Ksenia Olisova
Director of Medical Research
Director of Medical Research
Neonates with low birth weight account for up to 80% of neonatal deaths, and this number can be decreased if we apply appropriate fetal growth reference charts. Timely detection of small or large for gestational age babies allows doctors to adjust pregnancy management and delivery plan. Even though most of the expectant mothers choose to have a fetal size assessment scan during pregnancy, there is a lack of quality evaluation tool for biometric measurements, as well as a lack of uniform standard measurement protocol. This issue prevails not only in Taiwan but around the world, in addition to that there is no agreement between scientists, whether fetuses grow the same in different locations, or there are some genetic differences in growth.
Recently, a large international study was conducted to solve this dilemma and provide a uniform procedure for ultrasonographers to allow for within and between countries comparison. More than 300 researches from 18 countries coordinated from the University of Oxford with support from Bill and Melinda Gates Foundation established the International Fetal and Newborn Growth Consortium for the 21st Century, or INTERGROWTH-21st, a multidisciplinary network dedicated to improving perinatal health globally. One of the main projects for INTERGROWTH-21st became Fetal Growth Longitudinal Study. The study aimed to build international fetal growth standards with a prescriptive approach. The team of the researches started from the review of the methodology used in previous studies about fetal growth assessment, one of the main issues for those studies became a lack of quality control measurements. To address this issue, the INTERGROWTH team developed a rigorous protocol for quality evaluation.
Following the methodology proposed by the INTERGROWTH's fetal growth study, we conducted a series of pilot studies as a standardization exercise to assure intra-, interobserver variability. Standardization before the study guarantees that the participating sonographers will follow the same measurement procedure, and their measurements will be as close as possible. First of all, participating sonographers familiarized themselves with the detailed measurement protocol. The protocol presented a compilation of the protocol used by the INTERGROWTH study (BPD, OFD, HC, AC, FL) and several papers approved by ISUOG as practical guidelines for measuring CM, Vp, TCD, FiL, RL, NF, UtPI, UPI. Second, ultrasonographers recruited fifteen women, each woman was measured twice by each RT, meaning that for every biometric variable, we got four separate images. We tried to include as many biometric measurements as possible, but the results of the first trial were unsatisfying, as shown in Table 1. To check the inter-, intraobserver variability I calculated the intraclass correlation coefficients, first comparing RT1 and RT2 to assess interobserver variability, and, second comparing the measurement from the first scan to the measurement from the second. Another problem we faced, was the lengthy procedure, which could play a negative role on the returning rate for the study participants. We considered the variability reliable if intraclass correlation coefficient reached at least 90%. As we can see, for some measurements interrater variability was too high, meaning that the RTs could not reach perfect agreement between the measurements, which is unacceptable for the study.
Table 1 Intraclass correlation coefficients for comparison between RTs and between first and second scan for each RT
This result was frustrating, so I needed to come up with a practical solution. My first action was to reduce the number of measurements allowing sonographers to focus on the most important ones and decrease the time spent on each patient, which could also help to reduce patient dropout rate due to the tedious process. Another action was to review the measurement process and ask RTs to go through the protocol, assuring they were following the same procedure. They repeated the process once again with 15 patients. I reported data about each group of 5 women to control for the quality and reproducibility throughout the process, and the final report presents an excellent result. We changed the statistical method to the calculation of concordance correlation coefficient (CCC) as it better reflects the reliability of the measurements. On Figure 1 presented a graphical example of the CCC obtained during our last pilot study. The closer the dots to the line the better the concordance and it is clear, that the concordance between sonographers is perfect.
Figure 1 Concordance correlation coefficient, BPDoo
This pilot revealed the challenges of quality control for both researchers and participating sonographers. First of all, we needed to execute a smooth data collection process. We reached this goal using an online encrypted database in combination with the attached images. Second of all, the standardization exercise required the best performance from sonographers, so they needed to follow the study protocol without any deviations. And last but not least, continuous data monitoring is important to keep high-quality standards.
In conclusion, we hope that our experiences will help to achieve uniformity in the ultrasound biometric measurement acquisition. As we believe, following a uniform procedure used around Taiwan will allow launching a quality control system and comparing biometric parameters of the babies between different healthcare facilities around the island to provide the best perinatal care for the expectant mothers and their babies.
