Disease and Treatment Modeling

Last modifications: 02/08/2019 by FDM

• Prof. Annick Rousseau

• Dr. Aurelie Premaud
• Dr. Paul Carrier
• Prof. Véronique. Loustaud-Ratti 
• Prof. Marie Essig
• Prof. Franck Saint-Marcoux
• Dr. Claire Villeneuve
• Dr. Caroline Monchaud 
• Dr. Jean Debord
• Dr. Antoine Humeau
• Tom Nanga

Our experience in PK and PK/PD modeling as well as in biostatistics, led us to develop since 2011 a new research topic about disease progression modeling, based on cohort studies conducted in our unit or in the FHU SUPORT. In patient cohorts, we developed and externally validated a dynamic predictive score of short- and long-term kidney graft failure based on pre-transplantation and early post-transplantation factors (collected within the first year), the onset of de novo DSA, and acute rejection episodes. The development of a dynamic score which can be updated during patient follow-up was a new approach in transplantation. However, our score as well as other published scores predicting graft survival considered death as a censored event (at least when the patient died with a functional graft), which is unsatisfactory. Indeed, graft failure and death are competing risk events. We now aim to develop an overall approach integrating different levels of information related to graft and patient. In patients who died with a functioning graft, a major question is: was the evolution of their comorbidity(ies) independent of their status of transplanted patient and of the graft status? Conversely, comorbidities can participate in (i) the decline of graft function and (ii) the emergence of new comorbidities. The interaction between transplantation and comorbidities, and their joint evolution is a complex research question that no model currently takes into account; this question will be at the heart of our disease progression modeling project for the upcoming years.

The number of comorbidities increases with age and an increasing proportion of transplant patients are more than 60 years old. In France, the annual report of the French Biomedicine Agency (ABM) shows that in 2013, nearly 20% of renal transplantations were performed in patients over 65 years of age and about 10% of patients were aged at least 70. The number of liver transplant recipients aged from 56 to 65 years increased by 152% and now represents 40% of new enrollments compared to 27% in 2002. The number of subjects aged 66 and over increased by 450% in 11 years. Therefore, we can expect that the management of comorbidities will become increasingly important in renal and liver transplantation, which represents the treatment of choice for end-stage organ failure.

The final objective of this project is to model the risks of graft failure and death in kidney and in liver transplant recipients in parallel, which can be linked: 1) to transplantation; 2) to preexisting comorbidities or to comorbidities which emerged after transplantation; and 3) to risk factors linked to the association of comorbidities and transplantation. A survival model will be developed, considering the competitive risks of graft loss and death. This model will allow quantifying the benefit/risk ratio of an intervention or a therapeutic strategy by taking into account both transplantation and comorbidities. The idea is to develop survival models with competing risks combining: (i) structural equation modeling with latent variables; and (ii) mixed-effects models to describe the changes over time of renal function, cardiovascular parameters, and parameters characterizing the metabolic disturbances. The structural equation models will be used to clarify the causal pathways and to link the pejorative factors involved in the graft function deterioration and the development of comorbidities. The models will also take into account the complications associated with comorbidities. We will use mixture models with latent classes to explain and graduate the heterogeneity of both risks and survivals.

The study of comorbidies will include the stratification of risks in a dynamic approach (i.e. the risk category will be reassessed over time for each patient) into different levels (low, intermediate, high or increasing risk). Identifying as soon as possible patients with an increasing risk of worsening comorbidity or onset of de novo comorbidity is another main objective. The factors investigated in this step will involve: (i) the patient’s pre-transplant clinical history, (ii) indicators of graft function, (iii) comorbidities, (iv) factors linked to transplantation such as the administered treatments and risks induced by immunosuppression, (v) genetic polymorphisms and biomarkers predictive of the evolution of graft function identified in the biomarker projects developed in our unit (see below Theme 3). From our previous results in renal transplantation, the first emerging comorbidities studied will be new-onset diabetes, metabolic and cardiovascular diseases and cancer (particularly PTLD). Specifically in liver transplantation, one work is starting which aims to model renal failure as a key emerging comorbiditiy (see below § “Prevention of renal failure in liver transplant patients”).

There is a gap between the positive results obtained in clinical trials under controlled conditions and the effectiveness observed in the larger cohort of patients exposed to the drugs in real life conditions and in the longer term. Decision making based on results of randomized clinical studies has therefore shown its limits and alternative approaches are needed. In the present project, we will develop for the first time a system-based tool that will permit to anticipate unobserved favorable and unfavorable effects of treatments in a robust manner, without the need for systematic and intensive data collection. An integrated approach for decision making in clinical practice regarding the use of the ISD tacrolimus and mycophenolate will be proposed, based on mechanistic disease/PK/PD modeling and simulations, robust uncertainty quantification and benefit/risk analysis. Sensitivity analysis will be used to identify relevant components at the molecule, cell, tissue, organ and organism levels, preserving the mechanistic pathway connections between them. The resulting models should be able to predict treatment effects and patient outcome. The proposed approach is based on the use of information from different sources, including the literature and observational data from the unit clinical databases and prospective cohorts, for characterizing the context and further extrapolating findings from controlled clinical trials.

The present project aims to propose an integrated model of immune response after solid organ transplantation. The proposed models will cover all the relevant pathways from dosing regimen to therapeutic and toxic clinical responses. In the context of organ transplantation, the most important pathways of immune response and the mode of action of immunosuppressants have already been elucidated and their main determinants identified. However mechanistic/predictive models including these pathways are still lacking.

The potentially additive or synergistic effects of different drug combinations as frequently used in solid organ transplantation will be characterized in a quantitative manner. We have some encouraging preliminary results in regard to this: we have recently developed a mechanism-based PKPD model describing the simultaneous effects of the 3 most used IS drugs in liver transplantation namely tacrolimus, mycophenolate and methylprednisolone.

We will characterize drug exposure in a robust manner, using physiologically based pharmacokinetics (PBPK) modeling. To date, only one paper reports the results of a PBPK simulation for tacrolimus. Moreover, the validity of this model was only tested on blood concentrations collected in a small group of 10 patients.

Clinical trial simulations are already used in the context of drug development for study design optimization. This approach will be applied to ISD for the first time in order to help personalize treatment in special populations exemplified by pediatric patients. Appropriate software such as R and NONMEM will be used in an integrated manner to explore the influence of differences in relevant patient characteristics on clinical endpoints using simulations of different “what if?” realistic scenarios. Predictive models, real-life situations (e.g., compliance) and patient characteristics will be included in the exercise.

Phase III: Decision modeling: uncertainty quantification and benefit/risk assessment

A survey recently performed by the EMA reviewed the different approaches for balancing benefits and risks in decision-making about medicinal products. After exploring a range of qualitative and quantitative tools, the authors concluded that quantitative approaches are sufficiently comprehensive to enable the benefit-risk balance to be represented numerically by incorporating the value or utilities of favorable and unfavorable effects, along with probabilities representing the uncertainties of those effects.

We will use quantitative B/RA approaches with predictive modeling and simulations to assess the clinical effectiveness of immunosuppressive treatments in special populations, such as children. We will subsequently develop a decision tool (which will take the format of a dosing algorithm) based on treatment efficacy, safety, practicability, and feasibility in different groups of patients and at different periods post-transplantation.

The impact of uncertainty in the models on the decision taken will be evaluated using risk analysis. The overall uncertainty can source either from unexplained between-subject and/or between-occasion variability in parameters related to the disease, PK or PD processes, or from model misspecifications, analytical errors or protocol deviations (errors in sampling times, dosing, recording measures or outcomes, etc.). It should be noted that the ultimate decision (e.g. B/R analysis) can be influenced by the results of several modeling exercises (e.g. modeling of different independent criteria related to favorable and/or unfavorable effects). Sensitivity analysis will be carried out and “what-if” and “worse-case” scenarios tested to characterize the decision context.

Kidney failure is of major concern in patients with cirrhosis on the LT waiting list (hepatorenal syndrome and other disorders). Furthermore, after LT, chronic kidney disease (15%) is an under- and late-diagnosed independent predictive factor of patient and graft survival. Early identification of “at-risk” patients may allow specific care, for instance early access to LT and/or early conversion to mToR inhibitors after LT.

The first step to prevent renal failure in patients on the waiting list and after LT is to define tools aiming at detecting those at high-risk of renal dysfunction. This requires to:

• Develop a sensitive and specific tool to evaluate GFR in cirrhotic patients, which is currently lacking
• Identify urinary biomarkers of subclinical kidney lesions in patients at risk of hepatorenal syndrome or kidney failure before or after LT 
• Associate eGFR and biomarkers with clinical data (in particular from the FHU SUPORT “SCORED” database) 

The second step consists of a cohort survey of LT patients from the FHU (collaboration with Prof. E Salamé, Tours), in order to:

• Implement these “kidney” tools in patient follow-up
• Measure the impact of their use on LT survival and CKD progression, and on the choice of an immunosuppressive regimen more protective for the kidney.