| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 University of Chicago Medical Center, Chicago, Illinois
Correspondence and requests for reprints should be addressed to Mardi Gomberg-Maitland, M.D., M.Sc., Director of Pulmonary Hypertension, University of Chicago Medical Center, Section of Cardiology, MC2016, 5841 S. Maryland Ave., Chicago, IL 60637. E-mail: mgomberg{at}medicine.bsd.uchicago.edu
ABSTRACT
This article elaborates on the phases of drug development and trial design for pulmonary arterial hypertension, emphasizing the need for learning trials to assess efficacy signals before confirming trials. In this new era in pulmonary arterial hypertension, we need to reconsider innovative trial designs to maximize efficiency and pursue agents with the greatest potential for cure.
Key Words: clinical trials • drug development • pulmonary hypertension
Pulmonary arterial hypertension (PAH) is an angioproliferative vasculopathy resulting from abnormal endothelial and smooth muscle cell interactions. Idiopathic and familial PAH (formerly known as primary pulmonary hypertension [PPH]) occurs more often in women than in men, has a mean age of diagnosis of 35 years, and has a median survival of 2.8 years if left untreated (1, 2). Before the mid-1990s, there were no U.S. Food and Drug Administration–approved therapies. However, patients now have several approved treatment options, including epoprostenol, treprostinil, iloprost, bosentan, and ambrisentan (3–13). Despite these advances and continued epidemiologic understanding, there remains no cure for this disease. Two large referral-based cohorts in Europe and the United States have reported similar data regarding the etiology and severity of disease at referral (14, 15), but prognosis after initiation of therapy and the ability to "stage" the disease is less well studied. It is clear that this is an orphan disease with increased recognition and national attention, better understood pathophysiology, more available diagnostics, and the interest of industry. New therapeutics are still required, but advancements in therapy for this orphan disease will need to follow a different course due to a quickly changing landscape of clinical care and research.
Drug development and trial design goals in an era with no approved treatment are straightforward—that is, to prove that a therapy is safe and effective. After all, "something is likely better than nothing" (16). Traditional trial design in PAH has therefore focused on short-term (12–16 wk) improvement in the six-minute-walk distance (6MWD) rather than on long-term durability, without clear demonstration of vascular remodeling (17).
Currently, patients often present late in the disease course to tertiary centers (14, 15), but they are often evaluated by their local physicians earlier when they are "less sick." Traditional endpoints that adequately differentiated improvement in "very sick" patients may not be sufficient markers of disease progression with this new cohort. Other endpoints for exercise capacity, quality of life (18, 19), and measures of right ventricular (RV) function (20–22) may better assess the evidence of drug therapeutic activity early in drug development. With the growth of specialized centers and local physicians in rheumatology, pulmonology, and cardiology treating this disease, it is more difficult for physicians to wait until evidence from clinical trials is available before initiating therapy. This not only endangers the patient but also limits the availability of untreated patients for trial enrollment to test hypotheses (16). The classic trial designs may therefore no longer be feasible or ethical (16, 23–26). The feasibility of a placebo arm in future PAH trials is controversial, complicating our choices of trial design (27) (see also article by Halpern and colleagues in this symposium, pages 631–636) (48).
This article elaborates on the phases of drug development and trial design for these stages in PAH. The need for learning trials to assess efficacy signals before performing confirming trials will be emphasized. In the current era, we need to pursue agents with the greatest potential for cure and consider innovative trial designs to maximize efficiency and reduce risk to patients and expense. To illustrate the drug development process in PAH and drug development in general, I will trace the development of epoprostenol from preclinical studies to approval for use in PAH.
CLINICAL TRIALS
The goal of a clinical trial is to develop safe and effective therapies efficiently. The primary aim in early drug development is to find evidence for activity of a compound in a proportion of patients with a disease sufficient to warrant larger scale phase III multicenter trials. Early in development, investigators need to learn about the pros and cons of the drug to prepare for upcoming phases and determine if the drug is a suitable candidate (28). Understanding of biological mechanisms can advance this goal. Trials need to be tailored to phase of development, disease state, interpatient variability, and drug toxicities. With an orphan disease such as PAH, this process is fundamental, because subject enrollment will be limited as other therapeutic trials compete for a share of the total cohort of patients.
PRECLINICAL PHASE/PHASE I
Before initiation of clinical trials in humans, it is essential to assess the safety, dose, and pharmacologic and biologic properties of the drug in preclinical models of the disease. Although these trials do not always guarantee clinical efficacy, they are an essential step in the drug development process. With an understanding of the pharmacology and safety from these models, one proceeds to a phase I trial. The goals of these learning trials are to determine how well a drug is tolerated in a small number of people and to determine a safe dose, which may be the maximally tolerated dose (MTD) depending on the disease and goals (28). Phase I trials allow for the "learning" of answers to clinically relevant questions related to dosing that both the trialist and the practicing physician will ultimately need to know (28). For example, what is the appropriate initial starting dose? How soon will the intended effect start and last? Will tolerance develop? What happens if a subject misses a dose? Will the starting dose need to be altered and, if so, in what way? Is there a biologic indicator to guide dose adjustment? If investigators study a drug already approved for a different indication, dosing schemes need to be determined based on the approved dose and, if used in a combination trial, the pharmacology of the other agents (28).
Phase I trials typically test novel therapies that have not been evaluated in humans. Normal healthy volunteers or subjects with the disease of interest are the usual study populations for such trials. Phase I subjects may have failed all standard therapies, as is often the case in cancer trials. In advanced, refractory cancers, there is often no alternative and the risk–benefit ratio for finding the MTD may therefore be acceptable to the subject. In PAH trials, the investigator may also choose subjects who have failed all therapy and are not suitable for lung transplantation or those treated with stable background therapy. Safety and dosing data may be collected on combination therapy. Phase I studies in subjects with PAH who are not receiving any approved therapy are ethically controversial and are not recommended unless the trial is a short-term trial. The primary endpoints are usually safety, such as drug toxicity and adverse events, and pharmacology, such as pharmacokinetics. Efficacy indicators and biologic endpoints are common secondary endpoints.
The typical design of a phase I trial is dose escalation in small patient cohorts until a prespecified frequency of toxicities occur (Figure 1). This design limits exposure and maximizes safety. If an appropriate dose based on previous clinical trials for other disease indications is identified, the investigator may choose to individually dose escalate to an MTD. A step-up approach until toxicity is seen prevents "overshooting" the MTD. The sample size for this study design tends to be between 6 and 24 subjects. Although it is important to perform phase I trials of PAH therapies in normal subjects if there are no human data, it is equally important to determine safety and dosing in subjects with PAH, who may demonstrate different metabolic and side-effect profiles from healthy normal subjects.
|
PHASE II
Phase II studies aim to document evidence of drug efficacy and adverse events on which decisions whether to proceed to a confirmatory phase III trial are based. The exact phase II study design depends on the quality and adequacy of the preceding phase I trial. Typically, a range of doses is tested in phase II. If a number of drugs are tested for the same indication, phase II needs to efficiently determine which drugs are of highest priority for further study.
Phase II studies typically enroll subjects with the disease of interest often narrowed to a specific group of clinical (often with the highest probability of therapeutic response) or commercial/marketing interest. Such an "enriched population" provides the best chance to inform the decision of whether or not to proceed with phase III trials, which enroll more heterogeneous subjects representative of the "general diseased population." Negative results from a small phase II trial of a "nonenriched" study sample could otherwise shelve a drug that may indeed be effective when used clinically.
Phase II endpoints are often surrogate endpoints, which reflect the disease activity and are expected to predict the efficacy of therapy, or intermediate endpoints, which measure how a patient feels (quality of life), and are substitutes for an ultimate endpoint. Surrogate endpoints are useful for shorter term trials, whereas definitive endpoints often require longer investigations. In PAH, phase II studies are ideally suited to use novel endpoints that may signal efficacy, allowing for a rapid determination of favorable drug activity. Secondary endpoints are based on improving the understanding of the pharmacology, dose–response, and variability of the drug itself. New data may be collected regarding the understanding of the disease or specific types of subjects (targeted etiologies) with new biologic and functional endpoints. In PAH, laboratory markers of RV stress (e.g., N-terminal brain natriuretic peptide, brain natriuretic peptide, growth factors, imaging techniques of the RV or pulmonary artery) or measures of exercise capacity (e.g., Naughton-Balke treadmill test) are potential candidates (35).
Phase II—Study Design
The classic design of a phase II trial is a single-arm, two-staged design. The investigators determine a stopping rule for the first stage in which the trial terminates if the response rate is less than expected or not effective. In the second stage, the investigator determines the clinically relevant improvement in the primary endpoint and determines the necessary sample size to prove that the drug is effective to this level. Issues arise with this design in a multicenter trial because it assumes similar response and familiarity of the drug across institutions. Because the design does not have internal controls, it may be unclear if the drug is better than standard therapy. As with all trial design, the validity of the clinical trial is dependent on the predictive ability of the endpoint chosen (36). Flawed endpoints that do not reflect disease activity will yield spurious conclusions (36).
A randomized placebo-controlled phase II trial addresses many of these issues but requires a much larger sample size to determine efficacy and is more costly. In PAH, this trial design may pose difficulties because of the limited numbers of patients available for participation in trials, the ethical issues regarding assignment to placebo if patients are not on background therapy, and the use of non–evidence-based therapeutic combinations.
The randomized discontinuation trial is another phase II trial design, developed more than 30 years ago. An open-label period determines which subjects tolerate the MTD, which have a dramatic response, and which progress. Subjects are then randomized to active drug or placebo. If a subject progresses during the randomized phase of the trial, the investigator "unblinds" the subject. If the subject was receiving active therapy, he or she is withdrawn. If the subject was receiving placebo, he or she then receives active therapy (Figure 2) (37). This design works well if the open-label period is not expected to "cure" the patient and if the objective measures used to determine randomization are able to accurately define true responders with good sensitivity and specificity, conditions likely met in PAH. This design limits subject exposure to placebo and improves efficiency by selecting out patients who do not tolerate the drug due to adverse effects, noncompliance, or progression of disease during the open-label phase (37). On the other hand, this design assumes that withdrawal after the open-label phase will not cause dramatic clinical deterioration, which could be a concern in PAH. This trial design also assumes that investigators have chosen a minimally acceptable time on placebo and that the selected population reflects a broader group of subjects with the disease (37). If carefully designed to address these and other issues with appropriate endpoints, the randomized discontinuation trial can be an efficient, safe, and successful trial design in sick patient populations, such as in those with PAH (36, 38).
|
This trial showed that continued slow dose adjustment allowed for toleration of a higher dose and that a continued response on therapy allowed for further dose increases. The MTD was not clearly determined and further study was necessary. Acutely testing patients with epoprostenol identified those with pulmonary venoocclusive disease (subjects developed pulmonary edema) and thus alerted investigators to be more cautious with testing. Other adverse events included ascites and hemodynamic collapse, subsequent to heparin flushes of the catheter leading to acute medication bolus.
The authors found that changes in 6MWD did not correlate with hemodynamic changes, and they were unable to conclude that changes in exercise were dependent on either treatment. They surmised that the 6MWD may be "too dependent on other factors to be a useful variable to judge clinical response" (39), but future study may state otherwise. The long-term extension of this trial demonstrated improvement in the 6MWD and hemodynamics at both 6 and 18 months (40). Six subjects remained on therapy, four died, and eight received lung transplantation (5 survived). Mean pulmonary artery pressure and mean right atrial pressure inversely correlated with 6MWD. The new side effects observed were predominantly related to the continuous intravenous delivery system, including clotting of the system, pump interruption with subsequent clinical deterioration, and infection of the infusion line (40). On the basis of this study, the investigators designed a phase III trial.
PHASE III
Phase III trials establish both efficacy and the drug's role in clinical practice based on the previous phases of drug development. The typical design is a randomized, double-blind, placebo-controlled trial with the goal of proving that the new intervention is superior to conventional therapy. Phase III trials are larger, of longer duration, and may include a more diverse population compared with phase II trials. The trial may also expand knowledge of drug dosing.
The classic design is based on the "intention to treat" principle, by which patients in the trial are analyzed based on their initial random assignment to a treatment arm, regardless of their adherence with the entry criteria, the treatment they actually received, and withdrawal from treatment or deviation from the protocol (41). The design minimizes bias in estimating treatment effects but will be difficult to interpret if there is significant patient withdrawal from the trial or another cause for a large amount of missing data. Endpoints in phase III need to be clinically relevant, reproducible, well defined, and able to be measured in the desired time frame. Surrogate endpoints may be used, but they need to be well established and usually accompany established endpoints.
Phase III—Epoprostenol
The phase III trial of epoprostenol enrolled 81 subjects with PPH with New York Heart Association functional class III or IV to either epoprostenol with conventional therapy or conventional therapy alone because the use of placebo was deemed unethical (3). Endpoints included 6MWD, hemodynamics, quality of life, and survival. The authors found an improvement in 6MWD by 32 m in the active treatment group compared with a decrease of 15 m in the conventional therapy group (P < 0.003). Hemodynamic improvement was evidenced by a decrease in pulmonary artery pressure of 8% in the epoprostenol group versus an increase of 3% in the conventional group (P < 0.002) and the mean pulmonary vascular resistance decreased by 21% in the epoprostenol group versus an increase of 9% in the conventional group (P < 0.001). Eight patients died during the 12-week trial, all in the conventional therapy group (P = 0.003). The U.S. Food and Drug Administration approved epoprostenol for the treatment of patients with PAH on the basis of the results of this study.
PHASE IV
The last step in drug development is the postapproval trial, considered as phase IV. These trials increase understanding of risk and benefit, identify less common adverse events, refine dosing recommendations, better assess durability of the drug's effect, obtain pharmacoeconomic data, and allow for additional studies of clinical endpoints not assessed in phase III. All designs are possible, including withdrawal studies in the right setting with appropriate safeguards. Patients are enrolled to measure the response to drug withdrawal or drug dose reduction to illustrate efficacy or duration of benefit. Like all designs, by selecting a specified population all of whom tolerate the drug, the efficacy of the drug may be over- or underestimated upon withdrawal. In PAH, this design was mandated for the phase IV trial of subcutaneous treprostinil (42). The prospect of withdrawing effective therapy (intravenous epoprostenol) from patients with PAH receiving monotherapy evoked controversy in the community. Ultimately, the trial safely demonstrated the durability and efficacy of subcutaneous treprostinil.
ALTERNATIVE STUDY DESIGNS
The drug development path of epoprostenol for treatment of PAH did not follow the typical steps, likely because PAH is a rare disease that (at the time) had no approved therapies. Investigators, clinicians, and patients are now faced with new challenges, as we have approved therapies but have many unanswered questions concerning dosing, durability of effect, combination efficacy, tolerance, and long-term side effects. Utilization of nonclassical designs may help with future development and to expand knowledge of current therapeutics. The remainder of this article will focus on these designs.
Equivalence/Noninferiority Studies
Equivalence trials are designed to determine if a new therapy is neither better nor worse than a comparison therapy. The trial is designed in "reverse" of the typical superiority trial in that the trial aims to disprove (or reject) the null hypothesis that there is a difference between the two groups (41). The design requires a gold-standard comparator (based on historical trial data) and knowledge of a prespecifed clinically relevant difference. The margin or range of values (delta) that are to be accepted as similar ("the degree of unacceptable inferiority") is based on historic event rates, odds ratios, or confidence intervals (Figure 3) (41, 43–45).
|
|
The design of both equivalence and noninferiority trials has the potential for sloppy execution to bias the results toward the conclusion that the therapies are no different (away from the null hypothesis, which, in this case, is that the two therapies are different). Proper reporting of trial results is critical to support the trial conclusions. Suggested criteria for presentation are as follows: a comparison table of inclusion and exclusion criteria with the reported trial and the historic trial of the standard therapy; a detailed description of all patients screened, enrolled, withdrawn, and completed, including those who crossed over; a report of projected and actual patient-years with the impact of withdrawals and crossovers on exposure to initial assignment; a discussion of the rationale for the margins chosen; an account of the minimum number of events expected; and a comparison of the number of events observed in the trial compared with the number observed in previous trials (41).
Factorial Design
Factorial designs are best used when there is a biologic, clinical, or pharmacologic rationale for adding more than one drug together (Figure 5). The investigator may wish to study the interaction of two therapeutic approaches, if such an interaction exists. On the other hand, an investigator may want to test two different therapies simultaneously, which may be accomplished in an efficient manner in a factorial study. The benefits of this design are as follows: (1) the ability to test multiple hypotheses at once, (2) the potential demonstration that each therapy works alone and in combination, and (3) greatly reducing the sample size from that necessary to study each hypothesis individually. Each component of the trial is predetermined and therefore the P value does not need adjustment for multiple testing (46). To maintain this efficiency, it is critical that there is no interaction between the two or three therapies being studied (46). In addition, it must be safe to administer the therapies together, and it is best if the therapies have different mechanisms of action.
|
Crossover Design
Crossover designs also allow for trial efficiency and a smaller sample size compared with usual parallel trial designs. The design is commonly used in both phase II and phase III trials. Each treatment is given to each subject at different times and each subject's placebo period is compared with his/her own active drug period. Intrasubject variability is less than between-subject differences, thus greatly limiting the necessary sample size and reducing confounding.
The potential benefits of the crossover design come with some cost. This design is only useful for chronic diseases in which treatments provide measurable effects in a short time frame. Such trials have a longer duration with multiple assessments, increasing subject burden. The design also assumes a short time for washout, that there are no carry-over treatment effects, and that treatment does not completely change/cure the disease, which may not hold true for some PAH therapies. It also assumes that there are no treatment-by-period interactions, meaning that the order of treatment does not affect the result. Frequently, dropouts plague this design because the participants' burden is higher, the drug exposure is longer, and the subject may experience more toxicity. Dropouts may result in uninterruptible results. The uncertainty of washout time and the effects of drug withdrawal have mostly limited the use of this trial design in PAH. However, a phase II randomized clinical trial of antiplatelet therapy for PAH used a crossover design (47), demonstrating the usefulness of this approach if subjects are on background therapy, the trial duration is short, and early-phase studies are necessary to gather preliminary data and test feasibility for later-stage trials.
Drug development is an expensive, arduous process of learning that entails a team approach from phase I through phase IV. Clinical scientists and industry need to work together to design trials that answer the clinical and biologic questions while determining which therapies are best to proceed to approval. Trials in PAH starting from the preclinical phase through phases I and II should focus on learning to improve efficiency of the entire enterprise. Given the limited number of patients available to participate in investigations, all phases, including phase III trial designs, must face the challenge of gaining the most knowledge with the fewest patients placed at the least risk.
FOOTNOTES
M.G.-M. is supported by the Doris Duke Clinical Scientist Development award.
Support for this conference, including travel for each of the authors, was provided by unrestricted educational grants from Actelion Pharmaceuticals, Pfizer, Gilead Sciences, United Therapeutics, and Lung Rx, Inc.
Conflict of Interest Statement: M.G.-M. has received research grant support from Actelion, Co-Therix, Encysive, Gilead, Lilly/Icos, Pfizer, and United Therapeutics, and has served as a consultant and/or on Advisory Boards for Encysive, Gilead, Pfizer, and United Therapeutics.
(Received in original form March 4, 2008; accepted in final form April 11, 2008)
REFERENCES
Related articles in Proceedings of the American Thoracic Society:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
