Non-infectious fever predicts poor survival in azacitidine-treated high-risk, but not low-risk, myelodysplastic syndromes：a multicenter real-world cohort study (KOTOSG) Free

Growing evidence indicates that dysregulated inflammation—marked by specific genetic abnormalities and elevated inflammatory markers such as C-reactive protein—plays a significant role in the pathophysiology of myelodysplastic syndromes (MDS). Although some MDS patients experience a fever unrelated to infection, known as non-infectious fever (NIF), its clinical significance remains unclear. Consequently, we performed a multicenter, retrospective study to determine how often NIF occurs and its impact on overall survival (OS) in a real-world MDS patient cohort patients.

Patients newly diagnosed with MDS between January 2012 and December 2023 at nine institutions of the Kyoto Clinical Hematology Study Group were reviewed retrospectively. Diagnoses and subtypes followed the 2017 WHO classification. Using the Revised International Prognostic Scoring System (IPSS-R), patients with very low, low, or intermediate risk were classified as low risk (LR), whereas those with high or very high risk were classified as high risk (HR). The study cohort consisted of two groups: an LR-MDS cohort, comprising all patients with LR disease, and an HR-MDS cohort, defined as HR patients who received azacitidine (AZA).

NIF was defined as a fever ≥ 37.5 °C lasting at least 7 days without documented infection. NIF events were recorded if they occurred (i) before diagnosis, (ii) within 12 months after diagnosis in the LR-MDS cohort, or (iii) within 30 days after AZA initiation in the HR-MDS cohort. OS was measured from diagnosis for the LR-MDS cohort and from AZA initiation for the HR-MDS cohort until death from any cause. Univariable analyses used a time-dependent Cox model for NIF and standard Cox models for other covariates. Multivariable analysis was then performed with a Cox model adjusted for IPSS-R components, treating NIF as a time-dependent covariate. Leukemic transformation and non-leukemic mortality were analyzed as competing risks, and cumulative incidence functions were compared via Gray's test.

A total of 353 patients were assigned to the LR-MDS cohort and 276 to the HR-MDS cohort. In the LR-MDS cohort, median follow-up was 26.7 months, with a median age of 78 years. The HR-MDS cohort had a median follow-up of 11.3 months and a median age of 73 years. In LR-MDS, multilineage dysplasia was the most common subtype (36%), followed by single-lineage dysplasia (22%). Marrow blasts were ≤ 2% in 73% of patients, and 64% had good-risk karyotypes. Severe neutropenia, anemia, and thrombocytopenia were observed in 11%, 28%, and 13%, respectively. In contrast, the HR-MDS cohort was enriched for excess-blasts 2 (EB-2, 42%) and EB-1 (34%). Very-poor-risk cytogenetics were present in 50%, including −7/−7q (36%), del(5q) (24%), and +8 (22%). Severe neutropenia, anemia, and thrombocytopenia occurred in 42%, 57%, and 48%, respectively.

NIF was observed in 19 patients (5%) of the LR-MDS cohort and in 21 patients (8%) of the HR-MDS cohort. The median detection occurred 63 days before diagnosis in LR-MDS (range -26 to +338) and 1 day before starting AZA in HR-MDS (range -84 to +29). NIF did not predict prognosis in the LR-MDS cohort (univariate HR = 1.54, p = 0.212; multivariate HR = 1.19, p = 0.636), but it independently indicated worse survival in the HR-MDS cohort (univariate HR = 2.41, p < 0.001; multivariate HR = 3.21, p < 0.001). The rate of leukemic transformation was similar regardless of NIF status (p = 0.357). Conversely, the cumulative incidence of non-leukemic death was higher in NIF-positive patients (p < 0.001), with infections causing 56% of deaths in NIF-positive and 22% in NIF-negative cases (p = 0.004). Immunosuppressive therapy was administered to 57% of NIF-positive patients compared to 11% of NIF-negative patients. The median overall survival after NIF onset in the HR-MDS cohort was 5.7 months. Within this cohort, both NIF types were associated with poorer outcomes: NIF before or after AZA initiation was linked to worse overall survival versus NIF-negative patients (HR = 2.76, p = 0.002; and HR = 2.13, p = 0.028, respectively), with no significant difference between these subgroups (p = 0.519).

NIF serves as an independent predictor of poor survival related to infection in AZA-treated HR-MDS, though it has limited prognostic value in low-risk cases. These results highlight the importance of developing new treatments that address both inflammation and infection for HR-MDS patients exhibiting NIF.