Development of a nomogram for individual preterm birth risk evaluation

Development and validation of a nomogram to predict spontaneous preterm birth in singleton gestation with short cervix and no history of spontaneous preterm birth

Background

Spontaneous preterm birth (sPTB) stands as a leading cause of neonatal mortality. Consequently, preventing sPTB has emerged as a paramount concern in healthcare. Therefore, our study aimed to develop a nomogram, encompassing patient characteristics and cervical elastography, to predict sPTB in singleton pregnancies. Specifically, we targeted those with a short cervix length (CL), no history of sPTB, and who were receiving vaginal progesterone therapy.

Methods

A total of 568 patients were included in this study. Data from 392 patients, collected between January 2016 and October 2019, constituted the training cohort. Meanwhile, records from 176 patients, spanning November 2019 to January 2022, formed the validation cohort. Following the univariate logistic regression analysis, variables exhibiting a P-value less than 0.05 were integrated into a multivariable logistic regression analysis. The primary objective of this subsequent analysis was to identify the independent predictors linked to sPTB in the training cohort. Next, we formulated a nomogram utilizing the identified independent predictors. This tool was designed to estimate the likelihood of sPTB in singleton pregnancies, particularly those with a short CL, devoid of any sPTB history, and undergoing vaginal progesterone therapy. The C-index, Hosmer-Lemeshow (HL) test, calibration curves, decision curve analysis (DCA), and receiver operating characteristic (ROC) were used to validate the performance of the nomogram.

Results

Upon finalizing the univariate analysis, we progressed to a multivariable analysis, integrating 8 variables with P < 0.05 from the univariate analysis. The multivariable analysis identified 7 independent risk factors: maternal age (OR = 1.072; P < 0.001), cervical length (OR = 0.854; P < 0.001), uterine curettage (OR = 7.208; P < 0.001), GDM (OR = 3.570; P = 0.006), HDP (OR = 4.661; P = 0.003), C-reactive protein (OR = 1.138; P < 0.001), and strain of AI (OR = 7.985; P < 0.001). The nomogram, tailored for sPTB prediction, was grounded on these 7 independent predictors. In predicting sPTB, the C-indices manifested as 0.873 (95% CI, 0.827-0.918) for the training cohort and 0.916 (95%CI, 0.870-0.962) for the validation cohorts, underscoring a good discrimination of the model. Additionally, the ROC curves served to evaluate the discrimination of nomogram model across both cohorts. Calibration curves were delineated, revealing no statistically significant differences in both the training (χ2 = 5.355; P = 0.719) and validation (χ2 = 2.708; P = 0.951) cohorts as evidenced by the HL tests. Furthermore, the DCA underscored the model's excellence as a predictive tool for sPTB.

Conclusions

By amalgamating patient characteristics and cervical elastography data from the second trimester, the nomogram emerged as a visually intuitive and dependable tool for predicting sPTB. Its relevance was particularly pronounced for singleton pregnancies characterized by a short CL, an absence of prior sPTB incidents, and those receiving vaginal progesterone therapy.

A systematic review of prediction models for spontaneous preterm birth in singleton asymptomatic pregnant women with risk factors

Backgrounds

Spontaneous preterm birth (SPB) is a global problem. Early screening, identification, and prevention in asymptomatic pregnant women with risk factors for preterm birth can help reduce the incidence and mortality of preterm births. Therefore, this study systematically reviewed prediction models for spontaneous preterm birth, summarised the model characteristics, and appraised their quality to identify the best-performing prediction model for clinical decision-making.

Methods

PubMed, Embase, Cochrane Library, China National Knowledge Infrastructure, China Biology Medicine disc, VIP Database, and Wanfang Data were searched up to September 27, 2021. Prediction models for spontaneous preterm births in singleton asymptomatic pregnant women with risk factors were eligible for inclusion. Six independent reviewers selected the eligible studies and extracted data from the prediction models. The findings were summarised using descriptive statistics and visual plots.

Results

Twelve studies with twelve developmental models were included. Discriminative performance was reported in 11 studies, with an Area Under the Curve (AUC) ranging from 0.75 to 0.95. The AUCs of the seven models were greater than 0.85. Cervical length (CL) is the most commonly used predictor of spontaneous preterm birth. A total of 91.7% of the studies had a high risk of bias in the analysis domain, mainly because of the small sample size and lack of adjustment for overfitting.

Conclusion

The accuracy of the models for spontaneous preterm births in singleton asymptomatic women with risk factors was good. However, these models are not widely used in clinical practice because they lack replicability and transparency. Future studies should transparently report methodological details and consider more meaningful predictors with new progress in research on preterm birth.

Development of prognostic model for preterm birth using machine learning in a population-based cohort of Western Australia births between 1980 and 2015

Preterm birth is a global public health problem with a significant burden on the individuals affected. The study aimed to extend current research on preterm birth prognostic model development by developing and internally validating models using machine learning classification algorithms and population-based routinely collected data in Western Australia. The longitudinal retrospective cohort study involved all births in Western Australia between 1980 and 2015, and the analytic sample contains 81,974 (8.6%) preterm births (< 37 weeks of gestation). Prediction models for preterm birth were developed using regularised logistic regression, decision trees, Random Forests, extreme gradient boosting, and multi-layer perceptron (MLP). Predictors included maternal socio-demographics and medical conditions, current and past pregnancy complications, and family history. Class weight was applied to handle imbalanced outcomes and stratified tenfold cross-validation was used to reduce overfitting. Close to half of the preterm births (49.1% at 5% FPR, 95% CI 48.9%,49.5%) were correctly classified by the best performing classifier (MLP) for all women when current pregnancy information was available. The sensitivity was boosted to 52.7% (95% CI 52.1%,53.3%) after including past obstetric history in a sub-population of births from multiparous women. Around half of the preterm birth can be identified antenatally at high specificity using population-based routinely collected maternal and pregnancy data. The performance of the prediction models depends on the available predictor pool that is individual and time specific.

Validation of a Nomogram for Predicting Preterm Birth in Women With Threatened Preterm Labor: A Prospective Cohort Study in Iranian Tertiary Referral Hospitals

In this prospective cohort study, we aimed to investigate external validity of the Allouche's nomogram to predict preterm birth in symptomatic women in Iran. We employed six variables of cervical length, uterine contractions, rupture of membranes, vaginal bleeding, gestational age, and multiple pregnancy to draw the nomograms. These variables were examined in the first day of women's hospitalization and participants followed up until giving birth. The concordance index of area under the curve (AUC) was used for validation of the nomograms. Of the participants 10% gave birth within 48 hours and 29% before 34 weeks. The nomogram had sufficient accuracy in predicting birth within 48 hours (AUC 0.89 [95% CI 0.82-0.96]) and birth before 34 weeks (AUC 0.89 [95% CI 0.84-0.94]). The optimal risk threshold for nomogram predicting birth within 48 hours was 0.16. Use of these two nomograms, can improve the health of women and their neonates.

Cervical pessary plus vaginal progesterone in a singleton pregnancy with a short cervix: an experience-based analysis of cervical pessary's efficacy

BACKGROUND

Several studies were published about cervical pessary, with controversial results. These studies demonstrated that the patient follow-up after pessary insertion is very different between the study centers and the number of pessary insertions per center was often <30 cases. This study aims to determine cervical pessary performance in singleton pregnancies with a short cervix based on a single center learning curve.

METHODS

Between 2011 and 2018, 128 singleton pregnancies between 18 and 24 gestational weeks with a short cervix (<25 mm) were referred to our clinic. All cases were treated with progesterone, and when available in our supplies (due to low resources) cervical pessary was also offered. Three groups were created for statistical analysis: Group 1 (n = 33), treated with progesterone-only; Groups 2 and 3, treated with cervical pessary plus progesterone. Group 2 included the first cases (n = 30) of pessary, defined by a learning curve and cumulative sum analysis, while Group 3 included the subsequent 65 cases. The primary outcome was preterm birth (PTB) < 34 gestational weeks.

RESULTS

The learning curve was performed with all cases of pessary plus progesterone, and 30 patients were obtained as the number needed for learning, in our study with two operators. The PTB rate < 34 weeks was 27.3, 20, and 4.6% in groups 1, 2, and 3, respectively. There was no significant difference between Group 1 and 2 (OR 1.1; 95% CI 0.066 - 18.45; p = .945). When comparing Groups 1 and 3 there was a significant difference in PTB rates (OR 0.08; CI95% 0.01-0.42; p = .003). Considering Kaplan-Meyer Survival analysis, we can observe that the performance of progesterone alone (Group 1) was similar to Group 2 (progesterone + first 30 cases of pessary) (p = .432), but the performance of Group 3 (progesterone + subsequent 65 cases of pessary) and Group 1 shows a statistically significant difference (p = .011).

CONCLUSION

Learning curve and cumulative sum analysis determined that the application and surveillance of at least 30 patients is required to see significant improvements in the primary outcome of PTB < 34 weeks.

A Novel Nomogram for Predicting the Risk of Premature Delivery Based on the Thyroid Function in Pregnant Women

BACKGROUND

Maternal thyroid dysfunction and autoantibodies were associated with preterm delivery. However, recommendations for cutoff values of thyroperoxidase antibody (TPOAb) positivity and thyroid-stimulating homone (TSH) associated with premature delivery are lacking.

OBJECTIVE

To identify the pregnancy-specific cutoff values for TPOAb positivity and TSH associated with preterm delivery. To develop a nomogram for the risk prediction of premature delivery based on maternal thyroid function in singleton pregnant women without pre-pregnancy complications.

METHODS

This study included data from the International Peace Maternity and Child Care Health Hospital (IPMCH) in Shanghai, China, between January 2013 and December 2016. Added data between September 2019 and November 2019 as the test cohort. Youden's index calculated the pregnancy-specific cutoff values for TPOAb positivity and TSH concentration. Univariate and multivariable logistic regression analysis were used to screen the risk factors of premature delivery. The nomogram was developed according to the regression coefficient of relevant variables. Discrimination and calibration of the model were assessed using the C-index, Hosmer-Lemeshow test, calibration curve and decision curve analysis.

RESULTS

45,467 pregnant women were divided into the training and validation cohorts according to the ratio of 7: 3. The testing cohort included 727 participants. The pregnancy-specific cutoff values associated with the risk of premature delivery during the first trimester were 5.14 IU/mL for TPOAb positivity and 1.33 mU/L for TSH concentration. Multivariable logistic regression analysis showed that maternal age, history of premature delivery, elevated TSH concentration and TPOAb positivity in the early pregnancy, preeclampsia and gestational diabetes mellitus were risk factors of premature delivery. The C-index was 0.62 of the nomogram. Hosmer-Lemeshow test showed that the Chi-square value was 2.64 (P = 0.955 > 0.05). Decision curve analysis showed a positive net benefit. The calibration curves of three cohorts were shown to be in good agreement.

CONCLUSIONS

We identified the pregnancy-specific cutoff values for TPOAb positivity and TSH concentration associated with preterm delivery in singleton pregnant women without pre-pregnancy complications. We developed a nomogram to predict the occurrence of premature delivery based on thyroid function and other risk factors as a clinical decision-making tool.