Defining a Subgroup of Myelodysplastic Syndrome Patients With Very Poor‐Risk Cytogenetics Demonstrating a Relatively More Favorable Outcome After Allogeneic Hematopoietic Cell Transplantation

To the Editor,

1.

Allogeneic hematopoietic cell transplantation (Allo‐HCT) remains the only potentially curative strategy for selected patients with myelodysplastic syndromes/neoplasms (MDS). However, transplant outcomes remain strongly influenced by disease biology such as marrow blasts, cytopenias, and cytogenetics [1, 2]. Within the revised International Prognostic Scoring System (R‐IPSS), patients harboring more than three cytogenetic abnormalities are assigned to the very poor‐risk category, but this definition is mainly based on the number of abnormalities rather than the biological contribution of individual features such as monosomal karyotype (MK), 5q abnormalities (abn5q), 7q abnormalities (abn7q), 17p abnormalities (abn17p), and 3q26 abnormalities (abn3q26) [3]. We believe that this very poor‐risk population after allo‐HCT is not homogeneous and that specific cytogenetic abnormalities may refine prognosis beyond abnormality count alone.

This retrospective registry‐based study was conducted on behalf of the Chronic Malignancies Working Party of the European Society of Blood and Marrow Transplantation (EBMT) and included adult patients with very poor‐risk cytogenetics MDS transplanted between 2001 and 2020 from either matched sibling or unrelated donors. This study was approved by EBMT and all patients provided consent on the use of their personal information. To reduce procedural heterogeneity, second transplant, haploidentical and cord blood donors were excluded. A total of 580 patients met these criteria. Median age was 60 years; median time from diagnosis to allo‐HCT was 6 months; and median follow‐up was 52 months. Most patients received a reduced‐intensity conditioning regimen (65%); matched unrelated donors were used in 61%; and peripheral blood was the stem cell source in 93% (Table 1). Our cohort illustrates the highly biological heterogeneity hidden within this conventional risk group. MK was present in 78% of evaluable patients, abn7q in 60%, abn5q in 77%, abn17p in 34%, and abn3q26 in 14%, with substantial overlap between abnormalities. Hyperdiploid karyotype was identified in 16%, normal karyotype mosaicism in 54%, and more than one cytogenetic clone in 72%. In addition, 73% of patients had at least 4 abnormalities and 68% were classified as MDS‐IB2 (Increased Blasts).

TABLE 1.

Patients' characteristics.

Characteristic	N = 580 a	Missing
Median follow‐up (IQR, months) b	52 months (36–73)
Age at allo‐HCT (years)	60 (53, 66)	0 (0%)
Patient sex		0 (0%)
Male	342 (59%)
Female	238 (41%)
Interval from diagnosis to allo‐HCT (months)	6.0 (4.2, 8.8)	0 (0%)
Patient CMV		20 (3.4%)
Positive	352 (63%)
Negative	208 (37%)
Conditioning intensity		7 (1.2%)
Standard	198 (35%)
Reduced	375 (65%)
Donor type		0 (0%)
MSD/MRD	227 (39%)
MUD	353 (61%)
Stem cell source		0 (0%)
PB	541 (93%)
BM	39 (6.7%)
MDS stage		8 (1.4%)
MDS‐IB2	389 (68%)
Other	183 (32%)
Hyperdiploid karyotype	90 (16%)	24 (4.1%)
Monosomal karyotype	448 (78%)	5 (0.9%)
7q abnormalities	349 (60%)	3 (0.5%)
5q abnormalities	444 (77%)	4 (0.7%)
17p abnormalities	194 (34%)	5 (0.9%)
3q abnormalities	81 (14%)	4 (0.7%)
> 1 clone (clonal heterogeneity)	284 (72%)	187 (32%)
Normal karyotype mosaicism	210 (54%)	188 (32%)
Number of abnormalities		0 (0%)
< 4	155 (27%)
≥ 4	425 (73%)

Abbreviations: Allo‐HCT, allogeneic hematopoietic stem cell transplantation; BM, bone marrow; CMV, cytomegalovirus; MDS, myelodysplastic syndrome; MDS‐IB2, myelodysplastic syndrome with increased blasts 2; MRD, matched related donor; MSD, matched sibling donor; MUD, matched unrelated donor; PB, peripheral blood.

^a Median (Q1, Q3); n (%).

^b Calculated by reverse Kaplan–Meier method.

Overall outcomes after allo‐HCT were poor. At 5 years, progression‐free survival (PFS) was 19% [95% CI: 16%–23%], overall survival (OS) was 21% [95% CI: 17%–25%], relapse incidence (RI) was 56% [95% CI: 51%–60%], and non‐relapse mortality (NRM) was 25% [95% CI: 21%–29%]. Day‐100 cumulative incidence of Grades III and IV acute graft‐versus‐host disease (GvHD) was 13% [95% CI: 10%–15%], while the cumulative incidence of extensive chronic GvHD was 15% [95% CI: 12%–18%]. These translated into a 5‐year probability of GvHD/relapse‐free survival (GRFS) of 10% [95% CI: 8%–13%]. Relapse was the leading cause of death, accounting for 39% of deaths, followed by infection at 29% and GvHD at 17%. Overall, relapse rather than toxicity appeared to be the main barrier to durable benefit in this setting.

To dissect heterogeneity within this very poor‐risk cytogenetic cohort, a hierarchical classification was applied using cytogenetic abnormalities previously considered relevant in myeloid disease [4, 5]: hyperdiploid karyotype, MK, abn5q, abn7q, abn17p, and abn3q26. This produced six mutually assigned subgroups: isolated hyperdiploid karyotype without the other listed abnormalities (N = 23), MK without additional features (N = 21), abn7q without abn5q, abn17p, and abn3q26 (N = 36), abn5q without abn17p or abn3q26 (N = 229), abn17p without abn3q26 (N = 156), and abn3q26 irrespective of the presence of previous abnormalities (N = 77). This hierarchical classification significantly separated PFS, OS and RI (p < 0.001, p = 0.001 and p = 0.001, respectively). By contrast, it did not significantly affect NRM and both acute and chronic GvHD. These findings argue that the biological meaning of cytogenetic complexity depends not only on how many lesions are present, but also on which abnormalities. The most striking observation was that our six‐group model could be simplified into two clinically meaningful categories. Patients lacking abn5q, abn17p and abn3q26 had better outcomes than those harboring at least one of these abnormalities. In the favorable group, 5‐year PFS was 43% [95% CI: 32%–55%] and 5‐year OS was 46% [95% CI: 34%–58%], whereas in the unfavorable group, these fell to 16% [95% CI: 12%–19%] and 18% [95% CI: 14%–22%], respectively (p < 0.001) (Figure 1). These data suggest that not all patients currently labeled as having very poor‐risk cytogenetics experience similar dismal outcomes after allo‐HCT. Instead, the absence of abn5q, abn17p or abn3q26 identifies a subset with unexpectedly durable disease control after allo‐HCT.

FIGURE 1.

Kaplan–Meier curves of (A) progression‐free survival (PFS) and (B) overall survival stratified by categorized cytogenetic subgroup. Numbers of patients at risk are indicated below the time axis at specific timepoints. The shaded regions indicate 95% confidence intervals. p‐Values are calculated by the log‐rank test.

The burden of abnormalities also retained independent prognostic impact. Patients with 6 or more abnormalities had significantly worse 5‐year PFS compared with those with fewer than 6 abnormalities (12% [95% CI: 8%–16%] vs. 29% [95% CI: 23%–35%], p < 0.001), inferior 5‐year OS (14% [95% CI: 10%–18%] vs. 32% [95% CI: 25%–38%], p < 0.001), higher RI (62% [95% CI: 57%–68%] vs. 47% [95% CI: 40%–53%], p < 0.001) and lower 5‐year GRFS (7% [95% CI: 4%–10%] vs. 14% [95% CI: 10%–19%], p = 0.003). In contrast, NRM was not influenced by this threshold. In multivariate analysis, the number of abnormalities remained linearly associated with worse PFS, OS, RI and GRFS, with each additional abnormality increasing event risk by approximately 5%–7%. This observation supports the concept that cumulative genomic complexity contributes significantly to treatment failure after allo‐HCT, largely through relapse rather than excess NRM.

The multivariate analysis provides further support of the clinically utility of our simplified cytogenetic model (Tabe 2). Compared to the favorable group, the unfavorable group had worse PFS with a hazard ratio 1.73 (1.15–2.37, p < 0.001), higher RI with a hazard ratio of 2.33 (1.5–3.61, p < 0.001), inferior OS with a hazard ratio of 1.77 (1.27–2.47, p < 0.001) and worse GRFS with a hazard ratio of 1.4 (1.05–1.87, p = 0.021), while no significant effect was observed on NRM (Table 2). The absence of normal karyotype mosaicism was also associated with inferior PFS and OS, whereas clonal heterogeneity defined by the presence of more than 1 clone did not significantly influence major outcomes [6]. Older age adversely affected PFS and RI, and disease stage also retained prognostic value, while donor type was not associated with outcomes in our model [7]. Notably, when cytogenetic risk was further simplified by the presence or absence of abn5q alone, the effect on PFS remained comparable suggesting that abn5q may be a major driver of our unfavorable category (data noy shown).

TABLE 2.

Multivariate analysis using a Cox proportional hazard model in the entire cohort for PFS, OS, RI, NRM, and GRFS.

	p	HR	95% CI
PFS
Unfavorable	< 0.001	1.73	1.25–2.37
No NK mosaicism	0.027	1.41	1.04–1.92
No clonal heterogeneity	0.4	0.86	0.61–1.21
No. of abnormalities (cont.)	< 0.001	1.05	1.02–1.08
Age (per 10 year)	0.004	1.15	1.04–1.27
No excess of blasts	0.002	0.71	0.58–0.88
MUD vs. MSD	0.6	0.95	0.78–1.16
OS
Unfavorable	< 0.001	1.77	1.27–2.47
No NK mosaicism	0.018	1.45	1.06–1.98
No clonal heterogeneity	0.3	0.83	0.59–1.17
No. of abnormalities (cont.)	< 0.001	1.06	1.03–1.1
Age (per 10 year)	0.06	1.1	1–1.21
No excess of blasts	0.016	0.77	0.62–0.95
MUD vs. MSD	0.5	0.93	0.76–1.13
RI
Unfavorable	< 0.001	2.33	1.5–3.61
No NK mosaicism	0.15	1.32	0.91–1.91
No clonal heterogeneity	0.3	0.78	0.51–1.2
No. of abnormalities (cont.)	< 0.001	1.06	1.03–1.1
Age (per 10 year)	0.015	1.16	1.03–1.31
No excess of blasts	< 0.001	0.58	0.45–0.77
MUD vs. MSD	0.08	0.81	0.64–1.03
NRM
Unfavorable	0.6	1.15	0.72–1.85
No NK mosaicism	0.09	1.6	0.93–2.75
No clonal heterogeneity	0.9	1.03	0.58–1.81
No. of abnormalities (cont.)	0.2	1.03	0.98–1.09
Age (per 10 year)	0.13	1.14	0.96–1.34
No excess of blasts	0.9	1.03	0.73–1.47
MUD vs. MSD	0.12	1.34	0.93–1.93
GRFS
Unfavorable	0.021	1.4	1.05–1.87
No NK mosaicism	0.06	1.32	0.98–1.76
No clonal heterogeneity	0.4	0.86	0.62–1.19
No. of abnormalities (cont.)	0.004	1.04	1.01–1.07
Age (per 10 year)	0.1	1.08	0.99–1.18
No excess of blasts	0.17	0.87	0.71–1.06
MUD vs. MSD	0.9	0.99	0.82–1.19

Abbreviations: CI, confidence interval; cont., continuous; GRFS, graft‐versus‐host and relapse‐free survival; HR, hazard ratio; MSD, matched sibling donor; MUD, matched unrelated donor; NK, normal karyotype; No., number; NRM, nonrelapse mortality; OS, overall survival; PFS, progression‐free survival; RI, relapse incidence.

The adverse impact of abn5q, abn17p, and abn3q26 on outcomes has been demonstrated already. We hypothesize that advanced abn5q MDS with highly complex karyotype may be enriched by TP53 mutations [8, 9, 10], and that acquisition of abn17p may further intensify genomic instability [11]. Similarly, abn3q26 is linked to MECOM dysregulation and adverse biology [12]. Although molecular data were not available in the EBMT registry and therefore could not be formally incorporated, the observed cytogenetic patterns are consistent with the concept that certain lesions are surrogates for aggressive molecular programs that are relatively resistant to a graft‐versus‐leukemia effect [13]. This likely explains why the main disadvantage of our unfavorable group was higher relapse rather than higher NRM.

Our results are relevant to clinical decision‐making before and after allo‐HCT. We suggest that the current very poor‐risk cytogenetic category should not be automatically assigned to a hopeless dismal outcome after allo‐HCT, because a subset lacking abn5q, abn17p, or abn3q26 can achieve favorable long‐term survival. Conversely, patients harboring one or more of these abnormalities, especially in the setting of six or more abnormalities, appear to remain at very high risk of relapse after allo‐HCT and should therefore be considered for intensified disease control before transplantation and posttransplant strategies [14, 15, 16] including novel therapies as potential avenues for this subgroup.

This study also has limitations that should temper interpretation. It is retrospective, registry‐based, and spans two decades, during which transplant practices, supportive care, conditioning regimens, and disease classification evolved. Cytogenetic and clinical data were available, but molecular abnormalities, including TP53 mutations and other somatic mutations which are nowadays important to MDS risk assessment, were not captured. Some subgroup sample sizes were modest. Nonetheless, the consistency of the relapse‐driven signal across univariate and multivariate analyses supports the robustness of our main conclusion.

In summary, this EBMT retrospective study shows that MDS undergoing allo‐HCT with very poor‐risk cytogenetics do not constitute a homogenous prognostic group. The absence of abn5q, abn17p, and abn3q26 identifies a subgroup with better long‐term PFS and OS, whereas the presence of any of these abnormalities and increasing cytogenetic complexity predict relapse and inferior outcomes after allo‐HCT. These findings support refinement of cytogenetic risk assessment in transplant‐eligible MDS and provide a rationale for risk‐adapted peri‐transplant strategies, particularly for patients with abn5q, abn17p, or abn3q26‐driven complex disease.

Funding

The authors have nothing to report.

Ethics Statement

The study was approved by the board of the Chronic Malignancies Working Party of the EBMT.

Consent

Patients or legal guardians provided informed consent authorizing the use of their personal information for research purposes.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.