Optimal dose de-escalation trial designs for novel contraceptives in women
Introduction
One of the key steps in any pharmaceutical drug development program is the elucidation of the dose–response relationship. In dose finding trials the optimal dose has to be determined. On the one hand, choosing a dose too low might result in a failure in the subsequent phase III clinical trials due to inefficacy. On the other hand, choosing a dose too high might lead to unnecessary safety problems during subsequent drug development or even after the drug has being marketed; see, e.g. Bretz et al. (2008) for a comprehensive overview.
Dose finding for classical hormonal contraceptives is usually done by investigating ovulation inhibition in healthy female volunteers; see, e.g. Heger-Mahn et al. (2004). Ovulation inhibition is a broadly accepted surrogate endpoint since ovulation is an absolute precondition for conception. In addition ovulation inhibition can be rapidly and reliably assessed within two cycles of pill intake.
For novel non-ovulation-inhibition contraceptive approaches, e.g. methods impacting endometrial, cervical, or male reproductive parameters, such an approach is not feasible. These new strategies do not necessarily inhibit ovulation and yet no other surrogate endpoint for contraceptive efficacy is available. Similar to phase III clinical trials on contraceptives, the only way to assess the efficacy of such a product is the direct measurement of the contraceptive efficacy, by the Pearl Index (PI), i.e., the number of pregnancies per 100 women-years; see Gerlinger et al. (2003). The PI is the ultimate clinically relevant endpoint and in contrast to the surrogate parameter ovulation inhibition, it encompasses also all non-ovarian contraceptive effects, e.g. effects on cervical mucus sperm permeability, endometrial receptivity, fallopian tube function and all male reproductive factors. Phase III trials typically assess the contraceptive efficacy over a treatment period of one year.
A classical parallel group dose–response trial investing several doses including at least one ineffective dose would lead to a number of contraceptive failures, i.e., unplanned pregnancies, especially in women randomized to an ineffective dose. Such a dose–response trial design does not seem to be practicable nor does it seem acceptable from an ethical point of view. The strategy to identify the maximum tolerated dose (MTD) used frequently in early drug development, e.g., in oncology, is used as the basis for the concept. While the MTD is found by applying increasing doses, starting from a well-tolerated dose up to a point of intolerance, the approach presented here applies decreasing doses, starting from a presumed effective dose down to an ineffective dose. Thus, the least-effective dose (LED) is zeroed in. A trial design that minimizes the (absolute) number of unwanted pregnancies is presented. However, given the limited extend of our simulation study, further research is warranted.
Section snippets
Methods
In this section, the selection of dose-(de)escalation designs is described and discussed. In the oncology setting of dose escalation designs with the target to find the maximum tolerated dose (MTD), defined as a with chosen as highest acceptable probability for the event of interest (here: toxicity), a variety of designs have been proposed; see for example Ahn (1998), Gerke and Siedentop (2008), or Chevret (2006). The underlying dose-toxicity model assumes non-decreasing probabilities
Dose–response model and simulations
Each subject can be assigned to 1 out of 10 dose levels. The dose decrements follow a constant ratio of . The relationship between the dose and the probability of a success at dose level , is given by logistic functions of the form
The probabilities of success of each dose level comply with values for , given as (0.3901, 2.2488), (2.2835, 1.7856), and (2.8201, 2.4681) in scenarios 1, 2, and 3, respectively (see Table 1). The scenarios 1, 2,
Results
The primary objective of this simulation study is to assess properties of the aTDR and the eTDR designs and to compare them to CRML. Therefore, percentages are given of how often the true LED was determined and how often the recommendation deviated from the true LED. Furthermore, the average number of subjects necessary per run and the average number of pregnancies that occurred per run are displayed. In addition to the highest dose as starting dose (i.e. aTDR), the effect of more courageous
Discussion
A phase II dose-finding trial design for contraceptive approaches that minimizes the number of unwanted pregnancies is presented. The classical dose-escalation design of early drug-development for assessing tolerance and safety was used as a basis for a dose-de-escalation approach to assess efficacy. Of the seven methods initially investigated, only the aTDR, the eTDR and the CRML methods were suitable for our problem and could be included in our simulations. CRML was also implemented with
Conclusion
Dose-finding for contraceptives that do not consistently inhibit ovulation by counting the unintended pregnancies becomes feasible if an adaptive approach is taken instead of the classical parallel group design. Both CRML and eTDR are dose-finding designs that substantially reduce the expected number of pregnancies to less than 4 compared to 16.9 in the classical dose-finding design with a similar total sample size. However, this clear advantage comes at the price of a 5-fold increase in trial
References (23)
- et al.
Group up-and-down designs for dose-finding
Journal of Statistical Planning and Inference
(2006) - et al.
np1: a computer program for dose escalation strategies in phase I clinical trials
Computer Methods and Programs in Biomedicine
(2007) An evaluation of phase I cancer clinical trial designs
Statistics in Medicine
(1998)- et al.
Monitoring late-onset toxicities in phase I trials using predicted risks
Biostatistics
(2008) - et al.
Effective dose of nefopam in 80% of patients (ED80): a study using the continual reassessment method
British Journal of Clinical Pharmacology
(2007) - et al.
7: dose finding—a challenge in statistics
Biometrical Journal
(2008) Study design principles in the clinical evaluation of new drugs as developed by the chemotherapy programme of the National Cancer Institute
- et al.
Sequential designs for phase I clinical trials with late-onset toxicities
Biometrics
(2000) Statistical Methods for Dose-finding Experiments
(2006)- et al.
A random walk rule for phase I clinical trials
Biometrics
(1997)