Publications Archive

Clinical trials to clinical reality—
"Bridging the gap" between science and the patient

HIV+ Clinical trial volunteers live longer than HIV+ patients in routine care. How could careful scientific investigation fail to accurately predict outcomes in the clinic?

The media hype surrounding the success of protease inhibitor-containing regimens as reported at the XI World AIDS Conference in Vancouver has evolved into somber reports from Geneva on the realities of drug resistance, compliance managing, and adverse side effects in the general patient population. This "gap" is highlighted by a recent report which shows that participants of clinical trials live longer and progress more slowly to AIDS than the average patient in routine care (Ferrigno et al., 1997). Since the objective of clinical trials is to identify strategies that are effective in routine care, this disparity is cause for concern. How could carefully-controlled scientific investigation fail to accurately predict outcomes in the clinic?

At a satellite symposium of the XII World AIDS Conference sponsored by F. Hoffmann-La Roche Ltd., Dr. Melanie Thompson of the AIDS Research Consortium of Atlanta, Georgia, presented an excellent overview of the reasons why patients in clinical trials do better. She provided three general arguments. First, participants in many studies are not representative of the population in general. Second, trial sites extend services to patients that can factor into the level of care. Third, liberal analytical techniques can portray data in ways that appear to favor the regimen studied.

Selection criteria for trials are often biased towards patients likely to respond well to treatment. Inclusion criteria often require participants to be naive to antiretrovirals. Conditions that could make adverse effects more likely are frequently grounds for exclusion. The result is a relatively healthy population (even taking CD4 counts and initial viral load into account) that presents minimal drug resistance problems. There is also a self-selection effect; patients who volunteer tend to be more proactive with their treatment.

These selection biases imply that many of those who respond well in the context of a clinical trial would have responded equally as well in a routine treatment setting. There are also ways in which participation in a study actively improves the patient's care. Clinical trial sites provide diligent attention to volunteers, including follow-up phone calls, compliance monitoring, and prompt management of adverse effects as they arise. In general, sites have a vested interest in aggressively supporting their patients. This level of attention naturally has an impact on how participants handle their treatment regimen.

Another significant factor in evaluating an anti-HIV drug is the limit of detection of the viral load assay used in the study. The detection limits can vary by more than an order of magnitude, depending on which assay is used. For example, Chiron's "branched chain DNA" test can quantify viral load down to 400 copies per milliliter of plasma, although a new version of this test just released for investigational use can go down to 50 copies/mL. OrganonTeknika's NASBA test can detect down to 80 or to 400 copies/mL, depending on how much sample is used. Roche's FDA-approved Amplicor test can go down to 200 copies/mL, although a newer, experimental version of the test called the Ultra Sensitive Amplicor can quantify HIV at concentrations as low as 50 copies/mL.

A treatment regimen evaluated with a detection limit of 400 copies/mL might work well enough to be deemed successful, but the clinical utility of this "successful" regimen might be minimal. A recent report (Gunthard et al., 1998) has shown that there is evidence of low-level viral replication in patients whose viral load was less than 400 copies/mL, but higher than 20 copies/mL (measured with especially sensitive molecular techniques). However, patients with a viral load under 20 copies/mL showed no evidence of ongoing viral replication. Since even low-level replication is associated with the eventual development of drug resistance, one might expect substantial differences in the durability of patient responses to treatment when their viral load is below 20, as opposed to 400, copies/mL.

In a clinical trial, if a patient's viral load goes down to the assay's detection limit, the patient is considered successfully treated. If the patient began with a viral load only 10-fold higher than the detection limit, that patient may have experienced only a one log drop in viral load. As a result, the number of trial participants for whom a regimen is successful can depend substantially on the assay used to test viral load. Certainly, an average patient in routine care can achieve a degree of viral reduction consistent with that seen in studies and yet not go below detectable limits.

In addition, clinical trial reports utilize different methods of data analysis, which materially affects the reported outcome. "On-treatment" analyses consider only the patients that actually receive treatment for the full course of the study and tend to show the possible, not necessarily probable, benefit of the treatment. This means that this analysis technique ignores the data from patients who were non-compliant, withdrew due to side effects, or simply dropped out of the study. On the other hand, the more conservative "intent-to-treat" analysis includes all patients assigned to treatment groups, even those who had to drop out during the study for any reason. Therefore, clinical trial results which present data utilizing an "intent-to-treat" analysis will more accurately predict the clinical response of the general patient population.

After Ziagen (Abacavir), Sustiva (Efavirenz),and Adefovir (bis-POM PMEA) receive FDA approval later this year (as expected), it will be possible to prescribe anti-HIV drugs in 1028 possible triple therapy combinations. When you consider that 4, 5 and even 6 drug combinations are becoming widely prescribed, the number of possible drug regimen configurations grows logarithmically. While some of these combinations will be studied in phase IV post-marketing trials, it would be impossible to study them all. Against this backdrop, patients and their physicians are faced with the challenge of constructing the optimal regimen to meet the patient's needs. The gap between the published clinical literature and the usual clinical experience complicates this task.

There has been irrefutable progress in the clinical management of HIV disease in the last several years, and this progress can be understood in terms of the expanded possibilities for treatments and our improved comprehension of the biology of AIDS. Between Geneva and the XIII World AIDS Conference in South Africa, the challenge will be to complete the picture and translate that progress into better routine care. In addition to showing what's possible to achieve in the treatment of HIV in clinical trials, we look forward to seeing data from these studies that better predict what is probable in the thousands of patients who will take these drugs after the trials are completed.

References
Ferrigno L et al. (1997) Survival and progression of disease in HIV-infected individuals in interventional treatment studies as compared to an observational cohort. 37th Interscience Conference on Antimicrobial Agents and Chemotherapy. Abstract I-179. Gunthard H et al. (1998) Human immunodeficiency virus replication and genotypic resistance in blood and lymph nodes after a year of potent antiretroviral therapy. Journal of Virology, 72:2422-2428.