Search

Objectives: Real-world data (RWD) from the clinical care of patients captured through Electronic Health Records (EHRs) is a valuable resource for research. Understanding the relationship between characteristics and outcomes of patients treated in the real world and in oncology clinical trials by producing evaluable trial-like populations from RWD can advance clinical and regulatory knowledge. Methods: RWD from the Flatiron Health EHR-derived, de-identified, longitudinal database were compared with pooled patient-level data from the three randomized controlled trials (RCTs) supporting regular approval of CDK4/6 inhibitors for patients with previously untreated hormone receptor-positive, HER- metastatic breast cancer (mBC). The RCT group included patients who received aromatase inhibitor (AI) monotherapy (RCT-control), and an experimental group (RCT-experimental) that received a cyclin dependent kinase 4/6 (CDK4/6) inhibitor + AI. The real-world control group (rwCG) of patients initiating AI monotherapy was selected using the key eligibility criteria across the RCTs. Patients from the rwCG were matched to the RCT-control and the RCT-experimental patients, respectively. Matching (1:1) was performed through the propensity score (PS) method adjusting for baseline covariates of age, race, site of disease, ECOG PS, and metastatic disease (recurrent, De novo). Multiple imputation (MI) was adopted to impute missing ECOG PS in the rwCG due to the high percentage of missingness. Progression-free survival (PFS) and overall survival (OS) were analyzed for the following groups: A) rwCG and RCT- control B) rwCG and RCT-experimental and C) RCT-control and RCT-experimental. To assess the feasibility of RCT control arm replication, we assessed whether the trial replication hazard ratio (HR) estimates from analysis B were within the published trial estimates’ 95% confidence intervals (CIs). HR and their 95% CIs were estimated using Cox proportional hazard model. Results: A total of 1292 patients were selected from the EHR-derived database to comprise the rwCG and 1827 patients were pooled across the RCTs (1106 for RCT-experimental and 721 patients for RCT-control). With MI, 520 rwCG patients were matched to the RCT-control for analysis A, and 658 rwCG patients were matched to the RCT-experimental for analysis B. The results are summarized in Table 1. The point estimate of the PFS HR comparing the rwCG to the RCT-experimental was within the 95% CIs for three RCTs. For OS, the point estimate of the HR was within the 95% CI for MONALESSA-2 but not for PALOMA-2. Conclusion: PFS and OS appeared longer in the RCT-control than in the rwCG, and the difference was more pronounced in OS. While it appears that there is greater similarity for PFS than for OS based on the results of the matched analysis of RCT- experimental vs. rwCG, evaluation of PFS results are limited by substantial differences in assessment and outcome definitions for progression between RCT-control (RECIST) and rwCG. Despite PS matching, there are apparent differences between patients treated in RCTs and routine practice, highlighting the importance of clinical setting, trial selection, study design, and use of randomization. There are still outstanding feasibility questions on the evaluation of OS and further research is required to understand factors potentially impacting the outcomes between RCTs and RWD. Table 1: Estimated Treatment Effects in PFS and OS, rwCG vs RCT-control vs RCT-experimental Citation Format: Christy Osgood, Jiaxin Fan, Xin Gao, Catherine Keane, Jonathan Bryan, James P. Roose, Erik Bloomquist, Aracelis Torres, Shrujal Baxi, Nathan Nussbaum, Fatima Rizvi, Shenghui Tang, Irene Nunes, Julia Beaver, Donna R. Rivera, Lynn Howie, Prashni Paliwal, Laleh Amiri-Kordestani. Feasibility of generating an external control comparator using RWD by matching with previously conducted RCTs: CDK4/6 Inhibitors for the treatment of Metastatic Breast Cancer [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P2-13-03.

Background Adjusting for multiple biases usually involves adjusting for one bias at a time, with careful attention to the order in which these biases are adjusted. A novel, alternative approach to multiple-bias adjustment involves the simultaneous adjustment of all biases via imputation and/or regression weighting. The imputed value or weight corresponds to the probability of the missing data and serves to ‘reconstruct’ the unbiased data that would be observed based on the provided assumptions of the degree of bias. Methods We motivate and describe the steps necessary to implement this method. We also demonstrate the validity of this method through a simulation study with an exposure-outcome relationship that is biased by uncontrolled confounding, exposure misclassification, and selection bias. Results The study revealed that a non-biased effect estimate can be obtained when correct bias parameters are applied. It also found that incorrect specification of every bias parameter by +/-25% still produced an effect estimate with less bias than the observed, biased effect. Conclusions Simultaneous multi-bias analysis is a useful way of investigating and understanding how multiple sources of bias may affect naive effect estimates. This new method can be used to enhance the validity and transparency of real-world evidence obtained from observational, longitudinal studies.

Background and ObjectivesThe primary objective is to examine potential racial and ethnic (R/E) disparities in ambulatory neurology quality measures within the American Academy of Neurology Axon Registry. R/E disparities in neurologic US morbidity and mortality have been clearly documented. Despite these findings, there have been no nationwide examinations of how ambulatory neurologic care affects these negative health outcomes.MethodsThis was a retrospective nonrandomized cohort study of patients in the AAN Axon Registry. The Axon Registry is a neurology-specific outpatient quality registry that collects, reports, and analyzes real-world deidentified electronic health record (EHR) data. Patients were included in the study if they contributed toward one of the selected quality measures for multiple sclerosis, epilepsy, Parkinson disease, or headache during the study period of January 1, 2019–December 31, 2019. Descriptive analyses of patient demographics were performed and then stratified by race and ethnicity.ResultsThere were a total of 633,672 patients included in these analyses. Separate analyses were performed for race (64% White, 8% Black, 1% Asian, and 27% unknown) and ethnicity (52% not Hispanic, 5% Hispanic, and 43% unknown). The mean age ranged from 18 to 55 years, with 61% female and 39% male. Quality measures were chosen based on completeness of R/E data and were either process or outcomes focused. Statistically significant differences were noted after controlling for multiple comparisons.DiscussionThe large proportion of missing or unknown R/E data and low overall rate of performance on these quality measures made the relevance of small differences difficult to determine. This analysis demonstrates the feasibility of using the Axon Registry to assess neurologic disparities in outpatient care. More education and training are required on the accurate capture of R/E data in the EHR.

Background: Hispanic/Latinas (H/Ls) are more likely to be diagnosed with later staged breast cancer and to have poorer survival compared with White women. Mammography screening rates in H/Ls are lower than reported for Whites or African Americans. The goal of this study is to identify potential barriers to mammography screening in this understudied, at-risk population. Methods: We report baseline results of 1083 community-based Hispanic/Latinas who have been enrolled in a prospective study of mammography screening. Hispanic/Latinas, ages 40–75, seeking care at primary care clinics in 4 cities with the largest Hispanic populations in CT and no prior history of breast cancer or biopsy, were eligible for inclusion in our study. 78% of women who were recruited participated in an hour-long telephone interview, generally conducted in Spanish (90%). We collected detailed information on sociodemographic, access to care, health knowledge and belief variables. A follow-up interview conducted a minimum of 26 months after the baseline interview and review of radiology records will provide prospective information of mammography screening. Univariate and bivariate analyses of the baseline data, reported here, identify potential barriers to mammography screening. Results: Of the 1083 women recruited to date, 19% were born in the US, 54% in Puerto Rico, and 28% in other countries; almost 20% have been in the US for 10 years or less. Forty-four percent of women reported speaking and reading only Spanish and only 33% report that they speak English well or very well. These women were similar to the general population of H/Ls living in the Northeast in that they were of low socioeconomic status. Almost 50% of women reported an annual household income of $10,000 or less and 43% had less than a 12th grade education. The majority of women (78%) received some public insurance, primarily Medicaid, with 15.5% of our study population reporting no insurance coverage. Only 64% of women reported having a usual care provider, despite having been recruited from a primary care clinic. Less than half of the women (42%) reported their health as excellent or good, with a large proportion reporting significant comorbiity, particularly obesity and diabetes. Despite the large number of potential barriers, the self-reported history of mammography screening is somewhat higher than that reported for Hispanic/Latinas in national surveys at 74%, but comparable to the rates reported for Whites and African Americans. Conclusions: Given that prospective measures of mammography screening have consistently shown significantly lower rates than retrospectively reported screening, we are likely to see lower screening rates once we have followed our cohort for a minimum of 26 months. The baseline data suggest that the community-based population of Hispanic/Latinas living in CT have significant barriers that may potentially impact the adherence to mammography screening guidelines. Findings from this investigation will be critical in informing cancer prevention strategies for this under-resourced population. Citation Information: Cancer Epidemiol Biomarkers Prev 2011;20(10 Suppl):B105.