Validation analysis found that Lep/Cer (d18:1/25:0) ratio predicts PE early in gestation
When we evaluated the predictive performance of the Lep/Cer (d18:1/25:0) ratio in the validation cohort (20 women without PE with 55 samples, and 20 women with PE with 51 samples; see Table IV and Fig. 5), We found that among the 20 PE women, 5 and 13 had early- and late-onset PE, respectively. The dates of diagnosis were missing in the remaining 2 women.
We observed an increase in serum Lep (P < 0.0001; 1.71-fold) and a decrease in Cer (d18:1/25:0) (P < 0.0001; 0.70-fold) in PE women at 5 to 29 wks’ of gestation (Appendix 6), which resulted in an elevated Lep/Cer (d18:1/25:0) ratio (Fig. 6). Most notably, this ratio was a significantly better predictor of all types of PE (AUC = 0.887) than Lep (AUC = 0.809; P = 0.0006) or Cer (d18:1/25:0) (AUC = 0.790; P = 0.008) levels alone. In addition, the ratio performed well at wider gestational windows from 5 to 15 and 16 to 29 wks (AUC = 0.876 and 0.892, respectively) than the individual markers: Lep (AUC = 0.868; P = 0.4 and 0.824; P= 0.1, respectively) and Cer (d18:1/25:0) (AUC =0.868; P = 0.1 and 0.747; P = 0.02, respectively). These results were consistent with the univariate and multivariate results found in non-obese women (Appendix 7). The Lep/Cer (d18:1/25:0) ratio accounted for 31% of the variation in the PE outcome, higher than that with Lep (15%), Cer (d18:1/25:0) (15%), and pre-pregnancy body mass index (BMI) (8%). In a multivariate analysis, the explained variance of Lep was dominant early in gestation (5 to 15 wks), while Cer (d18:1/25:0) was dominant in mid-gestation (16 to 29 wks; Appendix 7B). The longitudinal profiling of Lep and Cer (d18:1/25:0) levels improved the predictive performance of Lep/Cer (d18:1/25:0) ratio. Moreover, the Lep/Cer (d18:1/25:0) ratio outperformed the sFlt-1/PlGF ratio in predicting impending PE, with a lower P (Fig. 6) and a higher AUC (P < 0.0001; Appendix 6D).
Time-to-event analysis at 5 to 25 wks (Fig. 7) compared the Lep/Cer (d18:1/25:0) ratio and the sFlt-1/PlGF ratio in predicting the impending PE. Among the 18 PE women with known diagnoses dates, 83% (15/18; 5 early-onset and 10 late-onset) were identified by the Lep/Cer (d18:1/25:0) ratio 11 or more weeks prior to their clinical diagnosis. In contrast, the sFlt-1/PlGF ratio only identified 22% (4/18; 1 early-onset and 3 late-onset) of PE women 11 or more weeks prior to the diagnosis. The Lep/Cer (d18:1/25:0) ratio was able to predict impending PE a median of 23.0 [95% confidence interval (CI): 12.8, 30.7] wks prior to the confirmatory diagnosis.
As shown in Fig. 8A, the Lep/Cer (d18:1/25:0) ratio correctly classified 85% (17/20) of women with impending PE and 90% (18/20) of pregnancies without PE at 5 to 25 wks, giving a sensitivity of 85% (17/20), a specificity of 90% (18/20), a positive predictive value (PPV) of 89% (17/19), and a negative predictive value (NPV) of 86% (18/21). In contrast, 40% (8/20) of women with subsequent PE and 45% (9/20) of women without PE were correctly classified by the sFlt-1/PlGF ratio, yielding a sensitivity of only 40% (8/20), a specificity of 45% (9/20), a PPV of 42% (8/19), and a NPV of 43% (9/21) (Fig. 8B). In addition, the Lep/Cer (d18:1/25:0) ratio had a higher AUC than the sFlt-1/PlGF ratio at 5 to 25 wks [0.92 (95% CI: 0.86, 0.98) vs. 0.52 (95% CI: 0.39, 0.64); P < 0.001].