stage increase	40	Mean 3.6 (NR; 0.5-17)	Age (older) Baseline ALT (elevated) Baseline Ishak (low) Baseline HAI (higher severity)	NR
Ryder 2004	≥1 Ishak stage increase	33	Median 2.5 (NR;1.9-9.4)	Age (older) Baseline Ishak (+fibrosis)	1.08a (1.03-1.11) 1.93a (1.3-9.0)
Tamaki 2013	Not defined	23	Mean 4.9 (2.9; NR)	ΔFIB-4 index/yr	3.7 (1.07-12.5)

ALT= alanine aminotransferase; APRI= aspartate aminotransferase to platelet ratio index; ETOH= alcohol; GGT= gamma-glutamyl transpeptidase; HAI= histologic activity index; HCV= hepatitis C virus; NR= not reported; RNA= ribonucleic acid;

^aRepresents adjusted relative risk (RR) instead of OR.

Table 4.

Outcomes and Predictors Evaluated and Summary of Results: Histologic Progression and Clinical Outcomes

Study	Outcomes Evaluated	% with Outcome	Years Follow-up (SD; range)	Predictors Significant on Multivariate Analysis	HR (95% CI)
Dienstaga 2011	1.Progression to cirrhosis	1.29%	Median 6 (NR;0.8-7)	Not performed
	2.Any clinical outcome: a) hepatic decompensation: ascites/variceal bleeding/ HE/SBP b)transplant c)HCC d)≥7 CTP score e)hepatic mortality f)overall mortality	2.31%
Everhart 2009	Combined outcome: ≥2 Increase in Ishak, hepatic mortality or hepatic decompensation (≥7 CTP score, ascites, variceal bleed, HE)	28%	Mean 3.5 (NR;NR)	Baseline Ishak (cirrhosis) HOMA2-IR (quartiles) Baseline steatosis (if cirrhosis) Mallory bodies	1.92 (1.12-3.28) 1.25 (1.08-1.45) 0.49 (0.35-0.70) 1.59 (1.10-2.31)
Fontana 2010	1.≥2 Increase in Ishak	1.34	Mean 4.25 (NR; NR)	1. Histologic progression Baseline platelet (per 50K, low) Baseline log HA	0.72 (0.57-0.91) 2.42 (0.27-4.47)
	2. Clinical Outcomes: a)Hepatic decompensation: ascites/variceal bleeding/HE/SBP b)HCC c)≥7 CTP score or d) overall mortality	2.15		2.Any clinical outcome Baseline bilirubin (elevated) Baseline INR (>1.0) Baseline albumin (low) Baseline logYKL-40	2.42 (1.42-4.13) 2.25 (1.30-3.89) 0.20 (0.10-0.38) 2.44 (1.28-4.63)
Ghany 2010	1.≥2 Increase in Ishak	1.28	Mean 3.5 (NR;NR)	1. Histologic progression Baseline BMI (high) Baseline platelets (low) Baseline steatosis	N R
	2.Clinical Outcomes: a)Hepatic decompensation: ascites/variceal bleeding/HE/SBP b) >7 CTP score or c)hepatic mortality	2.13		2.Any clinical outcome Baseline Log AST/ALT (high) Baseline bilirubin (elevated) Baseline albumin (low) Baseline platelets/ 50K (low)	3.34 (1.84-6.06) 1.82 (1.37-2.42) 0.20 (0.13-0.32) 0.59 (0.49-0.72)
Livingston 2010	1.≥1 Increase in Ishak 2.Hepatic Decompensation: ascites/esophageal varices/ HE/ coagulopathy	1.60 2.17	Mean 6.2 (NR;2.3-13.3)	Not performed

ALT= alanine aminotransferase; AST= aspartate aminotransferase; BMI= body mass index; CTP= Child-Turcotte-Pugh; CI= confidence interval; HA= hyaluronic acid; HE= hepatic encephalopathy; HCC= hepatocellular carcinoma; HCV= hepatitis C virus; HR= hazard ratio; INR= international normalized ratio; NR= not reported; RNA= ribonucleic acid; SBP= spontaneous bacterial peritonitis;

^aUnivariate significance not reported.

Incidence of Clinical Progression

Studies assessing risk factors for clinical progression (n=13) included several distinct outcomes. Four studies evaluating progression from compensated to decompensated cirrhosis reported an incidence between 13-40% over a range of follow-up of 2.3-14.4 years. ^{15, 24, 31, 34} No clear pattern was identified between length of follow-up or patient characteristics and rate of outcomes. Notably, the definition of decompensation varied across studies. Four studies evaluating the incidence of overall mortality reported incidences between 8-47%. The range of follow-up for these studies was 3.9-14.4 years, with a higher rate of outcomes reported in studies with longer duration of follow-up. ^{15, 27, 39, 40} The remaining studies used an aggregate outcome encompassing a broad range of clinical end points including decompensation, increase in Child-Turcotte-Pugh score, development of HCC, liver transplant, and liver related as well as overall mortality. The reported incidence of this aggregate outcome was 13-31% over a range of follow-up of 3.5-6.3 years. ^{19, 20, 23, 26, 36, 37}

Predictors of Histologic Progression

A detailed list of the predictors evaluated and the results of univariate analysis is provided in Supplement Tables 3, 4 and 5. For each study, the predictor variables were categorized as follows: 1) baseline clinical characteristics including demographics and relevant co-morbidities; 2) baseline laboratory results; 3) baseline histologic features; or 4) longitudinal laboratory and histology results.

All studies investigating predictors of histologic progression evaluated baseline clinical characteristics, baseline laboratory results and baseline histology results except for Tamaki et al who did not evaluate baseline histologic features.³⁸ Only half of the studies evaluated longitudinal variables which were predominantly serial aminotransferase levels. Longitudinal biopsy results such as changes in steatosis score or histologic activity index (HAI) were assessed in only five studies.^{16, 22, 28-30} The predictors that were most consistently evaluated are listed in Figure 2A. The most common clinical characteristics assessed were age, gender, HCV genotype, alcohol intake, body mass index (BMI) and biopsy interval, and the most common laboratory values evaluated were platelet count and ALT levels. Baseline histologic features were also frequently investigated predictors and were included in >70% of studies.

Figure 2.

List of variables identified to have significant predictive value for (A) histologic and (B) clinical progression

Multivariable analysis was performed in all but two studies.^{19, 31} Variables found to be independently predictive of histologic progression are listed in Tables 2 and 4. Among all the variables assessed, baseline steatosis was most consistently reported as independently predictive of subsequent fibrosis progression (significant on multivariate analysis in 6 of 21 studies) with an odds ratio (OR) [(95% confidence interval (CI)] of 4.8 (1.3-18.3) to 14.3 (2.1-111.1).^{12, 16, 18, 20, 24, 27} Notably, one study found that effect of baseline steatosis on fibrosis progression was dependent on baseline fibrosis stage.²⁰ Baseline Ishak/METAVIR fibrosis stage was the next most consistently identified independent predictor of histologic progression (significant on multivariable analyses in 5 of 21 studies).^{20, 25, 30, 33, 35} Only one of these studies reported the effect size, with adjusted relative risk of 1.93 (95% CI 1.3-9.0).³⁵ Figure 2A depicts the number of studies in which individual variables were significantly or not significantly predictive of histologic progression on multivariate analyses.

Predictors of Clinical Outcomes

All 13 studies examining predictors of clinical outcomes included baseline clinical characteristics and laboratory results (Supplement Tables 4 and 5). Baseline histology was assessed in only 8 studies though biopsies were performed in every study. Only 3 studies incorporated longitudinal data which consisted of serial laboratory values only.^{23, 24, 36} The predictors that were most consistently evaluated are listed in Figure 2B. The most common clinical characteristics assessed were age, gender, and BMI; the most common laboratory values evaluated were platelet count and ALT level.

Multivariable analysis was performed in all but two studies.^{19, 31} The variables found to be independently predictive of clinical progression are listed in Tables 3 and 4. Among the variables assessed, baseline platelet count was the most consistent independent predictor of clinical outcomes (significant on multivariate analysis in 6 of 13 studies) followed by age, baseline AST/ALT ratio, albumin and bilirubin (each significant in 4 studies).^{15, 24, 26, 36, 37, 39} Figure 2B depicts the number of studies in which individual variables were significantly or not significantly predictive of clinical outcomes in multivariate analyses.

Table 3.

Outcomes and Predictors Evaluated and Summary of Results: Clinical Outcomes

Study	Outcomes Evaluated	% with Outcome	Years Follow-up (SD; range)	Predictors Significant on Multivariate Analysis	HR (95% CI)
Cohorts with Patients with a Broader Range of Fibrosis
Ghany 2011	1.Decompensation: a)ascites b)variceal bleeding c)HE or d)SBP	1.13	Median 6.3 (NR; 1.4-8.7)	1.Decompensation Baseline Platelets ≤150 Baseline Bilirubin ≤0.8mg/dL Baseline AST/ALT ≤0.8 >15% decrease in platelets >15% increase in bilirubin >15% decrease in albumin	2.76(1.47-5.19) 0.37(0.18-0.75) 0.50(0.27-0.92) 2.29 (1.26-4.14) 2.62(1.37-5.00) 3.85(1.81-8.18)
	2. Hepatic mortality/ liver transplant	2.17		2.Hepatic Mortality/Transplant Baseline platelets ≤150 Baseline albumin ≤3.9 >15% increase in albumin 5-15% increase in AST/ALT	4.14 (2.29-7.47) 2.32 (1.33-4.06) 3.56 (1.82-6.97) 2.14 (1.16-3.96)
Giannini 2003	1 year overall mortality	25	≥ 1 (NR;NR)	Baseline AST/ALT >1.16 Baseline MELD >9 Baseline CTP score >7	NR
VanDerMeera 2012	Overall mortality	25	Median 8.1 (NR;NR)	Age (per year) Gender (male) Baseline Platelets per 10×109/L Log Baseline AST/ALT (per 0.1)	1.06 (1.03-1.09) 1.90 (1.10-3.29) 0.90 (0.86-0.95) 1.29 (1.11-1.50)
Vergniol 2011	Overall 5 year mortality	8	Median 3.9 (NR;NR)	Age (older) Treatment Liver stiffness FibroTest ActiTest	1.03 (1.01-1.04) 0.28 (0.19-0.42) 2.9 (2.0-4.3) 60 (14-255) 0.19 (0.07-0.53)
Cohorts Restricted to Patients with Cirrhosis
Bruno 2009	1.Decompensation: a)ascites b)variceal bleeding or c)HE	1.40	Median 14.4 (NR;0.9-19.5)	1.Decompensation HCV Genotype(1b vs. 2a/c) Esophageal varices Baseline Platelets <80 Baseline Bilirubin ≥1.2ml/dL AFP ≥10ng/ml HCC development	2.17 (1.31-3.59) 2.09 (1.33-3.30) 1.95 (1.08-3.51) 1.79 (1.16-2.76) 1.59 (1.09-2.32) 5.52 (3.77-8.09)
	2. Hepatic mortality	2.33		2.Hepatic Mortality Age (10 yr increase) Gender (Male) HCV Genotype(1b vs. 2a/c) Esophageal varices Creatinine (≥1.2mg/dl) MELD >10 Decompensation HCC development	1.61( 1.21-2.13) 1.87 (1.23-2.84) 2.37(1.33-4.22) 2.27 (1.41-3.66) 3.07 (1.65-5.73) 2.43 (1.57-3.76) 16.9 (9.97-28.6) 8.62 (5.57-13.3)
	3. Overall mortality	3.47		3. Overall Mortality Age (10 yr increase) Gender (male) HCV Genotype(1b vs. 2a/c) Esophageal varices MELD >10 AFP ≥1ng/ml Decompensation HCC development	1.63 (1.28-2.06) 1.88 (1.33-2.66) 1.83 (1.18-2.86) 2.19 (1.47-3.27) 2.15 (1.50-3.09) 1.62 (1.15-2.29) 7.08 (4.88-10.2) 3.80 (2.67-5.42)
Rincon 2013	Decompensation: a)ascites b)variceal bleeding or c)HE	29	Median 2.3 (NR; 0.2-9.2)	HVPG Baseline albumin	1.11 (1.05-1.17) 0.42 (0.22-0.82)
Sinn 2008	First occurrence of : a) ≥ 2 increase CTP score b)HCC c)SBP d)variceal bleed, e)HE or f)hepatic mortality	22	Median 4.6 (NR;1-12.6)	Age>55 Gender (Male) Diabetes Baseline Platelets <140 Baseline APRI>1	2.2 (1.4-3.6) 1.7(1.2-2.3) 1.8(1.3-2.7) 4.9(3.4-7.2) 5.4 (3.5-8.3)
Sinn 2013	Disease progression using same 2008 definition	14	Median 4.5 (NR;1-12.6)	Baseline ALT >26 (male) Baseline ALT> 23 (female) Baseline platelets (low, male) Baseline platelets (low, female)	5.35 (1.05-27.3) 4.40 (1.12-15.8) 0.98 (0.96-0.99) 0.97(0.96-0.98)

AFP= alpha-fetoprotein; ALT= alanine aminotransferase; APRI= aspartate aminotransferase to platelet ratio index; AST= aspartate aminotransferase; CTP= Child-Turcotte-Pugh; CI= confidence interval; HE= hepatic encephalopathy; HCC= hepatocellular carcinoma; HCV= hepatitis C virus; HVPG= hepatic vein pressure gradient; HR= hazard ratio; MELD= model for end-stage liver disease; NR= not reported; SBP= spontaneous bacterial peritonitis;

^aUnivariate significance not reported

Mathematical Prediction Models

Five studies provided prediction models, three for fibrosis progression and four for clinical outcomes (Supplement Table 6).^{23, 26, 32, 39, 40} Four of the models were derived from the HALT-C study. All the prediction models are primarily comprised of baseline laboratory results. Only one of the models incorporated longitudinal data. None of the models had been validated in external CHC cohorts and only two models reported the associated area under the receiver operating characteristic curve.^{23, 40}

Quality Assessment and Risk of Bias

Studies evaluating predictors of histologic progression were of varying quality, whereas studies investigating predictors of clinical outcomes or studies investigating combined outcomes were all of high quality except for one study.³¹ Six studies on histologic progression included a small number of patients with advanced fibrosis or cirrhosis on initial biopsy who were not able to progress according to the author's definition.^{17,18, 25, 28, 33, 38} Two studies evaluated select cohorts (Levine et al evaluated untreated Irish women who acquired HCV infection during pregnancy only, and Livingston et al evaluated only treatment naïve Alaska Native and American Indian persons) and were scored as having limited representativeness. ^{30, 31} The remaining studies were scored as being at least somewhat representative of the average patient with CHC in the community (Supplement Table 2).

Discussion

Although there is abundant literature on the topic of predictors of histologic and clinical outcomes for patients with CHC, only 29 studies met our inclusion criteria which captured studies with a longitudinal study design in broad patient populations. Within the 29 studies included, the incidence of outcomes varied widely: 16-61% during a median follow-up of 2.5-10 years for fibrosis progression; 13-40% over 2.3-14.4 years for hepatic decompensation; and 8-47% over 3.9-14.4 years follow-up for overall mortality. The wide range in incidence of outcomes highlights the heterogeneity in patient population evaluated, stage of liver disease at enrollment, duration of follow-up, and definition of outcomes. Interestingly, higher rates of outcomes did not clearly correlate with longer durations of follow-up or more advanced disease at baseline across studies, pointing to more complex underlying interactions driving outcomes. Although the incidence data were not conducive to providing consensus outcome rates, we were able to identify risk factors that have most consistently been associated with outcomes of interest. Baseline steatosis and fibrosis score were the most consistent predictors of fibrosis progression and baseline platelet count, AST/ALT ratio, albumin, bilirubin and patient age were the most consistent predictors of clinical outcomes.

The variables identified as being most predictive of outcomes were not unexpectedly markers of more advanced liver disease. Though the overall finding that patients with more advanced disease are at higher risk for adverse outcomes is not novel, our study is the first to systematically identify the specific risk factors from among the many markers of advanced liver disease that portends worse prognosis. For example, among the laboratory markers of more advanced liver disease, platelet count, bilirubin, albumin and AST/ALT ratio conveyed meaningful risk information whereas INR, AST, ALT and MELD score did not. Differences in study design made it difficult to identify clear cut-off values for each predictor aside from platelet count with values ≤150,000/uL consistently associated with worse prognosis. Furthermore, individual laboratory markers may be less reliable in predicting outcomes than panels of markers such as aspartate aminotransferase to platelet ratio index (APRI), FIB-4, Fibrotest and/or measurements of liver stiffness. The finding that patients with more advanced disease have greater risk of disease progression suggests there may be subsets of patients who are rapid progressors.Understanding whether some patients are destined to be rapid progressors and being able to identify these patients at an early stage will help target limited resources to treat those patients who will derive the most benefit. Though none of the existing predictive models have been externally validated, the model developed by Ghany and colleagues is most readily applicable in clinical practice as it is based on routinely available data and evaluates important liver-related clinical outcomes.²⁶

Examining the results in more detail yielded several useful insights. First, the finding of steatosis as a predictor of outcomes highlights a potential modifiable risk factor associated with disease progression. This is particularly relevant given the evolving obesity epidemic. Our data suggests that patients may benefit from aggressive lifestyle interventions in addition to other standard of care treatment for patients with CHC. The prognostic information gained from baseline liver biopsy results suggests that liver biopsies not only provide information regarding current staging of liver disease but also useful prognostic information. As performance of liver biopsies continue to decline, evaluating whether non-invasive assessment of fibrosis and steatosis will provide the same prognostic information would be important. Though only one study included in our review used an additional modality to assess liver fibrosis in conjunction with biopsy, this study showed that liver stiffness measurements were associated with overall mortality.⁴⁰

Our review also highlights several areas for improvement for future studies on predictors of disease progression in CHC. Analysis of the individual predictive value of each risk factor found that there was a notable lack of incorporation of longitudinal variables. In the few studies that did assess longitudinal data, these variables were usually restricted to laboratory values, predominantly AST and ALT levels. These models do not mirror clinical practice where assessments of risk of disease progression are based on the pattern of a patient's test results over time. Models restricted to only baseline data also cannot distinguish between patients with similar initial data but who go on to have distinct disease courses and outcomes. Future studies can also benefit from implementing standardized definitions and criteria for outcomes and employing a panel of investigators to adjudicate outcomes as the variability in definition of predictor and outcome variables was one of the biggest challenges.

There are other limitations to our review such as sample selection bias, sampling error, and misclassification bias in studies requiring paired biopsies. In the majority of studies biopsies were assessed by a single pathologist criteria for adequacy of biopsies was described in only 14 of 21 studies. Finally, the variability in duration of follow-up impacts not only incidence rates of outcomes, but also predictiveness of variables examined.

In summary, this systematic review demonstrated that while there is an abundance of literature on factors associated with histologic and/or clinical progression in CHC, there is a lack of longitudinal studies of representative, untreated, well characterized patients followed for a sufficiently long duration to allow the development of simple prediction models. Despite the limitations inherent to the existing literature, we were able to identify specific risk factors that have been consistently identified as being independently predictive of disease progression. By selecting studies consisting of broad patient populations and those that evaluated routinely obtained clinical data, our findings can be generalized to and applied in many clinical settings. From a policy standpoint, we have highlighted that it is possible to identify patients at higher risk for adverse outcomes. Policies that target costly new HCV therapies to these patients who would derive the most benefit will maximize their cost effectiveness. The availability of risk prediction tools that can be applied in the clinic will help physicians and patients decide whether to embark on HCV treatment now or to wait for more affordable treatment. These types of tools will be particularly important in resource-limited countries and must therefore be validated in broad patient populations.

Abbreviations

ALT

alanine aminotransferase

APRI

aspartate aminotransferase to platelet ratio index

ARR

adjusted relative risk

AST

aspartate aminotransferase

BMI

body mass index

CHC

Chronic hepatitis C

CI

confidence interval

DAA

direct-acting antiviral agents

HAI

histologic activity index

HALT-C

Hepatitis C Antiviral Long-term Treatment Against Cirrhosis

HBV

hepatitis B virus

HCC

hepatocellular carcinoma

HCV

hepatitis C virus

HE

hepatic encephalopathy

HIV

human immunodeficiency virus

HR

hazard ratio

IFN

interferon

OR

odds ratio

RNA

ribonucleic acid

SBP

spontaneous bacterial peritonitis

SVR

sustained virologic response