Designing and Implementing a Clinical Trial
One doesn’t discover new lands without consenting to lose sight of the shore for a very long time.
–Andre Gide (1869-1951).
Discovery consists in seeing what everybody has seen and thinking what nobody has thought.
–Albert Szent-Gyorgyi (1893-1986).
This chapter lays the foundation on which subsequent chapters in this section on clinical trials are built. As such, many topics discussed only briefly here are revisited in more detail in later chapters. Even so, a number of books devoted entirely to the in-depth discussion of these topics are provided in the additional readings at the end of this chapter.
OVERVIEW OF STEPS INVOLVED IN A SINGLE CLINICAL TRIAL
Before delving into the parts of a protocol and commenting on each of them, it seems worth briefly reviewing the overall process that is involved with the design, implementation, conduct, data collection, analysis, and interpretation of the results of a trial. This section presents a series of numbered events that present these steps, but not all of them are necessary in any one trial. In addition, these steps are rarely performed sequentially, both in the sense that some are conducted in parallel and also that some are conducted before those that precede them in this list.
Identifying the primary objective(s) of a trial. The primary objective(s) of a trial is the correct place where the planning of most clinical trials is initiated, but the specific objective(s) is not always easy to identify precisely and may require substantial discussion and thought to be able to be expressed correctly. This objective dictates what type of clinical trial design must be employed to best address the goal (e.g., to evaluate the tolerability of patients to X, to determine the safety of Y) or hypothesis posed as an objective (e.g., to determine if Drug X at Y dose in Z patients with G disease who are treated for W weeks can increase parameter O by 50% more than a group of similar patients treated with a placebo). To paraphrase the Mad Hatter in “Alice in Wonderland,” if you do not know where you are going, every path can look good.
Choosing a trial design to meet the objectives of the trial. While the design may not be difficult to choose, there may be many reasons why the ideal design cannot be used. A few of these reasons might be: limited resources, few interested investigators, limited patient numbers, or the requirement for too much time to conduct the trial [e.g., a mortality or survival study that would take five (or possibly more) years to conduct].
Determine the clinical endpoints and the tests/measures to assess the endpoints. This is much more of an issue than most new professionals realize. While a biomarker or partially validated surrogate endpoint may be the most desirable endpoint to measure, they are often not acceptable for trials that will be used to obtain regulatory approval (see Chapter 90).
Writing the protocol. This is often a very long process and may start with preparation of the protocol synopsis. If that is initially done, then it may be reviewed with the company’s scientific/medical advisory board, external consultants, and regulatory agencies as well. The full protocol has a large number of important components that are discussed in the next section.
Deciding if a contract research organization (CRO) will be used to help the company run the study. Different companies have different philosophies regarding the use of CROs to help them conduct trials. If a company has the staff available, it will usually plan, initiate, monitor, collect data, analyze and interpret the data from the trial entirely on its own. However, the staff are often busy with other studies, and the company may turn to a CRO for assistance to start another study. It is critical to understand that using CROs are not a process of “farming out part of the drug’s development” because the company maintains ultimate responsibility for what is done by the CRO and must monitor their performance. Using a CRO to assist the company in one or more parts of a clinical trial makes a great deal of sense, however, when its own resources are limited. CROs can be best thought as middlemen that are sometimes used to help a company in one, a few or many aspects of a specific clinical trial. CROs are generally considered as being somewhat more expensive than if the company runs the trial itself, but the added value is that the trial can be initiated earlier when the company does not have enough staff to do the work itself, and thus the company saves time, which is very critical in drug development. Other important reasons for using a CRO include the need for specialized expertise and access to investigators in locations where the company has
no experience or organizational capacity (e.g., in certain foreign countries).
Deciding on what functions the CRO will be asked to participate in. CROs offer a large menu of services and the company will choose anywhere from one to 50 or so different clinical services it wishes to have the CRO provide (see Chapter 74). Once the company has decided on the services it desires, it will invite a number of CROs to submit bids.
Obtaining bids from three or more CROs even though only one is to be used. Companies rarely ever bid a project to a single CRO, even when it has a preferred CRO it expects to give a contract to. The company will ask a number of CROs to bid on the contract and may use one CRO‘s bid against another to try to lower the overall cost of a trial if the company prefers to use a CRO that submitted a higher bid.
Choosing the CRO to help run the study. There are many ways a company may make their choice among the CROs, but often they will invite two or three CROs to make presentations as this practice will provide a great deal of additional information beyond what is included in the bid proposal itself.
Monitoring the actions of the CRO during initiation and conduct of the study. Because the CRO is hired by the company as the middleman, it might seem logical that it is not necessary for the company to monitor the CRO‘s activities, but this is not so. It is always important for a company to know how motivated and efficient and hard working the CRO actually is and if there are issues or even problems that they are unaware of. To understand these aspects of the CROs performance, they need to monitor closely the CRO itself. For small start-up companies, they can hire an outside consultant, firm, monitor, or even another CRO to conduct this monitoring. The principle that good past performance does not necessarily predict future performance must always be remembered (just as in financial investing). This results from high staff turnover at CROs and the competitive environment for patients who meet inclusion criteria in most therapeutic areas.
Having a liaison at the company to maintain close communications with the CRO. This will provide feedback on an ongoing basis to ensure that any issues or problems are immediately addressed. Often, weekly teleconferences are scheduled as “project meetings.”
Creating the case report forms (CRFs) to collect the data. These have to be well designed to help the sites understand the nature, quantity, and quality of data that are expected to be collected for a trial. They also allow the data management group to enter the data into computers and to capture the data needed by statisticians and clinicians who will analyze and interpret the trial’s results. Well-designed CRFs can improve the quality and completeness of data collection; whereas poorly designed CRFs may lead to significant rates of erroneous or missing data. The data collected may be submitted to the pharmaceutical company or the CRO electronically, by courier, telephone/interactive voice response system (IVRS), or may be faxed to the company on a frequent basis. There are often a few hundred pages of data forms for every patient enrolled. This may entail 20 or so pages per visit and involve up to 25 or so visits. In addition, there are special forms for use when needed (e.g., early termination form, severe adverse event form).
Interviewing and choosing the investigators and sites to run the trial. The time required for this activity varies enormously. The investigators may be chosen entirely by the company, the CRO, or both. Past experience, reputation, interest, and availability are among the important parameters used to choose the investigators. This may be a relatively easy or difficult exercise and depends on many factors. An important factor is the company’s existing relationships with known, trusted physician investigators who have trial sites and staff who are competent to conduct trials with all necessary equipment and are known to follow Good Clinical Practices.
Preparation of clinical supplies needed for the trial. This activity is often very complex. Imagine trying to prepare dialysis solutions containing one of four doses of the test drug or placebo in one of three bicarbonate solution concentrations. Imagine having to supply each patient with three gallons of solution three times a week and working with dialysis centers that have almost no storage room. On top of all of that, the empty gallon jugs have to be returned.
Beyond the preparation of supplies are issues related to labeling. It often takes time to design the labels to be affixed to the bottle and the carton that contains the bottle of pills (or ampoules, vials, etc.), as well as the shipping carton. The details for filling the bottles and developing the caps and any child resistant constraints must be specified. This last point may create problems if a blister pack or other packaging is needed and the patients themselves find it difficult to open the containers (e.g., those with arthritis, those with other problems preventing them from opening the blister pack). Many other issues are covered in other chapters, particularly in Section 9, such as small quantity manufacturing, blinding packaging, labeling, distribution to sites, storage, disposal, and scheduling.
Holding an investigators’ meeting. While it used to be considered essential to hold a face-to-face investigators meeting at which the CRFs and other protocol implementation pointers were reviewed in detail with the investigators and their study coordinators, at this time, more and more electronic web-based meetings are being held. In these meetings, the interactive program or even a DVD copy can be played for late-arriving investigators or those who are unable to attend the initial meeting, or for new company or site personnel.
Interacting with or Institutional Review Boards (IRBs) or Ethics Committees (ECs) as appropriate. Companies rarely interact with local IRBs directly (although they usually do interact directly with Central IRBs, see Chapter 81) and usually work through investigators by giving them the protocols, answers to commonly asked questions, a draft informed consent form that often is adopted by investigators, and any additional documents required. Questions or comments from IRBs are often sent by investigators to the sponsor for comments or response.
Negotiating and signing contracts with each of the sites or their institutions. This activity may take anywhere up to six months to complete and is one of the main reasons why many trials cannot be started shortly after the IRB has approved the protocol. Companies have created numerous ways to try and overcome this delay-causing event (e.g., creating a Master Service Agreement with the institution where each trial to be conducted is attached as a separate work order). The effort that a company expends to agree
on terms and to sign a contract is highly variable among sites. This has become a major issue in dealing with certain academic sites. Sometimes, it is necessary to negotiate two different agreements, one with the institution and one with the investigator.
Negotiating and signing contracts with other groups participating in a clinical trial. There are often numerous vendors and other individuals with whom the sponsor must have a contract. These groups include the CRO; central laboratories; specialized laboratories; Central IRBs; personnel who are members of a Data Safety Monitoring Board (DSMB); Scientific Advisory Board; Steering Committee; central readers of pathology, radiology, or other slides, X-rays, or other material (e.g., electroencepholograms); external monitors; consultants; IVRS groups; manufacturers; clinical packaging groups; and so forth.
Monitoring the clinical trial. This is considered as one of the most critical parts of the trial as it is necessary to ensure that the investigator is adhering to Good Clinical Practices and the protocol itself. This is often a difficult balance for the investigator, who is always trying to treat patient as medically appropriate as possible, and yet if all patients are not treated within the context and rules of the protocol, the data resulting from the trial may not be found acceptable by most regulatory agencies. The value of that trial’s data will be seriously compromised. Monitoring and auditing a clinical trial are discussed in Chapter 70.
Reacting to problems and issues that arise. Troubleshooting is always a challenge in any clinical trial and no one could ever anticipate all of the possible or likely issues and problems that could (and will) arise. Most authors try to identify the most likely issues in advance and to incorporate responses in the protocol itself, but other issues invariably arise that have to be dealt with on an immediate basis.
Preparing any protocol amendments needed. It would be wonderful if the protocol that often has taken many months or even a year (or possibly longer) to prepare was approved and initiated and the study completed on time, but unexpected events arise that lead to required (or important) changes to be made in the protocol. Some of these changes may have been required by the Food and Drug Administration (FDA). Each change must be sent to every IRB for their approval and every change is considered as a protocol amendment. If the changes are not of an urgent nature, companies try to wait for time to elapse in order to combine all of the changes they have accumulated into a single protocol amendment and then to notify every investigator, IRB, and the FDA about these changes.
Collecting all clinical data at the site and transmitting or sending it to the data management group. In some trials, this can be at the end of the trial, but that is rare. Most often, the data are sent to the sponsor or data management group on a periodic basis, such as every week. The way(s) the data have to be sent is specified in the protocol. This process is ideally done after all queries (i.e., questions of clarity, omission, or others) directed to the investigator on the data are addressed, although the data may go first to the monitor(s) to confirm that no queries exist or that they have been addressed.
Entering all clinical data into computers. Double data entry is required in almost all cases and any electronic differences between the two sets of data entry are reconciled and the data eventually are complete and almost error free.
Preparing software programs necessary to analyze the data. These programs include both existing ones that can be purchased [e.g., statistical analysis software (SAS) or the statistical package for the social sciences (SPSS)], as well as the statistical analysis software programs the company writes to execute the statistical analysis plan. There may be more specialized software for more complex analyses or graphs. In some cases, programs have to be developed for the specific trial. If the latter, then the time needed for that has to be built into the development program and should be scheduled so that this activity is not on the critical path.
Creating a statistical analysis plan. The statistical analysis plan is the blueprint of how the data will be presented (i.e., which tables, figures, and listings will be made) after they are entered into computers and then how the data will be analyzed. The data analyses must use the methods that regulatory agencies are expecting to see for the specific type of data presented, or a rationale needs to be given for the alternative methods or approaches used. The presentation of the analyses must also dovetail into the analyses and presentations of other pivotal and/or major supportive trials on that drug in order to put the data in the most appropriate clinical and/or analytical context and to ensure that the results are comparable. In some cases, if the data are from identical studies the total data can be pooled. In other cases, a meta-analysis may be appropriate (e.g., if many small trials have been conducted, particularly if only one large pivotal trial has been done).
Determining the skeleton (i.e., outline or shell) of the tables, figures, and listings. This step creates the ways that the trial’s data will be illustrated and presented to the statistician for analysis and to the clinician for interpretation. The tables, figures, and listings take a substantial amount of time to prepare and organize in a way that will best show the data obtained in a clinical trial.
Analyzing the data statistically. This step is not a simple one of turning an electronic crank or pushing a button, because after the data are analyzed, the statistician may find that the analysis was not the best one to analyze the data with and that another method must be used. Also, clinicians are likely to want to see additional analyses that the first draft suggests, and sometimes, this iteration goes back and forth between the clinician and statistician for an extended period. The questions might be along the lines of what would the data look like if we looked at the first half of the trial versus the second half, or it could involve many possible subgroup analyses of populations, such as what would the data look like if we compared the results from all patients who were taking a concomitant Drug X versus those who were not.
Preparing a statistical report. After all the statistical analyses are complete, it is necessary to prepare a highly detailed statistical report. Sometimes, this is combined with the clinical report so that a single combined study report is issued.
Interpreting the data clinically. The process of statistical analysis precedes the clinical interpretation of the data. The clinical interpretation is almost always done by clinicians involved in the study and not by the statisticians. The
“discussion” section of a paper or report is where the clinical interpretation is almost always presented.
Preparing a final medical report. This is usually prepared and follows formats that have been established by both the International Conference on Harmonisation (ICH) and the FDA. This may be combined with the statistical report or may be independent of it. Include the public health message, which reflects the medical need for a product to treat (or prevent or diagnose) a disease. Then include the medical value of the product being developed and discuss how the product addresses the medical need.
Polishing the final medical report. It is necessary to have the statistician and clinician who drafted the report, as well as various managers, review the report. Feedback procedures in a company should be used to ensure that the reports go through an extensive review process in order to yield the most accurate and complete data presentation, analyses, and interpretations.
Developing a publication strategy. This step and also overseeing the publishing of clinical data are important parts of the drug’s development. Most companies develop a highly sophisticated publication strategy and then work with their marketing and clinical colleagues (plus others when relevant such as preclinical and toxicology departments) to ensure that the data are presented in accordance with the strategy.
Writing a manuscript of the results. It is becoming more and more accepted, and even in some situations required, to publish results of a trial whether positive or negative. This is a controversial topic as there are often proprietary issues as well as competitive ones to consider, but there is little argument that once a product is marketed, the results of all past and future studies should be made public in one form or another.
Presenting clinical data at professional meetings as talks, abstracts, and poster presentations. The approaches to these events are designed to dovetail with the publication strategy, and are part of the rollout of the data from all companies on their new products.
Submitting the results to a registry of clinical trial results. There are a plethora of such registries today and this is a dynamic area that will continue to evolve over the next number of years (see Chapter 83).
DESIGNING A CLINICAL TRIAL PROTOCOL
While study designs are not specifically tied to any specific phases of development there are some designs that are more commonly used in some phases than in others. The general list of study designs that are possible to use are presented and variously classified in several sources (see Spilker 1991, Spilker 1997, and ICH 1996 in the Additional Readings section at the end of this chapter).
Components of a Protocol
The many parts of a protocol are listed in Table 61.1.
Not all of these are necessarily required in every protocol, but they should all be considered for discussion. Once a detailed protocol is available, much of it may be considered as “boilerplate” that can be inserted into most or all future protocols. Some of these will be definitions, such as the severity of adverse events (i.e., mild, moderate, severe), responders, or completers
(i.e., how much of the trial must the subject complete to be defined as a completer).
(i.e., how much of the trial must the subject complete to be defined as a completer).
Table 61.1 Parts of a protocol | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Table 61.2 Information to include in a primary objective | ||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Trial Objectives
As mentioned above, the place to begin a protocol is with the objectives. Even some large companies do not present these appropriately. The objective is a detailed and specific identification of the intentions of the trial and not a brief synopsis or summary such as “To study the safety and efficacy of Drug X.” The information to include in an objective is listed in Table 61.2 and an example is listed below.
An Example of a Clinical Trial Objective
To determine if Drug X given in a dose range of 100 to 300 mg once a day for four weeks will decrease Parameter Y more than placebo in 500 subjects with severe chronic Disease Z who have not responded to conventional therapy, using a randomized controlled trial where half of the patients are to be given a placebo.
Common Pitfalls in Designing a Primary Objective
Objectives are too often expressed in general terms and are incomplete. The FDA rarely comments on this issue as it is not against regulations, but may lead toward choosing an inappropriate trial design to adequately test the correct (i.e., full) objective. (It sometimes appears that the FDA‘s attitude about this issue is “Let the sponsor beware.”) Another common problem is that there are too often too many primary objectives. There should rarely be more than one or two. Safety is not always mentioned as an objective but, nonetheless, it is always one, whether mentioned specifically or not. No prospective trial ever is conducted without safety being an important concern and objective. Finally, some objectives are retrofitted to fit the design the author has chosen, which is always a dangerous tact to take.
Secondary Objectives
There are almost always a few or many secondary objectives included in a protocol. If the primary one is found to be correct (e.g., the drug is effective) then regulators are usually willing to accept the first or first two (possibly three) secondary endpoints for labeling purposes if they too are shown to be demonstrated. Thus, most (but not all) trials list the first several secondary objectives in order of importance to the company from a labeling perspective. In some cases, a company lists over 20 or 25 secondary objectives. While there are no guidelines or regulations against this practice, it is clearly unnecessary since most of these are simply indications of a desire to conduct a subgroup analysis of the data. In that case, it is unnecessary to list them in the protocol, as those analyses can always be performed with the data after the trial is completed.
Exploratory Objectives
Some trials list one or more exploratory objectives, often to determine if one of more endpoints has value to study in future trials. Some people refer to these as tertiary objectives.
How to Proceed after the Objectives Are Created
The next step is to determine the trial design that best addresses the primary objective(s). This may require a discussion among experts in both the therapeutic area and in clinical methodologies. The hierarchy of scientific and medical evidence must be considered in this discussion, i.e., (a) How definitive do we want the data to be? (b) Is an open label trial sufficient or do we want a higher level of evidence such as a randomized controlled trial? and (c) Do we want to include a placebo group or is another control appropriate? Additional information is found in the Users’ Guides to the Medical Literature: A Manual for Evidence-Based Clinical Practice (Guyatt and Rennie 2002) and in articles by Sackett et al. (1996), Straus et al. (2007), and Haynes (2007).
Types of Controls to Consider
Drug designs for clinical trials will almost always have a control group even if it is a historical control, a no-treatment control, or use of standard therapies that may or may not specify which are acceptable. Other controls are to use an active drug control, placebo control, or the actual test drug, but at a subtherapeutic dose. It is sometimes common to use two of these control groups in the same trial (e.g., analgesic trials that include an active control and a placebo group). In some cases, a cross-over design where the patient is their own control is desirable and provides strong results that are often quite definitive, assuming no carryover effects, period effects and that the patient returns to their baseline between the two arms of the crossover. If a historical control is to be used, the data may either be derived from the literature or each patient may serve as their own control, where the historical data are obtained from their medical records or from their recollection (e.g., how many asthmatic attacks did you have last year that required hospitalization?).
When an active drug control is used the design may be to compare the test drug against the active control without a placebo, in which case there are three basic design choices. A superiority trial is one in which the test drug is evaluated to see if it is statistically significantly (and clinically significantly) superior to the active drug. This takes fewer patients than if one conducts a noninferiority trial where one’s hypothesis is that the test drug is not statistically significantly inferior to the active control, and the third choice is to show that the test drug is equivalent to the active control, but this design requires the largest number of patients compared with the others. Finally, an active control may be included
in a study when the test drug is primarily being compared with a placebo, or possibly a lower dose of the test drug.
in a study when the test drug is primarily being compared with a placebo, or possibly a lower dose of the test drug.
Which Types of Controls Are Possible When Using a Placebo Is Unethical?
There are a variety of situations when it is deemed unethical to include a placebo group in a clinical trial. Under such conditions, it may be possible to bypass this constraint and to still have a well designed trial. Some of the alternatives to a placebo are shown in Table 61.3.
Even when a placebo is traditionally considered as unethical some trial designs that include a placebo can satisfy all ethical review groups. The major example is with a “fail-safe” design where patients who deteriorate (or do not improve) are removed from the trial.
Types of Blinds to Consider
The choices of blinds are thought of as limited to open-label designs, single blinds, and double blinds, but there is another, complete blind. In open-label trials, there is no blind. This usually allows a high degree of bias to enter the trial and leads to many false positive results. Single blind studies also have the same disadvantages as open-label designs and are similar in yielding approximately the same high level of false positive results, regardless of whether the investigator or patient is kept blind. Although use of the double blind is the industry standard, the degree to which the double blind is maintained varies tremendously because many trials cannot be adequately blinded (e.g., if a drug turns a subject’s urine a different color or if specific symptoms or clinical signs occur, it is not definitive that the subject is on active treatment but it is highly likely).
Table 61.3 Possible study designs when a placebo is deemed unethical | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
A complete blind (also referred to as a triple blind) is where the subject, investigator, and everyone who interacts directly with them (e.g., research coordinator, monitor, and others) are kept blind. If at all possible, this is the standard to use in clinical trials.
Various steps may be taken in the protocol to help ensure that the blind is maintained and not breached. Such steps include using blocks of more than three or four to randomize patients, and not allowing the person who interviews the patient for adverse events and/or adjusts the dose be the same person who evaluates the clinical status of the patient as improved or not. The DSMB may or may not be blind when they evaluate clinical data. There are many pros and cons to consider in deciding if they should be allowed to view unblinded data. (See Chapter 81 for a discussion of DSMBs.)
Selecting Inclusion Criteria (Zero-based Inclusion Criteria)
While every protocol has a section outlining inclusion and exclusion criteria both can be considered as inclusion criteria. Inclusion criteria can be expressed positively (i.e., a patient may be enrolled if …) or negatively (i.e., a patient may not be enrolled if …). The term inclusion criteria will be used in this chapter as referring to both inclusion and exclusion criteria. The overall goal of inclusion criteria is to obtain the appropriate balance of heterogeneity or homogeneity needed in the types of patients enrolled. Fewer inclusion criteria usually lead to a more heterogeneous population. An excessive number of inclusion criteria or a small number of highly restrictive criteria can lead to a trial’s inability to recruit patients and to its failure. In addition, the more narrow/restricted the study population, the less generalizable are the results.
The two most common mistakes made regarding inclusion criteria is to make some of them too restrictive (e.g., ages of 21 to 40) and to have too many of them. Overly restrictive inclusion criteria based on narrowly defined ranges may severely limit patient enrollment, sometimes making it difficult to enroll even a single patient in a trial. Delays in enrolling patients lead to wasted staff time, additional expenses in enrolling new trial sites (if that approach is used to increase enrollment), and modifications to the protocol with resubmission to the IRBs. The outcome is a delay in completing the trial, potential disapproval of the drug by regulators and, even if eventually approved, shortening of the patent life. Recruitment is the number one reason why trial times are prolonged, deadlines are missed and additional costs are incurred.
In an ideal world, it would be desirable to know how the cumulative effect of every major and minor inclusion criterion decreases the available patient population that could be enrolled. The wrong choice of such criteria in the trial will also unnecessarily complicate the collection and analysis of data, with regard
to the overall goal of the study. It is likely that many of the protocol’s inclusion criteria would be removed or modified by a sponsor if the effect on decreasing the pool of patients available for enrollment were known. In some situations, rigid inclusion criteria are essential and must be accepted regardless of their effect on enrollment. However, in others, the number of criteria could be minimized, and/or the acceptable ranges for each criterion relaxed, with little consequence to the ability to address the trial’s objectives or the product’s medical value. Inclusion criteria should be established to allow the maximal number of patients with acceptable characteristics to be eligible for enrollment.
to the overall goal of the study. It is likely that many of the protocol’s inclusion criteria would be removed or modified by a sponsor if the effect on decreasing the pool of patients available for enrollment were known. In some situations, rigid inclusion criteria are essential and must be accepted regardless of their effect on enrollment. However, in others, the number of criteria could be minimized, and/or the acceptable ranges for each criterion relaxed, with little consequence to the ability to address the trial’s objectives or the product’s medical value. Inclusion criteria should be established to allow the maximal number of patients with acceptable characteristics to be eligible for enrollment.
To minimize the number of patients who will be excluded from participation in a specific clinical trial, one should consider: the acceptable range of patient ages, the range of acceptable laboratory values for each analyte, the characterization of each patient’s current disease status, the history of their disease, prognostic factors, and the known safety profile inferred from preclinical and any earlier clinical data. Many, if not most, trials include some tests with the statement that patients must be “within normal range.” This is often too restrictive, as even in Phase 1 trials, few normal people have all their laboratory values in the normal range. Previous exposure to other pharmaceuticals, risk factors, and the type and severity of the patient’s disease are among other critical criteria that should be carefully evaluated. All inclusion criteria must be considered as to their impact on the size and the makeup of the trial’s potential population. When writing a clinical trial protocol, the golden rule is not to ask the usual question: “Which inclusion criteria will give the most desirable patient population for the clinical trial,” but rather “Which inclusion criteria are the minimal essential criteria to include in this clinical trial?”
Clinical trial designers with PhDs and other advanced life science degrees often have a predilection when they start to write protocols to include an excessive number of restrictive inclusion criteria. This practice often comes from their training and experience in science (even in the biological sciences) where many conditions of an experiment must be controlled in order for one to study the single parameter or characteristic of interest. While this approach is the appropriate and correct one to use in pharmacological, biochemical, or physiological studies, to name a few areas, this approach is not as easy or possible to apply to humans, where many parameters and conditions cannot be reliably controlled. As a result, inclusion criteria must be carefully reviewed by experienced clinicians to ensure that each is necessary and that restrictive ranges placed around those that are needed are appropriate for the trial being planned.
Zero-based Inclusion Criteria
The golden rule about inclusion criteria is to use the minimal number necessary and that no criterion should be included unless agreed to be essential or at least highly desirable to the success of the trial (i.e., in obtaining the desired population of patients). If it is known that an adequate number of subjects are available for enrollment, then this rule can be somewhat relaxed.
The best approach in creating a list of inclusion criteria is to start with a clean sheet of paper (or clean computer screen) and to only add those that are mandatory and follow the “need to have” rather than the “nice to have” principle. While it is easier to relax inclusion criteria during a trial than to try to make them more strict (usually not a wise decision), any relaxation will require a protocol amendment, submission of the amendment to each of the IRBs, and discussions with each of the investigators to ensure that they understand the change and the reason(s) for them. The amendments must also be submitted to regulatory agencies. Every protocol amendment takes time, effort, and money and is undesirable, often slowing development, so that efforts must be made to try and avoid this situation.
Choosing the right inclusion criteria has a major influence on the time to completion of a clinical trial and this time directly influences the time needed to submit a regulatory dossier. The choice of optimum inclusion criteria can come from lessons learned from previous programs and studies. Thus, good project management tools will help professional staff track the enrollment rate and identify any inclusion criteria that were problematical in previous trials with the same drug or device.
Too many authors of protocols take the last protocol used for that drug or product and seem to add-on or modify those inclusion criteria that appear to be of importance. This often leads to an excess number of criteria and the available population for enrollment is shrunk by every criterion that is added to the list.
Try to be creative in minimizing both the number and restrictiveness of the criteria. For example, what upper age cutoff is actually required as opposed to traditional? This is simply one criterion that is often made more restrictive than necessary.
Zero-based Procedures
The number of procedures to include in a clinical trial should also be addressed the same way as inclusion criteria (i.e., accept the principle that no procedures will be done in the trial simply because they were done in the last trial, but will be done when they are definitely necessary). This is the concept of zero-based procedures.
Recruitment, Retention, and Other Topics in a Protocol
The two topics of patient recruitment and retention are so critical to the success of a trial that many important principles have been presented in a separate chapter (see Chapter 72) and are not discussed further.
The list of topics covered in most protocols, as listed in Table 61.1 includes many issues and subjects that are not covered in this chapter due to lack of space. Some are discussed in other chapters in this section and many are discussed in Guide to Clinical Trials (Spilker 1991). This chapter only provides a high-level overview and introduction to other chapters and greater detail that is provided in other books and articles.
Diagrams and Illustrations to Include
While the majority of diagrams and illustrations are likely to be included in an operations manual (see that section of this chapter), there are usually a few such diagrams in the protocol. The most common one is the flow chart that shows how patients are randomized to treatment groups and are then treated and evaluated in the trial. The author believes that almost every protocol should include a diagram of each of the separate organizations involved in the trial and their role, such as those mentioned in this chapter. The relationships among these groups would provide important information relating to communications that is sometimes not clear to those involved. Many other types of diagrams or illustrations are possible, but are not as commonly included as a variety of tables, such as the Time and Events
Schedule. This table lists visits along the top horizontal header and specific tests that are to be run as rows along the vertical axis. Tables relating to patient groups and drug preparations are also quite common.
Schedule. This table lists visits along the top horizontal header and specific tests that are to be run as rows along the vertical axis. Tables relating to patient groups and drug preparations are also quite common.
Should Dropouts and/or Discontinued Subjects Be Replaced in a Clinical Trial?
Some protocols discuss the withdrawal or discontinuation of patients from a trial but do not distinguish between those who leave a study of their own volition (i.e., “dropouts”) and those who are “discontinued” by the investigator. This distinction is important, particularly when deciding whether a patient or volunteer who leaves a trial should be replaced by a new subject. There is no golden rule to address this issue nor any relevant guideline or regulation. Nonetheless, this matter is of great importance to ensure the number of completed patients is sufficient for “per protocol” statistical analyses. Therefore, it is an issue that should be addressed by protocol authors and clearly specified in the protocol.
The definitions of dropouts and discontinuers are quite straightforward. Reasons for replacement (if any are allowed) should be identified in the protocol and final medical report as well as stating which groups of subjects may be replaced. Patients who leave a trial due to an adverse event may leave on their own accord (i.e., as a dropout) or may be told to leave by the investigator (i.e., as a discontinuer). In neither case should they be replaced. However, replacements are often sought for subjects asked to leave the study for lack of cooperation or for other reasons unrelated to therapeutic outcome. There is a statistical basis for this approach, in that those with adverse events who are discontinued are fully accounted for as part of the study’s results relative to its objectives. Other common reasons for an investigator to discontinue a patient include lack of cooperation and learning after a patient began the study that he/she did not meet the inclusion criteria. In both of these situations, it would seem relevant to replace the patient, even though the patient was officially enrolled, randomized, and may have initiated therapy. Of course, there have been many trials when patients who were found not to meet inclusion criteria were not discontinued.
The author is unaware of any “official” or widely accepted definitions for dropouts and discontinuers, and has proposed those above. There are several kinds of study-leavers, including those who leave a study prematurely relative to the intended protocol-requirements: (a) those who quit on their own accord or volition for personal reasons, unrelated to therapy outcome; and (b) those who leave on their own because they have an adverse event or are unsatisfied with the anticipated efficacy and do not want to continue with the trial’s requirements. The other major category of those who leave early are those who the investigator asks to leave the trial for any of many reasons (e.g., noncooperation, poor compliance, adverse event, postentry realization the patient is ineligible, a placebo responder in a run-in study where every patient is put on placebo and those who demonstrate a positive response are discontinued).
The decision to enroll replacements may depend on the sensitivity of the trial goals to the number needed. Replacing non-cooperative patients is fairly logical, but if they do not cooperate (e.g., noncompliance with taking their drugs), it may be related to poor tolerance to the drug. Both efficacy and safety dropouts and discontinued subjects are expected as outcomes related to the trial design, and they may not have to be replaced, as long as there are enough patients left in the study to fulfill the overall goals. In double blind trials, the sponsor is unlikely to know how many patients will or will not be deemed successes in terms of efficacy or safety. Many companies have their own practices about which categories of dropouts or discontinued subjects they wish to replace and these are specified in their protocols.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
