Jerald F. Lawless

Obtaining valid real-world evidence about intervention effects from observational cohorts or administrative health records data is challenging. Visits to healthcare providers tend to occur more often during periods of increased disease activity and symptom exacerbation, or upon disease progression. Treatments likewise tend to change when it is apparent that disease activity has increased or a meaningful progression has occurred. This creates a dual problem in which patient visits are disease-related and treatments changes are driven by disease condition and clinical presentation. Disease-related visits and treatment by indication can produce a biased impression of the disease process in the target population and of the effects of treatment. We discuss how these challenges can be addressed through the use of joint models for the disease, marker, and treatment processes, as well as the observation (visit) process. Using illustrative multistate models, we demonstrate the biases that can arise from various types of analyses and show how estimators from fitting such joint models to persons with psoriatic arthritis can be used to gain scientific insights and address common questions about treatment effects.

In genome-wide association studies (GWAS), it is often desirable to test for interactions, such as gene–environment (G x E) or gene–gene (G x G) interactions, between single-nucleotide polymorphisms (SNPs, G’s) and environmental variables (E’s). However, directly accounting for interaction is often infeasible, because the interacting variable is latent or the computational burden is too large. For quantitative traits (Y) that are approximately normally distributed, it has been shown that indirect testing on GxE can be done by testing for heteroskedasticity of Y between genotypes. However, when traits are binary, the existing methodology based on testing the heteroskedasticity of the trait across genotypes cannot be generalized. In this paper, we propose an approach to indirectly test interaction effects for binary traits and subsequently propose a joint test that accounts for the main and interaction effects of each SNP during GWAS. The final method is straightforward to implement in practice—it simply involves adding a non-additive (i.e., dominance) term to standard GWAS additive models for binary traits and testing its significance. We illustrate the statistical features including type-I-error control and power of the proposed method through extensive numerical studies. Applying our method to the UK Biobank dataset, we showcase the practical utility of the proposed method, revealing SNPs and genes with strong potential for latent interaction effects.

Cohort studies of disease processes deal with events and other outcomes that may occur in individuals following disease onset. The particular goals are often the evaluation of interventions and estimation of the effects of risk factors that may affect the disease course. Models and methods of event history analysis and longitudinal data analysis provide tools for understanding disease processes, but there are numerous challenges in practice. These are related to the complexity of the disease processes and to the difficulty of recruiting representative individuals and acquiring detailed longitudinal data on their disease course. Our objectives here are to describe some of these challenges and to review methods of addressing them. We emphasize the appeal of multistate models as a framework for understanding both disease processes and the processes governing recruitment of individuals for cohort studies and the collection of data. The use of other observational data sources in order to enhance model fitting and analysis is discussed.

Regression analyses based on transformations of cumulative incidence functions are often adopted when modeling and testing for treatment effects in clinical trial settings involving competing and semi‐competing risks. Common frameworks include the Fine–Gray model and models based on direct binomial regression. Using large sample theory we derive the limiting values of treatment effect estimators based on such models when the data are generated according to multiplicative intensity‐based models, and show that the estimand is sensitive to several process features. The rejection rates of hypothesis tests based on cumulative incidence function regression models are also examined for null hypotheses of different types, based on which a robustness property is established. In such settings supportive secondary analyses of treatment effects are essential to ensure a full understanding of the nature of treatment effects. An application to a palliative study of individuals with breast cancer metastatic to bone is provided for illustration.

The HostSeq initiative recruited 10,059 Canadians infected with SARS-CoV-2 between March 2020 and March 2023, obtained clinical information on their disease experience and whole genome sequenced (WGS) their DNA. We analyzed the WGS data for genetic contributors to severe COVID-19 (considering 3,499 hospitalized cases and 4,975 non-hospitalized after quality control). We investigated the evidence for replication of loci reported by the International Host Genetics Initiative (HGI); analyzed the X chromosome; conducted rare variant gene-based analysis and polygenic risk score testing. Population stratification was adjusted for using meta-analysis across ancestry groups. We replicated two loci identified by the HGI for COVID-19 severity: the LZTFL1/SLC6A20 locus on chromosome 3 and the FOXP4 locus on chromosome 6 (the latter with a variant significant at P < 5E-8). We found novel significant associations with MRAS and WDR89 in gene-based analyses, and constructed a polygenic risk score that explained 1.01% of the variance in severe COVID-19. This study provides independent evidence confirming the robustness of previously identified COVID-19 severity loci by the HGI and identifies novel genes for further investigation.

Regression analyses based on transformations of cumulative incidence functions are often adopted when modeling and testing for treatment effects in clinical trial settings involving competing and semi-competing risks. Common frameworks include the Fine-Gray model and models based on direct binomial regression. Using large sample theory we derive the limiting values of treatment effect estimators based on such models when the data are generated according to multiplicative intensity-based models, and show that the estimand is sensitive to several process features. The rejection rates of hypothesis tests based on cumulative incidence function regression models are also examined for null hypotheses of different types, based on which a robustness property is established. In such settings supportive secondary analyses of treatment effects are essential to ensure a full understanding of the nature of treatment effects. An application to a palliative study of individuals with breast cancer metastatic to bone is provided for illustration.

To advance scientific understanding of disease processes and related intervention effects, study results should be free from bias and replicable. More broadly, investigators seek results that are transportable, that is, applicable to a perceived study population as well as in other environments and populations. We review fundamental statistical issues that arise in the analysis of observational data from disease cohorts and other sources and discuss how these issues affect the transportability and replicability of research results. Much of the literature focuses on estimating average exposure or intervention effects at the population level, but we argue for more nuanced analyses of conditional effects that reflect the complexity of disease processes.

Intensity‐based multistate models provide a useful framework for characterizing disease processes, the introduction of interventions, loss to followup, and other complications arising in the conduct of randomized trials studying complex life history processes. Within this framework we discuss the issues involved in the specification of estimands and show the limiting values of common estimators of marginal process features based on cumulative incidence function regression models. When intercurrent events arise we stress the need to carefully define the target estimand and the importance of avoiding targets of inference that are not interpretable in the real world. This has implications for analyses, but also the design of clinical trials where protocols may help in the interpretation of estimands based on marginal features.