Construction and validation of pregnancy risk prediction model for pregnant women with chronic kidney disease / 中华妇产科杂志

ABSTRACT

Objective:To construct the pregnancy risk prediction model of chronic kidney disease (CKD) pregnant women by analyzing their renal function and pregnancy outcome in the first trimester.Method:Totally 313 CKD women with 322 pregnancies who had deliveries in Peking University First Hospital from March 2009 to December 2018 were retrospectively analyzed. The history of kidney disease and renal function in the first trimester were collected, and the relationship between CKD and premature delivery, low birth weight infants, severe preeclampsia and fetal loss were analyzed.Result:Among 322 pregnancies with CKD, 120 (37.3%, 120/322) had adverse pregnancy outcomes. CKD stage, serum creatinine, urea, albumin, hemoglobin, 24-hour urine protein quantity and whether complicated with hypertension were independent predictors of adverse pregnancy outcome. A prediction model logit (P)=2.107+0.255×24-hour urine protein quantitative (g/24-hour)-0.107×albumin (g/L)+1.677×whether complicated with hypertension (1 or 0)+ 0.639×CKD stage was established. The area under curve value of the model was 0.812, the best threshold, sensitivity, specificity and Yoden index were 0.436, 0.658, 0.856 and 0.802, respectively.Conclusion:CKD stage, serum albumin, 24-hour urine protein quantity in the first trimester and hypertension are the main risk factors of adverse pregnancy outcome, which could predict the occurrence of adverse pregnancy outcome of CKD pregnant women and deserve further study.