Roche was requesting FDA approval for the first two uses (prognosis and therapeutic monitoring) for their PCR test kit, but acknowledged that there is not yet data to support claims for the third use. The Amplicor PCR test kit can reliably measure viral RNA particles in the range of 400 to 750,000 copies/mL. It has no cross-reactivity with viruses, bacteria, mycobacteria and protozoans which commonly affect HIV-infected patients. It can reliably detect 5-fold differences (about .7 log) with 95% confidence.
To support the claim for prognostic value, the company submitted data from patients in virology substudies of ACTG 116A (N=153, sub-study only) and 116B/117 (n=99, sub-study only). In ACTG 116A, baseline RNA was clearly associated with risk of an opportunistic infection or death (p=0.02) as was the log CD4+ count (p=0.0001). In ACTG 116B/117, however, only log CD4+ count and AIDS diagnosis at baseline were associated with risk of disease progression--not baseline RNA copy number. (This failure is very likely explained by genotype and phenotype alterations in the virus as a result of extended AZT exposure; median duration of prior AZT in 116B/117 was 14 months). A separate series of supportive data, however, presented by Drs. Robert Coombs and Douglas Mayers, convinced the committee that the Roche PCR test does have prognostic value.
With data from an ACTG combination study of saquinavir, AZT and ddC, (ACTG 229) Roche argued that its PCR test could reliably quantify RNA changes following initiation of antiretroviral therapy. (In the 229 study, for instance, patients taking triple drug combination therapy (AZT+ddC+SQV) had RNA reductions of about 0.8 log, whereas patients taking AZT+ddC had only a 0.3 log reduction, and patients taking SQV+AZT had only a 0.2 log reduction.)
In addition, Roche representatives presented data from another saquinavir study NV14256, which compared ddC and SQV monotherapy to the two drugs together in relatively heavily pre-treated patients. Patients taking saquinavir alone experienced only a 0.1 log reduction in RNA, while patients taking ddC had about a 0.3 log reduction. Patients taking the combination regimen, however, showed approximately a 0.6 log reduction in RNA. In both studies, 229 and NV14256, there was a strong correlation between HIV RNA levels as measured by the Roche Amplicor PCR test kit, and measurements using the quantitative PBMC microculture methodology.
Again citing the ACTG 116A data, Roche also argued that post-therapeutic changes in RNA levels could predict clinical outcome. Indeed in 116A, there was a significant correlation between log change in viral RNA at week 8, and clinical outcome (p=0.013). However again, in 116B/117, there was no significant correlation between these two variables.
Dr. Coombs gave an eloquent presentation, summarizing further data from ACTG studies, and discussing the criteria for distinguishing a valid surrogate marker. According to Dr. Coombs, no surrogate marker has ever predicted 100% of the treatment effect. So the difficulty, he explained, lies in establishing how much of the treatment effect can be predicted by the surrogate marker. This effort is further complicated by the fact that surrogates do not usually measure toxicity. Consequently, if a surrogate "really" predicts, for example, 50% of the treatment effect, but toxicity also reduces the treatment effect by 50%, then the surrogate will predict all of the treatment effect in that one study, making it appear much more useful as a predictive tool than it really is.
Finally, the committee agreed that the Roche PCR test could reliably quantify response to antiretroviral therapy. The committee also agreed, however, that Roche had not yet established the clinical significance of these viral RNA changes. If, for example, a patient had a high viral load and a high CD4+ count, no one knew what the patient should do.
The meeting ran late, discussion was very limited, and no formal vote was ever taken. As a consequence, many attendees left the meeting uncertain of the committees recommendations. Later calls to the FDA also suggested confusion as to the recommendation. According to Roche's press release, both the prognostic and therapeutic monitoring applications of the Amplicor test kit were recommended for approval. FDA has not yet acted on the committee's recommendation.