We read with interest the article by Soize et al. “Can early neurological improvement after mechanical thrombectomy be used as a surrogate for final stroke outcome?”[1] Based on their results, the authors concluded that early neurological improvement (ENI) 24 hours after thrombectomy is a straightforward surrogate of long-term outcome. However, all patients in this study were treated with conscious sedation (CS), and not general anesthesia (GA). The residual effects of GA may mask ENI and limit its utility as a surrogate for long-term outcome.[2]
We performed a similar analysis of patients enrolled in a prospective single-center registry. The ability of ENI to predict 3-month functional independence was assessed by the area under the receiver operating characteristic curve (AUC) and compared using the independent-samples Hanley test. Multivariable linear regression assessing the relationship between anesthetic technique and ENI was also performed. The analysis received ethics approval.
291 patients were treated with thrombectomy, with 261 (89.7%) procedures performed with GA, and 30 (10.3%) with CS. All patients were de-sedated and extubated more than 12 hours before 24-hour National Institutes of Health Stroke Scale assessment. 174 (59.8%) patients achieved 3-month functional independence. Baseline and procedural characteristics did not differ between GA and CS patients (all P>0.05). ENI demonstrated better prognostic ability in CS (AUC 0.91, 95% confiden...
We read with interest the article by Soize et al. “Can early neurological improvement after mechanical thrombectomy be used as a surrogate for final stroke outcome?”[1] Based on their results, the authors concluded that early neurological improvement (ENI) 24 hours after thrombectomy is a straightforward surrogate of long-term outcome. However, all patients in this study were treated with conscious sedation (CS), and not general anesthesia (GA). The residual effects of GA may mask ENI and limit its utility as a surrogate for long-term outcome.[2]
We performed a similar analysis of patients enrolled in a prospective single-center registry. The ability of ENI to predict 3-month functional independence was assessed by the area under the receiver operating characteristic curve (AUC) and compared using the independent-samples Hanley test. Multivariable linear regression assessing the relationship between anesthetic technique and ENI was also performed. The analysis received ethics approval.
291 patients were treated with thrombectomy, with 261 (89.7%) procedures performed with GA, and 30 (10.3%) with CS. All patients were de-sedated and extubated more than 12 hours before 24-hour National Institutes of Health Stroke Scale assessment. 174 (59.8%) patients achieved 3-month functional independence. Baseline and procedural characteristics did not differ between GA and CS patients (all P>0.05). ENI demonstrated better prognostic ability in CS (AUC 0.91, 95% confidence interval [CI], 0.80-1.00) than it did with GA treated patients (AUC 0.73, 95% CI, 0.67-0.80; P=0.008). Multivariable regression showed that GA was independently associated with attenuated ENI (P=0.03).
Our findings are in agreement with Soize et al, in that ENI does seem to predict long-term outcome in thrombectomy patients treated with CS, with comparable AUCs (0.91 and 0.93 respectively).[1] However, our results also suggest that ENI is worse at predicting long-term outcome following thrombectomy performed with GA. Furthermore, GA was independently associated with attenuated ENI, suggesting that GA might mask early neurologic recovery. These trends would appear to be in agreement with the SIESTA trial, which reported a greater likelihood of achieving 3-month functional independence in GA than CS patients, in the absence of significant differences in ENI.[3] Post-hoc analysis of SIESTA showed that propofol dose during thrombectomy was independently associated with reduced ENI.[2] The possible mechanisms for GA attenuating ENI may include the residual pharmacological effects of benzodiazepines, opioids, neuromuscular blockers, and intravenous or volatile anesthetic agents, or transient complications of endotracheal intubation such as ventilator-associated complications.[4]
[1] Soize S, Fabre G, Gawlitza M, et al. Can early neurological improvement after mechanical thrombectomy be used as a surrogate for final stroke outcome? J. Neurointerv. Surg. 2019;11(5):450-454.
[2] Schönenberger S, Uhlmann L, Ungerer M, et al. Association of Blood Pressure With Short- and Long-Term Functional Outcome After Stroke Thrombectomy: Post Hoc Analysis of the SIESTA Trial. Stroke. 2018;49:1451–1456.
[3] Schönenberger S, Uhlmann L, Hacke W, et al. Effect of Conscious Sedation vs General Anesthesia on Early Neurological Improvement Among Patients With Ischemic Stroke Undergoing Endovascular Thrombectomy: A Randomized Clinical Trial. JAMA. 2016;316:1986–1996.
[4] Sinclair RCF, Faleiro RJ. Delayed recovery of consciousness after anaesthesia. Continuing Education in Anaesthesia, Critical Care & Pain. 2006;6:114–118.
We read with interest the response to our manuscript on using machine learning to optimize elderly patient selection for endovascular thrombectomy (1). We acknowledge here, as the author reports, the limitation of SPOT being based on single center data, and the need for multicenter prospective validation of SPOT as next step in development. The author raises additional technical concerns that we do not necessarily view as applicable to this study.
First, we would like to stress the general limitations of artificial intelligence based techniques such as the overfitting and the data specific local optima problems. However, the specific comments brought by the author are not applicable in our case. First, studies on the number of events per predictor are applicable for logistic regressions (LRs) which is not used in the SPOT algorithm. In fact, our results show poor LR performance which is consistent with the rule of thumb of 1 to 10 referred to by the author. Hence, while serving as a good guidance for LR, the rule is not binding and more importantly it does not guarantee the generalization of the learned model. To further illustrate, classification models using convolutional neural networks have millions of parameters and are trained with datasets that, in most cases, do not have millions of samples in each group. However, these models have acceptable generalization capabilities and are tested using the data-split method. In SPOT, the model at its core is a regressi...
We read with interest the response to our manuscript on using machine learning to optimize elderly patient selection for endovascular thrombectomy (1). We acknowledge here, as the author reports, the limitation of SPOT being based on single center data, and the need for multicenter prospective validation of SPOT as next step in development. The author raises additional technical concerns that we do not necessarily view as applicable to this study.
First, we would like to stress the general limitations of artificial intelligence based techniques such as the overfitting and the data specific local optima problems. However, the specific comments brought by the author are not applicable in our case. First, studies on the number of events per predictor are applicable for logistic regressions (LRs) which is not used in the SPOT algorithm. In fact, our results show poor LR performance which is consistent with the rule of thumb of 1 to 10 referred to by the author. Hence, while serving as a good guidance for LR, the rule is not binding and more importantly it does not guarantee the generalization of the learned model. To further illustrate, classification models using convolutional neural networks have millions of parameters and are trained with datasets that, in most cases, do not have millions of samples in each group. However, these models have acceptable generalization capabilities and are tested using the data-split method. In SPOT, the model at its core is a regression model with continuous output. More importantly, while the overfitting concern is a valid one with the high area under a curve, SPOT was tested in an adequate fashion using data-splitting, a well-accepted validation scheme to test the generalization capability of the model, and thus detect overfitting. In fact, studies discussing events per variable rule for logistic regression use data-splitting as one of the validation method (2). The approach used for testing SPOT using a prospective data not part of the training is the most stringent approach to test a model. In fact, state of the art machine learning model evaluations use the data-splitting technique, and consider the performance of the testing data as the golden metric to judge upon the model’s generalization and over-fitting (3,4).
While we agree with the fact that machine learning are data hungry, the size of the dataset is highly dependent on the problem at hand. For example, in sparse regression models, the number of samples in the training dataset can be orders of magnitude smaller than that of the parameters. However, sparse regression models joined with proper training techniques are able to generalize to unseen data (5). Again, this ability is tested using the data-split method which was used to test SPOT.
Further, we emphasize that although SPOT returns a continuous output of mRS scores, the tool will specifically report grouped outcomes into mRS 0-2 and mRS 3-6. The decision to choose this dichotomy of outcomes was to be consistent with clinical trials that predominantly report functional independence as outcome measure to guide interventions even when elderly patients were included. In response to the concern about returning probability for outcomes, and as stated in the manuscript, when SPOT returns poor outcome prediction, its negative predictive value was 95.2% which represents the probability for a patient that screened negative to have a poor outcome and this probability is reported in the text. In the current form, SPOT does not return a probability for every mRS score.
Finally, while we agree that multicenter data is needed to additionally validate SPOT as a tool as stated in the manuscript, the current version of SPOT does meet technical requirements for a validated tool. We do stress again that “SPOT is designed to aid clinical decision of whether to undergo ET in elderly patients” (1), and not a stand-alone tool.
References
1. Alawieh A, Zaraket F, Alawieh MB, Chatterjee AR, Spiotta A. Using machine learning to optimize selection of elderly patients for endovascular thrombectomy. J Neurointerv Surg. 2019.
2. Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361-87.
3. Kohavi, Ron. "A study of cross-validation and bootstrap for accuracy estimation and model selection." Ijcai. Vol. 14. No. 2. 1995.
4. Arlot, Sylvain, and Alain Celisse. "A survey of cross-validation procedures for model selection." Statistics surveys4 (2010): 40-79.
5. Donoho, David Leigh, et al. Sparse solution of underdetermined linear equations by stagewise orthogonal matching pursuit. Department of Statistics, Stanford University, 2006.
Congratulations to Annika Keuler et al¹ on their experience with the wireless microcatheter technique preventing vessel perforations in endovascular thrombectomy. Based on their results, the authors conclude that in most cases of mechanical recanalization, the clot can be passed more safely with a wireless microcatheter. In our daily work, we also find the wireless microcatheter technique seems to reduce subarachnoid hyperdensity resulting from vessel perforations. However it seems difficult to confirm this correlation; the details of which will be discussed as follows. After reading and analyzing the article carefully, we have some opinions about the study which we would like to communicate with the authors because the conclusions of the paper directly relate to our clinical experience.
In the article, two radiological manifestations are defined as vessel perforations——contrast extravasation during angiography and angiographically occult ipsilateral circumscribed subarachnoid contrast extravasation which is identified by post-interventional CT scans. As confirmed by previous studies2-3, we agree with the authors on using immediate post-interventional CT examination to identify the subarachnoid hyperdensity due to intraoperative contrast extravasation. Based on their results, post-thrombectomy subarachnoid hyperdensity was observed on CT scans in 22 patients, in 18 of whom, the clot was passed using a microwire, and in the other four, using a wireless microcathete...
Congratulations to Annika Keuler et al¹ on their experience with the wireless microcatheter technique preventing vessel perforations in endovascular thrombectomy. Based on their results, the authors conclude that in most cases of mechanical recanalization, the clot can be passed more safely with a wireless microcatheter. In our daily work, we also find the wireless microcatheter technique seems to reduce subarachnoid hyperdensity resulting from vessel perforations. However it seems difficult to confirm this correlation; the details of which will be discussed as follows. After reading and analyzing the article carefully, we have some opinions about the study which we would like to communicate with the authors because the conclusions of the paper directly relate to our clinical experience.
In the article, two radiological manifestations are defined as vessel perforations——contrast extravasation during angiography and angiographically occult ipsilateral circumscribed subarachnoid contrast extravasation which is identified by post-interventional CT scans. As confirmed by previous studies2-3, we agree with the authors on using immediate post-interventional CT examination to identify the subarachnoid hyperdensity due to intraoperative contrast extravasation. Based on their results, post-thrombectomy subarachnoid hyperdensity was observed on CT scans in 22 patients, in 18 of whom, the clot was passed using a microwire, and in the other four, using a wireless microcatheter. The authors concluded that the complication rate for post-thrombectomy hyperdensity was significantly higher when a microwire was used to pass the clot. However, Omid Nikoubashman et al² report that post-interventional subarachnoid hyperdensities are associated with a long interval between clinical onset and recanalization, a long procedure time, and a high number of recanalization attempts. Perry P Ng et al ³ also mention that an increased number of stent retriever passes, distal device positioning, and presence of severe vasospasm were associated with post-thrombectomy subarachnoid hyperdensity. Additionaly, the interventional neuroradiologists used a microwire to pass the clot when first-pass microcatheter passage was not successful. There is confounding bias in this practice itself——It could be more difficult to pass the clot and need more stent retriever attempts in the microwire use group.
To sum up, hyperdensity on immediate post-thrombectomy CT scans, a manifestation of vessel perforation, is associated with many factors. The conclusion that the wireless technique can reduce post-thrombectomy hyperdensity would be more convincing if the authors ruled out the association between subarachnoid contrast extravasation and potential risk factors.
Yuan Ma, Pei-Cheng Li, Long Chen
Department of Interventional Radiology, The First Affiliated Hospital of Soochow University, Suzhou, China.
Correspondence to Dr. Long Chen, Department of Interventional Radiology, The First Affiliated Hospital of Soochow University, 188 Shizi Street,215006 Suzhou, China；lchen76@163.com
References:
1 Keulers A, Nikoubashman O, Mpotsaris A, et al. Preventing vessel perforations in endovascular thrombectomy: feasibility and safety of passing the clot with a microcatheter without microwire: the wireless microcatheter technique. J Neurointerv Surg 2018: 2018-14267.
2 Nikoubashman O, Reich A, Pjontek R, et al. Postinterventional subarachnoid haemorrhage after endovascular stroke treatment with stent retrievers. Neuroradiology 2014;56: 1087-1096.
3 Ng PP, Larson TC, Nichols CW, et al. Intraprocedural predictors of post-stent retriever thrombectomy subarachnoid hemorrhage in middle cerebral artery stroke. J Neurointerv Surg 2018;11: 127-132.
It is with great interest that we read the study of Alawieh et al(1), in which they developed a machine learning algorithm, called ‘SPOT’, to select stroke patients older than 80 years for endovascular therapy (EVT). Prediction modeling to optimize patient selection for EVT is an emerging topic of interest and we agree that predicting individual patient outcomes is increasingly important for decision making in medicine. However, we were surprised by the strong conclusions that were drawn by the authors, considering some serious limitations of the study.
First, the size of the training set is insufficient to develop a complex model with twelve predictor variables and many correlations. Only 22 patients had a good functional outcome, which means that the number of events per tested predictor variable is less than two. For the development of a reliable model, a sample size of at least ten events per variable is needed to minimize the risk of overfitting(2, 3). It has been suggested that even far more events per variable are needed to achieve stable predictions with machine learning techniques(4). Especially complex models developed on small sample sizes have a high risk of overfitting, resulting in unstable predictions and too optimistic model performance measures. The reported AUC of 0.92 is therefore very likely to be an overestimation.
Second, the SPOT algorithm provides a treatment advice based on the predicted outcome after treatment, without providing the...
It is with great interest that we read the study of Alawieh et al(1), in which they developed a machine learning algorithm, called ‘SPOT’, to select stroke patients older than 80 years for endovascular therapy (EVT). Prediction modeling to optimize patient selection for EVT is an emerging topic of interest and we agree that predicting individual patient outcomes is increasingly important for decision making in medicine. However, we were surprised by the strong conclusions that were drawn by the authors, considering some serious limitations of the study.
First, the size of the training set is insufficient to develop a complex model with twelve predictor variables and many correlations. Only 22 patients had a good functional outcome, which means that the number of events per tested predictor variable is less than two. For the development of a reliable model, a sample size of at least ten events per variable is needed to minimize the risk of overfitting(2, 3). It has been suggested that even far more events per variable are needed to achieve stable predictions with machine learning techniques(4). Especially complex models developed on small sample sizes have a high risk of overfitting, resulting in unstable predictions and too optimistic model performance measures. The reported AUC of 0.92 is therefore very likely to be an overestimation.
Second, the SPOT algorithm provides a treatment advice based on the predicted outcome after treatment, without providing the absolute probability of good functional outcome or the treatment benefit. Rational treatment decisions should be based on expected outcome with treatment compared to the expected outcome without treatment. A low likelihood of good outcome does not imply absence of treatment benefit. Besides that, many octogenarians might not be able to achieve complete recovery to functional independence, but an improvement from an mRS score of 4-5 to a score of 3 as a result of EVT can still be very relevant in clinical practice, and such outcome is not covered by the SPOT algorithm.
Well-developed prediction models may guide us in the selection of patients that benefit from treatment, but external validation in a large validation set is always needed before these models are implemented in everyday clinical care. The SPOT algorithm does not yet fulfill the minimum requirements for a well-developed and validated decision support tool. This means that an effective treatment may be withheld from patients who could benefit from it. Therefore, the SPOT algorithm should not yet be implemented in clinical care.
References
1. Alawieh A, Zaraket F, Alawieh MB, Chatterjee AR, Spiotta A. Using machine learning to optimize selection of elderly patients for endovascular thrombectomy. J Neurointerv Surg. 2019.
2. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996;49(12):1373-9.
3. Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361-87.
4. van der Ploeg T, Austin PC, Steyerberg EW. Modern modelling techniques are data hungry: a simulation study for predicting dichotomous endpoints. BMC Med Res Methodol. 2014;14:137.
We had an opportunity to read the article by Lakomkin et al regarding systematic literature review of LVO prevalence. Since one of our studies is part of this review we feel compelled to comment on the paper. We do appreciate the authors’ efforts in conducting this analysis which is important in understanding the burden of disease – but, with respect offer some criticisms. The major limitation of the paper which the authors recognize is the heterogeneity of the included studies. Unfortunately, this limitation is so critical that it yields unreliable information at best and misleading at worst.
The paper intends to study the prevalence of large vessel strokes. However, apart from a couple of population based studies in their review, the rest are a heterogenous mix describing an LVO rate from very selective cohorts of patients from single centers. Several are centered around validation of clinical scales for detecting LVOs. The key features of a population based study include a defined catchment population, access to a large part of that population and a reliable marker of disease. Without these a “prevalence” constitutes a report of a center’s experience of disease rate as it pertains to their patient intake. While still valuable it is not an estimation of the disease burden in the population that the center serves unless an overwhelming majority of that population comes to that center.
The authors determine an average rate of about 30% LVO amongst acute isch...
We had an opportunity to read the article by Lakomkin et al regarding systematic literature review of LVO prevalence. Since one of our studies is part of this review we feel compelled to comment on the paper. We do appreciate the authors’ efforts in conducting this analysis which is important in understanding the burden of disease – but, with respect offer some criticisms. The major limitation of the paper which the authors recognize is the heterogeneity of the included studies. Unfortunately, this limitation is so critical that it yields unreliable information at best and misleading at worst.
The paper intends to study the prevalence of large vessel strokes. However, apart from a couple of population based studies in their review, the rest are a heterogenous mix describing an LVO rate from very selective cohorts of patients from single centers. Several are centered around validation of clinical scales for detecting LVOs. The key features of a population based study include a defined catchment population, access to a large part of that population and a reliable marker of disease. Without these a “prevalence” constitutes a report of a center’s experience of disease rate as it pertains to their patient intake. While still valuable it is not an estimation of the disease burden in the population that the center serves unless an overwhelming majority of that population comes to that center.
The authors determine an average rate of about 30% LVO amongst acute ischemic stroke (AIS) patients based on their review. The critical factor here is the denominator, i.e. the number of AIS patients from which the LVO rate is derived. It can be misrepresentative to extrapolate disease rate to a larger denominator derived from a different methodology. For instance, the oft quoted denominator of about 700,000 ischemic strokes is based on population studies using very specific ICD discharge codes in well-defined populations. Examples include the GCNKSS and the BASIC projects. Unless a study uses the same discharge codes in its methodology for determining the AIS denominator it cannot transplant LVO percentages calculated from its selective cohort to the larger denominator and derive absolute numbers of disease. For instance, a 30% LVO rate in a cohort of patients assessed as having an ischemic stroke by the EMS is not the same as 30% LVO rate amongst all AIS based on specific ICD discharge codes. Using the percentage from one cohort and applying to a different denominator could yield inaccurate absolute numbers.
The study from our center that is included in this analysis was designed to capture the LVO incidence in a unique well defined population of which the vast majority (85%) was served by our hospital system based on each county data reported to the DHHR. Thus, similar to the studies used to derive the total AIS estimates (e.g. GCNKSS, BASIC) we had a defined population, had significant access to the population and used the same ICD discharge codes to determine the denominator as in those studies. We also used a reliable marker of LVO, i.e. CTA performed on every AIS patient as identified by these codes. In our first paper an LVO was defined as ICA-T, MCA and BA to restrict it predominantly to the occlusion sites considered as LVO in the major clinical trials. In our second paper, not included in this analysis, we separately determined an incidence of M2 occlusions and combined with ICA-T, M1, BA this yields a rate of 16% (95CI 14-17). In our population, this gives an incidence of 31 (95CI 26-35)/100,000 people/year. Our second paper also includes a chart showing the location of occluded sites for other vessels not considered as LVOs i.e. ACA, PICA, SCA, PCA etc.
The authors comment on our inclusion of TIA codes in the denominator. We did that because it had been part of previous population studies evaluating AIS incidences. TIA comprised at most 1% of our denominator and if we exclude these patients, the incidence of 16% LVO does not change by more than a percentage point. Another variable to consider in a single center report is that a tertiary level comprehensive center may get transfers of sicker stroke patients from referring hospitals which can further skew the denominator and hence the derived LVO rate.
In summary, it is important to differentiate between population studies and single center experiences – especially when compiling these in a systematic review. It is critical when extrapolating disease incidence from one center’s report to the national disease burden that the methodology of deriving the larger denominator is the same. Nonetheless we do appreciate this review and commend the authors on their efforts. This highlights the need for collaborative efforts to collect population data based on a uniform methodology. These efforts should include state and federal registers that collect health data based on specific disease codes. Such collective ventures can provide more realistic estimates of the LVO burden and help shape systems of care.
We would like to congratulate Nicholson et al. on their highly interesting work on the declining rate of SAH in the Irish population. This will certainly provide some very interesting points. Also in Germany there is - at least subjectively - the phenomenon of the declining rate of SAH. The authors can establish a clear correlation to the decline in the smoking rate. Now the question arises whether this is the only relevant correlation. In particular, it would certainly be necessary to investigate whether there has been an increased rate of detection of unruptered Aneurysma and an increasing rate of treatment of those during the study period and whether this may also have a relevant influence on the decrease in SAH.
Dear Editor,
we read with great interest the paper from Sallustio et al 1 regarding the use of new thromboaspiration catheter, AXS Catalyst 6 (Stryker Neurovascular, Mountain View, CA, USA), for endovascular treatment (EVT) of large vessel stroke (LVS) with A Direct Aspiration first Pass Technique (ADAPT)2.
In our center, a team composed by 4 vascular interventional radiologists, two physicians with certified experience in stroke treatment and two physicians with large carotid stent experience, and 4 stroke neurologist with large experience in intravenous thrombolysis, started to perform EVT in patients with LVS of anterior or posterior circulation from September 2017.
Given the wide availability of different systems of neurothrombectomy we decided to use AXS Catalyst 6 both for its technical features, as reported by Sallustio et al, both for its lower costs than the others available (6F SOFIA plus catheter, MicroVention, Tustin, CA, USA; the X Penumbra ACE catheters, Penumbra Inc., Alameda, CA, USA).
Between September 2017 and May 2018, 24 patients (72.1 ± 13.2 years old) affected by acute ischemic stroke with LVS underwent to EVT in our center. Median baseline NIHSS was 18 (range: 7-24). Intravenous thrombolysis was used in 5 patients.
The most frequent site of occlusion was the middle cerebral artery (MCA) (70.8%), while in 16.7% of cases was basilar artery. Tandem occlusions occurred in 12.5% of patients and the most frequent stroke etiolo...
Dear Editor,
we read with great interest the paper from Sallustio et al 1 regarding the use of new thromboaspiration catheter, AXS Catalyst 6 (Stryker Neurovascular, Mountain View, CA, USA), for endovascular treatment (EVT) of large vessel stroke (LVS) with A Direct Aspiration first Pass Technique (ADAPT)2.
In our center, a team composed by 4 vascular interventional radiologists, two physicians with certified experience in stroke treatment and two physicians with large carotid stent experience, and 4 stroke neurologist with large experience in intravenous thrombolysis, started to perform EVT in patients with LVS of anterior or posterior circulation from September 2017.
Given the wide availability of different systems of neurothrombectomy we decided to use AXS Catalyst 6 both for its technical features, as reported by Sallustio et al, both for its lower costs than the others available (6F SOFIA plus catheter, MicroVention, Tustin, CA, USA; the X Penumbra ACE catheters, Penumbra Inc., Alameda, CA, USA).
Between September 2017 and May 2018, 24 patients (72.1 ± 13.2 years old) affected by acute ischemic stroke with LVS underwent to EVT in our center. Median baseline NIHSS was 18 (range: 7-24). Intravenous thrombolysis was used in 5 patients.
The most frequent site of occlusion was the middle cerebral artery (MCA) (70.8%), while in 16.7% of cases was basilar artery. Tandem occlusions occurred in 12.5% of patients and the most frequent stroke etiology resulted to be cardioembolic (64%).
Time from stroke onset-to-arterial puncture was 213.5 ± 72.0 min
All procedures of patients with LVS that involved anterior circulation, were performed at first in conscious sedation, while 2 patients with involvement of the posterior circulation was treated under general anesthesia.
After catheterizing the common femoral artery with 8Fr introducer, a long sheath introducer (AXS Infinity LS Stryker Neurovascular, Mountain View, CA, USA) was positioned in the distal common carotid artery, proximal cervical ICA, or V1 segment of the vertebral artery.
Therefore, contrary to Sallustio et al, Catalyst 6 catheter was always advanced to the site of occlusion by creating a triaxial system, over a 0.014” guidewire (Transend Stryker, Neurovascular, Mountain View, CA, USA), and over a dedicated coaxial microcatheter (Offset, Stryker Neurovascular, Mountain View, CA, USA). Microcatheter over guidewire, was carefully advanced as close to the thrombus as possible, avoiding crossing it.
When Catalyst 6 catheter tip was touching the thrombus, guidewire and microcatheter were removed, and aspiration was started using always a dedicated vacuum pump.
When no-flow was obtained from aspiration system, Catalyst 6 catheter was slowly pushed forward, and aspiration continued during catheter removal until blood was obtained in the aspiration line or total removal with additional aspiration from the Infinity LS.
Using this approach, we experienced a median procedural time of 52 minutes (range: 20 min to 169 min) with slightly longer time compared to Sallustio et al, maybe because the use of a triaxial system that, at first, could be unwieldy. As assessable, in relation to the low number of cases, we did not find any differences in the procedural times between procedures performed by the different four vascular interventional radiologists.
During procedures we performed at least 3 attempts at revascularization using ADAPT technique before switching to stent retriever technique with Trevo stent (Stryker Neurovascular, Mountain View, CA, USA). We used stent retrieve technique associated with aspiration only in 2 cases (8.3 %). Time from door to reperfusion was 172.0 ± 45.2 min.
We obtained a successful recanalization (TICI 2b/3) in 21 patients (87.5%). Symptomatic intracranial hemorrhage occurred in 1 patient. Functional independence was obtained in 70.8% and mortality occurred in 3 patients.
Our technical success is comparable to results reported by Sallustio et al, although in a less numerous study population.
In our experience, this triaxial system, composed of AXS Infinity LS, Catalyst 6 and Offset microcatheter, surely facilitates performing EVT even when used by “non-expert thrombectomy performers”.
Offset microcatheter helps intermediate catheter ophthalmic passage, avoiding exaggerated thrusts of the catheter that could cause buckling of Catalyst 6, arterial spasms or dissections. Indeed, no arterial dissections or flow limiting arterial vasospasm were identified among 24 interventions.
However, using a microcatheter it could occur that it be passed through the clot and therefore might result in a rate of distal emboli. In our initial experience, we observed only 1 case of thrombus fragmentation with distal M2 branches embolization, due to passing by Offset microcatheter through the thrombus, resulting in TICI 2b revascularization.
In our experience, Catalyst 6 triaxial ADAPT technique was safe, technically feasible, rapid, and effective in patients with LVS.
In ischemic stroke care, fast reperfusion is essential to improve disability free survival and due to the serious paucity of thrombectomy performers, we believe that continuous development of technology of aspiration catheters and microcatheter may reduce procedure time allowing that the neurothrombectomy procedure may be within the reach of vascular interventional radiologists.
References
1. Sallustio, F. et al. Mechanical thrombectomy of acute ischemic stroke with a new intermediate aspiration catheter: preliminary results. J. Neurointerv. Surg. neurintsurg-2017-013679 (2018). doi:10.1136/neurintsurg-2017-013679
2. Turk, A. S. et al. ADAPT FAST study: a direct aspiration first pass technique for acute stroke thrombectomy. J. Neurointerv. Surg. 6, 260–264 (2014).
The paper by Farzin et al.[1] shows interesting results about measuring porosity of fully expanded flow diverter stents using (photographic) images of the stent being assessed. In their study, authors used 3 different methods and repeated measurements by different observers to assess the porosity of stents. According to their results, the variability when measuring porosity is so large that previous works assessing it should be questioned. On the other hand, they indicate that pore density seems to be more reliable and repeatable. The study highlights the difficulty of measuring such parameter in a controlled in vitro environment. After carefully reading the article, it became clear that the most reproducible way of measuring porosity, from the 3 options studied, was M3 (based on measuring the width and length of the struts and number of struts per reference square). Furthermore, some simple assumptions should improve the results and substantially reduce errors and variability:
1. Wire width: the value for wire width, indicated by the manufacturer, is likely to be more accurate. If this value is no to be trusted, at least in average, then the reproducibility of the manufacturing process could not be trusted. Measuring wire width directly on the images is likely to introduce error as it might be affected by reflection/refraction of light on the wire material and wire coating, as well as lens imperfections or optical aberrations in some cases.
2. Calculating poro...
The paper by Farzin et al.[1] shows interesting results about measuring porosity of fully expanded flow diverter stents using (photographic) images of the stent being assessed. In their study, authors used 3 different methods and repeated measurements by different observers to assess the porosity of stents. According to their results, the variability when measuring porosity is so large that previous works assessing it should be questioned. On the other hand, they indicate that pore density seems to be more reliable and repeatable. The study highlights the difficulty of measuring such parameter in a controlled in vitro environment. After carefully reading the article, it became clear that the most reproducible way of measuring porosity, from the 3 options studied, was M3 (based on measuring the width and length of the struts and number of struts per reference square). Furthermore, some simple assumptions should improve the results and substantially reduce errors and variability:
1. Wire width: the value for wire width, indicated by the manufacturer, is likely to be more accurate. If this value is no to be trusted, at least in average, then the reproducibility of the manufacturing process could not be trusted. Measuring wire width directly on the images is likely to introduce error as it might be affected by reflection/refraction of light on the wire material and wire coating, as well as lens imperfections or optical aberrations in some cases.
2. Calculating porosity per strut: using simple geometrical computations is possible measuring the amount of covered area divided by total area, namely porosity, for the rectangular region enclosed by the extremes of one strut (Figure 1 - https://www.dropbox.com/s/oxmbrbs2kv1fcov/Figure%201.pdf?dl=0). This area is equivalent to half the area of one stent cell (i.e. enclosed by 4 neighbouring struts) and the covered area is the rectangular region diagonal multiplied by the wire width, and then subtracting the overlapping region, called Area_open in Figure 1.
3. Calculating pore density per strut: equivalently, pore density can be computed for each strut. In this scenario, partial struts should not be considered. Each strut accounts for half a pore, then a simple division and transformation on the local porosity will yield a local estimation of pore density, as described in Figure 1.
4. Regional measurements are largely affected by FD shape and viewing angle: the further away from the point perpendicular position to the camera view point the measure is performed, the least reliable the measurement will be. Unavoidably, a reference rectangle will consider a region that is affected by this camera view point change. Under such conditions and without correcting the view angle position, only a small region of the FD nearby the region perpendicular to the camera should be considered. A correction to the proposed parameters, considering the angle between the mesh and the view point, should be considered.
The subject of FD porosity and its relation to aneurysm occlusion is recurrently mentioned the literature but so far this subject has not been properly addressed so far. Despite its relevance and implications in treatment selection and local hemodynamics [2], its computation in vivo has not been largely stuied. The work of Boulliot et al.[3] nicely addresses this matter on stents with multiple visible wires, providing an elegant approach to measuring porosity in treated patients as well as the means to simulate the behavior of flow diverters from their geometrical description. Fernandez et al. also proposed and validated a simulation technique with such capabilities[4]. Current medical imaging modalities (3DRA, Dyna-CT/Exper-CT, CTA) provide a resolution that allows visualizing individual wires in FD stents. The combination of improved imaging resolution, improved visibility of flow diverter wires in X-ray images, advanced imaging and simulation algorithms is likely to allow assessment of porosity and pore density in vivo, in the near future.
Ignacio Larrabide, DSc.
Pladema Institute - CONICET – UNICEN
Tandil, Argentina
Demetriud Lopes, PhD. Dr.
RUSH University Hospital
Chicago, IL, USA
References:
1 Farzin, B., Brosseau, L., Jamali, S., Salazkin, I., Jack, A., Darsaut, T. E., & Raymond, J. (2015). Flow diverters: inter and intra-rater reliability of porosity and pore density measurements. Journal of neurointerventional surgery, 7(10), 734-739.
2 Mut, F., & Cebral, J. R. (2012). Effects of flow-diverting device oversizing on hemodynamics alteration in cerebral aneurysms. American Journal of Neuroradiology, 33(10), 2010-2016.
3 Bouillot, P., Brina, O., Ouared, R., Yilmaz, H., Farhat, M., Erceg, G., ... & Pereira, V. M. (2016). Geometrical deployment for braided stent. Medical image analysis, 30, 85-94.
4 Fernandez, H., Macho, J. M., Blasco, J., San Roman, L., Mailaender, W., Serra, L., & Larrabide, I. (2015). Computation of the change in length of a braided device when deployed in realistic vessel models. International journal of computer assisted radiology and surgery, 10(10), 1659-1665.
I read with great interest the meta-analysis by Brinjikji et al.1 which evaluated outcomes after mechanical thrombectomy for acute ischemic stroke by using a balloon guiding catheter (BGC) device. In that study, the authors documented that patients who underwent mechanical thrombectomy with BGC had better clinical and angiographic outcomes than those without BGC. However, there were some issues which should be addressed and discussed.
First, the number of successful recanalizations, shown as 2b/3 of Thrombolysis In Cerebral Infarction (TICI) grade in Fig.3 in the article,1 might be not accurately described. The events of successful recanalization were noted in 113 of 149 in the BGC group and 133 of 189 in the non-BGC group according to Nguyen et al.2 However, the events were presented as 112 of 149 in the BGC group and 135 of 189 in the non-BGC group.1 Accordingly, the forest plot can be changed as in Fig. 1 below. Mechanical thrombectomy using BGC exhibited significantly higher successful recanalizations than did non-BGC use (OR, 1.710; 95% CI: 1.099-2.662). Second, there was no specific explanation for the publication bias of Fig. 4 in the result section.1 Although the authors reported a p value of 0.49 using Egger’s regression, we are not sure what publication bias meant to represent, successful recanalization or clinical outcome or other variables.
In this letter, we made a funnel plot for successful recanalization based on the revised number of events we h...
I read with great interest the meta-analysis by Brinjikji et al.1 which evaluated outcomes after mechanical thrombectomy for acute ischemic stroke by using a balloon guiding catheter (BGC) device. In that study, the authors documented that patients who underwent mechanical thrombectomy with BGC had better clinical and angiographic outcomes than those without BGC. However, there were some issues which should be addressed and discussed.
First, the number of successful recanalizations, shown as 2b/3 of Thrombolysis In Cerebral Infarction (TICI) grade in Fig.3 in the article,1 might be not accurately described. The events of successful recanalization were noted in 113 of 149 in the BGC group and 133 of 189 in the non-BGC group according to Nguyen et al.2 However, the events were presented as 112 of 149 in the BGC group and 135 of 189 in the non-BGC group.1 Accordingly, the forest plot can be changed as in Fig. 1 below. Mechanical thrombectomy using BGC exhibited significantly higher successful recanalizations than did non-BGC use (OR, 1.710; 95% CI: 1.099-2.662). Second, there was no specific explanation for the publication bias of Fig. 4 in the result section.1 Although the authors reported a p value of 0.49 using Egger’s regression, we are not sure what publication bias meant to represent, successful recanalization or clinical outcome or other variables.
In this letter, we made a funnel plot for successful recanalization based on the revised number of events we had corrected. However, the funnel plot showed an asymmetry indicative of possible publication bias. To resolve publication bias in deciding on successful recanalization, we trimmed one study. After correction of the forest plot, the adjusted OR was 1.430 (95% CI: 0.852-2.400), suggesting that there was no significant relationship between BGC use and a higher successful recanalization rate (Fig. 2). We think that heterogeneity across studies, in particular thrombus location, can explain the controversial results. Although we did not have accurate information about thrombus location in the studies, which were conference abstract,3-5 difference in posterior circulation may bias the result. For example, we made an additional forest plot based on published articles including two studies6,7 that were not enrolled in the previous meta-analysis1 (Fig. 3A). Our study also demonstrated a higher successful recanalization rate in the BGC group than in the non-BGC group (OR, 2.324; 95% CI: 1.228-4.396). However, possible publication bias was noted, and adjusted OR, after trimming two studies, was 1.630 (95% CI: 0.870-3.053) (Fig.3B). Consequently, we did subgroup analysis for studies that included only anterior circulation stroke patients. Subgroup analysis revealed that the BGC group exhibited a significantly higher successful recanalization rate without heterogeneity (OR, 3.187; 95% CI: 1.797-5.652, Fig. 4A). In addition, publication bias was not observed via the funnel plot (Fig. 4B) or Egger’s regression (p=0.979). Therefore, we think that clot location should be considered to interpret the results.
Per your comments, when using a BGC, some neurointerventionists including us (JHA and JPJ) can be reluctant to use BGC for patients with difficult arches or have concerns about larger 8Fr or 9Fr groin sheath due to complications. As your conclusion, we hope that further randomized trials can give an answer about the BGC feasibility during mechanical thrombectomy for acute stroke patients.
References
1. Brinjikji W, Starke RM, Murad MH, et al. Impact of balloon guide catheter on technical and clinical outcomes: A systematic review and meta-analysis. J Neurointerv Surg. 2018;10:335-9.
2. Nguyen TN, Malisch T, Castonguay AC, et al. Balloon guide catheter improves revascularization and clinical outcomes with the solitaire device: Analysis of the north american solitaire acute stroke registry. Stroke. 2014;45:141-5.
3. Nguyen TN, Castonguay AC, Nogueira RN, et al. Balloon guide catheter improved clinical outcomes, revascularization, and decreased mortality in Trevo thrombectomy. Analysis of the TREVO Stent Retreiver acute stroke (TRACK) Registry. Presentated at the Society of Vascular Interventional Neurology Conference. 2015
4. Zaidat O, Froehler MT, Aziz-Sulta MA, et al. Influce of balloon, conventional or distal catheters on angigraphic and clinical otucomes in the Stratis Registry. Stroke. 2017
5. Zaidat O, Liebeskind D, Jahan R, et al. Influce of balloon, conventional, or distal catheters on angigraphic and technical otucomes in STRATIS. J Neurointerv Surg. 2016.
6. Lee DH, Sung JH, Kim SU, et al. Effective use of balloon guide catheters in reducing incidence of mechanical thrombectomy related distal embolization. Acta Neurochir (Wien). 2017;159:1671-7.
7. Oh JS, Yoon SM, Shim JJ, et al. Efficacy of balloon-guiding catheter for mechanical thrombectomy in patients with anterior circulation ischemic stroke. J Korean Neurosurg Soc. 2017;60:155-64.
Figure legends
Figure 1. Revised forest plot of successful recanalization events according to the use of a balloon guiding catheter (BGC) during mechanical thrombectomy.
Figure 2. Funnel plots of the unadjusted and adjusted effect estimates after correction of publication bias using the “trim and fill” method for successful recanalization. The white circles indicate individual original studies, and the white diamond is the odds ratio (OR) and 95% confidence interval for the meta-analysis. The data point for imputed studies are highlighted in black, and the pooled effect by the recomputation is the black diamond.
Figure 3. A, Comparisons of mechanical thrombectomy for successful recanalization between the BGC group vs. the non-BGC group, based on published full-text articles. B, Funnel plots of the unadjusted and adjusted effect estimates after correction of publication bias using the “trim and fill” method.
Figure 4. A, Subgroup analysis of successful recanalization between the BGC group vs. the non-BGC group, based on published full-text articles that included only anterior circulation stroke. B, Funnel plots of the publication bias.
TO THE EDITOR: We read with interest the recent paper by Boned and colleagues.1 The authors conclude that “CT perfusion may overestimate final infarct core, especially in the early time window. Selecting patients for reperfusion therapies based on the CTP mismatch concept may deny treatment to patients who might still benefit from reperfusion”. We completely agree with this consideration, mainly when, as in this article, the core volume is assessed according to the classical CT perfusion (CTP) mismatch mean transit time (MTT)/cerebral blood volume (CBV)2 by measuring the lesion on CBV maps generated with a one-phase CT perfusion (CTP) acquisition protocol. In fact, it is well-known that a short CTP scan duration often produces a truncation of the perfusion curves resulting in an overestimation of CBV lesion that can frequently reverse.3 In addition, it has recently been demonstrated that relative cerebral blood flow (CBF) < 30% and time to peak of the residual function (Tmax) > 6 seconds is more reliable than CBV < 2.0 ml/100gr and relative MTT > 145% in identifying infarct core and ischemic penumbra at admission, respectively.4,5 As a consequence, the new CTP mismatch model Tmax/CBF was successfully used to include acute ischemic stroke (AIS) patients in the last trials showing the efficacy of endovascular treatment.6-9 We recently treated with combined intravenous thrombolysis and with mechanical thrombectomy patients imaged within 1.5 hour from symptom onset...
TO THE EDITOR: We read with interest the recent paper by Boned and colleagues.1 The authors conclude that “CT perfusion may overestimate final infarct core, especially in the early time window. Selecting patients for reperfusion therapies based on the CTP mismatch concept may deny treatment to patients who might still benefit from reperfusion”. We completely agree with this consideration, mainly when, as in this article, the core volume is assessed according to the classical CT perfusion (CTP) mismatch mean transit time (MTT)/cerebral blood volume (CBV)2 by measuring the lesion on CBV maps generated with a one-phase CT perfusion (CTP) acquisition protocol. In fact, it is well-known that a short CTP scan duration often produces a truncation of the perfusion curves resulting in an overestimation of CBV lesion that can frequently reverse.3 In addition, it has recently been demonstrated that relative cerebral blood flow (CBF) < 30% and time to peak of the residual function (Tmax) > 6 seconds is more reliable than CBV < 2.0 ml/100gr and relative MTT > 145% in identifying infarct core and ischemic penumbra at admission, respectively.4,5 As a consequence, the new CTP mismatch model Tmax/CBF was successfully used to include acute ischemic stroke (AIS) patients in the last trials showing the efficacy of endovascular treatment.6-9 We recently treated with combined intravenous thrombolysis and with mechanical thrombectomy patients imaged within 1.5 hour from symptom onset with NIHSS > 6, CT ASPECTS > 6, M1 occlusion and good collaterals on multi-phase CT Angiography (mCTA) and a hemispheric hypoperfusion on CTP maps. We analysed the one who achieved an angiographic result of TICI 3 within 3 hours from the onset of symptoms. In all these cases, there was no evidence of neurological deficits at neurological examination and ischemic lesion on non-contrast CT at 48 hours after admission. CTP was acquired with a two-phase imaging protocol. CTP maps were obtained using a commercially available software (Olea Sphere 3.0, Olea Medical, La Ciotat, France). In all the patients, an infarct core was found when we used CBV < 2.0 ml/100mg as a threshold value according to MTT/CBV mismatch. However, when the cut-off value was defined as relative CBF < 30 in agreement with Tmax/CBF mismatch, infarct core resulted substantially reduced in size or disappeared completely. These findings confirm that CTP performed early after stroke can overestimate infarct core since tissue survival thresholds may be instable in this stage and, then, more difficult to be recognized by rigid cut-off values. Nevertheless, our experience confirms that relative CBF < 30 is superior than CBV < 2.0 ml/100gr for detection of irreversible damaged tissue. Moreover, the use of a two-phase acquisition protocol and a fully-automated software allows to avoid data truncation affecting not only CBV but also CBF3 and to generate a reproducible tissue classification. Thus, these tools could represent a promising approach to limit the inaccuracies of CTP, making it more reliable in identifying infarct core at admission. As recently reported,10 the combination with information coming from mCTA could further improve the ability of CTP in predicting tissue fate in AIS patients and, therefore, in the selection of AIS patients for reperfusion therapies.
References
1. Boned S, Padroni M, Rubiera M, et al. Admission CT perfusion may overestimate initial
infarct core: the ghost infarct core concept. NeuroIntervent Surg 2017; 9: 66-9.
2. Wintermark M, Flanders AE, Velthuis B, et al. Perfusion-CT assessment of infarct core and penumbra: receiver operating characteristic curve analysis in 130 patients suspected of acute hemispheric stroke. Stroke 2006;37:979-85.
3. Copen WA, Deipolyi AE, Schaefer PW, et al. Exposing hidden truncation-related errors in acute stroke perfusion imaging. AJNR Am J Neuroradiol 2015; 36: 638-45.
4. Campbell BCV, Christensen S, Levi CR, et al. Cerebral blood flow is the optimal CT perfusion
parameter for assessing infarct core. Stroke 2011; 42: 3435-40.
5.Lin L, Bivard A, PhD; Christopher R. Levi CR, et. al. Comparison of computed tomographic and magnetic resonance perfusion measurements in acute ischemic stroke back-to-back quantitative analysis. Stroke 2014; 45: 1727-32.
6. Campbell BC, Mitchell PJ, Kleinig TJ, et al.. Endovascular therapy for ischemic stroke with perfusion-imaging selection. N Engl J Med 2015; 372: 1009-18.
7. Saver JL, Goyal M, Bonafe A, et al. Stent-retriever thrombectomy after intravenous t-PA vs t-PA alone in stroke. N Engl J Med 2015; 372: 2285-95.
8. Nogueira RG, Jadhav AP, Haussen DC, et al. Thrombectomy 6 to 24 Hours after stroke with a mismatch between deficit and infarct. N Engl J Med 2018; 378: 11-21.
9. Albers GW, Marks MP, Kemp S, et al. Thrombectomy for stroke at 6 to 16 hours with selection by perfusion imaging. N Engl J Med 2018 Jan 24. [Epub ahead of print].
10. d'Esterre CD, Trivedi A, Pordeli P, et al. Regional comparison of multiphase computed tomographic angiography and computed tomographic perfusion for prediction of tissue fate in Ischemic Stroke. Stroke 2017; 48: 939-945.
We read with interest the article by Soize et al. “Can early neurological improvement after mechanical thrombectomy be used as a surrogate for final stroke outcome?”[1] Based on their results, the authors concluded that early neurological improvement (ENI) 24 hours after thrombectomy is a straightforward surrogate of long-term outcome. However, all patients in this study were treated with conscious sedation (CS), and not general anesthesia (GA). The residual effects of GA may mask ENI and limit its utility as a surrogate for long-term outcome.[2]
We performed a similar analysis of patients enrolled in a prospective single-center registry. The ability of ENI to predict 3-month functional independence was assessed by the area under the receiver operating characteristic curve (AUC) and compared using the independent-samples Hanley test. Multivariable linear regression assessing the relationship between anesthetic technique and ENI was also performed. The analysis received ethics approval.
291 patients were treated with thrombectomy, with 261 (89.7%) procedures performed with GA, and 30 (10.3%) with CS. All patients were de-sedated and extubated more than 12 hours before 24-hour National Institutes of Health Stroke Scale assessment. 174 (59.8%) patients achieved 3-month functional independence. Baseline and procedural characteristics did not differ between GA and CS patients (all P>0.05). ENI demonstrated better prognostic ability in CS (AUC 0.91, 95% confiden...
Show MoreWe read with interest the response to our manuscript on using machine learning to optimize elderly patient selection for endovascular thrombectomy (1). We acknowledge here, as the author reports, the limitation of SPOT being based on single center data, and the need for multicenter prospective validation of SPOT as next step in development. The author raises additional technical concerns that we do not necessarily view as applicable to this study.
First, we would like to stress the general limitations of artificial intelligence based techniques such as the overfitting and the data specific local optima problems. However, the specific comments brought by the author are not applicable in our case. First, studies on the number of events per predictor are applicable for logistic regressions (LRs) which is not used in the SPOT algorithm. In fact, our results show poor LR performance which is consistent with the rule of thumb of 1 to 10 referred to by the author. Hence, while serving as a good guidance for LR, the rule is not binding and more importantly it does not guarantee the generalization of the learned model. To further illustrate, classification models using convolutional neural networks have millions of parameters and are trained with datasets that, in most cases, do not have millions of samples in each group. However, these models have acceptable generalization capabilities and are tested using the data-split method. In SPOT, the model at its core is a regressi...
Show MoreCongratulations to Annika Keuler et al¹ on their experience with the wireless microcatheter technique preventing vessel perforations in endovascular thrombectomy. Based on their results, the authors conclude that in most cases of mechanical recanalization, the clot can be passed more safely with a wireless microcatheter. In our daily work, we also find the wireless microcatheter technique seems to reduce subarachnoid hyperdensity resulting from vessel perforations. However it seems difficult to confirm this correlation; the details of which will be discussed as follows. After reading and analyzing the article carefully, we have some opinions about the study which we would like to communicate with the authors because the conclusions of the paper directly relate to our clinical experience.
Show MoreIn the article, two radiological manifestations are defined as vessel perforations——contrast extravasation during angiography and angiographically occult ipsilateral circumscribed subarachnoid contrast extravasation which is identified by post-interventional CT scans. As confirmed by previous studies2-3, we agree with the authors on using immediate post-interventional CT examination to identify the subarachnoid hyperdensity due to intraoperative contrast extravasation. Based on their results, post-thrombectomy subarachnoid hyperdensity was observed on CT scans in 22 patients, in 18 of whom, the clot was passed using a microwire, and in the other four, using a wireless microcathete...
It is with great interest that we read the study of Alawieh et al(1), in which they developed a machine learning algorithm, called ‘SPOT’, to select stroke patients older than 80 years for endovascular therapy (EVT). Prediction modeling to optimize patient selection for EVT is an emerging topic of interest and we agree that predicting individual patient outcomes is increasingly important for decision making in medicine. However, we were surprised by the strong conclusions that were drawn by the authors, considering some serious limitations of the study.
First, the size of the training set is insufficient to develop a complex model with twelve predictor variables and many correlations. Only 22 patients had a good functional outcome, which means that the number of events per tested predictor variable is less than two. For the development of a reliable model, a sample size of at least ten events per variable is needed to minimize the risk of overfitting(2, 3). It has been suggested that even far more events per variable are needed to achieve stable predictions with machine learning techniques(4). Especially complex models developed on small sample sizes have a high risk of overfitting, resulting in unstable predictions and too optimistic model performance measures. The reported AUC of 0.92 is therefore very likely to be an overestimation.
Second, the SPOT algorithm provides a treatment advice based on the predicted outcome after treatment, without providing the...
Show MoreWe had an opportunity to read the article by Lakomkin et al regarding systematic literature review of LVO prevalence. Since one of our studies is part of this review we feel compelled to comment on the paper. We do appreciate the authors’ efforts in conducting this analysis which is important in understanding the burden of disease – but, with respect offer some criticisms. The major limitation of the paper which the authors recognize is the heterogeneity of the included studies. Unfortunately, this limitation is so critical that it yields unreliable information at best and misleading at worst.
The paper intends to study the prevalence of large vessel strokes. However, apart from a couple of population based studies in their review, the rest are a heterogenous mix describing an LVO rate from very selective cohorts of patients from single centers. Several are centered around validation of clinical scales for detecting LVOs. The key features of a population based study include a defined catchment population, access to a large part of that population and a reliable marker of disease. Without these a “prevalence” constitutes a report of a center’s experience of disease rate as it pertains to their patient intake. While still valuable it is not an estimation of the disease burden in the population that the center serves unless an overwhelming majority of that population comes to that center.
The authors determine an average rate of about 30% LVO amongst acute isch...
Show MoreWe would like to congratulate Nicholson et al. on their highly interesting work on the declining rate of SAH in the Irish population. This will certainly provide some very interesting points. Also in Germany there is - at least subjectively - the phenomenon of the declining rate of SAH. The authors can establish a clear correlation to the decline in the smoking rate. Now the question arises whether this is the only relevant correlation. In particular, it would certainly be necessary to investigate whether there has been an increased rate of detection of unruptered Aneurysma and an increasing rate of treatment of those during the study period and whether this may also have a relevant influence on the decrease in SAH.
Dear Editor,
Show Morewe read with great interest the paper from Sallustio et al 1 regarding the use of new thromboaspiration catheter, AXS Catalyst 6 (Stryker Neurovascular, Mountain View, CA, USA), for endovascular treatment (EVT) of large vessel stroke (LVS) with A Direct Aspiration first Pass Technique (ADAPT)2.
In our center, a team composed by 4 vascular interventional radiologists, two physicians with certified experience in stroke treatment and two physicians with large carotid stent experience, and 4 stroke neurologist with large experience in intravenous thrombolysis, started to perform EVT in patients with LVS of anterior or posterior circulation from September 2017.
Given the wide availability of different systems of neurothrombectomy we decided to use AXS Catalyst 6 both for its technical features, as reported by Sallustio et al, both for its lower costs than the others available (6F SOFIA plus catheter, MicroVention, Tustin, CA, USA; the X Penumbra ACE catheters, Penumbra Inc., Alameda, CA, USA).
Between September 2017 and May 2018, 24 patients (72.1 ± 13.2 years old) affected by acute ischemic stroke with LVS underwent to EVT in our center. Median baseline NIHSS was 18 (range: 7-24). Intravenous thrombolysis was used in 5 patients.
The most frequent site of occlusion was the middle cerebral artery (MCA) (70.8%), while in 16.7% of cases was basilar artery. Tandem occlusions occurred in 12.5% of patients and the most frequent stroke etiolo...
The paper by Farzin et al.[1] shows interesting results about measuring porosity of fully expanded flow diverter stents using (photographic) images of the stent being assessed. In their study, authors used 3 different methods and repeated measurements by different observers to assess the porosity of stents. According to their results, the variability when measuring porosity is so large that previous works assessing it should be questioned. On the other hand, they indicate that pore density seems to be more reliable and repeatable. The study highlights the difficulty of measuring such parameter in a controlled in vitro environment. After carefully reading the article, it became clear that the most reproducible way of measuring porosity, from the 3 options studied, was M3 (based on measuring the width and length of the struts and number of struts per reference square). Furthermore, some simple assumptions should improve the results and substantially reduce errors and variability:
Show More1. Wire width: the value for wire width, indicated by the manufacturer, is likely to be more accurate. If this value is no to be trusted, at least in average, then the reproducibility of the manufacturing process could not be trusted. Measuring wire width directly on the images is likely to introduce error as it might be affected by reflection/refraction of light on the wire material and wire coating, as well as lens imperfections or optical aberrations in some cases.
2. Calculating poro...
I read with great interest the meta-analysis by Brinjikji et al.1 which evaluated outcomes after mechanical thrombectomy for acute ischemic stroke by using a balloon guiding catheter (BGC) device. In that study, the authors documented that patients who underwent mechanical thrombectomy with BGC had better clinical and angiographic outcomes than those without BGC. However, there were some issues which should be addressed and discussed.
Show MoreFirst, the number of successful recanalizations, shown as 2b/3 of Thrombolysis In Cerebral Infarction (TICI) grade in Fig.3 in the article,1 might be not accurately described. The events of successful recanalization were noted in 113 of 149 in the BGC group and 133 of 189 in the non-BGC group according to Nguyen et al.2 However, the events were presented as 112 of 149 in the BGC group and 135 of 189 in the non-BGC group.1 Accordingly, the forest plot can be changed as in Fig. 1 below. Mechanical thrombectomy using BGC exhibited significantly higher successful recanalizations than did non-BGC use (OR, 1.710; 95% CI: 1.099-2.662). Second, there was no specific explanation for the publication bias of Fig. 4 in the result section.1 Although the authors reported a p value of 0.49 using Egger’s regression, we are not sure what publication bias meant to represent, successful recanalization or clinical outcome or other variables.
In this letter, we made a funnel plot for successful recanalization based on the revised number of events we h...
TO THE EDITOR: We read with interest the recent paper by Boned and colleagues.1 The authors conclude that “CT perfusion may overestimate final infarct core, especially in the early time window. Selecting patients for reperfusion therapies based on the CTP mismatch concept may deny treatment to patients who might still benefit from reperfusion”. We completely agree with this consideration, mainly when, as in this article, the core volume is assessed according to the classical CT perfusion (CTP) mismatch mean transit time (MTT)/cerebral blood volume (CBV)2 by measuring the lesion on CBV maps generated with a one-phase CT perfusion (CTP) acquisition protocol. In fact, it is well-known that a short CTP scan duration often produces a truncation of the perfusion curves resulting in an overestimation of CBV lesion that can frequently reverse.3 In addition, it has recently been demonstrated that relative cerebral blood flow (CBF) < 30% and time to peak of the residual function (Tmax) > 6 seconds is more reliable than CBV < 2.0 ml/100gr and relative MTT > 145% in identifying infarct core and ischemic penumbra at admission, respectively.4,5 As a consequence, the new CTP mismatch model Tmax/CBF was successfully used to include acute ischemic stroke (AIS) patients in the last trials showing the efficacy of endovascular treatment.6-9 We recently treated with combined intravenous thrombolysis and with mechanical thrombectomy patients imaged within 1.5 hour from symptom onset...
Show MorePages