Use of autologous hematopoietic cell transplantation as initial therapy in multiple myeloma and the impact of socio-geo-demographic factors in the era of novel agents

Authors


  • Conflict of interest: Dr. Go was supported by the Gundersen Medical Foundation and Gundersen Center for Cancer and Blood Disorders.

Abstract

Very effective combination chemotherapy using novel agents has become available in multiple myeloma (MM). Its impact on the use of high-dose chemotherapy and autologous hematopoietic stem cell transplantation (AHCT) as part of initial therapy is unknown. Using the National Cancer Data Base, we studied the rate of upfront AHCT use among 137,409 newly diagnosed MM patients from 1998 to 2010 in the United States and determined whether disparity exists among various sociodemographic as well as geographic subgroups. Overall, 12,378 (9.0%) patients received AHCT as part of initial treatment. The use of upfront AHCT increased steadily from 5.2% in 1998 to 12.1% in 2010 (trend test, P < 0.001), with no sign of plateau. This was seen across all socio-geo-demographic subgroups except among patients treated in the Northeast where the rate fell from 8.7% in 1998 to 6.6% in 2010. In multivariable analysis, patients with the following characteristics were the least likely to receive AHCT (odds ratio): year of diagnosis from 1998 to 2003 before the era of novel agents (0.67), older age (0.35), Black race (0.58), Hispanic ethnicity (0.78), low level of education or annual household income (0.55), residence in a metro area (0.66), no or unknown medical insurance (0.30), treatment at a community cancer center (0.16), and treatment facility located in the Northeast region (0.54). Even after the introduction of novel agents, the rate of upfront AHCT in MM continues to increase annually. Significant disparities exist dependent on demographic, social, and geographic factors. Am. J. Hematol. 89:825–830, 2014. © 2014 Wiley Periodicals, Inc.

Introduction

The era of novel agents in multiple myeloma (MM) began in 1999 with the discovery that thalidomide was effective among patients who relapsed after high-dose chemotherapy and autologous hematopoietic cell transplantation (AHCT) [1]. This culminated with the US Food and Drug Administration approvals of several new drugs in the categories of immunomodulatory drugs (thalidomide [2006], lenalidomide [2006], and pomalidomide [2013]) and proteosome inhibitors (bortezomib [2003] and carfilzomib [2012]). Prior to the era of these novel agents, the early incorporation of AHCT after induction chemotherapy became routine when randomized trials collectively showed improved progression-free survival over chemotherapy alone, especially in the younger age group [2]. However, the timing of AHCT was controversial because whether it was performed upfront or in a delayed fashion, overall survival was similar [3]. Nevertheless, most experts and clinical practice guidelines continue to recommend upfront AHCT for eligible patients [4, 5].

With current novel drug combinations producing a response in nearly all patients, the role of AHCT has come into question again [6]. Some believe that there is an increasing trend toward delaying AHCT in recent years [7, 8], but population-based data to support this belief are lacking. Moreover, there are non-disease-related considerations that influence the decision whether to undergo AHCT, for example, medical comorbidities, sociodemographic factors, and geographic location. Data regarding the influence of these factors are notably limited.

We conducted this study of 137,409 MM patients using data from the National Cancer Data Base (NCDB) with two major objectives: (a) to compare the pattern of AHCT as initial therapy before (1998–2003) and during (2004–2010) the era of novel agents; and (b) to determine the effect of demographic and socioeconomic characteristics, as well as geographic location and type of treatment facility.

Methods

Data source

Data were obtained from the NCDB β-Participant-User File (PUF). The NCDB, a joint program of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society, is a nationwide oncology outcomes database for more than 1,500 CoC-accredited cancer programs in the United States and Puerto Rico. Some 70% of all newly diagnosed cases of cancer in the United States are captured at the institutional level and reported to the NCDB. The PUF is a Health Insurance Portability and Accountability Act compliant data file containing cases submitted to the NCDB. The PUFs are designed to provide investigators associated with CoC-accredited cancer programs with a data resource they can use to review and advance the quality of care delivered to cancer patients through analyses of cases reported to the NCDB [9]. The NCDB maintains a number of Web-based data applications that have been developed to promote access to NCDB data. These tools are for use by CoC-accredited cancer programs as a means by which to evaluate and compare the cancer care delivered to patients diagnosed and/or treated at their facility with that provided at state, regional, and national cancer facilities. Data elements are collected and submitted to the NCDB from CoC-accredited program registries using nationally standardized data item and coding definitions, as specified in the CoC's Facility Oncology Registry Data Standards, and nationally standardized data transmission format specifications coordinated through by the North American Association of Central Cancer Registries. This includes patient characteristics, cancer staging and tumor histological characteristics, type of first course treatment administered, and outcomes information. All types of malignant tumors (hematologic and solid) are reported. NCDB only collects data on the first course of treatment, including the type (systemic therapy, surgery, or radiation) and class (chemotherapy, immunotherapy, hormone therapy, or transplant) of oncologic therapy. It collects the date of the first day of the first course of treatment (unless observation is the management of choice) but not subsequent therapies. For example, the first day of induction chemotherapy is captured but not the dates of stem cell harvest, high-dose chemotherapy, or transplant even all of them are part of the first course or cycle of treatment.

Patients

This retrospective study used prospectively collected data and included all patients with MM (ICD-O codes: 9731 [plasmacytoma, NOS], 9732 [MM], and 9734 [plasmacytoma, extramedullary]) diagnosed from 1998 to 2010. We collected data at the time of MM diagnosis, including sociodemographics (age, sex, race, ethnicity, educational level, annual household income, primary insurance, and location of residence), comorbidities using Charlson–Deyo score (available only for patients diagnosed in 2003 and later), and treatment facility (type and location) [10].

Definitions

The primary outcome of the study was the receipt of AHCT in the upfront setting. AHCT was considered part of initial MM treatment if it was administered as part of the first course of chemotherapy. AHCT is considered part of the first course of treatment if the local cancer registrar determines from the clinician's notes that AHCT is a planned treatment to follow induction chemotherapy. It is not considered part of the first course of treatment if the AHCT is performed at the time of relapse or if AHCT is not a planned treatment after induction chemotherapy. All patients had a minimum of 1 year of follow-up. We defined the era of novel agents as from 2004 and beyond because the first novel agent that became commercially available, bortezomib, was approved by the US Food and Drug Administration in May 2003. There is generally a lag of several months before an approved drug becomes available for commercial use.

The educational level and annual household income are based on these measurements for each patient's area of residence estimated by matching the zip code of the patient recorded at the time of diagnosis against files derived from the year 2000 US Census data. For education, it is a measure of the number of adults in the patient's zip code who did not graduate from high school and is categorized as equally proportioned quartiles across all US zip codes. The level of education is categorized as high (<14%), middle-2 (14%–19.9%), middle-1 (20%–28.9%), or low (>29%). Household income is also categorized in quartiles: high (≥$46,000), middle-2 ($35,000–$45,999), middle-1 ($30,000–$34,999), or low (<$30,000).

With regards to treatment facility, the major categories are community, academic, and others. Among community cancer programs, a comprehensive designation is given if the facility accessions more than 500 newly diagnosed cancer cases each year. To be an academic comprehensive cancer program, the facility must provide postgraduate medical education in at least four program areas, including internal medicine and general surgery. The location of the facilities can be classified geographically either into nine divisions or four regions (Supporting Information Fig. 1). A full and detailed description of all the variables is available in the NCDB PUF data dictionary [11].

Figure 1.

Rate of autologous hematopoietic cell transplantation as initial therapy by year. *Cochran–Armitage trend test, P < 0.001. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Statistical analysis

We included all patients in the analysis. Patients with missing data were placed in a category called “unknown” within each variable. We calculated the annual rate of AHCT as part of initial treatment over time and determined the trend using the Cochran–Armitage test. Our large sample size yields a 90% power to detect a difference of at least 0.05% in the outcomes between patients who had AHCT as initial treatment versus those who did not. The alpha level (two-tailed test) was 0.05. The data were normally distributed. To study the relationships between variables, we used χ2 tests and independent samples t-tests, for categorical and continuous variables, respectively. We also used odds ratio from univariate simple logistic regression with 95% confidence interval to further assess those relationships. In multivariable analyses, we used multiple logistic regressions with binary logit model and reported adjusted odds ratio with 95% confidence interval to see the impact of predictors on the main outcome (AHCT as initial treatment). Backwards elimination was used at alpha level of 0.05 to eliminate insignificant variables from the model.

Results

From 1998 to 2010, there were a total of 137,409 MM patients. The median age at diagnosis was 68 years. Most patients were White, non-Hispanic men living in a metropolitan area who had attained a middle to high level of education and had an annual household income of above $35,000. Most were covered by Medicare insurance, had limited comorbidities, and received their care at community cancer facilities. Detailed descriptions of the sociodemographic factors as well as treatment facilities of the whole population and those who did and did not have AHCT as part of initial treatment are summarized in Table 1.

Table 1. Multiple Myeloma Patient Characteristics by AHCT Status
CharacteristicN = 137,409AHCT part of initial treatment?
Yes (n =12,378)No (n = 125,031)
  1. Data are presented as number of patients (%) unless otherwise noted.

  2. a

    Only for patients diagnosed in 2003 and later.

  3. AHCT, autologous hematopoietic cell transplantation; MM, multiple myeloma.

Age   
Mean (years ± SD)68 ± 12.657 ± 8.968.0 ± 12.4
<65 years55,812 (40.6)10,048 (81.2)45,764 (36.6)
65–75 years42,368 (30.8)2,269 (18.3)40,099 (32.1)
>75 years39,229 (28.5)61 (0.5)39,168 (31.3)
Sex   
Women62,907 (45.8)5,203 (42.0)57,704 (46.1)
Men74,502 (54.2)7,175 (58.0)67,327 (53.9)
Race   
White106,477 (77.5)10,158 (82.1)96,319 (77.0)
Black25,649 (18.7)1,702 (13.8)23,947 (19.2)
Asian2,219 (1.6)239 (1.9)1,980 (1.6)
Other1,449 (1.1)132 (1.1)1,317 (1.1)
Unknown1,615 (1.2)147 (1.2)1,468 (1.2)
Ethnicity   
Non-Hispanic118,702 (86.4)10,876 (87.9)107,826 (86.2)
Hispanic7,358 (5.4)646 (5.2)6,712 (5.4)
Unknown11,349 (8.2)856 (6.9)10,493 (8.4)
Payor   
Managed care33,965 (24.7)6,493 (52.5)27,472 (22.0)
Private14,713 (10.7)2,249 (18.2)12,464 (10.0)
Medicare72,212 (52.6)2,344 (18.9)69,868 (55.9)
Medicaid6,095 (4.4)579 (4.7)5,516 (4.4)
Veterans1,059 (0.8)133 (1.1)926 (0.7)
Uninsured4,079 (3.0)158 (1.3)3,921 (3.1)
Unknown5,286 (3.9)422 (3.4)4,864 (3.9)
Income   
≥$46,00048,469 (35.3)5,146 (41.6)43,323 (34.7)
$35,000–$45,99935,868 (26.1)3,243 (26.2)32,625 (26.1)
$30,000–$34,99924,698 (18.0)1,941 (15.7)22,757 (18.2)
<$30,00021,448 (15.6)1,241 (10.3)20,207 (16.2)
Unknown6,926 (5.0)807 (6.5)6,119 (4.9)
Education   
High43,975 (32.0)4,807 (38.8)39,168 (31.3)
Middle-230,190 (21.9)2,714 (21.9)27,476 (21.9)
Middle-130,729 (22.4)2,497 (20.2)28,232 (22.6)
Low25,582 (18.6)1,553 (12.6)24,029 (19.2)
Unknown6,933 (5.1)807 (6.5)6126 (4.9)
Residence   
Metro106,542 (77.5)9,406 (76.0)97,136 (77.7)
Urban20,700 (15.1)1,909 (15.4)18,791 (15.0)
Rural2,913 (2.1)253 (2.0)2,660 (2.1)
Unknown7,254 (5.3)810 (6.5)6,444 (5.2)
Charlson–Deyo scorea   
069,368 (77.3)8,039 (86.6)61,329 (76.3)
113,830 (15.4)1,002 (10.8)12,828 (16.0)
>26,505 (7.3)241 (2.6)6,264 (7.8)
Facility type   
Academic49,657 (36.3)8,205 (66.3)41,657 (33.2)
Comprehensive community64,822 (47.2)3,326 (26.9)61,496 (49.2)
Community20,233 (14.7)545 (4.4)19,688 (15.8)
Other2,492 (1.8)302 (2.4)2,190 (1.8)
Facility location   
Atlantic21,951 (16.0)1,913 (15.5)20,038 (16.0)
Great Lakes24,974 (18.2)2,531 (20.5)22,443 (18.0)
Midwest11,782 (8.6)1,559 (12.6)10,223 (8.2)
Mountain5,948 (4.3)730 (5.9)5,218 (4.2)
Northeast8,208 (6.0)513 (4.1)7,695 (6.2)
Pacific15,873 (11.6)1,638 (13.2)14,235 (11.4)
South9,191 (6.7)570 (4.6)8,621 (6.9)
Southeast28,417 (20.7)2,164 (17.5)26,253 (21.0)
Southwest11,065 (8.1)760 (6.1)10,305 (8.2)

Overall, 12,378 (9.0%) patients received AHCT as part of initial treatment (Table 1). The use of AHCT increased steadily over time from 5.2% in 1998 to 12.1% in 2010 (trend test P < 0.001; Fig. 1). In general, the rates of AHCT across all socio-geo-demographic subgroups increased year after year, although to varying levels, with one notable exception. The AHCT rate decreased among patients treated in the Northeast, with a rate of 8.7% in 1998 and 6.6% in 2010. Disparities in AHCT in year 2010 were most pronounced in the following patient subgroups compared with their peer subgroups: Blacks (8.8% vs. 10.1%–13.0% in other races), low income (8.3% vs. 10.3%–13.4% in other income levels), low education (8.6% vs. 11.9%–13.4% in other education levels), treated in the community (4.3%–6.4% vs. 21.4% in academic setting), and treatment facility located in the Northeast, Southwest, or West divisions (6.6%–8.6% vs. 11.0%–20.1% in the rest of the country). The specific AHCT rates over time according to subgroups are outlined in Table 2.

Table 2. Rate of Autologous Hematopoietic Cell Transplantation as Initial Therapy Over Time by Patient Characteristic
CharacteristicTransplant rate by year (%)
1998200020022004200620082010
  1. NA, not available.

All combined5.26.87.68.89.711.312.1
Age       
<65 years12.715.516.317.519.021.221.5
65–75 years1.02.73.55.46.17.59.4
>75 years0.20.20.20.10.10.10.3
Sex       
Women4.66.56.87.79.110.711.3
Men5.87.18.39.710.211.812.7
Race       
White5.67.28.09.310.312.013.0
Black3.74.85.86.56.68.78.8
Asian6.88.27.912.712.410.511.7
Other4.66.66.77.511.18.810.1
Unknown1.06.76.75.813.711.69.4
Ethnicity       
Non-Hispanic5.37.07.88.910.011.412.4
Hispanic4.07.56.99.17.511.211.3
Unknown5.34.26.27.48.29.88.4
Payor       
Managed care12.314.815.518.120.323.224.2
Private11.913.315.115.716.416.016.4
Medicare0.61.72.33.23.74.75.4
Medicaid6.77.78.612.18.89.912.1
Veterans7.02.47.57.814.419.217.3
Uninsured2.44.03.94.02.73.75.6
Unknown5.911.010.08.25.14.46.9
Income       
≥$46,0006.88.78.79.711.913.313.4
$35,000–$45,9995.06.73.58.59.011.712.5
$30,000–$34,9994.15.47.09.08.29.710.3
<$30,0003.44.05.16.95.97.48.3
Unknown6.99.29.48.913.912.015.9
Education       
High6.78.89.810.212.014.213.4
Middle-25.36.87.19.09.611.111.9
Middle-15.06.06.57.88.910.212.0
Low2.83.95.37.25.57.58.6
Unknown6.99.29.38.813.912.015.9
Residence       
Metro5.06.87.58.69.511.311.6
Urban5.26.47.910.310.011.014.3
Rural4.26.77.96.39.28.910.0
Unknown9.98.68.68.211.712.314.1
Charlson–Deyo score       
0NANANA9.911.112.413.4
1NANANA5.26.38.99.6
>2NANANA2.93.44.34.6
Facility type       
Academic8.713.414.315.518.219.921.4
Comprehensive Community3.43.64.85.24.76.66.4
Community1.42.32.03.22.92.44.3
Other12.17.31.710.717.217.10
Facility location       
Atlantic3.67.07.99.310.010.112.6
Great Lakes4.65.78.98.810.613.612.9
Midwest7.510.310.213.714.717.017.0
Mountain5.35.96.58.612.222.320.1
Northeast8.77.12.95.86.37.46.6
Pacific5.37.89.911.811.811.513.0
South4.03.75.96.76.46.87.2
Southeast5.16.67.17.18.58.911.0
Southwest6.27.25.36.15.48.28.6

We further looked into the sociodemographic composition of patients treated in areas with lower AHCT rates and compared it with the entire MM population. Patients treated in the South were likely to be Black (25.2% vs. 18.7%), to have low levels of education (39.2% vs. 18.6%) and income (35.6% vs. 15.6%), and to have received treatment at a nonacademic center (72.8% vs. 63.7%). Patients treated in the Southwest were more likely to be Hispanic (12.0% vs. 5.4%), to have low levels of education (34.3% vs. 18.6%) and income (29.3% vs. 15.6%), to be uninsured/with unknown insurance (14.1% vs. 6.9%), and to have received treatment at a nonacademic center (69.1% vs. 63.7%). On the other hand, patients treated in the Northeast were more likely to be White (90.2% vs. 77.5%), have high levels of education (46.2% vs. 32.0%) and income (53.6% vs. 35.3%), and have received treatment at an academic center (42.4% vs. 36.3%).

In order to determine the factors associated with the probability of receiving AHCT as part of initial treatment, we analyzed all the patients diagnosed from 1998 to 2010 using univariable and multivariable analyses. In univariable analysis, patients with the following characteristics were more likely to receive upfront AHCT: diagnosis in the era of novel agents (2004–2010), younger age, men, non-Black race, known or reported ethnicity, higher level of education, higher annual household income, residence in a metropolitan area, managed care medical insurance, lower Charlson–Deyo score, treatment at an academic facility, and treatment facility located in the Midwest or Western regions. In the multivariable analysis, we combined the income and education level into a composite variable due to multicollinearity between income and education. The Charlson–Deyo score was not initially included because data on comorbidities were not routinely collected by NCDB until 2003. Except for sex, all of the factors found to be significant in the univariable model continued to be predictive in the multivariable model (Table 3). The c-statistic for the model was 0.84.

Table 3. Univariable and Multivariable Analyses of Factors Influencing the Use of Autologous Hematopoietic Stem Cell Transplantation as Initial Therapy
VariableUnivariable analysis [odds ratio (95% CI)]Multivariable analysis [odds ratio (95% CI)]
  1. NS, not significant; ND, not done.

  2. a

    Only for patients diagnosed in 2003 and later.

Year of Diagnosis  
1998–2003ReferenceReference
2004–20101.54 (1.49–1.60)1.51 (1.45–1.57)
Age  
<65 yearsReferenceReference
65–75 years0.40 (0.38–0.43)0.35 (0.33–0.37)
>75 years0.04 (0.03–0.04)0.01 (0.008–0.014)
Sex  
MaleReferenceNS
Female0.85 (0.82–0.88) 
Race  
WhiteReferenceReference
Black0.67 (0.64–0.71)0.59 (0.55–0.62)
Asian1.15 (0.99–1.31)0.77 (0.66–0.89)
Other0.95 (0.79–1.14)0.69 (0.57–0.83)
Unknown0.95 (0.80–1.13)0.71 (0.60–0.86)
Ethnicity  
Non-HispanicReferenceReference
Hispanic0.95 (0.88–1.04)0.80 (0.73–0.87)
Unknown0.81 (0.75–0.87)0.86 (0.79–0.93)
Payor  
Managed careReferenceReference
Private0.76 (0.72–0.81)0.83 (0.78–0.88)
Medicare0.14 (0.14–0.15)0.57 (0.54–0.61)
Medicaid0.44 (0.41–0.49)0.47 (0.43–0.52)
Veterans0.61 (0.51–0.73)0.69 (0.57–0.84)
Uninsured0.17 (0.15–0.20)0.18 (0.15–0.21)
Unknown0.37 (0.33–0.41)0.41 (0.36–0.45)
Income  
HighReference 
Middle-20.84 (0.80–0.87)ND
Middle-10.72 (0.68–0.76) 
Low0.52 (0.49–0.55) 
Unknown1.11 (1.03–1.20) 
Education  
HighReferenceND
Middle-20.81 (0.77–0.85) 
Middle-10.72 (0.69–0.76) 
Low0.53 (0.50–0.56) 
Unknown1.07 (0.99–1.16) 
Income/educationND 
High Reference
Middle-2 0.92 (0.87–0.97)
Middle-1 0.75 (0.70–0.80)
Low 0.55 (0.50–0.60)
Unknown 1.1 (0.97–1.24)
Residence  
MetroReferenceReference
Urban1.05 (1.00–1.11)1.42 (1.34–1.51)
Rural0.98 (0.86–1.12)1.49 (1.28–1.72)
Unknown1.29 (1.20–1.40)1.07 (0.95–1.21)
Charlson–Deyo Scorea  
0ReferenceND
10.59 (0.56–0.64) 
>20.29 (0.26–0.34) 
Facility type  
AcademicReferenceReference
Comprehensive Community0.14 (0.13–0.15)0.16 (0.14–0.17)
Community0.28 (0.26–0.29)0.29 (0.28–0.31)
Other0.70 (0.62–0.79)0.85 (0.74–0.97)
Facility location  
MidwestReferenceReference
Northeast0.70 (0.66–0.74)0.55 (0.51–0.56)
South0.62 (0.59–0.65)0.66 (0.63–0.70)
West0.97 (0.92–1.03)0.94 (0.88–0.99)

To overcome the multicollinearity between age group 65–75 years and Medicare insurance and to determine the impact of Charlson–Deyo score in the multivariable model, we performed three sensitivity analyses: (1) including only patients enrolled in Medicare; (2) including only patients diagnosed from 2003 and after; and (3) including only patients aged <65 years. The results were similar (data not shown). The only exception is that among patients aged <65 years, men and women were equally likely to receive upfront AHCT.

Discussion

In this population-based study, we determined the trend of AHCT as part of initial therapy in the United States among patients with MM before and after the advent of novel agents and explored the impact of socio-geo-demographic factors. We show that from 1998 to 2010 there is a steady rise in the annual rate of AHCT utilization across almost all socio-geo-demographic subgroups with no sign of plateau. However, significant disparities exist in the AHCT rate in these defined subgroups, with the most pronounced disadvantage among those who are Blacks, have low income or low education, and treated in the community setting or in facilities located in the Northeast, South, or West. Because our study included 70% of the US cancer population of all ages, our results should be broadly generalizable.

A recent pattern of care study investigated a small sample population of patients (N = 1,976) who survived at least 12 months from MM diagnosis and resided in regions that were part of the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) cancer registries. The study found a marked change in the initial treatment of MM from a predominantly cytotoxic chemotherapy approach (70%) in 1999 to almost routine use of novel agents (75%) in 2007. It also showed an increase in the use of AHCT within 12 months of MM diagnosis from 11% to 22% during the same time period [12]. Another study from the Center for International Blood and Marrow Transplant Research studied the trends in utilization and focused on the outcomes of AHCT within 12 months of MM diagnosis (N = 23,725). Because not all transplant centers are members of Center for International Blood and Marrow Transplant Research and the database does not capture MM patients who did not undergo AHCT, the investigators estimated the numbers of newly diagnosed myeloma patients and those who received upfront AHCT from 1995 to 2010. They found increasing use of upfront AHCT over time (5-year eras) and a lower (unadjusted) rate of use among Blacks. However, they did not report the rates of AHCT use by year or sociodemographic subgroups and did not study geographic considerations. The main conclusion was that the outcome of MM patients after AHCT is improving in the era of novel agents, but it is unclear whether AHCT could be deferred without adversely affecting survival [13].

Our study expands on these findings, but focused on the rates and disparities of AHCT over time across the country using the NCDB, which captures over 70% of the cancer population in the United States. Despite the widespread use of novel agents in the initial treatment of MM, there is no sign that the rate of upfront AHCT is decreasing, contrary to perceptions [7, 8]. On the other hand, the trajectory appears to be that of continuous rise with doubling of the AHCT rate in the past decade (Fig. 1). This may be due in part to the three major MM guidelines endorsing upfront AHCT as an appropriate option [4, 5, 14]. The National Comprehensive Cancer Network Multiple Myeloma Clinical Practice Guideline suggests either AHCT or continuation of novel agents after induction therapy for all newly diagnosed patients, whereas the Mayo Stratification for Myeloma and Risk-Adapted Therapy recommends upfront AHCT for high and intermediate risk patients. In contrast, the general consensus of the International Myeloma Working Group is to perform upfront AHCT for all transplant eligible patients. Interestingly, a large retrospective single institution study from Mayo Clinic suggests similar overall survival among patients receiving immunomodulatory drug-based induction therapy regardless of upfront or delayed AHCT [15]. Currently, the role of upfront versus delayed AHCT remains an open question awaiting results from two on-going randomized trials [16].

Our analyses show that for the whole MM population, there was slightly over two-fold increase in the rate of AHCT from 1998 to 2010 (from 5.2% to 12.1%), which was uniform across all subgroups with few exceptions. The highest increase was among patients who were 65–75 years of age or the Medicare population, with a corresponding nine-fold increase (from 1.0% to 9.4%). This was primarily due to Medicare's initiation of coverage for patients with Salmon–Durie stage II and III MM that began in October 2000 [17]. Perhaps even more important is the fact that Medicare does not cover the cost of stem cell storage for a delayed AHCT. This may encourage both patients and physicians to default to upfront AHCT. In contrast, the rates of AHCT among patients being treated at facilities in the South and Southwest remained relatively stagnant, whereas those in the Northeast declined. The location of treatment facility remained a significant predictor of the likelihood of upfront AHCT even after adjusting for other sociodemographic variables and co-morbidity. Although the lower AHCT rates in the South and Southwest divisions of the country can be partly attributed to a higher proportion of disadvantaged sociodemographic subgroups, this cannot be said for the Northeast. In addition, relatively more patients in the Northeast were treated at academic facilities compared nationally. We are unaware of why such geographic disparity exists resulting in a decline in AHCT rate in the Northeast. A hypothetical explanation is a regional treatment philosophy favoring delayed AHCT.

Studies addressing disparity in the access of AHCT in hematologic malignancies are very limited [18]. To date, only one large study has included MM patients. This study used the actual AHCT utilization data from the Center for International Bone and Marrow Transplant Research but estimated the number and sociodemographic composition of MM patients using SEER and US Census data, respectively. It found that Whites were 1.7 times more likely compared with Blacks and men were 1.1 times more likely compared with women to receive AHCT. However, the study covered the era prior to novel agents (1997–2002), did not differentiate upfront versus delayed AHCT, and was limited by the lack of adjustments for other sociodemographic variables as well as geographic considerations [19]. Our study shows that the following characteristics were independently associated with lower rates of upfront AHCT: older age, racial and ethnic minorities, lower level of education or annual household income, non-managed care medical insurance, residence in a metro area, higher Charlson–Deyo score, treatment at a community facility, and treatment facility located outside of the Midwest or Western regions. In contrast, women are as likely to receive upfront AHCT as men. However, the latter is confined to those aged >65 years. Although most of these findings seem intuitive and are expected, this is the first study to provide supporting evidence in MM. An interesting finding is that we did not see a lower rate of upfront AHCT among patients residing in the rural areas. This suggests that access to AHCT per se may not be a barrier to the known rural–urban inequality in the US cancer mortality as previously reported [20].

Our study has several limitations. Because NCDB rely on reporting from local cancer registries using International Classification of Diseases for Oncology (ICD-O), which is based on local pathology review. As a result, there is no central pathology verification of MM diagnosis. Moreover, there is no ICD-O code for smoldering MM. Therefore, we were unable to analyze only patients who have active MM. There may be other variables (known and unknown) not captured in our registry data that may affect the use of AHCT and introduce bias in our analyses, for example, co-morbidities and access to the use of novel agents. NCDB did not collect comorbidity data until 2003. However, we performed a sensitivity analysis that included Charlson–Deyo score for patients diagnosed after year 2002 and found the same predictive variables. The recent pattern of care study from SEER suggests that the majority of patients in recent years had novel agents as part of their initial MM therapy [12]. Therefore, the lack of access to novel agents may not be a major reason for the rise in the AHCT rate. Because non-CoC-accredited cancer programs do not participate in NCDB, our results may not apply to the patient population in this group. We did not provide outcome data comparing upfront versus delayed transplant, as treatment beyond initial therapy is not captured by the NCDB. Moreover, NCDB does not have data on prognostic factors based on cytogenetics and fluorescent in situ hybridization studies.

Even after the introduction of novel agents, the rate of AHCT as initial therapy in MM continues to increase annually in the past decade. Although a major factor is Medicare's initiation of coverage for AHCT in MM, the rise in use of upfront AHCT is also seen in the non-Medicare population as well and across almost all subgroups. This is despite remaining uncertainties in the comparative survival outcome of early versus delayed AHCT. Significant disparities exist in the rate of upfront AHCT use depending on demographic, social, and geographic factors. Our study provides a comprehensive baseline assessment of upfront AHCT utilization patterns in MM across the United States. Although demographic and socioeconomic factors have general associations, we show that they are also independent predictive factors. Because the receipt of upfront AHCT is dependent on response to induction therapy, future studies investigating disparity in chemotherapy response rates, duration, and side effects (as a surrogate for tolerability) according to sex, age, and sex race may provide additional insight.

Ancillary