Flow Diversion in the Treatment of Intracranial Aneurysms: A Pragmatic Randomized Care Trial

DISCUSSION

FIAT is the first pragmatic trial showing improved efficacy outcomes with FD compared with other common management options. FIAT is also unique in showing that an endovascular innovation can be assessed in a randomized trial at the same time that it is introduced in clinical practice.^13,22 This change is an important breakthrough, for neurovascular devices have so far been approved on the basis of industry-led single-arm case series of ∼100 patients selected for regulatory approval purposes.^2,23,24 This is problematic, for clinical usage typically expands beyond the initial regulatory indications without reliable evidence that the innovation improves patient outcomes.²⁵

FIAT successfully addressed the multiple challenges that confront trials that assess the safe introduction of promising-but-potentially risky surgical innovations. One challenge is the diversity of patients, aneurysms, and clinical presentations. Before the availability of FD, these patients were treated using various ASMOs selected according to clinical presentation, individual patient anatomy, aneurysm location and morphology, local practice patterns, and individual preferences.¹⁰ The goal of treatment also varies, from relief of symptoms of mass effect to the prevention of life-threatening aneurysm ruptures in asymptomatic individuals. An explanatory attitude, looking for a signal that the innovation can work, fearing to lose that signal due to the noise associated with heterogeneity, might have called for multiple trials, each comparing FD in homogeneous patients eligible for only a single alternative treatment. However, this approach would not have addressed our main concern, which is whether FD actually improves patient outcomes in routine practice.²⁶

FIAT is also the prototype care trial.¹³ The guiding principle of the design is that patients are better protected when promising-but-unproven innovations are offered within a carefully controlled research context. This principle means that each item of the trial protocol must be reviewed to best protect the medical interest of participants.^13,22 Care trials are as inclusive as possible; protocols are flexible and allow the use of the treatment the patient would have otherwise received. The comparator intervention was prespecified before stratified randomized allocation, allowing valid comparisons within subgroups.

Care trials are fully integrated into clinical practice, and involve no extra risks, tests, visits, or cost compared with routine care. Although this design allowed us to conduct the trial without the support of industry or research agencies, the lack of funding discouraged many centers from participating. The data were collected by normal care personnel at the time of routine clinical follow-up visits. The main drawback of the approach, when rigorously applied, is that blinding becomes almost impossible. To mitigate that problem, we used a standardized, simplified mRS questionnaire¹⁸ and dichotomized results for a more difficult end point, death or dependency, which is more resistant to bias. This approach has been shown to be reliable.²⁷

The primary end point of the trial was a composite that attempted to capture in 1 judgment the comparative value of treatment in terms of safety and efficacy. This outcome combined clinical results (an independent functional score [mRS <3], which is the scale most frequently used in stroke trials) and the absence of clinical events signaling treatment failure during follow-up (such as aneurysm rupture, progressive mass effect, or retreatment of the index aneurysm). To account for the relatively short follow-up compared with lifetime protection against ruptures, which is the goal of treatment in many patients, the primary outcome included an angiographic result: complete or near-complete occlusion of the aneurysm. Angiographic outcomes are routinely used in practice to judge the success of therapy.²⁸ They are also the most common primary outcome of aneurysm trials.^20,28-33 Residual aneurysm was used to judge treatment failure because this category has been shown to be repeatable and its clinical significance is more constant across various raters, imaging modalities, and treatments.^20,21 The exploratory analyses we performed showed that using complete or near-complete occlusion as the angiographic outcome did not change the conclusions.

The overall morbidity and mortality of patients treated in FIAT are relatively high compared with these outcomes in other endovascular trials.^34-36 However, those comparisons are not valid because aneurysms considered for FD in FIAT were typically larger, more frequently had symptoms of mass effect, and many were difficult to treat by any and all methods. The primary safety outcome for all patients who received FD (26/209 or 12%; Online Supplemental Data) was at the upper limit of our initial estimate (10%) but similar that in to another randomized trial that compared FD with stent-assisted coiling.¹¹ In the Parent Artery Reconstruction for Large or Giant Cerebral Aneurysms Using a Tubridge Flow Diverter (PARAT) trial, the rates of death or stroke related to target vessels at 1-year follow-up were 14% and 17% in the stent-assisted coiling and FD groups, respectively.¹¹ Delayed complications, such as aneurysm ruptures, parenchymal hematomas at a distance from the aneurysm, and strokes from stent thrombosis were infrequent (10/209 or 4.8% of all patients having undergone FD in FIAT; Online Supplemental Data), but they remain a concern if FD is to be widely used.

The trial was all-inclusive, but not all patients with aneurysms were considered for FD, and results may not apply to all patients. To examine how study results could apply in clinical decision-making, one must examine prespecified subgroups of patients, even if the interaction tests were not statistically significant. The interaction test examines the potential influence of a variable on a single relative treatment effect, an implausible assumption in this trial comparing FD with as widely different options as conservative management and PVO. Nevertheless, subgroup results should always be interpreted with caution.

Examining the prespecified subgroups defined according to the comparator ASMO showed that FD was more effective than coiling with or without stent placement, but comparisons with PVO or surgical clipping were too few to draw firm conclusions.

Because of the pragmatic nature of the trial, conservative management was a potential prespecified ASMO (actually selected in 23 or 8% of patients in the RCT). This design choice may seem to unfairly favor FD. However, the alternative, to exclude these patients from the RCT, would have deprived the trial of the capacity to show this important advance provided by FD, ie, the ability to treat patients who otherwise could not be treated.

For prespecified aneurysm characteristics, FD was shown to be superior in the treatment of large, unruptured intradural carotid aneurysms (>10 mm), confirming the results of a previous trial.¹¹ For the initially approved regulatory indications for the use of FD (proximal carotid aneurysms), results were neutral, for they pull in opposite directions. Many large or giant extradural carotid aneurysms were treated with PVO, a treatment subgroup that was superior to FD (but not significantly) (Fig 2). Another trial showed FD to be less effective (but safer) than surgical occlusion and bypass in the treatment of complex anterior circulation aneurysms.¹² Surgical bypass was not used in FIAT, even in combination with PVO, and the RCT included only those patients with a circle of Willis that allowed safe carotid occlusion without bypass. PVO remains a good option for these selected patients.³⁷

Posterior circulation aneurysms have been shown in observational studies to have worse outcomes after FD than anterior circulation aneurysms, but FIAT randomized results show that ASMOs did not perform better for these difficult lesions (Fig 2).²⁵

There are a priori concerns with the use of antiplatelet regimens that come with the use of FD in patients with SAH from a recently ruptured aneurysm.³⁸ Too few patients with SAH were included in FIAT for conclusions to be drawn about FD use in these patients.

For the most frequently encountered small (<10 mm), unruptured, asymptomatic, intradural aneurysms, the primary efficacy outcome was similar for patients treated with FD or ASMO. Only by looking at the rate of complete occlusion in exploratory analyses can a signal possibly favoring FD be detected (Online Supplemental Data). FIAT results are in accordance with performance goals for FD of small and medium aneurysms that have recently been proposed (morbidity/mortality in 7.8% of patients [4.8%–11.4%]).³⁹ Subgroup safety data for coiling with or without stent placement (3%; 95% CI, 0%–8.5%) and for the corresponding stratified patients with FD (5%; 95% CI, 1.6%–11.4%) were comparable in FIAT, but a larger RCT would be needed for the ongoing concern of delayed complications with FD related to antiplatelet regimens and delayed stent thrombosis. Some believe that a 5%–8% risk may already be too high to justify the preventive treatment of small unruptured aneurysms.⁹ The pertinent question regarding small asymptomatic aneurysms is whether they should even be treated at all.⁴⁰ A pragmatic RCT has recently been launched to address this question.⁴¹

There are limitations to this study: Only 3 Canadian centers participated, limiting the generalizability of results. The recruitment period had to be increased to nearly 10 years, during which the indications for FD are likely to have changed. FIAT is the largest RCT on FD so far, but the number of patients remains small. On final analysis, safety results include the possibility of harm or benefit compared with other management options. Prerandomization resulted in 24 crossovers (8.6%) that dilute the contrast between groups. An aneurysm for which FD is considered a promising treatment is a vague definition that may be differently interpreted from one clinician to another and that may also change with time and experience. Clinical outcome assessment and core lab adjudications could not be masked to treatment allocation. Trial results were mainly driven by angiographic results, and the hard clinical outcome (death or dependency) accounted for a relatively small number of poor outcomes in both groups (13 versus 10 patients, including 9 versus 5 deaths). Potential bias from lack of blinding of mRS clinical assessors should not have significantly affected the results. The 12-month follow-up period was relatively short; too short to evaluate the effects of treatment on aneurysm rupture. In addition, this follow-up may not have given enough time for some FD-treated aneurysms to become occluded or for some recurrences after coiling to become apparent.⁴²