Clinical Trial Design
– including the determination of sample sizes
A Frequentist or Bayesian approach, study design selection, determination of sample sizes.
We provide support for classical and modern designs, such as parallel or cross-over designs, including Bayesian approaches, as well as application of ready-to-use software solutions (SAS, R or PASS), or implementation of simulations to determine the minimally required sample size.
By applying unique, study-specific simulations, we can reduce the sample size and consequently the costs of studies.
Data Management Support
of Clinical Trial Query Management
Data management is an extremely important aspect of clinical trials. For the most part, it is under-evaluated, as it exerts no visible impact on the outcome.
However, poor data management leads to weakness of analysis and reporting. Several steps of data management are required to assure perfect quality. Paper and electronic Case Report Forms (CRFs) may be considered as different approaches.
Data management has its own standard activities, which are independent of data collection methods. Whatever issues may arise (for example, query management, the handling of exemptions, documentation, and updates of the e-CRF, among others), a clear, consistent, 100% error-free database should be provided as a clinical study database.
SDTM and ADaM mapping
of Clinical Trial data
The way that we collect clinical data has a very human nature, as we are, after all, human.
Computers and standardised processes prefer a somewhat different data structure. While mapping the data into a Study Data Tabulation Model (SDTM), we map the original (raw) data into a format that maximally supports the process of analysis.
The reporting of clinical trials is also supported at database level. The preparation of ADaM (Analysis Data Model) datasets also involves the derivation and storage of all derived outcomes into the databases.
With the ADaM datasets, you also have all the outcomes described in the Protocol or Statistical Analysis Plan – even if it is difficult to read.
Evaluation and reporting are generally performed with the help of SAS.
This is a somewhat traditional tool, in that it is not easy to use, but you can do anything within its framework.
Those who have been working with SAS for decades, as we have, may have already built their own SASsolutions in the format of SASmacros. Output format does not matter: tables, listings or graphs can be created with the help of SAS.
The Evaluation of Clinical Trials
the Preparation of Tables, Listings and Graphs
The data that is collected in clinical trials should be tabulated in an appropriately formal way.
The required statistical activities are summarised in the Statistical Analysis Plan (SAP), while the associated table shell will indicate the exact format of the output. Essentially, there are three types of output:
- Tables are to provide answers to clinical questions.
- Graphs are to support the interpretation of the primary, secondary or safety outcomes.
- Listings provide transparency: they render possible a potential recheck of the results besides identifying and explaining some data related issues.
The Support of Post-Marketing Studies
/ Automated Reporting
With regard to Post-Marketing Studies, speed and cost are more critical.
Planimeter’s solution to this challenge is to provide automated evaluation and reporting.
How does it work?
The statistical report is generated in parallel with the statistical analysis – in programmatic way. When the final database is available, the statistical report can be generated from the
database directly. Do you need a subgroup analysis?
This is not a problem: the filter can be set (in five minutes) and the report can be generated (in a further five minutes). Would you like to communicate your results on the web (instead of a Word document)? This is not a problem: it only takes us five minutes.
The Preparation of
Integrated Safety / Efficacy Summaries
You may find yourself in a situation in which an integrated summary should be provided (either on efficacy or safety).
You have the specific study documentation (databases, programs,
and reports) but you need to pool them. We can support you in this activity with our standard steps:
- Planning: Do only the outcomes (meta-analysis) need to be pooled? Do the databases need to be pooled? Does a pooled database have to be analysed?
- Data Integration Plan: What are the questions? Which are the variables? How can we transform them to the same platform?
- Statistical Analysis Plan: A plan is certainly necessary for describing the required data manipulations and the desired outcomes.
- Data Management: Data migration and integration – the data should be transformed to a common view (coding, outliers, and missing value management, among others).
- Analysis: Running the programs, preparation of the final report.
R is an option in clinical trials.
While SAS is used on a daily basis in the pharmaceutical industry, it is a less known fact (from 2012) that, according to the United States Food and Drug Administration (FDA), application of R is accepted in drug
By means of R we can use a framework with more than 20 thousand (!) standalone solutions (packages) of specific mathematical problems.
R is extremely effective in the visualisation of the obtained results. Graphing with SAS has limitations, which is not the case with R.
R can easily be a part of your EDC (electronic data capture) systems to provide up-to-date status reports or to equip you with real time outcomes. Regular (safety) reporting is merely an option for the application of automated reporting techniques of R.
in the Pharmaceutical Industry
While Planimeter is a full-service CRO, we are also extremely
experienced in data science.
Our activities in this area are as follows:
- To search for publicly resources (databases, protocols,
initiations to join).
- To apply text mining approaches in finding connections in (billions of pages of) documentation.
- To support drug development with artificial intelligence solutions.
- To support marketing/sales processes with forecasting