In Response to: Dicpinigaitis AJ, Ogulnick JV, Mayer AS, Gandhi CD, Al‐Mufti F. Increase in Ruptured Cerebral Arteriovenous Malformations and Mortality in the United States: Unintended Consequences of the ARUBA Trial

We thank the authors for their engagement with our article. ¹ , ² The authors raise 2 primary issues with the study in question: (1) the use of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM) codes to identify cerebral vascular malformations, and (2) that the findings of our analysis are primarily due to discrepancies in coding before and after the transition to International Classification of Diseases, Tenth Revision, Clinical Modification (ICD‐10‐CM) billing codes in the 4th quarter of 2015.

Regarding the concern that the findings may be due to coding artifact in the ICD code transition timeline, we decided to perform a sensitivity analysis to evaluate quarterly arteriovenous malformation (AVM) rupture rates after the publication of the ARUBA (A Randomized Trial of Unruptured Brain Arteriovenous Malformations) trial but prior to the ICD‐10‐CM coding transition in the 4th quarter of 2015. We compared AVM rupture rates starting from the 3rd quarter of 2014 (to allow time for practice patterns to change following publication of ARUBA) to the 3rd quarter of 2015 (the last quarter utilizing ICD‐9‐CM codes). The rupture rates are as follows: 2014Q3 (13.3%), 2014Q4 (12.9%), 2015Q1 (17.8%), 2015Q2 (16.1%), and 2015Q3 (19.5%) (trend P=0.004). We then compared rupture rates between 2014 and 2015 post‐ARUBA within the ICD‐9 era (13.1 versus 17.8%; P<0.001). These data demonstrate that rates of cAVM hemorrhage increased in the period after the publication of the ARUBA trial and prior to the 4th quarter ICD‐10 coding transition, suggesting that even if a markedly increased rate was observed in the 4th quarter of 2015 due to ICD‐10‐CM coding artifact, that an effect was already in place and establishing itself.

The authors also raised concerns regarding the validity and accuracy of the code used to diagnose ruptured AVMs. The 2002 study by Berman et al investigates the use of ICD‐9‐CM billing codes to identify intracranial vascular malformations and concludes that the same codes utilized in our analysis for their identification achieved a sensitivity of 94% and a low false‐positive rate, and moreover, that registry admission rates were consistent with population disease detection rates. ³ The authors raise a concern that only 66% of pathologies identified with this coding schema corresponded to true brain AVMs, the other remaining 34% composed of fistulae and cavernous malformations, among other vascular anomalies. We contend that this is a well‐known and accepted limitation of ICD billing codes. To put these percentages into context, previous analyses evaluating ICD‐9‐CM coding for acute ischemic stroke admissions reported that only 61% of these admissions were correctly identified. ⁴ Moreover, the true issue in our opinion is not that only 66% percent of admission diagnoses were those of true brain AVMs in the ICD‐9 era, but that this rate would persist into the ICD‐10 era and maintain continuity between the eras before and after the billing code transition. At this point, there is no evidence to suggest or reason to believe that this same proportion would not carry over into the ICD‐10 era, even if there is a novel code to identify hemorrhage of a vascular malformation independent of one for an unruptured lesion.

Ultimately, we do not disagree with the authors that there are challenges associated with registry‐based research and the billing codes utilized to identify diagnoses of interest therein. However, we contend that the existence of ICD coding discrepancies and an effect of the ARUBA trial are not mutually exclusive, and for this reason we labeled our study as a hypothesis‐generating analysis and concluded that further inquiry is necessary to exclude other confounders contributing to this rise and to investigate the magnitude of this effect. ² Finally, it should be noted that a separate group of authors from the University of Miami performed a similar analysis as that of the present study using the National Inpatient Sample and they too concluded that rupture rates increased in the post‐ARUBA era, demonstrating that a national reduction in cAVM treatment was correlated with an increased rupture incidence. ⁵

We once again thank the authors for their time and effort in commenting on our manuscript. We feel it is important to scrutinize the scientific literature and that this inquiry has now further strengthened the conclusions of this manuscript.