SEARCH · Search PubMed

Results for “External validity”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

SourceResearch connectionFrom yearThrough year

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

At least 775 records · Page 43Linked to original sources

PubMed2026

Risk prediction models for blood transfusion in patients undergoing total hip and knee arthroplasty: a systematic review and meta-analysis.

OBJECTIVE: To systematically review and evaluate published risk prediction models for perioperative blood transfusion in patients undergoing total hip or knee arthroplasty (THA/TKA). METHODS: We systematically searched PubMed, Web of Science, the Cochrane Library, and Embase from inception to May 31, 2025. Two researchers independently screened the literature, extracted data, and assessed the risk of bias and applicability using the Prediction model Risk Of Bias Assessment Tool (PROBAST). The area under the receiver operating characteristic curve (AUC) values were pooled via a meta-analysis using Stata 18.0. RESULTS: d Fourteen studies containing 36 prediction models were included. The incidence of blood transfusion among THA/TKA patients ranged from 3.2% to 30.8%. Preoperative hemoglobin (Hb) level, tranexamic acid (TXA) use, operative duration, intraoperative blood loss, and age were the most frequently incorporated predictors. Model sensitivity ranged from 58% to 94.5%, and specificity ranged from 71.3% to 94%. Meta-analysis showed that the pooled AUC value of the 13 validated models was 0.87 (95% CI: 0.85-0.90), suggesting good discriminatory performance. All models were rated as having a high risk of bias. The applicability of four studies was rated as unclear. CONCLUSION: Although the included studies demonstrated promising discriminative ability of prediction models for blood transfusion in THA/TKA, all were assessed as having a high risk of bias using the PROBAST tool. Therefore, future research should prioritize the development of models with larger sample sizes, rigorous study designs, and multicenter external validation.

Humans↗

PubMed2003

The multidimensional scale of independent functioning: a new instrument for measuring functional disability in psychiatric populations.

The Multidimensional Scale of Independent Functioning (MSIF) is a new instrument for rating functional disability in psychiatric outpatients. The MSIF differs from other disability rating scales by providing discrete ratings of (1) role responsibility, (2) presence and level of support, and (3) performance quality. The MSIF, which consists of a semistructured interview and detailed rating anchors, was validated in 114 psychiatric outpatients. The instrument had good criterion, discriminative, interrater, and construct validity. Correlations between comparable ratings on the Social Adjustment Scale II (SAS II) ranged from 0.78 to 0.86. Nevertheless, redundancy analysis using canonical correlation demonstrated that, although the two instruments overlap, the MSIF contains information that is not contained in the SAS II. Furthermore, there was only modest shared variance with conceptually non-overlapping subscales in the SAS II. Interrater reliability (intraclass correlation coefficients) ranged from 0.74 to 1.00 for global and subscale scores. MSIF subscales performed as expected with respect to external validators such as hours of employment, earned income, supported versus nonsupported employment and housing, and mainstream versus nonmainstream educational status. MSIF global ratings were modestly correlated with IQ and psychopathology ratings, consistent with reports in the literature. Construct validity, estimated using Cronbach's alpha coefficient, was 0.72. The MSIF is a promising new instrument designed to circumvent several limitations with existing functional outcome instruments for longitudinal studies, intervention research, and services research.

Demography↗

PubMed2026

Metformin Adherence and Risk of Polyneuropathy in Type 2 Diabetes Mellitus: An International Matched Cohort Study with Independent Validation.

BACKGROUND: Metformin is a popular first-line glucose-lowering medication for type 2 diabetes mellitus (T2DM). Although metformin reduces the risks of various complications of diabetes, its potential to cause polyneuropathy by depleting vitamin B12 levels is concerning. This study investigated whether the adherence or discontinuation of metformin after adding-on a second-line antiglycemic agent increases the risk of polyneuropathy in patients with T2DM. METHODS: Data from TriNetX were obtained, and patients with T2DM who were receiving second-line antiglycemic agents were divided into metformin-adherent and metformin-nonadherent groups based on prescription claims data. Neuropathy incidence was evaluated using diagnostic claims and nerve conduction examinations. For independent confirmation and external validation of the primary findings, we used data from the National Health Insurance Research Database (NHIRD) of Taiwan. RESULTS: After matching, 58,027 patients were included in each group. Compared with metformin adherent patients, metformin nonadherent patients had a higher risk of polyneuropathy (adjusted hazard ratios [aHR] 1.26; 95% confidence interval [CI] 1.23-1.29; P < 0.001). Risks of diabetic foot ulcer, amputation, neuropathy-related medication use, and bone fracture were also higher among nonadherent patients. Sensitivity analyses confirmed the robustness of findings. In the validation NHIRD cohort (31,384 matched pairs), metformin nonadherence remained associated with increased polyneuropathy risk (aHR 1.25; 95% CI 1.10-1.42; P < 0.001). CONCLUSIONS: Metformin adherence in patients with T2DM who require second-line treatment may reduce the risk of polyneuropathy; vitamin B supplementation may enhance this benefit.

Humans↗

PubMed2026

Meniscal preservation in the age of biologics: toward a quantitative decision algorithm for personalized repair.

BACKGROUND: Despite advances in arthroscopic repair and biologic augmentation, surgical indication for meniscal tears remains heterogeneous. No standardized framework currently integrates biomechanical, clinical, and biological determinants to guide repair versus resection. PURPOSE: To develop a quantitative decision model-the Meniscal Preservation Score (MPS)-that unifies biomechanical and biological evidence to stratify reparability potential and standardize treatment selection in meniscal surgery. METHODS: A systematic evidence synthesis conducted in accordance with PRISMA 2020 reporting standards of studies published from 2000 to 2025 in PubMed, Embase, and Scopus identified key determinants of meniscal healing. Five consistent predictors-patient age, vascularity, tear morphology, associated pathology, and activity profile-were weighted through a two-round modified Delphi consensus among ten experienced knee surgeons. The resulting 0-9-point MPS was incorporated into a stepwise decision tree linking lesion morphology, biological context, and surgical strategy. Conceptual validation used 50 simulated cases and a retrospective cohort of 45 patients to test agreement between algorithm recommendations and expert surgical decisions. RESULTS: The MPS achieved 86% concordance with expert judgment in simulation and 84% agreement in clinical validation. In this retrospective exploratory cohort, cases in which surgical management was concordant with MPS recommendations demonstrated higher mean IKDC scores at 24 months and lower observed reoperation rates. These findings should be interpreted as associative rather than causal, as treatment allocation was not controlled and discordant cases may have represented inherently more complex pathology. CONCLUSION: The MPS represents an evidence-informed decision-support framework designed to systematize reparability assessment. While exploratory analyses suggest structural coherence with expert reasoning, prospective implementation and external validation are required before clinical adoption as a predictive tool. LEVEL OF EVIDENCE: conceptual model with exploratory validation.

Humans↗

PubMed2000

Athens Insomnia Scale: validation of an instrument based on ICD-10 criteria.

OBJECTIVES: To describe and validate the Athens Insomnia Scale (AIS). METHODS: The AIS is a self-assessment psychometric instrument designed for quantifying sleep difficulty based on the ICD-10 criteria. It consists of eight items: the first five pertain to sleep induction, awakenings during the night, final awakening, total sleep duration, and sleep quality; while the last three refer to well-being, functioning capacity, and sleepiness during the day. Either the entire eight-item scale (AIS-8) or the brief five-item version (AIS-5), which contains only the first five items, can be utilized. The validation of the AIS was based on its administration to 299 subjects: 105 primary insomniacs, 144 psychiatric patients and 50 non-patient controls. RESULTS: Regarding internal consistency, for both versions of the scale, the Cronbach's alpha was around 0. 90 and the mean item-total correlation coefficient was about 0.70. Moreover, in the factor analysis, the scale emerged as a sole component. The test-retest reliability correlation coefficient was found almost 0.90 at a 1-week interval. As far as external validity is concerned, the correlations of the AIS-8 and AIS-5 with the Sleep Problems Scale were 0.90 and 0.85, respectively. CONCLUSION: The high measures of consistency, reliability, and validity of the AIS make it an invaluable tool in sleep research and clinical practice.

Adolescent↗

PubMedDate not supplied

A validation study of the hospital anxiety and depression scale (HADS) in a Spanish population.

The present study aims to validate the Spanish version of the Hospital Anxiety and Depression Scale (HADS) and to determine the use of this tool for screening mood and anxiety disorders. Psychometric properties of the HADS were assessed in different groups of general medical outpatients attending the Hospital Clínic in Barcelona (N=385), and psychiatric diagnoses were made using DSM-IV criteria. A two-factor solution corresponding to the original two subscales of the HADS was found. The Spanish version of the HADS had good internal consistency and external validity, with favorable sensitivity and specificity in identifying cases of psychiatric disorder as defined by the Structured Clinical Interview for DSM-IV (SCID-I). The psychometric properties of the HADS and its brevity make it useful for screening for psychiatric disorders in the medically ill.

Adolescent↗

PubMed2000

Testing the application of a Western scientific theory of AIDS risk behavior among adolescents in Ethiopia.

OBJECTIVE: To test whether a theoretical scheme developed in the United States within the framework of Western science could be applied to a study of HIV risk behaviors among Ethiopian youths. METHODS: Informal interviews and focus group discussions were conducted to determine the relevance of particular AIDS-related risk and protective factors suggested by Jessor's theoretical framework and to generate additional risk and protective factors that were not suggested by a review of the (predominantly Western) literature on adolescent risk behavior. Data from informal interviews and focus group discussions were used to develop survey instruments and procedures for administering survey instruments A pilot study among 99 youths was conducted to examine the reliability and construct validity of the survey instrument. RESULTS: Based on information from focus group discussions and informal interviews, we confirmed the relevance of particular AIDS-related risk and protective factors. The definition of existing constructs was expanded and additional risk and protective factors were incorporated into the existing framework. Existing items and procedures for administering survey instruments were improved and new items were generated. The reliability of survey instruments was determined and improved whenever possible. CONCLUSIONS: We discuss the value of these preliminary steps for securing external validity by identifying theoretical constructs that are relevant to the population at hand and the use of a survey instrument that adequately captures these constructs and provides reliable information.

Acquired Immunodeficiency Syndrome↗

PubMed2005

Medical Record Linkage of Anonymous Registries without Validated Sample Linkage of the Dutch Perinatal Registries.

This paper describes the linkage of data from three Dutch Perinatal Registries: the Dutch National Midwife Registry, the Dutch National Obstetrics Registry and the Dutch National Pediatrics Registry, for the year of 2001. All these registries are anonymous and lack a common identifier. We used probabilistic and deterministic record linkage techniques to combine data from the mother, delivery and child involving to the same pregnancy. Records of singleton and twin pregnancies were linked separately. We have developed a probabilistic close method based on maximum likelihood methods to estimate the weights of individual linking variables and the threshold value for the overall weight. Probabilistic linkage identified 80% more links than a full deterministic linkage approach. External validation revealed an error rate of less than 1%. Our method is a flexible and powerful method to link anonymous registries in the absence of a gold standard.

Female↗

PubMed1999

Factors associated with exercise adherence among older adults. An individual perspective.

This paper reviews the literature concerning factors at the individual level associated with regular exercise among older adults. Twenty-seven cross-sectional and 14 prospective/longitudinal studies met the inclusion criteria of a mean participant age of 65 years or older. The findings are summarised by demographics, exercise experience, exercise knowledge, physiological factors, psychological factors, activity preferences and perceived social influences. In general, education and exercise history correlate positively with regular exercise, while perceived physical frailty and poor health may provide the greatest barrier to exercise adoption and adherence in the elderly. Social-cognitive theories identify several constructs that correlate with the regular exercise behaviour of older adults, such as exercise attitude, perceived behavioural control/self-efficacy, perceived social support and perceived benefits/barriers to continued activity. As well, stage modelling may provide additional information about the readiness for regular exercise behaviour among older adults. However, relatively few studies among older adults exist compared with middle-aged and younger adults. Further, the majority of current research consists of cross-sectional designs or short prospective exercise trials among motivated volunteers that may lack external validity. Future research utilising longitudinal and prospective designs with representative samples of older adults will provide a better understanding of significant causal associations between individual factors and regular exercise behaviour.

Aged↗

PubMed2006

Validation of the European proxy KIDSCREEN-52 pilot test health-related quality of life questionnaire: first results.

PURPOSE: The KIDSCREEN project aims to develop a European cross-cultural generic self-administered Health-Related Quality of Life (HRQoL) instrument for children and adolescents. Proxy measures HRQoL should be a useful and practical alternative to assess children's HRQoL. The KIDSCREEN pilot study involved 3988 children and 2526 child-proxy pairs in seven European countries (Austria, Switzerland, Germany, Spain, France, United Kingdom, and The Netherlands). The proxy instrument is based on the model developed from the children and adolescents reports. The aim of this study is to assess the psychometric properties of the proxy measure in terms of reliability and construct and external validity. METHODS: Confirmatory factor analysis (CFA) of the parent's data allows testing of the multidimensional structure of the proxy measure. Rasch analysis evaluates the scalability of each dimension. The mutltitrait-multimethod (MTMM) model assesses the trait validity through CFA. The agreement between children and proxies reports has been assessed using the Intraclass Correlation Coefficient (ICC). RESULTS: CFA indicates that the children's model is adequate to the parents' data. Reliability is satisfactory for every dimension (CFI = .957). For every dimension, Rasch analysis indicates that items form a unidimensional continuum. MTMM results confirm the trait validity of the instrument. Higher agreement was found for the physical well being dimension (ICC=.52) and school/cognitive functioning (ICC=.52). Adolescents showed higher agreement than the children, and girls higher than boys. Children with physical chronic health conditions showed higher agreement for several domains: physical and psychological well-being, social support, and school environment. CONCLUSIONS: Exploring different facets of validity showed satisfactory results. This new instrument provides a promising measure to further assess the relationships between youth and proxy reports.

Adolescent↗

PubMed2006

Development of a nomogram to predict probability of positive initial prostate biopsy among Japanese patients.

OBJECTIVES: Several nomograms for prostate cancer detection have recently been developed. Because the incidence of prostate cancer is lower among Asian men, nomograms based on Western populations cannot be directly applied to Japanese men. We, therefore, developed a model for predicting the probability of a positive initial prostate biopsy using clinical and laboratory data from a Japanese male population. METHODS: Data were collected from 834 Japanese male referrals who underwent initial prostate biopsies as individual screening. We analyzed age, total prostate-specific antigen (PSA) level, free/total PSA (f/t PSA) ratio, prostate volume, and digital rectal examination findings. Of these data, we randomly reserved 20% for study validation. Logistic regression analysis estimated relative risk, 95% confidence intervals, and P values. RESULTS: Independent predictors of a positive biopsy result included elevated PSA levels, decreased f/T PSA ratio, advanced age, small prostate volume, and abnormal digital rectal examination findings. We developed a predictive nomogram for an initial positive biopsy using these variables. The area under the receiver operating characteristic curve for the model was 81.8%, which was significantly greater than that of the prediction based on PSA alone (area under the receiver operating characteristic curve 67.8%). If externally validated, applying this model could reduce unnecessary biopsy procedures by 32% and reduce the overall need for prostate biopsies by 26%. CONCLUSIONS: In this study of a Japanese population, incorporating clinical and laboratory data into a prebiopsy nomogram significantly improved the prediction of prostate cancer compared with predictions based solely on the individual factors.

Aged↗

PubMed2000

Diagnosis of small-for-gestational-age fetuses between 24 and 32 weeks, based on standard sonographic measurements.

OBJECTIVE: To create and validate a formula using sonographic biometry measurements for the optimal diagnosis of small-for-gestational-age (SGA) fetuses between 24 and 32 weeks of gestation. METHODS: A logistic model using gestational age, femur diaphysis length, abdominal and head circumferences to diagnose SGA was set up in a first group of 64 fetuses born between 24 and 32 weeks (group I). A Receiver Operating Characteristic (ROC) curve was drawn. Our model was compared with standard single ultrasound measurements or combined into an estimated fetal weight (EFW) formula. An external validation was carried out on a second group of 183 fetuses (group II) from another maternity unit (ROC curve and comparisons). RESULTS: The area under the ROC curve was 0.91 in group I and 0.93 in group II. Using a 0.5 cut off point for our model yielded a sensitivity of 76% and specificity of 91% for group I. This model is more specific than most other measurement methods with a similar sensitivity. Using the same cut off point (0.5) in Group II, our model was more specific (98%) but less sensitive (66%) when compared with single ultrasound measurements and EFW formulae. By varying the cut off point, we were able to demonstrate that, for a similar sensitivity, our model had a higher specificity than single ultrasound measurements and had similar specificity to EFW formulae. CONCLUSION: The logistic model we set up was able to calculate an SGA risk score between 24 and 32 weeks of gestation in a population at high risk for elective delivery. The cut off point with a view to diagnosis can vary and makes it possible to give greater importance to the sensitivity or specificity depending on the clinical context.

Abdomen↗

PubMed2005

Large-scale randomised trials--a misguided approach to clinical research.

Large-scale randomised trials influence the clinical management of millions of patients throughout the world and are believed to be the most reliable source of evidence on which to base therapeutic decisions. But do they really deserve the accolades bestowed upon them? The decision to perform these studies implies that the treatment difference is expected to be small and, thus, that large numbers of patients are required to achieve statistical significance. This small treatment difference is a direct consequence of limited knowledge of the subject matter which precludes the formation of homogeneous classes of patients with respect to the outcome. In fact, the majority of patients recruited are redundant in the sense that they would not develop the outcome regardless of treatment and, hence, could not participate in testing the efficacy of the drug in question. The conventional view is that randomisation satisfactorily addresses the issues resulting from heterogeneous study populations. However, the statistical approach to causation--which ignores the fundamental features of causal inference characteristic of both everyday discourse and the scientific method--that is used in large-scale randomised trials fails to deliver reliable generalisations even if the many potential obstacles to internal validity are set aside. Moreover, the results of large-scale randomised trials are not open to independent verification. Given the enormous profits to be made from the long-term treatment of common chronic diseases, the absence of any satisfactory method to detect research fraud is of some concern. Fewer than 5% of patients given treatment on the basis of large-scale randomised trials derive any benefit whatsoever. Arguments based on the benefits of such treatment to the wider community of patients are weakened by the dubious external validity of these studies but, in any case, cannot be used to promote the treatment of individual patients. Interestingly, when patients are provided with detailed information about the size of the benefits, most decline treatment. This is hardly surprising as it is debatable whether or not such meagre treatment effects could have any meaning to an individual patient. Large-scale randomised trials continue to be regarded as the gold standard of clinical research. This, of course, merely reflects the ability of powerful vested interests--in particular the pharmaceutical industry--to defy the sound arguments which demonstrate that the methodology of these studies is deeply flawed. Sooner or later, though, common sense must prevail.

Bias↗

PubMed2026

Large-Scale Plasma Proteomics Enhances Prediction of Liver-Related Events Among Individuals With Prediabetes and Type 2 Diabetes: A Prospective Cohort Study in the UK Biobank.

OBJECTIVE: To develop a protein risk score (ProRS) for predicting liver-related events (LREs) in patients with diabetes and compare its predictive performance with the Fibrosis-4 Index (FIB-4) and an established polygenic risk score. RESEARCH DESIGN AND METHODS: This prospective cohort study included 13 516 individuals with prediabetes and type 2 diabetes (T2D) from the UK Biobank. Cox proportional hazards models and LASSO regression were applied to identify proteins associated with incident LREs and construct the ProRS. Predictive performance was assessed using Harrell's C-index, time-dependent area under the receiver operating characteristic curve, net reclassification improvement and integrated discrimination improvement. RESULTS: Over a median follow-up of 13.5 years, 171 (1.3%) incident LREs occurred. We identified 877 proteins associated with LRE risk, primarily enriched in inflammatory signalling, extracellular matrix remodelling and complement/coagulation cascades. In the training set, we developed a 24-protein ProRS (C-index, 0.842; 95% CI 0.797-0.884) that stratified individuals into low-, medium- and high-risk groups, with 10-year cumulative incidences of LREs of 0.2%, 1.2% and 14.2%, respectively. Compared with the low-risk group, the hazard ratio for LREs was 57.1 (95% CI 31.9-102) in the high-risk group. In the internal validation set, the ProRS model (C-index, 0.876; 95% CI 0.827-0.920) accurately predicted both short- and long-term LREs and outperformed FIB-4 index (C-index, 0.733; 95% CI 0.657-0.807) and polygenic risk score (C-index, 0.636; 95% CI 0.564-0.706). CONCLUSIONS: The protein risk score demonstrated superior performance compared with the FIB-4 index and the polygenic risk score in predicting incident LREs among individuals with prediabetes and T2D. The score allows stratification of individuals according to liver-related risk, though external validation in multi-ethnic cohorts is warranted.

Humans↗

PubMed2000

Estimation of sojourn time in chronic disease screening without data on interval cases.

Estimation of the sojourn time on the preclinical detectable period in disease screening or transition rates for the natural history of chronic disease usually rely on interval cases (diagnosed between screens). However, to ascertain such cases might be difficult in developing countries due to incomplete registration systems and difficulties in follow-up. To overcome this problem, we propose three Markov models to estimate parameters without using interval cases. A three-state Markov model, a five-state Markov model related to regional lymph node spread, and a five-state Markov model pertaining to tumor size are applied to data on breast cancer screening in female relatives of breast cancer cases in Taiwan. Results based on a three-state Markov model give mean sojourn time (MST) 1.90 (95% CI: 1.18-4.86) years for this high-risk group. Validation of these models on the basis of data on breast cancer screening in the age groups 50-59 and 60-69 years from the Swedish Two-County Trial shows the estimates from a three-state Markov model that does not use interval cases are very close to those from previous Markov models taking interval cancers into account. For the five-state Markov model, a reparameterized procedure using auxiliary information on clinically detected cancers is performed to estimate relevant parameters. A good fit of internal and external validation demonstrates the feasibility of using these models to estimate parameters that have previously required interval cancers. This method can be applied to other screening data in which there are no data on interval cases.

Aged↗

PubMed2026

Prognostic value of genes associated with metastasis and propionate metabolism in rectal cancer.

BACKGROUND: Research indicates that alterations in propionate metabolic pathways play a critical role in cancer development and invasion. Postoperative metastatic recurrence remains a major cause of mortality in patients with rectal cancer. However, propionate metabolism-related genes (PMRGs) in rectal cancer remain insufficiently characterized. Therefore, this study aimed to identify prognostic biomarkers associated with lymph node metastasis and propionate metabolism and construct a risk‑prediction model for rectal cancer via bioinformatic analyses. METHODS: The Cancer Genome Atlas-Rectum Adenocarcinoma (TCGA-READ) and GSE87211 datasets, together with a curated PMRGs gene set, were used in this study. Pearson correlation analysis was performed to assess associations between overlapping genes (differentially expressed genes between READ and normal tissues, as well as between N0 and N1-N2 stages) and PMRGs, leading to the identification of candidate genes. Functional enrichment analyses were subsequently conducted to characterize the biological roles of these candidates. Prognostic biomarkers were identified using univariate Cox regression combined with least absolute shrinkage and selection operator (LASSO) regression, and a prognostic model was constructed accordingly. Independent prognostic validation was then performed. In addition, immune checkpoint profiling and immunotherapy response analyses were conducted across risk subgroups. Single-gene Gene Set Enrichment Analysis (GSEA) was applied to elucidate the pathways associated with the identified biomarkers. Finally, drug sensitivity analyses were performed. RESULTS: A total of 157 candidate genes were identified through the analytical pipeline. Functional enrichment analysis indicated that these genes were primarily involved in inflammatory response regulation and tumor necrosis factor (TNF) signaling pathways. Five prognostic biomarkers were subsequently identified and incorporated into a predictive model. External validation using the GSE87211 cohort confirmed the robustness of the model. Risk score and disease status were identified as independent prognostic factors. Six immune checkpoint molecules exhibited differential expression between risk groups. Correlation analyses revealed that the risk score was positively associated with most immune checkpoint genes. Single-gene GSEA demonstrated that the biomarkers were mainly enriched in ribosomal biogenesis and cell adhesion molecule-related pathways. Furthermore, 51 therapeutic agents exhibited significantly different half-maximal inhibitory concentration (IC50) values between risk subgroups. CONCLUSIONS: This study identified five biomarkers (CCL24, IGFBP3, ODC1, PYGM, and VKORC1) associated with lymph node metastasis and propionate metabolism pathways, providing a potential foundation for prognostic prediction in patients with rectal cancer.

Rectal cancer↗

PubMed2026

A machine learning-based predictive model for radiosensitivity in nasopharyngeal carcinoma utilizing serum proteomics.

BACKGROUND: Nasopharyngeal carcinoma (NPC) remains highly sensitive to radiotherapy; however, radioresistance in a subset of patients leads to local recurrence and distant metastasis. Serum proteomics provides a minimally invasive approach to capturing dynamic physiological changes, and machine learning enables efficient construction of predictive models. This study aimed to develop and validate a serum proteomics–based machine-learning model for predicting radiotherapy sensitivity in nasopharyngeal carcinoma (NPC). METHODS: Pretreatment serum samples from newly diagnosed NPC patients were analyzed using SELDI-TOF-MS. Differentially expressed proteins between radiosensitive and radioresistant groups were identified using limma. GO and KEGG analyses were performed to explore functional enrichment. Twelve machine-learning algorithms were used to construct predictive models, and the top-performing models were optimized through feature selection. A Random Forest model with seven features was identified as the optimal model. External validation was performed using an independent cohort with ELISA-quantified protein levels. Model performance was assessed using Receiver operating characteristic curve (ROC), calibration analysis, decision curve analysis (DCA), and 10-fold cross-validation. SHapley Additive exPlanations (SHAP) analysis was applied for model interpretability, and the final model was deployed via a ShinyAPP. RESULTS: A total of 96 differentially expressed proteins were identified, which involved multiple function and signaling pathways. The Random Forest model demonstrated the best predictive performance, achieving an area under the curve (AUC) of 0.963 in the training set and 0.975 in the validation set. Cross-validation yielded an average AUC of 0.965. DCA indicated high clinical utility across a broad threshold range, and calibration curves showed good model agreement. Seven proteins (PLXND1, GSR, PGD, PTPRC, OR2T29, ACTG2, CHAD) were selected as final features. SHAP analysis provided global and individual-level interpretability. A web-based tool was developed to facilitate clinical application. CONCLUSION: This study establishes a robust serum proteomics–based machine-learning model capable of accurately predicting radiotherapy sensitivity in NPC. The model offers clinical interpretability and practical implementation, supporting personalized radiotherapy decision-making.

Humans↗

PubMed1989

Retrospective analyses for hypothesis generation. A commentary on the PACK trial (prevention of atherosclerotic complications with ketanserin).

Publication of the results of a large-scale randomized control trial (RCT) of the anti-serotonin drug ketanserin in patients with intermittent claudication offers an opportunity to examine the validity of retrospective subgroup analyses to generate hypotheses for further validation. This was prompted by an unanticipated adverse interaction which occurred in a subgroup of patients receiving both ketanserin and potassium-losing diuretics. The quality of the study ranked it in the 99th percentile of over 400 RCTs evaluated by a quality scoring system. The subgroup analysis resulted from the emergence during the study of a highly significant excess mortality in the patients on diuretics (relative risks 0.88 for those on ketanserin alone, 0.95 on potassium-sparing and 2.44 for those on potassium-losing diuretics (P = 0.007). External validity was evident from data in the literature indicating that a rare tendency of the many drugs that prolong the QT interval to cause torsade de pointes and fatal arrhythmias is exacerbated by hypokalemia of the degree caused by potassium losing diuretics. When the ketanserin and placebo treated patients also on potassium-losing diuretics are removed from the analyses there is a 23% reduction in endpoints which the power is no longer sufficient to detect as significant. There is also an apparent lag phase. Further study of the drug is clearly indicated.

Adult↗