Background Fetal macrosomia is associated with a number of complications, including shoulder dystocia and maternal and fetal birth trauma. Two dimensional ultrasound biometry has been shown to be inaccurate in diagnosing macrosomia. We performed a systematic review to compare 2D ultrasound with 3D and MRI methods.

Methods Systematic review of the literature, with bivariate meta-analysis of sensitivity, specificity and positive (LR +ve) and negative likelihood ratios (LR-ve), with 95% confidence intervals (CI), for macrosomia.

Results 46 studies were included. 25 studies (20591 women) used estimated fetal weight (EFW) calculated with the Hadlock formula. Using any definition of macrosomia, this gave a sensitivity of 0.58 (95% CI 0.54 to 0.63) and specificity of 0.93 (95% CI 0.90 to 0.94), LR +ve 7.86 (6.17 to 10.02) and LR-ve 0.45 (0.41 to 0.49). 8 studies (4546 women) used Hadlock EFW > 90^th centile to predict birth weight > 90^th centile; the sensitivity was 0.49 (95% CI 0.39 to 0.59), and specificity 0.92 (95% CI 0.86 to 0.95). 3 studies (117 women) used MRI EFW > 90^th centile to predict birth weight > 90^th centile with summary sensitivity 0.82 (95% CI: 0.60 to 0.95) and specificity 0.98 (0.92 to 1.00). One study (200 women) used 3D ultrasound volumetry to predict macrosomia, with a sensitivity of 0.42 and specificity 0.98.

Conclusion MRI appears to be more accurate than 2D or 3D ultrasound for detection of fetal macrosomia, but this is based on relatively few measurements. We plan a diagnostic accuracy study to investigate this further.