Streamlining non-CDISC decentralised trial data analysis and reporting

A pharma company, focused on developing a treatment for an unmet need in women’s health, embarked on a phase 2 proof of concept study in the treatment of moderate-to-severe hot flashes in peri- and post-menopausal women, within a one-week timeframe.

The study was evaluating both the safety and efficacy of the new treatment. It was executed as a fully decentralised clinical trial (DCT), incorporating innovative components such as eConsent, telehealth for medical assessment, and electronic patient-reported outcomes (ePROs) to assess clinical endpoints. Notably, no CDISC standards were required for data collection or data analysis. Also, traditional data management was not included in this study, as a third-party vendor handled database build and data collection without the customary data cleaning activities and standardised data transfers. ICON was specifically enlisted for expertise in statistical analysis and programming.

While not adhering to CDISC standards provided flexibility in statistical programming, it introduced challenges due to the non-standard data received from the vendor. The absence of data review by the vendor resulted in data inconsistencies. Additionally, frequent changes to study priorities by the client also led to significant shifts in timelines and scope.

The primary challenge involved collaborating with the data vendor to establish a standardised raw data transfer process. ICON worked with the vendor to create a Data Transfer Specification, enabling our statistical programming team to commence work on the non-CDISC datasets essential for analysis. Armed with insight into the structure of the raw data transfers, our statistical programming team was able to interrogate the data, uncovering a critical design flaw in the database build that did not allow trial participants to fully enter hot flash diary data under certain conditions.

Leveraging ICON’s standard TFL (Table, Figures, and Listings) shells, the statistical programming team designed ADaM-like analysis datasets, which powered our WorkBench reporting systems, and facilitated the generation of tables and listings through a metadata driven approach. This significantly reduced the time required to produce analysis outputs.

Effective collaboration between our statisticians and statistical programmers, coupled with the introduction of a Clinical Data Science Lead for data review, substantially improved data integrity. This approach successfully resolved the major design flaw identified in the database build during data analysis.

The development of a Data Transfer Specification in partnership with the data vendor ensured the availability of reusable SAS programming for future data transfers within the study, promoting efficiency and consistency.

Our decision to adopt a CDISC-like approach for analysis datasets and TFLs, along with the adaptability of our WorkBench statistical reporting application to read non-CDISC datasets, optimised our reporting process. The metadata-driven TFL approach saved time and resources, enabling a more focused allocation of resources toward data quality in the earlier stages of the study.

ICON’s efforts significantly enhanced the integrity of the analysis of the trial data while maintaining the client’s stringent timeline requirements. Our adaptable approach to resourcing ensured that we met the client’s needs effectively, delivering exceptional results in the face of unique challenges.