Proprietary nature of intravascular medical device coatings limits safety testing
Dear Dr. Albuquerque:
We are glad that our work has generated interest and discussion in the field [1]. Four years have elapsed since a need for updated device coating testing was officially announced [2], however complexities on the matter and persistent knowledge gaps limit safety studies of devices currently on the market for clinical intravascular use [3,4]. Standardized in vitro particulate generation testing is needed. However, available literature shows that preclinical device testing is not fully predictive of clinical response. Therefore, in vitro and animal studies cannot replace investigation in humans. Currently, lack of consensus on the following prevent meaningful testing in humans: I) optimal clinical testing methods; ii) definitions of permissible risk; iii) adverse cellular, organ, and temporal-specific effects of distinct coating biomaterials; and iv) effects of pre-existing comorbid conditions. Nevertheless, in vitro testing that does not incorporate clinical data has limited utility for safety guidance. Likewise, in vivo studies that do not incorporate biomaterial factors are incomplete. Thus, the proprietary nature of intravascular device coatings remains a significant limitation to clinical device testing and safety assurances. Growing data [2-6] suggest that it may be time for this to be addressed.
1. Chopra AM, Hu YC, Cruz JP. The Device Specific...
Proprietary nature of intravascular medical device coatings limits safety testing
Dear Dr. Albuquerque:
We are glad that our work has generated interest and discussion in the field [1]. Four years have elapsed since a need for updated device coating testing was officially announced [2], however complexities on the matter and persistent knowledge gaps limit safety studies of devices currently on the market for clinical intravascular use [3,4]. Standardized in vitro particulate generation testing is needed. However, available literature shows that preclinical device testing is not fully predictive of clinical response. Therefore, in vitro and animal studies cannot replace investigation in humans. Currently, lack of consensus on the following prevent meaningful testing in humans: I) optimal clinical testing methods; ii) definitions of permissible risk; iii) adverse cellular, organ, and temporal-specific effects of distinct coating biomaterials; and iv) effects of pre-existing comorbid conditions. Nevertheless, in vitro testing that does not incorporate clinical data has limited utility for safety guidance. Likewise, in vivo studies that do not incorporate biomaterial factors are incomplete. Thus, the proprietary nature of intravascular device coatings remains a significant limitation to clinical device testing and safety assurances. Growing data [2-6] suggest that it may be time for this to be addressed.
1. Chopra AM, Hu YC, Cruz JP. The Device Specific Nature of Polymer Coating Emboli: An Optimal Approach For Future Investigations Related to Polymer Embolism. Journal of Neurointerventional Surgery.
2. U.S. Food and Drug Administration Lubricious Coating Separation From Intravascular Medical Devices FDA Safety Communication. Silver Spring MD: FDA; 2015. Available at: https://wayback.archive-it.org/7993/20161022044037/http://www.fda.gov/Me.... Accessed September 11, 2019.
3. Mehta RI, Rai AT, Vos JA, et al. Intrathrombus polymer coating deposition: a pilot study of 91 patients undergoing endovascular therapy for acute large vessel stroke. Part I: Histologic frequency. Journal of NeuroInterventional Surgery Published Online First: 18 May 2019. doi: 10.1136/neurintsurg-2018-014684
4. Mehta RI , Mehta RI. Hydrophilic polymer embolism: implications for manufacturing, regulation, and postmarketsurveillance of coated intravascular medical devices. J Patient Saf 2018 [Epub ahead of print 19 Mar 2019].doi:10.1097/PTS.0000000000000473
5. Mehta RI , Mehta RI , Solis OE , et al. Hydrophilic polymer emboli: an under-recognized iatrogenic cause of ischemia and infarct. Mod Pathol 2010;23:921–30.doi:10.1038/modpathol.2010.74
6. Mehta RI , Mehta RI. Hydrophilic polymer embolism: an update for physicians. Am J Med 2017;130:e287–90.doi:10.1016/j.amjmed.2017.01.032
Sincerely,
Rashi I. Mehta, MD
West Virginia University
Department of Neuroradiology
Ansaar T. Rai, MD
West Virginia University
Department of Neuroradiology
James W. Simpkins, PhD
West Virginia University
Department of Physiology and Pharmacology
Center for Basic and Translational Stroke Research
Rockefeller Neuroscience Institute
Rupal I. Mehta, MD
University of Rochester
Center for Translational Neuromedicine
Acknowledgments: RIM (Rashi I Mehta) is supported by a grant from the National Institute of General Medical Sciences of the National Institutes of Health (5U54GM104942-03). RIM (Rupal I Mehta) is supported by a grant from the National Institute of Neurological Disorders and Stroke (K08NS089830).
Competing interests: ATR serves as a consultant for Stryker Corporation.
An increasing number of reports highlight polymer coating embolism as an iatrogenic complication of intravascular medical devices [1-3]. Autopsies, histologic evaluations of thrombectomy specimens, samples of captured debris, resected or biopsied tissues, are available methods used to study polymer emboli post investigative catherizations or interventional procedures. Reported data highlight the prevalence of this phenomenon and/or its clinicopathologic impacts, however, fall short of identifying higher-risk polymer emboli interventional devices. Consequently, an optimal approach for future investigations related to polymer coating embolism is required.
Mehta et. al investigate the histologic frequency of polymer emboli among patients who underwent endovascular thrombectomy for treatment of acute ischemic stroke due to large vessel occlusion by retrospectively evaluating thrombectomy specimens [2]. In this study, the reported frequency of polymer emboli includes the use of various devices and techniques among selected cases. However, literature highlights polymer coating embolism is device specific and dependent on coating integrity measured by particulates released [4]. Thus, the use of alternate devices with higher or lower particulate release for a given procedure may result in a large variation in incidence rates from reported results. Also, as mentioned by the authors, subsequent statistical correlations unless appropriately powered provide limited informatio...
An increasing number of reports highlight polymer coating embolism as an iatrogenic complication of intravascular medical devices [1-3]. Autopsies, histologic evaluations of thrombectomy specimens, samples of captured debris, resected or biopsied tissues, are available methods used to study polymer emboli post investigative catherizations or interventional procedures. Reported data highlight the prevalence of this phenomenon and/or its clinicopathologic impacts, however, fall short of identifying higher-risk polymer emboli interventional devices. Consequently, an optimal approach for future investigations related to polymer coating embolism is required.
Mehta et. al investigate the histologic frequency of polymer emboli among patients who underwent endovascular thrombectomy for treatment of acute ischemic stroke due to large vessel occlusion by retrospectively evaluating thrombectomy specimens [2]. In this study, the reported frequency of polymer emboli includes the use of various devices and techniques among selected cases. However, literature highlights polymer coating embolism is device specific and dependent on coating integrity measured by particulates released [4]. Thus, the use of alternate devices with higher or lower particulate release for a given procedure may result in a large variation in incidence rates from reported results. Also, as mentioned by the authors, subsequent statistical correlations unless appropriately powered provide limited information on higher-risk or culprit devices. Consequently, in addition to providing limited value, use of histologic evaluations of thrombi (or autopsies) to exclusively evaluate incidence rates may be time-consuming and have a high resource burden. Notably, this may not be the intent of the authors of this study who elude to an article series.
Few studies related to polymer coating embolism have included controls over procedures and specific devices used. For example, histologic evaluations of captured debris within cerebral protection devices during a mitral valve repair procedure highlighted polymer coating type material in 12 of 14 cases [5]. The source of polymer emboli was speculated to be from the mitral valve repair catheter and/or adjunct devices. In another study, intracranial polymer emboli incidence rates were determined for a branded hydrophilic coated guide sheath used for carotid and iliac stenting procedures respectively in Yucatan miniswine [6]. Since the experimental stents and stent delivery systems lacked any coating, impacts of polymer emboli were isolated to the coated guide sheaths, guide catheters and guidewires used during the procedures. These studies with controls over procedures and devices may be leveraged to understand an incidence rate from a specific set of interventional devices repetitively used during a given procedure. Notably however, even this approach is unable to provide device specific information or identify higher-risk interventional devices as the origin of emboli are difficult to determine.
Device type, coating composition-device substrate material combinations, coating application processes, coating thickness and degree of coating coverage are variables that impact polymer coating integrity on a device [4,7]. Particulate generation testing – the determination of a count, shape and size of particulates released from a device when used in an in-vitro vessel model – is the industry standard for evaluating polymer coating integrity from an intravascular device [8]. An intuitive correlation exists between particulates released in-vitro and the polymer emboli incidence rates [4]. Thus, particulate generation testing may be used to compare particulates released and understand relative incidence rates among intravascular devices. In-vitro particulate testing is an effective and efficient approach to determining higher-risk particulate release devices and may assist in identifying culprit interventional devices.
Based on the aforementioned study methodologies and outcomes, the following approach may be outlined for future investigations related to polymer coating embolism: a) Autopsy based studies that typically lack device information or controls over prior procedures should be limited to determining clinicopathologic impacts of polymer emboli. For these studies, estimating a particulate burden to correlate impacts with the quantity of polymer emboli for each area of assessment is relevant; b) Histologic evaluations of thrombi, captured debris, resected or biopsied tissues are also effective in determining localized disease processes. These methods may be used to determine incidence rates for a procedure if devices are consistently repeated for selected cases. For these studies (and postmortem investigations) combining incidence rates with particulate test data from devices will assist in categorizing embolic risk and determine higher risk devices; c) When clinical presentations preclude the consistent use of devices for a procedure, animal studies may be used to determine incidence rates. Extrapolating clinical impacts from healthy animals maybe acceptable, however, investigations that include relevant disease conditions are preferred; d) In-vitro particulate generation testing may be the optimal method to rank embolic risk and determine higher risk polymer emboli devices.
Inclusion and exclusion criteria for polymer emboli related investigations are essential for meaningful results. Patient history should be carefully evaluated as prior procedures may impact incidence rates. For this reason, patients with a history of multiple interventional procedures should be excluded from studies that attempt to determine polymer emboli frequencies. This patient subset may be included for studies determining potential impacts from polymer emboli for a worst-case scenario assessment. All investigations should include device specific information such as type, dimensions (e.g. length, diameter), brand, coating types (e.g. hydrophilic and/or hydrophobic) and coated dimensions. Other important parameters include patient baseline characteristics, procedural length, device in-dwell time (if available) and procedural outcomes.
Comparing particulate data among devices or reporting the magnitude of embolic burden may require assumptions to characterize total particulate volume. For the aforementioned studies, use of an efficient light obscuration methodology for particulate assessments that provides an equivalent circular diameter for particle areas may be used [9]. Combined with a uniform 1-micron height representing the typical lamellar nature of polymer particulates [3], a cylindrical volume calculation for each particle may be optimal. Summation of particle volumes may provide a total particulate burden per device or affected area of inspection.
The FDA continues to work with stakeholders to create tools which permit the standardization of particulate test methods and enable comparisons among devices [10]. Till these standards become available, investigations should include assumptions used to generate particulate data. In the future, an accumulation of procedural and device particulate data with associated incidence rates and clinicopathologic impacts may provide actionable input for regulators for setting device particulate limits. Given the sparse literature on this subject, more studies with controls over procedural parameters and devices used are required. Procedures with devices exposed to larger frictional forces (e.g. chronic total occlusions, aortic repair, or atherectomy), or aqueous environments for longer durations (e.g. percutaneous mechanical circulatory support) should be prioritized for future studies.
Acknowledgements, Funding Sources, Disclosures & Author Contributions
Acknowledgements: None. No persons other than the listed authors have made contributions to this manuscript.
Funding Sources: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Disclosures: None. Authors declare no current relationship with industry and no conflicts of interest.
Author Contributions: All authors have contributed to this manuscript.
References:
1. Chopra AM, Mehta M, Bismuth J, et al. Polymer coating embolism from intravascular medical devices - a clinical literature review. Cardiovasc Pathol 2017;30:45-54.
2. Mehta RI, Rai AT, Vos JA, Solis OE, Mehta RI. Intrathrombus polymer coating deposition: a pilot study of 91 patients undergoing endovascular therapy for acute large vessel stroke. Part I: Histologic frequency. J Neurointerv Surg. Epub ahead of print 18 May 2019. doi: 10.1136/neurintsurg-2018-014684.
3. Hickey TB, Honig A, Ostry AJ, et al. Iatrogenic embolization following cardiac intervention: postmortem analysis of 110 cases. Cardiovasc Pathol 2019;40:12-18.
4. Babcock DE, Hergenrother RW, Craig DA, Kolodgie FD, Virmani R. In Vivo Distribution of Particulate Matter from Coated Angioplasty Balloon Catheters. Biomaterials 2013;34(13):3196-205.
5. Frerker C, Schlüter M, Sanchez OD, et al. Cerebral Protection During MitraClip Implantation: Initial Experience at 2 Centers. JACC Cardiovasc Interv. 2016;9(2):171-9.
6. Stanley JR, Tzafriri AR, Regan K, et al. Particulates from Hydrophilic-Coated Guiding Sheaths Embolize to the Brain. EuroIntervention 2016;11(12):1435-41.
7. Work JW. Technical white paper: considerations for hydrophilic surface coatings on medical devices [internet]. Biocoat, Inc. Horsham, PA, USA. 2016. Available from www.biocoat.com. Accessed 22 Apr 2016.
8. Center for Devices and Radiological Health Recognized Consensus Standards. Recognition number 3-99: AAMI TIR42:2010 Evaluation of Particulates Associated with Vascular Medical Devices (cardiovascular) [Internet]. 2010. Available from: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfStandards/detail.cf.... Accessed Jan. 13, 2019.
9. AAMI TIR42:2010. Evaluation of Particulates Associated with Vascular Medical Devices. Arlington, VA: Association for the Advancement of Medical Instrumentation, 2010.
10. ASTM WK60510. New Test Method for Simulated Use Testing of Neurointerventional Device in Tortuous Vasculature. West Conshohocken, PA: American Society for Testing and Materials International, 2018.
Dear Editor,
We would like the thank Drs. Berndt, Zimmer, Kaesmacher, and Boeckh-Behrens for their interest in our study titled “Clot permeability and histopathology: Is a clot’s perviousness on CT imaging correlated with its histologic composition?” We read their letter with interest. The authors have been pioneers in stroke clot analysis and we greatly respect their academic rigor and expertise.
While we agree that there are certainly some methodological differences between our two studies, we do not believe that these are to blame for the differences in results. Rather, we feel that the observed differences in results between our studies could be due to differences in our patient populations.
Our group has previously shown that there is indeed a correlation between clot composition and etiology. In a recently published article in Stroke we found that large artery atherosclerosis clots were more likely to be platelet rich than those of a cardioembolic origin.1 To date, however, we have yet to find any definite correlation between etiology and RBC density or fibrin density, and we think it is too early to make any definite conclusions on the association between clot composition and etiology.
We agree that the association between perviousness, clot composition, etiology and clinical outcome is not conclusively clarified yet, and hence warrants further research, especially in a larger patient group in a multi-centric setting. Currently, our gr...
Dear Editor,
We would like the thank Drs. Berndt, Zimmer, Kaesmacher, and Boeckh-Behrens for their interest in our study titled “Clot permeability and histopathology: Is a clot’s perviousness on CT imaging correlated with its histologic composition?” We read their letter with interest. The authors have been pioneers in stroke clot analysis and we greatly respect their academic rigor and expertise.
While we agree that there are certainly some methodological differences between our two studies, we do not believe that these are to blame for the differences in results. Rather, we feel that the observed differences in results between our studies could be due to differences in our patient populations.
Our group has previously shown that there is indeed a correlation between clot composition and etiology. In a recently published article in Stroke we found that large artery atherosclerosis clots were more likely to be platelet rich than those of a cardioembolic origin.1 To date, however, we have yet to find any definite correlation between etiology and RBC density or fibrin density, and we think it is too early to make any definite conclusions on the association between clot composition and etiology.
We agree that the association between perviousness, clot composition, etiology and clinical outcome is not conclusively clarified yet, and hence warrants further research, especially in a larger patient group in a multi-centric setting. Currently, our group is collaborating with multiple centers across the United States and Canada in the Stroke Thromboembolism Registry of Imaging and Pathology and are in the process of performing histological and imaging analysis of over 1200 collected clots from stroke patients. Combining our findings with those of other ongoing multicenter studies may ultimately help us reach meaningful conclusions regarding the association between clot composition and imaging, outcomes and etiology.
References:
[1] Fitzgerald S, Dai D, Wang S, et al. Platelet-rich emboli in cerebral large vessel occlusion are associated with a large artery atherosclerosis source. Stroke 2019;50(7):1907-1910.
Recently, we have read with great interest the article by Benson et al. “Clot permeability and histopathology: is a clot’s perviousness on CT imaging correlated with its histologic composition?” [1].
It is pleasing, that research in the field of thrombus characterization by perviousness and its association to thrombus composition is emerging. Benson et al. report a higher clot perviousness for RBC rich clots in comparison to fibrin dominant thrombi [1]. These results stand in contrast to our previously published study [2], that shows an association between perviousness and fibrin rich clots. We furthermore validated those findings in a large collective by showing a relationship between perviousness and cardioembolic origin. Further research to this special topic is scarce. However, there is another experimental and therefore well controllable study on artificial clots, that showed a strong association of fibrin content and contrast agent uptake [3], similar as it has been shown for in vivo thrombi in our study [2].
Consequently, these contradictory results demand further explanations. In our opinion, the differing results might be caused by methodological differences, which we want to discuss.
First, thrombus localizations should be taken into account. Benson et al. used a collective of 57 thrombi with different thrombus locations (38 MCA, 6 ICA, 5 ICA/MCA, 3 basilar artery, 2 posterior cerebral artery, 2 ICA/MCA/ACA, 1 ICC/MCA). It is at least questiona...
Recently, we have read with great interest the article by Benson et al. “Clot permeability and histopathology: is a clot’s perviousness on CT imaging correlated with its histologic composition?” [1].
It is pleasing, that research in the field of thrombus characterization by perviousness and its association to thrombus composition is emerging. Benson et al. report a higher clot perviousness for RBC rich clots in comparison to fibrin dominant thrombi [1]. These results stand in contrast to our previously published study [2], that shows an association between perviousness and fibrin rich clots. We furthermore validated those findings in a large collective by showing a relationship between perviousness and cardioembolic origin. Further research to this special topic is scarce. However, there is another experimental and therefore well controllable study on artificial clots, that showed a strong association of fibrin content and contrast agent uptake [3], similar as it has been shown for in vivo thrombi in our study [2].
Consequently, these contradictory results demand further explanations. In our opinion, the differing results might be caused by methodological differences, which we want to discuss.
First, thrombus localizations should be taken into account. Benson et al. used a collective of 57 thrombi with different thrombus locations (38 MCA, 6 ICA, 5 ICA/MCA, 3 basilar artery, 2 posterior cerebral artery, 2 ICA/MCA/ACA, 1 ICC/MCA). It is at least questionable, if histological composition can be compared between different circulations as it is known that there are different pathophysiological mechanisms (e.g. more underlying stenosis in the posterior circulation) as well as different flow conditions that would influence thrombus evolution and consequently composition [4].
Second, and more important, using an in-homogenous collective for perviousness assessment would stringently lead to technical difficulties in data acquisition: to make measurements comparable, direct contact from fresh flooding contrast agent in CTA is required, preventing the problem of a stationary blood column, which can be observed in occlusions of the ICA, for example. From own experience, identifying exact thrombus location is challenging for cases of ICA occlusions, that complicates the perviousness assessment. To circumvent these risks for falsified measurements, we used in our study a homogenous collective of 32 MCA occlusions. It would be interesting if Benson et al. would reproduce their results in the subgroup of 38 MCA occlusions.
Third, technical issues about perviousness assessment should be discussed as they differ between the groups. We used a co-registration process for native CT and CTA images to exactly identify the thrombus, although the thrombus is not visible in the native CT scan. Benson et al. did not apply a co-registration process between native CT and CTA images, and they excluded patients, “if the inciting clot was too small to be visualized”. Especially fibrin-rich clots are not obviously visible on native CT, and they can be included in the analysis by using a co-registration process to avoid a systematical selection bias.
Fourth, we excluded patients because of non-occluding thrombi when contrast agent passed the thrombus. We think, that perviousness cannot be assessed adequately for these cases, because perviousness would be measured artificially high. Consequently, these cases also present with tendentially better clinical outcome. It would be interesting, if such cases are included in the study of Benson et al. and how they present.
From the majority of histological studies, an association between fibrin-rich clots and cardioembolic origin is known [5, 6]. This fact makes the perviousness assessment interesting as it can predict stroke cause early on (for detailed discussion see [2]). Benson et al. report a cardioembolic origin as most common (59.6% of thrombi). They also report that 66.7% appeared pervious, however with higher RBC density. It would be interesting if their data would show a possible correlation between histological thrombus composition and etiology, and, secondly, a correlation between perviousness and etiology.
Benson et al. based their hypotheses on a positive correlation between pervious clots and good clinical outcome, that seems plausible as the dependent tissue behind an occlusion is better supplied with blood and nutrients when the clot is pervious [7]. RBC-rich clots show better clinical outcome, that is, among others, based on an easier removal in mechanical thrombectomy [8]. Due to the predominant subgroup of cardioembolic, presumably fibrin-rich clots in the anterior circulation, these clots dominate the analyses. Possibly, the assumed tendency between perviousness and good clinical outcome is based on this predominant subgroup. Consequently, it would not contradict the possible association between higher perviousness and fibrin-rich clots.
In summary, association between perviousness, clot composition, etiology and clinical outcome is not conclusively clarified yet and warrants further research, especially in a larger patient group in a multi-centric setting.
Literature
1. Benson JC, Fitzgerald ST, Kadirvel R, Johnson C, Dai D, Karen D et al. Clot permeability and histopathology: is a clot's perviousness on CT imaging correlated with its histologic composition? J Neurointerv Surg. 2019. doi:10.1136/neurintsurg-2019-014979.
2. Berndt M, Friedrich B, Maegerlein C, Moench S, Hedderich D, Lehm M et al. Thrombus Permeability in Admission Computed Tomographic Imaging Indicates Stroke Pathogenesis Based on Thrombus Histology. Stroke. 2018;49(11):2674-82. doi:10.1161/STROKEAHA.118.021873.
3. Borggrefe J, Kottlors J, Mirza M, Neuhaus VF, Abdullayev N, Maus V et al. Differentiation of Clot Composition Using Conventional and Dual-Energy Computed Tomography. Clin Neuroradiol. 2017. doi:10.1007/s00062-017-0599-3.
4. Boeckh-Behrens T, Pree D, Lummel N, Friedrich B, Maegerlein C, Kreiser K et al. Vertebral Artery Patency and Thrombectomy in Basilar Artery Occlusions. Stroke. 2019;50(2):389-95. doi:10.1161/STROKEAHA.118.022466.
5. Sporns PB, Hanning U, Schwindt W, Velasco A, Minnerup J, Zoubi T et al. Ischemic Stroke: What Does the Histological Composition Tell Us About the Origin of the Thrombus? Stroke. 2017;48(8):2206-10. doi:10.1161/STROKEAHA.117.016590.
6. Boeckh-Behrens T, Kleine JF, Zimmer C, Neff F, Scheipl F, Pelisek J et al. Thrombus Histology Suggests Cardioembolic Cause in Cryptogenic Stroke. Stroke. 2016;47(7):1864-71. doi:10.1161/STROKEAHA.116.013105.
7. Santos EM, Marquering HA, den Blanken MD, Berkhemer OA, Boers AM, Yoo AJ et al. Thrombus Permeability Is Associated With Improved Functional Outcome and Recanalization in Patients With Ischemic Stroke. Stroke. 2016;47(3):732-41. doi:10.1161/STROKEAHA.115.011187.
8. Hashimoto T, Hayakawa M, Funatsu N, Yamagami H, Satow T, Takahashi JC et al. Histopathologic Analysis of Retrieved Thrombi Associated With Successful Reperfusion After Acute Stroke Thrombectomy. Stroke. 2016;47(12):3035-7. doi:10.1161/STROKEAHA.116.015228.
The authors describe two endovascular techniques for delivering IA chemotherapy to retinoblastoma patients. Technique A where a 1.2 Fr or 1.5 Fr micro catheter with continuous verapamil flush is advanced without a guide and technique B where a1.5 Fr or 1.7 Fr micro catheter is advanced within a 4 Fr catheter, through a 4 Fr sheath. We usually use a no sheath technique, using a 4Fr diagnostic catheter as a guide catheter for neonatal and pediatric cases. Most importantly, we do not use Echelon or Marathon micro catheters in neonatal and pediatric patients because their use is contraindicated per "instructions for use".
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.
Proprietary nature of intravascular medical device coatings limits safety testing
Dear Dr. Albuquerque:
We are glad that our work has generated interest and discussion in the field [1]. Four years have elapsed since a need for updated device coating testing was officially announced [2], however complexities on the matter and persistent knowledge gaps limit safety studies of devices currently on the market for clinical intravascular use [3,4]. Standardized in vitro particulate generation testing is needed. However, available literature shows that preclinical device testing is not fully predictive of clinical response. Therefore, in vitro and animal studies cannot replace investigation in humans. Currently, lack of consensus on the following prevent meaningful testing in humans: I) optimal clinical testing methods; ii) definitions of permissible risk; iii) adverse cellular, organ, and temporal-specific effects of distinct coating biomaterials; and iv) effects of pre-existing comorbid conditions. Nevertheless, in vitro testing that does not incorporate clinical data has limited utility for safety guidance. Likewise, in vivo studies that do not incorporate biomaterial factors are incomplete. Thus, the proprietary nature of intravascular device coatings remains a significant limitation to clinical device testing and safety assurances. Growing data [2-6] suggest that it may be time for this to be addressed.
1. Chopra AM, Hu YC, Cruz JP. The Device Specific...
Show MoreAn increasing number of reports highlight polymer coating embolism as an iatrogenic complication of intravascular medical devices [1-3]. Autopsies, histologic evaluations of thrombectomy specimens, samples of captured debris, resected or biopsied tissues, are available methods used to study polymer emboli post investigative catherizations or interventional procedures. Reported data highlight the prevalence of this phenomenon and/or its clinicopathologic impacts, however, fall short of identifying higher-risk polymer emboli interventional devices. Consequently, an optimal approach for future investigations related to polymer coating embolism is required.
Mehta et. al investigate the histologic frequency of polymer emboli among patients who underwent endovascular thrombectomy for treatment of acute ischemic stroke due to large vessel occlusion by retrospectively evaluating thrombectomy specimens [2]. In this study, the reported frequency of polymer emboli includes the use of various devices and techniques among selected cases. However, literature highlights polymer coating embolism is device specific and dependent on coating integrity measured by particulates released [4]. Thus, the use of alternate devices with higher or lower particulate release for a given procedure may result in a large variation in incidence rates from reported results. Also, as mentioned by the authors, subsequent statistical correlations unless appropriately powered provide limited informatio...
Show MoreDear Editor,
We would like the thank Drs. Berndt, Zimmer, Kaesmacher, and Boeckh-Behrens for their interest in our study titled “Clot permeability and histopathology: Is a clot’s perviousness on CT imaging correlated with its histologic composition?” We read their letter with interest. The authors have been pioneers in stroke clot analysis and we greatly respect their academic rigor and expertise.
While we agree that there are certainly some methodological differences between our two studies, we do not believe that these are to blame for the differences in results. Rather, we feel that the observed differences in results between our studies could be due to differences in our patient populations.
Our group has previously shown that there is indeed a correlation between clot composition and etiology. In a recently published article in Stroke we found that large artery atherosclerosis clots were more likely to be platelet rich than those of a cardioembolic origin.1 To date, however, we have yet to find any definite correlation between etiology and RBC density or fibrin density, and we think it is too early to make any definite conclusions on the association between clot composition and etiology.
We agree that the association between perviousness, clot composition, etiology and clinical outcome is not conclusively clarified yet, and hence warrants further research, especially in a larger patient group in a multi-centric setting. Currently, our gr...
Show MoreRecently, we have read with great interest the article by Benson et al. “Clot permeability and histopathology: is a clot’s perviousness on CT imaging correlated with its histologic composition?” [1].
Show MoreIt is pleasing, that research in the field of thrombus characterization by perviousness and its association to thrombus composition is emerging. Benson et al. report a higher clot perviousness for RBC rich clots in comparison to fibrin dominant thrombi [1]. These results stand in contrast to our previously published study [2], that shows an association between perviousness and fibrin rich clots. We furthermore validated those findings in a large collective by showing a relationship between perviousness and cardioembolic origin. Further research to this special topic is scarce. However, there is another experimental and therefore well controllable study on artificial clots, that showed a strong association of fibrin content and contrast agent uptake [3], similar as it has been shown for in vivo thrombi in our study [2].
Consequently, these contradictory results demand further explanations. In our opinion, the differing results might be caused by methodological differences, which we want to discuss.
First, thrombus localizations should be taken into account. Benson et al. used a collective of 57 thrombi with different thrombus locations (38 MCA, 6 ICA, 5 ICA/MCA, 3 basilar artery, 2 posterior cerebral artery, 2 ICA/MCA/ACA, 1 ICC/MCA). It is at least questiona...
The authors describe two endovascular techniques for delivering IA chemotherapy to retinoblastoma patients. Technique A where a 1.2 Fr or 1.5 Fr micro catheter with continuous verapamil flush is advanced without a guide and technique B where a1.5 Fr or 1.7 Fr micro catheter is advanced within a 4 Fr catheter, through a 4 Fr sheath. We usually use a no sheath technique, using a 4Fr diagnostic catheter as a guide catheter for neonatal and pediatric cases. Most importantly, we do not use Echelon or Marathon micro catheters in neonatal and pediatric patients because their use is contraindicated per "instructions for use".
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 MorePages