Optimización de tratamientos con irinotecan y capecitabina en pacientes con cáncer colorrectar basada en técnicas de farmacocinética e inteligencia artificial

Introduction: Colorectal cancer is the fourth cancer with the highest incidence and the second with the highest mortality rate, according to 2018 data. Adenocarcinoma is the fundamental form in which it occurs, occupying 90% of the cases. The drugs used to combat this disease are numerous and have narrow therapeutic ranges and strong adverse effects. Controlling plasma levels is essential to achieve optimal pharmacotherapy. Irinotecan and capecitabine have been the drugs analyzed in this work. These drugs are administered both in monotherapy and in combination, in schemes such as FOLFIRINOX or FOLFIRI for irinotecan and XELOX for capecitabine. Both drugs have an active metabolite responsible for the therapeutic action of the drug. These metabolites, in addition to the treatment effect, are responsible for a large part of the toxicities derived from the treatment, hence, their correct characterization is pharmacologically relevant. Hypothesis: The prevalence of colorectal cancer and the necessity to increase positive results in health generates the need to incorporate new tools, such as those included in the framework of artificial intelligence, which, together with other more classic ones such as pharmacokinetic and pharmacodynamic modeling, make it possible to facilitate optimal use of chemotherapy in routine care practice. Results: A compartmental pharmacokinetic model has been obtained for each of the drugs and their corresponding metabolites, making use of parametric and non-parametric methodologies. The precision indices of these models reach values of R2=0.964 for the case of irinotecan and R2=0.886 for capecitabine. Moreover, the capecitabine model has permitted to obtain optimal sampling times for this drug. On the other hand, models based on machine learning have been developed to predict irinotecan derived toxicities such as late diarrhea, leukopenia and neutropenia with success rates of 91, 76 and 75%, respectively. These models are based both on particular characteristics of each patient and on values of their pharmacokinetic parameters. Finally, the proposed models for irinotecan have been integrated in a software enabling their correct use in clinical practice. Conclusions: The kinetics of irinotecan and capecitabine and their corresponding metabolites have been correctly characterized, allowing an individualized adjustment of the concentrations over time of each of the patients. Moreover, in the case of irinotecan, the reconversion from the glucuronide metabolite to the active metabolite, due to enterohepatic reabsorption, has been characterized for the first time in the literature, which is a fundamental feature of this drug. The models based on artificial intelligence allow to correctly predict the possibility for a new patient to develop late diarrhea, leukopenia and neutropenia, by means of particular characteristics of that patient. A strong relationship between pharmacokinetic parameters and the studied toxicities for irinotecan has been demonstrated, for the areas under the plasma curves and the maximum concentrations of this drug are linked with the degree of toxicity. The developed software permits to apply the results obtained in this thesis in clinical practice