Incidence of myeloid neoplasms in Spain (2002–2013): a population-based study of the Spanish network of cancer registries

Comprehensive population-based data on myeloid neoplasms (MNs) are limited, mainly because some subtypes were not recognized as hematological cancers prior to the WHO publication in 2001, and others are too rare to allow robust estimates within regional studies. Herein, we provide incidence data of the whole spectrum of MNs in Spain during 2002–2013 using harmonized data from 13 population-based cancer registries. Cases (n = 17,522) were grouped following the HAEMACARE groupings and 2013-European standardized incidence rates (ASRE), incidence trends, and estimates for 2021 were calculated. ASRE per 100,000 inhabitants was 5.14 (95% CI: 5.00–5.27) for myeloproliferative neoplasms (MPN), 4.71 (95% CI: 4.59–4.84) for myelodysplastic syndromes (MDS), 3.91 (95% CI: 3.79–4.02) for acute myeloid leukemia, 0.83 (95% CI: 0.78–0.88) for MDS/MPN, 0.35 (95% CI: 0.32–0.39) for acute leukemia of ambiguous lineage, and 0.58 (95% CI: 0.53–0.62) for not-otherwise specified (NOS) cases. This study highlights some useful points for public health authorities, such as the remarkable variability in incidence rates among Spanish provinces, the increasing incidence of MPN, MDS, and MDS/MPN during the period of study, in contrast to a drop in NOS cases, and the number of cases expected in 2021 based on these data (8446 new MNs).

. Period of study, number of cases of myeloid neoplasms, and quality indicators of data provided by Spanish provinces. 1 21 .
Crude rate (CR) and age-standardized incidence rate using the 2013 European standard population (ASR E ) were calculated using population data provided by the National Statistics Institute (Instituto Nacional de Estadística-INE) 18 , and expressed per 100,000 person-years. Poisson regression models were used to analyze the overall incidence time trends and to estimate the annual percent change (APC). The number of cases in Spain for 2021 was determined by applying to the 2021 Spanish population 18 the age-specific rates estimated for the year 2021. The latter were obtained by applying the APC (period 2002-2013) to the last quinquennium of known incidence (i.e. 2009-2013). All analyses were performed using R software (version 3.6.1) 22 .

Ethics approval.
This study is based on data from cancer registries gathered in the Spanish network of cancer registries (REDECAN). The public health administration of each autonomous community/province* authorized the collection and use of this data for its analysis without requirement of informed consent and ethical approval, covered by the Spanish general and public health laws 14

Results
MNs accounted for 30.81% (n = 17,522) of all hematologic malignancies (n = 56,777) diagnosed in Spanish population covered by cancer registries between 2002 and 2013. The quality and completeness of each registry, together with the study period and the total cases are detailed in Table 1. Of the total, 96.3% of the cases had microscopic confirmation, 3.6% were NOS cases, and 1.9% were recorded exclusively by death certificate (DCO). In particular, 33.5% of cases were MPN, 29.8% MDS, 25.7% AML, 5.2% MDS/MPN, 2.3% acute leukemia of ambiguous lineage, and the remaining 3.6% were NOS cases. Table 2 shows the number of cases, median age and incidence rates of all MNs according to histological subtype. The overall CR was 13.97 (95% CI 13.77; 14.18), and the overall ASR E was 15.52 (95% CI: 15.29; 15.75), being 19.92 (95% CI 19.51; 20.33) in men and 12.39 (95% CI 12.12; 12.67) in women. There was a marked male predominance (9,650 cases in men (55.1%), sex ratio = 1.61), and the median age at diagnosis was 73 years (interquatile range (IQR) 60-81 years). Moreover, incidence increased markedly with age, reaching a maximum around 75-79 years in most subgroups (Fig. 1). ASR E of MNs by cancer registry are displayed in Fig. 2  Age group Age−specific rate  In line with previous studies, incidence of MNs was markedly higher in men than in women for most subtypes. Notable exceptions included MDS associated with isolated del(5q), AML and MDS therapy related, and essential thrombocythemia, already reported in the literature. Likewise, the incidence of all MNs increased with advancing age, being particularly marked in NOS cases, in which the incidence rose sharply from age 70 years. This might suggest a decline in the quality of the diagnostic workup in the elderly, who are less likely to receive aggressive diagnostic tests due to comorbidity and/or frailty, and may therefore receive a suboptimal treatment for their conditions 23 .
Regarding specific MNs subtypes, lower incidence rates for most entities were reported in our study in comparison to the most recent data provided by the HMRN 14 . This could be partly explained by the specialized nature of the HMRN (with all diagnoses made and coded by clinical specialists working at a single integrated hematopathology laboratory), and by the lack of concordance in the recording of progressions/transformations. www.nature.com/scientificreports/ In contrast, incidence rates of overall MPN, MDS, and MDS/MPN in our region were markedly higher in comparison to European 2,8,13 and US 9-11 datasets, most of them covering years before/close to the implementation of the ICD-O-3 and the 2001 WHO classification. As far as MPN are regarded, disparities were mainly attributed to polycythemia vera and essential thrombocythemia, while rates of chronic myeloid leukemia (consistently documented since 1970's with the identification of its causal chromosome transition), primary myelofibrosis, and mastocytosis were similar across studies. Such differences may be linked to the identification of the JAK2 mutation in 2005 24 and the derived 2008 WHO guidelines for MPN, whose impact is not documented in series covering only previous years. On the other hand, the incidence of AML, which is a long-established entity, was more homogeneous across different regions. Indeed, overall rates were consistent with European 2,8 and US 25 findings, as well as with smaller European series 13,26-28 , while slightly lower rates were reported in Canada 29 and Switzerland 30 . Karyotypic information was not available for many of our cases, and thus, the proportion of AML with cytogenetic abnormalities (14%) was slightly lower in comparison with more specific studies [31][32][33] . However, rates of AML with t(15;17) (q22;q12) were still higher compared to the European average, further supporting the hypothesis that such entity might be more prevalent among individuals with Spanish ancestry 34 . Finally, most of these studies included AML of ambiguous lineage within AML-NOS subgroups, although it is placed as a distinct category from AML since the introduction of the 2008 WHO classification. Further studies are warranted to clarify the epidemiology of these entities owing their clinical relevance.
We evidenced increasing incidence trends of MDS/MPN, MDS, and several MPN, previously reported in the literature and mostly linked to refinements in the diagnostic, classification, and registration practices. Within the latter, this was particularly seen in the three most frequent Philladelphia chromosome negative subtypes, and thus, may be linked to the implementation of screening for JAK2 mutation. In the same vein, Girodon et al. 35 documented an almost twofold increase in the incidence of essential thrombocythemia after 2005, but not in the remaining MPN subtypes. In agreement with the few European studies examining AML incidence trends 13,15 , we www.nature.com/scientificreports/ found a stable incidence of overall AML across the period of study. In contrast, an increasing trend was found in a Dutch pediatric study   36 and in Canada (1992-2010) 29 and US from 2009 to 2010 25 in the general population, the latter mainly attributed to changes in the registration of transformations in the Surveillance, Epidemiology, and End Results (SEER) program. Finally, NOS cases decreased remarkably across the period of study, which could be attributed both to a more specific clinical diagnosis and/or to an improved codification in Spanish cancer registries. The etiology of MNs, in line with most hematological malignancies, is still uncertain. Several subtypes have been consistently associated with treatments (i.e. radiation, alkylating agents or topoisomerase II inhibitors) 37 , while environmental epidemiological studies suggest a potential role of obesity, tobacco exposure, autoimmune disorders, and infections in myelodisplastic 38 or myeloproliferative diseases 39 . However, neither these factors, nor the genetic alterations currently described 40 , can explain the large variability in the incidence of these neoplasms 2 . In addition, drawing etiological hypothesis based on geographic heterogeneity in incidence rates is hampered by heterogeneity in accuracy and completeness in the registration of several subtypes. Several medical-claims-based studies have shown an underreporting of MNs [41][42][43][44] , namely MDS and MPN (which are often diagnosed and managed in an outpatient setting, and might be missed by surveillance systems relying on hospital registration), and among the elderly (in which diagnostic evaluation might not be as aggressively sought as in younger individuals). Indeed, we evidenced marked differences in incidence rates across Spanish provinces, with the highest incidence rates of MDS, MPN, and MDS/MPN reported in the Girona cancer registry, which has started several initiatives 15,45 to cope with these challenges. Following the example of the French Network of Cancer Registries (FRANCIM) 46 , training programs to improve the codification and registration of hematological neoplasm have been boosted in the REDECAN during the last few years, which are expected to start to bear fruits in future studies.
Since 2008, there have been numerous advances in the identification of genetic biomarkers associated with specific MNs, which led to the release of an updated WHO classification in 2016 5 . The impact of these changes will be noticeable within the next years, when they become routinely distinguished in clinical practice and consistently coded in cancer registries. The incorporation of these updates at a cancer registry level will be eased with the release of the ICD-O-3, second revision 47 , which is recommended for use from 2020. Further studies with contemporary data including these classification changes are warranted.
The number of expected MNs in 2021 depicts the present cancer burden of these malignancies in Spain. However, these data should be interpreted with caution due to several factors. First, some subtypes are extremely rare, making estimates less robust. Furthermore, the estimates provided herein do not reflect the impact of the new 2016 WHO classification 5 , nor that of coronavirus disease 2019 (COVID-19) 48 , as they are based on extrapolations of cancer data collected in previous years. Regarding the latter, although the full extent of the impact of the COVID-19 pandemic remains unknown, delays in cancer diagnosis are expected to cause a short-term decline in cases followed by an increasing incidence of advanced-stage diagnosis 49,50 . In addition, if, over the period 2002-2013, there had been an increase in the completeness in the registration of MN cases, with the corresponding positive effect on the APC, this would cause an overestimation in the number of cases predicted for the year 2021. Nonetheless, these results are still interesting for clinicians and public health specialists in evaluating the cost of management and new treatments for these pathologies, and to account for the gap between the expected and the observed cases after the COVID-19 pandemic.
Among the strengths of this population-based study is the large number of MNs that allowed us to assess and compare incidence rates not only for common but also for relatively rare entities. However, several limitations must be considered when interpreting our results. First, the changing classification and diagnostic criteria (and consequent heterogeneity in disease definitions across countries, clinical centers, and cancer registries) hamper the interpretation of our incidence rates and trends, as well as comparisons with previous studies. In addition, we cannot exclude the aforementioned underreporting of cases, particularly documented in MDS and MPN, and among the elderly. In addition, we lacked a centralized pathology and clinical review, which could have decreased the proportion of NOS cases and improved the quality of our data. This is particularly relevant for MDS, due to the poor inter-observed concordance in diagnosis and the numerous non-neoplastic conditions that can mimic such neoplasms 51,52 . Nevertheless, in spite of the unavoidable biases due to variability and variation in registration quality and coding practices, over 95% of cases had adequate morphology specification.
In conclusion, this study presents the first comprehensive population-based analysis of MNs incidence in Spain. It highlights some useful points for public health authorities, such as the increasing incidence of several subtypes, the remarkable variability in incidence rates (especially of MDS, MPN, and MDS/MPN) among provinces, and the number of cases expected in 2021 based on these data. The negative trend in the incidence of NOS cases suggests a more specific diagnosis and/or improvements in the registration of these cases across the study period, however, additional efforts should be made to improve the quality of MNs data in future studies.

Data availability
The dataset analyzed during the current study is not publicly available due to national regulations of cancer registry data. However, it is available anonymized from Dr. Rafael Marcos-Gragera (rmarcos@iconcologia.net) on reasonable request.