Reinventing the phases of clinical trials
Traditional clinical trials, in which a protocol is designed according to certain assumptions and strictly followed throughout the study, have numerous inefficiencies. Because there is no opportunity for researchers to adapt a protocol to reflect their increasing understanding of a drug as a trial progresses, the outcome may be inconclusive if the initial assumptions prove incorrect. The concept of reinventing the phases of clinical trials by certain adaptations made to the trial and/or statistical procedures of on-going clinical trials based on accrued data have been in practice for years in clinical research and development, thereby revolutionizing the way clinical research is being conducted. The potential benefits of adaptive trials are exciting, greatly enhancing the drug development process by requiring fewer patients and a reduced time to meet study endpoints. However, adaptive trials are challenging to implement. They must be designed carefully before they begin, with full scenario planning (including stoppage rules, decision trees and other important details) agreed beforehand with regulatory authorities. The purpose of this short review is to provide a comprehensive and unified presentation of the principles and methodologies in adaptive design and analysis with respect to adaptations made to the trial along with a well-balanced summary of current regulatory perspectives.
Keywords: Clinical trial, adaptive designs, statistical procedures, phases, regulatory
In recent years, the concept of reinventing the phases of clinical trials utilizing the adaptive clinical trial designs has been debated as an exciting alternative to revolutionize the way clinical research is conducted. The term implies making planned, well-defined changes in key clinical trial design parameters, during trial execution. Based on the data from that particular trial, it is possible to simultaneously achieve goals of validity, efficiency, and safety. Such a flexible design monitors the accruing efficacy data at administratively convenient intervals and makes important decisions concerning the future course of the study. Hence, adaptive design is defined as a design that allows adaptations or modifications to some aspects of the clinical trial such as statistical procedures of the trial after its initiation without undermining the validity and integrity of the trial . Adaptive designs are also used to allow mid study changes that are highly probable to maximize the success of the trial by preserving the Type I error rate and to enrich the trials with subgroup of patients having genomic profiles likely to respond or less likely to respond to toxicity . Adaptive designs have multiple steps and adaptations are made on the basis of updated information to exploit the probability of success of the clinical trial. To achieve validity, efficacy and safety through these adaptive designs, the various phases of clinical trials needed to be reinvented. Comparing adaptive designs with phases of classical trial design, the following reinventions could be placed phase-wise [Figure I and figure II]. The following text moves from phase I to phase IV of classical clinical trial design, being maneuvered from phase 0 to phase IV, thereby forecasting the advancements in the study designs.
Phase 0 (Human Microdosing):
Before the initiation of phase I clinical trial, a new experimental and efficient approach known as phase-0 or human micro-dosing studies has been developed. Central to this approach is the concept that ‘the best model for man is man’. It refers to the testing of only 1 percent (or 100 micrograms, whichever is smaller) of a pharmacological dose in humans, thereby facilitating the determination of the first dose for the subsequent Phase I clinical trial [3, 4]. Human micro-dosing is an innovative approach to speed up the selection of lead drug candidates by early characterization of the pharmacokinetics of new chemical entities such as clearance, volume of distribution, half-life etc. with advanced analytical techniques like accelerator mass spectroscopy (AMS) and LC/MS/MS, thereby offering a way of developing drugs in a faster, more cost effective and ethical way than ever before . Pharmacokinetic (PK) analysis of microdoses of drug, in low picogram to femotgram range, requires positron emission tomography (PET) technique that relies on the assessment and analysis of the radio isotopes incorporated into the drug under study . PET can also be used to provide pharmacodynamic (PD) information, for example receptor selectivity or occupancy profile, through the use of short half-life isotopes .
In Phase I studies, the primary objective is to determine the maximum tolerated dose that can be administered to a subject with some acceptable level of toxicity. Traditional approaches have relied on the “1-in-3” or 3+3 design which exposes groups of 3 subjects to a dose level and observes the response from the subjects. Based on the number of subjects demonstrating DLT (dose limiting toxicities), the next group of 3 subjects is treated at the same dose level or one level above or below . Such an approach is typically unreliable, unpredictable and likely to choose a lower therapeutically ineffective dose.
Adaptive dose finding design is used in early phase clinical trials to identify the minimum effective dose (MED) and or the maximum tolerable dose (MTD). The MTD is used to determine the dose used for the next trial. More advanced dose-escalation rules have been developed using modeling approaches such as the Continual Reassessment Method (CRM, 1999) and other accelerated escalation algorithms, that can reduce the sample-size and overall toxicity in a trial and improve the accuracy and precision of the MTD predictions [8, 9]. These are useful in controlling the Type I error .
Continual Reassessment Method, the very first Bayesian-based design for dose-finding studies in Phase I, has received much attention since its first proposal in 1990 . The method targets a specific probability of toxicity that is computed from both historical data and data from the current group of subjects treated. To implement the CRM, the probability of toxicity at each of the selected dose levels and the desired level of toxicity must be specified. The dose-toxicity curve is evaluated and the next group of subjects is treated at the dose for which the posterior mean is closest to the target probability of toxicity. Simulations show that the CRM is much more reliable than the 3+3 method. Widespread use of the CRM may have been limited due to the specialized software needed. Regardless, since its introduction, there have been several variants to the original CRM method, including the Modified CRM, Extended CRM [2 stage], Restricted CRM and Tri-CRM.
Whereas the CRM and its variants focus on safety, i.e. evaluating the MTD dose, an alternative innovative approach that combines Phase I and Phase II objectives, i.e. efficacy and toxicity has been explored. One such approach, based on Bayesian principles involves specifying 3 factors that are specified in consultation with the investigator :
- The upper limit for the probability of toxicity,
- The lower limit for the probability of efficacy and
- Three desirable targets for efficacy and toxicity
The first 2 limits are motivated by the desire to limit the risk of treating patients at a dose with either unacceptably high toxicity or unacceptably low efficacy.
Phase II (Proof-of-Concept Studies):
The proof-of-concept study is also open to innovative designs. Phase II proof-of-concept studies employ a small, targeted number of subjects to determine if there is enough evidence of clinical efficacy to warrant full-scale development.
Designs that address the limitations of traditional phase II designs include:
a)Drop-the-losers design (two stage design) permits the dropping of inferior treatment groups and the addition of new sections which is useful in phase II clinical development with respect to uncertainties in dose levels . It is affected by the selection criteria and decision rules. Drop- the-loser design is a two-stage design, wherein, the inferior arms will be dropped at the end of the first stage based on pre-specified criteria, and the winners will proceed to the next stage.
b)Adaptive seamless phase design addresses objectives that are achieved through separate trials in phase IIb and III of clinical development in a single trial . It is a two-stage design with a so-called learning stage and (phase IIb) and a confirmatory stage (phase III). It uses data from patients enrolled before and after the adaptation in the final analysis. However, the validity and efficiency of this design have been challenged as it does not give a well defined guide on the performance of combined analysis if there are similar objectives at different phases.
c)Two-stage Simon designs(minimax design) are optimal in the sense that the expected sample size is minimized if the regimen has low activity, subject to constraints upon the size of the type 1 and type 2 errors .These designs can also be utilized for pilot studies of new regimens to determine the toxicity endpoints. However, in this case, the hypothesis should be specified in terms of the probability of no toxic event.
Phase III (Confirmatory trial)
The goal of the late drug development phases (phase III) is to demonstrate an acceptable benefit/risk profile in a large and representative patient population . Failures and successes in phase III clinical trials have the greatest potential impact on R&D costs and company valuation and yet failure at this stage is unacceptably high.
In recent years, due to ethical, scientific and disastrous economic consequences of late stage drug failures; there has been a magnanimous interest in the development of novel predictive technologies to help researchers select only the most promising candidates for clinical development. One of these reinventive technologies include the use of Group sequential design, which permits the premature stopping of a trial due to safety, efficacy or futility or both with options of additional adaptations as a result of interim analysis. It is based on the assumption that the sequence of a normally distributed test statistic with unknown mean and known variance with proportional hazards accommodates the censored event times. If these proportional hazards are not met, then the designs and actual power may differ substantially from its nominal value . Various stopping boundaries based on different boundary functions for controlling an overall type I error rate are available in the literature [18, 19]. In recent years, the concept of two-stage adaptive design has led to the development of the adaptive group sequential design [20, 21]
Phase IV (Post Marketing Surveillance):
For phase IV studies, the adaptive design can also be applied when a combination of two approved drugs is used. This can again be done through simulations, wherein, the strategy starts with 5 combinations, inferior arms are dropped and the probability of selecting the best arm under different combinations of sample size is calculated at the interim & final stages. However, the situation is different because the drug has already been approved during phase IV clinical development, wherein, the focus is safety rather than efficacy .
Advanced Adaptive Designs
These designs could be tailor-made as per the requirements and can be incorporated in the respective phases [Figure II].
Adaptive randomization design permits the modification of randomization schedules based on varied and /unequal probabilities of treatment assignment in order to increase the probability of success. Frequently applied adaptive randomization processes include treatment - adaptive randomization, covariate- adaptive randomization and response-adaptive randomization [23-27]. Despite the fact that an adaptive randomization trial may increase the probability of success, it may not be feasible for use in a large trial or a trial with a relatively long treatment duration since the randomization of a given subject is dependent on the response of the previous subject.
Sample size re-estimation design is an adaptive design that tolerates sample size adjustment based on the interim observed data. This type of re-estimation can be done in a blinding or unblinding manner based on the size of treatment effect-size, conditional power and/or reproducibility probability [28-30]. The disadvantage to sample size re-estimation is the same as the original power analysis for its calculation before the study since it is performed by treating estimates of the study parameters, which are based on data observed at interim, as true values.
Biomarker-adaptive design entails biomarker qualification and standard, optimal screening design, model selection and validation. A prognostic biomarker provides information about the natural course of the disease in individuals who have or have not received the treatment under study and informs the clinical outcomes, independent of treatment. In contrary, a predictive biomarker informs the treatment effect on the clinical endpoint . On the whole, a biomarker-adaptive design can be used to select the right patient population, identify course of disease, early detection of disease, and assists in the development of personalized medicine.
Adaptive treatment-switching design permits the investigator to switch a patient’s treatment from an initial assignment to an alternative treatment if there is no evidence of efficacy or safety of the initial treatment [31, 32]. The disadvantage of this design in oncology treatments is the challenge regarding the estimation of survival in some of the patients. The percentage of patients that switch due to disease progression is high and this could lead to a change in the hypothesis to be tested.
Adaptive- hypothesis design permits changes in the hypothesis based on interim analysis results . The method is used before data lock and/or prior to data unblinding. Examples include a switch from superiority hypothesis to a non- inferiority hypothesis. The selection of the non-inferiority margin has a critical impact on sample size for achieving the desired power.
Multiple adaptive design is made up of different combinations of the designs above. Examples include:
a)The combination of adaptive group sequential design, drop-the- losers design and adaptive seamless trial design;
b)Adaptive dose escalation design and adaptive randomization .
Benefits with the reinvention approaches:
- Flexibility- In principle, there is no need to pre-specify the type and details of the adaptations in advance. Adaptive design allows for changes of essential features of the design based on data from inside or outside the ongoing trial without compromising on the type I error rate.
- Improve the power: Adaptive sequential procedures can improve the power when promising hypotheses are screened at earlier stages with small sample sizes to be investigated further at later stages with larger sample size at constrained costs.
- Sequential Information: The information is gathered sequentially.
- Goal based approach: The trial is stopped when goal of the experiment is achieved
- Efficient utilization of time and resources: More efficient use of time and resources leading to faster drug development
- Timely decision-making: Critical decisions are pushed earlier in the development process
- Faster detection of safety issues
- Patients are not unnecessarily exposed to non-efficacious drugs
Drawbacks and deficiencies:
- More complex statistics - For example, in gene expression or gene association studies the experimenter deals with a an extremely large number of markers (and associated hypothesis), but due to constraints on resources, sample sizes are rather small
- It does not reflect real practice
- It may not be flexible always
The Regulatory Approach:
Although the flexibility in design and analysis of clinical trials in early phases of the drug development is very attractive to clinical researchers and the sponsors, its use in late phase II or phase III clinical investigations has led to regulatory concerns regarding its limitation of interpretation and extrapolation from trial results. In the light of the above, the European Agency for the Evaluation of Medicinal Products (EMEA) published a concept paper [34, 35]. The paper focused upon the pre-requisites and conditions under which the methods could be acceptable in confirmatory phase III trials for regulatory decision making including the methods providing correct p-values, unbiased estimates, and confidence intervals for the treatment comparison(s) in an actual clinical trial.
Since safety is an issue, the following are some expectations from the regulatory bodies regarding clinical trial using adaptive design methods:
ü It should be prospectively planned
ü It should have valid statistical approaches on modification of design elements that have alpha control and can be defined in terms of ICH E-9
ü It should have valid point estimates and confidence interval estimates
ü It should build on experience from external trials
ü It should take a learn and confirm approach
ü It should include documentation of actual monitoring process, extent of compliance and potential effect on study results
Enabling technologies while Implementing and Managing Adaptive Designs for Clinical Trials:
Implementation of technology is critical because it empowers clinical teams with real-time information, and then enables them to plan and quickly implement seamless changes in response to that information. The key enabling technologies for adaptive trial design are interactive voice response (IVR), electronic data capture (EDC), and clinical trial material forecasting systems [Figure II].
Since most modifications in an adaptive trial design setting involve samples size, dosing, and/or randomisation, interactive voice response systems (IVRS) is a key enabling technology in this regard and find its clinical uses in the dynamic treatment allocation or automated dose titration .
Electronic data capture (EDC) is the concept of collecting patient clinical data directly from the investigative site electronically, and transmitting the data over the Internet, resulting in better quality data that is available much more quickly than paper-based alternatives .
Clinical Trial Material Forecasting Systems
Since the ability to plan appropriate drug supply strategies, in support of potential treatment arm modifications and study samples size adjustments, is critical to the success of many adaptive trials, clinical trial material forecasting technology is a useful vehicle to analyze and simulate patient enrolment and project time-phased patient demand of clinical supplies .
Since reinventing the phases of clinical trial by the aid of adaptive design is relatively new and can increase the outcome of clinical trials, as well as making more drugs available in the market, it is essential that the regulatory bodies, in conjunction with the scientists, pharmaceutical companies and patients work together to understand its use, work out the challenges of the process, with well-defined standard rules to be followed.
1.Chang M, Chow SC, Pong A. Adaptive design in clinical research, Journal of Biopharmaceutical Statistics 2006, 16, 299-309.
2.Hung HM, O’Neill RT, Wang SJ, Lawrence J. A regulatory view on adaptive/flexible clinical trial Design, Biometrical Journal 2006, 48(4), 565-573.
3.Arrowhead Publishers' new report: Innovations and Trends in Clinical Trials: Microdosing (Phase 0 Trials), Adaptive Trials, Phase IV Trials and the Role of Information Technology. Minneapolis, MN (PRWeb) March 20, 2007.
4.Kummar S, Rubinstein L, Kinders R, Parchment RE, Gutierrez ME, Murgo AJ, Ji J, Mroczkowski B, Pickeral OK, Simpson M, Hollingshead M, Yang SX, Helman L, Wiltrout R, Collins J, Tomaszewski JE, Doroshow JH. Phase 0 clinical trials: conceptions and misconceptions, Cancer Journal 2008, 14 (3), 133-137.
5.Lappin G. Kuhnj W.Jochemsen R, Kneer J, Chaudhary A, Oosterhuis B, Drijfhout WJ, Rowland M, Garner RC. Use of microdosing to predict pharmacokinetic at the therapeutic dose, Experience with five drugs, Clinical Pharmacology & Therapeutics 2006, 80, 213-215.
6.Lappin G and Garner RC Big physics, small doses – the use of AMS and PET in human microdosing of development drugs. Nature - Drug Discovery 2003, 2, 233-240.
7.Potter DM. Adaptive dose finding for phase I clinical trials of drugs used for chemotherapy of cancer, Statistics in medicine 2002, 21(13), 1805-1823.
8.Chevret S (ed.), Statistical methods for dose-finding experiments. John Wiley & Sons Ltd. West Sussex, England. 2006.
9.Chang M, Adaptive Design Theory and Implementation Using SAS. R. Chapman & Hall/CRC, Taylor & Francis, NC. 2007.
10.Frank B, Jason H, José P, Yi L. Dose Finding - A Challenge in Statistics, Biometrical Journal 2008, 50(4), 480-504.
11.O'Quigley J, Pepe M, Fisher L. Continual reassessment method: a practical design for phase I clinical trials in cancer, Biometrics 1990, 46, 33-48.
12.Thall PF, Cook J. Dose-finding based on toxicity-efficacy trade-offs, Biometrics 2004, 60, 684–693.
13.Michael WS, Allan R. Sampson. Drop-the-losers design: Binomial case Computational Statistics & Data Analysis 2009, 53(3), 586-595
14.Christopher J, Bruce WT. Adaptive seamless designs: selection and prospective testing of hypotheses. 2006 62nd Deming Conference on Applied Statistics., Atlantic City, NJ, USA(2006).
15.Richard S. Optimal Two-Stage Designs for Phase II Clinical Trials, Controlled Clinical Trials 1989, 10, 1-10.
16.Thaddeus HG, Jill FK, Cynthia AW, Joel SO, Brenda BC, Kathleen ER, Elizabeth AL, Julie AP, Charles WD. Challenges in the Transition to Model-Based Development, The AAPS Journal 2005, 7 (2), E488-E495.
17.Wathen JK, Thall PF. Bayesian adaptive model selection for optimizing group sequential clinical trials, Statistics in Medicine 2008, 27(27), 5586-5604.
18.Chow SC, Chang M. Adaptive Design Methods in Clinical Trials. Chapman and Hall/CRC Press, Taylor and Francis, New York, USA. 2006.
19.Jennison C, Turnbull BW. Meta-analysis and adaptive group sequential design in the clinical development process, Journal of Biopharmaceutical Statistics 2005, 15(4), 537-558.
20.Lehmacher W, Wassmer G, Adaptive sample size calculations in group sequential trials, Biometrics 1999, 55, 1286-1290.
21.Liu Q, Proschan MA, Pledger GW. A unified theory of two-stage adaptive designs, Journal of American Statistical Association 2002, 97, 1034-1041.
22.Shein CC, Mark C. Adaptive design methods in clinical trials – a review, Orphanet Journal of Rare Diseases 2008, 3, 11
23.Efron B. Forcing a sequential experiment to be balanced, Biometrika 1971, 58, 403-417
24.Lachin JM. Statistical properties of randomization in clinical trials, Controlled Clinical Trials 1988, 9, 289-311.
25.Rosenberger WF, Stallard N, Ivanova A, Harper CN, Ricks ML. Optimal adaptive designs for binary response trials, Biometrics 2001, 57, 909-913.
26.Hardwick JP, Stout QF. Optimal few-stage designs, Journal of Statistical Planning and Inference 2002, 104, 121-145.
27.Posch M, Bauer P. Adaptive two-satge designs and the conditional error function, Biometrical Journal 1999, 41, 689-696.
28.Cui L, Hung HMJ, Wang SJ. Modification of sample size in group sequential trials, Biometrics 1999, 55, 853-857.
29.Shih WJ. Group sequential, sample size re-estimation and two-stage adaptive designs in clinical trials: a comparison, Statistics in Medicine 2006, 25, 933-941.
30.Chung SC, Anderson K, Gallo P, Collins S. Sample size re-estimation: a review and recommendations, Drug Information Journal 2006, 40, 475-484.
31.Chang M, Chow SC. A hybrid Bayesian adaptive design for dose response trials, Journal of Biopharmaceutical Statistics 2005, 15, 667-691.
32.Branson M, Whitehead W. Estimating a treatment effect in survival studies in which patients switch treatment, Statistics in Medicine 2002, 21, 2449-2463.
33.Hommel G. Adaptive modifications of hypotheses after an interim analysis, Biometrical Journal 2001, 43, 581-589.
34.EMEA (2002). Point to Consider on Methodological Issues in Confirmatory Clinical Trials with Flexible Design and Analysis Plan. The European Agency for the Evaluation of Medicinal Products Evaluation of Medicines for Human Use. CPMP/EWP/2459/02, London, UK.
35.EMEA (2006). Reflection paper on Methodological Issues in Confirmatory Clinical Trials with Flexible Design and Analysis Plan. The European Agency for the Evaluation of Medicinal Products Evaluation of Medicines for Human Use. CPMP/EWP/2459/02, London, UK.
36.Bill Byrom. Managing the Medication Supply Chain Process Using Interactive Voice Response Systems, Life Science Today, 2002, 16-18.
37.Emam EE, Jonker E, Sampson M, Krleža-Jerić K, Neisa A. The Use of Electronic Data Capture Tools in Clinical Trials: Web-Survey of 259 Canadian Trials, Journal of Medical Internet Research 2009, 11(1), e8
38.Rosenberg WF. New Directions in Adaptive Designs, Statistical Science 1996, 11(2), 137-149.
Figure I: Schematic representation of the New Drug Development Process highlighting the duration, cost and number of subjects involved in each clinical trial phase
*Source: “Clinical Operations:Accelerating Trials, Allocating Resources and Measuring Performance” (http://www.ClinicalTrialBenchmarking.com), published by business intelligence firm Cutting Edge Information.
Figure II: Schematic representation of the various adaptive designs for clinical trials
Dr.Parloop A Bhatt
Assistant Professor, Department of Pharmacology, L.M. College of Pharmacy, Ahmedabad- 380009, Gujarat, India.
Manojkumar V Patel
Vibgyor Scientific Research Private Limited, Ahmedabad, Gujarat, India
Somsuvra B Ghatak
Vibgyor Scientific Research Private Limited, Ahmedabad, Gujarat, India
Dr. Keyur H Parikh
The Heart Care Clinic, Ahmedabad, Gujarat, India