A wild bootstrap approach for the Aalen-Johansen estimator

Comparison of nonparametric estimators of the expected number of recurrent events

We compare the performance of nonparametric estimators for the mean number of recurrent events and provide a systematic overview for different recurrent event settings. The mean number of recurrent events is an easily interpreted marginal feature often used for treatment comparisons in clinical trials. Incomplete observations, dependencies between successive events, terminating events acting as competing risk, or gaps between at risk periods complicate the estimation. We use survival multistate models to represent different complex recurrent event situations, profiting from recent advances in nonparametric estimation for non-Markov multistate models, and explain several estimators by using multistate intensity processes, including the common Nelson-Aalen-type estimators with and without competing mortality. In addition to building on estimation of state occupation probabilities in non-Markov models, we consider a simple extension of the Nelson-Aalen estimator by allowing for dependence on the number of prior recurrent events. We pay particular attention to the assumptions required for the censoring mechanism, one issue being that some settings require the censoring process to be entirely unrelated while others allow for state-dependent or event-driven censoring. We conducted extensive simulation studies to compare the estimators in various complex situations with recurrent events. Our practical example deals with recurrent chronic obstructive pulmonary disease exacerbations in a clinical study, which will also be used to illustrate two-sample-inference using resampling.

CASANOVA: Permutation inference in factorial survival designs

We propose inference procedures for general factorial designs with time-to-event endpoints. Similar to additive Aalen models, null hypotheses are formulated in terms of cumulative hazards. Deviations are measured in terms of quadratic forms in Nelson-Aalen-type integrals. Different from existing approaches, this allows to work without restrictive model assumptions as proportional hazards. In particular, crossing survival or hazard curves can be detected without a significant loss of power. For a distribution-free application of the method, a permutation strategy is suggested. The resulting procedures' asymptotic validity is proven and small sample performances are analyzed in extensive simulations. The analysis of a data set on asthma illustrates the applicability.

A connection between survival multistate models and causal inference for external treatment interruptions

Recently, treatment interruptions such as a clinical hold in randomized clinical trials have been investigated by using a multistate model approach. The phase III clinical trial START (Stimulating Targeted Antigenic Response To non-small-cell cancer) with primary endpoint overall survival was temporarily placed on hold for enrollment and treatment by the US Food and Drug Administration (FDA). Multistate models provide a flexible framework to account for treatment interruptions induced by a time-dependent external covariate. Extending previous work, we propose a censoring and a filtering approach both aimed at estimating the initial treatment effect on overall survival in the hypothetical situation of no clinical hold. A special focus is on creating a link to causal inference. We show that calculating the matrix of transition probabilities in the multistate model after application of censoring (or filtering) yields the desired causal interpretation. Assumptions in support of the identification of a causal effect by censoring (or filtering) are discussed. Thus, we provide the basis to apply causal censoring (or filtering) in more general settings such as the COVID-19 pandemic. A simulation study demonstrates that both causal censoring and filtering perform favorably compared to a naïve method ignoring the external impact.

Nonparametric tests for multistate processes with clustered data

In this work, we propose nonparametric two-sample tests for population-averaged transition and state occupation probabilities for continuous-time and finite state space processes with clustered, right-censored, and/or left-truncated data. We consider settings where the two groups under comparison are independent or dependent, with or without complete cluster structure. The proposed tests do not impose assumptions regarding the structure of the within-cluster dependence and are applicable to settings with informative cluster size and/or non-Markov processes. The asymptotic properties of the tests are rigorously established using empirical process theory. Simulation studies show that the proposed tests work well even with a small number of clusters, and that they can be substantially more powerful compared to the only, to the best of our knowledge, previously proposed test for this problem. The tests are illustrated using data from a multicenter randomized controlled trial on metastatic squamous-cell carcinoma of the head and neck.

Automated metabolic assignment: Semi-supervised learning in metabolic analysis employing two dimensional Nuclear Magnetic Resonance (NMR)

•

Automatic assignment of metabolites of 2D-TOCSY NMR spectra.

•

Semi-supervised learning for metabolic profiling.

•

Deconvolution and metabolic profiling of 2D NMR spectra using Machine Learning.

•

Accurate Automatic multicomponent assignment of 2D NMR spectrum.

Metabolomics is an expanding field of medical diagnostics since many diseases cause metabolic reprogramming alteration. Additionally, the metabolic point of view offers an insight into the molecular mechanisms of diseases. Due to the complexity of metabolic assignment dependent on the 1D NMR spectral analysis, 2D NMR techniques are preferred because of spectral resolution issues. Thus, in this work, we introduce an automated metabolite identification and assignment from ¹H-¹H TOCSY (total correlation spectroscopy) using real breast cancer tissue. The new approach is based on customized and extended semi-supervised classifiers: KNFST, SVM, third (PC3) and fourth (PC4) degree polynomial. In our approach, metabolic assignment is based only on the vertical and horizontal frequencies of the metabolites in the ¹H–¹H TOCSY. KNFST and SVM show high performance (high accuracy and low mislabeling rate) in relatively low size of initially labeled training data. PC3 and PC4 classifiers showed lower accuracy and high mislabeling rates, and both classifiers fail to provide an acceptable accuracy at extremely low size (≤9% of the entire dataset) of initial training data. Additionally, semi-supervised classifiers were implemented to obtain a fully automatic procedure for signal assignment and deconvolution of TOCSY, which is a big step forward in NMR metabolic profiling. A set of 27 metabolites were deduced from the TOCSY, and their assignments agreed with the metabolites deduced from a 1D NMR spectrum of the same sample analyzed by conventional human-based methodology.

Is there a role for statistics in artificial intelligence?

The research on and application of artificial intelligence (AI) has triggered a comprehensive scientific, economic, social and political discussion. Here we argue that statistics, as an interdisciplinary scientific field, plays a substantial role both for the theoretical and practical understanding of AI and for its future development. Statistics might even be considered a core element of AI. With its specialist knowledge of data evaluation, starting with the precise formulation of the research question and passing through a study design stage on to analysis and interpretation of the results, statistics is a natural partner for other disciplines in teaching, research and practice. This paper aims at highlighting the relevance of statistical methodology in the context of AI development. In particular, we discuss contributions of statistics to the field of artificial intelligence concerning methodological development, planning and design of studies, assessment of data quality and data collection, differentiation of causality and associations and assessment of uncertainty in results. Moreover, the paper also discusses the equally necessary and meaningful extensions of curricula in schools and universities to integrate statistical aspects into AI teaching.

Nonsmooth backfitting for the excess risk additive regression model with two survival time scales

We consider an extension of Aalen’s additive regression model that allows covariates to have effects that vary on two different time scales. The two time scales considered are equal up to a constant for each individual and vary across individuals, such as follow-up time and age in medical studies or calendar time and age in longitudinal studies. The model was introduced in Scheike (2001), where it was solved using smoothing techniques. We present a new backfitting algorithm for estimating the structured model without having to use smoothing. Estimators of the cumulative regression functions on the two time scales are suggested by solving local estimating equations jointly on the two time scales. We provide large-sample properties and simultaneous confidence bands. The model is applied to data on myocardial infarction, providing a separation of the two effects stemming from time since diagnosis and age.

Causal inference methods for small non-randomized studies: Methods and an application in COVID-19

The usual development cycles are too slow for the development of vaccines, diagnostics and treatments in pandemics such as the ongoing SARS-CoV-2 pandemic. Given the pressure in such a situation, there is a risk that findings of early clinical trials are overinterpreted despite their limitations in terms of size and design. Motivated by a non-randomized open-label study investigating the efficacy of hydroxychloroquine in patients with COVID-19, we describe in a unified fashion various alternative approaches to the analysis of non-randomized studies. A widely used tool to reduce the impact of treatment-selection bias are so-called propensity score (PS) methods. Conditioning on the propensity score allows one to replicate the design of a randomized controlled trial, conditional on observed covariates. Extensions include the g-computation approach, which is less frequently applied, in particular in clinical studies. Moreover, doubly robust estimators provide additional advantages. Here, we investigate the properties of propensity score based methods including three variations of doubly robust estimators in small sample settings, typical for early trials, in a simulation study. R code for the simulations is provided.

Nonparametric analysis of nonhomogeneous multistate processes with clustered observations

Frequently, clinical trials and observational studies involve complex event history data with multiple events. When the observations are independent, the analysis of such studies can be based on standard methods for multistate models. However, the independence assumption is often violated, such as in multicenter studies, which makes standard methods improper. This work addresses the issue of nonparametric estimation and two‐sample testing for the population‐averaged transition and state occupation probabilities under general multistate models with cluster‐correlated, right‐censored, and/or left‐truncated observations. The proposed methods do not impose assumptions regarding the within‐cluster dependence, allow for informative cluster size, and are applicable to both Markov and non‐Markov processes. Using empirical process theory, the estimators are shown to be uniformly consistent and to converge weakly to tight Gaussian processes. Closed‐form variance estimators are derived, rigorous methodology for the calculation of simultaneous confidence bands is proposed, and the asymptotic properties of the nonparametric tests are established. Furthermore, I provide theoretical arguments for the validity of the nonparametric cluster bootstrap, which can be readily implemented in practice regardless of how complex the underlying multistate model is. Simulation studies show that the performance of the proposed methods is good, and that methods that ignore the within‐cluster dependence can lead to invalid inferences. Finally, the methods are illustrated using data from a multicenter randomized controlled trial.

Dynamic inference in general nested case-control designs

Nested case-control designs are attractive in studies with a time-to-event endpoint if the outcome is rare or if interest lies in evaluating expensive covariates. The appeal is that these designs restrict to small subsets of all patients at risk just prior to the observed event times. Only these small subsets need to be evaluated. Typically, the controls are selected at random and methods for time-simultaneous inference have been proposed in the literature. However, the martingale structure behind nested case-control designs allows for more powerful and flexible non-standard sampling designs. We exploit that structure to find simultaneous confidence bands based on wild bootstrap resampling procedures within this general class of designs. We show in a simulation study that the intended coverage probability is obtained for confidence bands for cumulative baseline hazard functions. We apply our methods to observational data about hospital-acquired infections.

Relapse- and Immunosuppression-Free Survival after Hematopoietic Stem Cell Transplantation: How Can We Assess Treatment Success for Complex Time-to-Event Endpoints?

In most clinical oncology trials, time-to-first-event analyses are used for efficacy assessment, which often do not capture the entire disease process. Instead, the focus may be on more complex time-to-event endpoints, such as the course of disease after the first event or endpoints occurring after randomization. We propose "relapse- and immunosuppression-free survival" (RIFS) as an innovative and clinically relevant outcome measure for assessing treatment success after hematopoietic stem cell transplant (SCT). To capture the time-dynamic relationship of multiple episodes of immunosuppressive therapy during follow-up, relapse, and nonrelapse mortality, a multistate model was developed. The statistical complexity is that the probability of RIFS is nonmonotonic over time; thus, standard time-to-first-event methodology is inappropriate for formal treatment comparisons. Instead, a generalization of the Kaplan-Meier method was used for probability estimation, and simulation-based resampling was suggested as a strategy for statistical inference. We reanalyzed data from a recently published phase III trial in 201 leukemia patients after SCT. The study evaluated long-term treatment success of standard graft-versus-host disease prophylaxis plus a pretransplant antihuman T-lymphocyte immunoglobulin compared with standard prophylaxis alone. Results suggested that treatment increased the long-term probability of RIFS by approximately 30% during the entire follow-up period, which complements the original findings. This article highlights the importance of complex endpoints in oncology, which provide deeper insight into the treatment and disease process over time. Multistate models combined with resampling are highlighted as a promising tool to evaluate treatment success beyond standard endpoints. An example code is provided in the Supplementary Materials.

Nonparametric tests for transition probabilities in nonhomogeneous Markov processes

This paper proposes nonparametric two-sample tests for the direct comparison of the probabilities of a particular transition between states of a continuous time non-homogeneous Markov process with a finite state space. The proposed tests are a linear nonparametric test, an L ²-norm-based test and a Kolmogorov-Smirnov-type test. Significance level assessment is based on rigorous procedures, which are justified through the use of modern empirical process theory. Moreover, the L ²-norm and the Kolmogorov-Smirnov-type tests are shown to be consistent for every fixed alternative hypothesis. The proposed tests are also extended to more complex situations such as cases with incompletely observed absorbing states and non-Markov processes. Simulation studies show that the test statistics perform well even with small sample sizes. Finally, the proposed tests are applied to data on the treatment of early breast cancer from the European Organization for Research and Treatment of Cancer (EORTC) trial 10854, under an illness-death model.

Bootstrapping complex time-to-event data without individual patient data, with a view toward time-dependent exposures

We consider nonparametric and semiparametric resampling of multistate event histories by simulating multistate trajectories from an empirical multivariate hazard measure. One advantage of our approach is that it does not necessarily require individual patient data, but may be based on published information. This is also attractive for both study planning and simulating realistic real-world event history data in general. The concept extends to left-truncation and right-censoring mechanisms, nondegenerate initial distributions, and nonproportional as well as non-Markov settings. A special focus is on its connection to simulating survival data with time-dependent covariates. For the case of qualitative time-dependent exposures, we demonstrate that our proposal gives a more natural interpretation of how such data evolve over the course of time than many of the competing approaches. The multistate perspective avoids any latent failure time structure and sampling spaces impossible in real life, whereas its parsimony follows the principle of Occam's razor. We also suggest empirical simulation as a novel bootstrap procedure to assess estimation uncertainty in the absence of individual patient data. This is not possible for established procedures such as Efron's bootstrap. A simulation study investigating the effect of liver functionality on survival in patients with liver cirrhosis serves as a proof of concept. Example code is provided.

Confidence bands for multiplicative hazards models: Flexible resampling approaches

We propose new resampling‐based approaches to construct asymptotically valid time‐simultaneous confidence bands for cumulative hazard functions in multistate Cox models. In particular, we exemplify the methodology in detail for the simple Cox model with time‐dependent covariates, where the data may be subject to independent right‐censoring or left‐truncation. We use simulations to investigate their finite sample behavior. Finally, the methods are utilized to analyze two empirical examples with survival and competing risks data.

Factorial analyses of treatment effects under independent right-censoring

This paper introduces new effect parameters for factorial survival designs with possibly right-censored time-to-event data. In the special case of a two-sample design, it coincides with the concordance or Wilcoxon parameter in survival analysis. More generally, the new parameters describe treatment or interaction effects and we develop estimates and tests to infer their presence. We rigorously study their asymptotic properties and additionally suggest wild bootstrapping for a consistent and distribution-free application of the inference procedures. The small sample performance is discussed based on simulation results. The practical usefulness of the developed methodology is exemplified on a data example about patients with colon cancer by conducting one- and two-factorial analyses.