CREATIVITY IN CLINICAL TRIALS
Each month physicians participating in the expanded access protocols for ddI submit status reports on their patients to the drug sponsor and, in return, receive new supplies of the drug. The clinical and laboratory data enable the sponsor and government scientists to monitor the course of the protocols—especially with regard to safety concerns as described in Chapter 3. But the availability of these data also raises questions about the role of expanded access in the drug evaluation process. Can the information gathered through expanded access be used to speed or enhance the evaluation of AIDS drugs?
The answer to this question depends on a host of medical, scientific, and financial issues. Chapters 5 and 6 examine the costs of the parallel track approach for drug sponsors, patients, and third-party payers. This chapter explores the drive for innovation in clinical trials and the possibility of including expanded access protocols as part of the broad spectrum of drug evaluation mechanisms.
THE CHANGING ENVIRONMENT
Since the beginning of the AIDS epidemic, persons infected with HIV and their advocates have complained about the conservative nature of the drug development and evaluation process. Many have viewed the strict entry criteria for these clinical trials as an unreasonable barrier to participation. The problem grows worse as greater numbers of patients reach the advanced stages of HIV
This chapter is based on the presentations of Lincoln Moses, Lawrence Corey, Melanie Thompson, Marvin Zelen, Floyd J. Fowler, Susan Ellenberg, and Ellen Cooper.
infection. Each new opportunistic infection decreases the chances that a patient will qualify for entry into a desired clinical trial. One physician reports screening 80 to 100 patients to find 5 who qualified for any of the formal clinical trials of ddI. Another physician screened 275 patients to enroll 35 in the ddI trials. The extensive screening procedures slow patient accrual and lengthen the time required for completion of clinical trials. Moreover, when patients are excluded from a trial investigators lose the opportunity to learn anything from those patients, including information about drug-drug or drug-disease interactions, and how sicker patients respond to the drugs under study.
Some scientists also question the relevance of clinical trials conducted with a highly selected subgroup of the population. As HIV-infected patients live longer, their clinical histories become more diverse; they have different opportunistic infections and receive different combinations of drugs. Clinical trials that ignore these differences—that focus exclusively on a homogeneous group of patients—may not provide an accurate perspective on the drug's performance in the real world.
Most government scientists, academics, and patient representatives agree that there are many opportunities for greater creativity in the clinical trials process. Scientists are exploring new ways to modify the standard three-phase approach to drug evaluation—to improve efficiency without undermining the reliability of results. Other proposals include establishing preference trials, devising large-scale trials with broad eligibility requirements and limited data collection (similar to the International Studies of Infarct Survival in Europe), and gathering data through the parallel track.
Decades of experience indicate that conventional randomized controlled clinical trials (RCT) are the most reliable and informative way to obtain information about the safety and efficacy of an investigational drug. The IUDR story in Chapter 3 is just one example of a situation in which randomized study of a small group of patients revealed that a highly regarded treatment was not accomplishing its goal. The Cardiac Arrhythmia Suppression Trial, or CAST, provides another relevant case history.
The CAST effort involved two drugs, encainide and flecainide, both known to suppress irregular heartbeats in patients with heart disease. Many physicians believed that these drugs should be administered widely to patients who had suffered a heart attack (because deaths
following a heart attack often result from irregular heart rhythms) and that it would be unethical to conduct a placebo-controlled trial in which some patients would not receive the drugs. A randomized trial was begun, however, and the results were startling. The trial demonstrated that the drugs did, indeed, reduce irregular heartbeats compared with placebo, but they also increased mortality among patients who had symptomatic but not life-threatening rhythm abnormalities.
The strength of RCTs lies in the fact that they are structured to eliminate as many extraneous differences as possible between the groups being compared. This can be accomplished by applying very carefully devised inclusion and exclusion criteria, carefully following a protocol, and, of course, assigning patients randomly to treatment arms. These procedures are especially important in HIV infection, which has an erratic clinical course and many different patterns of illness.
The challenge for AIDS investigators is to retain the advantages of traditional clinical trials and at the same time reduce entry restrictions. The AIDS Clinical Trials Group has established a Protocol Evaluation Subcommittee to explore ways to make trials more efficient and flexible. The Statistical Working Group of the ACTG is also working to broaden entry criteria for clinical trials and to speed their progress. Scientists writing new protocols have been encouraged to reduce requirements for laboratory tests and to shorten reporting forms.
But there are limits to these approaches. Some research questions require a high level of technological expertise. On the other hand, the academic medical centers that possess this expertise may not have the facilities to provide care for the full spectrum of HIV-infected patients. (Chapter 7 explores the limitations of some traditional clinical trial sites in providing care for women and people of color.) The establishment of the Community Programs for Clinical Research on AIDS (CPCRA), described in Chapter 1, is based on the concept that HIV infection raises many different types of research questions. Some are best answered within the confines of a traditional clinical trial; others are more appropriate for large, simple randomized trials in which the bulk of the patient population is treated in the community. ''Low-technology" randomized trials, combined with expanded access programs, offer an opportunity to learn something from all segments of the population infected with HIV.
Before addressing the positive features of the large, low-technology trial, it is important to examine the pros and cons of one other experimental design, the preference trial. The preference trial is based on the concept that when treatments are tested only on select subsets of patients and a treatment effect is observed, doubt remains about how the treatment will work for most patients under less controlled circumstances. Therefore, proponents of preference trials argue that studies of broad ranges of patients under varying conditions are necessary to discover how well treatments really work.
Patients may also have very different "utilities" with regard to the risks and benefits of experimental therapies. For example, an HIV-infected patient may feel that the reduced energy level associated with one investigational drug is preferable to the nausea and vomiting associated with another. Alternatively, on a more serious plane, patients may feel that the risk of a very severe adverse reaction in the present is worth the potential benefits of a decade or more of disease-free survival. Supporters of preference trials believe that ethical research should take account of patients' values, a factor that randomized drug trials generally fail to consider.
These supporters propose trials in which all patients who might benefit from an experimental therapy are given an opportunity to choose between the new therapy and other treatment options. The investigator's role would be to provide patients with as much information as possible about the potential risks and benefits of each option, and then to collect data about the patient's health status and quality of life as the trial progresses. The underlying assumption is that individual values would lead patients with almost identical characteristics at the beginning of a trial to choose different treatment options. At the end of the study, comparisons of the outcomes associated with each choice would enable scientists to estimate the relative safety and efficacy of the investigational drug.
Supporters of such trials say they have many potential advantages: (1) patients would be more likely to comply with treatment regimens that they have selected themselves; (2) patient accrual would be rapid because no patients would be turned away; (3) the subject population could reflect the broad spectrum of HIV-infected patients; and (4) the process of informing patients about potential risks and benefits would mirror events in the "real world" of clinical practice.
Critics argue that previous experience in drug investigation indicates that nonrandomized trials do not give reliable results. Too many factors can influence drug choice and clinical outcome. For example,
rapid shifts in the popularity of certain underground drug therapies among HIV-infected persons demonstrate the power of a social network in influencing patient decision making. In addition, when patients choose their own treatment options, there may be important differences among the treatment groups that influence outcome. Even very sophisticated statistical techniques cannot control for all such effects; experience indicates, in fact, that many important factors are unknown or unquantifiable.
Scientists opposed to preference trials say it would be deceptive and unethical to tell patients in a nonrandomized study that they are contributing to advances in drug therapy because the data collected in such trials do not provide definitive answers to basic questions about drug efficacy. But scientists who are currently experimenting with preference trials answer that blind insistence on randomized controlled trials as the only appropriate method for evaluating drugs has limited our ability to discover the true value of most of the treatments available today—because such trials are costly, difficult to complete, and, when completed, apply only to the particular subset of patients who met eligibility criteria. They believe that there are ''good" and "bad" studies of all kinds and that alternatives to RCTs must be judged by how well they answer important questions and by the quality of their execution.
LARGE, SIMPLE RANDOMIZED TRIALS
Some of the positive features of the preference trial, such as broad patient participation, could also be achieved through large-scale, "low-technology" randomized trials similar to those developed in Europe to study cardiovascular diseases. These trials have involved endpoints that are easy to measure (such as survival or stroke) and limited data collection (the minimum possible to achieve satisfactory results). For example, in the International Studies of Infarct Survival (ISIS), tens of thousands of patients in many different health care settings were randomly assigned to groups to explore ways to increase survival after heart attacks. The second ISIS trial, ISIS-2, demonstrated that streptokinase and aspirin were both highly effective (compared with placebo) in reducing cardiovascular mortality after an acute myocardial infarction, and that the two agents together were significantly better than either agent alone. These effects were recognized despite the fact that patients had a wide range of background treatments (beta blockers, nitrates, and calcium channel
blockers) and prognostic variables. The large size of the study ensured comparability of the patient groups.
Most of the European trials have involved drugs with known toxicities administered on a one-time basis (so compliance was not an issue), but some scientists believe that the format could be adapted to study HIV-related disorders. Large-scale trials among AIDS patients with broad eligibility requirements and streamlined data collection might be appropriate for answering questions about the optimal dosage (quantity and schedule) of an antiviral drug, drug combinations, drug interactions (especially with regard to prophylactic agents and treatments for opportunistic infections), and drug resistance.
With careful planning that emphasizes a factorial design,1 virtually any HIV-infected patient who was willing to consent to randomization and who had access to a skilled primary care physician could be included in a trial of some kind. The idea would be to have an available trial for every AIDS patient. Data collection would focus primarily on such clinical endpoints as opportunistic infections, fevers, intractable diarrhea, HIV wasting syndrome, changes in stage of disease, important adverse reactions, and survival.
Registration could take place by telephone. A physician would call a central registration number and provide some initial data on the patient's medical history. The registrar would assign the patient to a protocol and, within 24 hours, send out patient consent forms, details of the protocol, drugs, and reporting forms.
Large-scale AIDS trials would require the cooperation of many different segments of the health care community. Designated members of CPCRA and other community-based research groups would be logical sources of health care providers for these trials, but the ultimate goal would be to include independent physicians and their nursing and laboratory support staffs. One scientist suggests creating a roster of community physicians who have demonstrated their interest in the research process by attending special AIDS workshops or seminars. Workshop topics might include available protocols, patient
consent, the collection of a core data set, staging, and ways to increase patient compliance.
For maximum efficiency, the link between specialists and primary care physicians could be maintained through a specialized computer network. The capability exists to develop extensive computer networks consisting of educational programs, bulletin boards with up-to-date information about ongoing trials (including special alerts about unexpected side effects), and electronic mail for direct access to medical consultants, nursing consultants, and data collection specialists.
One of the biggest challenges for the designers of large-scale AIDS trials would be to develop mechanisms for monitoring the quality of data. In addition to the usual data management strategies, some observers have suggested establishing auditing teams to visit participating physicians at random (primarily to compare patient records with information submitted to data collection centers).
DATA FROM THE PARALLEL TRACK
It is difficult to predict the ultimate importance of data collected through the parallel track and treatment INDs. Rapid implementation of large-scale randomized studies could greatly reduce the need for programs designed solely to increase access because most patients who could not qualify for current clinical trials would be eligible for trials with broadened eligibility requirements.
Whatever their size, parallel track programs are unlikely to provide substantial information about the efficacy of drug candidates. The value of data on efficacy depends on the existence of an adequate comparison or control group. The parallel track, as it is defined in the proposed PHS policy statement, does not make provisions for control groups of any kind. Although some researchers believe that important information on efficacy could be obtained using historical controls, many others feel that this method is flawed because of the dramatic changes in the treatment and prophylaxis of HIV-related
disorders, in the proportions of patients from different risk groups, and in disease manifestations within risk behavior groups.
Safety Data and Related Information
The situation is quite different with regard to safety data. If early parallel track and treatment IND programs remain the only alternatives to conventional clinical trials, they could play a vital role in the identification of important adverse reactions. Some believe that they also could provide some information about "real-world" drug interactions and drug resistance.
The value of data from an individual parallel track protocol would depend on the provisions made for data collection and for monitoring data quality. There are two basic types of data collection: event driven and regular reporting (according to a predetermined schedule). In event-driven reporting, physicians fill out reports only when they observe an outcome of interest, such as an unexpected adverse event. Event-driven reporting places the least possible burden on the practicing physician (because most patients proceed through treatment without experiencing a reportable event); however, the method does not provide any basis for determining the accuracy of estimated event rates. Lack of reports could mean either that no reportable events occurred or that physicians failed to comply with reporting procedures. Requiring physicians to make regular reports substantially reduces the problem of underreporting (although it does not eliminate it entirely). Regular reporting may be crucial for situations in which thousands of patients receive very early access to an experimental agent.
The quantity of data required would depend on the drug candidate. Expanded access programs involving drugs in the final stages of the evaluation process (such as the former treatment IND for AZT), or drugs that had been tested in other contexts, might require less data from participating physicians. Reporting requirements would be more stringent for drugs that did not have an established safety record.
Monitoring the quality of data from the parallel track as a whole could be very difficult, even with regard to reporting of adverse events. A possible strategy might be to use selected subgroups of parallel track participants (such as those described in Chapter 3) to make comparisons between safety data from the parallel track and safety data from corresponding clinical trials.
Government scientists and others caution against unrealistic expectations about the types of questions that could be answered through the parallel track. Some have expressed concern that excitement over nonrandomized expanded access protocols could detract from efforts to revitalize and improve randomized clinical trials. For example, the "safety valve" represented by the parallel track might relieve pressure to modify exclusion criteria or to take other actions that would make conventional trials more effective.
Widespread distribution of a drug through early expanded access programs also could lead to inordinate pressures to approve drugs for marketing before scientists have gathered adequate clinical evidence of safety and efficacy. The simple presence of a drug in the patient population could lead to a presumption of effectiveness that might be very hard to dispel. Such pressures might result in approval of a drug for a tightly defined patient group, such as patients intolerant to or failing standard therapies. Experience indicates, however, that once a drug is in the marketplace some physicians will use it in ways that are not supported by any data in the hope that it will have greater benefits (with acceptable toxicity) than standard treatments.