Background Endovascular treatment of basilar tip aneurysms is less invasive than microsurgical clipping, but requires closer follow-up.
Objective To characterize the additional costs associated with endovascular treatment of basilar tip aneurysms rather than microsurgical clipping.
Materials and methods We obtained clinical records and billing information for 141 basilar tip aneurysms treated with clip ligation (n=48) or endovascular embolization (n=93). Costs included direct and indirect costs associated with index hospitalization, as well as re-treatments, follow-up visits, imaging studies, rehabilitation, and disability. Effectiveness of treatment was quantified by converting functional outcomes (modified Rankin Scale (mRS) score) into quality-adjusted life-years (QALYs). Cost-effectiveness was performed using cost/QALY ratios.
Results Average index hospitalization costs were significantly higher for patients with unruptured aneurysms treated with clip ligation ($71 400 ± $47 100) compared with coil embolization ($33 500 ± $22 600), balloon-assisted coiling ($26 200 ± $11 600), and stent-assisted coiling ($38 500 ± $20 900). Multivariate predictors for higher index hospitalization cost included vasospasm requiring endovascular intervention, placement of a ventriculoperitoneal shunt, longer length of stay, larger aneurysm neck and width, higher Hunt-Hess grade, and treatment-associated complications. At 1 year, endovascular treatment was associated with lower cost/QALY than clip ligation in unruptured aneurysms ($52 000/QALY vs $137 000/QALY, respectively, p=0.006), but comparable rates in ruptured aneurysms ($193 000/QALY vs $233 000/QALY, p=0.277). Multivariate predictors for higher cost/QALY included worse mRS score at discharge, procedural complications, and larger aneurysm width.
Conclusions Coil embolization of basilar tip aneurysms is associated with a lower cost/QALY. This effect is sustained during follow-up. Clinical condition at discharge is the most significant predictor of overall cost/QALY at 1 year.
- basilar tip aneurysm
- cost effectiveness analysis
- clip versus coil
Statistics from Altmetric.com
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.
Basilar tip aneurysms (BTAs) represent a challenging neurosurgical entity for treatment owing to their close proximity to deep and critical structures, brainstem perforators, a deep operative corridor, and a limited number of cases even at centers of excellence.1 Microsurgical clip ligation remains a viable option for a number of select cases with comparable long-term clinical outcomes,2–4 but endovascular embolization has largely become the preferred treatment for aneurysms in this location,5 particularly with the development of stent-assisted techniques.6 Yet, aneurysm recanalization rates are higher with endovascular strategies, often requiring re-treatment.4 7 There is a shortage of investigation within cerebrovascular neurosurgery about the cost. Cost-effectiveness analyses,8 provide mixed results9–13 for guiding clip versus coil decisions, with few that account for potential re-treatment costs and none that are dedicated to the basilar tip location.
In this report, we evaluated clinical outcomes and total direct and indirect costs associated with the treatment of BTAs to perform a cost-effectiveness analysis. We sought to determine how cost effectiveness would be influenced by better clinical outcomes and higher re-treatment rates associated with endovascular embolization as compared with clip ligation.
After obtaining institutional review board approval, we retrospectively reviewed all cases of BTA treated at our institution from January 2005 through April 2017. We compiled the following information from the medical record: patient demographics; pre-existing chronic illness; smoking status; clinical presentation (including rupture status and Hunt-Hess grade if applicable); past medical history; hospital course; aneurysm morphology; treatment modality; residual neck or dome after initial treatment; procedural complications; and clinical outcomes at discharge, 3 months, 1 year, and most recent follow-up. Treatment types included ‘clip’—which included microsurgical clipping and bypass procedures—and ‘endovascular’—which included coil embolization, balloon-assisted coiling, and stent-assisted coiling. Primary outcomes of interest were the costs associated with the index hospitalization and the total overall costs associated with treatment through 1 year of follow-up divided by the cumulative quality-adjusted life-years (QALY) at 1 year (1-year cost/QALY). Separate analyses were performed for patients with ruptured and unruptured aneurysms.
All costs obtained were total actual costs for the hospitalization or clinic visit, including direct and indirect costs for facilities, personnel, supplies, and equipment; costs were obtained through the hospital billing department. In one case, only the ‘total charges’ were available, so the costs were calculated by using the average charge:cost ratio for the remaining cases. Outpatient follow-up costs (F/U) were obtained similarly and referenced to the clinical records. F/U costs included in the analysis were (1) neurosurgical clinic appointments, (2) cerebrovascular imaging studies (ie, CT angiography, magnetic resonance angiography, and digital subtraction angiography), and (3) other clinic visits related to the treatment of the BTA. Other clinic visits included appointments for headache, ophthalmologic treatments or procedures to deal with cranial neuropathies, ventriculoperitoneal shunt (VPS)-related admissions/revisions (if placed owing to subarachnoid hemorrhage from a ruptured BTA), and otolaryngology appointments and procedures for vocal cord dysfunction.
The average cost per imaging study at our institution was computed and used to estimate the costs of those studies that were performed at outside facilities where direct costs were not obtainable. Re-treatment hospitalization costs were collected for all patients requiring re-treatment of the index BTA at any time after index treatment. Post-hospitalization inpatient rehabilitation costs were collected for all patients who needed rehabilitation after the index hospitalization. Any patients with a modified Rankin Scale (mRS) score of ≥3 (ie, moderate or severe disability) had additional ‘disability’ costs added to the follow-up costs for the total duration of years that the mRS score remained at this level. This annual cost represents an averaged sum of both direct healthcare system costs and published measures of indirect costs of loss of productivity associated with a patient with a stroke in 1999, with annual adjustments based on the consumer price index.12 14 Final cost values were rounded to the nearest $100. We attempted to use only ’real' values for total cost (including both direct and indirect), with estimations only for costs related to disability.
Total costs at both 1 year and most recent follow-up were computed using the following equation:
Where INDEX is the cost of the index hospitalization, F/U, costs related to any follow-up visits and radiological studies, RETR, costs of re-treatment of the BTA after index treatment, REHAB, cost of inpatient rehabilitation, and DIS, the estimated cost of moderate or severe disability based on published rates as defined above.
Quality-adjusted life-year (QALY)
We assigned 1.0 QALY for 1 year after treatment without disability (ie, mRS score of 0 or 1), 0.75 QALY for mild disability (ie, mRS score 2), 0.50 QALY for moderate disability (ie, mRS score 3), and 0.25 QALY for severe disability (ie, mRS score 4 or 5).
Complications were categorized as either neurological or medical. Neurological complications included any new neurological deficit, and were classified as either transient or permanent. Medical complications included pneumonia, urinary tract infection, deep venous thrombosis, pulmonary embolus, acute respiratory distress syndrome, atrial fibrillation, bacteremia, Clostridium difficile colitis, and gastrointestinal hemorrhage. Technical complications included those that did not incur any neurological dysfunction, such as wound infection, vessel dissection or asymptomatic pseudoaneurysm, access site hematoma, cerebrospinal fluid leak, postoperative hematoma requiring evacuation, coil herniation into the parent vessel, intra-arterial thrombus formation without symptomatic cerebral infarct, and alopecia from radiation. Transient neurological complications included cranial nerve palsies, aphasia, or hemiparesis that resolved within 2 months of the index procedure. Permanent neurological complications included any permanent cranial neuropathy or stroke with neurological deficit(s) that persisted beyond 2 months or at the most recent follow-up.
Differences among various sample characteristics (table 1) were evaluated for statistical significance using Mann-Whitney U and Fisher’s exact tests, as appropriate. Differences among treatment modalities in hospitalization cost (table 2) were evaluated parametrically using t-tests, while differences in cost-related variables at follow-up (table 3) were evaluated non-parametrically using Mann-Whitney U tests. Both sets of analyses were corrected for multiple comparisons using the Holm-Bonferroni adjustment.15
The following 23 variables were tested in univariate and multivariate analyses as potential predictors of INDEX: treatment date, age at treatment, gender, discharge disposition, endovascular treatment, aneurysm height, aneurysm width, aneurysm neck, dome:neck ratio, aspect ratio, complications, Hunt and Hess grade, Fisher scale, history of hypertension, history of diabetes mellitus, history of hyperlipidemia, history of smoking, hydrocephalus requiring external ventricular drain, hydrocephalus requiring permanent VPS, asymptomatic cerebral vasospasm as diagnosed by transcranial Doppler sonography, symptomatic vasospasm requiring endovascular intervention, vasospasm-associated stroke, and total length of stay (LOS). The following 19 variables were tested in univariate and multivariate analyses as potential predictors of cost/QALY at 1 year: treatment date, age at treatment, gender, treatment modality (endovascular versus clip), residual neck or aneurysm dome, aneurysm height, aneurysm width, aneurysm neck, dome:neck ratio, aspect ratio, complications, VPS placement, tracheostomy, percutaneous gastrostomy, mRS score at presentation, mRS score at discharge, re-treatment, recurrence, and re-treated recurrence.
Multivariate models of prediction for INDEX (online supplementary tables 1,2) and cost/QALY at 1 year (online supplementary tables 3,4) were generated using a forward-selection, cross-validated linear regression approach. Although INDEX sufficiently met its distributional assumption of normality, cost/QALY first required a log-transformation, which thus implies that all resulting model coefficients are to be interpreted as multiplicative (rather than additive) changes. Initially, 19 variables were selected as potential predictors based on historical relevance, and then a multivariate model was built for each outcome in steps by successively entering the most statistically significant effect from the remaining set of predictors into the multivariate model, until no further effects could be included at a significance threshold of p<0.05. The resulting model was then checked for susceptibility to overfit using ‘k-fold cross-validation",16 in which a ‘prediction’ R2 value is calculated in an attempt to assess how the model would be expected to perform on new observations. Similar to jackknife analyses, the ’k-fold' approach is a variant of cross-validation in which the sample is randomly split into ’k' subsets, and each subset is held out as a validation set for the model constructed using the same algorithm on the remainder of the sample. The sum-of-squares statistics from the resulting ’k' models are then aggregated to generate a ‘prediction’ R2 value for the model built from the full sample, which can be interpreted in the same way as a regular R2 except in the context of new observations rather than based on the data that were used to construct the model.
Details of the overall cohort are displayed in table 1. Average follow-up was 33.3 months. Patients with unruptured aneurysms treated with clip had significantly longer LOS than those treated with coil (14.3 vs 3.4 days, p<0.001). Ruptured clip patients had higher rates of vasospasm and higher rates of endovascular intervention. While endovascular patients had higher rates of technical complications without any neurological sequelae (which included in the unruptured coil group intra-arterial thrombus (7), groin hematoma (5), vessel dissection (1), intraparenchymal hemorrhage (1), and alopecia (1)), clip patients had higher rates of transient neurological, permanent cranial nerve, and permanent neurological complications. Ruptured clip patients required permanent VPS more often. Unruptured clip patients were more likely to be discharged to either inpatient rehabilitation or a skilled nursing facility. Clinical outcomes were better in the endovascular group for both ruptured and unruptured aneurysms, but there were comparable rates of ‘good’ clinical outcomes (ie, mRS score 0–2) between clip and endovascular groups during follow-up. INDEX was higher for clip patients for both ruptured and unruptured aneurysms.
Table 2 shows INDEX for unruptured (top) and ruptured (bottom) aneurysms based on intervention type. Unruptured patients treated with clip had higher INDEX values than those treated with coil (either simple coiling or balloon-assisted coiling, p=0.007), stent-assisted coiling (p=0.018), and all combined endovascular interventions (p=0.001). Similarly, ruptured patients treated with clip had higher index costs than those treated with coil (p=0.002) and with all endovascular interventions combined (p=0.002). Temporal trends in endovascular and clip INDEX values are displayed in online supplementary figures 1 (unruptured) and 2 (ruptured). Unruptured aneurysm treatment demonstrated a steady but subtle increase for both clip and endovascular interventions, whereas ruptured aneurysm endovascular treatment remained relatively stable. Ruptured clip aneurysm treatment showed the highest trend in increased INDEX values. All comparisons remained significant (p<0.05) after adjusting for multiple comparisons, with the exception of unruptured clip versus SAC.
Table 3 displays the 1-year and most recent follow-up data for both average total cost of care, QALY, and cost/QALY. At 1 year, average total cost was significantly higher for clip patients than for endovascular patients ($109 000 vs $47 000 p=0.004), with comparable QALY values (0.89 vs 0.97, p=0.301), and a higher cost/QALY ($52 000/QALY vs $137 000/QALY, p=0.006). This relationship remained at the most recent follow-up, though without statistical significance. Given that endovascular treatment achieved both lower cost and higher QALY, no incremental cost-effectiveness ratio (ICER, a ratio used to compare the benefit in QALY observed with one treatment with the additional cost incurred with that treatment) was obtained. One-year costs for the ruptured cohort were comparable between clip and endovascular ($144 000 vs $118 000, p=0.327), with a significantly worse QALY (0.75 vs 0.84, p=0.037), and a non-statistically significant higher cost/QALY ($233,000/QALY vs $193,000, p=0.277). At the most recent follow-up, clip patients had higher total QALY, though this was due to longer follow-up, as clip patients had a lower average QALY per year. Ruptured patients treated with clip had higher total cost at the most recent follow-up compared with ruptured endovascular patients, but this did not achieve significance. Clip patients did have a lower cost/QALY, though this was also due to longer follow-up. All comparisons remained significant (p<0.05) after adjusting for multiple comparisons, with the exception of unruptured average cost at the most recent follow-up for clip versus endovascular, and ruptured average QALY at 1 year for clip versus endovascular.
Predictors of cost and cost-effectiveness
Univariate and multivariate predictors of INDEX are displayed in online supplementary tables 1 and 2 for patients with unruptured and ruptured aneurysms, respectively. Cross-validated R2 values are provided for the multivariate models, which represent the proportion of variability that is explained by the model. For unruptured patients, endovascular intervention was significant in univariate, but not in multivariate analysis as a predictor of higher INDEX. Vasospasm intervention (which occurred in only one unruptured patient owing to a new diagnosis of congenital vasculitis, not related to any rupture event) was the strongest predictor of higher INDEX (B=68.23, p=0.009), followed by placement of a VPS (B=35.52, p<0.001, which occurred in two patients), longer LOS (B=3.26, p<0.001), larger aneurysm neck (B=2.64, p<0.001), and later treatment date (B=2.41, p<0.001). For ruptured patients, clip was significantly predictive of higher INDEX in multivariate analysis (coil B=−8.42, p=0.02), as well as vasospasm intervention (B=17.53, p=0.045), vasospasm-related stroke (B=27.8, p<0.001), an associated complication (B=6.61, p<0.001), longer LOS (B=3.19, p=0.003), higher Hunt-Hess grade (B=4.05, p=0.041), later treatment date (B=2.91, p<0.001), and larger aneurysm width (B=2.34, p=0.019). History of hypertension (B=−6.8, p=0.009) and older age (B=−0.33, p<0.001) were predictive of lower INDEX.
Total cost/QALY for unruptured and ruptured aneurysms at 1 year after intervention is displayed in online supplementary tables 3 and 4, respectively. For unruptured aneurysms, a higher mRS score at discharge (Ratio=1.59, p<0.001) and associated complications (ratio=1.57, p=0.011) were predictive of higher cost/QALY in multivariate analysis, though treatment intervention was significant only in univariate analysis. For ruptured patients, higher mRS score at discharge (ratio=1.7, p<0.001) and larger aneurysm width (ratio=1.06, p=0.018) were predictive of higher cost/QALY in multivariate analysis.
Data extrapolation for long-term prediction
Figure 1 shows the total cost/QALY values (y axis) plotted for each patient with at least 1 year of follow-up, including both the 1-year and most recent follow-up data points, for unruptured (solid dot) and ruptured (open dot) clip (blue) and endovascular (red) cohorts. Linear models of best fit are applied to the unruptured (solid lines) and ruptured (dotted lines) data, with endovascular showing a statistically significantly lower pattern in the former (p=0.004), consistent with univariate results produced in online supplementary table 3. At no point during data extrapolation to 10 years did clip achieve lower cost/QALY then endovascular.
Here we present an updated account of clinical outcomes for a larger cohort of patients (141) at an average of 33.3 months' follow-up, with all associated complications. Additionally, we provide a cost-effectiveness analysis comparing endovascular therapy with open microsurgical obliteration for ruptured and unruptured BTAs. We have previously published details of a portion of this series (100 patients) showing clinical outcomes after microsurgical and endovascular treatment, including information on surgical approach, and comparable outcomes at 3 months and 1 year between the two groups (eg, similar rates of a ‘good clinical outcome’, defined as an mRS score of 0–2).3 Additionally, we have shown that BTA recurrence (or progression of residual aneurysm) and re-treatment are particularly problematic in this location as compared with other control locations, more so in ruptured cases, and those with endovascular treatment and a quicker time interval before re-treatment.17 In the current analysis, patients undergoing clip treatment have higher rates of permanent neurological morbidity (cranial neuropathy and otherwise), regardless of rupture status, but similar rates of medical complications, percutaneous gastrostomy and tracheostomy, and mortality (table 1). Accordingly, clip patients with unruptured aneurysms have longer LOS than the coil group, which contributes significantly to increased cost. Although rates of ‘good clinical outcome’ (eg, mRS score 0–2) are comparable statistically at every time point (including discharge, 3 months, 1 year, and most recent follow-up), endovascular treatment confers a benefit on average functional outcome scores that is sustained throughout all the follow-up in the ruptured cohort and through 3 months in the unruptured group.
Our data show that in addition to improvement in functional outcomes, endovascular intervention is significantly less expensive at the time of initial treatment (INDEX) than with clip, though not necessarily when a stent is used as an adjunctive device (tables 1 and 2). INDEX costs have increased over the past decade (online supplementary figures 1 and 2), consistent with inflation, so it comes as no surprise that a later treatment date predicted higher INDEX values in multivariate modeling for both ruptured and unruptured BTAs (online supplementary tables 1 and 2). Endovascular treatment yielded a superior average cost/QALY at 1 year and most recent follow-up for unruptured aneurysms, though only the former was statistically significant (table 3). The results were more comparable for ruptured aneurysms, though endovascular treatment was a predictor of lower INDEX in multivariate analysis, suggesting that poor clinical outcomes at discharge and larger aneurysm width play a stronger role in determining 1-year cost/QALY (online supplementary table 3). Finally, endovascular patients maintained a lower cost/QALY, regardless of rupture status, when the data were expanded to include most recent follow-up and re-treated recurrences, and extrapolated based on the trend (figure 1).
King et al first introduced cost-effectiveness analysis in the decision as to whether or not clip treatment of unruptured aneurysms was cost effective.18 Costs were estimated from Medicare data, and adjusted for inflation, using Markov modeling, but not accounting for potential re-treatments. The results nonetheless showed that the ICER for clipping an unruptured aneurysm in a middle-aged patient compared with observation was $24 200/QALY, well below the subjectively accepted threshold of $50 000/QALY, representing a ‘cost-effective’ strategy.19
Following the early introduction of endovascular therapy, Hoh et al analyzed the differences in costs associated with clip and endovascular treatment at the University of Florida, for both ruptured and unruptured aneurysms, using total hospital costs (similar to the current analysis, which includes both direct and indirect costs), and other cost metrics (including surgeon billing, hospital billing, surgeon collections, hospital collections, and payer data).9 Importantly, this was neither a cost-effectiveness analysis, nor a cost analysis that accounted for re-treatments later, but rather focused on INDEX values. Interestingly, clip patients with an unruptured aneurysm had lower hospital costs for both ruptured and unruptured aneurysms, despite longer LOS in the former. The authors concluded that this was probably owing to the higher cost of endovascular devices at that time. Larger aneurysm size and worse clinical grade (for ruptured cases) both predicted higher hospital costs, consistent with the current analysis.
Maud et al used clinical outcome data from the International Subarachnoid Aneurysm Trial (ISAT)20 and cost estimates from Premier Perspective Comparative Database for 2005–2006 to perform a cost-effectiveness analysis comparing surgery with endovascular therapy for ruptured aneurysms.12 The group also accounted for the costs associated with follow-up angiography, disability, and re-treatments. In this analysis, the group found higher costs and QALY at 1 year for the endovascular group compared with clip ($65 424/QALY vs $64 824/QALY, respectively). Accordingly, an ICER was obtained and came to $72 872/QALY for endovascular compared with clip, suggesting that endovascular therapy was not necessarily ‘cost-effective’ at 1 year—at least using costs from 2005. The authors astutely note that this ICER would probably decrease if the analysis was extended beyond 1 year, given that the majority of re-treatments for the endovascular group occurred within that period of time.
Despite including patients treated between 2005 and 2007, similar to the two aforementioned studies, our analysis conflicts with the notion that endovascular treatment costs more than clipping, at least in multivariate analysis for ruptured aneurysms. Moreover, the role that endovascular treatment plays in predicting cost-effectiveness seems less important (online supplementary tables 3 and 4), whereas aneurysm size and clinical condition at discharge appear more influential. More contemporary series do consistently identify lower INDEX costs in patients receiving endovascular treatment compared with clip for unruptured aneurysms.10 11
Further confusing the matter, are regional discrepancies in cost associated with care; Zygourakis et al identified the western United States region as an independent predictor of higher cost for craniotomy21 and spinal deformity surgery22 in multivariate modeling. Our results are consistent with those from Gonda et al,10 where a statewide database was used to track re-treatments and clinical outcomes from 1998 to 2005 for unruptured aneurysms. The group found higher re-treatment rates with endovascular interventions, but higher cumulative cost with clip treatment and re-treatments (secondary outcome), which was sustained throughout all of follow-up. In other words, endovascular treatment produced initially superior clinical outcomes, with higher re-treatment rates, but lower overall cumulative costs, and clinical outcomes that eventually equalized those in the clip group. Similar to this study, our analysis suggests no need to compute an ICER comparing clip and endovascular interventions, as endovascular treatment was associated with better clinical outcomes and lower costs, even when accounting for re-treatments.
Inherent to the design of this retrospective study is a selection bias, as complex aneurysms with wide necks (which was an independent predictor of higher INDEX cost for unruptured BTAs) were usually triaged to the clip group based on our institutional experience and known challenges in endovascular treatment of wide-necked ruptured BTAs.3 Thus, this does not represent a randomized or prospective study, and thus the results must be interpreted with caution. We do, however, capture a relatively robust, ‘real-world’ cohort of patients with a rare disease location, and extract costs associated with treatment (including both direct and indirect components), an outcome that is usually estimated from larger national databases. With this increased granularity comes an improved resolution for understanding the specific complications, clinical outcomes, health-related quality of life, and costs associated with BTAs.
Despite higher rates of re-treatment, endovascular therapy for BTAs produces a statistically significant improvement in clinical outcomes which, coupled with lower costs, yields a more cost-effective solution than microsurgical clipping. Larger aneurysm width, higher mRS score at discharge, and procedural complications are independent predictors of worse cost/QALY ratios at 1 year.
Contributors Conception or design of the work: IJA, ML, BG, LJK, LNS. Data collection: IJA, RS, CMK. Data analysis and interpretation: IJA, CMK, JB. Drafting the article: IJA, CMK, SL, RS. Critical revision of the article: ML, LJK, LNS. Final approval of the version to be published: IJA, RS, CMK, SL, JB, BG, ML, LJK, LNS.
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 BG: Viket Medical, Inc (equity interest), ML: National Institute of Neurological Disorders and Stroke, National Institutes of Health (grant), American Heart Association (grant), Stryker, Inc (unrestricted educational grant), Covidien (unrestricted educational grant), Medtronic (consulting), Minnetronix (consulting), Proprio (equity interest). LJK: National Institute of Neurological Disorders and Stroke, National Institutes of Health (grant), Microvention, Inc (consulting), Spi Surgical, Inc (equity interest). LNS: Spi Surgical, Inc (shareholder), Viket Medical, Inc (shareholder).
Ethics approval University of Washington institutional review board #5703.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Data is available under request.
Patient consent for publication Not required.