Impact of Treatment against Hepatitis C Virus on Overall Survival of Naive Patients with Advanced Liver Disease

Abstract

The beneficial effect of achieving a sustained virological response (SVR) after antiviral treatment against hepatitis C virus is well established. However, it remains unclear whether unsuccessful treatment (non-SVR) also improves patient survival, especially in patients with advanced liver fibrosis. We retrospectively evaluated the incidence of death or liver transplantation in the 427 naive patients with a Child-Pugh score of A and advanced fibrosis newly admitted to the Hospital Beaujon between 2000 and 2010. Patients were followed for a median time of 5.5 years. The baseline characteristics of untreated (n = 102) and treated (n = 325) patients were largely similar, and there was no evidence of a bias of indication. Treated patients received a combination of interferon and ribavirin and had an SVR rate of 32%. The incidence of death or liver transplantation per 100 person-years was 1.00, 3.20, and 5.44 in SVR, non-SVR, and untreated patients, respectively. After adjusting for baseline characteristics, the risk of death or liver transplantation was significantly lower in SVR than in non-SVR patients and in non-SVR than in untreated patients (hazard ratios, 0.35 and 0.51, respectively; P = 0.019 and 0.038, respectively). The effect of treatment in non-SVR patients was higher in patients who had a virological or a biochemical response than in those who did not have a virological or a biochemical response. The risk of death or liver transplantation was significantly lower in treated than in untreated patients. Moreover, there was a gradient of mortality between patients with SVRs, virological or biochemical responders, and untreated patients, suggesting that treatment, even in the absence of viral eradication, has a beneficial effect on survival.

INTRODUCTION

Chronic infection with hepatitis C virus (HCV) affects approximately 160 million people worldwide and is the leading cause of cirrhosis, hepatocellular carcinoma (HCC), and liver transplantation (1–3). The goal of treatment is to achieve a sustained virological response (SVR), defined by the absence HCV RNA 6 months after the end of treatment. Although viral eradication is thought to delay or even reverse the course of the disease (4), evaluation of the SVR effect is mostly based on retrospective cohorts, whose results are often hampered by a large rate of loss to follow-up and population or treatment heterogeneity. In particular, most studies which showed a beneficial effect of an SVR on clinical outcome used heterogeneous populations (5–9) and did not stratify the subjects on the basis of the stage of liver fibrosis, which remains a major determinant of survival in HCV-infected patients (10–13). Recently, two studies showing that an SVR is associated with a lower incidence of HCC and all-cause mortality in patients with advanced liver diseases have been published (14, 15). However, the number of patients was limited, and more data are still needed to confirm these findings.

Until recently, the reference therapy was pegylated interferon (peg-IFN) and ribavirin (RBV). In spite of the emergence of direct-acting antivirals, which are rapidly changing the landscape of HCV treatment, a large proportion of patients with advanced liver disease still do not achieve an SVR. For those non-SVR patients, it remains unclear whether treatment has any benefit on overall survival. In a population of 345 Japanese cirrhotic patients, only an SVR and not a non-SVR was found to significantly reduce the overall risk of death (16). Alternatively, in a recent study, cirrhotic non-SVR patients had improved survival compared to untreated patients. However, the number of untreated patients was small (n = 48 patients), and 35% of both untreated and treated patients had received a treatment before inclusion in the study, making it difficult to evaluate the intrinsic effect of treatment on survival (15). Overall the amount of data looking specifically at the effect of treatment in non-SVR patients remains very limited.

Here we used the HepAList database, a database with data for a large repertoire of patients admitted for care at Beaujon Hospital in the Paris area of France, to evaluate the benefit of treatment with or without an SVR on overall survival.

MATERIALS AND METHODS

Patients.

HepAList is a prospective cohort initiated in 1992 that records all patients admitted for care in the Hepatology Department of Beaujon Hospital (Paris area, France).

We included in our analysis all patients enrolled from 2000 to 2010 meeting the following criteria: they had a Metavir score of F3 (precirrhosis) or F4 (cirrhosis), as assessed by liver biopsy in the first 6 months following enrollment; they were HCV treatment naive; and they were positive for HCV RNA (as determined by a quantitative PCR assay and the qualitative Bayer Versant assay). Exclusion criteria included the presence of HCC at the baseline, the absence of a liver biopsy, a history of a Child-Pugh score of B or C, or decompensation. Moreover, we also defined as exclusion criteria the presence of HIV infection and any of the following active hepatitis diseases: hepatitis B virus infection (defined by positivity for HBsAg), autoimmune hepatitis, Wilson's disease, hemochromatosis, Budd-Chiari syndrome, primary biliary cirrhosis, and primary sclerosing cholangitis.

The cohort study was conducted in accordance with the ethical guidelines of the Declaration of Helsinki revised in 2013 and actual French legislation, and the protocol was approved by the Hospital Beaujon institutional review board.

Data collected at baseline.

Data on a large number of baseline characteristics were collected upon enrollment. This included sociodemographic information and clinical and blood parameters (Table 1). Patients were considered to have a history of alcohol intake if their daily intake was ≥30 g/day (males) or ≥20 g/day (females).

TABLE 1.

Baseline characteristics of 427 patients with advanced fibrosis according to antiviral therapy and response to therapy^a

Characteristic	All patients (n = 427)	SVR patients (n = 104)	Non-SVR patients (n = 221)	Untreated patients (n = 102)	P value
					SVR vs non-SVR patients	Non-SVR vs untreated patients	SVR and non-SVR vs untreated patients
No. (%) men	286 (67)	74 (71)	146 (66)	66 (65)	0.360	0.811	0.575
Median (IQR) age (yr)	50 (43–58)	50 (43–57)	49 (43–57)	51 (45–63)	0.790	0.020	0.016
Median (IQR) BMI	25 (23–28)	24 (23–28)	25 (23–27)	24 (22–27)	0.082	0.320	0.117
No. (%) patients with:
Alcohol intake	94 (23)	21 (20)	51 (24)	22 (23)	0.677	0.375	0.883
Diabetes	69 (16)	17 (16)	40 (18)	12 (12)	0.698	0.149	0.166
Arterial hypertension	85 (20)	20 (20)	47 (22)	18 (20)	0.672	0.450	0.512
Median (IQR) AST/ALT ratio	0.88 (0.69–1.15)	0.80 (0.62–1.00)	0.88 (0.70–1.13)	0.96 (0.77–1.22)	0.008	0.548	0.005
No. (%) of patients positive for anti-HBc	118 (28)	32 (31)	58 (26)	28 (27)	0.395	0.819	0.962
Median (IQR):
Albumin concn (g/liter)	42 (39–46)	43 (39–45)	42 (38–46)	42 (40–46)	0.224	0.297	0.490
Alpha-fetoprotein level (ng)	7 (4–12)	6 (4–9)	8 (5–12)	8 (4–12)	0.014	0.984	0.468
Bilirubin level (μmol/liter)	15 (11–19)	16 (12–18)	15 (11–18)	16 (11–19)	0.555	0.183	0.224
Creatinine concn (μmol/liter)	74 (68–79)	75 (70–82)	74 (68–80)	74 (65–75)	0.248	0.094	0.030
Platelet count (109/liter)	151 (131–183)	154 (134–187)	150 (128–181)	151 (112–176)	0.177	0.530	0.865
Prothrombin proportion (%)	83 (77–92)	84 (76–92)	83 (74–91)	83 (77–92)	0.686	0.174	0.193
HCV RNA load (log10 IU/ml)	6 (5.7–6.3)	5.9 (5.6–6.2)	6 (5.7–6.2)	6 (5.7–6.3)	0.424	0.862	0.496
No. (%) of patients infected with:					0.001	0.527	0.623
G1	238 (56)	43 (41)	136 (61)	59 (58)
A non-G1 genotype	189 (44)	61 (59)	85 (39)	43 (42)
No. (%) of infected patients with Metavir score of:					0.053	0.106	0.292
F3	215 (50)	59 (57)	100 (45)	56 (55)
F4	212 (50)	45 (43)	121 (55)	46 (45)

^aAbbreviations: ALT, alanine aminotransferase; anti-HBc, anti-hepatitis B core antigen; AST, aspartate aminotransferase; BMI, body mass index, calculated as weight (in kilograms) divided by height (in meters) squared; IQR, interquartile range; SVR, sustained virological response; non-SVR, nonsustained virological response. See Materials and Methods for the definition of the Metavir score. Baseline characteristics were compared between patients (SVR versus non-SVR patients, non-SVR versus untreated patients, and SVR and non-SVR versus untreated patients) using the Mann-Whitney test for continuous variables and the χ² test or Fisher exact test for categorical variables.

Anti-HCV treatment.

Anti-HCV therapy was interferon (standard, pegylated, alfa-2a, or alfa-2b interferon) in combination with ribavirin. Reasons for nontreatment were recorded and classified into medical contraindication, an anticipated lack of compliance, or refusal by the patient. Further, patients receiving treatment for less than half of the treatment duration according to the standard of care (i.e., less than 24 and 12 weeks for patients infected with HCV genotype 1 [G1] and non-G1 strains, respectively) were also considered untreated.

Patients with an SVR were defined as those with undetectable HCV RNA in serum 24 weeks after the end of treatment; otherwise, the patient was considered to have a non-SVR. Non-SVR patients were further categorized according to the virological and the biochemical response. Patients with undetectable HCV RNA at the end of treatment were considered transient virological responders (TVRs), and those with detectable HCV RNA were considered non-virological responders (NVRs). Patients with normal alanine aminotransferase (ALT) levels at the end of treatment and 6 months after treatment were considered sustained biochemical responders (SBRs), while patients not meeting this criterion were considered non-SBRs.

Study follow-up and endpoint.

The patient's date of enrollment was the first visit at the hospital, and the follow-up ended at the latest for all patients on 31 December 2011.

The endpoint was the time to death (liver or non-liver related) or liver transplantation. We considered, without a loss of generality (see Results), that the date of enrollment corresponded to the initiation of treatment, and this was considered the time of origination (time zero) in our analysis.

The living status of the patient at the end of the study period and the date of death, if relevant, were obtained by contacting the patient him- or herself or the patient's primary care physician. Because French law makes it mandatory for the local administration to send a copy of the death certificate of every French-born citizen to his or her municipality of birth, a request was sent to the vital statistics department of the municipality of each patient for whom information on the vital status of the patient was missing. Causes of death were classified as liver related (defined by hepatic failure, end-stage liver disease, or complications of portal hypertension, HCC, and liver transplantation), not liver related (other causes of death), or unknown. Information on liver transplantation was obtained by contacting the French biomedicine agency, which manages the national list of patients awaiting liver transplant.

Statistical analysis.

Results are reported as medians and interquartile ranges (IQRs) for continuous variables and as percentages for categorical variables. Differences were assessed using the Mann-Whitney U test for continuous variables and the chi-square or Fisher exact test for categorical variables.

The incidence of the event was stratified on treatment outcome (classified as an SVR, a non-SVR, and untreated) and reported per 100 person-years. Survival analysis was performed using Kaplan-Meier estimates, and differences were assessed using log-rank tests. In a second step, the non-SVR group was stratified according to the occurrence of a virological or a biochemical response.

The effects of the following baseline characteristics on survival were assessed using Cox univariate models: gender, alcohol intake, diabetes, hypertension, anti-HBc antigen, infection with HCV genotype 1, and Metavir score (categorical variables) and age, body mass index (BMI), the aspartate aminotransferase (AST)/ALT ratio, albumin concentration, alpha-fetoprotein concentration, serum bilirubin level, creatinine concentration, prothrombin proportion, platelet count, and HCV RNA load (continuous variables). Finally, the effect of treatment outcome on survival was analyzed using a multivariate Cox model that incorporated all covariates with P values of <0.05 in the univariate analysis. As suggested by Yoshida et al. (7), age and gender were incorporated in the multivariate model regardless of their P value in the univariate analysis because these covariates are strongly associated with life expectancy. Missing data were imputed at the mean of the distribution.

As a complement, a propensity score (PS) analysis was performed (17), in order to estimate the effect of treatment adjusted on the basis of the probability that the patient would be treated (18, 19) (see the Method section in the supplemental material) and therefore correct for a possible bias of indication (see Table S1 in the supplemental material).

All statistical tests were two-sided, and a P value of <0.05 was considered statistically significant. SAS software (version 9.3; SAS Institute, Cary, NC) was used for all statistical analyses.

RESULTS

Population characteristics.

From 2000 to 2010, 685 patients with advanced fibrosis were enrolled in the HepAList cohort. Among them, 258 were diagnosed at the baseline with HCC (n = 119), had a Child-Pugh score of B or C or a liver-related disease (n = 39), had HIV coinfection (n = 16), or had not had a liver biopsy (n = 84) (Fig. 1). Overall, 427 patients met the inclusion criteria and were included; the median age was 50 years (IQR, 43 to 58 years), 286 patients (67%) were men, 94 (28%) had a history of alcohol intake, 238 (56%) were HCV genotype 1 positive, and 212 (50%) had a Metavir score of F4. The overall median time between enrollment and treatment initiation was 5 months (IQR, 3 to 9 months) in both SVR and non-SVR patients (Table 1).

FIG 1.

Flow chart.

Further, 325 patients (76%) received at least one course of treatment and 102 (24%) patients were classified as untreated. Among the patients in the latter group, 43 patients had medical contraindication (psychiatric disorders, significant coronary heart disease, untreated thyroid diseases), 5 patients had anticipated compliance issues, 19 patients refused treatment, and 30 patients did not receive treatment for more than half of the treatment duration; in 5 patients the cause of nontreatment was not known.

Importantly, information on liver transplantation, death, and the date of death, if relevant, was obtained for all patients, and there were no missing data related to survival.

Response to anti-HCV treatment.

Among the 325 treated patients, 241 (74%) did not achieve an SVR during their first course of treatment; of these, 43 (18%) subsequently received at least one additional course of treatment and 21 (9%) eventually achieved an SVR. Finally, 104 patients were SVRs (after the first or the second course of treatment), 221 patients were non-SVRs, and 102 patients remained untreated. Among the non-SVR patients, 61 (28%) were transient virological responders and 160 (72%) were non-virological responders. Further, 62 (28%) were sustained biochemical responders and 159 (72%) were non-sustained biochemical responders.

The proportion of patients infected with G1 was significantly lower in the SVR than in the non-SVR group (41% versus 61%, respectively; P = 0.001). Although the difference was not statistically significant, the proportion of cirrhotic patients (Metavir score, F4) was also lower in the SVR than in the non-SVR group (43% versus 55%, respectively; P = 0.054). Regarding the untreated group, the proportions of cirrhotic or G1-infected patients were comparable to those observed in the non-SVR group (45% versus 55% and 58% versus 61%, respectively; P = 0.11 and 0.53, respectively). Besides cirrhosis and HCV genotype, baseline characteristics and comorbidities were largely similar among the SVR, non-SVR, and untreated patients (Table 1).

Survival analysis according to treatment outcome.

The overall median follow-up period was equal to 5.5 years (IQR, 3 to 8 years); and the median follow-up period was equal to 5.6 years (IQR, 2.8 to 8.2 years), 6 years (IQR, 3.4 to 8.4 years), and 3 years (IQR, 2 to 5.2 years) in SVR, non-SVR, and untreated patients, respectively. Forty-nine of the 427 enrolled patients died during the follow-up period: 3 in the SVR group, 27 in the non-SVR group, and 19 in the untreated group. Moreover, 19 patients underwent liver transplantation: 3 in the SVR group, 15 in the non-SVR group, and 1 in the untreated group (Table 2). This led to a total of 68 events (death or liver transplantation): 6 in the SVR group, 42 in the non-SVR group, and 20 in the untreated group. The cause of death or the reason for liver transplantation was liver related in 3 (50%), 30 (71%), and 17 patients (85%) in the SVR, non-SVR, and untreated groups, respectively (Fig. 1).

TABLE 2.

Incidence rates of death and liver transplantation per 100 person-years for SVR, non-SVR, and untreated patients^a

Patient group and outcome	No. of patients	No. of events	No. of PY	Rate/100 person-yr	95% CI
SVR	104	6	600	1.00	0.44–2.22
Death		3		0.05	0.04–0.05
LT		3		0.05	0.04–0.05
Non-SVR	221	42	1,310	3.20	2.36–4.33
Death		27		2.06	1.98–2.13
LT		15		1.14	0.57–1.71
Untreated	102	20	367	5.44	3.51–8.44
Death		19		5.17	4.93–5.40
LT		1		0.27	0.22–0.32

^aAbbreviations: PY, person-year; LT, liver transplantation; CI, confidence interval.

The incidence of death or liver transplantation per 100 person-years was 1.00, 3.20, and 5.44 in SVR, non-SVR, and untreated patients, respectively (P = 0.005 and 0.001 for comparisons of SVR versus non-SVR patients and non-SVR versus untreated patients, respectively). Further, the cumulative Kaplan-Meier estimates showed that the estimation of the occurrence of the event was significantly different in the SVR, non-SVR, and untreated groups (P = 0.005 and 0.001 for comparison of SVR versus non-SVR patients and non-SVR versus untreated patients, respectively). In a second step, the survival analysis was done by including only non-SVR and untreated patients. We found that there was a gradient in survival along both the virological and the biochemical responses (Fig. 2B and C), and the rate of survival was significantly higher in transient virological responders (TVRs) and in sustained biochemical responders (SBRs) than in non-virological responders (NVRs) and non-sustained biochemical responders (NSBRs), respectively (P = 0.003 and P = 0.001, respectively).

FIG 2.

Survival analysis with death or liver transplantation as the endpoint. (A) Kaplan-Meier curves of time to death or liver transplantation in the 427 patients with advanced fibrosis. The log-rank test was used to compare survival in SVR versus non-SVR patients and non-SVR versus untreated patients. (B) Kaplan-Meier curves of time to death or liver transplantation in untreated and non-SVR patients according to the virological response. The log-rank test was used to compare survival in transient virological responders versus non-virological responders (NVR) and NVRs versus untreated patients. (C) Kaplan-Meier curves of time to death or liver transplantation in untreated and non-SVR patients according to the biochemical response. The log-rank test was used to compare survival in sustained biochemical responders (SBR) versus non-SBRs.

Determinants of death or liver transplantation.

Univariate analysis using Cox proportional hazards regression showed that the risk of death or liver transplantation was lower in SVR than in non-SVR patients (hazard ratio [HR] = 0.31, 95% confidence interval [CI] = 0.13 to 0.45; P = 0.008) and in non-SVR than in untreated patients (HR = 0.41, 95% CI = 0.24 to 0.71; P = 0.001) (Table 3).

TABLE 3.

Univariate and multivariate analysis of factors associated with all-cause mortality or liver transplantation in 427 patients with advanced fibrosis according to antiviral therapy and response^a

Factor	Univariate analysis		Multivariate analysis
	HR (95% CI)	P valueb	HR (95% CI)	P valuec
Age	1.03 (1.01–1.05)	0.004	1.00 (0.98–1.03)	0.545
Gender (M vs F)	1.06 (0.66–1.74)	0.802	1.17 (0.67–2.04)	0.579
BMI (kg/m2)	0.97 (0.91–1.02)	0.298
Alcohol intake (yes vs no)	2.00 (1.22–3.27)	0.005	2.28 (1.23–4.22)	0.008
Diabetes (yes vs no)	1.25 (0.71–2.20)	0.424
Hypertension (yes vs no)	1.54 (0.91–2.63)	0.105
AST/ALT ratio	3.09 (1.91–4.99)	<0.001	1.96 (1.10–3.46)	0.020
Anti-HBc (yes vs no)	1.28 (0.73–2.21)	0.378
Serum albumin concn	0.85 (0.80–0.90)	<0.001	0.89 (0.82–0.96)	0.002
Alpha-fetoprotein concn	1.05 (1.03–1.08)	<0.001	1.05 (1.02–1.07)	0.006
Bilirubin concn	1.02 (1.00–1.05)	0.048	0.97 (0.93–1.00)	0.124
Creatinine concn	0.99 (0.09–1.01)	0.509
Platelet count	1.00 (1.00–1.00)	<0.001	1.00 (1.00–1.01)	<0.001
Prothrombin proportion	0.96 (0.94–0.98)	<0.006	1.03 (1.00–1.05)	0.035
HCV RNA load	0.99 (0.73–1.34)	0.954
Genotype (1 vs others)	0.77 (0.47–1.27)	0.318
Metavir score (F3 vs F4)	2.45 (1.46–4.13)	<0.008	1.55 (0.86–2.78)	0.138
SVR vs non-SVR	0.31 (0.13–0.74)	0.008	0.35 (0.15–0.84)	0.019
Non-SVR vs untreated	0.41 (0.24–0.71)	0.001	0.51 (0.27–0.96)	0.038

^aAbbreviations: HR, hazard ratio; CI, confidence interval.

^bUnivariate Cox proportional hazards regression analyses with statistically significant P values (P < 0.05).

^cMultivariate Cox proportional hazards regression analyses to adjust the HR of SVR, non-SVR, and untreated patients for all-cause mortality.

Baseline factors significantly associated with death or liver transplantation in multivariate Cox analysis were alcohol intake (HR = 2.28, 95% CI = 1.23 to 4.22; P = 0.008), the AST/ALT ratio (HR = 1.96, 95% CI = 1.10 to 3.46; P = 0.020), the alpha-fetoprotein concentration (HR = 1.05, 95% CI = 1.02 to 1.07; P = 0.006), and the prothrombin proportion (HR = 1.03, 95% CI = 1.00 to 1.05; P = 0.035) (Table 3). After adjusting on these baseline characteristics, the risk of death or liver transplantation was lower in SVR patients than in non-SVR patients (HR = 0.35, 95% CI = 0.15 to 0.84; P = 0.019) and in non-SVR than in untreated patients (HR = 0.51, 95% CI = 0.27 to 0.96; P = 0.038).

When stratification on propensity scores (PSs) was used, the treatment effect was slightly lower than that without the use of stratification on PSs, but it remained highly significant (HR = 0.28 versus 0.34, respectively; 95% CI = 0.16 to 0.49 versus 0.18 to 0.65, respectively; P = 0.001 in both cases). Similar results were found in a restricted analysis including non-SVR and untreated patients only (HR = 0.48 and 0.36, respectively; 95% CI = 0.25 to 0.93 and 0.20 to 0.63, respectively; P = 0.029 and P = 0.004, respectively). The results of the PS analysis are given in Table S2 in the supplemental material.

DISCUSSION

The objective of this study was to evaluate the effect of anti-HCV treatment, either successful or unsuccessful, on the survival of patients with advanced liver disease and to compare their survival with that of untreated patients. For that purpose, we retrospectively analyzed the overall survival reported in the HepAList database, a large repertoire of patients admitted for care at Beaujon Hospital (Paris area, France). We limited the heterogeneity of the population by including only patients naive for treatment of HCV infection who were enrolled between 2000 and 2010, who had a Metavir score of F3 or F4 (as assessed by biopsy), and who did not have or had no history of a Child-Pugh score of B or C. With this population, we found a pronounced gradient for the cumulative overall rate of death/liver transplantation among SVR, non-SVR, and untreated patients (Fig. 2).

The fact that SVR patients with advanced liver disease had a better overall survival than non-SVR or untreated patients has been reported in some studies (20, 21). However, only a small number of patients with advanced fibrosis was included in these studies, making it difficult to evaluate the effect of treatment in this specific population, which is characterized by both a lower SVR rate and a higher mortality rate. In a recent study by Van der Meer et al. (14), where only patients with Metavir scores of F3 or F4 were included, the incidence rate of all-cause mortality per 100 person-years was about three times lower in SVR patients than non-SVR patients (1.01 versus 2.93), consistent with our findings (1.00 versus 3.20). However, all patients included in this study were treated, and thus, the effect of unsuccessful treatment over the absence of treatment could not be investigated. To our knowledge, only one study, by Aleman et al. (15), specifically compared the incidence of death in non-SVR and untreated cirrhotic patients and found that mortality was significantly lower in non-SVR patients than untreated patients. Our sample size was comparable to that of the study of Aleman et al. (15) (total number of patients, 427 and 351, respectively), but the number of untreated patients was larger in our study (102 versus 48 in the study of Aleman et al. [15]); the follow-up times in the two studies were also comparable (means, 5.5 and 5.3 years, respectively). The fact that the incidence rate for overall death in non-SVR patients was higher in the study of Aleman et al. (15) than in our study (4.1 versus 3.20 per 100 person-years, respectively) might be due to the fact that only cirrhotic patients with a Metavir score of F4 were included in their study, whereas patients with Metavir scores of F3 and F4 were included in our study. Consistent with the hypothesis that the effect of treatment might depend on the fibrosis stage, the incidence rate of overall death in SVR patients was also larger in the study of Aleman et al. (15) and was equal to 1.8; i.e., it was substantially larger than that found in our study and in the study of Van der Meer et al. (14) (1.0 and 1.01, respectively). However, the incidence rates for untreated patients were largely similar in both studies (5.4 and 5.1, respectively).

Overall the difference in incidence rates between non-SVR and untreated patients was more pronounced in our study (3.20 versus 5.4, respectively; P = 0.038) than in the study of Aleman et al. (15) (4.1 versus 5.11, P = 0.11).

The estimation of the effect of treatment on survival is subject to various sources of bias. The most problematic one is the bias of indication, i.e., the fact that patients who did receive a treatment may have characteristics different from those who did not receive a treatment. Because peg-IFN treatment leads to a modest SVR rate, especially in patients with advanced liver disease, but may cause severe side effects, it is possible that patients who were not treated were also the ones with less favorable clinical profiles. In order to control this bias, we first limited the heterogeneity of the population by excluding all patients with Child-Pugh scores of B and C. Second, we examined retrospectively the medical files of all untreated patients and found that treatment was contraindicated because of the presence of liver diseases in only a small subset of untreated patients (n = 9 patients). Third, the survival analyses were adjusted on baseline factors, and our conclusions remained unchanged. Lastly, we verified that our results remained unchanged when using propensity scores (see Table S2 in the supplemental material). A second source of bias is the bias of survival, i.e., the fact that survival may be higher in treated patients because the effect of treatment can be estimated only in patients who survived until the time of treatment initiation. This bias could be corrected by including treatment as a time-dependent covariate in the analysis, as was done, for instance, in the study of Aleman et al. (15). However, here, this bias was minimal, since all patients received a treatment within the first 5 months of their inclusion in the cohort and only 21 patients achieved an SVR after a second course of treatment. Lastly, the heterogeneity of treatment was controlled because all patients received dual therapy with IFN and RBV. This is an important difference from previous studies, where cohorts were treated with several different therapeutic options.

In this study, we purposely focused on overall survival because this information could be obtained even if patients dropped out of the study. In fact, information on living status and the occurrence of liver transplantation was obtained for all patients, and therefore, there were no censored data. This would not have been the case if we had considered HCC or decompensation, which may be subject to interval censoring and, in addition, to a bias of diagnosis. Moreover, a previous analysis of the same cohort found that SVR patients had significantly lower chances of HCC and liver complications than non-SVR patients (22). Consistent with this finding, results from the HALT-C trial showed a lower incidence of HCC in patients randomized to long-term, low-dose pegylated interferon; however, the difference emerged only after 5 to 7 years of follow-up (23).

The mechanism by which IFN-RBV may improve overall survival, even though viral eradication is not achieved, is unclear. One possibility is that IFN may act in ways that are not directly related to its antiviral efficacy, in particular, through its immunomodulatory activity (24). Another possibility is that treatment leads to a reduction of liver inflammation and delays the natural course of the infection. This is supported by the fact that in our study, treated patients with a virological or a biochemical response had better overall survival than treated patients without a virological or a biochemical response. This result, consistent with the results of similar studies previously reported in the literature (21, 25, 26), reinforces the hypothesis that survival is causally related to the effect of treatment and that the better survival observed in non-SVR patients than untreated patients is not due to a bias of indication.

Here we observed that, after a delay of approximately 50 months, no death occurred in the non-SVR patients and the overall survival between non-SVR and untreated patients tended to be parallel after that delay (Fig. 1). This may suggest that a 3- to 4-year interval between two treatment courses may be relevant in order to maintain the protective effect of treatment.

In conclusion, IFN-RBV treatment, even if it does not lead to an SVR, was found to improve the overall survival in naive patients with advanced fibrosis. Further studies will be needed to evaluate whether this effect is IFN specific or also exists in the new era of IFN-free treatment.