Support Care Cancer DOI 10.1007/s00520-006-0154-2 ORIGINAL ARTICLE Mark Vincent George Dranitsaris Sunil Verma Cathy Lau Pere Gascon Simon Van Belle Heinz Ludwig Received: 15 May 2006 Accepted: 31 August 2006 # Springer-Verlag 2006 The development and validation of a prediction tool for chemotherapy-induced anemia in patients with advanced nonsmall cell lung cancer receiving palliative chemotherapy Abstract Background: In this study, the development and validation of a cycle-based prediction model for se- vere anemia [i.e., a hemoglobin (Hb) of ≤100 g/l] in patients with advanced nonsmall cell lung cancer (NSCLC) receiving palliative chemotherapy is described. Materials and methods: Data on 536 European patients who were receiving palliative chemother- apy were prospectively collected as part of the European Cancer Anemia Survey [Ludwig et al., Eur J Cancer 40:2293–2306, 2004]. The patient sample was then randomly divided into two-thirds model derivation and one-third validation sample. A third external sample consisting of 76 pa- tients treated at the Toronto Sunny- brook Regional Cancer Centre was separately used to externally validate the model. Multivariable logistic re- gression techniques were applied to develop the initial model. A risk scoring system based on the regres- sion parameters was then created ranging from 0 to 15. Results: Precycle Hb, patient body surface area, advanced age, poor performance status, recurrent or persistent disease, and the use of platinum or gemcita- bine-based chemotherapy were iden- tified as important predictors for anemia. A prechemotherapy risk score of ≥8 to <10 was identified as the optimal cut off to maximize the sensitivity (83.1%) and specificity (67.8%) of the prediction tool. Patients with a score of ≥8 would be considered at high risk for developing anemia after a particular cycle of chemotherapy. Discussion: We developed and validated an anemia prediction tool for advanced stage NSCLC patients receiving palliative chemotherapy. To make the model available for easy use and access, we have incorporated it on to our risk prediction website: http://www. PredictPatientEvents.com. It is hoped that this risk model will enhance patient care by optimizing the fre- quency of Hb testing and/or the use of preventative therapies. Keywords Anemia . Risk . Prediction . Lung cancer . Chemotherapy Introduction Lung cancer is among the commonest of the lethal malignancies throughout the world [22]. Nonsmall cell lung cancer (NSCLC) comprises about 80% of lung cancer and, at presentation, about 40% of patients have “distant” disease [20]. Some patients with earlier stage disease are unfit for or are unwilling to undergo radical curative-intent treatments; another group relapsed after prior combined modality therapies. Together, these patients, along with the bona fide metastatic patients, are classified pragmatically as “advanced.” This term implies incurability and they require some form of palliation. M. Vincent (*) London Regional Cancer Program, 790 Commissioners Rd. E., London, ON, N6A 4L6, Canada e-mail: mark.vincent@lhsc.on.ca Fax: +1-519-6858624 G. Dranitsaris . S. Verma Toronto Sunnybrook Regional Cancer Centre, Toronto, ON, Canada C. Lau Ortho Biotech Inc., Toronto, ON, Canada P. Gascon Hospital Clinic, Barcelona, Spain S. Van Belle Ghent University Hospital, Ghent, Belgium H. Ludwig Wilhelminenspital, Vienna, Austria