The paradigm for implementing pharmacokinetic monitoring of antineoplastic drugs in patients with osteosarcoma is represented by methotrexate. Given the high interocassion (intra-individual) variability in methotrexate clearance, monitoring must be performed at each administration cycle. For this drug there is an analytical technique of proven reliability as well as a PK/PD model, derived from a bicompartmental pharmacokinetic model, which allows personalized dose adjustment. The purpose of methotrexate Cp monitoring is to guarantee that the treatment is effective [by obtaining a Cmax of 1,000 µmol/L at the end of the infusion (8th hour)] and safe (by preventing the development of toxicity preventing Cp from remaining above 0.08 µmol/L for more than 96 hours)27. During pharmacokinetic monitoring it is essential to ensure proper hydration (3 L/m2), diuresis (300 mL/h) and urine alkalinization (pH ≥ 8) in order to promote adequate clearance and prevent the risk of precipitation and associated renal damage. Folinic rescue is always mandatory 12 hours after discontinuation of methotrexate and should be maintained for at least 60 hours or until Cp falls below 0.05 µmol/L. Likewise, when apparent half-life at 12 hours post-infusion is longer than 3 hours, the patient is considered to be at high risk of toxicity. To avoid this, rescue with colestipol should be initiated. Finally, the dosing regimen for the next cycle is established according to the Cp obtained during the clearance phase (24, 48 and 72 h post-infusion). Using this method means that patients with osteosarcoma treated with high-dose methotrexate reach the target Cmax in 70% of the cycles compared to 49% when fixed doses are administered28.

In the treatment of acute lymphoblastic leukemia (ALL), a correlation between methotrexate exposure and clinical effectiveness has been established. Evans et al. showed a relationship between the risk of relapse and methotrexate clearance in 108 children with ALL. Adjusting the methotrexate dose according to the patient's drug clearance profile during 24-hour infusion significantly improves clinical outcomes in this subpopulation of patients29.

However, in clinical practice, monitoring of methotrexate at high doses (> 500 mg/m2) is mainly applied in the prevention of toxicity rather than for adjusting the dose adjustment and optimizing the drug's effectiveness. Currently, delayed clearance of methotrexate is diagnosed when Cp are above 10 µmol/L at 24 hours, 1 µmol/L at 48 hours or 0.1 µmol/L at 72 hours30.

A recent study set about genotyping the organic anion polypeptide transporter 1B1 (OATP1B1) since this transporter allows uptake of methotrexate from circulating blood by hepatocytes. This transporter's genetic variants are associated with significant variability in drug clearance. The rs4149056 variant in particular is associated with decreased hepatic transport activity and has been shown to cause a moderate increase (13-26%) in the plasma clearance of the drug31,32.

Heterogeneity in the polymorphisms identified in the metabolism of 5-fluorouracil (5-FU), especially those related to dihydropyridine dehydrogenase (DPD), and their inherent intra- and inter-patient variability in patients receiving 5-FU continuous intravenous infusion, warrant pharmacokinetic monitoring of 5-FU. Nomograms, typically based on the establishment of a relationship between the Cp measured at a given time (or the AUC) and clinical response, have been used to individualize the dose of 5-FU to be administered. Thus, after determining the AUC 48 hours after the start of IV infusion, the appropriate dose for the patient is calculated to ensure the drug's effectiveness and safety as adjuvant and neoadjuvant therapy and as treatment for advanced disease in patients diagnosed with colorectal carcinoma (CRC) and squamous cell carcinoma of the head and neck (SCCHN)33. The AUC analysis has shown a good correlation with grade 3-4 toxicity, tumor response and survival, and several studies suggest better patient outcomes when a value between 20-24 mg-h/L is achieved34 (extended to 20-30 mg-h/L in continuous 46-hour infusions35). Using conventional dosing criteria for 5-FU based on body surface area, 20-30% of patients would achieve the therapeutic target for AUC and approximately 60% of patients would have lower exposure33. Pharmacokinetic monitoring of 5-FU in clinical practice has increased due to the incorporation of an automated immunoassay technique for the determination of 5-FU in plasma and has contributed to reducing variability in its exposure and toxicity rates (mainly diarrhea and hand-foot syndrome), although the results in terms of efficacy are more limited, with an acceptable increase in the response rate (33.7% vs. 18.3%)34 but poor survival improvement results in randomized clinical trials36.

The information provided by genetic analyses for the personalization of doses of fluoropyrimidines (5-FU and capecitabine) is complementary to monitoring exposure to 5-FU. Polymorphisms have been identified in the DPYD gene, which synthesizes the DPD involved in its metabolism and inactivation, resulting in a complete (0.1% of the population) or partial (3-8% of the population) deficiency of this enzyme. Different expert groups (ESMO, CPIC and DPWG)37–39 have recommended that, prior to treatment with fluoropyrimidines, the presence of four variants of the DPYD gene that have been shown to have a considerable functional and clinical impact should be analyzed: DPYD*2A and c. 1679T> G (complete deficiency), which result in complete DPD deficiency, and c.2846A> T and c.1236G> A, which result in partial deficiency. Based on the presence of these polymorphisms, recommendations are made regarding personalization of the 5-FU dose prior to initiation of treatment to reduce the potential of myelosuppression, gastrointestinal toxicity (diarrhea and/or mucositis) and neurotoxicity among other adverse events. Patients heterozygous for any of these 4 allelic variants (8% of the population) presented higher grade ≥ 3 toxicity with respect to native patients (39% vs. 23%; p = 0.0013), and DPYD genotype-guided dose personalization reduced the relative risk of developing this toxicity40. Nonetheless, it is estimated that only about 50% of cases with impaired DPD enzyme activity can be identified by these 4 variants.

The Spanish Medicines and Health Products Agency (AEMPS) has recently issued a safety note recommending genotype and/or phenotype testing for DPD deficiency in patients who are candidates for treatment with fluoropyrimidines as well as pharmacokinetic monitoring of 5-FU in patients with partial DPD deficiency. AEMPS also declared its use to be contraindicated in patients with complete DPD deficiency41.

Oral antineoplastics (OAs), which are used at fixed doses in most patients, present with high inter-individual variability in their pharmacokinetic parameters, which also include release and absorption phases that are highly influenced by various pathophysiological and/or external factors. Thus, the inter-individual variability in exposure when standard doses of tyrosine kinase inhibitors are administered is between 19 and 100%. Examples of sources of pharmacokinetic variability are the first-pass effect; the presence of interactions at a metabolic level (especially via cytochrome P450) and with food; the existence of polymorphisms in transporters; renal or hepatic insufficiency; and potential problems of non-adherence to treatment, which is especially important in long-term treatments. Consequently, when standard doses are used, exposure is very likely to be below (> 30%) or above (> 15%) the therapeutic target42.

Significant exposure-response and/or exposure-toxicity relationships have been demonstrated for most OAs43. However, despite compliance with the premises required for therapeutic monitoring44, the latter's usefulness in the context of dose personalization, or in the application of alternative interventions to ensure therapeutic concentrations and increase success rates, has not been demonstrated in prospective randomized studies. Nevertheless, for some OAs pharmacokinetic monitoring is considered useful as retrospective studies have associated it with a significant clinical benefit and the maintenance of drug plasma levels within the therapeutic range, and its implementation has been shown to be feasible in clinical practice (Table 3). A prospective study is currently underway of 23 OAs in a population of more than 600 patients (with historical controls) with the aim of determining the feasibility, effectiveness (proportion of patients below the therapeutic range, objective response rate according to RECIST 1.1, progression-free survival) and safety of dose personalization based on therapeutic monitoring45. Likewise, the lack of optimal standards or circuits for implementing pharmacokinetic monitoring as part of the therapeutic personalization of OAs in clinical oncology care is evident46,47.

Monoclonal antibodies (mAb) directed against cellular antigens have complex pharmacokinetic profiles characterized by the presence of multiple clearance pathways (proteolysis, internalization in effector cells and autoantibodies), with both dose- and time-dependent elimination half-live. Thus, clearance of the mAb may be lower at high doses than when lower doses are administered due to saturation of the antigenic receptors. On the other hand, when the antigen concentration is high, half-life decreases because the mAb binds to its epitope and is more rapidly eliminated from the body; as the antigen is depleted, clearance from plasma decreases and the plasma clearance half-life increases. Consequently, mAb clearance is a kinetic parameter that seems to play a major role in serum concentrations at the end of the dosing interval. Moreover, its high inter-individual variability (between 29 and 97%) seems to determine the risk of placing the patient outside the proposed therapeutic window62. Generally speaking, a high inter-individual variability has been observed in pharmacokinetic parameters following exposure to mAbs depending on various biometric (gender, weight or body surface area) and pathophysiological factors (glomerular filtration rate, performance status, tumor burden, histology, bone marrow infiltration, etc.). Thus, only 32% of the inter-individual variability in the rituximab clearance rate has been associated with the amount of circulating CD20 antigen, demonstrating the presence of other covariates that influence the drug's elimination63. On the other hand, the complexity is greater for cytotoxic mAb-drug conjugates or bispecific T-cell engagers (BiTEs) or multispecific mAbs. In addition to the complex pharmacokinetics of mAb, there are the characteristics of the cytotoxic agent and its release kinetics. Several studies have established significant relationships between exposure and response, which would justify the performance of prospective studies to validate the application of pharmacokinetic monitoring in clinical practice, since exposure is a predictive variable of the efficacy and/or toxicity of these targeted therapies64.

Immune checkpoint inhibitor mAbs (anti-CTLA-4, PD-1 and PD-L1) exhibit pharmacokinetic properties typical of mAbs along with a wide inter-individual variability in their kinetic parameters (30-50%) and therapeutic and toxic response65. Exposure-response relationships have been described for anti-CTL4 (ipilimumab and tremelimumab) and anti-PD-1 (nivolumab, pembrolizumab and cemiplimab) mAbs although limited data is available for anti-PD-L1 (atezolizumab, avelumab, durvalumab). For some of these mAbs, time-dependent clearance has been demonstrated (nivolumab, pembrolizumab and anti-PD-L1) and, in the case of atezolizumab, the high rate of immunogenicity observed could affect its pharmacokinetic properties. In any case, additional PK/PD studies for this group of mAbs would allow a more precise definition of the association between the concentration and/or clearance of these drugs and clinical response and would confirm the usefulness of monitoring them in clinical practice66. In addition, the correlation between weight and clearance observed in the case of ipilimumab, pembrolizumab and avelumab should be confirmed to legitimize certain claims related to the choice of a fixed dose versus a dose normalized to body weight. Variations in their nature (chimeric, humanized or human), type (IgG1-4) and even binding sites (monospecific, bispecific) and affinity may explain the differences between these mAbs’ various pharmacokinetic profiles (Table 4).

Asparaginase is used in chemotherapeutic regimens for the treatment of acute lymphoblastic leukemia (ALL) with increased cure rates, especially in the pediatric population. Achievement of optimal results with this therapy depends on complete and sustained depletion of serum asparagine. However, high inter-individual variability in exposure, differences in pharmacokinetic properties between different asparaginases and the formation of anti-asparaginase antibodies make it difficult to predict the degree of asparagine depletion that occurs after administration of a given dose. Under physiological conditions, the range of circulating asparagine concentrations is between 40 and 80 µM and, although there is no universal consensus, some investigators have established that asparagine deficiency is complete when plasma concentration is less than 0.1-0.2 µM. Thus, in a study in 214 children in first relapse of ALL, patients with serum asparagine levels < 1 µM on day 14 were more likely to achieve a second remission compared to patients with higher levels95. However, due to the rapid ex vivo metabolism of asparagine in the presence of asparaginase, and the complexity of the analysis, direct determination of asparagine concentration in vivo is challenging. The ALL therapeutic monitoring program should also include determination of asparaginase activity (≥ 100 IU/L; between 100-250 IU/L), and of anti-asparaginase antibodies, thus identifying situations of subclinical hypersensitivity and suboptimal enzyme activity (silent inactivation), personalizing current treatment, or evaluating the benefit of a change of formulation96. This would allow identification of situations where patients develop allergic-type reactions, but no enzymatic inactivation, and therefore do not require a change of formulation to ensure adequate treatment.

Busulfan represents the cornerstone of many myeloablative regimens. It is frequently administered in a two-hour infusion every 6 hours for 4 days. Monitoring of the therapy is recommended during the first dose to individualize subsequent doses by calculating the AUC from three plasma samples (end of infusion, 4 and 6 hours post-infusion). Currently, the therapeutic target is an AUC between 900-1,350 µM-min in pediatric patients and 900-1,500 µM-min in adults (four times higher for daily dosing schedules). AUC can be estimated by non-compartmental analysis, population pharmacokinetic models, or pooled sampling methods with a small number of samples97,98. Monitoring minimizes the incidence of hepatic veno-occlusive disease and neurologic toxicity, as well as graft failure and disease recurrence. Pediatric patients present with greater pharmacokinetic variability than adult patients. The variability is highest among newborns and infants < 9 kg, 40% of whom reach off-target AUC values, which makes monitoring particularly necessary in this subpopulation of patients99.

Carboplatin dosing is based on patients’ renal function (the main source of pharmacokinetic variability) and exposure to this drug (AUC) has been shown to correlate well with therapeutic response and toxicity in both pediatric and adult patients. Carboplatin monitoring is recommended mainly in patient subpopulations where exposure is more unpredictable, a positive impact of the treatment having been demonstrated for example in children with cancer on intensification regimes based on high carboplatin doses. In such cases, there is a higher risk of severe toxicity, and the degree of bioavailability may be more variable than in adult patients100. Likewise, monitoring can have a significant impact on anuric patients and those with renal failure, where non-renal clearance predominates. In obese patients, dosing based on total body weight (TBW) overestimates the renal clearance of carboplatin by more than 20%; conversely, ideal body weight (IBW) may underestimate clearance. The best weight value for carboplatin dosing in obese patients is the intermediate value between ideal and total weight (IBW*0.512x[TBW-IBW])101. Regarding myelotoxicity, population PK/PD models have described the impact of exposure to carboplatin on thrombocytopenia to be 1.45 times higher when administered in schedules with etoposide; 2.33-times higher with gemcitabine; and 0.764 times higher (protective effect) with paclitaxel102.

Several PK/PD studies have analyzed exposure-effectiveness relationships or exposure-toxicity relationships for taxanes such as paclitaxel and docetaxel. The usefulness of paclitaxel monitoring using Bayesian methods in reducing inter-individual variability has been demonstrated. However, although the time during which paclitaxel concentration exceeds 0.05 µM (42.7 µg/mL) may be considered a predictor of hematologic toxicity, neuropathy, and therapeutic response, further studies are required to confirm its clinical benefit99. In a prospective randomized controlled trial conducted to evaluate the feasibility and performance of docetaxel dosing based on AUC as a predictor of myelotoxicity and febrile neutropenia, inter-individual variability in the individualized treatment group was over 50% lower than in the body surface area-based dosing group103.

Irinotecan is another antineoplastic that is a good candidate to monitoring due to the high inter-individual variability in the metabolism of its active metabolite SN-38 mediated by UDP glucuronosyltransferase 1A1 (UGT1A1). In this case, a genetic approach is the most appropriate way of reducing the risk of toxicity (neutropenia and digestive toxicity). Several functional polymorphisms have been identified in the UGT1A1 gene (UGT1A1*28 in European, African and Latin populations and UGT1A1*6 in Asian populations mainly; UGT1A1*36 and UGT1A1*37 almost exclusively in African populations), leading to low enzyme expression and decreased glucuronidation activity, which have a significant impact on the incidence of toxicity. As a result, determination of these polymorphisms should be routinely carried out to identify patients at risk of severe toxicity where lower doses should be administered at initiation of treatment104,105. This strategy is especially useful in dose-intensive irinotecan regimens that have only been shown to be safe in patients homozygous for the UGT1A1*1 variant. The Royal Dutch Pharmacists Association - Pharmacogenetics Working Group has recommended an initial dose reduction of 30% for individuals homozygous for the UGT1A1*28 variant, or poor metabolizers, and a subsequent increase based on neutrophil count106. On the other hand, the Group of Clinical Onco-pharmacology (GPCO-Unicancer) and the National Pharmacogenetics Network (RNPGx) have recommended an initial dose reduction of 30% in UGT1A1*28 homozygous individuals for dosing schedules between 180 and 230 mg/m2 (if dose is lower than 180 mg/m2 no dose adjustment is proposed). In contrast, administration of the full dose in dosing schedules ≥ 240 mg/m2 is only recommended in patients homozygous for the UGT1A1*1 variant and in UGT1A1*1/*28 heterozygous patients in the absence of risk factors and under strict surveillance107.