Patients treated with multicycle chemotherapy can exhibit large interindividual heterogeneity of haematotoxicity. We describe how a biomathematical model of human granulopoiesis can be used to design risk-adapted dose-dense chemotherapies, leading to more similar leucopoenias in the population. Calculations were performed on a large data set for cyclophosphamide/doxorubicin/vincristine/prednisone (CHOP)-like chemotherapies for aggressive non-Hodgkin lymphoma. Age, gender, Eastern Cooperative Oncology Group performance status, lactate dehydrogenase and the degree of leucopoenia within the first therapy cycle were used to stratify patients into groups with different expected severity of leucopoenia. We estimated risk-specific bone marrow toxicities depending on the drug doses administered. These toxicities were used to derive risk-adapted therapy schedules. We determined different doses of cyclophosphamide and additional etoposide for patients treated with CHOP-14. Alternatively, the model predicted that further reductions of cycle duration were feasible in groups with low toxicity. We also used the model to identify appropriate granulocyte colony-stimulating factor (G-CSF) schedules. In conclusion, we present a method to estimate the potential of risk-specific dose adaptation of different cytotoxic drugs in order to design chemotherapy protocols that result in decreased diversity of leucopoenia between patients, to develop dose-escalation strategies in cases of low leucopoenic reaction and to determine optimal G-CSF support.