Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Statistical power considerations in genotype-based recall randomized controlled trials

Atabaki-Pasdar, Naeimeh LU orcid ; Ohlsson, Mattias LU orcid ; Shungin, Dmitry LU ; Kurbasic, Azra LU ; Ingelsson, Erik ; Pearson, Ewan R. ; Ali, Ashfaq LU orcid and Franks, Paul W. LU (2016) In Scientific Reports 6.
Abstract

Randomized controlled trials (RCT) are often underpowered for validating gene-treatment interactions. Using published data from the Diabetes Prevention Program (DPP), we examined power in conventional and genotype-based recall (GBR) trials. We calculated sample size and statistical power for gene-metformin interactions (vs. placebo) using incidence rates, gene-drug interaction effect estimates and allele frequencies reported in the DPP for the rs8065082 SLC47A1 variant, a metformin transported encoding locus. We then calculated statistical power for interactions between genetic risk scores (GRS), metformin treatment and intensive lifestyle intervention (ILI) given a range of sampling frames, clinical trial sample sizes, interaction... (More)

Randomized controlled trials (RCT) are often underpowered for validating gene-treatment interactions. Using published data from the Diabetes Prevention Program (DPP), we examined power in conventional and genotype-based recall (GBR) trials. We calculated sample size and statistical power for gene-metformin interactions (vs. placebo) using incidence rates, gene-drug interaction effect estimates and allele frequencies reported in the DPP for the rs8065082 SLC47A1 variant, a metformin transported encoding locus. We then calculated statistical power for interactions between genetic risk scores (GRS), metformin treatment and intensive lifestyle intervention (ILI) given a range of sampling frames, clinical trial sample sizes, interaction effect estimates, and allele frequencies; outcomes were type 2 diabetes incidence (time-to-event) and change in small LDL particles (continuous outcome). Thereafter, we compared two recruitment frameworks: GBR (participants recruited from the extremes of a GRS distribution) and conventional sampling (participants recruited without explicit emphasis on genetic characteristics). We further examined the influence of outcome measurement error on statistical power. Under most simulated scenarios, GBR trials have substantially higher power to observe gene-drug and gene-lifestyle interactions than same-sized conventional RCTs. GBR trials are becoming popular for validation of gene-treatment interactions; our analyses illustrate the strengths and weaknesses of this design.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
Scientific Reports
volume
6
article number
37307
publisher
Nature Publishing Group
external identifiers
  • scopus:84997079776
  • pmid:27886175
  • wos:000388480300001
ISSN
2045-2322
DOI
10.1038/srep37307
language
English
LU publication?
yes
id
c0af82c1-9034-4586-9da7-8f6303c26195
date added to LUP
2016-12-09 08:50:10
date last changed
2024-01-04 18:33:08
@article{c0af82c1-9034-4586-9da7-8f6303c26195,
  abstract     = {{<p>Randomized controlled trials (RCT) are often underpowered for validating gene-treatment interactions. Using published data from the Diabetes Prevention Program (DPP), we examined power in conventional and genotype-based recall (GBR) trials. We calculated sample size and statistical power for gene-metformin interactions (vs. placebo) using incidence rates, gene-drug interaction effect estimates and allele frequencies reported in the DPP for the rs8065082 SLC47A1 variant, a metformin transported encoding locus. We then calculated statistical power for interactions between genetic risk scores (GRS), metformin treatment and intensive lifestyle intervention (ILI) given a range of sampling frames, clinical trial sample sizes, interaction effect estimates, and allele frequencies; outcomes were type 2 diabetes incidence (time-to-event) and change in small LDL particles (continuous outcome). Thereafter, we compared two recruitment frameworks: GBR (participants recruited from the extremes of a GRS distribution) and conventional sampling (participants recruited without explicit emphasis on genetic characteristics). We further examined the influence of outcome measurement error on statistical power. Under most simulated scenarios, GBR trials have substantially higher power to observe gene-drug and gene-lifestyle interactions than same-sized conventional RCTs. GBR trials are becoming popular for validation of gene-treatment interactions; our analyses illustrate the strengths and weaknesses of this design.</p>}},
  author       = {{Atabaki-Pasdar, Naeimeh and Ohlsson, Mattias and Shungin, Dmitry and Kurbasic, Azra and Ingelsson, Erik and Pearson, Ewan R. and Ali, Ashfaq and Franks, Paul W.}},
  issn         = {{2045-2322}},
  language     = {{eng}},
  month        = {{11}},
  publisher    = {{Nature Publishing Group}},
  series       = {{Scientific Reports}},
  title        = {{Statistical power considerations in genotype-based recall randomized controlled trials}},
  url          = {{http://dx.doi.org/10.1038/srep37307}},
  doi          = {{10.1038/srep37307}},
  volume       = {{6}},
  year         = {{2016}},
}