It has become increasingly clear that social determinants impact the health outcomes of patients with multiple myeloma (MM). Within private-payer healthcare models, low-income, unmarried status, uninsured status, and living in low socioeconomic status (SES) index areas are all associated with poorer health outcomes.¹ Though the mechanism by which SES impacts outcomes is unknown, previous studies have highlighted disparities in accessing treatment within private-payer models. However, there is limited data examining the association of SES with MM-related health outcomes within a public-payer universal healthcare system.

Ontario, Canada, has a universal healthcare system that covers chemotherapy drug and administration costs for provincial funding regimens. Therefore, we performed a retrospective cohort study using the Institute for Clinical Evaluative Sciences (ICES) population-based administrative database to evaluate the association between socioeconomic marginalization and the likelihood of receiving treatment, and overall survival (OS) among treated patients in Ontario. Given the theoretical equal access to care, and data that MM outcomes are improved with universal versus private-payer healthcare in the United States, we hypothesized that socioeconomic marginalization would not affect the health outcomes of Ontarian patients with MM.

Patients diagnosed with MM between 2007 and 2018 were identified using the International Classification of Diseases (ICD) histology codes for MM, as described previously.² Patients untreated by 1-year post-diagnosis and lacking ICD codes for symptomatic end-organ dysfunction were presumed to have smoldering MM (SMM) and were excluded.³ The Ontario Marginalization (ONMARG) index at diagnosis was used as an area-level surrogate measure of SES. It is a validated composite score incorporating census data on material deprivation (income, education, employment, reliance on government assistance), residential instability (people per dwelling, marital status), ethnic concentration (recent immigrants, self-identified minorities), and dependency (age <15 or >65 years).⁴ The ONMARG index is reported in quintiles, with higher quintiles representing more marginalized areas. To simplify reporting, we grouped patients in ONMARG quintiles 2 to 4. Baseline demographic information (age at MM diagnosis, sex), diagnosis date, community size (urban if population ≥10 000, otherwise rural), and the modified Charlson-Deyo Comorbidity Index (CCI) was collected.

Logistic regression was used to assess the association between ONMARG and receipt of treatment 1 year post MM diagnosis. A liberal time period for treatment initiation was used to ensure that patients experiencing treatment delays were not incorrectly classified as untreated or having SMM. Treatment was defined as the receipt of a proteosome inhibitor (PI; bortezomib), immunomodulatory drug (IMID; lenalidomide), alkylator (melphalan, cyclophosphamide), or an autologous stem cell transplant (ASCT) as these encompassed all provincially funded first-line treatments during the study period. An interaction term between age (<70 vs. ≥70) and ONMARG was used in the multivariable model to assess whether age was an effect modifier. Cox proportional hazards regression was used to assess the association between ONMARG and OS. As poverty is more prevalent in rural settings and rurality is associated with barriers to accessing healthcare,⁵ we used an interaction term between ONMARG and community size to assess for effect modification. Similarly, ASCT is independently associated with improved survival in MM, but ASCT rates may differ in more marginalized populations, so an interaction term between ASCT and ON-MARG index was used in the multivariable Cox model. Statistical significance was defined as p < .05. The date of the last follow up was November 2, 2022. The study was approved by the ethics board of McMaster University and followed the data confidentiality and privacy guidelines of ICES.

After excluding 3359 patients with presumed SMM, our newly diagnosed MM cohort consisted of 9777 patients. At 1 year post diagnosis, 2242 (23%) of MM patients remained untreated. Among treated patients, the median time to treatment was 35 (IQR 16–85) days, with no significant difference in time to treatment between ONMARG groups (p = .630). Compared to treated patients, untreated patients tended to be older (median age 69 vs. 77 years), have more comorbidities (CCI ≥ 2 in 11% vs. 18%), and live in more marginalized areas (ONMARG Q5: 18% vs. 23%). Patients from more marginalized areas were less likely to receive treatment within 1 year of diagnosis even after adjusting for community size, time period of diagnosis, baseline CCI, sex, and age at diagnosis (ONMARG Q5 vs. Q1: OR 0.71, 95% CI 0.60–0.84; ONMARG 2–4 vs. Q1: OR 0.88, 95% CI 0.77–1.01; p < .001). Similarly, marginalization was associated with reduced odds of receiving an upfront ASCT (ONMARG Q5 vs. Q1: OR 0.82, 95% CI 0.67–1.01; ONMARG 2–4 vs. Q1: OR 0.74, 95% CI 0.64–0.85; p < .001). We further studied the likelihood of receiving treatment within 3 months of diagnosis and the results remained consistent even after adjusting for confounding. Neither age nor community size was the effect modifier on the association between ONMARG and the likelihood of receiving treatment (p_interaction = 0.086 and p_interaction = 0.357, respectively).

The median OS of MM patients treated versus untreated by 1 year post diagnosis was 4.3 (95% CI 2.1–4.5) versus 0.9 (95% CI 0.7–1) years, respectively. As treatment rates differed between ONMARG quintiles, we restricted our OS analyses to patients treated within 1 year of diagnosis (baseline characteristics are summarized in Table 1). Patients living in more marginalized areas tended to be older, female, and urban dwellers and had lower upfront ASCT rates (though this may relate to perceived age-related ASCT eligibility). The median OS of patients transplanted versus non-transplanted patients was 8.3 (95% CI 7.8–8.9) versus 2.9 (95% CI 2.8–3.0) years, respectively. Given that the association between ONMARG and survival differed among transplanted versus non-transplanted patients (p_interaction = 0.003), the multivariable analyses were stratified by transplant status. Among patients treated without an upfront ASCT, living in a marginalized area was not associated with survival after adjusting for age at diagnosis, sex, baseline CCI, time-period of diagnosis, and treatment with a PI/IMID (ONMARG Q5 vs. Q1: HR 1.03, 95% CI 0.91–1.15, p = .678; ONMARG 2–4 vs. Q1: HR 0.97, 95% CI 0.88–1.07, p = .553). Among transplanted patients, living in marginalized areas was associated with prolonged survival even adjusting for the same confounders (ONMARG Q5 vs. Q1: HR 0.82, 95% CI 0.67–1.01, p = .057; ONMARG 2–4 vs. Q1: HR 0.74, 95% CI 0.64–0.85, p < .001).

Our study suggests that socioeconomic marginalization is associated with a reduced likelihood of receiving early treatment for MM, refuting the common assumption that universal healthcare results in equitable health outcomes. However, even when the barrier to accessing treatment is overcome, marginalization was not associated with inferior survival. Interestingly, we found that transplanted patients living in more marginalized areas had a significantly longer OS compared to those living in less marginalized areas; we hypothesize that this may reflect a transplant referral bias, where perhaps only the minority of fit patients from marginalized patients were assessed for ASCT.

Though the mechanism by which SES impacts outcomes is not entirely clear, even in a universal healthcare system, a MM diagnosis has significant financial implications that may result in barriers to treatment, such as lost work productivity for patients and caregivers to facilitate treatment or toxicity management, or uncovered costs of supportive medication. Disparities in accessing MM treatment within a universal healthcare paradigm have also been demonstrated in a New Zealand population study which showed that ASCT rates were lower among Māori/Pasifika (Indigenous peoples) compared to Europeans/others.⁶ Though this study found that socioeconomic deprivation was associated with poorer survival, survival was assessed among all patients with MM, irrespective of treatment status; thus it's unclear if the inferior OS was predominantly due to low SES or lack of treatment.

Our study's strengths include our use of a large, comprehensive population database with minimal loss to follow-up. Additionally, standardized provincial treatment algorithms minimize treatment-related differences that could influence survival outcomes. Furthermore, SES is a complex and multidimensional construct, and the ONMARG index encompasses multiple social domains to provide a comprehensive measure of area-level SES. Study limitations include the lack of patient-specific demographic and disease data (such as cytogenetics or baseline staging) which can impact treatment decisions, prognosis, and health outcomes. While disease-specific data is included in MM registries, patients treated in community centers or untreated patients are systematically underrepresented in these registries, therefore leading to significant bias. Given the degree of missing data, we could not evaluate the role that race, or treatment at a community versus academic center, had on health outcomes. Lastly, though we used a validated algorithm, it is possible that some untreated MM patients may have had SMM due to the overlap in ICD codes. However, given the dismal survival among untreated MM patients in this study, we believe this cohort is truly comprised primarily of newly diagnosed MM patients and not SMM patients.

In conclusion, our study demonstrates that even in a universal healthcare system, disparities in accessing treatment for MM exist, especially among patients that are socioeconomically marginalized. However, once treated, marginalization was not associated with poorer survival in our cohort. Additional studies including qualitative studies with patients, caregivers, and healthcare professionals are needed to understand the patient-level and structural barriers that lead to inequitable healthcare access.

CONFLICT OF INTEREST STATEMENT

The authors have no relevant conflicts of interests to disclose.