Cancer Survival at a Comprehensive Cancer Center Compared With Surveillance, Epidemiology, and End Results (SEER) Estimates Authors: Ravi Salgia, Joseph Alvarnas, Vijay Trisal, Ann Vanderplas, Xiaochen Li, Javier Arias-Romero, Jeremy Fricke, Paul Frankel, Isa Mambetsariev, Hannah Rice, Harlan Levine, Marcel R.M. van den Brink Published: 2026 ABSTRACT ObjectiveWe compared outcomes across common cancer types at an NCI-designated Comprehensive Cancer Center (NCI-CCC) (City of Hope) vs the U.S. Surveillance, Epidemiology, and End Results (SEER) national database. Cancer Centers offer a comprehensive approach to prevention, diagnosis, and treatment through tertiary services and by providing coordinated, personalized care.MethodsPropensity score matching (PSM) methods were employed in a 1:5 ratio for patients with incident cancers from 2004 to 2020 treated at COH vs reported in SEER.ResultsThe COH cohort had significantly superior overall survival (OS) compared with the SEER cohort across all examined cancer types after multivariable adjustment for key baseline characteristics and in the PSM analyses. For the PSM analyses: non-small cell lung cancer (NSCLC) hazard ratio (HR) 0.73, breast cancer HR 0.83, prostate cancer HR 0.62, colorectal cancer HR 0.74, pancreatic cancer HR 0.75, acute myeloid leukemia HR 0.79, acute lymphoid leukemia HR 0.77, and multiple myeloma HR 0.70 (all P < .001). Notable findings included substantially lower mortality risk for patients with any of the studied advanced solid cancers.ConclusionWe found that cancer survival benefits were associated with patient care at an individual NCI-CCC employing state-of-the-art therapies and modern care. This work also highlights an opportunity to make adjusted survival outcomes more accessible to patients, complementing other types of comparative data now available. Further research could help to identify the drivers of observed survival advantages and promote the development of better guidelines and care models to improve outcomes for all patients. Introduction In 1971, the National Cancer Act established the National Cancer Institute (NCI) Cancer Centers Program; today across the U.S., there are 73 NCI-Designated Cancer Centers, including 57 Comprehensive Cancer Centers (CCC).1 These facilities offer a comprehensive approach to cancer prevention, diagnosis, and treatment through tertiary care services such as specialized imaging, complex surgeries, and innovative therapies and technologies available through interventional trials. Despite recent reductions in patient mortality, cancer remains the second leading cause of death in the U.S., with approximately 2.1 million new diagnoses and over 626,000 deaths expected in 2026, excluding nonmelanoma skin cancers.2 Reduced tobacco use, earlier detection, new treatments, and increased access to cancer care, screening, and therapy are among the factors that led to a 34% decline in mortality from 1991 to 20223 and increased rates of 5-year survival from 49% in the 1970s to 69% in the 2010s.4 However, prognoses still vary widely depending on cancer type and stage at diagnosis as cancers are highly complex and heterogeneous. For instance, while localized breast or prostate cancers are often curable, other advanced-stage cancers have a poorer outlook and may require more specialized care. The NCI-CCC model is driven by clinical practice specialization and multidisciplinary, disease-focused care teams, offering expertise informed by clinical trials and biomarker research, as well as treatment regimens tailored to a cancer’s tissue of origin and molecular context.5, 6, 7 Studies suggest that this model can improve communication, patient satisfaction, and clinical outcomes.8,9 Additionally, Cancer Centers are often earlier and more efficient adopters of modern approaches to precision oncology, which has known clinical practice gaps as demonstrated by a recent U.S. study.10 Evidence from that study suggests that patients may be underserved in some care settings such that they do not receive appropriate biomarker testing or, even among those who do, the correct treatment. In that study, 49.7% of patients with non-small cell lung cancer (NSCLC) did not receive biomarker testing results and, of the 50.3% who received testing results with actionable mutations, only 70.8% received the appropriate therapy.10 To evaluate the cancer care and survival outcomes at an NCI-CCC and in the population of the U.S. Surveillance, Epidemiology, and End Results (SEER) national database since 2004, we compared outcomes at our institution (City of Hope National Medical Center [COH]) with the data reported in SEER. Through subspecialized expertise and coordinated care delivery across our network of regional clinical sites, our model supports individualized, biomarker-driven treatment decisions aligned with clinical guidelines (which emerge from coordinated data, shared expertise, and decision support tools11) and allows for integration of clinical research into routine practice through timely referral to clinical trials.12 We hypothesized that care at a high-volume, specialized NCI-CCC employing state-of-the-art, evidence-based therapies and multi-disciplinary care may be associated with survival benefits for cancer. Currently, there is a lack of transparency and no national standard or national database for publicly reporting adjusted patient outcomes data in oncology, which makes it difficult for patients and families to find the highest value when seeking cancer care. However, it is already customary for centers to report outcomes for bone marrow transplantation procedures. Ideally, data reporting would be standardized across centers and include longitudinal and detailed treatment information, adverse events and toxicities, comorbidity history, end-of-life care, and comprehensive outcomes data. This study reports patient- and tumor-adjusted models for reporting survival outcomes that could be used as part of a collaborative, transparent model for all cancers. Methods Survival outcomes of patients treated at COH were compared with outcomes reported in the SEER national database. The SEER database collects incident cancer and survival data from national population-based cancer registries and covers approximately 46% of the U.S. population. The SEER registries collect data on patient demographics, primary tumor site, tumor morphology, stage at diagnosis, and patient vital status (alive/dead at end of study). This study was approved by the COH institutional review board (IRB 25219). Patients diagnosed with malignant cancer from 2004 to 2020 were identified from SEER and the COH Cancer Registry (CR). Patient-level SEER data were abstracted using SEER*Stat software, case listing option, from the SEER Research Data (17 registries, November 2023 submission).13 All patients queried from the COH CR had to have an abstraction status of complete and coded as an analytic case, as defined by the North American Association of Central Cancer Registries.14 Malignant tumors of the following common primary cancer sites and hematologic malignancies with concordant data availability were included: breast, colorectal, lung (non-small cell), prostate, pancreas, multiple myeloma, acute myeloid leukemia (AML), and acute lymphoid leukemia (ALL). These cancers had sufficient prevalence to allow for large enough sample sizes to support stable and interpretable results. The final cohort included patients with the following demographic, clinical, and survival characteristics available in both databases: age at diagnosis, gender, race, ethnicity, primary tumor site, American Joint Committee on Cancer (AJCC) stage (for solid tumors), cancer histologic type, diagnosis year, vital status, and survival time. The AJCC guideline in effect at time of diagnosis dictated tumor staging. The primary endpoint was overall survival (OS), defined from the date of diagnosis to the date of death from any cause or last contact. A study end date of December 31, 2021 was used for both COH and SEER cohorts to align with the SEER follow-up cut-off date. Baseline demographic and clinical characteristics by cancer type were summarized descriptively for each cohort. Multivariable adjustment for age group at diagnosis category per SEER standards (18-44, 45-54, 55-64, 65-74, 75+),15 gender, race, ethnicity, AJCC stage (for solid tumors), and diagnosis year was performed.15 In addition, to further reduce confounding from baseline differences and provide more representative data for survival graphs, propensity score matching (PSM) methods16, 17, 18 were performed. Propensity scores were estimated using logistic regression based on all these variables. COH patients were matched with SEER patients using nearest-neighbor matching on the logit of the propensity score without replacement at a 1:5 ratio. PSM was applied using the MatchIt package in R.18 Overall survival (OS) distributions in the COH and SEER cohorts were compared using the Kaplan-Meier method.19 Multivariable Cox proportional hazards regression methods comparing COH to SEER were performed separately for each primary cancer type, with adjustment for the same covariates. Adjusted hazard ratios (HRs) and 95% CIs (CIs) were reported from the Cox models. All reported P-values are 2-sided. Analyses were performed using R Version 4.4.3. Results Patient Characteristics The baseline characteristics of patients in the COH and SEER cohorts are summarized in Supplementary Table 1. A total of 27,318 patients from COH and 3,023,453 patients from SEER were included in the analysis across the 8 included cancer types. Compared with the SEER cohort, the COH cohort was younger (median: 62.0 vs 66.0 years), with a higher proportion in the 18-44 and 45-54 age groups and a lower proportion in the 75+ age group. The COH cohort also had a higher proportion of male patients (54.4% vs 49.7%), Asian or Pacific Islander patients (15.5% vs 7.4%), and Hispanic patients (18.6% vs 10.1%); a lower proportion of Black or African American patients (6.0% vs 12.0%); and a similar proportion of white patients (77.5% vs 79.4%). The COH cohort had a higher proportion of patients diagnosed at stage IV for NSCLC (50.5% vs 43.8%), breast cancer (7.1% vs 5.2%), and colorectal cancer (38.8% vs 20.8%), whereas the SEER cohort had a higher proportion of patients diagnosed at stage IV for prostate cancer (9.1% vs 7.8%) and pancreatic cancer (50.3% vs 48.6%). In this retrospective cohort study, the patient’s earliest occurring cancer during the study timeframe was selected. Patients in the COH and SEER cohorts were matched in a 1:5 ratio using propensity scores based on key baseline variables, as described above. After matching, the 2 cohorts were well balanced across demographic and clinical characteristics, supporting a fair comparison of clinical outcomes between populations across diverse cancer types (Table 1).Table 1Demographic and Clinical Characteristics of Patients in the Propensity Score-Matched CohortTable 1 dummy alt textCharacteristicsNSCLCBREASTPROSTATECOLORECTALPANCREASACUTE MYELOID LEUKEMIAACUTE LYMPHOID LEUKEMIAMULTIPLE MYELOMACOHSEERCOHSEERCOHSEERCOHSEERCOHSEERCOHSEERCOHSEERCOHSEER(N=2884)(N=14420)(N=7141)(N=35705)(N=8961)(N=44805)(N=2436)(N=12180)(N=958)(N=4790)(N=1921)(N=9605)(N=733)(N=3665)(N=2284)(N=11420)Age, years Mean (SD)67.7 (11.5)68.6 (11.7)57.1 (12.8)57.8 (13.1)64.0 (8.14)64.8 (8.48)61.2 (13.6)62.1 (14.0)66.0 (11.8)66.9 (11.5)56.5 (15.7)57.4 (16.2)44.7 (15.8)44.8 (16.3)59.4 (9.40)60.5 (9.81) Median [Min, Max]68.6 [22.3, 95.8]69.0 [18.0, 90.0]56.9 [19.0, 96.9]57.0 [18.0, 90.0]64.3 [36.8, 94.8]65.0 [30.0, 90.0]60.9 [18.1, 98.3]62.0 [18.0, 90.0]66.2 [18.9, 93.2]67.0 [18.0, 90.0]59.0 [18.2, 94.8]60.0 [18.0, 90.0]44.6 [20.0, 89.2]44.0 [20.0, 90.0]60.3 [24.1, 100]61.0 [18.0, 90.0]Age groups, years 18-4482 (2.8%)387 (2.7%)1137 (15.9%)5675 (15.9%)55 (0.6%)275 (0.6%)249 (10.2%)1243 (10.2%)31 (3.2%)147 (3.1%)402 (20.9%)2010 (20.9%)357 (48.7%)1834 (50.0%)131 (5.7%)623 (5.5%) 45-54279 (9.7%)1404 (9.7%)1864 (26.1%)9321 (26.1%)940 (10.5%)4695 (10.5%)504 (20.7%)2526 (20.7%)100 (10.4%)502 (10.5%)323 (16.8%)1595 (16.6%)146 (19.9%)692 (18.9%)476 (20.8%)2398 (21.0%) 55-64625 (21.7%)3136 (21.7%)2047 (28.7%)10240 (28.7%)3364 (37.5%)16822 (37.5%)625 (25.7%)3121 (25.6%)259 (27.0%)1305 (27.2%)506 (26.3%)2542 (26.5%)137 (18.7%)678 (18.5%)904 (39.6%)4532 (39.7%) 65-74990 (34.3%)4955 (34.4%)1362 (19.1%)6816 (19.1%)3632 (40.5%)18156 (40.5%)608 (25.0%)3037 (24.9%)326 (34.0%)1626 (33.9%)469 (24.4%)2353 (24.5%)72 (9.8%)360 (9.8%)687 (30.1%)3436 (30.1%) 75+908 (31.5%)4538 (31.5%)731 (10.2%)3653 (10.2%)970 (10.8%)4857 (10.8%)450 (18.5%)2253 (18.5%)242 (25.3%)1210 (25.3%)221 (11.5%)1105 (11.5%)21 (2.9%)101 (2.8%)86 (3.8%)431 (3.8%)Gender Female1484 (51.5%)7414 (51.4%)7141 (100%)35705 (100%)NANA1139 (46.8%)5702 (46.8%)449 (46.9%)2238 (46.7%)889 (46.3%)4487 (46.7%)343 (46.8%)1724 (47.0%)1004 (44.0%)5032 (44.1%) Male1400 (48.5%)7006 (48.6%)NANA8961 (100%)44805 (100%)1297 (53.2%)6478 (53.2%)509 (53.1%)2552 (53.3%)1032 (53.7%)5118 (53.3%)390 (53.2%)1941 (53.0%)1280 (56.0%)6388 (55.9%)Race White2058 (71.4%)10305 (71.5%)5262 (73.7%)26313 (73.7%)7558 (84.3%)37798 (84.4%)1772 (72.7%)8873 (72.8%)715 (74.6%)3587 (74.9%)1493 (77.7%)7525 (78.3%)616 (84.0%)3147 (85.9%)1707 (74.7%)8537 (74.8%) Black or African American120 (4.2%)596 (4.1%)408 (5.7%)2044 (5.7%)512 (5.7%)2557 (5.7%)117 (4.8%)581 (4.8%)59 (6.2%)300 (6.3%)78 (4.1%)385 (4.0%)19 (2.6%)92 (2.5%)314 (13.7%)1569 (13.7%) Asian or Pacific Islander685 (23.8%)3437 (23.8%)1384 (19.4%)6927 (19.4%)824 (9.2%)4115 (9.2%)518 (21.3%)2599 (21.3%)174 (18.2%)872 (18.2%)325 (16.9%)1615 (16.8%)88 (12.0%)401 (10.9%)245 (10.7%)1229 (10.8%) Others/Unknown21 (0.7%)82 (0.6%)87 (1.2%)421 (1.2%)67 (0.7%)335 (0.7%)29 (1.2%)127 (1.0%)10 (1.0%)31 (0.6%)25 (1.3%)80 (0.8%)10 (1.4%)25 (0.7%)18 (0.8%)85 (0.7%)Ethnicity Non-hispanic2593 (89.9%)12966 (89.9%)5225 (73.2%)26130 (73.2%)8078 (90.1%)40396 (90.2%)1917 (78.7%)9598 (78.8%)777 (81.1%)3875 (80.9%)1550 (80.7%)7716 (80.3%)386 (52.7%)1912 (52.2%)1708 (74.8%)8544 (74.8%) Hispanic291 (10.1%)1454 (10.1%)1916 (26.8%)9575 (26.8%)883 (9.9%)4409 (9.8%)519 (21.3%)2582 (21.2%)181 (18.9%)915 (19.1%)371 (19.3%)1889 (19.7%)347 (47.3%)1753 (47.8%)576 (25.2%)2876 (25.2%)Stage I611 (21.2%)3066 (21.3%)2922 (40.9%)14609 (40.9%)402 (4.5%)2000 (4.5%)345 (14.2%)1722 (14.1%)105 (11.0%)532 (11.1%)NANANANANANA II210 (7.3%)1043 (7.2%)2642 (37.0%)13213 (37.0%)6478 (72.3%)32401 (72.3%)409 (16.8%)2047 (16.8%)233 (24.3%)1162 (24.3%)NANANANANANA III607 (21.0%)3034 (21.0%)1066 (14.9%)5336 (14.9%)1381 (15.4%)6900 (15.4%)737 (30.3%)3691 (30.3%)154 (16.1%)770 (16.1%)NANANANANANA IV1456 (50.5%)7277 (50.5%)511 (7.2%)2547 (7.1%)700 (7.8%)3504 (7.8%)945 (38.8%)4720 (38.8%)466 (48.6%)2326 (48.6%)NANANANANANADiagnosis year2004-2008649 (22.5%)3254 (22.6%)1448 (20.3%)7238 (20.3%)3717 (41.5%)18590 (41.5%)551 (22.6%)2756 (22.6%)159 (16.6%)804 (16.8%)426 (22.2%)2134 (22.2%)138 (18.8%)690 (18.8%)364 (15.9%)1821 (15.9%)2009-2012636 (22.1%)3188 (22.1%)1577 (22.1%)7891 (22.1%)2688 (30.0%)13433 (30.0%)573 (23.5%)2867 (23.5%)208 (21.7%)1047 (21.9%)363 (18.9%)1823 (19.0%)144 (19.6%)746 (20.4%)472 (20.7%)2366 (20.7%)2013-2016708 (24.5%)3550 (24.6%)1949 (27.3%)9742 (27.3%)1351 (15.1%)6756 (15.1%)576 (23.6%)2895 (23.8%)262 (27.3%)1297 (27.1%)466 (24.3%)2318 (24.1%)200 (27.3%)994 (27.1%)641 (28.1%)3203 (28.0%)2017-2020891 (30.9%)4428 (30.7%)2167 (30.3%)10834 (30.3%)1205 (13.4%)6026 (13.4%)736 (30.2%)3662 (30.1%)329 (34.3%)1642 (34.3%)666 (34.7%)3330 (34.7%)251 (34.2%)1235 (33.7%)807 (35.3%)4030 (35.3%) Table 1Demographic and Clinical Characteristics of Patients in the Propensity Score-Matched CohortTable 1 dummy alt textCharacteristicsNSCLCBREASTPROSTATECOLORECTALPANCREASACUTE MYELOID LEUKEMIAACUTE LYMPHOID LEUKEMIAMULTIPLE MYELOMACOHSEERCOHSEERCOHSEERCOHSEERCOHSEERCOHSEERCOHSEERCOHSEER(N=2884)(N=14420)(N=7141)(N=35705)(N=8961)(N=44805)(N=2436)(N=12180)(N=958)(N=4790)(N=1921)(N=9605)(N=733)(N=3665)(N=2284)(N=11420)Age, years Mean (SD)67.7 (11.5)68.6 (11.7)57.1 (12.8)57.8 (13.1)64.0 (8.14)64.8 (8.48)61.2 (13.6)62.1 (14.0)66.0 (11.8)66.9 (11.5)56.5 (15.7)57.4 (16.2)44.7 (15.8)44.8 (16.3)59.4 (9.40)60.5 (9.81) Median [Min, Max]68.6 [22.3, 95.8]69.0 [18.0, 90.0]56.9 [19.0, 96.9]57.0 [18.0, 90.0]64.3 [36.8, 94.8]65.0 [30.0, 90.0]60.9 [18.1, 98.3]62.0 [18.0, 90.0]66.2 [18.9, 93.2]67.0 [18.0, 90.0]59.0 [18.2, 94.8]60.0 [18.0, 90.0]44.6 [20.0, 89.2]44.0 [20.0, 90.0]60.3 [24.1, 100]61.0 [18.0, 90.0]Age groups, years 18-4482 (2.8%)387 (2.7%)1137 (15.9%)5675 (15.9%)55 (0.6%)275 (0.6%)249 (10.2%)1243 (10.2%)31 (3.2%)147 (3.1%)402 (20.9%)2010 (20.9%)357 (48.7%)1834 (50.0%)131 (5.7%)623 (5.5%) 45-54279 (9.7%)1404 (9.7%)1864 (26.1%)9321 (26.1%)940 (10.5%)4695 (10.5%)504 (20.7%)2526 (20.7%)100 (10.4%)502 (10.5%)323 (16.8%)1595 (16.6%)146 (19.9%)692 (18.9%)476 (20.8%)2398 (21.0%) 55-64625 (21.7%)3136 (21.7%)2047 (28.7%)10240 (28.7%)3364 (37.5%)16822 (37.5%)625 (25.7%)3121 (25.6%)259 (27.0%)1305 (27.2%)506 (26.3%)2542 (26.5%)137 (18.7%)678 (18.5%)904 (39.6%)4532 (39.7%) 65-74990 (34.3%)4955 (34.4%)1362 (19.1%)6816 (19.1%)3632 (40.5%)18156 (40.5%)608 (25.0%)3037 (24.9%)326 (34.0%)1626 (33.9%)469 (24.4%)2353 (24.5%)72 (9.8%)360 (9.8%)687 (30.1%)3436 (30.1%) 75+908 (31.5%)4538 (31.5%)731 (10.2%)3653 (10.2%)970 (10.8%)4857 (10.8%)450 (18.5%)2253 (18.5%)242 (25.3%)1210 (25.3%)221 (11.5%)1105 (11.5%)21 (2.9%)101 (2.8%)86 (3.8%)431 (3.8%)Gender Female1484 (51.5%)7414 (51.4%)7141 (100%)35705 (100%)NANA1139 (46.8%)5702 (46.8%)449 (46.9%)2238 (46.7%)889 (46.3%)4487 (46.7%)343 (46.8%)1724 (47.0%)1004 (44.0%)5032 (44.1%) Male1400 (48.5%)7006 (48.6%)NANA8961 (100%)44805 (100%)1297 (53.2%)6478 (53.2%)509 (53.1%)2552 (53.3%)1032 (53.7%)5118 (53.3%)390 (53.2%)1941 (53.0%)1280 (56.0%)6388 (55.9%)Race White2058 (71.4%)10305 (71.5%)5262 (73.7%)26313 (73.7%)7558 (84.3%)37798 (84.4%)1772 (72.7%)8873 (72.8%)715 (74.6%)3587 (74.9%)1493 (77.7%)7525 (78.3%)616 (84.0%)3147 (85.9%)1707 (74.7%)8537 (74.8%) Black or African American120 (4.2%)596 (4.1%)408 (5.7%)2044 (5.7%)512 (5.7%)2557 (5.7%)117 (4.8%)581 (4.8%)59 (6.2%)300 (6.3%)78 (4.1%)385 (4.0%)19 (2.6%)92 (2.5%)314 (13.7%)1569 (13.7%) Asian or Pacific Islander685 (23.8%)3437 (23.8%)1384 (19.4%)6927 (19.4%)824 (9.2%)4115 (9.2%)518 (21.3%)2599 (21.3%)174 (18.2%)872 (18.2%)325 (16.9%)1615 (16.8%)88 (12.0%)401 (10.9%)245 (10.7%)1229 (10.8%) Others/Unknown21 (0.7%)82 (0.6%)87 (1.2%)421 (1.2%)67 (0.7%)335 (0.7%)29 (1.2%)127 (1.0%)10 (1.0%)31 (0.6%)25 (1.3%)80 (0.8%)10 (1.4%)25 (0.7%)18 (0.8%)85 (0.7%)Ethnicity Non-hispanic2593 (89.9%)12966 (89.9%)5225 (73.2%)26130 (73.2%)8078 (90.1%)40396 (90.2%)1917 (78.7%)9598 (78.8%)777 (81.1%)3875 (80.9%)1550 (80.7%)7716 (80.3%)386 (52.7%)1912 (52.2%)1708 (74.8%)8544 (74.8%) Hispanic291 (10.1%)1454 (10.1%)1916 (26.8%)9575 (26.8%)883 (9.9%)4409 (9.8%)519 (21.3%)2582 (21.2%)181 (18.9%)915 (19.1%)371 (19.3%)1889 (19.7%)347 (47.3%)1753 (47.8%)576 (25.2%)2876 (25.2%)Stage I611 (21.2%)3066 (21.3%)2922 (40.9%)14609 (40.9%)402 (4.5%)2000 (4.5%)345 (14.2%)1722 (14.1%)105 (11.0%)532 (11.1%)NANANANANANA II210 (7.3%)1043 (7.2%)2642 (37.0%)13213 (37.0%)6478 (72.3%)32401 (72.3%)409 (16.8%)2047 (16.8%)233 (24.3%)1162 (24.3%)NANANANANANA III607 (21.0%)3034 (21.0%)1066 (14.9%)5336 (14.9%)1381 (15.4%)6900 (15.4%)737 (30.3%)3691 (30.3%)154 (16.1%)770 (16.1%)NANANANANANA IV1456 (50.5%)7277 (50.5%)511 (7.2%)2547 (7.1%)700 (7.8%)3504 (7.8%)945 (38.8%)4720 (38.8%)466 (48.6%)2326 (48.6%)NANANANANANADiagnosis year2004-2008649 (22.5%)3254 (22.6%)1448 (20.3%)7238 (20.3%)3717 (41.5%)18590 (41.5%)551 (22.6%)2756 (22.6%)159 (16.6%)804 (16.8%)426 (22.2%)2134 (22.2%)138 (18.8%)690 (18.8%)364 (15.9%)1821 (15.9%)2009-2012636 (22.1%)3188 (22.1%)1577 (22.1%)7891 (22.1%)2688 (30.0%)13433 (30.0%)573 (23.5%)2867 (23.5%)208 (21.7%)1047 (21.9%)363 (18.9%)1823 (19.0%)144 (19.6%)746 (20.4%)472 (20.7%)2366 (20.7%)2013-2016708 (24.5%)3550 (24.6%)1949 (27.3%)9742 (27.3%)1351 (15.1%)6756 (15.1%)576 (23.6%)2895 (23.8%)262 (27.3%)1297 (27.1%)466 (24.3%)2318 (24.1%)200 (27.3%)994 (27.1%)641 (28.1%)3203 (28.0%)2017-2020891 (30.9%)4428 (30.7%)2167 (30.3%)10834 (30.3%)1205 (13.4%)6026 (13.4%)736 (30.2%)3662 (30.1%)329 (34.3%)1642 (34.3%)666 (34.7%)3330 (34.7%)251 (34.2%)1235 (33.7%)807 (35.3%)4030 (35.3%) Demographic and Clinical Characteristics of Patients in the Propensity Score-Matched Cohort Survival Outcomes by Cancer Type and Stratified Patients treated at COH had significantly longer OS compared with those in the SEER cohort across all examined cancer types, using a multivariable model with adjustment for key baseline characteristics in both the original and PSM populations. In the PSM analyses, the median OS estimates, when reached, were uniformly longer in the COH cohort compared with the SEER cohort. Adjusted hazard ratios (HRs) favored COH for NSCLC (HR 0.73, 95% CI: 0.70-0.76), colorectal cancer (HR 0.74, 95% CI: 0.70-0.79), pancreatic cancer (HR 0.75, 95% CI: 0.70-0.81), AML (HR 0.79, 95% CI: 0.74-0.84), ALL (HR 0.77, 95% CI: 0.69-0.87), and multiple myeloma (HR 0.70, 95% CI: 0.66-0.76). For breast and prostate cancer, median OS was not reached in either cohort; however, 5-year survival rates remained higher at COH compared with SEER (breast cancer: 87% vs 84%; prostate cancer: 93% vs 89%). Across all cancer types, the direction and magnitude of benefit favored COH, with P < .001 for all comparisons (Figure 1 and Table 2). Comparisons of the unmatched cohorts after multivariable adjustment for key baseline characteristics are provided in the Supplement (Supplementary Figure 1, Table 2). The direction and significance of the survival benefit was consistent across the original population and PSM population, supporting the robustness of the findings.Figure 1Overall survival comparisons between propensity score-matched City of Hope and SEER cohorts by Kaplan-Meier analysis for 8 cancer types. (A) NSCLC. (B) Breast cancer. (C) Prostate cancer. (D) Colorectal cancer. (E) Pancreas cancer. (F) Acute myeloid leukemia. (G) Acute lymphoid leukemia. (H) Multiple myeloma. CI = confidence interval; HR = hazard ratio; mOS = median overall survival; NR = not reached; NSCLC = non-small cell lung cancer; OS = overall survival.Figure 1 dummy alt textTable 2Comparison of Median Overall Survival and Hazard Ratios Between City of Hope and SEER Cohorts Across Cancer TypesTable 2 dummy alt textCancer TypeCohortBased on Original PopulationBased on PSM PopulationMedian OS (y), 95% CIHazard Ratio (COH vs SEER)Median OS (y), 95% CIHazard Ratio (COH vs SEER)NSCLCCOH2.20 (2.05, 2.33)0.67 (0.64-0.70, P < .001)2.20 (2.05, 2.33)0.73 (0.70-0.76, P < .001)SEER1.25 (1.25, 1.25)ref1.42 (1.42, 1.50)refBreast cancerCOHNR0.77 (0.73-0.81, P < .001)NR0.83 (0.78-0.88, P < .001)SEERNRrefNRrefProstate cancerCOHNR0.55 (0.53-0.58, P < .001)NR0.62 (0.58-0.65, P < .001)SEER16.0 (15.9, 16.0)refNRrefColorectal cancerCOH7.15 (6.38, 8.01)0.70 (0.66-0.74, P < .001)7.15 (6.38, 8.01)0.74 (0.70-0.79, P < .001)SEER6.83 (6.83, 6.92)ref5.17 (4.92, 5.50)refPancreatic cancerCOH1.14 (1.06, 1.22)0.71 (0.66-0.76, P < .001)1.14 (1.06, 1.22)0.75 (0.70-0.81, P < .001)SEER0.67 (0.67, 0.67)ref0.75 (0.75, 0.83)refAcute myeloid leukemiaCOH2.37 (2.09, 2.80)0.73 (0.69-0.78, P < .001)2.37 (2.09, 2.80)0.79 (0.74-0.84, P < .001)SEER1.00 (0.92, 1.00)ref1.50 (1.42, 1.50)refAcute lymphoid leukemiaCOH6.36 (4.78, 13.45)0.74 (0.66-0.83, P < .001)6.36 (4.78, 13.45)0.77 (0.69-0.87, P < .001)SEER2.42 (2.25, 2.58)ref3.67 (3.25, 4.08)refMultiple myelomaCOH9.4 (8.81, 10.05)0.64 (0.60-0.69, P < .001)9.40 (8.81, 10.05)0.70 (0.66-0.76, P < .001)SEER4.5 (4.50, 4.58)refrefCI = confidence interval; NE = not estimable; NR = not reached; OS = overall survival; PSM = propensity score-matching; Ref = reference.All hazard ratios are adjusted for age category, gender (except for breast and prostate cancer), race, ethnicity, AJCC stage (for solid tumors), and diagnosis year. PSM was propensity score matched, with the hazard ratio estimated in a model that retained adjustment for those key characteristics. Overall survival comparisons between propensity score-matched City of Hope and SEER cohorts by Kaplan-Meier analysis for 8 cancer types. (A) NSCLC. (B) Breast cancer. (C) Prostate cancer. (D) Colorectal cancer. (E) Pancreas cancer. (F) Acute myeloid leukemia. (G) Acute lymphoid leukemia. (H) Multiple myeloma. CI = confidence interval; HR = hazard ratio; mOS = median overall survival; NR = not reached; NSCLC = non-small cell lung cancer; OS = overall survival. Table 2Comparison of Median Overall Survival and Hazard Ratios Between City of Hope and SEER Cohorts Across Cancer TypesTable 2 dummy alt textCancer TypeCohortBased on Original PopulationBased on PSM PopulationMedian OS (y), 95% CIHazard Ratio (COH vs SEER)Median OS (y), 95% CIHazard Ratio (COH vs SEER)NSCLCCOH2.20 (2.05, 2.33)0.67 (0.64-0.70, P < .001)2.20 (2.05, 2.33)0.73 (0.70-0.76, P < .001)SEER1.25 (1.25, 1.25)ref1.42 (1.42, 1.50)refBreast cancerCOHNR0.77 (0.73-0.81, P < .001)NR0.83 (0.78-0.88, P < .001)SEERNRrefNRrefProstate cancerCOHNR0.55 (0.53-0.58, P < .001)NR0.62 (0.58-0.65, P < .001)SEER16.0 (15.9, 16.0)refNRrefColorectal cancerCOH7.15 (6.38, 8.01)0.70 (0.66-0.74, P < .001)7.15 (6.38, 8.01)0.74 (0.70-0.79, P < .001)SEER6.83 (6.83, 6.92)ref5.17 (4.92, 5.50)refPancreatic cancerCOH1.14 (1.06, 1.22)0.71 (0.66-0.76, P < .001)1.14 (1.06, 1.22)0.75 (0.70-0.81, P < .001)SEER0.67 (0.67, 0.67)ref0.75 (0.75, 0.83)refAcute myeloid leukemiaCOH2.37 (2.09, 2.80)0.73 (0.69-0.78, P < .001)2.37 (2.09, 2.80)0.79 (0.74-0.84, P < .001)SEER1.00 (0.92, 1.00)ref1.50 (1.42, 1.50)refAcute lymphoid leukemiaCOH6.36 (4.78, 13.45)0.74 (0.66-0.83, P < .001)6.36 (4.78, 13.45)0.77 (0.69-0.87, P < .001)SEER2.42 (2.25, 2.58)ref3.67 (3.25, 4.08)refMultiple myelomaCOH9.4 (8.81, 10.05)0.64 (0.60-0.69, P < .001)9.40 (8.81, 10.05)0.70 (0.66-0.76, P < .001)SEER4.5 (4.50, 4.58)refrefCI = confidence interval; NE = not estimable; NR = not reached; OS = overall survival; PSM = propensity score-matching; Ref = reference.All hazard ratios are adjusted for age category, gender (except for breast and prostate cancer), race, ethnicity, AJCC stage (for solid tumors), and diagnosis year. PSM was propensity score matched, with the hazard ratio estimated in a model that retained adjustment for those key characteristics. Comparison of Median Overall Survival and Hazard Ratios Between City of Hope and SEER Cohorts Across Cancer Types CI = confidence interval; NE = not estimable; NR = not reached; OS = overall survival; PSM = propensity score-matching; Ref = reference. All hazard ratios are adjusted for age category, gender (except for breast and prostate cancer), race, ethnicity, AJCC stage (for solid tumors), and diagnosis year. PSM was propensity score matched, with the hazard ratio estimated in a model that retained adjustment for those key characteristics. Similarly, patients in the COH cohort had significantly longer OS compared with those in the SEER cohort when analyzed by various stages of diagnosis for solid tumors (stage I-IV) or by age for hematologic malignancies (age <60 years or ≥60 years for AML and ALL; age <60 years, 60-70 years, and >70 years for multiple myeloma) (Figures 2 and 3, Supplementary Figures 2-7; Supplementary Tables 2 and 3). Hazard ratios demonstrating lower mortality risk not reaching significance included stage I prostate cancer (median OS not reached in COH cohort, 15 years in SEER cohort), stage II colorectal cancer (median OS 14.8 vs 13.2 years), stage I pancreatic cancer (median OS 3.77 vs 2.08 years), AML in patients <60 years (median OS 6.1 vs 7.8 years), and ALL in patients ≥60 years (median OS 2.02 vs 1.67 years). PSM hazard ratio analysis demonstrated that patients in the COH cohort diagnosed with stage IV solid tumors had lower mortality risk compared with the respective SEER patients: NSCLC HR 0.70 (95% CI: 0.65-0.74), breast cancer HR 0.77 (95% CI: 0.68-0.86), prostate cancer HR 0.58 (95% CI: 0.52-0.66), colorectal cancer HR 0.67 (95% CI: 0.61-0.73), and pancreatic cancer HR 0.72 (95% CI: 0.65-0.80) (all P < .001).Figure 2Overall survival comparisons between propensity score-matched City of Hope and SEER cohorts by Kaplan-Meier analysis for NSCLC stages I-IV. (A) Patients with stage I; (B) Patients with stage II; (C) Patients with Stage III; (D) Patients with Stage IV. CI = confidence interval; HR = hazard ratio; mOS = median overall survival; NR = not reached; OS = overall survival.Figure 2 dummy alt textFigure 3Overall survival comparisons between propensity score-matched City of Hope and SEER cohorts by Kaplan-Meier analysis for breast cancer stages I-IV. (A) Patients with stage I; (B) Patients with stage II; (C) Patients with Stage III; (D) Patients with Stage IV. CI = confidence interval; HR = hazard ratio; mOS = median overall survival; NR = not reached; OS = overall survival.Figure 3 dummy alt text Overall survival comparisons between propensity score-matched City of Hope and SEER cohorts by Kaplan-Meier analysis for NSCLC stages I-IV. (A) Patients with stage I; (B) Patients with stage II; (C) Patients with Stage III; (D) Patients with Stage IV. CI = confidence interval; HR = hazard ratio; mOS = median overall survival; NR = not reached; OS = overall survival. Overall survival comparisons between propensity score-matched City of Hope and SEER cohorts by Kaplan-Meier analysis for breast cancer stages I-IV. (A) Patients with stage I; (B) Patients with stage II; (C) Patients with Stage III; (D) Patients with Stage IV. CI = confidence interval; HR = hazard ratio; mOS = median overall survival; NR = not reached; OS = overall survival. Discussion In this study, we found that patients treated at COH had significantly better OS compared to patients in the SEER registry. This survival advantage was observed across 8 cancer types. Notable findings include lower mortality risk at COH for patients with several advanced-stage cancers, including stage IV NSCLC (HR 0.70) and stage IV breast cancer (HR 0.77), as well as for patients with hematologic malignancies including AML (HR 0.70 for patients aged ≥60 years) and multiple myeloma (HR 0.72 for patients aged ≥70 years). Several other studies have reported that care at a specialized NCI-CCC is associated with superior survival. In a systematic review and meta-analysis, Thamm et al. reported superior mortality and survival outcomes for adults with solid cancer, including 23% lower overall mortality risk, treated at an NCI-CCC vs a noncomprehensive treatment site.20 Wolfson et al. found that patients newly diagnosed with various adult-onset solid cancers in Los Angeles County (1998-2008) who were first treated at a care setting other than an NCI-CCC had a 20%-50% increased risk of mortality, adjusted for age at diagnosis, gender, race/ethnicity, stage of disease, socioeconomic status, payor, and distance to the nearest center.21 Bruno et al. found that U.S. patients treated for small lymphocytic lymphoma/chronic lymphocytic leukemia or Mantle cell lymphoma (2013-2022) had significantly longer survival when treated at an academic center vs a community clinic, adjusted for diagnosis year, age, gender, race/ethnicity, socioeconomic status, clinical practice size, insurance type, and disease-specific variables.22 Prior studies hypothesize several factors potentially leading to superior outcomes at a tertiary cancer center, including care quality, cutting-edge treatments and protocols, clinical trial enrollment, multidisciplinary disease management, systems for integrated supportive care both within the center and in the community, and initiatives implementing evidence-based strategies tailored to local populations.8,21, 22, 23, 24, 25, 26, 27 In 2024, Unger et al. reported that 21.6% vs <6% of patients are involved in clinical trials at NCI-CCCs compared with other academic centers, care networks, and community practices.28 Trials may offer personalized treatment strategies to benefit patients with relapsed, treatment-resistant, and persistent cancers. The evidence on how clinical trials affect survival is mixed. For instance, recent clinical trials of gemcitabine and abraxane for metastatic pancreatic cancer reported superior median OS (10 months29) vs our stage IV cohort, which includes patients who would not be eligible for a trial owing to poor performance status or other issues related to frailty or inability to receive therapy. In a 2013 population-based study, Chow et al. reported a modest survival benefit for patients with several solid organ cancers who were enrolled in clinical trials, but they postulated confounding factors such as advantaged patients being more able to enroll in trials.30 Similarly, in 2018, Ramalingam et al. observed longer survival in patients with stage IV lung cancer treated at an academic center vs community cancer sites.31 A specific improvement for patients with adenocarcinoma led them to speculate that recent treatment advances for that cancer may have been more available at the academic centers, contributing to the observed survival benefit. Nonetheless, recent meta-analyses by Iskander et al. in 2024 found that an apparent survival benefit for participants in clinical trials did not persist when controlling for confounding factors32 or that any survival benefit was small and potentially outweighed by toxicity.33 Advanced diagnostics including tumor profiling, biomarker testing, and advanced imaging allow treatment to be informed by the molecular characteristics of each cancer. Furthermore, molecular tumor boards, pyramidal decision-support structures, integrated data systems and diagnostic tools, and artificial intelligence (AI) support precision medicine.34, 35, 36, 37, 38 Patient-centered, coordinated care may also contribute to improved survival outcomes and quality of life. In our analysis, differences in survival were especially notable in cancers with generally poor prognoses, where expertise and early intervention may have a greater impact on outcomes. Finally, subspecialized clinical practice may also contribute to patient outcomes via availability of interventions such as stem cell transplantation and CAR T cell therapy39,40 and treatments guided by tumor-specific mutations and biology.41,42 As of early 2025, there were 9 Food and Drug Administration (FDA)-approved tumor-agnostic therapies, compared to hundreds of mutation-targeted and tumor-specific drugs.6,43 We hypothesize that NCI-CCCs are also well equipped to use tumor-agnostic therapies that nonetheless require deep understanding of specific tumor characteristics and genomics. We expect the benefits of CCCs to generally apply to all centers alike, such that our results are broadly similar to others. This study highlights a national gap in cancer outcomes data collection and sharing in existing registries. The absence of such data at the national level leaves patients and families with a key missing variable in finding the highest value and optimal outcomes for their cancer care. Our study, therefore, is intended to be a call to action to all cancer centers and cancer care providers to develop standardized, transparent, and appropriately risk-adjusted models that would allow for public reporting and center accountability for survival outcomes. A successful, government-mandated survival outcome model for blood and marrow transplant patients already exists, and we believe that it can be a blueprint for the collaborative creation of a similarly transparent model for survival outcomes for all cancers. This paper advances this much-needed process. Limitations Despite adjusting for key demographic and clinical variables, this study has several limitations related to those of the SEER dataset. This dataset lacks important variables affecting cancer survival, including comorbidities, insurance status (post-2019), performance status, organ function, socioeconomic factors and geolocation, educational status, treatment intent and detailed history (chemotherapy regimen, radiation dose, surgical quality, etc.), and molecular testing results, as well as institutional characteristics and care delivery metrics. Because we were unable to control for these variables, which were not included in the analysis, residual confounding is possible. Other databases such as National Cancer Database (NCDB) are also limited by lack of key data points, including comorbidities, socioeconomic factors, and molecular testing results. The SEER dataset includes a large, heterogeneous U.S. cohort of patients treated in differing care settings, whereas the COH cohort reflects the demographics in our Southern California catchment area and includes patients treated at a tertiary center. As the COH cancer registry did not capture the cause of death in a format directly comparable to SEER’s death-cause specific classification, all-cause death was utilized for both datasets to evaluate survival across tumors. However, it should be recognized that this approach may overestimate cancer-attributable mortality by including deaths unrelated to primary cancer, especially for cancers with low fatality rates or longer expected survival. Known disparities related to cancer treatment and survival outcomes in different settings, including rural vs urban treatment sites, community hospitals vs academic centers, and differing clinical trial access between populations, may therefore also affect the results. As Los Angeles and the region contribute data to the SEER program, there may be patient overlap that is unaccounted for that would impact the independence of the estimates, although the proportional contribution of COH data to SEER is small and unlikely to impact the results. Despite the mixed data on the benefits of clinical trial enrollment for survival, it is notable that our dataset includes all patients treated and therefore the survival rates may be lower than those observed in randomized clinical trials. These prior studies of clinical trial benefit, discussed above, suggest the importance of controlling for survival-associated confounders. Unfortunately, the existing databases reporting cancer treatment outcomes do not adequately report these variables, and we believe this further supports the need for standardization in data reporting. Referral bias is another potential confounder. Patients treated at a tertiary center may differ systematically and be healthier when they seek care, more likely to travel for care, have better functional status, or be more engaged in the healthcare system, compared with those treated in community settings who are included in the SEER dataset. In addition, certain disease populations, such as AML and ALL, are rare and clinically complex, and are frequently referred to tertiary care centers, potentially attenuating observable differences between datasets. While we used propensity score matching to balance measured characteristics between the 2 cohorts, unmeasured factors related to referral patterns could still introduce bias and may partially explain the early survival advantage for patients in the COH cohort. Furthermore, the COH dataset may differ from other NCI-CCCs as it includes a clinical network of coordinated care delivery beyond our academic cancer center. Despite these limitations, our results were similar in the adjusted and propensity score matched analyses and consistent with prior reports evaluating tertiary cancer centers. Conclusion Overall, our findings suggest that specialized cancer care at an NCI-CCC is associated with longer survival across a range of malignancies. Future studies could consider data from other sources such as the NCDB and other cancer centers, which may allow for comparison of the long-term trends in cancer survival in different settings and for analysis of more detailed center, patient, and treatment characteristics. Despite substantial advances in cancer care and technology in the last decade, public data have lagged. Genomics approaches and precision medicine offer great promise but also risk exacerbating existing disparities in patient outcomes. Follow-up studies to this analysis should spur further research to identify the drivers of observed survival advantages and promote the development of better guidelines and care models to improve outcomes for all patients. Funding Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under award number P30CA033572. This research was supported in part by the Deana and Steve Campbell Chief Physician Executive Distinguished Chair in Honor of Alexandra Levine, M.D., and the Morgan and Helen Chu Presidential Chair endowments to Marcel van den Brink. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. CRediT authorship contribution statement Ravi Salgia: Writing – review & editing, Writing – original draft, Visualization, Project administration, Conceptualization. Joseph Alvarnas: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Conceptualization. Vijay Trisal: Writing – review & editing, Visualization, Validation, Supervision, Conceptualization. Ann Vanderplas: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Formal analysis, Data curation. Xiaochen Li: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Formal analysis, Data curation. Javier Arias-Romero: Writing – review & editing, Writing – original draft, Visualization, Validation. Jeremy Fricke: Writing – review & editing, Writing – original draft, Visualization, Validation. Paul Frankel: Writing – review & editing, Visualization, Validation, Formal analysis. Isa Mambetsariev: Writing – review & editing, Writing – original draft, Visualization, Validation. Hannah Rice: Writing – review & editing, Writing – original draft, Visualization, Validation. Harlan Levine: Writing – review & editing, Validation, Supervision. Marcel R.M. van den Brink: Writing – review & editing, Writing – original draft, Supervision, Conceptualization. Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Ravi Salgia reports that financial support for this work was provided by the National Institutes of Health. Harlan Levine (HL) reports a relationship with Reimagine Care as a City of Hope (COH) representative (uncompensated/nonfinancial support). He also reports compensated consulting or advisory relationships with Laguna Health, Quantum Health Inc, Vida Health, and Active Life Scientific. Marcel RM van den Brink (MVDB) has received research support from Seres Therapeutics and holds equity or stock options in Seres Therapeutics and ThymoFox. He has received royalties from Wolters Kluwer and Juno and has intellectual property licensing with Juno Therapeutics. He has consulted for, received honoraria from, or participated in advisory boards for Seres Therapeutics, ThymoFox, Garuda Therapeutics, Inc., Novartis (spouse), Bristol Myers Squibb (spouse), Galapagos NV (spouse), and BeOne Medicines Ltd (spouse). He also reports board membership with Garuda Therapeutics, Inc., Novartis (spouse), Bristol Myers Squibb Company (spouse), Galapagos NV (spouse), BeOne Medicines Ltd (spouse), the Foundation Board of DKMS (a nonprofit organization; fiduciary role), and Smart Immune (chairman of the Scientific Advisory Board). The other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.