Intrapartum sonographic foetal weight estimation versus the actual birth weight at Pumwani Maternity hospital

Abstract

In this prospective cross-sectional study, 206 gravid mothers admitted at PMH's labour ward had obstetric U/S scans done within 72 hours before delivery. In each case, foetal biometric parameters (FL, AC, BPD and HC) were measured and used to calculate the foetal weight using nine known sonographic foetal weight estimating formulae. The actual birth weight of each newborn was taken soon after its delivery. The foetal weight estimates obtained were then compared to the actual birth weight with the aim of assessing the reliability of each of the nine sonographic foetal weight estimating formulae in our general African population. The mean foetal weight estimate by each of the nine formulae showed a significant positive linear correlation with the mean actual birth weight(R value range 0.58 - 0.81where R is Pearson's correlation coefficient). The strongest correlation between the mean foetal weight estimate and the mean actual birth weight was given by Hadlock 3' s formular (R = 0.81) while the weakest correlation was given by Warsof 1's formular (R= 0.58). Gestational age showed a positive linear correlation to both the mean actual birth weight and the mean foetal weight estimates. Sonographic gestational age showed a stronger correlation to the mean foetal weights (R = 0.61 - 0.74) than gestational age as assessed from the LMP (R = 0.42 - 0.48). The mean absolute percentage error of estimating the foetal weight ranged from 7.83± 0.43 to 12.78 ±0.71.Twenty five point seven percent (25.7%) to forty six point one percent (46.1%) and fifty one percent (51%) to seventy one point eight percent (71.8%) of the mean foetal weight estimates were within 5% and 10% respectively of the mean actual birth weight. Comb's formular had the lowest mean absolute percentage error (7.83± 0.43) and the highest percentage (46.1% and 71.8%) of its mean estimates falling within the 5% and 10% brackets respectively of the mean actual birth weight. Warsof2's formular had the highest mean absolute percentage error (12.78 ±0.71) and the lowest percentage (25.7% and 51%) of its mean estimates falling within the five and ten percent brackets respectively of the mean actual birth weight. In terms of foetal weight categorization, Comb's formular showed the least bias with its predicted foetal weight categories being in substantial agreement with the actual birth weight categories among the study population (kappa value =0.62). Predicted foetal weight categories by Warsof 2' s formular were only in fair agreement with the actual birth weight categories (Kappa value = 0.39) while Warsof l's formular underestimated the weight of all the babies in the excessive weight category. Both Warsof's formulae therefore performed poorly in predicting the foetal weight category in the studied Kenyan population (mainly from low and middle socioeconomic status). The predicted weight categories by the remaining seven formulae were in moderate agreement with the actual birth weight categories (Table 6). In terms of sensitivity, specificity and accuracy in categorizing foetal weight into the three categories (low, average and excessive weight category) all the nine formulae showed limitations in at least one aspect. Shepard's, Hadlock 2's and 3's formulae were however fairly consistent in all the foetal weight categories. Despite having moderate sensitivity in predicting weight in the excessive weight category, Comb's and Ott's formulae were fairly accurate in predicting the weight categories (Table 7).