Inter Observer Agreement of Amniotic Fluid Index Measurement Across Maternal BMI Groups at 36-40 Weeks

Abstract

One of the useful parameters in obstetric sonography is the amniotic fluid index (AFI). Indirectly evaluates fetal well-being, uteroplacental function and amniotic fluid regulation. AFI measurements are, however, dependent on the technique used by the operator and are subject to variation depending on the placement of the probe, scanning angle, the pressure of the transducer on the skin and the ability to detect fluid pockets. Further, maternal body mass index (BMI) can also be a problem as increased adipose tissue results in loss of strength of ultrasound beam and decreases spatial resolution, which makes it more difficult to view amniotic fluid pockets. This can cause an increase in the variation of opinions between observers. This is particularly important for clinical decision making during late pregnancy (36-40 weeks), but information on inter-observer agreement by BMI category in low and middle income countries such as Pakistan is scarce. The study was cross sectional observational study conducted for 4 months in Asfand Clinic, Rahim Yar Khan, Pakistan. A consecutive non-probability sample of 110 singleton pregnancies between 36-40 weeks of gestation were recruited. Sample size was determined with a standard formula (Z = 1.96, P = 0.20, d = 0.075) to yield n = 110 at 5% significance and 80% power levels and using a standard formula to calculate the Cronbach Reliability Coefficient, which was 78% for AFI. The four quadrant AFI was done by two independent expert sonographers (Observer 1 and Observer 2) with GE S6 and Aplio 300 ultrasound machines using a 3-5 MHz convex transducer and these observers were not aware of the results of each other's measurements. Participants were categorized by BMI: Normal (18.5-24.9 kg/m², n = 10), Overweight (25.0–29.9 kg/m², n = 92), and Obese (≥30.0 kg/m², n = 8). The Pearson r, ICC (two-way mixed absolute agreement), Cronbach's alpha, Bland-Altman analysis and One-way ANOVA were used for statistical analysis and were carried out in IBM SPSS Statistics v25. Pearson r value was 0.916 (p < 0.001); ICC was 0.915 (95% CI: 0.877–0.940); Cronbach's alpha value was 0.955, indicating good overall inter-observer agreement. Bland-Altman analysis showed that the systematic bias was negligible with mean bias of -0.013 cm (p = 0.900), and limits of agreement [-2.10, +2.07] cm, with 91.8% measurements lying within these limits. High values of ICC for the subgroups were also found for Normal (ICC = 0.905), Overweight (ICC = 0.897) and Obese (ICC = 0.996). The absolute differences in measurements between BMI groups were not significant (F = 1.117, df = 2,107; p = 0.331).