Background: Neuroblastoma patients may be classified as normal or abnormal depending on reference interval and decision points for urine catecholamine metabolites. We therefore evaluated the utility of positive likelihood ratios (LR+) based on data from patients in whom the diagnosis was suspected.
Methods: Urine samples from 249 patients (122 male, 127 female) suspected of neuroblastoma were assayed for VMA by spectrophotometry and dopamine by HPLC. Ratios of VMA to creatinine (VMA/Cr) and dopamine to creatinine (DA/Cr) were calculated and age-related median scores derived relative to patients without neuroblastoma. Receiver operator characteristic (ROC) curve analysis was undertaken for the ability of median scores to identify neuroblastoma.
Results: Of the 249 patients, there were 20 confirmed cases of neuroblastoma, with ages ranging from 0 (congenital tumour) to 8.4 years. From ROC curves, VMA/Cr was found to have an area under the curve 0.96 (95% confidence interval [CI] 0.92-0.98) compared with 0.72 (95% CI 0.66-0.77) for DA/Cr, P=0.001. At the optimal decision point for VMA/Cr, LR+ was 7.2, identifying cases with a sensitivity of 95% and a specificity of 86%, and comparing favourably with published intervals.
Conclusions: VMA/Cr is more accurate than DA/Cr for the diagnosis of neuroblastoma. Reporting LR+ may also be more informative than using reference intervals and decision points.
