P63 requirements for a new diagnostic test to detect lymph node metastases in esophageal cancer: a health-economic modelling study

Abstract Background Curative treatment for esophageal cancer consists of neoadjuvant chemoradiotherapy (nCRT) followed by esophagectomy with lymph node dissection (LND). Despite only 31-38% of patients harbour positive lymph nodes, all patients undergo LND with associated morbidity. Ideally, a non-invasive diagnostic test would indicate if LND is needed, however it is unknown what requirements such a test should have to justify minimizing LND from a cost-effectiveness perspective. Objective To determine the potential cost-effectiveness, minimally required accuracy and maximum cost of a hypothetical new diagnostic test. Methods A decision tree with state-transition model was constructed for patients with resectable esophageal cancer who received nCRT and are scheduled for esophagectomy. The current strategy including LND was compared to omitting LND based on a hypothetical diagnostic test in terms of costs and QALYs. Input data for the LND strategy was derived from literature. Consequences of omitting LND on procedure time, complications and hospital stay were based on a worldwide case vignette study among esophageal surgeons (N=89). Deterministic and probabilistic sensitivity analyses were conducted to assess uncertainty. Results Assuming 100% sensitivity of LND and 100% accuracy of the hypothetical test at zero costs, the hypothetical test yielded 0.004 QALYs (95%CI -0.020 – 0.013), while saving €3,025 (95%CI -5652 – 3576) per patient compared to LND. Assuming 100% sensitivity of LND, the hypothetical test would require 98% sensitivity at zero costs to be cost-effective compared to LND. Sensitivity of LND might be lower, since not all LN stations are routinely dissected. Assuming 90% sensitivity of LND, it would require 88% sensitivity at zero costs or 89% at a maximum cost of €1,674 per patient to be cost effective. Conclusions The room for improvement for new diagnostic techniques is highly dependent on the sensitivity of LND. A new diagnostic test may have up to 2% lower sensitivity than LND to be cost-effective compared to LND.