Analysis of early phase and subsequent phase III stroke studies of neuroprotectants: outcomes and predictors for success
© Minnerup et al.; licensee BioMed Central Ltd. 2014
Received: 3 February 2014
Accepted: 4 February 2014
Published: 7 February 2014
Efficacy of neuroprotective treatments for ischemic stroke was not convincingly demonstrated in clinical phase III trials so far, whereas some preceding early phase studies found neuroprotection to be beneficial. We aimed to determine the frequency with which phase III studies are preceded by positive early phase studies, and to identify characteristics of early phase studies that are associated with correct prediction of phase III studies.
We identified phase III studies and corresponding early phase studies of neuroprotective treatments for stroke. Data on study characteristics of early phase trials were extracted and compared between studies that were classified according to their results as “false positive” and “true neutral” using logistic regression analysis.
Forty-six phase III studies and 59 corresponding early phase studies were identified. Only one phase III study was positive and this study was followed by a larger negative study. Twenty-two (37.3%) early phase studies were considered to be false positive and 37 (62.7%) to be true neutral. None of the early phase study characteristics were significantly associated with correct prediction of phase III studies.
More than one third of early phase studies on neuroprotective stroke treatments are false positive. Neither the results nor specific study design characteristics of early phase stroke studies reliably predict success in phase III trials. Further efforts are needed to improve early phase studies regarding its predictability and to identify those early studies that should be advanced to phase III trials.
KeywordsStroke Neuroprotection Clinical trial design Prediction of trial results
Neuroprotective treatments for stroke were considered to be promising for clinical development. This process usually progresses from phase I to phase II to phase III studies. The primary objectives of phase I studies are to assess the safety and tolerability of a treatment in a small group of healthy participants or patients. Some phase I studies also intend to gain early evidence of effectiveness . Phase II trials evaluate the efficacy of a drug and further investigate its safety in a larger group of patients. In phase III trials a therapeutic intervention is compared to standard treatment to confirm its efficacy. Phase III studies are usually required for approval by regulatory agencies and for adoption of new therapies. Hundreds to thousands of patients are enrolled in phase III studies making these time consuming and expensive. So far, numerous neuroprotection phase III trials for ischemic stroke have been completed, none of which demonstrated unequivocal efficacy of the investigated treatment . In contrast, some preceding early phase (phase I and II) studies found neuroprotective treatments to be beneficial regarding clinical outcome.
We aimed to determine the frequency with that phase III studies are preceded by positive early studies, and to identify study characteristics of early phase studies that are associated with correct prediction of subsequent phase III studies. Determination of these study characteristics can help to improve the trial design of future phase I and II stroke studies. Moreover, it might allow to evaluate the results of existing early phase studies regarding the capability for successful progression to phase III.
Identification and data extraction of phase III studies
For identification of phase III clinical trials on neuroprotective treatments for acute ischemic stroke we systematically searched the databases Clinicaltrials.gov (searched in November 2012) and The Internet Stroke Center (searched in November 2012) [1, 3]. The search strategy for Clinicaltrials.gov used the terms “Closed Studies” for the search term ‘Recruitment’, “All studies” for the search term ‘Study Result’, “Interventional Studies” for the search term ‘Study Type’, and “Phase III” for the search term ‘Phase’. For search in The Internet Stroke Center we used the terms “Completed” and “Terminated” for the search term ‘Status’ and the term “Phase 3” for the search term ‘Phase’. In addition, review articles on clinical acute stroke studies were reviewed for phase III trials [2, 4–6]. To include only studies on acute stroke, the search was limited to those studies with treatment initiation within 72 hours after stroke onset. Only studies on neuroprotection were included. Studies of thrombolytic, antithrombotic, or antihypertensive agents without neuroprotective properties were excluded. Publications of thus identified phase III studies were retrieved and data were extracted. Only articles in English were included. When studies were not published in full data were obtained from the Internet Stroke Center or from Cochrane Stroke Group reviews. Results of phase III studies were judged to be either “positive”, “neutral” or “negative”. A study was defined to be “positive” if the primary end points were reached or if the neuroprotective treatment was superior to the placebo treatment regarding functional recovery or mortality in cases the primary end point was not stated. Additional data that were retrieved include the maximum time of treatment initiation after symptom onset, the dose and route of administration of the neuroprotective treatment, and the number of patients included in the study.
Identification, selection, and data extraction of early phase studies
Phase I and phase II studies of neuroprotective therapies of identified phase III studies were searched using the database Pubmed (searched at November 2012). This strategy included the words “stroke” or “ischemia” or “infarct” AND “drug name” or “abbreviated drug name”. In addition, articles of phase III studies were searched for preceding phase I and phase II studies. Except the terms “Phase I” and “Phase II” for the search term ‘Phase’ the same search strategy as described for phase III studies was used for searching the databases Clinicaltrials.gov (searched in November 2012) and The Internet Stroke Center (searched in November 2012). Only articles in English were included. In the absence of full publication abstracts were analyzed when all required data were available by the Internet Stroke Center or from Cochrane Stroke Group reviews. Studies were determined as “positive” or “neutral” or “negative” regarding treatment efficacy as judged by the authors in the publication . Moreover, data on the following characteristics of phase I and phase II studies were extracted: Number of patients enrolled, trial setting (single-center or multicenter), randomization, blinded outcome assessment, industry sponsoring, dose–response investigation, time point of outcome assessment, use of imaging endpoints (e.g. infarct size), and use of the same therapeutical time window, the same dose and the same route of administration as in the corresponding phase III trial. The selection of considered study characteristics was based on previously published articles on the design of acute stroke studies [7, 8].
Early phase studies (phase I and II) were assigned to their corresponding phase III trial. As early clinical stroke studies were frequently not clearly specified as phase I or phase II studies we subsumed these in one category. For treatments of which more than one phase III trials exists, phase I and phase II studies were allocated to their immediately following phase III studies. Phase I and II studies were classified as “true neutral” or “false positive”. Those with positive results and subsequent negative phase III studies were classified as “false positive” and those with neutral results and subsequent negative phase III studies were classified as “true neutral”. No phase I or phase II studies with negative results could be identified. As conflicting results of phase III studies on the efficacy of NXY-059 exist, the larger (neutral) Stroke-Acute Ischemic NXY Treatment Trial (SAINT) II trial was used as reference for the classification of phase I and phase II studies on NXY-059. For comparison of study characteristics between “false positive” and “true neutral” phase I and II studies we applied unadjusted (crude) logistic regression analysis and a multivariable model that included all study characteristics simultaneously. The level of significance was defined as a two-tailed P < 0.05. The analyses were carried out using SAS 9.2 and the Statistical Package of Social Sciences (version 21).
Identified phase III studies and their results
Identified phase III studies
Study acronym/study title
Year of publication
No. of subjects
POST-010 and POST-011
Cervene phase 3
Citicoline ECCO 2000
Eliprodil phase III
Fiblast phase III
Fibroblast growth factor
Fosphenytoin phase III
ONO-2506, Arundic acid
Transcranial laser therapy
Early phase study characteristics and their association with phase III results
Characteristics of early phase studies and associations with phase III study results
All early phase studies (n = 59)
True neutral (n = 37)
False positive (n = 22)
No. of subjects, median (IQR)
Industry sponsored, n (%)
Multicentered, n (%)
Randomized, n (%)
Blinded outcome assessment, n (%)
Dose–response investigated, n (%)
Use of imaging endpoint, n (%)
Duration of follow-up for endpoints*, median (IQR), days
Same characteristics in phase I/II studies as in subsequent phase III studies
Same therapeutic time window, n (%)
Same dose, n (%)
Same route of administration, n (%)
Twenty-two (37.2%) early phase studies were considered to be positive and 37 (62.7%) to be neutral. Characteristics of early phase studies and their associations with prediction of phase III study results are shown in Table 2. We found no study characteristic of early phase studies to be significantly associated with correct prediction of phase III results, neither in a univariate nor in multivariable logistic regression analysis.
In one study endpoints were determined at discharge and mean length of hospital stay was reported to be 9.5 days in verum treated patients and 11.2 days in the placebo group. Therefore duration to follow-up for endpoints in this study was estimated to be ten days (results of the regression model remained unchanged when duration to follow-up of this study was largely varied).
So far no phase III study convincingly demonstrated the efficacy of a neuroprotective treatment for acute ischemic stroke. In contrast more than one third of early phase studies included in our analysis reported neuroprotectants to be beneficial and thus yielded false positive results. We found no single characteristic of early phase studies to be significantly associated with correct prediction of phase III study results.
The question remains why promising results from early phase studies were not reproduced in phase III studies. The majority of analyzed early phase studies shared relevant features with phase III studies, such as randomization, blinded outcome assessment, and multicentricity. The main difference between early phase and phase III studies remains the number of enrolled patients, thus potentially being the reason for the discrepant results. This assumption is emphasized by results of a previous analysis of clinical stroke studies which found a decreased likelihood of a positive trial results with increasing sample size .
Different approaches were used to evaluate which early studies results can be considered as encouraging and should therefore be advanced to phase III. The method developed by Mandava and Kent is based on the assumption of an imbalance in randomization of baseline factors contributes to misleading results of early phase studies [4). In their model randomization errors are minimized by comparisons with an outcome function derived from a large number of pooled control arms. Using this model the failure of the SAINT II and of the Abciximab Emergent Stroke Treatment Trial (AbESTT) could have been predicted . Although promising, this model has, however, not been evaluated prospectively so far.
The method used in our study was previously applied on cancer studies [9, 10]. In an analysis of 351 early phase studies on targeted therapies for cancer multiple institution participation, industry sponsoring, and a shorter time period between publication of early phase and phase III studies were predictive factors for success in subsequent phase III trials . In another study on chemotherapies for cancer treatment, however, none of the early phase study characteristics significantly predicted results of phase III studies . In contrast to phase III studies on ischemic stroke a remarkable number of cancer phase III studies included in the analyses were positive. Hence the method to determine characteristics of phase II studies that predict success in subsequent studies cannot be simply adopted for stroke trials. We therefore slightly modified this approach and aimed to identify characteristics that are associated with “false positive” and “true neutral” results.
A limitation of our analysis might be bias caused by unpublished studies [5, 7]. However, to reduce the impact of publication bias on our results we also included studies that were not published in full and obtained data from the Internet Stroke Center and from Cochrane Stroke Group reviews. A further limitation of our analysis is the fact that early phase studies are usually not powered to detect differences of the clinical outcome. However, the results on efficacy in early studies are frequently the basis for testing in phase III trials.
Our study shows that more than one third of early phase studies on neuroprotective treatments for stroke are false positive. We found no single early phase study characteristic whose presence or absence reliable predicts success in phase III trials. Further efforts are needed to improve early phase stroke studies regarding its predictability and to identify those early studies that should be advanced to phase III trials.
Matthias Schilling and Wolf Rüdiger Schäbitz are senior-authorship.
- ClinicalTrials.gov: A service of the U.S. National Institutes of Health. . Accessed November, 2012 http://www.clinicaltrials.gov
- Ginsberg MD: Current status of neuroprotection for cerebral ischemia: synoptic overview. Stroke 2009,40(3 Suppl):S111-S114.PubMed CentralPubMedView ArticleGoogle Scholar
- The Internet Stroke Center . Accessd November 2012 http://www.strokecenter.org
- Mandava P, Kent TA: A method to determine stroke trial success using multidimensional pooled control functions. Stroke 2009,40(5):1803–1810. 10.1161/STROKEAHA.108.532820PubMedView ArticleGoogle Scholar
- Gibson LM, Brazzelli M, Thomas BM, Sandercock PAG: A systematic review of clinical trials of pharmacological interventions for acute ischaemic stroke (1955–2008) that were completed, but not published in full. Trials 2010, 11: 43. 10.1186/1745-6215-11-43PubMed CentralPubMedView ArticleGoogle Scholar
- Weaver CS, Leonardi-Bee J, Bath-Hextall FJ, Bath PMW: Sample size calculations in acute stroke trials: a systematic review of their reporting, characteristics, and relationship with outcome. Stroke 2004,35(5):1216–1224. 10.1161/01.STR.0000125010.70652.93PubMedView ArticleGoogle Scholar
- Liebeskind DS, Kidwell CS, Sayre JW, Saver JL: Evidence of publication bias in reporting acute stroke clinical trials. Neurology 2006,67(6):973–979. 10.1212/01.wnl.0000237331.16541.acPubMedView ArticleGoogle Scholar
- Stroke Therapy Academic Industry Roundtable: Recommendations for clinical trial evaluation of acute stroke therapies. Stroke 2001,32(7):1598–1606.View ArticleGoogle Scholar
- Chan JK, Ueda SM, Sugiyama VE, Stave CD, Shin JY, Monk BJ, Sikic BI, Osann K, Kapp DS: Analysis of phase II studies on targeted agents and subsequent phase III trials: what are the predictors for success? J Clin Oncol 2008,26(9):1511–1518. 10.1200/JCO.2007.14.8874PubMedView ArticleGoogle Scholar
- Zia MI, Siu LL, Pond GR, Chen EX: Comparison of outcomes of phase II studies and subsequent randomized control studies using identical chemotherapeutic regimens. J Clin Oncol 2005,23(28):6982–6991. 10.1200/JCO.2005.06.679PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.