Pharmacometrics and systems pharmacology for metabolic bone diseases

Abstract

Mathematical modelling and simulation (M&S) of drug concentrations, pharmacologic effects and the (patho)physiologic systems within which they interact can be powerful tools for the preclinical, translational and clinical development of drugs. Indeed, the Prescription Drug User Fee Act (PDUFA VI), incorporated as part of the FDA Reauthorization Act of 2017 (FDARA), highlights the goal of advancing model‐informed drug development (MIDD). MIDD can benefit development across many drug classes, including for metabolic bone diseases such as osteoporosis, cancer‐related and numerous rare metabolic bone diseases; conditions characterized by significant morbidity and mortality. A drought looms in terms of the availability of new drugs to better treat these devastating diseases. This review provides an overview of several M&S approaches ranging from simple pharmacokinetic to integrated pharmacometric and systems pharmacology modelling. Examples are included to illustrate the use of these approaches during the development of several drugs for metabolic bone diseases such as bisphosphonates, denosumab, teriparatide and sclerostin inhibitors (romosozumab and blosozumab).

1. INTRODUCTION

An essential part of drug development is the characterization of the compound's preclinical, translational and clinical pharmacokinetics and pharmacodynamics, which helps to identify the optimal dose regimen of new compounds for the next phases of drugs development. Regulatory authorities such as the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) have a set of guidelines that inform drug developers what experiments need to be undertaken to fulfil the regulatory requirements.1, 2, 3 These requirements include single‐ and multiple‐dose studies during which drug levels and effects are measured. Data analysis according to the guidelines was once restricted to model‐independent, non‐compartmental pharmacokinetic (PK) analysis and relationships between dose, exposure and effects were usually investigated using exploratory correlative analysis. The goal was often to see if relationships between dose, systemic exposure and effect were linear while providing a statistical test of the significance of the effect relative to placebo. In addition, efforts were made to see how these relationships held over various species in order to make initial dose recommendations for first in human phase 1 studies. These approaches required development studies and the analysis of their data, however, were often inefficient when determining effective dose regimens. In fact, because dose–systemic exposure and systemic exposure–effect relationships can be complex, the lack of fully integrating the complexity with the experimental observations only further complicates clinical optimization decisions. These complexities can include: saturable absorption, distribution, metabolism and/or excretion processes of the drug, difficulties assessing the drug's concentration at the site where it exerts its effect, lapses between initial drug binding to a receptor, DNA or enzyme and the final effect, and sensitivity changes attributable to disease‐specific changes or differences in pathways, tissue or organ‐level responses. Altogether, these complexities lead to challenging interpretations when identifying relationships between a drug's exposure and its effects. In these cases, model‐based analysis, including PK, exposure–response and systems pharmacology, support knowledge harvesting. When integrated together these analyses can help to unravel the complex relationships.1, 4, 5

The sixth iteration of the Prescription Drug User Fee Act (PDUFA VI), incorporated as part of the FDA Reauthorization Act of 2017 (FDARA), highlights the goal of advancing model‐informed drug development (MIDD). In particular, this Act recognizes the importance of supportive modelling and simulation for the following drug development decisions:

• Dose selection or estimation (eg, for dose/dosing regimen selection or refinement).

• Clinical trial simulation (eg, based on drug‐trial‐disease models to inform the duration of a trial, select appropriate response measures, predict outcomes).

• Predictive or mechanistic safety evaluation (eg, use of systems pharmacology/mechanistic models for predicting safety or identifying critical biomarkers of interest).

As part of the FDA's pursuit to encourage more informed therapeutic development programs, it has recently announced a Model‐Informed Drug Development (MIDD) Pilot Program to: “facilitate the development and application of exposure‐based, biological, and statistical models derived from preclinical and clinical data sources, referred to as MIDD approaches. MIDD approaches use a variety of quantitative methods to help balance the risks and benefits of drug products in development. When successfully applied, MIDD approaches can improve clinical trial efficiency, increase the probability of regulatory success, and optimize drug dosing/therapeutic individualization in the absence of dedicated trials.”6

A recent paper by the FDA sought to justify the use of a modelling and simulation (M&S) approach in the development of drugs for osteoporosis.7 An important point of that review was the FDA's optimistic view on the use of an M&S approach to better appreciate the applicability of various surrogate markers to predict fracture risk. In particular, bone turnover markers (BTMs) were noted by the FDA as a useful measure during the development of new drugs, albeit that the use of these markers was limited by the lack of harmonization and standardization. It was also noted that relationships between these markers, bone mineral density (BMD) and anti‐fracture efficacy need to be re‐investigated for each drug class with a new mechanism of action.

Expanding on that previous review, this current report will first describe several modelling and simulation approaches such as pharmacometric pharmacokinetic and pharmacodynamic (PK‐PD) modelling, systems pharmacology, and integrated pharmacometric and systems pharmacology (iPSP) modelling and subsequently discuss how these methods have been used for model informed development of individual (classes of) drugs for metabolic bone diseases. This summary of real‐world drug development examples will ideally benefit further development towards improving existing drugs and may serve as exemplars for the development of new drugs.

2. PHARMACOKINETICS

The choice of the various approaches for modelling and simulation may depend on a drug's PK and pharmacodynamics (PD) behaviour,5 as well as the nature of the clinical program's goals and outstanding development questions. These choices include compartmental and physiology‐based PK (PBPK) models. The first usually describes only serum or plasma pharmacokinetics and might include one or more compartments to describe multi‐phasic behaviour of the drug concentration–time profile following administration. The second, PBPK, follows a similar ideology, though it conceptualizes the compartments to real body compartments such as fat tissue(s) and/or specific organs of interest (eg, kidney and bone). PBPK models include a complete mass balance over time by using estimated organ volumes, blood perfusions and partition coefficients to describe the drug distribution to and from each of the assigned tissues in the body. Regardless of the model choice, these so‐called structural models are used to describe the general behaviour of each drug with respect to its time‐course and exposure in the body.

The average behaviour is one aspect of a drug's PK. Another concerns intra‐ and inter‐subject variability in the PK of drugs.5 Traditionally, this was analysed in a two‐stage fashion using the structural models, reanalysing each subject and taking the mean and median of each parameter to assess the variability. More recently, variability assessments have been characterized using statistical modelling techniques such as nonlinear mixed effect modelling and Bayesian estimation, which also often associate complex relationships between patient characteristics, such as genetic polymorphisms, body weight or renal function, and structural parameters. These associations provide “covariates” in the model that can serve as predictors of the PK variability therein.

In addition, techniques such as nonlinear mixed effect modelling also allow for information to be shared across the study population, meaning that the PK of a population can be described even if only a few variable data points were collected for each patient rather than a full PK profile.5 Once a population model has been described, Bayesian maximum a posteriori estimations can be used to assess the PK in an individual subject based on the population parameters, and only a limited number of data points. These individualized estimations can be used, for example, during drug monitoring in patient care, and also in clinical development programs when establishing the often complex relationships between the individual exposure and endpoint response(s). These endpoints often consider multiple safety and efficacy biomarkers and outcomes simultaneously to inform risk–benefit evaluations.

3. PHARMACOKINETICS AND PHARMACODYNAMICS (PK‐PD)

Despite our best bio‐analytical and modelling efforts, the exact concentration at the site of action is not always known. In addition, there is often a time lapse between having a drug at the site of action and the measured effect. And lastly, a disease can differ between patients as well as progress, or regress, within a patient, making it more challenging to describe relationships between dose, serum and tissue drug concentrations and the effects.

To some extent these challenges can be addressed by adding hypothetical effect compartments. Alternative approaches, however, include the development of more mechanistic pharmacodynamic (PD) models that, when linked to the PK models, allow for the comparison of simulated effects expected to result from alternative dose regimens. Within the bone field, indirect inhibitory effect models have been used to describe the relationship between drug levels of antiresorptive drugs and bone resorption markers.8 Other models within the bone field are far more complicated and try to encompass the relevant (patho‐)physiology of bone, bone cells and calcium and phosphate homeostasis, thereby resulting in a systems biology or 0systems pharmacology approach.9, 10 Other models have been more simplified while still including disease alterations.11

4. SYSTEMS PHARMACOLOGY

The use of multiscale systems pharmacology models in therapeutics development has been growing remarkably and has been supported by increased grant opportunities from the National Institutes of Health (NIH) and Cooperative Research and Development Agreements (CRADAs) from the FDA. Two National Institute of General Medical Sciences (NIGMS) workshops in 2008 and 2010 further promoted quantitative system pharmacology collaborations between the NIH, academia and industry. The scales involved in multiscale models span from subcellular to organ and whole body‐level interactions, and often include disease pathologies, PK and PK‐PD, and clinical outcomes to generate a full system‐level integration of the associated mechanisms and outcomes of a biologic system to include integration of experimental data through clinical outcomes. These models provide a workbench for exploring perturbations caused by diseases and their progressive states, senescence, gene polymorphisms and variants, and therapeutic intervention(s). In a wider context, these models describe multiple disease manifestations and can therefore bridge information, assumptions and information gaps across therapeutic areas to inform a range of research programs.

An attribute of these systems models is their extensibility, and so they can be reused and repurposed, particularly when integrated with more traditional pharmacokinetic and pharmacodynamics modelling. An example is a model originally developed to relate denosumab exposure with osteoclast proliferation.10 Physiologic representations were included to model bone mineral homeostasis and simultaneously describe the interrelated effects on calcium, phosphate, parathyroid hormone (PTH), vitamin D and bone formation (Figure 1). The model was integrated with additional PK models to describe pharmacologic effects of parathyroid analogue (teriparatide),10 calcium‐sensing receptor modulators,12 exogenous vitamin D,13 and sclerostin inhibition.14

Figure 1.

Schematic of physiologically‐based, multiscale systems pharmacology model. (Reprinted from Peterson & Riggs10)

Building out from a core model to include new complexities, either through signalling pathways or clinical outcomes is termed a “middle‐out” approach. This extensibility has been used to link BMD changes associated with bone marker changes,15 fracture risk associated with BMD,16 secondary disease effects with kidney failure progression,17 and pathophysiology associated with estrogen depletion.18 These examples have fed into drug development evaluations that include proof of mechanism and concept, dose‐ranging, disease state implications and off‐treatment effects. For example, an iPSP model was recently used to evaluate the balance between safety (BMD change) and efficacy (logistic regression of a symptom severity index) for GnRH modulating agents.18 In another case, external validation and use by the FDA of a QSP model during regulatory review of a proposed treatment in patients with hypoparathyroidism to recommend further dose regimen optimizations stands as a landmark use for this kind of systems model.19 Additional examples of multiscale systems model development and applications related to bone remodelling, including evaluations of bisphosphonate (BP) therapies, have been reported by Post et al.11 and Ross et al.20 Case examples supported through systems modelling are included and expanded below.

5. EXAMPLES

5.1. Bisphosphonates

The development of PK and PK‐PD models for BPs has traditionally been challenging because of the absence of sensitive assays to sufficiently quantify the drugs in serum, urine and bone to describe their long‐term PK and the relationship of these drug exposures with their long‐term antiresorptive effects.21 The complexity of models, as described above, is often dictated by the available data, e.g., more extensive PK results for some BPs enable both short‐ and long‐term PK‐PD models to be developed. For other BPs, where such data were not available, simpler (e.g., “K‐PD”) models were developed and validated.8 Central in both types of models has been the estimation of BP kinetics at the skeletal site, which is where the BPs exert their antiresorptive action through osteoclast uptake during bone resorption.22 Both inhibitory hyperbolic maximal effect (E_max) direct and indirect effect models have been used to link the PK of BPs to the bone resorption markers.8, 23 More recently, some efforts have begun to use iPSP models for BPs, which link BP PK, the effect on BTM, and the effects on cell populations, transcription factors, paracrine and endocrine factors such as runt‐related transcription factor 2 (RUNX2), receptor activator of nuclear factor kappa‐β ligand (RANKL), vitamin D metabolites, PTH and fibroblast growth factor 23 (FGF‐23), BMD (by DXA and HRpQCT) and how this all translates into osteoporotic fracture risk reduction.9, 10

Target levels for bone resorption and formation markers of effective BP therapy have been estimated as continuous suppression of at least 50% (for serum C‐terminal telopeptide, sCTX) and 40% (for N‐terminal propeptide of type I procollagen, PINP).24, 25 These targets have been used to develop alternate dosing regimens with intervals ranging from weekly, monthly and yearly.26, 27, 28, 29, 30 PK‐PD models were used to explore these various regimens and the FDA approved the alternative dose regimens based on phase 2 studies with BTMs and BMD measurements as surrogate endpoints. An example of the transition from daily to monthly oral ibandronate is given in Figure 2.28

Figure 2.

Simulations of serum CTX following daily (2.5 mg) and monthly (150 mg) oral administration. (Reprinted from Zaidi et al.28)

M&S has also been used to establish dosing aimed at minimizing the risk of side effects such as nephrotoxicity from bisphosphonates. Such work, conducted by the FDA, was essential in leading to the redesign of the dose regimen and infusion times for zoledronic acid in patients with renal impairment. This was especially important for patients with cancer and multiple myeloma, many of whom have impaired renal disease and were receiving monthly rather than a yearly dose regimen of this potent BP.31 M&S has also been used in the development of less dense dose regimens for patients with metastatic bone disease with continued suppression of bone resorption as target.32, 33

5.2. Cathepsin K inhibitors

M&S efforts for the cathepsin K inhibitors have helped the development of these compounds tremendously and have revealed significant differences between these antiresorptive drugs and the other major class of antiresorptive drugs, BPs. Significant differences exist in both the PK and PD of these drugs and BP; these differences have been quantified by M&S. Unlike BPs, cathepsin K inhibitors are metabolized.34 In addition, some cathepsin K inhibitors such as odanacatib are excreted via urine as well as bile and seem to undergo enterohepatic recirculation, which is reflected by a double peak after oral administration.34 However, observed double peaks after administration of odanacatib may also be related to circadian variability in endogenous levels of plasma proteins or lipoproteins to which the drug binds.35 Additionally, unlike BPs, cathepsin K inhibitors do not specifically target mineralized tissue. The PK of cathepsin K inhibitors have been described by various models, in most cases relatively simple one‐, two‐ or three‐compartment models,35, 36 that describe the PK of these compounds sufficiently well in most patients.

In contrast to BPs, the cathepsin K inhibitors do not need to be attached to bone in order to work. Instead of being taken up into the osteoclasts during osteoclast‐mediated bone resorption, it seems that it is the cathepsin K inhibitor concentration in the extracellular fluid surrounding area of the osteoclasts, and thereby the resorption pit, that determines the extent of inhibition of the enzyme that is secreted by the osteoclasts that degrades collagen type 1.37 And from various PK‐PD analyses using plasma concentration and decrease of bone resorption markers, it seems as if the concentrations of the cathepsin K inhibitors in the extracellular fluid at the bone correlate well with the plasma concentrations of the cathepsin K inhibitors. Consequently PK‐PD models for cathepsin K inhibitors can be relatively simple. These have consisted of a compartmental PK model to describe plasma PK and an inhibitory E_max model that linked bone resorption marker levels to the plasma concentration of the drug.37, 38 Interestingly, a recent paper by Ma et al.39 explored the relationship between the serum concentrations of three different cathepsin K inhibitors (odanacatib (MK‐0822), MK‐0674 and MK‐1256) and their effects on bone resorption markers in various animals and humans. They determined that a much better relationship existed between the unbound drug concentration and the antiresorptive action than between the total drug concentration and the antiresorptive action, which significantly enhanced cross‐species translation of PK‐PD of these compounds.

Probably the most challenging part of the PK‐PD M&S of cathepsin K inhibitors is the estimation of the wearing off of the antiresorptive effect of these drugs after treatment discontinuation. Similar to oestrogens and denosumab, but probably based on a different mechanism, bone resorption and formation increased rapidly immediately after cathepsin K inhibitor treatment termination. These levels rose to levels much higher than before treatment. This was accompanied by rapid bone loss that was maintained for a few months after which bone turnover returned to pretreatment levels. Stoch et al.37 were able to develop a PK‐PD model for odanacatib that adequately described this behaviour, both with respect to the BTMs and the effects on BMD. The model used one‐compartment disposition to describe the PK, an inhibitory indirect effect model to describe the effects on the urinary bone resorption marker, creatinine‐corrected urine type I collagen‐cross‐linked N telopeptide (NTX/Cr), as well as cellular models that describe osteoblasts and osteoclasts, and ultimately BMD. Unfortunately, the development of odanacatib was halted because of potential cardiovascular side effects. Otherwise these models would have been most useful in the further clinical development of this therapeutic candidate.

5.3. Denosumab

The interaction of receptor activator of NF‐κB (RANK) and its ligand (RANKL) is critical for the differentiation and survival of osteoclasts. Inhibition of RANKL, as through the administration of the monoclonal antibody denosumab, expectedly leads to a decrease in both osteoclastogenesis and osteoclast survival, and thereby a marked, and relatively rapid decrease in osteoclast function, as measured through bone markers such as sCTX. In addition to these changes in bone resorption markers, early clinical evaluations following administration of denosumab revealed a more delayed decline in osteoblast function (eg, as measured by bone‐specific alkaline phosphate [BSAP]), as well as transiently decreased serum calcium and increased PTH. To understand and quantify the interrelated mechanism behind these observed clinical changes, the previously mentioned multiscale systems pharmacology was developed (Figure 1).10 The modelling provided a mathematical trace of the physiologic response elicited by the osteoclast decline through RANKL inhibition. Namely, a decrease in osteoclast‐driven bone resorption leads to a decrease in transforming growth factor‐β (TGF‐β) activation and a subsequent decrease in osteoblast function. Mediation of this effect through TGF‐β is slower than the more rapid decline in osteoclasts caused by RANKL inhibition and so, although osteoblast function declines, it does so more slowly and to a lesser magnitude compared with the osteoclast effect.

Additionally, the early net increase in bone resorption changes resulted in a decrease in calcium exchange from bone to plasma leading to a transient decrease in serum calcium. The regulatory mechanism of calcium‐sensing in the PT gland thereby triggers an increased production of PTH, increasing plasma PTH and plasma calcitriol. These latter two responses serve to correct the effect of decreased bone calcium accretion to plasma calcium. The systems model, which included these feedback mechanisms, appropriately described the direction and magnitude of responses for each of these markers resulting from RANKL inhibition compared to available clinical observations over the time‐course.10 This included predicted changes following pre‐arranged cohorts with discontinuation and with dosage regimen switching (Figure 3); the systems model was therefore able to provide model‐informed support during the interpretation of dose‐ranging and Phase 3 dose selection for denosumab.

Figure 3.

Predictions from a systems pharmacology model (lines) and observed clinical measurements of lumbar spine bone mineral density (LSBMD) relative to time from start of denosumab therapy for several varied dosing regimens. (Reprinted from Peterson & Riggs15)

A later extension of the systems model was developed, enabling prediction of nonlinear changes in lumbar spine bone mineral density (LSBMD). Data for denosumab, dosed at several levels and regimens, was used for fitting the LSBMD component. BTM and LSBMD data extracted from the literature described on/off‐treatment effects of denosumab over 48 months.40 An indirect model linking bone markers to LSBMD was embedded in the existing model, reasonably predicting nonlinear increases in LSBMD during treatment (24 months); LSBMD declines following discontinuation and increases upon treatment reinstitution (Figure 3). This study demonstrated the utility of systems model extension to describe phenomena of interest not originally in a model, and the ability of this updated model to predict nonlinear longitudinal changes in the clinically relevant endpoint, LSBMD, with denosumab treatment.

5.4. Sclerostin inhibitors

Sclerostin has been discovered to be a key regulator of bone remodelling through its bone‐specific interaction within the Wnt signalling pathway. Sclerostin is secreted by osteocytes, which mature from osteoblasts and are embedded in bone, to downregulate bone formation through inhibition of Wnt signalling. Inhibition of sclerostin thereby prevents the sclerostin inhibition of Wnt. Administration of monoclonal antibodies designed as sclerostin inhibitors has led to increased markers of bone formation and decreased markers of resorption, thereby providing a net gain in bone calcification and an increase in BMD.41, 42 The decoupling of bone formation and resorption associated with this mechanism is a differentiation from other osteoporosis treatment mechanisms. Those other mechanisms are either entirely anabolic (formation and resorption increase concurrently, eg, intermittent PTH) or catabolic (formation and resorption decrease concurrently, eg, BPs, RANKL‐inhibition). Furthermore, because of the localized effects in bone, the Wnt‐mediated effects through sclerostin inhibition promise to limit off‐target effects in other tissues and organs.

Questions remain, however, in regard to how the linkage through sclerostin‐mediated osteocyte activity may affect, and be affected by, feedback regulation in bone remodelling.43 The impact of these feedback mechanisms on the ability to maintain efficacy through repeated anti‐sclerostin monoclonal antibody (mAb) administration remains an ongoing clinical question. Model‐based support may be suitable to identify favourable dosing regimens for sclerostin inhibitors and/or their combination with other agents (e.g., RANKL inhibition or BPs) to improve longer‐term treatment and maintenance of bone quality for further clinical investigations.

The same multiscale systems model10 was again used as basis for expansion. Osteocyte and sclerostin‐related components necessary to predict effects of sclerostin inhibition on clinical outcomes, eg, through bone markers and BMD measurements, were added. The model was expanded by adding components that described and incorporated emerging knowledge of the Wnt/β‐catenin signalling pathway. Available data reported from recent clinical studies41, 44, 45 were critical in enabling that extension. The updated model promoted understanding of how osteocyte signalling contributes to regulation of bone remodeling.14 The resulting expanded model (Figure 4) was used to simulate PINP, sCTX and BMD profiles that were qualified with clinical study reports with sclerostin mAb treatment (Figure 5). As described elsewhere in this Themed Issue (Tabacco et al.), these include a continuous remarkable increase of lumbar spine and total hip BMD and transient increases and decreases of PINP and sCTX, respectively (Figure 5). The resulting model provided a tool to explore Wnt pathway modification while incorporating the mechanisms of osteocyte‐derived feedbacks during bone remodelling. Model simulations pointed to differential effects from osteocyte‐driven changes in resorption activity that suggested the considerations for shorter dosing intervals for sclerostin mAbs should be considered for investigation in future clinical trials. In addition, the model‐based simulations suggested that combination therapies may be beneficial in osteoporosis patients to mitigate feedback effects.

Figure 4.

Schematic of the bone‐remodelling systems model. Intersection points of sclerostin signalling effects within the model are identified with numbers corresponding to description in the text. New model compartments are indicated with white text and shading and corresponding equation numbers. (Reprinted from Eudy et al.14)

Figure 5.

Simulated P1NP (A), CTx (B), lumbar spine BMD (C), and total hip BMD (D), (blue line) overlaying data from a clinical trial with blosozumab (red points). This qualification dataset was not used in constructing the model (n529, 31, 30, 30 for arms PBO, 180 mg Q2W, 180 mg Q4W, and 270 mg Q2W, respectively). (Reprinted from Eudy et al.14)

5.5. PTH and analogues

PTH plays a central role in the regulation of calcium homeostasis and bone health (Figure 1). A seemingly paradoxical effect occurs when PTH is transiently or chronically altered. For example, transient increases in PTH through once‐daily subcutaneous administration of PTH1‐84 or PTH1‐36 (teriparatide) provides an anabolic response in bone whereas chronic elevations, eg, through primary or secondary hyperparathyroidism, will lead to detrimental bone loss through net increase in bone catabolism. Similarly, chronic lowering of PTH, eg, through hypoparathyroidism, will also lead to net bone loss. Each of these effects, which were all recently reviewed in this journal,46 has been captured through mechanisms described in the reported systems pharmacology model.10 Notably, corresponding predictions of related clinical changes were also captured by the model, including the typical range of hyper‐ and hypo‐calcaemia associated with hyper‐ and hypo‐parathyroidism, respectively.

Therefore, in addition to the effects measured through bone markers and BMD, other organ functions can also be evaluated through systems models. For example, clinical investigations of once‐daily PTH administration for the treatment of hyperparathyroidism have revealed that, despite an apparent overall clinical benefit to this replacement therapy (eg, the transient elevations in PTH do provide beneficial increases in serum calcium), there is also an elevated risk of developing hypercalciuria when administering this treatment to patients with hypoparathyroidism. In an effort to better understand this effect, the FDA used an open‐source version of the Peterson and Riggs model to evaluate and propose alternative dosage considerations for PTH1‐84 treatment in this disease population.19 In remarking about this landmark use of the systems modelling during regulatory review, the lead medical reviewer stated during the advisory meeting: “We wanted to use the model to explain certain things that were seen in the trial. So it's interesting. It's thought provoking,” and he added that the use of the model was “hypothesis‐generating.”47 After further development the model might also be able to capture the transient state of the PTH receptor, which is relevant for newer PTH or PTHrP analogues such as abaloparatide.46

5.6. Estrogen

Estrogen loss, as during the menopause transition or through pharmacologic interventions (eg, via gonadotropin releasing hormone (GnRH) modulation), has long been associated with BMD loss. Systems modelling was again used to associate the mechanisms for this loss, as well as the magnitude and time‐course for these changes with varying magnitudes of estrogen.18 In the former case of estrogen loss through menopause transition, the modelling can be useful for determining expected changes, eg, in placebo‐treated patients during a clinical trial based on the postmenopausal age of the trial population. In the latter case of estrogen loss through GNRH modulation, the modelling was used to set a target window for oestradiol that would lead to minimal and acceptably low BMD loss while also providing symptom relief in patients with endometriosis. In a latter publication it was stated that “this work identified target levels for estrogen that would provide symptomatic pain relief with minimal impact on BMD. … targeting the GnRH pathway to achieve the desired range of serum estrogen levels would be difficult to achieve; therefore, the research program was halted before any compound entered the clinic.”48

5.7. Calcilytics

Recent research involving calcium‐sensing receptor antagonists (calcilytics) had focused on assimilating the transient spikes in PTH observed with subcutaneous PTH (1‐34, 1‐84) administration. The goal had been to provide a “by mouth” osteoporosis treatment comparable to PTH without invasive (eg, subcutaneous) dosing. Clinical investigations of orally administered calcilytics have yet to achieve this target profile, however, with typical BMD elevations of no more than 2–3% along with often notably elevated serum calcium. A model‐based approach to quantify the physiologic response to calcilytics was undertaken to inform development decisions for DS‐9194b, an orally administered investigational calcilytic. The existing physiologically‐based, multiscale systems pharmacology mathematical model10 was expanded to include a capacity‐limited PTH release pool. PK and PTH data (ronacaleret, JTT‐305) were used for this further model development. Results indicated a limit to the maximum achievable peak PTH response and described the characteristic persistent PTH elevation.

The modelling results were coupled with a modelled relationship between peak PTH and BMD; results suggested that average typical PTH maximum (PTH_max) from the investigated calcilytics (∼20 pM) equated to a mean BMD increase of <3%, consistent with the reported clinical investigations of these agents, whereas a PTH_max > 30 pM was considered necessary to provide an appreciable BMD increase.

PK, PTH and calcium data were prospectively collected from a single‐dose, first‐in‐human study including DS‐9194b administration (0–100 mg). PTH_max reached an apparent plateau (∼30 pM) as doses increased; this peak was well described by the model, as were the prolonged elevations at higher doses. Urine calcium excretion decreased with increased dose; this effect was inherently captured in the systems model through PTH effect on urine calcium excretion.

Overall, the model‐informed results supported that, although BMD elevation with the calcilytic administration routines evaluated was possible, the magnitude of elevation was unlikely to match that observed with existing exogenously administered PTH treatments. The systems model provided a physiologic explanation of maximal PTH response due to the capacity‐limited PT gland pool of PTH. Results can guide future considerations for calcilytic‐related therapies for osteoporosis or other PTH‐related disorders.12

5.8. Future

At present a drought looms in the development of new drugs for osteoporosis.49 The reasons are multiple, including a significantly increased risk for companies in terms of return on investment when developing drugs for osteoporosis. As a consequence, and helped by various initiatives from regulatory agencies,50 most drugs currently in development for metabolic bone diseases are for rare diseases. As described elsewhere in this Themed Issue, developing drugs for rare diseases, especially rare metabolic bone diseases, can be challenging and includes, eg, finding a sufficient number of patients to investigate the efficacy of a new drug (Hsiao et al.). As a consequence of these diseases often being inherited or associated with in utero mutations, these diseases sometimes include a focus on paediatric patients, as well. The use of translational PBPK and PK‐PD models, while applying an all‐encompassing pharmacometric strategy to the development of new drugs might benefit patients with these often devastating diseases through optimal use of all available data, including modelling of maturational effects in younger patients. The vast translational and clinical pharmacometric experience gathered during the development of drugs for more common metabolic bone diseases such as osteoporosis can thus be of immense help to the development of drugs for rare diseases. In turn, some drugs originally developed for rare diseases may be repurposed for more common applications (or vice versa). It therefore can be expected that a comprehensive M&S program for the development of drugs for rare metabolic bone diseases will also benefit the development of drugs for more common metabolic bone disease such as osteoporosis.

Integral to advancing the potential of these possible new drugs is a comprehensive understanding of the underlying system of cellular, tissue and organ‐level responses that they are purposefully, or unintentionally, affecting. Systems biology and pharmacology modelling, through the further incorporation of ’omics level information, offers a platform for quantifying disease‐level pathologies that result in disease‐associated manifestations; this in turn provides an opportunity to identify, understand and tailor treatment options for individual patients, ie based on a personalized medicine approach.51 A key to the continued development and validation of these models is their extensibility through persisting research efforts that expand with new data and emerging conceptions of metabolic bone diseases and mechanisms of drugs. These added inputs (eg, mechanical influences, regulatory/signalling functions and dysfunctions) and outputs (eg, bone quality measurements, site and bone type‐specific impacts, 3D imaging) widen our capabilities for targeting individualized medicine.

In all, early uptake of model‐informed decision support through comprehensive PK, PK‐PD and iPSP modelling and simulation is an encouraging method to expedite successful therapeutics development. These approaches combine multidisciplinary strengths, facilitate evaluations of subject‐level and population‐level responses for efficacy and safety assessments, lend insight into molecular and target‐level mechanisms, and allow predictive simulations of novel therapeutic interventions, including combination and switching regimens. These added efficiencies, added to already rigorous research and development efforts, promise to ensure that the correct medications for the correct patients at correct dosages are available as prescriptions for the next generation of metabolic bone disease therapeutics.

COMPETING INTERESTS

There are no competing interests to declare.

Riggs MM, Cremers S. Pharmacometrics and systems pharmacology for metabolic bone diseases. Br J Clin Pharmacol. 2019;85:1136–1146. 10.1111/bcp.13881