COMPLETE was a global, prospective, multicenter, single-arm, observational registry assessing the performance and safety of the Penumbra System in a patient population with AIS-LVO (ClinicalTrials.gov Identifier: NCT03464565). The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Inclusion criteria were: (1) patient age ≥18 years, (2) pre-stroke mRS 0–1, (3) Patient experiencing AIS secondary to intracranial LVO who are eligible for MT using the Penumbra System, (4) Planned frontline treatment with Penumbra System, and (5) Signed informed consent per Institution Review Board/Ethics Committee.

Exclusion criteria were: (1) Any comorbid disease or condition expected to compromise survival or ability to complete follow-up assessments through 90 days, and (2) Currently participating in an investigational (drug, device, etc.) clinical trial that will confound registry endpoints. Patients in observational, natural history, and/or epidemiological studies not involving intervention are eligible.

Participating hospitals maintained a screening and enrollment log of all AIS-LVO patients admitted into the hospital who were eligible for MT. Reason(s) for exclusion were recorded. Patients were considered enrolled once informed consent was obtained per Institutional Review Board/Ethics Committee and Penumbra System had been inserted into the body. Additional details on the COMPLETE Registry are reported separately (13).

Patients were either admitted directly to an enrolling hospital (direct) or transferred from an outside facility to the enrolling hospital (transfer). The decision for direct or transfer admission to a thrombectomy center was made by the stroke response procedures followed by the emergency services for each enrolling hospital and not dictated by the COMPLETE registry's protocol.

The choice of frontline treatment using direct aspiration only or aspiration combined with a 3D Revascularization Device was made by the treating physician. All procedures were conducted in accordance with routine care at each participating hospital and the Instructions for Use for each device. Available devices included: the Penumbra MAX, ACE, and JET Reperfusion Catheters; the 3D Revascularization Device; and the Penumbra Pump MAX, and ENGINE aspiration sources.

The primary efficacy endpoints were: angiographic revascularization of the occluded target vessel at immediate post-procedure as defined by a modified thrombolysis in cerebral infarction (mTICI) score of 2b or higher, and, good functional subject outcome at 90 days post-procedure as defined by a modified Rankin Score (mRS) 0–2. The primary safety endpoint was all-cause mortality at 90 days.

Secondary endpoints included: Incidence of device- and procedure-related Serious Adverse Events (SAEs) at ≤24 h, occurrence of embolization in previously uninvolved or new territories (ENT) as seen on the final control angiogram at the end of procedure, occurrence of symptomatic intracranial hemorrhages (sICH) at 24 h, vessel perforation, vessel dissection, length of hospital stay, and time metrics including onset to puncture, door to puncture and puncture to revascularization median times.

An independent imaging core lab reviewed pseudonymized angiography for mTICI scores by pass, embolization of new territory, pre-procedure CT for ASPECTS, CTA for clot location, and 24-h CT to assess hemorrhagic transformation using European Cooperative Acute Stroke Study (ECASS) classification.

Independent medical reviewers (IMRs) reviewed and adjudicated study endpoint events, including device related SAEs, neurovascular procedure related SAEs, sICH within 24 h, neurological deterioration events, and any deaths that occurred throughout the registry. Neurological deterioration was defined as a ≥4 point worsening of the National Institutes of Health stroke scale (NIHSS) score from baseline and sICH was defined as 24-h evidence of an ECASS defined ICH associated with a ≥4 point worsening of the NIHSS score from baseline.

Data was summarized using descriptive statistics. This included the number of observations, mean, standard deviation, median, and interquartile range (IQR) for continuous variables, and counts and percentages for discrete variables. Comparisons between direct and transfer patient groups were conducted using the t-test or Wilcoxon Rank-Sum Test for continuous variables and Fisher's Exact Test for categorical variables. All confidence intervals presented are Clopper-Pearson two-sided intervals. All statistical tests were performed using SAS version 9.4 (SAS Institute, Cary, NC) with α = 0.05.

Multivariable logistic regression models with an interaction term for onset-to-puncture time by cohort, were used to examine the variables associated with 90-day mRS. Candidate clinical predictors included cohort (direct vs. transfer), onset-to-puncture time (by 30 min intervals), age (by 10 year intervals), revascularization success (TICI2b-3 post-procedure), occlusion location, tPA (administered vs. not), baseline NIHSS (by 5 point intervals), hypertension (present vs. not), geographic region (US vs. Europe), baseline ASPECTS (<8 vs. ≥8). This model included 250 patients (direct n = 129, transfer n = 121) with witnessed stroke events and successful revascularization. Patients with onset-to-puncture times >600 min were statistical outliers and excluded.

Patients treated directly at a comprehensive stroke center had shorter onset-to-admission times (86.5 vs. 270.0 min), onset-to-puncture times (191.0 vs. 339.0 min), and onset to mTICI 2b-3 else final angiogram times (242.0 vs. 373.0 min) as compared with patients transferred secondarily (p < 0.0001 for all comparisons; Table 2; Figure 1). There was no significant difference in procedure time (puncture to mTICI 2b-3) between direct and transfer cohorts (Table 2); however, admission-to-puncture times were longer in the direct cohort, compared to transfer patients (p < 0.0001).

Out of 650 enrolled subjects, 613 had a 90-day follow-up assessment. Functional independence at 90 days (mRS 0–2) was significantly higher in the direct vs. transfer group [61.7% vs. 50.3% (+11.4%); p = 0.0056; Table 3; Figure 2]. There were no significant differences in post-procedural revascularization rates, nor in all-cause mortality at 90 days between direct and transfer cohorts.

Direct and transfer cohorts did not differ in rates of device- or procedure- related SAE within 24 h, nor in rates of procedural ENT, vessel dissection, or vessel perforation, and 24 h sICH (Table 3). The length of hospital stay was also similar between direct and transfer patients (6.0 vs. 7.0 days).

In the multivariable analysis, independent predictors of good functional outcome (mRS 0–2) included lower baseline NIHSS, baseline ASPECTS ≥8, IV-tPA administration, lower age, a distal occlusion location (e.g. M2–M4, instead of M1), and faster onset-to-puncture times (Table 4). For onset-to-puncture time, the odds of patients achieving good functional outcome decreased by 0.85 for every 30 min that treatment was delayed (p = 0.0180; Table 4). No interaction effect was observed between the direct admit cohort and onset-to-puncture time (p = 0.6612; Figure 3). The effect of direct admission on rates of mRS 0–2 did not persist after controlling for time from onset to puncture, nor other baseline and procedural covariates. For patients who were directly admitted, the odds of good functional outcome were higher [adjusted OR 2.17 (95% CI: 0.42, 11.3)] than for transfer patients; however, this effect was not significant (p = 0.3584).

Results from COMPLETE show that direct admission can significantly benefit functional outcome in AIS-LVO patients treated with frontline aspiration thrombectomy, and mirror two similar prospective registries in which mechanical thrombectomy was performed using stent retrievers (7), or a combination of stent-retriever or contact aspiration firstline (6). The magnitude of this effect was consistent across studies, with ~10% more patients achieving mRS 0–2 at 90 days in the direct vs. transfer cohorts: COMPLETE 61.7 and 50.3%, respectively; STRATIS 60.0 and 52.2% (7); and, ETIS 60.1 and 52.6% (6).

To further understand the benefit of direct admission on good functional outcome (mRS 0–2), a multivariable logistic regression was used to adjust for baseline and procedural characteristics (Table 4). In this analysis, the effect of direct admission was non-significant (p = 0.3584), suggesting that the benefits of direct admission occur indirectly, either from variation in tPA administration, baseline NIHSS and ASPECTS, occlusion location, or, based on predictive modeling presented in this study, from the time interval between stroke onset to puncture (Table 4). Furthermore, no significant interaction effect was found between patient cohort and onset-to-puncture time, supporting results that the benefits of direct admission can be attributed to delays in patient treatment (Table 4).

In the COMPLETE Study, the admission-to-puncture interval actually took significantly longer in the direct admission patients (87 min) compared with the transfer patients (48 min; p < 0.001; Table 3). In the ARTESp study (14), mechanical thrombectomy was initiated significantly faster for transfer patients following arrival at an endovascular-capable center, due to the availability of imaging. Although this did not compensate for the lost time incurred by secondary transport (direct admission was twice as likely to yield a favorable clinical outcome), this is consistent with the admission-to-puncture data from COMPLETE and the recognition that multiple variables can account for the improved clinical outcomes observed in direct admission patients.

Revascularization success (mTICI 2b-3), 90-day mortality, and rates of sICH, did not differ between direct vs. transfer cohorts in the COMPLETE registry. These results are again consistent with recent studies reporting similar revascularization and safety outcomes regardless of patient transfer status (6, 7).

Our results are further supported by recent meta-analyses reporting no significant differences found between the two transportation paradigms in the rate of sICH, mortality at 3 months, or in successful recanalization (15, 16).

For AIS-LVO patients eligible for mechanical thrombectomy, treatment delays can result in unfavorable clinical outcomes (17). Therefore, triage routing should focus on improving interval times and reducing time imaging and revascularization (18). For example, the interval between symptom onset and reperfusion might be reduced by bringing transfer patients directly to the angiosuite, a possibility supported by data showing intrahospital processing times were nearly twice as quick for transferred patients. Nevertheless, in the COMPLETE registry, patients who were secondarily transferred to an endovascular-capable center experienced longer onset-to-hospital admission times, longer onset-to-puncture times, and longer onset-to-revascularization times. When comparing the symptom onset-to-revascularization intervals between the direct (242.0 min) and transfer (373.0 min) cohorts, patients experienced a delay of 131 min if they were secondarily transported to an endovascular-capable center (Table 2).

This time delay is not surprising and compares to a median onset-to-revascularization delay in transfer patients of 109 min (311.5 vs. 202.0 min) in STRATIS (7), and of 100 min (219.0 vs. 319.0 min) for transferred patients in ETIS (6). When controlling for time from onset-to-treatment, the authors in STRATIS found no difference in outcome between direct and transfer patients, indicating for that registry population, that improved functional outcome was attributed solely to time delays associated with the transfer paradigm (7). Results from the COMPLETE registry corroborate this analysis, with direct admissions significantly associated with both faster onset-to-revascularization times, as well as better functional outcomes.

The results observed in COMPLETE are notable in that the overall effect on functional outcome remained comparable to registries in which transfer patients were treated approximately 1 to 2.5 h earlier than those in COMPLETE (6, 7). In COMPLETE, time from onset-to-treatment for the highest quartile was just over 9 h (547.0 min; Table 2); despite this, a significant benefit was still observed between direct and transfer cohorts. By contrast, controlled trials for late-presenting strokes (19, 20) reported equivalent benefits for both direct admissions and for transfer patients. Nevertheless, the authors note the selective nature of the patient population for penumbral mismatch and likely over-representation of candidates with favorable collaterals (19, 20), emphasizing the need for nuanced, controlled studies of transportation paradigms for AIS-LVO.

IV-tPA administration was an independent predictor of better functional outcomes for patients overall (Table 4), although there was no significant difference in IV-tPA administration across direct and transfer patients (Table 1).

Froehler et al. (7) remarks that although the ability to begin IV-tPA sooner is often cited as a reason to bring patients to a nearer hospital, direct-to-endovascular bypass may still be beneficial because it accelerates the start of endovascular therapy. For example, when modeling a bypass <20 miles to an endovascular-capable center, tPA is delayed by only 6.9 min, while endovascular treatment to begin 94 min sooner (7).

Nevertheless, the use of bridging thrombolysis when appropriate may alleviate the difference in transport paradigms, potentially by compensating for the negative effects of delayed reperfusion time: Zhao et al. (16) found in a subgroup analysis of that bridging thrombolysis resulted in similar outcomes for direct and transfer cohorts with respect to sICH, 90-day favorable functional outcome, 90-day mortality, and successful recanalization. In practice, however, the authors concur that for patients ineligible for IV thrombolysis, direct transport to the closest endovascular-capable center may be the best option (16).

Our results establish that AIS-LVO patients treated with frontline aspiration thrombectomy can also benefit significantly from direct routing to endovascular-capable centers. Further work should focus on in-field patient screening tools and controlled trials, e.g., RACECAT (NCT02795962), to provide crucial data to inform pre-hospital management systems for AIS-LVO. In the RACECAT trial, no differences were found between direct vs. transfer patient cohorts in Catalonia admitted for thrombectomy within 7 h of symptom onset (8). While RACECAT provides important data on the efficiency of stroke care in Catalonia, outcomes from the region's single, integrated, public-use healthcare network (which covers the entire Catalonian population of 7.5 million) (21) might not be generalizable to health systems elsewhere in the world, and raises vital questions regarding whether aspects of this system can be adapted to other urban and rural catchments. For instance, the onset-to-puncture times in RACECAT were only delayed by 56 min for transfer patients (onset-to-puncture for transfer was 270 vs. 214 min for direct) (8). By comparison, in our registry, onset-to-puncture time delay was 148 min (339 min for transfer vs. 191 min for direct, p < 0.0001) – nearly triple that of RACECAT. This much longer delay in onset-to-puncture time encountered in our registry may explain why in RACECAT, direct and transfer had equivalent outcomes, but in our registry, direct admission patients had significantly higher rates of good functional outcome (mRS 0–2) at 90 days. This point is supported by our multivariable logistic regression (Table 4) which found longer onset-to-puncture (per 30 min increase) to be a significant independent predictor of worse functional outcome at 90 days [OR 0.85 (95% CI 0.74, 0.97), p = 0.0180], whereas direct admission was not found to be a significant independent predictor [OR 2.17 (95% CI 0.42, 11.3), p = 0.3584].

An important consideration in design of COMPLETE compared to studies such as RACECAT, STRATIS, and DEFUSE-3, is that COMPLETE was a highly inclusive, real-world evaluation of global AIS-LVO treatment. COMPLETE was a global registry that included patients for aspiration thrombectomy with no restrictions on stroke location (anterior/posterior) or onset time. COMPLETE provides real-world data on outcomes based on local systems of care, transfer patterns, and time delays across broad geographies. This experience showed that under current systems of care, direct admission is appropriate for a majority of admitted strokes. However, each locality should develop their own stroke protocols based on the location of, and transit time required to reach a comprehensive stroke center or a stroke center with thrombectomy capability.

Additional research should focus on AIS-LVO detection to facilitate direct routing of patients to appropriate treatment centers. We concur with the need for well-designed, randomized-controlled trials to evaluate AIS-LVO systems of care based on regional access characteristics and patient populations (20, 22).

A limitation of this non-randomized study is the lack of a comparator arm. This study also did not track the decision making process for determining if a patient should be transferred or directly admitted, the distance from location of symptom onset to thrombectomy center (hub), any time metrics related to the initial institution (spoke) for transfer patients (e.g., onset to spoke arrival time, spoke door in to door out time, spoke departure to hub arrival time), nor time of IV-tPA administration, thus limiting possible analyses. Additionally, the conclusions of this analysis may not be widely applicable where primary and comprehensive stroke centers are not similarly accessible for all patients and direct admission may not be an option for many geographical areas.

Nevertheless, this large, prospective registry was core-lab adjudicated, and safety results were assessed by independent medical reviewers. This registry also enrolled patients from a large, heterogeneous population that included posterior lesions, and did not exclude patients based on presentation window.