05 Nov Stata Biostatistics and Epidemiology Virtual Symposium in 2022

1. Clinical Trial Design Simulation
Simple formulae for determining sample size and design operating characteristics are no longer available as clinical trial designs become more complex. Simulation is the preferred tool in the absence of formulae; however, trial statisticians are frequently unaware of how to conduct simulations. Trial simulations in Stata can be accomplished easily by writing many lines of code, but I introduce tacts, a new command that attempts to run the simulations in a single line of code. MAMS, trials with longitudinal outcomes, sample size reestimation, multiple-outcome trials with multiplicity correction, and adaptive randomization are all supported by this command. The command will also be able to collect simulation output and generate a results summary table. A complex and highly flexible syntax is required to handle the complexities of all of these trials.

2. Merlin Joint Models, Multistate Models, and More in Stata
As access to a diverse range of data types and the sheer volume of data increases, so do the challenges in statistics and data science. Multivariate data, which is sometimes measured multiple times and frequently necessitates the ability to model nonlinear relationships and hierarchical structures, is now confronting analysts. I’ll go over the merlin command in this talk, which aims to provide a very general framework for data analysis. Merlin is capable of fitting anything from a simple linear regression model or a Weibull survival model to a three-level logistic mixed-effects model or a multivariate joint model of multiple longitudinal outcomes (of various types) and a recurrent event and survival with nonlinear effects. I’ll use a single dataset to show Merlin’s full range of capabilities before diving into some new multistate survival modelling tricks.

3. Survival by First-Line Treatment Type and Timing of Progression in Patients With Follicular Lymphoma
In follicular lymphoma, disease progression within 24 months has emerged as a popular prognostic marker for overall survival (FL).

While it is clinically significant, it has inherent limitations due to the fixed time point of 24 months and potential variation by treatment type and comparison group choice. I will discuss some of the methodological limitations as well as preliminary findings from a large population-based cohort of FL patients in this talk. National register data was combined with detailed medical record data to create a unique cohort with detailed treatment and follow-up information. Using an illness-death modelling approach, we allow for time-varying progression and estimate relative rates as well as OS by first-line treatment and progression timing. Merlin and multistate Stata packages were used, and example code is provided. Our findings show that progression is associated with poorer survival after the 24-month mark, emphasising the importance of individualised management based on progression timing for the best care of FL patients.

4. Fitting the Cox Proportional Hazards Model to Interval-Censored Data
Interval-censored data are common in clinical, epidemiological, financial, and sociological studies in which the event or failure of interest is not observed at a specific time point but is known to occur within a time interval induced by periodic examinations.

The effects of potentially time-dependent covariates on failure time are expressed using the well-known Cox proportional hazards model, with a completely arbitrary failure time distribution. We consider nonparametric maximum-likelihood estimation with an arbitrary number of examination times for each study subject. We present an EM algorithm that employs very simple calculations and converges stably for any dataset, even those with time-dependent covariates. The resulting estimators for the regression parameters are consistent, asymptotically normal, and asymptotically efficient when using an easily estimated covariance matrix. In addition, we apply the EM algorithm and theoretical results to multivariate failure time data with multiple events per subject or study subject clustering. Finally, we provide examples based on real-world medical research.

5. Stata multivariate interval-censored Cox model
In this presentation, we describe the early results of Stata’s phase II Small Business Innovative Research (SBIR) grant, “Software for Cox Regression Analysis of Interval-Censored Data.” We present mvintcox, a new Stata prototype command for fitting multivariate interval-censored event-time models, and show how it can be used in a variety of real-world scenarios.

6. Stata Network Meta-Analysis: Principles and Applications
There is an urgent need to translate experimental interventions from research settings into clinical practise, which requires comparing the efficacy and safety of different interventions used to treat the same condition and selecting the most appropriate intervention (comparative effectiveness research – CER).

It has traditionally been used for pairwise meta-analysis to compare the effects of treatments based on head-to-head comparisons; however, data from direct comparisons is relatively limited, which has hampered knowledge translation and clinical decision making. Network meta-analysis (NMA) is an alternative analytical approach that includes not only direct comparisons but also indirect comparisons based on logical inference (and assumptions) from the network model in the meta-analysis. This method is becoming more popular because it improves how we use evidence in clinical decision making. In this lecture, we will introduce NMA definitions, relevant statistical concepts, and the frequentist NMA analytics process to be implemented using Stata with a practical example from the fibromyalgia literature.

7. Methodological Concerns and Strategies for Using and Measuring Race in Regression Analysis
There is a wealth of literature on racial and ethnic disparities in health, and the public health field is generally agreed that race is a social construct.

However, the majority of biostatistical and epidemiological researchers are unaware of how their perceptions of race and ethnicity influence their research designs and statistical analyses. Consider how racial and ethnic social constructions influence how biostatisticians and other researchers measure racial and ethnic phenomena, such as racial and ethnic disparities. This talk will go over different coding strategies as well as the potential of nested models for critical studies of race and racism using regression analysis.