A rapidly clinically translatable, closed-loop drug delivery system for personalization of drug dosing

Introduction::



While the modalities for treating cancer have advanced significantly over the past decade, the method of dosing chemotherapeutics has not changed, and as a result, the plasma concentration of chemotherapeutics can vary widely, leading to worse therapeutic outcome.(1–6) To date, chemotherapies are typically dosed based on body surface area (BSA), which does not take into account significant variabilities in pharmacokinetics (PK) in patients with the same BSA (Figure 1).(1–4) Moreover, within the same patient, the plasma concentration of drug administered via continuous infusion can vary significantly, in some patients with a 10-fold range, due to circadian rhythm-induced changes in expression of the drug metabolizing enzyme.(7) The DuBois & DuBois formula for calculating ideal BSA based on a patient’s height and weight was developed in 1916 by empirically fitting data from only nine people, and is still used clinically.(5) Moreover, the DuBois and DuBois formula was further simplified into the widely used Mosteller formula for estimating BSA in a highly cited NEJM paper in 1987, which results in additional variation to the BSA estimate in the DuBois & DuBois formula.(8) The result of BSA-based dosing is that patients may be overdosed or underdosed, resulting in toxicity or lack of efficacy, respectively. Moreover, chemotherapy induced toxicity can cause patients to switch, delay or stop their cancer treatment. Unfortunately, underdosing is often not recognized clinically, which could lead to tumor growth or metastasis.(1–4) This highlights the need to develop an intervention that can maintain the plasma concentration of drugs within its therapeutic window.





Materials and Methods::



We introduce a closed-loop drug delivery system that can determine the concentration of drug and adjust its infusion rate to the patient to reach the drug’s target concentration, under a range of pharmacokinetics (PK) conditions. We designed and built our closed-loop system using commercially available devices, which could enable rapid translation of our system to the clinic. We call our system the Closed-Loop AUtomated Drug Infusion regulAtor (CLAUDIA). We first investigate if CLAUDIA can accurately determine the concentration of 5-FU in spiked whole blood samples by following the 2018 FDA guidelines to industry for bioanalytics. We then investigate if CLAUDIA can control the concentration of 5-FU in vivo in rabbits according to multiple different concentration-time profiles, which could be useful for chronomodulated dosing of chemotherapies--where the dose of chemotherapy is altered according to the time of day when the cancer cells are most susceptible to death by the chemotherapy. We then perturb the PK of the rabbit by administering an inhibitor to DPD--the enzyme that metabolizes 5-FU--to model the PK variability observed across the population. We investigated the ability of CLAUDIA to control the concentration of 5-FU under these perturbed settings. Lastly, we performed a cost-effectiveness analysis, using a Markov model with three states (healthy patient, cancer patient, and progressed cancer patient), using estimates of transitions and costs from the literature.



Results, Conclusions, and Discussions::



We first demonstrated CLAUDIA is capable of detecting 5-FU concentrations over a clinically-relevant range of concentrations within 8-minutes. The relatively short time from blood sampling to measuring the concentration is critical because the shorter the delay time in the system, the better the performance of any closed-loop system.



We demonstrated that CLAUDIA is capable of controlling 5-FU concentrations according to a variety of concentration-time profiles (Figure 2). Importantly, the data demonstrate the if the dose determined by BSA is chosen correctly, the concentration of 5-FU can remain in the target window for the majority of the time (Figure 2A). However, if the dose is selected incorrectly, then the patient/rabbit can be either underdosed or overdosed (Figure 2B-C). In contrast, CLAUDIA kept the concentration of 5-FU near the target window in multiple different rabbits (Figure 2D-F).



Next, we sought to investigate the ability of CLAUDIA to control the concentration of 5-FU in the presence of a disturbance caused by administering an inhibitor to DPD, the enzyme that metabolizes 5-FU (Figure 3A). Importantly, we observed an ~7-fold change in 5-FU concentration across our animal experiments, which mimics the 8-fold PK variability of 5-FU seen clinically (Figure 2 and 3).(9) While all experiments in the BSA-based dosing group reached concentrations of >18 mg/L when the DPD inhibitor was added (target: 2.5-3 mg/L), when we allowed CLAUDIA to control the concentration of 5-FU, the concentration of 5-FU only exceeded 7 mg/L at one timepoint and it did not exceed 5 mg/L at any other timepoints (Figures 3E-G). Overall, the CLAUDIA group deviated much less from the target range (Figure 3H), increased the time in range (Figure 3I), and resulted in an area under the curve (AUC) much closer to the target AUC than BSA-based dosing (Figure 3J).



Lastly, our cost-effectiveness study found that CLAUDIA was cost-effective compared to BSA with an ICER of $92,500 per quality adjusted life year (QALY).



Overall, our results support the further development of CLAUDIA towards clinical translation, as it can control the concentration-time profile of 5-FU and may thereby improve the safety and efficacy of 5-FU for many patients.







Acknowledgements (Optional): :



We thank the members of the Langer and Traverso Labs for their support on this work. We thank Fenton Woods Laboratories for the use of the LC-TOF instrument used in these studies and for the support and guidance in pre-clinical analysis. This work was supported in part by: the National Science Foundation Graduate Research Fellowship Program under Grant No. 1745302 (L.D.), and the Karl van Tassel Career Professorship MIT, the Department of Mechanical Engineering, MIT (G.T.). Figure 1 was created with Biorender.com.



References (Optional): :




1. Lee, J. J., Beumer, J. H. & Chu, E. Therapeutic drug monitoring of 5-fluorouracil. Cancer Chemother. Pharmacol. 78, 447–464 (2016).



2. Cespedes Feliciano, E. M. et al. Body Composition, Adherence to Anthracycline and Taxane-Based Chemotherapy, and Survival After Nonmetastatic Breast Cancer. JAMA Oncol. (2019).



3. Kline, C. L. B. et al. Personalized dosing via pharmacokinetic monitoring of 5-fluorouracil might reduce toxicity in early- or late-stage colorectal cancer patients treated with infusional 5-fluorouracil-based chemotherapy regimens. Clin. Colorectal Cancer 13, 119–126 (2014).



4. White, J., Hu, P., Bonen, M., Bamat, M. K. & von Borstel, R. Point-of-care (POC) diagnostic assay for 5-fluorouracil (5-FU) quantitation to enable dose adjustment and detect dihydropyrimidine dehydrogenase (DPD) deficiency. J. Clin. Oncol. 29, e19562–e19562 (2011).



5. Du BOIS, D. & Du BOIS, E. F. Clinical Calorimetry: Tenth Paper. A formula to estimate the approximate surface area if height and weight be known. Arch. Intern. Med. XVII, 863–871 (1916).



6. DeRidder, L., Rubinson, D. A., Langer, R. & Traverso, G. The past, present, and future of chemotherapy with a focus on individualization of drug dosing. J. Controlled Release 352, 840–860 (2022).



7. Mosteller, R. D. Simplified Calculation of Body-Surface Area. N. Engl. J. Med. 317, 1098 (1987).



8. Metzger, G. et al. Spontaneous or imposed circadian changes in plasma concentrations of 5-fluorouracil coadministered with folinic acid and oxaliplatin: Relationship with mucosal toxicity in patients with cancer. Clin. Pharmacol. Ther. 56, 190–201 (1994).



9. Gamelin, E. et al. Correlation Between Uracil and Dihydrouracil Plasma Ratio, Fluorouracil (5-FU) Pharmacokinetic Parameters, and Tolerance in Patients With Advanced Colorectal Cancer: A Potential Interest for Predicting 5-FU Toxicity and Determining Optimal 5-FU Dosage. J. Clin. Oncol. (1999).