Article Text
PDF
Abstract
Background Novel thrombectomy strategies emanate expeditiously day-by-day counting on access system, clot retriever device, proximity to and integration with the thrombus, and microcatheter disengagement. Nonetheless, the relationship between native thrombectomy strategies and revascularization success remains to be evaluated in basilar artery occlusion (BAO).
Purpose To compare the safety and efficacy profile of key frontline thrombectomy strategies in BAO.
Methods Retrospective analyses of prospectively maintained stroke registries at two comprehensive stroke centers were performed between January 2015 and December 2019. Patients with BAO selected after MR imaging were categorized into three groups based on the frontline thrombectomy strategy (contact aspiration (CA), stent retriever (SR), or combined (SR+CA)). Patients who experienced failure of clot retrieval followed by an interchanging strategy were categorized as a fourth (switch) group. Clinicoradiological features and procedural variables were compared. The primary outcome measure was the rate of complete revascularization (modified Thrombolysis in Cerebral Infarction (mTICI) grade 2c–3). Favorable outcome was defined as a 90 day modified Rankin Scale score of 0–2.
Results Of 1823 patients, we included 128 (33 underwent CA, 35 SR, 35 SR +CA, and 25 switch techniques). Complete revascularization was achieved in 83/140 (59%) primarily analyzed patients. SR +CA was associated with higher odds of complete revascularization (adjusted OR 3.04, 95% CI 1.077 to 8.593, p=0.04) which was an independent predictor of favorable outcome (adjusted OR 2.73. 95% CI 1.152 to 6.458, p=0.02). No significant differences were observed for symptomatic intracranial hemorrhage, functional outcome, or mortality rate.
Conclusion Among BAO patients, the combined technique effectively contributed to complete revascularization that showed a 90 day favorable outcome with an equivalent complication rate after thrombectomy.
- thrombectomy
- angioplasty
- catheter
- device
- posterior fossa
Data availability statement
Data are available upon reasonable request.
Statistics from Altmetric.com
Introduction
Following the results of recent landmark randomized trials, endovascular thrombectomy (EVT) has been validated as the standard of care for anterior circulation stroke (ACS).1 Stent retriever (SR) thrombectomy, which was employed in more than 80% of patients within the MR CLEAN (Multicenter Randomized Clinical trial of Endovascular treatment for Acute ischemic stroke in the Netherlands), EXTEND-IA (Extending the Time for Thrombolysis in Emergency Neurological Deficits-Intra-Arterial), ESCAPE (Endovascular Treatment for Small Core and Proximal Occlusion Ischemic Stroke), and SWIFT PRIME (Solitaire With the Intention For Thrombectomy as Primary Endovascular Treatment for Acute Ischemic Stroke) trials achieved a significantly greater rate (58–88%) of successful recanalization (modified Thrombolysis in Cerebral Infarction (mTICI) grade 2b/3) and a proportionate higher rate of functional independence (modified Rankin Scale (mRS) score of 0–2) in 53–71% of patients compared with those treated with intravenous thrombolysis only.2
The introduction of soft tipped, atraumatic, highly trackable large bore intermediate catheters prompted the evolution of contact aspiration (CA) as frontline therapy in ACS.3 Similar angiographic results, analogous outcome, and comparable safety endpoints between CA and SR thrombectomy that were demonstrated in the ASTER (Contact Aspiration vs Stent Retriever for Successful Revascularization) and COMPASS (Aspiration thrombectomy versus stent retriever thrombectomy as first-line approach for large vessel occlusion) trials empowered incorporation of aspiration thrombectomy in the therapeutic armamentarium of stroke.4 5
As shown by Kang et al, the switching strategy between CA and SR resulted in a higher overall rate of successful recanalization compared with patients without the switching strategy (85.1% vs 73.8%, respectively).6 Currently, simultaneous distal aspiration with SR thrombectomy (combined SR +CA) has been broadly adopted with variable terminology to enhance the technical outcome and reduce potential clot fragmentation and distal embolization.7–10
To our knowledge, there are no current comparative studies regarding the safety and efficacy of foregoing revascularization techniques, particularly in the setting of basilar artery occlusion (BAO). Furthermore, swapping between techniques was customarily deemed as a failure with no separate exploration of those patients. We, therefore, investigated the influence of the frontline EVT technique on complete revascularization and outcome in MRI selected patients with BAO.
Material and methods
Study population
We conducted a retrospective analysis of two prospectively maintained stroke registries comprising stroke patients with large vessel occlusion, which were managed with EVT during the period between January 2015 and December 2019. The institutional review board approved the use of patient data for this research protocol and written informed consent was waived on the basis of the retrospective study design.
Patients fulfilling the following criteria were consecutively included: (1) occlusion of the basilar or intracranial V4 segment of the vertebral artery (VA) with a non-contributive or aplastic counterpart VA, as determined by time-of-flight MR angiography and confirmed by baseline DSA; (2) pre-stroke mRS score of ≤2; (3) baseline National Institutes of Health Stroke Scale (NIHSS) score of ≥4; (4) time interval between onset of symptoms or last time patient seen well until groin puncture of <24 hours; and (5) EVT performed with the latest generation of clot retriever devices. The choice of revascularization strategy was left to the discretion of the neurointerventionist in conformity with local institutional interdisciplinary stroke guidelines.
Among 1823 patients who had ischemic stroke and underwent EVT attempts, 142 consecutive patients with posterior circulation stroke were eligible for initial inclusion in the current study. We excluded all patients: (1) who underwent spontaneous satisfactory revascularization, as demonstrated at baseline angiography (n=1), (2) who were treated primarily by balloon angioplasty (n=1), or (3) who could not be assigned to a particular revascularization strategy because of procedural premature termination (target occlusion could not be accessed (n=7), clot penetration was not feasible (n=2), or significant vessel perforation attributable to micro guidewire navigation prior to commencing EVT (n=3)). Sixty-eight of 128 patients have been previously reported.11 12 These prior articles dealt with predictors of favorable outcome and mortality following EVT whereas here, we report on the impact of revascularization strategy on angiographic and clinical outcomes.
Clinical and treatment related variables
Patient demographics, cerebrovascular risk factors, baseline NIHSS, bridging therapy with intravenous tissue plasminogen activator, time intervals, clot location, and etiology were prospectively collected. Except for three cases diagnosed with intracranial vasculitis (n=2) and essential thrombocythemia, the etiology was determined based on angiographic appearance where three major causative mechanisms were depicted: (1) embolism attributable to tandem VA steno-occlusion defined as occlusion or high degree (>70%) of extracranial stenosis at the V1 segment of the VA; (2) thromboembolism with underlying intracranial atherosclerotic stenosis (ICAS) interpreted as severe (>70%) fixed focal intracranial stenosis overlaid with in situ thrombosis; and (3) embolism with no dubious atherosclerotic tandem or ICAS pathologies. Procedural technical details are illustrated in the online supplemental material.
Supplemental material
Outcomes measures
Technical efficacy endpoint consisted of first pass effect (FPE) rates (mTICI 3 after single device pass), successful, and complete recanalization rates defined as mTICI 2b/3 and mTICI 2c/3, respectively, on the final angiography in conformity with the BEST (Acute Basilar Artery Occlusion: Endovascular Interventions vs Standard Medical Treatment) trial and other prospective registries.13 14 The mTICI scores were independently evaluated by an experienced interventional neuroradiologist (MA) and neurologist (IM) who were blinded to functional outcomes. Disagreements were resolved by consensus. Clinical outcome was determined by the mRS score as favorable (mRS score of 0–2) or poor outcome (mRS score of 3–6). The mRS score was prospectively assessed by independent vascular neurologists or trained research nurses, unaware of the study group assignments, by face-to-face or structured telephone interviews.
A routine control non-contrast brain CT or MRI acquired 24 hours following EVT was scrutinized for hemorrhagic complications which were further classified according to ECASS-3 (European Cooperative Acute Stroke Study III). Symptomatic intracranial hemorrhage (sICH) was defined as any hemorrhagic event associated with an increase of at least 4 points in the NIHSS score within 24 hours.15
Statistical analysis
Statistical analyses were performed using SPSS V.24.0 (IBM Corporation, Armonk, New York). Continuous variables are expressed as mean (SD) or median (IQR) and compared respectively using the t test or the Mann–Whitney U test, based on the normality of the data distribution that was validated by histogram. Similarly, the Pearson χ2 test or Fisher’s exact test were used to contrast categorical variables, and are expressed as number (percentage).
First, differences were investigated between patients treated with a single approach (SR, CA), combined approach (SR +CA), and those treated with the switch strategy. Between groups, comparisons for continuous variables were made with the ANOVA or Kruskal–Wallis test, as appropriate. Categorical variables were compared using the χ2 test and Fischer exact test if the total frequency was less than 20 or the expected frequency was less than five in any cell.
Second, an initial univariate analysis was performed between clinicoradiological, procedural characteristics, and dichotomized revascularization rate (mTICI 2c/3 vs mTICI 0–2b) and 90 day functional outcome (mRS 0–2 vs 3–6). Third, between group confounding differences adding to variables with p<0.05 in the univariate analysis were selected for the complete revascularization multivariate logistic regression to assess the association between the EVT technique and complete revascularization primarily, and also favorable mRS at 90 days. All statistical tests were two sided and significance was set at p<0.05.
Results
Comparison between EVT techniques
Among 128 analyzed patients (83 (65%) men; mean age 66.9+11.8 years), 33 (26%) patients were treated with CA (CA group), 35 (27%) with SR (SR group), 35 (27%) with combined CA and SR (combined SR +CA group), and 25 (20%) with a heterogeneous frontline technique (20 patients with CA, 4 with SR, 1 with combined SR +CA) that underwent another revascularization technique (18 combined SR +CA, 5 SR, 2 CA) after failure of clot retrieval (switch group).
No clinically relevant differences were found for demographics, comorbidities, and baseline characteristics among the revascularization techniques, except for etiology, clot location, and procedural duration (table 1). As shown on baseline angiography, the target occlusion site was the distal basilar artery in 44% of patients (n=57) followed by the mid and proximal basilar artery in 23% (n=29) and 22% (n=28) of patients, respectively; the former was significantly more frequent in the SR group. Only 14 (11%) patients presented with intracranial V4 occlusion that was contrastingly underrepresented in the CA and SR groups.
Comparison of demographics, cerebrovascular risk factors, and procedural perspectives between the endovascular thrombectomy techniques in 128 patients with acute basilar artery occlusion
Patients in the switch group had a relatively lower proportion of embolic mechanisms (32% vs 79%, p<0.001) and higher ICAS proportion (36% vs 9%, p=0.01) compared with the CA group that were in agreement with the relatively greater number of device passes (median 3 (2–4) vs 1 (1–1), p<0.001; median 3 (2–4) vs 1 (1–2), p=0.001), slower revascularization in terms of increased median puncture-to-revascularization times (median 89 (56-124) vs 25 (18-52), p<0.001; median 89 (56-124) vs 45 (27-83), p<0.001) and higher rate of rescue maneuvers compared with the CA (40% vs 0%, p<0.001) and SR (40% vs 17%, p<0.05) groups, respectively.
Predictors of outcome
Overall, of 140 patients who underwent regular thrombectomy attempts, 83 (59.3%) patients achieved complete revascularization (mTICI 2c/3). The highest proportion of FPE (51%, 18/35) was noted in the combined SA+CA group, but it did not reach statistical significance. The combined SR +CA group achieved the greatest rate (80%) of complete revascularization (figure 1), which was synchronous with a lower rate of downstream and collateral embolization (6%, 2/35) compared with the CA (12%, 4/33), SR (20%, 7/35), and switch maneuvers (32%, 8/25) (p=0.04 (table 1).
Distributions of recanalization grading achieved by different thrombectomy techniques, as verified by the modified Treatment in Cerebral Infarction (mTICI) score on the final control angiography.
Combined SR+CA (adjusted OR (aOR) 3.04, 95% CI 1.077 to 8.593, p=0.04) along with rescue maneuver implementation (aOR 5.59, 95% CI 1.433 to 21.769, p=0.01), and puncture to revascularization interval (aOR 0.98, 95% CI 0.97 to 0.996, p=0.01) were independent predictors of complete revascularization. Complete revascularization was the only independent variable that was significantly associated with favorable outcome (aOR 2.73, 95% CI 1.152 to 6.458, p=0.02) in the multivariate logistic regression analysis following adjustment for relevant significant confounders (tables 2 and 3).
Baseline characteristics categorized based on recanalization and functional outcomes following endovascular thrombectomy
Multivariable logistic regression analysis for complete revascularization (modified Treatment in Cerebral Infarction grade 2c–3) and good clinical outcome (modified Rankin Scale score 0–2)
Periprocedural complications
Procedure related hemorrhagic complications among all attempted thrombectomies occurred in 11 of 140 cases (8%), which were due to subarachnoid hemorrhage traceable to microvascular perforations during micro guidewire navigation or difficulty in clot bypassing. sICH occurred in 5% (n=6/128) of patients while the mortality rate was 38% with no statistically significant differences between groups for safety outcomes except for the relatively higher rate of non-symptomatic parenchymal hemorrhage in the SR group.
Discussion
In this observational two center study, complete revascularization following EVT for BAO was the most significant predictor of a 90 day favorable outcome. Selection of the combined SR +CA technique as the frontline EVT strategy increased the rate of complete revascularization by 3.04-fold compared with single (CA or SR) or switch thrombectomy tactics.
EVT success and outcome
The precise relationship between revascularization and outcome is not entirely elucidated in BAO. In an early single center retrospective study, EVT was associated with a higher rate of complete revascularization, but with a similar outcome compared with intra-arterial fibrinolysis.16 A concurrent prospective international study (basilar artery international cooperation study (BASICS)) that enrolled 619 patients with BAO did not confirm the superiority of intra-arterial therapy over intravenous thrombolysis. However, the conclusions drawn might not be applicable to ongoing clinical practice because it was conducted more than a decade ago before the introduction of new generation clot retriever devices and novel thrombectomy approaches.17 Similarly, a recent Chinese multicenter randomized controlled trial (BEST) found no significant difference in favorable outcomes between patients treated with endovascular therapy and those receiving standard medical treatment only. However, the results in this study might be confounded by the premature termination owing to the poor adherence to the assigned study arm and the high crossover rate that was consistent with the prodigious benefit of thrombectomy in ACS and the poorer outcome in conservatively treated patients with BAO.13
Nevertheless, a recent prospective study in 39 BAO patients found that revascularization was a major predictor of functional outcome whereas failed revascularization induced a 13-fold increase in the risk of poor outcome.18 In an equivalent but larger prospective study in 829 BAO patients, EVT performed within 24 hours from symptomatic onset was associated with better functional outcome and less mortality compared with standard medical treatment alone.14
Revascularization grading and outcome
The principal anatomic dissimilarities between anterior and posterior flow systems include the disparate confluence of two VAs, potential collateral anastomosis, the relative paucity of nourishing vasa vasorum, and the overabundance of critical penetrating arteries. Therefore, the technical and clinical outcomes might be different from that of ACS.19 Several studies have demonstrated the superiority of mTICI 2c/3 over mTICI 2b in achieving a higher rate of early neurological improvement and better 90 day outcome in ACS.20 Correspondingly, our results showed that complete rather than successful revascularization was associated with favorable outcome in BAO, which is conceivable on account of the highly vital structures and the widely distributed side perforators that might leave a considerable deficit on occlusion. Moreover, devastating mass effect consequences with eventual mortality might result from mTICI 2b with single branch occlusion like the posterior inferior cerebellar artery infarct.21 22
EVT technique and revascularization
CA versus SR
Before the evolution in EVT devices, three small single center retrospective studies compared the efficacy of CA and SR in BAO in 31 (18 CA vs 13 SR),23 33 (20 CA vs 13 SR),24 and 50 (16 CA vs 34 SR) patients.25 No significant discrepancies were found with reference to mTICI 2b–3 rate and clinical outcome, however, CA had a faster procedural time and higher mTICI 2c–3 rate, in contrast with SR. Gory et al 26 further validated these results in 100 BAO patients where CA was associated with significantly higher odds (2.6-fold) of complete revascularization.
In contrast, in a later meta-analysis27 of 476 cases (193 CA, 283 SR), the rate of complete revascularization did not differ between the two devices but a higher rate of successful revascularization coupled with a lower incidence of collateral embolization was observed in the CA group.
Combined SR+CA versus CA
Maus et al 28 compared the combined SR+CA technique with CA (18 vs 21 patients, respectively) in BAO and observed a higher rate of first pass successful revascularization, FPE, lower number of passes, and a greater overall rate of complete revascularization in the combined SR+CA group. Another retrospective study showed that combined SR+CA was associated with a faster procedure and reduced thrombectomy attempts.29 These results are in line with a recent multicenter retrospective study that enrolled 353 patients with terminal internal carotid artery occlusions and showed better technical and functional outcomes when combined SR+CA was adopted as a frontline strategy compared with CA. The authors ascribed these results to the potential clot disruption and downstream embolization that might follow CA.30
CA versus SR versus combined SR+CA
A direct comparison between frontline EVT techniques (CA, SR, combined SR +CA) in BAO has been carried out in only one prospective study in 500 stroke patients but posterior circulation stroke was underrepresented (only about 11% of the total cohort). Procházka et al in this study found that CA and SR were associated with shorter puncture to revascularization times, higher revascularization rates, and subsequent better outcomes.31
A recent study demonstrated that CA could tightly grasp the proximal transverse face of the clot but it had no direct control on its rear end. Consequently, CA might be less successful in the management of erythrocyte rich clots which are associated with a high propensity for friability and downstream embolization. In contrast, SR interacts with the longitudinal aspect of the entire clot and therefore has a greater likelihood of seizing retrieval concomitant clot fragments. On the other hand, fibrin rich clots with their inherent higher friction coefficient tend to be less deformable which predisposes to poor device clot interaction, manifested by rolling over rather than integration within the struts of the SR.32
In agreement with Gory et al 26 and Maus et al,28 the combined SR +CA technique in our study expedited a higher rate of final mTICI 2c/3 with a similar frequency of periprocedural complications compared with the other revascularization strategies. The latter results might be explicated by the synergistic effect of clot trapping between a wedged aspiration catheter that tightly controls the proximal clot surface and the SR that completely engages the whole clot length which permits full clot extraction while reducing fragmentation and downstream embolization, respectively. These presumptions are in agreement with a previous study that analyzed data of 450 patients with ACS and showed significantly higher reperfusion rates in the combined SR +CA group concomitant with reduced attempts and a lower rate of secondary emboli.33 Similarly, Yeo et al found that BAO was especially more prone to distal embolization during thrombectomy, particularly with lack of flow arrest during retrieval, and residual antegrade flow from the contralateral VA.34 Phan et al in a recent meta-analysis found a higher rate (23.8%) of iatrogenic embolization in BAO compared with ACS. According to our results, the lowest rate of secondary embolism was observed when the combined SR +CA technique was used as a frontline technique which might underline the significance of the combined approach.35
Functional outcome
Two landmark randomized controlled trials (ASTER and COMPASS) contrasted the efficacy of CA and SR in ACS.4 5 The latter disclosed the non-inferiority of CA with regard to functional outcomes. In the ASTER trial, comparable rates of FPE (defined as mTICI 2c/3 after one pass with no adjuvant maneuvers) were obtained which was associated with better outcomes and less mortality. In parallel, the revascularization technique in our study had no direct impact on outcome; nonetheless, combined SR +CA exhibited a higher rate of final mTICI 2c/3 that sequentially ameliorated the functional outcome.
In BAO, apart from a small retrospective series24 which showed a higher frequency of favorable outcome (mRS 0–3) at discharge in the CA group, pooled results of mRS 0–2 at 3 months in a recent meta-analysis27 showed no significant difference between CA and SR.
Switch and rescue maneuvers
Various rescue maneuvers, such as switching to another EVT technique, angioplasty, adjuvant stenting, thrombolytics, and antiplatelet medications might be substantially efficacious to overcome the potential failure of front-line EVT.
Kang et al 36 in a retrospective study involving 955 patients with ACS suggested that CA necessitated a greater number of device passes and a higher rate of rescue therapy that implied longer procedural time and poorer functional outcome. Nonetheless, in a former study conducted by the same author to identify predictors of outcome in 212 patients with BAO, the tendency to shift to a rescue maneuver was equivalent between CA and SR.37 In contrast, Gory et al 26 indicated that despite the significantly higher rate of complete revascularization in the CA technique, it necessitated a greater number of rescue maneuvers.
In agreement with Gory et al,26 CA in our study showed the highest rate of a switch to another EVT technique but with the lowest rate of use of rescue maneuvers. This might be explained by the inability of CA to reduce the thrombus burden or fully retrieve the clot while simultaneously it showed the lowest proportion of atherosclerotic pathology that frequently compels further interventions. On the other hand, combined SR +CA warranted use of another EVT technique in only 3%, compared with 26% of rescue maneuvers owing to the greatest proportion of ICAS lesions.
Periprocedural adverse events
Periprocedural complication rates vary throughout the literature. Based on the ETIS registry, CA achieved more successful revascularization in M1 occlusion than SR and combined SR+CA. For intracranial internal carotid artery, combined SR+CA was associated with higher odds of successful revascularization compared with SR. Nonetheless, higher revascularization rates did not translate into improved patient outcomes. Moreover, combined SR+CA was associated with a higher likelihood of sICH and mortality compared with CA.38
According to the HERMES meta-analysis, the risk of sICH, parenchymal hematoma type 2, and mortality rate following SR mediated thrombectomy were approximately 4.4%, 5.1%, and 15.3%, respectively.1 On the other hand, the COMPASS trial5 demonstrated an equivalent rate of intracranial hemorrhage and mortality following SR and CA techniques which was comparable with the meta-analysis conducted by Texakalidis39 but with a higher risk of vasospasm in SR, and greater propensity of subarachnoid hemorrhage in the combined SR +CA group compared with the CA group.
Parallel assumptions were stated by Kang et al 37 and Gory et al 26 in BAO but with a lower rate of procedural complications (collateral embolization, arterial dissection, and perforation) in CA in the latter study. In agreement with previous former studies, no significant difference was observed in our study between EVT techniques except for a higher proportion of non-symptomatic intracranial hemorrhage in SR compared with CA and combined SR +CA groups.
Limitations
This study is one of the largest comparative studies between extensively used revascularization strategies in patients with BAO pre-selected by MRI. However, it has several limitations, including the retrospective design with its inherent limitations. Second, the EVT device was determined at the discretion of the neurointerventionist and so selection bias cannot be excluded. Although no formal statistical sample size calculation was performed, MRI pre-selected consecutive patients with BAO at two comprehensive stroke centers were enrolled which was satisfactory to demonstrate dissimilarities in revascularization and outcome between the devices analyzed. Finally, our study was not designed to determine the optimum etiology tailored EVT technique that would induce complete revascularization. However, the combined SR+CA technique was associated with complete revascularization (aOR 3.03, 95% CI 1.116 to 8.222, p=0.03) regardless of the etiology (aOR 0.87, 95% CI 0.529 to 1.421, p=0.57).
Conclusion
Compared with other EVT approaches in the setting of acute BAO, the selection of the combined SR+CA technique might be advantageous as a frontline strategy owning to a comparable safety profile and higher rates of complete revascularization which constitutes the major predictor of a 90 day favorable outcome. A prospective randomized controlled trial is needed to further determine the best revascularization strategy, particularly with reference to the etiology of ischemic stroke in BAO patients.
Supplemental material
Supplemental material
Data availability statement
Data are available upon reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
The institutional review board approved the use of patient data for the research protocol.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Collaborators RAMBO group investigators: Mohamed Abdelrady, Imad Derraz, Pierre-Henri Lefevre, Federico Cagnazzo, Carlos Riquelme, Gregory Gascou, Mehdi Mahmoudi, Lucas Corti, Nicolas Gaillard, Mourad Cheddad El Aouni, Douraied Ben Salem, Cyril Dargazanli, Julien Ognard, Isabelle Mourand, Caroline Arquizan, Jean-Christophe Gentric, and Vincent Costalat.
Contributors Study design: MA, ID, and JO. Acquisition, analysis, or interpretation of the data: all authors. Drafting of the manuscript: MA, JO, ID, P-HL, FC, GG, and CD. Statistical analysis: MA and JO. Supervision: VC, JC-G, IM, and CA. Guarantor: MA, JO, VC, JC-G.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial, or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.