Predictive model of response to epoetin-alfa in anemic patients with solid tumors receiving chemotherapy: Results from a multicenter prospective study

Abstract

20607 Background: The purpose of this prospective, multicenter study was to develop a model to predict hematopoietic response (HR) to epoetin alfa (EA) in anemic patients with solid tumors receiving chemotherapy (CT). Methods: Patients’ demographics, clinical and tumor characteristics, laboratory parameters and CT agents were collected just before the administration of each course of CT. Predictive factors for response (defined as an increase in Hb ≥ 2 g/dL or Hb level ≥ 12 g/dL during treatment) were tested in a univariate and multivariate analyses using logistic regression techniques. Finally a ROC analysis was undertaken to measure predictive accuracy. Results: A total of 409 pts (64% males) were included: a cohort of 205 pts received epoetin alfa 10,000 IU tiw and a second cohort of 204 pts received 40,000 IU qw. The response rate was 69.5%. Only four variables were independent predictive factors for HR: gender female (OR: 1.92; 95% CI: 1.05-3.49); percentage of underdose of EA respected to theoretical weight-based dosing [weekly dose administered/(450 IU qw*weigh)] (OR: 3.18; 95% CI: 1.2-8.42); Hb level at the begin of EA treatment (OR: 2.98; 95% CI: 2.2-4.03); change in Hb level after 3 weeks (OR: 2.96; 95% CI: 2.25-3.89). A statistical weight was assigned for each predictive factor on the basis of regression model beta coefficients to develop a predictive scoring system. The probability of HR was calculated over a range of total score (Table 1). A risk score of 5 was identified as being the optimal cut-off to maximize sensitivity (83.5%) and specificity (77.5%). Positive and negative predictive values were 89.2% and 67.5% respectively. The ROC analysis obtained an area under the ROC curve of 0.879 (95%CI: 0.845–0.912). Conclusion: This predictive model could be useful to identify patients at low probability of response to EPO in whom the increase of fixed dose of EPO adjusting to weight or the stop of EPO treatment could optimize the use of these molecules. Table 1. Score system and probability of response Predictive factor Score (0–9 points) Score Probability of response Gender: Female 1 0–2 < 15 % Percentage of underdose of EA: (weekly dose administered/450 IU*weight) < 90% 0 ≥ 90% 2 Hb at begin of treatment (g/dL): <8 0 3–4 30–50 % 8–8.9 1 9–9.9 2 10–10.9 3 Change Hb after 3 weeks (g/dL): < 0 0 5–7 60–80 % 0–0.5 0.5 0.6–1 1 1.1–1.5 1.5 >7 > 80 % 1.6–2 2 2.1–2.5 2.5 > 2.5 3 No significant financial relationships to disclose.