Skip to content

Exploring the potential for EHR-derived real-world data to reduce uncertainty in HTA decision-making: a case study of long-term survival outcomes


December 2021


Pittell, H, Kent, S, Groves, B, Mpofu, P, Baxi, SS, Copeland, A, Bargo, D, Adamson, BJS, Jonsson, P. . EU ISPOR. .


Pittell, H, Kent, S, Groves, B, Mpofu, P, Baxi, SS, Copeland, A, Bargo, D, Adamson, BJS, Jonsson, P

OBJECTIVES: Clinical trials are an important source of evidence for health technology appraisals (HTA). However, a key concern is uncertainty in survival due to immature data. This study investigates whether electronic health record (EHR)-derived data from the US may have the potential to reduce uncertainty in long-term outcomes, using NICE technology appraisal (TA) 531 as a case study.


METHODS: We selected patients with previously untreated, Stage IV NSCLC, with positive or unknown PDL1 status, who initiated first-line pembrolizumab monotherapy between October 2016 and December 2020 from the nationwide de-identified EHR-derived Flatiron Health database. We applied additional lab and ECOG eligibility criteria. Outcomes were overall survival from treatment start and treatment duration. Sensitivity analyses assessed a sub-group with known PDL1 status and a time horizon ending at NICE TA publication.


RESULTS: The study included 1109 patients (median age 72, 50% female). Real-world duration of therapy was median 5.0 months (95% Confidence Interval [CI]: 4.2-5.7). Median overall survival was 13.8 months (95% CI: 11.8-16.2) over the full study period. The sensitivity analysis excluding patients with missing PDL1 status found median overall survival of 14.9 months (95% CI: 12.5-17.6). The shorter time horizon (October 2016 to June 2018) estimated median overall survival of 13.1 months (95% CI: 10.8-NR).


CONCLUSIONS: In this case, EHR-derived data offered longer follow-up time (max 49 months) than the trial follow-up (max 22 months) used for extrapolation. This cohort had a median overall survival of 13.8 months while the trial (n=154) estimated 30.0 months and a similar Medicare claims analysis (n=3079) estimated 11.4 months. Real-world median age was 7-9 years older than the trial. Our study demonstrates that EHRs can be a source of mature data on specific cohorts of interest with potential to contextualize trial evidence and inform HTA-decision making.