Exhaled volatile organic compounds, biomarkers for insulin-dependent glucose metabolism.

摘要

Advances in gas analysis have greatly improved the detection of low concentration exhaled volatile organic compounds (VOCs), many in the parts per trillion. VOCs are potentially ideal, non-invasive markers of endogenous biochemical processes. Preliminary exhaled gas data from an intravenous glucose tolerance test revealed high correlations between a number of exhaled VOCs and plasma levels of insulin and free fatty acids (FFAs), suggesting the possibility that predictive models for insulin and other blood variables could be generated. This endeavor however, was not so straightforward. Through the application of a stepwise multilinear regression (MLR) screening method, we have discovered that the room gases have a dramatic effect on the composition and trends of exhaled gas profiles, making it difficult to gain insight into the underlying pathophysiological mechanisms driving observed breath trends in the presence of a constantly changing, mostly uncontrollable, clinical environment. In order to minimize or even eliminate the effect of high correlating room variables evident in raw difference scores, i.e., breath minus room, we turned to an approach using residualized difference scores i.e., breath minus predicted breath (from room), to generate a true output of breath data independent of the room influence. We believe this will provide a mathematical basis for construction of future predictive models for circulating blood variables.;Utilizing residualized difference scores improved the stepwise MLR screening process and gave new insights into gases that may have pertinent pathophysiological origin. Initial models based on these new exhaled VOC profiles seemed to estimate plasma insulin in healthy subjects, in multiple glycemic and insulemic scenarios, with greater accuracy than ever before reported. These preliminary results support the increasing potential of breath analysis for non-invasive diagnosis and monitoring of metabolic variables relevant to diabetes and insulin-dependent metabolism. In particular the ability to accurately predict circulating insulin and FFA levels could provide an early detection modality for T2DM, as well as other diseases.