There has been considerable interest in the role and possible predictive value of the circulating angiogenic factors sFlt1, sEng and PlGF for preeclampsia as evidenced by a growing literature since 2003. Most studies have focused on low-risk populations while very few have evaluated the utility of these factors in high-risk pregnant populations, and the latter reports are limited in size or have evaluated subjects with different reasons for being classified high-risk together 
. In our study, maternal circulating sFlt1, sEng and PlGF were significantly higher at study entry in subjects with multifetal gestations compared with other high-risk groups, and PlGF was significantly lower in the diabetic subjects. The odds of developing preeclampsia were significantly increased for each 2-fold elevation in sFlt1, sEng and the angiogenic factor ratio, and were significantly decreased for each 2-fold elevation in PlGF at study entry in subjects with multifetal gestations. Similarly, the odds of developing preeclampsia are significantly decreased by half for each 2-fold elevation in maternal PlGF among diabetics, and increased by 60% for each 2-fold elevation in sEng and the angiogenic factor ratio at study entry for subjects with chronic hypertension and diabetes respectively. While these odds ratios are statistically significant, the confidence intervals associated with them are wide, and the magnitude of the elevation in the angiogenic factors in early pregnancy is large, which does not lend confidence to the possibility of these factors individually as useful predictors of preeclampsia in these high-risk groups. However, these data do not preclude the possible utility of these factors in a multiple marker screen for the prediction of preeclampsia.
We also observed that the change in sFlt1 and sEng, and the change in the ratio of the angiogenic factors between study entry (7–26 weeks' gestation) and the second study visit (24–28 weeks' gestation) was associated with a significant increase in the odds of developing preeclampsia among these high-risk populations. These data are similar to that previously reported among low-risk patients, such that women who developed preeclampsia were more likely to show greater sequential changes in sFlt1 and sEng in early pregnancy, and these changes were most pronounced in preterm preeclampsia 
. These data suggest that perhaps the absolute concentration of these antiangiogenic factors is not as important as the relative change in these factors within an individual in early pregnancy, and that similar pathophysiologic mechanisms may be at work in low-risk and high-risk populations. While these data are intriguing, the absolute change in these factors associated with an increased risk of preeclampsia is large.
We observed that sFlt1 and sEng were statistically higher, and PlGF was significantly lower at 26 to 30 weeks' gestation or later during pregnancy in subjects who later develop preeclampsia in some but not all of the four high-risk groups. In addition, we observed that there were relatively few significant differences in the concentration of these angiogenic factors that precede the onset of the clinical syndrome of preeclampsia. In most cases, however, even when the differences are not significant, the relationship of mean values of these angiogenic factors is similar to that observed in low-risk women with higher sFlt1 and sEng and lower PlGF in women who later develop preeclampsia suggesting these factors may play a role in the pathophysiology of preeclampsia in these high-risk groups. Nonetheless, the small differences in these factors or their ratio make it unlikely they will be clinically useful as individual predictive markers of preeclampsia in these high-risk groups.
Among low-risk pregnant patients, sFlt1 has been reported to be significantly elevated by 33 to 36 weeks' gestation and >5 weeks before the clinical onset of preeclampsia 
. Similarly, maternal concentrations of sEng are significantly elevated at 25 to 28 weeks' gestation and 9 to 11 weeks before clinical onset of preterm preeclampsia, and 33 to 36 weeks' gestation and 9 to 11 weeks before the clinical onset of term preeclampsia 
. Maternal concentrations of PlGF are also reported to be significantly lower by 13 to 20 weeks' gestation among women who later develop preeclampsia compared with similar low-risk control subjects 
. In contrast to low-risk subjects, there has been limited investigation of the differences in these angiogenic factors across pregnancy among high-risk subjects. Moore Simas et al. evaluated the predictive value of maternal sFlt1 and PlGF between 22 and 36 weeks' gestation in a combination of several high-risk populations including patients with chronic hypertension and pre-gestational diabetes 
. This study reported that mean serum sFlt1 and sFlt1/PlGF ratio were higher at 22 weeks' gestation in subjects who developed early onset preeclampsia (<34 weeks) compared with uncomplicated controls, and that serum sFlt1 was significantly higher after 31 weeks' gestation in subjects with late onset preeclampsia (>34 weeks). The authors conclude that these angiogenic factors may be predictive of preeclampsia in high-risk populations. However, this study is limited by a small sample size, with 12 preeclamptic subjects total, 5 with early onset and 7 with late onset preeclampsia; as well as a significantly mixed population of high-risk groups (all 5 early onset preeclamptic subjects had chronic hypertension, 2 with pre-gestational diabetes, 1 with renal disease, and 2 with previous preeclampsia) making it difficult to conclude the significance of these factors within any single high-risk group. A study by the same group also reported that maternal serum sFlt1 and PlGF are significantly elevated in subjects with multifetal gestations as early as 22 weeks' gestation 
. These data are similar to the data reported in this study, as well as our finding that sEng is also significantly higher in all pregnant women with multifetal gestations. The explanation for the elevated angiogenic factors in women with multifetal gestations is likely related to the increased placental mass, however we lack sufficient information to provide a complete explanation. This same explanation was also proposed by Maynard et al. as well as reporting a significant correlation between circulating sFlt1 and placental mass in multifetal gestations (r
. In addition, a recent study by Sibai et al. investigated the predictive value of maternal sFlt1 and PlGF between 12 to 19 weeks and 24 to 28 weeks' gestation among 704 patients with previous preeclampsia and/or chronic hypertension (14.7% had both conditions) 
. The authors reported that PlGF concentrations were significantly lower at baseline and sFlt1 concentrations were higher at 24 to 28 weeks' gestation in subjects who later developed preeclampsia, and these differences were most noticeable among subjects who developed preeclampsia before 27 weeks' gestation. However, overall these differences were modest, and the sensitivities and positive predictive values were low suggesting that the clinical utility of these markers is limited.
In general, the patterns of angiogenic factor concentrations across pregnancy in this study (i.e. elevated sFlt1 and sEng and low PlGF among subjects who develop preeclampsia) are similar to those presented in studies of low-risk pregnant women. However, differences in these angiogenic factors between high-risk women who develop preeclampsia compared with high-risk women who do not develop preeclampsia appear more modest, similar to the data presented by Sibai et al 
. What then might be the possible reasons for these differences between studies involving low-risk pregnant women compared with high-risk pregnant women? One notable technical difference is a change in the formulation of the commercial kit for the measurement of sFlt1. Most prior studies and our study used the ELISA kit from the same company (R&D Systems; Minneapolis, MN). A formulation change in this kit made since the publication of several of the prior studies resulted in higher measured concentrations of sFlt1 than reported previously among low-risk pregnant women. However, the formulation of the commercial ELISA kits for sEng and PlGF have not changed, and the overall pattern of maternal serum sFlt1 concentrations in high-risk women is similar to that of low-risk women.
In this study we decided a priori to analyze these high-risk groups separately. Preeclampsia is a heterogeneous condition and it is quite likely that the path to preeclampsia might be different between these disparate high-risk conditions. Despite this, limitations in sample size are an unlikely explanation since the number of high-risk subjects within each group in this study is considerably greater than that presented in other studies of high-risk subjects, and the number of samples in this study was in many cases as numerous as those used in studies of low-risk women. However, despite the large number of subjects and samples in this study, we were not able to separate groups according to preterm severity criteria (<34 weeks' gestation) which is when the most pronounced differences in these angiogenic factors appear among low-risk subjects, because of an insufficient number of these preterm subjects in each of the high risk groups 
. An additional limitation of this study is the fact that the original parent study was not designed to predict preeclampsia. However, despite these limitations, all samples in this study were analyzed in a blinded fashion, the inter-assay variability for the measurement of each angiogenic factor was within acceptable limits and similar to that presented in other studies, the overall variability of the data particularly for sEng was low and similar to that presented in other studies, the majority of samples were thawed only one time (74%) or thawed two times (26%), and the storage time and handling of samples was similar to that of other studies including the CPEP cohort which reported similar values to those reported in this study  
. Importantly, aspirin was found to have no significant effect on the concentration of any of the angiogenic factors. It is possible that difficulty in determining the clinical diagnosis of preeclampsia in subjects with chronic hypertension and especially these and other subjects with preexisting proteinuria may have affected the results of this study. However, the diagnosis of preeclampsia in these subjects was critically evaluated by chart reviews in the initial parent study. In addition, this limitation provides additional rationale for maintaining separate diagnostic groups. The clinical diagnosis of preeclampsia in patients with previous preeclampsia or multifetal gestations was not confounded by preexisting conditions, yet even these high-risk groups had only modest differences in the measured angiogenic factors.
In conclusion, we have observed that the maternal concentration of the angiogenic factors sFlt1, sEng and PlGF are significantly higher among women with multifetal gestations compared with other high-risk groups including: pre-existing diabetes, chronic hypertension and previous preeclampsia. The sequential change in sFlt1, sEng and the ratio of these angiogenic factors in early pregnancy is associated with a significant increase in the odds of developing preeclampsia. However, these sequential changes need to be large. Cross-sectional analysis of maternal sFlt1, sEng and PlGF show modest significant differences of at least one of these factors during the third trimester in women who develop preeclampsia in all high-risk groups. Only sFlt1 was significantly higher 2 to 5 weeks before the clinical onset of preeclampsia in women with previous preeclampsia. The significant differences in sFlt1, sEng and PlGF evident in women at high-risk of preeclampsia may support a role for these factors in the pathophysiology of preeclampsia. However, we conclude that the concentration of these circulating angiogenic factors is unlikely to be useful in predicting preeclampsia in these high-risk populations.