HIV Articles  
Back 
 
 
HAART Associated With Accelerated Fibrosis Progression, study reports
 
 
  "Evaluating Liver Fibrosis Progression and the Impact of Antiretroviral Therapy in HIV and Hepatitis C Coinfection Using a Noninvasive Marker"
 
JAIDS Journal of Acquired Immune Deficiency Syndromes: Volume 44(4) 1 April 2007 pp 463-469
 
Al-Mohri, Huda MD*; Murphy, Tanya MSc; Lu, Ying MSc; Lalonde, Richard G MD; Klein, Marina B MD, MSc
 
"....Cumulative HAART did not protect against liver complications.... during 14 years of follow-up, APRI scores (non-invasive measure of liver fibrosis progression) changed significantly-to the greatest extent in patients with HIV-HCV but, interestingly, also in HIV monoinfection .....we used the APRI as a surrogate marker to evaluate the impact of HAART on the progression of liver fibrosis-the first longitudinal study to address this question. There was clearly a complex relationship between HAART and fibrosis. Although a higher CD4 cell count and better HIV control were indeed associated with lower rates of fibrosis progression as expected, HAART was additionally associated with increased progression in the APRI scores... HAART may accelerate progression to moderate to severe fibrosis but has less of an impact on further progression to end-stage liver disease ... cumulative d4T exposure was associated with an adjusted increase in the ln(APRI) in patients with HIV-HCV and in patients with HIV... during 14 years of follow-up, APRI scores changed significantly-to the greatest extent in patients with HIV-HCV but, interestingly, also in HIV monoinfection....
 
....Although a higher CD4 cell count and better HIV control were indeed associated with lower rates of fibrosis progression as expected, HAART was additionally associated with increased progression in the APRI scores. This finding is in contrast to those of a cross-sectional study14 and a case-control study, which concluded that early ARV therapy in coinfected patients may slow liver fibrosis progression,11 but is consistent with those of other reports.10,13 In exploratory analyses, the increase in the ln(APRI) score was significantly greater with PI-based regimens than with NNRTI-based regimens. This is in contrast to the findings of a cross-sectional study, which reported that HAART regimens containing nevirapine accelerated, whereas those containing PI slowed, fibrosis progression in coinfection.18 Furthermore, we found that exposure to d4T was additionally associated with an increase in the ln(APRI). Despite the association of HAART exposure with the change in the APRI score, this did not translate into an increased rate of hepatic outcomes. Given the long natural history of HCV-related liver disease, even longer follow-up may be required to observe sufficient numbers of hepatic events. Alternatively, HAART may accelerate progression to moderate to severe fibrosis but has less of an impact on further progression to end-stage liver disease....
 
...There are several possible reasons why HAART may not slow the rate of fibrosis. It is clear that HAART may worsen hepatic disease in coinfection through immune reconstitution or hepatotoxicity in the short term,24,25 but its long-term impact remains controversial. ARV therapy (particularly d4T and PIs) has increasingly been associated with important metabolic complications, including hepatic steatosis in individuals with and without viral coinfection.26,27 Steatosis alone may lead to cirrhosis.28 Indeed, the small but significant increase in APRI scores that occurred in HIV monoinfection may represent cumulative toxicity from ARVs. This finding raises the concern that liver-related complications may occur even among those without coinfection. Certainly, several other potential factors warrant consideration, including possible irreversibility of hepatic damage occurring before HIV treatment and problems with access or adherence to HAART in a population with high rates of substance abuse. HIV-HCV-coinfected injection drug users had an equal uptake of HAART as those with HIV alone and had a similar proportion of visits with suppressed HIV, suggesting that this was not a factor in our population (data not shown)...."
 
Summary:
The effects of highly active antiretroviral therapy (HAART) on progression of hepatic fibrosis in hepatitis C virus (HCV) coinfection with HIV are not well understood and are difficult to measure because of the need for repeated liver biopsy. We evaluated the evolution of a noninvasive measure of liver fibrosis, the alanine aspartyl transferase (AST)-to-platelet ratio index (APRI), longitudinally and determined its predictive value for hepatic outcomes in HIV-positive patients with and without HCV coinfection. A total of 673 HIV-positive patients without liver complications at baseline (540 with HIV only, 133 with HIV-HCV coinfection) were followed between 1991 and 2004 for a median of 4.6 years (3524 person-years). At baseline, HIV-HCV coinfection had a higher median APRI compared with HIV infection alone (0.59 vs. 0.33; P < 0.0001). The natural logarithm of the APRI [ln(APRI)] changed significantly over time, particularly among patients with HIV-HCV coinfection. The baseline ln(APRI) was predictive of liver complications (hazard ratio [HR] = 4.0, 95% confidence interval [CI]: 2.5 to 6.4 per log), as was HCV (HR = 4.5, 95% CI: 1.5 to 14). Cumulative HAART did not protect against liver complications, although it was significantly associated with progression of APRI scores in HIV-HCV coinfection and in HIV alone. In conclusion, the APRI may be a useful marker for longitudinal evaluation of the progression of liver disease in HIV-HCV coinfection.
 
Background
Liver fibrosis progression is more rapid in the context of hepatitis C virus (HCV)-HIV coinfection1-3 and is associated with lower CD4 cell counts.4,5 Thus, the risk of liver disease would be expected to decrease after the availability of highly active antiretroviral therapy (HAART). The opposite has been observed, however; HCV is an emerging cause of morbidity and mortality.6-9 Although this may simply be the unmasking of liver disease because of longer survival with HAART, other factors may be at play, including irreversibility of hepatic damage, incomplete immune recovery, and chronic hepatotoxicity related to antiretrovirals (ARVs). The true rate of hepatic fibrosis progression in coinfected patients and the effects of HAART are unknown. Most studies examining this question have been cross-sectional,2,10-13 have relied on modeled fibrosis progression rates,10,11,13,14 or have failed to take into account cumulative HAART exposure.10,15,16 Studies involving paired liver biopsies have been of small sample size.17 Results have therefore been contradictory, with some suggesting that HAART, especially that containing protease inhibitors (PIs), is associated with lower fibrosis rates,11,14,18 whereas others show no benefit.10,13 Ideally, studies involving sequential liver biopsies on a sufficient number of subjects would provide the best assessment of fibrosis progression. Multiple biopsies are impractical, costly, and risky, however. Noninvasive measures of liver fibrosis could therefore be useful. The alanine aspartyl transferase (AST)-to-platelet ratio index (APRI) has been validated as a good predictor of significant liver fibrosis in HCV monoinfection19,20 and in coinfection21,22 when compared with the gold standard, liver biopsy. For example, in coinfection, an APRI score >1.5 was 100% specific and 52% sensitive for predicting significant fibrosis (area under the receiver operating characteristic [ROC] curve of 0.85 ± 0.06).22 In this study, we evaluated the evolution of the APRI and its predictive value for hepatic outcomes in a cohort of HIV and HIV-HCV-coinfected patients with more than 10 years of follow-up and assessed the effects of HAART on liver disease progression.
 
METHODS
Subjects

The Montreal Chest Institute Immunodeficiency Service is a university-based HIV clinic that has maintained a computerized database on all patients since 1989. Clinical (including AIDS and other diagnoses and deaths), laboratory, and prescription information is prospectively entered. Changes in ARVs are reviewed at each visit by a physician or pharmacist and recorded. Subjects were HIV-seropositive (positive enzyme-linked immunosorbent assay [ELISA] with confirmatory Western blot test), attended the clinic at least twice between January 1991 and December 2004, were not diagnosed with cirrhosis at baseline, and had at least 2 concurrent measures of AST and platelets. Subjects were categorized as having chronic HCV (HIV-HCV) or HIV alone (HIV). Patients with HCV and chronic hepatitis B virus were excluded. HCV testing was performed using an ELISA, followed by confirmatory recombinant immunoblot assays. Repeat serologic testing for HCV was not routinely performed. Because HCV RNA was not consistently measured, it was not used to define the population. Patients were followed until July 2005 or were censored on their last clinic visit if lost to follow-up. Patients (n = 10) were censored when they received specific HCV therapy. The Institutional Research Ethics Board approved the study.
 
Baseline demographic, clinical, and laboratory variables and follow-up data for 6-month intervals were extracted from the database. The APRI was defined as: [100 X (AST/upper limit of normal)/platelet count (109/L)].20 Concurrent measures (taken a maximum of 7 days apart) of AST and platelets were used to calculate APRI values. For each 6-month interval, the most recent APRI value was used. If no measure was available in an interval, it remained missing for the analyses (ie, no imputation was made for missing data). CD4+ T-cell counts and plasma HIV viral load were measured using standard techniques.23 HAART was defined as at least 3 ARVs taken concurrently for more than 1 day. Cumulative exposure to HAART was determined by calculating the total number of days on HAART up to the last APRI measurement in each interval. Cumulative HAART did not include exposure before baseline, because these data were retrospectively acquired and considered less reliable. To account for prior ARV exposure, we analyzed a dichotomous term (yes [Y], no [N]) in adjusted models. Liver complications included decompensated cirrhosis, hepatic encephalopathy, esophageal varices, ascites, spontaneous bacterial peritonitis, hepatoma, and death attributable to liver cirrhosis. Diagnoses and deaths were verified by reviewing hospital charts and death certificates.
 
Analysis
An APRI score >1.5 was considered to represent significant fibrosis (corresponds to a biopsy score ≥F2). The natural logarithm of the APRI [ln(APRI)], which nearly normalizes the distribution, was used in all analyses. Baseline was defined as the date of first available APRI measure. Baseline characteristics in HIV and HIV-HCV were compared using the ƒÔ2 test for proportions and Wilcoxon-Mann-Whitney test for continuous variables. The trend in the ln(APRI) over time was estimated using repeated measures multiple linear regression (SAS PROC MIXED; SAS Institute, Cary, NC). The mean number of APRI values analyzed per patient was 8 (range: 2-29). The Kaplan-Meier survival method was used to estimate the cumulative incidence of developing an APRI score >1.5 in those with normal baseline scores and the cumulative incidence of liver complications in all subjects, stratified by HCV. Cox proportional hazards models were used to predict the development of an APRI score >1.5 and liver-related morbidity and mortality. We used forward selection to build adjusted models and forced baseline CD4 cell counts into all models. The proportionality of hazards was checked graphically and statistically. Cumulative time spent on HAART was included as a continuous time-dependent variable in regression models to analyze its effects. We also examined time on PI- and nonnucleoside reverse transcriptase inhibitor (NNRTI)-based regimens and exposure to stavudine (d4T). All final models were adjusted for age, gender, time since HIV diagnosis, and calendar year at baseline. In some analyses, the CD4 cell count and HIV RNA level were also modeled as time-dependent covariates accounting for baseline values. The final model selection was based on the overall goodness of fit as assessed by the Akaike information criterion (AIC). Data were analyzed using SAS software (version 8.02; SAS Institute).
 
RESULTS
Baseline demographic and clinical characteristics are shown in Table 1. In total, 673 HIV-positive patients were included: 540 (80%) were infected with HIV only, and 133 (20%) were infected with HIV-HCV. Patients with HIV-HCV were more likely to have acquired HIV infection through injection drug use (IDU). AST and alanine aminotransferase (ALT) levels were significantly higher (P < 0.0001) and platelet counts were lower (P < 0.01) in patients with HIV-HCV than in patient with HIV. The median duration of follow-up was 4.6 years (range: 0.4-14 years), representing 3524 person-years (p-y). By study end, 48 patients (7%) had died and 88 (13%) had not been seen within 1 year and were not known to be deceased. Mortality was much higher in patients with HIV-HCV: 12% (2.6/100 p-y) compared with 5.9% (1.1/100 p-y) in patients with HIV (P = 0.006 for comparison of rates).
 
Evolution of Alanine Aspartyl Transferase-to-Platelet Ratio Index Over Time At baseline, patients with HIV-HCV had higher median APRI values compared with patients with HIV (0.59 vs. 0.33; P < 0.0001; see Table 1). Similarly, a greater proportion of coinfected patients had APRI scores greater than 1.5 (P < 0.0001). APRI scores were not distributed normally: low scores were clustered at less than 2 and then rose to high levels once past this threshold. Therefore, the ln(APRI) was used to model changes in APRI over time. The group-specific distributions of the baseline ln(APRI) are shown in Figure 1. Only infection group, low CD4 cell count, and high HIV RNA level were associated with the baseline ln(APRI) after adjustment for other variables.
 
The evolution of the ln(APRI) over time is illustrated in Figure 2A. Because of potential concerns about introducing time-dependent confounding, models with and without time-updated CD4 cell counts and HIV RNA levels were compared. The addition of time-updated values improved the model fit but did not significantly change the slope estimates obtained. Therefore, the final models included both variables modeled as time-updated parameters. In these analyses, HCV infection, the baseline ln(APRI), male gender, and longer duration of HIV seropositivity all independently predicted a greater change in the ln(APRI) during follow-up (Table 2, column 2). In addition, lower concurrent CD4 cell count and higher concurrent HIV RNA level were associated with an increased slope of the ln(APRI).
 
The ln(APRI) increased significantly over time in both groups, with the greatest rise seen in HIV-HCV infection compared with HIV infection (adjusted β = 0.18 vs. 0.07 over 3 years). Among patients with a baseline APRI score ≦1.5 (n = 630), 32 (29%) of 111 patients with HIV-HCV compared with 61 (12%) of 519 patients with HIV developed an APRI score >1.5 during follow-up (P < 0.001). The cumulative incidence of an APRI score >1.5 was 0.53 and 0.26/100 p-y, respectively (log rank, P < 0.0001). The corresponding adjusted hazard ratios (HRs) are shown in Table 3. HCV and the baseline ln(APRI) were the only independent predictors for developing significant fibrosis.
 
Predictive Value of Alanine Aspartyl Transferase-to-Platelet Ratio Index for Hepatic Complications
Liver complications developed in 6 (4.5%) of 133 patients with HIV-HCV compared with 6 (1.1%) of 540 patients with HIV (P = 0.008). The types of liver complications were similar in both groups. The unadjusted rates of liver complications were 1.0/100 p-y (95% CI: 0.4 to 2.2) in patients with HIV-HCV versus 0.2/100 p-y (95% CI: 0.1 to 0.5) in patients with HIV (P = 0.007). Among those developing hepatic complications, the median time to develop complications was shorter for patients with HIV-HCV at 2.85 years than in patients with HIV at 3.96 years. The baseline ln(APRI) and HCV coinfection predicted the development of hepatic complications in univariate models (HR = 4.0, 95% CI: 2.5 to 6.4 and HR = 4.5, 95% CI: 1.5 to 14, respectively). Given the small number of clinical endpoints observed, we did not develop an adjusted model. Although associated with hepatic outcomes, AST or platelets themselves were much weaker predictors than the ln(APRI). For example, the HR for AST was 1.01 (95% CI: 1.00 to 1.02), and for platelets, it was 0.99 (95% CI: 0.98 to 1.0). These HRs did not change when AST and platelets were included in the model.
 
Effect of Highly Active Antiretroviral Therapy on Progression of Alanine Aspartyl Transferase-to-Platelet Ratio Index
At baseline, 165 (25%) patients were receiving HAART and 364 (54%) patients were ARV naive, of whom 284 (78%) began HAART during follow-up. The total number with any HAART exposure before or during the study, including those who had interrupted treatment, was 595 (88%). The median duration of HAART exposure was 0.36 years before and 3.4 years after (2502 p-y) cohort entry. Time on HAART, as a proportion of observation time, was similar in both groups (80%-83%). To try to differentiate the effects of HAART exposure from the effects of the passage of time on the progression of fibrosis, we modeled cumulative time on HAART and observation time not on HAART as interaction terms with the infection group (see Table 2, column 3). The results indicate that patients with HIV-HCV experienced an increase in the slope of the ln(APRI) of 0.12 over 3 years while not on HAART and an increase of 0.21 over 3 years while on HAART. In patients with HIV, there was no significant change in the slope of the ln(APRI) while not on HAART but an increase in slope of the ln(APRI) of 0.09 over 3 years while on HAART, suggesting that the change in the ln(APRI) was mostly attributable to HAART exposure in this group.
 
We further restricted the analysis to ARV-naive patients starting HAART during follow-up in an attempt to control for potential confounding by indication, prior ARV exposure, and calendar time effects. Although this reduced our sample size, the same general pattern was estimated. The adjusted increases in slope of the ln(APRI) were as follows: 0.08 (95% CI: 0.03 to 0.12) over 3 years of HAART exposure for patients with HIV and 0.27 (95% CI: 0.15 to 0.39) over 3 years of HAART exposure for patients with HIV-HCV. No significant changes in APRI were observed during non-HAART time in either group.
 
In some exploratory analyses, we examined the progression of the ln(APRI) according to time spent on PI-based versus NNRTI-based HAART regimens and the role of d4T. Before or during the study, a total of 481 (71%) patients had PI exposure (mean = 2.9 years, 1881 p-y) and 334 (50%) patients had NNRTI exposure (mean = 1.2 years, 827 p-y). The HIV-HCV group spent somewhat less observed time on NNRTI-based regimens than the HIV group (13% vs. 25%, respectively). Changes in the ln(APRI) according to time spent on PI-based and NNRTI-based regimens are shown in Figures 2C and 2B, respectively, stratified by infection group. The following adjusted changes in the ln(APRI) were observed: for PI-based regimens, 0.09 (95% CI: 0.05 to 0.13) over 3 years for patients with HIV and 0.22 (95% CI: 0.13 to 0.31) over 3 years for patients with HIV-HCV and for NNRTI-based regimens, 0.10 (95% CI: 0.05 to 0.16) over 3 years for patients with HIV and 0.12 (95% CI: -0.08 to 0.32) over 3 years for patients with HIV-HCV. PI-based estimates control for time on NNRTI-based regimens, and vice versa.
 
Overall, the mean d4T exposure was 1.6 years (1067 p-y), with 336 (50%) patients having some d4T exposure. Without considering HAART exposure, cumulative d4T exposure was associated with an adjusted increase in the ln(APRI) in patients with HIV-HCV and in patients with HIV (0.19 over 3 years [95% CI: 0.06 to 0.32] and 0.10 over 3 years [95% CI: 0.05 to 0.14], respectively), but including d4T exposure in the previous analyses did not affect the estimates for HAART or other variables.
 
Finally, to evaluate whether these observed patterns translated into an increased rate of significant fibrosis (APRI score >1.5) attributable to HAART use, we included a time-dependent cumulative HAART term in the previous Cox regression models. HAART exposure was not an independent predictor of significant fibrosis (HR = 1.07, 95% CI: 0.93 to 1.24).
 
DISCUSSION
Understanding the progression of liver fibrosis in the era of HAART is crucial for tackling the growing epidemic of HIV-HCV coinfection. A reliable noninvasive marker that could be used repeatedly over time would be an important contribution. We demonstrated that the APRI has these characteristics. First, it was highly correlated with the presence of HIV-HCV coinfection at baseline. Second, during 14 years of follow-up, APRI scores changed significantly-to the greatest extent in patients with HIV-HCV but, interestingly, also in HIV monoinfection. Third, the baseline APRI score was predictive of the development of significant fibrosis and liver-related morbidity and mortality.
 
Given these findings, we used the APRI as a surrogate marker to evaluate the impact of HAART on the progression of liver fibrosis-the first longitudinal study to address this question. There was clearly a complex relationship between HAART and fibrosis. Although a higher CD4 cell count and better HIV control were indeed associated with lower rates of fibrosis progression as expected, HAART was additionally associated with increased progression in the APRI scores. This finding is in contrast to those of a cross-sectional study14 and a case-control study, which concluded that early ARV therapy in coinfected patients may slow liver fibrosis progression,11 but is consistent with those of other reports.10,13 In exploratory analyses, the increase in the ln(APRI) score was significantly greater with PI-based regimens than with NNRTI-based regimens. This is in contrast to the findings of a cross-sectional study, which reported that HAART regimens containing nevirapine accelerated, whereas those containing PI slowed, fibrosis progression in coinfection.18 Furthermore, we found that exposure to d4T was additionally associated with an increase in the ln(APRI). Despite the association of HAART exposure with the change in the APRI score, this did not translate into an increased rate of hepatic outcomes. Given the long natural history of HCV-related liver disease, even longer follow-up may be required to observe sufficient numbers of hepatic events. Alternatively, HAART may accelerate progression to moderate to severe fibrosis but has less of an impact on further progression to end-stage liver disease.
 
There are several possible reasons why HAART may not slow the rate of fibrosis. It is clear that HAART may worsen hepatic disease in coinfection through immune reconstitution or hepatotoxicity in the short term,24,25 but its long-term impact remains controversial. ARV therapy (particularly d4T and PIs) has increasingly been associated with important metabolic complications, including hepatic steatosis in individuals with and without viral coinfection.26,27 Steatosis alone may lead to cirrhosis.28 Indeed, the small but significant increase in APRI scores that occurred in HIV monoinfection may represent cumulative toxicity from ARVs. This finding raises the concern that liver-related complications may occur even among those without coinfection. Certainly, several other potential factors warrant consideration, including possible irreversibility of hepatic damage occurring before HIV treatment and problems with access or adherence to HAART in a population with high rates of substance abuse. HIV-HCV-coinfected injection drug users had an equal uptake of HAART as those with HIV alone and had a similar proportion of visits with suppressed HIV, suggesting that this was not a factor in our population (data not shown).
 
Our study is subject to certain limitations. Foremost is the potential for cohort effects. Thus, we examined the effects of time in our models, taking into account calendar year and duration of HIV seropositivity. By studying cumulative exposure to HAART (overall and in a subset of patients who were treatment naive), we tried to examine the direct effect of HAART on the rate of change in APRI scores. Furthermore, by including time-updated values of CD4 cell counts and viral loads, we attempted to account for effects over and above potential immunologic or virologic benefits of HAART. Although we acknowledge that this modeling strategy can potentially introduce time-varying confounding, we found little evidence of this, because adjusting for time-updated variables had no overall impact on the results except to improve model fit. It is also difficult to ascertain the independent effects of individual drugs in combination HAART, given the potential interactions between drugs and the indication for their selection. Another important limitation of our study was the lack of data on alcohol use, which may be more prevalent among the HCV-infected persons. Interestingly, in an ongoing prospective HCV coinfection cohort, we have found that rates of alcohol use are lower than anticipated (58%)29 and comparable to reported use in HIV-infected individuals as a whole.30 Finally, inclusion in the HCV group was based on positive serology at baseline, because HCV RNA measures and repeat serologic testing were not routinely performed. Thus, some misclassification may have occurred. If any bias was introduced, however, it likely resulted in an underestimation of the progression rate in HCV infection.
 
The APRI is a surrogate for liver fibrosis, and although it seems to perform well, it may not be as accurate as serial liver biopsies. The APRI is subject to fluctuations, because many factors influence AST and platelet levels. It is notable that the APRI has a similar area under the curve (AUC) and positive predictive value for fibrosis as other noninvasive markers such as the FIB-4,31 Groupe d'Etude Multidisciplinaire sur les Pathologies Lies au Virus C (MULTIVIRC) equation,32 and Forns index,33 which do not incorporate the same parameters. We are not suggesting that the APRI replace liver biopsy for individual patient management at this time but, rather, that it may be a useful marker to study liver disease progression longitudinally in large populations. To validate our findings; tease out independent effects of various types of HAART regimens, including nucleoside backbones; and address other potential confounders, prospective studies in larger populations with longer extended follow-up are necessary.
 
In conclusion, despite prolonged exposure to HAART, HCV-coinfected persons remain at increased risk of progression to end-stage liver disease. Using the APRI as a surrogate, we found that HAART itself, particularly PI-based regimens, is associated with increased fibrosis progression. Our findings highlight the need to treat HCV infection specifically, because immune restoration from HAART alone cannot be relied on to improve outcomes in HCV coinfection.
 
 
 
 
  icon paper stack View Older Articles   Back to Top   www.natap.org