This is Part I of our two-part exposure-response blog series. This post will give a brief introduction to the Food and Drug Administration’s current guidance on exposure-response relationships and will provide some points to consider when designing these types of studies.
If a drug has any effect, at any dose (compared to placebo, a dose of 0), there is an exposure-response relationship. In fact, there are usually at least two exposure-response relationships, one for efficacy and one for toxicity. These relationships are critical to understanding how to use a drug most effectively, and therefore are important for developing the drug label (i.e., prescribing information). Quantifying these relationships for a new or modified drug intended for marketing is one of the purposes of exposure-response studies and is addressed in the FDA Guidance for Industry, “Exposure-Response Relationships — Study Design, Data Analysis, and Regulatory Applications.”
It should be noted that the FDA uses the term “exposure-response relationship” in the guidance document, not “dose-response relationship.” A dose-response is one type of exposure-response (dose being one way to measure exposure to a drug), and this can be a useful relationship to understand. But, as pharmacokineticists, we often find that dose is not the best predictor of response. Consider the problem of pediatric dosing: Giving a 100 mg dose to a 4 kg neonate would be expected to have a much different, and usually larger, effect than a 100 mg dose given to a 90 kg adult. Intrinsic in this principle is that it really doesn’t matter how big the pill is, but rather what the concentration of drug is at some effect site, and in turn, usually how much drug becomes bound to some receptor or enzyme at the effect site.
Therefore, dose is often a less than ideal measure of drug exposure. Indeed, the closer you can physically get to the effect site in your measure of exposure, the more predictive of response that measure is likely to be. Along the same lines, each step that the measurement is removed from the physical effect site will introduce some amount of variability in the relationship. This means that the ideal approach would be to measure drug concentration right at the effect site, but unfortunately, that usually isn’t feasible. As such, the goal becomes to get as close as practical to the effect site, which in most cases means analyzing the plasma. Often, we base exposure-response on plasma concentration data – a good balance between pharmacokinetic variability (e.g., between a 4 kg neonate and a 90 kg adult) and the practical barriers to measuring exposure at the effect site.
Examining Exposure-Response Relationships
The FDA guidance outlines two study design approaches for examining exposure-response relationships – randomly assigned doses and randomly assigned concentrations.
The more practical and commonly used approach is to randomly assign doses, although a key disadvantage of these studies is that they can be misleading unless care and proper analysis are employed. Consider, for example, that renal failure might simultaneously result in a higher plasma concentration of drug at a given dose due to decreased clearance and may also increase susceptibility to adverse effects. In this case, a natural, but incorrect conclusion might be that drug concentration is related to the observed adverse effects. In another example, a study titrating higher doses only to individuals who are less sensitive to the drug (e.g., nonresponders) might suggest a reduced response at higher doses (i.e., a bell-shaped dose-response or concentration-response curve), which would not be reflective of the true exposure-response relationship. These examples therefore represent potential sources of bias in dose-controlled studies. Importantly, even when data are appropriately analyzed, confounding of the concentration-response relationship may persist.
The potential for confounding due to unrecognized PK/response relationships or the random imbalance of influential factors in the way patients are selected to receive higher doses is not shared by concentration-controlled studies. In concentration-controlled studies, subjects are randomly assigned to predetermined levels of average plasma drug concentration. The target concentrations are achieved by an individualized PK controlled dosing scheme, with concentration-response relationships observed in the same individual over time (e.g., over a dosing interval). Such studies are therefore more defensible from a statistical standpoint and also remove some degree of variability, as discussed above. Despite this, these studies are not terribly common because the process is cumbersome and study results could have unanticipated labeling implications.
Study Design Considerations
There are a number of factors to consider when designing an exposure-response study. A few of these are highlighted in the following paragraphs.
First, the key to understanding exposure-response relationships, or any relationship, is to have an adequate domain of the exposure measure. The most practical way to get a wide domain is to study a broad range of doses in phase 1 and phase 2 studies. Indeed, failing to explore wide-enough dose ranges is a common misstep in early drug development, because a dose-response curve is difficult to draw without sufficient data points.
There is usually an expectation that most of the exposure-response curve, including exposures with little effect and those with near maximal effect, will be included in early phase studies. Including doses that are thought to be suboptimal can be a problem, either ethically in life-threatening conditions, or because of a perception that these are statistically “wasted patients” when it comes to early demonstrations of efficacy. The ethical issue is a real problem, but if a placebo arm is ethical in early phase studies, then a suboptimal dose arm will be, as well. The issue of statistical efficiency can best be addressed by rethinking how we demonstrate efficacy.
Demonstration of an exposure-response relationship is a valid statistical approach to proving efficacy, and in fact is explicitly endorsed in the FDA guidance, even for registration trials. In addition, if some of the variability can be removed (e.g., by using plasma concentration as the measure of exposure rather than the dose), then statistical power will usually be comparable to a simple T-test with a single dose vs. placebo, as well as being much more informative. The power analysis for such an exposure-response analysis is more complicated than for a simple T-test, however. Nuventra has experience using simulation methods to select an appropriate sample size to maximize statistical power. Learn more about Nuventra’s modeling and simulation services here.
There are several other key considerations when designing an exposure-response study. It is important, for example, to collect PK data in the same patients in which pharmacodynamic (PD) data are collected (when practical), which can help tease out exposure-response relationships. Ideally, at least some PK data should be collected in all patients, as this helps address the issue of PK variability. It also is important to collect accurate dosing data, usually for 3-4 half-lives prior to a clinical visit. At the clinical visit, it is helpful to collect a thorough, predose blood sample, have an accurate dosing time (observed at the clinic), and then take one or more blood samples after the observed dose. Remember, accurate dosing history is as important in pharmacokinetics as is the accurate recording of sampling times.
Finally, despite best efforts, it can sometimes be difficult to determine a concentration-response relationship. There can be several reasons for this, including: a limited range of exposures, variability among subjects, a lag time in effects, and a dose titration effect. But, be assured that if the drug is effective, then there is an exposure-response relationship, since a dose of zero has a different response than some non-zero dose.
The FDA’s exposure-response guidance states that “exposure-response information can support the primary evidence of safety and/or efficacy. In some circumstances, exposure-response information can provide important insights that can allow a better understanding of the clinical trial data (e.g., in explaining a marginal result on the basis of knowledge of systemic concentration-response relationships and achieved concentrations). Ideally, in such cases the explanation would be further tested, but in some cases this information could support approval.” In these words, FDA underscores the importance of understanding exposure-response relationships during drug development and also the key role that these insights can play in a successful marketing application.
Speak to a Senior Consultant today about designing your exposure-response study