BMC Medical Research Methodology最新文献

Background: When analyzing randomized controlled trials (RCTs) data, covariate adjustment is often employed to increase the precision of estimated treatment effects. Missing data in covariates, if not handled properly, can result in biased and inefficient estimates. However, the existing literature on handling missing covariate data is limited, and recommendations vary regarding a valid and efficient approach.

Methods: To help reconcile the seemingly inconsistent recommendations, we address two questions through methodological descriptions and simulated demonstrations. First, how should a multiple imputation (MI) model be specified for RCTs to best preserve the benefit of the randomization design? We consider three different approaches: MI with only baseline variables, "MI overall", and "MI by arm". Second, when and why will simple general strategies, such as grand mean imputation and the missing indicator method, perform as well as or better than MI in estimating treatment effects, and when and why do they fail?

Results: "MI by arm" has the potential to produce unbiased estimates for both the average and subgroup treatment effect (primary and secondary analyses) under the missing at random assumption. Strategies that capitalize on the randomization design, including MI with baseline variables, grand mean imputation, and the missing indicator method, may generate unbiased estimates for the average treatment effect (primary analysis) regardless of the missing data mechanism.

Conclusion: This article clarifies the assumptions and mechanisms by which different missing data strategies accommodate missingness in covariates and reconcile recommendations that sometimes appear contradictory in the literature. Under MAR, "MI by arm" produces unbiased estimates for both the average treatment effect and subgroup treatment effects. Leveraging the randomization design, "baseline-only MI", grand mean imputation, and the missing indicator method produce unbiased estimates for the average treatment effect, but biased subgroup treatment effects, regardless of the missing data mechanism.

Background: Establishing the efficacy of new treatments for rifampicin-resistant tuberculosis (RR-TB) is challenging due to the long-term clinical endpoints of two-year relapse-free survival. This study aimed to evaluate the effect of an alternative indicator of treatment response on sample size requirements and the use of a minimization strategy for randomization.

Methods: Sample size estimates were compared when based on the commonly used endpoint of the proportion of patients achieving stable culture conversion (SCC) at 12 weeks versus a novel but corresponding indicator of treatment response based on a model of changes in mycobacterial load (MBL) over time. The non-linear mixed effects model, calibrated using data from a RR-TB cohort in the same setting, included a longitudinal MBL decline, a probabilistic component for mycobacteria presence in sputum, and a time-to-event model for culture positivity. Data were simulated for a prespecified treatment effect to compare the power of detecting the treatment effect for various sample sizes when using the commonly used endpoint and alternative indicator of treatment response. Additionally, the impact of random patient allocation versus a minimization strategy for randomization on covariate imbalance was assessed.

Results: To achieve 80% power, 410 individuals were needed using the commonly used endpoint versus 110 participants when using the non-linear mixed effects model, corresponding to a 73% reduction in sample size. A small sample size results in high baseline covariate imbalance with random treatment group allocation, with a median relative imbalance of 0.104 for 110 participants versus 0.053 for 410 participants. This imbalance was reduced to 0.036 for 110 participants when an adaptive minimization procedure was implemented.

Conclusion: Using a model of mycobacterial burden changes over time as an alternative indicator of treatment response, combined with a minimization procedure during the randomization process, significantly reduced the sample size which could, if validated, enhance the efficiency of RR-TB clinical trial design.

Background: Quasi-experimental designs are a valid option to assess causal effects of public health interventions when randomized studies are unfeasible, but not widely used in Portugal. We identified and reviewed characteristics of studies employing quasi-experimental designs to evaluate causal effects of public health interventions in Portugal.

Methods: PubMed, Scopus, Web of Science and CINHAL were searched, alongside grey literature, reference mining and contact of authors of eligible studies. We extracted information on the intervention assessed, study design, outcomes assessed, statistical analysis and reporting guidelines.

Results: We identified 1143 studies; 25 were eligible. Studies assessed interventions in various areas, mainly healthcare services (28.0%), drugs/tobacco consumption policy (20.0%), and COVID-19 related restrictions (20.0%). Studies employed interrupted time series (56.0%) and difference-in-differences designs (44.0%). Analyses utilised regression-based models, namely linear (48.0%), negative binominal (20.0%) and logistic (12.0%). Studies analysed 53 outcomes, with two outcomes per study on average. No reporting guidelines were mentioned.

Conclusions: There is a limited number of studies using quasi-experimental designs to estimate the causal effects of public health interventions in Portugal, mainly interrupted time series and difference-in-differences. Training in this area might promote the adequate use and dissemination of quasi-experimental studies.

While platform trials have several benefits with their adaptive features, randomization challenges become of central relevance to the design and execution of a platform trial. This paper intends to address these challenges and explore some potential solutions. A platform type of clinical trial is a clinical trial design where multiple interventions are investigated simultaneously often against partly or fully shared controls, with new treatment arms added and completed treatment arms removed. Unequal allocation is often used in platform trials to improve statistical efficiency, deliver benefits to trial participants, and control the speed of enrollment in different treatment arms. Changes to the allocation ratio may be required after an interim analysis even when the number of treatment arms remains constant, for example, in a platform trial with response-adaptive randomization. To deliver the design efficiencies promised by the carefully optimized allocation ratio or simply to ensure a pre-determined allocation ratio, randomization methods that keep allocation proportions close to the target allocation ratio throughout randomization are helpful. Other situations commonly occurring in platform trials require special considerations for randomization methods and in some cases new classes of randomization methods. Such specific platform features include the requirement to accommodate differences in eligibility for different treatments, the need to ensure partial blinding with a 2-step randomization when mode of administration for different interventions is conspicuously different and full blinding is unfeasible, the objective to balance through dynamic randomization multiple prognostic factors or the need to accommodate limited drug supplies at the numerous trial centers, among others. The key to a successful execution of a complex randomization in the platform trial is the expert design of the Interactive Response Technology (IRT) system, where the system is built at the master protocol level and existing and potential randomization needs are incorporated from the outset. An additional, often overlooked, challenge when working with unequal allocation ratios and randomization methods to attain these, is the importance of preserving the unconditional allocation ratio at every allocation. Failure to do so might lead to a selection and evaluation bias even in double-blind trials, accidental bias, and reduced power of the re-randomization test.

Meta-analysis is a widely used method for synthesizing results from multiple studies across diverse fields. A central challenge in meta-analysis is assessing between-study inconsistency, which can arise from differences in study populations, methodological heterogeneity, or the presence of outliers. Conventional tools such as the [Formula: see text] and [Formula: see text] statistics could be limited in power, especially when the number of studies is small or when the between-study distribution deviates from normality. To address these limitations, we propose a family of alternative [Formula: see text]-like statistics and a hybrid test that adaptively combines their strengths. We also introduce new measures to quantify inconsistency based on these statistics. Simulation studies demonstrate that the hybrid test performs robustly across a wide range of inconsistency patterns, including heavy-tailed, skewed, and contaminated distributions. We further illustrate the practical utility of our methods using three real-world meta-analyses. These approaches offer more flexible and powerful tools for detecting and quantifying inconsistency in meta-analytic practice.

Background: Random survival forests (RSF) have emerged as valuable tools in medical research. They have shown their utility in modelling complex relationships between predictors and survival outcomes, overcoming linearity or low dimensionality assumptions. Nevertheless, RSF have not been adapted to right-censored data with recurrent events (RE).

Methods: This work introduces RecForest, an extension of RSF and tailored for RE data, leveraging principles from survival analysis and ensemble learning. RecForest adapts the splitting rule to account for RE, with or without a terminal event, by employing the pseudo-score test or the Wald test derived from the marginal Ghosh-Lin model. The ensemble estimate is constructed by aggregating the expected number of events from each tree. Performance metrics involve a concordance index (C-index) tailored for RE analysis, along with an extension of the mean squared error (MSE). A comprehensive evaluation was conducted on both simulated and open-source data. We compared RecForest against the non-parametric mean cumulative function and the Ghosh-Lin model.

Results: Across the simulations and application, RecForest consistently outperforms, exhibiting C-index values ranging from 0.60 to 0.82 and lowest MSE metrics.

Conclusions: As analysing time-to-recurrence data is critical in medical research, the proposed method represents a valuable addition to the analytical toolbox in this domain. The RecForest implementation is publicly available as an R package on CRAN.

Background: Master protocols leverage a common trial infrastructure for launching multiple sub-studies. Translational research aims to progress scientific discoveries toward public health impact, which depends on establishing an intervention's efficacy, effectiveness in real-world conditions, and successful strategies for implementation. While master protocols have been designed to improve the efficiency of clinical trials as sub-studies addressing a particular disease, their application with effectiveness-implementation hybrid studies is yet to be explored. The aim of this study was to develop recommendations for adapting mater protocol methods for effectiveness-implementation research.

Methods: A method of consultation with translational research networks was undertaken between January and December 2024. Consideration was given to the requirements for service providers to engage in translational research, and how master protocols could support effectiveness-implementation hybrid sub-studies. The underlying rationale for potential adaptations is provided with reference to implementation frameworks, discussion of advantages and disadvantages, and summary recommendations.

Results: Recommendations are proposed on establishing common trial infrastructure, aims and hypotheses, data collection, control groups, adaptive elements, and eligibility criteria. By leveraging cross-sectoral partnerships, co-producing research and dissemination, and incorporating adaptive elements, master protocols may offer a promising approach for accelerating progress along the translational research pipeline.

Conclusions: The adaptation of master protocols for hybrid sub-studies could enable evidence-based interventions to be more effectively implemented in routine care settings. The feasibility of master protocols for effectiveness-implementation research is yet to be tested, and further development in this area is needed to trial the proposed methodology.

Health expenditure is indicative of the financial burden of health care and serves as a yardstick of health system performance. However, health expenditure may be shaped by multiple factors such as prevalence of morbidity, income inequality and above all, unobserved heterogeneity such as disease severity. This study uses finite mixture models (FMM) to analyze health expenditure distribution based on a National Sample Survey (NSS) which is a nationally representative dataset. This exercise identifies three different class of health care users, acknowledging the heterogeneity within the expenditure distribution. The classes demonstrate variations in spending behavior and associated characteristics. It is observed that health spending is influenced by disease severity, age, gender, education, social group, and economic status. Notably, health expenditure for similar diseases varies significantly across three classes, with the highest expenditure observed in the third latent class. It also reaffirms the gender disparities in health spending irrespective of the class. Additionally, socio-economic status consistently affects health expenditure across classes. These findings underscore the importance of recognizing unobserved heterogeneity in health expenditure for the design of effective healthcare policies. In conclusion, there is a need to recognize the unobserved heterogeneity in health expenditure data and such a recognition that distinct classes within may have greater significance in designing better health care policies. Beyond health expenditure, this analytical framework can be adopted to other medical and public health research to identify the latent classes, thus offering a broader methodological value.