Consolidative autologous hematopoietic cell transplant (autoHCT) is the standard of care for transplant-eligible patients with multiple myeloma (MM) who achieve at least a partial response to induction therapy. Although MM remains incurable, autoHCT is associated with delayed disease progression and low treatment-related mortality for patients with MM [1–6]. Several large studies have demonstrated progression-free survival (PFS) and overall survival (OS) benefits of autoHCT in patients with MM compared to continuance of maintenance therapy without autoHCT [1–6]. The findings of these trials lead to consolidative autoHCT becoming the standard of care in this patient population. AutoHCT can be used as a front-line consolidative treatment or delayed until disease recurrence.

Despite survival benefits and some improvement in autoHCT utilization over time, overall utilization remains low, at approximately 10–40% [7–9]. The decision to proceed to autoHCT is complex, with disease characteristics and comorbidities affecting transplant utilization. For instance, a higher comorbidity index score has been associated with inferior survival [10]. In addition, sociodemographic factors have been associated with autoHCT use, with lower rates of autoHCT among those residing in lower socioeconomic status neighborhoods (nSES) and those insured by Medicare or Medicaid in comparison to patients with private health insurance [11, 12]. Despite the higher incidence and younger age at diagnosis of MM among Black/African Americans, they are more likely to be referred for autoHCT later in their disease course compared to non-Hispanic White patients [13,14]. Prior population-based studies have not fully captured autoHCT utilization and did not examine trends over time regarding disparities in the utilization of autoHCT [12,14].

To address this gap, we used a novel data linkage between the California Cancer Registry (CCR), the Center for International Blood and Marrow Transplant Research (CIBMTR), and the California Patient Discharge Database (PDD) [15] to describe autoHCT utilization in a population-based cohort of patients diagnosed with MM. We evaluated sociodemographic characteristics associated with autoHCT utilization in different treatment eras to inform efforts to mitigate treatment disparities.

Our study identified 29,109 patients diagnosed with MM between 1991 and 2016. The median follow-up time from diagnosis was 10.2 years (95% CI 10.0 – 10.4 years). Overall, using all three data sources, 18.9% of patients (n=5500) underwent an autoHCT. The CIBMTR captured 70.8% of autoHCTs (17.8% of all three sources, 36.7% CIBMTR and PDD, 10.2% CIBMTR and CCR, and 6.1% CIBMTR only). The remaining autoHCTs were captured from PDD only (19.0%), CCR only (4.7%), or PDD and CCR (5.5%) (Figure 2). The median time from diagnosis to transplant was 9.7 months (interquartile range: 6.9–16.2).

Overall, patients of non-Hispanic White, Black/African American, Hispanic, and Asian/Pacific Islander race/ethnicity comprised 57%, 14%, 20%, and 9% of the cohort, respectively (Table 1). A lower proportion of Black/African American patients (15.8%) received an autoHCT than other racial/ethnic groups. The utilization of autoHCT was higher among patients residing in higher SES neighborhoods (15.2% low, 18.5% intermediate, and 22.0% high nSES). The frequency of MM patients who received autoHCT increased in each treatment era, from 5.7% during 1991–1995 to 27.4% during 2011–2016. The proportion of patients receiving an autoHCT was higher among those with private/military health insurance (28.4%) than other types of health insurance. There was no significant difference between male versus female sex and receipt of autoHCT throughout the study period. Patients with no or fewer comorbidities were also more likely to receive autoHCT. Among those with 0 comorbidites, 20.8% underwent autoHCT versus 17.9% among those with 1–2 comorbidites and 10.9% for those with 3 or more comorbidities. Older age at diagnosis was also associated with lower overall autoHCT utilization. Patients aged <40, 40–49, 50–59, 60–69, and 70–79 had autoHCT rates of 37.3%, 39.9%, 33.8%, 19.3% and 2.1%, respectively. Residing in a rural versus urban address had no statistically significant association with autoHCT utilization overall and in each time era. The cumulative incidence of autoHCT at 12 months increased over time, from 2.9% in 1991–1995, to 7.2% in 1996–2000, 11.9% in 2001–2005, 12.2% in 2006–2010, and 18.9% in 2011–2016 (Figure 3).

As there were significant interactions between year of diagnosis with race/ethnicity and nSES, the models were stratified by treatment era. In multivariable models, we identified that Hispanic patients were less likely than non-Hispanic White patients to have an autoHCT in 1991–1995 (HR 0.58, CI: 0.37 – 0.90), but this difference was less pronounced in later years (HR 1.07, CI: 0.96–1.19 for 2011–2016) (Table 2). Black/African American patients had lower utilization throughout the study period, particularly after 2005 (HR 0.66, CI 0.55 – 0.79 in 2006–2010; HR 0.83, CI 0.72 – 0.95 in 2011–2016).

Increasing age at diagnosis was also associated with decreased utilization of autoHCTs throughout the study and in each time era. Patients over the age of 50 were less likely to undergo autoHCT overall in comparison to patients < 50 years of age (ages 50–59 HR 0.71, CI 0.67 – 0.71; age 60–69 HR 0.40, CI 0.37 – 0.43; age 70–79 HR 0.05, CI 0.04 – 0.05). The differences for patients residing in the lowest versus highest nSES became less pronounced over time, with the strongest association observed in 1991–1995 (HR 0.35, CI: 0.24 – 0.52). Compared with private insurance, patients with Medicare were less likely to undergo autoHCTs, but only after 2006 (HR 0.79, CI 0.65 – 0.97, 2006–2010; HR 0.70, CI 0.63 – 0.78, 2011–2016). Those with Medicaid were less likely to undergo autoHCTs over the entire study period, (HR 0.81, CI 0.72 – 0.91, 2011–2016). Patients who were never married or previously married were also less likely to undergo autoHCTs compared to married patients after 1996 (HR 0.71, CI 0.63–0.80, 2011–2016). Lastly, compared to patients with no comorbidities, patients with 1–2 comorbidities were less likely to have autoHCT during the treatment era from 1996 to 2000 (HR 0.74, CI 0.55–1.00) but not during other time periods, whereas those with ≥3 comorbidities were less likely to have autoHCT throughout the study (HR 0.71, CI 0.56–0.92, 2011–2016).

In this population-based study, we examined sociodemographic disparities in the utilization of autoHCT among patients diagnosed with MM across treatment eras. To improve the precision of autoHCT ascertainment during 1991–2016, we linked three high-quality data sources (CCR, CIBMTR, and PDD), allowing for a more comprehensive analysis of autoHCT usage across patient subgroups. AutoHCT increased steadily over time, but utilization remained low in 2011–2016 with a cumulative incidence of 18.9% within one year of diagnosis. We found sociodemographic disparities in utilization decreased over time for specific subgroups, including patients of Hispanic race/ethnicity and those residing in lower SES neighborhoods. In contrast, Black/African American patients were less likely to have autoHCT throughout each treatment era and this difference increased after 2005. Further, MM patients with Medicaid or Medicare health insurance or no health insurance were also less likely to receive autoHCT in each treatment era. Whereas married or previously married patients and patients < 50 years were more likely to undergo autoHCT in each time era from 1996 onward, we did not identify differences in autoHCT utilization in other demographic categories, including sex or rural/urban residence. Despite some improvements over time, our findings highlight persistent disparities in autoHCT utilization among MM patients in California.

In our study, CIBMTR and the PDD each captured the majority (≥70%) of transplants, but each source also identified HCTs not captured by the others. Prior studies assessing autoHCT utilization have relied on SEER/Medicare databases, which is limited to those >65 years, and the CIBMTR, which does not include the non-HCT population or capture all autoHCTs due to voluntary reporting [21]. Studies utilizing a cancer registry [15] or the National Cancer Database only capture upfront autoHCT, likely resulting in an under-ascertainment of overall utilization. More fully capturing autoHCT utilization in a population-based cohort of MM patients allows us to identify patterns of autoHCT utilization more accurately. In addition, our study encompasses a longer follow-up time compared to prior studies [8, 9, 11, 12, 14, 22, 23], informing trends over time and changes in disparities in autoHCT utilization.

While we observed increases in autoHCT utilization in more recent treatment eras, utilization was observed among only 18.9% of patients within 12 months of MM diagnosis. Results of studies demonstrating PFS and OS benefits of autoHCT in MM patients [1–6] likely led to more patients being referred and considered for autoHCT. Our findings are similar to other population-based studies, including our prior work observing autoHCT utilization in 23.9% of MM patients between 1998 and 2012 using CCR and PDD data [7]. Al-Hamadani et al. [11] showed a similar increase in upfront autoHCT from 5.2% to 12.1% in the National Cancer Database from 1998 to 2010, although autoHCT estimates were lower than observed in our study. Differences in autoHCT utilization over time also have been observed by race/ethnicity. Schriber et al. reported increased utilization across all racial/ethnic groups over time in CIBMTR data. However, this increase was lowest among Hispanic and Black/African American patients [9]. A SEER-Medicare analysis by Ailwadhi et al. also showed increase in autoHCT utilization between 2007 and 2009 for all racial/ethnic groups, except Black/African American patients [8], consistent with our findings.

We found that Black/African American patients were less likely than non-Hispanic White patients to receive autoHCT throughout each treatment era, a difference more pronounced after 2005. Contrary to prior studies, Hispanic patients were less likely to undergo autoHCT only in the earliest era [8, 9, 11, 14]. Schriber et al. showed that Hispanic patients had a lower comorbidity score and achieved better treatment response before autoHCT compared to non-Hispanic White and Black/African American patients [9]. One study also showed that once patients were evaluated at a transplant center, there were no differences in who proceeded to autoHCT based on ethnicity [22]. Thus, access to transplant centers rather than disease characteristics, such as high-risk cytogenetic features, response to treatment, and clinical comorbidities, may be impacting disparities seen among race/ethnicity for autoHCT in MM [24]. Other factors such as physician bias, referral bias, social factors, and cultural beliefs regarding transplantation should be investigated further [25]. Exploring these factors will lead to better understanding of the potential causes of racial/ethnic disparities in autoHCT utilization, with the goal of increasing autoHCT referrals/evaluations and ultimately receipt of transplant.

We also observed that patients residing in the lowest nSES neighborhoods were less likely to utilize autoHCT. This is similar to findings in prior studies that showed decreased use of autoHCT in those with lower annual household income or who resided in lower nSES areas [8, 11, 14]. Consideration to several factors should be given, including access to a transplant center, referral for autoHCT evaluation, financial burden to the patient and/or their family, and social barriers to transplant such as caregiver support and availability of transportation to a transplant center. The importance of these social barriers are highlighted by our finding that married and previously married patients were more likely to undergo autoHCT, as supported by prior studies showing higher autoHCT utilization [7] and improved OS in married patients with MM [26]. Unlike prior studies [8, 11, 14], our results showed differences by nSES decreased over more recent treatment eras. As such, further investigation into what factors changed over time may aid in determining how factors contributing to lower autoHCT utilization among disadvantaged subgroups can be further mitigated.

Consistent with prior studies [11, 23], MM patients with Medicare and Medicaid insurance or no health insurance were less likely to undergo autoHCT. Notably, we are the first to report no improvements over time in the likelihood of autoHCT with Medicare or Medicaid compared to private insurance, emphasizing the need to identify the barriers to transplant for these patients. Our findings warrant further evaluation of how Medicare and Medicaid insurance coverage impacts pre-transplant evaluation, access to a transplant center, time/frequency of referral to a transplant facility, and financial burden to the patient and their family.

This study has several limitations. The CCR lacks information on performance status, molecular data, disease stage and risk stratification, systemic therapy, or treatment responses, which inform autoHCT eligibility. In addition, we were unable to determine whether patients were referred to a transplant center, and thus cannot differentiate between barriers to autoHCT at the initial provider level and those identified at the transplant center and any changes over time. The strengths of our study include a novel linkage of three datasets, which better captured autoHCT in a population-based cohort; a large number of MM patients, which allowed for adequate power to compare between subgroups within treatment eras; a long median follow-up time, allowing for the capture of later autoHCT utilization after MM diagnosis; and comprehensive data on sociodemographic factors, permitting us to identify relevant disparities in autoHCT utilization.