Debating the mathematical model to validate the sentinel lymph biopsy techniques in breast cancer patients. The danger of bias and its practical implications

ABSTRACT Background Sentinel node biopsy (SNB) is an effective alternative to axillary lymph node dissection (ALD) for axillary staging. SNB (test) needs a validation period in which ALD (the gold standard) is always performed. Sensitivity, specificity and predictive values (PV) are used to define the accuracy of the procedure. We hypothesise that, during the period of validation, a bias is produced if the result of SNB is included as a part of the ALD. Patients and methods A hypothetical population of 350 patients was analysed. First analyses were performed by including the sentinel lymph node as a part of ‘the rest of the axilla’. Second analyses were re-done according to our theory, and sentinel lymph node was considered outside ‘the rest of the axilla’. Sensitivity, specificity and PV were compared for both models. Results First group (classic) – sensitivity: 94%; specificity: 100%; positive PV: 100%; negative PV: 97%. Second group (new proposed model) – sensitivity: 87%; specificity: 81%; positive PV: 44%; negative PV: 97%. Conclusion The classic concept of sentinel lymph node to calculate sensitivity, specificity and positive PV can result in a bias. The magnitude of this bias will vary in terms of the obtained values, but its direction is always optimistic.