The partial area under the summary ROC curve

Study question Can AI-based models reliably predict IVF outcomes by evaluating embryo quality? Summary answer The majority of AI-based models demonstrated high performance in predicting live birth, clinical pregnancy, and embryo ploidy status. What is known already Ensuring precision in embryo assessment is vital for optimizing IVF outcomes, yet traditional methods relying on human judgment remain subjective and introduce interobserver variability. Emerging AI-driven models integrating clinical, demographic, and laboratory data, beyond morphological and morphokinetics data, offer a more objective and comprehensive approach to evaluating embryo potential. Study design, size, duration A systematic review and meta-analysis building upon previous research, incorporating 21 studies published from 2022 onward. This continuation expands the evidence base on AI-driven prediction models in IVF/ICSI cycles, focusing on preimplantation embryos deemed suitable for transfer. The input data is diverse, incorporating demographics data, clinical and laboratory records, morphokinetics and morphological data. The primary outcomes assessed include the predictive performance for live birth, clinical pregnancy, and embryo ploidy status. Participants/materials, setting, methods A systematic literature search was conducted in the PubMed/Medline, Embase, and Web of Science databases, restricted to articles published in English up to January 2025. To ensure continuity with the previous meta-analysis, only studies published from 2022 onward were considered. The initial search identified 1319 studies, of which 57 were duplicates and 1241 were excluded for not meeting the inclusion criteria. After full-text screening and citation mining, 21 studies were deemed eligible for inclusion. Main results and the role of chance For live birth prediction, pooled sensitivity and specificity were 83% (95% CI: 71–90%) and 87% (95% CI: 59–97%), with PPV and NPV of 88% (95% CI: 48–98%) and 82% (95% CI: 76–87%). Clinical pregnancy prediction showed sensitivity of 80% (95% CI: 68–88%), specificity of 78% (95% CI: 69–85%), PPV of 80% (95% CI: 68–88%), and NPV of 72% (95% CI: 67–77%). For embryo ploidy, sensitivity and specificity were 74% (95% CI: 71–78%) and 73% (95% CI: 68–79%), with PPV and NPV both at 74%. Pooled estimates across all outcomes showed sensitivity of 74% (95% CI: 69–79%), specificity of 77% (95% CI: 72–82%), PPV of 78% (95% CI: 70–84%), and NPV of 74% (95% CI: 69–79%). The Area Under the Curve (AUC) of the Summary Receiver Operating Characteristics (SROC) curve was 0.809, and the partial AUC (pAUC) was 0.648, demonstrating consistent and reliable predictive performance across outcomes. Limitations, reasons for caution The limited number of studies meeting the inclusion criteria and the diverse designs used in developing AI models, which can increase heterogeneity, serve as limitations. Including women of all ages is another constraint, as advanced maternal age is linked to reduced IVF success rates. Wider implications of the findings The performance metrics suggest robust predictive capabilities of AI-based models in assisted reproduction. The latest results keep underscoring the potential of AI to complement traditional embryo evaluation methods, providing objective and reliable predictions for IVF outcomes. More research and increased collaboration among developers are essential for further improvements. Trial registration number No

BackgroundIn the recent years, machine learning algorithms have been more widely and increasingly applied in biomedical fields. In particular, their application has been drawing more attention in the field of psychiatry, for instance, as diagnostic tests/tools for autism spectrum disorder (ASD). However, given their complexity and potential clinical implications, there is an ongoing need for further research on their accuracy.ObjectiveThis study aimed to perform a systematic review and meta-analysis to summarize the available evidence for the accuracy of machine learning algorithms in diagnosing ASD.MethodsThe following databases were searched on November 28, 2018: MEDLINE, EMBASE, CINAHL Complete (with Open Dissertations), PsycINFO, and Institute of Electrical and Electronics Engineers Xplore Digital Library. Studies that used a machine learning algorithm partially or fully for distinguishing individuals with ASD from control subjects and provided accuracy measures were included in our analysis. The bivariate random effects model was applied to the pooled data in a meta-analysis. A subgroup analysis was used to investigate and resolve the source of heterogeneity between studies. True-positive, false-positive, false-negative, and true-negative values from individual studies were used to calculate the pooled sensitivity and specificity values, draw Summary Receiver Operating Characteristics curves, and obtain the area under the curve (AUC) and partial AUC (pAUC).ResultsA total of 43 studies were included for the final analysis, of which a meta-analysis was performed on 40 studies (53 samples with 12,128 participants). A structural magnetic resonance imaging (sMRI) subgroup meta-analysis (12 samples with 1776 participants) showed a sensitivity of 0.83 (95% CI 0.76-0.89), a specificity of 0.84 (95% CI 0.74-0.91), and AUC/pAUC of 0.90/0.83. A functional magnetic resonance imaging/deep neural network subgroup meta-analysis (5 samples with 1345 participants) showed a sensitivity of 0.69 (95% CI 0.62-0.75), specificity of 0.66 (95% CI 0.61-0.70), and AUC/pAUC of 0.71/0.67.ConclusionsThe accuracy of machine learning algorithms for diagnosis of ASD was considered acceptable by few accuracy measures only in cases of sMRI use; however, given the many limitations indicated in our study, further well-designed studies are warranted to extend the potential use of machine learning algorithms to clinical settings.Trial RegistrationPROSPERO CRD42018117779; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=117779

Several measures have been proposed to summarize the Receiver Operating Characteristic (ROC) curve, including the Projected Length of the Curve (PLC) and the Area Swept out by the Curve (ASC). These indices were first proposed by Lee (Epidemiology 1996; 7:605-611) to avoid certain deficiencies of the traditional Area Under the Curve (AUC) summary measure. More recently meta-analysis methods for assessing diagnostic test accuracy have been developed and the Summary Receiver Operating Characteristic (SROC) curve has been recommended to represent the performance of a diagnostic test. Some properties of the SROC curve were discussed by Walter (Statist. Med. 2002; 21:1237-1256). Here we extend that work to focus on properties of PLC and ASC in the context of SROC curve. Mathematical expressions for these two indices and their variances are derived in terms of the overall diagnostic odds ratio and the magnitude of inter-study heterogeneity in the odds ratio. Expressions for PLC and ASC and their variances are easily computed in homogeneous studies, and their values provide good approximations to the corresponding values for heterogeneous studies in most practical situations. General variances of PLC and ASC are derived by using delta methods, and are found to be smaller if the odds ratio is large. The methods are illustrated using data from two studies, the first being a meta-analysis on the detection of metastases in cervical cancer patients, and the second being a single study of HPV infection and pre-invasive cervical lesions.

Objective To assess the accuracy of contrast-enhanced EUS in differential diagnosis of benign and malignant pancreatic masses. Methods We systematically searched the Medline, PubMed, Web of Science, Embase, Cochrane Central Trials, CNKI and VIP databases for relevant studies published. Study selection, quality assessment and data extraction were performed by two reviewers independently. Meta-Disc （version 1.4） software was used to perform this meta-analysis for sensitivity, specificity, positive likelihood ratio （LR）, and negative LR. Pooling results were derived by using the fixed-effect model when significant heterogeneity was not present. The random-effect model was applied otherwise. A summary receiver-operating characteristic （SROC） curve was constructed. Furthermore, to explore the potential sources of heterogeneity, we used meta-regression to estimate the effect of the following characteristics of the studies on the diagnostic accuracy of contrast-enhanced EUS. In addition, the outliers were identified with the method described by Deville and a subgroup analysis was performed by excluding the outliers. We used Stata statistical software （version 10. 0） to assess the publication bias with the Begg-Mazumdar indictor and HarbordEgger indictor. Results Twelve studies involving 1139 patients were included. The pooled sensitivity ofcontrast-enhanced EUS for the differential diagnosis of pancreatic masses was 94% （95% CI, O. 91 ～0. 95）, the specificity was 89% （95% CI, O. 85 ～0. 92）, the positive LR was 8.09 （95% CI, 4. 47 ～ 14. 64）, and the negative LR was 0, 08 （95% CI, O. 06 ～ 0. 10）. The area under the curve （AUC） under SROC was 0. 9732 （ SE = 0.02）. The subgroup analysis by excluding two outliers provided a sensitivity of 93 % （95% C1, O. 91～0. 95） and a specificity of 93% （95% C1, O. 89～0. 95）. Additionally, the subgroup analysis showed that the heterogeneities were eliminated in pooled estimates when the outliers were excluded and the AUC under SROC was 0. 9745 （SE = 0. 02）. Moreover, no significant publication bias was found with the Begg-Mazumdar indictor （ P = 0.244） or the Harbord-Egger indictor （ P = 0.442 ）. Conclusion Contrastenhanced EUS is a valuable method in the differential diagnosis of pancreatic masses. Key words: Endoscopic ultrasonography ; Image enhancement ; Pancreatic neoplasms ; Meta-analysis

Objective To evaluate the diagnostic accuracy of miniprobe endoscopic ultrasonography(EUS)in assessment of tumor invasion depth in early gastric cancer. Methods We searched on the PubMed, Cochrane Library, Embase, CNKI, WanFang data, VIP data for relevant studies published.Stata 14.2 software was used to perform the Meta-analysis for sensitivity, specificity, positive likelihood ratio(LR), and negative LR, diagnostic odds ratio and 95% CI.A summary receiver-operating characteristic(SROC)curve was constructed, the area under the curve(AUC)was calculated, and the diagnostic value was evaluated.Furthermore, to explore the potential sources of heterogeneity, we used meta-regression to estimate the effect of the characteristics of the studies on the diagnostic accuracy of miniprobe EUS, and a subgroup analysis was performed.We assessed the publication bias using Deeks′Funnel plot. Results Sixteen studies with a total of 3 168 tumors were included in the analysis.The pooled sensitivity, specificity, positive likelihood ratio(LR), negative LR, diagnostic odds ratio and 95% CI and area under curve(AUC)were 0.85(95% CI: 0.78-0.90), 0.73(95% CI: 0.65-0.80), 3.2(95% CI: 2.4-4.1), 0.20(95% CI: 0.14-0.30), 16(95% CI: 9-26). The AUC of SROC curve was 0.85.No significant publication bias was found with Deeks′Funnel plot. Conclusion Miniprobe EUS can not accurately diagnose or exclude the M/SM1 layer invasion in early gastric cancer, but it still can greatly improve the diagnostic accuracy of invasion depth in early gastric cancer. Key words: Miniprobe endoscopic ultrasonography; Early gastric cancer; Invasion depth; Meta-analysis

Research Type: Level 3 - Retrospective cohort study, Case-control study, Meta-analysis of Level 3 studies Introduction/Purpose: The recently proposed Progressive Collapsing Foot Deformity (PCFD) classification system categorizes different deformity patterns into five classes (A, Hindfoot Valgus; B, Midfoot/Forefoot Abduction; C, Arch Collapse; D, Peritalar Subluxation; E, Ankle Valgus Instability). Currently, the decision-making process to determine the presence or absence of each deformity class relies on the surgeon's experience and interpretation of clinical and radiographic assessments, without established cut-off threshold values in the literature for PCFD measurements. This limits the reliability and reproducibility of the classification system. The goal of the current study was to establish cut-off threshold values for the most commonly used PCFD measurements assessing Class A to D deformities, using a large cohort of PCFD patients and controls. Methods: Two patient cohorts were included: a prospective cohort consisting of normally aligned and asymptomatic control volunteers (n=197 patients, 103 females/94 males, average age 41.7 years, average BMI 28.9) and a retrospective cohort of PCFD patients (n=321 patients, 136 females/185 males, average age 50.7 years, average BMI 29.8). All patients underwent Weight-Bearing Computed Tomography (WBCT) scans of the foot and ankle. Foot bone segmentation was performed semi-automatically using specialized software, followed by automatic measurements of various parameters including: Class A (Hindoot Moment Arm – HMA), Class B (Talonavicular Coverage Angle – TCA; Talus-First Metatarsal Angle Axial – TFMA-A), Class C (Talus-First Metatarsal Angle Sagittal – TFMA-S; Forefoot Arch Angle - FAA), Class D (Coverage Maps of the Subtalar Joint Middle-MF, Posterior Facets-PF; and Sinus Tarsi-ST). Receiver Operating Characteristic (ROC) curves, Youden's Indexes and Areas Under the Curve (AUC) were calculated for each measurement. P-values of 0.05 or less were considered significant. Results: The Cut-Off Threshold Values identified by the ROC curves using Youden’s Index for different PCFD measurements were as follows (all p-values < 0.0001): HMA of 13.9mm or more (AUC 85.3%) was considered diagnostic for Class A PCFD deformity. For Class B PCFD deformity, a TNCA of 38.7o or higher (AUC 84.4%) and TFMA-A of 20.3o or higher (AUC 82.1%) were considered diagnostic. For Class C PCFD deformity, a FAA of 8.7o or lower (AUC 83.9%) and TFMA-S of 18.7o or higher (AUC 82.9%) were diagnostic. For Class D Deformity, Coverage values for Middle Facet of 73.5% or lower (AUC 73.4%), Posterior Facet of 84.5% or lower (AUC 82.7%), and Sinus Tarsi of 25.7% or higher (AUC 84.4%) were diagnostic. Conclusion: In this diagnostic study involving 312 PCFD patients and 197 volunteer controls, we successfully determined the cut-off threshold values for key PCFD measurements related to Hindfoot Valgus (Class A), Midfoot/Forefoot Abduction (Class B), Arch Collapse (Class C), and Peritalar Subluxation (Class D) deformities. By utilizing a large cohort of patients and controls, we achieved high diagnostic accuracy for most measurements assessed (AUC of 80% or more) in identifying the presence or absence of these PCFD deformity classes. Our findings can potentially serve as normative threshold values to enhance the utilization, reliability, and reproducibility of the PCFD classification system. Figure 1 Scatter plots and ROC (Receiver Operating Characteristic) curves used to analyze parameters for the various deformity classifications in Progressive Collapsing Foot Deformity (PCFD). For the scatter plots, red indicates a normal control, green indicates an asymptomatic PCFD, and blue indicates a symptomatic PCFD.

Background: Approximately 10-20 million Americans have been clinically diagnosed with thyroid nodules. Shear wave elastography (SWE) and real-time elastography (RTE) are two primary forms of elastography for identifying the status of thyroid nodules. The aim of this study is to assess the performance of RTE and SWE in identifying malignant thyroid nodules. Methods: Relevant articles were systematically retrieved from PubMed, Embase and Cochrane Library. In order to evaluate the overall diagnostic accuracy, we have considered pooled sensitivity (SEN), specificity (SPE), area under the curve (AUC) and partial AUC with corresponding 95% confidence intervals (95%CIs). Stratified analyses by ethnicity (Caucasian, Asian), the number of malignant nodules (> 50, < 50), score system (elasticity scores: ES and strain ration: SR) and ES (> 4, < 4) were performed to explore potential sources of heterogeneity. All statistical tests were performed using the R 3.2.1 software package. Results: We analyzed 80 trials from 71 studies with 16,624 subjects (12,348 for SWE, 4,276 for RTE). The pooled results suggested that RTE is more accurate than SWE in diagnosing malignant cases (RTE: SEN= 0.829, 95%CI = 0.799-0.855, SPE = 0.828, 95%CI = 0.789-0.862, AUC = 0.889; SWE: SEN = 0.784, 95%CI = 0.732-0.828, SPE = 0.824, 95%CI = 0.766-0.871, AUC = 0.859). No significant difference was found in the subgroup analyses. Conclusion: Our findings revealed that RTE is superior to SWE in differentiating malignant and benign thyroid nodules. Nevertheless, more studies focusing on the diagnostic accuracy of RTE and SWE during different stages of thyroid nodules development should be carried out in the future.

Diagnostic value of ultrasound combined with computed tomography in papillary thyroid carcinoma and cervical lymph node metastasis: A meta-analysis

This study assessed the effects of food form and recreational resistance training (RT) on postprandial (PP) plasma glucose and insulin, energy expenditure (EE), and indices of appetite. 16 RT and 18 sedentary persons, age 62–84 y; physical activity 425 ± 33 vs. 272 ± 28 kcal/d* (*, p≤ 0.05); BMI 24.0 ± 0.6 vs. 25.7 ± 0.5 kg/m2*; and maximum strength 8.3 ± 0.3 vs. 6.7 ± 0.3 kg/kg FFM*, were studied. On separate days, each subject consumed 12.5% of their energy need as macronutrient‐matched meal replacement products in solid (S) or beverage (B) form. Parameters were assessed over 4 h. Independent of RT, glucose and insulin area under the curve (AUC) were higher for S (503 ± 292; 2809 ± 273) vs. B (−266 ± 240 mg/dL•240 min*; 1865 ± 179 uU/mL•240 min*). PPEE AUC was not different between S vs. B, and the respiratory exchange ratio AUC was lower for S (−11.54 ±−2.04) vs. B (−5.36 ± 3.07 •240 min*). Fullness AUC was higher for S (475 ± 96) vs. B (218 ± 94 AU•240 min*) and sedentary (450 ± 106) vs. RT (230 ± 118 AU•240 min*); these responses were independent and additive. Food form and RT did not influence hunger AUC, while desire to eat AUC was lower for S vs. B* and not influenced by RT. Meal replacement products consumed in solid and beverage forms should not be considered as metabolic and appetitive equivalents in older people who are sedentary or more physically active and use resistance training to achieve higher strength.Support: NIH R01 AG021911, The Solae Company

The skeleton is one of the favorable sites for the metastasis of almost all human malignant neoplasms. An accurate diagnosis of bone metastasis is crucial for the patient's staging and management. To investigate and compare diagnostic performance of 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) and bone scintigraphy (BS) for detection of bone metastasis in malignancies using meta-analysis. PubMed (Medline included) was searched for relevant articles. We assessed the methodological quality with Quality Assessment of Diagnosis Accuracy Studies (QUADAS) score tool, and used statistical software to obtain pooled estimates of sensitivity, specificity, diagnostic odds ratio (DOR), and summary receiver-operating characteristic (SROC) curve. Six studies met inclusion criteria. For 18F-FDG PET/CT, the pooled sensitivity and specificity were 0.934 and 0.975, respectively. The pooled positive likelihood ratio (LR+), negative likelihood ratio (LR-) and diagnostic odds ratio (DOR) were 34.990, 0.068 and 559.02, respectively. The area under the SROC curve was 0.9854. For BS, the pooled sensitivity, specificity, LR + , LR- and DOR were 0.706 (0.642-0.764), 0.911 (0.896-0.926), 13.982 (2.419-80.817), 0.319 (0.143-0.712), and 60.420 (21.393-170.64), respectively. The area under the SROC curve was 0.9386. The results indicate that 18F-FDG PET/CT do have both higher sensitivity and specificity than bone scintigraphy for detecting metastatic bone tumor. However, further research is needed to evaluate the diagnostic performance of 18F-FDG PET/CT and BS in each common malignancy.

Background: Non-Small Cell Lung Cancer (NSCLC) remains a challenging disease to manage with effectiveness. Early detection and precise monitoring are crucial for improving patient outcomes. Circulating tumor DNA (ctDNA) offers a non-invasive cancer detection and monitoring method. Emerging biomarkers, such as ctDNA methylation, have shown promise in enhancing diagnostic accuracy and prognostic assessment in NSCLC. In this review, we examined the current evidence regarding ctDNA methylation's role in NSCLC detection through a systematic review of the existing literature and meta-analysis. Methods: We systematically searched PubMed, Medline, Embase, and Web of Science databases up to 26 June 2024 for studies on the role of ctDNA methylation analysis in NSCLC patients. We included studies from 2010 to 2024 on NSCLC patients. We excluded case reports, non-English articles, studies on cell lines or artificial samples, those without cfDNA detection, prognostic studies, and studies with non-extractable data or mixed cancer types. Funnel plots were visually examined for potential publication bias, with a p value < 0.05 indicating bias. Meta-analysis was conducted using R packages (meta, forestplot, and mada). Combined sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR-), positive and negative predictive values, diagnostic odds ratio (DOR), and 95% confidence intervals (95% CI) were calculated. A summary receiver operating characteristic curve (SROC) and area under the curve (AUC) with related Standard Error (SE) were used to evaluate the overall diagnostic performance. Additionally, RASSF1A, APC, SOX17, SEPT9, and RARβ2 were analyzed, since their methylation was assessed in two or more studies. Results: From 38 candidate papers, we finally identified 12 studies, including 472 NSCLC patients. The pooled sensitivity was 0.62 (0.47-0.77) and the specificity was 0.90 (0.85-0.94). The diagnostic odds ratio was 15.6 (95% CI 9.36-26.09) and the area under the curve was 0.249 (SE = 0.138). The positive and negative predictive values were 5.38 (95% CI 3.89-7.44) and 0.34 (95% CI 0.22-0.54), respectively. For single genes, the specificity reached 0.83~0.96, except for RARβ2, but the sensitivity was relatively low for each gene. Significant heterogeneity across the included studies, the potential publication bias for specificity (p = 0.0231), and the need to validate the clinical utility of ctDNA methylation for monitoring treatment response and predicting outcomes in NSCLC patients represent the main limitations of this study. Conclusions: These results provide evidence of the significant potential of ctDNA methylation as a valuable biomarker for improving the diagnosis of NSCLC, advocating for its integration into clinical practice to enhance patient management.

Abies spectabilis (D.Don) Mirb. is an endemic Himalayan near-threatened coniferous species and the predominant treeline-forming species in the Indian Central Himalaya (Uttarakhand). The impact of climate change and anthropogenic activities is perilous to its habitat distribution. Accurate species habitat distribution is a prerequisite for efficient conservation planning. This vital information is still missing in the Indian Himalayan region for A. spectabilis. This study models the habitat suitability of A. spectabilis in Uttarakhand and discusses the management implications. Species occurrence records from primary and secondary sources were used with environment variables for habitat suitability modeling using the MaxEnt approach. Environment variables included bioclimatic (BCVs), topographic, edaphic, and anthropogenic variables. The BCVs from two global bioclimatic databases CHELSA (model 1) and WorldClim (model 2) were used. Models were validated using threshold-independent measures and further used to build habitat suitability maps. Both models performed optimally; however, model 2 had higher average values for the area under the curve (AUC) (0.980), partial AUC (0.977), and the AUC ratio (1.954) indicating higher predictive power. The precipitation of the driest month (BIO14) was the most important predictor variable under both models. The habitat suitability area for model 1 (16,324.57 km2) was five times greater than model 2 (3323.86 km2). Only 826.86 km2 area (model 2) was highly suitable for A. spectabilis. The habitat suitability is concentrated predominantly in a tight group in the northern region of Uttarakhand. Chamoli (8.43%), Rudraprayag (6.20%), and Uttarkashi (0.53%) were the top 3 districts with the highest percentage of high suitability regions under model 2. The results from the habitat suitability distribution of A. spectabilis suggest in situ conservation of present occurrences and monitoring of the regeneration process. Stringent monitoring in the protected areas and the highly suitable habitat regions modeled must be used for assisted regeneration and plantation-based habitat enhancement of this endangered species.

Clinical prediction rules could help identify children at risk of slowed growth after an episode of diarrheal illness, and these rules may be generalizable to all children, regardless of diarrhea status.

Introduction Currently, the Dukes and AJCC TNM staging systems are widely used to predict survival from colon cancer and assist with clinical decision making. The Memorial Sloan Kettering cancer Centre (MSKCC) nomogram has been developed as an adjunct to these, to predict the five and ten year recurrence free survival probability for patients who had curative resection for colon cancer. The aim of this study is to externally validate the MSKCC nomogram in the UK population, and determine its usefulness. Method The colon cancer database from a district general hospital in England was used to extract all patients who had a curative colon cancer resection. Inclusion criteria were all patients who had curative elective colon cancer resection over a five year period. Patients included in the study were followed up for up to ten years. Variables required for the nomogram were: age, sex, tumour location, CEA levels, T stage, positive lymph nodes, negative lymph nodes, tumour differentiation, lymphovascular invasion, perineural invasion and whether post-operative chemotherapy was given. Five and ten year predictions were calculated for each patient, and plotted against the actual recurrence using a Receiver Operator Characteristic (ROC) curve, and Area Under the Curve (AUC) was calculated for both the five and ten year nomogram. Two cut off points were identified for each nomogram, and this helped to stratify our cohort into three risk groups depending on risk of recurrence. Results 138 patients were included in the study. 49.2% were males with mean age of 69.47 years (SD= 10.81). Overall five year recurrence rate was 26.8% with a mean follow up of 60.24 months (SD= 38.6). 118 patients were included in the five year nomogram validation, and 102 patients were included in the ten year nomogram validation. A ROC curve was plotted for both the five and ten year nomogram and AUC was calculated. For the five year nomogram AUC was 0.673, and for the ten year nomogram AUC was 0.687. Two cut off points were identified for each nomogram and this allowed the groups to be stratified into low, medium and high risk for recurrence. Cox regression showed a significant difference between all groups for both nomograms. Conclusion The MSKCC colon cancer nomogram was validated in our cohort, but is recommended to be used in conjunction with AJCC TNM staging system. The use of risk groups to stratify patients may allow the nomogram to be used simply and effectively in clinical practice. Disclosure of interest None Declared.

BackgroundWearable devices enable the continuous collection of kinematic information, such as gait and postural control, in real-life environments, offering potential for the early identification and stratified management of fall risk in older adults. However, quantitative integrated evidence regarding their overall accuracy in predicting future falls is lacking. This systematic review and meta-analysis aims to evaluate the accuracy of wearable devices in predicting falls among older adults and to explore the potential influence of key study characteristics on predictive performance.MethodsA systematic search was conducted in PubMed, Embase, Web of Science, and the Cochrane Library from database inception to October 9, 2025. Using a bivariate random-effects model, we pooled sensitivity and specificity, calculated likelihood ratios, and fitted a summary receiver operating characteristic (SROC) curve to determine the area under the curve (AUC). Subgroup analysis and meta-regression explored potential sources of heterogeneity. The risk of bias was assessed with the PROBAST tool.ResultsA total of 20 studies were included. The pooled sensitivity was 0.55 (95% CI: 0.42–0.67), specificity was 0.89 (95% CI: 0.84–0.93), positive likelihood ratio was 5.2, negative likelihood ratio was 0.50, and diagnostic odds ratio was 10.39. The area under the summary receiver operating characteristic (SROC) curve was 0.85 (95% CI: 0.81–0.88). Subgroup and regression analyses indicated that studies employing machine learning modeling demonstrated superior overall discriminative ability (AUC = 0.90). Predictive performance may be influenced by factors such as population age structure, sample size, and sensor placement location.ConclusionWearable devices exhibit good discriminative ability for predicting future falls in older adults, characterized overall by high specificity and moderate sensitivity. They are more suitable as tools for early screening and risk stratification in community and institutional settings, thereby supporting decision-making regarding intervention priorities.Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/view/CRD420251274570.