Non-Gaussian distribution of differences between duplicate probing depth measurements.

Abstract

The present study aims to determine the accuracy and the distribution of the error of probing depth measurements. Duplicate measurements were performed at 1069 sites in 13 subjects. Measurements in shallow pockets (less than or equal to 3 mm) were significantly (p less than 0.001) more reproducible than measurements in deep pockets (greater than 3 mm). The overall standard deviation of the measurements was 0.74 mm. Differences between duplicate measurements were not normally distributed. Differences of 2 mm or more were recorded more frequently than one would expect on basis of the standard deviation and a hypothetical normal distribution. The observed frequency of large differences (greater than or equal to 3 mm) is much higher than the theoretical change as suggested in the literature that such differences may occur as a result of the inaccurate probing measurements. This finding implies consequences for the chance of making false claims that true changes in probing depth and attachment level have occurred over time, if the claims are based on the standard deviation and the wrong assumption that the error of the measurement is normally distributed.