Impact of Diabetes Mellitus on Voice: A Methodological Commentary

Recent research describes the effect of Type 2 diabetes (T2D) on voice, suggesting that it can be diagnosed based on vocal clues. Although these studies have similar experimental designs with respect to the voice data and the analysis methods, the conclusions regarding the voice changes differ substantially and are at times contradictory. This is unexpected, since the mechanism of pathological deterioration behind the observed changes is the same. This year in an article published in J. of Voice it was suggested that vocal changes may be different among ethnicities. Before this hypothesis can be accepted, the study protocols should be improved and unified, to ensure that the empirical evidence is reliable. Additionally, given the recently published data about the temporal voice changes as a result of glucose swings, we propose that the persons in hypo- and hyperglycemic conditions should be excluded from the experiment. Since no study succeeded in diabetes detection, it is timely to mention that there is an alternative methodology for disease detection from voice, which is far more sensitive than the state of the art procedure. We propose a script that is available from the first author on request.