S.F. Marshall, H.L. Elliott, P.A. Meredith.
The University Department of Medicine and Therapeutics, The Western Infirmary, Glasgow.
There are many examples of drugs for which clinical exposure leads to a preference for doses which are lower than those recommended when the drug is first marketed. This can be attributed to a number of factors but it appears increasingly likely that the reliance on clinical trials which use parallel dose designs contribute significantly to the problem in that they provide only poor information about the clinical useful range of the dose response curves. Sambol and Sheiner(l) have previously suggested that dose escalation in conjunction with population analysis provides more relevant information than the curve derived from the average response at each dose. The population approach has been evaluated in this study to define the dose response relationship for the HMG CoA inhibitor simvastatin with respect to cholesterol lowering and the effect on other lipoprotein sub-fractions.
Data were obtained from 41 hypertensive hypercholesterolaemic patients (10M:31f,57±6 yrs) who received placebo,10,20 and 40 mg simvastatin in random order each for 12 week treatment periods with measurement of serum total cholesterol and major lipid subfraction at 12 weekly intervals. Population analysis was undertaken using first order estimation with NONMEM. To determine if there was a graded response, a step model was compared to a step-linear model. The former assumes that the maximum response occurs at any given dose. The latter also models this response and has the additional term to describe the subsequent change in response with increasing dose. The more appropriate model was selected on the basis of a likelihood ratio test comparing the goodness of fits between the full and reduced models. When a graded response could be established as more appropriate, an Emax model was subsequently fitted to the data. The most appropriate error model was determined for intra and inter patient variability.
The step-linear model generally gave a superior fit over the step-model (except for triglycerides) and the Emax values for cholesterol and LDL were respectively
-2.74±0.19 mmol/1 and -2.60±0.18 mmol/1, while the corresponding D50 values were 5.05±0.85mg and 6.25±1.36mg. On the basis of the population response, 40mg simvastatin elicits 89% and 86% of the maximal expected total cholesterol and LDL responses respectively . A proportion error model for intrasubject variability gave the best fit for all models and modelling intersubject variability as either additive of proportional error did not significantly alter the objective function.
Using a population based approach it has proved possible to characterise the dose-response relationship for simvastatin with respect to cholesterol and LDL. It is apparent that in population terms 40mg simvastatin elicits less than 90% of the theoretical maximal. In contrast, based on this analysis 50% of the maximal response would be elicited by a dose significantly less than the lowest dose of 10mg used in this study. On the basis of this analysis it is probable that for many patients a satisfactory lipid lowering response would be obtained by a dose of 5mg simvastatin, a formulated dose which is not currently available in the united kingdom.
[1] Study designs for dose-ranging. Sheiner L.B. et al. J. Clin. Pharmacol. Ther. vol 46, 63-77,1989.
Reference: PAGE 3 () Abstr 858 [www.page-meeting.org/?abstract=858]
Poster: poster