Predicting Preeclampsia using sFlt-1:PlGF Ratio in Symptomatic Women

In January 2016 study (funded by Roche Diagnostics), Zeisler, et. al. examined the predictive value of the sFlt-1:PlGF ratio to predict the absence or presence of preeclampsia in women with suspected preeclampsia. The study included two groups of pregnant women with elevated blood pressure and suspected preeclampsia: a development cohort with 500 women (101 had preeclampsia or HELLP syndrome) and a validation cohort with 550 women (98 had preeclampsia).

Using the development cohort, the authors established an sFlt-1:PlGF ratio of 38 as a cutoff for predicting preeclampsia. Using the validation cohort, the authors used the sFlt-1:PlGF ratio of 38 to determine the ability to rule out preeclampsia within 1 week of presenting with symptoms and the ability to rule in preeclampsia within 4 weeks of presenting with symptoms. In the validation cohort, 15 women developed preeclampsia within 1 week (prevalence = 15/550= 2.7%) and the sFlt-1:PlGF ratio of 38 demonstrated a negative predictive value of 99.3% to rule out preeclampsia within one week of presenting with symptoms. In the validation cohort, 71 women developed preeclampsia within 4 weeks (prevalence = 71/550= 13%) and they reported a positive predictive value of 36.7% to predict preeclampsia within 4 weeks.  The authors conclude that an sFlt-1:PlGF ratio of 38 or lower can be used to predict the short term absence of preeclampsia in women who are suspected of having preeclampsia.

I have depicted their data below in a familiar 2×2 format.

One can conclude that the sFlt-1:PlGF ratio has an excellent negative predictive value to rule out preeclampsia in both 1 week and 4 weeks (99.3 and 94% respectively), but the positive predictive value is poor for both (9.2 and 36.7% respectively). However, one needs to think critically about the data. In both populations, despite the fact that the women have symptoms of preeclampsia, the prevalence of preeclampsia is still low (2.7% within 1 week and 13% within 4 weeks). Therefore, a marker with high positive predictive value is needed.

If we take the data from the within 1 week population in the table above and instead of using the sFlt-1:PlGF ratio, we flip a coin such that we get a sensitivity of 50% and specificity of 50% the negative predictive value is still 97% (see table below). The sFlt-1:PlGF ratio improves the negative predictive value by 3% over the flip of a coin. Why is this? This is because the pretest probability of not developing preeclampsia within one week, in these symptomatic women, is already 97.3%. The sFlt-1:PlGF ratio adds little to the negative predictive value.

In summary, it is clear that markers with a better positive predictive value are still needed to accurately predict women who are likely to develop preeclampsia even in a population of pregnant women with signs and symptoms.