Expert Insights

Expert Insights

5 Leading Factors to Consider When Converting Clinical Study Legacy Data

5 Leading Factors to Consider When Converting Clinical Study Legacy Data
08 July 2019

Work supporting biostats analysis of a clinical study can be complex and labor intensive; some of the heaviest lifting comes from an area that many people overlook – integrating legacy data.

Trials that are combined to create an integrated summary often have been performed over several years, across many countries, designs, and investigators. The work to pull these data together to create a single, comprehensive set of analysis‑ready data is perhaps the most critical undertaking in the integration process. Sponsors risk creating results that aren’t reflective of the true characteristics of the investigational product if data are combined incorrectly, inconsistently, or incompletely.

Following are 5 key factors to consider in planning your data integration strategy:

  1. What are the sources of your lab data?

Lab data clean-up can be one of the most time-consuming parts of the initial preparation. Data have likely come from a variety of sources: central labs, local labs, or even foreign labs that generate data in non-English languages. Each lab may have slightly (or drastically) different standards for reporting and transfer. The work to consolidate and clarify the sources of data, with specifics on their structures, parameters, normal ranges, and so on, needs to be a collaborative effort among Regulatory, Medical, and Biostatistics team members to determine what is necessary, what is appropriate, and what should be reported.

First, as data must be integrated from a variety of disparate sources, it is not uncommon to have multiple types of units and scales utilized in the legacy data. The programming team should be advised to run a data check on the units (not restricted to just lab data) used in the datasets. If they are not standardized in the source data, they need to be remapped and standardized at the SDTM level. Data exceptions can be noted in the programming specifications and data reviewer’s guide.

  1. How controlled is the terminology?

Over the dozen or more trials you may be integrating for safety, you may find an equal number of coding dictionaries being used. Between MedDra and WhoDRUG and the ever-evolving algorithms used to code medical data, it’s likely that you will need to spend some time standardizing your coding terminology. Where possible, drug development companies should try to match your historic studies against the coding dictionaries used for your pivotal trial, saving the effort of recoding what is likely your largest trial (and the one probably using the most recent coding dictionary.) Regardless of how you do it, you need to determine what versions you should use and allow enough time to complete this task. You don’t want to get to the finish line only to realize you need to recode, rerun, and rewrite your results.

  1. Is there a crossover design component in the data?

In cases of crossover designs or studies with an open-label extension, you may have populations of subjects who were treated with different treatments and/or multiple doses over the period of observation. When this happens, it’s important to ensure that subjects are counted correctly in their respective analyses.

Questions to ask might include:

  • What is the baseline period?
  • How long after treatment should adverse events (AEs) be considered treatment-emergent (especially if subjects are changing to another treatment)?
  • Is the AE analysis dataset structure set up to support the appropriate AE summary tables?

Even something as simple as study day needs to be reset in the open-label period to enable the proper summary of study results in corresponding outputs. In the Statistical Analysis Plan (SAP), do you have a rule defined for how to count the AEs across the study under all applicable treatments? How should the denominator be defined when percentages are present in AE summary tables?

  1. How do you harmonize the study schedules and summarize the data in a meaningful way: by visit or by shift tables?

The studies being integrated are likely to be on various schedules. Visit windows need to be defined in the SAP, and a windowing schedule that is comprehensive and meaningful in terms of summarizing the results needs to be determined. If a crossover component exists in the data, drug developers also need to take into account the reset baseline in the schedule.

For parameters that are normally summarized by scheduled visits, instead of presenting all windowed visits, it often makes more sense to summarize the post baseline outputs by worst values for the selected parameter to highlight the results in a more efficient way. Shift tables can be used in addition to simple descriptive summaries.

  1. How were exposure data collected and documented?

Exposure data are important as a lot of relationships are based on them (e.g., baseline, treatment duration, compliance, treatment-emergent adverse events, and so on). The quality of the exposure data can determine more than the outputs of the exposure summaries by themselves. The team should be aware of potential issues when they scrutinize and prepare the data for integration. They need to ask the following questions:

  • Was the dosing information collected on the Case Report Form (CRF) or outside the CRF? How reliable are the sources?
  • Are the dosing regimens consistent in the form and dose level across the studies to be integrated? Are they based on the same units? If not, what conversion needs to be done?
  • Are dates and/or times collected for all studies? Are there any missing dates/times? Is imputation needed?
  • Are the data sufficient for the calculation of compliance? If not, the studies for which drug accountability data were not available should be noted in the SAP.
  • Did the treatment design vary across studies? If there are various dose levels or crossover of treatment present in some studies, it should be specified in the SAP how exposure data will be calculated and summarized overall and separately for certain studies.

With careful attention to detail during the planning and conversion steps, your legacy data can augment rather than undermine the data from your pivotal studies during the integration process.

At Synteract, we offer extensive expertise in dealing with a wide range of legacy data issues. Contact us today to discover how we can help you!

Written by: Lan Liu, Manager, Biostatistics, Synteract

x

Submit RFI / RFP

..
By submitting your information to us through this webform, your business details will be added to our database. In accordance with our Privacy Notice, we may then contact you with marketing information about Synteract that might be of interest to you. We will never sell your details to third parties. You can unsubscribe from marketing communications at any time.

Required

x

Contact Synteract

Tell us how to stay in touch with you:

By submitting your information to us through this webform, your business details will be added to our database. In accordance with our Privacy Notice, we may then contact you with marketing information about Synteract that might be of interest to you. We will never sell your details to third parties. You can unsubscribe from marketing communications at any time.

Required

x

 

x

Download Resource

By submitting your information to us through this webform, your business details will be added to our database. In accordance with our Privacy Notice, we may then contact you with marketing information about Synteract that might be of interest to you. We will never sell your details to third parties. You can unsubscribe from marketing communications at any time.

*Required

x

To complete your download, please choose an option in the Internet Explorer pop-up window at the bottom of this browser.

x

Opt In

By submitting your information to us through this webform, your business details will be added to our database. In accordance with our Privacy Notice, we may then contact you with marketing information about Synteract that might be of interest to you. We will never sell your details to third parties. You can unsubscribe from marketing communications at any time.

*Required