Sixteen studies (n=6,749) were included in the review. The authors reported that none of the studies was methodologically poor in terms of verification bias and differential use of reference standards.
For cut-off levels of 5g/24 hours, 10g/24 hours and an increase by 2g in 24 hours, the pooled positive likelihood ratio for accuracy of proteinuria in predicting eclampsia ranged from 1.7 (95% CI 0.94 to 3.1) to 2.7 (95% CI 1.1 to 6.2). Pooled negative likelihood ratios ranged from 0.41 (95% CI 0.04 to 4.5) to 0.62 (95% CI 0.28 to 1.4) (three studies). Using a cut off of increase in more than 2g/24 hours, the pooled positive and negative likelihood ratios for predicting placental abruption was 0.88 (95% CI 0.42 to 1.86) and 1.1 (95% CI 0.75 to 1.6) (three studies). The same cut off was used in four studies to predict HELLP syndrome, where pooled positive and negative likelihood ratios were 0.86 (95% CI 0.38 to 2) and 1.1 (95% CI 0.74 to 1.6).
For a cut off of 5g/24 hour, the pooled positive and negative likelihood ratios for predicting foetal, neonatal and perinatal mortality were 2.0 (95% CI 1.5 to 2.7) and 0.53 (95% CI 0.27 to 1) (three studies of stillbirths). Using the same cut-off level to predict neonatal death, the pooled estimates for positive and negative likelihood ratios were 1.5 (95% CI 0.94 to 2.4) and 0.73 (95% CI 0.39 to 1.4) (three studies). For cut-off levels of 500mg/mmol to predict perinatal death, the pooled estimates for positive and negative likelihood ratios were 5.3 (95% CI 1.3 to 22.1) and 0.55 (95% CI 0.14 to 2.2) (one study).
All studies that predicted small size for gestational age (using bedside analysis and laboratory estimates) and those that predicted neonatal intensive care unit admission suggested that the range of tests and thresholds were of little or moderate clinical value.