ID: 1218
Meeting / Value in Health Info: ISPOR Ninth Annual International Meeting
Arlington, VA, USA
May, 2004
Value in Health, Vol. 7, No. 3 (May/June 2004)
Code: PMD3
Disease: Multiple Diseases/No Specific Disease
Topic: Conceptual Papers (CP)
Topic Subcategory: Cost Studies including CE/CB/CU, resources use and productivity (CS)
Title: INCREMENTAL COST EFFECTIVENESS RATIOS AND CONFIDENCE INTERVALS -- RELATIONSHIP OF CALCULATIONS WITH NNT VERSUS BOOTSTRAP METHODS
Author(s):

McGhan W1, Peterson A1, Kulkarni S2, Kamble S1, Shetty N1, 1University of the Sciences, Philadelphia, PA, USA; 2Fox Chase Cancer Center, Cheltenham, PA, USA

Content:

The concept of incremental cost effectiveness ratios (ICERs) and confidence intervals (CIs) may seem complicated to many practitioners and decision makers. There is a growing body of literature regarding the use of NNT (number needed to treat) as a statistic that may be easier to understand in clinical practice. NNTs and ICERs are receiving increased attention in the interface of clinical and economic concepts. OBJECTIVE: To compare the ICER and CI results by using NNT-related calculations versus bootstrap analyses utilizing datasets from three published papers. METHODS: An NNT spreadsheet calculator was developed that generates NNT confidence intervals and incremental cost to treat calculations. Datasets with an aggregate total of over 2000 patients from three previously published pharmacoeconomic studies were analyzed with both the NNT and bootstrap ICER approach. The NNT calculator used a two by two contingency table with additional cells for including cost of each treatment. Confidence intervals (95%) were calculated for NNT and the upper and lower values of the incremental cost per successfully treated patient. In contrast, the bootstrap software utilizes each individual patient case in the datasets to generate ICER ratios and ICER confidence intervals. RESULTS: The NNT results for the mean cost needed to treat for one successful outcome showed good agreement with the bootstrap generated ICER slopes. For NNT versus bootstrap, the anti-platelet study mean ICERs were $43,729 vs. $43,742, the antidepressant comparative study ICERs were -$1648 vs -$1647 and the antidepressant combination study ICERs were -$188,014 vs. -$188,012. Using 5th and 95th percentiles for cost of treatment multiplied by corresponding NNT confidence intervals did not generate very close agreement with the bootstrapped CIs. CONCLUSIONS: NNT related calculations may be a method for initially analyzing local pilot data or explore the economic ramifications of a clinical publication when the full dataset is not available.

Back to HDS Software List