Prospective Case Series in Clinical Trial Design--Proposal, and NTZ Example
Recently the drug NTZ was rejected by an FDA advisory committee--even though it is widely agreed that NTZ appears to be safe and may have value for treating cryptosporidiosis and other intestinal parasites. The rejection happened because the major clinical trial comparing NTZ to placebo in people with cryptosporidiosis had to be canceled recently, since very few patients volunteered. The FDA realized that it would be impossible to run such a trial in the U.S., and allowed UNIMED Pharmaceuticals, the drug's developer, to present what data it had. But this information--never intended to be the pivotal data for approving the drug--did not convince the FDA's analysts or the advisory committee that NTZ was or was not effective for treating this disease.We believe this debacle happened because in modern medical research--and especially in AIDS--only one kind of clinical trial is likely to be taken seriously: the randomized controlled trial, in which volunteers are randomly assigned to receive either the drug being studied, or a standard treatment for the illness (or placebo if there is no standard treatment, as with cryptosporidiosis). Certainly this kind of trial has unique advantages, and will continue to be important in medical research.
But what happens when a controlled trial is not feasible? Do we throw up our hands and either refuse to study the drug at all, or try to analyze data which was collected incidentally during medical care, with little or no thought for research? Today there is usually no other choice.
To help stimulate thinking about different kinds of clinical research, we suggest a different kind of trial, a prospective case series--an organized system of case studies in which the research team does the best it can for each patient, using any available treatments (approved or experimental) and absolutely first class documentation and data collection. The goal would not be statistical proof, but a document in which the physician-researchers tell what they have learned about how to treat the illness effectively, and use the data generated to explain why they have come to believe what they do. The full data itself could also be published, allowing others to suggest different interpretations or conclusions.
Design and Size of Study
A team of physicians and researchers would design the study. They would decide such issues as which patients can enroll, what drugs would be available, what laboratory and other tests could be provided, and how many volunteers could be accepted.
Unlike controlled trials which seek statistical answers, the studies we are suggesting do not have a minimum size which can be calculated mathematically. The number of volunteers might typically be ten to 40--but it might be as few as one, or it could depend on the resources available. Patients can be treated and studied as they arrive, and the researchers will not necessarily have to decide in advance on the total number of volunteers.
Entry Criteria
In the case of NTZ, one of the problems with the controlled trial (the one which had to be stopped because it did not recruit) was the time and difficulty of reaching a confirmed diagnosis of cryptosporidiosis. Researchers might avoid this problem by designing clear entry criteria for suspected cryptosporidiosis--or for diarrhea which is suspected to be caused by unidentified protozoa or bacteria. This approach might be the more practical, since NTZ is probably safe and is one of the broadest spectrum anti-parasite drugs known--and typically in medical practice, diagnosing exactly which pathogens are causing diarrhea can be difficult or impossible. Instead of waiting for an exact diagnosis, it may be better to treat certain patients empirically for a short time, and discontinue the drug if their symptoms do not improve. (Of course samples would be collected before treatment began, so that changes in microbiology could be recorded, and correlated with changes in symptoms.)
Treatment
The goal of the physicians and researchers running the trial would be to provide the best possible treatment for each volunteer all the time--using almost anything (of course with the consent of the volunteer, who would have at least one treating physician independent of the trial).
For example, controlled trials typically give a fixed dose of one drug for a predetermined time, then collect statistics on what fraction of patients benefited. In a prospective case series, the researchers would try to control the disease in every volunteer; drug doses could be increased or decreased, as appropriate for that individual. Or the study drug could be combined with other drugs (such as NTZ with azithromycin, which is also broad spectrum, and may have some activity against cryptosporidiosis).
With the volunteer's consent, the researchers might decide to stop treatment early to see if symptoms return, and start treatment again if they do. This kind of rechallenge can greatly increase confidence that the drug is (or is not) active in that particular person.
In some cases the researchers might decide that it was best for a volunteer not to use the study drug at all. The person could still stay in the trial and be monitored, could receive other treatments, and would be included in the analysis.
Data Collection
While the protocol for treatment would be flexible, the protocol for documentation and data collection would be exacting. Such information as laboratory tests, concomitant medications, and rationale for treatment strategies would be recorded at strictly defined times, and with pre-defined procedures for quality control.
Ethical Supervision
How could such a flexible study--where there is no rigid treatment protocol, and drugs can be changed at any time--be effectively controlled by an IRB (Institutional Review Board)?
One approach is to have the IRB approve a different kind of protocol which is really an "envelope," defining all the drugs, tests, and treatment strategies which are likely to be used in the trial. In exceptional cases where the researchers want to go outside the envelope, due to new information or to a volunteer's special circumstances, they would need permission either from the full IRB, or from a committee which the IRB had designated for that purpose.
Also, every volunteer should have a treating physician who is unconnected to the research project. This physician would have access to all the information gathered about the patient, and could advise the patient to withhold consent for any treatment or procedure.
Advantages
This kind of trial combines the best available medical care with excellent data collection for research. Here are some advantages:
The approach focuses directly on extending the standard of care as a whole--instead of testing just one drug at a time, due to commercial constraints. It provides a planned, coherent program of many case studies planned in advance and happening at once.
Recruitment should not be a problem, since volunteers will be cared for by a team of leading experts who can use all available information and treatments, who can change treatment at any time, and who are always trying for the best treatment for each individual.
Total cost should be small since relatively few volunteers are needed. And this kind of study can start quickly without waiting for extensive resources to be assembled, since no minimum number of subjects is needed, and useful information starts being generated with the first one. Also, results from the first few volunteers can provide a database to help justify additional funding or resources if necessary.
Results can be released as they are generated. There is no need for secrecy--unlike conventional controlled trials, where information learned early can threaten later adherence to the treatment protocol. Here it is intended that treatment will change as new information becomes available. The goal is always the best treatment of the individual, using all existing information toward that end.
Even if it fails to produce definitive or convincing proof of drug efficacy, a prospective case series can provide highly reliable information about what leading physicians are doing, what problems they face, and some of the results they are seeing--information which can help in the design of subsequent controlled trials.
This approach does not produce statistical proof like controlled trials do. But it may lead to comparable or even better confidence in the conclusions drawn. When a controlled trial reaches the p=.05 statistical significance level which is usually acceptable, there is (by definition) one chance in 20 that a worthless drug will pass as effective. In the design we suggest, well-informed physicians will gain experience with the drug under research conditions--including seeing how symptoms respond to repeated rechallenge, if appropriate. Although it is impossible to calculate an exact probability, it seems likely that after such a trial, a worthless drug may well have less than a one in 20 chance of being considered effective.
Instead of organizing study design around p-values, statistical analysis can focus on broader issues of using available information to assist decision making under uncertainty.
Medically, this kind of trial focuses on exactly the issues that treating physicians face. Treatment is contemporaneous, meaning that the strategies being tested are the ones which leading medical experts are using that day. In contrast, conventional controlled trials test strategies which were current when the trial was designed, often several years before. And they test treatment by protocol, which usually differs greatly from medical practice.
We believe these advantages justify serious attention to this kind of trial--a planned yet very flexible case series which is designed prospectively and run under research conditions. Such studies could be used when controlled trials are not feasible, as with NTZ today. Also, a prospective case series could rapidly provide practical information useful in designing randomized controlled trials which are well focused on provable and clinically relevant questions.
