Abstract
Most studies comparing the response of African American (AA) patients to other groups for treatment of hepatitis C virus (HCV) have been performed in academic centers in the setting of clinical trials. In general, AA patients have lower response rates that are only partially explained by unfavorable prognostic factors, such as infection with genotype 1, high viral load, insulin resistance, obesity, hypertension, and others. We examined response rates associated with these adverse prognostic factors within race/ethnicity-sex strata for a cohort of patients treated at a inner-city hospital-based, safety-net clinic. From 2004-2008, 88 evaluable AA and 30 other patients received standard therapy for HCV consisting of weekly injections of peg interferon alfa-2 plus oral weight-based ribavirin. The predicted probability of a favorable virologic outcome ranged from 89.0% (95% CI: 63.7% = 97.3%) for other females to 4.1% (95% CI: 0.9% - 17.9%) for AA females. Conclusion: Important interactions occur among race/ethnicity, sex, and co-morbidity in response to HCV therapy. Increasing co-morbidity places AA patients at greatest risk of non-response and raises the possibility that more aggressive treatment of co-morbidity might increase response rates in the vulnerable patients.

Introduction
Although the number of Americans with HCV infection has declined about ten-fold from the 1980s to 2001¹, estimates are that as many as four million persons have been infected with HCV. No vaccine is available to prevent HCV, unlike for Hepatitis A and B, and carriers can transmit the disease via blood and body fluids throughout the remainder of their life. HCV has an enormous effect on morbidity and mortality. Perhaps eighty percent of infected people in the United States show signs of chronic infection, which over a period of 5-20 years may lead to liver cirrhosis (20%) and or liver cancer². Moreover, approximately 25% of patients with cirrhosis will die from hepatic failure or require liver transplantation. To some extent, treatment of HCV may slow progress of fibrosis.

The current standard for treatment of HCV is pegalylated interferon alpha 2 (PEG-IFN) in combination with ribavirin (RBV)³, which has improved clinical efficacy over single agent PEG-IFN. The overall sustained viral response rates following combined PEG-IFN+RBV is 54 - 56% after a 48-week course of therapy. Patients with genotype 1 infection typically have a 40 - 50% overall likelihood of response to 48 weeks of therapy.⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹ Those with genotypes 2 or 3 infection can expect to experience an improved response to treatment, with a 75 - 80% SVR after only 24 weeks of treatment.⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹ Other conventional predictors of low responsiveness, such as high initial viral load or advanced liver fibrosis, affect the response to PEG-IFN less than they do with non-PEG IFN⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹

Although treatment is available for HCV, we lack a sound understanding of the complex interaction between race/ethnicity, comorbidity, and treatment response. For reasons that are not entirely clear, the sustained virological response (SVR) rate in AA patients was less than the response rate for other patients in two multicenter studies.¹², ¹³ To investigate these racial/ethnic differences more closely, we determined the overall antiviral response within a specified four year time period in a community based safety net setting, with emphasis on the interactions of race/ethnicity, sex and co-morbidity.

Methods
We monitored the antiviral response for the cohort of all patients with HCV genotype 1 entered into treatment in the Hepatitis Clinic at Cooper Green Mercy Hospital (CGMH), an inner-city public hospital in Birmingham, Alabama from January 2004 thru July 2008. Patients with hepatitis not due to HCV (HBV, alcoholic, autoimmune, others) were excluded. For all treated patients, we captured patient demographics, initial and subsequent viral loads, decision to treat, risk factors for drug response, and adverse effects. Before treatment was offered, all patients were engaged in an extensive education program including discussion with staff, formal classes, and brochures and videos illustrating the natural history of HCV, treatment options, anticipated response to treatment, and potential adverse effects. A directed search of the patient data was used to prepare this publication; no change in therapeutic decision was based upon the collection of this data. The CGMH Institutional Review Board (IRB) approved the use of anonymous demographic and incidence data collection for the purpose of preparing this publication.

Patients were offered treatment if their initial viral load was > 400,000 IU, they were symptomatic for HCV with chronic fatigue, weakness, abdominal pain, nausea), and they had elevated liver enzymes (AST > 2 times upper limit of normal). Patients not meeting these three criteria could still be considered for an individual treatment decision, if after an explanation of why they were not optimal candidates for treatment they still desired to be treated.

Patients received either PEG-IFN alpha 2a (180ug) or 2b (weight based) once per week, with weight based ribavirin twice a day (unless contraindicated). We did not show preference for the version of Peg-IFN offered other than that dictated by individual insurers. Patients were checked at 4 weeks of treatment when possible for a rapid virological response (RVR), and also at 12 weeks for an early virological response (EVR). Patients who were virus negative at 12 weeks of treatment had their medication continued for the complete 48 week duration. Patients who experienced at least a 2 log drop in viral load from baseline at 12 weeks of therapy but were not virus negative were continued to 24 weeks of therapy, and then if virus negative, were continued for 48 weeks of treatment, otherwise treatment was stopped.

Between three to six months after the end of a full course of treatment, a quantitative viral load was again measured (whenever the opportunity was available), to determine whether a SVR occurred. Not all patients made themselves available for all study time points, and not all patients started on therapy had finished therapy at the cutoff for collection of study data.

For this study, virologic response in patients was defined as meeting any of the following three criteria:

No detectable virus 3-6 months after stop of therapy (sustained viral response SVR);Completed treatment with undetectable virus at end of treatment (EOT) even though no follow up assessment of virologic status was made at 3-6 month post-treatment; or Still undergoing treatment at the time of data analysis with at least one favorable indicator of viral response: RVR or EVR.

We added the second and third criteria to reflect the real-world setting of a safety-net hospital. The second response option allowed us to include patients lost to follow-up after having responded up to EOT. The third criteria broadened the definition of Virologic Response to allow an overall picture of response rate in our patient population. Although these criteria are broader than those commonly used in randomized clinical trials, the more over-encompasing definition of response rate allowed us to analyze our single site data set to better understand within the purpose of our study the complex interactions between race/ethnicity, sex, and comorbidity with treatment response.

Non-response (NR) in patients was defined as either:

Did not experience at least a 2 log drop in virus titer at 12 weeks, or Experienced at least a 2 log drop in virus titer at 12 weeks, but who were not virus-negative at 24 weeks. Patients with detectable virus 3-6 months after cessation of therapy after an initial response (EOT) were classified as experiencing Relapse and were included for data analysis along with the non-responder group. Patients with a drop in viral load during treatment, followed by an increase while still on therapy were classified as experiencing Breakthrough, and also were were included for data analysis along with the non-responder group.

Patients were excluded from the analysis if they were:

Not treated,Lost to follow up,Discontinued treatment due to an adverse effect of medication (AE) before a determination of viral response could be made,Deceased for unrelated reason before a determination of viral response could be made.

Statistical Analysis
We first examined distributions and univariate statistics for the overall sample. Next we performed bivariate analyses to study the associations among patient characteristics and response status using two-sample t-tests and chi-square tests with p-values = 0.05 considered significant. Because of small cell sizes, exact tests of statistical significance were used for comparisons based on categorical independent variables.

The presence of diabetes mellitus (DM) and hypertension (HTN) were highly collinear, requiring these co-morbidities to be represented as a single, composite categorical variable. More specifically, we constructed a three-level categorical variable for the following states: (1) absence of both DM and HTN (no co-morbidity), (2) presence of either DM or HTN, and (3) presence of both DM and HTN.

Taking a dichotomous variable indicating virologic response as the patient-level outcome, multivariable logistic regression determined the independent associations for multiple clinical covariates. Because of the small number of outcomes, we exercised special caution to prevent over fitting. In particular, we omitted age and initial viral load from the final multivariable model because these variables were not significant in interim multivariable analyses. In addition, we entered the comorbidity score as a linear variable to conserve degrees of freedom and because the relationship between the comorbidity score and the outcome was monotonically decreasing. The power of the main logistic regression model to discriminate between the presence or absence of the outcome state (response to treatment) was examined with the c-statistic.

Based on the final multivariable model which included race/ethnicity, sex and the comorbidity score, we calculated the predicted probability of a favorable viral response. For each predicted probability, non-parametric re-sampling without replacement (bootstrapping) generated 95% bias-corrected and accelerated confidence intervals. Because of small cell sizes, predicted probabilities for certain patient profiles were not reported. All analyses were performed with STATA 10.0 SE.

Continue to full article and footnotes