Why Data Management and Statistics are intrinsically linked
When you initiate a clinical trial, you invest a lot of money, energy and resources to maximize the potential of your investigational product. You look far… far enough to believe that this product can definitely result in an approval by the Regulation authorities.
The approval is given on the basis of the final statistical analysis that demonstrates the safety and efficacy of your product.
How data management can maximize the statistical results?
Poor quality data can potentially ruin the whole clinical trial, even if the investigational product is innovative and highly performant! The way data are stored, the way the visits and forms are designed are critical and decisive factors that can impact positively or negatively the final results.
When data is extracted for analysis, the statistician needs specific formats and specific tables. Some examples will show you how those aspects are crucial:
- Typically, some data should be displayed as one row per patient, and some others as multiple rows per patient, for example the medications. How to handle the data if 10 rows (10 medications) are expected and some rows are missing? Here, the clinical data manager will program an occurrence format and not a standard table so that we never have blank rows.
- Also, the statistician may analyze a specific endpoint that is phrased differently for each patient. Which proportion of the patients experience migraines during the treatment? Sometimes, the study personnel will enter “migraine”, sometimes “headache”, sometimes “migraine attack” and sometimes we’ll also have typos. On a large scale, poor-quality data will not give the expected results. A smart EDC designer will think upstream in order to get accurate data downstream.
How Statistics can maximize data quality?
Similarly, Statistics can play a key role to improve data quality. How?
Statistics are a powerful tool to check data at any time during the study. Running statistical scripts on real-time data during the study conduct will identify potential issues in data collection: missing data, inconsistencies between 2 data points, wrong formats, unusual variability across sites, protocol violation, the use of local language instead of English, etc.
Those issues will be immediately addressed to the Data Manager and corrective actions will be taken (format changed, corrections asked to the site, monitoring strengthened) before they have a dramatic impact on the final outcomes.
Also, global statistics across the different sites will point out the enrollment difficulties for some of them and readjust the enrollment strategy.
These data analytics enable a precious oversight and are a strong value to the traditional monitoring activities that are done at the patient level only.
Statistician and Data Manager collaboration
The Data Manager and the Statistician must be involved from the very first stages of the trial design, when the protocol is being written. They will work side by side to think ahead and advise the Sponsor on the protocol and CRF design following strict SOPs along with technical and functional requirements of each party involved.
This is the right way of designing a clinical study with the best chance of success!
If you have any questions or suggestions, please contact me! email@example.com