Click for Complimentary Power Calculator
What is the Sequential Parallel Comparison Design?
SPCD is a clinical trial design paradigm which was proposed by Dr. Maurizio Fava and Dr. David Schoenfeld of Massachusetts General Hospital (“MGH”) in 2003. SPCD can increase signal detection by:
(a) utilizing at least two stages of treatment
(b) using results data twice from subjects who are placebo non-responders
(c) reducing placebo response
In comparison to a conventional parallel trial design with a specified “n” (sample size), use of SPCD with the same “n” can, over a wide range of treatment responses, increase the power of a trial - - typically by 10 - 25 percentage points. Similarly, if a particular power is specified, SPCD can allow a reduction of “n” - - typically by 20% - 50%. These benefits are attainable whether placebo response is high or low, as long as the drug or other therapy being tested does have some therapeutic benefit.
Stated another way: use of SPCD can (and has, in completed trials) provide a large reduction in the p-value. P-Value Reduction
SPCD Highlights -- Use in Corporate, Government and Non-Profit Sponsored Trials
1) The first pivotal trial using SPCD has completed enrollment. NCT01318434
2) Fourteen SPCD trials are completed, ongoing, or will start soon. Three are funded by NIH.
3) SPCD Trial Results
(a) Alkermes recently announced positive results from its Phase 2 SPCD study of ALKS 5461 for
treatment of Major Depressive Disorder, and indicated its plans to request a meeting with the
FDA and advance ALKS 5461 into a pivotal development program. NCT01500200
(b) Pamlab, Inc. funded a trial examining L-methylfolate as adjunct treatment for depression. With an
“n” of 75, the p-value was < 0.05. If a conventional design had been used, the p-value would
have been 0.12. NCT00955955
(c) Bristol-Myers Squibb funded a trial examining aripiprazole as adjunct treatment for depression. A
p-value of 0.17 was achieved, versus what would have been 0.86 if a conventional design had
been used. NCT00683852
4) SPCD can be particularly suitable for:
- pivotal trials with expected high placebo response
- proof-of-concept trials
- trials conducted under post marketing commitments
- trials in pediatric and adolescent populations
- orphan disease trials
Description of an SPCD Trial
An SPCD trial involves at least two double-blind stages of treatment, with Stage 2 commencing immediately at the conclusion of Stage 1. Neither subjects nor clinicians are aware of when Stage 1 ends and Stage 2 begins. Typically, all subjects from Stage 1 enter Stage 2, but only the subjects who are identified as not responding to placebo ("placebo non-responders", or "PNRs") during Stage 1 are included in the efficacy analysis of Stage 2. Therefore, SPCD involves enrichment in Stage 2. The other subjects from Stage 1 typically enter Stage 2 for blinding purposes and collection of useful data, but are not used in the efficacy analysis of Stage 2 per se. Therefore, data from both stages are utilized for the efficacy analysis, with response data of all subjects utilized at least once, and response data of PNRs utilized twice.
Stage 1 of an SPCD trial is aimed at: (1) comparing drug and placebo, as in a conventionally designed single stage trial (with the expectation that the drug-placebo difference will be of “normal size”) and (2) generating a cohort of PNRs. Stage 2 is aimed at comparing drug and placebo, as in a conventionally designed single stage trial design, but utilizing in the efficacy analysis individuals who were PNRs in Stage 1 (with the expectation that the drug-placebo difference in Stage 2 will be approximately the same or greater than in Stage 1).
Note: Here we refer to the active treatment as “drug”, but such treatment can also be a medical device or other type of therapy.
SPCD is a design which can include (i) the use of an “active comparator”, (ii) the study of a therapy which is adjunctive to a “standard of care” therapy, or (iii) multiple doses. The efficacy analysis of an SPCD trial can involve binary or continuous data. An SPCD trial may be preceded by a placebo lead-in stage, although typically sponsors have not used one.
The diagrams below illustrate two of the possible formats: Format 1, involving two randomizations, and Format 2, involving a single randomization. A more detailed explanation is set forth in the White Paper and the Slide Deck.
Preservation of Type 1 Error Rate
Over the past 10 years, many biostatisticians have reviewed SPCD and have determined that: (1) there are a number of efficient methods of aggregating the outcome data (categorical or continuous) that take into account the potential correlation of observations from subjects included in more than one stage, and (2) there are a number of valid test statistics that preserve the type 1 error rate. For example, in their original paper, Fava et al (2003) used a linear combination test. Other statistical tests have been proposed by Tamura and Huang (2007), Huang and Tamura (2010), Ivanova et al. (2011), Chen et al. (2011), Liu et al. (2012) and Doros et al. (2012). Comments with respect to the type 1 error rate include: Ivanova et al.: “All tests preserve the type 1 error rate rather well…”; Chen et al.: “…the weighted test statistic based on MMRM estimates appears to be the most robust test statistic for SPD-ReR in terms of type 1 error control, power performance, and estimation accuracy.”; Liu et al.: “From Table 2, it is seen that the simulated type 1 error rates are very close to the theoretical value a = .025.”; Doros et al.: “Our extensive simulations show that when compared with the other methods, our approach preserves the type 1 error even for small sample sizes and offers adequate power and the smallest mean squared error under a wide variety of assumptions.”
Identifying Placebo Non-Responders
When determining which patients to identify as PNRs in Stage 1 (who will then be included in the efficacy analysis of the enriched Stage 2), the appropriate definition of PNR depends upon the illness and therapy being studied. For example, for some illnesses, a PNR could be defined as a subject who "experienced an improvement in a relevant outcome measure of less than 50% during Stage 1, regardless of the level of severity". In another situation, a PNR could be defined as a subject who "had an illness severity rating (of a relevant outcome measure) of more than X at the end of Stage 1, regardless of the degree of improvement".
For illnesses such as Major Depressive Disorder, for which treatment outcome may be affected by a "Flooring Effect" (i.e., the ability to detect a signal is diminished when the level of illness severity is fairly low), experience has indicated that a suitable definition of PNR may involve dual criteria - - e.g., a PNR is “a subject who:
(a) During Stage 1 experienced a reduction in HAM-D score of less than 50% and
(b) Had a HAM-D score at the end of Stage 1 of 16 or higher."
Note: In SPCD trials which use a binary (i.e., categorical) outcome measure in Stage 1 and in Stage 2 (i.e., subjects are identified as responding or not responding to drug and to placebo), the definition of PNR used for enrichment (to determine which patients are included in the analysis of Stage 2) may or may not be the same as the definition of PNR used in the binary outcome measure itself. For example, the definition of PNR for enrichment may be defined as in “(a)” and “(b)” above, but the definition of PNR in terms of the outcome measure may be defined as in “(a)” only. Under these circumstances, the "count" of PNRs in Stage 1 could be somewhat larger than the number of subjects in Stage 2 who are included in the efficacy analysis.
The concept for all SPCD trials is that a Stage 1 PNR who is included in the efficacy analysis in Stage 2 should not have improved so much in Stage 1 that (i) further material improvement in Stage 2 is unlikely, or (ii) because the severity of the illness has significantly diminished, the natural course of the illness would involve further improvement, and therefore distinguishing drug versus placebo in Stage 2 is unlikely.
Why choose SPCD? An example of SPCD's economic benefits
Consider an example wherein a sponsor seeking 80% power will require 346 subjects using a traditional single stage trial design, with an expected total cost of approximately $10,400,000 ($30,000 per subject), for a new drug with anticipated profits of $500,000,000 per year. Based on reasonable assumptions, if SPCD were used, then only 199 subjects would be required and the benefits derived would be:
1. Direct Cost Savings of approximately $4,000,000 assuming the sponsor maintains power at 80%. Should the sponsor choose to increase the power to 90% by utilizing 266 subjects, there would still be a cost savings of approximately $2,000,000.
2. Additional Profit If Earlier Time-To-Market is Achieved of approximately $120,000,000 This figure takes into account the reduction in total trial time achievable because of the reduction in number of subjects as a result of using SPCD with 80% power.
3. Order-of-Magnitude Greater Benefit If A Failed Trial Is Avoided SPCD can provide significantly higher power. If this additional power allows a sponsor to avoid a trial failure, the cost of replicating the entire trial is avoided and, more importantly, the new therapy may be available to subjects (with revenues generated for the sponsor) 1 to 3 years earlier.
Completed SPCD Trials: P-Value Reduction Achieved
The following tables show results of two SPCD trials. The first trial, sponsored by Pamlab Inc., examined the use of L-methylfolate as adjunctive treatment to SSRI's, for patients with Treatment Resistant Depression. In this case, the difference in response between drug and placebo was almost the same size in both stages of the trial. Nonetheless, the p-value using SPCD was less than half of the p-value of the first stage alone The following tables show results of two SPCD trials. The first trial, sponsored by Pamlab Inc., examined the use of L-methylfolate as adjunctive treatment to SSRI's, for patients with Treatment Resistant Depression. In this case, the difference in response between drug and placebo was almost the same size in both stages of the trial. Nonetheless, the p-value using SPCD was less than half of the p-value of the first stage alone (i.e., the p-value which would have resulted in a conventional, single stage trial).
The second trial, “ADAPT-A”, was funded by Bristol-Myers Squibb. It examined the use of a low dose of Aripiprazole as adjunctive treatment to SSRI’s, also for patients with Treatment Resistant Depression. There was only a small difference in response between drug and placebo inthe first stage of the trial, which is equivalent to a conventional, single stage parallel trial. However, the difference was approximately 9 times larger in the second stage. A p-value of 0.17 was achieved, versus what would have been 0.94 if a conventional design had been used.
Illnesses with Significant Placebo Response
RCT Logic has identified over 40 illnesses that have been cited in Medical Journals and pharmaceutical industry related articles as having significant placebo response in clinical trials. Click to see a partial list of illnesses with significant placebo response. Note, however, the benefits of SPCD, in terms of reducing p-value, can be achieved whether placebo response is high or low.
SPCD Published Literature
Authors from FDA, Eli Lilly, Sanofi-Aventis, Johnson & Johnson and several academic institutions have published 14 articles solely or substantially about SPCD. In aggregate, through January 2013 over 140 articles have cited the original SPCD article by Fava et al published in 2003.
Selected publications focused solely or substantially on SPCD are listed in the following annotated bibliography. Notes indicate where authors have described SPCD using terms other than “Sequential Parallel Comparison Design” or “SPCD”, the name and acronym designated by Fava and Schoenfeld in the 2003 Patent Application and article.
Boessen R., Knol M., Groenwold R., Grobbee D., Roes K.: Increasing trial efficiency by early reallocation of placebo nonresponders in sequential parallel comparison designs: Application to antidepressant trials; Clin Trials 2012 9:578 DOI: 10.1177/1740774512456454
The authors refer to SPCD as “SPC”.
Chen Y., Yang Y., Hung H., Wang S.: Evaluation of performance of some enrichment designs dealing with high placebo response in psychiatric clinical trials; Contemporary Clinical Trials 32 2011; 592- 604.
The authors refer to SPCD Format 1 as the “SPD-ReR” and refer to SPCD Format 2 as the “SPD”.
Doros G., Pencina M., Rybin D., Meisner A., Fava M.: A Repeated Measures Model for Analysis of Continuous Outcomes in Sequential Parallel Comparison Design Studies; Statistics in Medicine 2013 DOI: 10.1002/sim.5728
Fava M., Schoenfeld D.: Several issued U.S. patents including Nos. 7,647,235; 7,840,419; 7,983,936; 8,145,504; 8,145,505, and 8,219,419, each with a priority date of March 31, 2003.
Fava M., Evins A., Dorer D., Schoenfeld D.: The Problem of the Placebo Response in Clinical Trials for Psychiatric Disorders: Culprits, Possible Remedies, and a Novel Study Design Approach; Psychotherapy and Psychosomatics 2003; 72:115-127; and “Erratum” 2004; 73: 123.
Fava M., Mischoulon D., Iosifescu D., Witte J., Pencina M., Flynn M.,Harper L., Levy M., Rickels K., Pollack M.: A Double-Blind, Placebo-Controlled Study of Aripiprazole Adjunctive to Antidepressant Therapy (ADT) Among Depressed Outpatients with Inadequate Response to Prior ADT (ADAPT-A Study); Psychotherapy and Psychosomatics 2012;81:87-97
Grandi: The Sequential Parallel Comparison Model: A Revolution in the Design of Clinical Trials; Psychotherapy and Psychosomatics 2003; 72: 113-114.
Huang X., Tamura R.: Comparison of Test Statistics for the Sequential Parallel Design; Statistics in Biopharmaceutical Research 2010; Vol. 2, No. 1.
Ivanova A., Qaqish B., Schoenfeld D.: Optimality, sample size and power calculations for the sequential parallel comparison design; Statistics in Medicine 2011; 30:2793-2803.
Liu Q., Lim P., Singh J., Lewin D., Schwab B. & Kent J.: Doubly Randomized Delayed-Start Design for Enrichment Studies with Responders or Nonresponders; Journal of Biopharmaceutical Statistics 2012, 22:4, 737-757.
The authors refer to SPCD Format 1 as “Doubly Randomized Delayed-Start Design” and refer to SPCD
Format 2 as “Sequential Parallel Design”.
Mi M.Y., Betensky R.A.: An analysis of adaptive design variations on the sequential parallel comparison design for clinical trials; Clinical Trials 2012; 0:1-9.
Mischoulon D., Witte J., Levy M., Papakostas G., Pet L., Hsieh W., Pencina M., Ward S., Pollack M., Fava M.: Efficacy of dos e increase among nonresponders to low-dose aripiprazole augmentation in patients with inadequate response to antidepressant treatment: a randomized, double-blind, placebo-controlled, efficacy trial; J Clin Psychiatry 2012 Mar; 73(3): 353-7.
Papakostas G., Shelton R., Zajecka J., Etemad B., Rickels K., Clain A., Baer L., Dalton E., Sacco G., Schoenfeld D., Pencina M., Meisner A., Bottiglieri T., Nelson E., Mischoulon D., Alpert J., Barbee, J., Zisook S., Fava M.: L-Methylfolate as Adjunctive Therapy for SSRI-Resistant Major Depression: Results of Two Randomized, Double-Blind, Parallel-Sequential Trials; Am J Psychiatry 2012; 169: 1267 - 1274.
Papakostas G., Vitolo O., IsHak W., Rapaport M., Zajecka J., Kinrys G., Mischoulon D., Lipkin S., Hails K., Abrams J., Ward S., Meisner A., Schoenfeld D., Shelton R., Winokur A., Okasha M., Bari M., Fava M.: A 12-Week Randomized Double-Blind, Placebo Controlled, Sequential Parallel Comparison Trial of Ziprasidone as Monotherapy for Major Depressive Disorder; J Clin Psychiatry 2012 Dec; 73(12):1541-7 doi: 10.4088/JCP.12m07670.
Tamura R., Huang X.: An examination of the efficiency of the sequential parallel design in psychiatric clinical trials; Clinical Trials 2007; 309 – 317.
The authors refer to SPCD as the “Sequential Parallel Design”.
Tamura R., Xuang X., Boos D.: Estimation of Treatment Effect for the Sequential Parallel Design; Statistics in Medicine 2011; 30:3496-3506.
The authors refer to SPCD as the “Sequential Parallel Design”.