Article Text


Original research
Long-term retreatment rates of cerebral aneurysms in a population-level cohort
  1. Taylor Daileda1,
  2. Farhaan S Vahidy2,
  3. Peng Roc Chen3,
  4. Hooman Kamel4,
  5. Conrad W Liang5,
  6. Sean I Savitz2,
  7. Sunil A Sheth2
  1. 1 UT Health McGovern School of Medicine, Houston, Texas, USA
  2. 2 Department of Neurology, Institute for Stroke and Cerebrovascular Disease, UT Health McGovern School of Medicine, Houston, TX
  3. 3 Department of Neurosurgery, Institute for Stroke and Cerebrovascular Disease, UT Health McGovern School of Medicine, Houston, TX
  4. 4 Department of Neurology, Weill Cornell Medical College, New York, USA
  5. 5 Department of Neurosurgery, Kaiser Permanente Fontana Medical Center, Fontana, California, USA
  1. Correspondence to Sunil A Sheth, Department of Neurology, Institute for Stroke and Cerebrovascular Disease, UT Health McGovern School of Medicine, Houston TX 77030, USA; ssheth{at}


Background The likelihood of retreatment in patients undergoing procedures for cerebral aneurysms (CAs) has an important role in deciding the optimal treatment type. Existing determinations of retreatment rates, particularly for unruptured CAs, may not represent current clinical practice.

Objective To use population-level data to examine a large cohort of patients with treated CAs over a 10-year period to estimate retreatment rates for both ruptured and unruptured CAs and explore the effect of changing treatment practices.

Methods We used administrative data from all non-federal hospitalizations in California (2005–2011) and Florida (2005–2014) and identified patients with treated CAs. Surgical clipping (SC) and endovascular treatments (ETs) were defined by corresponding procedure codes and an accompanying code for ruptured or unruptured CA. Retreatment was defined as subsequent SC or ET.

Results Among 19 482 patients with treated CAs, ET was performed in 12 007 (62%) patients and SC in 7475 (38%). 9279 (48%) patients underwent treatment for unruptured CAs and 10203 (52%) for ruptured. Retreatment after 90 days occurred in 1624 (8.3%) patients (11.2% vs 3.7%, ET vs SC). Retreatment rates for SC were greater in unruptured than in ruptured aneurysms (4.6% vs 3.1%), but the opposite was true for ET (10.6% vs 11.8%). 85% of retreatments were within 2 years of the index treatment. Retreatment was associated with age (OR=0.99, 95% CI 0.98 to 0.99), female sex (OR=1.5, 95% CI 1.3 to 1.7), Hispanic versus white race (OR=0.86, 95% CI 0.75 to 0.98), and ET versus SC (OR=3.25, 95% CI 2.85 to 3.71). The adjusted 2-year retreatment rate decreased from 2005 to 2012 for patients with unruptured CAs treated with ET (11% to 8%).

Conclusions Retreatment rates for CAs treated with ET were greater than those for SC. However, for patients with unruptured CAs treated with ET, we identify a continuous decline in retreatment rate over the past decade.

  • aneurysm
  • coil
  • subarachnoid
View Full Text

Statistics from


Cerebral aneurysms (CAs) are present in approximately 3% of the general population and are the most common non-traumatic cause of subarachnoid hemorrhage (SAH), a devastating disease from which only one-third of patients have a good neurological outcome.1 The highest quality data on the optimal treatment for CAs come from trials of ruptured aneurysms, including the International Subarachnoid Aneurysm Trial (ISAT) and the Barrow Ruptured Aneurysm Trial (BRAT).2 3 ISAT screened and enrolled patients between 1997 and 2001 and BRAT, between March 2003 and January 2007. While these studies reported improved clinical outcomes with endovascular therapy (ET) relative to surgical clipping (SC), they both raised concern about the durability of ET, as retreatment rates with ET were significantly higher than with SC.4

In the past decade since the conclusion of those trials however, there have been substantial advances in ET technology, such as the development of flow diversion, new intracranial stents, and higher-performance coils. These improvements, coupled with increased operator experience, may render the retreatment rates from trials of ruptured CAs conducted over 10 years ago obsolete and non-representative of current day clinical practice. In addition, these data cannot be readily extrapolated to treatments for unruptured aneurysms, and thus our understanding of retreatment rates for unruptured CAs is severely lacking.

Furthermore, SC of CAs has been in steady decline for the past 10 years, and operator experience with the technique is waning.5 6 This trend has been coupled with a migration of CA treatment away from tertiary referral hospitals, which were the centers that would have enrolled patients in the previous trials, and into community settings.7 These changes are of particular relevance as overall outcomes following both ET and SC have been reported to be better when these procedures are performed in higher-volume centers, by higher-volume practitioners.8–12 We hypothesized that the retreatment rates for CAs in current clinical practice may deviate significantly from those reported in clinical trials, particularly for patients with unruptured CAs. In this study, we use population-level data to examine a large cohort of patients with treated CAs over a 10-year period to estimate retreatment rates for both ruptured and unruptured CAs and explore the effect of changing CA treatment practices.


Study design and setting

We performed a retrospective cross-sectional study, using administrative claims data from the Healthcare Cost and Utilization Project State Inpatient Database on all discharges from non-federal acute care hospitals in California from 2005 to 2011 and in Florida from 2005 to 2014. These two states were chosen because they are large, demographically and socioeconomically diverse, with a mixture of urban and rural populations, located in geographically remote areas of the country. In addition, these two states provide administrative data that allow tracking of individual patients across visits over numerous years.13

Demographic and procedure information

Diagnoses and procedures were identified using International Classification of Diseases, ninth revision, clinical modification (ICD-9 CM) codes, and derived from prior relevant literature for consistency and comparison.9 12 14–16 We used ICD-9 CM codes 437.3, 747.81, or 442.81 for unruptured aneurysms and ICD-9 CM code 430 for SAH if listed as the primary or secondary diagnosis. ET was defined as 39.72, 39.75, 39.76, or 39.79 and SC as 39.51 or 39.52 in conjunction with a code for unruptured aneurysm or SAH as above. Patients were excluded if they died during the index hospitalization or had any diagnosis or treatment of arteriovenous malformation, arteriovenous fistula (747.81, 39.53, 92.30), or trauma (800–803.9 and 850–854.9). Index treatment was defined as the first occurrence of aneurysm treatment for each patient within our cohort. In addition, a composite score of comorbidities using the Charlson index was calculated.17 This index is a validated approach widely used by health researchers to measure burden of disease.18

Aneurysm retreatment was defined as the presence of any codes for ET or SC after the primary treatment. We separately analyzed patients who were retreated ‘early’ after their index treatment (within 90 days) and those treated ‘late’ (after 90 days) to exclude planned retreatments. This method is consistent with previous analyses of the ISAT data.4 We also determined 2-year rates of rupture of treated, previously unruptured aneurysms. All patients with treated aneurysms in our cohort up to 2012 were included in this analysis. Rupture was defined as the presence of an ICD-9 code for SAH as above within 2 years of the index treatment excluding admissions to rehabilitation centers (ICD-9 code V57 in the primary diagnosis position).

Statistical analysis

Demographic differences between patients treated with ET and SC were compared using univariate analyses (Wilcoxon rank sum test for continuous variables and χ2  for categorical data) and multivariate logistic regression, limited to those variables that showed a statistically significant difference in the univariate analysis. To analyze changes in retreatment rates over time, the likelihood of retreatment within 2 years of the index treatment was estimated by multivariate logistic regression models. A period of 2 years was chosen as the vast majority (85%) of retreatments had occurred by this time (see ’Results'), and to allow for sufficient data within our 10-year study period to analyze changes in retreatment rates. Thus, the retreatments studied in this 2-year analysis were for patients whose index treatment occurred in Florida between 2005 and 2012 and in California between 2005 and 2009. Only the first retreatment was considered for analyses in patients with multiple retreatments. Analyses were performed using StataMP 14 (StataCorp LLC, College Station, Texas, USA) and Prism 7 (GraphPad, La Jolla, California, USA) statistical software and reported as odds ratios and 95% confidence intervals (OR, 95% CI).


Among 19 482 patients treated for CAs during the 10-year study period, 7475 (38%) were treated with SC and 12 007 (62%) with ET (table 1). Median age was 57 (IQR 48–66), 13 855 (72%) were female, and 10 786 (58%) were white. As shown in table 1, hypertension was the most common comorbidity, present in 63% of patients. Univariate analysis demonstrated that older patients were more likely to be treated with ET than SC. Medical comorbidities, including renal failure and dyslipidemia, were associated more with ET than with SC. Smoking was associated with SC. ET was less likely in ethnic minorities, including black and Hispanic patients than in white subjects and in ruptured aneurysms relative to surgical clipping. In univariate analysis, ET was more common in Florida than in California (OR= 1.76, 95% CI 1.66 to 1.86). In multivariate analysis, age, sex, location of treatment (Florida vs California), dyslipidemia and year of index treatment remained significantly associated with ET compared with SC. Hispanic patients, patients with a history of smoking, and those presenting with SAH remained more likely to undergo SC (online supplementary table 1).

Table 1

Patient demographics by treatment type and univariate analysis

Retreatment occurred in 2299 (11.8%) patients (figure 1). There were 675 patients with early retreatment, occurring within 90 days of index treatment, who were excluded from further analysis to minimize the inclusion of patients with planned retreatments. Late retreatment, or retreatment after 90 days of index treatment, occurred in 1624 (8.3%) patients. Of these patients, 1381 (85%) underwent a single retreatment, 200 (12%) underwent two retreatments, and 3% underwent three or more. Retreatment after SC occurred in 279 (3.7%) patients and after ET in 1345 (11.2%) patients. Of these patients, 221 (16%) underwent surgical clipping after the initial ET.

Figure 1

Diagram outlining the profile of patients in the examined cohort.

Factors associated with retreatment in univariate analysis included age, with older patients less likely to undergo retreatment as shown in table 2. Female sex was associated with an increased retreatment rate compared with males (OR=1.45, 95% CI 1.29 to 1.65). Ethnic minorities were less likely to be retreated than white patients (Hispanic OR=0.86, 95% CI 0.75 to 0.98, Asian OR=0.68, 95% CI 0.52 to 0.87). Patients who had dyslipidemia, diabetes, hypertension, or renal failure were less likely to be retreated. The retreatment rate in patients who received ET was significantly higher than in those who received SC (11.2% vs 3.7%, OR=3.25, 95% CI 2.85 to 3.71).

Table 2

Univariate analyses of retreatment rates by patient demographics

Results of multivariate analysis of factors associated with retreatment are shown in table 3. The decreased likelihood of retreatment in older patients and ethnic minorities (Hispanic and Asian with the addition of black patients) was preserved, as was the increased likelihood of retreatment in women. After adjustment, medical comorbidities such as dyslipidemia, diabetes, hypertension and renal failure were no longer significant, nor was the distinction between California and Florida. SC compared with ET remained the most significant factor associated with retreatment, with an adjusted OR of 3.37 (95% CI 2.94 to 3.87).

Table 3

Multivariate logistic regression of covariates associated with retreatment

Among patients treated with SC, 4.6% (151/3300) of patients with unruptured CAs underwent retreatment, compared with 3.1% (128/4175) for ruptured. Among patients treated with ET, 10.6% (634/5979) of patients with unruptured CAs underwent retreatment, compared with 11.8% (711/6028) for ruptured. After adjusting for age, sex and ethnicity, retreatment remained more likely after SC for patients with unruptured compared with ruptured CAs (OR=1.47, 95% CI 1.14 to 1.89) but less likely for patients after ET (OR=0.82, 95% CI 0.73 to 0.92). Of the 8145 patients with unruptured aneurysms who underwent treatment up to 2012 in our cohort, 32 (0.4%) had subsequent SAH within 2 years.

Over the study period, the rate of SC decreased from 50.2% in 2005 to 19.4% in 2014, while the rate of ET increased from 49.8% in 2005 to 80.6% in 2014 (figure 2A). Retreatments after 90 days occurred most frequently within the first several years of index treatment. As shown in figure 2B, 85% of the instances of first retreatment occurred within the first 2 years.

Figure 2

Temporal trends in treatment and retreatments over the study period. (A) Yearly rates of surgical clipping (SC) vs endovascular treatment (ET) across the cohort over the study period. *Data available only for Florida. (B) Cumulative frequency of timing for first retreatment after 90 days. Red line indicates percentage of retreatments (85%) at 2 years after the initial treatment. (C) Adjusted probability of retreatment within 2 years for patients with unruptured cerebral aneurysms treated with endovascular coiling over the time period of the study. *Data available only for Florida.

We then analyzed the changes in retreatment rate over the 10-year study period. After adjusting for age, for unruptured aneurysms treated with ET, there was a significant decrease in the rate of retreatment within 2 years. Over the study period, the likelihood of retreatment for these patients decreased by 5% relative to the previous year (OR=0.95, 95% CI 0.90 to 0.99) as shown in figure 2C. This trend of decreasing retreatment rates, however, was not observed in aneurysms treated with SC (ruptured aneurysm OR=1.1, 95% CI 0.98 to 1.22 and unruptured aneurysm OR=1.93, 95% CI 0.84 to 1.03). Similarly, there was no decrease in retreatment rates for ruptured aneurysms treated with ET (OR=1.0, 95% CI 0.96 to1.05).


In this study of nearly 20 000 patients with treatments for CAs, we found that the likelihood of aneurysm retreatment varied by patient age, sex, and ethnicity. Retreatment rates for patients treated initially with ET were greater than for those treated with SC, with an absolute difference of over 7% for the cohort as a whole. With SC, retreatment rates for unruptured CAs were greater than for ruptured, but the opposite was true for ET. Over the course of the study, the use of ET relative to SC increased considerably, and the retreatment rate for unruptured aneurysms treated with ET decreased significantly. We report what we believe to be the first demonstration of continuous improvement in retreatment outcomes for unruptured CAs treated with ET.

The rates of endovascular coiling to treat cerebral aneurysms increased over the study period from 50% to over 80%, with rates of SC dropping to 20% by the end of the study. This finding, and the specific proportions that we identified, are consistent with findings in other population-level studies. A 2008 study by Andaluz et al used the Nationwide Inpatient Sample (NIS), which is a survey-design dataset that samples data from hospitals throughout the nation allowing for extrapolation to the rest of the country, from 1993 to 2003 to identify trends in CA therapy. The data from over 500 000 patients showed that the number of CAs treated with ET doubled from 12 000 (48%) in 1993 to over 24 000 (66%) in 2003, while the number of CAs treated with SC during the same time stayed the same around 13 000 (48% in 1993 to 66% in 2003).6 A 2012 study by Lin et al examined the NIS database more recently, from 1998 to 2007. Over their study period, there was a trend towards an increased use of endovascular coiling, with rates doubling in the period after ISAT was published in 2002.7 In this study, they report that 44.6% of CAs were treated with ET and 55.4% with SC in 2007, which is almost identical to our 2007 treatment rates of 44.5% and 55.5% rates for ET and SC, respectively.

Our study found that older patients, men, and Black or Asian patients were less likely to undergo retreatment. The decreased likelihood of retreatment in older patients and those with greater comorbidities may be secondary to increased procedural risk relative to their potentially limited life span. The sex and racial differences, however, are less readily explained. A follow-up of ISAT found decreased likelihood of retreatment with increased age and no significant influence of gender on retreatment.4 Using the NIS from 2001 to 2008, Brinjikji et al found that increased age, male gender, and non-white race were associated with an increased risk of mortality.14 However, because we excluded patients who died after their index treatment, these findings do not explain the decreased retreatment rate in these patients in our cohort. Other possibilities include sex- or race-related variations in aneurysm location and morphology.19 Another and potentially more likely explanation may be disparities in access and re-access to care.20

Retreatment rates in our cohort for patients with ruptured CAs were 3.7% after SC and 11.2% after ET, and are comparable to those found in ISAT and BRAT. Long-term follow-up of patients in ISAT found retreatment rates of 0.9% after SC and 9.0% after ET, after excluding early retreatments within 90 days, as in our study.4 BRAT, which did not exclude early retreatments reported 1-year retreatment rates of 4.5% after SC and 10.6% after ET. However, we found that retreatment rates for unruptured aneurysms were significantly different for both ET and SC relative to ruptured CAs, with higher retreatment rates for SC and lower retreatment rates for ET. Our findings suggest that retreatment rates and procedural outcomes reported in these trials cannot be extrapolated to treatments of unruptured CAs.

In addition, we were able to demonstrate a continuous decline in the retreatment rate of unruptured aneurysms treated with ET. One possible explanation for this improvement may be an increase in practitioner experience, a finding that has been shown previously in both cerebral aneurysm treatments and several other surgical fields.21–23 In the case of ruptured aneurysms, over this time period there has also been a dissemination of ETs from large academic centers to smaller non-teaching hospitals. Lin et al found that while large teaching hospitals performed the majority of ETs for ruptured CAs from 1998 to 2002, from 2003 to 2007 this difference was no longer present.7 Thus, with a shift of these treatments away from high-volume centers, the relative consistency in retreatment rates for ruptured CAs being performed more frequently at lower-volume centers may be a result of gains in practitioner experience over time.

Furthermore, over the time period examined in our cohort, an increasing number of unruptured CAs were being detected and treated.15 As we demonstrate here, an increasing proportion of these CAs were being treated with ET relative to SC. Thus, increasing numbers of CAs that in the earlier years of the cohort would have been thought to be either more suitable for SC or equally suitable for both techniques were being treated with ET. Thus, with greater numbers of aneurysms previously felt to be less suitable for ET being treated with ET, it is possible that the retreatment rate for ET would have increased over the course of the study. However, the observed decline in retreatment for unruptured CAs and unchanged retreatment rate in ruptured CAs treated with ET implies that improvements in ET have kept pace with the increasing complexity of treatments.

Our study has a number of limitations, many of which are inherent in the use of aggregate data. Retreatment rates of aneurysms are probably related to aneurysm-specific characteristics, including size, morphology, and location; these factors, however, are not captured in administrative databases. Thus, there may be bias in the results as the selection of treatment method (ET vs SC) is heavily influenced by these factors. However, in this study we analyze re-retreatment rates by treatment type over time. And while aneurysm characteristics may influence treatment type as well as retreatment rates, it is unlikely that the types of aneurysms treated in aggregate over the nearly 20 000 patients and multiyear study period changed significantly.

In addition, we are unable to determine whether the retreatment that was performed was a planned retreatment, treatment of a different aneurysm, or retreatment of the initially treated aneurysm. To deal with this limitation, we followed the precedent used in the ISAT study and implemented a ‘washout period’ of 90 days. In addition, comparisons within treatment types (SC or ET), as well as comparisons from year to year, should not be disturbed by this limitation as the two techniques and yearly cohorts should be affected similarly. Another limitation is the possibly varied follow-up that these patients received, particularly with imaging surveillance for recurrence. In addition, there is probably variability in the threshold at which individual practitioners determine whether an aneurysm needs to be retreated. These limitations, however, are inherent in almost all aneurysm follow-up studies, including high-quality trials such as ISAT, which relied on patient self-reporting surveys to identify retreatments.24 Also, because ICD-9 coding for ET does not differentiate between different modalities, such as primary coiling versus stent-assisted coiling, we are unable to make any definitive statements about the relative durability of these treatment subtypes. Finally, accurate determination of clinical outcomes including neurological disability is not feasible using such a dataset. Given these limitations, prospective registries collecting detailed reporting of clinical presentations, treatment techniques, in-hospital complications, and longer term clinical outcomes will be needed to deal with these problems in greater detail.

In this study of a large modern population-level cohort of patients with treated CAs, SC resulted in fewer retreatments than ET. Over a 10-year period, we found increasing use of ET relative to SC and for patients with unruptured aneurysms, differential retreatment rates for SC and ET relative to ruptured CAs. In addition, we discovered that rates of retreatment for patients with unruptured CAs have been declining steadily over the past decade. We demonstrate, for what we believe is the first time, population-level evidence of improvements in retreatment outcomes for unruptured CA treatments with ET.


View Abstract


  • Contributors SAS had full access to all the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis. Study concept and design: SAS, HK, TD, FSV, PRC, SIS, CWL. Acquisition of data: SAS, TD, FSV. Analysis and interpretation of data: SAS, TD, HK, FSV, PRC, SIS, CWL. Drafting of manuscript: SAS, TD. Critical revision of the manuscript for important intellectual content: PRC, SIS, CWL, FSV. Statistical analysis: SAS, FSV. Administrative, technical, or material support: SAS, FSV. Study supervision: SAS, FSV.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement N/A. The data used in this analysis are already publicly available.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.