Pharmaceutical Statistics最新文献

Appropriately performing Global Sensitivity Analysis (GSA) and refining models through the estimation of key parameters on individual data are fundamental steps in PBPK modeling, yet they remain insufficiently addressed in current practice. The main aim of this work is to establish a computational framework linking PhysPK with Python, enabling the application of the Two-stage δ GSA and individual parameter estimation via the iterative two-stage (ITS) method to the semi-mechanistic PBPK model Phys-DAT. The Two-Stage δ GSA was implemented to assess the impact of parameter uncertainty and correlations on key pharmacokinetic (PK) endpoints, AUC, C_max, and T_max. The most influential parameters were identified for the subsequent individual estimation by using the ITS method. Six simulated scenarios were designed by combining different sampling schedules (rich vs. sparse) and virtual sub-populations (real-case, best-case, worst-case), each reflecting specific variability patterns. Three optimization algorithms (Nelder-Mead, Powell, BFGS) were compared. Estimation performance was evaluated using Average Fold Error (AFE), Absolute AFE (AAFE), and Percentage Estimation Error (PEE). The Two-Stage δ GSA successfully identified the volume of distribution, clearance, and gastric emptying rate constant as the most influential parameters. Overall, estimation performance of the individual PK parameters and PK endpoints was provided. Most estimations yielded AFE and AAFE values between 0.8 and 1.25. Nelder-Mead showed the highest accuracy and precision. Both sampling strategy and individual variability impacted estimation quality. This work demonstrates the feasibility and value of combining correlation-aware GSA with individual parameter estimation in a semi-mechanistic PBPK framework. The integration of the Two-Stage δ GSA into PhysPK represents a major extension of the platform capabilities, providing a powerful tool to guide model simplification through dimensionality reduction of parameter space and support individual parameter estimation, especially under data-constrained conditions.

Motivated by the analysis of data from a clinical trial on patients with early breast cancer, we propose in this paper a new joint model that uses a Tobit partly linear mixed model for longitudinal measurements which are bounded in an interval and have a nonlinear relationship with the observation times and a semiparametric mixture cure model that incorporates a B-spline baseline hazard for survival times with cure proportion. A procedure is developed for estimating parameters in the proposed model using the partial likelihood and Laplace approximation. Additionally, a method of random weighting is proposed to compute the variances of the parameter estimators. The performance of the proposed model and the inference procedures is evaluated through simulation studies and data from the clinical trial that motivated this study.

The purpose of an exploratory clinical trial is to determine whether a new treatment is worth evaluating in subsequent trials. These trials often assessed the efficacy and safety of a single-arm design with binary outcomes. In cancer therapy, time-to-event may be the primary endpoint. In such cases, the frequentist method is typically used to determine sample size. The Bayesian method has recently attracted attention from the perspective of using prior information and introducing early termination criteria. We propose a sample size determination method based on Bayesian power using the posterior probability and prior predictive probability of the hazard ratio (parameter), assuming that the survival functions of historical control and new treatment have proportional hazards. The prior information of the parameter is expressed as the analysis prior and the uncertainty of the parameter is expressed as the design prior. To conduct the clinical trial efficiently, we extended the study design to include early termination criteria. The simulation results showed that the sample size decreased when using an informative analysis prior, whereas it increased when using a design prior that accounted for variance, thus allowing for a more conservative sample size design while taking advantage of the available prior information.

The advancement of precision medicine hinges on accurately tailored diagnostic strategies yet estimating reliable confidence intervals (CIs) for the maximal partial Youden Index under verification bias presents considerable challenges, especially within critical false positive rate (FPR) ranges (e.g., (0, 0.1), (0.05, 0.2)) vital for specific clinical applications. While previous work established the partial Youden Index framework, and methods like full imputation (FI), mean score imputation (MSI), inverse probability weighting (IPW), and semiparametric efficient (SPE) address verification bias, robustly integrating these for the partial index across demanding FPRs has needed further development. This paper significantly advances this area by adapting and applying these four bias-correction techniques to estimate the partial Youden Index and its confidence interval (CIs) under verification bias. We systematically evaluate their performance with the proposed (bootstrap-based, MOVER) CI construction approaches. Extensive simulations demonstrate distinct method-specific patterns across verification proportions and FPR ranges, revealing the complexities in achieving reliable estimates. Bootstrap-based CIs exhibit greater robustness to model misspecification, a common clinical uncertainty, while analytical CIs often face undercoverage issues. A cardiovascular disease biomarker analysis corroborates these findings, showing Blood Pressure's superior discriminatory capability compared to Pulse Rate. Operating under the Missing at Random (MAR) assumption, these results offer crucial, updated guidance for CI estimation in diagnostic studies with incomplete verification, providing significant value where precise evaluation in specific FPR regions is paramount and complete verification is unfeasible. Our findings enhance the statistical foundation for diagnostic test evaluation, extending beyond previous work by comprehensively addressing the partial Youden Index with these updated verification bias correction and CI formula applications.

Clustered competing-risks data often arise in clinical studies, such as multi-center clinical trials, where the occurrence of an event within a cluster hinders the observation of other types of events. The correlation resulting from clustering can be modeled using random effects. These competing-risks data have usually been analyzed using hazard-based models, rather than survival times themselves. Hao et al. proposed a cause-specific joint accelerated failure time (AFT) random-effect modeling approach for analyzing the clustered competing-risks data, which is easy to interpret. In this article, we propose a variable selection method for fixed effects using a penalized h-likelihood (HL) procedure in the joint AFT competing-risk model. Simulation studies were conducted to evaluate the performance of the proposed variable selection procedure, which concluded that the penalized methods of SCAD and HL are more appropriate than that of LASSO. The proposed method is illustrated with two real clinical datasets.

In cluster-correlated data, the number of observations in a cluster can be associated with the outcome from that cluster. This phenomenon is known as informative cluster size which can occur in cluster-randomized clinical trial data. Several studies have found that ignoring the issue of informative cluster size can produce biased results in the analysis of clustered data. Most of the existing methods for addressing informative cluster size are suited to continuous outcomes. However, ordinal outcomes and covariates are often encountered in clustered data obtained from large clinical studies. The existing methods for ordinal association testing in clustered data can produce suboptimal results in the presence of informative cluster size. In this article, we propose a new nonparametric method for testing marginal association between ordinal variables in clustered data that can account for informative cluster size. Through simulated data analyses, we show that our new test outperforms the existing alternatives in accurately identifying significant marginal ordinal associations in the presence of informative cluster size. Even if the cluster size is not informative, the performance of our method is comparable to the existing methods. Additionally, we demonstrate the usefulness of our proposed method through an application to a real-world cluster-randomized clinical trial data.

This manuscript advocates for the implementation of multiple-sequence cross-over designs in early-phase clinical trials by investigating the bias in within-subject variance present in paired and AB/BA cross-over clinical trial designs. While the advantages of adding additional sequences to mitigate confounding effects are well established, the authors noted a lack of mathematical discussion regarding the estimation of random effects in early-phase trials-an important consideration for planning subsequent studies. The manuscript illustrates the importance of multiple-sequence designs by analysing data obtainable from a paired and AB/BA cross-over design for a normally distributed variable. It reveals that the residual mean square error from these two designs serves as an unbiased estimator of within-subject variability only under the rare conditions of no subject-by-treatment interaction and equal variances in both test and reference treatments. This implies that while paired or AB/BA cross-over design might be suitable for early pharmacological studies, it should not be relied upon solely for sample size calculations in late-stage studies due to its limited interpretative potential.

Adjustment for "super" or "prognostic" composite covariates has become more popular in randomized trials recently. These prognostic covariates are often constructed from historical data obtained from previous clinical trials or registries by fitting a predictive model of the outcome on the raw covariates. A natural question that we have been asked by applied researchers is whether this can be done without the historical data: can the prognostic covariate be constructed or derived from the trial data itself, possibly using different folds of the data, before adjusting for it? Here we clarify that such "within-trial" prognostic adjustment is nothing more than a form of targeted maximum likelihood estimation (TMLE), a well-studied procedure that typically improves the power of trial analyses. We therefore argue that there is no reason to reinvent the wheel: within-trial prognostic score adjustment should be referred to as TMLE, without qualification.

Well-controlled clinical trials employ careful processes to reduce bias, often blinding investigators and sponsors to prevent knowledge of study outcomes and potential operational bias. Quality assurance of outcomes is also ensured through designation of unblinded data managers and statisticians, so that complex adaptive designs with multiple interim analyses can be executed. Our approach addresses potential ad-hoc requests by the Data and Safety Monitoring Board (DSMB) for monitoring safety, efficacy, and ethical oversight. A novel approach utilizing current trial data is proposed to predict trial outcomes for blinded decision-makers without unblinding those that should stay blinded. Bayesian predictive power, a trial prediction method, is employed and illustrated on simulated data. This study presents an approach for presenting updated Bayesian predictive power in complex adaptive designs, exemplified by the Hyperbaric Oxygen Brain Injury Treatment (HOBIT) trial. Simulation examples motivated from the trial demonstrate the utility of Bayesian predictive power in predicting trial outcomes and sample size distribution, aiding in resource allocation and decision-making with different reports for blinded and unblinded teams. Bayesian predictive power calculations offer valuable insights into future trial behavior for both blinded and unblinded groups, aiding in guidance during trial conduction. The approach outlined in this short communication can be applied to various trial designs.

For generic drugs, a three-way crossover bioequivalence (BE) study is often used to compare two generic (T) formulations against the common brand-name (R) formulation. Adjustment for multiplicity in equivalence testing, however, is little researched. The new ICH M13A guidance mentioned multiplicity control for equivalence but did not recommend any specific methods. In this paper, we evaluate the applicability of traditional multiplicity adjustment methods in equivalence testing. We propose three revised methods (Bonferroni, Holm, and Hochberg) which are applied on not only p-values but also the more commonly used confidence intervals in equivalence testing. We also apply the 'two-at-a-time' rule as recommended by regulatory agencies and incorporate the correlation among test statistics in simulation. All of these are advances compared to current multiplicity control methods for equivalence. Simulation shows that our proposed methods in a three-way crossover study greatly improve power and reduce needed sample size compared to conducting two two-way crossover studies, control the family-wise error rate at a desired level, and only slightly increase the required sample size compared to no alpha adjustment. Therefore, we recommend our revised Bonferroni, Holm, or Hochberg method in a three-way crossover design when assessing the BE of 2 Ts to 1 R.