Retrospective Observational Study Comparing Efavirenz to Nevirapine and PI Regimens in 800 Patients in Two Clinics in London

"Comparison of First Line HAART Success in a Large Cohort Analysis of ARV Naïve Patients (PI vs NNRTI)"
     Matthews, Sabin, Mandalia, Lampe, Phillips, Nelson, Bower, Johnson, Gazzard, Chelsea & Westminster Hosp. London; Royal free Hospital, London

At the 7th BHIVA in April researchers from England reported on a retrospective analysis (a look back at patient records) in two hospitals in London (Chelsea & Westminster and Royal free Hospital). This type of study has certain limitations compared to a randomized prospective clinical trial. The purpose of this study was to compare efavirenz, nevirapine and single & double protease inhibitor regimens for superiority in patients who had never before had treatment. The study looked to see which regimen was more successful in achieving undetectable viral load (<500 copies/ml) by 6 months after starting therapy.

From an observational database of over 8000 patients from the two hospitals, 1109 patients were identified for this analysis (726 from Chelsea & Westminster, 383 from Royal Free). 221 patients were excluded due to lack of follow-up viral load test results leaving a total of 888 patients for this analysis. 55% or 484 received a PI regimen for initial therapy while 45% or 404 patients received a NNRTI (NVP or EFV) regimen to begin their first regimen of HIV therapy. Dual PI regimens were included if ritonavir was used. Patients taking saquinavir hard gel were excluded, patients had to have a minimum of 6 months follow-up, and CD4 and HIV viral load test results within 6 months of starting HAART. 237 patients received NVP and 167 received EFV.

The study evaluated outcomes with two analyses: time to initial response <500 copies/ml and a "composite failure" endpoint. They did not report the Kaplan-Meier curve showing time to failure (which regimen reached undetectable first). And they did not define what they considered in the composite failure endpoint, but presumably it included not reaching undetectable, rebound above 500 copies/ml, lost to follow-up, discontinuation due to side effects/toxicities, and withdrawals for other reasons.

In their analysis, they considered a number of variables that may have affected outcome: age, gender, ethnicity, treatment center, year of therapy start, year of HIV infection, prior AIDS defining event, baseline CD4, viral load and drug therapy.

The final report at BHIVA compares NVP to EFV to all protease inhibitors. This is one of the study limitation. In other words. they did not compare individual protease inhibitors to NVP or EFV. All PI regimens were lumped together. So you don’t have a comparison of EFV to indinavir, nelfinavir, or to ritonavir+indinavir.

Most PI regimens consisted of either IDV or NFV: 200 taking indinavir, 172 NFV, 50 RTV, 41 RTV/IDV, and 21 Fortovase. Interestingly, the researchers reported drug prescriptions in 1997 showed about 70% of patients received a PI regimen vs 20% in 1999, showing a sizable change towards using more NNRTIs. Patients receiving a PI had more prior AIDS defining events before starting therapy, while for EFV vs NVP it appeared to be about 19% vs 15%, respectively. So, patients receiving a PI may have been more sick. CD4s and viral load at baseline were 169 CD4s and 5.3 log (200,000 copies/ml) for PI, 183 CD4s and 5.05 log (112,000 copies/ml) for EFV, and 217 CD4s and 5 log (100,000 copies/ml) for NVP. So, CD4s and viral load were a little better for NVP prior to starting therapy, and a little worse for PI regimens.

RESULTS

Baseline viral load was a significant predictor of viral load suppression. The higher a person’s viral load the less likely to achieve undetectable. The authors said white ethnicity and age varied significantly but they didn’t show the data.

Baseline Viral Load
<4.71, RH 1.00
4.71-5.15 log, Hazard Ratio (RH) 0.90
5.16-5.57 log, 0.75 RH
>5.58 log 0.68 RH

There was considerably more ddI use in the NVP arm than in either the EFV or PI arms (50% vs 20%). There was more 3TC use in the PI and EFV arms than in the NVP arm (75-80% vs 52%). The types of side effects and toxicities patients experienced were not reported. If a toxicity or side effect specifically related to one NRTI was experienced, this was not reported. Hepatoxtoxicity was not reported. And adherence was not reported. These are some of the limitations of this analysis. There are a number pluses and minuses when using a randomized clinical trial (RCT) vs an observational retrospective study. One striking example is that patients are not randomized in the observational retrospective analysis and doctors generally place them on the regimen they feel may be most approriate for the patient. Patients in a RCT may all be relatively similar in terms of disease stage and other factors, but when looking at an entire cohort or patient group in a clinic the patient’s disease stage will be very variable, and the regimen selected for that patient may be because of their disease stage & other reasons.

In general RCTs are considered the best way to collect data and have less bias, but using a retrospective observational cohort can have some benefits. It may be easier and quicker to collect data this way. And the drug manufacturers may not want to conduct a RCT comparing their drug to another.

The authors will be presenting this study again in Buenos Aires and may evaluate response by comparing individual protease inhibitors to EFV & NVP.

At 6 months on therapy (estimate of initial response by Kaplan-Meier at 6 months) the cumulative % of patients with <500 copies/ml: 92% on EFV, compared to 83% for NVP, and 79% for all PI regimens (<0.0001). When the 3 study regimens were evaluated by looking at the composite failure endpoint and treatment switches were considered as failures, 208 patients (43%) receivi ng a PI regimen, 35 (21%) receiving an EFV regimen, and 93 (39%) receiving a NVP regimen reached the composite endpoint at 6 months (<0.001). When treatment switches were ignored: 127 (26.2%) in PI arm, 19 (11.4%) in EFV arm, and 52 (21.9%) in NVP arm reached composite endpoint failure at 6 months (<0.001). So, it appears to me that there were more treatment switches due to "composite endpoint" failure in the PI & NVP arms.

The authors concluded that EFV had an advantage by all analyses at 6 months; PI and NVP regimens performed equivalently (but, they did not compare individual PI drugs to NVP or EFV regimens). They also said better initial EFV response may reflect less short term toxicity (or maybe the doctors & patients were able to manage certain side effects better), and an observational cohort may be susceptible to bias. And they said that boosted PI regimens using RTV are increasing in use.

In the program abstract they said: "a comparison of NVP vs EFV suggested a treatment benefit for EFV (relative hazard [RH] 0.81 for NVP); in multivariate analysis, EFV was associated with increased likelihood of success over both PI and NVP regimens (RH 0.71 for PI, p=0.007); RH 0.78 fot NVP, p=0.04), after controlling for all variables including start year and treatment center. They reported no effect was seen for baseline CD4, age, sex, ethnicity or prior AIDS defining event.