EVALUATION OF EXPANDED ACCESS PROGRAMS
Scientists have only begun to formulate the questions that will need to be addressed to determine whether the benefits of very early access to AIDS drugs exceed the risks. Experience with the treatment IND program and with other forms of early access may help guide the evaluation process.
TREATMENT INVESTIGATIONAL NEW DRUGS
The treatment IND regulations issued by the FDA in May 1987 were greeted with great enthusiasm by persons with AIDS and their advocates. They viewed the regulations as a dramatic shift in FDA policies that would allow hundreds or thousands of patients to gain access to investigational drugs. Three years later, however, many AIDS activists consider the program a failure. Although six treatment INDs for AIDS-related drugs have been approved, the activists say that the treatment IND accomplishes too little, too late.
The underlying problem, described in Chapter 1, is that patients and FDA officials had very different expectations about what the treatment IND would accomplish. FDA officials regarded the rules as an opportunity to make drugs available to patients with life threatening conditions and no treatment alternatives, but only after the acquisition of clinical evidence that a drug was relatively safe and probably effective. Early charts produced by the FDA showed that approval of a treatment IND would be most likely for drugs nearing the end of the traditional phase 2 clinical trial; or, if the condition
This chapter is based on the presentations of Robert Temple, Jay Lipner, Lawrence Corey, Raphael Dolin, Bernard Lo, Jerome Birnbaum, and Susan Ellenberg.
was ''serious" but not life-threatening, the treatment IND might be approved during the phase 3 trial. AIDS patients, on the other hand, expected the treatment IND to make drugs available at a very early stage of development.
Problems with the treatment IND demonstrate the importance of early and consistent communication. Efforts to include patients, primary care physicians, and AIDS activists in the planning process might prevent similar problems from arising with the new parallel track program.
Drug regulators have indicated their willingness to begin parallel track programs concurrently with the beginning of phase 2 trials. Patients and their physicians must understand, however, that scientists may have very little information about the potential adverse effects of a drug at that time.
Consider, for example, a phase 1 trial involving 20 patients. Statisticians explain that even if no serious toxicities were observed in those 20 patients, the upper 95 percent confidence limit for the true serious toxicity rate would be 17 percent (as many as 1 in 6 patients might experience a severe toxic reaction). Even if the phase 1 trial were expanded to 40 patients with no adverse effects, the true toxicity rate could be as high as 9 percent.
If physicians observed one serious reaction among 20 patients in a phase 1 trial, the observed rate would be 5 percent, but the true rate could be as high as 25 percent. A patient considering enrollment in a parallel track protocol might feel very differently about a 1-in-20 chance of a severe adverse reaction than about a 1-in-4 chance.
Examples from the Past
A study in the early 1970s of a potential treatment for herpes encephalitis provides a more graphic illustration of some of the problems that can arise with expanded access protocols. Herpes encephalitis is a life-threatening infection of the brain caused by the herpes simplex virus. In the late 1960s and early 1970s, about half a dozen case reports in the medical literature described treatment of this disease with a new drug called 5-iododeoxyuridine (IUDR). Scientists had a strong rationale for IUDR's antiviral effects and there were no treatment alternatives, so the FDA quickly approved an
open-label IND. Infectious disease specialists came to regard IUDR as the treatment of choice for herpes encephalitis and administered it to more than 70 patients under the open-label IND.
After several years, however, investigators decided to take a closer look at the drug's effects. Despite the objections of some early pioneers in the field, they organized a small placebo-controlled, randomized study. (The pioneers had believed that a placebo-controlled trial would be unethical because of existing clinical evidence in favor of the drug.) The results of the placebo-controlled study were striking. Of the 12 patients who received IUDR, 9 had some evidence of serious toxicity (an estimated 3 to 5 died as a direct result of the drug's adverse effects on bone marrow). Moreover, autopsy results in 4 patients who received IUDR showed that the patients had virus in the brain at the end of therapy. This example shows that even those who are acknowledged experts in a medical field may have difficulty predicting the benefit/risk ratio for a new investigational drug.
The Target Population
Physicians enrolling patients in expanded access protocols also should understand the potential impact of demographic differences on the outcome of clinical trials. These differences are important first in interpreting risk based on phase 1 trials and later in comparing phase 2 clinical trials with corresponding parallel track protocols.
Conventional clinical trials generally have very specific requirements with regard to the health status of prospective subjects. Scientists want to be able to see the effects of an investigational drug with a minimum of interference from other drugs, from confounding diseases, or from other risk factors. In HIV infection, age and overall health status affect survival and the rate of progression of disease; nutrition, past smoking history, and past occupational status affect the frequency of disseminated opportunistic infections (especially fungal and mycobacterial infections). To control for these differences among patients, differences that could confound the findings of a study, most investigators attempt to make their study populations as homogeneous as possible. In contrast, parallel track protocols enable thousands of patients with diverse backgrounds and medical histories to get access to drugs after minimal testing in a highly selected subgroup. The incidence of adverse effects in patients weakened by repeated battles with Pneumocystis carinii and other microorganisms, or by the effects of intravenous drug abuse, could be very different from that observed
in the relatively healthier population characteristic of the phase 2 trial.
The recent controversy over ddI illustrates some of the problems that can arise when people try to make direct comparisons between expanded access programs and conventional phase 2 trials. In March 1990, a spokesman for Bristol-Myers Squibb told a reporter from the New York Times that the death rate among patients in the parallel track protocol for ddI was 10 times higher than the death rate among patients enrolled in phase 2 trials of the drug. Of the almost 8,000 patients enrolled in the expanded access program, 290 had died; in contrast, only 2 of 700 patients in the phase 2 trials had died.
These figures were used to illustrate the dangers of expanded access, and physicians received hundreds of telephone calls from worried patients. Yet initial reviews of the data indicated that most or even all of the disparity might be due to the fact that patients in the expanded access program were sicker to begin with than patients in the clinical trials. With the exception of six deaths from pancreatitis (five in the expanded access program and one in the phase 2 trial), the deaths seemed to result from natural progression of the disease. (However, complete data from both protocols have not yet been published.)
The expanded access protocols for ddI were designed for patients who could not meet eligibility requirements for participation in the clinical trials. The majority have been intolerant of or unresponsive to zidovudine (AZT). Many cannot participate in the conventional trials because they are taking medications such as gancyclovir to fight severe opportunistic infections. Others are too sick to undertake the time and travel commitments necessary for participation in a traditional drug trial. All of these characteristics are indicative of progressive disease.
The proposed policy statement on the parallel track developed within the Public Health Service indicates that all physicians participating in a parallel track protocol should be required to report safety data (the collection of efficacy data depends on the specific protocol; see Chapter 4). Some observers have suggested that this requirement is too stringent—that mandatory reporting should not be part of a program designed primarily to increase access to therapy for desperately ill patients with no treatment alternatives. They worry that even minimal reporting requirements will discourage participation
by physicians and by drug sponsors (who assume financial responsibility for data collection).
Although these concerns are genuine, they represent a failure to appreciate the risks involved in widespread use of minimally tested agents. Failure to request reports of serious events in the parallel track could delay recognition of severe side effects for months or even years, increasing potential risks for patients enrolled in both research and open protocols.
IMPACT ON CONVENTIONAL RANDOMIZED TRIALS
In addition to worries about safety, critics of the parallel track have expressed great concern about its effects on enrollment in traditional clinical trials. Randomized, controlled trials provide definitive information about the relative risks and benefits of new therapies. Sponsors must have such information to seek approval for marketing from the FDA. Marketing, in turn, is the most efficient way to make a drug available to large numbers of people.
The critics are concerned that patients will enter the parallel track to avoid the uncertainty of a randomized trial. They refer to news articles about individuals who have sought outside help to identify drugs they receive in blinded studies as evidence that patients will not participate in a traditional clinical trial if other options are available. Supporters of the parallel track believe that current screening mechanisms are sufficient to separate people who are eligible for clinical trials from those who are eligible for the parallel track. They suggest that improved patient education and better trial design (see Chapter 4) will ensure patient accrual in the randomized trials and, at the same time, allow patients who are ineligible for clinical trials to receive experimental drugs through expanded access programs.
The experience with ddI highlights some early mistakes. Bristol-Myers Squibb received the exclusive license for ddI from the National Cancer Institute (NCI) in mid-1988. During the late summer and early fall of 1988, the NCI and Bristol-Myers Squibb sponsored four separate phase 1 trials of the drug. These trials produced a reasonable expectation that ddI was efficacious against HIV and that it was safe enough to allow expansion of a clinical program.
The company worked with the NCI, the National Institute of Allergy and Infectious Diseases (NIAID), and the FDA to develop appropriate phase 2/3 trials. At the FDA's request, the company submitted a proposal for a treatment IND; it also worked closely with AIDS patients and their representatives to develop a prototype
parallel track protocol. The FDA approved the phase 2/3 trials and the expanded access protocols on September 28, 1989. The first patient was placed on the expanded access protocol on October 12. The drug was shipped to the first phase 2 investigators on October 11, but actual enrollment of patients did not begin until October 20.
Many observers believe that the lag time between the start-up of the parallel track and the beginning of the phase 2/3 trials may have distorted the enrollment process. Patients who were eligible for the clinical trials entered the parallel track because ddI was not available any other way (excessive expectations for ddI created by the news media may have been partly responsible for this problem). An important lesson for future parallel track programs has been learned, that is, to ensure that enrollment in clinical trials precedes or coincides with the release of drugs through the parallel track protocol.
To this day, much controversy remains over the question of whether or not the availability of ddI through the parallel track has adversely effected accrual to the three clinical trials of ddI (protocol 116, a comparison of AZT versus ddI in recently diagnosed AIDS and ARC patients; protocol 117, a comparison of AZT versus ddI in patients who have been on AZT for longer than 12 months; and protocol 118, a dose-escalating trial of ddI in AIDS-intolerant persons). Views range from the belief that the expanded access program had very little effect on accrual to the trials, to the belief that the expanded access program has had a disastrous effect. As of January 1991, there were more than 14,000 patients receiving ddI through expanded access programs and approximately 1,600 patients participating in the trials, which are about 75 percent filled. Most scientists associated with the trials seem to feel that some eligible patients must have been diverted from the trials by expanded access. They assert that, although the rates of accrual to the ACTG trials of ddI have been similar to rates for trials of other AIDS drugs, without the competing availability of ddI through expanded access these trials would have recruited patients much more quickly.
THE PARALLEL TRACK EXPERIMENT
Although expanded access has been part of the drug evaluation system for many years, the formal parallel track approach is new. The proposal for the approach states that the entire parallel track program for AIDS drugs should be regarded as a pilot test. Many investigators would like to see mechanisms built into the program to answer questions about differential toxicity rates and about the impact
of the parallel track on clinical trials. The Public Health Service plans to organize a working group to assess the parallel track concept, but their task will be quite difficult unless data collection and certain evaluation strategies are incorporated into parallel track protocols from the beginning.
The extent of data collection efforts to be included in the parallel track is a complex issue. Ideally, basic demographic data and some clinical indicators of efficacy and toxicity should be collected in a format similar to that used for conventional clinical trials. It may be difficult, however, to convey state-of-the-art staging information to the broad spectrum of physicians who wish to enroll patients in the parallel track. Given that the primary goal of the parallel track is to make drugs accessible to desperately ill patients, sponsors must strive to obtain basic information without discouraging physician participation.
One option may be to arrange for a subset of patients in the parallel track to be followed by persons or institutions who are also involved in conventional trials of a candidate drug. Academic medical centers, CPCRA groups (members of the Community Programs for Clinical Research on AIDS), and other community-based research groups could be recruited for such a task. Data collection on these patients would be more comprehensive than that required for other parallel track participants but not as extensive as that specified for the ACTG trials. This strategy would permit comparisons of outcome and toxicity on two different levels: (1) between subjects enrolled in clinical trials and the subgroup of parallel track participants followed in a fairly rigorous fashion and (2) between the subgroup and the larger population of parallel track patients. The results could give government scientists, drug sponsors, and HIV-infected patients and their physicians a foundation for evaluating the parallel track program.