Predictive factors for preeclampsia in pregnant women: a Receiver Operation Character approach

Machine Learning Algorithms Versus Classical Regression Models in Pre-Eclampsia Prediction: A Systematic Review

PURPOSE OF REVIEW

Machine learning (ML) approaches are an emerging alternative for healthcare risk prediction. We aimed to synthesise the literature on ML and classical regression studies exploring potential prognostic factors and to compare prediction performance for pre-eclampsia.

RECENT FINDINGS

From 9382 studies retrieved, 82 were included. Sixty-six publications exclusively reported eighty-four classical regression models to predict variable timing of onset of pre-eclampsia. Another six publications reported purely ML algorithms, whilst another 10 publications reported ML algorithms and classical regression models in the same sample with 8 of 10 findings that ML algorithms outperformed classical regression models. The most frequent prognostic factors were age, pre-pregnancy body mass index, chronic medical conditions, parity, prior history of pre-eclampsia, mean arterial pressure, uterine artery pulsatility index, placental growth factor, and pregnancy-associated plasma protein A. Top performing ML algorithms were random forest (area under the curve (AUC) = 0.94, 95% confidence interval (CI) 0.91-0.96) and extreme gradient boosting (AUC = 0.92, 95% CI 0.90-0.94). The competing risk model had similar performance (AUC = 0.92, 95% CI 0.91-0.92) compared with a neural network. Calibration performance was not reported in the majority of publications. ML algorithms had better performance compared to classical regression models in pre-eclampsia prediction. Random forest and boosting-type algorithms had the best prediction performance. Further research should focus on comparing ML algorithms to classical regression models using the same samples and evaluation metrics to gain insight into their performance. External validation of ML algorithms is warranted to gain insights into their generalisability.

Risk factors and prediction model for new-onset hypertensive disorders of pregnancy: a retrospective cohort study

Background and aims

Hypertensive disorders of pregnancy (HDP) is a significant cause of maternal and neonatal mortality. This study aims to identify risk factors for new-onset HDP and to develop a prediction model for assessing the risk of new-onset hypertension during pregnancy.

Methods

We included 446 pregnant women without baseline hypertension from Liyang People's Hospital at the first inspection, and they were followed up until delivery. We collected maternal clinical parameters and biomarkers between 16th and 20th weeks of gestation. Logistic regression was used to determine the effect of the risk factors on HDP. For model development, a backward selection algorithm was applied to choose pertinent biomarkers, and predictive models were created based on multiple machine learning methods (generalised linear model, multivariate adaptive regression splines, random forest, and k-nearest neighbours). Model performance was evaluated using the area under the curve.

Results

Out of the 446 participants, 153 developed new-onset HDP. The HDP group exhibited significantly higher baseline body mass index (BMI), weight change, baseline systolic/diastolic blood pressure, and platelet counts than the control group. The increase in baseline BMI, weight change, and baseline systolic and diastolic blood pressure significantly elevated the risk of HDP, with odds ratios and 95% confidence intervals of 1.10 (1.03-1.17), 1.10 (1.05-1.16), 1.04 (1.01-1.08), and 1.10 (1.05-1.14) respectively. Restricted cubic spline showed a linear dose-dependent association of baseline BMI and weight change with the risk of HDP. The random forest-based prediction model showed robust performance with the area under the curve of 0.85 in the training set.

Conclusion

This study establishes a prediction model to evaluate the risk of new-onset HDP, which might facilitate the early diagnosis and management of HDP.

Externally validated prediction models for pre-eclampsia: systematic review and meta-analysis

OBJECTIVE

This systematic review and meta-analysis aimed to evaluate the performance of existing externally validated prediction models for pre-eclampsia (PE) (specifically, any-onset, early-onset, late-onset and preterm PE).

METHODS

A systematic search was conducted in five databases (MEDLINE, EMBASE, Emcare, CINAHL and Maternity & Infant Care Database) and using Google Scholar/reference search to identify studies based on the Population, Index prediction model, Comparator, Outcome, Timing and Setting (PICOTS) approach until 20 May 2023. We extracted data using the CHARMS checklist and appraised the risk of bias using the PROBAST tool. A meta-analysis of discrimination and calibration performance was conducted when appropriate.

RESULTS

Twenty-three studies reported 52 externally validated prediction models for PE (one preterm, 20 any-onset, 17 early-onset and 14 late-onset PE models). No model had the same set of predictors. Fifteen any-onset PE models were validated externally once, two were validated twice and three were validated three times, while the Fetal Medicine Foundation (FMF) competing-risks model for preterm PE prediction was validated widely in 16 different settings. The most common predictors were maternal characteristics (prepregnancy body mass index, prior PE, family history of PE, chronic medical conditions and ethnicity) and biomarkers (uterine artery pulsatility index and pregnancy-associated plasma protein-A). The FMF model for preterm PE (triple test plus maternal factors) had the best performance, with a pooled area under the receiver-operating-characteristics curve (AUC) of 0.90 (95% prediction interval (PI), 0.76-0.96), and was well calibrated. The other models generally had poor-to-good discrimination performance (median AUC, 0.66 (range, 0.53-0.77)) and were overfitted on external validation. Apart from the FMF model, only two models that were validated multiple times for any-onset PE prediction, which were based on maternal characteristics only, produced reasonable pooled AUCs of 0.71 (95% PI, 0.66-0.76) and 0.73 (95% PI, 0.55-0.86).

CONCLUSIONS

Existing externally validated prediction models for any-, early- and late-onset PE have limited discrimination and calibration performance, and include inconsistent input variables. The triple-test FMF model had outstanding discrimination performance in predicting preterm PE in numerous settings, but the inclusion of specialized biomarkers may limit feasibility and implementation outside of high-resource settings. © 2023 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Risk factors associated with attendance at postpartum blood pressure follow-up visit in discharged patients with hypertensive disorders of pregnancy

BACKGROUND

This study aims to investigate the risk factors for not returning to postpartum blood pressure (BP) follow-up visit at different time points in postpartum discharged hypertensive disorders of pregnancy (HDP) patients. Likewise, females with HDP in China should have a BP evaluation continuously for at least 42 days postpartum and have BP, urine routine, and lipid and glucose screening for 3 months postpartum.

METHODS

This study is a prospective cohort study of postpartum discharged HDP patients. Telephone follow-up was conducted at 6 weeks and 12 weeks postpartum, the maternal demographic characteristics, details of labor and delivery, laboratory test results of patients at admission, and adherence to BP follow-up visits postpartum were collected. While logistic regression analysis was used to analyze the factors associated with not returning to postpartum BP follow-up visit at 6 weeks and 12 weeks after delivery, the receiver operating characteristic (ROC) curve was drawn to evaluate the model's predictive value for predicting not returning to postpartum BP visit at each follow-up time point.

RESULTS

In this study, 272 females met the inclusion criteria. 66 (24.26%) and 137 (50.37%) patients did not return for postpartum BP visit at 6 and 12 weeks after delivery. A multivariate logistic regression analysis identified education level of high school or below (OR = 3.71; 95% CI = 2.01-6.85; p = 0.000), maximum diastolic BP during pregnancy (OR = 0.97; 95% CI = 0.94-0.99; p = 0.0230)and delivery gestational age (OR = 1.12; 95% CI = 1.005-1.244; p = 0.040)as independent risk factors in predicting not returning to postpartum BP follow-up visit at 6 weeks postpartum, and education level of high school or below (OR = 3.20; 95% CI = 1.805-5.67; p = 0.000), maximum diastolic BP during pregnancy (OR = 0.95; 95% CI = 0.92-0.97; p = 0.000), delivery gestational age (OR = 1.13; 95% CI = 1.04-1.24; p = 0.006) and parity (OR = 1.63; 95% CI = 1.06-2.51; p = 0.026) as risk factors for not returning to postpartum BP follow-up visit at 12 weeks postpartum. The ROC curve analysis indicated that the logistic regression models had a significant predictive value for identify not returning to BP follow-up visit at 6 and 12 weeks postpartum with the area under the curve (AUC) 0.746 and 0.761, respectively.

CONCLUSION

Attendance at postpartum BP follow-up visit declined with time for postpartum HDP patients after discharge. Education at or below high school, maximum diastolic BP during pregnancy and gestational age at delivery were the common risk factors for not returning for BP follow-up visit at 6 and 12 weeks postpartum in postpartum HDP patients.

First trimester preeclampsia screening and prediction

Preeclampsia is a major cause of maternal and perinatal morbidity and mortality. Early-onset disease requiring preterm delivery is associated with a higher risk of complications in both mothers and babies. Evidence suggests that the administration of low-dose aspirin initiated before 16 weeks' gestation significantly reduces the rate of preterm preeclampsia. Therefore, it is important to identify pregnant women at risk of developing preeclampsia during the first trimester of pregnancy, thus allowing timely therapeutic intervention. Several professional organizations such as the American College of Obstetricians and Gynecologists (ACOG) and National Institute for Health and Care Excellence (NICE) have proposed screening for preeclampsia based on maternal risk factors. The approach recommended by ACOG and NICE essentially treats each risk factor as a separate screening test with additive detection rate and screen-positive rate. Evidence has shown that preeclampsia screening based on the NICE and ACOG approach has suboptimal performance, as the NICE recommendation only achieves detection rates of 41% and 34%, with a 10% false-positive rate, for preterm and term preeclampsia, respectively. Screening based on the 2013 ACOG recommendation can only achieve detection rates of 5% and 2% for preterm and term preeclampsia, respectively, with a 0.2% false-positive rate. Various first trimester prediction models have been developed. Most of them have not undergone or failed external validation. However, it is worthy of note that the Fetal Medicine Foundation (FMF) first trimester prediction model (namely the triple test), which consists of a combination of maternal factors and measurements of mean arterial pressure, uterine artery pulsatility index, and serum placental growth factor, has undergone successful internal and external validation. The FMF triple test has detection rates of 90% and 75% for the prediction of early and preterm preeclampsia, respectively, with a 10% false-positive rate. Such performance of screening is superior to that of the traditional method by maternal risk factors alone. The use of the FMF prediction model, followed by the administration of low-dose aspirin, has been shown to reduce the rate of preterm preeclampsia by 62%. The number needed to screen to prevent 1 case of preterm preeclampsia by the FMF triple test is 250. The key to maintaining optimal screening performance is to establish standardized protocols for biomarker measurements and regular biomarker quality assessment, as inaccurate measurement can affect screening performance. Tools frequently used to assess quality control include the cumulative sum and target plot. Cumulative sum is a sensitive method to detect small shifts over time, and point of shift can be easily identified. Target plot is a tool to evaluate deviation from the expected multiple of median and the expected median of standard deviation. Target plot is easy to interpret and visualize. However, it is insensitive to detecting small deviations. Adherence to well-defined protocols for the measurements of mean arterial pressure, uterine artery pulsatility index, and placental growth factor is required. This article summarizes the existing literature on the different methods, recommendations by professional organizations, quality assessment of different components of risk assessment, and clinical implementation of the first trimester screening for preeclampsia.

Predictors of Postpartum Persisting Hypertension Among Women with Preeclampsia Admitted at Carlos Manuel de Cèspedes Teaching Hospital, Cuba

PURPOSE

We established the prevalence and predictors of persisting hypertension in women with preeclampsia admitted at the Carlos Manuel de Cèspedes Teaching Hospital in Cuba so as to guide the health-care providers in early identification of the patients at risk for timely intervention.

PATIENTS AND METHODS

A three-year prospective cohort study was conducted between March 2017 and March 2020. A cohort of 178 women diagnosed with preeclampsia at the hypertension unit of Carlos Manuel de Cèspedes Teaching Hospital were recruited. Interviewer administered questionnaires and laboratory and ultrasound scan result forms were used to collect the data. Binary logistic regression was conducted to determine the predictors. All data analyses were conducted using STATA version 14.2.

RESULTS

Forty-five (27.8%) of the studied 162 patients were still hypertensive at 12 weeks postpartum. Maternal age of 35 years or more (aRR=1.14,95% CI:1.131-4.847, p=0.022), early onset preeclampsia (before 34 weeks of gestation) (aRR=7.93, 95% CI:1.812-34.684, p=0.006), and elevated serum creatinine levels of more than 0.8mg/dl (aRR=1.35, 95% CI:1.241-3.606, p=0.032) were the independent predictors of persisting hypertension at 12 weeks postpartum.

CONCLUSION

Recognition of these predictors and close follow-up of patients with preeclampsia will improve the ability to diagnose and monitor women likely to develop persisting hypertension before its onset for timely interventions.

A risk factor-based predictive model for new-onset hypertension during pregnancy in Chinese Han women

BACKGROUND

Hypertensive disorders of pregnancy (HDP) is one of the leading causes of maternal and neonatal mortality, increasing the long-term incidence of cardiovascular diseases. Preeclampsia and gestational hypertension are the major components of HDP. The aim of our study is to establish a prediction model for pregnant women with new-onset hypertension during pregnancy (increased blood pressure after gestational age > 20 weeks), thus to guide the clinical prediction and treatment of de novo hypertension.

METHODS

A total of 117 pregnant women with de novo hypertension who were admitted to our hospital's obstetrics department were selected as the case group and 199 healthy pregnant women were selected as the control group from January 2017 to June 2018. Maternal clinical parameters such as age, family history and the biomarkers such as homocysteine, cystatin C, uric acid, total bile acid and glomerular filtration rate were collected at a mean gestational age in 16 to 20 weeks. The prediction model was established by logistic regression.

RESULTS

Eleven indicators have statistically significant difference between two groups (P < 0.05). These 11 factors were substituted into the logistic regression equation and 7 independent predictors were obtained. The equation expressed including 7 factors. The calculated area under the curve was 0.884(95% confidence interval: 0.848-0.921), the sensitivity and specificity were 88.0 and 75.0%. A scoring system was established to classify pregnant women with scores ≤15.5 as low-risk pregnancy group and those with scores > 15.5 as high-risk pregnancy group.

CONCLUSIONS

Our regression equation provides a feasible and reliable means of predicting de novo hypertension after pregnancy. Risk stratification of new-onset hypertension was performed to early treatment interventions in high-risk populations.

The relationship between body mass index and preeclampsia: A systematic review and meta-analysis

Background

One of the causes of maternal and fetal mortality and morbidity is pregnancy-induced hypertension, the most common form of which is preeclampsia that causes many complications for mother and fetus.

Objective

The aim of this systematic review and meta-analysis was to determine the relationship between body mass index (BMI) and preeclampsia in Iran.

Materials and Methods

Using valid keywords in the SID database, PubMed, Scopus, data obtained from all the articles, which were reviewed in Iran between 2000 and 2016, were combined using the meta-analysis method (random-effects model) and analyzed using STATA version 11.1.

Results

A total number of 5,946 samples were enrolled in 16 studies with the mean BMI values of 25.13, 27.42, and 26.33 kg /m2 in the healthy, mild, and severe preeclamptic groups, respectively.

Conclusion

The results of this study revealed that there is a significant relationship between BMI and the risk of preeclampsia, so it can be said that BMI may be one of the ways to diagnose preeclampsia.

The Maternal Near Miss Incidence Ratio with WHO Approach in Iran: A Systematic Review and Meta-Analysis

Background:

Maternal near miss (MNM) is one of the important criteria for checking the quality of care in maternal health. This systematic review and meta-analysis study was conducted in 2017 to evaluate the incidence ratio of MNM using the World Health Organization approach in Iran.

Materials and Methods:

This study was designed based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist for systematic reviews, and Web of Science and PubMed databases were searched systematically, which, respectively, yielded 171 and 137 papers published before June 9, 2017. To include papers written in Persian by Iranian scholars, Google Scholar database was searched and 542 papers were retrieved. Finally, 12 papers which had covered the topic more appropriately were included in the study. Random-effects meta-analysis was used to pool the incidence ratio. Heterogeneity was explored using formal tests and subgroup analyses, then the study quality was also explored.

Results:

The pooling of overall potentially life-threatening conditions ratio was I2 (97.60%, p < 0.001, ratio = 2.50/1000 live births [LBs] [95% CI: 2.00-3.00]), which is divided into two indicators: severe complication ratio (2.40/1000 LBs) and critical intervention ratio (2.54/1000 LBs). The pooling of overall life-threatening conditions ratio was I2 (95.10%, p < 0.001, ratio = 0.86/1000 LBs [95% CI: 0.64-1.07]).

Conclusions:

The incidence ratio of MNM needs more attention in Iran. Therefore, it is necessary to identify the factors related to MNM and then implement suitable strategies to reduce the risk factors of the maternal morbidity and improve the quality of maternal care in facilities.

External Validation and Clinical Usefulness of First Trimester Prediction Models for the Risk of Preeclampsia: A Prospective Cohort Study

INTRODUCTION

This study assessed the external validity of all published first trimester prediction models for the risk of preeclampsia (PE) based on routinely collected maternal predictors. Moreover, the potential utility of the best-performing models in clinical practice was evaluated.

MATERIAL AND METHODS

Ten prediction models were systematically selected from the literature. We performed a multicenter prospective cohort study in the Netherlands between July 1, 2013, and December 31, 2015. Eligible pregnant women completed a web-based questionnaire before 16 weeks' gestation. The outcome PE was established using postpartum questionnaires and medical records. Predictive performance of each model was assessed by means of discrimination (c-statistic) and a calibration plot. Clinical usefulness was evaluated by means of decision curve analysis and by calculating the potential impact at different risk thresholds.

RESULTS

The validation cohort contained 2,614 women of whom 76 developed PE (2.9%). Five models showed moderate discriminative performance with c-statistics ranging from 0.73 to 0.77. Adequate calibration was obtained after refitting. The best models were clinically useful over a small range of predicted probabilities.

DISCUSSION

Five of the ten included first trimester prediction models for PE showed moderate predictive performance. The best models may provide more benefit compared to risk selection as used in current guidelines.

Renin-angiotensin-aldosterone system gene polymorphisms in gestational hypertension and preeclampsia: A case-control gene-association study

Pregnancy-induced hypertension (PIH, including preeclampsia [PE] and gestational hypertension [GH]) and cardiovascular diseases (CVDs) have some metabolic changes and risk factors in common. Many studies have reported associations between single nucleotide polymorphisms (SNPs) of renin-angiotensin-aldosterone system (RAAS) genes and CVDs (particularly hypertension), and their findings have provided candidate SNPs for research on genetic correlates of PIH. We explored the association between hypertension-related RAAS SNPs and PIH in a Chinese population. A total of 130 cases with PE, 67 cases with GH, and 316 controls were recruited. Six candidate SNPs of the RAAS system were selected. Multiple logistic regression analysis adjusting for maternal age, fetal sex, and gestational diabetes mellitus showed significant associations between angiotensinogen (AGT) rs3789678 T/C and GH (p = 0.0088) and between angiotensin II receptor type 1 (AGTR1) rs275645 G/A and PE (p = 0.0082). The study population was further stratified by maternal age (<30 and ≥30 years), and stratified and crossover analyses were conducted to determine genetic associations in different age groups. Our findings suggest that the impacts of different SNPs might be affected by maternal age; however, the effect of this potential gene-age interaction on PIH needs further exploration.

Prevalence and Predictive Factors of Sexual Dysfunction in Iranian Women: Univariate and Multivariate Logistic Regression Analyses

BACKGROUND

Female sexual dysfunction, which can occur during any stage of a normal sexual activity, is a serious condition for individuals and couples. The present study aimed to determine the prevalence and predictive factors of female sexual dysfunction in women referred to health centers in Ilam, the Western Iran, in 2014.

METHODS

In the present cross-sectional study, 444 women who attended health centers in Ilam were enrolled from May to September 2014. Participants were selected according to the simple random sampling method. Univariate and multivariate logistic regression analyses were used to predict the risk factors of female sexual dysfunction. Diffe rences with an alpha error of 0.05 were regarded as statistically significant.

RESULTS

Overall, 75.9% of the study population exhibited sexual dysfunction. Univariate logistic regression analysis demonstrated that there was a significant association between female sexual dysfunction and age, menarche age, gravidity, parity, and education (P<0.05). Multivariate logistic regression analysis indicated that, menarche age (odds ratio, 1.26), education level (odds ratio, 1.71), and gravida (odds ratio, 1.59) were independent predictive variables for female sexual dysfunction.

CONCLUSION

The majority of Iranian women suffer from sexual dysfunction. A lack of awareness of Iranian women's sexual pleasure and formal training on sexual function and its influencing factors, such as menarche age, gravida, and level of education, may lead to a high prevalence of female sexual dysfunction.

Inadvertent use of bevacizumab in pregnant women with diabetes mellitus type 1

BACKGROUND

The use of vascular endothelial growth factor (VEGF) inhibitors may cause fetal harm and systemic side effects in the mother, so these drugs are contraindicated in pregnancy. We report a case of inadvertent administration of two intravitreal bevacizumab injections in a woman with diabetes mellitus type 1, 5 days before ovulation (±3 days) and during the 5th gestational week, respectively. The patient had a past history of both miscarriage and requirement for cesarean section for preeclampsia.

METHODS

Observational case report.

RESULTS

The patient did not have any drug-related adverse event and delivered a healthy full-term infant, reaching all developmental milestones appropriately during infancy.

CONCLUSIONS

Intravitreal drug injections did not result in any detectable adverse event in the mother and infant although she had a significant past obstetric history. However, there have been no studies evaluating the effects of bevacizumab in pregnant women and suggesting that intravitreal drug injection in this patient is safe. Until more is known about this, it seems reasonable to avoid treatment with this drug a few weeks before and during pregnancy.

Predicting reintubation, prolonged mechanical ventilation and death in post-coronary artery bypass graft surgery: a comparison between artificial neural networks and logistic regression models

INTRODUCTION

In coronary artery bypass (CABG) surgery, the common complications are the need for reintubation, prolonged mechanical ventilation (PMV) and death. Thus, a reliable model for the prognostic evaluation of those particular outcomes is a worthwhile pursuit. The existence of such a system would lead to better resource planning, cost reductions and an increased ability to guide preventive strategies. The aim of this study was to compare different methods - logistic regression (LR) and artificial neural networks (ANNs) - in accomplishing this goal.

MATERIAL AND METHODS

Subjects undergoing CABG (n = 1315) were divided into training (n = 1053) and validation (n = 262) groups. The set of independent variables consisted of age, gender, weight, height, body mass index, diabetes, creatinine level, cardiopulmonary bypass, presence of preserved ventricular function, moderate and severe ventricular dysfunction and total number of grafts. The PMV was also an input for the prediction of death. The ability of ANN to discriminate outcomes was assessed using receiver-operating characteristic (ROC) analysis and the results were compared using a multivariate LR.

RESULTS

The ROC curve areas for LR and ANN models, respectively, were: for reintubation 0.62 (CI: 0.50-0.75) and 0.65 (CI: 0.53-0.77); for PMV 0.67 (CI: 0.57-0.78) and 0.72 (CI: 0.64-0.81); and for death 0.86 (CI: 0.79-0.93) and 0.85 (CI: 0.80-0.91). No differences were observed between models.

CONCLUSIONS

The ANN has similar discriminating power in predicting reintubation, PMV and death outcomes. Thus, both models may be applicable as a predictor for these outcomes in subjects undergoing CABG.