Noninvasive fetal sex determination using cell-free fetal DNA: a systematic review and meta-analysis

Abstract

Context: Noninvasive prenatal determination of fetal sex using cell-free fetal DNA provides an alternative to invasive techniques for some heritable disorders. In some countries this testing has transitioned to clinical care, despite the absence of a formal assessment of performance.

Objective: To document overall test performance of noninvasive fetal sex determination using cell-free fetal DNA and to identify variables that affect performance.

Data sources: Systematic review and meta-analysis with search of PubMed (January 1, 1997-April 17, 2011) to identify English-language human studies reporting primary data. References from review articles were also searched.

Study selection and data extraction: Abstracts were read independently to identify studies reporting primary data suitable for analysis. Covariates included publication year, sample type, DNA amplification methodology, Y chromosome sequence, and gestational age. Data were independently extracted by 2 reviewers.

Results: From 57 selected studies, 80 data sets (representing 3524 male-bearing pregnancies and 3017 female-bearing pregnancies) were analyzed. Overall performance of the test to detect Y chromosome sequences had the following characteristics: sensitivity, 95.4% (95% confidence interval [CI], 94.7%-96.1%) and specificity, 98.6% (95% CI, 98.1%-99.0%); diagnostic odds ratio (OR), 885; positive predictive value, 98.8%; negative predictive value, 94.8%; area under curve (AUC), 0.993 (95% CI, 0.989-0.995), with significant interstudy heterogeneity. DNA methodology and gestational age had the largest effects on test performance. Methodology test characteristics were AUC, 0.988 (95% CI, 0.979-0.993) for polymerase chain reaction (PCR) and AUC, 0.996 (95% CI, 0.993-0.998) for real-time quantitative PCR (RTQ-PCR) (P = .02). Gestational age test characteristics were AUC, 0.989 (95% CI, 0.965-0.998) (<7 weeks); AUC, 0.994 (95% CI, 0.987-0.997) (7-12 weeks); AUC, 0.992 (95% CI, 0.983-0.996) (13-20 weeks); and AUC, 0.998 (95% CI, 0.990-0.999) (>20 weeks) (P = .02 for comparison of diagnostic ORs across age ranges). RTQ-PCR (sensitivity, 96.0%; specificity, 99.0%) outperformed conventional PCR (sensitivity, 94.0%; specificity, 97.3%). Testing after 20 weeks (sensitivity, 99.0%; specificity, 99.6%) outperformed testing prior to 7 weeks (sensitivity, 74.5%; specificity, 99.1%), testing at 7 through 12 weeks (sensitivity, 94.8%; specificity, 98.9%), and 13 through 20 weeks (sensitivity, 95.5%; specificity, 99.1%).

Conclusions: Despite interstudy variability, performance was high using maternal blood. Sensitivity and specificity for detection of Y chromosome sequences was greatest using RTQ-PCR after 20 weeks' gestation. Tests using urine and tests performed before 7 weeks' gestation were unreliable.

Figure 1. Study Size by Publication Year

Included studies (N=57) are listed in eTable 2.

Figure 2. Literature Search

^aMost common: review study, marker not sex-specific, DNA extraction methodology.

Figure 3. Sensitivity and Specificity for 57 Included Studies (80 Data Sets) for Predicting Fetal Sex, With Summary Receiver Operator Curve

Included studies are listed in eTable 2. Diagonal dotted lines indicate the Q* index (point at which sensitivity=specificity on the curve). Indeterminate results were considered false-positive or false-negative depending on the true fetal sex. The curve shown in the right panel was fit to the data (including outliers) using the Moses-Shapiro-Littenberg model. Right panel is an enlarged view of the blue tinted region shown in the left panel.