In this retrospective study, we identified that the level of SCr could predict the risk of adverse pregnancy outcomes as well as the number of concurrent adverse pregnancy outcomes. Specifically, when using the GA-specific SCr distribution16the predictive power of adverse pregnancy outcomes was more robust based on beta coefficients and their p-values, compared to the raw distribution of SCr (Fig. 3). Furthermore, we identified that the SCr groups (Q1adj.–Q4adj.) determined by GA-specific SCr distribution exhibited a nonlinear U-shaped relationship with the risk of adverse pregnancy outcome. Park et al.9 reported the results consistent with our findings that midline GFR had a nonlinear relationship with adverse pregnancy outcomes.
Several studies have shown that the exclusive use of SCr-based equations could lead to misclassification of kidney function during pregnancy.5.16. To overcome this limitation, Park et al.9 only analyzed mid-term GFR values estimated on the basis of SCR levels and their relationship to adverse pregnancy outcomes. Additionally, Harel et al.16 collected approximately 362,000 RCS results from approximately 240,000 women and generated the GA-specific RCS distribution. We could not find the GA-specific GFR distribution and we could only obtain the GA-specific SCr distribution16, so we designed the linear regression or logistic regression models based on SCr levels, not GFR values. Direct application of results obtained from the Canadian population to the population of Korean pregnant women may produce inaccurate results due to genetic and ethnic differences. However, we considered this distribution to be generalized and unbiased results because it was obtained from over 360,000 results.16. Moreover, it was evident to provide a better prediction of unfavorable pregnancy outcomes of Korean women when using the Canadian-based distribution. Future studies to organize Korea-specific SCR distribution based on GA are needed. Moreover, it is crucial to establish the GA-specific GFR distribution.
SCr value depends on muscle mass or body composition, such as age, gender, and ethnicity17. Due to many factors that negatively impact accurate GFR estimation, the Kinney Disease Improving Global Outcomes (KDIGO) guidelines recommend confirming CKD status in a specific population initially determined based on the SCr-estimated GFR (eGFRCR) using an alternative method, such as GFR estimated by serum cystatin C (eGFRcys)18. Our study also focused on the inaccurate estimation of renal function measured by CRS in a specific population (ie.16 may improve the prediction of adverse pregnancy outcomes. During pregnancy, despite renal hyperfiltration, maternal serum cystatin C is known to be more stable than SCr before the third trimester19. Serum cystatin C is known to be elevated during the third trimester of pregnancy; however, its underlying mechanism is not converged19,20,21. In the general population, cumulative evidence has supported that eGFRcys is a good alternative method for estimating GFR, especially in the subgroup with low amount of muscle: elderly population, vegetarians and people with muscle atrophy, chronic disease or limb amputation22. Additionally, most studies generally suggested that a combined equation of creatinine and cystatin C provided more accurate estimates and greater precision than GFR estimated from creatinine or cystatin C alone.22.23. Taken together, we suggest that kidney function in pregnant women should be measured via several serum indices, such as creatinine and cystatin C.
Adverse pregnancy outcomes, such as preeclampsia and PTB, have been considered the leading causes of perinatal morbidity and mortality worldwide24.25. Therefore, establishing a model for predicting adverse pregnancy outcomes is crucial to help minimize adverse perinatal outcomes.26.27. Establishing the prediction model included two main tasks: one is feature selection and the other is parameter optimization using multiple machine learning methods.28. This study focused on feature selection. In addition, our results were obtained from previously validated findings16also called prior knowledge, indicating that our study analyzed data based on the Bayesian approach or transfer learning29. Our results indicated a nonlinear relationship between SCR levels and adverse pregnancy outcomes; hence machine learning methods (e.g. support vector machine, random forest or deep learning)30 that can process data with characteristics of nonlinear or complex relationships between entities30 are needed to screen patients with high-risk adverse pregnancy outcomes.
In the present study, multidisciplinary experts (laboratory, database administrator, obstetrician and computer scientist) performed several tasks to build a database to identify risk factors for adverse pregnancy outcomes. Initially, obstetricians are responsible for literature-based review to select candidate features. Later, the lab worker, database administrator, and computer scientist constructed an initial dataset using the automated platform, including candidate variables. Next, the obstetrician and laboratory technician manually updated the patient’s medical history and pregnancy outcomes.
This study has several limitations. First, because the WSCH is a tertiary hospital, the prevalence of pregnant women with adverse birth outcomes is high compared to the general population. To obtain generalized results, the multi-institutional study or registry is necessary, and this study could motivate the multicenter approach. Second, several biomarkers could not be analyzed because the database (eg, electronic health records and automated platform) is not designed to compile them. For example, we could not analyze data such as body mass index (BMI) or waist circumference because they were not recorded at the time of the initial pregnancy assessment. Also, due to the retrospective design, we could not control the re-examination of pregnant women’s blood tests, so most of the women examined had only one measurement point in time (Fig. 1). . Future study bolstered by these limitations could identify valuable pathophysiological signatures related to decreased renal function and blunted hyperfiltration, as well as adverse pregnancy outcomes.
We implemented prior knowledge6 obtained from the huge scale of data and observed that high SCR levels were significantly related to the risk of adverse pregnancy outcomes. This study is a fundamental task of developing an algorithm to predict the possibility of adverse pregnancy outcomes based on pregnant women’s SCr levels adjusted for weeks of gestation.