New tool developed to predict success of hepatitis C therapy in HIV/HCV-co-infected patients

Spanish investigators have developed a non-invasive diagnostic tool that can accurately predict which HIV-positive patients who are co-infected with hepatitis C virus will successfully respond to treatment for hepatitis C. Their findings are reported in the November 15^th edition of Clinical Infectious Diseases.

Called Prometheus, the diagnostic tool can be downloaded for free here.

Four variables were included in the investigators’ model. Two of these – hepatitis C genotype and viral load – are well-known predictors of treatment outcome. The researchers also included two novel diagnostic markers: liver stiffness and genetic variations near the IL28B gene.

All these markers can be monitored without the need for an invasive liver biopsy, which the investigators believe makes their model especially attractive for use in routine care.

It is estimated that approximately 20% of HIV-positive individuals are co-infected with hepatitis C. In the era of modern HIV treatment, liver disease caused by hepatitis C is a leading cause of illness and death in co-infected patients.

Treatment for hepatitis C is available. Involving pegylated interferon and ribavirin, its aim is an undetectable hepatitis C viral load six months after the completion of therapy. This is often called a sustained virological response (SVR), and only a third of co-infected patients achieve this outcome.

Poorer outcomes are seen in patients infected with hepatitis C genotypes 1 and 4 (as opposed to 2 and 3), and in individuals with a higher hepatitis C viral load at the start of therapy.

Recently, genetics have also been identified as predicting treatment response. Improved outcomes were seen in patients with a variation in the polymorphism rs12979860 close to the IL28B gene. Patients with the CC genotype were twice as likely as individuals with the CT or TT genotype to have a successful treatment response.

Assessment of liver stiffness using FibroScan has also been shown to be an accurate way of monitoring liver damage in patients with hepatitis C. This method of monitoring involves placing a small device against the patient’s skin and therefore, unlike liver biopsy, does not require surgery.

Therefore, the investigators developed a predictive model including these four variables.

Their main study population included a cohort of 159 co-infected patients who received hepatitis C therapy at the Hospital Carlos III, Madrid, between November 2004 and December 2008. To validate their results, the diagnostic tool was also tested on 86 co-infected patients who received treatment at two other clinics.

The two cohorts were generally well matched. However, individuals in the validation population had a higher baseline hepatitis C viral load (p < 0.001) and lower liver stiffness scores (p = 0.03).  Overall, approximately 55% of patients in both cohorts had a sustained virological response.

However, the investigators stress that they excluded “nonadherent patients and patients who discontinued therapy because of side-effects and only evaluated the subjects who had completed the course of therapy”. Moreover, 99% of patients were male and the average body mass index (BMI) was well within the normal range (23 kg/m²).

In both groups of patients, hepatitis C genotype, liver stiffness, the CT or TT polymorphism and hepatitis C viral load were predictive of treatment outcome.

The strongest predictor of outcome was CT or TT polymorphism versus CC (odds ratio [OR] = 0.170; 95% CI, 0.065 to 0.442; p < 0.001).

 This was followed by baseline hepatitis C viral load (OR, 0.186; 95% CI, 0.091 to 0.381; p < 0.001), genotype 1 and 4 versus 2 and 3 (OR, 0.212; 95% CI, 0.078 to 0.577; p = 0.002), and increased liver stiffness (OR, 0.920; 95% CI, 0.865 to 0.977; p = 0.007).

Using these data, the investigators developed their tool to predict treatment response.

They used the control cohorts to test the accuracy of their model at three cut-off points. These assessed different degrees of sensitivity and specificity.

Using the lowest point (highest sensitivity and lowest negative predictive value), a lack of treatment response was correctly predicted in 89% of patients.

A similar accuracy – 90% – was seen when the investigators used the highest cut off (highest specificity and positive predictive value).

An intermediate cut off (highest sensitivity and specificity) accurately predicted a successful treatment response in 83% of patients and a lack of response in 79%.

Therefore, they assessed the accuracy of their model as between “excellent” and “good”.

The investigators believe that their tool “may be of great value for making adequate therapeutic decisions for HIV-HCV-coinfected patients”.

For example, use of current hepatitis C therapy could be encouraged for patients with the greatest chance of responding to treatment.

Those with a low chance of sustained virological response could be counselled to “wait for new direct-acting antiviral agents against HCV”.

Although the investigators are encouraged by the accuracy of their model, they nevertheless caution, “as with any diagnostic tool used in clinical practice, misclassification may occur, and understanding the limits of estimated outcomes with predictive indexes is important before its widespread use.”