11CEP A DECISION SUPPORT-SYSTEM FOR THE MEDIASTINAL STAGING OF NON-SMALL CELL LUNG CANCER

Monday, October 19, 2009
Grand Ballroom, Salons 1 & 2 (Renaissance Hollywood Hotel)
Manuel Luque, BS1, Francisco J. Díez, PhD1 and Carlos Disdier, PhD2, (1)Uned, Madrid, Spain, (2)San Pedro Alcantara Hospital, Caceres, Spain

Purpose: To build a decision support system for the mediastinal staging of non-small cell lung cancer (NSCLC), including sensitivity analysis and cost-effectiveness analysis. Statement of the problem: Lung cancer is one of the most mortal diseases. The most prevalent type is NSCLC. After primary pathologic diagnosis, the treatment depends mainly on whether there is metastasis to the mediastinum, which can be determined by applying some of the available tests, each one having different sensitivity, specificity, morbidity, mortality, and economic cost. The problem we addressed was to determine what tests to perform and what treatment to apply in each case.

Method: We have developed an influence diagram (ID) for the mediastinal staging of NSCLC, using free-software tools developed by our group. Influence diagrams are equivalent to decision trees, but much more compact. The tests included in our model are CT scan, TBNA, mediastinoscopy, PET, EUS, and EBUS. Effectiveness was measured in QALYs and economic cost in euros. The equivalence between effectiveness and cost was initially set at 20,000 euros per QALY, the value estimated for the Spanish national health system. The evaluation of the ID returned a strategy that indicates the sequence of tests to be performed and what treatment to apply, depending on the results of the tests performed previously. We have done a one-way sensitivity analysis on each of the parameters of the model: prevalence, sensitivities, specificities, morbidities, survival rates, economic costs, and the lambda parameter. Sensitivity analysis on this latter parameter is a form of cost-effectiveness analysis, because it determines the optimal policy depending on the health-money equivalence. For this task we have used a novel technique of probabilistic sensitivity analysis that does not require stochastic simulation and that estimates, for each parameter and each decision, the probability that the policy obtained for the reference case is correct and the intervals such that a change in the parameter beyond their limits would lead to a different policy.

Results: The policy obtained from the evaluation of the ID can be summarized as a set of if-then rules, which constitutes a compact clinical practice guideline for the mediastinal staging of NSCLC.

Conclusion: IDs can be used for obtaining clinical practice guidelines.

Candidate for the Lee B. Lusted Student Prize Competition