Patient and Treatment Characteristics
Data were obtained from NCDB for 101,067 patients with BMs diagnosed between 2010 and 2015. Patients excluded included those who were M0 stage or were missing information related to the M stage, facility type, and treatment variables (8,434) (Fig. 1a). The final analysis included 93,633 patients, among whom, 31,579 (34.09%) were treated at academic facilities, and 61,054 (65.91%) were treated at non-academic treatment facilities. There were 1,317 facilities, among which 226/1,317 (17.2%) were academic facilities, and the remaining 1,091/1,317 (82.8%) were non-academic facilities. On average, each academic facility treated 23.3 cases/year, while each non-academic facility treated 9.3 cases/year. The median age of the entire study population was 65 with a range of (40–90) years. The median age of diagnosis of the patients treated at academic facilities was 64 (40–90), while the median age at diagnosis of the patients treated at non-academic facilities was 66 (40–90). The majority of the patients were White, from high income-level areas, had insurance, did not receive surgery of the primary site, received RT, had a comorbidity score of zero, were diagnosed between 2010–2013, and had NSCLC. The trend of receiving treatment at academic and non-academic hospitals over time is illustrated in Fig. 1b.
Outcomes
Younger age, black race, non-white non-black, living in urban areas, living in areas with a high proportion of people with no high school degree, living in areas with high income-level, having comorbidity score of zero, receiving surgery of the primary tumor, RT, chemotherapy, immunotherapy, and diagnosis in 2014 or after were more likely to be treated at academic facilities compared to non-academic facilities. Patients belonging to areas with income <$35,000 were 16% less likely to receive treatment at an academic facility (OR: 0.839, CI: 0.811–0.868) compared to their counterparts from the areas with income => $35,000. Patients from rural counties were 27% less likely to receive treatment at academic facilities (OR: 0.726, CI: 0.657–0.803) compared to patients from urban areas. Patients who did not undergo surgery of the primary site tumor were 10% less likely (OR: 0.90, CI: 0.83–0.98), patients who did not receive chemotherapy were 4% less likely (OR: 0.96, CI: 0.93–0.99), patients who did not receive RT were 4% less likely (OR: 0.96, CI: 0.93–0.99), and patients who did not receive immunotherapy were 10% less likely (OR: 0.90, CI: 0.83–0.97) to receive treatment at academic facilities compared to their counterparts. The characteristics of the patients and the OR of factors associated with receiving treatment at academic facilities are provided in Table 1.
Table 1
Multivariable logistic analysis with the probability of being treated in academic hospital in BMs patients
Variable | Academic 31,579 (34.1%) | Non-academic 61,054 (65.9%) | Total 92,633 | OR (95% CI) | P |
Age at diagnosis, Median (range) | 64.00 (40–90) | 66.00 (40–90) | 65.00 (40–90) | 0.99 (0.99–0.99) | 0.001 |
Sex | Male | 15,998 (50.7) | 31,130 (51.0) | 47,128 (50.9) | Reference | |
Female | 15,581 (49.3) | 29,924 (49.0) | 45,505 (49.1) | 1.02 (0.99–1.05) | 0.23 |
Race | White | 24,790 (79.4) | 53,343 (87.9) | 78,133 (85.0) | Reference | |
Black | 5005 (16.0) | 5,596 (9.2) | 10,601 (11.5) | 1.94 (1.86–2.03) | 0.001 |
Other | 1,429 (4.6) | 1,785 (2.9) | 3,214 (3.5) | 1.58 (1.47–1.70) | 0.001 |
Unknown | 355 | 330 | 685 | | |
Education | >=13% NHD | 15,228 (48.3) | 28,622 (47.0) | 43,850 (47.4) | 1.07 (1.04–1.11) | 0.001 |
< 13% NHD | 16,276 (51.7) | 32,304 (53.0) | 48,580 (52.6) | Reference | |
Unknown | 75 | 128 | 203 | | |
Income | >=$35,000 | 17,473 (55.5) | 32,571 (53.5) | 50,044 (54.2) | Reference | |
< 35,000 | 14,005 (44.5) | 28,319(46.5) | 42,324 (45.8) | 0.84 (0.81–0.87) | 0.001 |
Unknown | 101 | 164 | 265 | | |
Place of Living | Urban | 30,510 (98.2) | 57,724 (97.3) | 88,234 (97.6) | Reference | |
Rural | 547 (1.8) | 1,597 (2.7) | 2,144 (2.4) | 0.73 (0.66–0.80) | 0.001 |
Unknown | 522 | 1,733 | 2,255 | | |
Insurance Status | Yes | 29,057 (96.5) | 56,998 (97.3) | 86,055 (94.4) | Reference | |
No | 1,934 (3.5) | 3,195 (2.7) | 5,129 (5.6) | 1.04 (0.98–1.11) | 0.17 |
Unknown | 588 | 861 | 1,449 | | |
Charlson/Deyo Score | 0 | 21,032 (66.6) | 36,986 (60.6) | 58,018 (62.6) | Reference | |
1 | 7,132 (22.6) | 16,130 (26.4) | 23,262 (25.1) | 0.82 (0.79–0.84) | 0.001 |
>=2 | 3,415 (10.8) | 7,938 (13.0) | 11,353 (12.3) | 0.81 (0.77–0.85) | 0.001 |
Primary site surgery | Yes | 1,061 (3.4) | 1,604 (2.6) | 2,665 (2.9) | Ref | |
No | 30,518 (95.6) | 59,450 (97.4) | 89,968 (97.1) | 0.90 (0.83–0.98) | 0.02 |
Chemotherapy | Yes | 17,687 (56.0) | 32,419 (53.1) | 50,106 (54.1) | Reference | |
No | 13,892 (44.0) | 28,635 (46.9) | 42,527 (45.9) | 0.96 (0.93–0.99) | 0.007 |
Radiation Therapy | Yes | 23,373 (74.0) | 43,737 (71.6) | 67,110 (72.5) | Reference | |
No | 8,206 (26.0) | 17,317 (28.4) | 25,523 (27.5) | 0.96 (0.93–0.99) | 0.02 |
Immunotherapy | Yes | 1,217 (3.9) | 1,811 (3.0) | 3,028 (3.3) | Ref | |
No | 30,362 (96.1) | 59,243 (93.0) | 89,605 (96.7) | 0.90 (0.83–0.97) | 0.009 |
Year of Diagnosis | 2010–2013 | 20,224 (64.0) | 40,100 (65.7) | 60,324 (65.1) | 0.93 (0.90–0.96) | 0.001 |
2014–2015 | 11,355 (36.0) | 20,954 (34.3) | 32,309 (34.9) | Reference | |
Primary Cancer Type | Breast | 14,141 (4.5) | 2,522 (4.1) | 3,936 (4.3) | 0.69 (0.62–0.77) | 0.001 |
NSCLC | 21,250 (67.3) | 39,575 (64.8) | 60,825 (65.7) | 0.76 (0.70–0.83) | 0.001 |
SCLC | 4,406 (14.0) | 10,378 (17.0) | 14,784 (15.9) | 0.64 (0.59–0.70) | 0.001 |
Other lung | 1,779 (5.6) | 4,450 (8.0) | 6,229 (6.7) | 0.65 (0.59–0.72) | 0.001 |
Melanoma | 1,224 (3.9) | 1,825 (2.3) | 3,049 (3.3) | 1.02 (0.92–1.15) | 0.67 |
Colorectal | 393 (1.2) | 726 (1.2) | 1,119 (1.2) | 0.74 (0.64–0.86) | 0.001 |
Renal cell | 1,113 (3.5) | 1,578 (2.6) | 2,691 (2.9) | Ref | |
NHD: no high school degree |
Table 2
Univariable and multivariable Cox proportional regression analysis of factors associated with OS in BMs patients
Variable | Univariable analysis | Multivariable analysis |
Hazard Ratio (95% CI) | P | Hazard Ratio (95% CI) | P |
Age at diagnosis (continuous) | 1.02 (1.02–1.03) | 0.001 | 1.01 (1.01–1.01) | 0.001 |
Facility Type | Academic | 0.81 (0.80–0.82) | 0.001 | 0.85 (0.84–0.87) | 0.001 |
Non-academic | Ref | | | |
Sex | Male | Ref | | Ref | |
Female | 0.82 (0.81–0.83) | 0.001 | 0.86 (0.85–0.88) | 0.001 |
Race | White | Ref | | Ref | |
Black | 0.95 (0.93–0.97) | 0.001 | 0.94 (0.91–0.96) | 0.001 |
non-white non-black | 0.69 (0.67–0.72) | 0.001 | 0.73 (0.70–0.76) | 0.001 |
Education | >=13% NHD | 1.07 (1.05–1.08) | 0.001 | … | |
< 13% NHD | Ref | | | |
Income | >=$35,000 | Ref | | Ref | |
<$35,000 | 1.12 (1.10–1.13) | 0.001 | 1.06 (1.04–1.07) | 0.001 |
Place of Living | Urban | Ref | | Ref | |
Rural | 1.08 (1.04–1.134) | 0.004 | …. | |
Insurance Status | Insured | Ref | | | |
Not insured | 1.04 (1.01–1.07) | 0.02 | 1.11 (1.08–1.15) | 0.001 |
Charlson/Deyo Score | 0 | Ref | | Ref | |
1 | 1.25 (1.24–1.28) | 0.001 | 1.14 (1.13–1.16) | 0.001 |
>=2 | 1.50 (1.47–1.53) | 0.001 | 1.23 (1.20–1.26) | 0.001 |
Primary Site Surgery | Yes | Ref | 0.001 | Ref | 0.001 |
No | 2.24 (2.14–2.34) | | 2.13 (2.03–2.23) | |
Chemotherapy | Yes | Ref | | Ref | |
No | 2.37 (2.34–2.40) | 0.001 | 2.19 (2.15–2.22) | 0.001 |
Radiation Therapy | Yes | Reference | | Reference | |
No | 1.62 (1.59–1.64) | 0.001 | 1.25 (1.22–1.27) | 0.001 |
Immunotherapy | Yes | Ref | | Ref | |
No | 1.91 (1.83–1.99) | 0.001 | 1.45 (1.39–1.52) | 0.001 |
Year of Diagnosis | 2010–2013 | 1.09 (1.07–1.10) | 0.001 | 1.06 (1.05–1.08) | 0.001 |
2014–2015 | Ref | | Ref | |
Primary Cancer Type | Breast cancer | 0.72 (0.68–0.76) | 0.001 | 0.74 (0.70–0.78) | 0.001 |
NSCLC | 1.09 (1.04–1.13) | 0.001 | 1.06 (1.02–1.11) | 0.009 |
SCLC | 1.20 (1.15–1.25) | 0.001 | 1.24 (1.19–1.30) | 0.001 |
Other lung | 2.26 (2.16–2.38) | 0.001 | 1.39 (1.32–1.46) | 0.001 |
Melanoma | 1.01 (0.95–1.07) | 0.77 | 0.78 (0.74–0.83) | 0.001 |
Colorectal cancer | 1.20 (1.11–1.29) | 0.001 | 1.26 (1.166–1.36) | 0.001 |
Renal cell | Ref | | Ref | |
NHD: no high school degree Ref = reference |
Based on the Kaplan Meier curves, patients who received treatment at an academic facility had significantly improved OS with an absolute median OS benefit of 1.61 (6.18, CI: 6.05–6.31 vs. 4.57, CI: 4.50–4.63; p < 0.001) months compared to patients who were treated at non-academic facilities (Fig. 2a). The 1-year and 2-year survival rates were 32% (CI: 31%-32%) and 16% (CI: 16%-17%) in patients treated at academic hospitals vs. 24% (CI: 24%-25%) and 11% (CI: 10%-11%) in patients treated at non-academic hospitals. The median OS of patients treated at academic hospitals was longer compared to patients who were treated at community hospitals among most of the treatment options. The KM curves by treatment options are provided in Figure (3a-4d).
In the univariate Cox regression analysis, receiving treatment at academic hospitals, younger age, female sex, black race, non-white non-black race, living in areas with income =>$35,000, living in areas with high education level, living in the rural areas, having insurance, comorbidity score of 0, surgery of the primary cancer type, chemotherapy, RT, immunotherapy, diagnosis between 2014 and 2015 and primary cancer type of breast (renal cancer) were all associated with improved OS.
In the multivariable Cox regression analysis adjusted for the age of diagnosis, race, sex, income level, education, place of living, insurance status, surgery of the primary site, chemotherapy, RT, immunotherapy, year of diagnosis, and primary cancer type, receiving treatment at an academic facility was associated with significantly improved OS (HR: 0.85, CI: 0.84–0.87; p < 0.001) compared to receiving treatment at a non-academic facility. Other variables associated with significantly improved OS were young age, female sex, black race, non-white non-black race, having insurance, living in an area with an income level of =>$35,000, comorbidity score of zero, receiving surgery of the primary site, receiving chemotherapy, RT, immunotherapy, diagnosis in 2014 or after, and primary cancer type of breast, and melanoma (compared to renal cell). The findings stayed significant after stratifying by comorbidity score and age of diagnosis. To make sure that the findings of our study are not affected by immortal time bias, we conducted an analysis restricted to only patients who received all of the first course treatment at the reporting facility. Treatment at an academic center still remained significantly associated with improved OS compared to treatment at a non-academic center (HR: 0.819, CI: 0.802–0.836; p < 0.001). The survival benefit of receiving treatment at an academic center became more significant compared to treatment at a non-academic center, an indication that our results underestimated the improved OS associated with receiving treatment at an academic center.
We also performed the stratified analysis by treatment and compared the OS of patients treated at academic hospitals vs. non-academic hospitals. Among patients who only received brain RT, receiving treatment at an academic facility was associated with significantly improved OS (HR: 0.84, CI: 0.82–0.87; p < 0.001) compared to receiving treatment at a non-academic facility. Treatment at academic hospitals was associated with improved OS compared to treatment at non-academic hospitals among patients who only received RT to other than brain (HR: 0.88, CI: 0.83–0.93; p < 0.001), patients who received chemotherapy plus brain RT (HR: 0.85, CI: 0.83–0.87; p < 0.001), and patients who received surgery of the primary site plus chemotherapy plus brain RT (HR: 0.75, CI: 0.64–0.88; p < 0.003). The HR of academic vs. non-academic stratified by various treatment combinations is provided in Table 3. We further compared survival outcomes for academic comprehensive cancer programs (ACCPs), integrated network cancer programs (INCPs), comprehensive community cancer programs (CCCPs), and community cancer programs (CCPs) to check if our findings stand beyond the academic vs. non-academic facilities. Patients treated at academic comprehensive cancer programs had significantly improved OS compared to each of the other types of facilities (Fig. 2b). The findings stayed significant when stratified by comorbidity score and age of diagnosis (data not shown).
Table 3
Multivariable Cox analysis of academic vs. non-academic (reference) stratified by treatment combinations
Combinations | N (%) | Multivariable Cox analysis HR ((95% CI) | |
Only brain RT | 17,879 (19.7) | 0.84 (0.82–0.87) | 0.001 |
Only other RT | 5,790 (6.4) | 0.88 (0.83–0.93) | 0.001 |
Only chemotherapy | 7,033 (7.8) | 0.83 (0.78–0.88) | 0.001 |
Only surgery | 335 (0.4) | … | |
Chemotherapy plus brain RT | 31,341 (34.5) | 0.85 (0.83–0.87) | 0.001 |
Chemotherapy plus other RT | 9,663 (10.7) | 0.86 (0.82–0.90) | 0.001 |
Surgery plus brain RT | 660 (0.7) | 0.78 (0.64–0.94) | 0.009 |
Surgery plus other RT | 78 (0.1) | … | |
Surgery plus chemotherapy | 280 (0.3) | 0.74 (0.55–0.99) | 0.046 |
Surgery plus chemotherapy plus brain RT | 1,063 (1.2) | 0.75 (0.64–0.88) | 0.003 |
Surgery plus chemotherapy plus other RT | 210 (0.2) | … | |
No surgery no chemotherapy no RT | 16,398 (18.0) | 0.91 (0.88–0.95) | 0.001 |
RT: radiation therapy; The results of only surgery, surgery plus other RT, and surgery plus chemotherapy plus other RT were not significant. |