The most frequently used paradigm for drug administration is "the same dose works for everyone". In this approach, pharmaceutical companies obtain the safety and efficacy data of drugs in a specific population and recommend the needed dose to reach the therapeutic window; that means achieving the plasmatic concentrations at which the majority of the population will present the desired effect with an acceptable incidence of adverse reactions.¹

While this approach is effective for drugs with a wide therapeutic window, where toxic concentrations are distant from therapeutic levels, it falls short for drugs with a narrow therapeutic window.² Thus, neglecting inter-individual variability may result in therapeutic failure, either by adverse reactions or lack of efficacy.³

Reports from the United States of America indicate that up to 23% of the current pharmacologic treatments result in adverse effects⁴, and up to 17% of the treatments are considered inefficient. Both situations result in a high cost for the patient and health systems. Therefore, alternative approaches, such as population pharmacokinetic (PopPK) modeling, are necessary to advance personalized medicine and ensure patients receive tailored treatments that meet their needs.

Several tools have been developed to assess the plasma concentration variability and factors affecting drug exposure to investigate and understand the variability of the pharmacologic responses between patients. One of the most promising approaches is the study of PopPK and the development of pharmacokinetic (PK) mathematical models. These latter use the power of computer technology to describe, examine, and comprehend the impact of various factors on PK variability and quantify the extent to which they influence plasma concentrations.⁵

Hemophilia A is the most common hereditary disorder of hemostasis and affects 1 out of 5,000 males; it is calculated that it occurs in more than 400,000 males worldwide. Quick and stable pharmacologic control in these patients is essential to avoid further complications, such as articular bleeding, which can lead to chronic articular disease, or intracranial bleeding that might complicate to seizures, paralysis, or cognitive problems.⁶

PK is a branch of Pharmacology that studies a drug's absorption, distribution, metabolism, and excretion over time. In other words, it is the study of the drug's transit from its entry into the body until its elimination.⁷

PK studies are commonly conducted to understand the body's exposure to the drug after administration and to determine how those four processes occur. Researchers recruit healthy volunteers or patients who receive the drug under investigation to conduct PK studies. Blood samples are taken at strategically planned intervals to determine plasma concentrations of the drug and study the four processes above.⁸ Time of evaluation through these studies may range from hours for drugs with a short half-life, which is quickly eliminated, to days for drugs with a long half-life.

While PK studies are essential to understand a drug's transit in the body, they can be challenging and costly from a logistic standpoint. In contrast, PopPK aims to study absorption, distribution, metabolism, and excretion at a population level. That involves calculating averages and variability of these PK processes and identifying factors for PK variability in a population to establish associations between the patients' characteristics and plasmatic concentrations.

A larger number of samples is required to study these associations. However, PopPK employs strategies such as limited sampling, which entails recruiting more patients while obtaining only one or two samples for each patient. Such strategies simplify logistics, significantly reduce the cost of PK studies, and allow the creation of PopPK models with "real-world patients" beyond rigid protocols. For these reasons, PopPK offers a better description of the population's variability, and its use is becoming increasingly common.⁹

At first glance, "population pharmacokinetics" might suggest that individual patients are disregarded. However, the individual's significance in population models is reemphasized by the inclusion of "variability". Each individual contributes information that helps identify tendencies in drug exposure changes.

Population models may encompass covariates that have demographic information, age, sex, weight, height, clinical data (concomitant diseases, concomitant drugs), and even genetic factors.¹⁰

Understanding the factors that impact a drug's absorption, distribution, metabolism, and excretion at individual and population levels helps us estimate the dose necessary to reach the therapeutic window. That is particularly important as concentrations exceeding the therapeutic window are associated with an increased risk of having adverse reactions. In contrast, concentrations lower than the established therapeutic window (subtherapeutic concentration) may not produce the desired effect. Both scenarios can indeed lead to therapeutic failures.

Development of Population Pharmacokinetic Models

To study plasmatic concentrations at a population level, PopPK models have been developed using the particular characteristics of each patient. These models are non-linear mixed-effects models. These are called "non-linear" because the relationship between the dependent variable (plasmatic concentrations) and the other variables is non-linear. For example, values over 30 mL/min in creatinine clearance are not associated with a reduction in the clearance of certain drugs. However, values under 30 mL/min show an exponential decrease in the drugs excreted by the renal way. They are also called "mixed effects" because these models have two kinds of parameters: fixed effects, which do not vary among the population, and random effects, which vary among the population.¹¹

The first step in developing a PopPK model is calculating the structural parameters, which depend on the drug's characteristics and route of administration. Those parameters are the absorption constant Ka (for drugs not administered intravenously), the distribution volume (Vd, with one or more compartments depending on the drug), and the clearance (CL), reflecting the drug elimination from the body.

Once the population structural parameters are obtained, the covariates can be added to the model. Covariates are intrinsic or extrinsic factors affecting positively or negatively one or more structural parameters such as age, size, weight, kidney function, or administered formulation.

The model determines the significance of these variables by using mathematical and statistical criteria. When the model shows a good fit, adequately describing the drug's transit through the body and including significant factors impacting PK, it must pass through a validation process.¹¹

Model validation ensures that the model adequately describes PK in patients who participated in model development (internal validation) and in patients who were not part of the study to develop the model (external validation). That provides confidence in using the model's calculations in clinical practice.

Implementing mathematical models in clinical practice is relatively new, so doctors are often unfamiliar with them. However, it is essential to note that these models are increasingly being developed for drugs requiring precise dosing. They help understand not only variants playing a role in drug exposure but also the extent to which those variants affect all PK processes.¹¹

One of the most promising applications from a clinical perspective is the use of simulations based on Bayes' theorem to personalize doses of narrow therapeutic window drugs.

Bayes' Theorem and Its Usefulness in Pharmacology

Determining the appropriate dose of narrow therapeutic window drugs involves starting drug administration with the amount recommended in the brochure or clinical practice guidelines. Then, depending on the therapeutic monitoring (determination of plasmatic concentrations), observation of the lack of pharmacologic effect, or occurrence of adverse reactions, the decision to increase or reduce doses is made.

However, this "trial and error" approach often results in patients not starting with the optimal dose, which can negatively impact their health.¹² One of the most promising applications of PK models is using Bayes' theorem to estimate the PK of these drugs. PopPK models, which describe the factors associated with the variability of plasmatic concentrations and calculate these factors' variability in the population, enable using Baye's theorem for dose personalization.

Bayes' theorem, known as the "causes probability" formula, is a base for calculating PK parameters. This theorem is based on the logic that the probability of making an observation depends on the factors involved in that observation and the known variability of those factors. For example, a Bayesian estimate can tell us the probability of a plasmatic concentration being above the therapeutic window if we know which factors influence plasmatic concentrations, such as renal failure or co-administration of inducers or inhibitors of metabolism. To apply Bayes' theorem in PK, a PopPK model is needed. Using these tools together can eliminate the "trial and error" approach, and the therapeutic effect can be achieved as soon as possible through dose personalization.¹²

To make the Bayesian estimates, a patient's therapeutic monitoring data can be combined with the PopPK model equations that describe how a drug behaves in the population. These estimates can determine the drug's past and future plasma concentrations based on the patient's current plasmatic concentrations and associated factors with previously selected confidence intervals. This process has immense therapeutic value, allowing accurate individualization of the patient's dose based on their clinical and demographic characteristics.

As a result, the patient is kept within the therapeutic window for long periods, ensuring a suitable therapeutic effect and the risk of adverse reactions.^11,12 This approach represents a significant advance in precision medicine, acknowledging differences and moving away from the "one dose for all" paradigm. It is particularly relevant in diseases in which pharmacologic treatment is essential for keeping good health and a positive prognosis for the patient.

Factor VIII PK Models and Bayesian Estimates in the Treatment of Hemophilia

Current guidelines for hemophilia A treatment recommends prophylaxis to prevent severe bleeding in joints that could result in arthropathy.⁶ It has been shown that proper peak levels (C_max) of FVIII significantly reduce physical activity-related joint bleeding. At the same time, an area under the curve (AUC) is associated with reduced subclinical bleeding levels.¹³

For this reason, keeping suitable plasmatic concentrations of factor VIII (FVIII) for each patient is fundamental. Although a traditional PK approach seems theoretically possible, it is not easy in the real world. The International Society of Thrombosis and Hemostasis (ISTH) recommends 9 to 11 samples in adults and five in children for the therapeutic monitoring of FVIII, with a washout period of 72 hours between subsequent doses.¹⁴ That is not only logistically complicated but also expensive. Following this kind of monitoring procedure for many treated patients is often impractical.

The conjoint use of pharmacokinetic models and Bayesian estimates is promising since it allows the study of FVIII's PK with only a limited number of samples. The first studies that utilized PK models with hemophilia A patients showed a strong correlation between the limited sampling and the PK estimates.¹⁵

While these studies have shown the efficacy of the PK models and Bayesian estimates in enhancing the treatment and prognosis of these patients, their application in the daily clinical environment still needs user-friendly tools that are easy for doctors to use.

It should be noted that hospitals and medical offices typically do not have access to qualified personnel for the use of complex mathematical models or powerful computer tools. Consequently, the availability of adequately validated and user-friendly devices for applying PopPK models for factor VIII that provide reliable results for individualizing and adjusting doses is a significant advancement, such as MyPKFiT^®.

Translational Medicine: MyPKFiT^® as a Tool for Dose Personalization

One of science's biggest challenges is transferring laboratory knowledge to hospitals and medical offices, where patients can benefit from the latest advances. Translational medicine aims to integrate different disciplines, resources, experiences, and techniques to promote improvements in prevention, diagnosis, and therapy.¹⁶ It encompasses all areas involved in developing and implementing medicine to enhance patient care in every possible aspect.

In the context of PopPK models, it is necessary to establish a connection that bridges the gap between the mathematical baggage and the prescribing doctor. The aim is to provide a simple, automated, and friendly tool that enables doctors to use models in their daily practice without disrupting the consultation process. Furthermore, involving patients in using this innovative strategy is highly favorable. Adopting innovative technological devices can empower patients to become active participants in their treatment, help them understand and engage with their therapy, and increase their satisfaction with the medical attention they receive.¹⁷

Therefore, to fulfill this need in PK, software was developed that simplifies and streamlines the use of PopPK models for doctors, providing a practical solution that can be easily integrated into their professional practice. By using this tool, doctors can individualize and adjust doses more precisely and personally, improving patient treatment outcomes.

When treating hemophilia A, the MyPKFiT^® software has proven to be a valuable tool for implementing factor VIII mathematical models into patient care. Developed by Shire in Massachusetts, USA, MyPKFiT^® is a user-friendly software that any doctor can access with an internet connection. It operates online and contains preloaded PopPK models of factor VIII.

Doctors can enter patient data and clinical information through a simple interface, and the software uses the PopPK model and Bayes' theorem to calculate and present the estimated PK profile. That allows doctors and patients to simulate different administration regimes and choose the most suitable one to ensure the patient can decide which is the most convenient regimen so that the patient is always within the therapeutic window.

By simplifying these models, MyPKFiT^® empowers doctors to make informed decisions about patient care without disrupting their consultation dynamics. Additionally, involving patients in the decision-making process and allowing them to see favorable results from their treatment can improve patient satisfaction and adherence to treatment.

In addition, MyPKFiT^® also offers a patient application that keeps the patient informed and involved in their pharmacological treatment. The PopPK model calculations provide the patient with information about their treatment, and the software includes a diary to record adverse effects or bleeding. The patient can efficiently share this information with their treating doctor.

The recent MyPKFit implementation has yielded interesting results in clinical practice. Studies have demonstrated that using tools like MyPKFiT^® significantly optimizes FVIII dosing.

A comparative study conducted in 2018 observed that PK model-guided dosing with MyPKFiT^® significantly reduced the annual bleeding rate and joint bleeding rate compared to traditional dosing schemes.¹⁷ In the same study, estimations based on PopPK models reduced annual FVIII consumption in 18 patients, while 14 required a dose increase.¹⁷ These results emphasize that a fixed dose scheme is unsuitable in FVIII and highlight the usefulness of MyPKFiT in maintaining an adequate treatment scheme on individual PK parameters for each patient.¹⁸

In another recent study, the PopPK model using MyPKFiT suggested reconsidering the prophylactic regime in 20 out of 39 participants, with five patients recommended to switch to the extended half-life drug. Following the MyPKFiT^® recommendations, a reduction in the annual bleeding and joint bleeding rates was observed.¹⁹ These results emphasize the importance of individualized treatment and the effectiveness of MyPKFiT^® in achieving this.

Furthermore, in 2023, Bao-Lai Hua et al. published a real-world study where the PK data of patients assessed by extensive sampling was compared with the PK data of patients with sparse sampling and MyPKFit. The results demonstrated that myPKFit can provide adequate dose estimates to maintain FVIII level above the target threshold at a steady state in Chinese patients.²⁰

Analogously, Rakmanotham et al. demonstrated in 15 patients that by using low-dose prophylaxis with PK-based adjustment (myPKFit) for six months, the annualized bleeding rates were reduced by a mean difference of -11.1 and a p-value of < 0.001.²¹

PopPK models involve the mathematical description of drug behavior in a population based on individual patient data. To estimate the drug's PK parameters, these models consider patient-specific factors, such as age, weight, genetic variations, and disease characteristics. By doing this, PK models can help to optimize dosing strategies, individualize treatments, and predict factor VIII levels.²²

As PopPK models rely on patient-specific data, one of their biggest strengths is that, unlike clinical trials where the patients are carefully selected, PK models incorporate "real-world evidence" that enhances the accuracy and robustness of the data. Using real-world data can provide insights into the influence of additional factors not initially captured in clinical trials, thus improving drug safety and effectiveness even when drugs that present substantial variability are used.

In the context of factor VIII replacement therapy, PopPK models have proven invaluable in optimizing dosing strategies for patients with hemophilia A. These models have allowed researchers and clinicians to analyze large FVIII PK data datasets and identify the drug's average PK parameters, considering patient-specific factors. By incorporating these factors into dosing calculations, PopPK models have significantly improved therapeutic outcomes, leading to better control of bleeding episodes also reducing the risk of drug-related adverse effects; thus, providing an enhanced quality of life for individuals with hemophilia A.

One of the main challenges to using the PopPK model in clinical practice is that the clinician needs an easy-to-understand tool that facilitates patient information input with a friendly user interface (UI). Tools like MyPKFiT^® at a doctor's office supply the clinician and the patient with tools; that was unimaginable in the past. The software can help doctors to personalize a patient's treatment based on computational simulations with high reliability, which was unthinkable a few years ago.

We are shifting from the traditional "one dose fits everyone" approach to precision medicine. This approach aims to supply" the proper medication for the right patient, at the correct dose, at the right moment"²², essential for narrow therapeutic window drugs or treatments requiring quick and stable therapeutic effects.

In that case, the usefulness of PopPK models and Bayesian estimates is indisputable. It is essential to note that PopPK models are already part of clinical studies for recently developed drugs. This knowledge can be translated to clinical practice by tools such as MyPKFiT.

Implementing precision medicine strategies optimizes drug administration in cases where the "one same dose for everybody" approach is not accurate. This latter has several advantages for health personnel, health systems, and, most importantly, patients. It eliminates the trial-and-error method of finding a suitable dose, improving patient quality of life.

The benefits of precision medicine are particularly evident in the case of hemophilia A, where optimal treatment reduces the bleeding rate and prevents the development of more severe symptoms like arthropathy.

Available data indicate that personalized and precise treatment significantly improves patients' quality of life. Furthermore, MyPKFiT includes an application that keeps patients informed and involved in pharmacological treatment. That allows them to access information about their treatment and register adverse effects or bleeding, and share them efficiently with their treating doctor. The software has shown usefulness for maintaining an adequate treatment scheme based on individual PK parameters for each patient.

However, while PopPK models have shown great promise in optimizing FVIII therapy, several challenges remain. The complexity of FVIII metabolism, the interplay between genetic variations, and the need for timely and accurate data collection pose significant hurdles. Efforts between clinicians, researchers, and the pharmaceutical industry are needed to standardize data collection protocols and expand data-sharing initiatives to create comprehensive and representative real-world datasets.

Looking ahead, the availability of friendly software that allows clinicians to input the patient's data and integrate advanced technologies such as machine learning and artificial intelligence can further enhance the utility of PopPK models in FVIII and other therapies.²³ By leveraging big data analytics, predictive algorithms, and real-time monitoring, clinicians can make more informed treatment decisions, reduce dosing errors, and improve patient outcomes.