Journal of Biopharmaceutical Statistics最新文献

The primary purpose of an oncology single-arm trial is to evaluate the effectiveness of anticancer agents and make a go/no-go decision while maintaining patient safety. We propose a flexible Bayesian optimal phase II design with futility and efficacy stopping boundaries for single-arm clinical trials, named the BOP2-FE design. The proposed BOP2-FE design allows for early stopping of efficacy when the observed antitumor effect is sufficiently higher than the null hypothesis value in the interim looks and retains the benefits of the original BOP2 design, such as explicitly controlling the type I error rate while maximizing power, accommodating different types of endpoint, flexible number of interim looks, and stopping boundaries calculated before the start of the trial. Simulation studies show that the BOP2-FE design reduces the total sample size under the alternative hypothesis while strictly controlling the type I error rate and providing a similar statistical power to the original BOP2 design and a higher statistical power than another existing design.

In medical diagnostic studies involving a transitional intermediate stage of disease progression, the Youden index offers a valuable summary measure for evaluating test accuracy across three diagnostic groups. However, ignoring covariate effects may lead to misleading assessments. To address this, we incorporate covariate information using linear regression models with normally distributed errors, enabling maximum likelihood estimation of the covariate-adjusted Youden index and its corresponding optimal cut-off points. We further develop several types of confidence intervals for these parameters, including generalized confidence intervals, Bayesian credible intervals, and bootstrap-based intervals. The finite-sample performance of the proposed estimators and interval procedures is evaluated via Monte Carlo simulations. Finally, we apply our methods to a diabetic dataset to illustrate their practical utility.

When animals are used in a preclinical experiment, ethical concerns may arise regarding animal welfare. The 3Rs principles were developed to guide more humane animal research practices. This article specifically addresses the reduction aspect of the 3Rs. Under our proposed framework, the preclinical experiment is conducted sequentially, and at every stage of the experiment we examine the outcome and decide whether to stop early for efficacy or futility. Compared to traditional methods in the literature, which typically only check for efficacy, the proposed method has the potential to further reduce the number of animals needed in an experiment. The proposed design requires specifying loss functions at every stage of the experiment. These functions may be directly related to the actual cost of conducting the study or can be calibrated to reflect the prior belief that the drug will be effective. Decisions are made based on minimizing the posterior expected loss. We evaluate the design methodology through simulation studies that involve two-arm experiments with either binary or continuous endpoints. Additionally, we also provide examples taken from real preclinical experiments.

The mean survival time (MST) is usually estimated as the area under the curve of the estimated survival function obtained using the Kaplan-Meier method. However, when the maximum observed survival time is censored, the MST cannot be estimated because the survival function does not reach zero. In such cases, parametric and hybrid methods are used to estimate the MST. The parametric method assumes a probability distribution throughout the entire time and has been evaluated in several studies. The hybrid method combines two approaches: it first applies the Kaplan-Meier method up to a specified time point and then extrapolates the survival curve beyond this point using a parametric distribution. Evaluation of the performance of the hybrid method is limited to a few data-generating mechanisms and analysis models. This study evaluated the performance of the parametric and hybrid methods through numerical experiments, assuming nine probability distributions for the data-generating mechanism and 16 analysis models. The bias and root mean square error of the generalized gamma model and the Royston-Parmar models with the log(-log) link function tended to be smaller than those of the other analysis models, even when the assumed probability distribution of the analysis model was inconsistent with that of the data-generating mechanism when the sample size is relatively large. Overall, the performances of the parametric and hybrid methods were comparable across all the data-generating mechanisms.

A randomized clinical trial with multiple experimental groups and one common control group is often used to speed up development to select the best experimental regimen or to increase the chance of success of clinical trials. Most of the time, multiple dose levels of an experimental drug or multiple combinations of one experimental drug with other drugs comprise multiple experimental groups. Because the experimental drug appears in multiple comparisons with a shared control group, multiple testing adjustments to control the family-wise type I error rate are needed. We extend the stepwise over-correction (SOC) method that is applied to a multi-arm trial with a response rate as its endpoint to a multi-arm trial where time to event is the primary endpoint and confidence interval of the hazard ratio determines the statistical significance. We provide the formula of the bias of the maximum treatment effect estimate toward the true treatment effect between the selected experimental group and the shared control group. We aim to use the bias-corrected estimate for the inference of treatment effects in multi-arm trials on the full alpha level and demonstrate a completely new type of reject region. This approach does not require us to split alpha level among the multiple comparisons or to specify the test order ahead of time. The type I error control and the power enhancement of the proposed approach are both held.

Propensity score-integrated Bayesian dynamic borrowing methods offer an effective approach for covariate adjustment when using external data to augment randomized controlled trials (RCTs). However, identifying the correct propensity score model can be challenging due to unknown treatment selection processes, potentially leading to model misspecification and biased estimates. To improve robustness to model misspecification, we propose an innovative Bayesian inference procedure that incorporates multiply robust weights into the construction of informative power priors. Specifically, we specify a set of candidate propensity score models to derive multiply robust weights, balancing covariates between the current data and external data. The weighted external data is then incorporated into the analysis using a Bayesian power prior method. We further extend this approach to leverage multiple external datasets. Simulation studies indicate that when the set of postulated propensity score models include a correctly specified model, the proposed method achieves desirable operating characteristics, including low bias, low root mean squared error (RMSE), controlled type I error rate at the predetermined nominal level, and high statistical power. This method also provides a robust strategy for researchers who may have a difficult time developing or selecting a single propensity score model.

This research explores the application of marginal structural models (MSMs) in evaluating the causal treatment effect of active intervention versus control on overall survival in randomized clinical trials (RCTs) allowing for control arm patients to switch to active intervention after disease progression. When MSMs are applied in RCTs under this type of treatment switching setting, the question of interest and model specifications differ from both observational studies and from RCTs where patients in both arms are permitted to take alternative treatments after disease progression. A violation of structural positivity may result as an undesired consequence if MSM model weights are constructed using data directly from both arms. This research proposes a two-step approach to avoid this issue. Through simulation studies, it is demonstrated that the proposed approach allows for MSM to be used for analyzing survival data to detect causal active treatment effects under this one-way treatment switching setting. Additionally, estimation for the causal effect of the active intervention as the next line (post-disease progression) therapy can also be obtained from the MSM approach. A case study is presented to illustrate the application of MSMs under this type of treatment switching setting.

The overlap coefficient ($\mathrm{OVL}$) quantifies the similarity between two distributions through the overlapping area of their distribution functions. It has been discussed in the literature in a variety of different contexts. One approach for testing the bioequivalence of treatments is to measure the overlap of the distributions of individual responses to therapy. In some situations, covariates can significantly influence distributional overlap. This paper develops a covariate-specific $\mathrm{OVL}$ estimator using linear regression with a possible Box-Cox transformation. Bootstrap-based confidence intervals for the covariate-specific $\mathrm{OVL}$ are proposed and evaluated through extensive simulations. The methodology is illustrated using fingerstick post-prandial blood glucose measurements as a biomarker for diabetes patients adjusted for age.

Youden index is one of the broadly used measurements to assess the accuracy of the diagnostic test under consideration. In real medical diagnostic studies, verification of the true disease status might only be partially available due to ethical and cost considerations, and the drawbacks of gold-standard tests. Therefore, statistical evaluation of the diagnostic accuracy of a test based only on data from subjects with verified disease status is typically biased. Youden indices for the assessment of accuracy and optimal cutoff point(s) selection in diagnostic tests classifying two disease stages and three disease stages have been proposed without considering this verification bias. In this article, we develop novel confidence intervals for three-class Youden index to correct verification bias under the assumption that the true disease status, if missing, is missing at random (MAR). The proposed methods provide a comprehensive guide to dealing with the verification bias in diagnostic test accuracy studies and lead to a better choice of diagnostic tests.