Epidemiology of myelodysplastic syndromes and chronic myeloproliferative disorders in the United States, 2001-2004, using data from the NAACCR and SEER programs

Myelodysplastic syndromes (MDSs) comprise morphologically distinct disorders characterized by dysplastic and ineffective hematopoiesis. Although historically MDS has not been defined as a cancer, MDS results from the clonal expansion of an hematopoietic progenitor and progresses to acute myeloid leukemia in approximately 30% of patients.¹  Incidence rates for MDS and chronic myeloproliferative disorders (CMDs) in the United States were unavailable prior to the addition of these stem cell malignancies to central cancer registries in 2001. Description of national incidence rates provides an important baseline for future studies of secular trends and allows for the examination of rates by selected demographic factors to define risk profiles of these hematologic malignancies in the American population.

Estimated incidence rates of MDS for the United States in 2001 to 2003 were recently published based on initial data reported from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program.²  Incidence rates increased with age and were higher among males than females,²  although rates were not analyzed for MDS subtypes or for CMD. SEER covers approximately 26% of the US population, and given the recent incorporation of MDS to cancer registry surveillance, we sought to extend national estimates to 33 additional geographic areas that are not included in SEER and assess possible patterns in diagnostic recognition that might affect reporting frequency. To investigate MDS and CMD incidence overall and by disease subtype, evaluate trends in incidence by demographic factors, estimate total numbers of cases expected to be diagnosed in the entire United States, and examine incidence and survival data updated through 2004, we conducted an extensive analysis of data obtained through both SEER and the North American Association of Cancer Registries (NAACCR), based on more than 40 000 patients, the largest number ever available for analysis of MDS and CMD.

Frequencies of reported MDS, CMD, and chronic myelomonocytic leukemia (CMML) cases were obtained from US state and regional population-based cancer registries. (Although CMML is classified as an MDS subtype in the French, American, British [FAB] system, CMML is not coded as MDS in the system by which cancer cases are reported to central cancer registries. Therefore, CMML was included as a disease entity separate from MDS in the present analysis.) These cancer registries collect information on new cancer diagnoses through the National Cancer Institute's SEER Program, Centers for Disease Control Prevention's National Program of Cancer Registries (NPCR) Program, or both.³  All cancer registries are members of NAACCR. The source of data in this paper for all analyses, except survival, is NAACCR's research data file, Cancer in North America (CINA); the December 2005 submission of the CINA Deluxe 1995 to 2003 file was used, and diagnosis years from 2001 to 2003 were included. To be included in the CINA deluxe database, registry data are required to meet specific certification criteria, details of which have been previously described.⁴ 

In the CINA Deluxe database, SEER data are composed of 9 state areas (California, Connecticut, Hawaii, Iowa, New Mexico, Utah, Kentucky, Louisiana, and New Jersey) and 3 metropolitan areas (Atlanta, Detroit, and Seattle). The other state registries in the CINA Deluxe file include Alabama, Alaska, Arizona, Colorado, Delaware, Washington, DC, Florida, Georgia, Idaho, Illinois, Indiana, Maine, Massachusetts, Michigan, Minnesota, Missouri, Montana, Nebraska, Nevada, New Hampshire, New York, North Carolina, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, South Dakota, Texas, Washington, West Virginia, and Wisconsin. Since the SEER metropolitan areas of Atlanta, Detroit, and Seattle are also included in the Georgia, Michigan, and Washington registries, we excluded these 3 metropolitan areas from their respective states, thus avoiding double counting of patients and facilitating the comparison of incidence rates for MDS and CMD across 3 groups: SEER registries, NAACCR excluding SEER registries, and total registries.

Patients diagnosed between 2001 and 2003 were identified by International Classification of Diseases for Oncology Third Edition (ICD-O-3) codes (MDS: 9980-9989; CMD: 9950-9964; and CMML: 9945). ICD-O-3 includes codes for both topography and morphology and was implemented for data collection in cancer registries worldwide in 2001.⁵  Although ICD-O-3 was developed by the World Health Organization (WHO), there are subtle differences between ICD-O-3 and the “WHO classification” of myeloid neoplasms, published in 1997 by WHO in conjunction with the European Association of Hematopathologists and the Society for Hematopathology.^6,7  As previously described,²  the WHO classification includes the following MDS subtypes: refractory anemia (RA), RA with ringed sideroblasts (RARS), refractory cytopenia with multilineage dysplasia (RCMD), RCMD with ringed sideroblasts (RCMD-RS), RA with excess blasts-1 (RAEB-1), RAEB-2, MDS associated with isolated 5q deletion, and MDS, unclassified. The ICD-O-3 similarly includes RA, RARS, RAEB (RAEB-1 and -2 combined), MDS associated with isolated 5q deletion, RCMD, and MDS, not otherwise specified. Additional codes exist in ICD-O-3 for RAEB in transformation (retained from the original FAB classification) and therapy-related MDS. Categories of myeloproliferative disease (or CMDs in ICD-O-3) also differ from the WHO classification. The WHO classification includes 7 CMD categories: BCR/ABL⁺ chronic myelogenous leukemia (CML), chronic neutrophilic leukemia, hypereosinophilic syndrome, polycythemia vera, chronic idiopathic myelofibrosis (synonymous with myelosclerosis with myeloid metaplasia), essential thrombocythemia, and CMD, unclassifiable. The same categories are included in ICD-O-3 with the exception of BCR/ABL⁺ CML, which is grouped with the leukemias in ICD-O-3. Incidence rates for CML have been previously reported from SEER,⁸  and are therefore not included in this report.

Patterns in case reporting were described for SEER, NAACCR minus SEER registries, and all registries combined, including the distribution of patients by reporting source and diagnostic confirmation of the malignancy. Incidence rates were expressed as the number of new primary cancers per 100 000 persons at risk per year and age-adjusted according to the 2000 US standard population based on 5-year age groups.⁹  To investigate patterns in MDS and CMD risk, incidence rates were stratified by demographic characteristics, including year of diagnosis, sex, age, and race (white, black, and other) and ethnicity (Hispanic vs non-Hispanic). (There were too few patients with CMML to conduct stratified analyses.) At the time of this analysis, data from 2004 were available from SEER Program but not from the CINA Deluxe file. Therefore, all analyses of incidence were based on the years 2001 through 2003. Statistical significance of the differences in age-adjusted incidence rates was assessed using analysis of variance (ANOVA).

To examine the national burden of disease, numbers of newly diagnosed patients with MDS, CMD, and CMML in 2004 were estimated for the entire US population based on the respective incidence rates calculated from the CINA Deluxe file for 2001 through 2003 and population estimates obtained from the US Census Bureau for 2004.¹⁰  Specifically, MDS, CMD, and CMML incidence rates, stratified by 5-year age groups and sex, were multiplied by corresponding stratified census counts and divided by 100 000. Numbers of cases were then summed across all age groups and both sexes to obtain the total projected count for 2004.

To evaluate patterns in survival, 3-year relative survival rates were calculated using the SEER limited-use database consisting of SEER-17 registries, for which data are available from 2001 to 2004.¹¹  Relative survival rate is calculated by comparing observed survival with expected survival from a set of people with the same characteristics as the patient cohort with respect to age, race, sex, and calendar period.¹²  We used patients with MDS, CMD, and CMML diagnosed in 2001 through 2003 and followed through 2004 for 3-year relative survival rate calculations. Survival rates were stratified by sex, age and race, and 95% confidence interval (CI) are presented for each group. Survival by MDS subtype within SEER has been previously described² ; thus, subtype analyses are presented only for CMD. The SEER*Stat program (version 6.2.4; National Cancer Institute) was used for all analyses in this paper, except for the statistical significance testing with ANOVA, which was conducted using SAS (Cary, NC).

Reporting characteristics are presented in Table 1 for patients with MDS and CMD ascertained by SEER registries, by NAACCR registries that were not included in SEER (referred from here on out as “NAACCR minus SEER”), and for all patients combined. Between 2001 and 2003, 7076 and 17 722 patients with MDS were reported to SEER and NAACCR minus SEER registries, respectively, for a total of 24 798 patients with MDS overall. Roughly two-thirds as many patients with CMD were reported during the same time period (n = 16 119). As compared with patients with CMD, a greater proportion of patients with MDS were confirmed by positive histology in both SEER registries and NAACCR minus SEER registries (88% and 79% of patients with MDS and CMD, respectively, for SEER; 81% and 66% for NAACCR minus SEER). Conversely, a greater proportion of patients with CMD than with MDS were confirmed by a positive laboratory test (Table 1). The proportion of patients confirmed only by clinical diagnosis or other means was similar for MDS and CMD in both SEER and NAACCR minus SEER registries (4%-5%). Among patients with CMD and MDS reported to SEER registries, 92% were reported by a hospital or laboratory, as compared with 93% to 95% of patients with CMD and MDS in NAACCR minus SEER registries.

Age-adjusted incidence rates for MDS and CMD are presented in Table 2 for individual US states and geographic regions represented in NAACCR. MDS rates ranged from 1.5 per 100 000 in North Carolina to 5.6 per 100 000 in Detroit, while CMD rates ranged from 0.8 per 100 000 in Delaware to 4.1 per 100 000 in Idaho. The average annual age-adjusted incidence rate for MDS in 2001 through 2003 was 3.3 per 100 000, based on 24 798 cases reported by SEER and NAACCR minus SEER registries combined (Table 3). A slight but statistically significant increase in MDS incidence rates was observed with calendar year, ranging from 3.1 per 100 000 in 2001 to 3.5 per 100 000 in 2003 in all registries combined. Data for 2004 were available only for SEER at the time of this analysis, and the increase in MDS incidence continued through 2004, with an incidence rate of 3.8 per 100 000 based on 2720 reported patients (P < .05). Age-adjusted incidence of MDS was significantly higher among males (4.4 per 100 000) than females (2.5 per 100 000; P < .05; Table 3). A sharp increase in MDS incidence rates was observed with age, especially among the elderly: rates were 5 times greater among those aged 80 years and older (35.5 per 100 000) as compared with those aged 60 to 69 years (7.1 per 100 000). While no significant differences in MDS incidence rates were observed by race, non-Hispanics had a statistically significant increased risk of MDS compared with Hispanics. MDS incidence rates were highest among whites and non-Hispanics. Similar trends in MDS incidence rates with calendar year, sex, age, and race/ethnicity were observed between SEER and NAACCR minus SEER registries.

Incidence rates for CMD were lower than those for MDS, with an average annual age-adjusted incidence rate of 2.1 per 100 000, based on 16 119 reported patients in 2001 through 2003 (Table 4). No differences in incidence rates were observed by calendar year. (The incidence of CMD for 2004 in SEER was 2.1 per 100 000, based on 1561 patients.) CMD incidence rates were significantly higher in males (2.5 per 100 000) than females (1.8 per 100 000; P < .05). CMD incidence rates also significantly increased with age, although to a lesser extent than MDS incidence rates. Among individuals aged 80 years and older, the CMD incidence rate was 13.3 per 100 000. Statistically significant differences in CMD incidence rates were observed by race in all registries combined, with whites being at highest risk. No differences were observed by ethnicity. Trends in CMD incidence were similar in SEER and NAACCR minus SEER registries.

Frequencies and age-adjusted incidence rates are presented by disease subtype in Table 5. The distributions of disease subtypes were similar for SEER and NAACCR minus SEER registries. Refractory anemia (RA) comprised 16.8% and 13.4% of patients with MDS reported to SEER and NAACCR minus SEER registries, respectively. RA with sideroblasts and RA with excess blasts were the next most commonly reported MDS subtypes. More than half of all patients with MDS reported in SEER and NAACCR registries combined were of unspecified subtype. Polycythemia vera comprised 45% of patients with CMD, and 24% patients had essential thrombocythemia. Approximately 20% of patients with CMD were of unspecified subtype. The age-adjusted incidence rate for CMML was approximately one-tenth that of MDS, at 0.3 per 100 000.

Applying the observed incidence rates from combined SEER and NAACCR registries to the total US Census population estimates for 2004, the estimated numbers of patients with MDS, CMD, and CMML diagnosed in the total US population for 2004 were 9730, 6328, and 1039, respectively (Table 6). Approximately 7000 patients with MDS and 3000 patients with CMD were estimated to occur in individuals aged 70 years and older.

Trends in survival based on SEER data are presented in Table 7. Relative to the general population, 3-year survival with MDS was poorer than survival with CMD (45% vs 80%, respectively), while the worst 3-year survival was observed for CMML (21%). Males experienced poorer 3-year survival than did females for MDS, CMD, and CMML, a difference that was statistically significant for MDS (males, 41%; females, 50%). Survival for patients with MDS and CMD decreased with age: among those younger than 50 years of age at diagnosis, relative 3-year survival was greater than 60% and greater than 90% for MDS and CMD, respectively, while relative 3-year survival among those 80 years and older dropped to 37% and 66% for MDS and CMD, respectively. Although a similar inverse association between age at diagnosis and survival was observed for CMML, patients in the youngest age category for which there were reportable data (50-59 years) still experienced an extremely poor prognosis, with a relative 3-year survival of 33%. The 3-year survival was greatest for the most common CMD subtypes, polycythemia vera (n = 1615 patients; 88% 3-year survival) and essential thrombocythemia (n = 1028 patients, 92% 3-year survival). Similar survival was observed among 464 patients with hypereosinophilic syndrome (85%), while the poorest survival among patients with CMD was observed for myelosclerosis with myeloid metaplasia (synonymous with primary myelofibrosis; n = 464 patients; 53% 3-year survival) and CMD, not otherwise specified (n = 730; 63% 3-year survival).

The current analysis is the first to be based on NAACCR data, encompassing approximately 82% of the US population.¹³  Based on more than 40 000 observations, average annual age-adjusted incidence rates in the United States were highest for MDS, followed by CMD and CMML, for the years 2001 through 2003, with corresponding rates of 3.3, 2.1, and 0.3 per 100 000, respectively. Incidence rates were similar whether they were based on SEER or NAACCR minus SEER registries, with results from NAACCR minus SEER registries confirming the positive trends in MDS incidence with increasing age and male sex previously reported from SEER.² 

MDS may be misdiagnosed and/or underreported to population-based cancer registries. Elderly patients presenting to primary care physicians with anemia may not be assessed for a possible MDS diagnosis, and the likelihood of accurate MDS diagnoses may vary, depending on pathology expertise, physician, and patient characteristics. Although 88% of MDS diagnoses were confirmed by positive histology or laboratory test in NAACCR and SEER registries, 56% of diagnoses were of unspecified subtypes. Therefore, the observed subtype distribution may not be representative of the true MDS patient population if some subtypes were more likely to be characterized than others. No information or estimates are available on the number of patients in whom a possible diagnosis of MDS is not investigated by bone marrow studies.

Incidence rates for MDS increased with calendar year in 2001 to 2003 among the combined registry data, with the most recent 2004 SEER data indicating a rate of 3.8 per 100 000. Since MDS became a reportable malignancy only in 2001, it is possible that the increase in incidence rates over time reflect acclimation of those involved in the reporting process to the new guidelines resulting in rising capture rates. For example, only 4% of patients with MDS in NAACCR were reported by physicians' offices in 2001 through 2003. Since MDS is more commonly diagnosed and managed outside of hospitals compared with other cancers, it is possible that many of these cases are unreported to population-based registries. Although independent laboratories are also responsible for reporting MDS cases to local registries, the completeness of case reporting by out-of-state laboratories is unknown. As recently described by De Roos and colleagues,¹⁴  case-finding methods may affect the completeness of MDS case reporting; registries that rely on passive case-finding (ie, cases are reported to the registry by hospitals and other diagnostic facilities) may not capture as many patients with MDS as those that use active case-finding methods (ie, surveying billing, pathology, and cytogenetic and other laboratory testing records). Surveys of private physicians' offices and central referral laboratories are needed to estimate the proportion of patients with MDS not captured by hospital registries. Nevertheless, the incidence rate of MDS estimated for the United States in 2001 through 2003 (3.3 per 100 000) is remarkably similar to those previously reported from European countries,¹⁵  including England and Wales (3.6 per 100 000),¹⁶  Germany (4.1 per 100 000),¹⁷  Sweden (3.6 per 100 000)¹⁸  and France (3.2 per 100 000).¹⁹  Therefore, if substantial underreporting of MDS to cancer registries exists, the phenomenon is most likely not isolated to the United States. As clinicians, laboratories, and cancer registrars become more accustomed to reporting and recording MDS cases, incidence rates may continue to rise in the upcoming years.

In this first report of population-based incidence rates of CMD for the United States, demographic risk factors for CMD were similar to MDS, including older age, male sex, and white race. Although the CMD incidence rates did not increase with calendar year in 2001 through 2004 as they did for MDS, underreporting of CMD to population-based registries cannot be excluded. A recent analysis of medical claims data estimated the prevalence of polycythemia vera and essential thrombocythemia in the United States to be 136 000 patients as of 2003,²⁰  far greater than what would be expected based on the incidence rates reported here from SEER and NAACCR.

The 3-year relative survival was greater in patients with CMD (80%) than patients with MDS (45%), even among those aged 80 years and older at diagnosis. In contrast, survival with CMML was extremely poor (21%), even among younger patients. These data suggest that treatments with potential to alter the natural history of disease or curative strategies such as hematopoietic stem cell transplantation should be considered for patients with CMML who are appropriate candidates.

In conclusion, continued surveillance of MDS, CMD, and CMML through population-based registries will be useful for investigating trends in incidence and survival so that future prevention and treatment strategies may be developed. Concurrent assessment of potential misdiagnosis and underreporting of these malignant conditions is paramount for the elucidation and interpretation of these rates and trends.

The authors are grateful to Holly Howe at NAACCR and acknowledge Matthew Hayat for assistance with preliminary analyses.

Contribution: A.F.L., B.K.E., L.R., W.D.M., S.S.S., M.S., and D.E.R. designed the research; N.H. analyzed the data, and D.E.R. wrote the paper with contributions from A.F.L., B.K.E., L.R., W.D.M., S.S.S., M.S., and N.H.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Dana E. Rollison, Moffitt Cancer Center, 12902 Magnolia Drive, Tampa, FL 33612; e-mail: dana.rollison@moffitt.org.