Reply

S. Daya

Department of Obstetrics and Gynecology, Mcmaster University, 1200 main Street West, hamilton, Ontario, Canada L8N 3Z5. E-mail: dayas@mcmaster.ca

Thank you for the opportunity to respond to the letter sent by Dr Dickey (Dickey, 2003Go) in which he included comments on my paper on the pitfalls in the design and analysis of efficacy trials in subfertility (Daya, 2003Go). I have several points to make.

First, it appears that Dr Dickey is confusing the analytical and inferential phases of clinical trials. It is well known that the purpose of a clinical trial on efficacy evaluation is to estimate the true (but unknown) effect of the experimental intervention when compared with the control intervention. In this context, the larger the sample size the more precisely will the researchers be able to obtain this estimate of truth. The analytical part of the trial is directed towards determining how likely (or not) the estimate observed in the trial is the result of a chance finding; the lower the likelihood (probability), the lower the play of chance and the higher the likelihood that what was observed represents a true finding. Next, the magnitude of the effect of the intervention is addressed. Owing to sampling variability, the size of the treatment effect will vary considerably. Hence, larger trials, performed repeatedly, are more likely to provide a robust estimate of the size of the treatment effect. These issues of statistical significance and size of treatment effect are addressed by the analytical step in the clinical trial.

In the inferential step, the researcher tries to make sense of the results as they apply to the population from which the sample was derived, i.e. what do the results of the study indicate about the effect of treatment in the population? What conclusions can one make about the value of the experimental intervention in the patients, identified by the inclusion criteria. Thus, statistical significance (an analytical aspect) has to be differentiated from clinical significance or relevance (an inferential aspect) of efficacy evaluation by clinical trials. The former relies on strict methodological criteria, whereas the latter requires individual judgment. These two aspects are not mutually exclusive, as Dr Dickey implies, but rather complementary.

The second point of apparent misunderstanding by Dr Dickey is the concept of ‘unit of analysis’. In his letter, Dr Dickey uses the term ‘unit of analysis’ to mean outcome measure (e.g. live birth, pregnancy and so on), rather than the unit (or subject) that was randomized, which is the accepted definition. The experimental intervention is administered to the subject (selected randomly so that bias can be minimized) and compared with the control intervention administered to another subject (also selected randomly). The outcomes are then compared between the two groups of subjects (i.e. the unit of analysis is the subject). Thus, in ART trials of two interventions, it is the subject (i.e. female patient) who is being randomized and so the outcome must pertain to her (as is the case with pregnancy, live birth and even number of oocytes retrieved). The use of implantation rate as an outcome measure is inappropriate because the denominator is no longer the total number of subjects (the unit of analysis) but the total number of embryos transferred. However, because the embryos were not randomly allocated they cannot be used as the unit of analysis.

The third point, also related to the unit of analysis, pertains to the sample size that is calculated on the basis of assumptions made about the expected event rates (such as pregnancy rates) in the experimental and control groups. Whenever the sample studied is much larger than required, there is a higher likelihood that smaller differences in outcome events between the two groups will become statistically significant. Thus, switching the focus away from pregnancy rate to implantation rate, the denominator chosen is no longer the subject but the number of embryos transferred, a number that is much larger than the number of subjects studied. It is not surprising, therefore, that one finds significant differences in implantation rates (embryos as denominator) even though pregnancy rates (subject as denominator) are not significantly different. This fact, no doubt, is the reason why many investigators include this outcome when comparing treatments in ART cycles.

The fourth point pertains to the outcome measures that should be used. Depending on the objective of the trial, as identified by the research question, the investigator will choose a primary outcome (on which the sample size calculation is based) and one or more secondary outcomes. In the field of infertility treatment, because there is no consensus on the key outcome measures that should be utilized, several outcomes have been reported, including ovulation rate, pregnancy rate (both clinical and ongoing), live birth rate, numbers of oocytes and so on. The focus for most patients, investigators and other stakeholders is pregnancy, and until consensus can be reached on how this outcome should be reported, it may be prudent to provide data on all three relevant pregnancy outcomes viz clinical pregnancy, ongoing pregnancy and live birth rate (Daya, 2003Go). Such reporting also implies that data will be provided on numbers of miscarriage, ectopic pregnancy and stillbirth. The concern about the high rates of multiple pregnancy can be addressed, in part, by selecting singleton live birth as the outcome of choice (Daya, 2003Go). This initiative is supported by the increasing call for single embryo transfer in women undergoing treatment with ART and should become the standard for efficacy evaluation studies.

In summary, useful and valid inferences can be made only when trials are executed with high methodological quality and rigour, and analysed with the necessary attention to detail, as it pertains to the randomized subject who is the unit of analysis, using appropriate and clinically relevant outcome measures.


    References
 Top
 References
 
Daya, S. (2003) Pitfalls in the design and analysis of efficacy trials in subfertility. Hum. Reprod., 18, 1005–1009.[Free Full Text]

Dickey, R.P. (2003) Clinical as well as statistical knowledge is needed when determining how subfertility trials are analysed. Hum. Reprod., 18, 2495–2496.[Free Full Text]





This Article
Extract
FREE Full Text (PDF )
Alert me when this article is cited
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Request Permissions
Google Scholar
Articles by Daya, S.
PubMed
PubMed Citation
Articles by Daya, S.