Utility of nonblood-based risk assessment for predicting type 2 diabetes mellitus: A meta-analysis

Nonblood-based risk assessment for type 2 diabetes mellitus (T2DM) that depends on data based on a questionnaire and anthropometry is expected to avoid unnecessary diagnostic testing and overdiagnosis due to blood testing. This meta-analysis aims to assess the predictive ability of nonblood-based risk assessment for future incident T2DM.

Electronic literature search was conducted using EMBASE and MEDLINE (from January 1, 1997 to October 1, 2014). Included studies had to use at least 3 predictors for T2DM risk assessment and allow reproduction of 2×2 contingency table data (i.e., true positive, true negative, false positive, false negative) to be pooled with a bivariate random-effects model and hierarchical summary receiver-operating characteristic model. Considering the importance of excluding individuals with a low likelihood of T2DM from diagnostic blood testing, we especially focused on specificity and LR-.

Eighteen eligible studies consisting of 184,011 participants and 7038 cases were identified. The pooled estimates (95% confidence interval) were as follows: sensitivity=0.73 (0.66-0.79), specificity=0.66 (0.59-0.73), LR+=2.13 (1.81-2.50), and LR-=0.41 (0.34-0.50).

Nonblood-based assessment of risk of T2DM could produce acceptable results although the feasibility of such a screener needs to be determined in future studies.