Article Text
Abstract
Background Flow modulation is the newest endovascular technique for treatment of cerebral aneurysms.
Objective To investigate changes in aneurysm treatment practice patterns in the USA.
Methods From the 2007 to 2016, the National Inpatient Sample databases, hospital discharges associated with unruptured aneurysms (UA), and/or ruptured aneurysms (RA) having undergone surgical clipping (SC) and/or endovascular treatments (EVT) were identified using the International Classification of Diseases codes. Patient demographics, hospital characteristics, and clinical outcomes were reviewed. Five year subgroup analyses were performed for treatment differences.
Results A total of 39 282 hospital discharges were identified with a significant increase in EVT (UA: SC n=7847 vs EVT n=12 797, p<0.001; RA: SC n=8108 vs EVT n=10 530, p<0.001). Hospitals in the South demonstrated the most significant EVT use regardless of aneurysm status (UA: SC n=258.5±53.6 vs EVT n=480.7±155.8, p<0.001; RA: SC n=285.6±54.3 vs EVT n=393.3±102.9, p=0.003). From 2007 to 2011, there was no significant difference in the mean number of cases for the treatment modalities (UA: SC n=847.4±107.7 vs EVT n=1120.4±254.1, p=0.21; RA: SC n=949.4±52.8 vs EVT n=1054.4±219.6, p=0.85). Comparatively, from 2012 to 2016, significantly more UA and RA were treated endovascularly (UA: SC n=722.0±43.4 vs EVT n=1439.0±419.2, p<0.001; RA: SC n=672.2±61.4 vs EVT n=1051.6±330.2, p=0.02).
Conclusions As technological innovations continue to advance the neuroendovascular space, the standard of care for treatment of cerebral aneurysms is shifting further towards endovascular therapies over open surgical approaches in the USA.
- aneurysm
- flow diverter
- coil
- stent
Statistics from Altmetric.com
Introduction
Approximately 3–5% of Americans harbor a cerebral aneurysm (CA). Every year, an estimated 30 000 CA rupture in the USA. Outcomes after a ruptured aneurysm (RA) vary widely, from full recovery to irreversible brain damage to death.1 Prehospital mortality from a RA ranges from 10% to 25%, while the 30 day mortality rate is approximately 45%.2 3
Traditionally, CA were treated with open microsurgical clipping beginning in the 1930s. The invention of Gugliemi coils in the 1990s ushered in the field of neuroendovascular with minimally invasive alternatives.4 In 2002, the landmark study International Subarachnoid Aneurysm Trial (ISAT) reported that patients with RA treated with endovascular coiling had a lower mortality rate and superior clinical outcome after 12 months of follow-up than those treated with open surgery.5 This study revolutionized the clinical practice management of RA. In 2012, Lin et al evaluated national treatment trends for CA in the USA and found that during the 5 years post ISAT, the per cent of RA treated with endovascular therapy increased to 42.9% (2002–2007) from 9.3% (1998–2001) while the per cent treated with open surgery decreased to 57.1% (2002–2007) from 90.7% (1998–2001).6 While the focus of the ISAT was on RA, the less invasive nature of endovascular therapy (EVT) over open surgical clipping (SC) in RA also led to increasing endovascular use for unruptured aneurysms (UA). The neurointerventional space has continued to evolve with improved technologies in coiling, new stents, catheters and, more recently, an alternative technique for treating aneurysms based on braided mesh devices. Within this context, we investigated changes in national trends since 2007 using the National Inpatient Sample (NIS) database.
Methods
The Agency for Healthcare Research and Quality and the US Department of Health and Human Services support the Healthcare Cost and Utilization Project (HCUP).7 NIS databases, 2007 to 2016 (most recent year available), were purchased from the HCUP Central Distributor (Rockville, Maryland, USA) to investigate the national trends in CA treatment since the 2012 Lin et al publication.6 8
Patients were selected based on their diagnosis followed by their treatment method. Prior to October 1, 2015, the International Classification of Diseases, ninth revision, clinical modification (ICD-9-CM) was used. Thereafter, the tenth revision, clinical modification (ICD-10-CM) and procedure coding system (ICD-10-PCS) were used. The ICD-9-CM diagnostic code 437.3 (cerebral aneurysm, non-ruptured) was used to identify patients with UA and codes 430.0 (subarachnoid hemorrhage) and 431.0 (intracerebral hemorrhage) were used to identify patients with RA, consistent with Lin et al.6 The ICD-10-CM codes used after October 1, 2015 are listed in online supplementary table 1S.9 Patients coded as having UA and RA in one hospitalization were placed in the RA group, consistent with Lin et al.6 The ICD-9-CM procedure codes 39.51 (clipping of aneurysm) defined patients who underwent open surgery while 39.72 (endovascular (total) embolization or occlusion of head and neck vessels) and 39.79 (other endovascular procedures on other vessels) defined patients who underwent endovascular therapy. In addition, patients with both 39.52 (other repair of aneurysm) and 88.41 (arteriography of cerebral arteries) without open surgery were also considered as having undergone endovascular therapy to be consistent with Lin et al’s analysis.6 The ICD-10-PCS codes for treatment methods after October 1, 2015 are listed in online supplementary table 2S.10
Supplemental material
Patient demographics, hospital characteristics, and clinical outcomes except Elixhauser comorbidity score, were collected directly from the NIS database. The NIS provides 29 of the original 30 Elixhauser comorbidity indicators (cardiac arrhythmia is excluded). The Elixhauser comorbidity score was calculated using the Walraven point system developed in 2009 and validated for use with NIS databases by Thompson et al.11 12 Patients treated with more than one modality were analyzed separately. Five year subgroup analyses were performed for treatment differences given the introduction of new endovascular techniques with the Food and Drug Administration (FDA) approval of the pipeline embolization device (PED; Medtronic Neurovascular, Irvine, California, USA) flow diverter in April 2011.13
Statistical tests included the χ2 test, Fisher’s exact test, Student’s t test, one way ANOVA, and two way ANOVA. Statistical programs used were IBM SPSS Statistics Premium GradPack V.26 and GraphPad Prism 8.
Results
Single treated cerebral aneurysms
From 2007 to 2016, 39 282 hospital discharges were associated with single treated CA (online supplementary tables 3S and 4S). In the UA group, 7847 (38.0%) patients underwent SC and 12 797 (62.0%) underwent EVT. The χ2 analysis showed a significant shift toward EVT compared with expected projections based on the 1998–2007 data shown by Lin et al (SC n=7597 (51.5%) and EVT n=7168 (48.5%), p<0.001).6 In the RA group, 8108 (43.5%) patients underwent SC and 10 530 (56.5%) underwent EVT. Similarly, χ2 analysis revealed a significant shift towards EVT based on projections from the 1998–2007 data set (SC n=14 514 (72.1%) and EVT n=5620 (27.9%), p<0.001).6 Patient demographics and hospital characteristics are shown in table 1.
Total numbers of UA and RA treated with SC or EVT each year (1998–2016), combining the 1998–2006 data presented in Lin et al, are shown in figure 1A,B, respectively.6 From 2007 to 2016, mean number of UA and RA treated with EVT was significantly higher than with SC (UA: EVT n=1279.7±367.4 vs SC n=784.7±101.8, p<0.001; RA: EVT n=1053.0±264.4 vs SC n=810.8±155.8, p=0.02).
In the 5 year subgroup analyses, similar numbers of total UA and RA were treated (online supplementary table 5S). From 2007 to 2011, similar numbers of each treatment modality were performed. In contrast, figure 1C,D for the 2012–2016 period after introduction of the PED, shows a significant increase in endovascular use for both UA (p<0.001) and RA (p=0.02). This corresponded to a 9.7% increase in EVT for UA (56.9% to 66.6%) and an 8.4% increase for RA (52.6% to 61.0%). Online supplementary tables 6S and 7S provide patient and hospital characteristics for the subgroup analyses of UA and RA, respectively.
To assess if the shift in treatment modality was immediate after 2011, we examined the number of cases treated from 2012 to 2013. For UA, significantly more cases were treated with EVT (p=0.02), suggesting this shift was immediate (red arrow in figure 1A). In contrast, for RA during the same time frame, similar numbers of cases were treated with each modality (p=0.21). It was not until 2015 when there was significantly more RA treated with EVT (63.9%) compared with 2014 (EVT 55.7%) (p<0.001), suggesting the shift was delayed (red arrow in figure 1B).
Hospital regional differences
For UA treatments, the South demonstrated the most significant EVT use (p<0.001) (figure 2A). The Midwest also demonstrated significantly more EVT use (p=0.04). In comparison, the Northeast and West exhibited no significant differences between SC and EVT numbers (Northeast: p=0.052; West: p=0.14). For RA treatments, only the South exhibited significantly higher EVT use (p=0.003) (figure 2B). Other regions demonstrated no significant differences between treatment modalities (Northeast: p=0.09; Midwest: p=0.84; West: p>0.99).
In the 5 year subgroup analyses, figure 2C demonstrates that there were significantly more EVT treatments for UA in the Midwest, South, and Northeast from 2012 to 2016 than from 2007 to 2011 (online supplementary table 6S). In contrast, figure 2D shows for RA treatments, only the South demonstrated a significantly higher EVT use from 2012 to 2016 (p<0.001) compared with 2007–2011 (online supplementary table 7S).
Hospital type
Table 1 presents data on hospital type. Few data were missing from the NIS (UA: SC n=63 and EVT n=60; RA: SC n=24 and EVT n=84). From 2007 to 2016, urban teaching hospitals treated significantly more UA with EVT (p<0.001). This difference was due to an increase in EVT beginning in 2012 (2007–2011: p=0.08; 2012–2016: p<0.001) (online supplementary table 6S). In other hospital types, similar numbers of SC and EVT were performed from 2007 to 2016 (table 1).
Similarly, from 2007 to 2016, urban teaching hospitals treated significantly more RA with EVT (p<0.001) (table 1). This difference was due to an increased use of EVT after 2011 (2007–2011: p=0.67; 2012–2016: p<0.001) (online supplementary table 7S). In other hospital types, similar numbers of SC and EVT were performed from 2007 to 2016 (table 1).
Hospital size
The same patients who had missing data for hospital type also had missing data for hospital size. From 2007 to 2016, large hospitals treated significantly more UA with EVT (p<0.001). This difference was consistent over the decade (2007–2011: p=0.03; 2012–2016: p<0.001) (online supplementary table 6S). In medium and small sized hospitals, similar numbers of cases were treated with either treatment modality over the decade (table 1).
For RA, from 2007 to 2016, significantly more EVT were performed in large hospitals (p<0.001). This difference was due to an increased use of EVT after 2011 (2007–2011: p=0.56; 2012–2016: p<0.001) (online supplementary table 7S). For small and medium sized hospitals, similar numbers of cases were treated with either treatment modality over the decade (table 1).
Inhospital mortality
Mortality rates from 2007 to 2016 based on aneurysm status and treatment modality are shown in figure 3A, with 0.06% of the data missing. For UA, a similar mortality rate was found between the treatment modalities (p=0.90). The two deceased groups had a similar mean age (p=0.33) and mean Elixhauser comorbidity scores (p=0.51). In contrast, for RA, a significantly higher mean mortality rate was associated with EVT (12.9%±1.6%) than SC (11.3%±0.7%) (p=0.01). The deceased EVT group had a significantly older patient population with a mean age of 61.2±15.3 years compared with 58.2±14.0 years for SC (p<0.001), and a significantly higher Elixhauser comorbidity score (mean 5.36±6.70 compared with 4.67±5.96 for SC; p=0.02).
In the 5 year subgroup analyses, the mean mortality rate remained stable over the decade for UA (p=0.52) and RA (p>0.99). From 2012 to 2016, there was no significant difference in the mean mortality rate between the treatment modalities for UA and RA (UA: p>0.99; RA: p=0.25) (online supplementary tables 6S and 7S).
Over the decade, mortality rate was lower at urban teaching hospitals compared with other location/teaching hospital types for RA. There was no relationship between mortality rate and hospital size for both UA and RA (online supplementary table 8Sa,b).
With respect to treatment modality, over the decade, mortality rate was lower at urban teaching hospitals compared with urban non-teaching hospitals for RA treated surgically. For RA treated endovascularly, mortality rate was lower at urban teaching hospitals compared with other location/teaching hospital types. There was no relationship between mortality rate and hospital size for both UA and RA (online supplementary table 9Sa,b).
Length of stay
Data on length of stay (LOS) are shown in table 1, excluding 0.005% of missing data. From 2007 to 2016, SC was associated with a longer LOS for UA (p<0.001) and RA (p<0.001).
In the 5 year subgroup analyses, LOS was consistent over the decade for UA and RA, with SC associated with a significantly longer LOS (UA 2007–2011: p<0.001; UA 2012–2016: p<0.001; RA 2007–2011: p<0.001; RA 2012–2016: p=0.002) (online supplementary tables 6S and 7S).
Total charges
Data on total charges (TC) are shown in table 1, excluding 2.11% of missing data. From 2007 to 2016, for UA, similar TC was found between SC and EVT (p=0.90). By comparison, for RA, a significantly higher mean TC was associated with EVT (p=0.003).
In 5 year subgroup analyses, UA treated with SC was associated with a significantly higher mean TC from 2007 to 2011 (p=0.003) but similar mean TC from 2012 to 2016 (online supplementary table 6S). For RA, there was no significant difference in mean TC over the decade (2007–2011: p=0.20; 2012–2016: p=0.71) (online supplementary table 7S).
Doubled treated cerebral aneurysm
From 2007 to 2016, a total of 630 hospital discharges were associated with double treated cerebral aneurysms. In the UA group, 118 patients underwent surgical clipping first (SCF) and 22 underwent endovascular therapies first (EVTF). In the RA group, 324 underwent SCF and 166 underwent EVTF. Patient characteristics were analyzed but not shown because HCUP prohibits the publication of cell tables with n<10. Similar patient demographics were observed between aneurysm status and treatment modality.
The total numbers of UA and RA treated with SCF and EVTF each year over the decade are shown in figure 4A,B, respectively. Regardless of aneurysm status, the mean number of cases treated with SCF was significantly higher than with EVTF (UA: p=0.007; RA: p=0.002).
In the 5 year subgroup analyses, significantly fewer cases were treated after 2011 (UA: p=0.03; RA: p=0.11). The decrease was due to fewer aneurysms treated with SCF (UA: p<0.001; RA: p<0.001). From 2007 to 2011, significantly more aneurysms were treated with SCF (UA: p<0.001; RA: p<0.001) (figure 4C,D). Comparatively, the 2012–2016 period showed similar numbers of cases treated with SCF and EVTF (UA: p=0.39; RA: p>0.99) (figure 4C,D).
Inhospital mortality
Mortality rates for each year based on aneurysm status and treatment modality are shown in figure 3B. Overall, the mean mortality rate was significantly lower for UA than RA (n=1.8%±3.2% and n=13.3%±6.6%, respectively, p<0.001). Similar mean mortality rates were observed between the treatment modalities for UA (p=0.10) and RA (p=0.24).
Length of stay
From 2007 to 2016, the length of hospitalization was similar between the treatment modalities for UA (p=0.78) and RA (p=0.503).
In the 5 year subgroup analyses, LOS was consistent over the decade for UA and RA, with similar LOS between SCF and EVTF (UA 2007–2011: p=0.71; UA 2012–2016: p=0.14; RA 2007–2011: p=0.77; RA 2012–2016: p=0.26).
Total charges
A small (1.59%) proportion of data were missing. For UA, similar mean TC was found between SCF and EVTF (p=0.11). In contrast, a significantly higher mean TC was associated with RA treated with EVTF (p<0.001). From 2007 to 2011, UA treated with EVTF were associated with a significantly higher mean TC (p=0.03). In contrast, from 2012 to 2016, similar mean TC was found between the treatment modalities for UA (p=0.70). For RA, during 2007–2011, similar mean TC was found between SCF and EVTF (p=0.17) whereas during 2012–2016, EVTF was associated with a significantly higher mean TC (p<0.001).
Discussion
Analysis of the national treatment trends for CA previously focused on the 1998–2007 period after the birth of the endovascular space with the Gugliemi coils invention in the 1990s and the subsequent 2002 ISAT landmark study.4–6 The publication of Lin et al demonstrated for RA a national trend in the 1998–2007 time frame towards declining numbers of SC (1998: n=1656 vs 2007: n=1027) and increasing EVT (1998: n=100 vs 2007: n=970). For UA, there was the beginning of a shift towards increasing EVT (1998: n=92 vs 2007: n=1233) but a steady number of SC (1998: n=668 vs 2007 n=750).6 Subsequently, Brinjiki et al performed an analysis of the NIS from 2001 to 2009 to identify age related trends in the treatment and outcomes specific to ruptured cerebral aneurysms. They found, similarly, a significant increase in the number of patients with RA treated with endovascular coiling across all age groups and each increasing age group was associated with an increased proportion of patients treated with coiling.14
Since the early 2000s, the neuroendovascular space has continued to evolve and, more recently, ushered a new era of flow diversion techniques as alternatives to coiling. Our study provides an indepth analysis of changes in practice patterns since 2007. With the FDA approval of the PED flow diverter in 2011 as an alternative endovascular treatment for UA located along the internal carotid artery, we hypothesized an increased use of EVT over SC for UA, beginning in 2012. Indeed, we found a shift towards EVT use after 2011 for UA. Interestingly, we also found a significant, although smaller, shift towards EVT use for RA, which was unexpected given that the PED was approved for treating UA. Perhaps the continued successful usage of EVT for UA influenced the practice patterns for RA.15 16 Our data support this explanation, given a delay was observed in a paradigm shift for RA, beginning more so after 2015 compared with the immediate effect for UA in 2012 (figure 1A,B). For aneurysms that underwent both treatment modalities, we noted a significant reduction in the number of surgical first cases beginning in 2012. This observation may be attributed to the significant shift towards endovascular approaches.
Our analysis demonstrated that the shift in clinical practices was not uniform across the USA. Overall, the South displayed the greatest adoption of EVT. This may be explained by the US census data on the estimated population within each region, demonstrating a significantly higher population density in the South, which may lead to the much higher volume observed compared with the other regions.17 While the Northeast, Midwest, and South all performed more EVT than SC for UA beginning in 2012, the West lagged significantly behind, perhaps secondary to the West demonstrating a higher percentage of aneurysm treatments performed at urban non-teaching hospitals and smaller hospitals (online supplementary tables 10Sa,b and 11Sa,b). Urban teaching hospitals and large hospitals treated most of the aneurysms with EVT modalities. For other hospital types/sizes, the trend towards an increased use of EVT was evident beginning in 2012 although not significant. Collectively, the data demonstrate a continued shift towards EVT.
While Lin et al found a 1.2% mortality rate associated with UA and a 13.7% mortality rate associated with RA from 1998 to 2007, our analysis revealed lower mortality rates for UA (0.4%) and RA (12.2%) from 2007 to 2016.6 The reduction was most likely multifactorial (eg, increased surgeon experience and technological advances). Clinical outcomes for RA showed a higher mortality rate associated with EVT. This could be explained by a significantly older patient population with a mean age of 61.2±15.3 years for EVT compared with 58.2±14.0 years for SC (p<0.001), and a significantly higher Elixhauser comorbidity score (mean 5.36±6.70 for EVT compared with 4.67±5.96 for SC; p=0.02). Our data for UA showed that the mortality rate remained stable over the decade, suggesting that perhaps the learning curve with the introduction of flow diversion and PED was not steep or difficult to overcome, among potential other factors such as improvements in catheter access products.
Analysis of LOS and total hospital charges during the 2007–2016 period demonstrated shorter LOS for aneurysms treated with EVT, consistent with the analysis of Hoh et al using NIS 2002–2006.18 Interestingly, unlike Hoh et al, surgical cases were not associated with higher total hospital charges.18 The inconsistencies could be due to differences in the number of participating states in the HCUP each year (eg, more states represented in 2007–2016), the type of data provided by each state, and the presentation of the NIS.19
Limitations of this study include the retrospective nature with known inherent flaws of missing data that impede comprehensive data analysis. Data collected were limited to those provided by the NIS. For example, NIS does not allow for specific differentiation of endovascular modalities. As such, the data analysis provided attempts to offer indirect correlations based on the timeline of FDA approval of the PED in 2011 in comparison with pre-existing (prior to 2011) devices, such as stents, coils, and balloons, where no significant changes in volume were observed in the 2007–2011 subgroup analyses. In addition, the validity of the data may be dependent on the validity of the hospital reporting. Inaccurate coding may also contribute to study limitations. The transition from ICD-9 to ICD-10 adds another layer of possible misinformation. Data from the NIS commonly lags approximately 2 years, limiting the analysis up to 2016. Since then, an additional flow diverter, the Surpass Streamline Flow Diverter (Stryker Neurovascular, Fremont, California, USA), and the Woven EndoBridge (WEB) Aneurysm Embolization System (MicroVention/Sequent Medical, Inc, Irvine, California, USA) flow disruptor both received approval from the FDA in 2018.20 21 Most recently in December 2019, the FDA approved another flow diverter, Flow Re-Direction Endoluminal Device (FRED; MicroVention, Irvine, California, USA).22 Nevertheless, the NIS remains the preferred tool to investigate national trends as it is the largest representative sample available in the USA.
Conclusion
Treatment paradigms for CA continue to shift towards EVT. The introduction of flow diversion technology in April 2011 may have contributed to the rise in EVT since then. Clinical practices are not uniform across the USA. Open surgery is associated with a longer LOS; however, total charges are similar between the two. Inhospital mortality rates have remained stable. Based on the updated clinical practice patterns presented in this paper, and the recent FDA approval of additional new aneurysm treatment devices, such as the WEB, Surpass and FRED, the standard of care for treatment of CA will likely continue to evolve in parallel with further technological innovations in EVT.
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @dralexandercoon, @drlimeilin
Correction notice Since the online publication of this article, it was noticed that extra zeros were erroneously added to data during the production process. In the Results section of the Abstract, EVT n=11200.4 has been changed to n=1120.4; n=10540.4 to n=1054.4; n=14390.0 to n=1439.0 and n=10510.6 to n=1051.6. Additionally in the Results section of the article, n=12790.7 has been changed to n=1279.7 and n=10530.0 to n=1053.0.
Contributors ASW collected and analyzed the data, drafted the original manuscript, and revised the paper. JKC critically reviewed the paper. GPC and ALC revised the paper. L-ML conceptualized and supervised the study, and revised the paper.
Funding This study was partially funded by Western University Summer Research Fellowship Award to ASW.
Competing interests GPC is a consultant/proctor for Stryker Neurovascular, MicroVention and Medtronic. ALC is a consultant/proctor for Stryker Neurovascular, Medtronic, and MicroVention, and share holder for InNeuroCo. L-ML is a consultant/proctor for Stryker Neurovascular and a consultant for MicroVention and Cerenovus.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.