First-trimester predictive models for adverse pregnancy outcomes—a base for implementation of strategies to prevent cardiovascular disease development

Introduction This study aimed to establish efficient, cost-effective, and early predictive models for adverse pregnancy outcomes based on the combinations of a minimum number of miRNA biomarkers, whose altered expression was observed in specific pregnancy-related complications and selected maternal clinical characteristics. Methods This retrospective study included singleton pregnancies with gestational hypertension (GH, n = 83), preeclampsia (PE, n = 66), HELLP syndrome (n = 14), fetal growth restriction (FGR, n = 82), small for gestational age (SGA, n = 37), gestational diabetes mellitus (GDM, n = 121), preterm birth in the absence of other complications (n = 106), late miscarriage (n = 34), stillbirth (n = 24), and 80 normal term pregnancies. MiRNA gene expression profiling was performed on the whole peripheral venous blood samples collected between 10 and 13 weeks of gestation using real-time reverse transcription polymerase chain reaction (RT-PCR). Results Most pregnancies with adverse outcomes were identified using the proposed approach (the combinations of selected miRNAs and appropriate maternal clinical characteristics) (GH, 69.88%; PE, 83.33%; HELLP, 92.86%; FGR, 73.17%; SGA, 81.08%; GDM on therapy, 89.47%; and late miscarriage, 84.85%). In the case of stillbirth, no addition of maternal clinical characteristics to the predictive model was necessary because a high detection rate was achieved by a combination of miRNA biomarkers only [91.67% cases at 10.0% false positive rate (FPR)]. Conclusion The proposed models based on the combinations of selected cardiovascular disease-associated miRNAs and maternal clinical variables have a high predictive potential for identifying women at increased risk of adverse pregnancy outcomes; this can be incorporated into routine first-trimester screening programs. Preventive programs can be initiated based on these models to lower cardiovascular risk and prevent the development of metabolic/cardiovascular/cerebrovascular diseases because timely implementation of beneficial lifestyle strategies may reverse the dysregulation of miRNAs maintaining and controlling the cardiovascular system.


Introduction
MiRNAs are small non-coding RNAs (18-25 nucleotides) that regulate gene expression at the post-transcriptional level (Lai, 2002;Bartel, 2004).Increased miRNA expression results in the degradation of mRNAs or blockage of translation of potential target genes.Conversely, upregulation of potential target genes results from decreased miRNA levels.An altered miRNA expression profile usually contributes to the pathophysiology of the disease and may be used for the diagnosis and/or the assessment of prognosis of the disease (Piletič and Kunej, 2016;Wang et al., 2016;Condrat et al., 2020).
Afterwards, we identified multiple independent risk factors predisposing to the development of pregnancy-related complications such as maternal age and body mass index (BMI) at early stages of gestation, nulliparity, confirmed diagnosis of autoimmune disease, infertility treatment using assisted reproductive technology, presence of chronic hypertension, presence of thrombophilia gene mutations, history of pregnancyrelated complications (PE, HELLP, SGA, FGR, and preterm birth) in previous pregnancy (ies), history of miscarriage (before 20 gestational weeks), and occurrence of diabetes mellitus in first-degree relatives (Hromadnikova et al., 2024;Hromadnikova et al., 2023a;Hromadnikova et al., 2023b;Hromadnikova et al., 2022e;Hromadnikova et al., 2022d;Hromadnikova et al., 2023c).
Subsequently, we involved these maternal clinical characteristics in miRNA-based predictive models, which increased the detection rate of pregnancies at high risk of adverse pregnancy outcomes (Hromadnikova et al., 2024;Hromadnikova et al., 2023a;Hromadnikova et al., 2023b;Hromadnikova et al., 2022e;Hromadnikova et al., 2022d;Hromadnikova et al., 2023c).In addition, we added firsttrimester screening for PE and/or FGR and spontaneous preterm birth, both determined using the FMF algorithm (Tan et al., 2018), to the predictive models for GH, PE, HELLP syndrome, FGR, SGA, and GDM, as these two independent variables slightly increased the detection rates.
Currently, we focused on the development of efficient, costeffective, early predictive models for identifying adverse pregnancy outcomes based on a selection of six miRNAs (miR-181a-5p, miR-20a-5p, miR-146a-5p, miR-574-3p, miR-1-3p, and miR-16-5p), whose altered expression was a common phenomenon shared between multiple pregnancy-related complications (Table 1).These miRNAs were combined with maternal clinical characteristics previously identified as the risk factors for a complicated gestational course (Table 2).

Inclusion and exclusion criteria
Pilot and validation studies were performed.Sample size calculation was used to calculate the minimal required sample size of subjects for analyses.
All procedures were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and the Helsinki Declaration of 1964 and its later amendments.All the included patients provided informed consent for participation in the study.The Ethics Committee of the Third Faculty of Medicine, Charles University, granted initial approval for this study (Implication of placental-specific miRNAs in maternal circulation for diagnosis and prediction of pregnancy-related complications, date of approval: 7 April 2011).Ongoing approval for the study was obtained from the Ethics Committee of the Third Faculty of Medicine, Charles University (Long-term monitoring of complex cardiovascular profiles in mother, fetus, and offspring descending from pregnancy-related complications, date of approval: 27 March 2014) and the Ethics Committee of the Institute for the Care of the Mother and Child, Charles University (Long-term monitoring of complex cardiovascular profiles in mother, fetus, and offspring descending from pregnancyrelated complications, date of approval: 28 May 2015, number of approval: 1/4/2015).Informed consent is a complex process as it involves attaining consent for collecting peripheral blood samples at the beginning of pregnancy.In addition, it also includes gaining consent for collecting peripheral blood samples at the onset of pregnancy-related complications and collecting placental samples during childbirth in case of the onset of pregnancy-related complications.
Briefly, total RNA enriched for small RNAs was isolated from whole peripheral venous blood (EDTA) using a mirVana miRNA isolation kit (Ambion, Austin, United States of America).mRNAs of miRNAs of interest were reverse transcribed into complementary DNA (cDNA) using miRNA-specific stem loop primers and TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems, Branchburg, United States of America).Reverse transcription was performed in a total reaction volume of 10 µL.
Subsequently, 3 µL of cDNA was mixed in a total reaction volume of 15 µL with specific primers, TaqMan MGB probes (the components of TaqMan MicroRNA Assays), and the components of the TaqMan Universal PCR Master Mix (Applied Biosystems, Branchburg, United States of America).Real-time RT-qPCR was performed on a 7,500 Real-Time PCR System under standard TaqMan PCR conditions described in the TaqMan guidelines.The miRNA gene expression was determined using the comparative Ct method (Livak and Schmittgen, 2001).The normalization factor (Vandesompele et al., 2002) (geometric mean of Ct values of selected endogenous controls: RNU58A and RNU38B) was used to normalize the miRNA gene expression data.

Statistical analysis
Predictive models for adverse pregnancy outcomes were constructed using logistic regression and receiver operating characteristic (ROC) curve analyses (MedCalc Software bvba, Ostend, Belgium).ROC curves displayed the areas under the curves (AUC), the cut-off points associated with sensitivities, specificities, positive and negative likelihood ratios (LR+, LR-), and sensitivities at 10.0% false positive rate (FPR) (MedCalc Software bvba, Ostend, Belgium).Initially, all independent variables (selected miRNAs and maternal clinical characteristics) and dependent variables (diagnoses, for example preeclampsia -1, normal term pregnancies -0) were entered into the logistic regression models for particular pregnancy-related complications.
Subsequent ROC curve analyses were applied (MedCalc Software bvba, Ostend, Belgium), where the predictive probabilities gained from logistic regression analyses were saved and next used as the new variables and the diagnoses (for example preeclampsia -1, normal term pregnancies -0) acted as the classification variables in ROC curve analyses.

Analysis of MiRNA-target interactions
The miRWalk database (http://mirwalk.umm.uni-heidelberg.de/) and disease ontology module (http://mirwalk.umm.uniheidelberg.de/diseases/)were used to provide information on the predicted and/or validated targets of miRNAs.Pregnancy-related complications, such as preeclampsia, HELLP syndrome, placental insufficiency, and GDM were available in the miRWalk database.The only common targets associated with pregnancy-related complications, cardiovascular risk factors (obesity, hypertension, atherosclerosis, prediabetes syndrome, and diabetes mellitus), and cardiovascular and cerebrovascular diseases (myocardial infarction, cerebral infarction, systolic and diastolic heart failure, and heart, cardiovascular, and cerebrovascular diseases as a whole) were reported.

The cost-effective first-trimester predictive models for adverse pregnancy outcomes
The cost-effective first-trimester predictive models for adverse pregnancy outcomes were based on the combinations of a minimum number of miRNA biomarkers with jointly altered expression during the early gestational stages.In addition, maternal clinical characteristics identified as the risk factors for adverse pregnancy outcomes were added into the predictive models.
Predictive models based on the combinations of these miRNAs and selected maternal clinical characteristics identified as the risk factors for appropriate adverse pregnancy outcomes in our previous studies showed higher detection rates at 10.0% FPR (GH, 62.65%; PE, 78.79%; HELLP syndrome, 85.71%; FGR, 58.54%; SGA, 70.27%; GDM on therapy, 78.95%; late miscarriage, 84.85%; and preterm delivery in the absence of the above-mentioned pregnancy-related complications, 45.28%) (Hromadnikova et al., 2022d;Hromadnikova et al., 2022e;Hromadnikova et al., 2023a;Hromadnikova et al., 2023b;Hromadnikova et al., 2023c;Hromadnikova et al., 2024) (Table 4).In the case of stillbirth, maternal clinical characteristics need not be added to the predictive model because the detection rate of cases was high only when using a combination of appropriate miRNAs.
More advanced predictive models, which included the results of first-trimester screening for PE and/or FGR and spontaneous preterm birth using the FMF algorithm, increased the detection rates of various adverse pregnancy outcomes at 10.0% FPR (GH: 69.88% cases; PE: 83.33% cases; HELLP syndrome: 92.86% cases; FGR: 73.17% cases; SGA: 81.08% cases; GDM on therapy: 89.47% cases; and preterm delivery in the absence of the above-mentioned pregnancy-related complications: 51.89% cases).In the case of late miscarriage, the detection rate remained the same at a FPR of 10.0% (84.85%) (Table 4).

Mutual comparison of individual firsttrimester predictive models
Only one of six joint miRNAs (miR-181a-5p) was dysregulated at the early gestational stages in pregnancies developing GH.MiR-181a-5p was upregulated in 22.89% of cases with 10.0% FPR (Hromadnikova et al., 2022a).A predictive model based on a  (Hromadnikova et al., 2024).A more advanced GH predictive model based on the combination of the first-trimester expression profile of miR-181a-5p and seven maternal clinical characteristics (adding the results gained from the first-trimester screening for PE and/or FGR and spontaneous preterm birth, both using the FMF algorithm) slightly increased the detection rate to 69.88% cases at 10.0% FPR (Hromadnikova et al., 2024).The predictive power for GH can only be improved using this approach.
The HELLP syndrome predictive model previously demonstrated by our group based on the combination of six miRNAs (AUC 0.903, p < 0.001) (Hromadnikova et al., 2023a;Hromadnikova, 2022f) reached a detection rate of 78.57% at 10.0% FPR.When this model was expanded for the same selected maternal clinical characteristics representing risk factors for HELLP syndrome, the predictive power significantly increased: six miRNAs +6 clinical variables (85.71% cases at 10.0% FPR, AUC 0.979, p < 0.001) and six miRNAs +7 clinical variables (92.86% cases at 10.0% FPR, AUC 0.975, p < 0.001) (Hromadnikova et al., 2023a;Hromadnikova, 2022f).The HELLP syndrome predictive model based on three out of six miRNAs common to adverse pregnancy outcomes and the same maternal clinical characteristics (six variables or seven variables) reached similar detection power (85.71% cases at 10.0% FPR, AUC 0.970, p < 0.001; 92.86% cases at 10.0% FPR, AUC 0.969, p < 0.001) as similar models with six miRNA biomarkers and may be considered the most cost-effective first-trimester predictive model for HELLP syndrome.
nine variables) reached a slightly lower detection power (58.54% cases at 10.0% FPR, AUC 0.815, p < 0.001; 73.17% cases at 10.0% FPR, AUC 0.860, p < 0.001) than similar models with a higher number of miRNA biomarkers.However, it may still be considered the most cost-effective first-trimester predictive models for FGR, irrespective of disease severity and time of disease onset.
Similarly, the most cost-effective first-trimester predictive model for SGA, which had already been presented, is based on the combination of four out of six miRNAs common to adverse pregnancy outcomes and five maternal clinical characteristics (81.08% cases at 10.0% FPR, AUC 0.922, p < 0.001) (Hromadnikova et al., 2023b).Another SGA predictive model containing eight miRNAs and five maternal clinical characteristics showed a slightly higher detection rate (89.19% cases at 10.0% FPR, AUC 0.956, p < 0.001) (Hromadnikova et al., 2023b).The combination of only four miRNAs (75.68% cases at 10.0% FPR, AUC 0.868, p < 0.001) (Hromadnikova et al., 2022b) or the combination of only eight miRNAs (83.78% cases at 10.0% FPR, AUC 0.926, p < 0.001) (Hromadnikova, 2023a) substantially impacted the SGA detection rate.The implementation of maternal clinical variables slightly increased the SGA detection rate.
A previously demonstrated predictive model for GDM requiring the administration of appropriate therapy by our group based on the combination of only three miRNAs (AUC 0.731, p < 0.001) (Hromadnikova et al., 2022d;Hromadnikova 2023b) reached a detection rate of 30.0%cases at 10.0% FPR.When this model was extended to the same selected maternal clinical characteristics representing risk factors for GDM, the predictive power was significantly increased: 3 miRNAs +3 clinical variables (78.95% cases at 10.0% FPR, AUC 0.949, p < 0.001) and 3 miRNAs +7 clinical variables (89.47% cases at 10.0% FPR, AUC 0.957, p < 0.001) (Hromadnikova et al., 2022d).The predictive model for GDM requiring administration of appropriate therapy based on 1 out of six miRNAs common to adverse pregnancy outcomes and the same maternal clinical characteristics (3 variables or seven variables) reached the same detection power (78.95% cases at 10.0% FPR, AUC 0.949, p < 0.001; 89.47% cases at 10.0% FPR, AUC 0.957, p < 0.001) as the similar models with a higher number of miRNA biomarkers and may be considered as the most costeffective first-trimester predictive model for GDM requiring administration of appropriate therapy.
A previously demonstrated predictive model for late miscarriage by our group, based on the combination of only six miRNAs (AUC 0.941, p < 0.001) (Hromadnikova et al., 2023c), reached a detection rate of 79.41% at 10.0% FPR.Four of these miRNAs, dysregulated at early gestational stages in pregnancies affected by late miscarriage, were common to adverse pregnancy outcomes.The combination of only these four miRNAs was insufficient to predict the occurrence of late miscarriage (52.94% cases at 10.0% FPR, AUC 0.828, p < 0.001).The predictive model based on four miRNAs common to adverse pregnancy outcomes was further expanded to include maternal clinical characteristics (maternal age and BMI at early gestational stages, confirmed diagnosis of autoimmune disease, infertility treatment using assisted reproductive technology, presence of non-autoimmune hypothyroidism, presence of uterine fibroids or abnormal-shaped womb, history of miscarriage(s) in previous gestation(s), and presence of thrombophilia gene mutations) to increase the detection power of late miscarriage.Since the predictive power for late miscarriage significantly increased, this model can also be utilized as a cost-effective model (84.85% cases at 10.0% FPR, AUC 0.936, p < 0.001).Alternatively, this model may be extended to the results of first-trimester screening for PE and/or FGR using the FMF algorithm; however, the detection rate of pregnancies with late miscarriage remained the same as that of the model without this variable (84.85% cases at 10.0% FPR, AUC 0.935, p < 0.001).
Predictive models based on the combinations of only two miRNAs common to adverse pregnancy outcomes (91.67% cases at 10.0% FPR, AUC 0.951, p < 0.001) (Hromadnikova, 2022f;Hromadnikova et al., 2023a) or six miRNAs commonly associated with adverse pregnancy outcomes (91.67% cases at 10.0% FPR, AUC 0.967, p < 0.001) were sufficient to predict the later occurrence of stillbirth cost-effectively.Maternal clinical characteristics were not included in the stillbirth predictive models.A previously introduced predictive model for stillbirth based on a combination of 11 dysregulated miRNAs at the early gestational stages achieved a slightly higher detection power (95.83% cases at 10.0% FPR, AUC 0.986, p < 0.001) (Hromadnikova, 2022f;Hromadnikova et al., 2023a).
Previously demonstrated predictive models for preterm delivery (PPROM or PTB) in the absence of other pregnancy-related complications by our group, based on the combinations of six miRNAs (AUC 0.812, p < 0.001) or 12 miRNAs (AUC 0.818, p < 0.001) (Hromadnikova et al., 2022c;Hromadnikova, 2023a), reached a detection rate of 52.83% at 10.0% FPR.Extension of the models based on miRNA expression profiles for the same selected maternal clinical characteristics representing risk factors for preterm delivery in the absence of other pregnancy-related complications increased the predictive power significantly: six miRNAs +5 clinical variables (69.81% cases at 10.0% FPR, AUC 0.874, p < 0.001), MiRNA-target interactions-Common targets of gestational diabetes mellitus, cardiovascular risk factors, cardiovascular and cerebrovascular diseases.Search for interactions between miR-20a-5p and genes using the miRWalk database and disease ontology module revealed several common targets associated with gestational diabetes mellitus, cardiovascular risk factors, cardiovascular and cerebrovascular diseases.

Analysis of MiRNA-target interactions
Numerous predicted and/or validated targets of miRNAs that predict the occurrence of PE have been associated with cardiovascular risk factors and cardiovascular and cerebrovascular diseases (Figures 1A, B).In case of HELLP syndrome, only one common target (CD40LG, the gene encoding the CD40 ligand) associated with cardiovascular risk factors and cardiovascular diseases was identified (Figure 2).Placental insufficiency, usually manifested clinically as preeclampsia and/or fetal growth restriction, has several common miRNA targets associated with cardiovascular risk factors and cardiovascular and cerebrovascular diseases (Figure 3).MiR-20a-5p, a biomarker used solely to predict the occurrence of GDM requiring appropriate therapy, also showed several common targets associated with cardiovascular risk factors and cardiovascular and cerebrovascular diseases (Figure 4).

Discussion
Currently, no first-trimester predictive algorithm for GH, HELLP syndrome, SGA, GDM, late miscarriage, and stillbirth is available.Novel efficient cost-effective modalities for predicting these pregnancy-related complications at the early gestational stages have been proposed.The proposed approach is based on the combinations of selected maternal clinical characteristics and a minimum number of miRNA biomarkers, which play key roles in cardiovascular system maintenance and control and pathogenesis of cardiovascular diseases and whose altered expression was also observed at early gestational stages in pregnancies with adverse outcomes.
At present, the first-trimester algorithm used by the majority of fetal medicine centres developed by the Fetal Medicine Foundation (FMF) calculates the risks for the development of early PE (before 34 gestational weeks) and FGR (before 37 gestational weeks).The risks are calculated on the basis of knowledge of maternal history, BMI, mean arterial blood pressure (MAP), serum levels of pregnancyassociated plasma protein-A (PAPP-A) and placental growth factor (PIGF), and mean uterine artery pulsatility index (UtA-PI) (O´Gorman et al., 2016;O´Gorman et al., 2017;The Fetal Medicine Foundation, 2023;Tan et al., 2018;Mazer Zumaeta et al., 2020).Using the predictive models based on six miRNA biomarkers and selected maternal clinical characteristics, the detection rate of PE increased 2.50 times and the detection rate of FGR 2.61 times when compared with the first-trimester screening for PE and/or FGR using the FMF algorithm.Moreover, using the proposed approach any subtype of PE and FGR regardless of the severity of the disease (mild and severe PE) and time of disease onset can be detected.
In addition, we demonstrated that numerous predicted and/or validated targets of miRNAs used to predict the occurrence of pregnancy-related complications in the first trimester of gestation were associated with several cardiovascular risk factors and cardiovascular and cerebrovascular diseases.
Based on this evidence, we suggest initiating preventive programs for pregnancies at risk of developing pregnancy-related complications as early as possible with the aim of lowering cardiovascular risk and the consequent development of metabolic, cardiovascular, and cerebrovascular diseases.The dysregulation of miRNAs involving in cardiovascular system maintenance and control may still be reversible via the timely implementation of beneficial lifestyle strategies.
Consecutive large-scale retrospective and prospective analyses are needed to verify the reliability of predictive models based on the combinations of the minimum number of miRNA biomarkers common to adverse pregnancy outcomes and maternal clinical characteristics to differentiate between pregnancies with normal and abnormal courses of gestation at early gestational stages.Gynecologists and obstetricians could have a feasible, cost-effective way of identifying pregnancies at risk of adverse pregnancy outcomes at disposal at early gestational stages if satisfactory discrimination power could be achieved.
The dysregulated miRNAs associated with cardiovascular system maintenance and control may be reversed back to normal via the timely implementation of beneficial lifestyle strategies, which may reduce or delay potential cardiovascular risk in mothers.

TABLE 2
Maternal clinical characteristics representing risk factors for adverse pregnancy outcomes involved in first-trimester predictive models.

TABLE 3
Characteristics of selected MiRNAs.

TABLE 4
Predictive models for adverse pregnancy outcomes based on the combinations of MiRNA biomarkers with jointly altered expression during early gestational stages and maternal clinical variables representing risk factors for adverse pregnancy outcomes.

TABLE 4 (
Continued) Predictive models for adverse pregnancy outcomes based on the combinations of MiRNA biomarkers with jointly altered expression during early gestational stages and maternal clinical variables representing risk factors for adverse pregnancy outcomes.

TABLE 4 (
Continued) Predictive models for adverse pregnancy outcomes based on the combinations of MiRNA biomarkers with jointly altered expression during early gestational stages and maternal clinical variables representing risk factors for adverse pregnancy outcomes.HELLP, haemolysis, elevated liver enzymes and low platelets syndrome; FGR, fetal growth restriction; SGA, small-for-gestational-age; GDM, gestational diabetes mellitus; PTB, spontaneous preterm birth; PPROM, preterm prelabor rupture of membranes.