Abstract
Background
In recent years new medicines used in systemic anticancer therapy (SACT) have been successful in reducing cancer mortality. There are concerns, however, regarding the potential impact of socioeconomic position on patients’ access to these treatments. This study aims to assess the extent of socioeconomic differences in the initiation, type, and intensity of SACT for cancer patients, and to identify differences that could indicate barriers to treatment.
Methods
We conducted a cohort study of adult patients diagnosed with solid cancers in the North Denmark Region 2008 to 2020 using the Danish Cancer Registry. Socioeconomic position was categorized using income, highest attained education, cohabitation status, ethnicity, and distance to specialized treatment. Outcomes included initiation of SACT, intravenous treatments, different SACT regimens, cost of SACT, only curative therapy, and use of monoclonal antibodies. Analyses were adjusted for age, year, comorbidity, and stage and were stratified by sex, cancer group, and treatment intent.
Results
We identified 42 364 patients of whom 12 792 (30.2%) initiated SACT. Socioeconomic differences were observed with lower SACT initiation among patients with lower educational attainment, low income, or living alone. These patient groups also had decreased accumulated cost, intravenous treatments, and fewer different SACT regimens, particularly for gynecological, lung, and upper gastrointestinal cancers. The observed differences were more pronounced among patients treated with palliative intent.
Conclusion
This study indicates socioeconomic disparities in SACT utilization, with lower use among disadvantaged patients primarily within palliative treatment regimens. Some of the observed differences warrant further investigation into the complex socioeconomic mechanisms involved.
Key words: socioeconomic position, cancer, systemic anticancer therapy, health care disparities, inequality, immunotherapy
Graphical abstract
Highlights
-
•
Low socioeconomic position is suspected to limit access to cancer treatment.
-
•
Variations in treatment initiation and type was found in the North Denmark Region.
-
•
Disadvantaged patients show a consistent pattern of reduced treatment use.
-
•
Socioeconomic differences were most pronounced in palliative intended therapy.
Introduction
Mortality rates are declining across various solid cancers. This trend is in part due to advancements in systemic anticancer therapy (SACT), such as the introduction of novel chemotherapies, immunotherapies, and targeted therapies. Research has shown, however, that the use and access to these therapies, to some degree, differ by the socioeconomic position (SEP) of the patient.1, 2, 3 Countries with universal tax-funded health care often have equal access to treatment mandated by law and therefore any inequality in access would run counter to this intent.4
Across health care systems, socioeconomic characteristics such as educational background, personal network, labor market attachment, and income are strongly associated with cancer survival.5, 6, 7, 8, 9 These survival differences are partially attributable to socioeconomic differences in prognostic factors such as lifestyle, comorbidities, or disease stage at diagnosis, but there is evidence that these differences may be additionally explained by disparities in treatment.9, 10, 11
Among the few previous studies in Denmark comparing SACT use in relation to SEP, a study from 2017 showed that patients with acute myeloid leukemia who had short education were less likely to receive induction chemotherapy, even when accounting for differences in clinical factors such as white blood cell count and cytogenetic risk groups.12 Another study showed that living alone and a long distance to the treating hospital were associated with not receiving the guideline-recommended curative treatment in non-small-cell lung cancer.13 International meta-analyses have shown disparities in the use of both chemotherapy and targeted therapies for cancer patients with low SEP.3,14
One of the main obstacles for research concerning disparities in SACT use is the lack of individual-based information on SEP and medicine administrations. England was among the first countries to systematically gather nationwide data on the use of SACT, and assessing inequalities in access is described as one of the main purposes of this data collection.15
This study included patients diagnosed with solid tumors in the North Denmark Region (2008-2020), identified through the Danish Cancer Registry (CAR) and linked with a regional database of SACT administrations. By combining these data with individual-level indicators of SEP from administrative registries, we had a unique opportunity to study the complex mechanisms between SEP and SACT utilization in a pan-cancer setting. This includes the application of multiple indicators for both SEP and measures of SACT utilization, and the impact of potential explanatory factors such as cancer site, treatment intent, stage, and comorbidity. We did this with the aim of assessing the extent of socioeconomic differences in the initiation, type, and intensity of SACT, and to identify differences across groups that could indicate barriers to treatment.
Methods
Study design and inclusion criteria
This population-based cohort study included adult patients (age ≥18 years) diagnosed with a solid cancer (excluding nonmelanoma skin cancer) between 2008 and 2020 in the North Denmark Region, which covers a population of ∼600 000 or 10% of the Danish population. Patients were identified using CAR16 and linked to other administrative registries using a civil registration system (CRS) identifier. Information on SACT administrations from 1 January 2008 to 31 December 2021, was retrieved from the local database ARIA OIS for Medical Oncology (MedOnc) used for prescribing antineoplastic treatment of solid cancers.
Data on age, sex, date of death, date of diagnosis, and TNM stage at diagnosis were obtained from CAR. The study was limited to patients diagnosed between 1 January 2008 and 31 December 2020, to avoid left-truncated data, and to ensure a minimum 1 year of follow-up, thereby reducing bias from right-censored data. As there was a small degree of missing data on some SEP variables (<1%) patients with missing SEP data were excluded from models using those specific variables.
Patients were grouped by primary cancer diagnosis according to the International Classification of Diseases 10th Revision (ICD-10). The groups were: breast cancer, brain/central nervous system (CNS) cancer, head and neck cancer, lower gastrointestinal cancer, upper gastrointestinal cancer, lung cancer (including thoracic malignancies), gynecological cancer, urologic cancer, and a group of ill-defined and rare cancers. Note that ‘rare’ should be understood in the context of receiving SACT. Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2026.100691, presents the specific ICD-10 codes for each cancer group. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for observational studies in epidemiology17 and the Guidance for Reporting Oncology real-World evidence (GROW) guidelines.18
Socioeconomic factors
Individual and household socioeconomic data were retrieved from Statistics Denmark. These were aggregated into measures of SEP following standard methodology19 and using the SEPLINE guidelines.20 Highest attained education before diagnosis was obtained from the Danish Education Registry and divided into three levels, short: mandatory primary and lower secondary school (≤9 years, ISCED 0-2); medium: optional upper secondary school and vocational education (10-12 years, ISCED 3-4); and long: bachelor’s degree, business academy education, or higher educational level (>12 years, ISCED 5-8).
Income was obtained from the Danish Income Registry, using the equated disposable household income in the calendar year before the year of first SACT. We divided income groups based on income quartiles for all Danish residents stratified by age and calendar year. Cohabitation was based on data from the CRS, and we defined non-cohabiting as a person residing at an address with only one registered adult. Ethnicity was based on data from CRS, and categorized individuals with Danish origin, i.e. individuals who had at least one parent with Denmark as country of birth as Danish. Information on residence was obtained from the Danish Address Registry and distance to specialized treatment was defined as the direct distance from the central hospital to the center of the parish of residence 6 months before first treatment. Distance was divided into: <25 km, 25-100 km, and ≥100 km.
Treatment
SACT was defined as medical agents in the Anatomical Therapeutic Chemical (ATC) classification system group L (antineoplastic and immunomodulating agents), excluding L02 (endocrine therapies) due to inconsistent data.
Treatment was quantified using five measures. (i) Accumulated cost of SACT (continuous variable) using data on the cost of individual SACT administrations [oral, intravenous (I.V.), and subcutaneous]. Expenditures for SACT were obtained from the hospital pharmacy database, and individual treatments were priced by using a yearly average price-per-milligram by ATC code. Costs were accumulated at the patient level using constant prices in EUR. (ii) Number of I.V. treatments received (integer variable). (iii) Number of different SACT regimens received (integer variable), identified using predefined treatment regimens in the clinical SACT registry. (iv) Curative intended therapy (binary variable), an indicator of whether patients only received regimens categorized as adjuvant/curative, as opposed to regimens categorized as metastatic/palliative or a mix of curative/adjuvant and metastatic/palliative. (v) Monoclonal therapy (binary variable), defined as patients who received one or more administrations of monoclonal antibodies as SACT. Monoclonal antibodies were defined as ATC group L01F (monoclonal antibodies and antibody–drug conjugates).
Statistical analysis
Associations between SEP and cost of treatment were estimated as ratios from multivariable linear models using a log(x+1) transformation of the dependent variable. For discrete outcomes (I.V., regimens) rate ratios (RR) were estimated using Poisson regression. For binary outcomes (initiation, curative, monoclonal) we estimated odds ratios (OR) using logistic regression. Table 1 summarizes the outcome measures and models. In all analyses we estimated the controlled direct effect by adjusting for confounders (sex, age at diagnosis, and year of diagnosis), and potential mediators [cancer site, Elixhauser comorbidity index, and TNM (tumor–node–metastasis) stage]. A hypothesized directed acyclic graph (DAG) of the associations can be found in Supplementary Figure S1, available at https://doi.org/10.1016/j.esmorw.2026.100691.
Table 1.
Outcome measures and statistical modeling framework for SACT
| Label | Description | Model | Interpretation | Cohort |
|---|---|---|---|---|
| Initiated SACT | Patients diagnosed 2008 to 2020 who initiated SACT (binary) | Logistic regression | Odds ratio (OR) | Diagnosed cohort |
| Cost | Accumulated cost of SACT in EUR (continuous) | Linear model using log(x+1) as dependent variable (log-level model) | Ratio/multiplier | Initiated SACT cohort |
| I.V. | Number of I.V. treatments received (count) | Poisson regression | Rate ratio (RR) | |
| Regimens | Number of different SACT regimens received (count) | Poisson regression | Rate ratio (RR) | |
| Curative | Only curative intended therapy (binary) | Logistic regression | Odds ratio (OR) | |
| Monoclonal | Any monoclonal antibody therapy ATC group L01F (binary) | Logistic regression | Odds ratio (OR) |
ATC, Anatomical Therapeutic Chemical; SACT, systemic anticancer therapy.
This approach of estimating controlled direct effect was chosen, as we were interested in the effect of SEP not explained by these important predictors for SACT. In sensitivity analyses, however, we estimated the total effect of SEP on SACT, adjusting for confounders only. Additionally, in observational studies measures of treatment are at risk of bias due to right-censoring and death as a competing event. To account for this, we carried out a sensitivity analysis including patients conditioned on a minimum of 1-year survival and with treatment data restricted to 1 year following diagnosis. Estimates were presented with 95% confidence intervals (CIs) and significant results in figures are marked by P < 0.01. All analyses were carried out in R 4.2.2,21 forest plots were generated using ‘forestmodel’,22 and standardization was carried out using the ‘stdReg’ library.23
Results
Between 2008 and 2020 a total of 42 364 individuals were diagnosed with a solid cancer in the North Denmark Region, as presented in Table 2. Among all diagnosed patients there was an overweight in lower income groups with 30.5% of patients in the lowest income quartile compared with 17.5% in the highest. Within the study period 30.2% of patients initiated SACT. The highest initiation rate was in brain/CNS cancers (71.5%) and the lowest for ill-defined and rare cancers (6.6%). Rates for each cancer group are provided in Supplementary Table S2, available at https://doi.org/10.1016/j.esmorw.2026.100691.
Table 2.
Patient characteristics
| Characteristics | Diagnosed cohort Total (n = 42 364) |
Initiated SACT cohort | ||
|---|---|---|---|---|
| Total (n = 12 792) | Women (n = 7223) | Men (n = 5569) | ||
| Age (years), median (IQR) | 69 (61-77) | 65 (56-72) | 63 (54-71) | 67 (60-73) |
| Age group, no. (%) | ||||
| 18-44 | 2147 (5) | 752 (6) | 589 (8) | 163 (3) |
| 45-64 | 12 508 (30) | 5286 (41) | 3359 (47) | 1927 (35) |
| 65-74 | 13 860 (33) | 4636 (36) | 2245 (31) | 2391 (43) |
| ≥75 | 13 849 (33) | 2118 (17) | 1030 (14) | 1088 (20) |
| TNM stage, no. (%) | ||||
| I | 10 728 (25) | 1738 (14) | 1448 (20) | 290 (5) |
| II | 6 232 (15) | 2053 (16) | 1478 (20) | 575 (10) |
| III | 6421 (15) | 3553 (28) | 1696 (23) | 1857 (33) |
| IV | 8811 (21) | 3839 (30) | 1847 (26) | 1992 (36) |
| Not recorded | 10 172 (24) | 1609 (13) | 754 (10) | 855 (15) |
| Year of diagnosis, no. (%) | ||||
| 2008-2011 | 12 271 (29) | 3336 (26) | 1945 (27) | 1391 (25) |
| 2012-2016 | 16 154 (38) | 5082 (40) | 2890 (40) | 2192 (39) |
| 2017-2020 | 13 939 (33) | 4374 (34) | 2388 (33) | 1986 (36) |
| Cancer group, no. (%) | ||||
| Breast | 6035 (14) | 2670 (21) | 2654 (37) | 16 (0) |
| Lung | 7593 (18) | 3316 (26) | 1582 (22) | 1734 (31) |
| Lower gastrointestinal | 6543 (15) | 2434 (19) | 1044 (14) | 1390 (25) |
| Upper gastrointestinal | 3660 (9) | 1419 (11) | 487 (7) | 932 (17) |
| Urology | 9578 (23) | 780 (6) | 90 (1) | 690 (12) |
| Head and neck | 1102 (3) | 443 (3) | 88 (1) | 355 (6) |
| Brain/CNS | 699 (2) | 500 (4) | 192 (3) | 308 (6) |
| Gynecological | 2441 (6) | 919 (7) | 919 (13) | — |
| Ill-defined and rare cancers | 4713 (11) | 311 (2) | 167 (2) | 144 (3) |
| Education, no. (%) | ||||
| Short | 17 504 (41) | 4749 (37) | 2769 (38) | 1980 (36) |
| Medium | 16 116 (38) | 5281 (41) | 2728 (38) | 2553 (46) |
| Long | 7578 (18) | 2552 (20) | 1622 (22) | 930 (17) |
| Disposable income, No. (%) | ||||
| Q1 (low) | 12 915 (30) | 3762 (29) | 2190 (30) | 1572 (28) |
| Q2 | 11 761 (28) | 3657 (29) | 2102 (29) | 1555 (28) |
| Q3 | 10 164 (24) | 3154 (25) | 1730 (24) | 1424 (26) |
| Q4 (high) | 7419 (18) | 2198 (17) | 1187 (16) | 1011 (18) |
| Cohabitation, no. (%) | ||||
| Living alone | 14 629 (35) | 3621 (28) | 2201 (30) | 1420 (25) |
| Cohabiting | 27 655 (65) | 9130 (71) | 5003 (69) | 4127 (74) |
| Distance to specialized treatment, no. (%) | ||||
| <25 km | 13 101 (31) | 3944 (31) | 2284 (32) | 1660 (30) |
| 25-100 km | 23 394 (55) | 7202 (56) | 4037 (56) | 3165 (57) |
| ≥100 km | 5145 (12) | 1525 (12) | 844 (12) | 681 (12) |
| Ethnicity, no. (%) | ||||
| Danish | 41 148 (97) | 12 353 (97) | 6943 (96) | 5410 (97) |
| Non-Danish | 1216 (3) | 439 (3) | 280 (4) | 159 (3) |
| Days to first SACT, median (IQR) | — | 42 (29-66) | 42 (29-62) | 43 (29-74) |
| Treatment intent, no. (%) | ||||
| Combination | — | 1734 (14) | 989 (14) | 745 (13) |
| Only curative | — | 5346 (42) | 3637 (50) | 1709 (31) |
| Only palliative | — | 5712 (45) | 2597 (36) | 3115 (56) |
| Monoclonal antibody therapy, no. (%) | ||||
| No | — | 10 028 (78) | 5564 (77) | 4464 (80) |
| Yes | — | 2764 (22) | 1659 (23) | 1105 (20) |
| I.V. treatments, mean (SD) | — | 19.2 (25.6) | 20.1 (25.3) | 18.1 (25.9) |
| Antineoplastic regimens, mean (SD) | — | 1.8 (1.3) | 1.9 (1.4) | 1.8 (1.1) |
| Cumulated cost EUR, mean (SD) | — | 11 000 (26 600) | 12 200 (30 200) | 9400 (21 000) |
IQR, interquartile range; I.V., intravenous; SACT, systemic anticancer therapy; SD, standard deviation; TNM, tumor–node–metastasis.
The cohort that initiated SACT included 12 792 patients, of whom 56% were women and 44% were men. The median age for women was 63 years [interquartile range (IQR) 54-71 years] and for men 67 years (IQR 60-73 years); this difference was mainly driven by the age profiles of sex-specific cancer types. The median time from diagnosis to first SACT was 42 days (IQR 29-66 days). The largest cancer groups were lung cancer (26%), breast cancer (21%), and lower gastrointestinal cancer (19%). The distributions of SEP across sex were relatively similar, but there were more women with long educations (22% versus 17%) and who were non-cohabiting (30% versus 25%) compared with men. The median follow-up time, calculated using the reverse Kaplan–Meier method across cancers, was 6.9 years (95% CI 6.8-7.1 years).
Differences in SACT initiation
The adjusted ORs for initiating any SACT across SEP are presented in Figure 1. Long education, high income, and cohabiting were positively associated with initiating SACT: OR 1.27, 95% CI 1.18-1.37; 1.43, 95% CI 1.32-1.54; and 1.61, 95% CI 1.53-1.71, respectively. There was no significant association for non-Danish origin (OR 0.89, 95% CI 0.77-1.03), and no clear pattern in distance to specialized treatment. Similar associations between SEP and SACT initiation were observed across cancer sites (see Supplementary Figure S2, available at https://doi.org/10.1016/j.esmorw.2026.100691).
Figure 1.
Effect of socioeconomic position on initiation of SACT. Estimated OR of initiating SACT across all cancers by socioeconomic exposure in separate models. The models are adjusted for age, sex, calendar year of diagnosis, cancer group, TNM (tumor–node–metastasis) stage, and comorbidities. CI, confidence interval; OR, odds ratio; SACT, systemic anticancer therapy.
Differences in SACT use
Association of SEP and SACT for the patients who received treatment are found in Figure 2A. For CIs and P values, see Supplementary Table S3, available at https://doi.org/10.1016/j.esmorw.2026.100691. For the total cohort, increased I.V. administrations were observed for those with long education, high income, cohabiting status, and Danish origin. For patients with Danish origin and patients cohabiting, an association with increased use of monoclonal antibodies was seen: OR 1.46, 95% CI 1.12-1.89 and OR 1.13, 95% CI 1.02-1.25, respectively. For the cost of SACT, we found positive associations for patients cohabiting (ratio 1.29, 95% CI 1.17-1.41) and those with high income (ratio 1.29, 95% CI 1.13-1.46). Long education (RR 1.05, 95% CI 1.02-1.07), high income (RR 1.09, 95% CI 1.05-1.13), and cohabiting (RR 1.07, 95% CI 1.05-1.10) were associated with a higher number of different SACT regimens. No clear pattern, except for a negative association with distance, was observed for receiving solely curative intended treatments (OR 0.82, 95% CI 0.71-0.95). For data on all levels of SEP for the total cohort see Supplementary Figures S3 and S4, available at https://doi.org/10.1016/j.esmorw.2026.100691.
Figure 2.
Effect of socioeconomic position on treatment of patients who initiated SACT. Estimates of association between SEP (exposure, y-axis) and treatment measure (outcome, x-axis) presented by (A) total cohort, (B) sex, (C) treatment intent, and (D) cancer group. For each group, number of patients/patients deceased is displayed in parenthesis. The models are adjusted for age, sex, year of diagnosis, stage, comorbidities, and cancer diagnosis. Estimates are colored by significance at a 1% level. For the SEP measures references are short education, first quantile (low) income, non-cohabiting, distance >100 km, and non-Danish origin. For CI and P values see Supplementary Table S3, available at https://doi.org/10.1016/j.esmorw.2026.100691. CNS, central nervous system; I.V., intravenous; OR, odds ratio; RR, risk ratio; SACT, systemic anticancer therapy; SEP, socioeconomic position.
Differences in SACT use across groups
In Figure 2B the analyses are stratified by sex. For men, the number of I.V. treatments was increased for high income (RR 1.18, 95% CI 1.07-1.29) and cohabiting (RR 1.17, 95% CI 1.09-1.24). For women, similar associations were observed, although the magnitudes were lower than in men: RR 1.08, 95% CI 0.99-1.17 and RR 1.09, 95% CI 1.03-1.16, respectively. For women with a long education, we observed a significant increase in accumulated cost (ratio 1.30, 95% CI 1.12-1.50) and number of regimens (RR 1.08, 95% CI 1.04-1.13); these associations were not observed for men.
In Figure 2C the analyses are stratified by treatment intent. The observed socioeconomic differences in SACT utilization were primarily present among patients receiving SACT with palliative intent. In the groups of patients who exclusively received SACT with curative intent, no significant differences were observed across SEP. For patients treated with a combination, i.e. they received both adjuvant and palliative regimens, we observed that patients with Danish origin or cohabiting were more likely to receive more regimens and more I.V. administrations compared with patients with non-Danish origin or living alone.
In Figure 2D the analyses are stratified by cancer group. Within groups, most of the significant associations with SEP were found in gynecological, lung, and upper gastrointestinal cancers. Specifically, estimates for cohabitation and accumulated costs were: ratio 2.31, 95% CI 1.53-3.50; ratio 1.58, 95% CI 1.37-1.83; and ratio 1.43, 95% CI 1.17-1.74, respectively. Estimates were highest within income, education, and cohabitation for the outcomes of cost, number of regimens and I.V. A significant decrease for the high SEP group was observed for lung and upper gastrointestinal cancers in relation to distance to specialized treatment.
For monoclonal therapy use in breast cancer the odds for Danish women were double those of non-Danish women (OR 2.02, 95% CI 1.51-2.53), correlating with an increase in accumulated cost (ratio 1.63, 95% CI 1.16-2.30). There was also a pattern of patients with non-Danish origin having reduced regimens and I.V. administrations in lower gastrointestinal and urologic cancers.
Correlated measures and differences in survival
In this study, the individual measures of SEP correlate to some degree, but also represent distinct concepts. Similarly, there is a correlation between some of the measures of treatment. To clarify these relationships, we have calculated the Spearman correlations, which can be found in Supplementary Figure S5, available at https://doi.org/10.1016/j.esmorw.2026.100691.
To quantify the differences in survival across SEP the differences in standardized 5-year survival probabilities were estimated. The difference in standardized 5-year survival for patients initiating SACT by SEP can be found in Supplementary Table S4, available at https://doi.org/10.1016/j.esmorw.2026.100691. Kaplan–Meier plots for overall survival from diagnosis and from first palliative treatment across SEP variables are available in Supplementary Figures S6 and S7, available at https://doi.org/10.1016/j.esmorw.2026.100691.
Sensitivity analyses
Considering the total effect of SEP on treatment (without adjusting for potential mediators), stronger or similar associations were observed (see Supplementary Figure S8 and Supplementary Table S5, available at https://doi.org/10.1016/j.esmorw.2026.100691). Across all cancers, larger socioeconomic differences in the use of monoclonal therapy and curative treatment were present compared with models adjusting for potential mediators. This pattern, however, mostly disappears when stratifying by cancer group, except in breast cancer where a strong association remains between SEP and curative SACT.
A second cohort, conditioning on a minimum of 1-year survival and restricting the measures of treatment to this period, is available in Supplementary Table S6, available at https://doi.org/10.1016/j.esmorw.2026.100691. In analyzing this cohort, we see a weaker pattern compared with the original analysis. The results are available in Supplementary Figure S9 and Supplementary Table S7, available at https://doi.org/10.1016/j.esmorw.2026.100691. Most effects, however, remained in the hypothesized direction that low SEP leads to reduced use of SACT.
Discussion
This study investigated socioeconomic differences in SACT utilization among patients with solid cancers in a pan-cancer approach. Across multiple indicators of SEP, we observed differences in initiation of SACT, and among those who initiated SACT, we found differences in the type and intensity of treatment. These differences were primarily found among patients who received palliative intended treatment. We found that patients with short education, low disposable household income, non-cohabiting status or non-Danish origin in several ways had a lower use of SACT. These findings align with previous studies on the influence of SEP in cancer treatment.9, 10, 11, 12, 13
Cohabitation status was found to be an important determinant for initiating SACT across all solid cancers. This association persisted, though diminished, when adjusting for differences in comorbidity and stage at diagnosis. A possible explanation for this is the risk of treatment-related side-effects that can be more difficult to manage for patients living alone. These patients and patients who are frail or otherwise at known risk for treatment-related toxicity may not be offered or may decline to initiate SACT. Additionally, the level of social support, structurally and emotionally, differs depending on living arrangements, personal resources, and network. Cohabitation is a known determinant of 1- and 5-year cancer survival rates in Denmark.1 According to a study by Østgård et al.,24 cohabitation status in acute myeloid leukemia influences treatment decisions favoring cohabiting patients. Another study found living alone was associated with decreased likelihood of standard treatment among non-small-cell lung cancer patients compared with cohabiting patients.13 Likewise, cohabiting cancer patients have been found to be favored in admittance to specialized palliative care in Denmark.25
For the group that initiated SACT most of the differences were observed in treatment among patients treated with a palliative intent. Possible explanations for this could be less standardized treatment and fewer or more ambiguous guidelines compared with treatment with a curative intent. The role of patient preferences also becomes increasingly important as the uncertainties around benefits and side-effects increase in the palliative setting.
In relation to sex, we observed that men with low SEP had a lower utilization of I.V. treatments and received fewer unique SACT regimens relative to women with low SEP. Previous research has shown that women to a larger extend seek and utilize health care services compared with men,26 which may partly explain this observation. Similarly, a meta-analysis of lung cancer patients showed that women were diagnosed at an earlier stage and had a higher probability of using inpatient cancer-care services.27
Another factor observed in this study was that the effect of SEP seemed to vary across type of malignancies, with patients treated for gynecological, lung, and upper gastrointestinal cancers showing the largest differences in treatment between high SEP and low SEP. These cancers are generally aggressive, something that likely influences the way SACT is used. Another plausible explanation could be an indirect association with lifestyle factors such as smoking or alcohol consumption in lung and upper gastrointestinal cancer, which additionally correlates with SEP and may reduce tolerance for SACT. For patients with gynecological cancers, an explanation is less obvious, but it could be related to the fact that the group covers both endometrial and ovarian cancers which are known to vary in risk factors.28
The study used monoclonal antibodies given as SACT as a way of distinguishing type of treatment. Monoclonal antibodies are a relatively new and often expensive treatment. Previous research has found reduced use of monoclonal antibodies in lung cancer patients with short educational background and/or low income.2 In this study we did not find reduced use of monoclonal antibodies across income or education, but we did find a difference for patients of non-Danish origin, mainly due to treatment of patients with breast cancer. The observed difference could be related to the patient’s ability to report side-effects when using monoclonal antibodies where non-native speakers could be challenged. A recent study in Denmark showed that, even in the presence of professional interpreters, miscommunication happens between cancer patients of non-Danish origin and the clinical oncologists.29 This may lead to differences in treatment choices compared with patients of Danish origin. The difference could also be associated with tumor characteristics as some research has shown breast cancer subtype variation across different ethnicities,30 which could lead to differences in appropriate SACT.
With regards to ethnicity, aside from differences in the use of monoclonal antibodies, we observed differences in the number of SACT regimens and the number of I.V. treatments in patients treated for urologic and lower gastrointestinal cancers. One potential explanation for this observation could be differences in diagnostics related to a form of inconvenience or cultural barrier related to intimate examinations, i.e. colonoscopies and prostate exams. A study from England found that stigma around cancer and cancer screening was higher in men and individuals with ethnic minority backgrounds.31
Cancer stage at diagnosis and comorbidity are crucial factors in determining SACT choice and intent. Some of the differences in treatment across SEP might be caused by the fact that low SEP patients often present with a more advanced stage or comorbidity at a diagnosis.32 This means the effect of SEP on treatment could be partly mediated through these factors. When comparing the controlled direct effect, i.e. accounting for the mediating effect of stage and comorbidity, with the total effect presented in the sensitivity analysis the estimates were reduced to some degree but still increased. While this indicates that socioeconomic differences in the use of SACT are partially attributable to differences in cancer stage and comorbidity, there are still other mechanisms behind the observed socioeconomic differences that warrant further investigation.
Strengths and limitations
Investigating the causal mechanisms behind differences across SEP groups given observational data is difficult and requires cautious interpretation. Differences could be caused by inequality in access, but they might also be explained by differing preferences for treatment, reflect differences in perceived fitness for a given treatment not sufficiently captured in the adjustment for stage and comorbidity, or differences in tumor histology and thereby the relevant SACT. In this study we did not know whether the decision for SACT was made by the clinician or the patient. It is likely that preferences for treatment and treatment intensity vary across SEP.
Additionally, a challenge was death as a competing risk. Treatment influences the risk of death, and death influences the potential for treatment. We attempted to remedy this in the conditioned sensitivity analysis. In that analysis we saw that by conditioning on 1-year survival and only including treatment within the first year, the estimated differences were smaller. This analysis, however, is susceptible to another set of biases as we do not know what happened to patients with short survival, nor the effect of treatment administered after the first year. One explanation is that this excludes much of the palliative treatment, something that likely influences the observed differences.
In the models the estimates of controlled direct effects rely on well measured mediators, and we know that stage at diagnosis and comorbidity are crude measures, which might bias analyses through residual mediation. We also know that lifestyle influences cancer type, even within cancer groups. For example, in oropharynx squamous cell carcinoma the treatment response and survival varies greatly according to HPV/P16 status and smoking status.33 Such differences can lead to differences in treatment and are a source of bias in the results. Furthermore, the potential for confounding by unmeasured health states remains a concern. Pre-existing conditions independent of socioeconomic origin, such as chronic disabilities, could result in biased estimates of the effect of SEP on treatment patterns.
In the main analysis, five SEP measures in relation to six SACT measures were investigated across several strata. This raises the issue of multiple testing, which is the increase in the chance of false positives. We have compensated for this by denoting significance at a 1% level instead of the conventional 5%, however, the exact method for solving this is debatable. Conventional corrections, e.g. Bonferroni correction, would be too restrictive given that both the outcomes and the exposures correlate.
A strength is the population-based approach with detailed data on both treatment and socioeconomic indicators. This analysis allows for comparisons across cancer groups and helps create hypotheses for further examinations of the socioeconomic differences in treatment.
Conclusion
This study highlights the presence of socioeconomic differences in relation to initiation, type, and intensity of SACT. Within patients who received SACT most of these differences were related to palliative intended treatment. Although differences are found in several areas, it is important to note that in most we found small or negligible differences, a result we believe highlights a clinical practice that to a large degree is standardized and aims to be non-discriminatory regarding patient background. In areas where we did find differences, some could be ascribed to differences in stage and comorbidity before diagnosis. Some differences, however, are worth investigating independently.
Overall, the results emphasize the critical role of social support for cancer patients in navigating the system and their treatment, as living alone was associated with lower initiation and use of SACT. We advocate for increased focus on identifying and addressing the needs of these patient groups, as it is a crucial step in reducing disparities in outcomes.
Acknowledgments
Funding
This work was supported by the Danish Cancer Society (R303-A17377) and institutional support from the Department of Hematology, Department of Oncology, Clinical Cancer Research Center, Aalborg University Hospital and Center for Clinical Data Science, Aalborg University and Aalborg University Hospital.
Disclosure
The authors have declared no conflicts of interest.
Data sharing
The participants of this study did not give consent for their data to be shared publicly, therefore due to the sensitive nature of the research sharing data is not possible.
Supplementary data
References
- 1.Dalton S.O., Schüz J., Engholm G., et al. Social inequality in incidence of and survival from cancer in a population-based study in Denmark, 1994–2003: summary of findings. Eur J Cancer. 2008;44(14):2074–2085. doi: 10.1016/j.ejca.2008.06.018. [DOI] [PubMed] [Google Scholar]
- 2.Gupta A., Omeogu C., Islam J.Y., et al. Socioeconomic disparities in immunotherapy use among advanced-stage non-small cell lung cancer patients: analysis of the National Cancer Database. Sci Rep. 2023;13(1):8190. doi: 10.1038/s41598-023-35216-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Norris R.P., Dew R., Sharp L., et al. Are there socio-economic inequalities in utilization of predictive biomarker tests and biological and precision therapies for cancer? A systematic review and meta-analysis. BMC Med. 2020;18(1):282. doi: 10.1186/s12916-020-01753-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Princippapir om prioritering for sygehuslægemidler. 2016. https://medicinraadet.dk/media/1oxfmhsz/ad-pkt-4-folketingets-7-principper-for-prioritering-af-sygehuslaegemidler_final-a.pdf Available at.
- 5.Rachet B., Ellis L., Maringe C., et al. Socioeconomic inequalities in cancer survival in England after the NHS cancer plan. Br J Cancer. 2010;103(4):446–453. doi: 10.1038/sj.bjc.6605752. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kogevinas M., Marmot M.G., Fox A.J., Goldblatt P.O. Socioeconomic differences in cancer survival. J Epidemiol Community Health. 1991;45(3):216–219. doi: 10.1136/jech.45.3.216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Dalton S.O., Olsen M.H., Johansen C., Olsen J.H., Andersen K.K. Socioeconomic inequality in cancer survival – changes over time. A population-based study, Denmark, 1987–2013. Acta Oncol. 2019;58(5):737–744. doi: 10.1080/0284186X.2019.1566772. [DOI] [PubMed] [Google Scholar]
- 8.Nielsen L.H., Kristensen D.T., Jakobsen L.H., et al. Socioeconomic status and overall survival among patients with hematological malignant neoplasms. JAMA Netw Open. 2024;7(3) doi: 10.1001/jamanetworkopen.2024.1112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Afshar N., English D.R., Milne R.L. Factors explaining socio-economic inequalities in cancer survival: a systematic review. Cancer Control. 2021;28 doi: 10.1177/10732748211011956. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Woods L.M., Rachet B., Coleman M.P. Origins of socio-economic inequalities in cancer survival: a review. Ann Oncol. 2006;17(1):5–19. doi: 10.1093/annonc/mdj007. [DOI] [PubMed] [Google Scholar]
- 11.Fiva J.H., Hægeland T., Rønning M., Syse A. Access to treatment and educational inequalities in cancer survival. J Health Econ. 2014;36:98–111. doi: 10.1016/j.jhealeco.2014.04.001. [DOI] [PubMed] [Google Scholar]
- 12.Østgård L.S.G., Nørgaard M., Medeiros B.C., et al. Effects of education and income on treatment and outcome in patients with acute myeloid leukemia in a tax-supported health care system: a national population-based cohort study. J Clin Oncol. 2017;35(32):3678–3687. doi: 10.1200/JCO.2017.73.6728. [DOI] [PubMed] [Google Scholar]
- 13.Langballe R., Jakobsen E., Iachina M., et al. Who are the vulnerable lung cancer patients at risk for not receiving first-line curative or palliative treatment? Acta Oncol. 2023;62(10):1301–1308. doi: 10.1080/0284186X.2023.2252581. [DOI] [PubMed] [Google Scholar]
- 14.Forrest L.F., Adams J., Wareham H., Rubin G., White M. Socioeconomic inequalities in lung cancer treatment: systematic review and meta-analysis. PLoS Med. 2013;10(2) doi: 10.1371/journal.pmed.1001376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Bright C.J., Lawton S., Benson S., et al. Data resource profile: The Systemic Anti-Cancer Therapy (SACT) dataset. Int J Epidemiol. 2020;49(1) doi: 10.1093/ije/dyz137. 15–15l. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Gjerstorff M.L. The Danish Cancer Registry. Scand J Public Health. 2011;39(suppl 7):42–45. doi: 10.1177/1403494810393562. [DOI] [PubMed] [Google Scholar]
- 17.STROBE guidelines v4. https://www.equator-network.org/wp-content/uploads/2015/10/STROBE_checklist_v4_combined.pdf Available at.
- 18.Castelo-Branco L., Pellat A., Martins-Branco D., et al. ESMO Guidance for Reporting Oncology real-World evidence (GROW) ESMO Real World Data Digit Oncol. 2023;1 doi: 10.1016/j.esmorw.2023.10.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Krieger N., Williams D.R., Moss N.E. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health. 1997;18:341–378. doi: 10.1146/annurev.publhealth.18.1.341. [DOI] [PubMed] [Google Scholar]
- 20.Hjorth C.F., Kjærulff T.M., Thomsen M.K., Cronin-Fenton D., Dalton S.O., Olsen M.H. SEPLINE: Socioeconomic position in epidemiological research—a National Guideline on Danish Registry Data. Clin Epidemiol. 2025;17:593–624. doi: 10.2147/CLEP.S520772. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.R Core Team R: A Language and Environment for Statistical Computing. 2023. https://www.R-project.org/ Available at.
- 22.Kennedy N. forestmodel: Forest Plots from Regression Models. 2020. https://cran.r-project.org/web/packages/forestmodel/ Available at.
- 23.Sjölander A. Regression standardization with the R package stdReg. Eur J Epidemiol. 2016;31(6):563–574. doi: 10.1007/s10654-016-0157-3. [DOI] [PubMed] [Google Scholar]
- 24.Østgård L.S.G., Nørgaard M., Medeiros B.C., et al. Associations between cohabitation status, treatment, and outcome in AML patients: a national population-based study. Blood. 2018;131(24):2730–2733. doi: 10.1182/blood-2017-11-814327. [DOI] [PubMed] [Google Scholar]
- 25.Adsersen M., Thygesen L.C., Neergaard M.A., et al. Cohabitation status influenced admittance to specialized palliative care for cancer patients: a nationwide study from the Danish Palliative Care Database. J Palliat Med. 2019;22(2):164–172. doi: 10.1089/jpm.2018.0201. [DOI] [PubMed] [Google Scholar]
- 26.Redondo-Sendino Á., Guallar-Castillón P., Banegas J.R., Rodríguez-Artalejo F. Gender differences in the utilization of health-care services among the older adult population of Spain. BMC Public Health. 2006;6:155. doi: 10.1186/1471-2458-6-155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Rana R.H., Alam F., Alam K., Gow J. Gender-specific differences in care-seeking behaviour among lung cancer patients: a systematic review. J Cancer Res Clin Oncol. 2020;146(5):1169–1196. doi: 10.1007/s00432-020-03197-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Cramer D.W. The epidemiology of endometrial and ovarian cancer. Hematol Oncol Clin North Am. 2012;26(1):1–12. doi: 10.1016/j.hoc.2011.10.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Birkelund L., Dieperink K.B., Sodemann M., Lindell J.F., Steffensen K.D., Nielsen D.S. Communicating without a shared language: a qualitative study of language barriers in language-discordant cancer communication. J Health Commun. 2024;29(3):187–199. doi: 10.1080/10810730.2024.2309357. [DOI] [PubMed] [Google Scholar]
- 30.Wan D., Villa D., Woods R., Yerushalmi R., Gelmon K. Breast cancer subtype variation by race and ethnicity in a diverse population in British Columbia. Clin Breast Cancer. 2016;16(3):e49–e55. doi: 10.1016/j.clbc.2015.09.005. [DOI] [PubMed] [Google Scholar]
- 31.Vrinten C., Gallagher A., Waller J., Marlow L.A.V. Cancer stigma and cancer screening attendance: a population based survey in England. BMC Cancer. 2019;19(1):566. doi: 10.1186/s12885-019-5787-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Ammitzbøll G., Levinsen A.K.G., Kjær T.K., et al. Socioeconomic inequality in cancer in the Nordic countries. A systematic review. Acta Oncol. 2022;61(11):1317–1331. doi: 10.1080/0284186X.2022.2143278. [DOI] [PubMed] [Google Scholar]
- 33.Olsen M.H., Frederiksen K., Lassen P., et al. Association of smoking, comorbidity, clinical stage, and treatment intent with socioeconomic differences in survival after oropharyngeal squamous cell carcinoma in Denmark. JAMA Netw Open. 2022;5(12) doi: 10.1001/jamanetworkopen.2022.45510. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.