Obstet Gynecol 2022 03;139(3):400-406
Departments of Obstetrics and Gynecology, Brown University, Providence, Rhode Island, University of Texas Medical Branch at Galveston, Galveston, Texas, Northwestern University, Chicago, Illinois, Columbia University, New York, New York, University of Utah Health Sciences Center, Salt Lake City, Utah, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, University of Alabama at Birmingham, Birmingham, Alabama, The Ohio State University, Columbus, Ohio, Duke University, Durham, North Carolina University of Colorado School of Medicine, Anschutz Medical Campus, Aurora, Colorado, Case Western Reserve University, Cleveland, Ohio, University of Texas Health Science Center at Houston and Children's Memorial Hermann Hospital, Houston, Texas, Stanford University, Stanford, California, University of Texas Southwestern Medical Center, Dallas, Texas University of Pennsylvania, Philadelphia, Pennsylvania, University of Pittsburgh, Pittsburgh, Pennsylvania, Madigan Army Medical Center, Joint Base Lewis-McChord, Washington, and Washington University in St. Louis, St. Louis, Missouri; the Departments of Pediatrics, University of Alabama at Birmingham, Birmingham, Alabama, and Duke University, Durham, North Carolina; the George Washington University Biostatistics Center, Washington, DC; and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, Bethesda, Maryland.
Objective: To develop and internally validate a noninvasive method for the prediction of congenital cytomegalovirus (CMV) infection after primary maternal CMV infection.
Methods: We conducted a secondary analysis of a multicenter randomized placebo-controlled trial of CMV hyperimmune globulin to prevent congenital infection. Women were eligible if they had primary CMV infection, defined as detectable plasma CMV-specific immunoglobulin (Ig)M and CMV-specific IgG with avidity less than 50% before 24 weeks of gestation or IgG seroconversion before 28 weeks, and were carrying a singleton fetus without ultrasonographic findings suggestive of CMV infection. Antibody assays were performed in a single reference laboratory. Congenital infection was defined as CMV detection in amniotic fluid, neonatal urine or saliva, or postmortem tissue. Using backward elimination, we developed logit models for prediction of congenital infection using factors known at randomization. The performance of the model was assessed using leave-one-out cross-validation (a method of internal validation).
Results: Of 399 women enrolled in the trial, 344 (86%) had informative data for this analysis. Congenital infection occurred in 68 pregnancies (20%). The best performing model included government-assisted insurance, IgM index 4.5 or higher, IgG avidity less than 32%, and whether CMV was detectable by polymerase chain reaction in maternal plasma at the time of randomization. Cross-validation showed an average area under the curve of 0.76 (95% CI 0.70-0.82), indicating moderate discriminatory ability. More parsimonious one-, two-, and three-factor models performed significantly less well than the four-factor model. Examples of prediction with the four-factor model: for a woman with government-assisted insurance, avidity less than 32%, IgM index 4.5 or higher, and detectable plasma CMV, probability of congenital infection was 0.69 (95% CI 0.53-0.82); for a woman with private insurance, avidity 32% or greater, IgM index less than 4.5, and undetectable plasma CMV, probability of infection was 0.03 (95% CI 0.02-0.07).
Conclusion: We developed models to predict congenital CMV infection in the presence of primary maternal CMV infection and absence of ultrasonographic findings suggestive of congenital infection. These models may be useful for patient counseling and decision making.