Racial variation in tumor stage at diagnosis among Department of Defense beneficiaries

Abstract

Introduction

Tumor stage at diagnosis often varies by racial/ethnic group, possibly due to inequitable healthcare access. Within the Department of Defense (DoD) Military Health System, beneficiaries have equal healthcare access. This study aimed to determine if tumor stage differed between whites and blacks for breast, cervical, colorectal and prostate cancers, which have effective screening regimens, based on data from the DoD’s Automated Cancer Tumor Registry from 1990–2003.

Methods

Distributions of tumor stage (localized vs. non-localized) between whites and blacks in the military were compared stratified by sex, active duty status, and age at diagnosis. Logistic regression was used to further adjust for age, marital status, year of diagnosis, geographic region, military service branch and tumor grade. Distributions of tumor stage were then compared between the military and general populations.

Results

Racial differences in the distribution of stage were significant only among non-active duty beneficiaries. After adjusting for covariates, earlier stages of breast cancer after age 49 and prostate cancer after age 64 were significantly more common among white than black non-active duty beneficiaries (p<0.05), although the absolute difference for prostate cancer was minimal. Racial differences in stage for cervical and colorectal cancers were not significant after adjustment. Compared to the general population, the racial differences in the military were similar or slightly attenuated.

Conclusion

Racial disparities in stage at diagnosis were apparent in the DoD’s equal access healthcare system among older non-active duty beneficiaries. Socioeconomic status, supplemental insurance, cultural beliefs and biological factors may be related to these results.

Introduction

Cancer is the second leading cause of death in the US,¹ resulting in an estimated 569,490 deaths in 2010.² Earlier tumor stage at diagnosis, which can be achieved by screening of some cancers, is associated with improved outcome. Cancer screening tests have been established for breast, cervical, colorectal, and prostate cancers. In the general US population, however, many of these tumors are still diagnosed at later stages, particularly among blacks. As a result, blacks tend to have lower relative survival rates than whites.³

The reasons for these racial disparities are complex and likely arise due to a combination of factors. The most often cited reasons for the worse statistics among blacks are lower socioeconomic status and decreased insurance coverage resulting in limited access to quality healthcare.^4–8 However, other factors such as cultural beliefs^4,7,9 and genetic or other biological variations^6,10 have also been implicated.

The Department of Defense (DoD)’s Military Health System provides a unique opportunity to study potential racial disparities because equal healthcare access is provided without regard to race or socioeconomic status. A recent survey of DoD beneficiaries indicated that self-reported cancer screening rates were higher, and racial/ethnic disparities in cancer screening rates were lower, in comparison to the general US population.¹¹ It is unclear if these findings translate to decreased racial disparities in tumor stage at diagnosis. Previous studies among combined DoD active and non-active duty beneficiaries have found evidence of racial variation by tumor stage at diagnosis for colorectal and breast cancers.^12–14 However, it was not known if racial if racial differences varied by active duty status or persisted after adjusting for other covariates.

The aim of this study was to conduct a DoD-wide comparison of tumor stage at diagnosis between whites and blacks for four cancers (breast, cervical, colorectal, and prostate). These cancer sites were selected because racial variations in stage at diagnosis observed in the general population may be partially attributable to access to the established screening tests. Comparisons were made between white and black beneficiaries by sex, active duty status, and age at diagnosis because active duty personnel may be monitored more closely and because cancer screening practices vary with age.¹⁵ Data on racial differences by tumor stage at diagnosis in the general US population were obtained for comparison. It was hypothesized that tumor stage at diagnosis would not differ significantly within the military by race after covariate stratification and adjustment and that any racial differences observed in the military would be less than in the general population.

Methods

The Automated Cancer Tumor Registry (ACTUR) was established in 1986 and is the data collection and clinical tracking system for all cancer cases diagnosed or treated at military treatment facilities among DoD beneficiaries, including active-duty military personnel, retired military personnel, and their dependents. Certified cancer registrars at each facility enter and maintain the ACTUR data according to state and federal guidelines. The registry includes information on demographic factors (e.g., age, race, sex and geographic location), diagnostic factors (e.g., date of diagnosis), and tumor characteristics (e.g., histology, stage and grade). The anatomic sites of the cancers were categorized using the first International Classification of Diseases for Oncology (ICD-O) for cases diagnosed from 1986 to 1991,¹⁶ the second edition ICD-O-2 for cases diagnosed from 1992 to 2001,¹⁷ and the third edition ICD-O-3 for cases diagnosed since 2001;¹⁸ all cases were recoded using the ICD-O-3 codes by ACTUR personnel.

For the purposes of this study, registry data for cases aged 18 years or older and diagnosed from 1990 to 2003 were included. Although all data submitted to ACTUR are reviewed and verified for accurate diagnoses, cases diagnosed between 1986 and 1989 were excluded to minimize the possibility of incomplete ascertainment. Procedures were developed with reference to national and state cancer registry guidelines^19,20 to identify and consolidate duplicate records so that only one record existed for each primary cancer. Tumor stage at diagnosis was determined by combining two variables: “SEER Summary Stage 1977” for cases diagnosed from 1990 to 2000 and “SEER Summary Stage 2000” for cases diagnosed from 2001 to 2003. These variables are described and used by the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute.^21,22 When there were multiple records per tumor at the time of diagnosis with different tumor stage codes, the information was chosen as follows: 1) if surgery information was available, then the code associated with the most definitive surgery was selected; and 2) if no surgery information was available, then the most advanced stage code was selected. Cancer-specific age categories were based on the American Cancer Society (ACS) cancer screening recommendations: breast cancer screening via mammography starting at age 40 years; cervical cancer via Pap smear between 18 and 69 years; and colorectal and prostate cancer starting at age 50 years.¹⁵ There are generally few older active duty beneficiaries, especially those who are Medicare eligible (age 65+); however, there are many older non-active duty beneficiaries, which allowed for further age stratification among this latter group. Breast, prostate and colorectal cancers were thereby investigated separately among individuals older than 64 years. Breast cancer was also investigated among women aged 50 to 64 years; the average age at menopause is 50 years and breast cancer etiology is known to vary by menopausal status. Cervical cancer was investigated stratified at age 30 years because ACS recommendations change at this age from annually/bi-annually to every 2–3 years, depending on the method used. Individuals aged 65 to 69 years were not investigated separately for cervical cancer due to small numbers.

For comparison to the general US population, data for cases diagnosed from 1990 to 2003 were obtained on breast and cervical cancers among women and colorectal cancers among both sexes from the nine original SEER registries (Connecticut, Iowa, New Mexico, Utah, Hawaii, Detroit, San Francisco-Oakland, Atlanta, and Seattle-Puget Sound), which cover about 10% of the US population.²³ As in ACTUR, tumor site was based on ICD-O-3 codes. Stage at diagnosis was determined for each case according to “SEER Historic Stage A” codes (in situ, localized, regional, distant, unstaged; in situ cancers were not included in this analysis).²⁴ SEER Historic Stage A codes are comparable to SEER Summary Stage codes (for localized vs. regional/distant stages) that were used by ACTUR for breast, prostate and cervical cancers. Codes for colorectal cancer stages differ somewhat in that colorectal tumors with “invasion of/through serosa” are classified as localized in SEER and regional in ACTUR.²⁵ Stage comparisons of prostate cancer between the two populations were not conducted because SEER Historic Stage A codes for prostate cancer group localized and regional cases in a single category.²⁶

The distributions of tumor stage between whites and blacks within the same population were compared, stratified by sex, cancer, categorical age at diagnosis and active duty status at diagnosis. For active-duty members only univariate analyses were conducted using Chi-square tests or Fisher’s exact tests because sample sizes were relatively small. For non-active-duty members and the SEER population, odds ratios (ORs) and 95% confidence intervals (CIs) comparing tumor stage (localized vs. regional/distant) between blacks and whites while adjusting for continuous age, marital status (Married, Single/Divorced/Separated/Widowed, Unknown), year of diagnosis, geographic region (North, South, West, Unknown),²⁷ tumor grade (I, II, III/IV, Unknown) and military service branch (Army, Navy, Air Force, Marine, Other; ACTUR only) were calculated if cell frequencies for the binary tumor stage variable were at least 10. All analyses were conducted using SAS software, version 9.1 (SAS Institute Inc., Cary, NC), and the two-sided significance level was set at p<0.05.

The protocol for this study was reviewed and approved by the National Naval Medical Center Institutional Review Board (IRB), and the Armed Forced Institute of Pathology IRB.

Results

For descriptive purposes, the demographic characteristics of the included cancer cases in the military by duty status and in the general population are presented by cancer site in Table 1. Among active duty beneficiaries, there were 782 breast, 133 cervical, and 69 colorectal cancers with known tumor stage diagnosed from 1990 to 2003 among women and 817 prostate and 588 colorectal cancers included among men. No significant differences in tumor stage distribution for any of these cancers were observed between white and black active duty beneficiaries (Table 2).

Table 1.

Distribution of demographic variables among cancer cases in the Department of Defense’s Automated Central Tumor Registry (ACTUR) and the general population (SEER) by cancer site, 1990–2003.

Cancer site	Demographic variable		ACTUR				SEER
			Active Duty		Non-Active Duty
			N	%	N	%	N	%
Breast	Age*	18–39	407	52%	1,591	10%	12,608	6%
		40+	375	48%	14,715	90%	210,344	94%
	Race	White	487	62%	14,302	88%	203,353	91%
		Black	295	38%	2,004	12%	19,599	9%
Cervix	Age*	18–29	39	29%	145	12%	1,048	9%
		30–69	94	71%	1,051	88%	10,496	91%
	Race	White	107	80%	1,059	89%	9,742	84%
		Black	26	20%	137	11%	1,802	16%
Colorectum	Age*	18–49	565	86%	948	10%	11,480	7%
		50+	92	14%	8747	90%	147,393	93%
	Race	White	500	76%	8,432	87%	143,675	90%
		Black	157	24%	1,263	13%	15,197	10%
	Sex	Women	69	11%	3,679	38%	78,896	50%
		Men	588	89%	6,016	62%	79,977	50%
Prostate**	Age*	18–49	413	51%	423	2%
		50+	404	49%	21444	98%
	Race	White	623	76%	17,894	82%
		Black	194	24%	3,973	18%

^*Age categories were cancer site specific based on American Cancer Society cancer screening recommendations.

^**Prostate cancer in SEER was not analyzed.

Table 2.

Racial comparison of tumor stage at diagnosis for selected cancers among active duty Department of Defense healthcare beneficiaries by sex, cancer site and age at diagnosis, Automated Central Tumor Registry 1990–2003.

Sex	Site	Age at diagnosis	Stage*	White		Black		p-value**
Women	Breast	18–39	Localized	112	49%	90	50%	0.89
			Regional/Distant	115	51%	90	50%
		40+	Localized	162	62%	62	54%	0.13
			Regional/Distant	98	38%	53	46%
	Cervix	18–29	Localized	27	84%	4	57%	0.14
			Regional/Distant	5	16%	3	43%
		30–69	Localized	67	89%	16	84%	0.69
			Regional/Distant	8	11%	3	16%
	Colorectum	18–49	Localized	17	46%	6	26%	0.12
			Regional/Distant	20	54%	17	74%
		50+	Localized	3	43%	1	50%	1.00
			Regional/Distant	4	57%	1	50%
Men	Prostate	18–49	Localized	242	88%	119	86%	0.61
			Regional/Distant	33	12%	19	14%
		50+	Localized	280	80%	47	84%	0.54
			Regional/Distant	68	20%	9	16%
	Colorectum	18–49	Localized	140	37%	42	34%	0.62
			Regional/Distant	242	63%	81	66%
		50+	Localized	39	53%	5	56%	1.00
			Regional/Distant	35	47%	4	44%

^*SEER Summary Stage 1977: cases diagnosed 1990–2000 and SEER Summary Stage 2000: cases diagnosed 2001–2003.

^**Two-sided p-value from a Chi-square test or Fishers exact test (expected cell frequencies <5) comparing the distribution between whites and blacks.

Among non-active duty beneficiaries, there were 16,306 breast, 1,196 cervical, and 3,679 colorectal cancers included among women and 21,867 prostate and 6,016 colorectal cancers among men (Table 3). Significant racial variation was observed among both sexes. White women older than 49 years of age were less likely to have nonlocalized breast cancers than were black women, after adjustment for age, marital status, year of diagnosis, geographic region, military service branch and tumor grade. This racial difference increased with age (18–39: OR=1.07; 40–49: OR=0.85; 50–64: OR=0.77; 65+: OR=0.63; p-interaction<0.01). A significant racial difference in stage distribution was observed for prostate cancer among men age 65 years or older (adjusted OR=0.78; 95% CI=0.68–0.91), but the actual variation in the distribution was small, suggesting that statistical significance was the result of the large sample size. After adjustment, no significant racial differences in stage distribution were observed for cervical or colorectal cancers.

Table 3.

Racial comparison of tumor stage at diagnosis for selected cancers among non-active duty Department of Defense healthcare beneficiaries (ACTUR) and individuals in the general US population (SEER) by sex, cancer site and age at diagnosis, 1990–2003.

Sex	Site	Age	Stage*	ACTUR					SEER
				White		Black		OR** (95% CI)	White		Black		OR** (95% CI)
Women	Breast	18–39	Localized	606	50%	193	51%	1.07 (0.85–1.36)	5,533	53%	919	44%	0.71 (0.64–0.79)
			Regional/Distant	605	50%	187	49%		5,005	47%	1,151	56%
		40–49	Localized	1,491	58%	271	52%	0.85 (0.70–1.03)	19,914	60%	2,311	51%	0.76 (0.71–0.81)
			Regional/Distant	1,059	42%	253	48%		13,509	40%	2,202	49%
		50–64	Localized	4076	67%	450	58%	0.77 (0.66–0.90)	41,701	65%	3,551	55%	0.73 (0.69–0.77)
			Regional/Distant	2036	33%	329	42%		22,579	35%	2,941	45%
		65+	Localized	3297	74%	202	63%	0.63 (0.49–0.81)	67,086	71%	3,963	61%	0.72 (0.68–0.76)
			Regional/Distant	1132	26%	119	37%		28,026	29%	2,562	39%
	Cervix	18–29	Localized	99	78%	15	83%	--	768	84%	101	77%	0.63 (0.37–1.07)
			Regional/Distant	28	22%	3	17%		148	16%	31	23%
		30–69	Localized	645	69%	81	68%	1.01 (0.65–1.57)	5,286	60%	837	50%	0.74 (0.66–0.84)
			Regional/Distant	287	31%	38	32%		3,540	40%	833	50%
	Colorectum	18–49	Localized	128	30%	47	34%	1.30 (0.83–2.04)	1,655	37%	307	32%	0.71 (0.60–0.84)
			Regional/Distant	301	70%	90	66%		2,759	63%	663	68%
		50–64	Localized	486	39%	88	37%	0.92 (0.68–1.26)	5,281	40%	853	38%	0.81 (0.73–0.89)
			Regional/Distant	749	61%	147	63%		7,765	60%	1,413	62%
		65+	Localized	634	43%	52	33%	0.71 (0.49–1.02)	22,460	42%	1,860	39%	0.84 (0.78–0.89)
			Regional/Distant	853	57%	104	67%		31,008	58%	2,872	61%
Men	Prostate	18–49	Localized	190	84%	157	80%	0.70 (0.39, 1.27)
			Regional/Distant	36	16%	40	20%
		50–64	Localized	5743	81%	1707	82%	1.14 (0.99, 1.31)
			Regional/Distant	1336	19%	380	18%
		65+	Localized	8904	84%	1369	81%	0.78 (0.68, 0.91)
			Regional/Distant	1685	16%	320	19%
	Colorectum	18–49	Localized	77	26%	24	29%	0.96 (0.52, 1.76)	1,895	36%	292	32%	0.79 (0.67–0.93)
			Regional/Distant	221	74%	60	71%		3,301	64%	608	68%
		50–64	Localized	864	39%	132	36%	0.85 (0.67, 1.08)	7,640	42%	897	37%	0.77 (0.70–0.85)
			Regional/Distant	1363	61%	237	64%		10,669	58%	1,502	63%
		65+	Localized	1221	44%	112	40%	0.78 (0.60, 1.01)	21,665	44%	1,573	40%	0.81 (0.76–0.87)
			Regional/Distant	1535	56%	170	60%		27,578	56%	2,357	60%

^*ACTUR: SEER Summary Stage 1977: cases diagnosed 1990–2000 and SEER Summary Stage 2000: cases diagnosed 2001–2003. SEER: SEER Historic Stage A.

^**Odds Ratio and 95% Confidence Interval comparing local vs. regional/distant tumors between blacks vs. whites adjusted for continuous age, marital status, year of diagnosis, geographic region, tumor grade and military service branch (ACTUR only). Not calculated for prostate cancer in SEER because Historic Stage A codes combine local and regional cases.

--Not calculated if frequency was <10

There were 222,952 breast, 11,544 cervical, and 78,896 colorectal cancers included among women and 79,977 colorectal cancers included among men in the general population (Table 3). The percentages of cancers that were localized were significantly higher among whites than blacks for all four cancers, even after adjustment. Racial differences did not vary significantly by age.

When compared to the general population, the racial difference in percentage of cancers that were localized among non-active duty beneficiaries tended to be similar or slightly attenuated. For example in SEER, among women aged 18 to 39 years, 53% of white and 44% of black women were diagnosed with local breast cancers (difference= 9%) compared to 50% and 51% of their respective counterparts in ACTUR (difference= −1%). Comparisons between the general population and the active duty military population were not conducted because no significant racial differences were observed among the latter population.

Discussion

The distribution of tumor stage by race did not appear to vary greatly among active duty beneficiaries for the four cancers studied but small sample sizes may have resulted in insufficient power to detect true differences. In contrast, racial differences were observed among non-active duty beneficiaries, particularly among older individuals with breast or prostate cancers. Nonetheless, the racial differences observed in the military appear to be similar or attenuated in comparison to the general population; the racial difference in percentage of cancers that were localized was smaller for breast among young non-active duty female beneficiaries in comparison to young women in the general population.

Our results among non-active duty beneficiaries were in agreement with a previous DoD-wide breast cancer study, although the calendar years differed.¹⁴ Our findings further indicated that significant racial differences exist after covariate adjustment and occur for age groups for whom mammography is recommended. It is unclear why racial differences in breast cancer stage increased with age in the military but not in the general population. After covariate adjustment our findings indicated that stage at diagnosis for colorectal cancer did not differ significantly between whites and blacks, which is contradictory to previous unadjusted studies among DoD beneficiaries.^12,13 This inconsistency may have resulted from differences in study populations as one of the previous studies¹² was based on registry data from one DoD medical center whereas our study was DoD-wide. Geographical variation in utilization and physician practices have been observed in the Department of Veterans Affairs (VA) healthcare system,²⁸ which is composed of military veterans. Additionally, our study adjusted for covariates whereas the previous studies^12,13 did not. To our knowledge, stage at diagnosis of prostate and cervical cancers among DoD beneficiaries have not been previously summarized and compared by race.

Larger racial disparities were observed among older non-active duty individuals. We can only speculate as to what the reasons are. The disparity may be due to increased physical or cultural barriers to healthcare utilization and variations in healthcare options among older individuals. These barriers could result in a greater tendency for black individuals to forgo more frequent care and/or postpone seeking medical care until after symptoms appear, thus resulting in a greater proportion of older black beneficiaries diagnosed at later stages of cancer. Although our data did not contain information on the use of cancer screening tests and could not show any racial differences in cancer screening, perceived cancer risk and cancer screening awareness have been observed to vary by race.^29,30 Racial variations in other health insurance (i.e., through employment, spouse or Medicare) may also influence healthcare seeking behaviors. Additionally, it is possible that racial differences in genetics or other biological factors might result in variations in tumor aggressiveness³¹ and thus differences in tumor stage at diagnosis.

The main strength of this study was the use of data from a healthcare system that is based on equal access to assess racial variations. Limitations of the study included the small numbers of specific cancer cases by race among active duty beneficiaries. Such numbers may have limited the power to detect racial differences. Secondly, the grouped categorization of localized and regional prostate cancers in the SEER data precluded us from comparing the magnitude of the racial difference in tumor stage between the two populations for this cancer site. There is also the possibility of under-reporting in ACTUR. Although DoD policies require cancer cases to be reported to ACTUR, some military treatment facilities, especially small ones, might not have reported their cancer patients. While the extent of the under-reporting is not known, differential reporting of tumor stage by race would have to exist to explain the observed differences. Additionally, there is the possibility of selection bias of the cases that were included in the study because beneficiaries have to be seen at a DoD medical center to be reported to ACTUR; beneficiaries with other health insurance may seek care elsewhere. There would have to be differential selection by stage and race for this to account for the observed differences. If whites have greater access to other healthcare and are diagnosed elsewhere at earlier stages, it is possible that our findings could represent an underestimate of the true racial difference in tumor stage at diagnosis among non-active duty military beneficiaries. Finally, for colorectal cancer, differences in tumor stage criteria between the two populations may partially explain the higher percentage of localized tumors in the general population; therefore, results should be interpreted with caution. However, unless the percentage of tumors with “invasion of/through serosa” differed by race, the racial difference in percentage of localized tumors in the two populations should be comparable.

In conclusion, racial disparities in tumor stage at diagnosis were not observed in the DoD Military Health System among active duty beneficiaries, but disparities were apparent among older non-active duty beneficiaries, with whites generally being diagnosed with earlier stage breast and prostate cancers than blacks. Although the DoD system is based on equal access, racial variation in socioeconomic status and supplemental insurance may still affect tumor stage at diagnosis. As such, more studies are needed to assess the independent impact of these and other possible factors.