SEARCH · PubMed Health

Results for “External validation”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 127 records · Page 7Linked to original sources

PubMedDate not supplied

[Theory and practice in medical specialization. I. An instrument for measuring learning strategies].

We present the development and validation of a measurement instrument intended to estimate the degree of vinculation between the theoretical and practical learning activities of medical residents in their usual working conditions in hospitals. The main reason for residents to read medical literature is to find support to their decisions when treating patients. Based on this perspective we designed a self-applied questionnaire that explores diverse circumstances in which the vinculation between theory and practice may be expressed. This instrument was validated through rounds of experts in terms of its construction and content. Its external validity was explored with two groups of internal medicine residents with a different degree of vinculation between theory and practice: one high and the other with a low vinculation. In addition, we designed a guide for the direct observation of the theoretical and practical clinical learning activities in order to estimate its concordance with the results of the questionnaire. The questionnaire was able to discriminate the group differences and showed a satisfactory concordance with the information provided by direct observation. A copy of the questionnaire is available by request to the authors. We conclude that in its present stage of development, the instrument has shown internal and external validity and may be used to explore the process training of medical residents.

Education, Medical↗

PubMed2026

Data-centric, robust, and explainable multimodal deep learning for clinical decision support: A systematic review.

PURPOSE: Multimodal deep learning is increasingly proposed for clinical decision support (CDS) under a "data-centric" framing that prioritizes label quality, missing-modality robustness, distribution shift, calibration, and explainability. Prior reviews have examined multimodal medical AI, CDS, and data-centric methods separately, but none address their intersection. We mapped the modalities, fusion strategies, and data-centric and explainability techniques used in this recent literature, quantified how often each is implemented rather than merely mentioned, assessed deployment-relevant evidence (external validation, clinical-outcome measurement, equity), and formally appraised study-level risk of bias. METHODS: Following the PRISMA 2020 statement (PROSPERO CRD420261427815; registered retrospectively), we screened 150 records and included primary, clinical, multimodal studies that applied machine or deep learning to a decision-support task and reported at least one quantitative result. Two reviewers screened and extracted data with consensus adjudication. Each study was coded against pre-specified operational definitions, separating implemented or empirically evaluated techniques from those only mentioned. Study-level risk of bias was assessed with PROBAST + AI. Synthesis was narrative. RESULTS: Thirty-one studies met inclusion; 30 (97%) were published between 2024 and 2026, with a median of three modalities (range 2-6), most commonly structured EHR (71%) and imaging (39%). Data-centric techniques were frequently reported (74-84% across label-noise, distribution-shift, calibration, missing-modality and class-imbalance handling; equity 61%). However, external validation was reported in only 4/31 studies (13%), a clinical or provider outcome in 3/31 (10%), and no study reported routine deployment. Overall risk of bias was high in 27/31 studies (87%), driven by the analysis domain. CONCLUSION: Within this recent, self-selected slice of the field, technical robustness and explainability techniques are widely reported but rarely validated out-of-distribution or against clinical outcomes, and the underlying evidence is at high risk of bias. Progress requires external multi-site validation, clinical-outcome measurement, formal bias appraisal, and adherence to AI reporting standards (e.g., TRIPOD + AI) before deployment can be justified.

Deep Learning↗

PubMed1999

Representativeness and response rates from the Domestic/International Gastroenterology Surveillance Study (DIGEST).

BACKGROUND: The Domestic/international Gastroenterology Surveillance Study (DIGEST) examined the prevalence of upper gastrointestinal symptoms among the general population in 10 countries, and the impact of these symptoms on healthcare usage and quality of life. This report discusses the validation of the DIGEST sample and reviews the response rates from the survey. METHODS: External validation of the DIGEST sample was conducted by comparing the age, age by gender and annual household incomes of the sample with census-derived data. A comparison was also made between Psychological General Well-Being Index (PGWBI) scores from study subjects in the Scandinavian countries and the USA and the total sample population norms. RESULTS: Under- and oversampling, defined as > or =5% difference from the population norms, was evident in eight out of 10 countries, but no systematic bias was evident. The final distribution of the sample by gender was 51% female and 49% male. Although differences in PGWBI scores were noted between DIGEST subjects and population norms, these differences were <0.30 standard deviations--markedly below the difference considered as relevant for the PGWBI. Response for the survey in individual countries ranged from 17% in the USA to 61% in Norway, with a survey-wide rate of 27%. The overall response rate, including primary non-respondents, was 13.4%. The majority of nonresponse (51.4%) was attributed to failure to establish contact with the subjects, with 41.7% of subjects declining to be interviewed and the remaining 6.9% of subjects not meeting the age and sex criteria used for the survey. CONCLUSIONS: The DIGEST sample exhibited good external validity, providing a foundation for comparison between data derived from individual countries in the survey.

Adult↗

PubMed2026

Beyond predictive performance: A systematic review and critical methodological appraisal of AI/ML and conventional modelling strategies in breast, colorectal, and pancreatic Cancer.

BACKGROUND: Predictive modelling for cancer risk, treatment-related complications, and survival is central to precision oncology. Conventional logistic regression (LR) and Cox proportional hazards (CoxPH) regression remain widely used but are limited when modelling nonlinear interactions, high-dimensional imaging features, and multimodal clinical-metabolic predictors. Artificial intelligence (AI) and machine learning (ML) methods offer expanded capability through automated feature extraction, ensemble learning, and flexible survival modelling, but the evidence on when AI/ML adds value over conventional models across cancer sites and predictive tasks remains fragmented. OBJECTIVE: To systematically evaluate the methodological performance, validation strategies, and translational limitations of AI/ML models compared with conventional statistical models in published predictive-modelling studies for breast, colorectal, or pancreatic cancer. METHODS: PubMed, Scopus, and Web of Science were searched for studies published between January 2019 and March 2025. Two reviewers independently conducted title-and-abstract screening, full-text eligibility assessment, and PROBAST risk-of-bias assessment. Sixty-five studies (n = 907,567 participants) were narratively synthesised by cancer site, predictive task, model family, comparator, validation strategy, predictor modality, and calibration or explainability reporting. RESULTS: The 65 studies comprised breast cancer (n = 35), colorectal cancer (n = 21), and pancreatic cancer (n = 9). AI/ML superiority over LR and CoxPH was task- and data-dependent. CNN- and U-Net-based models predominated in imaging and body-composition tasks, tree-based ensembles consistently outperformed LR for tabular perioperative complication prediction, and CoxPH remained competitive, and in the largest pancreatic risk study, superior to XGBoost (C-index 0.802 vs 0.723) in well-structured datasets. PROBAST analysis-domain risk was moderate in 54 of 65 studies (83%), driven by limited external validation, sparse calibration reporting (11/65), and few decision-curve analyses (7/65). CONCLUSION: AI/ML adds the most methodological value in imaging-derived feature extraction and nonlinear perioperative prediction, while conventional regression remains preferable in large, structured datasets with linear predictors. Clinical translation requires standardised body-composition definitions, external validation, calibration assessment, decision-curve analysis, and explainability, in line with TRIPOD+AI and CLAIM standards.

Humans↗

PubMed1987

How generalizable are the effects of smoking prevention programs? Refusal skills training and parent messages in a teacher-administered program.

This study investigated both substantive and methodological issues associated with school-based smoking prevention programs. Substantive issues included the efficacy of a refusal skills training curriculum and of parent messages mailed to students' homes. Methodological issues included the effects of assigning classrooms versus entire schools to experimental conditions and determination of the effects of attrition on internal and external validity. Results revealed differential impact for different subgroups of adolescents. The refusal skills program produced lower rates of smoking than the control condition for students who were smokers at the pretreatment assessment but may have produced detrimental effects among males who were nonsmokers at pretest. The provision of parent messages did not affect outcome. Method of assignment (schools versus classrooms) failed to produce significant effects, and attrition did not affect internal validity. However, the above differential findings, as well as the impact of attrition on external validity, raise questions concerning the generalizability of smoking prevention programs.

Adolescent↗

PubMedDate not supplied

[Interobserver and intraobserver variation: a problem of validity in epidemiologic studies of arterial pressure].

In carrying out blood pressure epidemiologic studies there may be different factors that can affect internal and external validity and thus eliminate the inferential process. As part of the Hypertension and Risk Factors Associated Study conducted in March 1987 in Cuajimalpa de Morelos, Mexico City, 23 nursing students were standardized on the blood pressure auscultatory method using a sound picture and measuring intraobserver and interobserver agreement through intraclass correlation coefficient. Even though initial standardization sessions showed difficulties in the use of instruments and in the reading of blood pressure levels, final K (kappa) values measuring interobserver agreement increased from 0.25 to 0.86. Omega values measuring intraobserver agreement fluctuated between 0.86 and 0.98. This epidemiologic technique is proposed in order to improve internal and external validity of blood pressure studies.

Blood Pressure↗

PubMed1996

The WHO (Ten) Well-Being Index: validation in diabetes.

BACKGROUND: In a European trial in 8 countries, the subjective well-being of patients on alternative forms of treatment for insulin-dependent diabetes was compared using the 28-item WHO Well-Being Questionnaire, covering four dimensions of depression, anxiety, energy and positive well-being. The objective of the analysis reported here has been to identify the items of the WHO questionnaire which belong to an overall index of negative and positive well-being. METHODS: Adult patients at 10 study centres in 8 countries who had been on insulin for at least 2 years were invited to participate in a randomised, cross-over trial to compare insulin pump treatment with injection therapy. At each phase, patients completed questions on well-being and general health. Internal validity of the well-being index was evaluated by Cronbach's alpha and Loevinger's and Mokken's homogeneity coefficients, as well as factor analysis. External validity was evaluated by comparisons with results of the general assessment questions and by the ability to discriminate between the alternative forms of treatment. RESULTS: 358 patients had sufficient data for analysis. Ten items were found to constitute a valid index of well-being with respect to internal and external validity. Coefficients of homogeneity were acceptable and there was evidence for both concurrent and discriminant validity. CONCLUSIONS: The WHO (Ten) well-being index includes negative and positive aspects of well-being in a single uni-dimensional scale. Its advantage lies in its ability to show overall change along the continuum of well-being, thus facilitating comparisons between patient groups and treatments. It is not specific to diabetes, and therefore may be useful as a disease-independent index of well-being in a broad range of health care studies.

Adaptation, Psychological↗

PubMed2026

Artificial Intelligence for Diagnosis, Risk Stratification, and Prognosis of Neuroblastoma - A Systematic Review and Meta-Analysis.

PURPOSE: To synthesizes evidence on artificial intelligence (AI) performance in neuroblastoma (NB) diagnosis, risk stratification, prognosis, and genomic characterization. MATERIALS AND METHODS: A systematic review and meta-analysis was conducted following PRISMA 2020 guidelines (PROSPERO: CRD42024539475) across five databases. Meta-analyses used random-effects models with logit-transformed Area Under the Curve (AUCs) and cluster-robust standard errors. AI models were classified as Machine Learning Models (MLM) or Hybrid Nomograms (HN) based on their construction methodology. RESULTS: Of 3,742 articles identified, 53 were included. MLMs demonstrated higher point estimates than radiologists in differential diagnosis (AUC: 0.87 vs. 0.83), though this difference was not statistically significant and carried substantial uncertainty. HNs achieved stronger performance in risk stratification (AUC: 0.87). AI-derived nomograms (AUC: 0.9) and gene signatures (AUC: 0.8) outperformed conventional prognostic markers descriptively. Chemotherapy response prediction remained below clinical utility thresholds across all model types. Only 33.9% of models reported calibration and 24.5% underwent external validation. CONCLUSIONS: AI demonstrates proof-of-concept across multiple NB clinical domains. However, clinical adoption remains premature given persistent gaps in external validation, calibration, dataset size, and pediatric-specific model development. Future studies should test these models prospectively in multicenter pediatric cohorts, ideally through COG or SIOPEN, using shared definitions for diagnosis, risk group, treatment response, and survival outcomes.

Humans↗

PubMed1991

Subject attrition in prevention research.

Subject attrition threatens the internal validity of substance abuse prevention studies because differences in the rate of attrition and the substance use behavior of remaining subjects in the different conditions could account for any differences found in substance use rates. Attrition threatens the external validity of prevention studies because, to the extent that study dropouts are different from remaining subjects, the results of the study may not be generalizable to study dropouts. Analysis of these threats to the validity of prevention studies should be routinely conducted. However, studies of alcohol and drug abuse prevention have generally failed to report or analyze subject attrition. Smoking prevention studies have more frequently reported attrition, and they have recently begun to analyze the degree to which attrition may affect the internal and external validity of the study. Evidence thus far suggests that differences in attrition across conditions do occur occasionally. The evidence is substantial that study dropouts are systematically more likely to smoke, to use other substances, and to score highly on other risk-taking measures.

Alcoholism↗

PubMed2026

Research progress and application prospects of multi-omics integration strategies in precision risk stratification of type 1 diabetes mellitus.

Type 1 diabetes (T1D) is a chronic metabolic disease mediated by autoimmunity. Its pathogenesis involves complex interactions between genetic susceptibility and environmental factors. Conventional T1D risk stratification primarily relies on genetic markers, islet autoantibodies, and glycemic indicators. Although these biomarkers remain indispensable in current clinical practice, they are often insufficient when used alone to accurately identify ultra-early high-risk individuals, predict disease progression rates, or support individualized preventive strategies. Consequently, more comprehensive molecular approaches are needed to improve precision risk stratification. In recent years, the rapid development of multi-omics technologies has provided new strategies for precise risk stratification of T1D. This narrative review critically evaluates how multi-omics integration strategies can improve precision risk stratification throughout the T1D disease continuum by integrating complementary molecular information from genomics, transcriptomics, proteomics, metabolomics, epigenomics, and the microbiome. Particular emphasis is placed on stage-specific biomarker discovery, multi-omics data integration frameworks, artificial intelligence-assisted prediction models, biomarker validation, and the opportunities and challenges associated with clinical translation. Current evidence suggests that integrated multi-omics approaches have the potential to improve risk prediction accuracy, distinguish heterogeneous disease trajectories, identify individuals at imminent risk of progression, and provide biologically informed targets for precision intervention. However, important challenges remain, including data harmonization, external validation, model interpretability, cost-effectiveness, and integration into routine clinical screening programs. Future research should prioritize prospective multicenter cohorts, standardized analytical pipelines, externally validated prediction models, and clinically interpretable multi-omics frameworks to facilitate the translation of precision risk stratification into routine T1D prevention and management.

Humans↗

PubMed2026

Ensemble DNA methylation clock demonstrates Immune-metabolic aging signatures associated with mortality.

Aging is a multifactorial process that is best described in terms of the progressive acquisition of multiple layers of phenotypic changes, such as epigenetic modifications, inflammation, and metabolic dysregulation. DNA methylation clocks have been extensively used to construct epigenetic clocks based on the DNAm profiles that can be used to estimate biological age and predict age-associated outcomes. Nevertheless, the vast majority of clocks constructed so far have been based on linear models, which are unlikely to fully account for the heterogeneity and non-linearity of survival-related DNAm signatures. In this work, we constructed a heterogeneous stacked ensemble survival model based on DNAm data obtained from the Framingham Heart Study. We first identified 190 CpG loci using elastic net Cox regression and subsequently constructed a survival prediction model based on the fusion of five complementary survival models by means of a neural network meta-learner. The prediction power of the survival model was evaluated in an external validation cohort, where we observed strong performance for predicting all-cause mortality that significantly exceeded PhenoAge and was statistically comparable to GrimAge. These performance estimates were derived in cohorts of European ancestry and externally validated in postmenopausal women aged 50-79 years, and should therefore be interpreted as applicable only to demographically similar populations.

Humans↗

PubMed1997

The reliability of upper limb anthropometry in older Chinese people.

OBJECTIVE: To evaluate the validity of the Durnin-Womersley equations and to derive our local predictive equations for body fat from upper limb skinfold thicknesses in older Chinese people in Hong Kong. To evaluate the validity of mid-arm circumference and corrected arm muscle area in predicting lean tissue mass in the same population. DESIGN: Comparison of fat percentages predicted by Durnin-Womersley (D-W) equations with those estimated by Dual energy X ray absorptiometry (DXA). Predictive equations derived from regression between upper limb skinfold thicknesses and fat percentages estimated by DXA were similarly evaluated in internal and external validation groups. Mid-arm circumference (MAC) and corrected arm muscle area (CAMA) were correlated with the limb lean tissue mass, body lean tissue mass and fat percentage. SUBJECTS: 354 female and 263 male, apparently well, community dwelling subjects, aged 69-82 y; of which 40 subjects of each sex were randomly selected from the study population for internal validation of the local predictive equations; 60 female and 33 male hospital medical outpatients, aged 61-87 y, were recruited for external validation. MEASUREMENTS: Triceps and biceps skinfold thicknesses, mid-arm circumference, body mass index, fat percentages, limb and whole body lean tissue masses estimated by Hologic QDR-2000 bone densitometer. RESULTS: Fat percentages calculated by D-W equations were significantly different from those estimated by DXA (average difference -2.4 (s.d. 4.8)% and +2.1 (5.2)% in females and males respectively). The corresponding differences for our local predictive equations were not significant (-0.9 (4.7)% and -0.5 (5.0)% in females and males respectively). There was a trend of under-estimation of body fat with increasing fatness. In the hospital medical outpatients, there was a significant difference between fat percentages predicted by our equation and those by DXA in female (-2.9(5.3)%), but not in male (+0.3(4.3)%) subjects. In males, MAC correlated with limb and body lean tissue masses as well as with fat percentage (r = 0.60, 0.68, 0.65 respectively). CAMA correlated similarly well with lean tissue masses but was more independent of fat percentage (r = 0.61, 0.65, 0.44 respectively). In females, both MAC and CAMA correlated poorly with limb and body lean tissue masses. Moreover, MAC correlated well with fat percentage (r = 0.80). CONCLUSION: Upper limb skinfold thicknesses measurement is a valid means of predicting body fat in older Chinese people. Local predictive equations were more reliable that D-W equations. They were, however, subject to errors at the extreme ends of body fatness and in the presence of disease. In older females, MAC and CAMA were not reliable in predicting lean tissue mass, but MAC could be used to predict fat percentages. In older males, CAMA was more reliable than MAC in predicting lean tissue mass.

Absorptiometry, Photon↗

PubMed2026

Catecholaminergic polymorphic ventricular tachycardia mediated by ryanodine receptor 2: a validated risk stratification.

BACKGROUND AND AIMS: Patients with catecholaminergic polymorphic ventricular tachycardia (CPVT) are at risk for potentially life-threatening arrhythmic events (AEs) even while treated with β-blockers. The aim was to develop a model for individualized prediction of AEs in patients with RYR2-mediated CPVT on β-blocker monotherapy. METHODS: The derivation and independent validation cohorts included 743 and 129 patients, respectively. AEs were defined as arrhythmic syncope, appropriate implantable cardioverter-defibrillator shock, sudden cardiac arrest (SCA), and sudden cardiac death. Near-fatal or fatal AEs (nf/fAEs) included all AEs except for arrhythmic syncope. Prediction models using Cox regression were developed and internally and externally validated. RESULTS: A total of 102 (13.7%) patients in the derivation cohort and 24 (18.6%) patients in the validation cohort experienced ≥1 AE over a median follow-up of 5.1 [interquartile range (IQR), 7.7] and 2.4 (IQR, 4.4) years, respectively. Predictors of AE were arrhythmic syncope or SCA prior to diagnosis and age at β-blocker initiation. In the derivation and validation cohorts, the optimism-corrected C-indices of the models for AE were 0.67 [95% confidence interval (CI) 0.62-0.72] and 0.59 (95% CI 0.48-0.71), respectively. For nf/fAEs, ventricular arrhythmia severity before β-blocker initiation was a fourth independent predictor, and C-indices of the models in the derivation and validation cohorts were 0.74 (95% CI 0.68-0.80) and 0.60 (95% CI 0.47-0.72), respectively. In the derivation cohort, calibration slopes were 1.00 (95% CI 0.59-1.41) for AE and 1.00 (95% CI 0.69-1.32) for nf/fAE. CONCLUSIONS: These externally validated risk prediction models using clinical parameters accurately distinguished CPVT patients on β-blocker monotherapy at low and high risk for future AEs while treated with β-blockers. These models provide guidance for implementation of clinical management therapies to prevent AEs in patients with CPVT.

Humans↗

PubMed1998

Die Vision einer pragmatischen klinischen Forschung oder das Ende der Diskussion über < > und < >

The Concept of Pragmatic Clinical Research or the End of Discussion about 'Placebo' and 'Specific Effects&rsquoThe reorientation of clinical research towards the questions of treatment benefit (beyond the question of treatment efficacy) and of how much clinical trials represent actual practice (external validity) is the timely path to clinical research questions of real interest and importance. Postmodern 'anything goes' makes it possible to also consider thus far looked down on placebo effects as valuable, however, it requires the precise documentation of the external validity of such effects. Not disease as such, but the disease context, not therapy as such, but the therapy context, not the patient as such, but the patient context, not a test as such, but the test situation have become the important focuses of clinical research. In respect to test results current medicine has to recognize its illiterate mystification of allegedly 'objective' and 'hard' data. The patient context can determine whether an 'efficacious' therapy is beneficial or harmful, and thus, it is the proper definition of the patient context which makes medicine scientific, no matter how 'objective' or 'subjective' the effect of therapy is. The consideration of the therapy context leads to the important distinction between efficacy and effectiveness (or benefit), and it becomes intelligible that the randomised controlled trial in its traditional design as the placebo-controlled double-blind trial is limited to the evaluation of an agent theory. The evaluation of treatment effectiveness requires more pragmatic trials which study treatment operations and not isolated components and which may even compare entire treatment strategies. Pragmatic clinical trials, in future, will not only allow the study of 'pathogenesis blockers'., but also the study of 'salutogenetic' interventions working with the formation of the host. The focus of attention and research in the new school of evidencebased medicine with clinical epidemiology as its basic science (if not superficially understood as mere literature medicine) has long ago been identified as illness as the product of host, disease and environment. The dispute about 'placebo' and 'specific effects', in the meantime, has become obsolete.

Journal Article↗

PubMed1989

Meaning in life depth in the active married elderly.

To discover if elderly people have developed a deeper meaning in life than younger individuals, a sample of active married elderly people was compared to a group of younger adults. Two dimensions of meaning in life depth were investigated. The first was a self-suitability measure indicating comfort with one's own meaning, measured by Crumbaugh and Maholick's (1969) Purpose in Life Test. The second was an external validation measure derived from a statement about their own strongest meaning in life, written by the participants and rated for depth by two outside judges. The older group scored significantly higher than the younger adults on the self-suitability measure and significantly lower on the external validation measure. Such results could mean that toward the end of life we are better able to appreciate life's beauty though less able to communicate our depth of appreciation to others. An alternative interpretation of the results is that the elderly participants were engaging in self-deception.

Aged↗

PubMed2026

An Exosomal Signature for Preoperative Detection of Occult Liver Metastasis in Pancreatic Cancer.

IMPORTANCE: Early liver metastasis (early-LiM) after pancreatectomy represents an aggressive biological phenotype of pancreatic ductal adenocarcinoma (PDAC) and is associated with markedly poor survival. Reliable preoperative biomarkers to identify occult hepatic micrometastasis remain lacking. OBJECTIVE: To develop and externally validate a circulating exosomal microRNA (exo-miRNA)-based machine learning model for preoperative detection of occult early-LiM in PDAC. DESIGN, SETTING, AND PARTICIPANTS: This multicenter retrospective case-control study included 3 phases: genome-wide discovery using exo-miRNA sequencing (discovery cohort), model development (training cohort), and independent external validation (2 validation cohorts). The study took place at 4 medical centers in China, Japan, and South Korea. A total of 372 patients were enrolled between 2011 and 2024. Data were analyzed from July 2024 to November 2025. EXPOSURES: Circulating plasma-derived exosomal miRNA expression profiles. MAIN OUTCOMES AND MEASURES: The primary outcome was early-LiM, defined as liver recurrence within 6 months after curative-intent resection. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and survival outcomes were assessed using Kaplan-Meier analysis. RESULTS: Among 372 patients with PDAC (median [IQR] age, 67 [59-73] years; 229 [61.6%] male and 143 [38.4%] female; median follow-up among survivors, 969 days),early-LiM was associated with significantly worse overall survival compared with other recurrence patterns (median OS, 9.1 months vs 26.6-31.8 months; log-rank P < .001). A 7-exo-miRNA extreme gradient boosting model demonstrated discrimination in the training cohort (AUC, 0.899; 95% CI, 0.822-0.976) and maintained performance in external testing cohorts (AUC, 0.876; 95% CI, 0.846-0.951 and AUC, 0.862; 95% CI, 0.744-0.981). The exo-miRNA panel score remained an independent identifier of early-LiM in multivariable analysis (odds ratio, 26.49; 95% CI, 18.45-55.28; P < .001) and stratified overall survival (log-rank P < .001). Decision curve analysis suggested improved net clinical benefit compared with conventional clinicopathologic variables. CONCLUSION AND RELEVANCE: In this multicenter study, a circulating exo-miRNA-based machine learning model enabled preoperative detection of occult early liver metastasis risk in PDAC. These findings support the potential of exosomal biomarkers to inform biology-guided treatment sequencing and warrant prospective validation.

Journal Article↗

PubMed1998

Threats to the validity of clinical trials employing enrichment strategies for sample selection.

Subject selection and exclusion criteria employed in typical clinical effectiveness trials of investigational new drugs have two fundamental aims: (1) to ensure that patients entering a study are truly suffering from the condition the drug is intended to treat and (2) to maximize the likelihood that the study will detect an effect of the drug if, in fact, one exists. Typical protocol selection criteria not only specify exacting procedures for establishing and documenting the diagnosis of those recruited for a study but also seek to increase, relative to the prevalence in the general population, the proportion of individuals in the sample likely to respond to pharmacological treatment. Because it is ordinarily impossible to learn prior to extensive clinical experience with a new drug which, if any, patient characteristics reliably predict a consistent treatment response, strategies for sample "enrichment" typically operate by excluding patients (for example, those with very advanced and/or complicated illness, those with serious concomitant illness, those at the extremes of age, those with very mild illness, and so forth) in whom a dependable response to treatment seems unlikely on logical and/or generic grounds. Some studies use positive strategies for sample "enrichment." In studies evaluating drugs intended to treat recurrent episodes of psychiatric illnesses, many protocols recommend selective recruitment of patients with a history of meaningful positive responses to antipsychotic treatment during prior episodes. Sample selection procedures of these kinds impose limits on the generalizability of a study's results (i.e., external validity), but the use of nonrandom patient samples is ordinarily held to have no effect on the internal validity of the results. In short, studies employing highly selected patient samples are, despite their limited external validity, regularly accepted as valid sources of evidence bearing on a drug's effectiveness. There are exceptions, however; this paper describes one in which the use of a seemingly innocuous sample enrichment maneuver proved highly damaging to the ultimate credibility of an important multicenter trial. In particular, exposure to an experimental treatment during an open qualification phase may invalidate drug-placebo comparisons made during a later randomized, blinded, controlled phase. Our review of the trial also reveals that the enrichment maneuver employed probably failed to accomplish its intended aims, selecting patients whose improvements on the outcome variable may be as reasonably ascribed to chance as to drug effect. This is all the more surprising because the method of sample enrichment employed has much in common with those long recommended in the clinical trial literature.

Aged↗

PubMed2026

Machine learning vs. traditional methods for predicting postoperative cardiac complications after non-cardiac surgery: a systematic review and Bayesian network meta-analysis.

INTRODUCTION: Accurate prediction of peri-operative cardiac complications is critical to optimise pre-operative decision-making. Traditional risk prediction scores, such as the Revised Cardiac Risk Index, show only modest discrimination. Machine learning can model complex, non-linear relationships but their predictive performance compared with traditional scores remains unclear. METHODS: We performed a systematic review and Bayesian network meta-analysis. The primary outcome was postoperative adverse cardiac events following non-cardiac surgery. Prediction models were assessed relative to the Revised Cardiac Risk Index. As many studies evaluated multiple versions of each model type, the highest performing ('best version') and lowest performing ('worst version') results were analysed. Models were ranked using the surface under the cumulative ranking curve (SUCRA). RESULTS: Thirteen studies evaluating 54 models and 927,113 patients were included. Machine learning approaches generally outperformed traditional risk scores. Automated machine learning ranked highest (SUCRA 96.6) showed the greatest improvement in the best version analysis (mean difference (MD) 0.28 (95%CrI 0.16-0.40)) and remained superior in the sensitivity analysis (MD 0.30 (95%CrI 0.14-0.45)). Gradient boosting models showed superior performance over the Revised Cardiac Risk Index across analysis (best version: MD 0.20 (95%CrI 0.14-0.26), worst version: MD 0.18 (95%CrI 0.12-0.25), SUCRA 82.4). The Gupta Perioperative Risk for Myocardial Infarction or Cardiac Arrest score outperformed the Revised Cardiac Risk Index in the best version analysis (MD 0.16 (95%CrI 0.01-0.32)). Between-study heterogeneity was low. None of the included studies externally validated their machine learning models and only six were judged to be at low risk of bias. DISCUSSION: Most machine learning models showed better discrimination than traditional risk scores, with automated machine learning and gradient boosting models ranking highest. However, study quality, calibration reporting and absence of external validation limit immediate clinical adoption. Prospective, multicentre evaluation is required before integration of these models into peri-operative practice.

Humans↗