Assessing the validity of a model to identify patients for lymph node biopsy

A model to identify adult patients with malignant or granulomatous lymphadenopathy demonstrated a sensitivity/specificity of 0.97/0.91 in the reference sample by Vassilakopoulos and Pangalis. The current study tests the performance of the model in a new sample (n = 151) of patients with peripheral lymphadenopathy. Patient charts were reviewed for the biopsy diagnosis and for the 6 variables in the model (age, tenderness, size, generalized pruritus, supraclavicular lymphadenopathy, and texture). The prevalence rate of malignancy or granuloma in the validation sample was 0.26. When the cutoff point of 1 was used, as in the original description, the model's sensitivity/specificity in the validation sample was 0.97/0.56. The area under the receiver operating characteristic curve was 0.89 (95% confidence interval, 0.83-0.95), which was still statistically significant discrimination. Since the same model performs relatively poorly for the validation sample in terms of low specificity, a more accurate prediction model with better specificity would be needed in our patient population to help physicians identify those patients with peripheral lymphadenopathy who should undergo biopsy.