Predictive score of haematological toxicity in patients treated with linezolid.

Abstract

The aims of our study were to determine the factors associated with developing haematological toxicity (HT) in patients taking linezolid (LZD), to develop a predictive model of HT in these patients, and to evaluate factors associated with 30-day mortality. This was an observational retrospective cohort study of patients treated for at least 5days with LDZ in 2015. Demographic, clinical and analytical data were collected. Development of HT was defined as a 25% platelet count decrease between the basal count and the 1-week lab test. Five hundred forty-nine patients were finally included, mean age was 73.3 (SD 15.4) years, and 303 (55.2%) were men. One hundred seventy-five (30.1%) patients achieved HT criteria during treatment with LZD and 41 (7.5%) died. The final model included the presence of cerebrovascular disease (2 points), moderate or severe liver disease (2 points), renal failure (2 points) and basal platelet count less than 90,000/mm3 (8 points). This new model showed an AUC of 0.711 (IC 95% 0.664-0.757; p < 0.001) to predict the development of HT. The probability of HT based on this classification was 6.2, 29.9 and 76.5% for low (0-4 points), intermediate (5-10 points) and high risk (>10 points), respectively. The independent variables associated with 30-day mortality were metastatic solid tumor, lymphoma, age >75years and HT. This score could help in the identification of patients with high risk for HT and assess the use of an antibiotic other than LZD, an important issue considering its relation with 30-day mortality observed in our study.