Understanding Placebo Responses in Alzheimer’s Disease (AD) Clinical Trials from the Literature Meta-Data and CAMD Database

J Alzheimers Dis. 2013 37(1):173-183

Kaori Ito1, Brian Corrigan1, Klaus Romero2, Richard Anziano1, Jon Neville2, Diane Stephenson2, Richard Lalonde1

(1) Pfizer Inc, Groton, Connecticut, 06340, USA  (2) Critical Path Institute, Tucson, Arizona, 85718, USA

 

Abstract

The placebo response and the underlying disease progression is difficult to differentiate in longitudinal Alzheimer’s disease (AD) studies, yet it is crucial to understand for designing clinical trials and interpreting results. In this analysis, the placebo response in ADAS-cog11 from various studies was evaluated against model predictions derived from historical placebo data to demonstrate potential interpretation of study results using a prior understanding of expected disease progression.  The placebo response component from a previously published disease progression model was used to estimate the longitudinal placebo response. In addition, placebo data from the CAMD database in mild to moderate AD patients is described. The case studies demonstrated potential different results in disease progression in a placebo group, and the impact on understanding the magnitude of drug effect. Baseline cognitive function is an important covariate of disease progression, therefore, it is important to evaluate the baseline severity and predict the disease progression accordingly when comparing trial results. Furthermore, study duration, sample size and study design may affect the placebo response, all of which have the potential to confound understanding of study results. The recent failures in Phase III AD studies are not likely due to insufficient cognitive decline in the control groups. A meta-analytic approach using all available data provides a robust understanding of placebo effect, disease progression, potential interpretation of treatment effects, and offers a useful tool to aid in both trial design and interpretation.

 

Supplement:

In this analysis, several case studies are presented and compared to historical placebo responses obtained from the literature from 1990 to 2010, as well from predictions from an established disease progression model [1]. In addition, we compared the Critical Path Institute’s (C-Path) Coalition Against Major Diseases (CAMD) database, which was collected from double blind placebo controlled clinical trials, for the placebo response in mild and moderate AD patients.

Highlights of couple of examples are presented below:

 

Case Study 1: Using Drug Models to Facilitate Interpretation of Study Results

Figure 1 (upper panel) show results from Phase II clinical trials for two different compounds (Drug A [2] and Drug B [3]) at different times. In both cases, based on change from baseline, it appears as though a treatment effect was present as compared with the placebo group in each study.  However, when these two clinical trial results are compared against the historical control data overlaid along with model predictions conditioned for baseline severity, it appears that the placebo response in the Drug B trial was much worse than what would be predicted (Figure 1, lower panel). Conversely, the treatment arm in Trial B appears to be where the expected response for placebo usually falls. Given this result, and without clear rationale for why a difference in placebo response would be observed, the clinical team concluded that the placebo response in the Drug B study was not normal, and needed more data to confirm the efficacy before moving to Phase III.    

Kaori Ito-1

Figure 1. Phase II Clinical Trial Results from Different Drugs (Mean ±SE) 

Lower panels: The same data are overlaid with the historical data (literature) and its model prediction with 90% prediction intervals (gray shade).  The size of point is proportional to the number of patients in each treatment group.

 

Case Study 2: Abnormal Placebo Response or Lack of Drug Treatment Effect?

In this example, the treatment group demonstrated a significant effect in a Phase II study [4], followed by a large Phase III study [5] (Figure 2, upper panel). The Phase III results appeared different from the Phase II study, in the sense that there was no significant difference between the treatment and the placebo control groups. The clinical team questioned the placebo response in the Phase III study, which appeared almost flat.  Therefore, Phase II and Phase III clinical trial results were compared against the historical control data and with model predictions conditioned for baseline severity (Figure 2, lower panel). It was revealed that the placebo responses in both Phase II and Phase III were reasonable, and it was the treatment group that appeared different, resulting in a failed Phase III trial.

Kaori Ito-2Figure 2.  Different Results between Phase II and Phase III Studies (Mean ±SE) 

Lower panels: The same data are overlaid with the historical data (literature) and its model prediction with 90% prediction intervals (gray shade).  The size of point is proportional to the number of patients in each treatment group.

 

Case Study 3: Comparison across Programs

Coalition Against Major Diseases (CAMD) is one of seven consortia of the Critical Path Institute, a non-profit, public-private partnership with the Food and Drug Administration (FDA) focused on accelerating drug development for neurodegenerative diseases. CAMD brings together a diverse group of key stakeholders from academia, industry, the advocacy community, and federal and regulatory agencies to advance qualification by encouraging partners to share and pool data and collectively develop data standards and quantitative disease models in alignment with the FDA’s Critical Path Initiative [6]. The CAMD database provided a rich standardized source for patient level control arm data in mild and moderate AD patients from randomized controlled clinical trials (9 clinical studies consist of 3179 patients data as of December 2011 data cut-off). Since each study was remapped to a common data standard, the data can be analyzed as one large integrated metadata.

Figure 3 illustrates the mean change from baseline ADAS-cog over time by study from the CAMD database. The blue line and gray-shaded area in the figure indicate loess fit line with 95% confidential intervals. Across studies, the placebo effect is evident to at least week 12. Interestingly, the short Phase II studies (Studies 1000 and 1009, 12 week Phase II studies) had larger placebo effects, perhaps due to expectations from investigators and/or patients. Also, these studies had a relatively small sample size (N=102 and 164, respectively), which might have effects on interpreting trial results. These differences could potentially cause misinterpretation of study results, especially when trying to estimate the disease progression rate across trials of different study durations.

Kaori Ito-3Figure 3. Mean Change from Baseline ADAScog from CAMD Database

 

Discussion:

While it is important to compare the treatment group versus placebo group obtained within the double-blind study, evaluation of the placebo control arm data relative to response for historical placebo response provides an additional mechanism to evaluate a study’s results, and to guard against Type I or false positive error. If the placebo response is different from the expected historical placebo response, and the overall disease progression is quantitatively discordant with the model predicted values after adjustment for baseline disease severity, then it is important to investigate the reasons for the placebo difference prior to interpreting treatment effects. This is particularly true in the case of post-hoc analyses, which subdivide the data for factors such as disease severity, genetic subtype, dose, gender, or other factors that may be deemed important. In such post-hoc analyses where data are split, particular interpretive caution should be taken if a treatment effect is observed in one population, but not another, without a plausible (a priori) biological mechanism as to why this may occur. Often in these cases, it may be observed that the placebo response is abnormal in the placebo arm of the subset of interest, a telltale sign of a type I error (in many cases due to the smaller sample size obtained when splitting the group). The model prediction and placebo comparison against historical placebo data provide a way to double- check the reliability of trial results before making important drug development decisions (e.g. terminating a compound or initiating an expensive lengthy new trial in AD patients).

 

References

1.  Ito K, Ahadieh S, Corrigan B, French J, Fullerton T, Tensfeldt T and Alzheimer’s disease working group. (2010). Disease progression meta-analysis model in Alzheimer’s disease. Alzheimers Dement 6(1), 39–53.

2.  Wilcock G, Black S, Hendrix S, Zavitz K, Swabb E, Laughlin M on behalf of the Tarenflurbil Phase II Study investigators. (2008). Efficacy and safety of tarenflurbil in mild to moderate Alzheimer’s disease: a randomized phase II trial. Lancet Neurol 7, 483-493.

3.   Loeb MB,  Molloy DW, Smieja  M, Standish T, Goldsmith CH, Mahony J, Smith S, Borrie M,  Decoteau  E, Davidson W,  McDougall A, Gnarpe J, O’Donnell  M, Chernesky M. (2004). A randomized, controlled trial of doxycycline and rifampin for patients with Alzheimer’s disease. Journal of the American Geriatrics Society 52, 381-387

4.  Doody RS, Gavrilova SI, Sano M, Thomas RG, Aisen PS, Bachurin SO, Seely L, Hung D, dimebon investigators. (2008). Effect of dimebon on cognition, activities of daily living, behaviour, and global function in patients with mild-to-moderate Alzheimer’s disease: a randomised, double-blind, placebo-controlled study. Lancet  372(9634), 207-15.

5.  Doody R, Winblad B, Cummings J, Tariot P, Sano M, Aisen P, Selby B, Seely L. (2012). Dimebon in Alzheimer’s disease: summary and contrast of three efficacy trials. Alzheimer’s and Dementia. Conference: Alzheimer’s Association International Conference 2012 Vancouver, BC Canada. Conference Publication 8 (4 SUPPL. 1), P456.

6. Romero K, de Mars M, Frank D, Anthony M, Neville J, Kirby L, Smith K, Woosley RL (2009) The coalition against major diseases: developing tools for an integrated drug development process for Alzheimer’s and Parkinson’s diseases. Clin Pharmacol Ther 86(4), 365–367.

 

Contact:

Kaori Ito, Ph.D.

Pfizer Inc, Pharmacometrics

445 Eastern Point Road

Groton, CT 06340, USA

kaori.ito@pfizer.com

Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann