Introduction
Several anti-epileptic drugs (AEDs) are currently available as alternative and competing interventions for epilepsy treatment. In the literature almost all information regarding new AEDs is obtained from a large amount of clinical trials based on a comparison of one AED with placebo, and on few studies comparing a new AED with an older one. Such an approach provides only a partial fragment of the whole picture: it may be well established that new AEDs are more effective and better tolerated than placebo, but real comparative values of both efficacy and safety of new AEDs as a whole are difficult to be evaluated. In other terms, knowing that a new AED is better than placebo may be useful, but ideally one should have the possibility to know how all the different options rank against each other and how big these differences are in effect size between all the available options.1
This commentary focuses on the designs of randomised controlled trials (RCTs) of new AEDs as treatment for focal epilepsy. Only RCTs studying drugs that are already marketed are considered. Limits of these trials, with particular focus on placebo-controlled designs, are discussed and strategies to overcoming them proposed.
Interpretation of epilepsy RCTs: which standpoint has been adopted?
When critically interpreting designs of epilepsy RCTs, it is important to consider which was the main standpoint upon which RCTs were conducted.
From the standpoint of the regulator, licensing decisions are based upon an overall assessment of both efficacy and tolerability. Regarding the former, a strong evidence to support efficacy of a new AED is required, provided a significant difference is found in favour of the new drug over placebo, and replicated in another study. Regarding tolerability, the regulator considers safety data obtained from placebo-controlled trials and from open label extension studies providing further patient years of exposure. In conclusion, licensing decisions are based upon an overall assessment of risk and benefit.
From the standpoint of the prescribing clinician, patient and policy makers, the results of RCTs should have a direct impact on pragmatic, daily clinical decisions: in real-life conditions, practice choices are made between drugs rather than between drug and placebo.
The choice of a RCT design is therefore influenced by the standpoint that has been adopted, so that an explanatory trial may be for example, preferred over a pragmatic design. Pragmatic trials asking whether a drug works in real conditions have therefore a study design which may allow to obtain answers that matter to both the prescribing clinician and the patient. On the other hand, explanatory trials asking whether a drug works under ideal conditions are useful for understanding questions of efficacy but are of limited value for telling us whether we should administer a drug to a wide variety of patients and/or in a wide variety of circumstances.
Clinical trials using placebo as comparator
New AEDs have shown to be superior to placebo for epilepsy treatment in RCTs. A recent study2 evaluated the geometry of networks representing the conclusions of a recent analysis with multiple comparisons between different treatments in several randomised trials;3 37 clinical trials with second-generation AEDs for the treatment of focal epilepsy were included. None of them performed head-to-head comparisons, but all used placebo as comparator.
In many of these trials the choice of placebo as comparator is debatable, and in some cases may even represent a treatment preference bias. Such a predilection towards a placebo-controlled comparison may be due to several factors: (i) when some of these trials were first designed, there was not yet evidence regarding efficacy for the studied drug; (ii) sponsor's choice; and (iii) regulatory requirements.2 Regarding the last aspect, it is noteworthy that the revision of paragraph 29 of the Declaration of Helsinki in 2000 suggested that the best possible treatment should be used as active comparator.4 Avoiding direct head-to-head comparisons is neither scientifically nor ethically acceptable, and a new AED should be tested against the AED previously considered most active, rather than against placebo or a clearly inferior agent. On the other hand, regarding regulatory aspects, most add-on studies are compared with placebo as usually they are drug company trials used for licensing purposes.
Placebo-controlled trials of AEDs are therefore difficult to justify and often leave the clinician unsatisfied: showing that an AED is better than 'nothing' only tells us that the intervention has effect, but it does not help to select optimal therapy.5
Properly conducted head-to-head comparisons, although unpopular with industry as there is a risk that such trials would fail to find a significant difference between treatment groups, should nevertheless be considered much more reliable and methodologically adequate than placebo-controlled trials or indirect comparisons. However, as currently designed, even head-to-head comparison trials have significant limitations for informing clinical decisions in terms of patient selection and choice of outcomes (e.g. too short follow-up periods).6 Moreover, as for trials of any type, limitations due to poor internal validity (i.e. methodologically poor designed studies) may exist, and one should be well aware of them, carefully reading and critically interpreting the results, following for example the comprehensive methodological procedure outlined by the Cochrane Collaboration and currently adopted in its systematic revisions.
RCTs comparing new AEDs with older AEDs
Regarding the treatment of focal epilepsy, to date there are few clinical trials comparing new AEDs with older ones, whereas studies comparing a new AED with another one are indeed very rare.
The main reason for that may be due to the fact that industries may be reluctant to conduct trials on new AEDs with the risk of failing to find a significant differences in terms of efficacy and tolerability between groups.
Head-to-head AED trials may be of limited utility where equivalence (or non-inferiority) is found, but they should nevertheless be considered as the most relevant source of data to inform clinical decisions.6
A recent meta-analysis of RCTs comparing lamotrigine and carbamazepine monotherapy showed that lamotrigine is significantly better tolerated than carbamazepine, although a nonsignificant trend suggests that carbamazepine is superior regarding time to remission and seizure freedom at 6 months.7 Results of this meta-analysis are partially limited by following aspects: (i) included studies enrolled also patients with generalised epilepsies, which may be exacerbated by carbamazepine (population choice bias/selection bias); (ii) all included trials were sponsored by the manufacturers of lamotrigine (sponsorship bias); and (iii) follow up and duration of included studies were too short to measure clinically important efficacy outcomes such as time to 12-month remission. Authors of this meta-analysis concluded therefore that current industry-sponsored trials fail to adequately inform clinical practice and that clinically relevant longer-term outcomes are required before concluding about the role of lamotrigine in the treatment of epilepsy.
The subsequent unblended pragmatic RCT Standard and New Antiepileptic Drugs (SANAD), to date the largest and probably the best-designed randomised trial ever conducted in epilepsy patients, compared carbamazepine with newer AEDs (gabapentin, lamotrigine, oxcarbazepine and topiramate).8 Despite some criticisms,9 SANAD trial had a very innovative pragmatic design, aimed to get a high external validity: SANAD assessed the policies of starting treatment with one of a number of drugs, but the protocol was not prescriptive and allowed clinicians to use the titration and maintenance doses they thought best for the patient. As a matter of fact, a prescriptive protocol would have diminished the pragmatic value of the trial and prevented clinicians from using their own judgement to gain the optimum effect from a treatment for the patient in front of them.
Multiple-treatment meta-analysis: a way of overcoming the limits of placebo-controlled RCTs?
Ideally, direct head-to-head comparisons of new alternative and competing AEDs should be available to get the whole picture of each treatment, but usually RCTs have not such a direct-comparison randomised design. In order to fill this gap, indirect comparisons have been conducted between AEDs that have not been compared directly with each other.
Such indirect comparisons may therefore produce a ranking efficacy of different drugs included in the analysis, thus allowing to answer some questions that are of paramount importance for daily practice and decision-making: what is the probability that each new AED is the best? Are there any alternatives, and if so, how do they rank?
Multiple-treatment meta-analysis (MTMA), also known as mixed-treatment comparison meta-analysis or network meta-analysis, is a promising quantitative method of collecting, integrating and representing data from all the available comparisons, synthesising information obtained from a network of trials.10 In other terms, MTMA, which actually should be considered as an extension of systematic reviews, may be useful in order to interpret the available evidence from networks of randomised trials, thus going beyond focusing on simple pairwise comparisons.
Comparisons may be either direct or indirect, the former occurring when treatments are compared within a randomised trial (trials of A vs. B), whereas the latter when treatments are compared between trials evaluating how effective they are compared with a common (or intermediate) comparator treatment (trials of A vs. C and of C vs. B). MTMA is a method that permits not only to combine direct and indirect evidence on the relative effectiveness of two interventions (A and B), but also to retain both the randomisation and the identity of each intervention. As a consequence, the main intrinsic limitation of these indirect comparisons is that dose and population equivalence are difficult to be reached.
A recent meta-analysis performing an indirect comparisons between second-generation AED for partial epilepsy treatment was conducted by Otoul et al. in 2005.3 This study evaluated efficacy and tolerability of add-on levetiracetam compared with other new AEDs (gabapentin, lamotrigine, oxcarbazepine, tiagabine, topiramate and zonisamide), performing adjusted indirect comparisons between levetiracetam and each AED. At the doses tested, levetiracetam was more effective than gabapentin and lamotrigine and equally well tolerated; it had a significantly lower withdrawal rate than topiramate and oxcarbazepine, with comparable efficacy. Although it did not differ significantly from the other AEDs, numerical trends favouring levetiracetam were obtained in response rate and in withdrawal rate (tiagabine, zonisamide). Authors concluded that add-on levetiracetam has a favourable responder and/or withdrawal rate compared with other new AEDs in patients with partial epilepsy with doses used in clinical trials. This study confirms the results of a previous meta-analysis with indirect comparisons between drug and placebo in which, considering efficacy and tolerability, levetiracetam come out best followed by zonisamide and oxcarbazepine.11
A subsequent multiple-treatment comparison study summarised all individual patient data from 20 epilepsy monotherapy randomised trials comparing at least two out of eight AEDs.12 Results favoured carbamazepine, lamotrigine and oxcarbazepine for partial onset seizures, and valproate or phenytoin for generalised onset tonic clonic seizures.
To date there is no MTMA evaluating data from epilepsy add-on therapy randomised trials.
Historical control design studies
Very recently the International League Against Epilepsy raised the possibility of using historical data from treated and control (untreated or undertreated) patients who have been enrolled in monotherapy trials in newly diagnosed epilepsy.
Performing a systematic review, French et al. advocated use of a virtual placebo historical control as an alternative to pseudoplacebo. This review showed that patients randomised to suboptimal treatments in all previous trials had similar outcomes, thus setting the conditions for the build-up of a dataset of historical controls against which response to investigational AEDs could be compared in future trials.13 To demonstrate efficacy with this study design, it would be sufficient to show that the exit rate in a group of patients converted to monotherapy with a full dosage of the AED of interest is significantly below the exit rates previously observed in historical controls.14 Despite some debates regarding the definition of 'significantly' and some methodological concerns related to conversion-to-monotherapy and to historical controls,14 this study design might have several advantages: all patients receive a promising AED at dose(s) that are expected to be fully effective. Moreover, using this kind of design there is no need for a parallel group on suboptimal treatment or on placebo, which represents an ethical problem, given the risk of lesions and mortality possibly associated with epileptic seizures. As a consequence, such a study design may reduce patient risk that would result from exposure to pseudoplacebo.13
Conclusions
An alarming lack of well-designed, properly conducted epilepsy RCTs exists, and most of those currently available have significant methodological problems limiting their applicability (external validity) to clinical relevant questions.15
In RCTs evaluating efficacy and safety of new AEDs the choice of placebo as comparator should be discouraged, in favour of an active comparator (head-to-head comparison), but always being aware that even within these designed studies biases may occur. Moreover, as currently designed, these head-to-head comparison trials may have significant limitations for informing clinical decisions in terms of patient selection and choice of inadequate or clinically not relevant outcomes.
Multiple-treatment meta-analysis may provide information on a ranking efficacy of new AEDs, thus answering several relevant questions regarding daily practice and decision-making. Although not free from concerns, historical design trials might have several advantages in that all patients receive a promising AED at dose(s) that are expected to be fully effective and eliminate the need for a parallel group on suboptimal treatment. All the presented strategies to overcome the lack of head-to-head comparisons can't anyway be considered as a substitute for properly conducted direct-comparison randomised trials, which remain the most relevant source of data to inform clinical decisions.
References