Among childhood cancer survivors, cardiac disease is the leading non-malignant cause for morbidity and mortality [24]. As the vast majority of children diagnosed with cancer are currently cured [25], the prevention of treatment-related toxicities plays a key role. For DOX and other anthracyclines the relationship of PK measures (e.g. AUC, cmax) and treatment outcome has not been definitively established. Nevertheless, the reduction of variability in treatment intensity holds promise to better balance tumour efficacy and the risk of toxicity, in particular late cardiac effects. Seemingly arbitrary thresholds for dose modification and conversion rules as part of empirically-derived treatment regimens along with age-dependent differences in individual PK substantially contribute to this variability. Protocol optimisation is needed and might offer the possibility to increase the safety of DOX administration.
A population PK model-based strategy to adapt DOX doses in young children has been described by our group before [16]. A weak point of this method, however, is the lack of evidence for the target PK parameter. Adapting the DOX dose with the aim to achieve more uniform drug exposure across the age range considers that AUC is presumably the most relevant determinant for therapy efficacy [26]. It neglects, however, the higher organ toxicity of DOX in young children and the role of the peak concentration for cardiac toxicity [3, 8–10]. In a situation when definite clinical evidence is lacking a Delphi consensus procedure may help to sharpen the rationale and pharmacological goals of DOX dose modifications in children with cancer. In contrast to open group discussions this approach permits collecting individual opinions and transforming opinions into a group consensus without being influenced by a single opinion leader [27]. Though evidence is scarce, the Delphi procedure allowed clarifying the pharmacological goals of dose modifications and formulating a standardised dosing concept, based on the collective knowledge and opinion of clinical experts. It should be clearly stated that collective opinion must not be erroneously confused with scientific evidence and should not be seen as indisputable fact. A Delphi procedure does not create new knowledge but rather seeks to make optimal use of already existing knowledge [27].
However, with the consented a priori dose adaptation a consistent strategy applicable to all treatment regimens has become available. The Delphi panel confirmed the initially proposed concept which individualises absolute DOX doses based on patient characteristics (age and BSA) that are predictive for DOX PK. Aiming at uniform drug exposure among children treated according to the same protocol thereby appears to be most reasonable as systemic drug exposure has been widely used as a surrogate marker for dose adaptations [28]. An appropriate dosing equation is available based on the population PK model for DOX (cf. formulas 1 & 2 in the “Methods” section) [16]. In this way optimised treatment regimens may allow for a rational choice of the DOX dose in paediatrics, ideally improving the safety of DOX application. As an extension of the dosing concept described in [16], the experts also recommended an additional reduction of peak levels in very young children by prolonged infusion, thus taking into account the presumed influence of peak levels on cardiotoxicity and the higher cardiac risk of very young patients. In conclusion, modifications of treatment strategies in young children should therefore be based on two aspects, adjustment of the dose and of the infusion duration.
As a prerequisite for the proposed dosing concept the target AUC that should serve as reference needs to be specified. In our example we used the AUC expected for a ‘standard’ 18-year-old boy (i.e. an adult patient) as a reference (fig. 3), as this seems to be straightforward. However, other targets might be even more appropriate. For instance, a target AUC based on the median clearance of a representative patient population has been used for renal function-based carboplatin dosing [29]. Apart from that, the consented prolongation of infusion time in younger children as a measure to reduce peak concentrations might be opposed by clinical practicability (i.e. practicability in an ambulatory care setting) and patient convenience. In addition, the exact influence of infusion time on peak concentrations also requires further investigation.
Constraining the range of DOX doses and infusion times that are applied in current protocols may offer an opportunity to prevent extreme AUC values and, maybe more important, peak concentrations. As described above, a plethora of studies investigated the potentially beneficial impact of prolonged infusion (i.e. lower peak concentration) on cardiac outcome [8–11, 14]. Based on a systematic review of the existing literature, Loeffen and colleagues recommended a DOX infusion duration of at least 1 h in paediatric cancer patients [30]. However, this conclusion does not take into account the administered dose and its impact on cmax. Additionally, some evidence is available pointing to an increased risk of heart failure with a higher maximal anthracycline dose within one week [31]. The avoidance of very short infusion times on the one hand or very high DOX doses on the other hand thus represents a potential measure to reduce the risk of long-term cardiac side effects. This has been unanimously consented by the expert panel but some disagreement arose from the question whether target ranges could be uniformly defined across different tumour types. In contrast to the large variety in DOX administration, there is no data that clearly demonstrate that different tumour entities indeed need specific peak concentrations or drug exposure. Yet, in multi-agent combination chemotherapy regimens adequate DOX therapy intensity will be influenced by the particular combination of chemotherapeutic drugs. Obviously, more research on the dose-concentration-effect relationships in different tumour types is needed to support the establishment of pharmacologically meaningful thresholds and the selection of the most appropriate doses and schedules.
The approach presented herein underlines the value of population PK modelling for treatment optimisation. The DOX population PK model was used to illustrate the complex interplay of dose modifications and PK relationships. Moreover, it provides an opportunity to translate the consented dosing goals into alternative dosing algorithms. It has to be mentioned that the validity of the model-based approach is limited by the small number of patients below the age of one year recruited in the EPOC-MS-001-Doxo trial (N = 4) with the youngest child being 2.5 months old. Thus, uncertainties of model-based predictions are highest in this age group. Similar is true for highly obese paediatric patients. For a routine use of any model-based dosing recommendation two requirements are thus mandatory. Firstly, it is necessary to further validate the population PK model by assessing its predictive performance in a new patient population which should include relevant numbers of infants and young children [32]. Secondly, the consented dosing concept needs to be validated in a prospectively-designed clinical trial assessing its suitability to target a predefined drug exposure.
One may criticize that the Delphi expert panel was rather small to draw meaningful conclusions. However, standards for panel sizes have not yet been established and in the past, Delphi studies have been performed with virtually any panel size. With similar trained experts a small expert panel may be used with sufficient confidence [33]. Despite the small number of participants, agreement among the experts was strong with relatively little variation for most of the questions. The obtained consensus reflects the perspectives of both relevant paediatric study groups and clinical centres. Nonetheless, further discussion with clinical experts on the findings and potential implementations is highly welcome.
As suggested by fig. 3, a relatively small reduction in variability of drug exposure can be expected though individualisation of the DOX dose with respect to age and BSA. Large variability is a long-known characteristic of DOX PK. In adults, substantial inter-patient variations of AUC despite standardisation of the dose based on BSA were observed and differences in dose-normalised peak concentrations of more than 10-fold between children with ALL were reported in a study by Frost et al. [34–36]. Adaptive administration of chemotherapeutics based on plasma concentration measurements could provide an opportunity to further reduce variability in drug exposure. Individual PK parameters can be easily predicted based on a few plasma concentration measurements using a Bayesian forecasting approach [28]. It has been shown before that adaptive dosing of chemotherapeutics can result in a narrower and more accurate exposure range compared with standard BSA-based dosing and can positively impact therapeutic outcome [37, 38]. However, in the past several studies revealed unpredictable differences in individual DOX PK between consecutive administrations [34, 39]. In a study by Hempel et al. in paediatric ALL and non-Hodgkin lymphoma patients intra-individual deviations in peak concentration ranged from 3.5 % to 198 % [39]. In accordance, population PK analysis of data from the EPOC-MS-001-Doxo trial found high intra-individual variability on the central volume of distribution [15]. Due to the high intra-individual variability Hempel et al. concluded that dose individualisation based on monitoring of peak concentrations will not be feasible. In contrast, in a population PK analysis in adults and children older than three years intra-individual variability of DOX clearance accounted only for 13 % [40]. As drug elimination might be less affected from intra-individual variability adaptive dosing approaches aiming to better control variability in drug exposure could indeed be promising. In fact, within the Delphi process the expert panel members acknowledged that therapeutic drug monitoring might be beneficial at least for defined paediatric patient populations.
Nevertheless, pre-analytical variability affects the uncertainties of pharmacokinetic models and model-based predictions. Further, the implementation of drug monitoring and adaptive dosing approaches in clinical routine is hampered by considerable technical effort and logistical requirements. The development of miniaturised monitoring tests and their delivery to the point-of-care is crucial to overcome these limitations [41].