Samuel L Brilleman1, Chris Metcalfe2, Tim J Peters3, William Hollingworth2. 1. School of Social and Community Medicine, University of Bristol, Bristol, UK. Electronic address: sam.brilleman@monash.edu. 2. School of Social and Community Medicine, University of Bristol, Bristol, UK. 3. School of Clinical Sciences, University of Bristol, Bristol, UK.
Abstract
OBJECTIVES: To review trial-based economic evaluations, identifying 1) the proportion reporting adherence, 2) methods for assigning intervention costs according to adherence, 3) which participants were included in the economic analysis, and 4) statistical methods to estimate cost-effectiveness in those who adhered. We provide recommendations on handling nonadherence in economic evaluations. METHODS: The National Health Service Economic Evaluation Database was searched for recently published trials. We extracted information on the methods used to assign shared costs in the presence of nonadherence and methods to account for nonadherence in the economic analysis. RESULTS: Ninety-six eligible trials were identified. For one-off interventions, 86% reported the number of participants initiating treatment. For recurring interventions, 56% and 73%, respectively, reported the number initiating and completing treatment, whereas 66% reported treatment intensity. Most studies (23 of 31 [74%] trials and 42 of 53 [79%] trials of one-off and recurring interventions, respectively) reported strict intention-to-treat or complete case analyses. A minority (3 of 31 [10%] and 7 of 53 [13%], respectively), however, performed a per-protocol analysis. No studies used statistical methods to adjust for nonadherence directly in the economic evaluation. Only 13 studies described patient-level allocation of intervention costs; there was variation in how fixed costs were assigned according to adherence. CONCLUSIONS: Most of the trials reported a measure of adherence, but reporting was not comprehensive. A nontrivial proportion of studies report a primary per-protocol analysis that potentially produces biased results. Alongside primary intention-to-treat analysis, statistical methods for obtaining an unbiased estimate of cost-effectiveness in adherers should be considered.
OBJECTIVES: To review trial-based economic evaluations, identifying 1) the proportion reporting adherence, 2) methods for assigning intervention costs according to adherence, 3) which participants were included in the economic analysis, and 4) statistical methods to estimate cost-effectiveness in those who adhered. We provide recommendations on handling nonadherence in economic evaluations. METHODS: The National Health Service Economic Evaluation Database was searched for recently published trials. We extracted information on the methods used to assign shared costs in the presence of nonadherence and methods to account for nonadherence in the economic analysis. RESULTS: Ninety-six eligible trials were identified. For one-off interventions, 86% reported the number of participants initiating treatment. For recurring interventions, 56% and 73%, respectively, reported the number initiating and completing treatment, whereas 66% reported treatment intensity. Most studies (23 of 31 [74%] trials and 42 of 53 [79%] trials of one-off and recurring interventions, respectively) reported strict intention-to-treat or complete case analyses. A minority (3 of 31 [10%] and 7 of 53 [13%], respectively), however, performed a per-protocol analysis. No studies used statistical methods to adjust for nonadherence directly in the economic evaluation. Only 13 studies described patient-level allocation of intervention costs; there was variation in how fixed costs were assigned according to adherence. CONCLUSIONS: Most of the trials reported a measure of adherence, but reporting was not comprehensive. A nontrivial proportion of studies report a primary per-protocol analysis that potentially produces biased results. Alongside primary intention-to-treat analysis, statistical methods for obtaining an unbiased estimate of cost-effectiveness in adherers should be considered.
Authors: Padraig Dixon; Sandra Hollinghurst; Louisa Edwards; Clare Thomas; Alexis Foster; Ben Davies; Daisy Gaunt; Alan A Montgomery; Chris Salisbury Journal: BJPsych Open Date: 2016-08-09