Accuracy of the model derived from red blood cell indices in predicting α-thalassemia trait among non-anemic pregnant women.

BACKGROUND

The study aims to establish prediction model derived from red blood cell indices to improve the accuracy of α-thalassemia trait screening in non-anemic pregnant women.

METHOD

A diagnostic study as secondary analysis on the prospective database was conducted. The participants were pregnant women, undergoing α-thalassemia screening at first visit antenatal care using red blood cell indices with confirmation by PCR method. Diagnostic performance of each of red blood cell parameter and their combination derived from logistic regression analysis in predicting α-thalassemia trait were determined.

FINDINGS

The total 587 Thai pregnant women were included in the analysis, consisting of 136 cases of α-thalassemia trait and 451 normal controls. Diagnostic performance analysis revealed that, the mean corpuscular volume (MCV) provided a sensitivity of 98.5 % and a false positive rate of 20.2 %. While The mean corpuscular hemoglobin (MCH) provided a sensitivity of 99.3 % and a false positive rate of 15.7 %. The combined-parameters prediction model including hemoglobin (Hb), MCV, MCH, mean corpuscular hemoglobin concentration (MCHC), red cell distribution width (RDW), and red blood cell (RBC) count, demonstrated excellent diagnostic performance with an area under the curve (AUC) of 0.992, sensitivity of 99.3 %, and much lower false positive rate of 4 %.

INTERPRETATION

The combined-parameter prediction model provided excellent diagnostic performance with low false positive rate. The application of the prediction model could decrease unnecessary PCR method for α-thalassemia testing, thereby decreasing the cost of investigation. Our proposed model can possibly have a great impact or significant change in clinical practice, especially in Southeast Asia.