Extramedullary Hematopoiesis in Myelodysplastic Syndromes: A Systematic Literature Review

: Extramedullary hematopoiesis is rarely seen in patients with myelodysplastic syndromes, and its clinical characterizations are not well-deﬁned. Here, we systematically reviewed the published literature to summarize the clinical manifestations, treatments, and long-term outcomes of biopsy-proven extramedullary hematopoiesis in patients with myelodysplastic syndromes. We included 41 patients, and ring sideroblasts were the most common myelodysplastic subtype (30.6%). Extramedullary hematopoiesis was typically symptomatic on presentation due to local compression, frequently involving the liver or spleen (36.6%), or the paravertebral region (24.4%). Notably, ring sideroblasts were predominantly seen in patients with non-hepatosplenic involvement (38.5 vs. 6.7%, p = 0.034). Interventions, when required, usually included surgery (36.8%) or radiation (13.2%), which led to symptomatic improvement in 55.5% of patients. The median overall survival of the current cohort was 7 months. The current study conﬁrms the rarity of extramedullary hematopoiesis as a complication of myelodysplastic syndromes; however, its outcomes in response to systemic modern therapies require further investigation.


Introduction
Extramedullary hematopoiesis (EMH) in adults is uncommon and typically pathological.Its underlying mechanisms might include the abnormal trafficking of hematopoietic progenitor cells in the context of ineffective bone-marrow hematopoiesis and aberrant cytokine milieu [1].It is primarily seen in patients with primary myelofibrosis or thalassemia; however, its rare association with other hematological disorders has long been recognized [2,3].For example, epidural EMH causing spinal cord compression was first reported in a patient fulfilling the diagnosis of myelodysplastic syndrome (MDS) dating back to the 1980s [4], when MDS was first proposed to be a separate entity of myeloid neoplasms [5].Most subsequent studies describing EMH in the setting of MDS were isolated case reports.Intriguingly, MDS was the most frequent associated condition with EMH (~13%) in a recent single-center retrospective case series from Mayo Clinic, including patients with biopsy-proven EMH in the absence of myeloproliferative neoplasms, although details of these MDS cases were not described [6].Overall, the current understanding of this rare condition is limited.We systematically review the available literature to describe the clinical features, management, and long-term outcomes of patients with EMH in MDS.
Hemato 2022, 3 570 its current classification [7].The reference lists of the retrieved records were also reviewed to identify additional cases.Duplicated records were excluded first.Two authors (C.W and Y.S.) independently screened the title and abstract of each record, and full texts of relevant records were reviewed by both authors to reach a consensus for inclusion or exclusion.To collect a more homogeneous group of patients, we focused on MDS patients with EMH proven by tissue biopsy, and excluded those solely diagnosed by imaging studies.Cases with individual patient information were included for data synthesis.Exclusion criteria were an inappropriate type of record (review or records not in the English language), an incorrect setting (wrong diagnosis or nonhuman studies), or a correct diagnosis without information at the patient level (Figure 1).

Methods
Records were identified by searching the PubMed and Embase databases between January 1982 and May 2022, using the following terms: "extramedullary hematopoiesis" OR "extramedullary haematopoiesis" AND "myelodysplasia" OR "myelodysplastic syndromes".The classification system of MDS was first introduced by the French-American-British group in 1982, and the World Health Organization used it as a starting point to create its current classification [7].The reference lists of the retrieved records were also reviewed to identify additional cases.Duplicated records were excluded first.Two authors (C.W and Y.S.) independently screened the title and abstract of each record, and full texts of relevant records were reviewed by both authors to reach a consensus for inclusion or exclusion.To collect a more homogeneous group of patients, we focused on MDS patients with EMH proven by tissue biopsy, and excluded those solely diagnosed by imaging studies.Cases with individual patient information were included for data synthesis.Exclusion criteria were an inappropriate type of record (review or records not in the English language), an incorrect setting (wrong diagnosis or nonhuman studies), or a correct diagnosis without information at the patient level (Figure 1).

Results and Discussion
Of the 41 cases from 30 records [4, included in the current study, the median age at EMH diagnosis was 70 years (interquartile range (IQR) 64-76), and 71.1% were male (Table 1).Patients were typically anemic, and the median level of hemoglobin was 8.2 g/dL (IQR 7.4-9.4,n = 20).MDS was often diagnosed before EMH onset (73.0%) with a median time interval of 13 months (IQR 4-36), and four cases had a reported temporal relationship between prior growth factor use (recombinant erythropoietin, n = 3; granulocyte-colony stimulating factor, n = 1) and subsequent EMH onset.Those with ring sideroblasts were this cohort's most common MDS subtype (30.6%).Bone-marrow blast percentage, cytogenetics, and mutational profile were not reported in most cases, and among the five patients with reported risk stratification with the revised international

Results and Discussion
Of the 41 cases from 30 records [4, included in the current study, the median age at EMH diagnosis was 70 years (interquartile range (IQR) 64-76), and 71.1% were male (Table 1).Patients were typically anemic, and the median level of hemoglobin was 8.2 g/dL (IQR 7.4-9.4,n = 20).MDS was often diagnosed before EMH onset (73.0%) with a median time interval of 13 months (IQR 4-36), and four cases had a reported temporal relationship between prior growth factor use (recombinant erythropoietin, n = 3; granulocyte-colony stimulating factor, n = 1) and subsequent EMH onset.Those with ring sideroblasts were this cohort's most common MDS subtype (30.6%).Bone-marrow blast percentage, cytogenetics, and mutational profile were not reported in most cases, and among the five patients with reported risk stratification with the revised international prognostic scoring system, four had lower-risk diseases.The presentation of EMH was either symptomatic (65.9%), typically related to local compression, or incidental, first suspected by imaging studies.The majority had antemortem diagnoses based on tissue biopsies, while four patients were diagnosed postmortem with autopsies.The most frequently involved organs were the liver and/or spleen (36.6%), and the most common non-hepatosplenic site was the paravertebral region (24.4%),particularly at the thoracic level.MDS with ring sideroblasts was predominantly seen in patients with non-hepatosplenic EMH when compared to those with hepatosplenic involvement (38.5 vs. 6.7%, p = 0.034).More importantly, ring sideroblasts were shown in the extramedullary tissue of four such cases when iron staining was reported, and one Hemato 2022, 3 571 patient had a concordant SF3B1 K666R mutation in his bone marrow and extramedullary tissue.Age, gender, or levels of hemoglobin were not different between these two groups.1 Continuous variables are presented as median (interquartile range), and categorical variables are reported as number (percentage).The number of patients with available information is provided in the parentheses; 2 Other sites involved with EMH include the lungs and pleura (n = 4), lymph nodes (n = 3), mediastinum (n = 2), skin (n = 2), central nervous system (n = 2), adrenal gland (n = 1), pericardium (n = 1), and pelvic space (n = 1).
EMH was observed without specific intervention in more than half of the cases (52.6%).When treatment was required, the most common interventions included surgery (36.8%) and radiation (13.2%).About 55.5% patients had symptomatic improvement after treatment.Of the four patients with EMH presumably related to growth factor use, discontinuation or dose reduction resulted in an improvement in EMH.Regarding the underlying MDS, it was mainly managed supportively with transfusion (63.0%) or recombinant erythropoietin use (18.5%).With a median follow-up of 16 months, 19 patients died at the last follow-up of their respective reports, and the median overall survival of the current cohort was 7 months.Four cases had secondary acute myeloid leukemia transformation (median interval from EMH onset, 11 months) during follow up.
To our knowledge, this is the only systematic review focusing on the rare association between MDS and EMH, which is a compensatory process to ineffective bone-marrow hematopoiesis contributed by the extramedullary seeding of abnormal hematopoietic progenitor cells.Nonetheless, this association could be extended to pre-MDS conditions and may not be as rare as we thought, given that an EMH case was recently reported with underlying clonal hematopoiesis with indeterminate potential, both of which sharing an ASXL1 mutation [37], implying that the diagnosis of idiopathic cases or those associated with unexplained cytopenia may need to be revised if next-generation sequencing is available to look for myeloid mutations.We confirmed that the most frequently involved organs were the liver and/or spleen, which play a role in embryonic hematopoiesis.MDS with ring sideroblasts was the most common subtype in the current cohort, and was first significantly associated with non-hepatosplenic EMH, which requires further validation because SF3B1-mutated MDS, classically linked to ring sideroblasts, is a distinct disease subtype, and its association with EMH has not been well-established in prior studies [38].In addition, the presence of ring sideroblasts in the EMH, and a shared SF3B1 mutation between EMH and bone marrow suggest that EMH in MDS is derived from underlying clonal myeloid progenitors rather than a reactive process.EMH-directed treatments, mainly surgery and/or radiation, are usually given to those affected significantly by local compression.However, the long-term prognosis of the current cohort was poor, considering that our study included cases spanning the last four decades, and most patients had not received modern MDS therapy.Further studies are needed to better define this condition from both the clinical and genetic standpoints, and investigate its natural history and outcomes in the current era.

Figure 1 .
Figure 1.Flow diagram of record selection.

Figure 1 .
Figure 1.Flow diagram of record selection.

Table 1 .
Characteristics of MDS patients with EMH.