First-in-Human (FIH) studies are designed and conducted based on nonclinical data submitted to and reviewed by regulatory authorities. For the very first trial in humans, nonclinical toxicology, pharmacology, pharmacokinetics, in vitro assays, etc. are conducted with the investigational compound and are used (among other purposes) to establish a “maximum recommended starting dose” (MRSD). There are two categories of approaches that can be used to help choose this dose: model-independent approaches and model-based approaches.
Model-independent approaches use nonclinical data either with or without additional in vitro data to determine the appropriate dose in humans by using allometric scaling. Allometry, in its simplest form, uses pharmacokinetic (PK) parameters which are scaled as a function of body mass. Allometry uses relatively straightforward techniques that do not require sophisticated mathematical models. Model-independent approaches estimate a human equivalent dose (HED) using some measure of drug effect in pre-clinical species. Three commonly used measures for estimating the HED are the:
- No adverse effect level (NOAEL)
- Minimal anticipated biological effect level (MABEL)
- Pharmacologically active dose (PAD)
To estimate the MRSD using the NOAEL approach, the dose level with no adverse effects is identified from in vivo studies and converted to the HED using allometric scaling with the addition of a safety factor. Similarly, the PAD can be identified from appropriate pharmacodynamic (PD) models. This dose can then be converted to the HED using allometric scaling. In the MABEL approach, the lowest dose level with activity is identified by incorporating additional in vitro or in vivo pharmacology data and then converted to the HED using allometric scaling and a safety factor. All three methods are relatively easy to calculate but rely heavily on allometric scaling to account for differences in species.
Model-based approaches utilize models constructed with nonclinical data so that predictions can be extrapolated across species. PK models and physiologically based pharmacokinetic (PBPK) models are two examples of models that can be used to make FIH dose predictions. PK models are extrapolated across species by using similar allometric scaling exponents applied to key PK parameters such as volume of distribution and clearance. PBPK models, on the other hand, are built using detailed descriptions of physiology and are extrapolated across species by using species-specific descriptions of physiological differences in blood flows and organ volumes (among other things). Both methods allow for human physiology to be directly incorporated into the model, but can vary by complexity, costs, and time.
Key Considerations to Determine FIH Dose
Determining the FIH dose for clinical studies is critical. Knowing how to select a safe starting dose and knowing what factors need to be taken into consideration is extremely important. Here we cover six key considerations related to determining the FIH dose for clinical studies.
- Accounting for uncertainty in the initial dose prediction
- Observing a weak dose-response relationship in the initial FIH study data
- Determining FIH dose for biologics compared to small molecules
- Validating model-based approaches for a FIH dose in a novel molecule
- Determining the best method if allometric scaling is not appropriate
- Incorporating multiple activity metrics into the FIH dose prediction
Accounting for Uncertainty
Some uncertainty in a FIH study is to be expected. Adaptability and flexibility in both study design and dosing regimen can be important for success. Once the initial FIH dose is determined (using the method of choice) a 10-fold safety factor is typically added to account for uncertainty. It is also not uncommon to add an additional safety factor especially if the molecule is first in its class or has a novel mechanism of action. To get an even more robust picture of the uncertainty that might be present, it is worthwhile to estimate the FIH dose using multiple methods. The model-independent approaches are a great place to start since they are simple to calculate and are also endorsed by the FDA. Model-based approaches are the next type of method to consider as they use more detailed information for the FIH dose prediction.
Observing a Weak Dose-Response Relationship
Establishing both safety and efficacy levels is essential when determining the MRSD. If there does not appear to be a strong dose-response relationship among the active doses in pre-clinical studies this may mean that the selected doses are on the higher end of the exposure-response relationship. Exposure-response models typically take the form of a sigmoidal curve which has saturation at very high doses and very low doses. For doses that are too high, there may appear to be little to no change in the response given increasingly higher doses. Depending on the margin between the NOAEL and efficacious dose, it may be important to consider exploring lower doses in additional pre-clinical studies before predicting the FIH dose. Another source for a weak dose-response relationship could be uncertainty in the response. There may be a relationship present but it could be potentially obscured by patient-specific covariates. In this case, a model-independent approach is less reliable because underlying mechanisms are not taken into account. To address this, it may be beneficial to consider building a model that can explain the observed variability.
Biologics Compared to Small Molecules
Although biologics (such as monoclonal antibodies) and small molecules are different, the general principles for determining a FIH dose are similar. Additional factors such as complexity of the physiological environment, target-mediated drug disposition, and complexity relating to the target (i.e., membrane-bound, soluble, etc.) need to be taken into consideration for biologics. In vitro PD data, such as receptor occupancy and target binding can prove useful in determining the FIH dose for biologics.
Validating Model-Based Approaches in a Novel Molecule
The most common approach for model validation is analyzing diagnostic plots to make sure that the model can reproduce expected behavior. However, it is challenging to validate a model being used to estimate a FIH dose without having the necessary human PK data for comparison. In particular, these models tend to be fit-for-purpose and only include aspects relevant to the drug. To counteract the uncertainty that comes with using a fit-for-purpose model before human PK data has been collected, it may be beneficial to consider a variety of models to see if predictions change. If there does not appear to be agreement between models, the most conservative approach would be to start with the lowest predicted dose.
Best Method if Allometric Scaling is Not Appropriate
If allometric scaling is not appropriate, this assumes that exposure and toxicity of the drug do not correlate well to a measure of body size (i.e., surface area or body mass). There are many reasons why this may be the case. For example, since the expression of transporters and metabolizing enzymes do not scale with body size, if either one of these pathways heavily influences the PK of a drug, allometric scaling may provide misleading dose projections. Since all model-independent approaches rely on allometry for calculating the HED, they are unlikely to provide a reliable FIH dose prediction in these scenarios. Model-based approaches such as PBPK modeling may be more useful as they can account for species differences directly. Although the model-independent approaches may not be appropriate in determining the MRSD, the HED can still provide additional estimates that can reduce uncertainty in the final FIH dose.
Incorporating Multiple Activity Metrics
Model-independent approaches (such as PAD and MABEL) and model-based approaches can all incorporate activity information when projecting a FIH dose. It is recommended to predict a FIH dose for each relevant activity metric separately. The MRSD could then be established by choosing either the lowest projected dose or a dose based on activity metrics most representative of the drug’s mode of action from both a safety and efficacy perspective. Alternatively, if all predicted doses are similar then there is a higher degree of confidence in the dose projection, and this could potentially mean that a safety factor beyond the traditional 10-fold is not necessary. Take care not to rely solely on the metrics that provide the best safety profile of the drug. It is important to consider the results from all available metrics of activity.
FIH dose prediction is a challenging endeavor that has inherent uncertainty. The best way to counteract this uncertainty is by choosing the best method for the compound of interest and, more often than not, this involves combining methods from both model-independent approaches and model-based approaches. Understanding the strength of each method as well as the particularities of the compound is essential to success in determining a safe and effective FIH dose in a Phase 1 clinical study. At Allucent, we have the experience and expertise to help you design and execute an effective FIH study including determining the optimal FIH dose. Additionally, our team of pharmacometricians has experience successfully using modeling and simulation to determine FIH dose. Contact us to learn more about FIH study design and FIH dose selection services.