| |
Antiretroviral treatment interruption leads to progression of liver fibrosis in HIV-hepatitis C virus co-infection
|
| |
| |
AIDS:
24 April 2011
Thorpe, Juliaa; Saeed, Sahara; Moodie, Erica EMb; Klein, Marina Ba; for the Canadian Co-infection Cohort Study (CTN222)
"ART interruption was associated with an increased risk of fibrosis progression in HIV-HCV co-infection that was only partially accounted for by HIV viral load and CD4+ T-cell counts. Our findings suggest that liver disease progression observed in ART-treated co-infected patients is partly due to the consequences of treatment interruptions."
"Our findings are consistent with the observation from the SMART study that co-infected patients who interrupted ART experienced a greater risk of nonopportunistic disease-related death"
"During 6 years of follow-up, 10% of patients in this cohort reported interrupting ART for at least 14 days and the majority continued their interruption for at least 6 months"
"To the best of our knowledge, this is the first study to specifically examine the impact of treatment interruption on liver fibrosis progression among co-infected persons in the clinical setting."
"A possible explanation for the association of interruption with developing liver fibrosis even after accounting for changes in CD4+ T-cell counts and HIV RNA may be that interruption increases inflammatory processes. A biomarker substudy of SMART found that among co-infected patients, those with impaired liver function were in a pro-inflammatory state that is associated with excess risk of non-AIDS-related death. Interruption of ART further intensified this pro-inflammatory state [35]. Furthermore, Macias et al. have recently reported that increased inflammatory activity independently predicted fibrosis progression in subsequent liver biopsies in co-infected patients [9]."
Abstract
Objective: Despite potential negative consequences, HIV/hepatitis C virus (HCV) co-infected patients may discontinue antiretroviral treatment (ART) for several reasons. We examined the impact of ART interruption on liver fibrosis progression in co-infected adults, using the aspartate aminotransferase-to-platelet ratio index (APRI) as a surrogate marker of liver fibrosis.
Method: Data were analyzed from a multisite prospective cohort of 541 HIV-HCV co-infected adults. ART interruption was included as a time-updated variable and defined as the cessation of all antiretrovirals for at least 14 days. The primary endpoint was the development of an APRI score at least 1.5. Time-dependent Cox proportional hazards regression and inverse probability-of-treatment weighting (IPTW) in a marginal structural model were used to evaluate the association of baseline and time-varying covariates with developing significant fibrosis.
Results: Patients were followed for a median of 1.02 years; 10% (n = 53) interrupted ART and 10% (n = 53) developed significant fibrosis. After accounting for potential confounders, including CD4+ T-cell count, HIV viral load, baseline APRI score, age and gender, the hazard ratio for ART interruption was 2.52 (95% confidence interval 1.20-5.28). Use of IPTW resulted in a similar effect estimate, suggesting that mediation by time-varying confounders was negligible.
Conclusion: ART interruption was associated with an increased risk of fibrosis progression in HIV-HCV co-infection that was only partially accounted for by HIV viral load and CD4+ T-cell counts. Our findings suggest that liver disease progression observed in ART-treated co-infected patients is partly due to the consequences of treatment interruptions.
Discussion
Interruption of ART has been shown to lead to a greater risk of nonopportunistic disease-related death in randomized trials, particularly among HIV-HCV co-infected participants [22]. To the best of our knowledge, this is the first study to specifically examine the impact of treatment interruption on liver fibrosis progression among co-infected persons in the clinical setting. We found a significant, harmful effect of interruption on the development of fibrosis after accounting for clinical factors. The effects of risk behaviors such as injection drug and alcohol use were also investigated. Our findings were further supported by analyses using the more stringent criteria of an APRI score at least 2.0 or clinical liver events. Although these results did not achieve statistical significance, this was likely due to limited power. The association between ART interruption and outcome was also evaluated using IPTW and was not found to be statistically significant; however, the point estimate was similar to that obtained using Cox regression, suggesting that mediation by time-varying confounders was actually negligible. Baseline APRI score was also found to be predictive of developing liver fibrosis, implying that among patients with the same ART interruption status, those with an elevated baseline APRI were at a greater risk of an outcome than those with a baseline APRI score less than 0.5.
During 6 years of follow-up, 10% of patients in this cohort reported interrupting ART for at least 14 days and the majority continued their interruption for at least 6 months. A study conducted in British Columbia, Canada, showed that 19% of HIV-infected participants interrupted ART in 2006 [34]. This was greater than the proportion of patients who reported interrupting in our cohort but may be partly due to the fact that the CCC consisted of patients who were in care and therefore likely more stable than the co-infected population as a whole.
Our findings are consistent with the observation from the SMART study that co-infected patients who interrupted ART experienced a greater risk of nonopportunistic disease-related death [22]. Although only one participant randomized to interrupt therapy died of liver-related causes, if SMART had not been terminated prematurely, it is likely that more liver-related deaths would have occurred. The use of significant liver fibrosis as the endpoint in the current analysis rather than liver disease-related death allowed for the relationship between interruption and liver disease progression to be studied in greater detail.
A possible explanation for the association of interruption with developing liver fibrosis even after accounting for changes in CD4+ T-cell counts and HIV RNA may be that interruption increases inflammatory processes. A biomarker substudy of SMART found that among co-infected patients, those with impaired liver function were in a pro-inflammatory state that is associated with excess risk of non-AIDS-related death. Interruption of ART further intensified this pro-inflammatory state [35]. Furthermore, Macias et al. have recently reported that increased inflammatory activity independently predicted fibrosis progression in subsequent liver biopsies in co-infected patients [9].
A strength of our study was the use of data from the CCC which specifically recruited patients who are extremely marginalized, access various models of care and who have diverse risk profiles. Many of these groups, in particular IDU, are typically under-represented in randomized controlled trials, limiting the generalizability of their results [36]. In addition, our study included data about IDU during follow-up which has not been considered in previous cohort studies that investigated the consequences of interruption in HIV-infected patients [37-39].
The development of significant liver fibrosis was measured using the APRI score which has been validated in HIV-HCV co-infected patients [23] but is not without limitations. It has been shown that interruption of ART may lead to a greater risk of thrombocytopenia [40]. As a result of the fact that APRI is affected by fluctuations in platelets, it is possible that changes in APRI in those who interrupted may be attributed to changes in platelets that are not necessarily due to the progression of liver fibrosis. However, among 21 patients who had a measurement of platelets performed before and during the interruption, the median change was relatively small and the inter-quartile range suggests that the platelet count actually increased during an interruption in several participants. Similarly, in theory fluctuations in AST due to increased HCV replication following treatment interruption or on resumption of ART due to hepatotoxicity may occur [41] and impact APRI measurements; however, in this cohort AST elevations postinterruption were found to be very modest. Finally, many included patients had interrupted for more than 6 months arguing against acute changes in APRI driving our findings. It would be ideal in future studies to examine other potential noninvasive markers of fibrosis such as hyaluronic acid [42] or transient elastography to validate our findings [43].
A potential limitation of this analysis is the possibility of residual confounding, caused by a relatively crude measurement of alcohol consumption using a variable that only indicated whether or not participants consumed any alcohol during the previous 6 months. This may not have allowed for binge drinking or consistently heavy consumption to be distinguished from less harmful drinking habits. Had this more detailed information been included, the estimated effect might have decreased.
Finally, data regarding ART interruptions were based on patient recall so it is possible that the number of interruptions reported underestimates the true number. This may be attributed to poor recall and hesitation to report interrupting treatment as it would not be condoned by their physicians. In addition, the relatively low sensitivity of the APRI in detecting individuals with liver fibrosis may have resulted in fewer outcomes being documented than actually occurred. Had both of these factors been absent, these analyses may have been better powered to detect the association between ART interruption and the development of liver fibrosis. Therefore, the estimated hazard ratio may have been slightly underestimated. Finally a significant proportion of patients were already on a treatment interruption at the time of inclusion. It is therefore possible that we have underestimated the impact of interruptions on fibrosis progression by excluding individuals who progressed before cohort inclusion. However, very few patients who were excluded for APRI more than 1.5 at baseline had a history of treatment interruption prior to baseline so it is unlikely to have had a major impact on our findings.
Results
Description of the cohort
Of the 912 HIV-infected individuals enrolled in the CCC at the time of analysis, 732 had virologic evidence of HCV infection and were treated with ART. After censoring patients who began HCV treatment and excluding those who were missing APRI at baseline (n = 14) or had significant fibrosis at baseline (n = 160), 541 study participants remained (Fig. 1). The median duration of follow-up was 1.02 years [interquartile range (IQR) 0.5-1.78]. During 760 person-years of follow-up, 10% (n = 53) of participants achieved an APRI score at least 1.5 and 10% (n = 53) interrupted ART (Table 1). Overall, the median age, baseline HIV RNA, nadir CD4+ T-cell count and highest viral load did not differ between included patients and those who were excluded. Among patients who were excluded for an APRI at least 1.5 at baseline, 15% (n = 24) interrupted ART during follow-up.
As seen in Table 1, the median duration of follow-up and nadir CD4+ T-cell count were significantly greater among those who interrupted during follow-up, compared to those who did not. ART regimens immediately prior to interruption were similar to those taken by participants who did not interrupt therapy. The lower CD4+ T-cell count and greater HIV viral load at baseline among those who interrupted may be explained by the fact that 34 of 53 patients who reported interrupting ART during follow-up were on an interruption during the first 6-month time interval (32 of these patients began their interruption more than 6 months prior to baseline, and two patients began their interruption within the first 6 months of the baseline visit). The median baseline platelet count among those who interrupted was found to be significantly lower than among those who did not interrupt however, in both groups it was still well above the normal cut-off of 150 x 109/l. In addition, baseline APRI scores did not differ between the two groups and were well below the cut-off used to identify the presence of significant fibrosis.
The total number of interruptions that occurred was 55. Two patients interrupted twice. The median duration of treatment interruption during follow-up was 180 days (IQR 65-310). Among the 53 patients who interrupted, 38% (n = 20) resumed ART during follow-up (two patients interrupted and resumed ART twice during follow-up), whereas the remaining interruptions were ongoing at the last clinic visit. The proportion of visits at which patients reported interrupting ART peaked in 2004 (15% of visits). Among 21 patients who had a measurement of platelets and AST performed before and during the interruption, the median change in platelets was -2 (IQR -34 to 24) x 109/l and the median change in AST was 21 (IQR 1-38) U/l. The median number of days between measurements was 217 (IQR 169-352).
Impact of antiretroviral treatment interruption on developing liver fibrosis
Univariate time-dependent Cox regression analyses revealed a significant harmful effect of ART interruption, higher HIV viral load and baseline APRI at least 0.5 on the development of liver fibrosis. In contrast, higher CD4+ T-cell counts were protective against achieving an APRI at least 1.5. The effect of age, gender, active IDU, alcohol use, duration of HIV and HCV infections, ART regimen, nadir CD4+ T-cell count, highest recorded HIV viral load and HBV co-infection on the outcome were not statistically significant (Table 2).
After adjustment, the effect of ART interruption on liver fibrosis was slightly attenuated but still statistically significant (hazard ratio 2.52, 95% CI 1.20-5.28). A baseline APRI score at least 0.5 was strongly associated with progression of fibrosis. The effect of CD4+ T-cell count, HIV viral load, age and gender were not statistically significant. When those who began HCV treatment were not excluded or censored from the analysis (n = 556), a similar effect of ART interruption on development of liver fibrosis was observed (hazard ratio 2.68, 95% CI 1.26-5.69).
Adjustment for time-varying confounders affected by prior interruption
In addition, the effect of ART interruption on achieving an APRI score at least 1.5 was estimated using IPTW in a marginal structural model (Table 3). Baseline APRI score at least 0.5 was significantly associated with developing liver fibrosis. The OR for the effect of interruption on experiencing an outcome did not reach statistical significance (OR 2.60, 95% CI 0.61-6.18) but was similar to the hazard ratio obtained from the multivariate Cox proportional hazards model, suggesting that mediation by time-varying confounders that are also affected by treatment was negligible in the Cox model.
Investigating the impact of antiretroviral treatment interruption on developing liver disease
The impact of ART interruption on the development of an APRI score at least 2.0 or a clinical liver disease diagnosis was assessed using the same multivariate Cox proportional hazards model as above. After excluding those classified as having an outcome at baseline, the analyzed cohort for this analysis was composed of 554 study participants. An APRI score at least 2.0 or a liver disease diagnosis was achieved by 9% (n = 48) of participants. The effect of interruption on the outcome was similar to that seen when an APRI score at least 1.5 was used as the endpoint; however, the hazard ratio did not reach statistical significance in the multivariate model (hazard ratio 2.12, 95% CI 0.87-5.16).
Methods
Study design, population and setting
Data were obtained from the Canadian HIV-HCV Co-infection Cohort (CCC) which has been described elsewhere [27]. In brief, recruitment began in 2003 and as of December 2009, 912 patients were enrolled from 16 sites across Canada. Eligible patients were at least 16 years old with documented HIV infection [HIV-seropositive by enzyme-linked immunosorbant assay (ELISA) with western blot confirmation] and chronic HCV infection or evidence of HCV exposure (HCV-seropositive by ELISA with recombinant immunoblot assay II or enzyme immunoassay confirmation, or HCV RNA positive). The study was approved by the research ethics boards of all participating institutions.
All patients included in this analysis had virologic evidence of active HCV infection (HCV RNA positive, limit of detection: 60 IU/ml, Roche Cobas Amplicor assay) and had at least one cohort visit between April 2003 and December 2009. Because HCV treatment may alter progression of liver fibrosis, patients receiving treatment at baseline were excluded (n = 17) and those who began therapy during follow-up were censored at that point (n = 46). Sensitivity analyses including those who received HCV treatment were performed. Only ART-treated participants were included and for those who began ART during follow-up, the baseline visit was considered to be the one at which ART was commenced. As the outcome of interest was development of an APRI score at least 1.5 (equivalent to a liver biopsy score of F2 or greater), participants with significant fibrosis (APRI score ³1.5) at baseline were excluded. Patients were censored when an outcome occurred or at their last clinic visit prior to December 2009. The analyzed cohort was composed of 541 study participants (Fig. 1).
Exposure and covariate assessment
After providing informed consent, participants underwent an initial evaluation followed by study visits approximately every 6 months. During each visit, patients completed a questionnaire and results from routine blood tests were extracted from laboratory reports by research personnel.
ART was defined as taking at least three antiretrovirals concurrently for at least 30 days. The exposure of interest, ART interruption, was defined as the cessation of all antiretrovirals for at least 14 days. Using information extracted from questionnaire responses, the occurrence of an interruption was assessed for each 6-month interval and was included as a time-updated variable in analyses. Therefore, if participants restarted ART after an interruption, they would be coded as not being on an interruption for the interval in which they resumed treatment and for each subsequent interval until such time as they interrupted in future.
The following variables were investigated as confounders of the relationship between ART interruption and liver fibrosis: age, gender, time since HIV diagnosis and duration of HCV infection, ART regimen, prior (failed) treatment for HCV, hepatitis B virus (HBV) co-infection, current injection drug use (IDU) or alcohol consumption, time-updated CD4+ T-cell count and HIV RNA, as well as nadir CD4+ T-cell count, highest recorded HIV viral load and APRI score at baseline. Baseline APRI scores were divided into three categories: APRI < 0.5, which is indicative of the absence of fibrosis and was used as the reference category, 0.5 ≤ APRI < 1.00, and 1.00 ≤ APRI < 1.5. The date of HIV diagnosis was the date of first HIV test positive test. The duration of HCV infection was determined using the date of HCV seroconversion, if known, or year of first IDU or blood product exposure as a proxy of HCV infection [28]. CD4+ T-cell count and HIV viral load were missing for 2% (n = 27 visits) and 3% (n = 39 visits) of visits respectively. For these visits, the value from the next closest visit was carried forward or backward.
Outcome assessment
The outcome of interest was the development of significant liver fibrosis, defined as an APRI score at least 1.5. This cut-off has been shown to be predictive of significant fibrosis [23,24,29,30]. A sensitivity analysis was performed using clinical liver disease diagnoses or an APRI score at least 2.0, which is predictive of cirrhosis. The presence of any one of the following was considered to be an end-stage liver disease diagnosis: cirrhosis, ascites, esophageal varices, spontaneous bacterial peritonitis, portal hypertension, encephalopathy or hepatocellular carcinoma. Significant liver fibrosis was chosen as the primary outcome of interest rather than the more advanced endpoint of end-stage liver disease because few participants were expected to experience such an outcome during the relatively short follow-up period.
Concurrent measures of AST and platelets were used to calculate APRI scores using the following formula: APRI = 100 x [AST (U/l)/upper limit of normal]/platelet count (109/l) [30]. An APRI score could not be calculated due to missing AST or platelet count at 2% (n = 26 visits) of follow-up visits. For these visits, it was assumed that an outcome did not occur.
Statistical methods
Data were analyzed using R statistical software version 2.9.2. Baseline and follow-up characteristics were compared between those who interrupted ART and those who did not using the χ2 or Fisher's exact test for categorical variables and the Wilcoxon-rank sum test for continuous variables. Significance tests were carried out assuming two-sided alternative hypotheses and 0.05 level of significance.
Multivariate time-dependent Cox proportional hazards regression models were constructed and included covariates that had statistically significant hazard ratios in univariate analyses as well as variables that were determined a priori to be clinically significant. The final multivariate model included ART interruption, time-updated CD4+ T-cell count and HIV viral load, age, gender and baseline APRI. Robust variance estimation was used for all Cox regression analyses to account for the correlation of data contributed by the same participant at multiple visits.
Adjusting for time-varying confounders, such as CD4+ T-cell count and HIV viral load that are also intermediate variables in the causal pathway using a standard Cox model could result in a biased estimate for the effect of ART interruption on development of significant fibrosis (Fig. 2) [31]. In order to appropriately adjust for these variables, inverse probability-of-treatment weighting (IPTW) was used in a marginal structural model [32]. Time-updated CD4+ T-cell count, HIV viral load and age were accounted for through the IPTW process. The model included ART interruption as well as gender and baseline APRI score as regressors. Odds ratios (ORs) were estimated using weighted pooled logistic regression which approximates Cox regression when the risk of an event is less than 10% per person-time interval [33]. In this case, the risk of achieving an APRI score at least 1.5 was 3.5% per 6-month interval. Nonparametric percentile-based bootstrap confidence intervals (CIs) were calculated by resampling with replacement from the observed data.
|
|
| |
| |
|
|
|
