Background/Aims: Circular RNAs (circRNAs) are transcribed prevalently in the genome; however, their potential roles in multiple cardiovascular diseases, particularly preeclampsia (PE), are not yet well understood. This study investigated the expression profiles of circRNAs and explored circRNA-mediated pregnancy-associated plasma protein A (PAPP-A) expression as a potential biomarker for PE before 20 weeks of pregnancy. Methods: A nested case-control two-phase screening/validation study was performed in pregnant women before 20 weeks of gestation (before clinical diagnosis) at Guangzhou Women and Children’s Medical Center from 2012 to 2015. In the screening phase, circRNA expression profiles of blood cells were assessed using a human circRNA microarray, which was designed to detect simultaneously 5396 circRNAs, in 5 patients with PE and 5 age- and gestational week-matched controls. In the validation phase, 18 circRNAs in blood cells predicted by bioinformatics tools were validated by quantitative reverse transcription PCR in a cohort of 60 patients (PE and age-, gestational week-, and sample storage time-matched controls). Then, we examined the involvement of circRNAs in PE-related pathways via interactions with miRNAs by multiple bioinformatics approaches. Bioinformatics analysis predicted that hsa_circ_0004904 and hsa_circ_0001855 miRNA sponges directly target PAPP-A. PAPP-A was verified in the serum of the same cohort of patients using an enzyme-linked immunosorbent assay. Finally, we combined PAPP-A with circRNAs to create a novel preclinical diagnostic model for PE with logistic regression and evaluated the efficiency of this model with receiver operating curve analysis. Results: Volcano plot analysis using various parameters showed that circRNAs were differentially expressed among both groups (P < 0.01, fold change > 2). In the screening phase, we found that 2178 circRNAs were differentially expressed between the control and PE groups, in which 884 circRNAs were downregulated and 1294 circRNAs were upregulated in the PE group compared with the control group. In the validation phase, two circRNAs, hsa_circ_0004904 and hsa_circ_0001855, were significantly upregulated in PE patients compared with healthy pregnant women (P < 0.05). PAPP-A expression levels, related to the two circRNAs based on bioinformatics prediction, were increased in the PE group compared with the control group. The area under the curve of the combined model was 0.94 in the predicted PE subjects. Conclusions: This is the first study to report circRNA profiling in patients with PE prior to the onset of symptoms. Our data suggested that hsa_circ_0004904 and hsa_circ_0001855 combined with PAPP-A might be promising biomarkers for the detection of PE. Moreover, circRNAs may provide new insights into the potential mechanisms underlying the pathophysiology of PE.