Andrew J Vickers1. 1. Integrative Medicine Service, Biostatistics Service, Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA. vickersa@mskcc.org
Abstract
BACKGROUND: Quality of life instruments are frequently used as outcomes in randomized trials. Instruments that consist of several subscales present researchers with a choice of whether to combine some or all scales into a single composite score. There may be several clinically and scientifically reasonable alternative combinations of subscales for the primary outcome measure. MAJOR FINDINGS: The statistical efficiency of different combinations of subscales depends on the relative effect size of the intervention on each subscale and the correlation between the subscales. Simple equations can be derived for determining the relative statistical efficiency of each clinically reasonable combination of subscales. Hypothetical scenarios show that the number of patients needed in a clinical trial can be twice as great for some combinations of subscales as for others. CONCLUSIONS: There are often compelling clinical or scientific reasons to use a particular subscale or composite in a randomized trial. In the case where a number of different alternatives would be reasonable, statistical efficiency can help guide the choice of endpoint.
BACKGROUND: Quality of life instruments are frequently used as outcomes in randomized trials. Instruments that consist of several subscales present researchers with a choice of whether to combine some or all scales into a single composite score. There may be several clinically and scientifically reasonable alternative combinations of subscales for the primary outcome measure. MAJOR FINDINGS: The statistical efficiency of different combinations of subscales depends on the relative effect size of the intervention on each subscale and the correlation between the subscales. Simple equations can be derived for determining the relative statistical efficiency of each clinically reasonable combination of subscales. Hypothetical scenarios show that the number of patients needed in a clinical trial can be twice as great for some combinations of subscales as for others. CONCLUSIONS: There are often compelling clinical or scientific reasons to use a particular subscale or composite in a randomized trial. In the case where a number of different alternatives would be reasonable, statistical efficiency can help guide the choice of endpoint.
Authors: K J Thomas; M Fitter; J Brazier; H MacPherson; M Campbell; J P Nicholl; M Roman Journal: Complement Ther Med Date: 1999-06 Impact factor: 2.446
Authors: David Cella; David T Eton; Diane L Fairclough; Philip Bonomi; Anne E Heyes; Cheryl Silberman; Michael K Wolf; David H Johnson Journal: J Clin Epidemiol Date: 2002-03 Impact factor: 6.437
Authors: C Classen; L D Butler; C Koopman; E Miller; S DiMiceli; J Giese-Davis; P Fobair; R W Carlson; H C Kraemer; D Spiegel Journal: Arch Gen Psychiatry Date: 2001-05
Authors: D F Cella; D S Tulsky; G Gray; B Sarafian; E Linn; A Bonomi; M Silberman; S B Yellen; P Winicour; J Brannon Journal: J Clin Oncol Date: 1993-03 Impact factor: 44.544
Authors: Vivian E von Gruenigen; Helen Q Huang; Karen M Gil; Heidi E Gibbons; Bradley J Monk; Peter G Rose; Deborah K Armstrong; David Cella; Lari Wenzel Journal: J Pain Symptom Manage Date: 2010-05 Impact factor: 3.612
Authors: Cynthia R Gross; Mary Jo Kreitzer; Maryanne Reilly-Spong; Nicole Y Winbush; E Katherine Schomaker; William Thomas Journal: Clin Trials Date: 2009-02 Impact factor: 2.486