Influence of socioeconomic factors on survival after breast cancer—A nationwide cohort study of women diagnosed with breast cancer in Denmark 1983–1999

Authors


Abstract

The reasons for social inequality in breast cancer survival are far from established. Our study aims to study the importance of a range of socioeconomic factors and comorbid disorders on survival after breast cancer surgery in Denmark where the health care system is tax-funded and uniform. All 25,897 Danish women who underwent protocol-based treatment for breast cancer in 1983–1999 were identified in a clinical database and information on socioeconomic variables and both somatic and psychiatric comorbid disorders was obtained from population-based registries. We used Cox proportional hazards models to estimate the association between socioeconomic position and overall survival and further to analyse breast cancer specific deaths in a competing risk set-up regarding all other causes of death as competing risks. The adjusted hazard ratio (HR) for death was reduced in women with higher education (HR, 0.91; 95% confidence interval (CI), 0.85–0.98), with higher income (HR, 0.93; 95% CI, 0.87–0.98) and with larger dwellings (HR, 0.90; 95% CI, 0.85–0.96 for women living in houses larger than 150 m2). Presence of comorbid disorders increased the HR. An interaction between income and comorbid disorders resulting in a 15% lower survival 10 year after primary surgery in poor women with low-risk breast cancer having comorbid conditions (∼65%) compared to rich women with similar breast cancer prognosis and comorbid conditions (∼80%) suggests that part of the explanation for the social inequality in survival after breast cancer surgery in Denmark lies in the access to and/or compliance with management of comorbid conditions in poorer women. © 2007 Wiley-Liss, Inc.

Overall survival of breast cancer patients has improved but unfortunately not all groups of women share these benefits equally. Although affluent women have a higher incidence of breast cancer than socially deprived women, several studies, using individual and area-based socioeconomic measures, have shown consistently that deprived women with breast cancer have poorer survival.1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 Social inequalities in survival may involve differences in the timing of diagnosis, in the biological characteristics of the tumour, in treatments applied or in patient-specific factors, such as psychosocial factors or presence of comorbid conditions.14, 15 Comorbidity in breast cancer patients limits treatment options, increases the risk of death from other causes,16, 17 and adversely affects survival.16, 18 However, the extent to which socioeconomic differences in comorbidity explain disparities in survival of breast cancer is not clear.

In an earlier study, among postmenopausal Danish women, we observed, an independent effect of education, disposable income and living in rural areas on the risk of being diagnosed with a high risk breast cancer (as defined by tumour size > 20 mm, 1 or more tumour positive axillary lymph nodes, ductal tumour with grade of malignancy II or III and receptor negative status).19 Consequences of such differences in tumour progression at diagnosis include not only the need for more aggressive and radical cancer treatment but possibly also shorter survival. The aim of the present study is, in a population based setting, to compare the relative importance of a range of socioeconomic factors, and comorbid disorders on overall and breast cancer specific survival in a population-based sample of almost 26,000 women treated for breast cancer.

Material and methods

Identification of breast cancer patients

The study population consisted of all 31,770 women identified in the files of the Danish Breast Cancer Cooperative Group (DBCG) with a primary invasive breast cancer diagnosed between January 1, 1983 and December 31, 1999 and who were less than 70 years of age at the time of diagnosis. The DBCG has registered breast cancer patients in Denmark since 1977, and conducted protocol-based randomised trials of surgery, radiation, chemotherapy or endocrine therapy in patients with primary invasive breast cancer.20 The registry contains information on 95% of all Danish women below 75 years of age diagnosed with breast cancer over the period and each record contains detailed information about prognostic factors and the database holds continuously updated information on relapse-free interval and localisation of first recurrence.

We excluded women who did not undergo per protocol treatment, in total 4,886 women or 15% of the study population, because of the following reasons: disseminated or metastasising disease (N = 812), bilateral breast cancer (N = 564), contraindication for surgery (N = 376), technically inoperable (N = 318), surgery not according to protocol (N = 982), misclassified as to adjuvant treatment (N = 808), missing consent (N = 579) or other reasons (N = 447).

Tumour size was categorised into 1–10, 11–15, 16–20, 21–30, 31–50 or ≥51 mm, number of retrieved lymph nodes as 1–3, 4–9, 10–14, ≥15, number of positive lymph nodes as 0, 1–3, 4–9, ≥10, tumour grade and type as grade I, grade II, grade III, non-ductal or unknown, receptor status as positive, negative or unknown and protocol version as 1982 or 1989/1999. Of the 26,884 women who underwent treatment according to protocol, we excluded 159 women because of missing information on tumour size and 30 women because of missing information on retrieved or positive axillary lymph nodes, leaving 26,695 women with information on tumour characteristics and treatment.

Information on socioeconomic position

Information on socioeconomic characteristics of the women with breast cancer was obtained by data linkage to the population based Integrated Database for Labor Market Research administered by Statistics Denmark since 1980. The core variables in the database are derived by linkage with the Central Population Registry, covering all persons in Denmark, all companies with more than 1 employee, the taxation authorities, the Registry for Education Statistics and the Registry Relating to Unemployment. We obtained information on the individual level about a number of demographic and socioeconomic variables for the end of the year of breast cancer surgery. Further, we identified spouses, cohabiters and children aged 0–17 years of all women with breast cancer at the time of breast cancer surgery. In total we identified 22,313 partners (those married or cohabiting with the index-persons) and for each of these, we obtained the same socioeconomic and demographic information. From the Building and Dwelling Register, which contains information on exact address codes on all Danish persons, we obtained information on size, type and tenure of dwelling.

Highest attained education was categorised as basic school/high school, vocational training, higher education and unknown; job position as higher functionaries/self-employed, lower functionaries, skilled workers, unskilled workers, not in the work force (unemployed and other economically inactive—predominantly housewives) and pensioners (retirement and disability); disposable income adjusted for number of persons in household and deflated according to the 2000 value of the Danish crown (DKK) was categorised as <100,000 DKK/year, 100,000–130,000 DKK/year, 130,000–165,000 DKK/year, >165,000 DKK/year; housing tenure as owner-occupied or rental; and size of dwelling as 0–99, 100–124, 125–149, ≥150 m2. Demographic variables included age at time of diagnosis categorised in 5-year intervals, cohabitation status as: single or living with partner, children aged 0–17 years living at home as: none, 1, 2–5 and degree of urbanisation as capital area, capital suburban area, provincial cities and rural areas. Because of the missing information in 1 or more of the socioeconomic variables, 492 women (2%) were excluded.

Information on comorbid disorders

By linking the personal identification number to the files of the Danish National Patient Registry we obtained a full history of diseases leading to hospitalisation or outpatient visits for each cohort member from 1978 through 2000. The information in the Registry includes hospital and department, dates of admission and discharge, up to 20 diagnoses per hospitalisation, coded according to modified versions of the International Classification of Diseases revisions 8 (ICD-8) and 10 (ICD-10).21 We used the Charlson Index to classify comorbid disorders, as measured by hospitalisations with the diseases in question from 1978 through to 6 months prior to the breast cancer surgery. This scale provides an overall score of comorbidity based on a composite of values weighted by level of severity assigned for a total of 19 selected conditions scored from 1 to 6.22 On the basis of the accumulated sum of scores, the comorbidity index was grouped into scores of 0, 1, ≥2. To be consistent with the criteria of no previous malignant disease in DBCG we excluded 307 women (1%) because of the registration of any tumour in the National Patient Registry prior to the breast cancer surgery.

The Psychiatric Case Register contains data on all admissions to Danish psychiatric in-patient facilities since 1969, and, since 1995, information from out-patient contacts.23 There is no fee for psychiatric treatment in Denmark, and no private psychiatric facilities exist. The diagnostic system used during the study period was the ICD-8 up to 1993 and ICD-10 thereafter. By linkage to the Registry we identified all persons with any hospitalisation or out-patient contact for affective disorders by ICD-8 codes 296.09 through 296.99, 298.09-19, 300.19, 300.4 and ICD-10 codes F30 through F39. Further, all hospitalisations or out-patient contacts for schizophrenia were identified by ICD-8 code 295 and ICD-10 codes F20 and F25.

Information on vital status and causes of death

Information on date of death and underlying cause of death was obtained through linkage with the Cause of Death Register. This registry contains a computerised version of the information on the death certificates with up to 3 notified causes of deaths covering the period from 1974 through to December 31, 2001. The cause of death was coded according to ICD-8 up to 1993 and according to ICD-10 hereafter. Causes of death were classified as breast cancer deaths or non-breast cancer deaths. Through the files of the Central Population Registry it was possible to update the follow-up of all the breast cancer patients to September 30, 2005 for the analysis of overall survival.

Statistical analysis

Overall survival was defined as time from date of surgery until date of death. The date of last follow-up was September 30, 2005 and all patients known to be alive at that date were censored. Curves illustrating overall survival were computed for each covariate, using the Kaplan–Meier method. Differences in overall survival within the subgroups of the covariates were analysed using the log–rank test statistics.

The association between socioeconomic position and overall survival was evaluated using multivariate Cox proportional hazards models. The proportional hazard assumption was for each covariate verified graphically by means of plot of log minus log of the survival density function versus log time. By visual evaluation it was assessed whether the curves were approximately linear and parallel. If a covariate was found to violate the proportional hazard assumption it was included in the analyses as a stratification variable. Putative interactions between comorbid disorders and the socioeconomic or demographic factors, respectively, were entered into the model, one at the time. Using a Wald test it was tested if any of the interactions were significant. The final model included all tumour characteristics, socioeconomic and demographic variables of interest with receptor status and tumour grade and type entered as stratification variables.

Breast cancer specific deaths were analysed in a competing risk set-up where all other causes of death were regarded as competing risks. The date of last follow-up in this analysis was December 31, 2001. Curves showing the cumulative incidences for the 2 competing causes of death were plotted for each covariate, and differences within the subgroups of the covariates were analysed for each event, using the log-rank test statistics. The Cox proportional hazard model was adjusted for the same covariates as for the overall survival analyses. In addition to the Wald test for test of significance of each covariate within each competing causes of death, a likelihood ratio test was performed testing whether the effect across the 2 causes of death were significantly different.

To illustrate absolute socioeconomic differences in breast cancer survival and the importance of comorbid conditions we classified all breast cancers into 2 prognostic groups: high-risk or low-risk (low risk tumours being tumour size less than 20 mm, no positive lymph nodes, Grade 1, ductal or unknown and estrogen receptor positive or unknown and high risk all other tumours) and further classified presence of comorbid conditions, both as measured by the Charlson Index and depression or schizophrenia as a dichotomous variable, and finally used the quartiles of disposable income to categorize into rich (highest quartile) and poor (lowest quartile). Using a Cox proportional hazard model stratified by those 3 new parameters and adjusted for the remaining variables we plotted survival curves with the value of the adjusted factors set to the reference level for the 8 strata of rich and poor women with high or low risk breast cancer and with or without presence of comorbid disorders.

Patients not allocated into protocol (N = 4,886; 15%) were not included in the analyses. Consequently the issue of bias by selection was addressed. The differences for patients allocated and not allocated into protocols were investigated using contingency tables and χ2-square tests, and differences in overall survival for the 2 patients groups were investigated using Kaplan–Meier plots and log-rank tests.

Results

The descriptive characteristics of the breast cancer patients by disposable income are shown in Table I. Differences in tumour characteristics across income groups were only slight; however there were more small-size tumours (10 mm or less) and a higher number of retrieved lymph nodes among women with higher income. In regard to the socioeconomic and demographic factors, more women with lowest income had only basic school education, were unemployed or retired at the time of diagnosis as well as inhabited rental housing and lived in smaller dwellings. Further, more women with lowest income lived in rural areas, had a comorbidity score of 1 or more (9% vs. 6%) and had young children living at home compared to women with highest income (24% vs. 11%).

Table I. Descriptive Characteristics of the 25,897 Women Diagnosed with Breast Cancer in Denmark 1983–1999 by Disposable Income1
CharacteristicTotal (N = 25,897) (%)<100,000 DKK (N = 6,624) (%)100,000–129,999 DKK (N = 6,728) (%)130,000–164,999 DKK (N = 5932) (%)>164,999 DKK (N = 6,613) (%)
  • 1

    Household disposable income in Danish crowns (DKK) after taxation and interest adjusted for number of persons in household.

  • 2

    Children aged 0–17 years.

  • 3

    Presence of these disorders as defined in the Charlson Index was defined as an in- or out-patient contact with one of the below diagnosis from 1978 to half a year prior to the diagnosis of breast cancer. Score 1: Myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcer disease, mild liver disease, diabetes type1, diabetes type2. Score 2: Hemiplegia, moderate to severe renal disease, diabetes with end organ damage type1 or type 2. Score 3: Moderate to severe liver disease. Score 6: AIDS; all cancer related diagnoses were excluded since first primary cancers only are in DBCG.

Age
 <45 years1819212012
 45–49 years1610141921
 50–54 years179131926
 55–59 years1711161921
 60–64 years1720201412
 65–69 years16311697
Tumour size
 0–10 mm1614161817
 11–15 mm2119212122
 16–20 mm2223222223
 21–30 mm2425252422
 31–50 mm1314131212
 >50 mm55554
No. of retrieved lymph nodes
 1–367665
 4–94146434235
 10–143029303033
 15+2218212228
No. of positive lymph nodes
 05756575758
 1–32728282727
 4–91212111111
 10+44443
Histologic grade and type
 Grade I2726282828
 Grade II3537353632
 Grade III1616161616
 Non-ductal1817181719
 Unknown43444
Receptor status
 Negative1715181719
 Positive5854575862
 Unknown2531252519
Protocol version
 19823446363321
 1989/996654646779
Occupation
 Higher functionary1812111831
 Lower functionary188162424
 Skilled worker725813
 Unskilled worker138151711
 Unemployed1522151210
 Pensioner3049382011
Education
 Basic school/high-school4759554330
 Vocational training3021293434
 Higher education187111933
 Unknown/unregistered613543
Cohabitation status
 Living with partner7360677687
 Single2740332313
Children living at home2
 08076757989
 1121113148
 2–59131273
Housing tenure
 Owner-occupied6661586879
 Rental3439423220
Size of dwelling
 0–99 m23747463520
 100–124 m22121212219
 125–149 m21813161922
 >149 m22419172339
Degree of urbanicity
 Capital area1313121415
 Capital suburbs148141620
 Provincial cities3840403836
 Rural areas3439323230
Charlson Index3
 09291909394
 167755
 2+22221
Depressive disorder
 Ever34322
 Never9796979898
Schizophrenia
 Ever1111<1
 Never99999999100

At the end of follow-up at October 1, 2005, with a median follow-up of 12.6 years, a total of 10,505 women had died (41%). Table II presents overall survival adjusting for tumour characteristics, socioeconomic and demographic variables. There was a better survival with higher socioeconomic position, with a lower hazard ratio (HR) for death in women with higher education (HR, 0.91; 0.85–0.98, compared to only basic school or high school education), with higher income (HR, 0.93; 0.87–0.98 and HR, 0.89; 0.83–0.95, in the 2 highest income groups compared to the lowest income group), living with a partner, being a house-owner (HR, 1.05; 1.00–1.11 in women living in rented housing) and with larger dwellings (HR, 0.90; 0.85–0.96 for women living in houses larger than 150 m2). Also, presence of comorbid disorders prior to the breast cancer surgery increased the HR for death with increasing comorbidity score (HR, 1.31, 95% confidence interval (CI), 1.21–1.41 and 2.19, 95% CI, 1.94–2.47 for comorbidity scores 1 and 2+, respectively). There was, however, an interaction between comorbidity and disposable income (p = 0.007), with the HR for death increasing with comorbidity score for the 3 lowest income groups but with no association between comorbidity and death in women within the highest income group (data not shown). Presence of depression as well as schizophrenia increased the HR of dying (HR, 1.19; 1.06–1.32 and HR, 1.43; 1.21–1.70, respectively).

Table II. Hazard Ratios with 95% Confidence Intervals for Death Due to All Causes of Breast Cancer In A Cohort of 25,897 Women Diagnosed With Breast Cancer in Denmark, 1983–1999
 Hazard ratio95% Confidence limitsp-value
  • p-values from Wald's test. All results were stratified by receptor status and tumour grade and type.

  • 1

    Household disposable income in Danish crowns (DKK) after taxation and interest adjusted for number of persons in household.

  • 2

    Children aged 0–17 years.

  • 3

    Presence of these disorders as defined in the Charlson Index was defined as an in- or out-patient contact with one of the below diagnosis from 1978 to half a year prior to the diagnosis of breast cancer. Score 1: Myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcer disease, mild liver disease, diabetes type1, diabetes type2. Score 2: Hemiplegia, moderate to severe renal disease, diabetes with end organ damage type1 or type 2. Score 3: Moderate to severe liver disease; Score 6: AIDS; all cancer related diagnoses were excluded since first primary cancers only are in DBCG.

Age  <0.0001
 −44 years1  
 45–49 years0.880.81–0.95 
 50–54 years1.030.95–1.12 
 55–59 years1.201.10–1.30 
 60–64 years1.241.14–1.36 
 65–69 years1.451.31–1.60 
Tumour size  <0.0001
 1–10 mm0.640.59–0.69 
 11–15 mm0.750.71–0.80 
 16–20 mm0.890.84–0.94 
 21–30 mm1  
 31–50 mm1.211.14–1.28 
 51– mm1.481.37–1.61 
Lymph nodes retrieved  <0.0001
 0–31.121.04–1.21 
 4–91  
 10–140.810.77–0.85 
 15–0.670.63–0.72 
Positive lymph nodes  <0.0001
 01  
 1–31.571.50–1.65 
 4–93.062.89–3.24 
 10–145.224.79–5.69 
Protocol version  <0.0001
 19821.181.12–1.24 
 1989/19991  
Highest attained education  0.0343
 Basic school/high school1  
 Vocational training0.960.91–1.01 
 Higher education0.910.85–0.98 
 Unknown1.020.95–1.11 
Occupation  <0.0001
 Higher salaried employee0.890.82–0.96 
 Lower salaried employee0.820.76–0.88 
 Skilled worker0.780.69–0.88 
 Manual worker0.880.82–0.95 
 Unemployed or housewife0.910.85–0.97 
 Pensioner1  
Disposable income1  0.0018
 <100,0001  
 100,000–129,9990.990.94–1.04 
 130,000–164,9990.930.87–0.98 
 165,000–0.890.83–0.95 
Cohabiting  0.0419
 Single1  
 Living with partner0.950.91–1.00 
Children living at home2  0.1373
 01  
 10.940.86–1.01 
 2–51.020.92–1.12 
Housing tenure  0.0774
 Owner occupied1  
 Rental1.051.00–1.10 
Size of dwelling  0.0107
 0–99 m21  
 100–124 m20.960.91–1.01 
 125–149 m20.920.87–0.99 
 150– m20.900.85–0.96 
Urbanity  0.0295
 Capital area1.081.01–1.14 
 Surburbian area1.010.95–1.07 
 Provincial cities1  
 Rural areas0.970.93–1.02 
Charlson Index3  <0.0001
 None1  
 11.311.21–1.41 
 2+2.191.94–2.47 
Depression  0.0023
 No1  
 Yes1.191.06–1.32 
Schizophrenia  <0.0001
 No1  
 Yes1.431.21–1.70 

The analysis of competing causes of deaths included a restricted follow-up period ending December 31, 2001 with a total of 8,309 deaths reported (32%) and a median follow-up of 9.1 years (Table III). All covariates were significantly associated with both breast cancer specific death and death due to other causes except for lymph nodes retrieved (log-rank test, p = 0.38), grade of malignancy and type (p = 0.52) and protocol version (p = 0.58) for death due to other causes and degree of urbanicity (p = 0.69) and depression (p = 0.11) for death due to breast cancer (data not shown). The multivariate analysis showed that the younger women died more from their breast cancer whereas the HR due to other causes increased strongly with age. All tumour characteristics were associated with the HR of dying of breast cancer whereas only a high number of positive lymph nodes had an effect on death of other causes (HR, 1.74; 1.30–2.33). In regard to the socioeconomic variables, there was an association between education, occupation, disposable income and housing tenure and the HR of dying of other causes. In general, however, the risk estimates were in the same direction and magnitude as that of the overall survival analysis, but failing to reach statistical significance for breast cancer deaths. Living with a partner was as well as living with young children at home reduced the HR of death due to other causes whereas these associations could not be observed for breast cancer deaths. Comorbid disorders influenced both the HR of dying from breast cancer and from other causes, although the association was much stronger for death of all causes, with HR of 2.24 in women with comorbidity score 1 (95% CI, 1.94–2.59) and 5.77 with score 2 or more (95% CI, 4.77–6.97). Presence of depression or schizophrenia was only significantly associated with risk of dying from other causes (1.30, 95% CI, 1.04–1.63 and 2.13, 95% CI, 1.55–2.92, respectively).

Table III. Hazard Ratios with 95% Confidence Intervals for Breast Cancer Specific Death and for Death Due to Othercauses in a Cohort of 25,897 Women Diagnosed with Breast Cancer in Denmark, 1983–1999
 Dead breast cancer (N = 6,455)Dead other causes (N = 1,855)LR test p-value
Hazard ratio95% Confidence limitsp-valueHazard ratio95% Confidence limitsp-value
  • p-values from Wald's test. LR test: likelihood ratio test for each parameter group. All results were stratified by receptor status and tumour grade and type.

  • 1

    Household disposable income in Danish crowns (DKK) after taxation and interest adjusted for number of persons in household.

  • 2

    Children aged 0–17 years.

  • 3

    Presence of these disorders as defined in the Charlson Index was defined as an in- or out-patient contact with one of the below diagnosis from 1978 to half a year prior to the diagnosis of breast cancer. Score 1: Myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcer disease, mild liver disease, diabetes type1, diabetes type2. Score 2: Hemiplegia, moderate to severe renal disease, diabetes with end organ damage type1 or type 2. Score 3: Moderate to severe liver disease. Score 6: AIDS; all cancer related diagnoses were excluded since first primary cancers only are in DBCG.

Age  <0.01  <0.0001<0.0001
 –44 years1  1   
 45–49 years0.850.77–0.93 1.220.92–1.62  
 50–54 years0.970.88–1.08 1.761.33–2.34  
 55–59 years1.000.90–1.10 2.501.89–3.30  
 60–64 years0.980.87–1.09 2.962.23–3.92  
 65–69 years1.010.89–1.14 4.063.04–5.43  
Tumour size  <0.0001  0.06<0.0001
 1–10 mm0.520.47–0.58 0.850.73–1.00  
 11–15 mm0.670.62–0.73 1.000.87–1.15  
 16–20 mm0.800.75–0.86 1.090.95–1.24  
 21–30 mm1  1   
 31–50 mm1.231.15–1.32 1.050.89–1.24  
 51– mm1.521.39–1.67 1.150.88–1.50  
Lymph nodes retrieved  <0.0001  0.27<0.0001
 0–31.241.13–1.36 0.890.75–1.05  
 4–91  1   
 10–140.760.71–0.81 0.950.84–1.07  
 15–0.590.54–0.64 0.890.75–1.04  
Positive lymph nodes  <0.0001  <0.001<0.0001
 01  1   
 1–31.991.87–2.12 0.940.84–1.05  
 4–94.283.99–4.60 1.140.96–1.36  
 10–147.706.93–8.55 1.741.30–2.33  
Protocol version  <0.0001  0.470.10
 821.191.12–1.27 1.050.92–1.20  
 89/991  1   
Highest attained education  0.32  0.110.55
 Basic school/high school1  1   
 Vocational training0.970.91–1.03 0.930.82–1.05  
 Higher education0.930.85–1.01 0.800.66–0.96  
 Unknown1.000.90–1.11 0.990.85–1.15  
Occupation  0.05  <0.0001<0.01
 Higher salaried employee1.010.92–1.12 0.720.60–0.88  
 Lower salaried employee0.890.81–0.98 0.650.54–0.79  
 Skilled worker0.890.75–1.04 0.570.36–0.91  
 Manual worker0.960.87–1.06 0.640.53–0.78  
 Unemployed/housewife0.970.89–1.06 0.740.64–0.87  
 Pensioner1  1   
Disposable income1  0.06  0.010.06
 <100,0001  1   
 100,000–129,9991.010.95–1.08 0.840.75–0.95  
 130,000–164,9990.940.87–1.01 0.850.74–0.98  
 165,000–0.920.84–1.00 0.790.66–0.93  
Cohabiting  0.95  <0.001<0.01
 Single1  1   
 Living with partner1.000.94–1.07 0.830.75–0.92  
Children living at home2  0.15  0.010.01
 01  1   
 10.930.85–1.03 0.820.63–1.06  
 2–51.030.92–1.16 0.550.37–0.82  
Housing tenure  0.47  0.090.60
 Owner occupied1  1   
 Rental1.030.96– 1.10 1.110.98–1.25  
Size of dwelling  0.24  0.240.26
 0–99 m21  1   
 100–124 m20.970.90–1.04 1.010.88–1.15  
 125–149 m20.960.88–1.04 0.870.74–1.03  
 150– m20.920.85–1.00 0.900.77–1.06  
Urbanity  0.54  0.010.01
 Capital area1.020.94–1.11 1.271.11–1.45  
 Surburbian area1.030.95–1.11 0.950.82–1.10  
 Provincial cities1  1   
 Rural areas0.970.92–1.03 0.960.85–1.08  
Charlson Index3  <0.0001  <0.0001<0.0001
 None1  1   
 11.100.98–1.23 2.241.94–2.59  
 2+1.521.25–1.85 5.774.77–6.97  
Depression  0.15  0.020.28
 No1  1   
 Yes1.120.96–1.31 1.301.04–1.63  
Schizophrenia  0.15  <0.0001<0.01
 No1  1   
 Yes1.200.94–1.52 2.131.55–2.92  

Figure 1 provides a graphical illustration of overall survival by income and prognostic factors, i.e., high- versus low-risk breast cancers and presence versus absence of comorbid disorders. Among the poor, having comorbid conditions influence survival more than among the rich, regardless of having high-risk or low risk breast cancer. 10-year survival is 15% lower among poor women with low-risk breast cancers and comorbid conditions (∼65%) compared to rich women with similar breast cancer prognosis and comorbid conditions (∼80%). Further, the survival curve for poor low-risk breast cancer patients with comorbid conditions deviates from the convex shape of the other low-risk breast cancer survival curves and approach the concave shape of the high-risk breast cancer survival curves.

Figure 1.

Adjusted overall survival after breast cancer by income and comorbidity status in Danish women diagnosed with breast cancer, 1983–1999. Low risk: tumour size less than 20 mm, no positive lymph nodes, Grade 1, ductal or unknown and estrogen receptor positive or unknown; high risk: all other tumours. Rich: highest quartile of disposable income; poor: lowest quartile of disposable income; comorbidity: conditions included in Charlson Index, affective disorders or schizophrenia; survival curves adjusted for age, number of retrieved lymph nodes, protocol version, education, occupation, cohabitation status, number of children living at home, housing tenure and degree of urbanicity.

The 4,886 women who did not go into protocol and consequently were excluded from analysis had a worse prognostic profile, i.e., with more tumours larger than 20 mm (55.3% vs. 41.4%), more positive lymph nodes (34.5% vs. 15.7% had 4 or more positive lymph nodes) as well as more grade III or unknown tumours (35.5% vs. 19.9%). Further, there were, i.e., more women with the lowest disposable income (31.4% vs. 25.9%) and more living alone (34.8% vs. 27.6%) among the women who did not undergo per protocol treatment. Finally, this group had a higher comorbidity burden, with more women scoring higher on the Charlson Index (4.8% vs. 2.9% with score 2+) and more with depression (3.5% vs. 2.8%) or schizophrenia (2.2% vs. 1.0%). There was, however, no difference in age distribution between those included or excluded in the final sample. The overall survival was better for the group of women included in the analysis compared to that of the 4,886 women who did not undergo per protocol treatment as well as the 680 women who were excluded because of unknown socioeconomic or prognostic variables (p < 0.0001).

Discussion

Our study demonstrates a significant socioeconomic difference in overall survival among almost 26,000 women treated for breast cancer in Denmark between 1983 and 1999. When important clinical prognostic factors were accounted for, a better overall survival remained in women with higher education, higher income and an active labour market affiliation, as well as measures of material living standards, such as house ownership and size of housing. However, these socioeconomic factors had a greater influence on non-breast cancer survival than on breast cancer-specific survival.

Analysis of overall survival may seem superfluous when it was possible to analyse breast cancer-specific and competing-causes survival. However, the misclassification that can occur between the latter 2 categories does not occur with all-cause death. Also, all-cause survival to some extent captures combined effects,24, 25 because it is likely that in some cases comorbidity and breast cancer are not mutually exclusive but in combination contribute to shortened survival. Thus, survival to all-cause death serves as a useful outcome for summarising the overall impact of socioeconomic position on the cohort of breast cancer patients.

In Denmark it seems that socioeconomic factors had a stronger influence on non-breast cancer survival. Although the breast cancer specific survival differences were of similar size to those observed in the overall survival analysis most failed to reach statistical significance. We used the Charlson Index to account for comorbid disorders with a known influence on survival in breast cancer patients.22 We operationalised the Charlson Index through the use of in- or out-patient hospital contacts and we had the possibility to include a time period of at least 4.5 years before the time of surgery for breast cancer for identifying the presence of comorbid conditions in all women prior to their primary surgery for breast cancer. The disorders included in the Charlson Index are generally of such serious nature that if present, it would have led to hospital contact at some point in time. However, there might be clinically important differences in tumour progression because of the individual characteristics of the breast cancer patients, such as nutritional status, physical activity, weight, smoking or alcohol use, some of which may be influenced by socioeconomic position and might lead to less severe comorbidities not quantified in the Charlson Index. However, comparing the survival between the socioeconomic groups on an absolute scale, being poor and also having severe comorbid conditions entails a long term prognosis as bad as that of having a high-risk breast cancer, thus underpinning a necessity for aggressive management of comorbid disorders in breast cancer care especially among the more disadvantaged women.

Although earlier studies have reported on social inequality in survival after breast cancer in societies with uniform tax-funded health care systems designed to provide equitable access to care3, 6, 8, 10 this has never before been investigated in a Danish setting. These studies have utilised area-based3, 8 and/or individual measures of social position.6, 10 In a study of 10,865 women with breast cancer in the UK diagnosed between 1982 and 1993 socioeconomic position was measured both as individual social class and by area-based deprivation groups.10 The authors observed a social inequality in both overall survival and breast cancer specific survival, which was only partly explained by stage and morphological type. Another study from the UK estimated that estrogen receptor status and treatment factors accounted for some 20% of the difference in survival between patients of high and low socioeconomic position (measured by area of residence).8 In line with our findings, however, the survival effect for death from any cause was stronger than for breast cancer as underlying cause of death.10 Further, overall survival was significantly related to both education, annual income and occupation in a recent population based study of 15,021 breast cancer patients in Stockholm, Sweden,12 a difference, however, mostly attributable to differences in non-breast cancer mortality. In that study they used stage, since histopathological data were unavailable among patients with inoperable disease. Conversely, another Swedish study observed a 37% higher risk of death due to breast cancer in women of low socioeconomic status (HR high vs. low 0.73; 95% CI, 0.54–0.99) when adjusting for tumour size and lymph node involvement, age, cohabitation status and parity.11 None of these previous studies have included measures of comorbid conditions, which had a strong effect on especially non-breast cancer survival, but also on breast cancer-specific mortality in the present data.

We used the date of primary breast cancer surgery as the entry point in our study as this information is uniformly recorded and available for all women registered in DBCG. Information on date of diagnosis, biopsies or whether the woman had participated in mammography screening was not available to us but although there might be a differential delay between diagnosis and time of surgery not accounted for in the present study still the date of surgery is the date where the tumour characteristics are clarified, the breast cancer is staged and the subsequent treatment is determined based on the pathological characteristics. A strength of our study is the use of histopathological stage, a better indicator of extent of disease compared to clinical stage only. As a result we excluded all women who did not undergo per protocol treatment for their breast cancer (15%), because of the fact that their clinical prognostic variables were not available in all cases and that these patients did not undergo the standardised treatment otherwise applied. The available descriptive information does strongly indicate that we hereby excluded a group of more advanced cases. Further, our exploratory analysis of this group indicates a stronger social gradient on almost all factors investigated, as well as a worse overall survival. Despite of the exclusion of this less advantaged group, social inequality still remain in survival after breast cancer surgery in Denmark.

In conclusion, we found social inequality in survival after breast cancer surgery in the uniform tax-funded Danish health care system even when accounting for the existence of comorbid disorders. The social inequality was strongest, however, for causes of death other than breast cancer. The existence of comorbid conditions had a strong independent effect on both breast cancer-specific and overall survival, and the striking differences in long term survival of breast cancer patients with lowest and highest income probably suggest poorer access to, management of and/or compliance with treatment of comorbid conditions among poorer women with breast cancer. No matter the reasons, life is shorter among poor women treated for breast cancer in Denmark.

Ancillary