Disparities in Ovarian Cancer Care Quality and Survival According to Race and Socioeconomic Status

Abstract

Background

The relationship between racial and socioeconomic status (SES) disparities and the quality of epithelial ovarian cancer care and survival outcome are unclear.

Methods

A population-based analysis of National Cancer Data Base (NCDB) records for invasive primary epithelial ovarian cancer diagnosed in the period from 1998 to 2002 was done using data from patients classified as white or black. Adherence to National Comprehensive Cancer Network (NCCN) guideline care was defined by stage-appropriate surgical procedures and recommended chemotherapy. The main outcome measures were differences in adherence to NCCN guidelines and overall survival according to race and SES and were analyzed using binomial logistic regression and multilevel survival analysis.

Results

A total of 47 160 patients (white = 43 995; black = 3165) were identified. Non-NCCN-guideline-adherent care was an independent predictor of inferior overall survival (hazard ratio [HR] = 1.43, 95% confidence interval [CI] = 1.38 to 1.47). Demographic characteristics independently associated with a higher likelihood of not receiving NCCN guideline-adherent care were black race (odds ratio [OR] = 1.36, 95% CI = 1.25 to 1.48), Medicare payer status (OR = 1.20, 95% CI = 1.12 to 1.28), and not insured payer status (OR = 1.33, 95% CI = 1.19 to 1.49). After controlling for disease and treatment-related variables, independent racial and SES predictors of survival were black race (HR = 1.29, 95% CI = 1.22 to 1.36), Medicaid payer status (HR = 1.29, 95% CI = 1.20 to 1.38), not insured payer status (HR = 1.32, 95% CI = 1.20 to 1.44), and median household income less than $35 000 (HR = 1.06, 95% CI = 1.02 to 1.11).

Conclusions

These data highlight statistically and clinically significant disparities in the quality of ovarian cancer care and overall survival, independent of NCCN guidelines, along racial and SES parameters. Increased efforts are needed to more precisely define the patient, provider, health-care system, and societal factors leading to these observed disparities and guide targeted interventions.

Despite improvements in cancer care during the past two decades, persistent disparities exist in the United States with regard to receipt of recommended guideline therapy and survival. Disparities in cancer care and outcomes have been linked to race and ethnic classification, socioeconomic status (SES), educational level, geographic locale, culture, and health system factors (1–3). For many types of cancer, blacks have lower stage-specific survival than whites (4).

In the United States, there are 22 000 new cases of ovarian cancer diagnosed and more than 15 000 disease-related deaths annually (5). Racial disparities in ovarian cancer have been documented with respect to stage of diagnosis, extent of treatment, and survival (6–9). There is more limited data on ovarian cancer disparities associated with SES characteristics (10–13). Previous studies have been limited by analyses of race without the context of SES indicators, correlation of race with either treatment or survival but not both outcomes in the same population, an absence of validated standards of quality care, or small study groups that may not be representative of the broader population. As a result, accurately quantifying the magnitude of ovarian cancer disparities and characterizing those sociodemographic characteristics associated with limited access to care, substandard care, and worse clinical outcomes remain important prerequisites to developing corrective measures.

The National Cancer Data Base (NCDB) is an oncology outcomes database of approximately 70% of all newly diagnosed patients with cancer in the United States. Previous studies of ovarian cancer using the NCDB have found correlations between both race and payer status and receipt of surgery as initial therapy but have not benchmarked treatment to validated quality outcome measures (12–14). The objective of this study was to examine disparities according to race and SES indicators in the quality of epithelial ovarian cancer care and survival outcome. For the purposes of this study, adherence to National Comprehensive Cancer Network (NCCN) guideline therapy was selected as the measure of quality of cancer care and considered the therapeutic standard that the majority of ovarian cancer patients should be provided.

Methods

Case Ascertainment and Definition of Variables

This study received exempt status from the Institutional Review Board of Washington University. All NCDB cases of invasive malignant epithelial ovarian cancer classified as white or black diagnosed between January 1, 1998, and December 31, 2002, were selected from the NCDB using topography code C56.9; individual subjects and facilities were de-identified before delivery of the public use file. Records were included if the tumor corresponded to one malignant or the first of two or more independent malignant primary tumors and if either the pathological or clinical staging was known. Histology codes were sorted into serous, mucinous, endometrioid, clear-cell, mixed, and undifferentiated tumors. Nonepithelial ovarian cancer, borderline cancers, and tumors of nonovarian origin were excluded. Tumor grade was consolidated into two categories: well/moderately differentiated and poorly differentiated/undifferentiated/anaplastic. A new overall staging variable was constructed based on 1) complete pathologic staging or 2) clinical staging if pathologic staging was missing or improperly reported.

The average annual hospital ovarian cancer volume was ranked into quartiles (1–6 cases, 7–14 cases, 15–25 cases, and ≥26 cases). Zipcode of residence was matched against year 2000 US Census data and US Department of Agriculture data to estimate median household income, the percentage of residents with a college degree, and the continuum of rural/urban residence. Payer status was consolidated into six categories. Private insurance included fee-for-service, health maintenance organization, or independent physician association. Patients with managed care insurance, TRICARE, or other military coverage were grouped under a single category (managed care). Patients were classified as having Medicare if they had Medicare with or without supplemental coverage. Patients with Medicaid, public health service insurance, and other federal insurance programs were consolidated into a single category (Medicaid). Patients without insurance coverage were classified as “not insured/self pay,” and the remaining patients were classified as “insurance status unknown.”

Statistical Analysis

The first outcome variable was adherence to NCCN guidelines for treatment for ovarian cancer and was based on recommendations for surgical and chemotherapy treatment according to the time period of diagnosis (1998–2000, 2001–2002) (15–17). Surgical treatment was considered adherent to NCCN guidelines if any of the following procedures were recorded: 1) partial or total unilateral or bilateral salpingo-oophorectomy with omentectomy, with or without hysterectomy; 2) debulking or cytoreductive surgery with colon (including appendix) and/or small intestine resection and/or partial resection of urinary tract (not incidental); or 3) anterior, posterior, total, or extended pelvic exenteration. For tumors that were International Federation of Gynecology and Obstetrics stage I to IIIB, examination of pelvic and/or para-aortic lymph nodes was required for adherent care. NCCN-specified administration of multiagent chemotherapy was considered adherent care. The second outcome variable was 5-year overall survival. In all analyses, age at diagnosis was modeled as a continuous variable.

Differences between white and black patients according to demographic and clinical variables and deviations from NCCN guideline care, stratified by race and insurance payer, were examined using the χ² test. The independent effects of race, SES indicators, tumor characteristics, and healthcare system factors on adherence to NCCN guidelines for overall treatment were examined using binomial logistic regression. Survival analysis was performed using the Kaplan–Meier method and log rank test. The Cox proportional hazards model, taking into consideration clustering within treating facilities, was used to evaluate the independent effect of all variables on 5-year survival. The appropriateness of the proportional hazards assumptions were validated by examining plots of the deviance residuals from the fit model. Statistical significance was set to P less than .05, and all analyses were performed using SAS version 9.2 (SAS Institute, Inc, Cary, NC). All statistical tests were two-sided.

Results

Cohort Characteristics

The final study population consisted of 47 160 patients: 43 995 patients were white, and 3165 patients were black (Table 1). Statistically significant differences between white and black patients were noted for age, median household income, education, urban/rural residence, and payer status.

Table 1.

Cohort characteristics of 47160 cases of epithelial ovarian cancer in the period from 1998 to 2002*

Characteristic	White (n = 43 995)	Black (n = 3165)	P†
Age, mean (SD), y	62.4 (13.8)	61.3 (14.1)
Adherence to NCCN guidelines for treatment, No. (%)			<.001
Adherent	19 304 (43.9)	1128 (35.6)
Nonadherent	24 691 (56.1)	2037 (64.4)
Percentage with college degree, No. (%)			<.001
<9%	19 321 (43.9)	792 (25.0)
9%–12.9%	11 767 (26.8)	847 (26.8)
13%–20.9%	6401 (14.6)	699 (22.1)
≥21%	4130 (9.4)	682 (21.6)
Not available	2376 (5.4)	145 (4.6)
Median household income 2000, No. (%)			<.001
≥$46 000	17 319 (39.4)	519 (16.4)
$35 000–$45 999	12 130 (27.6)	690 (21.8)
<$35 000	12 172 (27.7)	1811 (57.2)
Missing	2374 (5.4)	145 (4.6)
Primary payer at diagnosis, No. (%)			<.001
Private insurance	8825 (20.1)	480 (15.2)
Medicare/Medicare with supplements	17 622 (40.1)	1275 (40.3)
Managed care/ TRICARE/military	12 764 (29.0)	725 (22.9)
Medicaid/federal insurance programs/public health service	1504 (3.4)	287 (9.1)
Not insured/self pay	1392 (3.2)	249 (7.9)
Insurance status unknown	1888 (4.3)	149 (4.7)
Tumor stage, No. (%)			<.001
Stage I	7814 (17.8)	359 (11.3)
Stage II	3641 (8.3)	214 (6.8)
Stage III	19 753 (44.9)	1294 (40.9)
Stage IV	12 787 (29.1)	1298 (41.0)
Tumor grade, No. (%)			<.001
Well/moderately differentiated (referent)	11 949 (27.2)	767 (24.2)
Poorly/undifferentiated/anaplastic	22 969 (52.2)	1489 (47.1)
Missing	9077 (20.6)	909 (28.7)
Tumor histology, No. (%)			<.001
Serous	21 335 (48.49)	1418 (44.80)
Mucinous	3093 (7.03)	257 (8.12)
Endometrioid	5744 (13.06)	300 (9.48)
Clear-cell	2437 (5.54)	89 (2.81)
Mixed	345 (0.78)	21 (0.66)
NOS/undifferentiated	11 041 (25.10)	1080 (34.12)
Facility type, No. (%)			<.001
Community cancer program	5579 (12.68)	323 (10.21)
Comprehensive community cancer program	20 284 (46.11)	1097 (34.66)
Academic/research program (includes NCI-designated comprehensive cancer centers)	18 132 (41.21)	1745 (55.13)
Urban rural continuum 2003, No. (%)			<.001
Counties in metro areas of ≥1 million population	20 641 (46.92)	1912 (60.41)
Counties in metro areas of 250 000–1 million population	9020 (20.50)	535 (16.90)
Counties in metro areas of <250 000 population	4623 (10.51)	241 (7.61)
Urban population of ≥20 000, adjacent to a metro area	2182 (4.96)	100 (3.16)
Urban population of ≥20 000, not adjacent to a metro area	809 (1.84)	30 (0.95)
Urban population of 2500–19 999, adjacent to a metro area	2130 (4.84)	133 (4.20)
Urban population of 2500–19 999, not adjacent to a metro area	1114 (2.53)	33 (1.04)
Completely rural or <2500 urban population, adjacent to a metro area	432 (0.98)	29 (0.92)
Completely rural or <2500 urban population, not adjacent to a metro area	3044 (6.92)	152 (4.80)
Number of reasons for nonadherence to NCCN guidelines for treatment, No. (%)‡			<.001
0	19 304 (43.88)	1128 (35.64)
1	15 745 (35.79)	1140 (36.02)
2	7754 (17.62)	780 (24.64)
3	1192 (2.71)	117 (3.70)
Hospital ovarian cancer volume, No. of cases per year			<.001
1–6	10 970 (24.93)	772 (24.39)
7–14	10 986 (24.97)	882 (27.87)
15–25	10 985 (24.97)	835 (26.38)
≥26	11 054 (25.13)	676 (21.36)
Total	43 995 (100.00)	3165 (100.00)

* NCCN = National Comprehensive Cancer Network; NCI, National Cancer Institute; NOS, not otherwise specified.

† χ² test was used for categorical variables; t test was used for continuous variables. All P values are two-sided.

‡ Nonadherence to surgical treatment guidelines, chemotherapy guidelines, and/or missing pathological reporting.

Adherence to NCCN Guideline Care

The distributions of subjects by elements of nonadherent care stratified by race and payer are shown in Table 2. NCCN guideline adherent care was administered in 43.9% of white patients and 35.6% of black patients (P < .0001). Black patients were statistically significantly less likely to receive proper chemotherapy and proper surgery. Managed care and private insurance had the highest rates of adherent care, whereas not insured/self pay (42.3%) and Medicare (35.3%) had the lowest rates of adherence. Across all payer categories, black patients with ovarian cancer were less likely than white patients to receive NCCN guideline-adherent overall care, surgical treatment, and proper chemotherapy.

Table 2.

Distribution of patients according to treatment elements of National Comprehensive Cancer Network guideline nonadherent care stratified by race, payer, and a bivariable combination of race/payer

Characteristic	Chemotherapy					Surgery and lymph node examination					Overall treatment
	Proper		Improper		P*	Proper		Improper		P*	Proper		Improper		P*
	No.	%	No.	%		No.	%	No.	%		No.	%	No.	%
Race					<.001					<.001					<.001
White	28847	65.6	15148	34.4		27062	61.5	16933	38.5		19304	43.9	24691	56.1
Black	1879	59.4	1286	40.6		1623	51.3	1542	48.7		1128	35.6	2037	64.4
Insurance payer					<.001					<.001					<.001
Private insurance	6673	71.7	2632	28.3		6175	66.4	3130	33.6		4639	49.9	4666	50.2
Medicare/ Medicare with supplements	11042	58.4	7855	41.6		10119	53.6	8778	46.5		6678	35.3	12219	64.7
Managed care/ TRICARE/military	9490	70.4	3999	29.7		9202	68.2	4287	31.8		6788	50.3	6701	49.7
Medicaid/federal insurance programs/public health service	1287	71.9	504	28.1		1083	60.5	708	39.5		847	47.3	944	52.7
Not insured/self pay	1083	66.0	558	34.0		946	57.7	695	42.4		694	42.3	947	57.7
Insurance status Unknown	1151	56.5	886	43.5		1160	57.0	877	43.1		786	38.6	1251	61.4
Race and insurance payer
Whites with private insurance	6356	72.0	2469	28.0	0.005	5899	66.8	2926	33.2	<.001	4438	50.3	4387	49.7	<.001
Blacks with private insurance	317	66.0	163	34.0		276	57.5	204	42.5		201	41.9	279	58.1
Whites with Medicare/Medicare with supplements	10372	58.9	7250	41.1	<.001	9578	54.4	8044	45.7	<.001	6320	35.9	11302	64.1	<.001
Blacks with Medicare/Medicare with supplements	670	52.6	605	47.5		541	42.4	734	57.6		358	28.1	917	71.9
Whites with managed care/ TRICARE/military	9019	70.7	3745	29.3	0.001	8750	68.6	4014	31.5	<.001	6476	50.7	6288	49.3	<.001
Blacks with managed care/ TRICARE/military	471	65.0	254	35.0		452	62.3	273	37.7		312	43.0	413	57.0
Whites with Medicaid/federal insurance programs/public health service	1100	73.1	404	26.9	0.006	929	61.8	575	38.2	0.010	731	48.6	773	51.4	.01
Blacks with Medicaid/federal insurance programs/public health service	187	65.2	100	34.8		154	53.7	133	46.3		116	40.4	171	59.6
Whites with not insured/self pay	929	66.7	463	33.3	0.134	815	58.6	577	41.5	0.081	598	43.0	794	57.0	.20
Blacks with not insured/self pay	154	61.9	95	38.2		131	52.6	118	47.4		96	38.6	153	61.5
Whites with insurance status unknown	1071	56.7	817	43.3	0.472	1091	57.8	797	42.2	0.007	741	39.3	1147	60.8	.03
Blacks with insurance status unknown	80	53.7	69	46.3		69	46.3	80	53.7		45	30.2	104	69.8

* χ² test used for statistical analysis. All P values are two-sided.

Logistic regression analysis revealed that age, stage of disease, histologic subtype, and annual hospital case volume were statistically significantly associated with receipt of NCCN guideline-adherent care (Table 3). Median household income was positively associated with receipt of adherent care, whereas education level was not a statistically significant predictor. Black race (odds ratio [OR] = 1.36, 95% confidence interval [CI] = 1.25 to 1.48), Medicare payer status (OR = 1.20, 95% CI = 1.12 to 1.28), and not insured payer status (OR = 1.33, 95% CI = 1.19 to 1.49) were independently associated with a higher likelihood of not receiving NCCN guideline-adherent care.

Table 3.

Logistic regression model (binomial logistic regression) of demographic, clinical, pathological, and treatment variables associated with National Comprehensive Cancer Network guideline nonadherent care*

Risk factor	No.	%	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Patient Characteristics
Age, mean, SD, y	62.31	13.8	1.03 (1.02 to 1.03)	1.02 (1.02 to 1.03)
Race
White	43 995	93.3	1.00 (referent)	1.00 (referent)
Black	3165	6.7	1.41 (1.31 to 1.52)	1.36 (1.25 to 1.48)
Proportion with college degree-2000
<9%	20 113	42.7	1.00 (referent)	1.00 (referent)
9%–12.9%	12 614	26.8	1.10 (1.05 to 1.15)	0.96 (0.90 to 1.01)
13%–20.9%	7100	15.1	1.18 (1.12 to 1.25)	1.03 (0.96 to 1.10)
≥21%	4812	10.2	1.22 (1.15 to 1.30)	1.00 (0.92 to 1.09)
Missing	2521	5.4	1.05 (0.97 to 1.15)	0.24 (0.01 to 4.74)
Median household income - 2000
≥$46 000	17 838	37.8	1.00 (referent)	1.00 (referent)
$35 000–$45 999	12 820	27.2	1.12 (1.07 to 1.18)	1.05 (0.99 to 1.11)
<$35 000	13 983	29.7	1.26 (1.21 to 1.32)	1.09 (1.02 to 1.16)
Missing	2519	5.3	1.09 (1.00 to 1.18)	4.63 (0.23 to 91.68)
Primary payer at diagnosis
Private insurance	9305	19.7	1.00 (referent)	1.00 (referent)
Medicare/Medicare with supplements	18 897	40.1	1.82 (1.73 to 1.91)	1.20 (1.12 to 1.28)
Managed care/TRICARE/military	13 489	28.6	0.98 (.093 to 1.04)	1.03 (0.97 to 1.09)
Medicaid/federal insurance programs/public health service	1791	3.8	1.11 (1.00 to 1.23)	1.08 (0.96 to 1.20)
Not insured/self pay	1641	3.5	1.36 (1.22 to 1.51)	1.33 (1.19 to 1.49)
Missing: insurance status unknown	2037	4.3	1.58 (1.44 to 1.75)	1.80 (1.61 to 2.00)
Tumor Characteristics
Tumor stage
Stage IA	4464	9.5	1.00 (referent)	1.00 (referent)
Stage IB	511	1.1	1.10 (0.90 to 1.35)	1.13 (0.91 to 1.39)
Stage IC	3198	6.8	0.80 (0.73 to 0.88)	0.78 (0.70 to 0.86)
Stage IIA	833	1.8	1.20 (1.02 to 1.42)	1.06 (0.89 to 1.27)
Stage IIB	1198	2.5	0.87 (0.76 to 1.00)	0.73 (0.64 to 0.85)
Stage IIC	1824	3.9	0.91 (0.81 to 1.02)	0.74 (0.66 to 0.84)
Stage IIIA	1529	3.2	1.22 (1.07 to 1.39)	0.92 (0.80 to 1.05)
Stage IIIB	1999	4.2	1.24 (1.10 to 1.39)	0.98 (0.86 to 1.11)
Stage IIIC	17 519	37.2	0.28 (0.26 to 0.30)	0.21 (0.19 to 0.23)
Stage IV	14 085	29.9	0.76 (0.71 to 0.82)	0.40 (0.37 to 0.43)
Tumor grade
Well/moderately differentiated	12 716	27.0	1.00 (referent)	1.00 (referent)
Poorly/undifferentiated/anaplastic	24 458	51.9	0.77 (0.74 to 0.80)	0.91 (0.87 to 0.96)
Missing	9986	21.2	2.07 (1.96 to 2.19)	1.95 (1.83 to 2.09)
Tumor histology
Serous	22 753	48.3	1.00 (referent)	1.00 (referent)
Mucinous	3350	7.1	2.09 (1.94 to 2.26)	1.39 (1.27 to 1.51)
Endometroid	6044	12.8	1.49 (1.40 to 1.57)	1.10 (1.03 to 1.18)
Clear-cell	2526	5.4	1.65 (1.51 to 1.79)	1.36 (1.23 to 1.49)
Mixed	366	0.8	1.46 (1.19 to 1.80)	1.56 (1.25 to 1.95)
NOS/undifferentiated	12 121	25.7	2.97 (2.83 to 3.11)	2.11 (2.00 to 2.22)
Hospital ovarian cancer volume, no. of cases per year
1–6	11 742	24.9	1.00 (referent)	1.00 (referent)
7–14	11 868	25.2	0.65 (0.62 to 0.69)	0.78 (0.73 to 0.82)
15–25	11 820	25.1	0.45 (0.43 to 0.48)	0.59 (0.55 to 0.62)
≥26	11 730	24.9	0.36 (0.34 to 0.38)	0.47 (0.45 to 0.50)
Total	47 160	100.0

* CI = confidence interval; NOS = not otherwise specified; OR = odds ratio; SD = standard deviation.

Survival Analysis

The 5-year overall survival for all patients was 38.4% (95% CI = 38.0% to 38.9%). Statistically significant differences in overall 5-year survival rates were observed according to both race and payer stratified by adherence to NCCN guidelines. The 5-year overall survival rates for white patients receiving NCCN guideline-adherent care (41.4%) and nonadherent care (37.8%) were statistically significantly better compared with black patients receiving NCCN guideline-adherent care (33.3%) and nonadherent care (22.5%; P < .0001) (Figure 1A). The directionality and statistical significance of these trends in 5-year overall survival stratified by race and NCCN guideline adherence were the same regardless of facility type or annual hospital ovarian cancer volume (Supplementary Figures 1 and 2, available online). Statistically significant differences in 5-year overall survival were also observed according to payer status. Those with private insurance and managed care payer status had statistically significantly improved 5-year overall survival outcomes compared with patients with Medicaid, those who were not insured/self pay, and patients with Medicare for both NCCN guideline-adherent (Figure 1B) (P < .0001) and nonadherent (data not shown) care.

Figure 1.

Overall survival (OS) probability for patients with invasive primary epithelial ovarian cancer from the National Cancer Data Base stratified by adherence to National Comprehensive Cancer Network (NCCN) guideline therapy and race and payer category. Survival analyses were performed using the Kaplan–Meier method and two-sided log rank test. A) Data from all patients (n = 47 160) were analyzed according to race and adherence and nonadherence to NCCN guideline care. The 5-year overall survival was 41.4% (95% confidence interval [CI] = 40.6% to 42.1%) for adherent whites, 33.3% (95% CI = 30.4% to 36.2%) for adherent blacks, 37.8% (95% CI = 37.1% to 38.4%) for nonadherent whites, and 22.5% (95% CI = 20.6% to 24.4%) for nonadherent blacks (two-sided P < .001). B) Data from all patients receiving NCCN guideline-adherent therapy (n = 20 432) were analyzed according to payer category. The 5-year overall survival was 46.3% (95% CI = 44.8% to 47.8%) for patients with private payer category, 47.3% (95% CI = 46.0% to 48.5%) for patients with managed care payer category, 31.6% (95% CI = 30.4% to 32.7%) for patients with Medicare payer category, 35.8% (95% CI = 32.4% to 39.3%) for patients with Medicaid payer category, and 42.4% (95% CI = 38.4% to 46.4%) for patients with not insured/self pay payer category (two-sided P < .001). The number of patients at risk in each group at various time points are listed below the graphs.

Multilevel survival analysis revealed that tumor stage, grade, histological subtype, and annual hospital ovarian cancer case volume were all statistically significantly predictive of overall survival (Table 4). Non-NCCN-guideline-adherent care was an independent predictor of inferior overall survival (hazard ratio [HR] = 1.43, 95% CI = 1.38 to 1.47). The negative effect of Medicare payer status on overall survival was attenuated although still statistically significant. A median household income less than $35 000 was also independently negatively associated with survival (HR = 1.06, 95% CI = 1.02 to 1.11). However, of the sociodemographic factors evaluable in this study, the three strongest predictors of a worse overall survival outcome, after controlling for other factors, including NCCN guideline-adherent care, were black race (HR = 1.29, 95% CI = 1.22 to 1.36), Medicaid payer status (HR = 1.29, 95% CI = 1.20 to 1.38), and not insured payer status (HR = 1.32, 95% CI = 1.20 to 1.44). Each of these characteristics was associated with an approximately 30% increase in the risk of death.

Table 4.

Multilevel survival analysis*

Risk factor	No.	%	Unadjusted HR (95% CI)	Adjusted HR (95% CI)
Treatment Characteristics
Adherence to NCCN guidelines for treatment
Yes	20 432	43.3	1.00 (referent)	1.00 (referent)
No	26 728	56.7	1.33 (1.29 to 1.37)	1.43 (1.38 to 1.47)
Patient Characteristics
Age, mean, SD, y	62.31	13.8	1.04 (1.04 to 1.04)	1.03 (1.03 to 1.03)
Race
White	43 995	93.3	1.00 (referent)	1.00 (referent)
Black	3165	6.7	1.50 (1.43 to 1.59)	1.29 (1.22 to 1.36)
Proportion with college degree-2000
<9%	20 113	42.7	1.00 (referent)	1.00 (referent)
9%–12.9%	12 614	26.8	1.16 (1.12 to 1.20)	1.08 (1.04 to 1.12)
13%–20.9%	7100	15.1	1.19 (1.15 to 1.24)	1.08 (1.04 to 1.13)
≥21%	4812	10.2	1.21 (1.16 to 1.27)	1.05 (0.99 to 1.11)
Missing	2521	5.4	0.99 (0.94 to 1.05)	0.66 (0.56 to 0.77)
Median household income - 2000
≥$46,000	17 838	37.8	1.00 (referent)	1.00 (referent)
$35 000–$45 999	12 820	27.2	1.12 (1.09 to 1.16)	1.00 (0.97 to 1.04)
<$35 000	13 983	29.7	1.28 (1.24 to 1.33)	1.06 (1.02 to 1.11)
Missing	2519	5.3	1.02 (0.96 to 1.07)	1.48 (1.25 to 1.75)
Primary payer at diagnosis
Private insurance	9305	19.7	1.00 (referent)	1.00 (referent)
Medicare/Medicare with supplements	18 897	40.1	2.14 (2.07 to 2.22)	1.07 (1.02 to 1.11)
Managed care/TRICARE/military	13 489	28.6	1.03 (0.99 to 1.07)	1.02 (0.98 to 1.06)
Medicaid/federal insurance programs/public health service	1791	3.8	1.49 (1.39 to 1.60)	1.29 (1.20 to 1.38)
Not insured/self pay	1641	3.5	1.39 (1.28 to 1.50)	1.32 (1.20 to 1.44)
Missing: insurance status unknown	2037	4.3	1.45 (1.34 to 1.58)	1.05 (0.93 to 1.17)
Tumor Characteristics
Tumor stage
Stage IA	4464	9.5	1.00 (referent)	1.00 (referent)
Stage IB	511	1.1	1.44 (1.15 to 1.81)	1.61 (1.29 to 2.02)
Stage IC	3198	6.8	1.50 (1.33 to 1.69)	1.59 (1.41 to 1.78)
Stage IIA	833	1.8	2.20 (1.87 to 2.60)	2.23 (1.89 to 2.63)
Stage IIB	1198	2.5	2.52 (2.18 to 2.91)	2.46 (2.13 to 2.85)
Stage IIC	1824	3.9	3.61 (3.19 to 4.08)	3.58 (3.17 to 4.04)
Stage IIIA	1529	3.2	5.72 (5.10 to 6.42)	5.34 (4.76 to 6.00)
Stage IIIB	1999	4.2	6.02 (5.43 to 6.69)	5.65 (5.06 to 6.30)
Stage IIIC	17 519	37.2	8.15 (7.45 to 8.90)	8.79 (8.00 to 9.66)
Stage IV	14 085	29.9	14.26 (13.03 to 15.60)	12.45 (11.34 to 13.68)
Tumor grade
Well/moderately differentiated	12 716	27.0	1.00 (referent)	1.00 (referent)
Poorly/undifferentiated/anaplastic	24 458	51.9	1.78 (1.72 to 1.84)	1.11 (1.07 to 1.15)
Missing	9986	21.2	3.09 (2.97 to 3.22)	1.32 (1.27 to 1.37)
Tumor histology
Serous	22 753	48.3	1.00 (referent)	1.00 (referent)
Mucinous	3350	7.1	0.77 (0.72 to 0.82)	1.78 (1.67 to 1.90)
Endometroid	6044	12.8	0.41 (0.39 to 0.44)	0.85 (0.81 to 0.90)
Clear-cell	2526	5.4	0.53 (0.49 to 0.57)	1.47 (1.35 to 1.59)
Mixed	366	0.8	0.73 (0.62 to 0.85)	1.00 (0.85 to 1.17)
NOS/undifferentiated	12 121	25.7	1.72 (1.67 to 1.78)	1.31 (1.26 to 1.35)
Hospital ovarian cancer volume, no. of cases per year
1–6	11 742	24.9	1.00 (referent)	1.00 (referent)
7–14	11 868	25.2	0.85 (0.82 to 0.89)	0.96 (0.92 to 1.00)
15–25	11 820	25.1	0.79 (0.75 to 0.82)	0.93 (0.89 to 0.97)
≥26	11 730	24.9	0.74 (0.70 to 0.78)	0.92 (0.88 to 0.97)
Total	47 160	100.0

* Hazard ratios were computed using Cox proportional hazards model, taking into consideration clustering within treating facilities. CI = confidence interval; HR = hazard ratio; NCCN = National Comprehensive Cancer Network; NOS = not otherwise specified; SD = standard deviation.

Discussion

Ovarian cancer is the fifth leading cause of cancer-related death among US women and accounts for more deaths than all other gynecologic cancers combined. Although treatment advances have improved the expected survival outcome of the general population of women with ovarian cancer over the past 30 years, this improvement in survival has not been universal across racial, ethnic, and socioeconomic groups (7). Women with ovarian cancer from racial or ethnic minorities and socioeconomically disadvantaged populations suffer a disproportionately greater mortality burden (6–8,12,14,18–22). The literature is divided, however, on whether race and SES indicators are independent negative prognostic factors for ovarian cancer survival. Many of these studies have had limited numbers of black patients and inconsistently controlled for the effects of tumor characteristics, treatment, and SES variables that might also influence survival. The objective of this study was to examine disparities in the quality of epithelial ovarian cancer care and survival outcome according to race and socioeconomic indicators using the resources of the NCDB.

Ultimately, disparities in healthcare based on race and SES can be reframed as fundamental issues of healthcare quality under the assumption that high-quality care should be universally accessible and administered irrespective of one’s phenotypic features or socioeconomic station. The literature examining disparities in the quality of ovarian cancer care, benchmarked to a validated national standard, is limited. In 2005, Harlan et al. studied 7134 patients with 11 different cancer types from the Surveillance Epidemiology and End Results (SEER) Patterns of Care database and found that whites received NCCN guideline cancer therapy more often than blacks and Hispanics (3). This study did not link receipt of guideline therapy to survival and included only 504 patients with ovarian cancer. The only other study linking ovarian cancer care to specific treatment guidelines was reported by Harlan et al. and described 1167 ovarian cancer patients from the SEER database in 1991 and 1996 (10). These investigators examined trends in surgery and chemotherapy according to recommendations from the 1994 National Institutes of Health Consensus Development Conference on ovarian cancer and found that 55.2% of whites received recommended therapy compared with just 43.6% of blacks. This study, however, included only 118 black patients and did not link treatment to survival outcome.

The current dataset draws from approximately 70% of the ovarian cancer cases diagnosed in the United States during a time interval of standardized contemporary treatment. The resulting perspective is unique and offers the first large-scale, population-based analysis benchmarking racial and SES disparities in ovarian cancer care to a quality process measure—NCCN guideline therapy—that has been simultaneously validated and linked with survival outcome. In this dataset, black race was independently associated with 36% increased likelihood of not receiving NCCN guideline-adherent care. Furthermore, after adjusting for the effects of other variables, including receipt of NCCN guideline therapy, black patients experienced a 29% increase in the risk of death compared with whites. The contribution of reduced access to expert care to the race-based survival gap is difficult to ascertain. Although previous investigators have reported that blacks have reduced access to high-volume ovarian cancer surgeons, the observed survival disparity is adjusted for the effect of hospital case volume (23,24).

Disentangling the observed race-based ovarian cancer survival gap from differences in the quality of care administered is a complex problem. For example, in an early study using the NCDB, Parham et al. found that blacks with ovarian cancer were more than twice as likely as whites to not receive recommended treatment and had a 30% increased risk of death (14). This analysis was only adjusted for age, stage, and residential income and did not correct for the type or quality of treatment or other demographic variables. More recently, Chase et al. found that black race was independently predictive of receiving initial chemotherapy for ovarian cancer rather than primary surgery; however, the associated effect on survival was not examined (13). In an analysis of 4262 ovarian cancer patients from the SEER database, Du and colleagues found that although only 50.2% of blacks received recommended chemotherapy treatment, compared with 64.7% of whites, there was no difference in survival after adjusting for tumor characteristics, treatment, and sociodemographic factors (25). Data supporting the hypothesis that differences in ovarian cancer treatment could account for all or most of the race-based survival differences come from single institution studies, population-based studies, and cooperative group trial data showing that when equal treatment is administered, race-based survival disparities are largely mitigated or even eliminated (25–31). In contrast, the current analysis revealed a persistent and statistically significant negative survival impact associated with black race after controlling for NCCN guideline-adherent care and suggests that there may be subtle but important survival determinants (eg, amount of residual disease, chemotherapy dose intensity, access to high-volume surgeons) within the initial treatment program that are not captured by the algorithm for adherence to NCCN guidelines (23,30). Alternatively, unmeasured factors along the clinical care continuum impacting survival (eg, second-line therapy) may be unequally distributed according to race (32). It is unlikely that the presence of potential underlying biological differences between races, including variation in BRCA mutation status, would be of a magnitude that would account for the observed differences in quality of care and survival (33).

The influence of SES indicators, such as education level, employment status, and household income, as independent predictors of ovarian cancer treatment and survival has been inconsistent. In several studies, an increasing level of Census-area poverty has been associated with both a lower likelihood of receiving recommended treatment and worse ovarian cancer survival (18,22,34). In contrast, others have failed to identify poverty level, household income, or education as predictive of either treatment or outcome (14,26,35). In the current study, a median annual household income less than $35 000 was statistically significantly associated with a lower likelihood of receiving proper treatment and worse overall survival independent of race, although the effects were modest.

The type of health insurance is both a health system factor and an individual-level measure of SES and has been linked to expenditure on cancer treatment, suggesting that payer status may influence receipt of appropriate care (3,36). In this analysis, patients with Medicare and not insured/self pay payer status were statistically significantly less likely to receive NCCN guideline-adherent care and experienced an approximately 30% increased risk of death. These data confirm findings of previous studies of ovarian cancer in which nonprivate or uninsured payer status was highly predictive of receiving nonstandard treatment and death (12,13,33,37). Interestingly, in the 2003 study by Harlan et al., the lack of private insurance was an impediment to receiving appropriate ovarian cancer therapy for both black and Hispanic patients but not whites (10). Two population-based studies and one single institution study, however, have found that payer status was not predictive of ovarian cancer treatment (3,28,38).

There are several limitations of this study that must be considered in interpreting the data. First, the dataset contained incomplete information for some characteristics, which may have introduced selection bias in determining the final study population. Second, the NCDB is subject to potential errors in reporting, although rigorous validation of data accuracy is performed through internal monitoring, site surveys, and review of data quality (39,40). Third, the NCDB does not contain information on surgeon specialty, the extent of residual disease, specific chemotherapy agents, platinum dose intensity/cumulative dose, or treatment of recurrent disease. It is possible that differences in these parameters could account for a proportion of the observed survival disparities. A fourth potential limitation is that the NCCN guideline adherence decision algorithm incorporated both initial surgery and initial chemotherapy (neoadjuvant chemotherapy). To the extent that neoadjuvant chemotherapy has been associated with a worse survival outcome, associations between race and payer status and the frequency of neoadjuvant chemotherapy could be reflected as disparities in survival (12,13). A fifth limitation of this study is that information on medical comorbidity and performance status was not available for the 1998 to 2002 study cohort. As a result, neither NCCN guideline adherence nor survival could be adjusted for the presence of medical comorbidity. Although unmeasured differences in medical comorbidity likely accounts for a portion of the observed disparities according to race and payer status, it is improbable that this would explain the full extent of differences in treatment quality and survival (13,25).

This is the largest, most comprehensive dataset reported to date that analyzes ovarian cancer disparities using a validated quality process measure linked to a meaningful clinical outcome. These data highlight and quantify clinically significant disparities in the quality of ovarian cancer care and overall survival, independent of NCCN guidelines, along racial and SES parameters in the United States and indicate that not all segments of the population have benefitted equally from improvements in ovarian cancer care. It is also apparent that factors other than adherence to NCCN guideline care influence ovarian cancer survival for racial minorities and the socioeconomically disadvantaged. Increased efforts are needed to more precisely define the patient, provider, healthcare system, and societal factors leading to these observed disparities and develop an informed platform from which targeted interventions can be designed to reduce and ultimately eliminate ovarian cancer disparities among persons from all racial groups and socioeconomic strata.

Funding

REB was supported in part by the Queen of Hearts Foundation .

The sponsor had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication.

This paper received the Hugh Barber Best Scientific Abstract Plenary Presentation at the Society of Gynecologic Oncology Annual Meeting on Women’s Cancer, Austin, Texas, March 27, 2012.