Refine
Year of publication
Document Type
- Article (29)
- Other (5)
- Conference Proceeding (3)
- Postprint (3)
Language
- English (40)
Is part of the Bibliography
- yes (40)
Keywords
- PBPK (3)
- Aluminium (2)
- EGFR (2)
- Induced neutropenia (2)
- Intensive care (2)
- Model (2)
- Myelosuppression (2)
- Prediction (2)
- Renal function (2)
- inflammatory bowel disease (2)
- infliximab (2)
- mAb disposition (2)
- Absorption kinetics (1)
- Al-26 (1)
- Aluminium adjuvants (1)
- Approximate approximations (1)
- Blood coagulation network (1)
- Cauchy problem (1)
- Cell-level kinetics (1)
- Classical compartment model (1)
- Clearance induction (1)
- Continuous renal replacement therapy (1)
- Control theory (1)
- C‐ reactive protein remission (1)
- D. discoideum (1)
- Dosing algorithm (1)
- Error control/adaptivity (1)
- Extravasation rate-limited tissue model (1)
- FcRn salvage mechanism (1)
- Gestagenic drug (1)
- Global sensitivity analysis (1)
- Immunodeficient mice models (1)
- In vitro dissolution (1)
- Model order reduction (1)
- Nonlinear systems (1)
- ODE with random initial conditions (1)
- Ordinary Differential Equation model (1)
- PBTK (1)
- Pharmacokinetic models (1)
- Pharmacokinetics/Pharmacodynamics (1)
- Pharmacokinetics/pharmacodynamics (1)
- Pregnancy (1)
- Protein binding (1)
- Risk assessment tool (1)
- Rothe method (1)
- SHBG (1)
- Target attainment (1)
- Therapeutic proteins (1)
- Toxicokinetic modelling (1)
- Toxicokinetics (1)
- Unscented Kalman Filter (1)
- Unspecific antibody clearance (1)
- WKB ansatz (1)
- actin dynamics (1)
- acute severe (1)
- adrenal insufficiency (1)
- amoeboid motility (1)
- association (1)
- asymptotic approximation (1)
- beta-Lactam (1)
- beta-Lactams (1)
- binding (1)
- body mass index procedure (1)
- body surface area (1)
- cell migration (1)
- cell-level kinetics (1)
- certolizumab pegol (1)
- chemical master equation (1)
- cortisol (1)
- crohn's disease (1)
- disease activity (1)
- dried blood spots (1)
- drug monitoring (1)
- fat-free mass (1)
- global sensitivity analysis (1)
- hybrid model (1)
- infliximab dosing (1)
- joint normal distribution (1)
- keratocytle-like motility (1)
- linical databases (1)
- migration (1)
- modes of (1)
- monoclonal antibodies (1)
- multiscale analysis (1)
- oral anticancer drugs (1)
- ordinary differential equation (1)
- partial averaging (1)
- pediatrics (1)
- personalised medicine (1)
- pharmacokinetic (1)
- pharmacokinetic models (1)
- pharmacokinetics (1)
- pharmacometrics (1)
- red blood cells (1)
- regimen (1)
- targeted antineoplastic drugs (1)
- therapeutic (1)
- therapeutic proteins (1)
- tyrosine kinase inhibitors (1)
- ulcerative colitis (1)
Quantifying uncertainty, variability and likelihood for ordinary differential equation models
(2010)
Background
In many applications, ordinary differential equation (ODE) models are subject to uncertainty or variability in initial conditions and parameters. Both, uncertainty and variability can be quantified in terms of a probability density function on the state and parameter space.
Results
The partial differential equation that describes the evolution of this probability density function has a form that is particularly amenable to application of the well-known method of characteristics. The value of the density at some point in time is directly accessible by the solution of the original ODE extended by a single extra dimension (for the value of the density). This leads to simple methods for studying uncertainty, variability and likelihood, with significant advantages over more traditional Monte Carlo and related approaches especially when studying regions with low probability.
Conclusions
While such approaches based on the method of characteristics are common practice in other disciplines, their advantages for the study of biological systems have so far remained unrecognized. Several examples illustrate performance and accuracy of the approach and its limitations.
Quantifying uncertainty, variability and likelihood for ordinary differential equation models
(2010)
Background: In many applications, ordinary differential equation (ODE) models are subject to uncertainty or variability in initial conditions and parameters. Both, uncertainty and variability can be quantified in terms of a probability density function on the state and parameter space. Results: The partial differential equation that describes the evolution of this probability density function has a form that is particularly amenable to application of the well- known method of characteristics. The value of the density at some point in time is directly accessible by the solution of the original ODE extended by a single extra dimension (for the value of the density). This leads to simple methods for studying uncertainty, variability and likelihood, with significant advantages over more traditional Monte Carlo and related approaches especially when studying regions with low probability. Conclusions: While such approaches based on the method of characteristics are common practice in other disciplines, their advantages for the study of biological systems have so far remained unrecognized. Several examples illustrate performance and accuracy of the approach and its limitations.
We propose a novel strategy for global sensitivity analysis of ordinary differential equations. It is based on an error-controlled solution of the partial differential equation (PDE) that describes the evolution of the probability density function associated with the input uncertainty/variability. The density yields a more accurate estimate of the output uncertainty/variability, where not only some observables (such as mean and variance) but also structural properties (e.g., skewness, heavy tails, bi-modality) can be resolved up to a selected accuracy. For the adaptive solution of the PDE Cauchy problem we use the Rothe method with multiplicative error correction, which was originally developed for the solution of parabolic PDEs. We show that, unlike in parabolic problems, conservation properties necessitate a coupling of temporal and spatial accuracy to avoid accumulation of spatial approximation errors over time. We provide convergence conditions for the numerical scheme and suggest an implementation using approximate approximations for spatial discretization to efficiently resolve the coupling of temporal and spatial accuracy. The performance of the method is studied by means of low-dimensional case studies. The favorable properties of the spatial discretization technique suggest that this may be the starting point for an error-controlled sensitivity analysis in higher dimensions.
The human immunodeficiency virus (HIV) can be suppressed by highly active anti-retroviral therapy (HAART) in the majority of infected patients. Nevertheless, treatment interruptions inevitably result in viral rebounds from persistent, latently infected cells, necessitating lifelong treatment. Virological failure due to resistance development is a frequent event and the major threat to treatment success. Currently, it is recommended to change treatment after the confirmation of virological failure. However, at the moment virological failure is detected, drug resistant mutants already replicate in great numbers. They infect numerous cells, many of which will turn into latently infected cells. This pool of cells represents an archive of resistance, which has the potential of limiting future treatment options. The objective of this study was to design a treatment strategy for treatment-naive patients that decreases the likelihood of early treatment failure and preserves future treatment options. We propose to apply a single, pro-active treatment switch, following a period of treatment with an induction regimen. The main goal of the induction regimen is to decrease the abundance of randomly generated mutants that confer resistance to the maintenance regimen, thereby increasing subsequent treatment success. Treatment is switched before the overgrowth and archiving of mutant strains that carry resistance against the induction regimen and would limit its future re-use. In silico modelling shows that an optimal trade-off is achieved by switching treatment at & 80 days after the initiation of antiviral therapy. Evaluation of the proposed treatment strategy demonstrated significant improvements in terms of resistance archiving and virological response, as compared to conventional HAART. While continuous pro-active treatment alternation improved the clinical outcome in a randomized trial, our results indicate that a similar improvement might also be reached after a single pro-active treatment switch. The clinical validity of this finding, however, remains to be shown by a corresponding trial.
During preclinical development of a gestagenic drug, a significant increase of the total plasma concentration was observed after multiple dosing in pregnant rabbits, but not in (non-pregnant) rats or monkeys. We used a PBPK modeling approach in combination with in vitro and in vivo data to address the question to what extent the pharmacologically active free drug concentration is affected by pregnancy induced processes. In human, a significant increase in sex hormone binding globulin (SHBG), and an induction of hepatic CYP3A4 as well as plasma esterases is observed during pregnancy. We find that the observed increase in total plasma trough levels in rabbits can be explained as a combined result of (i) drug accumulation due to multiple dosing, (ii) increase of the binding protein SHBG, and (iii) clearance induction. For human, we predict that free drug concentrations in plasma would not increase during pregnancy above the steady state trough level for non-pregnant women.
Cell-level kinetic models for therapeutically relevant processes increasingly benefit the early stages of drug development. Later stages of the drug development processes, however, rely on pharmacokinetic compartment models while cell-level dynamics are typically neglected. We here present a systematic approach to integrate cell-level kinetic models and pharmacokinetic compartment models. Incorporating target dynamics into pharmacokinetic models is especially useful for the development of therapeutic antibodies because their effect and pharmacokinetics are inherently interdependent. The approach is illustrated by analysing the F(ab)-mediated inhibitory effect of therapeutic antibodies targeting the epidermal growth factor receptor. We build a multi-level model for anti-EGFR antibodies by combining a systems biology model with in vitro determined parameters and a pharmacokinetic model based on in vivo pharmacokinetic data. Using this model, we investigated in silico the impact of biochemical properties of anti-EGFR antibodies on their F(ab)-mediated inhibitory effect. The multi-level model suggests that the F(ab)-mediated inhibitory effect saturates with increasing drug-receptor affinity, thereby limiting the impact of increasing antibody affinity on improving the effect. This indicates that observed differences in the therapeutic effects of high affinity antibodies in the market and in clinical development may result mainly from Fc-mediated indirect mechanisms such as antibody-dependent cell cytotoxicity.
Cell-level kinetic models for therapeutically relevant processes increasingly benefit the early stages of drug development. Later stages of the drug development processes, however, rely on pharmacokinetic compartment models while cell-level dynamics are typically neglected. We here present a systematic approach to integrate cell-level kinetic models and pharmacokinetic compartment models. Incorporating target dynamics into pharmacokinetic models is especially useful for the development of therapeutic antibodies because their effect and pharmacokinetics are inherently interdependent. The approach is illustrated by analysing the F(ab)-mediated inhibitory effect of therapeutic antibodies targeting the epidermal growth factor receptor. We build a multi-level model for anti-EGFR antibodies by combining a systems biology model with in vitro determined parameters and a pharmacokinetic model based on in vivo pharmacokinetic data. Using this model, we investigated in silico the impact of biochemical properties of anti-EGFR antibodies on their F(ab)-mediated inhibitory effect. The multi-level model suggests that the F(ab)-mediated inhibitory effect saturates with increasing drug-receptor affinity, thereby limiting the impact of increasing antibody affinity on improving the effect. This indicates that observed differences in the therapeutic effects of high affinity antibodies in the market and in clinical development may result mainly from Fc-mediated indirect mechanisms such as antibody-dependent cell cytotoxicity.
The chemical master equation (CME) is the fundamental evolution equation of the stochastic description of biochemical reaction kinetics. In most applications it is impossible to solve the CME directly due to its high dimensionality. Instead, indirect approaches based on realizations of the underlying Markov jump process are used, such as the stochastic simulation algorithm (SSA). In the SSA, however, every reaction event has to be resolved explicitly such that it becomes numerically inefficient when the system's dynamics include fast reaction processes or species with high population levels. In many hybrid approaches, such fast reactions are approximated as continuous processes or replaced by quasi-stationary distributions in either a stochastic or a deterministic context. Current hybrid approaches, however, almost exclusively rely on the computation of ensembles of stochastic realizations. We present a novel hybrid stochastic-deterministic approach to solve the CME directly. Our starting point is a partitioning of the molecular species into discrete and continuous species that induces a partitioning of the reactions into discrete-stochastic and continuous-deterministic processes. The approach is based on a WKB (Wentzel-Kramers-Brillouin) ansatz for the conditional probability distribution function (PDF) of the continuous species (given a discrete state) in combination with Laplace's method of integral approximation. The resulting hybrid stochastic-deterministic evolution equations comprise a CME with averaged propensities for the PDF of the discrete species that is coupled to an evolution equation of the related expected levels of the continuous species for each discrete state. In contrast to indirect hybrid methods, the impact of the evolution of discrete species on the dynamics of the continuous species has to be taken into account explicitly. The proposed approach is efficient whenever the number of discrete molecular species is small. We illustrate the performance of the new hybrid stochastic-deterministic approach in an application to model systems of biological interest.
Despite the success of highly active antiretroviral therapy (HAART) in the management of human immunodeficiency virus (HIV)-1 infection, virological failure due to drug resistance development remains a major challenge. Resistant mutants display reduced drug susceptibilities, but in the absence of drug, they generally have a lower fitness than the wild type, owing to a mutation-incurred cost. The interaction between these fitness costs and drug resistance dictates the appearance of mutants and influences viral suppression and therapeutic success. Assessing in vivo viral fitness is a challenging task and yet one that has significant clinical relevance. Here, we present a new computational modelling approach for estimating viral fitness that relies on common sparse cross-sectional clinical data by combining statistical approaches to learn drug-specific mutational pathways and resistance factors with viral dynamics models to represent the host-virus interaction and actions of drug mechanistically. We estimate in vivo fitness characteristics of mutant genotypes for two antiretroviral drugs, the reverse transcriptase inhibitor zidovudine (ZDV) and the protease inhibitor indinavir (IDV). Well-known features of HIV-1 fitness landscapes are recovered, both in the absence and presence of drugs. We quantify the complex interplay between fitness costs and resistance by computing selective advantages for different mutants. Our approach extends naturally to multiple drugs and we illustrate this by simulating a dual therapy with ZDV and IDV to assess therapy failure. The combined statistical and dynamical modelling approach may help in dissecting the effects of fitness costs and resistance with the ultimate aim of assisting the choice of salvage therapies after treatment failure.
Amoebae explore their environment in a random way, unless external cues like, e. g., nutrients, bias their motion. Even in the absence of cues, however, experimental cell tracks show some degree of persistence. In this paper, we analyzed individual cell tracks in the framework of a linear mixed effects model, where each track is modeled by a fractional Brownian motion, i.e., a Gaussian process exhibiting a long-term correlation structure superposed on a linear trend. The degree of persistence was quantified by the Hurst exponent of fractional Brownian motion. Our analysis of experimental cell tracks of the amoeba Dictyostelium discoideum showed a persistent movement for the majority of tracks. Employing a sliding window approach, we estimated the variations of the Hurst exponent over time, which allowed us to identify points in time, where the correlation structure was distorted ("outliers"). Coarse graining of track data via down-sampling allowed us to identify the dependence of persistence on the spatial scale. While one would expect the (mode of the) Hurst exponent to be constant on different temporal scales due to the self-similarity property of fractional Brownian motion, we observed a trend towards stronger persistence for the down-sampled cell tracks indicating stronger persistence on larger time scales.
The structure, interpretation and parameterization of classical compartment models as well as physiologically-based pharmacokinetic (PBPK) models for monoclonal antibody (mAb) disposition are very diverse, with no apparent consensus. In addition, there is a remarkable discrepancy between the simplicity of experimental plasma and tissue profiles and the complexity of published PBPK models. We present a simplified PBPK model based on an extravasation rate-limited tissue model with elimination potentially occurring from various tissues and plasma. Based on model reduction (lumping), we derive several classical compartment model structures that are consistent with the simplified PBPK model and experimental data. We show that a common interpretation of classical two-compartment models for mAb disposition-identifying the central compartment with the total plasma volume and the peripheral compartment with the interstitial space (or part of it)-is not consistent with current knowledge. Results are illustrated for the monoclonal antibodies 7E3 and T84.66 in mice.
Broad-spectrum antibiotic combination therapy is frequently applied due to increasing resistance development of infective pathogens. The objective of the present study was to evaluate two common empiric broad-spectrum combination therapies consisting of either linezolid (LZD) or vancomycin (VAN) combined with meropenem (MER) against Staphylococcus aureus (S. aureus) as the most frequent causative pathogen of severe infections. A semimechanistic pharmacokinetic-pharmacodynamic (PK-PD) model mimicking a simplified bacterial life-cycle of S. aureus was developed upon time-kill curve data to describe the effects of LZD, VAN, and MER alone and in dual combinations. The PK-PD model was successfully (i) evaluated with external data from two clinical S. aureus isolates and further drug combinations and (ii) challenged to predict common clinical PK-PD indices and breakpoints. Finally, clinical trial simulations were performed that revealed that the combination of VAN-MER might be favorable over LZD-MER due to an unfavorable antagonistic interaction between LZD and MER.
Immunodeficient mice are crucial models to evaluate the efficacy of monoclonal antibodies (mAbs). When studying mAb pharmacokinetics (PK), protection from elimination by binding to the neonatal Fc receptor (FcRn) is known to be a major process influencing the unspecific clearance of endogenous and therapeutic IgG. The concentration of endogenous IgG in immunodeficient mice, however is reduced, and this effect on the FcRn protection mechanism and subsequently on unspecific mAb clearance is unknown, yet of great importance for the interpretation of mAb PK data. We used a PBPK modelling approach to elucidate the influence of altered endogenous IgG concentrations on unspecific mAb clearance. To this end, we used PK data in immunodeficient mice, i.e. nude and severe combined immunodeficiency mice. To avoid impact of target-mediated clearance processes, we focussed on mAbs without affinity to a target antigen in these mice. In addition, intravenous immunoglobulin (IVIG) data of immunocompetent mice was used to study the impact of increased total IgG concentrations on unspecific therapeutic antibody clearance. The unspecific clearance is linear, whenever therapeutic IgG concentrations, i.e. mAb and IVIG concentrations are lower than FcRn; it can be non-linear if therapeutic IgG concentrations are larger than FcRn and endogenous IgG concentrations (e.g., under IVIG therapy). Unspecific mAb clearance of immunodeficient mice is effectively linear (under mAb doses as typically used in human). Studying the impact of reduced endogenous IgG concentrations on unspecific mAb clearance is of great relevance for the extrapolation to clinical species, e.g., when predicting mAb PK in immunosuppressed cancer patients.
Paclitaxel is a commonly used cytotoxic anticancer drug with potentially life-threatening toxicity at therapeutic doses and high interindividual pharmacokinetic variability. Thus, drug and effect monitoring is indicated to control dose-limiting neutropenia. Joerger et al. (2016) developed a dose individualization algorithm based on a pharmacokinetic (PK)/pharmacodynamic (PD) model describing paclitaxel and neutrophil concentrations. Furthermore, the algorithm was prospectively compared in a clinical trial against standard dosing (Central European Society for Anticancer Drug Research Study of Paclitaxel Therapeutic Drug Monitoring; 365 patients, 720 cycles) but did not substantially improve neutropenia. This might be caused by misspecifications in the PK/PD model underlying the algorithm, especially without consideration of the observed cumulative pattern of neutropenia or the platinum-based combination therapy, both impacting neutropenia. This work aimed to externally evaluate the original PK/PD model for potential misspecifications and to refine the PK/PD model while considering the cumulative neutropenia pattern and the combination therapy. An underprediction was observed for the PK (658 samples), the PK parameters, and these parameters were re-estimated using the original estimates as prior information. Neutrophil concentrations (3274 samples) were over-predicted by the PK/PD model, especially for later treatment cycles when the cumulative pattern aggravated neutropenia. Three different modeling approaches (two from the literature and one newly developed) were investigated. The newly developed model, which implemented the bone marrow hypothesis semiphysiologically, was superior. This model further included an additive effect for toxicity of carboplatin combination therapy. Overall, a physiologically plausible PK/PD model was developed that can be used for dose adaptation simulations and prospective studies to further improve paclitaxel/ carboplatin combination therapy.
Background: Severe bacterial infections remain a major challenge in intensive care units because of their high prevalence and mortality. Adequate antibiotic exposure has been associated with clinical success in critically ill patients. The objective of this study was to investigate the target attainment of standard meropenem dosing in a heterogeneous critically ill population, to quantify the impact of the full renal function spectrum on meropenem exposure and target attainment, and ultimately to translate the findings into a tool for practical application. Methods: A prospective observational single-centre study was performed with critically ill patients with severe infections receiving standard dosing of meropenem. Serial blood samples were drawn over 4 study days to determine meropenem serum concentrations. Renal function was assessed by creatinine clearance according to the Cockcroft and Gault equation (CLCRCG). Variability in meropenem serum concentrations was quantified at the middle and end of each monitored dosing interval. The attainment of two pharmacokinetic/pharmacodynamic targets (100% T->MIC, 50% T->4xMIC) was evaluated for minimum inhibitory concentration (MIC) values of 2 mg/L and 8 mg/L and standard meropenem dosing (1000 mg, 30-minute infusion, every 8 h). Furthermore, we assessed the impact of CLCRCG on meropenem concentrations and target attainment and developed a tool for risk assessment of target non-attainment. Results: Large inter-and intra-patient variability in meropenem concentrations was observed in the critically ill population (n = 48). Attainment of the target 100% T->MIC was merely 48.4% and 20.6%, given MIC values of 2 mg/L and 8 mg/L, respectively, and similar for the target 50% T->4xMIC. A hyperbolic relationship between CLCRCG (25-255 ml/minute) and meropenem serum concentrations at the end of the dosing interval (C-8h) was derived. For infections with pathogens of MIC 2 mg/L, mild renal impairment up to augmented renal function was identified as a risk factor for target non-attainment (for MIC 8 mg/L, additionally, moderate renal impairment). Conclusions: The investigated standard meropenem dosing regimen appeared to result in insufficient meropenem exposure in a considerable fraction of critically ill patients. An easy-and free-to-use tool (the MeroRisk Calculator) for assessing the risk of target non-attainment for a given renal function and MIC value was developed.
As a potentially toxic agent on nervous system and bone, the safety of aluminium exposure from adjuvants in vaccines and subcutaneous immune therapy (SCIT) products has to be continuously reevaluated, especially regarding concomitant administrations. For this purpose, knowledge on absorption and disposition of aluminium in plasma and tissues is essential. Pharmacokinetic data after vaccination in humans, however, are not available, and for methodological and ethical reasons difficult to obtain. To overcome these limitations, we discuss the possibility of an in vitro-in silico approach combining a toxicokinetic model for aluminium disposition with biorelevant kinetic absorption parameters from adjuvants. We critically review available kinetic aluminium-26 data for model building and, on the basis of a reparameterized toxicokinetic model (Nolte et al., 2001), we identify main modelling gaps. The potential of in vitro dissolution experiments for the prediction of intramuscular absorption kinetics of aluminium after vaccination is explored. It becomes apparent that there is need for detailed in vitro dissolution and in vivo absorption data to establish an in vitro-in vivo correlation (IVIVC) for aluminium adjuvants. We conclude that a combination of new experimental data and further refinement of the Nolte model has the potential to fill a gap in aluminium risk assessment. (C) 2017 Elsevier Inc. All rights reserved.
Background: Infliximab (IFX), an anti-TNF monoclonal antibody approved for the treatment of inflammatory bowel disease, is dosed per kg body weight (BW). However, the rationale for body size adjustment has not been unequivocally demonstrated [1], and first attempts to improve IFX therapy have been undertaken [2]. The aim of our study was to assess the impact of different dosing strategies (i.e. body size-adjusted and fixed dosing) on drug exposure and pharmacokinetic (PK) target attainment. For this purpose, a comprehensive simulation study was performed, using patient characteristics (n=116) from an in-house clinical database.
Methods: IFX concentration-time profiles of 1000 virtual, clinically representative patients were generated using a previously published PK model for IFX in patients with Crohn's disease [3]. For each patient 1000 profiles accounting for PK variability were considered. The IFX exposure during maintenance treatment after the following dosing strategies was compared: i) fixed dose, and per ii) BW, iii) lean BW (LBW), iv) body surface area (BSA), v) height (HT), vi) body mass index (BMI) and vii) fat-free mass (FFM)). For each dosing strategy the variability in maximum concentration Cmax, minimum concentration Cmin (= C8weeks) and area under the concentration-time curve (AUC), as well as percent of patients achieving the PK target, Cmin=3 μg/mL [4] were assessed.
Results: For all dosing strategies the variability of Cmin (CV ≈110%) was highest, compared to Cmax and AUC, and was of similar extent regardless of dosing strategy. The proportion of patients reaching the PK target (≈⅓ was approximately equal for all dosing strategies.
Understanding and reducing complex systems pharmacology models based on a novel input-response index
(2018)
A growing understanding of complex processes in biology has led to large-scale mechanistic models of pharmacologically relevant processes. These models are increasingly used to study the response of the system to a given input or stimulus, e.g., after drug administration. Understanding the input–response relationship, however, is often a challenging task due to the complexity of the interactions between its constituents as well as the size of the models. An approach that quantifies the importance of the different constituents for a given input–output relationship and allows to reduce the dynamics to its essential features is therefore highly desirable. In this article, we present a novel state- and time-dependent quantity called the input–response index that quantifies the importance of state variables for a given input–response relationship at a particular time. It is based on the concept of time-bounded controllability and observability, and defined with respect to a reference dynamics. In application to the brown snake venom–fibrinogen (Fg) network, the input–response indices give insight into the coordinated action of specific coagulation factors and about those factors that contribute only little to the response. We demonstrate how the indices can be used to reduce large-scale models in a two-step procedure: (i) elimination of states whose dynamics have only minor impact on the input–response relationship, and (ii) proper lumping of the remaining (lower order) model. In application to the brown snake venom–fibrinogen network, this resulted in a reduction from 62 to 8 state variables in the first step, and a further reduction to 5 state variables in the second step. We further illustrate that the sequence, in which a recursive algorithm eliminates and/or lumps state variables, has an impact on the final reduced model. The input–response indices are particularly suited to determine an informed sequence, since they are based on the dynamics of the original system. In summary, the novel measure of importance provides a powerful tool for analysing the complex dynamics of large-scale systems and a means for very efficient model order reduction of nonlinear systems.