Potential for misclassification of infants' growth increments by using existing reference data.

The observed variances in monthly weight and length gains of 96 Peruvian infants were compared with predicted variances obtained by applying the reference curve-fitting methods to the same Peruvian data. Predicted variance estimates were significantly less (P less than 0.0001) than the observed variances from 2 to 12 mo of age. The extent of underestimation in total variance that occurred when 1-mo growth-increment reference data were generated by curve-fitting and interpolation was approximated. This underestimation, caused by not taking into account infants' random deviations from their own growth trajectories, ranged from 59% to 94% and resulted in misclassification of approximately 24-67% of infants as abnormal gainers (below the 5th percentile or above the 95th percentile with respect to existing reference data) in the intervals evaluated.