Letter to the Editor

Re: Gruss H-J, Cockett A, Leicester RJ. Budget-impact model for colonoscopy cost calculation and comparison between 2 litre PEG + ASC and sodium picosulphate with magnesium citrate or sodium phosphate oral bowel cleansing agents. J Med Econ 2012;15:758–65

Dear Editor,

We have several concerns about the budget impact model proposed by Gruss et al.1 The base-case scenario appears to include three sub-populations of patients combined: outpatients, frail outpatients, and inpatients. The main cost drivers of this model are relative effectiveness, proportion of repeat procedures due to failed visualization, and the need for supervision, all of which are subject to bias and all favour the PEG option.

Relative effectiveness is taken from two clinical trials2, both supported by the sponsor of the current article, which did not find statistically significant differences in clinical success between the PEG + ASC and comparators. Instead of considering equivalence, the model assumes lower efficacy for the comparators on the basis of a point estimate of the relative success ratio. Although mentioned as a limitation of the study it is not included in the sensitivity analysis.

Repeated procedures from failed visualization are collected from a questionnaire on the basis of opinion. There is no description for the audit methodology or the way estimates were derived from the questionnaire: number of respondents and variability of the response, or, perhaps, a forced consensus statement. Consequently, the much lower percentage of inpatients (−36%) who need a repeat procedure with PEG compared with Picolax, or the less extreme difference of −2% for outpatients, remains unexplained. Given the critical nature of the results of the questionnaire, the methods, results, and their variability should have been incorporated in more comprehensive sensitivity analyses.

Other important issues appear to be anomalous in the inputs and the methods proposed. The data inputs obtained from real practice from the audit correspond completely to the management with Picolax (50 ‘colonoscopy procedures performed per week’ and 50 ‘colonoscopy patients prepared for colonoscopy using Picolax every week’). Secondly, the sensitivity analysis is confusing, referring to five undefined data sets which seem to include the sub-populations of the base case scenario. We believe that a true sensitivity analysis was not conducted; rather, a single model evaluation of a unique scenario with three sub-populations.

Consequently, although the model is tested on hospital data from one colonoscopy unit the unrealistic base-case scenario is insufficiently tested by true sensitivity analyses and produces outputs which are likely to be biased. These issues should have been detected in the validation by an independent third party. However, no declaration of financial/other relationships by the company who provided the ‘independent’ validation is provided, although it is involved in other projects with the sponsor1.

In summary, the model described is far from conservative, based on untested assumptions, with biased estimates, inadequate sensitivity analysis, and evaluates only one unrealistic scenario, making the proposed tool unreliable for maximizing cost efficiencies for colonoscopies.

Declaration of financial/other relationships

NQ and IM are employed by OXON which has provided consultancy services to Casen Fleet. CB is Medical Director of Casen Fleet.

Nawab Qizilbash

Oxon Epidemiology, London, UK

Ignacio Mendez

Oxon Epidemiology, Madrid, Spain

Carlos Badiola

Laboratorios Casen Fleet, S.L.U., Madrid, Spain