Drug development for humans starts with pre-clinical studies (in which the drug is tested in animals), followed by clinical trials. Clinical trials are commonly classified into phases (I, II, III and IV). Phase I trials are conducted in a small group of subjects (usually 10’s to 100’s of people), who are typically healthy volunteers, except in some toxic drugs3 (often, but not always intended for cancer treatment). Mainly they aim at studying the drug’s safety and determination of a safe dose range. In some rare cases they are preceded by very small trials (typically 10 subjects), intended to help researchers in deciding whether a new drug should be tested in Phase I.
Since the ultimate goal of clinical research is to draw inferences from the findings of a clinical trial, the statistical reasoning (in addition to clinical reasoning) is crucial, as it allows forming reasonable conclusions from collected information. Therefore, the statistician plays an important role in almost every stage of clinical research4, from the planning stage up to and including the analysis and interpretation of the results.
A general introduction of how the followings questions are answered can be found here: General Introduction on Drug Development
1. What are the responsibilities of a statistician in this phase?
A statistician working in phase I clinical trials is responsible for determining the most suitable design and its analysis. This task includes sample size calculation, randomization list, modeling safety, determining the statistical approaches and, in general, managing all the statistical aspects for correctly organizing, executing and analyzing a phase I clinical trial.
2. With whom do statisticians collaborate in this phase?
In phase I clinical trials, the statistician mainly collaborates with the principal investigator, data managers, statistical programmers, sponsors/CROs, regulatory authorities and pharmacologists.
3. What are the major challenges emerging from these collaborations?
The statistician has to explain in a comprehensive manner the statistical aspects of his/her work. Moreover, the statistician has to ensure that the conclusions are properly drawn, putting them in the right context (e.g., link the conclusions to the sample size).
Apart from correctly communicating statistical aspects of the clinical trial, statisticians also collaborate to improve the quality of the clinical trial data.
4. What are major ethical challenges in this phase?
Ethical challenges are a major issue when it comes to all phases of trials and particularly when exposing individuals to a drug or a substance for the very first time. In this phase the main interest lies in evaluating the safety of a new therapeutic agent and, as a corollary, its toxicity. It goes without saying that the safety of participants should be ensured and all precautionary measures should be taken when administering the new drug. Overdosing, i.e. exposing individuals to very high doses can be very dangerous and in some cases can even lead to death. However, it should not be neglected that being safe can also mean administering sub-therapeutic doses, which can be equally hazardous in certain diseases, such as cancer.
5. What are the major statistical challenges in this phase?
Statistical challenges mainly associated with Phase I trials regard correct model specification, dose selection and sample size calculations. On the one hand, it is argued that identifying the optimal dose of a treatment can be difficult, since relatively few subjects are exposed to each dose level and time for observing any adverse events is quite short, after the experimental drug has been administered. What is more, sample sizes are typically selected based on budget and practicality and as a result they may not be large enough to allow firm conclusions to be reached. On the other hand, some respondents claim that there are not any major statistical challenges in Phase I trials and that they are quite straightforward.
6. Which innovative study designs are particularly important in this phase?
Experts from different fields support the view that adaptive and crossover designs are particularly important in Phase I clinical trials. Other methods that are listed as innovative in this area are the modified toxicity probability interval method, the continual reassessment method, Bayesian approaches and combining toxicity and efficacy endpoints.
7. Which statistical analyses are most important in this phase?
While there is a huge variety of developed statistical methods for Phase I studies, there are some of them that have been shown to be particularly important in practice. Among these, mixed effect models, non-linear and hierarchical models and ANOVA are particularly important. At the same time, statisticians from the pharmaceutical industry list graphical representation and descriptive statistics as an important step in the statistical analysis as well.
8. How often are Bayesian methods used?
Due to small sample sizes in Phase I clinical trials the correct target dose can be quite hard to identify. Thus, methods to increase accuracy are needed. While in the frequentist approach the prior knowledge about the drug is not formally taken into account, Bayesian approach allows the incorporation of this knowledge. Therefore, Bayesian methods seem to be a natural choice. Superior characteristics of these methods have been shown through simulations (e.g. for the continual reassessment method6 and escalation with overdose control). Despite this fact, Bayesian methods still seem to be quite rarely applied in practice and their application mainly depends on established practice in particular statistical groups in companies and research institutions. Interestingly, it appears that statisticians from academia tend to answer more often that Bayesian methods are used compared to ones from pharmaceutical companies. At the same time, researchers from academia mainly agree that Bayesian methods are crucially important and have quite natural place in Phase I clinical trials due to the limited number of subjects.
9. What are the statistical topics that are particularly “hot” in this phase?
There have been extensive research and publications done on Phase I dose-finding methods over the past few decades. The classical statistical approaches are generally believed to be enough for purposes of some pharmaceutical companies. However, there are still some new emerging topics in the area. Some of these are: dose-escalation models for combination therapies, prior elicitation and default prior investigation in the Bayesian framework. Although some of these questions are partly covered in the literature, proper implementation will still take some time.
10. Are new methods used regularly in practice?
In practice, newly published methods are not regularly used. The main reason for that is that investigators and regulatory authorities are reluctant to use new methods, as they often prefer proven standard techniques. Another reason is the lack of consensus among statisticians.
11. What is the connection between Phase I and Phase II studies?
Phase I and Phase II studies are connected in a sense that only drugs that pass the safety requirements in Phase I move to Phase II, and the dose used in proof of concept studies (Phase IIa) is taken from Phase I. Some experts, though, are of the opinion that the information about toxicity from Phase I trials is not sufficiently integrated when designing subsequent phases.
“This blog was written by Arsenio, Jose, Maria-Athina and Pavel”