Introduction

Confirmatory Phase 3 clinical trials typically involve the demonstration of efficacy and safety of a new intervention based on comparisons between a treated group and a control group, which may be placebo or another active treatment. Guidelines on clinical trial design generally state that new interventions should be compared with the best available therapy, and that a placebo comparator should only be used where no effective treatment exists [1]. This creates an ethical dilemma and clinical trial design challenge when regulatory authorities require large-scale, randomised, Phase 3 trials that are placebo controlled (in the absence of an approved treatment option).

As an example, intravenous immunoglobulin (IVIg) has been considered first-line therapy in chronic inflammatory demyelinating polyradiculoneuropathy (CIDP) for several years, based on evidence from small trials that IVIg is beneficial [2]. In the absence of an approved treatment for CIDP, gaining regulatory approval for the use of IVIg in CIDP required the conduct of large-scale, placebo-controlled Phase 3 trials.

A randomized, double-blind, controlled design (RCT) is the gold standard for confirmatory trials. Strengths of the RCT design include reduction or elimination of patient selection bias, minimization of imbalance of covariates such as prognostic factors, and enhanced reliability of statistical analysis due to the introduction of randomization [3]. Yet the RCT design also presents methodological and ethical challenges, including dilemmas in choosing the most suitable comparator, in ensuring an honest null hypothesis, and in designing studies that are not only acceptable ethically to physicians and patients but also meet the rigorous standards for Phase 3 trials required by regulatory authorities to support licensure. The latter challenge is the focus of this article and revolves around the concept of clinical equipoise.

Clinical equipoise

The ethics of clinical research require a state of clinical equipoise to exist between the comparator treatments. The term “clinical equipoise” was introduced in 1987 by Freedman [4], who defined it as a state of genuine uncertainty on the part of the clinical investigator about the relative therapeutic merits of the treatment arms in a trial. An investigator who has good reason to believe that one treatment is superior is obliged ethically to offer that treatment to patients.

While the results of RCTs invariably show that treatments differ in their effects, clinical equipoise at the start of a trial and throughout its duration protects patients from knowingly being exposed to inferior treatments. Progress in clinical medicine relies heavily on the willingness of patients to take part in clinical trials and evidence shows that they are only willing to participate in RCTs if there is an acknowledgment of expert uncertainty and if clinical equipoise exists [5]. Without clinical equipoise, investigators may be unwilling to risk their patients being randomized to a treatment arm (or placebo) that data indicate may be inferior, while patients are unlikely to enroll for the same reason.

The ethical debate surrounding clinical equipoise and RCTs

The principle of clinical equipoise as the basis of clinical research ethics has come under increasing criticism in recent years [6–10]. Miller and colleagues argue that clinical research and therapeutic practice are distinct activities with different goals, and are governed by different ethical principles [8, 9, 11, 12]. In their view, physicians in clinical practice have a moral obligation to provide patients with optimal care, whereas investigators in clinical trials have a primary duty to increase scientific knowledge, which may be at the expense of their secondary duty—to prevent harm to experimental subjects. This view has been regarded as unsatisfactory by many because it requires an implausible moral dissociation whereby trial investigators must ignore the professional obligations that they have as physicians [13].

The question arises as to whether clinical equipoise can ever exist in late-stage development trials. Some evidence for the new treatment under test always exists before a confirmatory RCT is conducted, including data from in vitro and animal experiments, data from uncontrolled clinical studies, evidence for the same treatment in other diseases, and evidence for similar treatments in the same disease [14–16]. Clinical equipoise may also be challenged by the accumulation of data during an RCT. To maintain clinical equipoise, investigators are usually prevented from looking at the accumulating data during the study, and an independent data monitoring committee may be tasked with stopping or modifying the trial if the accumulating data indicate that this is necessary. This approach is generally effective, with the notable exception of situations where test drugs become identifiable by virtue of having noticeably different side-effect profiles.

A novel response-conditional crossover trial design to ease concerns about lack of equipoise

When clinical equipoise is in question, as is undoubtedly the case with placebo-controlled trials, drug manufacturers and regulatory bodies share an obligation to utilize clinical trial designs that remove investigators’ ethical objections and protect patients while providing the regulators with appropriate evidence to grant approval. Without this imperative, there is a risk of patients being denied access to an effective treatment for a condition where no equivalent treatment option exists.

Response-adaptive clinical trial designs utilize outcome data that accumulate as the trial progresses to assign more patients to the better treatment arm. An alternative novel approach is to use a response-conditional crossover study design. This study design was adopted for a randomized, double-blind, placebo-controlled, pivotal, Phase 3 trial of 10% caprylate-chromatography purified immune globulin intravenous (IGIV-C; Gamunex®, Talecris Biotherapeutics, Research Triangle Park, NC, USA) for the treatment of CIDP [17]. This response-conditional crossover design differed from a typical crossover design and a response-adaptive design in a number of ways, as outlined in Table 1.

Table 1 A comparison of the response-conditional crossover design with (a) a typical crossover design and (b) a response-adaptive design
Full size table

At the start of the IGIV-C CIDP efficacy (ICE) study, there was a lack of clinical equipoise. Three of four small, short-term, placebo-controlled studies had suggested that IVIg was efficacious in patients with CIDP [2]. Subsequently, a meta-analysis concluded that IVIg improved disability in patients with CIDP for at least 2–6 weeks compared with placebo, and had similar efficacy to plasma exchange and oral prednisolone [2]. Furthermore, IVIg was being used in several countries to treat patients with CIDP (even though the labelling for IVIg did not include CIDP as an indication), and IVIg was recommended as a first-line treatment option for CIDP in clinical practice guidelines [18–20]. Because of the lack of equipoise, investigators were unwilling to expose patients to long-term placebo treatment. Therefore, a trial had to be designed in which exposure to placebo was minimized.

The prospectively designed, randomized, double-blind, placebo-controlled ICE study included an initial treatment period incorporating response-conditional rescue crossover that is the focus of this paper and an extension phase (Fig. 1). After screening, patients were randomised in a 1:1 ratio to receive either IGIV-C or placebo. Patients randomized to IGIV-C received a loading dose of 2 g/kg over 2–4 days followed by a maintenance infusion of 1 g/kg over 1–2 days every 3 weeks for up to 24 weeks. Albumin (0.1%) was used as the placebo.

Fig. 1
figure 1

The ICE study—a response-conditional crossover trial design [17]. Reprinted from [17] with permission from Elsevier

Full size image

During this first period, patients either remained on their randomized treatment or switched to the alternative treatment depending on their treatment responses [17]. Nonresponders or “rescued” patients received the alternative treatment for up to 24 weeks and were withdrawn from the study if no improvement was seen after one infusion of the alternative treatment or if improvement was not maintained during the crossover period. Patients who showed maintained improvement and completed 24 weeks of treatment in either the first or crossover periods were eligible for re-randomization in a blinded 24-week extension phase. During the extension phase, patients were again withdrawn from the trial if improvement was not maintained.

This response-conditional crossover trial design enabled the inclusion of a placebo arm to meet regulatory requirements, while minimizing ethical concerns about placebo treatment (the duration of exposure to treatment was short for any therapy that did not provide sustained improvement in the patient’s condition) [17]. Importantly, the design was valid for the primary endpoint: completion of the first period without crossover (IGIV-C responders 54.2% vs placebo responders 20.7%; p = 0.0002). The results from the crossover period provided verification of these findings (IGIV-C 57.8% vs placebo crossover completers 21.7%; p = 0.005). Overall, the study design reduced patient exposure to the inferior treatment in favour of the agent with superior efficacy (see Table 2).

Table 2 Treatment exposure in the ICE study [17]
Full size table

The response-conditional crossover trial design has some limitations. It requires investigators to be vigilant to ensure that crossover is applied correctly to avoid patients remaining on therapy to which they are not responding. In addition, due to differences in duration of drug exposure, the safety data need to be adjusted to provide incidences of adverse events per infusion.

A greater limitation is the reduction in the data that can be collected during the latter stages of the study, due to enhanced patient crossover or drop-out (withdrawal). In the case of the ICE study, secondary efficacy endpoints included changes in grip strength and nerve conductions from baseline to the end of the first period (up to week 24). The results for these secondary endpoints were supportive of the results for the primary endpoint, but as patients could be crossed-over to the alternative therapy at various time points during the first period, the number of patients providing data to week 24 was reduced. Furthermore, the number of patients in the extension phase was reduced because patients who were crossed over to the alternative treatment were discontinued if they failed to improve and maintain improvement based on the adjusted INCAT disability score.

Despite the limitations discussed above, the response-conditional crossover trial design appears to provide a rigorous evaluation of the drug under test. The ICE study results [17] confirmed the efficacy of IGIV-C in treating CIDP observed previously [2] and resulted in successful licensure in the US, Canada, Europe, and elsewhere.

Conclusions

The development of innovative trial designs may help to ease concern about the lack of clinical equipoise in clinical trials while providing the clinical trial data required to support regulatory approvals. In the case of the ICE study, a novel response-conditional crossover design addressed concerns about lack of equipoise raised by physicians interested in trial participation. The design minimized patient exposure to the inferior treatment and proved acceptable to both investigators and regulatory authorities. We conclude that trial sponsors should collaborate with investigators, experts in the field, and regulatory authorities at the protocol design stage to ensure that any proposed study design minimizes ethical concerns regarding lack of equipoise.