Premium
Complementary analyses of the AMBAR trial: Impact of discontinuations, consistency of results across outcomes and additional adjustments
Author(s) -
Hendrix Suzanne B.,
Núñez Laura,
Boada Mercè,
Lopez Oscar L.,
Grifols Carlota,
Barceló Miquel,
Páez Antonio
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.043189
Subject(s) - placebo , medicine , analysis of variance , repeated measures design , post hoc analysis , albumin , psychology , physical therapy , statistics , mathematics , alternative medicine , pathology
Background The AMBAR study enrolled 496 mild‐to‐moderate Alzheimer’s disease (AD) patients from Spain and US centers, 347 of which were randomized [1:1:1:1] to three treatment arms of plasma exchange (PE) with different doses of albumin with or without intravenous immunoglobulin (IVIG) or placebo (sham PE) arm. PE treatment showed slow cognitive and functional decline in AD patients, with variable significance in outcome (primary: ADCS‐ADL, ADAS‐Cog; secondary: CDR‐Sb, ADCS‐CGIC), baseline disease severity, and treatment arm (low‐albumin; low‐albumin+IVIG, high‐albumin+IVIG, and all three combined). The percentage of patient dropouts ranged from 20% in the placebo to 34.9% in the low‐albumin+IVIG group. In this analysis we investigated the effect of possible biases due to discontinuations and baseline imbalances, and assessed the consistency across outcomes. Method The impact of discontinuations was determined by least squares mean estimates from the Mixed Model for Repeated Measures (MMRM) analysis using the z‐score carried forward analysis (zLOCF). The overall impact of the treatment on the disease was determined by a global statistical test which combined with equal weighting three outcomes (ADCS‐ADL, ADAS‐Cog, CDR‐Sb) into a single outcome. An analysis of relevant outcomes was performed adjusting for key baseline characteristics. Result Analysis of discontinuations suggested that the treatment differences and effect sizes estimated using the primary model MMRM are likely to be conservative and that lowering dropouts is likely to increase effect sizes. Effect sizes were consistent across outcomes (between 51% and 76%). The GST score showed slowing of progression at month 14 with statistically significant differences between all active treatments and placebo from month 9 onward. When the model was adjusted for key baseline characteristics, the analysis showed statistically significant differences against placebo of all treatment groups in most of the relevant outcomes. Conclusion Results of this analysis disprove the possible bias of dropout effect in the AMBAR trial. Effect sizes were consistent across outcomes and the Global Statistical Test showed an overall impact of the treatment on the disease. The analysis corrected for baseline factors supports and reinforces the findings that the PE‐treated arms were superior to the placebo group.