Chronic Myeloid Leukemia (CML) is a hematological cancer characterized by the unregulated proliferation of immature myeloid cells in the bone marrow. This leads to a displacement of the normal hematopoiesis with a lethal course if untreated. The introduction of Tyrosine Kinase Inhibitors (TKIs) as a specific therapy has significantly influenced the CML treatment and represents the current first line treatment option for affected patients. Several clinical trials confirmed that TKI treatment can be stopped for some well responding patients without the occurrence of a relapse. It is not yet possible to prospectively identify those patients. Also, the mechanisms leading to a relapse or a treatment-free remission still remain unclear. However, recent clinical trials suggest that an immunological component plays an important role in the long-term disease control. Aim of this work was the expansion of a mathematical CML model by an anti-leukemic immune component to improve the description of the disease behavior of CML patients while treated with TKI and after stopping treatment. A mathematical proof of concept analysis of the model mechanisms leading to relapse or treatment-free remission was performed. Also, it was examined whether such a model can reliably predict the occurrence of a relapse. The BCR-ABL time courses of 21 TKI-treated CML patients, for whom TKI-therapy had been stopped as a clinical intervention, were used in this work. While the adaption of a simplified ODE-model without the immune component could describe the BCR-ABL/ABL decrease during therapy, the model failed to describe a treatment-free remission. Only by expanding the model with an individual immune component, it was possible to correctly reproduce the BCR-ABL/ABL time courses during TKI-treatment and after treatment cessation. Also, an estimation of the individual parameters was only possible by fitting the model to the complete BCR-ABL/ABL time course (including measurements after treatment cessation). Significant differences of the immune parameters determined in this way between relapsing and non-relapsing patients signaled an important influence of the immune component on the relapse behavior. A detailed mathematical analysis of the identified relapse behavior attractor landscapes also suggested that the available patients can be grouped in three general classes (A–C) corresponding to their individual immune response. Certain patients presented an insufficient immune response (class A) and thus, consistently relapsed after stopping treatment as they were unable to prevent a renewed proliferation of residual leukemic cells after treatment cessation. In contrast, some patients showed a sufficient immune response. While patients with a sufficient and strong immune response (class B) required only a minimal treatment duration in the simulations to retain a remission after stopping treatment, the prevention of a relapse for patients with a sufficient and weak immune response (class C) was only possible if an optimal balance between leukemia abundance and immunological activation was achieved before treatment cessation. It could be shown that this balance can theoretically be achieved by an individual titrated and narrowly adapted treatment duration and intensity. Since estimations of the model parameters could only be obtained if the complete data (including post-cessation measurements) were available, it was not possible to predict the individual relapse behavior. Therefore, a 12-months TKI dose reduction simulation was performed for each patient and the resulting BCR-ABL/ABL changes within this period were analyzed. A correlation between the BCR-ABL/ABL increase and the clinical relapse behavior suggested that the BCR-ABL/ABL changes from such system perturbation yields the required information for predictions of the relapse behavior. These simulation results are in qualitative and quantitative agreement with clinical data of the DESTINY trial (NCT01804985). It could be shown that a mathematical CML model is capable of describing the treatment response and relapse behavior of CML patients by incorporation of an immunological control component. Thus, this work supports the results of recent clinical trials which suggest an important role of immune cells for the maintenance of a treatment-free remission. Since a prediction of the individual relapse behavior cannot be obtained from BCR-ABL measurements before stopping treatment, this work highlights the need for further research in this clinical field. Moreover, the results of this work suggest that clinical trials with treatment interventions like a TKI-dose reduction could provide the information required for those predictions.