ユーザーIDをお忘れですか? or パスワードをお忘れですか?

メニュー 戻る Poster-Presentations-Details

P07: A Unique Use of Real-World Data in a Complex Private-Public Collaborative Partnership to Support a New Drug Application

Poster Presenter

      Nancy Ann Sherman

      • Clinician, Post-Approval Clinical Development, Biopharmaceuticals Group
      • Pfizer Inc
        United States


The RAMPART Medical Records Project (MRP) aimed to review, extract and analyze comprehensive real-world data from medical records of participants in a randomized clinical trial, the RAMPART study, to support a revised new drug application (NDA) for midazolam for treatment of status epilepticus.


From 2014 to 2017, we obtained >1000 medical records of RAMPART study participants and, following redaction of personal information, a comprehensive review of unstructured medical record data was conducted to extract all safety and other relevant information to support the revised NDA.


RAMPART was a randomized, controlled trial comparing intramuscular (IM) midazolam given via auto-injector vs intravenous (IV) lorazepam for pre-hospital status epilepticus. Results showed IM midazolam was more effective and as safe as IV lorazepam. RAMPART was designed as an NIH-funded academic study and was exemplary in meeting clinical research needs. However, safety data collection focused on targeted events most relevant to the condition under study rather than the comprehensive safety data collection typically required for the drug approval process (ie the NDA). RAMPART was pivotal for the new indication sought; however, on initial NDA review, FDA requested additional patient-level safety data, which could only be obtained from the study participants’ medical records. To support the RAMPART MRP, a collaboration was created with the NDA sponsor (Pfizer Inc), the RAMPART lead academic institution (University of Michigan) and several US government agencies. A dedicated Program Plan (analogous to a clinical trial protocol) was developed to describe the rationale, methodology, processes, procedures and guidelines to review, extract, and analyze safety data of interest, including creating a case report form and database to capture the unstructured data. In total, 1020 medical records were reviewed from 890 participants. The resulting data was analyzed and included in a revised NDA. Analyses were supportive of IM midazolam use for this new indication, revealing no previously unrecognized adverse drug effects, nor any statistically significant relationships between the primary efficacy outcome (seizure termination before emergency department arrival without rescue medication) and covariates such as prior medication usage, medical history, or study drug dose. Based on the comprehensive safety data, 771 (86.6%) participants had a safety event and 495 (55.6%) a serious safety event. The proportion of patients with safety events was similar for the 2 study drugs.


Real-world data collected retrospectively in the RAMPART MRP complemented data collected prospectively in the RAMPART study and facilitated FDA approval of IM midazolam for pre-hospital treatment of status epilepticus in September 2018. While targeted retrospective data collection is widely employed, comprehensive capture of unstructured safety data from medical records presents significant challenges, particularly for the accurate identification of adverse safety events. Our experience may be useful to others considering such unstructured medical record data collection to support regulatory submissions. The RAMPART MRP required a highly complex hybrid model of data review and collection. A unique methodology, with detailed description of processes and procedures was developed and implemented. A complex public–private partnership undertook the project, and key stakeholders of these collaborative partners worked closely over several years; problem-solving as challenges and barriers were identified and mitigated. The main challenge was that, unlike a typical chart review, outcomes of interest were not pre-defined, hence intensive review of each page of the medical records was required. Many medical records were lengthy, with unclear or incomplete information but unlike a clinical trial, querying an investigator or treating physician when data were unclear, missing or discrepant was not possible. For any similar project, success requires a clearly defined methodology to ensure data collection accuracy, consistency and robustness, as well as effective mitigations to address potential challenges and barriers. In summary, this creative use of comprehensive, unstructured real-world medical record data can be instrumental in supporting an NDA submission. In addition, the RAMPART MRP demonstrated that complex private–public partnerships between industry sponsors, academic institutions and government agencies are possible, and can be collaborative, effective and successful.