Statistical Methods in Medical Research最新文献

There has been a renewed interest in identifying heterogenous treatment effects (HTEs) to guide personalized medicine. The objective was to illustrate the use of a step-by-step transparent parametric data-adaptive approach (the generalized HTE approach) based on the G-computation algorithm to detect heterogenous subgroups and estimate meaningful conditional average treatment effects (CATE). The following seven steps implement the generalized HTE approach: Step 1: Select variables that satisfy the backdoor criterion and potential effect modifiers; Step 2: Specify a flexible saturated model including potential confounders and effect modifiers; Step 3: Apply a selection method to reduce overfitting; Step 4: Predict potential outcomes under treatment and no treatment; Step 5: Contrast the potential outcomes for each individual; Step 6: Fit cluster modeling to identify potential effect modifiers; Step 7: Estimate subgroup CATEs. We illustrated the use of this approach using simulated and real data. Our generalized HTE approach successfully identified HTEs and subgroups defined by all effect modifiers using simulated and real data. Our study illustrates that it is feasible to use a step-by-step parametric and transparent data-adaptive approach to detect effect modifiers and identify meaningful HTEs in an observational setting. This approach should be more appealing to epidemiologists interested in explanation.

Multi-armed multi-stage designs evaluate experimental treatments using a control arm at interim analyses. Incorporating response-adaptive randomisation in these designs allows early stopping, faster treatment selection and more patients to be assigned to the more promising treatments. Existing frequentist multi-armed multi-stage designs demonstrate that the family-wise error rate is strongly controlled, but they may be too conservative and lack power when the experimental treatments are very different therapies rather than doses of the same drug. Moreover, the designs use a fixed allocation ratio. In this article, Fisher's least significant difference method extended to group-sequential response-adaptive designs is investigated. It is shown mathematically that the information time continues after dropping inferior arms, and hence the error-spending approach can be used to control the family-wise error rate. Two optimal allocations were considered. One ensures efficient estimation of the treatment effects and the other maximises the power subject to a fixed total sample size. Operating characteristics of the group-sequential response-adaptive design for normal and censored survival outcomes based on simulation and redesigning the NeoSphere trial were compared with those of a fixed-sample design. Results show that the adaptive design attains efficient and ethical advantages, and that the family-wise error rate is well controlled.

The stepped wedge design is an appealing longitudinal cluster randomised trial design. However, it places a large burden on participating clusters by requiring all clusters to collect data in all periods of the trial. The staircase design may be a desirable alternative: treatment sequences consist of a limited number of measurement periods before and after the implementation of the intervention. In this article, we explore the relative efficiency of the stepped wedge design to several variants of the 'basic staircase' design, which has one control followed by one intervention period in each sequence. We model outcomes using linear mixed models and consider a sampling scheme where each participant is measured once. We first consider a basic staircase design embedded within the stepped wedge design, then basic staircase designs with either more clusters or larger cluster-period sizes, with the same total number of participants and with fewer total participants than the stepped wedge design. The relative efficiency of the designs depends on the intracluster correlation structure, correlation parameters and the trial configuration, including the number of sequences and cluster-period size. For a wide range of realistic trial settings, a basic staircase design will deliver greater statistical power than a stepped wedge design with the same number of participants, and in some cases, with even fewer total participants.

Survival models with cure fractions, known as long-term survival models, are widely used in epidemiology to account for both immune and susceptible patients regarding a failure event. In such studies, it is also necessary to estimate unobservable heterogeneity caused by unmeasured prognostic factors. Moreover, the hazard function may exhibit a non-monotonic shape, specifically, an unimodal hazard function. In this article, we propose a long-term survival model based on a defective version of the Dagum distribution, incorporating a power variance function frailty term to account for unobservable heterogeneity. This model accommodates survival data with cure fractions and non-monotonic hazard functions. The distribution is reparameterized in terms of the cure fraction, with covariates linked via a logit link, allowing for direct interpretation of covariate effects on the cure fraction-an uncommon feature in defective approaches. We present maximum likelihood estimation for model parameters, assess performance through Monte Carlo simulations, and illustrate the model's applicability using two health-related datasets: severe COVID-19 in pregnant and postpartum women and patients with malignant skin neoplasms.

Pilot and feasibility studies are routinely used to determine whether a definitive trial should be pursued; however, the methodologies used to assess feasibility endpoints are often basic and are rarely informed by the requirements of the planned future trial. We propose a new method for analyzing feasibility outcomes which can incorporate relationships between endpoints, utilize a preliminary study design for a future trial and allow for multiple types of feasibility endpoints. The approach specifies a Joint Feasibility Space (JFS) which is the combination of feasibility outcomes that would render a future trial feasible. We estimate the probability of being in the JFS using Bayesian methods and use simulation to create a decision rule based on frequentist operating characteristics. We compare our approach to other general-purpose methods in the literature with simulation and show that our approach has approximately the same performance when analyzing a single feasibility endpoint but is more efficient with more than one endpoint. Feasibility endpoints should be the focus of pilot and feasibility studies. The analyses of these endpoints deserve more attention than they are given, and we have provided a new, effective method their assessment.

Propensity score estimation is often used as a preliminary step to estimate the average treatment effect with observational data. Nevertheless, misspecification of propensity score models undermines the validity of effect estimates in subsequent analyses. Prediction-based machine learning algorithms are increasingly used to estimate propensity scores to allow for more complex relationships between covariates. However, these approaches may not necessarily achieve covariates balancing. We propose a calibration-based method to better incorporate covariate balance properties in a general modeling framework. Specifically, we calibrate the loss function by adding a covariate imbalance penalty to standard parametric (e.g. logistic regressions) or machine learning models (e.g. neural networks). Our approach may mitigate the impact of model misspecification by explicitly taking into account the covariate balance in the propensity score estimation process. The empirical results show that the proposed method is robust to propensity score model misspecification. The integration of loss function calibration improves the balance of covariates and reduces the root-mean-square error of causal effect estimates. When the propensity score model is misspecified, the neural-network-based model yields the best estimator with less bias and smaller variance as compared to other methods considered.

Simulation models of smoking behaviour provide vital forecasts of exposure to inform policy targets, estimates of the burden of disease, and impacts of tobacco control interventions. A key element of useful model-based forecasts is a clear picture of uncertainty due to the data used to inform the model, however, assessment of this parameter uncertainty is incomplete in almost all tobacco control models. As a remedy, we demonstrate a Bayesian approach to model calibration that quantifies parameter uncertainty. With a model calibrated to Australian data, we observed that the smoking cessation rate in Australia has increased with calendar year since the late 20th century, and in 2016 people who smoked would quit at a rate of 4.7 quit-events per 100 person-years (90% equal-tailed interval (ETI): 4.5-4.9). We found that those who quit smoking before age 30 years switched to reporting that they never smoked at a rate of approximately 2% annually (90% ETI: 1.9-2.2%). The Bayesian approach demonstrated here can be used as a blueprint to model other population behaviours that are challenging to measure directly, and to provide a clearer picture of uncertainty to decision-makers.

In stopping the spread of infectious diseases, pathogen genomic data can be used to reconstruct transmission events and characterize population-level sources of infection. Most approaches for identifying transmission pairs do not account for the time passing since the divergence of pathogen variants in individuals, which is problematic in viruses with high within-host evolutionary rates. This prompted us to consider possible transmission pairs in terms of phylogenetic data and additional estimates of time since infection derived from clinical biomarkers. We develop Bayesian mixture models with an evolutionary clock as a signal component and additional mixed effects or covariate random functions describing the mixing weights to classify potential pairs into likely and unlikely transmission pairs. We demonstrate that although sources cannot be identified at the individual level with certainty, even with the additional data on time elapsed, inferences into the population-level sources of transmission are possible, and more accurate than using only phylogenetic data without time since infection estimates. We apply the proposed approach to estimate age-specific sources of HIV infection in Amsterdam tranamission networks among men who have sex with men between 2010 and 2021. This study demonstrates that infection time estimates provide informative data to characterize transmission sources, and shows how phylogenetic source attribution can then be done with multi-dimensional mixture models.

In this article, we propose bivariate small area estimation methods for proportions based on the logit-normal mixed models with latent spatial dependence. We incorporate multivariate conditional autoregressive structures for the random effects under the hierarchical Bayesian modeling framework, and extend the methods to accommodate non-sampled regions. Posterior inference is obtained via adaptive Markov chain Monte Carlo algorithms. Extensive simulation studies are carried out to demonstrate the effectiveness of the proposed bivariate spatial models. The results suggest that the proposed methods are more efficient than the univariate and non-spatial methods in estimation and prediction, particularly when bivariate spatial dependence exists. Practical guidelines for model selection based on the simulation results are provided. We further illustrate the application of our methods by estimating the province-level heart disease rates and dyslipidemia rates among the middle-aged and elderly population in China's 31 mainland provinces in 2020.

We conducted a systematic comparison of statistical methods used for the analysis of time-to-event outcomes under various proportional and non-proportional hazard (NPH) scenarios. Our study used data from recently published oncology trials to compare the Log-rank test, still by far the most widely used option, against some available alternatives, including the MaxCombo test, the Restricted Mean Survival Time difference test, the Generalized Gamma model and the Generalized F model. Power, type I error rate, and time-dependent bias with respect to the survival probability and median survival time were used to evaluate and compare the performance of these methods. In addition to the real data, we simulated three hypothetical scenarios with crossing hazards chosen so that the early and late effects "cancel out" and used them to evaluate the ability of the aforementioned methods to detect time-specific and overall treatment effects. We implemented novel metrics for assessing the time-dependent bias in treatment effect estimates to provide a more comprehensive evaluation in NPH scenarios. Recommendations under each NPH scenario are provided by examining the type I error rate, power, and time-dependent bias associated with each statistical approach.