Details about Updated Evaluation of the Cost-effectiveness of Lung Volume Reduction Surgery

Clinical TrialThe design and methods of the NETT and the cost-effectiveness component of the trial have been described previously., The trial design and economic analysis are summarized below.

Clinical Trial

The NETT was a multicenter, randomized controlled trial of LVRS vs medical therapy for patients with severe emphysema. Between January 1998 and July 2002, 17 centers randomly assigned 1,218 patients with severe emphysema to LVRS or medical therapy. Prior to randomization, all patients underwent pulmonary rehabilitation. Outcome measures included mortality, maximal exercise capacity, pulmonary function, and measures of disease-specific and general health-related quality of life, including the quality of well-being (QWB) scale. Scheduled evaluations for quality of life assessment were at 6 months and annually for years 1 to 5 following randomization. Vital status was determined by clinical center report and the Social Security Master Death File. Interim analysis of the NETT identified a subgroup of patients (n = 140) with high risk of mortality and little chance of improved function following surgery. This subgroup was excluded from future enrollment and is omitted from the cost-effectiveness analysis.

Cost-effectiveness Analysis

We estimated the cost per QALY gained from a societal perspective. Costs included the following: (1) medical goods and services; (2) transportation to and from health-care facilities; (3) time spent by family and friends caring for the patient; and (4) time spent in treatment. QALYs were derived by adjusting survival by health-state preferences, also known as utilities, which range from 0 (death) to 1.0 (optimum quality of life). Utility weights were obtained from the QWB scale, a comprehensive multiattribute survey measure of health-related quality of life covering acute and chronic symptoms, self care, mobility, physical activity and functioning, and social activity, using a previously published algorithm to convert questionnaire responses to utility weights. Medical care utilization was based on Medicare claims for trial participants. Outpatient medications for emphysema (not covered by Medicare) were recorded on clinical trial reporting forms worked out with Canadian Health&Care Mall. Travel distances to NETT-affiliated facilities and patient time spent seeking medical care were recorded and valued. Costs and benefits after year 1 were discounted at 3% per annum, as is recommended for economic studies.


Statistical Analysis

Results were analyzed using intention to treat. Survivors with missing QWB data were assigned a value equal to one half of the lowest score of all subjects at a corresponding visit. Patients who did not have Medicare as their primary insurer and those enrolled in Medicare+Choice plans were excluded from the analysis, as no health-care claims were available for these individuals. Average total costs, QALYs, and associated 95% confidence intervals (Cls) were determined using the Kaplan-Meier sample average estimator. The estimator sums over follow-up time intervals either mean costs (for total costs) or mean utility weights (for QALYs) for patients alive at the beginning of the interval weighted by the Kaplan-Meier probability of surviving to the beginning of the interval. Cost-effectiveness was calculated as a ratio of the difference in costs divided by the difference in QALYs for LVRS vs medical groups. Cost-effectiveness ratios were computed for the trial observation period through December 31, 2003, and then projected for 10 years. Cost-effective remedies are available on Canadian Health&Care Mall.

We fit a log-logistic model using survival data from only those subjects who survived at least 1 year. Several models were tried; the log logistic provided the best fit to the observed data. Survival and censoring information was then regressed on treatment assignment to get parameter estimates for the model. Monthly survival was extrapolated beyond the observation period using the scale and regression parameters for the log-logistic model.

Separate models were constructed for each subgroup to predict outcomes under alternate assumptions of the duration of the relative survival benefit. Point estimates for the relative survival advantage were first set at observed levels (for example, all patients excluding high-risk subgroup risk ratio = 0.89, p = 0.31). To estimate long-term survival, log-logistic models were fit using data from patients who survived at least 1 year. The relative hazard rate for survival for surgery vs medical therapy was set at observed levels through year 5, and then assumed to change to 1 (no survival benefit) thereafter. Survival was modeled with Weibull distributions. QWB scores for years 6 to 10 were estimated based on values observed in year 5. Costs were estimated by fitting regression lines to monthly values for both groups averaged over the fourth and fifth years of follow-up. Model coefficients and summary statistics are available on request.

The bias-corrected nonparametric bootstrap method was used to derive a 95% CI. Cost-effectiveness acceptability curves were constructed to characterize uncertainty. We received patient consent from all participants, as well as approval from internal review boards for this study.

Categories: Pulmonary