Survival of Multiple Myeloma Patients in Epic Cosmos Confounded By Death Reporting

Background: Data obtained from the Surveillance Epidemiology and End Results (SEER) registry, a public database sponsored by the National Cancer Institute (NCI) report age-adjusted mortality rates 2000-2022 decreasing with the most recent rates for Non-Hispanic Blacks (NHB) at 5.6 per 100,000 and Non-Hispanic Whites (NHW) at 2.6 per 100,000, age-adjusted to the 2000 US Std Population. Non-Hispanic Asians / Pacific Islanders had the lowest mortality rate at 1.4 per 100,000 using SEER*Explorer updated 27-Jun-2024 (seer.cancer.gov). Two reports regarding race and survival in patients with myeloma using the Flatiron database showed survival amongst NHB to be equivalent (Buck, T et al, 2024, PMID 38031762) or better (Maignan K et al, 2022, PMID 35440047) than NHW. We endeavored to find if survival amongst different multiple myeloma groups in Epic Cosmos was closer to SEER, Flatiron, or in between.

Methods: Data used in this study came from Epic Cosmos, a dataset created in collaboration with a community of Epic health systems representing more than 259 million patient records from over 1,548 hospitals and 35.4K clinics from all 50 states and Lebanon. We identified myeloma patients by isolating the appropriate ICD10, ICD9 and SNOMED codes from billed admitting, final, or linked diagnosis. Patients must have received anti-myeloma treatment [this was a surrogate for the date of diagnosis], always have less than 2 years between encounters, and age greater than 20 years old. Study start date based on the availability of diagnoses in Cosmos start on 1-Jan-2012 and an end date of 1-Jan-2024.

A 1:1 control group pool from active patients between 2012-2024, no history of a blood cancer, and the first face-to-face outpatient encounter with a calculated eGFR within +/- 6 months; this is the index encounter. From this pool, patients were matched by birthdate, sex, social vulnerability index (within 10 percentile points), eGFR within 5 mL/min, first encounter date, and race (combined native American, Asian, and Hawaiian into a separate other category).

Results: 74,295 multiple myeloma patients were identified of which 69.5% were white, 19.7% were black, 1.6% Asian, and 7% multi-racial. 55.8% were male, and 94.5% were not Hispanic or latino. The age distribution was as follows: 1.4% below 40 years old, 5.2% 40-50, 16.6% 50-60, 32.4% 60-70, and 31.3% 70-80, 12.2% 80-90, and 1% greater than 90 years old. When race and sex were combined, 45.7% were white males, 32.3% were white females, 10.4% black men, and 11.6% black women.

Kaplan-Meier curves will be displayed at the meeting of myeloma patients by age decile, social vulnerability percentile, race with ethnicity, and race with sex. Surprisingly white men had the shortest survival, white women and black men next, and the best survival for black women, the median being approximately 2.5 years longer than white men. Because this was in stark contrast to SEER data, we discovered that of the 31.5% (23,424) patients reporting death – 4.2% white patients reported death to their Epic site compared to only 2.7% black patients, 2.1% of multi-racial patients, and 1.5% of Asian patients.

The median survival of the myeloma patients was 7.5 years, while the median survival had not been reached in the 73,660 patients in the matched control group - approximately 75% survival at 7.5 years.

Conclusions: While similar in size to published SEER analyses, death was reported twice as often to Epic sites by white patients, creating what is thought to be an artificial survival advantage for black patients. Survival analyses in Epic Cosmos [and similar datasets] may be confounded by race and other medical access surrogates.