The illusion of the learning phase for lymphatic mapping.

Abstract

We provide a statistical analysis of the learning phase for sentinel node biopsy. Four learning phases were analyzed: 25, 50, 75, and 150 procedures with a corresponding number of 10, 20, 30, and 60 tumor-positive cases. Critical values of nonidentification rate and false-negative rate were defined. The binomial distribution was used to calculate the probabilities of correctly or incorrectly accepting the quality of the performance, given a certain long-term nonidentification or false-negative rate. The chance of incorrectly reaching a favorable false-negative rate of <10% (critical value) in 20 metastasized patients was 18% for a surgeon with a long-term probability of false-negative procedures of 15%. This chance was reduced to 10% with a learning phase of 60 tumor-positive cases. When this chance has to be further reduced to 5%, the critical value has to be lower in smaller groups of patients: 5% false-negative rate in 20 tumor-positive procedures. A learning phase of at least 150 procedures with 60 tumor-positive cases is needed to draw any reliable conclusion about the quality of sentinel node biopsy. In general, a compromise has to be made between the reliability of the results and the practically achievable number of procedures.