The impact of socioeconomic status on survival after cancer in the United States

Findings from the National Program of Cancer Registries Patterns of Care Study


  • Disclaimers: The authors declare no financial conflicts of interest in this work. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

  • Supported by cooperative agreements between the Centers for Disease Control and Prevention; the states of California, Colorado, Illinois, Louisiana, New York, Rhode Island, and South Carolina.



Understanding the ways in which socioeconomic status (SES) affects mortality is important for defining strategies to eliminate the unequal burden of cancer by race and ethnicity in the United States.


Disease stage, treatment, and 5-year mortality rates were ascertained by reviewing medical records, and SES was determined by analyzing income and education at the census tract level for 4844 women with breast cancer, 4332 men with prostate cancer, and 4422 men and women with colorectal cancer who were diagnosed in 7 U.S. states in 1997.


Low SES was associated with more advanced disease stage and with less aggressive treatment for all 3 cancers. The hazard ratio (HR) for 5-year all-cause mortality associated with low SES was elevated after a diagnosis of breast cancer when the analysis was adjusted for age (HR, 1.59; 95% confidence interval [CI], 1.35-1.87). Adjustment for mediating factors of race/ethnicity, comorbid conditions, cancer stage, and treatment reduced the association. The age-adjusted mortality risk associated with low SES was elevated after a diagnosis of prostate cancer (HR, 1.33; 95% CI, 1.13-1.57), and multivariate adjustments for mediating factors also reduced that association. There was less association between SES and mortality after a diagnosis of colorectal cancer. For all 3 cancer sites, low SES was a much stronger predictor of mortality among individuals aged <65 years and among individuals from racial/ethnic minority groups.


The current results indicated that low SES is a risk factor for all-cause mortality after a diagnosis of cancer, largely because of a later stage at diagnosis and less aggressive treatment. These findings support the need to focus on SES as an underlying factor in cancer disparities by race and ethnicity. Cancer 2008. © 2008 American Cancer Society.

Disparities by race and ethnicity are well documented among cancer patients in the United States, particularly in studies comparing African Americans with non-Hispanic whites.1-9 Although the incidence rates for many sites are lower among racial and ethnic minority groups than among non-Hispanic whites, both Hispanics and African Americans have a higher risk of mortality after a diagnosis of cancer.2, 10 Adjustment for clinical factors, such as disease stage at diagnosis and treatment, seems to explain some, but not all, of these disparities in survival.11-16 Although there appear to be biologic differences by race or ethnicity in the characteristics of certain cancers, socioeconomic factors clearly also are important determinants of racial and ethnic disparities in survival.17-27 Numerous studies have examined cancer disparities by race/ethnicity, but few have examined the independent contribution of socioeconomic status (SES) to these disparities across multiple cancer sites. In the current study, we measured the association between SES and mortality after a diagnosis of cancer in a large population-based sample of individuals who were diagnosed with cancers of the breast, colon, rectum, or prostate.


This is an analysis of data from the Breast, Colon, and Prostate Cancer Data Quality and Patterns of Care (POC) Study of the National Program of Cancer Registries (NPCR).28 In brief, the NPCR POC Study is a collaborative inquiry by 7 state registries (California, Colorado, Illinois, Louisiana, New York, Rhode Island, and South Carolina) and is funded and coordinated by the Centers for Disease Control and Prevention (CDC). The purpose of the POC Study is to assess the adequacy of cancer treatment in the United States and to relate treatment to outcomes.

The POC Study reabstracted information from both inpatient and outpatient sources for a population-based sample of cancer cases from a single calendar year (1997) to determine how often patients were receiving treatment consistent with consensus guidelines and how treatment was related to survival. The POC Study also was designed to compare 5-year all-cause mortality between this large U.S. sample and a similar population-based sample of cancer cases from countries across Europe (the CONCORD Study).29 In each collaborating state, random samples of cases diagnosed in 1997 were selected for 3 sites: breast (International Classification of Disease for Oncology-2nd edition [ICD-O2] codes C50.0-C50.9), prostate (ICD-O2 code C61.9), and colorectal (ICD-O2 codes C18.0-C20.9). Localized breast and prostate cancer cases and regional colon cancer cases were over sampled to increase the power to assess compliance with evidence-based treatment guidelines. In each state, study personnel abstracted information from both hospital medical records and outpatient sources on patient and tumor characteristics and on the treatment received. Outpatient treatment sources were identified by using a snowballing technique, which began with the hospital record and added outpatient surgical, radiation therapy, chemotherapy, oncology, and primary care settings as those treatment providers were identified. Comorbidities for serious chronic conditions other than the index cancer were ascertained by using the method described by Charlson et al.30 All-cause mortality was ascertained through 5 years after the date of diagnosis both from the record sources used to collect treatment information and from searches of each state's vital records system and the National Death Index. Data privacy considerations precluded the submission of cause-specific mortality from all of the collaborating states, so the current analysis is limited to total mortality.

For this analysis, the adequacy of cancer treatment was assessed for breast cancer and colorectal cancer but not for prostate cancer, because 1997 consensus guidelines were available for only breast and colorectal cancer treatment.31 For breast cancer, adequacy was defined by the receipt of adjuvant radiation therapy for localized cases after a lumpectomy, by the receipt of chemotherapy for regional cases, and by the receipt of antiestrogen therapy when the cancer had tested positive for estrogen or progesterone receptors. For individuals with regional-stage colon cancer, adequacy was defined by the receipt of adjuvant chemotherapy.

The SES of each patient was assessed by using previously validated methods that included the education and income levels of the census tract of residence, as ascertained by the year 2000 Census.32-38 Both education and income were classified into 2 levels (<25% vs ≥25% of adults aged ≥25 years with less than a high school education and <20% vs ≥20% of households with incomes below the Federal Poverty Level).39 Each patient was then classified as living in a census tract with neither low education nor low income (65% of cases), with only 1 of those indicators of low SES (20% of cases), or with both of those indicators (15% of cases).

For each site (breast, prostate, or colorectal), first, we examined differences in disease stage and treatment across the 3 levels of SES. Then, we compared 5-year all-cause mortality across the 3 SES categories, first with univariate proportional hazards modeling and then with multivariate models that were adjusted progressively for mediating factors coded as sets of dummy variables: race/ethnicity (non-Hispanic white, Hispanic white, African American, or all others), stage (in situ, localized, regional, distant, or unknown), and treatment (did not receive consensus treatment, did receive consensus treatment, or unknown). Sex and subsite were covariates only for the colorectal analyses (subsites were colon and rectum). Treatment was a covariate only for the breast and colorectal analyses. Hazard ratios (HRs) were estimated by proportional hazards analysis using age as a fixed covariate. Participants were censored in the analysis when death occurred, when they were lost to follow-up, or at 5 years after diagnosis. We also conducted sensitivity analyses in which we assumed that all individuals who were censored because of loss to follow-up had survived for the full 5 years. All analyses were conducted using SAS software (version 9.1). This study was approved by the Institutional Review Board at each collaborating state health department and university and at the CDC.


Among 4844 breast cancer cases, 4422 colorectal cancer cases, and 4332 prostate cancer cases, over half were aged >65 years, and approximately 80% were non-Hispanic whites (Table 1). Among the colorectal cancer cases, 85% involved the colon, and 15% involved the rectum. African Americans and Hispanics were more likely than non-Hispanic whites to have lived in low-SES areas (Table 2). Among individuals aged <65 years, those who lived in low-SES areas were more likely not to have health insurance; and, among individuals aged ≥65 years, those in the low-SES category were more likely to have comorbidities. Living in a low-SES area was associated with more advanced stage at diagnosis for breast cancer and prostate cancer but was unrelated to stage at diagnosis for colorectal cancer (Table 3).

Table 1. Characteristics of 13,598 Patients With Cancer Diagnosed in 1997 in 7 U.S. States From the Patterns of Care Study
CharacteristicCancer site: No. of patients (%)
  1. HS indicates high school; SES, socioeconomic status.

Total4844 (100)4422 (100)4332 (100)
Age, y
 <501130 (23)359 (8)89 (2)
 50–641535 (32)1010 (23)1339 (31)
 65–741141 (24)1320 (30)1842 (42)
 ≥751038 (21)1733 (39)1062 (24)
 Women4844 (100)2254 (51)0 (0)
 Men0 (0)2168 (49)4332 (100)
 Non-Hispanic white3995 (82)3562 (81)3431 (79)
 Hispanic white192 (4)176 (4)149 (3)
 African American512 (11)568 (13)641 (15)
 Other/unknown145 (3)116 (3)111 (3)
 Private2242 (46)1332 (30)1468 (34)
 Medicare1837 (38)2487 (56)2090 (48)
 Medicaid/welfare149 (3)114 (3)67 (2)
 Not insured136 (3)103 (2)43 (1)
 Other/unknown480 (10)386 (9)664 (15)
Comorbidity score
 04186 (86)3234 (73)3573 (82)
 1530 (11)850 (19)587 (14)
 ≥2128 (3)338 (8)172 (4)
 Urban3246 (67)2961 (67)2855 (66)
 Nonurban1583 (33)1447 (33)1447 (33)
 Unknown15 (0.3)14 (0.3)30 (1)
Poverty level of census tract
 Poverty (≥20% in poverty)1188 (24)1183 (27)1036 (24)
 Not in poverty (<20% in poverty)3641 (75)3225 (73)3266 (75)
 Unknown15 (0.3)14 (0.3)30 (0.7)
Education level of census tract
 Undereducated (<25% HS)1244 (26)1324 (30)1181 (27)
 Not undereducated (≥25% HS)3585 (74)3084 (70)3121 (72)
 Unknown15 (0.3)14 (0.3)30 (0.7)
Joint SES level of census tract
 Neither poverty nor undereducated3148 (65)2688 (61)2789 (64)
 Either poverty or undereducated930 (19)933 (21)809 (19)
 Both poverty and undereducated751 (16)787 (18)704 (16)
 Unknown15 (0.3)14 (0.3)30 (0.7)
Table 2. Relation Between Socioeconomic Status of Residence and Personal Characteristics of Patients
CharacteristicCensus tract SES: No. of patients (%)*
Neither in poverty nor undereducatedEither in poverty or undereducatedBoth in poverty and undereducated
  • SES indicates socioeconomic status.

  • *

    Column percentages may not add to 100% because of rounding.

Age, y
 <501018 (12)287 (11)269 (12)
 50-642514 (29)745 (28)598 (27)
 65-742733 (32)848 (32)704 (31)
 ≥752360 (27)792 (30)671 (30)
 Women4495 (52)1409 (53)1172 (52)
 Men4130 (48)1263 (47)1070 (48)
 Non-Hispanic white7678 (89)2000 (75)1270 (57)
 Hispanic white227 (3)122 (5)167 (7)
 African American494 (6)473 (18)737 (33)
 Other/unknown226 (3)77 (3)68 (3)
Comorbidity score
 Age <65 y
  03177 (90)910 (88)730 (84)
  1308 (9)102 (10)118 (14)
  ≥247 (1)20 (2)19 (2)
 Age ≥65 y
  03900 (77)1222 (75)1007 (73)
  1864 (17)305 (19)262 (19)
  ≥2329 (6)113 (7)106 (8)
 Age <65 y
  Private2624 (74)652 (63)468 (54)
  Medicare239 (7)107 (10)105 (12)
  Medicaid/welfare78 (2)64 (6)95 (11)
  Not insured87 (2)76 (7)91 (11)
  Other/unknown504 (14)133 (13)108 (12)
 Age >65 y
  Private859 (17)221 (13)202 (15)
  Medicare3700 (73)1226 (75)1007 (73)
  Medicaid/welfare50 (1)25 (2)17 (1)
  Not insured8 (0.2)10 (0.6)8 (0.6)
  Other/unknown476 (9)158 (10)141 (10)
Table 3. Relation Between Socioeconomic Status and Stage at Diagnosis*
Type of cancerCensus tract SES: No. of patients (%)
Neither in poverty nor undereducatedEither in poverty or undereducatedBoth in poverty and undereducated
  • SES indicates socioeconomic status.

  • *

    Data are presented as the number of patients (%) within each SES group. These stage distributions are affected by the over sampling of patients with localized breast and prostate cancer and patients with regional colon cancer in the study.

  • P values for the association between stage and poverty (chi-square test): P<.0001 for breast cancer, P = .0009 for prostate cancer, and P = .60 for colorectal cancer.

 In situ391 (12)81 (9)78 (10)
 Local1984 (63)568 (61)429 (57)
 Regional609 (20)217 (23)173 (23)
 Distant91 (3)33 (4)43 (6)
 Unknown/unstaged73 (2)31 (3)28 (4)
 In situ131 (5)51 (5)47 (6)
 Local697 (26)237 (25)199 (26)
 Regional1339 (50)437 (47)380 (48)
 Distant410 (15)160 (17)128 (16)
 Unknown/unstaged111 (4)48 (5)33 (4)
 Local2171 (78)585 (72)508 (72)
 Regional261 (9)81 (10)67 (10)
 Distant126 (4)47 (6)45 (6)
 Unknown/unstaged231 (8)96 (12)84 (12)

Treatment differed by SES for each of the 3 cancer sites. Women who lived in low-SES areas who were diagnosed with localized breast cancer were more likely to have undergone mastectomy than women who were not from low-SES areas (Table 4). Among women who underwent lumpectomy, those from low-SES areas were less likely to have received adjuvant radiotherapy than those from high-SES areas (60% vs 77%; P≤.001). Among women with regional-stage breast cancer, those from low-SES areas were marginally less likely to have received adjuvant chemotherapy than those from high-SES areas (63% vs 68%; P = .16). Among women with estrogen/progesterone receptor-positive breast cancer, those from low-SES areas also were marginally less likely to have received antiestrogen therapy (52% vs 58%; P = .08). Among patients who were diagnosed with regional-stage colon cancer, those from low-SES areas were less likely to have received adjuvant chemotherapy than those from high-SES areas (50% vs 56%; P = .02). Finally, men who resided in low-SES areas were less likely to have been treated by either prostatectomy or radiation (considered the 2 standard first-line therapies) than men from high-SES areas (67% vs 78%; P≤.0001) and were more likely to have received hormone therapy (12% vs 6%; P≤.0001).

Table 4. Relation Between First Course of Treatment and Socioeconomic Status
First course of treatmentCensus tract SES: No. of patients (%)
Neither in poverty nor undereducatedEither in poverty or undereducatedBoth in poverty and undereducated
  1. SES indicates socioeconomic status; ER, estrogen receptor; PR, progesterone receptor.

Breast cancer
 Localized stage
  Mastectomy750 (38)278 (49)199 (46)
  Lumpectomy1228 (62)289 (51)229 (53)
  Localized stage and lumpectomy
  Radiation944 (77)191 (66)137 (60)
  No radiation269 (22)90 (31)84 (37)
  Unknown15 (1)8 (3)8 (4)
 Regional stage
  Chemotherapy415 (68)134 (62)109 (63)
  No chemotherapy183 (30)79 (36)60 (35)
  Unknown11 (2)4 (2)4 (2)
 ER- or PR-positive
  Hormone therapy1052 (58)276 (54)210 (52)
  No hormone therapy714 (39)204 (40)184 (46)
  Unknown55 (3)27 (5)8 (2)
 All patients
  Received standard care2046 (65)572 (62)437 (58)
  Did not receive standard care1026 (33)323 (35)279 (37)
  Unknown76 (2)35 (4)35 (5)
Colorectal cancer
 Regional stage
  Chemotherapy753 (56)221 (51)189 (50)
  No chemotherapy541 (40)199 (46)176 (46)
  Unknown45 (3)17 (4)15 (4)
 All patients
  Received standard care1991 (74)669 (72)563 (72)
  Did not receive standard care541 (20)199 (21)176 (22)
  Unknown156 (6)65 (7)48 (6)
Prostate cancer
 Localized stage
  Prostatectomy843 (39)201 (34)166 (33)
  Radiation therapy851 (39)227 (39)175 (34)
  Chemotherapy1 (0.1)0 (0)0 (0)
  Hormone therapy135 (6)48 (8)61 (12)
  Other treatment2 (0.1)0 (0)0 (0)
  Unknown or watchful waiting339 (16)109 (19)106 (21)

Mortality was higher among individuals living in low-SES areas than among those in high-SES areas, as reflected by increased 5-year HRs (Table 5) and cumulative mortality rates (Fig. 1). For breast cancer, models that were adjusted only for age showed substantially increased risk for women in the lowest SES areas (HR, 1.59; 95% confidence interval [CI], 1.35-1.87); adjusting for race/ethnicity reduced this association somewhat (HR, 1.33; 95% CI, 1.11-1.58); and further adjustment for stage, comorbidities, and treatment reduced this association further (HR, 1.16; 95% CI, 0.97-1.38). In the fully adjusted breast cancer models, the association between mortality and SES was more apparent among women other than non-Hispanic whites (HR, 1.52 for all others, 1.11 for non-Hispanic whites).

Figure 1.

The relation between socioeconomic status (SES) of area of residence and mortality within 5 years after cancer diagnosis in the Patterns of Care Study.

Table 5. Relation Between Socioeconomic Status and Cancer Mortality by Race/Ethnicity
VariableCensus tract SES: HR (95% CI)
Neither in poverty nor undereducatedEither in poverty or undereducatedBoth in poverty and undereducated
  • SES indicates socioeconomic status; HR, hazard ratio; CI, confidence interval.

  • *

    Numbers of patients are somewhat smaller than in the total study sample (Table 1), because individuals with missing variables were not included in the multivariate models.

  • Race/ethnicity-specific HRs were adjusted for all factors except race/ethnicity.

  • Age-specific HRs were adjusted for all factors.

  • §

    Sex-specific HRs were adjusted for all factors except sex.

Breast cancer: 935 Deaths/4828 patients [19.4%]*538/3141 [17.1%]203/925 [22%]192/748 [25.7%]
 Adjustment factors
  Age1.0 (Reference)1.38 (1.18-1.62)1.59 (1.35-1.87)
  Age, race/ethnicity1.0 (Reference)1.28 (1.09-1.51)1.33 (1.11-1.58)
  Age, race/ethnicity, comorbidity1.0 (Reference)1.26 (1.07-1.48)1.28 (1.07-1.52)
  Age, race/ethnicity, comorbidity, stage1.0 (Reference)1.11 (0.94-1.31)1.17 (0.98-1.39)
  Age, race/ethnicity, comorbidity, stage, treatment1.0 (Reference)1.11 (0.94-1.31)1.16 (0.97-1.38)
 Race/ethnicity-specific full model
  Non-Hispanic whites1.0 (Reference)1.02 (0.84-1.23)1.11 (0.90-1.38)
  All others1.0 (Reference)1.68 (1.17-2.40)1.52 (1.09-2.10)
 Age-specific full models
  Age <65 y1.0 (Reference)1.65 (1.25-2.18)1.25 (0.92-1.69)
  Age ≥65 y1.0 (Reference)0.92 (0.75-1.13)1.08 (0.87-1.35)
Colorectal cancer: 2251 Deaths/4419 patients [50.9%]*1326/2686 [49.4%]508/932 [54.5%]410/787 [52.1%]
 Adjustment factors
  Age, sex1.0 (Reference)1.19 (1.07-1.32)1.10 (0.98-1.23)
  Age, sex, race/ethnicity1.0 (Reference)1.17 (1.05-1.29)1.03 (0.92-1.17)
  Age, sex, race/ethnicity, comorbidity1.0 (Reference)1.18 (1.06-1.31)1.03 (0.91-1.16)
  Age, sex, race/ethnicity, comorbidity, stage1.0 (Reference)1.15 (1.04-1.28)1.10 (0.98-1.24)
  Age, sex, race/ethnicity, comorbidity, stage, treatment1.0 (Reference)1.14 (1.03-1.27)1.09 (0.97-1.23)
 Race/ethnicity-specific full models
  Non-Hispanic whites1.0 (Reference)1.11 (0.99-1.24)1.10 (0.95-1.26)
  All others1.0 (Reference)1.39 (1.08-1.80)1.28 (1.02-1.62)
 Age-specific full models
  Age <65 y1.0 (Reference)1.37 (1.11-1.68)1.38 (1.08-1.77)
  Age ≥65 y1.0 (Reference)1.07 (0.95-1.21)1.03 (0.90-1.18)
 Sex-specific full models§
  Men1.0 (Reference)1.09 (0.94-1.26)0.99 (0.84-1.18)
  Women1.0 (Reference)1.20 (1.03-1.38)1.20 (1.01-1.42)
Prostate cancer: 921 Deaths/4322 patients [21.3%]*527/2784 [18.9%]207/806 [25.7%]184/703 [26.2%]
 Adjustment factors
  Age1.0 (Reference)1.44 (1.23-1.70)1.33 (1.13-1.57)
  Age, race/ethnicity1.0 (Reference)1.35 (1.15-1.59)1.17 (0.98-1.40)
  Age, race/ethnicity, comorbidity1.0 (Reference)1.33 (1.13-1.57)1.16 (0.97-1.39)
  Age, race/ethnicity, comorbidity, stage1.0 (Reference)1.22 (1.03-1.44)1.14 (0.95-1.37)
 Race/ethnicity-specific full model
  Non-Hispanic whites1.0 (Reference)1.13 (0.94-1.38)1.25 (1.00-1.56)
  All others1.0 (Reference)1.65 (1.16-2.35)1.23 (0.88-1.71)
 Age-specific full models
  Age <65 y1.0 (Reference)1.59 (1.03-2.45)1.05 (0.65-1.70)
  Age ≥65 y1.0 (Reference)1.21 (1.01-1.45)1.16 (0.96-1.41)

For colorectal cancer, there was little association between SES and mortality overall (HR, 1.10; 95% CI, 0.98-1.23), but there was a substantial difference in the SES association with mortality by age (HR, 1.38 among those aged <65 years, but 1.03 among those aged ≥65 years). There was also a substantial difference in the association by race/ethnicity (HR, 1.28 among those who were other than non-Hispanic whites but only 1.10 among non-Hispanic whites).

For prostate cancer, the association between low SES and mortality in models that were adjusted only for age (HR, 1.33; 95% CI, 1.13-1.57) became statistically insignificant when they were adjusted for race/ethnicity (HR, 1.17; 95% CI, 0.98-1.40), and further adjustment for stage and treatment did not alter the association appreciably (HR, 1.14; 95% CI, 0.95-1.37) (Table 5). Sensitivity analyses in which we assumed that all individuals who were lost to follow-up but not known to be dead had survived the full 5 years did not substantially change any of these relations.


This study indicates that there are disparities by SES (as measured by socioeconomic characteristics of the neighborhood of residence) in cancer stage, treatment, and subsequent mortality in the United States. In this population-based analysis from 7 states, a substantial proportion of the SES difference in 5-year mortality after the diagnosis of cancer was attributable to a later stage of cancer at diagnosis and to less aggressive approaches to treatment for patients living in low-SES areas.

There is evidence that cancer presents with a more aggressive phenotype among patients from minority racial/ethnic groups.17-20 Therefore, biologic differences may explain some of the associations between low SES and both stage and mortality that we observed in this study. Social factors, however, seem to be more important than biologic factors in explaining racial/ethnic cancer disparities. There is a larger racial disparity in cancer mortality in the United States, where access to medical care is tied to economic status, than in Canada, where access is more universal.21 Within the United States, studies of cancer mortality in health maintenance organization settings and in the Department of Defense and Veterans Administration healthcare systems indicate much smaller differences in cancer survival rates between racial/ethnic groups than are observed in the general population.22-26 Medicare is nearly universal for Americans after age 65 years. Therefore, it is important to note that, in the current analysis, the SES-related disparities were much less apparent among individuals aged ≥65 years, who generally are afforded access to cancer screening and treatment services through Medicare regardless of their SES. We also observed that the association between low SES and increased mortality was more apparent among racial/ethnic minorities, which suggests that interacting forces between SES and race/ethnicity contribute to cancer disparities.

The current analysis was limited to all-cause mortality, because cause-specific mortality could not be contributed by all collaborating states. Therefore, the mortality differences we observed by SES may have been caused in part to SES differences in causes of death other than cancer. Adjusting for differences in chronic conditions that were prevalent at the time of cancer diagnosis probably does not account completely for this possibility. Total mortality is the usual standard, however, for outcomes in cancer surveillance and population-based research, especially for outcomes up to 5 years, when most deaths are because of cancer. (In this study, cancer accounted for 61% of the deaths in the first 5 years after diagnosis among the states that were able to report causes of death.) In further research on SES differences, other endpoints, such as disease-free survival and cause-specific mortality, should be a priority.

Although individual measures of SES may have been preferable to the neighborhood-level measures that we used to define the SES of study patients, it has been demonstrated that geographically based estimates like those used in this study are useful markers of social influence on cancer outcomes, because neighborhood-level SES factors, in combination with individual characteristics, can affect access to healthcare and its acceptability.40, 41 Both income and education are informative indicators of SES; however, currently, income may be a stronger correlate of lower health status than education.38

The current results suggest that low SES is a substantial adverse prognostic factor after a diagnosis of breast cancer. Factors tied to later stage at diagnosis and less aggressive cancer treatment for women living in low-SES areas accounted for a substantial proportion of this association. Of the 3 cancers we examined, breast cancer was the 1 with the best evidence base for the benefits of early detection and treatment. Mammographic screening increases 5-year survival, as does adherence to consensus recommendations regarding adjuvant chemotherapy for regional disease and antiestrogen therapy for cancers that test positive for estrogen receptor.42 Higher mortality after a diagnosis of breast cancer in low-SES areas well may reflect a combined effect of several factors, including sociocultural and behavioral factors tied to risk of disease; educational and access factors tied to early detection; and economic, sociocultural, and healthcare structural problems tied to cancer treatment.

We also observed adverse effects of low SES on survival after prostate cancer, but those associations were substantially weaker than those we observed for breast cancer. The evidence base for benefits from screening and treatment options is developed far less for prostate cancer than for breast cancer.

SES was not a substantial risk factor for mortality after a diagnosis of colorectal cancer across all ages, but we observed that it was a risk factor for mortality among individuals aged <65 years. This finding may be because of less access to these treatments among individuals who are disadvantaged socioeconomically. That this disparity disappeared after age 65 years may reflect an effect of Medicare coverage. In colon cancer, the main benefit of screening is reduction in the risk of new cases by the removal of adenomas, which cannot be assessed in a case-only study of this type.43

Clearly, access to both clinical preventive services and proper cancer treatment are major determinants of outcomes after cancer. Therefore, adequacy of health insurance is a major cancer risk factor in the United States.44 Despite numerous studies of this type demonstrating that socioeconomic factors affect outcomes, most cancer surveillance systems continue to report data on incidence, stage, survival, and mortality stratified only by race/ethnicity. Race and ethnicity, thus, are being used in large part as proxy measures for SES-related factors, such as education, income, and health insurance, which are the modifiable factors in the equation. Because it is increasingly apparent that a substantial proportion of the disparities in cancer defined by race/ethnicity can be attributed to SES, and because SES can be estimated now quite easily either by enhancing medical record and tumor record systems or by using geocoding methods like as those we used in the current study, now it would be prudent for cancer surveillance systems to routinely report cancer incidence and mortality data by SES levels.45–47 Better information on how cancer outcomes are related to SES is needed if we are to properly identify and address the root causes of racial and ethnic cancer disparities in the United States.