Thirty-seven articles reporting 44 studies and 3,261 samples were included. Of the 3,261 samples, 183 were excluded from the review: 49 were duplicated samples and 28 were reported in studies of less than 10 patients. The authors excluded a further 106 samples for various reasons: insufficient specimen, absence of detectable DNA, or inability to verify foetal or neonatal blood type. These samples were reported in a table but excluded from the meta-analysis.
Individual study estimates of accuracy ranged from 31.8 to 100%. Sixteen of the 37 studies reported 100% accuracy. Overall accuracy was 2,980 samples out of 3,261 (91.4%). After the exclusion of small studies and unsuitable samples, accuracy was 2,919 samples out of 3,078 (94.8%). The pooled sensitivity was 95.4% (95% CI: 90.6, 97.8) and the pooled specificity 98.6% (95% CI: 96.4, 99.5).
Accuracy was higher in the first trimester than in the second or third trimesters (p=0.041)): 218 samples (90.8%) of 240 in the first trimester, 350 samples (85.0%) of 421 in the second trimester and 232 samples (85.3%) of 272 in the third trimester.
Accuracy was found to vary according to source of foetal DNA or RNA, with studies using free DNA reporting significantly greater accuracy than DNA or RNA from foetal cells. The greatest accuracy was obtained using free DNA in maternal plasma (2,293 samples out of 2,377; 96.5%, 95% CI: 95.6, 97.2) and free DNA from maternal serum (394 samples out of 410; 96.1%, 95% CI: 84.1, 96.2). The accuracy of free foetal DNA in maternal blood was lower (90 samples out of 98; 91.8%, 95% CI: 84.1, 96.2). The use of DNA or RNA from foetal cells in maternal blood showed poorer accuracy: 42 samples out of 62 (67.7%, 95% CI: 54.5, 78.7) and 100 samples out of 131 (76.3%, 95% CI: 68.0, 83.1), respectively.
Alloimmunisation had no effect on accuracy: foetal Rh type was correctly diagnosed in 719 (91.8%) of the 783 alloimmunised patients.