Background:

Socioeconomic status (SES) is an important factor in cancer survival; however, results are heterogeneous and linked to characteristics of the study population and health care system. This population-based cohort study evaluates the association between individual-level socioeconomic and demographic factors and cancer survival for the first time in Belgium.

Methods:

From the Belgian Cancer Registry, we identified 109,591 patients diagnosed between 2006 and 2013 with one of eight common cancer types. Information on treatment, socioeconomic parameters, and vital status were retrieved from multiple data sources and linked using a unique personal identification number. The outcome was 5-year observed survival. Associations between survival and socioeconomic and demographic factors were assessed using multivariable Cox proportional–hazard regression models.

Results:

Lower income, unemployment, and living alone were all associated with worse cancer survival. These associations were most pronounced for certain lifestyle-related cancer types (e.g., head and neck cancers) and those with good to moderate prognosis (e.g., colorectal and female breast cancer).

Conclusions:

These results indicate that, despite a comprehensive and nationwide health insurance program in which equity in rights and access to health care are pursued, SES is associated with disparities in cancer survival in Belgium.

Impact:

This population-based study with individual-level socioeconomic information of more than 100,000 patients with cancer identifies patient groups that may be at highest risk for socioeconomic disparities in cancer survival. Reasons behind the observed disparities are multiple and complex and should be further examined. Health policy interventions should consider the observed deprivation gap to plan targeted actions.

Overall, socioeconomic status (SES) is associated with inequalities in healthcare usage, morbidity, and mortality. Cancer is no exception. Numerous studies have shown an association between socioeconomic (SE) and sociodemographic (SD) factors and cancer risk, cancer mortality, and cancer survival. However, these associations are not homogeneous across regions and ethnical groups, nor across cancer types (1). The effects on risk and prognosis are heterogeneous and not yet fully understood (2, 3).

As for cancer survival, a poorer prognosis for lower SE groups has consistently been observed. Despite a global improvement in cancer survival, mainly through advancements in diagnostics and treatments, the gap between deprived and affluent patients has increased over time (4–6).

In Belgium so far, most studies focused on the impact of SE and SD factors on cancer mortality (7–9). Although crucial for public health and healthcare policies, mortality rates depend on both the risk of and survival from a disease. In addition, mortality rates are subject to the accuracy of death certification, which can be challenging, especially in older patients representing a large proportion of cancer cases (10). In this study, we will therefore focus on the association between SES and survival among patients with cancer.

SES is a complex, multifaceted concept consisting of several dimensions (11). So far, numerous studies concerning cancer-related outcomes operationalized SES either by a single variable (often income as a proxy for material wealth, educational attainment reflecting chances in early life, or unemployment rate as a proxy for area-deprivation; ref. 12), or by a composite deprivation index to group patients into SE groups (13). Although strongly related, different SE and SD factors tap into different causal pathways on how SES influences the onset and course of diseases, highlighting the importance of considering multiple SE and SD indicators separately.

The objective of this population-based cohort study is to assess, for the first time in Belgium, the association between cancer survival and different dimensions of SES, more specifically, household income, employment, and marital status. Second, this study also aims to explore if observed associations can be (partly) explained by specific clinical characteristics such as tumor characteristics, stage at diagnosis, comorbid conditions, and detailed primary treatment information.

Patient selection and data sources

Data were retrieved from three databases and linked through a unique personal identification number assigned to all residents of Belgium (1:1 linkage): the Belgian Cancer Registry, the Crossroads Bank for Social Security (CBSS), and the reimbursement databases of the health insurance companies gathered by the Intermutualistic Agency (IMA).

Patients diagnosed with cancer between 2006 and 2013, ages 25 years and older at time of diagnosis and registered in the IMA database were identified using the Belgian Cancer Registry, a nationwide register that holds information on all primary invasive cancer diagnoses since 2004 (14). Using the 10th International Classification of Diseases (ICD-10), eight cancer types were included in this study: Colon (C18–19), rectal (C20), female breast (C50), lung (C34), ovarian (C56), head and neck (C00-C14,C30-C32), pancreatic (C25) and stomach (C16.1-C16.9) cancer. These sites were selected because they represent high-burden cancers for which treatment data were reliably available. For each of these cancer types, a random sample (2/3 if less than 10,000 cases present in the registry, otherwise 1/2) was drawn from the total patient population and included in the study. Stratified sampling was performed by age quartiles (for each cancer type) and sex (if applicable). Single random sampling was applied using the PROC SURVEYSELECT procedure in SAS 9.3 (SAS Institute Inc.). From the Belgian Cancer Registry, we extracted information on patient (age and sex) and tumor (stage and subtype) characteristics. For head and neck cancer and stomach cancer, subtypes were defined on the basis of the primary tumor topography. Histological subtypes were considered for female breast cancer, as well as lung, ovarian, and pancreatic cancer. Topography and morphology codes are available in Supplementary Table S1. A combined TNM stage, where the pathological stage prevails over the clinical, except for cases diagnosed with clinical stage IV (i.e., category cM1 for most cases), was considered (14).

From the reimbursement databases of the health insurance companies, covering 97% of the Belgian population, we extracted detailed information on primary tumor-directed treatment schemes [i.e., charged nomenclature codes for surgery and radiotherapy, and Anatomical Therapeutic Chemical (ATC; ref. 15) codes for chemotherapy], and data on comorbidities (chronic respiratory and cardiovascular comorbidities and diabetes mellitus) treated before cancer diagnosis (16). As IMA data do not directly refer to a specific diagnosis, timeframes around the cancer diagnosis were used to assess primary cancer treatment: surgery was included if undergone within one month before until 6 months after diagnosis, chemo- and radiotherapy were included if administered within one month before diagnosis until 6 months after diagnosis or surgery (if applicable; ref. 17).

Vital status information was extracted from CBSS. SE and SD data were extracted from the same database, including household income, employment, and marital status. The CBSS integrates socioeconomic data from several Belgian social security institutions and public entities, for research and administrative purposes. The majority of data are registered on a quarterly basis for all inhabitants of Belgium (18).

To avoid the impact of cancer on the patient's SE status, the retrieved data referred to the patient's situation in the year before diagnosis (trimester of cancer diagnosis excluded). Income was available at the household level by increments of 5,000€. Weights were calculated for each household based on its composition: 1.0 to the first adult; 0.5 to the second and each subsequent person ages 12 years and over; 0.3 to each child under 12 years of age. The household income was then divided by the sum of the weightings (equivalized) to yield a representative income, according to the OECD-modified scale (19). Low-, middle-, and high-income households were defined using cutoffs at p25 and p50 and considering the active (patients younger than 65 years and non-retired) and non-active (ages 65+ years or retired) cancer population separately (20). Employment status among (potentially) active patients was classified as employed, unemployed and job-seeking, and unemployed and not job-seeking. The latter category consisted of people who could not work due to incapacity/disability or due to other reasons (8).

De jure marital status was divided into the following categories: Living together, separated, single and widowed, and was available for the period 2009–2013 only.

Follow-up and outcome analyses

The start of follow-up for each patient was the date of cancer diagnosis. Vital status information was available until July 1, 2016. The outcome was observed survival during the 5-years after diagnosis. Patients whose observation duration was shorter than the maximum time for which survival probability was calculated (e.g., some patients had an early diagnosis but were lost to follow-up because they were no longer found at CBSS; n = 175) were censored at the date of last information on vital status (21).

Depending on the SE or SD factor, analyses were performed considering the whole-study cohort (marital status), the (potentially) active population (household income and employment status) or the non-active population (household income). Because of the different nature of household income in the active and non-active population (i.e., salary and pension, respectively), analyses were performed separately.

Following univariable analyses, the relationships between survival and household income, employment, and marital status were assessed using multivariable Cox proportional-hazards regression models, adjusting for age (continuous), gender (if applicable, binary), tumor subtype (except for colon and rectal cancer, categorical), stage at diagnosis (categorical), chronic cardiovascular disease (binary), chronic respiratory disease (binary), diabetes mellitus (binary), and primary treatment scheme (at baseline, categorical). Cases with unknown stage were included in the study and considered as a distinct category. Furthermore, cases with non-specified or less-common subtypes were combined into a separate category.

Results were presented as hazard ratios (HR) with 95% confidence intervals (95% CI). To fulfil proportional hazard assumptions, interactions with time-points and two-way interactions for variables with significant effects in the model were considered under the threshold of alpha = 0.01 using a backward interaction selection procedure (22).

Sensitivity analyses

First, the association between 5-year observed survival and individual, household and neighborhood-level income was assessed in uni- and multivariable Cox proportional–hazards regression models. Second, to assess the potential mediator effect of stage at diagnosis and comorbid conditions, adjusted survival estimates were calculated without considering those adjustment factors. Likewise, the effect of treatment on survival estimates was assessed by removing this adjustment factor from the multivariable regression models.

All analyses were conducted in SAS 9.3 (SAS Institute Inc.).

Descriptive analysis

Patient and tumor characteristics by cancer type are shown in Table 1. A total of 109,591 patients (female breast, n = 33,644; colon, n = 20,149; rectum, n = 8,174; head and neck, n = 7,304; lung, n = 27,668; ovary, n = 3,394; pancreas, n = 4,737; stomach, n = 4,521) with a median follow-up time of 3.0 years (interquartile range, 0.9–5.0 years) were included. Of these, 60,206 patients (54.9%) were women and 38,168 of all patients (34.8%) were younger than 65 years old and non-retired at time of diagnosis. The percentage of patients treated for chronic respiratory and cardiovascular comorbidities or diabetes mellitus before the cancer diagnosis were 7.8%, 49.2%, and 13.0%, respectively. Furthermore, 59,715 patients (54.5%) died within 5 years after diagnosis. This percentage varied largely by cancer type: From 21.8% for breast cancer to 90.3% for patients with pancreas cancer.

Table 1.

Patient and tumor characteristics by cancer type, Belgium 2006–2013.

Breast cancerLung cancerColon cancerRectal cancerHead and neck cancersPancreatic cancerStomach cancerOvarian cancer
(n = 33,644)(n = 27,668)(n = 20,149)(n = 8,174)(n = 7,304)(n = 4,737)(n = 4,521)(n = 3,394)
Patient characteristics  N (%)  N (%) N (%) N (%)  N (%)  N (%)  N (%)  N (%) 
Gender 
 Male  —  20,907 (75.6) 11,567 (57.4) 5,283 (64.6)  5,871 (80.4)  2,648 (55.9)  3,109 (68.8)  — 
 Female  33,644 (100.0)  6,761 (24.4) 8,582 (42.6) 2,891 (35.4)  1,433 (19.6)  2,089 (44.1)  1,412 (31.2)  3,394 (100.0) 
Age and professional status at diagnosis 
 <65 years and nonretired (active)  17,243 (51.3)  7,714 (27.9) 3,989 (19.8) 2,208 (27.0)  3,565 (48.8)  1,152 (34.3)  1,017 (22.5)  1,280 (37.7) 
 65+ years or retired (nonactive)  16,401 (48.8)  19,954 (72.1) 16,160 (80.2) 5,966 (73.0)  3,739 (51.2)  3,585 (75.7)  3,504 (77.5)  2,114 (62.3) 
Mean age (standard deviation)  61.7 (14.3)  68.7 (10.7) 72.1 (12.0) 69.5 (12.1)  63.1 (10.8)  70.0 (11.7)  71.5 (12.8)  66.1 (14.1) 
Comorbiditya 
 Chronic respiratory  1,054 (3.1)  4,635 (16.8) 1,137 (5.6) 470 (5.8)  631 (8.6)  257 (5.4)  277 (6.1)  122 (3.6) 
 Chronic cardiovascular  12,787 (38.0)  15,375 (55.6) 11,473 (56.9) 4,183 (51.2)  3,229 (44.2)  2,765 (58.4)  2,552 (56.5)  1,511 (44.5) 
 Diabetes mellitus  2,982 (8.9)  3,799 (13.3) 3,189 (15.8) 1,162 (14.2)  727 (10.0)  1,277 (27.0)  757 (16.7)  351 (10.3) 
Tumor characteristics  N (%)  N (%N (%) N (%)  N (%)  N (%)  N (%)  N (%) 
Subtypeb 
 IDC 25,410 (75.5) ADC 10,717 (38.7) — — HP 904 (12.4) ADC 3,689 (77.9) GOJ 1,353 (29.9) ADC 2,638 (77.7) 
 ILC 4,426 (13.2) LCLC 1,169 (4.2) — — LX 2,285 (31.3) PNET 332 (7.0) GB 1,338 (26.6) Other 756 (22.3) 
 Other 3,808 (11.3) SCLC 4,132 (14.9) — — OC 2,055 (28.1) Other 716 (15.1) Other 1,830 (40.5)   
   SCC 7,602 (27.5)   OP 2,060 (28.2)       
   Other 4,048 (14.6)           
Combined stagec 
 I  13,782 (41.0)  4,309 (15.6) 3,151 (15.6) 2,046 (25.0)  1,322 (18.1)  385 (8.1)  868 (19.2)  610 (18.0) 
 II  11,791 (35.1)  1,943 (7.0) 6,115 (30.4) 1,797 (22.0)  870 (11.9)  1,065 (22.5)  669 (14.8)  196 (5.8) 
 III  4,092 (12.2)  5,421 (19.6) 5,167 (25.6) 2,136 (26.1)  1,018 (13.9)  359 (7.6)  772 (17.1)  961 (28.3) 
 IV  1,878 (5.6)  10,318 (37.3) 3,800 (18.9) 1,329 (16.3)  2,971 (40.7)  1,786 (37.7)  1,157 (25.6)  660 (19.5) 
 X (unknown)  2,101 (6.2)  5,677 (20.5) 1,916 (9.5) 866 (10.6)  1,123 (15.4)  1,142 (24.1)  1,055 (23.3)  967 (28.5) 
Primary treatment scheme 
 Surgery only  902 (2.7)  5,017 (18.1) 10,014 (49.7) 2,338 (28.6)  1,340 (18.4)  514 (10.9)  1,240 (27.4)  512 (15.1) 
 Surgery + (neo-)adjuvant  27,800 (82.6)  6,009 (21.7) 7,539 (37.4) 4,406 (53.9)  1,465 (20.1)  674 (14.2)  884 (19.6)  1,791 (52.8) 
 Chemo ± RT ± HT  4,110 (12.2)  10,216 (36.9) 1,335 (6.6) 768 (9.4)  4,016 (55.0)  2,140 (45.2)  1,046 (23.1)  737 (21.7) 
 No tumor-directed treatment  832 (2.5)  6,426 (23.2) 1,261 (6.3) 662 (8.1)  483 (6.6)  1,409 (29.7)  1,351 (29.9)  354 (10.4) 
Breast cancerLung cancerColon cancerRectal cancerHead and neck cancersPancreatic cancerStomach cancerOvarian cancer
(n = 33,644)(n = 27,668)(n = 20,149)(n = 8,174)(n = 7,304)(n = 4,737)(n = 4,521)(n = 3,394)
Patient characteristics  N (%)  N (%) N (%) N (%)  N (%)  N (%)  N (%)  N (%) 
Gender 
 Male  —  20,907 (75.6) 11,567 (57.4) 5,283 (64.6)  5,871 (80.4)  2,648 (55.9)  3,109 (68.8)  — 
 Female  33,644 (100.0)  6,761 (24.4) 8,582 (42.6) 2,891 (35.4)  1,433 (19.6)  2,089 (44.1)  1,412 (31.2)  3,394 (100.0) 
Age and professional status at diagnosis 
 <65 years and nonretired (active)  17,243 (51.3)  7,714 (27.9) 3,989 (19.8) 2,208 (27.0)  3,565 (48.8)  1,152 (34.3)  1,017 (22.5)  1,280 (37.7) 
 65+ years or retired (nonactive)  16,401 (48.8)  19,954 (72.1) 16,160 (80.2) 5,966 (73.0)  3,739 (51.2)  3,585 (75.7)  3,504 (77.5)  2,114 (62.3) 
Mean age (standard deviation)  61.7 (14.3)  68.7 (10.7) 72.1 (12.0) 69.5 (12.1)  63.1 (10.8)  70.0 (11.7)  71.5 (12.8)  66.1 (14.1) 
Comorbiditya 
 Chronic respiratory  1,054 (3.1)  4,635 (16.8) 1,137 (5.6) 470 (5.8)  631 (8.6)  257 (5.4)  277 (6.1)  122 (3.6) 
 Chronic cardiovascular  12,787 (38.0)  15,375 (55.6) 11,473 (56.9) 4,183 (51.2)  3,229 (44.2)  2,765 (58.4)  2,552 (56.5)  1,511 (44.5) 
 Diabetes mellitus  2,982 (8.9)  3,799 (13.3) 3,189 (15.8) 1,162 (14.2)  727 (10.0)  1,277 (27.0)  757 (16.7)  351 (10.3) 
Tumor characteristics  N (%)  N (%N (%) N (%)  N (%)  N (%)  N (%)  N (%) 
Subtypeb 
 IDC 25,410 (75.5) ADC 10,717 (38.7) — — HP 904 (12.4) ADC 3,689 (77.9) GOJ 1,353 (29.9) ADC 2,638 (77.7) 
 ILC 4,426 (13.2) LCLC 1,169 (4.2) — — LX 2,285 (31.3) PNET 332 (7.0) GB 1,338 (26.6) Other 756 (22.3) 
 Other 3,808 (11.3) SCLC 4,132 (14.9) — — OC 2,055 (28.1) Other 716 (15.1) Other 1,830 (40.5)   
   SCC 7,602 (27.5)   OP 2,060 (28.2)       
   Other 4,048 (14.6)           
Combined stagec 
 I  13,782 (41.0)  4,309 (15.6) 3,151 (15.6) 2,046 (25.0)  1,322 (18.1)  385 (8.1)  868 (19.2)  610 (18.0) 
 II  11,791 (35.1)  1,943 (7.0) 6,115 (30.4) 1,797 (22.0)  870 (11.9)  1,065 (22.5)  669 (14.8)  196 (5.8) 
 III  4,092 (12.2)  5,421 (19.6) 5,167 (25.6) 2,136 (26.1)  1,018 (13.9)  359 (7.6)  772 (17.1)  961 (28.3) 
 IV  1,878 (5.6)  10,318 (37.3) 3,800 (18.9) 1,329 (16.3)  2,971 (40.7)  1,786 (37.7)  1,157 (25.6)  660 (19.5) 
 X (unknown)  2,101 (6.2)  5,677 (20.5) 1,916 (9.5) 866 (10.6)  1,123 (15.4)  1,142 (24.1)  1,055 (23.3)  967 (28.5) 
Primary treatment scheme 
 Surgery only  902 (2.7)  5,017 (18.1) 10,014 (49.7) 2,338 (28.6)  1,340 (18.4)  514 (10.9)  1,240 (27.4)  512 (15.1) 
 Surgery + (neo-)adjuvant  27,800 (82.6)  6,009 (21.7) 7,539 (37.4) 4,406 (53.9)  1,465 (20.1)  674 (14.2)  884 (19.6)  1,791 (52.8) 
 Chemo ± RT ± HT  4,110 (12.2)  10,216 (36.9) 1,335 (6.6) 768 (9.4)  4,016 (55.0)  2,140 (45.2)  1,046 (23.1)  737 (21.7) 
 No tumor-directed treatment  832 (2.5)  6,426 (23.2) 1,261 (6.3) 662 (8.1)  483 (6.6)  1,409 (29.7)  1,351 (29.9)  354 (10.4) 

Abbreviations: ADC, adenocarcinoma; GB, gastric body; GOJ, gastro–esophageal junction; HP, hypopharynx; HT, hormonal therapy (only considered for breast cancer); IDC, invasive ductal carcinoma; ILC, invasive lobular carcinoma; LCLC, large-cell lung cancer; LX, larynx; OC, oral cavity; OP, oropharynx; PNET, primitive neuro-ectodermal tumor; RT, radiotherapy; SCC, squamous cell carcinoma; SCLC, small-cell lung cancer.

aComorbidity data were extracted from the administrative databases of the health insurance companies and referred to patients treated before cancer diagnosis.

bSubtypes are based on morphology codes, except for head and neck cancers and stomach cancer for which topography codes were considered.

cCombined stage: The pathological stage prevails over the clinical, except for cases diagnosed with clinical stage IV.

Male predominance was most pronounced in head and neck (80.4%) and lung cancer (75.6%). Mean age at diagnosis differed strongly by cancer type, with female breast cancer (51.3%) and head and neck cancers (48.8%) being most often diagnosed in the (potentially) active population, whereas colon cancer affected mostly the older, non-active population (80.2%). The percentage of patients with chronic respiratory comorbidity was highest in the lung cancer cohort (16.8%), whereas cardiovascular comorbidities were present in almost half of the patients across the different cancer types, except breast cancer. Diabetes was largely associated with age, resulting in higher proportions for patients with colon (15.8%), pancreatic (27.0%), and stomach cancer (16.7%). Stage distribution varied across cancer types; whereas the majority of breast cancer cases were diagnosed at an early stage, more than one third of lung, pancreas, and head and neck cancers were diagnosed at stage IV of the disease. Missing information on stage at diagnosis ranged from 6.2% in patients with breast cancer to 28.5% in patients with ovarian cancer.

About 12% of patients did not receive primary tumor-directed treatment. This percentage differed largely by cancer type, being lowest for patients with breast cancer (2.5%) and highest for patients with pancreatic (29.7%) and stomach cancer (29.9%). More than 80% of patients with breast cancer underwent surgery in combination with neo- and/or adjuvant therapy. Surgery alone was performed in almost half (49.7%) of the patients diagnosed with colon cancer and 28.6% of patients with rectal cancer.

Table 2 shows SE and SD factors according to patient characteristics. Patients with low and/or middle income were more strongly represented in the older, non-active population. The proportion of patients with high income was higher for women (38.9%) compared with men (31.2%), mainly because of the high proportion of female patients with breast cancer with high income (37.9%). Furthermore, the percentage of patients affected by comorbidities was lower for high-income patients.

Table 2.

Patient characteristics by socioeconomic and sociodemographic factors, Belgium 2006–2013.

Household incomeEmployment statusMarital statusa
LowMiddleHighWorkingJob-seekingDisabled/otherRetiredCohabitatingSeparatedSingleWidowed
Patient characteristics N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) 
Overall 35,681 (32.6) 35,065 (32.0) 38,845 (35.4) — — — — 39,656 (56.2) 10,147 (14.4) 6,037 (8.6) 14,695 (20.8) 
Age and status at diagnosis 
 <65 years and nonretired (active) 9,959 (26.1) 11,234 (29.4) 16,975 (44.5) 25,604 (67.1) 3,252 (8.5) 9,312 (24.4) — 14,424 (59.0) 5,694 (23.3) 3,678 (15.1) 640 (2.6) 
 65+ years or retired (nonactive) 25,722 (36.0) 23,831 (33.4) 21,870 (30.6) — — — 71,423 (100.0) 25,232 (54.7) 4,453 (9.7) 2,359 (5.1) 14,055 (30.5) 
Mean age (standard deviation) 69.5 (12.6) 67.8 (13.0) 63.8 (13.1) 51.6 (7.6) 52.4 (6.7) 55.6 (7.5) 74.6 (8.1) 65.9 (12.1) 61.6 (10.7) 58.7 (15.2) 77.9 (9.4) 
Sex 
 Male 17,622 (35.7) 16,357 (33.1) 15,406 (31.2) 8,419 (63.5) 1,190 (9.0) 3,652 (27.5) 36,124 (50.6) 21,666 (68.6) 4,009 (12.7) 2,486 (7.9) 3,432 (10.9) 
 Female 18,059 (30.0) 18,708 (31.1) 23,439 (38.9) 17,185 (69.0) 2,062 (8.3) 5,660 (22.7) 35,299 (49.4) 17,990 (46.2) 6,138 (15.8) 3,551 (9.1) 11,263 (28.9) 
Comorbidityb 
 Chronic respiratory 3,100 (8.7) 3,108 (8.9) 2,375 (6.1) 847 (3.3) 133 (4.1) 858 (9.2) 6,745 (9.4) 3,380 (8.5) 845 (8.3) 356 (5.9) 1,152 (7.8) 
 Chronic cardiovascular 18,445 (51.7) 18,294 (52.2) 17,136 (44.1) 5,608 (21.9) 683 (21.0) 3,091 (33.2) 44,493 (62.3) 20,196 (50.9) 4,144 (40.8) 2,132 (35.3) 10,286 (70.0) 
 Diabetes mellitus 5,070 (14.2) 4,947 (14.1) 4,227 (10.9) 1,233 (4.8) 205 (6.3) 1,003 (10.8) 11,803 (16.5) 5,531 (13.9) 1,065 (10.5) 597 (9.9) 2,592 (17.6) 
Household incomeEmployment statusMarital statusa
LowMiddleHighWorkingJob-seekingDisabled/otherRetiredCohabitatingSeparatedSingleWidowed
Patient characteristics N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) 
Overall 35,681 (32.6) 35,065 (32.0) 38,845 (35.4) — — — — 39,656 (56.2) 10,147 (14.4) 6,037 (8.6) 14,695 (20.8) 
Age and status at diagnosis 
 <65 years and nonretired (active) 9,959 (26.1) 11,234 (29.4) 16,975 (44.5) 25,604 (67.1) 3,252 (8.5) 9,312 (24.4) — 14,424 (59.0) 5,694 (23.3) 3,678 (15.1) 640 (2.6) 
 65+ years or retired (nonactive) 25,722 (36.0) 23,831 (33.4) 21,870 (30.6) — — — 71,423 (100.0) 25,232 (54.7) 4,453 (9.7) 2,359 (5.1) 14,055 (30.5) 
Mean age (standard deviation) 69.5 (12.6) 67.8 (13.0) 63.8 (13.1) 51.6 (7.6) 52.4 (6.7) 55.6 (7.5) 74.6 (8.1) 65.9 (12.1) 61.6 (10.7) 58.7 (15.2) 77.9 (9.4) 
Sex 
 Male 17,622 (35.7) 16,357 (33.1) 15,406 (31.2) 8,419 (63.5) 1,190 (9.0) 3,652 (27.5) 36,124 (50.6) 21,666 (68.6) 4,009 (12.7) 2,486 (7.9) 3,432 (10.9) 
 Female 18,059 (30.0) 18,708 (31.1) 23,439 (38.9) 17,185 (69.0) 2,062 (8.3) 5,660 (22.7) 35,299 (49.4) 17,990 (46.2) 6,138 (15.8) 3,551 (9.1) 11,263 (28.9) 
Comorbidityb 
 Chronic respiratory 3,100 (8.7) 3,108 (8.9) 2,375 (6.1) 847 (3.3) 133 (4.1) 858 (9.2) 6,745 (9.4) 3,380 (8.5) 845 (8.3) 356 (5.9) 1,152 (7.8) 
 Chronic cardiovascular 18,445 (51.7) 18,294 (52.2) 17,136 (44.1) 5,608 (21.9) 683 (21.0) 3,091 (33.2) 44,493 (62.3) 20,196 (50.9) 4,144 (40.8) 2,132 (35.3) 10,286 (70.0) 
 Diabetes mellitus 5,070 (14.2) 4,947 (14.1) 4,227 (10.9) 1,233 (4.8) 205 (6.3) 1,003 (10.8) 11,803 (16.5) 5,531 (13.9) 1,065 (10.5) 597 (9.9) 2,592 (17.6) 

aMarital status was only available for incidence years 2009–2013. Overall percentages by income and marital status category are provided in the first row.

bComorbidity data were extracted from the administrative databases of the health insurance companies and referred to patients treated before cancer diagnosis.

The proportion of patients not working because of disability or other reasons before diagnosis was higher for men (27.5%) compared with women (22.7%). Within the non-active population, almost one third of patients (30.5%) were widowed at time of diagnosis and 38.4% of patients in the active population lived alone, being single or separated. Patients who lived with a partner at time of diagnosis were the most highly represented group in both the active (59.0%) and non-active population (54.7%).

The proportion of patients with comorbid conditions was higher in patients either living with a partner or widowed, which represents the majority of the non-active population. In other words, the presence of comorbidities was largely age-dependent. Indeed, the proportion of patients in the non-active population presenting a comorbidity of any type was higher compared with those in the active population. The percentage of females living alone, for any reason, was 53.8% compared with 31.4% in males.

SES and cancer survival

Adjusted HR estimates, including 95% CI, patients at risk and the number of deaths during follow-up are shown in Fig. 1. Unadjusted and adjusted HR estimates with 95% CI are provided in Supplementary Table S2.

Figure 1.

Adjusted hazard ratios (HR) by socioeconomic (SE) and sociodemographic (SD) factors and cancer type: female breast, colon, rectal, and head and neck cancers (A); lung, ovarian, pancreatic, and stomach cancers (B). HR estimates are provided with 95% confidence intervals (CI) and were adjusted for age, sex (if applicable), stage, subtype (if applicable), chronic respiratory and cardiovascular comorbidities, diabetes mellitus, and primary treatment scheme. Active: patients ages less than 65 years at diagnosis and nonretired; Nonactive: patients ages at least 65 years or retired. Marital status was only available for incidence years 2009–2013.

Figure 1.

Adjusted hazard ratios (HR) by socioeconomic (SE) and sociodemographic (SD) factors and cancer type: female breast, colon, rectal, and head and neck cancers (A); lung, ovarian, pancreatic, and stomach cancers (B). HR estimates are provided with 95% confidence intervals (CI) and were adjusted for age, sex (if applicable), stage, subtype (if applicable), chronic respiratory and cardiovascular comorbidities, diabetes mellitus, and primary treatment scheme. Active: patients ages less than 65 years at diagnosis and nonretired; Nonactive: patients ages at least 65 years or retired. Marital status was only available for incidence years 2009–2013.

Close modal

In the active population, survival was lower for patients with a low or middle household income compared with those with high income. The association was most pronounced for colon (HRlow, 1.41; 95% CI, 1.23–1.62; HRmiddle, 1.32; 95% CI, 1.16–1.50), breast (HRlow, 1.57; 95% CI, 1.38–1.77; HRmiddle, 1.17; 95% CI, 1.04–1.32) and head and neck cancers (HRlow, 1.85; 95% CI, 1.65–2.07; HRmiddle, 1.44; 95% CI, 1.27–1.63). For patients diagnosed with rectal, lung, pancreatic, and stomach cancer, survival was lower for patients with low income. No association between survival and income was found for ovarian cancer. Within the non-active population, low (HRlow, 1.08; 95% CI, 1.02–1.14) and middle-income categories (HRmiddle, 1.08; 95% CI, 1.02–1.13) showed a statistically significant lower survival for colon cancer, whereas for head and neck (HRlow, 1.24; 95% CI, 1.12–1.37) and pancreatic cancer (HRmiddle, 1.13; 95% CI, 1.04–1.23), only patients with low income or middle income seemed to have a poorer survival compared with high-income patients, respectively.

Not working was associated with worse survival for all cancer types, except ovarian cancer. Both job-seeking and not working because of disability/incapacity or other reasons were found to be statistically significantly associated with poorer survival compared with being employed, with the exception of stomach cancer (only job-seeking). The strongest association was observed in head and neck cancers (HRnot-working, 1.78; 95% CI, 1.61–1.97; HRjob-seeking, 1.73; 95% CI, 1.51–1.98), followed by rectal (HRnot-working, 1.65; 95% CI, 1.38–1.98; HRjob-seeking, 1.63; 95% CI, 1.23–2.18), colon (HRnot-working, 1.46; 95% CI, 1.28–1.67; HRjob-seeking, 1.64; 95% CI, 1.34–2.02) and female breast cancer (HRnot-working, 1.69; 95% CI, 1.51–1.89; HRjob-seeking, 1.50; 95% CI, 1.25–1.81). For lung (HRnot-working, 1.25; 95% CI, 1.18–1.32; HRjob-seeking, 1.15; 1.05–1.25) and pancreatic cancer (HRnot-working:1.31 [1.11–1.55], HRjob-seeking, 1.46; 95% CI, 1.16–1.83), this association was less pronounced, though statistically significant.

Marital status was also associated with cancer survival. Overall, living alone was associated with worse survival than being married or cohabitating. We found worse survival for single, separated, and widowed patients across all cancer types, except for stomach (only separated and widowed), ovarian (only singles), pancreatic (only singles and widowed), and rectal cancer (only separated and singles). The largest differences were observed for head and neck cancer (HRseparated, 1.35; 95% CI, 1.22–1.49; HRsingle,1.53; 95% CI, 1.34–1.75; HRwidowed, 1.51; 95% CI, 1.32–1.73), colon cancer (HRseparated, 1.28; 95% CI, 1.16–1.40; HRsingle, 1.37; 95% CI, 1.24–1.53; HRwidowed, 1.10; 95% CI, 1.03–1.18) and female patients with breast cancer (HRseparated, 1.23; 95% CI, 1.10–1.37; HRsingle, 1.35; 95% CI, 1.19–1.52; HRwidowed, 1.16; 95% CI, 1.07–1.27). Again, for patients with lung cancer (HRseparated, 1.15; 95% CI, 1.09–1.20; HRsingle, 1.16; 95% CI, 1.09–1.24; HRwidowed, 1.14; 95% CI, 1.09–1.19), the associations were less pronounced, though statistically significant.

Sensitivity analyses

The association between survival and income was most pronounced for individual income, followed by household, and neighborhood-level income, respectively. This association was higher in the active compared with the non-active population (Supplementary Table S3). Survival models without stage at diagnosis and comorbid conditions did not result in substantial changes in the survival estimates (Supplementary Table S4), whereas the association between SES and observed survival was less pronounced, though statistically significant, when removing primary treatment as adjustment factor. This was especially the case for employment and marital status (Supplementary Table S5).

Evidence of SES-associated disparities in cancer survival have been reported in many countries and for many cancer types (6, 7). The present study is the first to investigate the association between SE and SD factors and cancer survival at the individual-level in Belgium, including more than 100,000 patients diagnosed with 8 common cancer types between 2006 and 2013. Besides patient age, gender and tumor characteristics, survival estimates were adjusted for three major comorbidities (chronic respiratory disease, chronic cardiovascular diseases, and diabetes mellitus) and primary cancer treatment scheme.

Being a complex societal concept, SES consists of cultural, social, and economic dimensions. No single variable adequately captures SES (23, 24). Composite deprivation indexes integrate multiple dimensions of SES, but neither distil the contribution of each factor nor the variation in association (25). To date, most studies that focused on cancer survival considered SES at the area-level (26, 27). However, the size of the geographic unit, and consequently the population heterogeneity, in addition to the index construct, can have major effects on the observed inequalities. As a result, area-level SES may not correspond to the individual SES (28, 29). By investigating multiple SE and SD dimensions separately at the individual-level, this study aimed to provide novel insights into how these factors are associated individually with cancer survival in Belgium, without ecological bias (30), and correcting for differences in case-mix.

As a proxy for standard of living, we included a weighted income indicator accounting for differences in household composition and size, which yielded a representative income by household. The research results showed that income was significantly associated with survival for most cancer types, especially in the younger, active population. Financial difficulties could be associated with lifestyle habits that negatively affect health, and furthermore, could alter the optimal and timely use of healthcare services. In addition, although the healthcare system in Belgium is largely tax-funded (31), more affluent patients might disproportionally benefit from advancements in medicine through financial and social connections, and resulting knowledge (11).

Several studies have separately examined the associations of individual and neighborhood deprivation with cancer survival (or mortality). Consequently, it has been suggested that both are independently associated, and could also interact (10, 29, 32, 33). In a Supplementary Analysis, we observed that the association between survival and income was most pronounced for individual income, followed by household, and neighborhood-level income, respectively.

Not-working was negatively associated with cancer survival. Importantly, this association persisted after controlling for stage and comorbid conditions, limiting at least partially the potential health selection bias (34). However, important job-related aspects were not available (i.e., job type, occupational exposures, psychological, and physical factors). We assessed the potential mediator effect of stage at diagnosis and comorbid conditions by removing those adjustment factors in the analyses, resulting in no substantial changes in the survival estimates.

Many studies have suggested that living together/cohabitating has beneficial effects for most patients with cancer, independent of age, which could be assigned to stronger social support and healthier lifestyle (35). In agreement with international studies (3, 23, 36), our study showed that being single, divorced, or widowed affected cancer survival negatively for most cancer types.

In the present study, differences in cancer survival according to SES were most pronounced for head and neck, female breast, and colorectal cancer. These Belgian results correspond with international studies reporting strong associations between SES and survival among patients diagnosed with a cancer with good to intermediate overall prognosis (such as female breast cancer), and certain life-style–related cancer types (e.g., head and neck cancers; refs. 25, 37, 38). The association between SE and SD factors and survival among patients with lung cancer were statistically significant but less pronounced compared with other lifestyle-related cancers, which confirms results of previous studies in other populations (35), suggesting that for cancers with poor prognosis, such as lung and pancreatic cancer, there might be less potential impact of SES (1, 39–41).

Although screening programs in Belgium exist for breast and colorectal cancer, information on participation in screening was not available. For both cancer types, SE and SD factors were strongly associated with survival in this study. In line with international observations, a lower uptake of screening for lower SES groups might also be possible in Belgium (42). Nevertheless, other researchers have observed SE inequities in cancer survival for both screen-detected and non–screen-detected groups, independent of stage at diagnosis (38).

Reasons behind the observed disparities are multiple and complex and should be further examined. The absence of survival disparities according to SES in clinical trials (43–45) suggests that lower survival among deprived patients in the general population, as observed in our study, may be due to healthcare factors. Delay in diagnosis and treatment, thoroughness of diagnostic investigation, access to optimal care and complications of treatment might be different between deprived and affluent patients in a real-world situation (1, 46). Besides healthcare-related factors, differences in health-consciousness, health-seeking behaviors, or social support and connections might play a role and can be presumed to be more important in the current study given the high-level and accessible healthcare setting in our country. Health policy interventions need to consider the observed deprivation gap to plan targeted actions and to further promote equity in health, health awareness, and treatment adherence.

Our study presented several limitations. Income is associated with age and gender. It is less stable than other dimensions of SES, for example, educational attainment, and cannot be set on equal terms with wealth or self-assessed financial status, especially in the older population. Data on education, occupation, and de facto living situation would have provided additional insights on how SES interplays with cancer survival. Unfortunately, these data were not available. The integration of all three SE or SD factors into one model would have provided information on how these factors interact with each other and survival; however, these models would have lacked other dimensions of SES, and hence still provided a partial view of the relationship between SES and cancer survival. We therefore chose to study each SE or SD factor separately, taking clinical characteristics and treatment into account. A cross-tabulation of household income, employment, and marital status is provided in Supplementary Material (S6).

Origins and mechanisms leading to inequalities in health are multifaceted; besides clinical patient and tumor characteristics, residual confounding from lifestyle factors is likely present. No information regarding alcohol and tobacco consumption, nutrition, physical activity, obesity, psychosocial and mental health, social support, health consciousness or healthcare seeking behaviors was available (20, 34, 47–49). Furthermore, information on molecular markers, which affect available treatment options and prognosis in different cancer types, were unavailable.

A high proportion of patients with pancreatic cancer presented with diabetes mellitus. However, the relationship between diabetes and pancreatic cancer is complex and intertwined: Although diabetes is an important risk factor for pancreatic cancer, new onset diabetes may be an early symptom of yet undetected pancreatic cancer (50).

The treatment data refer to the received primary treatment scheme, which does not necessarily reflect the intended treatment. Furthermore, treatment was not considered as a time-varying covariable but at baseline (51). However, the effect of treatment as an additional adjustment factor was assessed in sensitivity analyses; the association between SES and survival remained statistically significant but was less pronounced, especially for employment and marital status.

Several studies observed trends in cancer survival disparities according to SES (5, 52, 53). In our study, sensitivity analyses considered the inclusion of the year of diagnosis as an additional adjustment factor. However, this variable was non-significant and excluded from the final analyses. Future research should consider a larger observation period to perform trend analyses.

The number of patients with cancer who died within 5 years after diagnosis varied largely by cancer type and patient characteristics. For some cancer types (i.e., ovary, pancreas and stomach), low numbers of death in some categories necessitate cautious interpretation of results.

Despite a comprehensive and nationwide health insurance program where equity in rights and access to healthcare are pursued, SES is a significant determinant of cancer survival in Belgium. Although differences in patient and tumor characteristics at least partially explained the association between SE and SD factors and cancer survival, household income, employment, and marital status remained associated with cancer survival for most cancer types. The magnitude of the association between SE and SD factors and cancer survival differed by cancer type and was most prominent for certain lifestyle-related cancer types (e.g., head and neck cancers) and those with good to moderate prognosis (e.g., female breast cancer and colorectal cancer). By providing new and substantial aspects, this study aims to build a more comprehensive picture of cancer epidemiology in Belgium and to provide new insights on how to tackle SE disparities in cancer survival.

M. Rosskamp reports grants from Kom op tegen Kanker during the conduct of the study. J. Verbeeck reports grants from Kom op tegen Kanker during the conduct of the study. S. Gadeyne reports other from Vrije Universiteit Brussel during the conduct of the study. F. Verdoodt reports grants from Kom op tegen Kanker during the conduct of the study. H. De Schutter reports grants from Kom op tegen Kanker during the conduct of the study. No disclosures were reported by the other authors.

M. Rosskamp: Conceptualization, software, formal analysis, funding acquisition, validation, methodology, writing-original draft, project administration, writing-review and editing. J. Verbeeck: Data curation, formal analysis, writing-review and editing. V. Sass: Validation, writing-review and editing. S. Gadeyne: Conceptualization, supervision, validation, methodology, writing-review and editing. F. Verdoodt: Conceptualization, resources, supervision, funding acquisition, validation, methodology, project administration, writing-review and editing. H. De Schutter: Conceptualization, resources, supervision, funding acquisition, validation, writing-review and editing.

M. Rosskamp, J. Verbeeck, F. Verdoodt, and H. De Schutter had been awarded a grant from Kom op tegen Kanker (KOTK). The authors would like to thank KOTK for funding this research; the Crossroads Bank for Social Security, Intermutualistic Agency (IMA), and eHealth for data preparation, delivery, and linkage; Gilles Macq and Geert Silversmit (Belgian Cancer Registry) for their statistical expertise; and Hanna Peacock (Belgian Cancer Registry) for proofreading the article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Skyrud
KD
,
Bray
F
,
Eriksen
MT
,
Nilssen
Y
,
Møller
B
. 
Regional variations in cancer survival: impact of tumour stage, socioeconomic status, comorbidity and type of treatment in Norway
.
Int J Cancer
2016
;
138
:
2190
200
.
2.
Akinyemiju
T
,
Meng
Q
,
Vin-Raviv
N
. 
Race/ethnicity and socio-economic differences in colorectal cancer surgery outcomes: analysis of the nationwide inpatient sample
.
BMC Cancer
2016
;
16
:
715
.
3.
Dalton
SO
,
Schüz
J
,
Johansen
C
,
Kjaer
SK
,
Engholm
G
,
Storm
HH
, et al
Social inequality in cancer incidence and survival in Denmark
.
Eur J Cancer
2008
;
44
:
2074
85
.
4.
Coleman
MP
,
Rachet
B
,
Woods
LM
,
Mitry
E
,
Riga
M
,
Cooper
N
, et al
Trends and socioeconomic inequalities in cancer survival in England and Wales up to 2001
.
Br J Cancer
2004
;
90
:
1367
73
.
5.
Tervonen
HE
,
Aranda
S
,
Roder
D
,
You
H
,
Walton
R
,
Morrell
S
, et al
Cancer survival disparities worsening by socio-economic disadvantage over the last 3 decades in new South Wales, Australia
.
BMC Public Health
2017
;
17
:
691
.
6.
Rachet
B
,
Ellis
L
,
Maringe
C
,
Chu
T
,
Nur
U
,
Quaresma
M
, et al
Socioeconomic inequalities in cancer survival in England after the NHS cancer plan
.
Br J Cancer
2010
;
103
:
446
53
.
7.
Vanthomme
K
,
Van den Borre
L
,
Vandenheede
H
,
Hagedoorn
P
,
Gadeyne
S
. 
Site-specific cancer mortality inequalities by employment and occupational groups: a cohort study among Belgian adults, 2001–2011
.
BMJ Open
2017
;
7
:
e015216
.
8.
Renard
F
,
Devleesschauwer
B
,
Gadeyne
S
,
Tafforeau
J
,
Deboosere
P
. 
Educational inequalities in premature mortality by region in the Belgian population in the 2000s
.
Arch Public Health
2017
;
75
:
44
.
9.
Haagedorn
P
,
Vandenheede
H
,
Vanthomme
K
,
Gadeyne
S
. 
Socioeconomic position, population density and site-specific cancer mortality: a multilevel analysis of Belgian adults, 2001–2011
.
Int J Cancer
2017
;
142
:
23
35
.
10.
Ellis
L
,
Woods
LM
,
Estève
J
,
Eloranta
S
,
Coleman
MP
,
Rachet
B
. 
Cancer incidence, survival and mortality: explaining the concepts
.
Int J Cancer
2014
;
135
:
1774
82
.
11.
Kogevinas
M
,
Pearce
N
,
Susser
M
,
Boffetta
P
, editors.
Social inequalities and cancer. IARC Scientific Publication No. 138
.
International Agency for Research on Cancer
:
Lyon
,
France
, 
1997
.
12.
Rutherford
MJ
,
Hinchliffe
SR
,
Abel
GA
,
Lyratzopoulos
G
,
Lambert
PC
,
Greenberg
DC
. 
How much of the deprivation gap in cancer survival can be explained by variation in stage at diagnosis: an example from breast cancer in the East of England
.
Int J Cancer
2013
;
133
:
2192
200
.
13.
Singh
GK
,
Jemal
A
. 
Socioeconomic and racial/ethnic disparities in cancer mortality, incidence, and survival in the United States, 1950–2014: over six decades of changing patterns and widening inequalities
.
J Environ Public Health
2017
;
2017
:
2819372
.
14.
Henau
K
,
Francart
J
,
Silversmit
G
,
Vandendael
T
,
Pieters
G
,
Xicluna
J
, et al,
editors
. 
Cancer burden in Belgium 2004–2013
.
Brussels, Belgium
:
Belgian Cancer Registry
; 
2015
.
Available from:
https://kankerregister.org/media/docs/publications/BCR_publicatieCancerBurden2016_web160616.pdf.
15.
WHO Collaborating Centre for Drug Statistics Methodology
.
Guidelines for ATC classification and DDD assignment
.
Norwegian Institute of Public Health
.
[updated 2019 Dec 16; cited 2020 March 20]. Available from
: http://www.whocc.no/atc_ddd_index/.
16.
Jegou
D
,
Dubois
C
,
Schillemans
V
,
Stordeur
S
,
De Gendt
C
,
Camberin
C
, et al
Use of health insurance data to identify and quantify the prevalence of main comorbidities in lung cancer patients
.
Lung Cancer
2018
;
125
:
238
44
.
17.
Rosskamp
M
,
Macq
G
,
Nackaerts
K
,
Praet
M
,
Van Eycken
L
,
Van Meerbeeck
JP
, et al
Real-life treatment practices for malignant pleural mesothelioma in Belgium
.
Lung Cancer
2018
;
125
:
258
64
.
18.
Crossroads Bank for Social Security (CBSS/KSZ/BCSS)
.
[cited 2020 June 30]. Available from
: https://www.ksz-bcss.fgov.be/en.
19.
Hagenaars
AJM
,
de Vos
K
,
Zaidi
MA
.
Poverty statistics in the late 1980s: research based on micro-data
.
Luxembourg
,
Luxembourg
:
Office for Official Publications of the European Communities
; 
1994
.
20.
Hvidberg
L
,
Pedersen
AF
,
Wulff
CN
,
Vedsted
P
. 
Cancer awareness and socio-economic position: results from a population-based study in Denmark
.
BMC Cancer
2014
;
14
:
581
.
21.
De Schutter
H
,
Adam
M
,
Giusti
F
,
Schillemans
V
,
Gendt
CD
,
Jegou
D
, et al,
editors.
Cancer survival in Belgium 2004–2008
.
Brussels, Belgium
:
Belgian Cancer Registry
; 
2012
.
22.
Thomas
L
,
Reyes
EM
. 
Tutorial: survival estimation for Cox regression models with time-varying coefficients using SAS and R
.
J Statist Software
2014
;
61
:
1
23
.
23.
Sharpe
KA
,
McMahon
AD
,
Raab
GM
,
Brewster
DH
,
Conway
DI
. 
Association between socioeconomic factors and cancer risk: a population cohort study in Scotland (1991–2006)
.
PLoS ONE
2014
;
9
:
e89513
.
24.
Geyer
S
,
Hemström
O
,
Peter
R
,
Vågerö
D
. 
Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice
.
J Epidemiol Community Health
2006
;
60
:
804
10
.
25.
Auluck
A
,
Walker
BB
,
Hislop
G
,
Lear
SA
,
Schuurman
N
,
Rosin
M
. 
Socio-economic deprivation: a significant determinant affecting stage of oral cancer diagnosis and survival
.
BMC Cancer
2016
;
16
:
569
.
26.
Syriopoulou
E
,
Bower
H
,
Andersson
TML
,
Lambert
PC
,
Rutherford
MJ
. 
Estimating the impact of a cancer diagnosis on life expectancy by socio-economic group for a range of cancer types in England
.
Br J Cancer
2017
;
117
:
1419
26
.
27.
Jansen
L
,
Eberle
A
,
Emrich
K
,
Gondos
A
,
Holleczek
B
,
Kajüter
H
, et al
Socioeconomic deprivation and cancer survival in Germany: an ecological analysis in 200 districts in Germany
.
Int J Cancer
2014
;
134
:
2951
60
.
28.
Tervonen
HE
,
Morrell
S
,
Aranda
S
,
Roder
D
,
You
H
,
Niyonsenga
T
, et al
The impact of geographic unit of analysis on socioeconomic inequalities in cancer survival and distant summary stage—a population-based study
.
Aust NZ J Public Health
2017
;
41
:
130
6
.
29.
Singer
S
,
Bartels
M
,
Briest
S
,
Einenkel
J
,
Niederweiser
D
,
Papsdorf
K
, et al
Socio-economic disparities in long-term cancer survival—10 year follow-up with individual patient data
.
Support Care Cancer
2017
;
25
:
1391
9
.
30.
Robert
SA
,
Strombom
I
,
Trentham-Dietz
A
,
Hampton
JM
,
McElroy
JA
,
Newcomb
PA
, et al
Socioeconomic risk factors for breast cancer: distinguishing individual- and community-level effects
.
Epidemiology
2004
;
15
:
442
50
.
31.
Vrijens
F
,
Renard
F
,
Camberlin
C
,
Desomer
A
,
Dubois
C
,
Jonckheer
P
, et al
Performance of the Belgian health system—Report 2015—health services research (HSR)
.
Brussels, Belgium
:
Belgian Health Care Knowledge Centre (KCE)
; 
2016
.
32.
Chang
CM
,
Su
YC
,
Lai
NS
,
Huang
KY
,
Chien
SH
,
Chang
YH
, et al
The combined effect of individual and neighborhood socioeconomic status on cancer survival rates
.
PLoS ONE
2012
;
7
:
e44325
.
33.
Hagedoorn
P
,
Vandenheede
H
,
Vanthomme
K
,
Willaert
D
,
Gadeyne
S
. 
A cohort study into head and neck cancer mortality in Belgium (2001–2011): are individual socioeconomic differences conditional on are deprivation?
Oral Oncol
2016
;
61
:
76
82
.
34.
Møller
H
,
Coupland
VH
,
Tataru
D
,
Peake
MD
,
Mellemgaard
A
,
Round
T
, et al
Geographical variations in the use of cancer treatments are associated with survival of lung cancer patients
.
Thorax
2018
;
73
:
530
7
.
35.
Varlotto
JM
,
Voland
R
,
McKie
K
,
Flickinger
JC
,
De Camp
MM
,
Maddox
D
, et al
Population-based differences in the outcome and presentation of lung cancer patients based upon racial, histological, and economic factors in all lung patients and those with metastatic disease
.
Cancer Med
2018
;
7
:
1211
20
.
36.
Randi
G
,
Altieri
A
,
Gallus
S
,
Chatenoud
L
,
Montella
M
,
Franceschi
S
, et al
Marital status and cancer risk in Italy
.
Prev Med
2004
;
38
:
523
8
.
37.
Beckmann
KR
,
Bennett
A
,
Young
GP
,
Cole
SR
,
Joshi
R
,
Adams
J
, et al
Sociodemographic disparities in survival from colorectal cancer in South Australia: a population-wide data linkage study
.
BMC Health Serv Res
2016
;
16
:
24
.
38.
Woods
LM
,
Rachet
B
,
O'Connell
D
,
Lawrence
G
,
Coleman
MP
. 
Impact of deprivation on breast cancer survival among women eligible for mammographic screening in the West Midlands (UK) and New South Wales (Australia): women diagnosed 1997–2006
.
Int J Cancer
2016
;
138
:
2396
403
.
39.
Quaglia
A
,
Lillini
R
,
Mamo
C
,
Ivaldi
E
,
Vercelli
M
,
SEIH Working Group
. 
Socio-economic inequalities: a review of methodological issues and the relationships with cancer survival
.
Crit Rev Oncol Hematol
2013
;
85
:
266
77
.
40.
Mackenbach
JP
,
Kulhánová
I
,
Bopp
M
,
Deboosere
P
,
Eikemo
TA
,
Hoffmann
R
, et al
Variations in the relation between education and cause-specific mortality in 19 European populations: a test of the "fundamental causes" theory of social inequalities in health
.
Soc Sci Med
2015
;
127
:
51
62
.
41.
Rubin
MS
,
Clouston
S
,
Link
BG
. 
A fundamental cause approach to the study of disparities in lung cancer and pancreatic cancer mortality in the United States
.
Soc Sci Med
2014
;
100
:
54
61
.
42.
Hoeck
S
,
van de Veerdonk
W
,
De Brabander
I
,
Kellen
E
. 
Does the Flemish colorectal cancer screening programme reach equity in FIT uptake?
Eur J Public Health
2019
;
29
:
1108
14
.
43.
Nur
U
,
Rachet
B
,
Parmar
MKB
,
Sydes
MR
,
Cooper
N
,
Lepage
C
, et al
No socioeconomic inequalities in colorectal cancer survival within a randomised clinical trial
.
Br J Cancer
2008
;
99
:
1923
8
.
44.
Abdel-Rahman
ME
,
Butler
J
,
Sydes
MR
,
Parmar
MKB
,
Gordon
E
,
Harper
P
, et al
No socioeconomic inequalities in ovarian cancer survival within two randomized clinical trials
.
Br J Cancer
2014
;
111
:
589
97
.
45.
Gardy
J
,
Dejardin
O
,
Thobie
A
,
Eid
Y
,
Guizard
AV
,
Launoy
G
. 
Impact of socioeconomic status on survival in patients with ovarian cancer
.
Int J Gynecol Cancer
2019
;
29
:
792
801
.
46.
Forrest
LF
,
Adams
J
,
Wareham
H
,
Rubin
G
,
White
M
. 
Socioeconomic inequalities in lung cancer treatment: systematic review and meta-analysis
.
PLoS Med
2013
;
10
:
e1001376
.
47.
Morris
M
,
Iacopetta
B
,
Platell
C
. 
Comparing survival outcomes for patients with colorectal cancer treated in public and private hospitals
.
Med J Aust
2007
;
186
:
296
300
.
48.
Pokhrel
A
,
Martikainen
P
,
Pukkala
E
,
Rautalahti
M
,
Seppä
K
,
Hakulinen
T
. 
Education, survival and avoidable deaths in cancer patients in Finland
.
Br J Cancer
2010
;
103
:
1109
14
.
49.
Van den Borre
L
,
Deboosere
P
. 
Investigating self-reported health by occupational group after a 10-year lag: results from the total Belgian workforce
.
Arch Public Health
2018
;
76
:
68
.
50.
Li
D
. 
Diabetes and pancreatic cancer
.
Mol Carcinog
2012
;
51
:
64
74
.
51.
Falcaro
M
,
Carpenter
JR
. 
Correcting bias due to missing stage data in the non-parametric estimation of stage-specific net survival for colorectal cancer using multiple imputation
.
Cancer Epidemiol
2017
;
48
:
16
21
.
52.
Lyratzopoulos
G
,
Barbiere
JM
,
Rachet
B
,
Baum
M
,
Thompson
MR
,
Coleman
MP
. 
Changes over time in socioeconomic inequalities in breast and rectal cancer survival in England and Wales during a 32-year period (1973–2004): the potential role of health care
.
Ann Oncol
2001
;
22
:
1661
6
.
53.
Rachet
B
,
Woods
LM
,
Mitry
E
,
Riga
M
,
Cooper
N
,
Quinn
MJ
, et al
Cancer survival in England and Wales at the end of the 20th century
.
Br J Cancer
2008
;
99
:
S2
S10
.

Supplementary data