DOI: 10.1213/00000539-199909000-00064

,  

,  

PMID: 10475338

Issn Print: 0003-2999

Publication Date: 1999/09/01


Print

The Importance of Dose-Response in Study Design

Steven L. Shafer

    loading  Checking for direct PDF access through Ovid

Excerpt

The recent manuscript by Varrassi et al. (1) compared two groups of subjects, one group received ketorolac 30 mg and the other group received propacetamol 2 g, each given at tracheal extubation and 6 h postextubation. The outcomes were indistinguishable. The authors concluded that “propacetamol and ketorolac, combined with patient-controlled analgesia morphine, show similar analgesic efficacy after gynecologic surgery.” They could just as well have said that the two drugs “show similar lack of analgesic efficacy,” a statement with very different connotations. The study does not tell us whether propacetamol is an effective analgesic.
We can learn from their example. Many things were done correctly here. The authors used a double-blinded, double-dummy, randomized parallel-group design. They chose a sensitive and well validated endpoint: morphine sparing. They had a large number of subjects (200), a consistent surgical operation (abdominal hysterectomy), and multiple centers. All of these are strengths in the study design.
Unfortunately, they compared a single regimen of propacetamol with a single regimen of ketorolac. Dose-response is the basis of pharmacology. All drugs have dose-response relationships, even if some are noisy or very steep. To make meaningful comparisons between two drugs, we must compare their dose-response relationships, ideally for both therapeutic and toxic effects. It is almost impossible to draw valid conclusions from a study that compares a single dose (or single regimen) of one drug with a single dose (or single regimen) of another drug. For example, let us assume that the authors found greater morphine sparing in the propacetamol group. Had this been the case, all we would know is that 2 g of propacetamol, given post extubation and again 6 h later, was better than 30 mg of ketorolac given at the same times. We would not know anything about the intrinsic safety and efficacy of propacetamol versus ketorolac. Maybe ketorolac is far superior, but must be given according to a different schedule.
The study by Varrassi et al. (1) found no difference between the groups. However, there was no placebo group, which would have provided the response to a 0 dose for both the propacetamol and ketorolac groups. Because the study lacked this simple control, the data cannot show whether propacetamol and ketorolac were analgesic. We do not know if there was any morphine sparing, because the study does not establish morphine usage in the absence of the study drugs. A perfectly valid interpretation of the data is that neither drug was an effective analgesic.
As a clinician, I believe the authors’ conclusions. Because I think ketorolac is a good analgesic, I interpret these results as showing that propacetamol is as good as ketorolac. However, this falls well short of scientific proof. The reason we do blinded, randomized, adequately controlled clinical trials is to support claims independent of the beliefs of the investigators or readers.
One can argue that adding placebo controls and multiple doses (or multiple dosing regimens) increases the study size and, hence, makes the study much more costly. However, failure to study dose-response relationships can produce an even more costly result: a multicenter study with 200 patients that cannot prove that the drug is an effective analgesic.
    loading  Loading Related Articles