BMC Medical Research Methodology最新文献

Background: Although there are numerous studies exploring predictors of clinical trial failure, no comprehensive review of their methodological specificities and findings exists. We performed a scoping review with the aim of exploring the methodological approaches and findings of studies analysing predictors of clinical trial failure.

Methods: The Ovid Medline and Embase databases were systematically searched from inception to December 13, 2024, for studies employing frequentist statistics or machine learning (ML) approaches to assess predictors of trial failure across multiple clinical trials. A generalized linear model (GLM) was employed to assess the impact of certain methodological factors (failure and non-failure definitions, study types included and trial phases included) on reported failure proportions. To estimate the effects of the predictors included in the model on failure proportions, odds ratios (OR) with 95% confidence interval (95% CI) were calculated from model coefficients.

Results: The literature search identified 17,961 records, 81 of which were included in the review. Most of the studies used Clinicaltrials.gov data (73 studies, 90.1%). Frequentist statistics were used to analyze predictors of trial failure in 73 studies (90.1%), and remaining 8 studies employed ML techniques (9.9%). The GLM showed a 27.5% deviance reduction, indicating that certain methodological factors substantially contribute to observed differences in failure proportions. Studies including trials with both completed and ongoing statuses when calculating failure proportions had lower odds of failure compared to those just including completed statuses (OR = 0.44, 95% CI: 0.29-0.67, p < 0.001).

Conclusions: There has been a recent expansion of ML approaches, potentially signaling the beginning of a paradigm shift. Methodological variations substantially influence reported failure proportions, implicating the need for adoption of standardized definitions of failure and calculation approach. We recommend categorizing terminated and withdrawn studies as failed and completed ones as non-failed.

Background: Clinical trials for rare epilepsies face substantial methodological and ethical challenges. Small and heterogeneous populations, coupled with limited validated outcome measures, often render traditional designs underpowered and unable to capture outcomes that are meaningful to patients and their families. Innovative approaches-such as decentralised, adaptive, and participatory designs-offer potential solutions but have rarely been systematically applied in this context.

Methods: We employed an exploratory three-phase sequential mixed-methods design to co-design a patient-centred clinical trial protocol for rare epilepsies in Australia. An expression-of-interest process recruited 40 participants, equally distributed across four advisory groups: patients, clinicians, researchers, and industry representatives. Phase 1 surveys collected demographic information, trial preferences, digital literacy, and perspectives on participation. Phase 2 comprised semi-structured interviews, analysed using reflexive thematic analysis, to identify themes relevant to trial design. In Phase 3, a consensus-driven process involving four structured online workshops with a multidisciplinary subcommittee translated these findings into a trial protocol recommendation.

Results: Four priorities emerged: (1) decentralised trial models to improve access and inclusion, particularly through home-based care; (2) embedding cultural safety and systems integration to support diversity; (3) prioritising outcomes beyond seizure reduction, including quality of life and patient-reported measures; and (4) improving communication and accessibility through digital innovation. These insights informed recommendations for an ethics-approved protocol emphasising inclusivity, feasibility, and real-world relevance.

Conclusions: This study demonstrates the feasibility of participatory co-design in developing rare epilepsy trial protocols. Embedding patient perspectives and adopting innovative methodologies can enhance scientific rigour, build trust, and strengthen the clinical and policy impact of rare disease research.

Background: Epidemiological studies that rely on biomarkers of exposure typically estimate each subject's exposure from measurements on that individual. If repeated measurements of biomarkers of exposure are obtained on an individual, they are typically averaged. This averaging helps to reduce error from within-person variability if average exposure is a better measure of the biologically effective dose than the instantaneous one. However, these analyses then often ignore the residual within-person variation in the averages of measurements. Not considering this variation can bias effect estimates and lead to inaccurate risk assessment.

Methods: We developed software ("calculators") that help design studies of continuous and binary outcomes that rely on biomarkers of exposure. An independent panel of experts was employed to peer review the models and answer questions regarding their use and best practices for the design of epidemiology studies that utilize biomonitoring data for the exposure assessment.

Results: Web-based tools were developed to estimate the required sample sizes, number of repeated measurements, and the trade-offs between power and bias in simple linear and logistic regression models under classical (independent, additive, normally distributed, homogeneous variance) measurement error assumptions. Application of the calculators was illustrated in case studies of investigation of the associations between urinary levels of bisphenols during pregnancy and fetal growth, and urinary levels of triclosan and neurodevelopment in children. Best practices are recommended for the design of epidemiology studies that utilize biomonitoring data for the exposure assessment.

Conclusions: Calculators have been developed and vetted by a panel of experts. They are designed to estimate sample size (number of individuals sampled and number of samples per individual), power and bias in epidemiological studies that use biomonitoring to assess each subject's exposure in the presence of classical measurement errors. These user-friendly tools account for measurement error and allow researchers to design more accurate and appropriately powered studies, ultimately improving quality of public health research.