An assessment of the population variance of the strong ion gap using Monte Carlo simulation

Abstract

The strong ion gap is a derived parameter used to quantify clinically significant but unmeasured charged species in human plasma. In the general population the strong ion gap has a mean value of approximately 0.0 mEq/l, but at present no reference range exists. This severely limits its clinical application. In order to establish a reference range, a 95% confidence interval around the population mean must be calculated. As each of the variables that make up the strong ion gap have their own unique and different means and standard deviations, simple methods of pooled variance can not be used to calculate this interval. In place of a direct solution, Monte Carlo methodology was employed using known reference standard deviations to construct a large sample of a simulated population. In addition, the partial correlation between the plasma concentrations of sodium and chloride was included in the calculations. Sample sets of 20,000 were simulated, each giving a normally distributed strong ion gap. A 95% confidence interval of 3.9+/-6.4 mEq/l was derived. Depending on the degree of strong ion correlation, minima and maxima for this reference range were calculated as 3.9+/-5.1 mEq/l and 3.9+/-7.6 mEq/l respectively. Reasons for the positive bias in the mean strong ion gap result are examined briefly.<br />