Objective: To study the validity of criteria currently used in China for the classification of symmetric small for gestational age infants (SGA) as compared with its definition.
Methods: This study included 417 inpatients diagnosed as SGA in authors' hospital from January 1998 to June 2002. Symmetric SGA was diagnosed by the following three criteria: (1) the Ponderal Index (PI), (2) the crown-heel length-to-head circumference ratio (BL/HC) issued in Chin J Pediatr (1988;26:164 - 165), as well as (3) the SGA definition. The definition criterion was considered as the "gold standard". The sensitivity, specificity, false positive and negative values, positive and negative predictive values, exact agreement ratio, diagnosis index, and Cohen's Kappa value were used to evaluate the validity and agreement of the methods of PI and BL/HC. Receiver Operating Characteristic (ROC) analysis was used to evaluate the validity of the diagnosis.
Results: Of 417 SGA infants, 376 (90.17%), 376 (90.17%) and 187 (44.84%) subjects were diagnosed as symmetric type with PI, BL/HC and the definition criteria, respectively. (2) The agreement rate and Kappa value between PI and BL/HC was 80.82% and -0.093 (SEM 0.026), respectively. And the agreement rates between PI or BL/HC and the definition criterion were 49.88% and 50.84%, respectively. As compared with the definition criterion, the PI and BL/HC methods had sensitivities of 91.8% - 96.4%, specificities of 9.3% - 25.9%, positive predictive values of 45.8% - 51.1%, negative predictive values of 72.7% - 82.8%, diagnosis indices of 4.9% - 17.7% and Kappa values of 0.070 - 0.167. (3) The areas under the ROC curves in full-term and preterm infants by PI method were 0.635 (95% CI, 0.573 - 0.697) and 0.698 (95% CI, 0.622 - 0.725), respectively. PI cutoffs at 2.47 in full-term SGA, at 2.43 in preterm SGA, and BL/HC cutoff at 1.43 produced the maximum diagnosis indices that were 24.7%, 39.6% and 33.7%, respectively. When the PI at 2.50 (full-term), PI at 2.31 (preterm) and BL/HC values at 1.46, the sensitivity closed mostly to the specificity. The sensitivities and specificities in full-term and preterm infants were 59.4% and 59.3%, 65.3% and 65.5%, and 66.3% and 65.5%, respectively.
Conclusion: In the classification of SGA, the results showed a poor agreement between PI or BL/HC and the definition criterion. The results suggested that the current cutoffs of PI and BL/HC might not be appropriate for the diagnosis of symmetric SGA. Low AUC suggested that PI and BL/HC could not give a valid diagnosis at any cutoffs.