|Home | About | Journals | Submit | Contact Us | Français|
With the expansion of elective abdominal aortic aneurysm (AAA) repair after the introduction of endovascular aneurysm repair (EVAR), there is a concern that even with a lower operative mortality there could be increasing number of aneurysm related deaths. To evaluate this, we looked at national trends in AAA repair volume as well as mortality after intact and ruptured AAA repair encompassing the introduction of EVAR.
Patients with intact or ruptured AAA undergoing open repair or EVAR and all those with a diagnosis of ruptured AAA were identified within the 1993–2005 Nationwide Inpatient Sample database using ICD-9 diagnosis and procedure codes. The number of repairs, number of rupture diagnoses without repair, number of deaths, and associated mortality rates were measured for each year of the database. Outcomes (mean annual volumes) were compared from the pre-EVAR era (1993–1998) to the post-EVAR era (2001–2005).
Since introduction, EVAR increased steadily and accounted for 56% of repairs, yet only 27% of the deaths for intact repairs in 2005. The mean annual number of intact repairs increased from 36,122 in the pre-EVAR era to 38,901 in the post-EVAR era while the mean annual number of deaths related to intact AAA repair decreased from 1,693 pre-EVAR to 1,207 post-EVAR (P < .0001). Mortality for all intact AAA repair decreased from 4.0% to 3.1% (P < .0001) pre and post-EVAR but open repair mortality was unchanged (open repair 4.7% to 4.5%, P= .31; EVAR 1.3%). During the same time periods, mean annual number of ruptured repairs decreased from 2,804 to 1,846 and deaths from ruptured AAA repairs decreased from 2,804 to 1,846 (P < .0001). Mortality for ruptured AAA repair decreased from 44.3% to 39.9% (P < .0001) pre and post-EVAR (open repair 44.3% to 39.9%, P< .001; EVAR 32.4%). The overall mean annual number of ruptured AAA diagnoses (9,979 to 7,773, P < .0001) and overall mean annual deaths from a ruptured AAA decreased post-EVAR (5,338 to 3,901, P < .0001).
Since the introduction of EVAR, there has been a significant decrease in the annual number of deaths from both intact and ruptured AAA. This coincided with an increase in intact AAA repair after the introduction of EVAR and a decrease in ruptured AAA diagnosis and repair volume.
The introduction of endovascular aortic aneurysm repair (EVAR) has resulted in a significant change in the treatment of infra-renal aortic aneurysms. The first EVAR was reported by Juan Parodi in Argentina in 1990.1 The widespread use of this technique did not come about until the beginning of the next decade after clinical device trials were completed and FDA approval was gained in 1999. Subsequently, a procedure code for ICD-9-CM was developed in 2000 (International Classification of Diseases 9th Revision, Clinical Modification).
Previous studies have demonstrated that EVAR reduces the rate of complications and mortality compared to open repair in spite of the fact that EVAR patients as a group are older with more comorbidities.2–4 In randomized controlled studies, perioperative mortality was 4.6% after open repair and 1.2%–1.6% after EVAR.3,4 In most series, EVAR accounts for 40–60% of all elective AAA repair.2,5–9 In the Medicare population, EVAR volume surpassed open repair volume in 2003 for the first time and continues to increase for elective AAA repair.2 More recently, EVAR has also been increasingly used for ruptured AAA repair and in small series appears to result in better perioperative survival.10–13
The purpose of this study is to evaluate the impact of EVAR on the annual volume of aneurysm repair, and the overall annual number of aneurysm related deaths (in both elective and rupture repair), as well as its impact on rupture occurrence.
The Nationwide Inpatient Sample (NIS) from the years 1993 to 2005, the most recent year available, was used for this study. The NIS is maintained by the Healthcare Cost and Utilization Project (HCUP) of the Agency for Healthcare Research and Quality. The database is a 20% all-payer sample of hospital stays and contains sampling weights to allow for stratified calculation of total population estimates. The database was initialized in 1988 and has been modified over the years to adjust for changes in the population. Updated sampling weights reflecting these alterations to allow for comparison across years have been made available by HCUP and were used for the current study. The initial years of the database from 1988 to 1992 constituted the first release of the database and were excluded from the current study as significant alterations were made to the database between 1992 and 1993.14
The database was queried with SAS (version 9.1, SAS Institute, Cary, NC) using ICD-9 diagnosis codes for intact AAA (441.4, 441.9) and ruptured AAA (441.3, 441.5). Patients with a procedure code for open repair (38.25, 39.44) and EVAR (39.71) were included. Patients with both procedures performed within the hospitalization were recorded as EVAR patients and were assumed to be either pure EVAR cases with coding errors or conversions from EVAR to open repair. To allow for trend analysis and calculation of total volumes for the entire time period, we included those endovascular repairs done prior to the implementation of the ICD-9-CM code. For this, we queried the database for the ICD-9-CM code of 39.90 (insertion non-drug eluting, non-coronary stent) if coupled with a primary diagnosis of one of the above aneurysm diagnosis codes. Patients less than 18 years were excluded. Totals were calculated per year of hospitalization. Intact AAA repair (open and EVAR), ruptured AAA repair (open and EVAR), and all ruptured AAA diagnoses (repaired and unrepaired) were recorded. The primary outcomes were in-hospital death and the respective mortality rates for each group. Age, gender, race (white vs other), length of stay, discharge disposition (home vs rehab or other facility), and hospital charges were also measured.
Statistical analysis was performed using survey analysis programs with Stata (Stata Statistical Software: Release 8. College Station, TX: StataCorp LP). Total population estimates for each subgroup are reported by applying the sampling weight for each observation within Stata calculations. Means and standard deviations are reported for parametric data and median values and ranges for non-parametric data. Statistical significance was assigned as a p-value of < .05. Comparisons between cohorts were carried out using the Wilcoxon-ranked sum for nonparametric continuous data, Student’s t-test for parametric continuous data, and chi-square for categorical and count data. Groups were stratified by repair method as well as by intact versus ruptured status. Annual number of AAA repairs and AAA related deaths from the time period representative of the pre-endovascular era (1993 to 1998) were compared to the time period during which endovascular repair coding had been fully incorporated (2001–2005). The years 1999–2000 were omitted from this comparison as FDA approval was first attained in 1999 and the ICD-9-CM code was introduced in October of 2000. Mean annual population totals rather than cumulative totals are reported for this subanalysis as the year ranges are unequal. As an additional test, the slope of the linear regression line from the pre-EVAR time period was compared to the post-EVAR time period by analysis of the interaction effect.
Study approval was obtained from the Institutional Review Board at Beth Israel Deaconess Medical Center. Data use agreements for use of the NIS data were made with HCUP.
Over half a million aneurysms (555,577 intact and ruptured) were repaired over the time from 1993 to 2005 with 50,261 deaths attributable to aneurysm repair. Overall repairs (41,831 in 1993 to 41,185 in 2005) stayed relatively stable due to an increase in elective repair that was offset by a decrease in ruptured aneurysm repair (Figure 1). Over this time, annual AAA repair related deaths decreased by 42% (4,477 to 2,618) (Figure 2). In the post-EVAR era (years 2001–2005), there were fewer mean annual repair related deaths (3,063) than in the pre-EVAR era (years 1993–1998) (4,496) (P < .0001).
Over the time period from 1993 to 2005 there were a total of 482,625 intact AAA repairs performed with an average of 37,125 repairs performed per year (Table I). Of the total repairs, 19,131 patients died (4.0% mortality). The number of intact AAA repair-related deaths per year decreased by 43% from 1993 to 2005 (1,775 to 1,013, P < .0001) (Figure 2).
The average annual number of repairs per year rose after the establishment of EVAR. In the pre-EVAR era, the mean annual number of repairs performed for intact AAA was lower than in the post-EVAR era ( 36,122 vs 38,901, P< .0001) (Table II). Patients were older in the post-EVAR era. There was a greater proportion of patients older than 80 years in the post-EVAR cohort as well as a decreased proportion of white patients. Males made up the majority of the repair population and there was no difference in this percentage pre or post-EVAR.
The mean annual number of deaths associated with intact AAA repair was lower in the post-EVAR era despite the increase in the number of repairs (Table II). Mortality was significantly decreased (4.7% to 3.1%, P < .0001). The mortality associated with open repair was equivalent before and after EVAR introduction (4.7% versus 4.5%, P= .31). Median length of stay was decreased by 2 days in the later time period for all repairs. Open repairs had a smaller decrease in length of stay comparing pre and post-EVAR eras (8 days vs 7 days, P < .0001). Discharge to home was more likely in the pre-EVAR era (89.3% vs post-EVAR 87.6%, P< .0001) with a more prominent decrease in home discharges after open repair (89.3% vs 82.8%, P < .0001).
The number of endovascular repairs performed eclipsed the number of open repairs by the year 2004 (Figure 3). In 2005, EVAR accounted for 56% of intact AAA repairs but only 27% of the deaths.
In 2001–2005, patients undergoing EVAR were on average 2 years older than those undergoing open repair (Table I). Mortality was 1.3% for EVAR and 4.5% for open repair. In the year 2005, overall mortality was 2.7%, the lowest among all years analyzed. Median length of stay was shorter after EVAR and more patients were discharged to home. Median hospital charges were higher with EVAR.
The total number of admissions for a diagnosis of ruptured AAA decreased by 30% from 9,807 to 6,921 per year from 1993 to 2005, with the greatest rate of decline after the introduction of EVAR (P < .0001) (Figure 4). The mean annual number of ruptured AAA diagnoses pre-EVAR was 9,979 versus 7,773 post-EVAR (p<.0001).
The percentage of patients who had repair of their ruptured aneurysm was 63% overall. Patients not undergoing repair of a ruptured AAA were older and more likely to be female compared to those undergoing repair (Table III).
The mean annual number of deaths associated with a diagnosis of ruptured AAA was 5,338 pre-EVAR and 3,891 post-EVAR (P < .0001).
Over the time period from 1993 to 2005 there were a total of 72,952 ruptured AAA repairs performed with an average of 5,612 repairs performed per year (Table IV). The number of repairs for ruptured AAA decreased by 35% from 6,091 in 1993 to 3,966 in 2005, again with a more significant rate of decline post-EVAR (P < .0001) (Figure 4).
The mean annual number of ruptured AAA repairs decreased from 6,335 pre-EVAR to 4,667 post-EVAR (p<.0001) (Table V). Mean age was similar between time periods, however a greater proportion of octogenarians underwent repair in the post-EVAR era.
Of the total repairs for ruptured aneurysms, 43% of patients died. The number of ruptured AAA repair-related deaths per year decreased over time from 2,702 in 1993 to 1,605 in 2005 (P < .0001) (Figure 4). The mean annual number of deaths associated with ruptured AAA repair was significantly lower post-EVAR (1,846 vs 2,804, P<.0001) along with a decrease in mortality (39.9% vs 44.3%, P< .0001). Open repair mortality was lower in the post-EVAR era (40.8% vs 44.3%, P< .001). Length of stay was unchanged however discharge to home was more likely in the pre–EVAR time period.
In the latest year available (2005), EVAR was performed in 17% of ruptured AAA repairs. Mortality with EVAR decreased from 42.9% in 2001 to 30.3% in 2005 (P< .0001).
During the post-EVAR era (2001–2005), mortality was 32.3% after EVAR for ruptured aneurysms and 40.8% after open repair (Table IV). Length of stay was shorter after EVAR and patients were more likely be discharged home. Total hospital charges were similar.
Patients admitted with a diagnosis of ruptured AAA who did not undergo repair made up 37% (3,645) of all ruptured AAA diagnosis pre-EVAR and 40% (3,105) of those post-EVAR. Mortality (without repair) pre-EVAR and post-EVAR were 69.6% and 65.9% respectively (P < .001).
The overall number of AAA related deaths (intact repair, ruptured repair, ruptured unrepaired) from 1993 to 2005 was 79,955. From 1993 to 2005, the number of annual deaths decreased by 38%, with the mean annual number of deaths post-EVAR decreasing to 5,108 from 7,031 (P < .0001). In addition, there was a continued downward trend with 4,498 deaths in 2005.
Abdominal aortic aneurysm is a disease of the elderly. With the aging of the US population, it would be expected that the volume of AAA repair as well as AAA related mortality would increase. Our current study shows that after the introduction of EVAR there has been an increase in elective repairs and a decrease in ruptured AAA as well as a decrease in procedure related mortality for both intact and ruptured AAA. This has led to a decrease in overall AAA related deaths despite an unchanged mortality for elective open repair.
Heller et al. examined trends in AAA related mortality from the National Hospital Discharge Survey and found no improvement in the number of aneurysm repair deaths from 1979–1997. They also found an unchanged incidence of ruptured aneurysms and ruptured aneurysm repair.15 Cowan et al. found stable rates of repair using NIS data up to 2003.16 This background, along with the current finding that open repair mortality remains the same, suggests that the decrease in annual AAA-related deaths of 38% seen in this study is not a continuing effect of medical advancement in general, but a result of the new technique of repair.
The increased age of those repaired in the post-EVAR era (as well as EVAR vs open) suggests the expansion to an older and perhaps sicker patient population that may not have been considered for open surgery but were still at risk for rupture. The shift of repair to EVAR has driven an increase in intact aneurysm repair volume and a subsequent reduction in the number of ruptures overall. In this NIS population, EVAR volume was slightly greater than open repair volume in 2004 and accounted for 56% of repairs by 2005. In the Medicare population, EVAR volume overtook open repair volume by 2003.2 This difference is likely due to the age difference of the datasets given that EVAR patients (and Medicare patients) are typically older. Although EVAR now is more common than open repair, with its low elective mortality, its contribution to elective deaths was only 27% in 2005.
We previously have reported that, even when patient populations are matched closely to control for confounders, EVAR has a lower in-hospital mortality rate in the US Medicare population than open repair (1.2% vs 4.8%). This difference is still significant, 1.3% vs 4.6%, when comparing the entire (unadjusted) Medicare population even though EVAR patients are older with more comorbidities.2 The outcome of this lower mortality as we see here is an overall decrease in population deaths as EVAR becomes the favored repair.
As with intact AAA repair, we see a lower mortality with EVAR for ruptured aneurysms compared to open repair (32% vs 43%). Less can be concluded from this finding in a retrospective study as there is the potential for selection bias that cannot be assessed with this administrative database. With proper utilization, however, it is believable that the method could lead to overall improved outcomes. Our finding that mortality within the US population has decreased to just under 40% for ruptured repair shows progress compared to prior studies for the past 5 decades.17 Institutional ‘EVAR first’ programs have been promising with lower mortality using EVAR for ruptured aneurysm repair.11,18,19 Mehta et al. reported the results of a hospital- wide initiative to facilitate EVAR for ruptured AAA. Their program resulted in an 18% mortality after EVAR with 47% of patients receiving an endovascular graft rather than open repair when presenting with ruptured aneurysms.11
Cowan et al. found that mortality associated with repair of ruptured aneurysms decreased from 1993 to 2003 for open repair (46.5% to 40.7%).16 Dillavou et al. reported outcomes from the same time period using a 5% inpatient sample from the Medicare population. They reported an unchanged mortality for ruptured repair overall (male average 44.2%, female average 52.8%).20 With the inclusion of more recent years of data, we show that ruptured repair mortality has decreased from an annual average of 44.3% prior to EVAR to 39.9% for all repairs and to 40.8% for open repair. The decrease in open repair mortality over this time indicates that EVAR is not entirely responsible for the decreased mortality of ruptured aneurysm repair, however there is a contribution that could be expected to increase as volume continues to rise.
Another of the promising findings of this study is that hospital admission for the diagnosis of a ruptured AAA has decreased since EVAR. In the Medicare population, Dillavou et al. found ruptured aneurysms to decline from 7300 in 1994 to 5640 in 2003 (23%).20 We report a 30% decrease within our time frame with a rate of decrease that was more rapid after the introduction of EVAR. Given that rupture repair deaths are the larger contributor to overall aneurysm repair deaths, the population benefit is substantial.
Overall mortality associated with a diagnosis of ruptured AAA without repair was lower than expected, raising questions about the accuracy of the diagnosis. This highlights a limitation of the database in that it is reliant upon accurate coding of conditions in order identify cases. Pairing concomitant procedures within the hospitalization increases the accuracy of the diagnosis. It is likely that some patients are admitted with an initial diagnosis of ruptured AAA and an alternative diagnosis is ultimately determined. In this database, the admitting diagnosis is retained as well as any subsequent final diagnoses. This should not impact rupture repair rate and mortality calculations, nor should it impact intact repair. We allowed ruptured AAA diagnosis without repair as an endpoint due to the fact that the observed trends in diagnoses and related deaths mirrored those seen in ruptured AAA with repairs (Figure 4) and there was no identifiable reason why coding accuracy would change over time.
The limitations of the current study include the data source along with its retrospective design. The NIS is designed to analyze heathcare trends and outcomes and as such, it is ideal for a study of this nature, however the database is reliant upon accurate and uniform coding and relies upon sampling weights to derive total population estimates. The weights are designed to control for sampling bias and are based upon hospital region and patient characteristics. Analysis of only actual NIS cases without the utilization of the sampling weights resulted in the same outcome. The NIS has undergone multiple changes since its introduction that include changes in state participation as well as data element inclusion. We used the published supplemental trend file weights to discount any effect these changes may have in comparisons across years.14 Additionally, administrative data do not include anatomic data such as AAA diameter or extent (infrarenal vs pararenal), so stratification along those criteria is not possible.
The inclusion of the peripheral stent code in combination with a primary diagnosis of AAA was made in order to capture some of the EVARs performed before the introduction of the specific procedure code. We believe that this still underestimates the true number of EVARs performed in the transition period however, and thus those years were excluded for the comparative analyses in order for more accurate conclusions to be reached.
There are other factors that may have an effect on aneurysm repair and ruptures in the US today including health care patterns and risk factor prevalence. Increased patient and physician awareness of AAAs as a result of screening programs may have an impact upon the number of patients presenting with rupture.21 In 2004 the Society for Vascular Surgery Consensus Statement recommended ultrasound screening for patients greater than 50 years with a family history of AAA, or for men age 60 to 85 years and women 60 to 85 years with cardiovascular risk factors.22 Screening for AAA did not become a benefit offered by Medicare until January of 2007, and then only for male smokers or patients with a family history of aneurysm at the time of their welcome to Medicare visit.
Risk factors including smoking and hypertension have been associated with the diagnosis or rupture of AAA.23,24 Smoking has decreased in the US population over the past four decades by 50% and from 1993–2005 the rates have decreased from 25% to 21%.25 This may account for some of the observed decrease in ruptures and may decrease the prevalence of AAA over time. Hypertension was shown to have an unchanged prevalence from 1999 to 2006 (28–30%). In the year 2005–2006, 68% of hypertensive adults in the US used antihypertensive medication, however only 64% of those achieved an adequate blood pressure goal.26,27 Less than half of patients entering a large multicenter trial for infrainguinal bypass were using beta-blockers or lipid-lowering therapy.28
With new technology the threshold for repair may be lowered to include older, sicker patients who were not candidates for open repair yet were still at risk for rupture. Additionally, the threshold may be lowered for smaller diameter aneurysms although these data cannot confirm any potential benefit in these subgroups.
The introduction of EVAR has led to an increase in elective AAA repair with lower mortality. There has been a coincident decrease in AAA rupture as well as a decrease in total aneurysm related deaths.
Harvard-Longwood Research in Vascular Surgery Program. NIH grant: 5 T32HL007734 (NIH-NHLBI).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
No competing interest declared