Introduction

The neurointerventional (NI) endeavor is a field of marked contrasts. Complex intracranial aneurysms, disabling strokes, and a variety of other neurovascular diseases are successfully treated through minimally invasive means. The technical and geographic growth of the field has seen a similar maturation in academic support and its resultant data. In that vein, over these past few decades single-center case series have been replaced by larger multicenter observational cohorts and ultimately by randomized controlled trials (RCTs). One of the more marked and important contrasts in the NI field exists within these RCTs—not, as many might assume, conflicting results but, instead, the quality of the underlying studies.

The pages of JNIS have chronicled these studies and we will not review the particulars of them again here. However, as with other spaces in the medical field, we have our share of both high-quality1–3 and low-quality4 studies. Some trials that the NI community collectively viewed as subpar were able to secure publication in lead journals, speaking to great variability of not just studies but review and publication. Indeed, JNIS has featured an article on some of the challenges of peer review.5

Sheth and colleagues published a recent comment called ‘Watching but not waiting: vascular neurology perspective on the disparate regulatory pathway for stroke’ in JNIS.6 They made a number of important points that should prompt conversation in our circles. The paper started with a discussion of how vascular neurologists were paying close attention to the completion of the regulatory pathway of the Watchman device for non-valvular atrial fibrillation as a cause of ischemic stroke.7 The authors lament the lack of involvement of vascular neurologists in the design of the Watchman trial and its subsequent approval pathway. We support this perspective from our neurology colleagues.

Non-inferiority versus superiority trial design

Sheth et al go on to explain that cardiology trials are typically non-inferiority in design whereas NI trials are more often superiority trials. We think it important to emphasize that both approaches are, however, used by each of the disciplines. While superiority trials aim to test the hypothesis that a new therapeutic intervention is better than an existing one, non-inferiority trials seek to test the hypothesis that the new intervention is not ‘much worse’ than the existing one. The main advantage of non-inferiority trials is that the overall therapeutic benefit of a new treatment needs to be similar to—not better than—the standard of care for the study to be considered ‘positive’. Formally, this kind of design is indicated in situations in which effective therapies for the disease of interest already exist, and especially when these therapies are considered to be powerful. In this setting, it can be complicated to demonstrate that a new intervention is more efficacious than the existing one. However, the new intervention may be better in certain facets: improved cost-effectiveness, better safety profile, or higher compliance rates. If this is the case, showing that the new intervention is ‘not worse’ than the existing one will suffice to justify its use. This is a key point about non-inferiority trials: factors other than biological efficacy must exist—and should have been appropriately tested—for the implementation of this approach to make sense.

From a statistical standpoint, a non-inferiority trial involves determining the maximal difference in therapeutic effect between the existing and new interventions that will be accepted in order to call the new intervention non-inferior to the existing one. Importantly, the scenario in which the new intervention is better than the existing one is not even considered. As a result, this type of trial involves deciding on the maximal difference between therapeutic effects that will be accepted as non-inferior and then looking only at the appropriate side of the CI for the difference between both interventions. While convoluted at first glance, this formal statement can be easily grasped through a familiar example. Imagine that we are considering a non-inferiority trial to compare therapeutic effects between a new endovascular device (ED) and intravenous recombinant tissue plasminogen activator (tPA) for acute ischemic stroke of all mechanisms. After looking at existing data, we decide that the maximum difference in therapeutic effects that we will accept between ED and tPA is 2%, or that we will label ED as non-inferior if effect_tPA − effect_EV <0.02. Needless to say, in the harshest scenario, the accepted difference between both treatment strategies would be 0%, meaning that effect_tPA − effect_ED = 0. Note that the scenario in which ED outperforms tPA, which would imply that effect_tPA − effect_ED <0, is not even being considered. After executing this trial, we would only look at the right bound of the CI for the difference in effects between both treatment strategies and conclude that ED is non-inferior if this bound lies to the left of the predefined effect difference (figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Superiority versus non-inferiority designs. Hypothetical results of a study that evaluates 90-day post-stroke modified Rankin scale, dichotomized as 0–2 vs 3–6. The figure presents risk differences between tPA (existing therapy) and EV (new intervention). The blue arrows represent the areas within which the results are considered significant; gray lines are 95% CIs. In non-inferiority studies we only evaluate the right bound of the CI for the difference in effects and conclude that EV is non-inferior if this bound lies to the left of the predefined effect difference (in this case 2%). EV, endovascular therapy; tPA, tissue plasminogen activator.

The field of cardiology frequently uses non-inferiority designs because two important conditions for their implementation are usually present: (1) the existing treatment for the disease of interest is considered to be powerful; and (2) the new intervention outperforms the existing one in some facet other than therapeutic efficacy. How does this impact the NI field? Physicians involved in the care of patients with stroke should engage in discussing if these two conditions are met for existing or future EV treatments. If this is the case, the non-inferiority approach should be considered and should probably be the approach of choice. Trying to demonstrate superiority in such circumstances would lead to an expected null result, with the consequent effects of a negative trial result.

Thrombolysis in myocardial infarction

While trial design is crucial, with the ever-growing number of meaningful trials in the NI space, one could ask about optimized approaches to effectively accomplishing them. Cardiovascular medicine provides an insight. In 1984 Dr Eugene Braunwald established the Thrombolysis in Myocardial Infarction (TIMI) study group in Boston, Massachusetts, headquartered at Brigham and Women's Hospital and affiliated with Harvard Medical School. Beyond developing the TIMI risk score for ischemic events and death in patients with unstable angina or non-ST elevation myocardial infarction, the TIMI group conducts a broad range of meaningful clinical trials in patients with cardiac disease. Indeed, the effect of TIMI has been profound even within our field; the TIMI grade for post-treatment flow in the coronary arteries has been adapted by our community for stroke treatment patients8—Thrombolysis in Cerebral Infarction, the very familiar TICI scale, now the modified TICI scale.

TIMI is an Academic Research Organization (ARO); in fact, it is the oldest cardiovascular ARO in the USA. The TIMI organizational structure provides in-depth knowledge and experience in the key aspects of a clinical trial performance, attracting large pharmaceutical and medical device companies as well as governmental agencies like the National Institutes of Health (NIH) who trust TIMI to handle their most important trials. Involving TIMI provides more than expertise; it provides additional credibility to any results that are derived from the trials, as TIMI is independent of the sponsoring entity. To that end, building from the pivotal findings of TIMI 1 demonstrating the utility and practice changing implications for intravenous tPA in ST elevation myocardial infarction, TIMI has now conducted over 60 clinical trials.9

This foresight is not unique to cardiology. Our neurology colleagues in stroke care have, through the NIH, recently created the NIH StrokeNet,10 currently headed by a national coordinating center. This organized network of 25 regional centers across the USA (in total over 200 hospitals) is designed to provide the infrastructural support for evaluating new stroke therapeutics. The network aims to conduct clinical trials to advance acute stroke treatment, stroke prevention, and recovery and rehabilitation following a stroke. A smaller role is for StrokeNet to provide an educational platform for those routinely involved in stroke care. To apply for trial consideration, one does not need to be part of the StrokeNet system, access that provides an avenue for those in NI work to potentially scale up meritorious projects. The promise of such an enterprise is vast.

Is it time for TICI to be more than a flow score?

The growth trend for NI trials will almost certainly continue, and likely accelerate. Gratifyingly, a recent spate of trials related to ischemic stroke demonstrated convincingly that our area could support positive trials. Registries (eg, STRATIS, NASA, and TRACK) are active, and have proved useful for analysis of our delivery of stroke care but do not readily allow prospective evaluation of challenging clinical questions—often the very questions that would have a major impact on the field. Currently, the majority of NI trials are supported by industry. This is understandable given our evolutionary phase as a field. Industry typically relies on a combination of internal resources or clinical research organizations to provide trial-related services. The benefits of a centralized TIMI-like group would inure almost immediately and manifest as enhanced credibility of NI studies. By cultivating expertise in a single organization, efficiencies of scale occur with academic leadership, global trial management, biostatistical support, and the many other elements of successfully completing clinical trials.

We are under no illusion that such an endeavor would be easy. Many immediate questions come to mind. Who would initiate it? Where would centralized control reside? Who would participate? How would it be funded? Industry leaders would need to be comfortable with a centralized third party that might control the data and subsequent publications. Moreover, there are costs to running an organization as exemplary as TIMI, and cardiovascular medicine has a broader available population of patients, practitioners, and possible industry supporters than NI and cerebrovascular medicine. Nonetheless, we believe that there are real costs to our space that are incurred by not having such a central repository for trial development and execution. To that end, we ask whether it is time for TICI to take on an additional meaning and be the name of more than merely a flow score. In 1984 Dr Braunwald took that risk and, to paraphrase George Bernard Shaw, asked, “Why not?”. Should we—the cerebrovascular community—not do the same in 2015?