Dose Response Studies

The dose response of a drug is important in pharmacology, pharmacokinetics, toxicology and clinical research. Dose response studies may be part of larger research to develop new treatments or to supplement existing knowledge of a drug whose benets may have already been established. The study should aim to address the following key questions. 1. Is there evidence of a drug effect? 2. What dose is required to produce a response different from the control response? 3. What is the nature of the dose response relationship? (eg. Does doubling the dose double the response?) 4. What is the optimal dose? (Is there a level at which a drug becomes harmful, or below which there is no benet or harm?) Parallel Designs Typically, patients are randomised to one of several active dose groups or placebo in the same way as with a RCT. However, if the efcacy of a particular treatment has been established it may be unethical to have a placebo group. Equally, a placebo group may be unnecessary if the placebo response rate is very low (such as spontaneous cure from cancer). Dose titration is often used, especially in early drug development, where the safety is of greater concern. Trial patients are started at low doses, which may be increased at intervals to achieve a response. It may require fewer patients to assess a drug effect since within-patient variability is used, but dose and time effects cannot be completely separated, which may obscure the real drug effect. Titration studies are most appropriate when the ultimate use of the drug will be in a titration scheme. Crossover Designs Crossover studies aim to reduce between-subject variability (heterogeneity) by exposing the study participants to both the drug and the placebo at different times. In this situation, the differences in response are more meaningful because each participant acts as his own control. In the example design, patients are allocated to one of four sequences, during which they are crossed over to different doses each week. Study participants are increasingly exposed to the study drug, making the design safer and more appealing to clinicians. The use of the placebo group in each period maintains the double blind nature of the trial and allows for the separation of treatment and time period effects.
Period Week 1 Week 2 Week 3 Week 4 Group 1 Placebo 10mg 20mg 40mg Group 2 10mg Placebo 20mg 40mg Group 3 10mg 20mg Placebo 40mg Group 4 10mg 20mg 40mg Placebo 75% lowest dose 50% middle dose 25% highest dose 75% highest dose

Table 1. Dose escalation crossover study

Factorial Designs Factorial designs are useful for studying two factors simultaneously to nd optimal dosage regimens, or studying the effects of two drugs to be used in combination. Examples of possible designs below. Equal numbers are usually allocated to each group, but the statistical power may be improved by increasing the size of the placebo and highest dose-by-frequency group. Because of the higher number of groups, factorial designs require larger samples.
Dose Route Oral IV Placebo 10mg 20mg 40mg Dose 0 10mg 20 mg Tables 2&3. Factorial designs for comparing two variables in a dose response study. 0 Duration of IV infusion 15 m 30 m 60 m

Key issues with dose response studies Often studies aim to identify the minimum effective dose for a drug. A particular dose may produce a statistically signicant (p<0.05) difference from placebo. This does not necessarily translate to a clinically important response. For a minimum effective dose study, it is essential that there is a placebo group. It is considered wise to choose doses that are equally spaced on an ordinal (1,2,3,4) or log (1n,2n,3n,4n) scale and to choose a dose that is suspected to be ineffective. This will make it easier to generate a dose response curve. Sample size calculations may be more complex because of multiple groups, but large samples may not be as important if the effectiveness of the drug has already been well demonstrated. (See Statistics support 2010-3-30 for more information on sample size.) Dose response studies generally assume that the drug is effective, so they use a one-sided (or onetailed) hypothesis, which doesnt require as large a sample size. (Two tailed means that you are testing whether the drug is better or worse than placebo, or two possible reasons to reject your null hypothesis). Two sided tests require a larger sample size because they are testing two hypotheses rather than one. Because of the different pharmacokinetics in individual patients, some trials propose monitoring blood concentration rather than doses. There are inherent challenges in doing this accurately.

This information was largely gleaned from the following article: Dose response studies I. Some Design Considerations. Stephen J Ruburg. Journal of Biopharmaceutical Statistics, 5(1) 1-14 (1995).