Surgical and endovascular treatment of unruptured cerebral aneurysms at university hospitals

Methods.

The University Health System Consortium (UHC) is an organization of 80 health centers with involvement in medical education. UHC maintains a database of all hospital discharges for member hospitals that includes 2.1 million discharges per year.16 The UHC database was searched from January 1, 1994 to June 30, 1997 for primary diagnosis of unruptured cerebral aneurysm (ICD-9 437.3) and surgical clipping (ICD-9 procedure code 39.51) or repair by other means including coil embolization and wrapping (ICD-9 39.52). Sixty member universities contributed unruptured aneurysm cases to the database (see Appendix).

To determine the accuracy of coding in the UHC database, we compared a detailed review of records at the University of California, San Francisco (UCSF), the largest contributor to the database, with the coded data for the years 1994 to 1996. Cases of treated unruptured cerebral aneurysms were identified through computerized search of medical records and review of clinical databases in the Division of Neurointerventional Radiology and the Neurovascular Service. Medical records were reviewed independently by a clinical nurse specialist and one of the authors (S.C.J.), both blinded to UHC coding, and information on accuracy of diagnosis, procedure performed, length of stay, and discharge status were recorded. The Rankin Score at discharge was determined from record review. Charges were independently obtained through the hospital billing database.

Information from medical record review was compared with UHC data from the same patients, linking records by year of birth and medical record numbers. From this review (described below), strategies for reducing misclassification in the UHC analysis were developed.

Surgically wrapped aneurysms were occasionally coded with the ICD-9 code 39.52 used to identify endovascularly treated aneurysms. Because we were interested in comparing surgical with endovascular therapy, procedures were grouped by operating physician (figure), and cases coded as 39.52 in which the operating physician performed more than 50% surgical clippings (4.2% of total cases) were reanalyzed as surgical cases.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Patient categorization and selection. *Cases reclassified based on the predominant procedure type of the treating physician.

It is common for patients undergoing coiling to need follow-up embolization to obliterate the aneurysm completely (occurred in 11% of cases). Surgically treated patients need repeat procedures less often. Because follow-up procedures involve less risk than primary procedures, accurate comparison of the total risk for a treatment approach requires the linking of primary procedures to follow-up procedures (creating a single course of treatment). Failure to identify courses completely would bias results in favor of coiling. To correct this, all admissions for a given patient were clustered by sorting records by university hospital, year of birth, and medical record number (80% identity required). These records were then combined so that each represented the entire treatment course for unruptured aneurysms during the study period (see figure). For these combined entries (2.9% of total), length of stay and charges were taken as the sum of all hospitalizations, and discharge status was taken as the most adverse. In this way, multiple admissions for treatment of an unruptured aneurysm (occurring more commonly for coiled aneurysms) were treated as a single event with the most adverse consequence of treatment taken as discharge status. When two different treatments were undertaken, the first was coded as the primary procedure to parallel an intention-to-treat analysis.

An adverse outcome was defined as an in-hospital death or transfer to a nursing home or rehabilitation hospital at discharge. Patients admitted from nursing homes were excluded from analysis of discharge to rehabilitation hospitals or nursing homes as it was presumed they would return to these hospitals even if the procedure was uncomplicated.

Statistical analyses were performed using the Stata statistical package (version 5.0, Stata Corporation, College Station, TX). For univariate comparisons of continuous variables, we used Student’s t-test for normally distributed variables and Wilcoxon’s rank-sum test for skewed results (i.e., length of stay and hospital charges). For discrete variables, Fisher’s exact test was used for comparisons with a cell number less than 10 and Pearson’s chi-square test was used for all others. Multivariable analyses were performed with logistic regression for dichotomous outcomes and linear regression for continuous outcomes, including all admission variables in the model (age, sex, race, transfer admission, emergency room admission, and year). To account for the effects of the correlation between observations within a given university hospital, the generalized estimating equations approach was used for the logistic regression. Compound symmetry (equal correlation between all inter-center observations) was chosen as the covariance structure.17

Results.

Discussion.

The findings of this study must be viewed with several limitations in mind, shared by other analyses of large, retrospective, multipurpose databases. First, there is a potential problem with misclassification of the outcomes, the procedures, and the covariates. We have tried to reduce this problem by comparing a subset of the data with detailed medical record review, and subsequently altering variable definitions to reduce misclassification. Certainly, there is residual misclassification, and it may be that UCSF data in the UHC database is better than the data provided by other institutions. Nonetheless, based on our record review at UCSF, we believe that misclassification is not a major problem when all admissions for a given patient are considered a single treatment course and when procedure is defined by the predominant procedure of the operating physician.

Second, the outcome variables have several important limitations. For this study, the only available clinical outcome variables were in-hospital mortality and transfer to a nursing home or rehabilitation facility. Transfer to another facility at discharge was highly correlated with disability at time of discharge assessed by Rankin Score from hospital record review. However, surgically clipped patients may recover more slowly but more completely, and this would not be apparent in our analysis. Nevertheless, the outcomes are clearly defined, reliable, and probably important, providing an initial insight into differences between treatment complications.

Third, there may be inadequate control of important confounding variables.18 Patients treated with surgical clipping may have more comorbid disease, larger aneurysms, or more acute compressive symptoms than those treated endovascularly. We included all available potential risk factors for poor outcome in the multivariable models, including age, gender, race, and admission source (emergency room, transfer, or elective), but other potentially important risk factors could not be controlled because they were not available in the database. Because many patients with unruptured aneurysms are asymptomatic and tend to by younger (average age 52 in this study), patient characteristics may be less important as confounders, but aneurysm characteristics may have influenced decisions about which treatment to employ and could have important effects on the outcomes.15,19 However, treatment decisions appeared to be dictated more by institutional preference than case-specific characteristics, as coil embolization was not offered at many institutions and predominated at a few, and this would tend to reduce selection bias.

Fourth, studies of this type can only track events that occur during hospitalization. Therefore, the success of therapy in terms of reduction in aneurysm rupture rates cannot be measured. It is possible that endovascular coil embolization is ineffective at treating aneurysms, and this would not be apparent from our data. However, reported rates of rupture in the literature have been extremely low for unruptured aneurysms treated by surgical clipping7,15 and coil embolization.20-22 Given recent data from the International Study of Unruptured Intracranial Aneurysms suggesting rupture rates are much lower than previously believed,23 this might be expected even with ineffective treatment. More persuasively, however, ruptured aneurysms also have very low rerupture rates after treatment with either modality,21,24-29 supporting the effectiveness of both therapies because aneurysm rerupture rates are high.30 Therefore, defining perioperative complications may be more important for planning treatment.

Despite the limitations of this study, the finding of relatively low risks of in-hospital mortality and discharge to nursing home or rehabilitation facility for endovascularly treated compared to surgically treated unruptured aneurysms is dramatic and provocative. Also, the study has advantages over the case series reported in the literature. By comparing these two treatment modalities head-to-head and performing risk adjustment, we can be more confident that differences in definitions of potential confounders and outcomes are not responsible for the observed effect. In addition, we have removed a potential bias of practicing physicians to underestimate important complications. Finally, centers that do not specialize in these procedures are included, providing a more realistic estimate of the performance of these techniques in the community. However, it was necessary to focus the analysis on academic medical centers because endovascular coil embolization is rarely performed outside this setting; nontertiary hospitals have higher mortality rates for aneurysm surgery10 and including them could have negatively biased our assessment of surgery.

Our results are similar to those reported in the literature. A recent meta-analysis of studies published between 1966 and 1996 on clipping of unruptured aneurysms found an overall mortality of 2.6%,9 very close to the 2.3% found in this study. Morbidity was 10.9% in a meta-analysis,9 lower than the 18.5% adverse outcome rate we found, but this difference may simply be due to more strict definitions of complications in the studies included in the meta-analysis. Some studies have defined morbidity as a permanent deficit or disability with dependence, neither of which is necessary in our definition of adverse outcome, defined as in-hospital death or discharge to a nursing home or rehabilitation hospital. For endovascular coil embolization, fewer data are available in the literature. For unruptured aneurysms, small case series have demonstrated mortality of 0 to 12% and morbidity 0 to 16%, with the higher rates in patients who had failed surgical attempts.20-22,31 A more recent case series of 187 subarachnoid hemorrhage cases without neurologic deficits on presentation found 0.5% mortality and 7.5% deterioration in coiled patients,28 more similar to the 0.4% mortality and 10.6% adverse outcomes we found in unruptured aneurysms. Less experience with the technique and selection of more difficult cases for coiling may account for the differences in the earlier series.

This study suggests that perioperative mortality and need for rehabilitation/nursing home transfer at discharge are lower for unruptured aneurysm patients treated by endovascular coiling compared with surgical clipping. Hospital charges and length of stay were also significantly lower in the endovascular cases. The results must be interpreted with caution because of the nature of the study. Rather than providing evidence for recommending coil embolization, the results should be seen as further impetus to compare the pervasive treatment modality, surgical clipping, with the new technology of endovascular coil embolization. Clearly, more definitive data are required with studies comparing these therapies in patients who could have received either treatment, either randomizing patients to clipping versus coiling or, if randomization is not possible, adjusting for all known potential risk factors including patient comorbidities and aneurysm location and size.

Appendix

The following institutions, in alphabetical order by state, contributed unruptured aneurysm cases to the database:

University of Alabama, University of Southern Alabama; University of Arizona; University of Arkansas; University of California Davis, UCLA, UCSD, UCSF, Stanford; University of Colorado; Yale; Georgetown; University of Florida; Emory University; Medical College of Georgia; Loyola University, University of Chicago, University of Illinois at Chicago; Indiana University; University of Iowa; University of Kansas; University of Kentucky; University of Massachusetts; University of Michigan; Fairview University of Minnesota; University of Missouri, St. Louis University; University of Nebraska; Robert Wood Johnson University; Columbia-Presbyterian, New York University, SUNY Brooklyn, SUNY Syracuse, SUNY Stony Brook; University of Eastern Carolina, University of North Carolina, Wake Forest; Medical College of Ohio, Ohio State University, University of Cincinnati, Case Western Reserve; University of Oklahoma; Oregon Health Sciences University; Allegheny University, University of Pennsylvania, Pennsylvania State University, University of Pittsburgh, Thomas Jefferson University, Lehigh Valley; Medical University of South Carolina; Vanderbilt; Memorial Hermann Hospital, University of Texas, Galveston; University of Utah; Medical College of Virginia, University of Virginia; University of Washington; West Virginia University; Froedtert Memorial Hospital, University of Wisconsin.