Abstract
Hypomethylating agents (HMA) showed overall survival (OS) benefits in patients with higher-risk myelodysplastic syndromes (HR-MDS) in clinical trials. We conducted a retrospective cohort study of Surveillance, Epidemiology, and End Results (SEER)-Medicare data of patients ≥66 years diagnosed with refractory anemia with excess blasts (RAEB), a proxy for HR-MDS, in 01/2001–04/2004 (pre-period) or 01/2006–12/2011 (post-period). Association between post-period diagnosis and OS was examined using propensity scores (PS)-matched samples.
Among 1,876 RAEB patients, median OS was 9 months and 30.8% received HMAs (3.6% in pre-period; 43.0% in post-period) with no association between post-period diagnosis and OS. In the top PS quartile, post-period diagnosis was associated with a 74% lower risk of death (Hazard ratio [HR]=0.26, 95%-CI: 0.10–0.69, P=0.007), while outcomes were worse in the lowest PS quartile (HR=2.80, 95%-CI: 1.06–7.36, P=0.037).
HMA lead to a 3-month OS benefit for patients most likely to receive HMA but not for unselected RAEB cohort.
Keywords: Myelodysplastic syndromes, survival, effectiveness, hypomethylating agents (HMA), azacitidine, SEER-Medicare
Introduction:
Therapeutic options for patients in the United States (US) with higher risk myelodysplastic syndrome (HR-MDS) expanded considerably with the approval of the two hypomethylating agents (HMAs) azacitidine (AZA) and decitabine (DEC) in 2004 and 2006, respectively.1–3 AZA was subsequently shown to prolong overall survival (OS) by a median of 9.5 months when compared to conventional care among HR-MDS patients.2 As such, it would be expected that OS of HR-MDS patients would have improved in the post-HMA era. Surprisingly, recent registry studies from Spain, the Mayo Clinic in the USA, and the US National Cancer Database reported no improvement in OS in HR-MDS patients over the last 15 years despite the increasing use of HMAs over time.4–6 A study using SEER-Medicare data comparing OS rates for all patients with MDS before (2001–2003) and after the approval of AZA and DEC (2007–2010) showed that the OS for MDS patients had not improved substantially.7 When analysis was restricted to patients with refractory anemia with excess blasts (RAEB), a proxy for HR-MDS, an improvement in OS of only 3 months was observed.7 In fact, several studies in the real-world setting showed a median OS for HR-MDS patients treated with HMA of 11–17 months.4,8,9 This suggests that the median OS of 24.5 months observed with AZA in the AZA-001 trial might have reflected a highly selected sample of clinical trial participants and is not generalizable to routine daily practice. Alternatively, the limited benefit of HMA on a population level might be due to a potential underuse of HMA therapy in patients with HR-MDS which might dilute the observed population-level benefits of HMA and underestimate the potential benefits to individual users.10
One of the challenges of observational studies of the effects of medical treatments is the potential for selection bias. In the case of HR-MDS, clinicians may use information on disease severity, other clinical indicators, and/or patient characteristics related to healthcare access or social support, to assess which patients are most likely to adhere to and benefit from HMA therapy. As a result, estimates of the HMA treatment effects may be biased by selection of patients who would have better OS even in the absence of HMA treatment. To limit the potential for selection bias, we took advantage of the natural experiment of approval and market entry of HMAs in the US (2004–2006). We estimated an HMA treatment propensity model in the post-HMA period which was used to generate treatment propensity scores for the whole sample and matched samples across the pre-HMA and post-HMA periods. We stratified further to identify the OS benefit of HMA availability for RAEB patients most likely and least likely to receive HMA treatment. We hypothesized that the HMA benefit would be greatest for those with the highest treatment propensity, with little effect for those least likely to receive treatment.
Methods
Data Source and Study Sample
We conducted a retrospective cohort study using the Surveillance, Epidemiology and End Results (SEER)-Medicare database, which links two large population-based sources of data to Medicare beneficiaries with cancer. The SEER database captures clinical, demographic, and cause of death information on a longitudinal cancer patient cohort from registries in 18 population-based registries in the US. These data are linked to Medicare enrollment and claims files, which contain detailed individual-level information on covered health care services from the beginning of Medicare eligibility until death or the latest available claims.
Our study cohort included Medicare beneficiaries diagnosed with RAEB (International Classification of Disease for Oncology, 3rd edition code 9983) in January 2001-April 2004 (pre-period) or January 2006-December 2011 (post-period). Diagnosis date was assigned as the first day of the SEER diagnosis month. Other inclusion criteria were: 1) age ≥66 years at diagnosis; and 2) continuous enrollment in Medicare Parts A and B, and no Medicare Advantage enrollment from 12 months before diagnosis until death or end of follow-up on 12/31/2011 (diagnosis in 2001–2003) or 12/31/2013 (diagnosis in 2004–2011). We excluded patients if they had a previous diagnosis of MDS, their diagnosis was reported from death certificate or autopsy only, had a missing diagnosis date, or survived less than 28 days (minimum required to capture at least one HMA cycle). Among those with any HMA treatment, we further excluded patients receiving HMA before their cancer diagnosis or with <3 days of HMA in their first treatment cycle. Detailed sample selection criteria are shown in Figure 1. This study was determined to be “Not human subjects research” by Yale’s Human Investigations Committee.
Figure 1: Illustration of the patient selection process for the study cohort.
HMA – hypomethylating agent, MDS – myelodysplastic syndrome, RAEB – refractory anemia with excess blasts
Key Variables
The study outcome was OS as calculated from 28 days after diagnosis (to accommodate the uncertainty in diagnosis day within the month) until death or censoring at the end of the study period. HMA treatment was defined by the presence of HMA claims (AZA or DEC) after diagnosis, reflecting at least 3 days of treatment within a 28-day period. Supportive care, including erythropoietin stimulating agents (ESAs) was assessed from claims 12 months before through the diagnosis date; allogeneic stem cell transplantation (alloSCT) was assessed during the entire study period. As disease-specific measures of MDS severity, such as cytogenetics and blood/bone marrow blast proportions, were not available in the dataset, we used transfusion receipt and hospitalization for bleeding or infections as proxy indicators for greater disease severity. Transfusions of red blood cells (RBC) and platelets were identified from claims eight weeks prior to four weeks after diagnosis. We recorded claims for hospitalizations with a diagnosis of bleeding or infection from six months prior, through one month after diagnosis month (eight months total). All treatments and therapies were identified using the Healthcare Common Procedural Coding System (HCPCS) (Supplementary table 1).
Other covariates included patient health status, as assessed by the number of comorbid conditions within 12 months before the diagnosis date based on the approach developed by Elixhauser et al.,11 and demographic characteristics. We required the relevant diagnosis codes to be present in either a single inpatient claim or ≥ 2 physician or outpatient claims, with at least two that were ≥30 days apart for the comorbidity to be considered present. Health status also included disability status, a claims-based proxy for performance status, which was assessed using claims one year prior to diagnosis.12,13 Patient characteristics included age, sex, race/ethnicity, marital status, metropolitan statistical area (MSA) residence, and Medicaid dual enrollment. Median household income and percentage of adults with high school education or less that were linked at the census tract level were used as proxy for a patient’s socioeconomic status.
Statistical Analysis
We described sample characteristics (distribution within categorical variables, and median and interquartile range (IQR) of continuous variables) for the sample overall and stratified by the pre-HMA and post-HMA periods. Chi-square test and Wilcoxon Rank-Sum test were used to compare the distribution of categorical and continuous variables, respectively. To generate treatment propensity scores (PS) we estimated a multivariable logistic regression model among RAEB patients diagnosed in the post-HMA approval period. The model included demographic characteristics (age, race/ethnicity, sex, marital status, MSA residence, Medicaid dual enrollment, area median income and educational levels) and health status (comorbidity, disability status). To improve predictive ability, we allowed all possible interactions between the model covariates and used backward selection, keeping all variables with a p≤0.40. The final model had a C-index=0.790 and a non-significant Hosmer-Lemeshow goodness-of-fit test statistic (p=0.954). The estimated coefficients were used to generate PS in both the pre-HMA and post-HMA subgroups, and matching patients from the two samples in a 1:2 ratio. We estimated Cox proportional hazard models that included a post- versus pre-HMA indicator. After testing for and rejecting the proportional hazards assumption, we added a post-HMA*time interaction term. In sensitivity analyses we censored all observations at 2 years, 1 year, and 6 months, to further examine the effect of treatment over time. The matched sample was stratified into PS quartiles, reflecting the likelihood of treatment in the post-period. This process is illustrated in Figure 2. Primary analysis of the effect of HMA was conducted by comparing survival between pre and post-period patients who fell within the top PS quartile. We estimated a parallel analysis on the bottom PS quartile. We used SAS version 9.4 (SAS Institute, Inc., Cary, NC) to conduct all analyses, with two-sided statistical tests and an alpha of 0.05.
Figure 2:
Illustration of Propensity Score Matching and Stratification Process
Results
We identified 1,876 patients with a median age of 79 years (IQR: 73–84) with 89.7% being white, 59.5% male and 3.6% Hispanic (Table 1). Ninety-six percent of the sample died during the follow-up period. The median follow-up time from 28 days after diagnosis was 9 months (IQR: 4–21), and the median OS was 9 months (95% CI: 8–10). Five hundred and seventy-eight patients (30.8%) received HMAs during the follow-up period. Among patients diagnosed in the pre-approval period, only 3.6% received HMA treatment, which increased to 43.0% in the post-approval period. The median duration from diagnosis to HMA initiation was 1 month (IQR: 1–4). Four percent of patients received platelet transfusion, 10.4% received RBC transfusion, and 32.0% had a prior non-MDS cancer diagnosis. Only 19 patients (1.0%) received alloSCT during follow-up. (Table 1 and Supplementary table 2).
Table 1.
Sample characteristics overall and by pre-post HMA period, original and matched samples
| Original Sample | Propensity Matched Sample | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Overall | Pre-Period | Post-Period | P-value | Standardized Difference | Overall | Pre-Period | Post-Period | P-value | Standardized Difference | |
| % | % | % | % | % | ||||||
| N | 1876 | 581 | 1295 | 9.48 | 1684 | 570 | 1114 | 1.97 | ||
| HMA | <.001 | |||||||||
| No | 69.2 | 96.4 | 57.0 | 68.9 | 96.3 | 54.9 | ||||
| Yes | 30.8 | 3.6 | 43.0 | 31.1 | 3.7 | 45.1 | ||||
| MDS Diagnosis Year | ||||||||||
| 2001 | 8.3 | 26.9 | 9.0 | 26.5 | ||||||
| 2002 | 8.1 | 26.2 | 8.9 | 26.3 | ||||||
| 2003 | 10.9 | 35.3 | 11.9 | 35.3 | ||||||
| 2004 | 3.6 | 11.7 | 4.0 | 11.9 | ||||||
| 2005 | ||||||||||
| 2006 | 12.0 | 17.5 | 11.1 | 16.8 | ||||||
| 2007 | 11.4 | 16.5 | 10.8 | 16.3 | ||||||
| 2008 | 11.6 | 16.8 | 11.0 | 16.7 | ||||||
| 2009 | 12.3 | 17.8 | 11.6 | 17.5 | ||||||
| 2010 | 11.1 | 16.1 | 11.0 | 16.7 | ||||||
| 2011 | 10.6 | 15.3 | 10.6 | 16.0 | ||||||
| Age Group | .51 | .67 | ||||||||
| 66–69 | 10.8 | 10.8 | 10.8 | 0.10 | 10.7 | 10.4 | 11.0 | −1.947 | ||
| 70–74 | 19.3 | 21.2 | 18.5 | 6.81 | 19.7 | 21.1 | 18.9 | 5.28 | ||
| 75–79 | 22.9 | 21.5 | 23.6 | −4.88 | 22.8 | 21.6 | 23.4 | −4.431 | ||
| 80+ | 47.0 | 46.5 | 47.2 | −1.42 | 46.8 | 47.0 | 46.7 | 0.68 | ||
| Sex | .83 | .90 | ||||||||
| Female | 40.5 | 40.1 | 40.6 | −1.05 | 40.2 | 40.0 | 40.3 | −0.62 | ||
| Male | 59.5 | 59.9 | 59.4 | 1.05 | 59.8 | 60.0 | 59.7 | 0.62 | ||
| Race | .098 | .16 | ||||||||
| White | 89.7 | 91.4 | 88.9 | 8.44 | 90.0 | 91.4 | 89.2 | 7.36 | ||
| Other | 10.3 | 8.6 | 11.1 | −8.44 | 10.0 | 8.6 | 10.8 | −7.36 | ||
| Hispanic | .64 | .57 | ||||||||
| Non-hispanic | 96.4 | 96.7 | 96.3 | 2.38 | 96.9 | 97.2 | 96.7 | 2.98 | ||
| Hispanic | 3.6 | 3.3 | 3.7 | −2.38 | 3.1 | 2.8 | 3.3 | −2.98 | ||
| Marital Status | .95 | .90 | ||||||||
| Married | 55.7 | 55.4 | 55.8 | −0.67 | 55.8 | 55.4 | 55.9 | −0.98 | ||
| Unmarried | 36.7 | 37.2 | 36.5 | 1.35 | 36.6 | 37.2 | 36.3 | 1.92 | ||
| Other | 7.6 | 7.4 | 7.7 | −1.21 | 7.7 | 7.4 | 7.8 | −1.67 | ||
| % Adults w/ < HS education | .71 | .29 | ||||||||
| <33% | 31.9 | 32.0 | 31.8 | 0.43 | 32.3 | 32.5 | 32.2 | 0.49 | ||
| 33%-66% | 55.0 | 55.8 | 54.6 | 2.36 | 54.3 | 56.0 | 53.5 | 4.95 | ||
| ≥ 66% | 13.2 | 12.2 | 13.6 | −4.09 | 13.4 | 11.6 | 14.3 | −8.04 | ||
| Income | .34 | .29 | ||||||||
| <$33,000 | 20.1 | 22.2 | 19.2 | 7.54 | 20.0 | 22.1 | 18.9 | 7.84 | ||
| $33,000–40,000 | 14.1 | 14.6 | 13.8 | 2.31 | 14.1 | 14.6 | 13.9 | 1.85 | ||
| $40,000–50,000 | 21.0 | 21.0 | 21.0 | −0.01 | 20.8 | 21.2 | 20.6 | 1.65 | ||
| ≥ $50,000 | 44.8 | 42.2 | 46.0 | −7.77 | 45.1 | 42.1 | 46.6 | −9.03 | ||
| MSA Status | .83 | .94 | ||||||||
| MSA | 83.6 | 83.3 | 83.7 | −1.08 | 83.4 | 83.3 | 83.5 | −0.40 | ||
| Non-MSA | 16.4 | 16.7 | 16.3 | 1.08 | 16.6 | 16.7 | 16.5 | 0.40 | ||
| Medicaid Dual Enrollment | .063 | .19 | ||||||||
| No | 88.1 | 90.2 | 87.2 | 9.51 | 89.0 | 90.4 | 88.2 | 6.83 | ||
| Yes | 11.9 | 9.8 | 12.8 | −9.51 | 11.0 | 9.6 | 11.8 | −6.83 | ||
| Elixhauser Comorbidity Index | <.001 | .002 | ||||||||
| None | 31.2 | 38.2 | 28.0 | 21.76 | 32.7 | 38.4 | 29.8 | 18.25 | ||
| 1 to 2 | 39.1 | 36.3 | 40.3 | −8.22 | 39.1 | 36.3 | 40.5 | −8.58 | ||
| more than 3 | 29.7 | 25.5 | 31.7 | −13.73 | 28.2 | 25.3 | 29.7 | −9.98 | ||
| Disability | .17 | .58 | ||||||||
| Not disabled | 88.9 | 90.4 | 88.2 | 7.04 | 89.8 | 90.4 | 89.5 | 2.84 | ||
| Disabled | 11.1 | 9.6 | 11.8 | −7.04 | 10.2 | 9.6 | 10.5 | −2.84 | ||
| Hospitalization for infection or bleeding | .099 | .17 | ||||||||
| No | 79.6 | 81.9 | 78.6 | 8.34 | 80.4 | 82.3 | 79.4 | 7.22 | ||
| Yes | 20.4 | 18.1 | 21.4 | −8.34 | 19.6 | 17.7 | 20.6 | −7.22 | ||
| RBC transfusion | .20 | .14 | ||||||||
| Naïve | 89.6 | 90.0 | 89.3 | 2.21 | 89.8 | 90.4 | 89.6 | 2.54 | ||
| User | 7.7 | 6.5 | 8.3 | −6.58 | 7.6 | 6.3 | 8.3 | −7.48 | ||
| Dependent | 2.7 | 3.4 | 2.4 | 6.23 | 2.6 | 3.3 | 2.2 | 7.22 | ||
| Prior Platelet transfusion | .036 | .061 | ||||||||
| No | 96.0 | 97.4 | 95.4 | 11.02 | 96.6 | 97.7 | 96.0 | 10.07 | ||
| Yes | 4.0 | 2.6 | 4.6 | −11.02 | 3.4 | 2.3 | 4.0 | −10.07 | ||
| Prior non-MDS cancer diagnosis | <.001 | .004 | ||||||||
| No | 68.0 | 73.8 | 65.4 | 18.41 | 69.0 | 73.5 | 66.7 | 14.92 | ||
| Yes | 32.0 | 26.2 | 34.6 | −18.41 | 31.0 | 26.5 | 33.3 | −14.92 | ||
| ESA use prior to first HMA | .13 | .11 | ||||||||
| No | 89.4 | 87.8 | 90.1 | 89.9 | 88.2 | 90.8 | ||||
| Yes | 10.6 | 12.2 | 9.9 | 10.1 | 11.8 | 9.2 | ||||
| Bone marrow transplantation | .15 | .20 | ||||||||
| No | >98.6 | >98.1 | 98.8 | >98.5 | >98.1 | 98.7 | ||||
| Yes | <1.4 | <1.9 | 1.2 | <1.5 | <1.9 | 1.3 | ||||
Propensity Score Matching
The treatment PS model among post-approval patients included all demographic characteristics, health status, and disease severity measures. After the PS matching, 570 out of the 581 pre-period and 1114 out of 1295 post-period patients were matched (total matched=1684). The sum of standardized differences suggested that the overall distribution of covariates between these two samples was balanced (Table 1). Because Elixhauser Comorbidity Index and proportion with prior non-MDS cancer diagnosis remained unbalanced after matching (p-value<0.05), all survival models included controls for these two variables.
Matched patients were stratified into PS quartiles (Table 2). Among patients in the top PS quartile (PS>0.62, n=423), 77.1% of those in the post-period group received HMA, compared with <7% in the pre-period cohort. In contrast, only 12.6% of patients in the bottom PS quartile received HMA treatment in the post-period.
Table 2.
HMA Treatment Receipt Overall and by Propensity Score Quartile
| Treatment Propensity | ||||||
|---|---|---|---|---|---|---|
| N | Overall | Q1 (Bottom) | Q2 | Q3 | Q4 (Top) | |
| N | 1684 | 1684 | 421 | 422 | 418 | 423 |
| Pre-HMA | 570 | 3.7 | <7.0 | <7.0 | <7.0 | <7.0 |
| Post-HMA | 1114 | 45.1 | 12.6 | 37.9 | 52.7 | 77.1 |
| Total | 31.1 | * | * | * | * | |
Cell percentages suppressed per Centers for Medicare & Medicaid Services restrictions on reporting. Abbreviations: HMA, hypomethylating agents.
Survival
Kaplan-Meier analyses indicated no difference in OS across treatment quartiles in the pre-period (Supplementary Figure 1). For the entire PS-matched patient cohort, being in the post-period did not affect OS compared to being in the pre-period (adjusted Hazard Ratio [aHR]=1.00, 95% confidence interval [CI]=0.64–1.59) or with treatment propensity (aHR=0.69, 95%CI=0.18–2.65) [Table 3]. However, within the top PS quartile, patients diagnosed in the post-period had a 74% lower risk of death (aHR=0.26, 95% CI=0.10–0.69, p=0.007). This survival benefit diminished over time given the significant time interaction effect (aHR=1.22, 95% CI=1.02–1.47, p=0.033). While comorbidity was not significantly associated with survival, patients with a prior non-MDS cancer diagnosis showed a 1.49 times increased risk of death (95%CI=1.01–2.20, p=0.044). Results from the parallel models on the bottom PS quartile group indicated significantly higher risks of death compared to the entire study cohort (aHR=2.80, 95%CI=1.06–7.36, p=0.037) and this effect was consistent over time (time interaction aHR=0.83, 95%CI=0.68–1.03, p=0.087). Neither comorbidity nor prior non-MDS cancer diagnosis was associated with survival in this subgroup. Results from the sensitivity analyses are consistent with the time-dependent nature of the treatment effect, with larger effects with shorter follow-up time (Supplementary Table 3).
Table 3.
Estimated Associations between Pre-Post HMA Period and Overall Survival
| Overall | Top PS Quartile | Bottom PS Quartile | ||||||
|---|---|---|---|---|---|---|---|---|
| aHR (95% CI) | P-value | aHR (95% CI) | P-value | aHR (95% CI) | P-value | |||
| Post vs. Pre | 1.00 (0.64–1.59) | .99 | 0.26 (0.10–0.69) | .007 | 2.80 (1.06–7.36) | .037 | ||
| Post vs. Pre * Time | 0.99 (0.90–1.09) | .82 | 1.22 (1.02–1.47) | .033 | 0.83 (0.68–1.03) | .087 | ||
| PS logit | 0.69 (0.18–2.65) | .59 | ||||||
| Comorbidity | ||||||||
| None | Ref | Ref | Ref | |||||
| 1 to 2 | 0.98 (0.80–1.20) | .83 | 1.02 (0.66–1.58) | .93 | 1.14 (0.74–1.75) | .55 | ||
| 3 or more | 1.26 (0.99–1.59) | .063 | 1.54 (0.90–2.63) | .12 | 1.33 (0.82–2.16) | .25 | ||
| Prior non-MDS cancer diagnosis | ||||||||
| No | Ref | Ref | Ref | |||||
| Yes | 1.15 (0.96–1.37) | .14 | 1.49 (1.01–2.20) | .044 | 1.15 (0.83–1.61) | .40 | ||
Abbreviations: HMA, hypomethylating agents; PS, propensity score; aHR, adjusted hazard ratio; ref, reference; MDS, myelodysplastic syndromes.
Discussion
In this large, retrospective population-based study of patients with RAEB, we did not find an OS benefit in the entire patient cohort following the introduction of HMA. However, we observed a statistically significant OS advantage for patients most likely to receive HMA treatment in a PS matched analysis. Within the high treatment propensity quartile, patients diagnosed in the post-HMA period had an initial 74% lower risk of death when compared with patients diagnosed in the pre-period, although the benefit of HMAs diminished over time. In contrast, those least likely to receive HMAs fared much worse in the post-period compared to the pre-period, and the effect was not time-dependent. The diminishing therapeutic effect of HMAs has been documented in several studies previously and HMA failure develops in the majority of patients within 2 years which is also reflected by the converging survival curves in our analysis.14,15 While HMAs are not a curative treatment modality in MDS, the prolonged survival, delay of progression to acute myeloid leukemia, and the alleviation of symptoms related to bone marrow failure and supportive blood transfusions, are all valuable therapeutic goals in this elderly patient population.
The results of this study differ from prior observational studies comparing OS for MDS patients diagnosed in the pre- and post-HMA period. In a Spanish registry study of 821 HR-MDS patients ineligible for alloSCT, 30.6% of patients were treated with AZA but did not gain a survival benefit compared to conventional care (supportive care only, conventional chemotherapy, lenalidomide, immunosuppression).4 In US Medicare patients, Zeidan et al. showed that RAEB patients in the post-HMA era had a 3-month OS benefit and lower risk of death (Hazards ratio = 0.88, 95% CI: 0.80–0.96; p = 0.01) compared to the pre-HMA period.7 Benefits in terms of hematologic improvement and transfusion needs have also been reported in an US-based, community-based, prospective registry of both lower- and higher-risk MDS patients.16 However, in these studies the proportion of MDS patients receiving HMA is low, even those with higher risk disease, raising concern that undertreatment may dilute the observed benefit from HMAs. This is further supported by a physician survey in the US showing that only 18% of newly-diagnosed MDS patients received HMAs.17 Our study also confirms the low uptake of HMA therapy in RAEB patients following their market entry. More recently, Ma et al. conducted a retrospective review of an electronic medical records database of 5162 MDS patients and showed that only 12.1% of all MDS patients received HMAs as their first line of treatment with younger patients being more likely to be treated with HMAs.18 We therefore focused in this study on the subgroup with the highest PS scores and found that over 77% of the post-period group received HMAs. While the treatment effect may still be diluted to some extent, as not all patients received HMAs, the statistical signal is much stronger than in earlier studies. One concern that arises with these results is whether physicians are selecting patients for HMA treatment who are healthier and would have better OS compared to the patients in the lower PS quartiles independent of HMA treatment. In this case, we would expect OS to be better for patients diagnosed in the pre-period who are in the top compared to other PS quartiles. However, this is not supported by our analysis. Hence, we conclude that HMAs had a positive OS benefit for this group of RAEB patients.
While we expected patients in the top PS quartile to have improved OS in the post-period, our result regarding patients in the bottom PS quartile, showing substantially worse survival in the post-period, was unexpected. As noted previously, OS in this group did not differ from those in the other quartiles in the pre-period, and only 12.6% received HMAs in the post-period. In an ad hoc analysis we compared OS in the pre-post period after dropping the pairs where one member received HMAs and found no change in the results. Hence, there is no evidence that HMAs accelerated death in those low-propensity patients who received it. The reasons for this finding are speculative especially as there was no difference in the rate of comorbidities and prior malignancies between the top and bottom PS quartile.
Our analytic approach, a hybrid of propensity score and instrumental variable (IV) analysis, was crafted to account for an environment where there was increased registry reporting of MDS that may have altered the patient mix in the post-period compared to the pre-period, while taking advantage of the natural experiment of HMA market entry. The PS matching across time periods addressed the former issue, in that pre-post pairs are matched based on observable characteristics that affected treatment likelihood. By focusing on the top PS quartile, the patient group most likely to receive treatment, we paralleled the IV estimate of local area treatment effect – relevant to those most likely to be influenced by the “instrument,” diagnosis during the post-period.19,20
Despite our findings of a significant OS benefit associated with HMAs for patients with RAEB, the magnitude of effect is small especially when compared to the AZA-001 trial.2 However, the 9.5 months OS benefit and the median OS of 24.5 months have not been seen in various subsequent clinical trials with HMAs and real-world analyses.4,21–23 With a median survival of 335 days in the pre-period and 429 days in the post-period, the incremental OS was 94 days, slightly more than 3 months, which is in line with prior studies in patients with RAEB.1,7 Potential explanations for the modest response rates seen in our study are the low rate of patients in our study who proceeded to allogeneic stem cell transplant (1%), the high rate of secondary and therapy-related MDS, and the advanced median age, all of which have been shown to be an adverse prognostic factor in MDS.24–26 It is also important to note that RAEB and HR-MDS should not be used interchangeably although RAEB has previously been used as a proxy for HR-MDS in claims-based studies. Therefore, differences in the patient population in our study compared to other studies could have contributed.2,4,21–23 It is also possible that the use of DEC, which has not been shown to improve OS over best supportive care in randomized clinical trials, might have diluted the magnitude of the OS benefit we observed with HMAs. While it remains controversial whether the lack of documented OS in DEC trials in contrast to the AZA trials is a biological or a clinical trial design-related phenomenon, multiple laboratory and clinical studies including our own indirect comparison of the two agents indicate that the differences in impact of the two HMAs on OS are minimal.7,23,27 Indeed, many experts and clinical practice guidelines consider the two agents equivalent and recommend both of them for use in this setting. Furthermore, we previously reported that DEC was used as the first HMA in only 22% of older SEER-Medicare beneficiaries with HR-MDS who received an HMA while the vast majority (78%) used AZA as the first agent.23
A related concern is whether alternative and suboptimal schedules of AZA that are commonly used in the USA (e.g. 5-day schedules or lower dosing) might have diluted the OS benefit with the drug at the population level compared to the only schedule that has shown an OS advantage in clinical trials (AZA at 75mg/m2/day for 7 consecutive days each 28-day cycle). We did not account for this issue directly. However, several registry reports from Europe have shown significantly shorter OS with AZA (13–17 months) among patients with HR-MDS even when used on the approved 7-day schedule in most patients, suggesting that this is not a major contributor to this observation.4,8 Lastly, provider experience using agents like HMA and the small volumes of patients per community provider for a rare cancer like MDS might all impact the ability to translate the full benefit of these agents at the population level. We however analyzed this issue in a recent study and found that the provider experience with HMAs and their volume of MDS patients were not significantly associated with the OS achieved with the use of these agents.28
Despite the large sample size of RAEB patients, and the robust study design, our study is subject to other limitations. As in other claims-based studies examining treatment with HMAs,6,7,23 we relied on the WHO histologic classifications to identify RAEB patients which is not equivalent to HR-MDS. We lacked clinical markers such as blast count or hemoglobin levels to confirm the HR diagnosis based on the International Prognostic Scoring System (IPSS), instead using diagnoses and procedures on the claims to construct proxy measures for severity of illness. We defined HMA treatment by the presence of claims for AZA or DEC with at least 3 days of treatment. As responses to HMA therapy can take several cycles to develop, premature discontinuation of HMAs might have been contributing to the modest OS benefit seen in our study.28–30 We acknowledge that there may be other unobserved factors that could influence both treatment and outcomes. It is important to note, that our study only included patients with RAEB and results may not be generalizable to patients with other MDS subtypes, who may experience a different benefit from HMA therapy. Finally, our study methods cannot explain the unexpected finding that patients in the bottom PS quartile fared substantially worse in the post-period compared to the pre-period. A better understanding of these issues is vital for optimal HMA patient selection and administration.
Conclusion
The increase in HMA treatment from 3.6% of patients diagnosed in the pre-period to 43.0% in the post-period did not lead to an OS benefit in the full cohort of patients with RAEB. However, in the top PS quartile, the patient subgroup most likely to receive HMA, we observed a 75% lower risk of death and a 3-month OS improvement in the post-HMA era. These observations support that the OS advantage associated with HMAs are very limited and provide strong rationale for continued clinical trial efforts in the frontline management of patients with RAEB to improve upon the current standard of HMA monotherapy.31
Supplementary Material
Acknowledgements:
AMZ is a Leukemia and Lymphoma Society Scholar in Clinical Research and is also supported by a NCI’s Cancer Clinical Investigator Team Leadership Award (CCITLA). This research was partly funded by the Dennis Cooper Hematology Young Investigator Award (AZ), and was in part supported by the National Cancer Institute of the National Institutes of Health under Award Number P30 CA016359. The collection of the California cancer incidence data used in this study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute (NCI)’s Surveillance, Epidemiology and End Results (SEER) Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention’s National Program of Cancer Registries, under agreement #U55/CCR921930-02 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the author(s) and endorsement by the State of California, Department of Public Health, the National Cancer Institute, and the Centers for Disease Control and Prevention or their Contractors and Subcontractors is not intended nor should be inferred. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare and Medicaid Services; Information Management Services, Inc.; and the SEER Program tumor registries in the creation of the SEER-Medicare database. The interpretation and reporting of the SEER-Medicare data are the sole responsibility of the authors.
Declaration of conflicts of interest: AJD reports grants from Celgene during the conduct of the study; and a family member with personal fees from Abbvie, Jazz Pharmaceuticals, Kyowa Hakko Kirin, Tolero Pharmaceuticals, and Daiichi Sankyo to family members outside of the submitted work. RW received research funding from Celgene Corp. NAP received research funding (institutional) from Boehringer Ingelheim, Astellas Pharma, Daiichi Sankyo, Sunesis Pharmaceuticals, Celator, Pfizer, Astex Pharmaceuticals, CTI BioPharma, Genentech, AI Therapeutics, Samus Therapeutics, Arog Pharmaceuticals and Kartos Therapeutics. NAP received research funding from Celgene. NAP had a consultancy with and received honoraria from Alexion, Pfizer, Agios Pharmaceuticals, Blueprint Medicines, Incyte, Novartis and Celgene. SFH received research funding (institutional) from Celgene, TG Therapuetics, DTRM, Genentech. SFH reports personal fees from Celgene, personal fees from Pharmacyclics, personal fees from Genentech, personal fees from Bayer, outside the submitted work; SDG has consulted for and receives research funding from Celgene; personal fees from Abbvie, Jazz Pharmaceuticals, Kyowa Hakko Kirin, Tolero Pharmaceuticals, and Daiichi Sankyo outside of submitted work. XM received research funding from Celgene Corp, which supported the development of this manuscript, and consulted for Celgene and Incyte. AMZ received research funding (institutional) from Celgene, Acceleron, Abbvie, Otsuka, Pfizer, Medimmune/AstraZeneca, Boehringer-Ingelheim, Trovagene, Incyte, Takeda, ADC Therapeutics. AMZ had a consultancy with and received honoraria from AbbVie, Otsuka, Pfizer, Celgene, Ariad, Agios, Boehringer-Ingelheim, Novartis, Acceleron, Astellas, Daiichi Sankyo, Trovagene, BeyondSpring, and Takeda. The other authors have no conflicts of interest to declare.
Footnotes
Data sharing: SEER-Medicare data cannot be shared by the authors as directed by the SEER-Medicare data use agreement. Data may be requested directly from the National Cancer Institute. However, we are open to sharing our methodology and analytical approaches upon request.
References
- 1.Lubbert M, Suciu S, Baila L, et al. Low-dose decitabine versus best supportive care in elderly patients with intermediate- or high-risk myelodysplastic syndrome (MDS) ineligible for intensive chemotherapy: final results of the randomized phase III study of the European Organisation for Research and Treatment of Cancer Leukemia Group and the German MDS Study Group. J. Clin. Oncol. 2011;29(15):1987–1996. [DOI] [PubMed] [Google Scholar]
- 2.Fenaux P, Mufti GJ, Hellstrom-Lindberg E, et al. Efficacy of azacitidine compared with that of conventional care regimens in the treatment of higher-risk myelodysplastic syndromes: a randomised, open-label, phase III study. Lancet Oncol. 2009;10(3):223–232. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kantarjian H, Issa JP, Rosenfeld CS, et al. Decitabine improves patient outcomes in myelodysplastic syndromes: results of a phase III randomized study. Cancer. 2006;106(8):1794–1803. [DOI] [PubMed] [Google Scholar]
- 4.Bernal T, Martinez-Camblor P, Sanchez-Garcia J, et al. Effectiveness of azacitidine in unselected high-risk myelodysplastic syndromes: results from the Spanish registry. Leukemia. 2015;29(9):1875–1881. [DOI] [PubMed] [Google Scholar]
- 5.Gangat N, Patnaik MM, Begna K, et al. Survival trends in primary myelodysplastic syndromes: a comparative analysis of 1000 patients by year of diagnosis and treatment. Blood Cancer J. 2016;6(4):e414–e414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Al-Kali AZ D; Bartley AC et al. Outcome of Myelodysplastic Syndromes over Time in the US: A National Cancer Data Base Study from 2004–2013. Blood. 2016;128:3604. [DOI] [PubMed] [Google Scholar]
- 7.Zeidan AM, Wang R, Gross CP, et al. Modest improvement in survival of patients with refractory anemia with excess blasts in the hypomethylating agents era in the United States. Leuk Lymphoma. 2017;58(4):982–985. [DOI] [PubMed] [Google Scholar]
- 8.Dinmohamed AG, van Norden Y, Visser O, et al. Effectiveness of azacitidine for the treatment of higher-risk myelodysplastic syndromes in daily practice: results from the Dutch population-based PHAROS MDS registry. Leukemia. 2015;29:2449. [DOI] [PubMed] [Google Scholar]
- 9.Murthy VH, Krumholz HM, Gross CP. Participation in cancer clinical trials: race-, sex-, and age-based disparities. JAMA. 2004;291(22):2720–2726. [DOI] [PubMed] [Google Scholar]
- 10.Zeidan AM, Wang R, Davidoff AJ, et al. Disease-related costs of care and survival among Medicare-enrolled patients with myelodysplastic syndromes. Cancer. 2016;122(10):1598–1607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med. Care. 1998;36(1):8–27. [DOI] [PubMed] [Google Scholar]
- 12.Davidoff AJ, Gardner LD, Zuckerman IH, Hendrick F, Ke X, Edelman MJ. Validation of disability status, a claims-based measure of functional status for cancer treatment and outcomes studies. Med. Care. 2014;52(6):500–510. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Davidoff AJ, Zuckerman IH, Pandya N, et al. A novel approach to improve health status measurement in observational claims-based studies of cancer treatment and outcomes. J. Geriatr. Oncol. 2013;4(2):157–165. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Zeidan AM, Stahl M, Hu X, et al. Long-term survival of older patients with MDS treated with HMA therapy without subsequent stem cell transplantation. Blood. 2018;131(7):818–821. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Santini V How I treat MDS after hypomethylating agent failure. Blood. 2019;133(6):521–529. [DOI] [PubMed] [Google Scholar]
- 16.Grinblatt DL, Sekeres MA, Komrokji RS, Swern AS, Sullivan KA, Narang M. Patients with myelodysplastic syndromes treated with azacitidine in clinical practice: the AVIDA® registry. Leuk. Lymphoma. 2015;56(4):887–895. [DOI] [PubMed] [Google Scholar]
- 17.Sekeres MA, Schoonen WM, Kantarjian H, et al. Characteristics of US patients with myelodysplastic syndromes: results of six cross-sectional physician surveys. J. Natl. Cancer Inst. 2008;100(21):1542–1551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ma X, Steensma DP, Scott BL, Kiselev P, Sugrue MM, Swern AS. Selection of patients with myelodysplastic syndromes from a large electronic medical records database and a study of the use of disease-modifying therapy in the United States. BMJ Open. 2018;8(7):e019955–e019955. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Harris KM, Remler DK. Who is the marginal patient? Understanding instrumental variables estimates of treatment effects. Health Serv. Res. 1998;33(5 Pt 1):1337–1360. [PMC free article] [PubMed] [Google Scholar]
- 20.Ertefaie A, Small DS, Flory JH, Hennessy S. A tutorial on the use of instrumental variables in pharmacoepidemiology. Pharmacoepidemiol. Drug Saf. 2017;26(4):357–367. [DOI] [PubMed] [Google Scholar]
- 21.Zeidan AM, Sekeres MA, Garcia-Manero G, et al. Comparison of risk stratification tools in predicting outcomes of patients with higher-risk myelodysplastic syndromes treated with azanucleosides. Leukemia. 2016;30(3):649–657. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Sekeres MA, Othus M, List AF, et al. Randomized Phase II Study of Azacitidine Alone or in Combination With Lenalidomide or With Vorinostat in Higher-Risk Myelodysplastic Syndromes and Chronic Myelomonocytic Leukemia: North American Intergroup Study SWOG S1117. J. Clin. Oncol. 2017;35(24):2745–2753. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Zeidan AM, Davidoff AJ, Long JB, et al. Comparative clinical effectiveness of azacitidine versus decitabine in older patients with myelodysplastic syndromes. Br. J. Haematol. 2016;175(5):829–840. [DOI] [PubMed] [Google Scholar]
- 24.Greenberg PL, Stone RM, Al-Kali A, et al. Myelodysplastic Syndromes, Version 2.2017, NCCN Clinical Practice Guidelines in Oncology. J. Natl. Compr. Canc. Netw. 2017;15(1):60–87. [DOI] [PubMed] [Google Scholar]
- 25.Zeidan AM, Shallis RM, Wang R, Davidoff A, Ma X. Epidemiology of myelodysplastic syndromes: Why characterizing the beast is a prerequisite to taming it. Blood Rev. 2019;34:1–15. [DOI] [PubMed] [Google Scholar]
- 26.Benton CB, Khan M, Sallman D, et al. Prognosis of patients with intermediate risk IPSS-R myelodysplastic syndrome indicates variable outcomes and need for models beyond IPSS-R. Am. J. Hematol. 2018;93(10):1245–1253. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zeidan AM, Stahl M, DeVeaux M, et al. Counseling patients with higher-risk MDS regarding survival with azacitidine therapy: are we using realistic estimates? Blood Cancer J. 2018;8(6):55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Zeidan AM, Hu X, Zhu W, et al. Association of provider experience and clinical outcomes in patients with myelodysplastic syndromes receiving hypomethylating agents. Leuk. Lymphoma. 2019:1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Silverman LR, McKenzie DR, Peterson BL, et al. Further analysis of trials with azacitidine in patients with myelodysplastic syndrome: studies 8421, 8921, and 9221 by the Cancer and Leukemia Group B. J. Clin. Oncol. 2006;24(24):3895–3903. [DOI] [PubMed] [Google Scholar]
- 30.Silverman LR, Fenaux P, Mufti GJ, et al. Continued azacitidine therapy beyond time of first response improves quality of response in patients with higher-risk myelodysplastic syndromes. Cancer. 2011;117(12):2697–2702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Zeidan AM, Stahl M, Sekeres MA, Steensma DP, Komrokji RS, Gore SD. A call for action: Increasing enrollment of untreated patients with higher-risk myelodysplastic syndromes in first-line clinical trials. Cancer. 2017;123(19):3662–3672. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.