Category: Statistics

Scand J Clin Lab Invest. 2026 May 31:1-14. doi: 10.1080/00365513.2026.2681040. Online ahead of print.

ABSTRACT

In large datasets, conventional t-tests may identify statistically significant but practically trivial differences because statistical significance increases with sample size. A log-adjusted t-statistic, defined as an empirical sample-size-aware modification of the classical t-statistic, was evaluated to reduce this oversensitivity. Performance was assessed by Monte Carlo simulations of two-sample comparisons across sample sizes from 10 to 50,000 and effect sizes from δ = 0 to 1.0, and by application to a real clinical laboratory dataset comprising 464,145 participants. In simulations, the ${\text{log}}_{10}\left(\text{Df}\right)$-adjusted statistic showed null rejection rates close to 0.05 across sample sizes, whereas the classical t-test became increasingly oversized at very large n. The adjustment was more conservative for small effects (δ = 0.2-0.4) while high rejection rates were retained for larger effects (δ = 0.6-1.0). In the real-data analysis, several sex differences that were highly significant by the classical t-test had small effect sizes and yielded reference p-values above the conventional 0.05 threshold after adjustment; platelet count (Cohen’s d = 0.13) changed from p < 10^-300 to reference p = 0.052, and potassium (d = 0.05) from p = 10-51 to reference p = 0.104. In contrast, larger effects such as hematocrit (d = 0.83) and HDL cholesterol (d = 0.77) continued to yield reference p-values below that threshold. These reference p-values were compared with the conventional α = 0.05 threshold for illustrative purposes only and were not intended to imply formal Type I error control. These findings suggest that the log-adjusted t-statistic may serve as a useful empirical decision aid for interpreting large clinical laboratory datasets by attenuating sample-size-driven significance while preserving detection of substantively meaningful effects.

PMID:42218778 | DOI:10.1080/00365513.2026.2681040

J Sports Sci. 2026 May 31:1-12. doi: 10.1080/02640414.2026.2673237. Online ahead of print.

ABSTRACT

The shared genetic architecture linking physical activity, sedentary behavior, and osteoporosis risk remains unclear. We investigated the genetic basis, pleiotropic effects, and causal relationships between moderate-to-vigorous physical activity (MVPA), leisure screen time (LST), and heel estimated bone mineral density (eBMD). Leveraging summary statistics from genome-wide association studies of European individuals (MVPA: N = 606,820; LST: N = 526,725; eBMD: N = 426,824), we conducted a genome-wide cross-trait analysis. A significant global genetic correlation was observed for MVPA and eBMD ($r_g$ = 0.13, P = 7.97 × 10^-11), but not for LST and eBMD ($r_g$ = $–$ 0.02, P = 0.34). Two specific genomic regions showed evidence of local genetic correlation. Cross-trait meta-analysis identified 90 pleiotropic loci, of which 20 were novel. Transcriptome-wide association studies revealed 42 shared genes. Mendelian randomization suggested a causal relationship between genetically predicted MVPA and eBMD (beta = 0.07, 95%CIs = 0.01-0.14, P = 0.03), but not for LST (beta = 0.01, 95%CIs = $–$ 0.04-0.05, P = 0.81). Our findings demonstrate a shared genetic basis and pleiotropic effects between MVPA and eBMD, highlighting their intrinsic link and supporting MVPA’s role in osteoporosis prevention.

PMID:42218761 | DOI:10.1080/02640414.2026.2673237

JAMA Health Forum. 2026 May 1;7(5):e261649. doi: 10.1001/jamahealthforum.2026.1649.

ABSTRACT

IMPORTANCE: Nursing home residents with dementia are often unnecessarily hospitalized at the end of life. Institutional Special Needs Plans (I-SNPs) are a type of Medicare Advantage plan for long-term nursing home residents that use advanced practice clinicians to manage care. Studies have demonstrated the effectiveness of the original and largest I-SNP operated by UnitedHealthcare (UHC), but there has been minimal evaluation of non-UHC I-SNPs, which have driven recent growth, nor specific focus on end-of-life outcomes.

OBJECTIVE: To examine the association of I-SNP enrollment with end-of-life outcomes for nursing home residents with dementia, separately for UHC and non-UHC I-SNPs.

DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study used 2010 to 2022 Medicare data on 1.4 million long-stay nursing home residents with dementia who died between 2013 and 2022. Facility-level and patient-level selection bias were addressed with cross-temporal propensity score matching and difference-in-differences models. Both the direct effects of I-SNP enrollment, as well as the indirect (ie, spillover) effects on nonenrollees residing in nursing homes offering I-SNPs were assessed. Variation in these relationships by I-SNP maturity was also examined. Data were analyzed from November 2024 to April 2026.

EXPOSURE: Four I-SNP exposure categories: UHC I-SNP enrollment and spillover; non-UHC I-SNP enrollment and spillover.

MAIN OUTCOMES AND MEASURES: Hospitalization and hospice use in the last month of life.

RESULTS: The study cohort included 1 415 265 long-stay nursing home residents with dementia who died between 2013 and 2022. The unadjusted hospitalization rate in the last 30 days of life was 27.7%. UHC I-SNP enrollment was associated with a 9.0-percentage point (pp) reduction in hospitalization (95% CI, -10.3 pp to -7.7 pp) while non-UHC I-SNP enrollment was associated with a 5.9-pp reduction (95% CI, -8.4 pp to -3.5 pp). The spillover effect on nonenrollees in nursing homes offering a UHC I-SNP was a 1.7-pp (95% CI, -2.4 pp to -1.1 pp) decline in hospitalizations; the spillover effect in non-UHC nursing homes was not statistically significant. Similar trends appeared with hospitalization in the last 3 days of life, intensive care unit admission, and mechanical ventilation, but there was no association with hospice use. The reduction in hospitalizations increased in the 3 years after nursing home I-SNP adoption, then plateaued.

CONCLUSIONS AND RELEVANCE: In this retrospective cohort study, I-SNP enrollment was associated with significantly fewer hospitalizations for nursing home residents with dementia at the end of life, with effect sizes larger for UHC vs non-UHC I-SNPs. Plan maturity and volume are likely important factors impacting success.

PMID:42218737 | DOI:10.1001/jamahealthforum.2026.1649

Brief Bioinform. 2026 May 4;27(3):bbag270. doi: 10.1093/bib/bbag270.

ABSTRACT

Genomic selection leverages genome-wide markers and phenotypes to predict breeding values, with its effectiveness largely dependent on the accuracy of genomic prediction (GP) models. However, GP methods often struggle to capture inter-individual variability and are limited by the curse of dimensionality, where the number of single-nucleotide polymorphisms (SNPs) far exceeds the sample size. To address these challenges, we present CLCNet (Contrastive Learning and Chromosome-aware Network), a novel deep learning framework that integrates contrastive learning and chromosome-aware feature modeling. CLCNet comprises two key components: (i) a contrastive learning module that enhances the model’s ability to capture fine-grained, genotype-dependent phenotypic differences among individuals, and (ii) a chromosome-aware module that captures structured feature selection at both chromosome and genome levels, thereby distilling the most informative SNPs. We evaluated CLCNet across 4 crop species, covering 10 agronomically important traits, and compared it with a diverse set of classical linear, machine learning, and deep learning models. CLCNet achieved superior prediction performance, with statistically significant improvements in Pearson correlation coefficient, ranging from 0.34% to 12.19% over baseline, together with reduced mean squared error. Performance gains were more pronounced for traits with moderate linkage disequilibrium (LD; r2 = 0.21-0.36) and high heritability (h2 > 0.66), such as those in maize, rapeseed, and soybean. For cotton traits characterized by high LD (r2 = 0.74) and lower heritability (h2 < 0.50), CLCNet maintained robust performance without degradation. Overall, these results demonstrate that CLCNet is an effective framework for improving genomic prediction accuracy and holds strong potential for practical applications in plant breeding.

PMID:42218721 | DOI:10.1093/bib/bbag270

Brief Bioinform. 2026 May 4;27(3):bbag262. doi: 10.1093/bib/bbag262.

ABSTRACT

Gene set analysis evaluates the collective impact of groups of genes on an outcome of interest, such as disease occurrence. By incorporating biological knowledge through predefined gene sets, this approach enhances the interpretability of results and improves statistical power compared with gene-wise analyses. In the context of time-to-event data, existing methods are limited and fail to account for potentially strong correlations within gene sets. Given the strong performance of the Generalized Berk-Jones (GBJ) statistic, which effectively incorporates correlation within the test statistic, we adapted this method to the time-to-event framework using a Cox model. We then compared its performance with established methods, including the Cauchy, Harmonic Mean, Wald test, global test, and global boost test. We further benchmarked these methods in two different real-world datasets: gliomas and breast cancer. Our proposed method, sGBJ, shows an overcontrol of Type I error, leading to reduced statistical power compared with other methods in numerical studies particularly when the number of genes is greater than or equal to the number of observations. The Wald test and global boost test generally exhibited the highest power, except in very high-correlation settings for the global boost test, while the Wald test could not adjust for confounders in current implementations.

PMID:42218714 | DOI:10.1093/bib/bbag262

Brief Bioinform. 2026 May 4;27(3):bbag273. doi: 10.1093/bib/bbag273.

ABSTRACT

Quantitative trait locus (QTL) mapping is a powerful approach to uncover the genetic architecture of complex traits. Traditional biparental populations, though widely used, offer limited genetic diversity and resolution, making them less effective for robust genetic architecture dissection. In contrast, multiparent advanced generation intercross (MAGIC) populations possess more allelic diversity, enabling the detection of intricate gene interactions including dominance and epistasis in QTL mapping. However, QTL mapping in MAGIC populations encounters some great challenges, particularly in inferring parental origin of QTL and estimating their effects unbiasedly. We tackled these challenges by developing a novel QTL mapping method for pure-line MAGIC populations, accompanied by the software QTLNetwork-MP. This method extends the mixed linear model framework by integrating a Markov chain-based algorithm and orthogonal QTL effect decomposition. This allows for the estimation of QTL effects, including additive and additive-additive epistasis as well as environmental interactions for MAGIC populations derived from four-way or eight-way crosses. Simulation studies show that QTLNetwork-MP achieves high statistical power and well-controlled false discovery rates across varying heritability, population size, and parent number. We applied QTLNetwork-MP to map QTLs for seed size in an eight-parent recombinant inbred line MAGIC cowpea population, identifying three significant additive QTLs and accurately estimating their genetic effects, validating the value of the method in practical application.

PMID:42218710 | DOI:10.1093/bib/bbag273

Ann Pharmacother. 2026 May 31:10600280261447498. doi: 10.1177/10600280261447498. Online ahead of print.

ABSTRACT

BACKGROUND: Heart failure (HF) with mildly reduced ejection fraction (HFmrEF) and heart failure with preserved ejection fraction (HFpEF) constitute a substantial clinical burden with limited therapeutic options. While sodium-glucose cotransporter 2 (SGLT2) inhibitors are established therapies for heart failure with reduced ejection fraction, their efficacy and safety profile in HFmrEF and HFpEF warrant comprehensive synthesis.

OBJECTIVE: To systematically evaluate the impact of SGLT2 inhibitors on cardiovascular (CV) outcomes, functional capacity, quality of life, and safety in patients with HFmrEF and HFpEF.

METHODS: Eligible studies compared SGLT2 inhibitors with placebo or standard care in patients with left ventricular ejection fraction (LVEF) > 40%. The primary outcome was defined as CV adverse events, primarily including the composite of first hospitalization for heart failure (HHF) or CV mortality (CV death), first HHF, CV death, all-cause mortality. Secondary outcomes included Kansas City Cardiomyopathy Questionnaire (KCCQ) scores, 6-minute walk test distance (6MWTD), and echocardiographic parameters.

RESULTS: Eighteen randomized controlled trials comprising 18 774 patients (SGLT2 inhibitors group: 9564; control group: 9210) were included. Meta-analysis showed that SGLT2 inhibitors significantly reduced the risk of the primary composite endpoint (odds ratio [OR]: 0.71; 95% CI: 0.66-0.75, P < 0.00001) and HHF (OR: 0.69; 95% CI 0.64-0.74; P < 0.00001). However, reductions in CV death (HR: 0.92; 95% CI 0.84-1.00; P = 0.05) and all-cause mortality (OR: 0.92; 95% CI 0.84-1.02; P = 0.10) did not reach statistical significance. Subgroup analyses indicated consistent benefits across New York Heart Association classes, body mass index, and concomitant mineralocorticoid receptor antagonists use, with pronounced efficacy observed in patients with renal impairment (estimated glomerular filtration rate ≥60 mL/min/1.73 m²). Furthermore, SGLT2 inhibitors significantly improved KCCQ-total symptom score, 6MWTD, E/e‘ ratio.

CONCLUSION: SGLT2 inhibitors effectively improve the risk of composite CV death and HHF in patients with HFmrEF and HFpEF, while concurrently improving functional status and quality of life. These findings support the use of SGLT2 inhibitors as a foundational therapy for these populations, although their independent effect on mortality endpoints requires further elucidation.

PMID:42218708 | DOI:10.1177/10600280261447498

Trauma Violence Abuse. 2026 May 31:15248380261451837. doi: 10.1177/15248380261451837. Online ahead of print.

ABSTRACT

Parenting programs are a key strategy for reducing child maltreatment and strengthening parents’ protective factors. However, evidence for their effectiveness remains inconclusive, and few reviews have examined results by outcome domains or program types. This article aims to analyze the effectiveness of parenting programs in preventing or reducing child maltreatment and improving parenting and family dynamics. We conducted a systematic review of randomized controlled trials (RCTs) published between 2013 and 2023, through the Web of Science, Scopus, Education Resources Information Center (ERIC), MEDLINE, and PsycINFO databases. Data synthesis included narrative synthesis and vote counting based on direction of effect, with binomial tests. Twenty-three studies involving 3,997 participants were included. Analysis by outcome domain indicated significant improvements in parenting practices (p = .004), parenting stress and emotional regulation (p = .006), family functioning (p < .001), and parental risk factors (p = .004). No significant effects were detected for child maltreatment reports (p = .063), child abuse potential (p = .219), parenting knowledge and sense of competence (p = .180), or child functioning (p = .063). Regarding program type and theoretical approach, parent training (p < .001) and interventions with cognitive-behavioral therapy (CBT) components (p < .001) showed significant effectiveness, while home visiting (p = .070) and interventions that combined CBT with other approaches (p = .063) did not reach statistical significance. Parenting programs may be effective strategies for preventing child maltreatment, although effects are limited and not always sustained over time. More RCTs with robust designs are needed to strengthen the evidence base and demonstrate effectiveness in protecting children.

PMID:42218702 | DOI:10.1177/15248380261451837

JCO Oncol Pract. 2026 May 31:101200OP2600433. doi: 10.1200/OP-26-00433. Online ahead of print.

ABSTRACT

PURPOSE: A prior national survey of U.S. hematology/oncology (H/O) fellowship curricula demonstrated substantial heterogeneity and limited protected didactic time. Since then, artificial intelligence (AI), including large language models (LLM) and ambient tools, has become increasingly integrated into trainee education and clinical practice. We conducted a multi-center survey to assess the use of AI among H/O fellows.

METHODS: H/O fellows were recruited via program leadership to complete an anonymous survey adapted from our prior study, with added questions on AI education, attitudes, and clinical use. Responses were collected via REDCap and summarized using descriptive statistics.

RESULTS: A total of 118 H/O fellows responded from 18 of 30 invited U.S. H/O fellowship programs (60%), primarily from academic centers (94%), with an even distribution across fellowship training years. Most fellows (74%) reported using AI tools. Other commonly used resources included NCCN guidelines (92%), UpToDate (86%). Only 8% reported receiving formal AI training. Most fellows viewed AI as useful for education (93%) and were confident using it for learning (74%); 92% anticipated increased use and 82% desired formal training. LLMs were most commonly used to clarify concepts (86%), summarize literature (83%), and explore emerging research (75%). AI-assisted documentation was the most frequent clinical application (51%). Reported barriers included (in order of highest concern) accuracy, lack of formal training, data privacy, and unclear ethical or institutional guidelines.

CONCLUSIONS: AI is widely used and valued by current H/O fellows, yet formal training during fellowship remains limited. These findings highlight the need for structured education on effective, safe, and ethical AI use to support clinical integration.

PMID:42218658 | DOI:10.1200/OP-26-00433

Health Informatics J. 2026 Apr-Jun;32(2):14604582261456889. doi: 10.1177/14604582261456889. Epub 2026 May 31.

ABSTRACT

We introduce a novel geostatistical framework to map and predict pedestrian fatalities in Cali, Colombia (2008-2010), addressing critical gaps in research on built environment risk factors particularly in low- and middle-income countries. We triple stack: (1) a grid-based global moving window reducing the modifiable areal unit problem, redistributing events and population; (2) a Poisson-based land use regression incorporating built environment data; and (3) Poisson kriging addressing small number issues. Nine built environment variables were significant predictors of pedestrian fatalities. Health centers had the highest risk increase (54%), while parks and population density were protective. The triple-stack model markedly improved risk prediction, sharpening hotspot delineation and reducing background noise; at high population densities, predictive errors were as low as 0.26 deaths. The framework is a transferable tool for other urban contexts; thus, we provide method decision-making support helping users select between a simplified, codeless, GUI based platform or an advanced, custom-coded approach. Our results demonstrate gains in model predictive accuracy may be primarily attributable to the first two layers, the grid and LUR. Although, Poisson kriging is traditionally used to address the “small number problem,” the first two layers may substantially stabilize rates, resulting in more realistic, interpretable risk maps.

PMID:42218637 | DOI:10.1177/14604582261456889