Machine Learning Prediction Models for Gestational Diabetes Mellitus: Meta-analysis

In the era of precision therapy, early classification of breast cancer (BRCA) molecular subtypes has clinical significance for disease management and prognosis. We explored the accuracy of machine learning (ML) models for early classification of BRCA molecular subtypes through a systematic review of the literature currently available. We retrieved relevant studies published in PubMed, EMBASE, Cochrane, and Web of Science until 15 April 2022. A prediction model risk of bias assessment tool (PROBAST) was applied for the assessment of risk of bias of a genomics-based ML model, and the Radiomics Quality Score (RQS) was simultaneously used to evaluate the quality of this radiomics-based ML model. A random effects model was adopted to analyze the predictive accuracy of genomics-based ML and radiomics-based ML for Luminal A, Luminal B, Basal-like or triple-negative breast cancer (TNBC), and human epidermal growth factor receptor 2 (HER2). The PROSPERO of our study was prospectively registered (CRD42022333611). Of the 38 studies were selected for analysis, 14 ML models were based on gene-transcriptomic, with only 4 external validations; and 43 ML models were based on radiomics, with only 14 external validations. Meta-analysis results showed that c-statistic values of the ML based on radiomics for the identification of BRCA molecular subtypes Luminal A, Luminal B, Basal-like or TNBC, and HER2 were 0.76 [95% confidence interval (CI): 0.60-0.96], 0.78 (95% CI: 0.69-0.87), 0.89 (95% CI: 0.83-0.91), and 0.83 (95% CI: 0.81-0.86), respectively. The c-statistic values of ML based on the gene-transcriptomic analysis cohort for the identification of the previously described BRCA molecular subtypes were 0.96 (95% CI: 0.93-0.99), 0.96 (95% CI: 0.93-0.99), 0.98 (95% CI: 0.95-1.00), and 0.97 (95% CI: 0.96-0.98) respectively. Additionally, the sensitivity of the ML model based on radiomics for each molecular subtype ranged from 0.79 to 0.85, while the sensitivity of the ML model based on gene-transcriptomic was between 0.92 and 0.99. Both radiomics and gene transcriptomics produced ideal effects on BRCA molecular subtype prediction. Compared with radiomics, gene transcriptomics yielded better prediction results, but radiomics was simpler and more convenient from a clinical point of view.

BackgroundHIV treatment interruption remains a significant barrier to achieving global HIV/AIDS control goals. Machine learning (ML) models offer potential for predicting treatment interruption by leveraging large clinical data. Understanding how these models were developed, validated, and applied remains essential for advancing research.MethodsWe searched databases including the PubMed, BMC, Cochrane Library, Scopus, ScienceDirect, Lancet, and Google Scholar, for studies published in English from 1990 to September 2024. Search terms covered HIV, machine learning, treatment interruption, and loss to follow-up. Articles were screened and reviewed independently, and data were extracted using the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) tool. Risk of bias was assessed with Prediction model Risk Of Bias Assessment Tool (PROBAST). The Preferred Reporting Items for Systematic reviews and Meta-analysis (PRISMA) guidelines were followed throughout.ResultsOut of 116,672 records, 9 studies met the inclusion criteria and reported 12 ML models. Random Forest, XGBoost, and AdaBoost were predominant models (91.7%). Internal validation was performed in all models, but only two models included external validation. Performance varied, with a mean area under the receiver operating characteristic curve (AUC-ROC) of 0.668 (standard deviation (SD) = 0.066), indicating moderate discrimination. About 75% of models showed a high risk of bias due to inadequate handling of missing data, lack of calibration, and the absence of decision curve analysis (DCA).ConclusionsML models show promise for predicting HIV treatment interruption, particularly in resource-limited settings. Future research should prioritize external validation, robust missing data handling, and decision curve analysis and include sociocultural predictors to improve model robustness.Systematic review registrationPROSPERO CRD42024578109.Supplementary InformationThe online version contains supplementary material available at 10.1186/s44263-025-00184-4.

12074 Background: Skeletal-related events (SREs) are detrimental clinical events in bone metastasis from advanced cancer, defined by pathologic fracture, spinal cord compression, and inevitable surgical or radiational intervention to the bone. Given their negative impact on quality of life and prognosis and interpatient heterogeneity in the SRE risk, accurate identification of patients with high SRE risk is critical. Methods: The patient-level data from three randomized clinical trials that administered zoledronic acid (ZA) to patients with bone-metastatic breast cancer, castration-resistant prostate cancer (CRPC), and other types of cancer were analyzed. (N = 460, 452, and 315, respectively) Machine learning (ML) models to predict SREs within 18 months (breast cancer), 12 months (CRPC), and 9 months (other cancers) were developed based on more than 40 baseline clinical and laboratory data. Seven ML algorithms and five feature selection methods were utilized to develop multiple models. The ML model with the best performance was identified based on the F1 score and the area under the receiver operating characteristic curve (AUC-ROC) for each cancer type and interrogated for important features with Shapley additive explanations. Lastly, the ML models’ ability to stratify patients by the cumulative SRE risk was evaluated by calculating hazard ratio (HR) with Cox-proportional hazards models. Results: Among the multiple ML models developed with different algorithms and feature selection methods, the model developed utilizing the random forests algorithm and the Boruta method for selecting features demonstrated the best performance in all types of cancer (F1 0.70, 0.67, and 0.67, and AUC-ROC 0.72, 0.68, and 0.73 for breast cancer, CRPC, and other cancers, respectively). In the ML model for breast cancer, performance status (PS), history of SRE, serum alkaline phosphatase (ALP), history of anti-neoplastic surgery, radiation therapy, and pathologic fracture were included as important features. Serum ALP, albumin, sodium, Gleason scores, and geographic regions were shown to be relevant in the CRPC model. For other cancers, serum ALP, albumin, total protein, phosphorus, red blood cell count, white blood cell count, visceral metastases, and history of arthritis were incorporated in the ML model. The ML model prediction successfully stratified the patients for cumulative SRE risk in all three cohorts. (HR with 95% confidence interval: 2.43 [1.86 – 3.18], 1.92 [1.51 – 2.45], and 3.06 [2.29 – 4.09] for breast cancer, CRPC, and other types of cancer, respectively). Conclusions: ML models incorporating baseline clinical and laboratory data can identify patients with bone metastasis on ZA harboring a high SRE risk.

To develop a machine-learning (ML) model for prediction of shoulder dystocia (ShD) and to externally validate the model's predictive accuracy and potential clinical efficacy in optimizing the use of Cesarean delivery in the context of suspected macrosomia. We used electronic health records (EHR) from the Sheba Medical Center in Israel to develop the model (derivation cohort) and EHR from the University of California San Francisco Medical Center to validate the model's accuracy and clinical efficacy (validation cohort). Subsequent to application of inclusion and exclusion criteria, the derivation cohort included 686 singleton vaginal deliveries, of which 131 were complicated by ShD, and the validation cohort included 2584 deliveries, of which 31 were complicated by ShD. For each of these deliveries, we collected maternal and neonatal delivery outcomes coupled with maternal demographics, obstetric clinical data and sonographic fetal biometry. Biometric measurements and their derived estimated fetal weight were adjusted (aEFW) according to gestational age at delivery. A ML pipeline was utilized to develop the model. In the derivation cohort, the ML model provided significantly better prediction than did the current clinical paradigm based on fetal weight and maternal diabetes: using nested cross-validation, the area under the receiver-operating-characteristics curve (AUC) of the model was 0.793 ± 0.041, outperforming aEFW combined with diabetes (AUC = 0.745 ± 0.044, P = 1e-16 ). The following risk modifiers had a positive beta that was > 0.02, i.e. they increased the risk of ShD: aEFW (beta = 0.164), pregestational diabetes (beta = 0.047), prior ShD (beta = 0.04), female fetal sex (beta = 0.04) and adjusted abdominal circumference (beta = 0.03). The following risk modifiers had a negative beta that was < -0.02, i.e. they were protective of ShD: adjusted biparietal diameter (beta = -0.08) and maternal height (beta = -0.03). In the validation cohort, the model outperformed aEFW combined with diabetes (AUC = 0.866 vs 0.784, P = 0.00007). Additionally, in the validation cohort, among the subgroup of 273 women carrying a fetus with aEFW ≥ 4000 g, the aEFW had no predictive power (AUC = 0.548), and the model performed significantly better (0.775, P = 0.0002). A risk-score threshold of 0.5 stratified 42.9% of deliveries to the high-risk group, which included 90.9% of ShD cases and all cases accompanied by maternal or newborn complications. A more specific threshold of 0.7 stratified only 27.5% of the deliveries to the high-risk group, which included 63.6% of ShD cases and all those accompanied by newborn complications. We developed a ML model for prediction of ShD and, in a different cohort, externally validated its performance. The model predicted ShD better than did estimated fetal weight either alone or combined with maternal diabetes, and was able to stratify the risk of ShD and neonatal injury in the context of suspected macrosomia. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.

COPD is a major global health burden, associated with high rates of mortality, readmissions and cardiovascular disease (CVD) complications. Predictive models, including statistical and machine learning (ML) approaches, have been developed to support risk stratification and clinical decision-making. This review assesses their performance and generalisability. A systematic search of EMBASE, MEDLINE and PubMed identified studies published since 2015 evaluating predictive models for COPD-related outcomes. Studies were screened using predefined criteria, and model performance was synthesised via meta-analysis. Pooled area under the curve (AUC) values were calculated for each model type. Risk of bias was assessed using the Prediction model Risk Of Bias ASsessment Tool (PROBAST). Of 3 488 records screened, 37 studies met inclusion criteria: 20 focused on mortality, 14 on readmissions and six on CVD complications. Statistical models had a pooled AUC of 0.787 (95% CI 0.755-0.816). For mortality, statistical models outperformed or matched ML models (AUC 0.801 versus 0.760; p=0.1195), while ML models outperformed statistical approaches for readmissions (AUC 0.812 versus 0.758; p=0.4423). CVD outcomes showed a pooled AUC of 0.810 (95% CI 0.749-0.859). External validation often reduced ML model performance, raising concerns about overfitting. ML models improve readmission prediction but offer no consistent advantage for mortality, where statistical models perform similarly. ML models face generalisability challenges due to overfitting. Future work should emphasise real-world validation and hybrid approaches to enhance interpretability and clinical applicability in COPD care.

We aimed to evaluate the diagnostic performance of neuroimaging-based machine learning (ML) models for non-invasive prediction of H3K27M mutations in diffuse midline gliomas (DMG). Following PRISMA-DTA guidelines, we searched four databases up to May 2025 to identify eligible studies evaluating neuroimaging-based ML for predicting H3K27M mutations in DMG. Study quality was assessed using PROBAST+AI and GRADE. Bivariate random-effects models were used to pool sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Sixteen studies were included, comprising 2,357 patients in internal validation cohorts and 1,792 patients in external validation cohorts. MRI-based ML models showed strong diagnostic performance in internal validation, with pooled sensitivity of 0.86 (95% CI: 0.79-0.91), specificity of 0.82 (95% CI: 0.75-0.87), and AUC of 0.91 (95% CI: 0.88-0.93). For PET/CT-based ML models, pooled sensitivity was 0.58 (95% CI: 0.44-0.71) and pooled specificity was 0.65 (95% CI: 0.46-0.81), with an AUC of 0.61 (95% CI: 0.57-0.66). MRI-based ML showed significantly higher sensitivity (Z = 3.71; P < 0.01) and AUC (Z = 11.42; P < 0.01) than PET/CT-based ML. Subgroup analysis indicated that MRI-based deep learning (DL) models outperformed conventional machine learning (cML) algorithms (P = 0.01). Models using DNA sequencing as the reference standard showed higher specificity than those using immunohistochemistry (P < 0.001). MRI-based ML demonstrates high accuracy and generalizability for non-invasive H3K27M prediction in DMG, seemingly outperforming current PET/CT-based ML. The adoption of DL architectures and DNA sequencing as the reference standard may further improve performance, supporting the clinical utility of MRI-based ML for molecular stratification. https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=1078673, identifier CRD420251078673.

P125. Development of a novel ensemble machine learning algorithm for prediction of complications and readmission after anterior cervical spinal fusion

Introduction: Early and accurate diagnosis of transthyretin amyloid cardiomyopathy (ATTR-CM) is known to improve prognosis. However, it is sometimes challenging to distinguish ATTR-CM from other cardiac conditions that cause left ventricular hypertrophy (LVH), such as hypertrophic cardiomyopathy (HCM), hypertensive LVH (h-LVH), and aortic stenosis (AS). The Mayo ATTR-CM score is commonly used to distinguish ATTR-CM from these mimickers. Herein, we constructed a machine learning (ML) model and compared its performance to the Mayo score. Hypothesis: A ML model based on readily available clinical variables will outperform the Mayo score in distinguishing ATTR-CM from other cardiac conditions that cause LVH. Methods: We performed a multicenter case-control study of 1,605 patients, including 303 ATTR-CM cases and 1,302 controls – i.e., 935 with HCM, 331 with h-LVH, and 36 with AS. We allocated the former two-thirds of each disease group to the training set and the latter one-third to the prospective test set for validation. We developed a main ML model in the training set using XGBoost with 18 clinical variables including demographics, comorbidities, laboratory values, and echocardiographic data. We examined its performance in the test set using the area under the receiver-operating-characteristic curve (AUC) and decision curve analysis (DCA). In addition, we developed a simplified ML model using only the top 5 variables identified by feature importance and tested its performance. Results: The top 5 most important variables were age, LV diameter in systole, LV posterior wall thickness in diastole, LV ejection fraction, and left atrial diameter ( Figure 1 ). The main ML model (AUC: 0.96, 95% CI: 0.93–0.98) outperformed the Mayo score (AUC: 0.84, 95% CI: 0.79–0.89, DeLong’s P &lt;0.0001; Figure 2 ) in the test set. DCA demonstrated that the ML model provided greater net benefit across all threshold probabilities ( Figure 3 ). The simplified 5-variable model also outperformed the Mayo score (AUC: 0.94, 95% CI: 0.91–0.97, DeLong’s P &lt;0.0001). Conclusions: In this multicenter study, we developed ML models that effectively distinguish ATTR-CM from its mimickers that cause LVH using readily available clinical variables. Given the simplicity and accuracy of our models, the present study serves as an important initial step toward the development of a robust, data-driven tool that could be readily automated and integrated into electronic medical record system to screen for ATTR-CM.

PD27-01 DEVELOPMENT OF A MACHINE LEARNING (ML) MODEL TO AUTOMATICALLY AND PRECISELY IDENTIFY KIDNEY STONES FROM URETEROSCOPY VIDEO RECORDINGS

We aimed to identify existing hypertension risk prediction models developed using traditional regression-based or machine learning approaches and compare their predictive performance. We systematically searched MEDLINE, EMBASE, Web of Science, Scopus, and the grey literature for studies predicting the risk of hypertension among the general adult population. Summary statistics from the individual studies were the C-statistic, and a random-effects meta-analysis was used to obtain pooled estimates. The predictive performance of pooled estimates was compared between traditional regression-based models and machine learning-based models. The potential sources of heterogeneity were assessed using meta-regression, and study quality was assessed using the PROBAST (Prediction model Risk Of Bias ASsessment Tool) checklist. Of 14,778 articles, 52 articles were selected for systematic review and 32 for meta-analysis. The overall pooled C-statistics was 0.75 [0.73-0.77] for the traditional regression-based models and 0.76 [0.72-0.79] for the machine learning-based models. High heterogeneity in C-statistic was observed. The age (p = 0.011), and sex (p = 0.044) of the participants and the number of risk factors considered in the model (p = 0.001) were identified as a source of heterogeneity in traditional regression-based models. We attempted to provide a comprehensive evaluation of hypertension risk prediction models. Many models with acceptable-to-good predictive performance were identified. Only a few models were externally validated, and the risk of bias and applicability was a concern in many studies. Overall discrimination was similar between models derived from traditional regression analysis and machine learning methods. More external validation and impact studies to implement the hypertension risk prediction model in clinical practice are required.

Background and purpose — External validation of machine learning (ML) prediction models is an essential step before clinical application. We assessed the proportion, performance, and transparent reporting of externally validated ML prediction models in orthopedic surgery, using the Transparent Reporting for Individual Prognosis or Diagnosis (TRIPOD) guidelines. Material and methods — We performed a systematic search using synonyms for every orthopedic specialty, ML, and external validation. The proportion was determined by using 59 ML prediction models with only internal validation in orthopedic surgical outcome published up until June 18, 2020, previously identified by our group. Model performance was evaluated using discrimination, calibration, and decision-curve analysis. The TRIPOD guidelines assessed transparent reporting. Results — We included 18 studies externally validating 10 different ML prediction models of the 59 available ML models after screening 4,682 studies. All external validations identified in this review retained good discrimination. Other key performance measures were provided in only 3 studies, rendering overall performance evaluation difficult. The overall median TRIPOD completeness was 61% (IQR 43–89), with 6 items being reported in less than 4/18 of the studies. Interpretation — Most current predictive ML models are not externally validated. The 18 available external validation studies were characterized by incomplete reporting of performance measures, limiting a transparent examination of model performance. Further prospective studies are needed to validate or refute the myriad of predictive ML models in orthopedics while adhering to existing guidelines. This ensures clinicians can take full advantage of validated and clinically implementable ML decision tools.

&lt;p&gt;Background: The prediction of aneurysm treatment outcomes can help to optimize the treatment strategies. Machine learning (ML) has shown positive results in many clinical areas. However, the development of such models requires expertise in ML, which is not an easy task for surgeons.&lt;/p&gt;&lt;p&gt;Objectives: The recently emerged automated machine learning (AutoML) has shown promise in making ML more accessible to non-computer experts. We aimed to evaluate the feasibility of applying AutoML to develop the ML models for treatment outcome prediction.&lt;/p&gt;&lt;p&gt;Methods: The patients with aneurysms treated by endovascular treatment were prospectively recruited from 2016 to 2020. Treatment was considered successful if angiographic complete occlusion was achieved at follow-up. A statistical prediction model was developed using multivariate logistic regression. In addition, two ML models were developed. One was developed manually and the other was developed by AutoML. Three models were compared based on their area under the precision-recall curve (AUPRC) and area under the receiver operating characteristic curve (AUROC).&lt;/p&gt;&lt;p&gt;Results: The aneurysm size, stent-assisted coiling (SAC), and posterior circulation were the three significant and independent variables associated with treatment outcome. The statistical model showed an AUPRC of 0.432 and AUROC of 0.745. The conventional manually trained ML model showed an improved AUPRC of 0.545 and AUROC of 0.781. The AutoML derived ML model showed the best performance with AUPRC of 0.632 and AUROC of 0.832, significantly better than the other two models.&lt;/p&gt;&lt;p&gt;Conclusions: This study demonstrated the feasibility of using AutoML to develop a high-quality ML model, which may outperform the statistical model and manually derived ML models. AutoML could be a useful tool that makes ML more accessible to the clinical researchers.&lt;/p&gt;

Metabolic dysfunction-associated steatotic liver disease (MASLD) can progress to metabolic dysfunction-associated steatohepatitis (MASH) and liver fibrosis, contributing to a heavier global health burden. Non-invasive diagnostic tools developed using machine learning (ML) and deep learning (DL), two representative artificial intelligence algorithms, are increasingly being explored for MASH and its related fibrosis assessment. This study aimed to compare the diagnostic performance of different ML and DL models and identify the top-performing models for diagnosing MASH and associated liver fibrosis. A systematic review and meta-analysis were conducted across PubMed, Web of Science, Embase and Cochrane Library from inception to May 18, 2025. Pooled area under the receiver operator characteristic curve (AUROC) values with 95 % confidence interval (CI) were calculated. Accuracy, specificity, sensitivity, positive predictive values, and negative predictive values were also recorded. Of 4,314 studies initially identified, 106 met the inclusion criteria, with 35 studies (ML: n=28; DL: n=7) providing data for analysis. Logistic Regression and Neural Network are the most commonly algorithms applied in ML and DL, respectively. The pooled AUROCs for diagnosing MASH were 0.833 (95 %CI: 0.806-0.860) for ML models and 0.841 (95 %CI: 0.782-0.900) for DL models. Light Gradient Boosting Machine (LightGBM) and ResNet50 were the best-performing models for diagnosing MASH within ML and DL algorithms, respectively, achieving corresponding AUROCs of 0.920 (95 %CI: 0.916-0.924) and 0.960 (95 %CI: 0.951-0.969). For fibrosis diagnosis, ML models had a pooled AUROC of 0.826 (95 %CI: 0.792-0.860), with Categorical Boosting (CatBoost) achieving the highest AUROC of 0.960 (95 %CI: 0.950-0.970). DL models yielded the pooled AUROC of 0.875 (95 %CI: 0.816-0.934) for fibrosis diagnosis. Both ML and DL models demonstrated strong diagnostic performance for MASH and liver fibrosis, with DL achieving marginally higher AUROCs. AI-driven approaches show promise in MASLD management.

Predictive models play a critical role in enhancing medication safety in clinical practice. While multiple models for adverse drug reactions (ADRs) in adults have demonstrated potential clinical utility, pediatric-specific predictive models remain understudied, with limited comprehensive evaluation of their methodological quality and reporting standards. To map the landscape of existing risk prediction models for ADRs specifically developed or validated for pediatric patients, and to describe their characteristics, methodological quality, and reporting completeness. A systematic search was conducted in Embase, PubMed, CNKI, Wanfang, VIP, and SinoMed for studies on pediatric ADR prediction models. The information from included studies was evaluated using the Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist, the risk of bias was assessed with the Prediction Model Risk of Bias Assessment Tool (PROBAST), and adherence to Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) guidelines was reviewed. Predictive ability of the included model, including the area under the receiver operating characteristic curve (AuROC), sensitivity, and specificity, was also reported. Out of 12,667 screened studies, 7 articles (describing 10 models) met the inclusion criteria. Study designs included case-control, nested case-control, prospective cohort, and cross-sectional studies. Logistic regression was the primary modeling method, with one study using machine learning. Common methodological limitations included unreported handling of missing data and univariable predictor screening. Model discrimination (AuROC) ranged from 0.63-0.97, with sensitivity and specificity between 52.02-98.50% and 33.33-98.79%, respectively. TRIPOD adherence varied (62.16-86.49%), with notable reporting deficiencies in blinding, sample size justification, intervention details, model usage instructions, and supplementary materials. No models underwent external validation. Existing pediatric ADR prediction models are limited by methodological and reporting shortcomings. Future research should focus on rigorous model development and external validation to ensure generalizability across diverse clinical settings.

Machine learning (ML) models provide more choices to patients with diabetes mellitus (DM) to more properly manage blood glucose (BG) levels. However, because of numerous types of ML algorithms, choosing an appropriate model is vitally important. In a systematic review and network meta-analysis, this study aimed to comprehensively assess the performance of ML models in predicting BG levels. In addition, we assessed ML models used to detect and predict adverse BG (hypoglycemia) events by calculating pooled estimates of sensitivity and specificity. PubMed, Embase, Web of Science, and Institute of Electrical and Electronics Engineers Explore databases were systematically searched for studies on predicting BG levels and predicting or detecting adverse BG events using ML models, from inception to November 2022. Studies that assessed the performance of different ML models in predicting or detecting BG levels or adverse BG events of patients with DM were included. Studies with no derivation or performance metrics of ML models were excluded. The Quality Assessment of Diagnostic Accuracy Studies tool was applied to assess the quality of included studies. Primary outcomes were the relative ranking of ML models for predicting BG levels in different prediction horizons (PHs) and pooled estimates of the sensitivity and specificity of ML models in detecting or predicting adverse BG events. In total, 46 eligible studies were included for meta-analysis. Regarding ML models for predicting BG levels, the means of the absolute root mean square error (RMSE) in a PH of 15, 30, 45, and 60 minutes were 18.88 (SD 19.71), 21.40 (SD 12.56), 21.27 (SD 5.17), and 30.01 (SD 7.23) mg/dL, respectively. The neural network model (NNM) showed the highest relative performance in different PHs. Furthermore, the pooled estimates of the positive likelihood ratio and the negative likelihood ratio of ML models were 8.3 (95% CI 5.7-12.0) and 0.31 (95% CI 0.22-0.44), respectively, for predicting hypoglycemia and 2.4 (95% CI 1.6-3.7) and 0.37 (95% CI 0.29-0.46), respectively, for detecting hypoglycemia. Statistically significant high heterogeneity was detected in all subgroups, with different sources of heterogeneity. For predicting precise BG levels, the RMSE increases with a rise in the PH, and the NNM shows the highest relative performance among all the ML models. Meanwhile, current ML models have sufficient ability to predict adverse BG events, while their ability to detect adverse BG events needs to be enhanced. PROSPERO CRD42022375250; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=375250.