Cancer Incidence among Patients of the United States Veterans Affairs (VA) Healthcare System: 2010 Update

The VA Central Cancer Registry

The origin of the VACCR and its cancer registration procedures has been previously described.² The VACCR contains information on VA cancer cases that were diagnosed in 1995 or later. VACCR strives to capture all cancer cases occurring within the VA healthcare system. It has been estimated that VACCR captures nearly 90% of VA healthcare system cancer cases.¹⁶ Cancer registrars at VA medical facilities across the country manually abstract case data using the North American Association of Central Cancer Registries (NAACCR) standards.¹⁷ In keeping with the NAACCR criteria for reportable cases, the VACCR includes cases that are diagnosed outside of the VA healthcare system if they subsequently receive care within VA.

For this analysis, data were obtained from VACCR for all incident cases diagnosed in the VA (or diagnosed outside of the VA healthcare system, but subsequently treated within the VA) with a cancer diagnosis date in calendar year 2010.

Descriptive Variables

Descriptive variables included anatomic site and stage of the cancer, and the patient’s geographic location, race, and sex. To determine stage at diagnosis, we used pathological TNM stage (primary tumor(T), regional lymph nodes(N), distant metastasis(M)) first, followed by clinical TNM stage when the pathological stage was unknown or missing.¹⁸

Within the VA, geographic location is described by Veterans Integrated Service Networks (VISN). To simplify analysis, each case’s reporting VISN was aggregated into four larger geographic regions – Northeastern (VISNs 1, 2, 3, and 4), Southern (VISNs 5, 6, 7, 8, 9, 16, and 17), Midwestern (VISNs 10, 11, 12, 15, and 23), and Western (VISNs 18, 19, 20, 21, and 22). These regions are not evenly distributed with regard to the geographic land mass, general population, or VA patient volume.

Primary anatomic site was defined according to International Classification of Diseases for Oncology (ICD-O-3) site and histology (i.e. type) codes from 2003.¹⁹ ICD-O-3 codes are commonly used in pathology reports and cancer registries to capture information about the topography and morphology of neoplasms. The unit of analysis throughout this report is the specific cancer diagnosis and not unique patient. Therefore, any patients with multiple cancer diagnoses in the year 2010 would also have multiple records in the database and could potentially be included in the analysis several times.

Eligibility Criteria

There were 3,687 non-invasive or in situ cancers excluded from analysis. The six cancer types most affected by exclusion of in situ cases were urinary bladder (40%, n=1,473), melanoma of the skin (31%; n=1,158), colon (8%, n=311), rectum (3%, n=104), cervix (3%, n=97), and breast (3%, n=93). Among the in situ breast cancers, the majority were diagnosed among females (n=88). The remaining 46,170 invasive cancers were included in the analysis.

While 46,170 invasive cancers were included in the overall analysis and reporting of incidence cancers, additional analysis-specific exclusions were made and are described as table footnotes. Specifically, cases with non-binary or missing sex (e.g., not either male or female) (n=4) were excluded from sex-specific analyses. Similarly, cases missing information regarding race (e.g., missing, unknown, or not reported race) (n=548) were excluded from race-specific analyses. One case with missing geographic information was excluded from analyses focused on geographic location (n=1). Stage information was not applicable (n=4,667), or missing, unknown or invalid (n=3,978) for a total of 8,645 cases which were designated as “No Stage” for stage-specific analyses.

Underlying VA Patient Population

To calculate standardized estimates of incidence, we collected information on the number of VA healthcare system users in calendar year 2010 from the Veterans Health Administration (VHA) Support Service Center Capital Assets (VSSC) current enrollment database. A VA user was defined as someone who: 1) enrolled in VA healthcare; and 2) received care at any VA location(s).

Statistical Analysis

Frequency distributions of incident cancer diagnoses were evaluated by primary anatomical site, sex, race, and geographic region. For the most commonly occurring cancers, we also created a frequency distribution by stage at diagnosis. Cancer frequency trends within the VA population were examined descriptively and informal comparisons were made with the larger U.S. cancer population. Due to the categorical nature of these data, the characteristics of cancer cases were summarized using frequencies and proportions. These were compared by sex and race at the organ system level using a chi-square test when a minimum of 25 cases occurred in each group. To protect veterans’ privacy, cells with five or fewer cases were redacted. Standard methods of age-adjustment to the 2010 estimated U.S. population²⁰ were used in support of comparisons with the general U.S. population, as follows: 1) within each age category, we calculated the crude cancer incidence within the VA based on the 2010 population of VA healthcare system users; 2) within each age category, we calculated the expected cancer frequency for the 2010 Estimated Population as (crude incidence in VA) × (2010 Estimated Population); 3) summed the expected cancer frequencies over age categories; and 4) estimated the age-adjusted cancer incidence by dividing the summed frequencies by the 2010 Estimated Population total. Statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC) and R version 3.2.2 using the ggmap and qmap packages.

Demographic Characteristics

In calendar year 2010, approximately 5,894,299 patients used the VA healthcare system. A total of 49,857 of these patients were reported in the VACCR with a cancer diagnosis. After excluding in situ diagnoses, 46,170 cases remained for analysis. Among male/female cancer diagnoses (n=46,166), 97% (n=44,836) occurred in males, while 3% (n=1,330) were in females. The median age of VA patients newly diagnosed with cancer in 2010 was 65 years (10th percentile = 55 years, 90th percentile = 81 years). Among cancer cases with known race (n=45,622), approximately 79% (n=36,308) occurred among white patients, compared to almost 20% (n=8,744) among black patients and just over 1% (n=570) among patients of other minority races.

Overall Incident Cases

The three most frequently occurring cancers in 2010 were prostate (29%, n=13,438), lung/bronchus (18%, n=8,216), and colon/rectum (8%, n=3,793). Cancers of the kidney/renal pelvis (4%, n=1,770) and melanoma of the skin (4%, n=1,733) are fourth and fifth, respectively, but tied for the fourth position in terms of percent. Liver and intrahepatic bile duct (3%, n=1,567) and urinary bladder (3%, n=1,546) cancers closely followed. The top three most frequently occurring cancers in the VA are the same as those observed among U.S. males (prostate, lung/bronchus, colon/rectum).¹⁴

Incident Cases by Sex

Incident cases by sex are shown in Table 1. The most frequently diagnosed cancer among Veteran males was prostate cancer, accounting for 30% (n=13,438) of male cancer diagnoses (Figure 1). Among female Veterans, breast cancer was the most commonly diagnosed, accounting for about 30% (n=402) of female cancers. For both sexes, cancers of the lung/bronchus (male, 18%, n=8,019; female, 15%, n=197) and colon/rectum (male, 8%, n=3,705; female, 7%, n=88) occur second and third most frequently, respectively. This is consistent with national non-VA data.¹⁴ Kidney and renal pelvis is fourth in ranking for males within VA (n=1,733), comprising approximately 4% of VA male cancer diagnoses. It is followed closely by melanoma and urinary bladder (4%, n=1,674; and 3%, n=1,536, respectively). For females in the VA, the fourth ranked cancer was uterine corpus (6%, n=75). Melanoma (4%, n=59) and thyroid (4%, n=53) cancers were also common among females. This is similar to national non-VA data, where thyroid is the fifth most commonly diagnosed cancer among females.¹⁴ Chi-square tests indicated statistically significant differences by sex for each organ level, with the exception of melanoma of the skin, and hematologic malignancies (e.g., lymphoma, myeloma, leukemia, and other and unspecified primary sites).

Incident Cases by Race

Table 2 presents incident cases by cancer site and racial group. The most striking difference was in prostate cancer. For all racial groups, prostate was the most common cancer, yet the proportions varied substantially by racial group (white 26%, n=9,515; black 41%, n=3,573; other minority 30%, n=169). Additionally, among white, black, and other minority Veterans there were differences in the diagnosis of cancers of the lung/bronchus (19% vs. 15% vs. 13%, respectively). Melanoma incidence also varied among Veterans of white, black, and other minority races (5% vs. <1% vs. 2%). Statistically significant racial differences were observed for many organ levels including head and neck, digestive and respiratory systems, bone marrow, and genital and urinary systems, among others (Table 2) There were no significant racial differences for endocrine, soft tissue, and central nervous systems.

Age Distribution

The median age at diagnosis for VA cancers in 2010 was 65 years; the median age at diagnosis for all SEER cancer sites during a similar time period was also 65 years.²¹ Within age categories, most cancers were diagnosed in Veterans aged 55–64 years, accounting for approximately 41% of cancer diagnoses (Table 3). After adjusting the age structure of the VA population to that of the 2010 U.S. Estimated Population, we estimated that the age-adjusted cancer incidence is 508 cases per 100,000 person-years. This is slightly higher than the approximately 473 cases per 100,000 person-years reported for all of the US.¹⁴

Because female users of the VA healthcare system are, on average, younger than male users²², we also evaluated differences in age at diagnosis by sex (results not shown). The median age at diagnosis among males and females was 65 and 57 years respectively. National SEER estimates from 2009–2013 depict an overall median age at diagnosis of 65 years for both sexes (66 years for males and 65 years for females).²¹ The median age at diagnosis varied slightly according to anatomic site (data not shown). For prostate cancer, the median age at diagnosis in the VA was 64 years and the SEER national estimate is 66 years.²¹

The median age of VA lung cancer cases is 66 years; the SEER median age at diagnosis for both sexes combined and for males is 70 years.²¹ For VA colon/rectum cancer patients, median age at diagnosis was 66 years (66 years for males, 60 years for females), while the SEER estimate was 68 years for both sexes combined (66 years for males and 70 for females).²¹

Geographic Distribution

The US Veteran population is not evenly distributed geographically. Florida, Texas, and California have the largest Veteran populations.²³ Among Veterans with cancer reported in the VACCR and with geographic location data (n=46,169), approximately 42% (n=19,351) were in the South, 22% (n=10,080) were in the West, 21% (n=9,887) were in the Midwest, and 15% (n=6,851) were in the Northeast (Figure 2; Figure 3).

Stage Distribution of Common Cancers

Stage at diagnosis dictates appropriate treatments and is an important predictor of survival. Stage at diagnosis was examined for the three most common cancer sites: prostate, lung/bronchus, and colon/rectum (Figure 4). In 2010, fewer than half (38%, n=3,122) of VA lung/bronchus cancers were metastatic at diagnosis (TNM Stage 4I), compared with 55% with distant disease in the SEER population.¹⁴

Cancers of the colon/rectum also tended to be diagnosed at earlier stages in the VA healthcare system (Stage 1I 31%, n=1,177; Stage 2II 22%, n=836, Stage 3III 18%, n=681). Approximately 19% (n=725) of colon/rectum cancers were diagnosed were diagnosed at Stage 4 IV in the VACCR (identical to distant disease in SEER).¹⁴ Prostate cancers were also diagnosed at similar stages for Veterans using the VA healthcare system compared with national data. Rates of metastatic/distant prostate cancer diagnosis were slightly higher in the VA than SEER (7% vs. 4%, respectively).¹⁴