Statistics in Medicine最新文献

A dynamic treatment regime (DTR) is a sequence of treatment decision rules tailored to an individual's evolving status over time. In precision medicine, much focus has been placed on finding an optimal DTR which, if followed by everyone in the population, would yield the best outcome on average; and extensive investigations have been conducted from both methodological and applied standpoints. The purpose of this tutorial is to provide readers who are interested in optimal DTRs with a systematic, detailed, but accessible introduction, including the formal definition and formulation of this topic within the framework of causal inference, identification assumptions required to link the causal quantity of interest to the observed data, existing statistical models and estimation methods for learning the optimal regime from the data, and application of these methods to both simulated and real data.

Seasonality plays a crucial role in the transmission dynamics of many infectious diseases, contributing to periodic fluctuations in disease incidence. The previously developed geographically dependent individual-level model (GD-ILM) has been effective in modeling infectious diseases, but does not incorporate seasonal effects, limiting its ability to capture seasonal trends. In this study, we extend the GD-ILM by introducing a seasonally varying transmission component, allowing the model to account for periodic fluctuations in infection risk. Our approach integrates a seasonally forced infection kernel to model periodic changes in transmission rates over time, leading to a novel spatiotemporal kernel. To facilitate efficient and reliable parameter estimation in this high-dimensional setting, we employ the Monte Carlo expectation conditional maximization algorithm. We apply our model to individual-level influenza A data from Manitoba, Canada, examining spatial and seasonal infection patterns to identify high-risk regions and periods, and thus informing targeted intervention strategies. The proposed model's performance is further validated through comprehensive simulation studies. Simulation results confirm that models omitting seasonal components lead to biased spatial parameter estimates under various disease prevalence conditions. To support reproducibility and practical application, we developed the SeasEpi R package publicly available on the comprehensive R archive network (CRAN), which implements the seasonal GD-ILM framework and provides tools for model fitting, simulation, and evaluation. The seasonal GD-ILM offers a more accurate framework for modeling infectious disease transmission by integrating both spatial and seasonal dynamics. It supports more accurate risk assessment and enhances public health responses by enabling timely and location-specific interventions based on seasonal transmission patterns.

Replication is essential to reliable and consistent scientific discovery in high-throughput experiments. Quantifying the replicability of scientific discoveries and identifying sources of irreproducibility have become important tasks for quality control and data integration. In this work we introduce a novel statistical model to measure the reproducibility and replicability of findings from replicate experiments in multi-source studies. Using a nested copula mixture model that characterizes the interdependence between replication experiments both across and within sources, our method quantifies reproducibility and replicability of each candidate simultaneously in a coherent framework. Through simulation studies, an ENCODE ChIP-seq dataset and a SEQC RNA-seq dataset, we demonstrate the effectiveness of our method in diagnosing the source of discordance and improving the reliability of scientific discoveries.

Recently, adaptive seamless phase II/III designs (ASDs) have gained attention because they improve the efficiency of drug development. In an ASD, a phase II trial, which explores the dose-response relationships and identifies treatments for the phase III trial, is combined with a phase III trial, which aims to demonstrate the efficacy and safety of the selected treatment arms in a single trial. This study focused on ASD, which selects treatment groups based on the short-term outcomes observed early in the trial, and involves a confirmatory outcome as the long-term outcome. The method based on the combination test, which considers treatment group selection based on short-term outcomes, tends to be conservative. In other words, it controls the type I error rate more strictly than necessary as the correlation decreases among outcomes in phases II and III. To address this issue, we proposed an adaptive seamless phase II/III randomization test that can appropriately consider the correlation between outcomes based on a randomization distribution, where phase II has a binary outcome and phase III has overall survival. Based on the simulation study, the proposed method improved conservatism owing to the correlation among outcomes and controlled the type I error rate around the nominal level. In addition, the power of this method tended to be higher than that of the method based on the combination test in most scenarios. Overall, the proposed method can increase the probability of trial success compared with conventional phase III designs.

Risk prediction is a key component of survival analysis across various fields, including medicine, public health, economics, engineering, and others. The fundamental concern of risk prediction lies in the joint distribution of risk factors and the time to event. The recent success of survival analysis has already been extended to dynamic risk prediction, which incorporates multiple longitudinal observations into predictive models. However, existing methods often rely on parametric model assumptions or discretely approximate survival functions, potentially introducing more bias in predictions. To address these limitations, we introduce a deep neural network featuring a novel output layer termed the Smooth Monotonic Output Layer (SMOL). This model avoids discretization as well as parametric model assumptions. At its core, SMOL takes a general vector as the input and constructs a monotonic, differentiable function via B-splines. Employing SMOL as the output layer allows for direct, nonparametric estimation of monotonic functions of interest, such as survival and cumulative distribution functions. We performed extensive experiments utilizing data from the Cardiovascular Disease Lifetime Risk Pooling Project (LRPP), which harmonized individual data from multiple longitudinal community-based cardiovascular disease (CVD) studies. Our results demonstrate that the proposed approach achieves state-of-the-art accuracy in predicting individual-level risk for atherosclerotic CVD.

Binary observations are often repeated to improve data quality, creating technical replicates. Several scoring methods are commonly used to infer the actual individual state and obtain a probability for each state. The common practice of averaging replicates has limitations, and alternative methods for scoring and classifying individuals are proposed. Additionally, an indecisive response might be wiser than classifying all individuals based on their replicates in the medical context, where 1 indicates a particular health condition. Building on the inherent limitations of the averaging approach, three alternative methods are examined: the median, maximum penalized likelihood estimation, and a Bayesian algorithm. The theoretical analysis suggests that the proposed alternatives outperform the averaging approach, especially the Bayesian method, which incorporates uncertainty and provides credible intervals. Simulations and real-world medical datasets are used to demonstrate the practical implications of these methods for improving diagnostic accuracy and disease prevalence estimation.

In including random effects to account for dependent observations, the odds ratio interpretation of logistic regression coefficients is changed from population-averaged to subject-specific. This is unappealing in many applications, motivating a rich literature on methods that maintain the marginal logistic regression structure without random effects, such as generalized estimating equations. However, for spatial data, random effect approaches are appealing in providing a full probabilistic characterization of the data that can be used for prediction. We propose a new class of spatial logistic regression models that maintain both population-averaged and subject-specific interpretations through a novel class of bridge processes for spatial random effects. These processes are shown to have appealing computational and theoretical properties, including a scale mixture of normal representation. The new methodology is illustrated with simulations and an analysis of childhood malaria prevalence data in Gambia.

The presence of intermediate confounders, also called recanting witnesses, is a fundamental challenge to the investigation of causal mechanisms in mediation analysis, preventing the identification of natural path-specific effects. Common alternatives (such as randomizational interventional effects) are problematic because they can take non-null values even when there is no mediation for any individual in the population. A promising alternative to natural path-specific effects was outlined in a recent article based on replacing recanting witnesses by draws from their conditional distribution. In this manuscript we formally develop these parameters (which we call recanting twin effects) into a viable alternative to natural effects for mediation analysis in the presence of intermediate confounding. Our contributions include (i) proposing a falsification procedure to test whether the observed data are compatible with intermediate confounding by a given intermediate variable, (ii) showing that recanting twin effects are equal to natural effects at the individual level in the absence of intermediate confounding, (iii) showing that recanting twin effects can be interpreted in agential frameworks such as the recently proposed separable effects, in addition to the non-agential framework in which they were originally outlined, and (iv) developing non-parametric efficiency theory including deriving the efficiency bound and non-parametric efficient estimators that can accommodate high-dimensional confounders through the use of data-adaptive estimation methods. We present an application of the methods to evaluate the role of new-onset anxiety and depressive disorder in explaining the relationship between gabapentin/pregabalin prescription and incident opioid use disorder among Medicaid beneficiaries with chronic pain.

A dynamic treatment regime is a sequence of decision rules that map available history information to a treatment option at each decision point. The optimal dynamic treatment regime seeks to make these decisions to maximize the expected outcome of interest. Most existing methods assume population homogeneity. In many complex applications, ignoring latent heterogeneous structures may compromise estimation, highlighting the necessity of exploring heterogeneous structures during the estimation of optimal treatment regimes. We propose heterogeneous Q-learning that facilitates the estimation of optimal dynamic treatment regimes using a concave pairwise fusion penalized approach. The proposed method employs an alternating direction method of multipliers algorithm to solve the concave pairwise fusion penalized least squares problem in each stage. Simulation studies demonstrate that our proposed method outperforms the standard Q-learning method, and it is further illustrated through a real data analysis from the China Rural Hypertension Control Project (CRHCP) study group.

Phase I clinical trials aim to identify the maximum tolerated dose (MTD), a task that becomes challenging in rare disease due to limited patient recruitment. Traditional dose-finding designs, which assign one dose per patient, require a sufficient sample size that may be infeasible for rare disease trials. To address these limitations, we propose the patient retreat in dose escalation (PRIDE) scheme, which integrates intra-patient dose escalation and considers intra-patient correlations by incorporating random effects into a Bayesian hierarchical framework. We further introduce PRIDE-FA (flexible allocation), an extension of PRIDE with a flexible allocation strategy. By allowing retreated patients to be assigned to any dose level based on trial needs, PRIDE-FA improves resource efficiency, leading to greater reductions in required sample size and trial duration. This paper incorporates random effects into established dose-finding designs, including the calibration-free odds (CFO) design, the Bayesian optimal interval (BOIN) design, and the continual reassessment method (CRM) to account for intra-patient correlations when each patient may receive multiple doses. Simulation studies demonstrate that PRIDE and PRIDE-FA significantly improve the accuracy of MTD selection, reduce required sample size, and shorten trial duration compared to existing dose-finding methods. Together, PRIDE and PRIDE-FA provide a robust and efficient framework for phase I clinical trials with rare diseases.