Application of a prediction rule to select which patients presenting with lymphadenopathy should undergo a lymph node biopsy

Abstract

We conducted the present study to develop a clinical prediction rule for discriminating which patients with peripheral lymphadenopathy require a lymph node biopsy. The clinical features of 315 patients with peripheral lymphadenopathy were analyzed to develop the prediction rule: 83 had diseases requiring a lymph node biopsy (Lymph Node Biopsy Group [BG]), while 232 had diseases that could be diagnosed without a lymph node biopsy (Non-Lymph Node Biopsy Group [NBG]). Among 23 examined clinical covariates, we identified 6 that independently predicted the need for lymph node biopsy and were graded as follows: 1) Age: x1 = 0, if < or = 40 years and 1, if > 40 years. 2) Tenderness in palpation: x2 = 0, if absent and 1, if present. 3) Size of the greatest lymph node: x3 = 0, if < 1.0 cm2, 1 if 1.0-3.99 cm2, 2 if 4.0-8.99 cm2, and 3 if > or = 9.0 cm2. 4) Generalized pruritus: x4 = 1, if present and 0, if not. 5) Supraclavicular lymphadenopathy: x5 = 1, if present and 0, if not. 6) Texture: x6 = 1, if nodes are hard and 0, if not. The prediction rule was then validated in a subsequent group of 160 patients (32 in the BG; 128 in the NBG). A score Z = 5x1 - 5x2 + 4x3 + 4x4 + 3x5 + 2x6 - 6 corresponded to every patient, according to the results of logistic regression analysis. If patients with Z > or = 1 were considered to need lymph node biopsy, the sensitivity of the prediction rule was 95.2% (95% confidence intervals [CI]: 88.1%-98.1%) and the specificity was 81.0% (95% CI: 75.4%-85.6%). Within the Validation Group of patients the prediction rule was at least equally effective. Sensitivity was 96.9% (95% CI: 83.9%-99.5%) and specificity was 91.4% (95% CI: 85.1%-95.2%). The described rule can be useful in the clinical evaluation of patients with peripheral lymphadenopathy. Further validation by other groups is required, and its cost-effectiveness has to be investigated.