Epilepsy Research 45 (2001) 101–103

Monotherapy comparative trials: equivalence and differencesin clinical trials

David Chadwick *Uni�ersity of Li�erpool & Walton Centre for Neurology & Neurosurgery, Lower Lane, Li�erpool, L9 7LJ, UK

Abstract

Investigators are familiar with the concept of clinical trial design to detect differences. Equivalence studies set outto define how similar treatments might be. They become important for two reasons: (1) similarities may be asclinically important as differences and (2) in therapeutic areas with drugs of proven efficacy already marketed, it maybe ethically difficult to perform placebo-control studies. If equivalence is rigorously defined (the minimal clinicallyimportant difference), they could represent a route to licensing (cancer chemotherapy and antibiotics). Equivalencewill be defined by the limits of 95% confidence intervals for a comparison being constrained within pre-defined limits.As such, studies need larger numbers of subjects rather than difference studies, and intention-to-treat analysis may nolonger be the more conservative approach. © 2001 Elsevier Science B.V. All rights reserved.

Keywords: Monotherapy trials; Comparative trials; Equivalence; Differences; Confidence intervals; Antiepileptic drugs

www.elsevier.com/locate/epilepsyres

1. Introduction

Whilst regulatory agencies may only feel confi-dent to attribute efficacy where the test treatmentshows a superiority with statistical significanceagainst a control group, in clinical practice equiv-alence for an important outcome may providevery useful information influencing choices oftherapy. If a new antiepileptic drug (AED) wasequivalent in efficacy to the standard treatmentbut carried only half the risk of withdrawal foradverse effects, we would probably wish to use thenew treatment in preference to the old. We would,however, want to be confident about equivalence

of efficacy. Equivalence is not, however, the sameas a failure to detect a difference, which is com-monly due to inadequate powering of a study.

2. Requirements

Clinical trials can be structured and powered toestablish equivalence for an outcome. The firstrequirement is that a clinical definition of equiva-lence is provided. This may be best understoodfrom pharmaceutical bioequivalence. In order togain a licence for a generic drug, bioequivalenceshould be demonstrated. This is usually acceptedwhen the bioavailability of the generic compoundis within �20% of the standard agent. Such widemargins may be unacceptable for clinical equiva-

* Tel.: +44-151-529546l; fax: +44-151-5295465.E-mail address: d.w.chadwick@liv.ac.uk (D. Chadwick).

0920-1211/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.

PII: S 0920 -1211 (01 )00228 -5

D. Chadwick / Epilepsy Research 45 (2001) 101–103102

Fig. 1. How big a difference must we exclude?

equivalence once we had excluded that the newtreatment was no worse by a pre-specified amountthan the standard, i.e. demonstrated non-inferior-ity. Clinical trials would fulfil these requirementsif the 95% confidence intervals of a comparisonbetween standard and new treatment wereconfined within either 0.9–1.1 (in the case of aratio) or plus or minus 10% in the case of adifference (Fig. 1), for a hazard ratio for time to a12 month remission of seizures. More practically,equivalence might be accepted if we can be confi-dent of excluding the new treatment as being nomore than 10% worse than the standard. Definingthe clinically important difference which has beenexcluded needs to be a source of debate for clini-cians and will be determined by the outcomes thatare used in an individual clinical trial.

It is important to recognise that confidencelimits around equivalence could be defined byeither a risk difference (where equivalence is 0) ora ratio (where equivalence is 1). There are impor-tant differences to be recognised here, because theexpected event rate for an outcome will alter therelationships between the magnitude of differenceand ratio for an outcome (see Fig. 2). While therecommendations of the Commission onAntiepileptic Drugs (1998) are for equivalence tobe constrained within 20%, this margin is almostcertainly too generous to be meaningful and does

lence. For mortality in a common condition, thetolerance limits might be even more tightly con-strained: for example, for a condition as commonas stroke, differences of below 5% may be impor-tant for comparison against a treatment such asaspirin, which is readily available and cheap. Inepilepsy, it might be reasonable to accept that twotreatments are clinically equivalent if we had ex-cluded a greater than 10% difference for an im-portant outcome, such as seizure-free rates.Practically if we are comparing a new and astandard treatment we might accept practical

Fig. 2. The effects of two different event rates (65 and 90%) on the magnitude of rate difference for non-inferiority (by permissionof W. Samermann).

D. Chadwick / Epilepsy Research 45 (2001) 101–103 103

not indicate whether reliance should be placed ondifference or a ratio.

Equivalence studies differ from conventionalstudies in number of important respects (Jones etal., 1996):� They are dependent on confidence intervals,

not point estimates of difference.� Intention-to-treat analysis is no longer conser-

vative–protocol correct populations should becompared.

� They require larger number of patients.

3. Discussion

Equivalence studies are well accepted in thearea of bioequivalence of pharmaceutical prepara-tions and will become more and more importantin neurology as effective treatments become avail-able in more therapeutic areas. Already in thefield of cancer chemotherapy, placebo-controlledmonotherapy studies cannot ethically be under-taken and the efficacy of new treatments may bedefined by equivalence studies. We are perhapsarriving at the situation in motorneurone diseasewhere a treatment, riluzole, has shown to have amodest effect on survival (Riviere et al., 1998),and in MS where �-interferon may benefit relaps-ing and secondary progressive disease (EuropeanStudy Group on Interferon beta-1b in Secondary

Progressive MS, 1998; PRISMS Study Group,1998). There needs to be serious thought given byregulatory agencies as to whether epilepsy is nowan area where all monotherapy trials need to beactively controlled with efficacy accepted whenthe criteria for rigorously defined equivalence hasbeen met (Chadwick and Privitera, 1999).

References

Chadwick, D., Privitera, M., 1999. Placebo-controlled studiesin neurology: where do they stop? Neurology 52, 682–685.

European Study Group on Interferon beta-1b in SecondaryProgressive MS, 1998. Placebo-controlled multicentre ran-domized trial of interferon beta-1b in the treatment ofdisability in secondary progressive multiple sclerosis.Lancet 352, 1491–1497.

ILAE Commission on Antiepileptic Drugs, 1998. Consider-ations on designing clinical trials to evaluate the place ofnew antiepileptic drugs in the treatment of newly diag-nosed and chronic patients with epilepsy. Epilepsia 39,799–803

Jones, B., Jarvis, P., Lewis, J.A., Ebbutt, A.F., 1996. Trials toassess equivalence: the importance of rigorous methods.Br. Med. J. 313, 36–39.

Riviere, M., Meininger, V., Zeisser, P., Munsat, T., 1998. Ananalysis of extended survival in patients with amyotrophiclateral sclerosis treated with riluzole. Arc. Neurol. 55,526–528.

PRISMS Study Group, 1998. Randomized double-blindplacebo-controlled study of interferon beta-1a in relapsing-remitting multiple sclerosis. Lancet 352:1498-1504.

.