Human Immunodeficiency Virus/Hepatitis C Virus (HCV) Co-infected Patients With Cirrhosis Are No Longer at Higher Risk for Hepatocellular Carcinoma or End-Stage Liver Disease as Compared to HCV Mono-infected Patients

Patients and Definition of Baseline

Patients were recruited from two French national prospective cohorts of the ANRS.

CirVir-ANRS CO12 is a cohort of patients with compensated biopsy-proven cirrhosis attributed to viral B or C hepatitis, recruited in 35 clinical centers specialized in hepatology from March 2006 until June 2012.14 Hepavih-ANRS CO13 is a cohort of HIV/HCV co-infected patients recruited from 2005 to 2016 from 27 clinical centers specialized in management of HIV infection.15 In both cohorts, patients with cirrhosis have follow-up every 6 months for hepatocellular carcinoma (HCC) surveillance whereas patients without cirrhosis have yearly follow-up. Both protocols were approved by the Ethics Committee and conformed to ethical guidelines of the 1975 Declaration of Helsinki. All patients gave written informed consent to participate in the cohorts. The full CirVir and Hepavih protocols are available on the ANRS website (http://anrs.fr).

Inclusion criteria in this substudy were as follows: age older than 18 years; histologically proven cirrhosis, whatever the time of biopsy; negative hepatitis B surface antigen; absence of previous complications of cirrhosis (particularly ascites, gastrointestinal hemorrhage, or HCC); patients in Child-Pugh class A stage; and absence of severe uncontrolled extrahepatic disease resulting in estimated life expectancy of less than 1 year. Patients without liver biopsy and those with an F3/F4 Metavir score from the Hepavih cohort on their liver biopsy were excluded from the analysis.

For patients from the Cirvir cohort, baseline corresponded with date of inclusion. For patients from Hepavih eligible for this substudy, baseline was defined by the date of inclusion in this cohort if the cirrhosis was histologically proven before inclusion, otherwise by the date of liver biopsy performed during follow-up.

Follow-Up and Definitions of Endpoints

Patients were seen by physicians every 6 months, and the usual clinical and biological data were recorded. Examination by Doppler ultrasound (US) was performed every 6 months. In cases of focal liver lesion detected by US, a diagnostic procedure using contrast-enhanced imaging (computed tomography scan or magnetic resonance imaging), serum alpha-fetoprotein (AFP) assay, and/or guided biopsy was performed according to 2005 American Association for the Study of Liver Diseases (AASLD) guidelines.16 HCC diagnosis was thus established either by histological examination performed by an experienced pathologist or by using probabilistic noninvasive criteria (mainly dynamic imaging showing early arterial hypervascularization and portal washout) according to the different periods of time (before and after 2011). When HCC diagnosis was established, treatment was determined using a multidisciplinary approach according to AASLD guidelines for HCC.16

Patient follow-up was maintained from the date of cohort inclusion until December 2015 or until the date of event, meaning that they were censored at the earliest at the: date of last follow-up, date of liver transplantation (LT), and date of liver-related or non-liver-related death in accord with the studied event. All events occurring during follow-up, liver-related or not, were recorded based on information obtained from medical files of patients from each center. In particular, all incidental cases of HCC and all episodes of liver decompensation (LD) encompassing ascites, hepatic encephalopathy, and gastrointestinal bleeding were described, as well as their severity, management according to international recommendations,16 and outcome. Causes of death were assessed by an independent validation committee. Sustained virological response (SVR) following an anti-HCV regimen was defined by undetectable HCV RNA 12 weeks after treatment cessation.

Statistical Analyses

Continuous variables were described by means ± SD or medians (interquartile range; IQR) and categorical data by numbers (percentages). Baseline characteristics were compared between the two groups of patients according to HIV coinfection status using the Student t test or Mann-Whitney U test for continuous variables. Categorical variables were compared using the χ² test or Fisher's exact test, as appropriate.

Predictors associated with outcomes were assessed using uni- and multivariate Cox proportional regression models. All features were assessed according to their baseline levels, with the exception of HIV treatments (cART) and SVR. These variables were included as time-dependent covariates in all models because treatments may start during follow-up and SVR may occur at different time points. Fixed cART values were used for patients receiving cART at baseline (cART = 1) and those who had never received cART (cART = 0). Patients without cART at baseline who received cART during follow-up had their cART status switched from 0 to 1 at the start of treatment. Similarly, fixed SVR values were used for patients who had never experienced SVR (SVR = 0) and for those with SVR at the time of their inclusion (SVR = 1). Non-SVR patients at inclusion who achieved SVR during follow-up had their status switched from non-SVR to SVR; this took into account the end of treatment, which led to undetectable HCV RNA as the time point for setting SVR values from 0 to 1. No reinfection or relapse, as defined by detectable HCV RNA in a patient who had previously achieved SVR, was observed during the follow-up. Variables in the multivariate models were defined after backward step-wise elimination of nonsignificant variables at a P level <0.05.

Age-adjusted cumulative incidence curves according to HIV status were built for HCC, LD (ascites, digestive bleeding and encephalopathy), overall mortality, and liver- and non-liver-related mortality at the mean age of the entire population (56.9 years), with estimates of incidence determined using a Fine-Gray competing risk regression model computing age-adjusted subhazard ratios (SHRs) along with their 95% CIs. Competing events were defined as follows: overall death for HCC and liver decompensation, LT for overall mortality, non-liver-related death for liver-related mortality, and liver-related death for non-liver-related mortality. As a sensitivity analysis, age-adjusted cumulative incidence of all events was also computed using a Cox proportional hazard regression model to compare results from competing risk analysis.

Crude cumulative incidence of overall, liver-related, and non-liver-related death were determined by calculating the cumulative incidence function, and the influence of HIV status on these outcomes was evaluated by a Gray test and a Fine-Gray model.

To take into account the differences between HIV/HCV co-infected patients and HCV mono-infected patients at baseline, a propensity-score–matching analysis was performed. This propensity score was built using a logistic regression model. The model included all the covariables for which a significant difference (P < 0.05) was found between the two groups: age, sex, bilirubin, alanine aminotransferase (ALT), body mass index (BMI), mode of contamination, HCV RNA, tobacco consumption, alcohol consumption, diabetes, and HCV genotype. Pairs of patients with and without HIV co-infection were created (1:1 matching) based on the closest logit of propensity scores and a caliper width of a 0.2 SD of the logit of this propensity score. It allowed matching 95 pairs of patients with similar characteristics. In this subpopulation, incidence curves of occurrence of HCC, liver failure, and overall and specific (liver- and non-liver-related) survival curves were constructed using the Kaplan-Meier method, according to HIV status. Effect of HIV co-infection on complications in this subpopulation was assessed by a Cox regression model, without and with adjustment on virological response, accounting for the clustered structure of the data between HCV and HIV/HCV patients.

Results are expressed as the risk of events for HIV/HCV co-infected patients compared to HCV mono-infected patients.

All statistical analyses were performed using Stata software (version 13.0; StataCorp LP, College Station, TX). A P value ≤0.05 was considered statistically significant.

Baseline Characteristics of Patients

A total of 1,671 patients with cirrhosis and 142 HIV/HCV co-infected patients with cirrhosis were extracted from both cohorts and considered in the present analyses (see flow-chart, Fig. 1). Among the 1,671 patients from the CirVir cohort, 365 were excluded because of HBV co-infection, lack of data on HIV co-infection, or duplicate with HIV/HCV cohort. Among the 142 patients from the Hepavih cohort, 20 were excluded because of absence of liver biopsy with Metavir grade F4, HBV co-infection, or past history of liver complication before liver biopsy procedure (HCC, LD, or LT). Consequently, final analyses were performed on 1,458 patients, comprising 1,253 HCV mono-infected patients with cirrhosis and 175 HIV/HCV co-infected patients with cirrhosis, including 53 HIV/HCV patients from the CirVir cohort (Fig. 1).

At baseline (Table 1), HIV/HCV patients were younger (P < 0.001), were more frequently males (P < 0.001), and had more frequently HCV genotype 3 (P = 0.006) and consumption of alcohol and tobacco (P < 0.001) than HCV mono-infected patients. Overall, 92.3% of HIV/HCV patients were on cART and 80.4% had an undetectable HIV. Both groups had similar rates of HCV eradication at baseline (16.6% vs. 19.9% in HCV mono-infected patients; P = 0.30). At the end of follow-up, the rate of patients with SVR did not differ either between both groups (47.0 % vs. 52.0% in HCV mono-infected patients; P = 0.22; Table 1), as well as the percentage of patients treated and cured with direct-acting antiviral (DAA) therapy (17 of 77 [22.1%] vs. 167 of 637 [26.2%]; P = 0.49).

Incidences of Liver-Related Events (HCC, LD)

At the end of follow-up in December 2015, median follow-up was 55.7 and 59.3 months in HIV/HCV co-infected and HCV mono-infected patients, respectively, and did not differ between both groups (P = 0.10).

During this time frame, 192 (13.4%) patients (179 HCV, 14.3%; 13 HIV/HCV, 7.4%) developed an HCC and 225 (15.4%) (206 HCV, 16.4%; 19 HIV/HCV, 10.9%) developed at least one episode of LD (Tables 2 and 3). Median delay for occurrence of an HCC was 33.2 months and did not differ between HIV/HCV co-infected and HCV mono-infected patients (36.6 vs. 32.4 months, respectively; P = 0.90). However, HCC presentation at diagnosis was more severe in HIV/HCV co-infected patients as reflected by the larger size of the main tumor (P = 0.023) and the higher frequency of tumoral portal thrombosis (P = 0.011) than in HCV mono-infected patients. This was observed despite a similar delay since the last normal liver US echography in both groups (6.3 [IQR, 5.6-8.2] vs. 6.5 months [IQR, 5.6-9.2]; P =0.66) and a time ≥7 months since last imaging examination before HCC diagnosis in 30.8% and 40.2%, respectively (P = 0.50; Table 2). Moreover, HIV/HCV patients were younger than HCV mono-infected patients at the time they developed HCC (median, 52.9 vs. 61.8 years). Although the percentages of patients outside Milan were not significantly different between the two groups, a univariate analysis of the predictors for presenting HCC outside Milan showed that only a younger age (OR, 0.96; 95% CI, 0.92-0.99; P = 0.026) was associated with this outcome.

After adjustment for age, HCC 5-year cumulative incidence was similar in HIV/HCV patients and in HCV mono-infected patients (8.5% vs. 13.2%; SHR = 0.63; 95% CI, 0.35-1.13; P = 0.12; Fig. 2A).

Median delay for occurrence of LD was 32.8 months and did not differ between HIV/HCV co-infected and HCV mono-infected patients (30.1 vs. 35.5 months, respectively; P = 0.17; Fig. 2B).

After adjustment for age, 5-year cumulative incidence of LD was similar in HIV/HCV patients and in HCV mono-infected patients (12.8% for HIV/HCV vs. 15.6% for HCV; SHR = 0.81; 95% CI, 0.49-1.33; P = 0.40).

In addition, cumulative incidences of HCC and LD assessed by the Kaplan-Meier method and Cox regression gave similar results.

Supporting Tables S1 and S2 summarize results of univariate analyses. In multivariate models (Table 4), HIV/HCV co-infection with or without cART intake was not associated with a higher risk of liver complications. Meanwhile, absence of SVR, severity of disease as assessed by platelet count <100,000/mm³ or albumin ≤35 g/L, and elevated gamma-glutamyl transferase (GGT) were significantly associated with occurrence of liver complications.

Mortality and Causes of Death

One hundred eighty-nine patients died during the follow-up (23 HIV-HCV co-infected patients [13.1%] vs. 166 HCV mono-infected [13.2%]).

Median delay for occurrence of a death was 46.3 months.

Overall causes of death were HCC or cholangiocarcinoma (suppress CCA; n = 48; 25.4%), other liver-related disease (n = 47; 24.9%), non-liver-related death (n = 97; 39.2%), and unknown (n = 20; 10.6%; Table 3).

Overall crude mortality rate was similar in both groups (12.9% for HIV/HCV vs. 10.8% HCV at 5 years) with an SHR of 1.11 (95% CI, 0.72-1.72; P = 0.64) as well as crude liver-related and non-liver-related mortality rates (liver-related, 5.1% for HIV/HCV vs. 5.4% for HCV at 5 years, P = 0.36; non-liver-related, 6.5% for HIC/HCV vs. 4.7% for HCV at 5 years, P = 0.32).

However, HIV/HCV co-infected patients died at a younger age—10 years in median—than HCV mono-infected patients (47.3 vs. 56 years; P < 0.001).

After adjustment for age, HIV/HCV co-infected patients presented a higher risk of overall mortality (SHR = 1.88; 95% CI, 1.15-3.06; P = 0.011; Fig. 3A) and non-liver-related mortality (SHR = 2.44; 95% CI, 1.18-5.07; P = 0.016; Fig. 3C), compared to HCV mono-infected patients.

Similar findings were obtained by the Kaplan-Meier method and Cox regression models.

Results of univariate analyses for overall mortality are reported in Supporting Table S3. In multivariate models (Table 4), HIV/HCV co-infection was associated with a higher risk of overall death. The other factors associated with overall death were absence of SVR, age, and severity of disease as assessed by platelets count <100,000/mm³ and albumin ≤35 g/L (Table 4).

A sensitivity analysis for the predictors of HCC, LD, and mortality was performed with exclusion of the few patients not receiving cART at baseline. The corresponding results are presented in Supporting Tables S4-S7.

Use of a Propensity Score

In order to take into account the differences between HCV mono-infected patients and HIV-HCV co-infected patients at baseline, we used a propensity score constructed from the regression coefficients of all the variables included in the model.

This propensity score enabled the matching of 95 patients from each group with the same score and similar baseline characteristics (Supporting Table S8). Similar results were observed in this subpopulation as in the entire study population. Five-year incidence of HCC and of hepatic decompensation did not differ significantly between HIV/HCV and HCV mono-infected patients (hazard ratio [HR] = 0.73; P = 0.47 and HR = 0.88; P = 0.74, respectively; Fig. 4A,B). Overall survival remained shorter in HIV/HCV co-infected patients than in HCV mono-infected patients (HR = 3.08; P = 0.011; Fig. 4C). Adjusted for SVR status (HR = 0.20; P = 0.043), the effect of HIV co-infection on overall survival persisted (HR = 3.02; P = 0.015).