Assessing a composite end point for new tocolytics in clinical trials: Data from 4 US integrated delivery networks.

PURPOSE: A composite end point (CE) measuring neonatal benefit was created for use in tocolytic randomized controlled trials with rates assessed using data from one referral hospital. The goal of this study was to assess wider generalizability of the CE, using data from multiple integrated delivery networks, creating a cohort of linked mother-neonate pairs to understand neonatal outcomes in a broad population. METHODS: Retrospective data on births (2001-2012) were collected from 4 US integrated delivery networks in the COMparative effectiveness PAtient Safety and Surveillance (COMPASS) Research Network, and linked mother-neonate pairs were identified. The CE was analyzed for all in-hospital singleton neonates at ≥24 weeks of gestational age (GA) born to mothers aged ≤45 years at a referral hospital or hospital with >2000 annual births. RESULTS: The CE analyses included 56 485 eligible mother-neonate pairs; frequency of the CE decreased from 89% to 66% between GA weeks 24 and 29 and further decreased to <14% for infants born >34 weeks of GA. Composite end point rates were 20% to 30% lower at 24 to 30 weeks of GA in COMPASS compared with Medical University of South Carolina but were similar by 31 weeks. CONCLUSIONS: The COMPASS Network enabled evaluation of the CE across a large population demonstrating that the CE findings could be replicated beyond a single hospital and the potential for lower CE frequency. Based on this, an adaptive design was adopted for randomized controlled trials, specifically sample size reestimation to mitigate against the risk of lower outcome rates, highlighting the use of real-world data in drug development.

Characterizing phenotypes and outcomes of drug-associated liver injury using electronic medical record data.

PURPOSE: To evaluate the incidence, phenotypes, and outcomes of drug-associated liver injury identified in electronic medical record (EMR) data using standardized criteria for drug-induced liver injury (DILI). METHODS: This retrospective cohort study used EMR data from a large integrated healthcare system. Study inclusion required 18 years of age or older, ≥1 prescription fill for any of 14 medications associated with hepatotoxicity between 1 January 2003 and 30 June 2009, and ≥12 months of membership prior to the drug exposure. Patients with underlying non-drug causes of liver injury were excluded to minimize capture of liver injury events unrelated to drugs. Drug-associated liver injuries were identified by liver chemistry elevations temporally associated with drug use based on standardized criteria for DILI. Cases were classified by clinical pattern and severity. Outcomes of liver transplant and all-cause and liver-related death were examined. RESULTS: A total of 1 053 979 drug exposures were identified in 601 125 patients. We identified 265 drug-associated liver injuries (32.8 per 100 000 persons) occurring in 250 patients. Isoniazid exhibited the highest incidence rate of 606 per 100 000 persons. Of the 265 cases, 41% were mild; 12% exhibited moderate drug-associated liver injury (with concomitant ALT ≥ 5× ULN and bilirubin ≥2× ULN); and 17% exhibited coagulopathy, ascites, encephalopathy, or other organ failure. Last, seven cases (3%) were associated with death, and there were no liver transplants. CONCLUSIONS: Study results align with earlier prospective studies, supporting the value of standardized methodology to identify drug-associated liver injury in the EMR. These methods can potentially enhance safety and clinical outcomes.