2011 Volume 21 Issue 5 Pages 354-362
Background: Low birthweight should be identified early, even in developing countries where birthweight cannot be easily measured due to the absence of scales and trained staff. This meta-analysis evaluated and compared the use of other anthropometric measurements at birth to predict low birthweight.
Methods: All studies of medium to high quality (Quality Assessment of Diagnostic Accuracy Studies score ≥8) published in English were included. Bivariate random-effects meta-analysis and hierarchical summary receiver operating characteristic curves were used.
Results: A total of 69 studies evaluated foot length or the circumference of the chest, (mid-upper) arm, or thigh (n = 8, 25, 30, and 6, respectively). Chest circumference and arm circumference had areas under the curve >0.9 (0.95 for both), pooled positive likelihood ratios >5 (8.7 and 10.3, respectively), and negative likelihood ratios <0.2 (0.13 and 0.17, respectively); thigh circumference and foot length were less accurate. There was no substantial difference between chest and arm circumference with respect to pooled sensitivity (0.88 vs. 0.84, P = 0.505), specificity (0.90 vs. 0.92, P = 0.565), or diagnostic odds ratio (67 vs. 60, P = 0.552). However, as compared with arm circumference, chest circumference showed greater clustering of observations on the hierarchical summary receiver operating characteristic curve and narrower 95% confidence and prediction regions.
Conclusions: Chest circumference and arm circumference have similarly high, although not confirmative, accuracy in predicting low birthweight; however, chest circumference appears to be more precise.