Menu 返回 Poster-Presentations-Details

T-13: Protocol Deviation Reporting and Tracking Without a Deviations Case Report Form

Poster Presenter

Kia Bryant

Epidemiologist
Centers For Disease Control and Prevention (CDC)
United States

Objectives

To develop a method using existing data to create the Standard Data Tabulation Model (SDTM)-formatted protocol deviation (DV) data domain required by the US Food and Drug Administration (FDA) for submission of data from clinical trials done under an Investigational New Drug Application (IND).

ORAL PRESENTATION: 12:50PM

Method

We reviewed the study protocol for potential deviations. We mapped deviations to data fields on study case report forms. We wrote code in Statistical Analysis Software (SAS v9.4) to populate the DV domain from primary study data.

Results

Tuberculosis Trials Consortium (TBTC) Study 31/AIDS Clinical Trials Group (ACTG) 5349 (S31) is a large (2,500 persons) phase 3 clinical trial evaluating 4- and 6-month regimens for treatment of drug-susceptible tuberculosis (TB), being done under IND. From review of the S31 protocol, we identified 57 potential deviations identifiable from primary study data. Over 3 months, we wrote and refined code drawing on quality assurance findings to ensure identification of the correct participants and deviations. Output of the SAS code is in the SDTM-DV domain format, and contains details to send to clinical trial site staff for confirmation and resolution of deviations. The DV domain code captures 99% of all deviations. Deviations not captured through this automated method are reported to the Sponsor via email and tracked manually. Based on the DV domain, queries were emailed to sites to confirm the deviation or resolve data deficiencies. In some cases, sites were not aware of the deviation, or had not reported the deviation to the Sponsor. The most common deviations involved submission of incomplete study dose logs; most were resolved after submission of complete data. When data entry errors were found, sites were able to update in real time, instead of waiting until exhaustive data cleaning during preparation for data analysis, thereby improving data accuracy. Focused training sessions incorporating knowledge gained from working with the DV domain were delivered to clinical sites with guidance on best practices for data submission to minimize deviations and queries. Following these trainings the number of deviations related to incomplete study dose logs dropped from approximately 41.5 per month to 35 per month.

Conclusion

We used primary study data to identify protocol deviations and create the DV domain without requiring a deviation case report form (CRF). Our SAS code identified thousands of deviations, without requiring site staff to complete an additional CRF. Thus, clinical trial site staff time can devote time to the many other required study procedures. One limitation is that our SAS program identified only deviations for which primary data are reported. There are infrequent deviations that must be reported manually, yet this burden of work is minimal, requiring on average less than one hour of Sponsor and site staff time when an event occurs. In addition to creating the required DV data domain, this use of primary study data to identify deviations allowed the study Sponsor to focus quality assurance (QA) and data cleaning activities on prioritized data related to protocol compliance and on accurate reporting of study results. We were able to inform sites about deviations, resolve missing or inconsistent data, and gain a deeper understanding of challenges experienced at clinical trial sites. As our process relies on a SAS program with weekly feedback to sites, site staff are provided timely information to assist in modifying local procedures to minimize future deviations. Additionally, we use knowledge gained during communication with trial sites about deviations to tailor training messages, creating a cycle of quality improvement. QA activities are resource intensive for both Sponsor and clinical site staff; use of the DV domain to prioritize QA tasks increases the utility of the deviation code while populating the required DV data domain.