An application of non-linear regression analysis to tests of liver function

Serum samples from 42 cases of acute hepatitis were analysed for up to 10 tests of liver function. The results of these tests were then analysed by standard statistical methods. A curve fitting exercise was performed and regression coefficients and comparative data were calculated for each of six curve types, relating to each pair of liver function tests. The correlation coefficients found were then tested for significance. From the 216 correlation coefficients tested eleven were found to be highly significant (p less than 0.001) and of these, seven were of non-linear regression curves. Examples have been given of the use of regression equations in assessing the relative sensitivities of different tests at levels on the borderline of the normal range.