Statistical Methods in Medical Research最新文献

Death among subjects is common in observational studies evaluating the causal effects of interventions among geriatric or severely ill patients. High mortality rates complicate the comparison of the prevalence of adverse events between interventions. This problem is often referred to as outcome "truncation" by death. A possible solution is to estimate the survivor average causal effect, an estimand that evaluates the effects of interventions among those who would have survived under both treatment assignments. However, because the survivor average causal effect does not include subjects who would have died under one or both arms, it does not consider the relationship between adverse events and death. We propose a Bayesian method which imputes the unobserved mortality and adverse event outcomes for each participant under the intervention they did not receive. Using the imputed outcomes we define a composite ordinal outcome for each patient, combining the occurrence of death and the adverse event in an increasing scale of severity. This allows for the comparison of the effects of the interventions on death and the adverse event simultaneously among the entire sample. We implement the procedure to analyze the incidence of heart failure among geriatric patients being treated for Type II diabetes with sulfonylureas or dipeptidyl peptidase-4 inhibitors.

Docosahexaenoic acid (DHA) supplementation has proven beneficial in reducing preterm births. However, the challenge lies in addressing nonadherence to prescribed supplementation regimens-a hurdle that significantly impacts clinical trial outcomes. Conventional methods of adherence estimation, such as pill counts and questionnaires, usually fall short when estimating adherence within a specific dosage group. Thus, we propose a Bayesian finite mixture model to estimate adherence among women with low baseline red blood cell phospholipid DHA levels (<6%) receiving higher DHA doses. In our model, adherence is defined as the proportion of participants classified into one of the two distinct components in a normal mixture distribution. Subsequently, based on the estimands from the adherence model, we introduce a novel Bayesian adaptive trial design. Unlike conventional adaptive trials that employ regularly spaced interim schedules, the novelty of our proposed trial design lies in its adaptability to adherence percentages across the treatment arm through irregular interims. The irregular interims in the proposed trial are based on the effect size estimation informed by the finite mixture model. In summary, this study presents innovative methods for leveraging the capabilities of Bayesian finite mixture models in adherence analysis and the design of adaptive clinical trials.

Adaptive enrichment allows for pre-defined patient subgroups of interest to be investigated throughout the course of a clinical trial. These designs have gained attention in recent years because of their potential to shorten the trial's duration and identify effective therapies tailored to specific patient groups. We describe enrichment trials which consider long-term time-to-event outcomes but also incorporate additional short-term information from routinely collected longitudinal biomarkers. These methods are suitable for use in the setting where the trajectory of the biomarker may differ between subgroups and it is believed that the long-term endpoint is influenced by treatment, subgroup and biomarker. Methods are most promising when the majority of patients have biomarker measurements for at least two time points. We implement joint modelling of longitudinal and time-to-event data to define subgroup selection and stopping criteria and we show that the familywise error rate is protected in the strong sense. To assess the results, we perform a simulation study and find that, compared to the study where longitudinal biomarker observations are ignored, incorporating biomarker information leads to increases in power and the (sub)population which truly benefits from the experimental treatment being enriched with higher probability at the interim analysis. The investigations are motivated by a trial for the treatment of metastatic breast cancer and the parameter values for the simulation study are informed using real-world data where repeated circulating tumour DNA measurements and HER2 statuses are available for each patient and are used as our longitudinal data and subgroup identifiers, respectively.

Researchers often use outcome-dependent sampling to study the exposure-outcome association. The case-control study is a widely used example of outcome-dependent sampling when the outcome is binary. When the outcome is ordinal, standard ordinal regression models generally produce biased coefficients when the sampling fractions depend on the values of the outcome variable. To address this problem, we studied the performance of survey-weighted ordinal regression models with weights inversely proportional to the sampling fractions. Through an extensive simulation study, we compared the performance of four ordinal regression models (SM: stereotype model; AC: adjacent-category logit model; CR: continuation-ratio logit model; and CM: cumulative logit model), with and without sampling weights under outcome-dependent sampling. We observed that when using weights, all four models produced estimates with negligible bias of all regression coefficients. Without weights, only stereotype model and adjacent-category logit model produced estimates with negligible to low bias for all coefficients except for the intercepts in all scenarios. In one scenario, the unweighted continuation-ratio logit model also produced estimates with low bias. The weighted stereotype model and adjacent-category logit model also produced estimates with lower relative root mean square errors compared to the unweighted models in most scenarios. In some of the scenarios with unevenly distributed categories, the weighted continuation-ratio logit model and cumulative logit model produced estimates with lower relative root mean square errors compared to the respective unweighted models. We used a study of knee osteoarthritis as an example.

In the search for effective treatments for COVID-19, the initial emphasis has been on re-purposed treatments. To maximize the chances of finding successful treatments, novel treatments that have been developed for this disease in particular, are needed. In this article, we describe and evaluate the statistical design of the AGILE platform, an adaptive randomized seamless Phase I/II trial platform that seeks to quickly establish a safe range of doses and investigates treatments for potential efficacy. The bespoke Bayesian design (i) utilizes randomization during dose-finding, (ii) shares control arm information across the platform, and (iii) uses a time-to-event endpoint with a formal testing structure and error control for evaluation of potential efficacy. Both single-agent and combination treatments are considered. We find that the design can identify potential treatments that are safe and efficacious reliably with small to moderate sample sizes.

We are addressing the problem of estimating conditional average treatment effects with a continuous treatment and a continuous response, using random forests. We explore two general approaches: building trees with a split rule that seeks to increase the heterogeneity of the treatment effect estimation and building trees to predict $Y$ as a proxy target variable. We conduct a simulation study to investigate several aspects including the presence or absence of confounding and colliding effects and the merits of locally centering the treatment and/or the response. Our study incorporates both existing and new implementations of random forests. The results indicate that locally centering both the response and treatment variables is generally the best strategy, and both general approaches are viable. Additionally, we provide an illustration using data from the 1987 National Medical Expenditure Survey.

There is a growing interest in clinical trials that investigate how patients may respond differently to an experimental treatment depending on the basis of some biomarker measured on a continuous scale, and in particular to identify some threshold value for the biomarker above which a positive treatment effect can be considered to have been demonstrated. This can be statistically challenging when the same data are used both to select the threshold and to test the treatment effect in the subpopulation that it defines. This paper describes a hierarchical testing framework to give familywise type I error rate control in this setting and proposes two specific tests that can be used within this framework. One, a simple test based on the estimated value from a linear regression model with treatment by biomarker interaction, is powerful but can lead to type I error rate inflation if the assumptions of the linear model are not met. The other is more robust to these assumptions, but can be slightly less powerful when the assumptions hold.

Clustered current status data frequently occur in many fields of survival studies. Some potential factors related to the hazards of interest cannot be directly observed but are characterized through multiple correlated observable surrogates. In this article, we propose a joint modeling method for regression analysis of clustered current status data with latent variables and potentially informative cluster sizes. The proposed models consist of a factor analysis model to characterize latent variables through their multiple surrogates and an additive hazards frailty model to investigate covariate effects on the failure time and incorporate intra-cluster correlations. We develop an estimation procedure that combines the expectation-maximization algorithm and the weighted estimating equations. The consistency and asymptotic normality of the proposed estimators are established. The finite-sample performance of the proposed method is assessed via a series of simulation studies. This procedure is applied to analyze clustered current status data from the National Toxicology Program on a tumorigenicity study given by the United States Department of Health and Human Services.

Restricted cubic splines (RCS) allow analysts to model nonlinear relations between continuous covariates and the outcome in a regression model. When using RCS with the Cox proportional hazards model, there is no longer a single hazard ratio for the continuous variable. Instead, the hazard ratio depends on the values of the covariate for the two individuals being compared. Thus, using age as an example, when one assumes a linear relation between age and the log-hazard of the outcome there is a single hazard ratio comparing any two individuals whose age differs by 1 year. However, when allowing for a nonlinear relation between age and the log-hazard of the outcome, the hazard ratio comparing the hazard of the outcome between a 31- and a 30-year-old may differ from the hazard ratio comparing the hazard of the outcome between an 81- and an 80-year-old. We describe four methods to describe graphically the relation between a continuous variable and the outcome when using RCS with a Cox model. These graphical methods are based on plots of relative hazard ratios, cumulative incidence, hazards, and cumulative hazards against the continuous variable. Using a case study of patients presenting to hospital with heart failure and a series of mathematical derivations, we illustrate that the four methods will produce qualitatively similar conclusions about the nature of the relation between a continuous variable and the outcome. Use of these methods will allow for an intuitive communication of the nature of the relation between the variable and the outcome.

Predicting patient survival probabilities based on observed covariates is an important assessment in clinical practice. These patient-specific covariates are often measured over multiple follow-up appointments. It is then of interest to predict survival based on the history of these longitudinal measurements, and to update predictions as more observations become available. The standard approaches to these so-called 'dynamic prediction' assessments are joint models and landmark analysis. Joint models involve high-dimensional parameterizations, and their computational complexity often prohibits including multiple longitudinal covariates. Landmark analysis is simpler, but discards a proportion of the available data at each 'landmark time'. In this work, we propose a 'delayed kernel' approach to dynamic prediction that sits somewhere in between the two standard methods in terms of complexity. By conditioning hazard rates directly on the covariate measurements over the observation time frame, we define a model that takes into account the full history of covariate measurements but is more practical and parsimonious than joint modelling. Time-dependent association kernels describe the impact of covariate changes at earlier times on the patient's hazard rate at later times. Under the constraints that our model (a) reduces to the standard Cox model for time-independent covariates, and (b) contains the instantaneous Cox model as a special case, we derive two natural kernel parameterizations. Upon application to three clinical data sets, we find that the predictive accuracy of the delayed kernel approach is comparable to that of the two existing standard methods.