Category: Statistics

bioRxiv [Preprint]. 2026 Feb 7:2026.02.04.703900. doi: 10.64898/2026.02.04.703900.

ABSTRACT

Detecting a sex difference in response to a treatment or intervention, often reported as a ‘sex-specific effect,’ requires statistical comparison of the response across sex. Here, we investigated analytical approaches used to test for such effects in the behavioral and brain sciences. Of 200 recent articles containing terms such as ‘sex-specific’ or ‘gender-dependent’ in their titles, only 24% presented appropriate evidence supporting the claim: the effect was compared statistically across sex and results consistent with the claim were reported. In most articles (58%), no test was conducted that could have supported the title claim. Only 15% of studies on non-human animals supported the claim with appropriate evidence, which was significantly less frequently than studies on human participants (34%; p = 0.002). The use of appropriate analytical approaches was unrelated to journal rank or the citation impact of the article. We conclude that claims of sex/gender-dependent effects in the behavioral and brain sciences are only infrequently supported by appropriate evidence.

PMID:41676574 | PMC:PMC12889599 | DOI:10.64898/2026.02.04.703900

bioRxiv [Preprint]. 2026 Feb 6:2026.02.03.703561. doi: 10.64898/2026.02.03.703561.

ABSTRACT

Directed evolution is a powerful and widely used technique for protein engineering, and reducing the cost of iterated experimental observations has become a major priority for practitioners. A number of recent efforts to use machine-learning-based predictors to improve sequence selection have led to remarkable improvements in efficiency, but the sparse data at each experimental iteration restricts these approaches to extremely simple models. Adapting large-scale pre-trained protein language models using experimental data offers an alternative that we show productively leverages the strong inductive biases of the natural distribution of protein sequences to navigate high-dimensional, combinatorially large fitness landscapes. Our approach uses a general-purpose “post-training” algorithm grounded in statistical physics that employs quantitative experimental rankings to directly produce a sampler for diverse, high fitness sequences with fewer data points than competing methods. The resulting adapted protein language model can itself be studied and interpreted, shedding further light on the biophysical characteristics of highly fit sequences and their properties.

PMID:41676563 | PMC:PMC12889597 | DOI:10.64898/2026.02.03.703561

bioRxiv [Preprint]. 2026 Feb 6:2026.02.04.703796. doi: 10.64898/2026.02.04.703796.

ABSTRACT

Recent advances in deep learning have led to the development of sequence-to-omics (S2O) models that predict molecular phenotypes directly from DNA sequences. Here, we systematically evaluate the utility of these models, e.g., AlphaGenome, Borzoi, Enformer, and Sei, for improving the reproducibility of genetic fine-mapping across expression quantitative trait loci (eQTL) datasets from Genotype-Tissue Expression (GTEx), Trans-Omics Precision Medicine (TOPMed), and Multi-Ancestry Analysis of Gene Expression (MAGE) projects. We show that purely statistical fine-mapping often yields high replication failure rates (RFRs), but integrating S2O model predictions substantially reduces RFRs and enhances the accuracy of prioritizing SNPs replicated in other consortia. We describe a generalized framework for functionally informed fine-mapping that combines traditional posterior inclusion probabilities (PIPs) from statistical fine-mapping methods with scores from S2O models to generate functionally informed PIPs (fiPIPs) that improve reproducibility. Our findings demonstrate that S2O models, particularly newer ones like AlphaGenome and Borzoi, enable robust identification of replicated variants across consortia, highlighting their promise for scalable, functionally aware genetic mapping.

PMID:41676556 | PMC:PMC12889643 | DOI:10.64898/2026.02.04.703796

BMC Med. 2026 Feb 11. doi: 10.1186/s12916-026-04623-3. Online ahead of print.

ABSTRACT

BACKGROUND: We tested whether inflammation indexed by soluble tumor necrosis factor receptor-1 (sTNFR1) is related to cognitive decline. We examined serum sTNFR1 with cognition in the Health and Retirement Study (HRS) and cerebrospinal fluid (CSF) sTNFR1 with tau pathology and magnetic resonance imaging (MRI)-based atrophy in the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Finally, we used Mendelian randomization (MR) to assess associations between genetically proxied sTNFR1 and regional brain volumes.

METHODS: Data were from HRS (2016-2020; N = 6028) and ADNI (N = 287). In HRS, serum sTNFR1 was log-transformed (quartiles); in ADNI, CSF sTNFR1 was analyzed. Global cognition included word recall, serial 7 s, and counting backwards. In ADNI, cognition was measured by the Clinical Dementia Rating-Sum of Boxes (CDR-SB); CSF total tau/phosphorylated tau and longitudinal MRI regional volumes were analyzed. Associations were estimated with linear and linear mixed-effects models adjusted for demographic, clinical, and genetic covariates including apolipoprotein E ε4 (APOE ε4). Incident mild cognitive impairment (MCI)/dementia was modeled with cause-specific Cox and Fine-Gray models. Incremental prediction used optimism-corrected change in area under the curve (AUC; ΔAUC), net reclassification improvement (NRI)/integrated discrimination improvement (IDI), calibration, and decision curve analysis. MR used genome-wide association study (GWAS) statistics to test effects of genetically proxied sTNFR1 on MRI-derived regional volumes.

RESULTS: In HRS (follow-up 4 years), higher serum sTNFR1 was associated with lower baseline cognition and faster decline in global cognition (β = – 0.16/year). Higher sTNFR1 predicted MCI/dementia (Cox HR ≈ 1.17; Fine-Gray sHR ≈ 1.14); among cognitively normal individuals, risk was elevated (OR = 1.30; 95% CI, 1.03-1.63). Adding sTNFR1 to 2- and 4-year prediction models conferred small discrimination gains after internal validation (ΔAUC ≤ 0.003) and minimal or inconsistent net clinical benefit. In ADNI, higher CSF sTNFR1 was associated with greater CSF total tau and phosphorylated tau, and predicted accelerated caudate atrophy. Exploratory MR suggested a nominal association with reduced right inferior temporal volume, limited by instruments.

CONCLUSIONS: sTNFR1 is associated with cognitive decline and tau-related selective neurodegeneration, but provides limited incremental predictive value beyond established risk factors; external validation and replication are warranted.

PMID:41673623 | DOI:10.1186/s12916-026-04623-3

BMC Public Health. 2026 Feb 11. doi: 10.1186/s12889-026-26315-6. Online ahead of print.

ABSTRACT

BACKGROUND: Oral health during pregnancy is critical for both maternal and neonatal outcomes, yet awareness and preventive behaviors remain suboptimal. This study evaluated the effect of a web-based educational intervention, grounded in the Health Belief Model (HBM), on caries-preventive oral health behaviors in pregnant women.

METHODS: In a quasi-experimental design, 66 pregnant women in Bushehr, Iran, were randomly assigned to intervention and control groups. The intervention group received a multimedia web-based education program based on HBM constructs, while the control group received routine care. Data on knowledge, HBM constructs, and preventive behaviors were collected before and three months after the intervention using validated questionnaires. Statistical analyses included repeated measures ANOVA to compare changes over time between groups.

RESULTS: Post-intervention, the intervention group showed significant improvements in knowledge and most HBM constructs (perceived susceptibility, severity, benefits, barriers, and self-efficacy) compared to controls (p < 0.05). Although preventive oral health behaviors increased significantly within the intervention group (p = 0.022), between-group differences in behavior change were not statistically significant (p = 0.171).

CONCLUSION: The web-based educational program effectively enhanced pregnant women’s knowledge and health beliefs regarding oral health but did not produce a statistically significant improvement in preventive behaviors compared to routine care. Integrating HBM-based web education offers a flexible, cost-effective approach to promote oral health awareness during pregnancy, though further strategies may be needed to translate knowledge gains into sustained behavioral change.

PMID:41673612 | DOI:10.1186/s12889-026-26315-6

BMC Pulm Med. 2026 Feb 11. doi: 10.1186/s12890-025-04038-5. Online ahead of print.

ABSTRACT

BACKGROUND: Alterations in the central nervous system of patients with asthma have been reported. However, the intricate relationship between asthma and cortical structure remains unclear and challenging to determine. To determine the association between asthma and changes in brain cortical structure, we conducted a two-sample Mendelian randomization study.

METHODS: Genome-wide association studies summary data of asthma in 408,442 participants from the UK Biobank were used to identify genetically predicted asthma. Uncorrelated (r² < 0.001) single nucleotide polymorphisms (SNPs) were selected as instrumental variants. Cortical thickness and surface area were obtained from 51,665 patients from the ENIGMA Consortium as outcomes in this study. Inverse-variance weighted was employed as the primary estimate whereas MR Pleiotropy RESidual Sum and Outlier, MR-Egger, and weighted median were used to detect heterogeneity and pleiotropy.

RESULTS: At the global level, no statistically significant association was observed between asthma and global surface area or cortical thickness. In region-specific analyses, nominal significant associations were identified between asthma and the surface area of the parahippocampal gyrus, both with global weighting (β = 5.07 mm², 95% CI: 0.80 to 9.33, P = 0.0198) and without global weighting (β = 5.92 mm², 95% CI: 0.77 to 11.07, P = 0.0242), as well as with the cortical thickness of the caudal anterior cingulate gyrus, both with global weighting (β = -0.01 mm, 95% CI: -0.02 to -0.001, P = 0.0243) and without global weighting (β = -0.01 mm, 95% CI: -0.02 to -0.001, P = 0.0259). However, none of these associations remained statistically significant after adjustment for multiple comparisons.

CONCLUSIONS: This two-sample Mendelian randomization analysis did not identify a statistically significant association between asthma and global cortical metrics. While nominal associations were observed between asthma and changes in parahippocampal and caudal anterior cingulate, these findings did not remain significant after correction for multiple comparisons and should be interpreted as exploratory. Further research is needed to elucidate potential neuroanatomical alterations associated with asthma.

PMID:41673596 | DOI:10.1186/s12890-025-04038-5

BMC Infect Dis. 2026 Feb 11. doi: 10.1186/s12879-026-12836-3. Online ahead of print.

NO ABSTRACT

PMID:41673590 | DOI:10.1186/s12879-026-12836-3

BMC Cardiovasc Disord. 2026 Feb 11. doi: 10.1186/s12872-026-05594-2. Online ahead of print.

ABSTRACT

BACKGROUND: Despite statin therapy, achieving target low-density lipoprotein cholesterol (LDL-C) levels remain suboptimal in high-risk patients with atherosclerotic cardiovascular disease (ASCVD). This study evaluated efficacy and safety of early addition of ezetimibe (EZ) with atorvastatin (AS), prior to reaching the maximally tolerated dose of statin, in very high-risk patients.

METHODS: This phase 4 (NCT05761444), multicenter, randomized, open-label, active-controlled study enrolled patients (≥ 30 years) with very high-risk of ASCVD. Eligible patients had LDL-C ≥ 70 mg/dL with low/moderate intensity statin monotherapy or statin-naïve or not been on stable statin regimen prior to enrollment. Patients were randomized 1:1 to EZ10/AS40 mg combination therapy or AS40 mg statin alone for 12 weeks. Primary endpoint was percentage change in LDL-C from baseline to week 6.

RESULTS: Patients (N = 137) received EZ/AS (n = 67) or AS (n = 70) once a day. The EZ/AS lipid-lowering effect was statistically greater than AS monotherapy at week 6 (LSMD: -21.2; P < 0.0001) and week 12 (LSMD: -16.0; P < 0.0001). At week 12, higher proportions of patients who received EZ/AS achieved target LDL-C < 55 mg/dL (55.0% vs. 15.4%; P < 0.0001) and LDL-C < 70 mg/dL (85.0% vs. 58.5%; P = 0.0009) than in AS group. Higher reduction from baseline was observed for lipid parameters in EZ/AS group than AS monotherapy. Incidence of adverse events were comparable between EZ/AS and AS groups.

CONCLUSIONS: Early combination of EZ with AS, rather than a stepwise approach, significantly reduced LDL-C levels and improved LDL-C reduction target achievement compared to AS monotherapy in very high-risk patients with dyslipidemia with no new safety issues.

TRIAL REGISTRATION: NCT05761444; Registration date: March 9, 2023.

PMID:41673582 | DOI:10.1186/s12872-026-05594-2

BMC Pregnancy Childbirth. 2026 Feb 11. doi: 10.1186/s12884-026-08760-8. Online ahead of print.

ABSTRACT

BACKGROUND: Gestational diabetes mellitus (GDM), a prevalent pregnancy complication, can adversely impact both maternal and neonatal health post-delivery due to poor glycemic control. Although numerous studies highlight the close association between blood lipids and GDM, the specific correlation between lipid profiles and the prognosis of GDM remains ambiguous.

METHODS: This study was conducted at the Women and Children’s Hospital of Ningbo University, with data collected between December 2017 and December 2018. A total of 841 GDM participants were categorized into good glycemic control (GGC, N = 524) and poor glycemic control (PGC, N = 317) groups based on ACOG and ADA criteria. Blood lipid indices were analyzed using univariate and multivariate logistic regression analyses, with further stratification by maternal age, parity, and pre-pregnancy BMI.

RESULTS: In the third trimester, the PGC group exhibited significantly lower levels of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), and high-density lipoprotein cholesterol (HDL-c) (P < 0.05). Multivariate analysis revealed that a 1 mmol/L increase in HDL-c during late pregnancy reduced the risk of PGC by 54% (OR = 0.46, 95% CI = 0.24-0.89), while a higher triglyceride (TG)/HDL-c ratio increased the risk by 23% (OR = 1.23, 95% CI = 1.02-1.48). Stratified analysis confirmed that among women aged ≤ 35 years, multiparas, and those who were overweight or obese, higher HDL-c served as a protective factor, whereas a greater TG/HDL-c ratio posed a significant risk.

CONCLUSION: These findings underscore the critical role of blood lipid metabolism in maintaining glycemic control in GDM, particularly in high-risk populations, such as multiparas and overweight/obese women.

PMID:41673574 | DOI:10.1186/s12884-026-08760-8

BMC Geriatr. 2026 Feb 11. doi: 10.1186/s12877-026-07100-x. Online ahead of print.

ABSTRACT

BACKGROUND: This study evaluates the impact of the SOCAV program on promoting self-direction of individuals with dementia in nursing homes and examines the behavioral changes in care staff when trained and coached to implement the program. SOCAV is a longitudinal training and coaching intervention designed to enhance person-centered care and support self-direction in individuals with dementia.

METHODS: This mixed-methods longitudinal pilot study was conducted in two long-term care units and one day activity center of a nursing home. Participants included 61 individuals with dementia, 48 caregivers, and 85 care staff members. Qualitative data were collected from reflective coaching diaries, while quantitative data were gathered using the Canadian Occupational Performance Measure (COPM) from the perspectives of individuals with dementia, their caregivers, and care staff.

RESULTS: Thematic analysis revealed four positive changes in care staff: acceptance of reflective coaching, focus on self-direction, improvement in communication and interaction, and enhanced team collaboration. Statistically significant improvements in COPM performance and satisfaction were observed in individuals with dementia, as reported by all stakeholders, providing quantitative validation of the observed behavioral changes adopted by the care staff. Initial resistance from staff, linked to fears of criticism and discomfort with change, diminished through supportive peer coaching and emphasis on positive experiences.

CONCLUSIONS: The SOCAV program was associated with meaningful improvements in care staff attitudes and behaviors that support self-direction among individuals with dementia. Key facilitators, including peer coaching, team cohesion, and the realization of staff impact on residents’ self-direction, were instrumental in establishing these changes. Future research is needed to assess potential benefits of the SOCAV program in other settings and populations.

PMID:41673572 | DOI:10.1186/s12877-026-07100-x