In their interesting paper published in the British Journal of Clinical
Pharmacology, Mackey et al. present an analysis aiming at characterizing
resonance (i.e., neutrophils oscillations) in young patients treated by
cyclic chemotherapy [1]. The authors used further Quantitative
Systems Pharmacology (QSP) modeling applied to granulopoïesis to
demonstrate that timing of chemotherapy could impact on the dynamics of
neutrophils. Consequently, the authors suggest that model-informed
scheduling could help limiting hematological toxicity, e.g. by delaying
supportive G-CSF therapy. The issue of controlling drug-induced side
effects, especially hematological toxicities with cytotoxics, is
critical in many respects. Pancytopenia can be rapidly life-threatening,
especially in frail patients. When they do not directly lead to
toxic-death, such severe toxicities frequently oblige practitioners to
postpone or discontinue chemotherapy or associated radiation therapy,
thus affecting clinical outcome and survival eventually. In addition,
severe neutropenia with sepsis require antibiotics which history of use
is suspected now to compromise the efficacy of immune checkpoint
inhibitors, probably because of disruption of gut microbiota [2]. In
addition, it is now acknowledged that prolonged lymphopenia can have
deleterious effect as well, especially in the era of immunotherapy. For
instance, high neutrophils-to-lymphocytes basal ratios have been
repeatedly associated with poor response to immunotherapy because it
could promote immune desert at the tumor level [3]. Altogether,
developing strategies to control or reduce the risk of drug-induced
hematological toxicities is therefore a major concern in clinical
oncology, especially because as stated by the authors, cytotoxics are
still today the backbone of most treatments of solid tumors and
hematological malignancies. Developing in silico approaches as
decision-making tools to optimize anticancer therapies has been a rising
trend for decades now in clinical oncology [4].
Pharmacokinetically-guided regimen with Bayesian adaptive dosing
procedures have been already proposed for several years to tailor the
administration of a variety of cytotoxics and oral targeted therapies
[5]. However, implementing adaptive dosing strategies in routine
clinical setting remains challenging. Real-life precision medicine
requires indeed mathematical models that are kept simple enough to allow
proper identification of their parameters. This is a prerequisite for
being easily applied prospectively in actual patients next, and not to
be used solely as part of retrospective in silico modeling. This
calls for using primarily top-down approaches, such as compartmental
analysis prior to developing pharmacokinetics/pharmacodynamics (PK/PD)
models likely to help oncologists to determine the optimal dosing and
scheduling of a given drug to a given patient. More intricate modeling
and QSP approaches are appealing strategies which are unfortunately
impaired by their intrinsic complexity, which has made them unfit for
routine use at bedside so far (Figure 1). Conversely, phenomenological
modeling could in many respects look like an over-simplistic, suboptimal
strategy, often mocked as being black-boxes simply linking an output to
a given input. In turn, such models are more likely to be actually used
in real-life setting, not despite the fact that they are black-boxes
simply linking an output to a given input, but precisely because they
are black-boxes simply linking an output to a given input. For instance,
the Friberg model, as cited by the authors, is a simplified
representation of hematopoiesis, using a semi-mechanistic, compartmental
description of the proliferation and dynamics of the maturation of blood
progenitors. Because of its simplicity, this top-down approach has
allowed the Friberg model to be extensively used over the last 15 years,
both by academics and pharmaceutical companies, to describe the
myelosuppressive effects of a variety of cytotoxics [6]. Countless
phenomenological models have been derived from the Friberg model ever
since. For instance, the Meille model is based upon a similar simplified
and semi-mechanistic representation of hematopoiesis and granulopoiesis,
which encapsulates additionally a PK/PD model describing effects of
supportive G-CSF administration on blood cells progenitors [7]. When
further combined with another phenomenological model for
antiproliferative efficacy, it was used next to build an original
constraint-model determining the optimal dosing and scheduling of
densified chemotherapy combo plus G-CSF support. Once calibrated with
pre-defined acceptable levels of hematological toxicity and desired
level of tumor shrinkage, this mathematical-driven regimen was finally
tested prospectively in metastatic breast cancer patients and showed
excellent performances such as prolonged overall survival in heavily
pretreated patients with fully controlled hematological toxicities
[8]. Importantly, the prospective use of such a mathematical model
was made possible thanks to a first identification step with few blood
samples taken from patients when treatment starts, providing individual
data on drugs PK profile and blood counts . This critical step allowed
fine tuning of individual PK/PD model parameters in real-time, thus
ensuring optimal, personalized dosing next. Transposing such
model-driven regimen at bedside seems to be only achievable when the
whole modeling strategy is primarily built upon the parsimony principle,
so as to be able to identify next individual parameters from sparse,
routine data collected from patients in a real-life setting. Of note, no
in-depth understanding of biological mechanisms can be provided by such
models – dosing and scheduling of anticancer agents and G-CSF are
connected to efficacy and toxicity endpoints through phenomenological
black-boxes, not multi-scales models providing biological explanations
of the phenomenon. In contrast, QSP models, such as the ones developed
by the Craig group in this BJCP issue, offer appealing mechanistic
features, thus allowing a better understanding of all the underlying
mechanisms at play to explain pharmacodynamic endpoints. The downside is
that the Mackey model is complex. It is based indeed upon more than 30
parameters, a large number of them having been fixed from the literature
and thus are dependent on the variability and possible biases of the
experiments used for their identification [9]. Furthermore, in
contrast to standard approaches in pharmacometrics using nonlinear
mixed-effects (NLME) modeling, inter-individual variability of the
parameters is not quantified. The issue with such a large number of
parameters is that the practical identifiability from sparse individual
data collected at bedside is expected to be poor, resulting in a large
uncertainty for quantitative model predictions in real-life practice.
Nevertheless, the qualitative observation of the resonance effect in
neutrophils time dynamics induced by the administration of periodic
chemotherapies, highlighted by Mackey et al., should prompt modelers to
include such pivotal phenomenon, including in their phenomenological
representations of hematological toxicity. This new concept could be
easily integrated and quantitatively tested in NLME models by means of
inter-occasion variability (IOV) between administrations. Such IOV could
allow to describe the resonance effect and improve model predictions. A
covariate analysis could be performed to identify potential factors
explaining the origin of this observation. Furthermore, as stated by the
authors, to avoid this deleterious phenomenon and to reduce
cytotoxic-induced hematological toxicities with subsequent negative
impact on tumor immunity, it is an absolute prerequisite to optimize
empirical chemotherapy regimens and G-CSF support administrations, based
on robust PK/PD compartmental modeling.
Altogether, the Mackay study highlights how the very way we use
anticancer agents does matter, and how a same drug can have
diametrically different pharmacodynamics effects, depending on its
scheduling. In this respect, this work is an important contribution to
the field of Precision Medicine in Oncology, by suggesting that there is
much room left to improve standard use of canonical cytotoxics and thatin silico strategies could help to achieve a better way to
administrate anticancer drugs. To transform this theoretical concept
into a practical decision-making tool for oncologists, mathematicians
and modelers must compose with the issue of parameters identifiability,
and what kind of accessible individual data are actually made available
in routine patients. As long as these issues are nor fixed, modeling
will remain an elegant but theoretical field disconnected from bedside
practice. To the ever-rising complexity of cancer biology and the
amazing amount of knowledge regarding in-depth mechanisms of action of
drugs (such as molecular signaling pathways, genetic and epigenetic
regulations affecting targets or key-proteins involved in
pharmacodynamics responses), the temptation to implement all this
knowledge into super-models should be resisted. Indeed for a proper and
rapid in silico -to-bedside transposition, we believe that in the
era of Precision Medicine, the more complex is a phenomenon, the simpler
should be the mathematical model describing it.
Competing Interest: the authors declare none
Funding: none