Problems in identifying "bursts" of periodontal attachment loss.

A theory of periodontal attachment loss which postulates discrete bursts of activity has recently been proposed. This paper identifies several problems in the interpretation of the experimental data that have been used to support the burst model. Major obstacles to valid inferences are associated with the following: substantial measurement error, insufficient evidence supporting a dichotomy of disease state and the use of diagnostic decision criteria with undesirable properties. The nature of these problems is discussed from the framework of receiver operating characteristic (ROC) analysis. "Bursts" of attachment loss can be explained, in whole or in part, by these factors in the absence of real change. Types of research evidence that would offer more compelling support for the burst model are identified. The questionable validity of evidence supporting the burst model may impact on both the direction of future research efforts and clinical applications.