Association Between Insurance Status at Diagnosis and Overall Survival in Chronic Myeloid Leukemia: A Population-Based Study

Abstract

Background

Chronic myeloid leukemia (CML) can be treated effectively with tyrosine kinase inhibitor (TKI) therapy directed at BCR-ABL, but access to care, medication cost, and adherence may be barriers to treatment. We sought to determine whether insurance status at diagnosis influences CML patient outcomes.

Methods

We used the Surveillance, Epidemiology, and End Results Program (SEER) database and identified 5784 patients age 15 or older, diagnosed with CML between 2007 and 2012, with insurance status documented at diagnosis. Primary outcomes were 5-year overall survival. Covariates of interest included age at diagnosis, race, ethnicity, sex, county-level socioeconomic status, and marital status. Overall survival was evaluated by log-rank test and Kaplan-Meier estimates.

Results

Among patients age 15 to 64, insurance status was associated with overall survival (p<0.001), where being uninsured or having Medicaid was associated with worse 5-year OS compared to insured patients (uninsured 72.7%, Medicaid 73.1%, insured 86.6%). For patients over age 65, insurance had less of an impact on OS (p=0.07), with similar 5-year OS rates between patients with Medicaid and those with other insurance (40.2% vs. 43.4%). In multivariable analysis of patients age 15-64, both uninsured (HR 1.93, p<0.001) and Medicaid patients (HR 1.83, p<0.001) had an increased hazard of death compared to insured patients; age under 40, female sex, and married persons also had a lower hazard of death.

Conclusion

Our findings suggest that CML patients under age 65 who are uninsured or have Medicaid have significantly worse survival compared to patients with other insurance coverage.

Introduction

The advent of tyrosine kinase inhibitors (TKIs) that target the aberrant BCR-ABL kinase in patients with chronic myeloid leukemia (CML) ushered in a new era of targeted therapy in cancer care. CML was previously associated with a median survival of only 5 years, but more recently the outcomes of clinical trials treating patients with frontline TKI therapy report a long-term survival of approximately 90%, the majority in deep molecular remissions.^{1, 2} Indeed, in some countries the life expectancy of older patients receiving effective CML therapy may near that of the general population.³ There are a number of agents directed at BCR-ABL which are approved by the FDA for the treatment of CML, including imatinib, dasatinib, and nilotinib in the up-front setting, as well as bosutinib and ponatanib.

These significant advances are contingent upon access and adherence to effective and prolonged TKI therapy, taken once or twice daily, for an indefinite period of time. Indeed, non-adherence, even when manifesting in as few as 10% missed doses, is associated with worse molecular response rates.^4–6 In the United States, a number of factors may be associated with non-adherence to oral chemotherapies, including younger age, depression, adverse effects from therapy,^{4, 6, 7} as well as insurance coverage and amount of copayment.⁸ Each of the approved frontline TKIs has a cost of approximately $100,000 per year, a price that may limit access to these highly effective therapies.⁹ The amount that an individual pays depends largely on the extent that individual insurance plans defray patient out-of-pocket costs, as well as availability of and eligibility for financial support programs.

Insurance status impacts the outcomes of a number of malignancies. For certain solid tumors, uninsured patients or those with Medicaid coverage have higher rates of cancer-specific death, present with later-stage disease, and are less likely to receive cancer-directed surgery or radiation therapy.^10–12 The impact of insurance status on outcomes has not been as clear in hematologic malignancies, such as acute leukemia.^13–15 Moreover, to date, no such studies have specifically investigated CML, a malignancy for which access to therapy has a potentially high impact on outcomes due to the availability of highly effective and tolerable oral agents.

For the current analysis, we utilized the Surveillance, Epidemiology, and End Results (SEER) database, an accepted standard in population cancer outcomes, to identify patients with CML, who had insurance information documented at the time of their diagnosis. We hypothesized that patient insurance status at diagnosis would impact overall survival rates and other patient outcomes.

Methods

Data Source and Study Population

We identified CML cases through the Surveillance, Epidemiology, and End Results Program (SEER) 18 Registries research database (1973-2012), released April 2015, based on the November 2014 submission, from the National Cancer Institute. We included CML cases from SEER registries in Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound, Utah, Los Angeles, San Jose- Monterey, Rural Georgia, Alaska Native, Greater California, Kentucky, Louisiana, New Jersey and Georgia. SEER captures approximately 28% of the U.S. population and is meant to be broadly representative of population demographics, income, and education. In accordance with the policy of our institution for SEER database research, Institutional Review Board approval was not deemed necessary.

We included patients age 15 or older at the time of CML diagnosis. CML cases were identified according to ICD-O-3 codes 9863, chronic myeloid leukemia, and 9875, chronic myelogenous leukemia, BCR-ABL1 positive. We excluded patients with ICD-O-3 code 9876, atypical chronic myeloid leukemia, BCR-ABL1 negative, as these patients are not routinely treated with TKI therapy. Patients diagnosed at autopsy, or by death certificate only, were excluded from this analysis. We also included only those patients with known follow-up. Because SEER first started reporting patient insurance information in 2007, we only included those whose initial diagnosis was between 2007 and 2012, and who had a documented insurance status in the database at the time of diagnosis.

We classified insurance status according to whether patients were uninsured at diagnosis, had Medicaid coverage at diagnosis, or had another form of insurance at the time of diagnosis, which could include Medicare, military coverage, or private payers such as health maintenance organizations (HMOs), preferred provider organizations (PPOs) and managed care (coded in SEER as “Insured” or “Insured/No Specifics”). Patients with both Medicaid and Medicare are coded as Medicaid in SEER and treated as such in this analysis. We excluded patients without known insurance status from further analysis. Because patients age 65 or older are eligible for Medicare, we excluded patients in this age group whose insurance status was reported as uninsured (n=16). Marital status was categorized as married (including domestic partner and common law marriage), single, and separated (including divorced or widowed).

Estimates of socioeconomic status (SES) were performed using county attributes from the Census 2009-2013 American Community Survey 5-year data files, provided through the NCI SEER program SEER*Stat software. County-level estimates of median household income, percentage with < 9^th grade education, percentage with < high school education, and percentage living below the poverty line were assigned to patients according to their state-county recode. We identified Medicaid copayments according to state of residence using data from the Kaiser Family Foundation based on 2010 estimates (The Kaiser Family Foundation's State Health Facts. Data Source: KCMU Medicaid Benefits Database. Prepared by Health Management Associates for the Kaiser Commission on Medicaid and the Uninsured (KCMU), “Medicaid Benefits: Prescription Drugs.”).

Analytical Approach

Patient characteristics were compared using chi-square and Fisher's exact test for categorical variables and the Wilcoxon rank sum test for continuous variables. Median follow-up was calculated according to the method of Schemper and Smith.¹⁶ Overall survival (OS) was defined as the time from CML diagnosis to death, with censoring at the time last known alive in SEER according to the vital status recode. OS was analyzed according to insurance status using the method of Kaplan and Meier and the log rank test. Cox regression was performed separately within each age group to calculate predictors of overall survival, with variables of interest including insurance status, age, race, Hispanic ethnicity, sex, marital status, Medicaid copay, and county-level socioeconomic status. Patients without recorded marital status (n=388) were censored in the multivariate analysis; however, when these patients were included using a separate category of “unrecorded marital status” our findings did not significantly change. We dichotomized age for the multivariable analysis (<40 vs. ≥40 among patients under 65, and <75 vs. ≥75 for patients 65 or older). We dichotomized county-level variables at the median values for the entire cohort: household income ($56,460), poverty rates (13.9%), ninth grade education rates (6.4%), high school education rates (14.4%), and unemployment rates (9.9%). Forest plots were created using the %ForestMacro. Statistical analyses were performed using SAS software (Cary, NC). All p-values are considered significant at the two-sided 0.05 significance level.

Results

We identified 5784 patients aged 15 or older in the SEER database who were diagnosed with CML between 2007 and 2012 and had insurance status documented at the time of diagnosis. Of these, 337 were uninsured, 785 had Medicaid coverage, and 4662 had either private insurance or Medicare. The entire CML population was 57.4% male, and 81% white, 12% black, and 7% other races. Of the cohort, marital status was reported as single for 1205 (22.3%), married for 3114 (57.7%), and separated, divorced, or widowed for 1077 (20.0%) patients.

Non-Medicare Population (Age 15-64)

We identified 3626 patients aged 15-64 at the time of CML diagnosis. Of this group, 321 (8.9%) were uninsured, 595 (16.4%) had Medicaid coverage, and 2710 (74.7%) had insurance at the time of their diagnosis. Among this age group, patients who were uninsured or who had Medicaid had a median age of 44 and 45, respectively; while those who were insured had a median age of 50 (p<0.001). Insurance status was associated with race (p<0.001), Hispanic ethnicity (p<0.001), and marital status (p<0.001) in this cohort (Table 1). More patients in the uninsured or Medicaid groups were of non-white race, Hispanic ethnicity, or single. Insurance status varied according to the SEER registry of residence (P<0.0001). For registries with at least 20 patients, the percentage of uninsured among younger patients ranged from 0% (Hawaii, 0 of 45) to 18.73% (Greater Georgia, 47 of 251 patients) and Medicaid ranged from 7.1% (New Jersey, 28 of 294) to 24.3% (New Mexico, 11 of 70) (Supplementary Table 1).

Table 1. Patient Demographics.

	Age 15-64 (3626 patients)				Age 65+ (2142 patients)
	Uninsured	Medicaid	Insured	p-value	Medicaid	Insured	p-value
Total, n (%)	321 (8.9%)	595 (16.4%)	2710 (74.7%)		190 (8.9%)	1952 (91.1%)
Age, median (range)	44 (18-64)	45 (15-64)	50 (15-64)	<0.001	75 (65-97)	76 (65-102)	0.54
Gender, n (%)				0.04			0.007
Male	203 (63%)	328 (55%)	1603 (59%)		86 (45%)	1087 (56%)
Female	118 (37%)	267 (45%)	1107 (41%)		105 (55%)	865 (44%)
Race, n (%)				<0.001			<0.001
White	231 (72%)	402 (68%)	2112 (78%)		131 (69%)	1724 (89%)
Black	68 (21%)	114 (19%)	313 (12%)		26 (14%)	144 (7%)
American Indian	2 (1%)	25 (4%)	13 (0.5%)		2 (1%)	7 (0.4%)
Asian, Pacific Islander	15 (5%)	48 (8%)	215 (8%)		30 (16%)	65 (3%)
Unknown	5 (2%)	6 (1%)	57 (2%)		1 (0.5%)	12 (0.6%)
Hispanic Ethnicity, n (%)				<0.001			<0.001
Non-Hispanic	234 (73&percnt;)	420 (71&percnt;)	2341 (86&percnt;)		151 (79&percnt;)	1813 (93&percnt;)
Hispanic	87 (27%)	175 (29%)	369 (14%)		39 (21%)	139 (7%)
Marital Status, n (%)				<0.001			<0.001
Single	141 (46%)	281 (50%)	588 (23%)		41 (23%)	153 (9%)
Married/partner	123 (40%)	194 (34%)	1652 (65%)		67 (37%)	1069 (59%)
Divorced/separated/widowed	45 (15%)	90 (16%)	288 (11%)		72 (40%)	578 (32%)
County Level SES
Median Household Income	$56,240 (22.3-105.9K)	$56,240 (21.2-98.6K)	$57,660 (21.2-106.4K)	<0.001	$56,240 (21.7-98.6K)	$56,320 (27.8-106.4K)	0.49
Percentage with < 9th Grade Education	6.7%	7.5%	5.8%	<0.001	8.3%	5.8%	<0.001
Percentage <High School education	15.8%	16.4%	13.8%	<0.001	17.1%	13.8%	<0.001
Percent living below poverty line	16.1%	16.7%	13.7%	<0.001	16.1%	13.9%	0.01
Percent unemployed	10.2%	10.5%	9.9%	<0.001	10.2%	10.0%	0.26

We compared county-level metrics of socioeconomic status according to patient insurance status. Patients who were insured lived in counties with a higher median household income ($57,660), lower rates of persons with <9^th grade or <high school education, fewer persons living under the poverty line (13.7%, vs. 16.1% and 16.7%), and less unemployment (Table 1). Patients who were uninsured lived in states with a higher average Medicaid copay ($1.78) than patients who were on Medicaid ($1.51) or insured ($1.52) (p=0.01).

Medicare Population (Age ≥ 65)

We identified 2142 patients age 65 and over at CML diagnosis; 190 (8.9%) had Medicaid coverage and 1952 (91.1%) had other insurance, including Medicare. Within the older age group, Medicaid patients were more often female, of non-white race or Hispanic ethnicity, and less often married (Table 1). Insurance status varied according to the SEER registry of residence among those who are Medicare eligible (P<0.0001). Of registries with at least 20 patients, the percentage of Medicaid patients ranged from 0% (Utah, 0 of 31) to 19.9% (Los Angeles, 33 of 166 patients) (Supplementary Table 2).

Medicaid patients were significantly more likely to reside in counties with a larger population below the poverty line (16.1% vs 13.9%, p=0.01), and without a high school education (16.8% vs 13.8%, p<0.001). There was no difference in county-level unemployment rates or median household income according to insurance status.

Patient Outcomes

The median follow-up for survivors was 32 months. Among patients aged 15 to 64, being uninsured or having Medicaid was associated with worse survival compared to insured patients (p<0.001). At 5-years after diagnosis, patients with insurance at time of diagnosis had an 86.6% overall survival rate, while only 72.7% of uninsured patients and 73.1% of Medicaid patients were alive at 5-years (Figure 1A). For patients age 65 or older, there was no statistically significant difference in OS between patients with Medicaid and those with other forms of insurance (40.2% vs. 43.4%) (Figure 1B).

Figure 1.

Survival of patients (A) age 15-64 and (B) age 65 or older, according to insurance status at diagnosis. Among patients age 15-64, survival was worse among uninsured or Medicaid patients compared to those who had insurance (p<0.001). There was no significant difference in survival between patients with Medicaid and those with insurance for patients age 65 or older at diagnosis (p=0.05)

We performed Cox regression, separately within each age group, and adjusted for insurance status, age, race, ethnicity, sex, marital status, Medicaid copay, percentage living below the poverty line, percentage with <9^th and <high school education, median household income, and unemployment rates. For patients age 15-64, being uninsured or having Medicaid at the time of diagnosis was associated with an increased hazard for death, where insured patients are the comparator (uninsured HR 1.93, p<0.001; Medicaid HR 1.83, p<0.001, Figure 2). There was worse survival among patients age 40 or older compared to those younger than 40 (HR 2.44, p<0.001), and among male compared to female patients (HR 1.31, p=0.018). In addition, compared to married persons, single individuals had worse survival (HR 1.65, p<0.001). There was no association between outcomes and county-level socioeconomic metrics (Table 2), and there was no significant association between SEER registry residence and survival when adjusting for insurance status, marital status, age, and patient sex.

Figure 2.

Forest plot of patient characteristics on hazard of mortality among CML patients. Among younger patients ages 15 to 64 (top), having Medicaid, being uninsured, and being age 40 or male, male, or single, was associated with worsened survival. Among those age 65 or older (bottom), being 75 or older, male, and divorced/widowed was associated with worsened survival. LCL, lower confidence limit; UCL, upper confidence limit. Due to limited patient numbers American Indian (n=49) and unknown race (n=81) were excluded from this analysis.

Table 2. Multivariate Predictors of Mortality.

	Age 15-64 (3626 patients)			Age 65+ (2142 patients)
	Hazard Ratio for Death	95% CI	p-value	Hazard Ratio for Death	95% CI	p-value
Insurance status
Insured	1.00	-	-	1.00	-	-
Medicaid	1.83	1.39-2.40	<0.001	1.26	1.00-1.64	0.07
Uninsured	1.93	1.40-2.66	<0.001	n/a	n/a	n/a
Age Above Cut-off (40/75y)*	2.44	1.82-3.26	<0.001	2.57	2.19-3.01	<0.001
Sex
Female	1.00	-	-	1.00	-	-
Male	1.31	1.05-1.63	0.02	1.17	1.01-1.36	0.04
Race
White	1.00	-	-	1.00	-	-
Black	1.15	0.86-1.53	0.34	0.86	0.65-1.13	0.29
American Indian**	indeterminate		0.96	1.27	0.41-3.98	0.68
Asian/Pacific Islander	0.96	0.62-1.48	0.85	0.99	0.71-1.39	0.99
Hispanic ethnicity
Non-Hispanic	1.00	-	-	1.00	-	-
Hispanic	0.84	0.61-1.16	0.29	0.84	0.63-1.11	0.26
Marital Status+
Married	1.00	-	-	1.00	-	-
Single	1.65	1.29-2.13	<0.001	1.15	0.89-1.48	0.29
Separated/Divorced/Widowed	1.30	0.94-1.79	0.11	1.35	1.16-1.59	<0.001
Medicaid Copayment	1.03	0.94-1.14	0.50	0.99	0.93-1.05	0.73
Above Median Poverty Line	1.03	0.69-1.54	0.88	1.06	0.83-1.36	0.61
Above Median Household Income	0.98	0.69-1.40	0.91	1.02	0.83-1.27	0.82
Above Median % 9th Grade Education	1.16	0.83-1.62	0.38	0.97	0.78-1.20	0.76
Above Median % High School Education	1.05	0.73-1.52	0.80	1.20	0.94-1.53	0.15
Above Median Unemployment	1.01	0.78-1.32	0.92	1.02	0.86-1.20	0.92

^*Among patients age 15-64, age was compared between those younger than 40 and those 40 years of age or older; among patients over 65, patients were compared for those younger than 75 and those 75 years or older.

^**The hazard ratio was indeterminate for the American Indian race variable.

⁺Patients without recorded marital status (n=388) were censored from the multivariate analysis.

For patients age 65 or older at diagnosis, there was a trend toward worse survival in the Medicaid population that did not reach statistical significance (HR 1.26, p=0.073). Patients age 75 or older had worse survival compared to those younger than 75 (HR 2.57, p<0.001), and male patients had worse survival compared to females (HR 1.17, p=0.04). In the older patient population, compared to married patients, divorced/widowed/separated patients had worse survival (HR 1.35, p<0.001), although single patients did not have worsened outcomes (Table 2). Similar to patients under age 65, after adjusting for insurance status, marital status, age, and patient sex, there was no impact of SEER registry residence on survival.

Discussion

Among newly diagnosed CML patients in the U.S., insurance status at the time of diagnosis impacts overall survival at the population level, specifically among patients who are under age 65. In this age group, patients who were uninsured or who had Medicaid insurance had an approximately twofold hazard for death, compared to patients with other insurance. Among the Medicare eligible population, age 65 and older, insurance status did not have as large an impact on overall survival, although the overall outcomes of this group were generally poorer, similar to previous reports.^17–19 While not significant, the survival of Medicaid patients age 65 and older was marginally lower than insured patients through the duration of the analysis (Fig 1B, p=0.07). Interestingly, we found significant differences in survival according to whether patients are married or not at diagnosis, and male sex in both age groups was associated with worse survival. Together, our findings highlight the importance of considering payment structures when assessing broader population outcomes, including in highly treatable diseases such as CML.

Insurance programs in the United States vary widely, and are the primary means of offsetting the cost of therapies for those with cancer. Younger patients commonly receive insurance either through employer-based or direct purchase plans, which provide coverage to approximately 67% of insured individuals. Government plans include Medicare, provided for patients aged 65 and older and for certain disease states, and Medicaid, offered to those at high risk due to various socioeconomic criteria. Approximately 10% of the population, until recently, remained without insurance.²⁰ Although such programs successfully extend access to care among the US population, it has been shown that even modest variations in plan characteristics, such as small differences in Medicaid copays, may reduce the use of prescription medications, and increase emergency room visits.²¹ In our study, the Medicaid population had outcomes rivaling the uninsured population, both inferior when compared to insured patients. Interestingly, in the Medicare-eligible population over age 65, there was no significant difference in survival between Medicaid patients and other insured patients. One potential reason may be that Medicare is reported along with other insurance carriers, while Medicaid is reported separately in SEER. We were therefore unable to determine whether there is any difference in outcomes among Medicare patients and patients with private insurance. If Medicare patients do more poorly, this would bias toward the null and could have impacted our findings.

Given the high cost of TKI therapy, patient assistance programs have been developed by manufacturers as well as organizations to help defray this expense. However, these programs may have restrictions that exclude certain populations, including elderly patients on Medicare plans and undocumented immigrants. Indeed, previous research has identified cost-sharing as a potential barrier to TKI initiation and adherence among CML patients on Medicare,^{8, 22} and patients with higher co-pays are more likely to stop treatment within the first 6 months.⁸ In 2007 it was estimated that noncitizens comprise approximately 20% of the U.S. uninsured population;²³ undocumented immigrants comprise a disproportionate percentage of the uninsured, and following the Affordable Care Act (ACA) may represent up to 25% of the total uninsured population.^{24, 25} In addition, given the chronic nature of CML treatment, expenses associated with monitoring and follow-up can limit adherence.^{22, 26} These factors may have contributed to the poor outcomes seen both among the nonelderly uninsured population, and the insured patients over age 65 in this analysis, many of whom would have Medicare insurance.

Several previous studies have explored the impact of insurance status on survival in other cancers. Particularly among younger adults with solid tumors, insurance status has been associated with later stages of disease at presentation, lower rates of surgical or radiation therapy, and higher rates of cancer-specific mortality.^{10–13, 15} The impact of insurance status on leukemia outcomes has been less clear, in part because there is less information about treatment and disease characteristics in many population-based datasets. In one study of younger patients with AML, being uninsured or having Medicaid was associated with worsened survival ¹³. However, in other studies of myelodysplastic syndromes, acute lymphoblastic leukemia, and acute myeloid leukemia, there was not a clear association between insurance status and outcomes.^{14, 15}

One challenge with assessing the impact of disparities on acute leukemias is the heterogeneity of these diseases, which has substantial influence on treatment response and patient survival. Chronic phase CML, in contrast, is a relatively uniform disease, characterized and defined by the Philadelphia chromosome, with highly effective therapies available to those who can afford it either on their own or through insurance or assistance programs. Therefore, insurance coverage is an important consideration in their treatment, and variations in insurance plans or being uninsured may have a particularly high impact on patient outcomes given the high cost of therapy. Indeed, there are concerns that the cost of TKI therapies for CML is unsustainable, and may threaten access to effective treatment and the stability of our national health care systems in general.⁹ A generic version of imatinib is now available in the United States, which may provide an alternative agent with similar efficacy,^{27, 28} though it is not yet known whether this will improve access to such agents.

It is also not clear whether insurance status itself influences patient outcomes directly, for instance, through maintaining access to effective medications, physician visits, and general medical care and follow-up, or whether insurance status is a proxy for other variables that influence patient outcomes. In particular, patients with Medicaid may have other conditions that contribute to worsened survival; however, some studies suggest that uninsured persons not enrolled in Medicaid may actually have fewer comorbidities.²⁹ We attempted to control for some aspect of the latter by including several county-level socioeconomic metrics in our analysis, as well as state-level Medicaid copayment rates at the end of the study period. These variables did not appear to have a significant impact on patient outcomes compared to insurance status itself. Insurance coverage varies between states,³⁰ but patient residence itself, as defined by SEER registry, did not appear to significantly impact survival when accounting for other variables including insurance status. Nonetheless, the county or registry-level metrics in this analysis are indirect measurements; we did not have access to individual level SES factors which could play a role in the findings we report here.

The impact of marital status on patient outcomes has been noted in other cancers,^{15, 31–34} though this has not been well described in leukemias, including CML, to date. Our results also support the finding that being married at the time of a cancer diagnosis is associated with improvement in survival. This was seen both among younger patients, where married individuals had improved survival compared to those who were single, and among older patients, where married individuals had improved survival compared to those who were divorced, widowed, or separated. The impact of marital status may relate to enhanced social support during and after cancer therapy,^{31, 35} treatment adherence,^{36, 37} and may also impact individual access to high quality care and delays in presentation.^{12, 38, 39}

Our findings should also be interpreted within the context of the SEER registry, which, although robust and frequently used in population-based research, has some limitations. SEER is comprehensive within its regions and representative of the US population, but insurance plans themselves may differ widely between states, and it is possible that our findings may not apply to other states with different insurance models. Moreover, this study population predates the ACA, and it is not known whether changes in insurance access would alter our findings.^40–42 This will be an area of future investigation once more data is available, given data showing that out-of-pocket cancer treatment expenses have declined with the passage of the ACA.⁴³ SEER is limited in that specific chemotherapy information is not available, including oral chemotherapies. We also do not have specific data on adherence to any such treatment regimens. Lastly, SEER does not provide information on the phase of CML at presentation; although most patients with CML present with chronic phase disease, it is possible that delays in or diminished access to care lead to more uninsured patients presenting with accelerated or blast phase disease; however, this would not necessarily explain the similar results seen in the Medicaid population. In one analysis of 149 CML patients from an urban hospital with an underserved population, 96% of patients nonetheless presented with chronic phase disease.⁴⁴

The example of chronic phase CML is the paragon for highly effective targeted cancer treatment, and yet our findings suggest that such favorable outcomes may not be optimally realized in certain patient populations, in part related to their insurance coverage. This is particularly notable because it is younger patients for whom the impact appears to be greatest. Although CML affects relatively few patients each year, challenges in treating this patient cohort highlight barriers to providing effective care to the uninsured and underinsured that may be relevant across a broad spectrum of cancers. In particular, as more targeted, efficacious, and also more costly cancer therapies are developed for various malignancies, it will be increasingly important to ensure all patients can access highly active therapies. Therefore, when treating patients with CML, as well as other highly treatable cancers, resources may be needed to ensure access to treatment, and to devise better strategies for suboptimal responses.