Abstract
Therapy-related myelodysplastic syndromes (t-MDS) and acute myeloid leukemias (t-AML) are dramatic late complications of therapies with DNA-damaging agents, with 5-year survival rates around 10%.
We retrospectively evaluated 86 patients (pts) diagnosed with t-MDS (N=42), t-MDS/AML (N=13), or t-AML (N=31), according to the International Consensus Classification. Inclusion criteria were diagnosis between 2007-2025 and prior exposure to chemotherapy (CT), radiotherapy (RT), and/or immunotherapy (IT) for previous cancer. We estimated overall survival (OS) and cumulative incidence of t-MDS/t-AML related mortality, considering disease- and treatment-related deaths as events, and censoring patients at the time of first cancer relapse occurring after t-MDS/t-AML onset. Uni- and multivariable Fine and Gray models identified predictors of mortality, including age, prior malignancy type, high-dose (HD)-CT, RT or IT exposure, achieving first-line complete remission (CR1), candidacy or receipt of alloSCT, cytogenetic risk factors (complex karyotype [CK], del[5q], del[7q], TP53 mutation/del[17p]), Revised-International Prognostic Scoring System (R-IPSS), and European Leukemia Net (ELN) 2022 risk categories. Starting our analysis in 2007, molecular data were insufficient for inclusion.
Median latency from primary cancer to t-MDS/t-AML onset was 87.1 months; median age at diagnosis 65.6 years. Prior malignancy was solid (29%) or hematologic (71%); 36% had received also HD-CT, 44% also RT. High-risk cytogenetics included CK (43%), del(5q) (37%), del(7q) (35%), TP53/del(17p) (27%). ELN risk was adverse in 65% t-AML pts, while R-IPSS was high/very-high in 61%. Among 39/86 candidates, 24 underwent alloSCT. HD-CT exposure showed a non-significant trend toward longer latency (118 vs 71.3 months; p=.068) and was associated with higher frequency of t-MDS (71% vs 36%, p=.003) and greater CK frequency (66% vs 31%, p=.005). Baseline characteristics were similar across diagnosis periods (2007–2017 vs 2018–2025) and primary malignancy types. Two-year OS was 35.1% and differed significantly between pts either undergoing or not alloSCT (61.7% vs 21.2%, log-rank p=.0001). Cumulative incidence of t-AML/t-MDS-related mortality at 3, 6, and 24 months was 9.6%, 27%, and 63%, respectively. Univariable analysis showed higher t-MDS/t-AML related mortality in pts with age ≥65 years (HR 4.28; 95%CI 2.28–8.02; p<.001), CK (HR 2.03; 95%CI 1.13–3.66; p=.018), del(5q) (HR 2.43; 95%CI 1.39–4.23; p=.002), del(7q) (HR 1.97; 95%CI 1.09–3.57; p=.025), TP53/del(17p) (HR 2.77; 95%CI 1.53–5.04; p<.001), high/very-high R-IPSS (HR 6.26; 95%CI 2.2–17.8; p<.001) and adverse ELN risk (HR 9.61; 95%CI 1.13–81.6; p=.038). Protective factors included achieving CR1 (HR 0.06; 95%CI 0.01–0.43; p=.004), candidacy (HR 0.49; 95%CI 0.28–0.86; p=.012) or receipt of alloSCT (HR 0.33; 95%CI 0.17–0.62; p<.001). Diagnosis period (2007-2017 vs 2018-2025) did not impact disease-related mortality (p=.474). Interaction models between age-alloSCT and CK-alloSCT showed that alloSCT did not alter adverse prognostic impact of older age but reversed the negative effect of CK (HR 3.02 without alloSCT vs HR 0.72 with alloSCT; interaction p=.034). Multivariable analysis confirmed age ≥ 65 (HR 2.79; 95%CI 1.18–6.63; p=.02) and CK (HR 1.85; 95%CI 1.02–3.35; p=.044) as independent mortality predictors, but not alloSCT (p>.05). Results were similar when t-AML-related mortality (including MDS/AML) was analyzed separately.
Our data confirm poor outcomes in t-MDS/t-AML (disease-related mortality 63% at 2 years) without improvement in the last two decades. HD-CT exposure selects for an MDS-predominant, cytogenetically complex phenotype with prolonged latency to myeloid neoplasm development. Older age and CK were the strongest independent disease-related mortality predictors, overshadowing the protective role of alloSCT found in univariable analyses. In interaction models, the negative prognostic effect of age outweighed transplant benefit, whereas the negative prognostic impact of CK was successfully reverted by alloSCT. Although proceeding to alloSCT represent a positive selection factor per se, reflecting younger age, chemosensitive or low-blast count disease, our data may still support the beneficial role of alloSCT in this setting, irrespective of high-risk cytogenetics. With current available treatments the best option for our patients remains to limit the occurrence of therapy-related malignancies.
