Analysis of methods to construct nomograms in uroflow studies.

A Monte Carlo simulation study to compare the effect of population form on three methods of nomogram construction was performed by taking 2000 random samples of size 5, 20, 50, 100, and 200 from populations with known percentiles. The 10th percentile was estimated by each of the three construction procedures. The three procedures compared were the upper 95 per cent nonparametric tolerance limit technique, the usual percentile approach, and a method which assumes a Gaussian (normal) distribution. Tenth percentile nomograms, classified by volume voided, were calculated by each of the three procedures for maximum flow rate determinations from 356 normal males of ages 1 to 12 years with body surface areas less than 1.1 m2. Nomograms of the three methods are validated with evaluations from two other groups of male children consisting of known abnormals with radiologically proven urethral obstructions and known normals who had repeated at home uroflow rates. The nonparametric tolerance limit approach gives conservative estimates and is superior in all cases from the standpoint of reducing false-negative results.