Antiretroviral Therapy
(Part XXIV)

Viral Load and Genotypic Resistance Assays

Validation of HIV-1 RNA and CD4 count as surrogate markers in the CEASAR trial

The CEASAR trial examined the effect of adding lamivudine (3TC), loviride (an experimental non-nucleoside reverse transcriptase inhibitor), 3TC plus loviride or placebo to a zidovudine (AZT) containing regimen. The addition of a 3TC containing regimen to AZT had a dramatic effect in delaying disease progression.

In this analysis, Montaner and colleagues studied the value of a change in viral load and/or CD4 cell count in predicting clinical outcomes. This sub-study analyzed 794 patients. Surprisingly, the treatment induced effects on CD4 and HIV-1 RNA levels at weeks 8 to 28 predicted most, if not all, of the eventual clinical outcome (predictive value 100%, lower 95% confidence interval 56%) [Montaner, abstract 61].

Comment: Until recently, "body count" studies, in which survival was the only relevant outcome, was the gold standard in large scale clinical trials. The results from this analysis indicate that studies designed to analyze surrogate endpoints only are probably sufficient, and that large, expensive clinical endpoint studies are no longer necessary. Licensure of new drugs by the Food and Drug Administration need not bdepend on evidence that the drug delays disease progression or prolongs life.

Detection of HIV-1 RNA in the plasma of patients in whom HIV-1 RNA is undetectable using commercial assays

Current standard of care is to treat HIV aggressively, with an explicit goal of attaining "undetectabe" levels of plasma HIV 1 RNA (typically defined as a viral load less than 200 to 500 copies/ml). The natural history of patients who achieved undetectable levels remains unclear.

In a study reported by Natarjan and colleagues, an ultrasensitive viral load assay (RT-PCR, lower level of detection 50 copies/ml) was used to better characterized patients with 5 months of documented viral load levels less than 500 copies/ml (Chiron, bDNA assay) or 400 copies/ml (Roche, RT-PCR assay). Most patients were receiving aggressive combination therapy. Of the 72 patients studied with the ultrasensitive assay, 50 had detectable levels (>50 copies/ml) [Natrajan et al, abstract 51].

Comment: Plasma HIV-1 RNA can be readily detected in the majority of patients with so-called "undetectable levels" using current assays. This suggests that many patients have persistent low level viral replication, despite the use of combination therapy. Whether these patients are more likely to eventually fail a given regimen remains to be determined. Biologically, one can assume that a patient with a viral load < 50 copies/ml is more likely to achieve long-term viral suppression than a patient with a viral load of 50 to 500 copies/mL. Once these ultrasensitive assays are commercially available, clinicians should be able to identify patients at high risk for failing a given regimen, and therefore be able to intervene before high level resistance develops.

The durability of response to protease inhibitor therapy is predicted by viral load

Clinicians clearly need tests which will allow them to predict who will fail a protease inhibitor, so modifications can be made before high level resistance develops. Kempf and his colleagues from Abbott reviewed clinical data from several ritonavir studies to determine which virological markers predict a durable response to protease inhibitor therapy.

To be eligible for their initial analysis, patients must have reached a documented viral load "nadir" (i.e., their lowest viral load), and then rebounded with a virus containing at least two resistance mutations. Of the 29 patients studied, 6 received ritonavir monotherapy, 16 received zidovudine (AZT) plus ritonavir, and seven received the combination of ritonavir and saquinavir. Baseline RNA did not predict how long the patient would respond to a given therapy. The absolute value of the viral load nadir was highly predictive of the durability of viral suppression (R2=0.885): the lower the nadir, the longer the response. Patients who achieved levels below the level of quantification (< 200 copies/ml) had the longest response.

To confirm this observation, Kempf and colleagues looked at 202 subjects who received at least 30 days of ritonavir or ritonavir plus AZT, and who had at least a 0.5 log decrease in viral load. Subjects who discontinued therapy were excluded form the analysis. Of the 152 patients eligible for analysis, baseline viral load at the time ritonavir was initiated did not predict duration of maximal response. Once again, subjects who achieved a viral RNA < 200 copies/ml had a durable response, while those who never went below 1000 copies/ml rebounded back toward baseline quickly. Most subjects reached their nadir by week 12.

Notably, many patients who went below 200 copies did eventually rebound. It is unclear if more sensitive assays (lower limit of detection 20 copies/mL) could more accurately predict outcome [Kempf, abstract 62].

Comment: This data suggest that the viral load at week 12 of therapy with a protease inhibitor is highly predictive of outcome. Therefore, clinicians may want to modify or change therapy at week 12 if the viral load remains greater than 200 copies/ml.

Genotypic or phenotypic susceptibility testing may not predict clinical responses to indinavir

Should genotypic or phenotypic assays be performed routinely in the clinic? This critical question was another focus of the meeting. Several very promising assays are in development. Should they be used? Unfortunately, there is no convincing evidence that genotypic or phenotypic assays reliably predict response to subsequent therapy.

Condra and colleagues at Merck presented several case studies illustrating the limits of genotypic assays. In their experience with indinavir (Crixivan), subjects often rebound with clear genotypic changes but a normal phenotype. In each case, genotype predicted failure to indinavir prior to the development of detectable phenotypic resistance. Although the presence of phenotypic resistance likely predicts failure, it is unclear if "phenotypic susceptibility" predicts success.

Does a lack of genetic information for resistance predict success? Again, Condra presented case studies of a few subjects who stopped therapy after failing indinavir. Wild-type virus rebounded quickly, and the patient's genotype reverted to normal. Later, when indinavir was restarted, failure occurred rapidly, despite the absence of genotypic resistance at the time indinavir was started. Unfortunately, resistant variants often exist at very low levels, well below the level of detection. These minor "quasi-species" are selected for rapidly in the presence of a drug [Condra, abstract 47]. Therefore, the patient's prior treatment history may be a better predictor of outcome than genotype.

Comment: Genotypic and phenotypic assays are undergoing evaluation in several clinical studies. Until further data is available, viral load data and the patient's prior antiretroviral history should be the most important factors in developing therapeutic strategies.