A Prognostic Model for Predicting the Impact of Comorbidities on Survival of Patients with Myelodysplastic Syndromes.

Abstract

Myelodysplastic syndromes (MDS) occur mainly in older persons, and these patients are likely to have comorbidities. We studied the impact of comorbidities on non-leukemic death (NLD) and overall survival (OS) in MDS patients with the aim of developing a specific prognostic index. Eight hundred forty consecutive patients receiving a diagnosis of MDS at Policlinico San Matteo, Pavia, Italy, between 1992 and 2006 were retrospectively evaluated. One or more comorbidities were present in 455/840 (54%) patients: the older the age, the higher their prevalence (P<0.001). Cardiac disease was observed in 25% of patients, liver disease in 16%, diabetes in 11%, prior solid tumor in 10%, nephropathy and pulmonary disease in 4%. Non-leukemic causes of death included cardiac failure (63%), infection (24%) and hemorrhage (7%). In a Cox analysis with age, sex, WHO category, cytogenetics and transfusion-dependency as time-dependent covariates, the presence of one or more comorbidities significantly affected both the risk of NLD (HR=1.91, P=0.001) and OS (HR=1.51, P=0.01), while it did not influence the risk of leukemic progression. The negative effect of comorbidities on OS was more evident in patients without excess of blasts (HR=1.8 P=0.007), while it retained a borderline significance in patients with more advanced disease (P=0.05). By including comorbidities as distinct entities in multivariable analysis, cardiac failure, liver or pulmonary disease, and solid tumors were found to independently affect the risk of NLD (HR=3.7, HR=2.08, HR=2.07 HR=2.23, respectively; P values from <0.001 to 0.033). Based on results of uni- and multivariable analysis, we developed a prognostic model for predicting the effect of comorbidities on NLD and OS. For each comorbidity, risk scores were estimated from the coefficients of the Cox regression. This MDS-specific comorbidity index (MDS-CI) allowed us to identify 3 groups of patients with different probability of NLD and OS (HR 2.78, P<0.001; HR 1.67 P=0.001), and provided a better stratification than the available non MDS-specific indices. Focusing on WPSS categories [J Clin Oncol 2007; 25:3503–10], MDS-CI significantly stratified survival of patients with very-low, low and intermediate risk groups (P<0.001), while it had no effect in high and very-high risk groups. We then investigated the relationship between transfusion-dependency, secondary iron overload and comorbidities. Heart failure (28% vs. 18% P=0.001) and cardiac death (69% vs 55% P=0.03) were significantly more frequent in transfusion-dependent patients. In a Cox analysis with time-dependent covariates, transfusion-dependent patients showed an increased risk of NLD (HR=2.12 P=<0.001), heart failure (HR 1.34 P=0.03), and cardiac death (HR 2.99 P=0.01). The development of secondary iron overload significantly affected the risk of NLD and OS (HR=1.25 and 1.16 respectively, P<0.001), and this effect was maintained after adjusting for transfusion burden. Iron overload specifically increased the risk of developing heart failure (HR=1.17, P<0.001). In summary, the presence of non-hematological comorbidities significantly worsens the survival of MDS patients. Transfusion-dependency and secondary iron overload are associated with an increased risk of cardiac complications and cardiac death. The MDS-CI might be a useful tool for clinical decision making in patients with myelodysplastic syndromes.