Founded in 2004, PHUSE is an independent, not-for-profit organization, providing the industry with the premier platform for creating and sharing ideas, tools, and standards around data, statistical and reporting technologies; it is a community of professionals who are passionate about advancement of clinical information. Since its inception, PHUSE has expanded from its roots as a conference for European statistical programmers to a global membership organization and platform for the discussion of topics encompassing the work of data managers, biostatisticians, statistical programmers and eClinical IT professionals.
Intego Group has been a proud supporter of PHUSE over the years and this year we are once again headed to Amsterdam for their 2019 EU event! This year we will have 8 presenters delivering 5 presentations between November 10-13th
Below are a list of presenters from Intego Group who will be participating at the event, along with a description of their presentation
Sub-population Detection Using Graph-based Machine Learning
Sergey, Iryna and Kostiantyn will be speaking on exploratory analysis and how it is used for investigating clinical data by building statistical models, defining end-points, and determining significant covariates. The expected outcome is the identification of a sub-population of patients most responsive to treatment under the study. A graph-based approach to visualize complex relationships in clinical datasets can be an effective solution for sub-population detection. In this approach each node on the graph corresponds to a single patient, while similar patients are connected with an edge. As datasets may include a large number of participants, visual exploration on the graph may be challenging or even misleading. This paper describes machine-learning algorithms applied to automatic detection of sub-populations of similar patients using a graph- based community search. The computational experiment was performed on a clinical study with 1,041 participants. A novel approach to Topological Data Analysis was used to extract graphs from the dataset to further perform a community search using several algorithms.
Multiple imputation as a valid way of dealing with missing data
Vadym will be discussing how missing data appears in every study. In terms of clinical trials it could be a potential source of bias. Missing data in clinical trials may emerge due to various reasons, e.g. some patients could be prematurely discontinued from the study or could miss planned visits while remaining in the study. Every reasonable effort should be made to obtain the protocol-required data for all the study assessments that are scheduled for all the enrolled patients.
Multiple imputation provides a useful and effective way for dealing with missing data. This process results in valid statistical inferences that properly reflect the uncertainty due to missing values.
This paper reviews methods for analyzing missing data, including basic approach and applications of multiple imputation techniques. It presents SAS (PROC MI and PROC MIANALYZE) and R (MICE package) procedures for creating multiple imputations for incomplete multivariate data, analyzes and compares results from multiple imputed data sets.
Making data mapping process easier and smarter with SAS and R
Andrii will be discussing how a Case Report Form or CRF is a tool, used in clinical trials to collect the patients’ data and deliver it to the regulatory affairs in a standard format (SDTM). The CRF design may vary, thus every trial involves a unique SDTM mapping process.
This paper describes macros that read the data from the PDF template of the CRF. The macros search for the CRF fields with the list of predefined and fixed values, convert them into SAS datasets using PDFtools package in R, and compare to SDTM controlled terminology data, provided by CDISC. This functionality helps to reduce the time of mapping the raw data.
The result of the macros’ execution is the SAS code that can be directly used for correct mapping of CRF values to the SDTM’s.
The macros can check the SDTM data to make sure they do not contain typos or unexpected values.
May Your Data Pool Become Super Cool
Anastasiia and Daniil will be talking about how regulatory authorities require both ISS and ISE submissions for new drug applications. The idea of this analysis is quite simple: pool data from a number of studies and provide the statistical analysis – which is most likely to have already been done for each of the studies – in order to receive more accurate statistical results. At first glance, it does not sound like a difficult task, so the timelines for the ISS and ISE are usually tough. However, the process of pooled data harmonization needs a thorough and careful approach. This paper will describe some steps that may help to understand the process of aggregated dataset creation, as well as to show some tips and tricks to deal with possible difficulties. It contains general technical aspects of the pooled data handling to make it ready for analysis. In addition, the paper will describe a macro which combines SDTM datasets for ISS and ISE together.
Implementation of SDTM IG v.3.3 for Neurological Therapeutic Area
Daryna will be discussing how a number of new domains included in the recently released SDTM IG v.3.3 such as nervous system findings, domains for biofluid biomarkers, device domains, PROs etc will be studied based on Therapeutic Area User Guides for Neurology as well as the best practices to implement standards for the study.
• Introduction to TAUGs in Neurological Therapeutic Area in general;
• Types of new domains, examples of raw data and how it should be mapped according to new standards;
• Comparison of its implementation (in Multiple Sclerosis, Alzheimer’s TAUGs etc in Neurology) on SDTM level;
• Pros and cons;
• Contradictions between TAUGs (if any).
To learn more about Intego Group and our involvement in statistical programming and clinical studies, be sure to stop by Booth #16 and meet the team!