Modelling chemotherapy effects on granulopoiesis

Abstract

Background: Although the growth-factor G-CSF is widely used to prevent granulotoxic side effects of cytotoxic chemotherapies, its optimal use is still unknown since treatment outcome depends on many parameters such as dosing and timing of chemotherapies, pharmaceutical derivative of G-CSF used and individual risk factors. We showed in the past that a pharmacokinetic and -dynamic model of G-CSF and human granulopoiesis can be used to predict the performance of yet untested G-CSF schedules. However, only a single chemotherapy was considered so far.

Results: Model assumptions proved to be feasible in explaining granulotoxicity of 10 different chemotherapeutic drugs or drug-combinations applied in 33 different schedules with and without G-CSF. Risk groups of granulotoxicity were traced back to differences in toxicity parameters.

Conclusion: We established a comprehensive model of combined G-CSF and chemotherapy action in humans which allows us to predict and compare the outcome of alternative G-CSF schedules. We aim to apply the model in different clinical contexts to optimize and individualize G-CSF treatment.

Figure 1

Structure of the granulopoiesis model under chemotherapy and G-CSF treatment. Boxes represent major cell- or cytokine compartments of the model: S = haematopoietic stem cells, CG = granulopoietic progenitors, PGB = granulopoietic precursors, MGB = maturing granulopoietic precursors in bone marrow, GRA = mature granulocytes in blood, LY = lymphocytes. We modelled two G-CSF derivatives (fil = Filgrastim, peg = Pegfilgrastim). Arrows represent cell/cytokine fluxes and interactions. CX represents the strength of chemotherapy (see below).

Figure 2

Construction of toxicity functions exemplified by the high-CHOEP regimen and its toxic effect on stem cells. The Heavyside functions for a single application of the combination of Cyclophosphamide, Doxorubicin and Vincristin (at time point 0) and for single Etoposide (applications at time points 0, 1 and 2) are determined (first column), delayed (second column) and multiplied with the corresponding stem cell toxicity factors (third column). Finally, the functions were added resulting in the overall toxicity function.

Figure 3

Toxicity of CHOP chemotherapy. Time courses of normalised cell counts of different cell stages after a single application of CHOP chemotherapy. A: CHOP effect only on stem cells. B: CHOP effect only on PGB. C: CHOP effect on all cell stages.

Figure 4

Simulation results for CHOP-21, CHOP-14, CHOEP-21, CHOEP-14, younger patients. We show simulated cell counts for CHOP and CHOEP chemotherapy. Dots represent patient medians, grey lines represent interquartile range of patient data, black squares correspond to chemotherapy administrations, “+” correspond to days with G-CSF-injections. Clinical data originate from our collaborating clinical trials group [39], see Table 2.

Figure 5

Simulation results of selected complex chemotherapies. We present results for the two breast cancer therapies, EC-T and ETC with Filgrastim on days 3–10 (first row). Note that in these schedules, chemotherapeutic drugs differ between cycles: For EC-T the drugs epirubicine and cyclophosphamide where applied in combination in the first four cycles. The single drug paclitaxel was applied for the last four cycles. For ETC, the single drug epirubicine was applied in three cycles followed by three cycles of paclitaxel and three cycles of cyclophosphamide. We also present two therapies of advanced Hodgkin’s lymphoma, BEACOPP-21 and BEACOPP escalated with Filgrastim on days 8–15, in which multiple drugs are administered at different time points per cycle (second row). Dosages can be found in Table 1. Dots represent patient medians, grey lines represent interquartile range of patient data, squares represent chemotherapy administrations, “+” denote G-CSF-injections.

Figure 6

Correlation of stem cell toxicity and peripheral toxicity. We determined cumulative toxicities of stem cells and mature blood cells for each drug or drug combination considered. Toxicities are expressed in terms of AOC of normalized cell counts applying the steady-state value 1 as threshold. AOC is calculated over 28 days. Only a single injection of chemotherapy was simulated for this purpose. The unit of AOC is “d”. A good correlation between stem cell and peripheral toxicity can be observed (Spearman’s r = 0.88). Toxicity relations between schedules are plausible. Note that schedules may be attributed to different groups of patients.

Figure 7

Validation scenarios. We compare model results with clinical data from validation scenarios: A: CHOP with Pegfilgrastim 6000 μg on day 2 (data: [37]), B: ESHAP with Pegfilgrastim on day 5 (data: [45]), C: CHOP with 480 μg Filgrastim on cycle-days 6–12, for elderly patients, data: [43]. Dots represent patient medians, squares represent the chemotherapy administrations, + are time points with G-CSF-injections.

Figure 8

Analysis of CHOP with Filgrastim (age > 60 years). Simulated cell counts for CHOP-14, with Filgrastim 480 μg on days 5–13 (black line), days 2–8 (grey line) and days 3–12. Filgrastim at day 2–8 results in particularly low leukocyte counts. Better results are obtained by the Filgrastim application on days 3–12 or days 5–13.

Figure 9

Modified Filgrastim schedules for CHOP-14 in elderly patients. Predicted WBC AOC (applying a threshold of 4000/μl, calculated over 84 days, unit of AOC is 1000/μl*d) under CHOP-14 with Filgrastim: G-CSF injections start at day 7 in each cycle. We modified the number of Filgrastim injections and its doses. The color scale on the right corresponds to AOC values.