FDA Success Stories and CDE Recommendations: How MIDD Accelerates Drug Discovery?

With the encouragement of innovative drug policies and the normalization of generic drug collection, in less than 10 years, the transformation and upgrading of traditional domestic pharmaceutical enterprises and the vigorous development of new innovative pharmaceutical enterprises have sprung up, and more and more enterprises have joined the competition track of innovative drugs.

How do you make drug creation easy? MIDD is a good solution to this problem.

Model-guided drug development (MIDD) can further improve the success rate of research and development on the basis of traditional experience. There have been many successful cases (including the well-known osimertinib and pembrolizumab), and CDE has also been encouraged and recommended in recent years, which is an essential and important means for future drug research and development. At present, the technology innovation companies in the forefront are Jingtai Technology, Wangshi Wisdom, Fimer Valley, Chuangteng Technology, Tige Medicine, Andu biology, etc., please see below.

25% of Class 1 innovation drug development is in use. What is MIDD?

MIDD is "Model-guided drug development" (slightly different from MBDD---- model-based drug development in the early years), which uses modeling and simulation techniques to integrate and quantify physiological, pharmacological, and disease process information to guide new drug development and decision-making.

The application of modeling and simulation in drug development and its lifecycle management involves multiple aspects, covering all stages from non-clinical to clinical research and post-marketing clinical reevaluation.

In the drug discovery stage, it can be used to discover and prove the binding ability of targets to candidate drugs; In the preclinical stage, some in vitro and in vivo correlations can be made to evaluate the safety and efficacy of FIH, and further calculate the dose of FIH. In the clinical development stage, it can be used to optimize the dosing regimen (such as dose climbing design, dosing interval, etc.), to recommend the dose for subsequent clinical trials, to provide supporting evidence for the package insert, and to reduce the need to carry out confirmatory clinical studies of the same dose as much as possible.

According to some research statistics, among the Class 1 innovative drugs approved by NMPA in 2018, about 1/4 used MIDD-related research methods, and another 1/3 were required to carry out MIDD-related analysis in post-marketing studies. In 2019, about 70% of the newly approved anti-tumor drugs by NMPA have carried out Pop PK and other studies, and the utilization rate of imported drugs is significantly higher than that of domestic drugs.

What are the relevant guidelines issued by the CDE?

In August 2020, CDE issued a notice on public solicitation of opinions on the "Technical Guidelines for Model-guided Drug Development (Draft)". At the end of 2020, CDE issued the Technical Guidelines for Model-Guided Drug Research and Development, which was officially implemented.

CDE in July 2022 release of "model to guide drug research and development (MIDD)" in the practice of new drug research and development enterprise survey questionnaire, which emphasizes "clinical pharmacology research is of great significance for new drug research and development", "the application of modeling and simulation technology is more and more widely, to improve the efficiency of new drug research and development, and guide decision-making plays an important role", In other words, model-guided drug research and development has a relatively large weight in the development of innovative drugs. Some of the relevant guidelines are shown in the table below.

Studies have confirmed that almost all the varieties submitted for new drug registration include the research content of MIDD, such as 9291 (osimertinib) and K drug (pembrolizumab), according to the review report of innovative drugs published by FDA.

➣ 9291 (Mr Bush for)

Osimertinib is mainly metabolized by the liver (CYP3A4/5 enzyme), with a small proportion of kidney clearance. In vitro transporter experiment data show that osimertinib is a breast cancer resistance protein inhibitor, suggesting that DDI may occur when the drug is combined with CYP3A enzyme substrate, inducer, inhibitor or BCRP substrate. Therefore, the DDI study of osimertinib with itraconazole (CYP3A strong inhibitor), rifampicin (CYP3A strong inducer), simvastatin (CYP3A substrate) and rosuvastatin (BCRP substrate) was carried out. In the above process, we used Simcyp software (PBPK modeling and simulation platform) to establish PB PK-DDI model, and finally developed a DDI prediction and simulation research strategy based on PBPK model. The data of in vitro and partial clinical trials were fully utilized, and non-essential clinical trials were exempted.

➣ K medicine (paabo single resistance method)

In the early stage of FIH, the Imax model was established to estimate that 1mg/kg of FIH could achieve in vitro receptor saturation. Thus, the maximum recommended starting dose of FIH was determined and the KEYNOTE-001 study was initiated.

The PK/PD tumor growth inhibition model of pembrolizumab was constructed using preclinical mouse data and extended to humans. The results showed that when the dose of pembrolizumab was 2mg/kg, the receptor occupancy rate of pembrolizumab exceeded 95%, and the probability of achieving a greater than 30% reduction in tumor volume reached a plateau, suggesting that 2mg/kg can be used as an effective dose in clinical trials. This dose level (2mg/kg) was used in a larger randomized controlled trial in patients with advanced melanoma and NSCLC.

The PopPK model analysis showed that the PK curve of pembrolizumab was consistent with that of classical therapeutic mabs, showing limited volume of distribution, low clearance, and low variability, and that intrinsic and extrinsic factors had no clinically meaningful effect on pembrolizumab exposure.

The results showed that the PK variability of the two dosing regimens was similar, and further confirmed the accuracy of the regimen predicted by the PopPK model. Ultimately, the FDA approved the application for a shift from weight-based to fixed-dose pembrolizumab administration.

In addition to the above representative cases, there are many other classic cases of FDA, such as:

1) Nesiritide, the FDA issued a letter of non-approval in 1999, suggesting that the applicant should optimize the dose to minimize the expected side effects and achieve the expected efficacy quickly. Through the E.R model, the applicant should finally simulate the dose that can achieve the best benefit/risk, and the applicant should choose this dose method for clinical trials. After the applicant submitted the trial results to the FDA, which confirmed their similarity to simulations, the drug was approved by the FDA in 2001 for the treatment of acute heart failure.

2) Zoledronicacid, which the FDA has shown through modeling analysis to be associated with nephrotoxicity, and after discussion with the applicant, recommended dose adjustment based on AUC for patients with mild-to-moderate renal insufficiency, and added this recommendation to the label.

Cases of MIDD literature in academia in recent years

In addition to the above successful cases based on regulation, there are also many good research results in the literature. Here are some examples.

➣ population pharmacokinetic model

Structural models typically include absorption, disposition, and pharmacodynamic models. The common absorption models include zero-order absorption model, first-order absorption model, mixed model, gradual absorption model and special model. The disposition model refers to the traditional compartmental model in pharmacokinetics, including the first, second, and third compartment models. Pharmacodynamic models include linear models or Sigmoid models. The most important parameters of the pharmacodynamic model include the maximum effect of the drug (Emax) and the concentration of the drug at half of the maximum effect (EC50).

Example 1: The model supported the preclinical pharmacokinetics evaluation of Lorlatinib

Journal of Pharmaceutical Sciences (2022), using a series of transgenic mice to avoid the influence of transporters and enzymes on drug pharmacokinetics, and then taking Lorlatinib, the corresponding PK data were measured. Through the corresponding population pharmacokinetic modeling method, it was finally inferred that Lorlatinib was mainly dissolved in the stomach of mice, and the absorption rate in the intestine was reduced. The absorption curves of Lorlatinib submitted to the FDA used a mixed model, and the final model fit curves were also very similar.

➣ pharmacokinetics/pharmacodynamics model

PK and PD model is a comprehensive research in vivo pharmacokinetic process and the effects on the kinetics of quantitative indicators, described the pharmacokinetic and pharmacodynamic time, concentration of drugs, drug effect to study the relationship between the three organically unifies in together, contribute to a more comprehensive and accurate understanding of the drug's effect on dose (concentration) and the laws of time and change. The PK-PD model has a wide range of applications, such as active drug screening, phase I clinical maximum tolerance measurement determination, mechanism of action research, clinical guidance of drug use, optimization of drug administration, preparation evaluation, and so on.

Example 2: Prospective clinical validation of propofol PK-PD

Most existing models of clinical research data from a specific population, strictly speaking, the use of these models is limited to the crowds, and to conduct a broader propofol, PK and PD model aims in children and adults with general anesthesia, elderly, obese adult subjects prospective validation of the model, finally confirmed in the clinical anesthesia, The overall predictive performance of the PK-PD model in this study was better (favoring population PK-PD).

➣ based on physiological pharmacokinetic model

Physiological pharmacokinetics (PBPK) model can simulate the time-dependent changes of drugs in various tissues and blood through the physicochemical properties of drugs and the data obtained from in vitro tests. In addition, the PK behavior of drugs in animals can be extrapolated between species to predict the PK of humans. The PBPK model takes into account the process factors such as drug absorption, distribution and elimination, and describes the exposure of drugs in various tissues and organs and their changes over time. Thus, the PBPK model allows early prediction of local organ tissue concentrations, which can be correlated with pharmacodynamic models assessing the response of any given tissue.

Example 3: Studies on the drug formation of Danirixin free base and various salts

The PBPK model was used to investigate the free base and various salts of Danirixin (GSK1325756), a drug currently under clinical investigation for the treatment of chronic obstructive pulmonary disease. The PBPK model confirmed that the free base bioavailability was reduced in the presence of a proton pump inhibitor (AUC decreased to 42% of the original, AUC decreased to 42% of the original). No equivalent reduction in PK exposure was observed with hydrobromate, and hydrobromate exposure was increased compared with free base (1.32-fold increase in AUC and 1.44-fold increase in Cmax). The simulation results were consistent with subsequent clinical trial results in that hydrobromate reduced the variability of drug exposure. Moreover, exposure was not affected by the combination of proton pump inhibitors, and the simulation results provide support for clinical studies of hydrobromate.

Domestic MIDD status: Jingtai Technology, Fanmogu, Taige Medicine, Andu Biology

In the past 10 to 20 years, a number of technology-based innovative companies have emerged in the domestic drug research and development industry, such as Jingtai Technology, which is well known by the industry for its deep cultivation of solid state drug research, and Wangshi Wisdom, which is the direction of AI small molecule drug research and development. Famervalley, Chuangteng Technology, and some clinical CRO companies such as Tige Pharmaceutical, Andu Biological, etc., focus on model-guided drug development.

In addition to the above technology-based companies with strong focus on technology direction, regulatory departments, hospitals and universities have always published corresponding research, but the overall trend is academic; For model-guided drug research and development, enterprises that pay more attention to the specific implementation of projects are still in the early contact stage, and tend to cooperate, solve some problems, and conduct preliminary exploration in related fields.

For example, the enterprise users of Fanmo Valley are as follows: Hutchison Whampo, Baiji, Dizhe Pharmaceutical, Hualin Pharmaceutical, Howson Pharmaceutical, Kelun Pharmaceutical, Luoxin Pharmaceutical, Huadong Pharmaceutical, Taiwan Central, Pharmatech, Hechuan Pharmaceutical, Hong Kong InSilico, Shengsu New Drug, Kanglong Huacheng and so on.

Barriers to widespread use of MIDD

Although the CDE has issued relevant guiding principles, the degree of awareness and recognition is still not high. At the same time, the research and development of traditional drugs based on experience is still the mainstream of domestic new drug research and development, and capital is still following the highly successful leaders.

In addition, the implementation of MIDD requires multidisciplinary comprehensive talents based on awareness and technology parallel, including quantitative clinical pharmacology, statistics, program engineering, data management, etc., and requires deep cooperation with clinical PI.

Moreover, existing models need to be constantly updated and evolved, and the study of models is still in its early stage, so the development and application of various models need to be studied deeply, which is definitely not a simple writing of metaphysics.

Despite the layers of obstacles, MIDD to join the whole life cycle of new drug research is imperative, and validated by a large number of successful cases, need before clinical trials begin, even as a candidate drugs into the development period, will be included in the part of the MIDD plan work, constantly using experience, from the perspective of computation for the chance of success.

We will finally make it easy to create drugs.

Any work that can increase the probability of success in the development of new drugs, medicine people can not easily give up......

With the encouragement of innovative drug policies and the normalization of generic drug collection, in less than 10 years, the transformation and upgrading of traditional domestic pharmaceutical enterprises and the vigorous development of new innovative pharmaceutical enterprises have sprung up, and more and more enterprises have joined the competition track of innovative drugs.


How do you make drug creation easy? MIDD is a good solution to this problem.


Model-guided drug development (MIDD) can further improve the success rate of research and development on the basis of traditional experience. There have been many successful cases (including the well-known osimertinib and pembrolizumab), and CDE has also been encouraged and recommended in recent years, which is an essential and important means for future drug research and development. At present, the technology innovation companies in the forefront are Jingtai Technology, Wangshi Wisdom, Fimer Valley, Chuangteng Technology, Tige Medicine, Andu biology, etc., please see below.


25% of Class 1 innovation drug development is in use. What is MIDD?


MIDD is "Model-guided drug development" (slightly different from MBDD---- model-based drug development in the early years), which uses modeling and simulation techniques to integrate and quantify physiological, pharmacological, and disease process information to guide new drug development and decision-making.


The application of modeling and simulation in drug development and its lifecycle management involves multiple aspects, covering all stages from non-clinical to clinical research and post-marketing clinical reevaluation.


In the drug discovery stage, it can be used to discover and prove the binding ability of targets to candidate drugs; In the preclinical stage, some in vitro and in vivo correlations can be made to evaluate the safety and efficacy of FIH, and further calculate the dose of FIH. In the clinical development stage, it can be used to optimize the dosing regimen (such as dose climbing design, dosing interval, etc.), to recommend the dose for subsequent clinical trials, to provide supporting evidence for the package insert, and to reduce the need to carry out confirmatory clinical studies of the same dose as much as possible.


According to some research statistics, among the Class 1 innovative drugs approved by NMPA in 2018, about 1/4 used MIDD-related research methods, and another 1/3 were required to carry out MIDD-related analysis in post-marketing studies. In 2019, about 70% of the newly approved anti-tumor drugs by NMPA have carried out Pop PK and other studies, and the utilization rate of imported drugs is significantly higher than that of domestic drugs.


What are the relevant guidelines issued by the CDE?


In August 2020, CDE issued a notice on public solicitation of opinions on the "Technical Guidelines for Model-guided Drug Development (Draft)". At the end of 2020, CDE issued the Technical Guidelines for Model-Guided Drug Research and Development, which was officially implemented.


CDE in July 2022 release of "model to guide drug research and development (MIDD)" in the practice of new drug research and development enterprise survey questionnaire, which emphasizes "clinical pharmacology research is of great significance for new drug research and development", "the application of modeling and simulation technology is more and more widely, to improve the efficiency of new drug research and development, and guide decision-making plays an important role", In other words, model-guided drug research and development has a relatively large weight in the development of innovative drugs. Some of the relevant guidelines are shown in the table below.


Studies have confirmed that almost all the varieties submitted for new drug registration include the research content of MIDD, such as 9291 (osimertinib) and K drug (pembrolizumab), according to the review report of innovative drugs published by FDA.


➣ 9291 (Mr Bush for)


Osimertinib is mainly metabolized by the liver (CYP3A4/5 enzyme), with a small proportion of kidney clearance. In vitro transporter experiment data show that osimertinib is a breast cancer resistance protein inhibitor, suggesting that DDI may occur when the drug is combined with CYP3A enzyme substrate, inducer, inhibitor or BCRP substrate. Therefore, the DDI study of osimertinib with itraconazole (CYP3A strong inhibitor), rifampicin (CYP3A strong inducer), simvastatin (CYP3A substrate) and rosuvastatin (BCRP substrate) was carried out. In the above process, we used Simcyp software (PBPK modeling and simulation platform) to establish PB PK-DDI model, and finally developed a DDI prediction and simulation research strategy based on PBPK model. The data of in vitro and partial clinical trials were fully utilized, and non-essential clinical trials were exempted.


➣ K medicine (paabo single resistance method)


In the early stage of FIH, the Imax model was established to estimate that 1mg/kg of FIH could achieve in vitro receptor saturation. Thus, the maximum recommended starting dose of FIH was determined and the KEYNOTE-001 study was initiated.


The PK/PD tumor growth inhibition model of pembrolizumab was constructed using preclinical mouse data and extended to humans. The results showed that when the dose of pembrolizumab was 2mg/kg, the receptor occupancy rate of pembrolizumab exceeded 95%, and the probability of achieving a greater than 30% reduction in tumor volume reached a plateau, suggesting that 2mg/kg can be used as an effective dose in clinical trials. This dose level (2mg/kg) was used in a larger randomized controlled trial in patients with advanced melanoma and NSCLC.


The PopPK model analysis showed that the PK curve of pembrolizumab was consistent with that of classical therapeutic mabs, showing limited volume of distribution, low clearance, and low variability, and that intrinsic and extrinsic factors had no clinically meaningful effect on pembrolizumab exposure.


The results showed that the PK variability of the two dosing regimens was similar, and further confirmed the accuracy of the regimen predicted by the PopPK model. Ultimately, the FDA approved the application for a shift from weight-based to fixed-dose pembrolizumab administration.


In addition to the above representative cases, there are many other classic cases of FDA, such as:


1) Nesiritide, the FDA issued a letter of non-approval in 1999, suggesting that the applicant should optimize the dose to minimize the expected side effects and achieve the expected efficacy quickly. Through the E.R model, the applicant should finally simulate the dose that can achieve the best benefit/risk, and the applicant should choose this dose method for clinical trials. After the applicant submitted the trial results to the FDA, which confirmed their similarity to simulations, the drug was approved by the FDA in 2001 for the treatment of acute heart failure.


2) Zoledronicacid, which the FDA has shown through modeling analysis to be associated with nephrotoxicity, and after discussion with the applicant, recommended dose adjustment based on AUC for patients with mild-to-moderate renal insufficiency, and added this recommendation to the label.


Cases of MIDD literature in academia in recent years


In addition to the above successful cases based on regulation, there are also many good research results in the literature. Here are some examples.


➣ population pharmacokinetic model


Structural models typically include absorption, disposition, and pharmacodynamic models. The common absorption models include zero-order absorption model, first-order absorption model, mixed model, gradual absorption model and special model. The disposition model refers to the traditional compartmental model in pharmacokinetics, including the first, second, and third compartment models. Pharmacodynamic models include linear models or Sigmoid models. The most important parameters of the pharmacodynamic model include the maximum effect of the drug (Emax) and the concentration of the drug at half of the maximum effect (EC50).


Example 1: The model supported the preclinical pharmacokinetics evaluation of Lorlatinib


Journal of Pharmaceutical Sciences (2022), using a series of transgenic mice to avoid the influence of transporters and enzymes on drug pharmacokinetics, and then taking Lorlatinib, the corresponding PK data were measured. Through the corresponding population pharmacokinetic modeling method, it was finally inferred that Lorlatinib was mainly dissolved in the stomach of mice, and the absorption rate in the intestine was reduced. The absorption curves of Lorlatinib submitted to the FDA used a mixed model, and the final model fit curves were also very similar.


➣ pharmacokinetics/pharmacodynamics model


PK and PD model is a comprehensive research in vivo pharmacokinetic process and the effects on the kinetics of quantitative indicators, described the pharmacokinetic and pharmacodynamic time, concentration of drugs, drug effect to study the relationship between the three organically unifies in together, contribute to a more comprehensive and accurate understanding of the drug's effect on dose (concentration) and the laws of time and change. The PK-PD model has a wide range of applications, such as active drug screening, phase I clinical maximum tolerance measurement determination, mechanism of action research, clinical guidance of drug use, optimization of drug administration, preparation evaluation, and so on.


Case 2: Prospective clinical validation of propofol PK-PD


Most existing models of clinical research data from a specific population, strictly speaking, the use of these models is limited to the crowds, and to conduct a broader propofol, PK and PD model aims in children and adults with general anesthesia, elderly, obese adult subjects prospective validation of the model, finally confirmed in the clinical anesthesia, The overall predictive performance of the PK-PD model in this study was better (favoring population PK-PD).


➣ based on physiological pharmacokinetic model


Physiological pharmacokinetics (PBPK) model can simulate the time-dependent changes of drugs in various tissues and blood through the physicochemical properties of drugs and the data obtained from in vitro tests. In addition, the PK behavior of drugs in animals can be extrapolated between species to predict the PK of humans. The PBPK model takes into account the process factors such as drug absorption, distribution and elimination, and describes the exposure of drugs in various tissues and organs and their changes over time. Thus, the PBPK model allows early prediction of local organ tissue concentrations, which can be correlated with pharmacodynamic models assessing the response of any given tissue.


Example 3: Studies on the drug formation of Danirixin free base and various salts


The PBPK model was used to investigate the free base and various salts of Danirixin (GSK1325756), a drug currently under clinical investigation for the treatment of chronic obstructive pulmonary disease. The PBPK model confirmed that the free base bioavailability was reduced in the presence of a proton pump inhibitor (AUC decreased to 42% of the original, AUC decreased to 42% of the original). No equivalent reduction in PK exposure was observed with hydrobromate, and hydrobromate exposure was increased compared with free base (1.32-fold increase in AUC and 1.44-fold increase in Cmax). The simulation results were consistent with subsequent clinical trial results in that hydrobromate reduced the variability of drug exposure. Moreover, exposure was not affected by the combination of proton pump inhibitors, and the simulation results provide support for clinical studies of hydrobromate.



Domestic MIDD status: Jingtai Technology, Fanmogu, Taige Medicine, Andu Biology


In the past 10 to 20 years, a number of technology-based innovative companies have emerged in the domestic drug research and development industry, such as Jingtai Technology, which is well known by the industry for its deep cultivation of solid state drug research, and Wangshi Wisdom, which is the direction of AI small molecule drug research and development. Famervalley, Chuangteng Technology, and some clinical CRO companies such as Tige Pharmaceutical, Andu Biological, etc., focus on model-guided drug development.


In addition to the above technology-based companies with strong focus on technology direction, regulatory departments, hospitals and universities have always published corresponding research, but the overall trend is academic; For model-guided drug research and development, enterprises that pay more attention to the specific implementation of projects are still in the early contact stage, and tend to cooperate, solve some problems, and conduct preliminary exploration in related fields.


For example, the enterprise users of Fanmo Valley are as follows: Hutchison Whampo, Baiji, Dizhe Pharmaceutical, Hualin Pharmaceutical, Howson Pharmaceutical, Kelun Pharmaceutical, Luoxin Pharmaceutical, Huadong Pharmaceutical, Taiwan Central, Pharmatech, Hechuan Pharmaceutical, Hong Kong InSilico, Shengsu New Drug, Kanglong Huacheng and so on.


Barriers to widespread use of MIDD


Although the CDE has issued relevant guiding principles, the degree of awareness and recognition is still not high. At the same time, the research and development of traditional drugs based on experience is still the mainstream of domestic new drug research and development, and capital is still following the highly successful leaders.


In addition, the implementation of MIDD requires multidisciplinary comprehensive talents based on awareness and technology parallel, including quantitative clinical pharmacology, statistics, program engineering, data management, etc., and requires deep cooperation with clinical PI.


Moreover, existing models need to be constantly updated and evolved, and the study of models is still in its early stage, so the development and application of various models need to be studied deeply, which is definitely not a simple writing of metaphysics.


Despite the layers of obstacles, MIDD to join the whole life cycle of new drug research is imperative, and validated by a large number of successful cases, need before clinical trials begin, even as a candidate drugs into the development period, will be included in the part of the MIDD plan work, constantly using experience, from the perspective of computation for the chance of success.


We will finally make it easy to create drugs.


Any work that can increase the probability of success in the development of new drugs, medicine people can not easily give up......