|Home | About | Journals | Submit | Contact Us | Français|
To describe short and long-term survival of patients with descending thoracic aortic aneurysms (TAA) following open and endovascular repair (TEVAR).
Using Medicare claims from 1998–2007, we analyzed patients who underwent repair of intact and ruptured TAA, identified using a combination of procedural and diagnostic ICD-9 codes. Our main outcome measure was mortality, defined as peri-operative mortality (death occurring before hospital discharge or within 30 days), and five year survival, using life-table analysis. We examined outcomes across repair type (open repair. or TEVAR) in crude, adjusted (age, sex, race, procedure year, and Charlson comorbidity score), and propensity-matched cohorts. Overall, we studied 12,573 Medicare patients who underwent open repair, and 2,732 patients who underwent TEVAR. Peri-operative mortality was lower in patients undergoing TEVAR as compared to open repair for both intact (6.1% versus 7.1%, p=0.07) and ruptured TAA (28% versus 46%, p<0.0001). However, patients with intact TAA selected for TEVAR had significantly worse survival than open patients at one year (87% open, 82% TEVAR, p=0.001) and five years (72% open, 62% TEVAR, p= 0.001). Further, in adjusted and propensity-matched cohorts, patients selected for TEVAR had worse 5-year survival than patients selected for open repair.
While peri-operative mortality is lower with TEVAR, Medicare patients selected for TEVAR have worse long-term survival than patients selected for open repair. The results of this observational study suggest that higher risk patients are being offered TEVAR, and that some do not benefit based on long-term survival. Future work is needed to identify TEVAR candidates unlikely to benefit from repair.
Given the significant risk involved with descending thoracic aortic aneurysm (TAA) repair, as well as the incidence of medical co-morbidities in the patients who present with these lesions, surgeons have struggled to balance the probability of death from aneurysm rupture with the risk of surgical intervention1–4. The introduction of thoracic endovascular repair (TEVAR) has further complicated this relationship5–7, as this less-invasive intervention has expanded the pool of patients who are physiologic candidates for surgery. Subsequently, single-center, regional, and national studies have described a significant increase of utilization of TEVAR, with a significant short-term benefit in peri-operative mortality5–10.
However, long-term survival following TAA repair in real-world clinical practice remains uncertain. While several reports have described similar mid-term survival following open repair and TEVAR, these studies have been limited to single institution series from centers of excellence, or results from industry-sponsored trials and registries 11–14, and only one clinical trial has reported five-year outcomes Further, little population-based data describing TEVAR and open repair is available that allows examination of survival differences for distinct TAA, especially across rupture status at presentation, or by repair type5.
Therefore, we sought to compare peri-operative and long-term mortality in the contemporary, real-world practice of open repair and TEVAR. We studied all patients undergoing open repair and TEVAR in recent years in the United States Medicare population, and examined the short term and long survival with these procedures.
We used the Medicare Physician/Supplier file and the Medicare Denominator file for these analyses, for the years 1998–2007. First, we created a cohort of patients who underwent a broad range of thoracic aortic procedures, as defined by the ICD-9 codes in Table 115. Next, we selected from this cohort those patients who underwent TEVAR and open repair of TAA. As outlined in Figure 1, we eliminated all claims that were not contained in the Medicare Denominator File, as well as patients not at least age 66. We required patients to have at least one year of Medicare eligibility prior to surgery, and used this time period to establish comorbidities and construct patient-specific Charlson scores16. The Charlson score contains 19 categories of comorbidity, based on ICD-9 diagnosis and procedure codes, and has been validated in a variety of settings for use in administrative datasets17, 18. Details regarding the components of the Charlson score and their assigned ICD-9 codes are outlined in Supplementary Table 1. Further, we eliminated any claim wherein the patient’s Medicare health insurance number (HIC) had changed during the study period, because this would not allow survival analysis.
In addition to the procedural codes for TAA, we required that each patient have a diagnosis code for TAA. Initially, we studied five distinct diagnoses, as indicated by their respective ICD-9 diagnosis codes (Table 1). We excluded any patient with diagnosis codes for ascending thoracic aortic aneurysm, or with procedural codes for cardiopulmonary bypass occurring with circulatory arrest. Further, we excluded patients with thoraco-abdominal aneurysms, thoracic aortic dissections, and “other” aortic pathology from our analysis, as these entities are clinically distinct from TAA. Lastly, we also excluded patients with ICD9 procedural codes that may indicate the presence of “debranching” or other procedures to extend endovascular landing zones, such as 39.24 (aorto-renal bypass), 39.25 (aorta-iliac-femoral bypass). In situations wherein we encountered more than one procedure (open repair or TEVAR) per patient, we assigned the patient to the first procedure performed.
Lastly, we examined the effect of changes in practice pattern over time in two ways. First, in prior work10, we demonstrated that between 1998 and 2003, a significant increase in the use of TEVAR occurred. Before 2003, fewer than 10% of all intact TAA were repaired using TEVAR. After 2003, more than 10% of all intact TAAs were repaired with TEVAR, and this rate grew to 27% by 2007. Therefore, to avoid any bias introduced by the era (before or after introduction of TEVAR), we generated a binary variable (“era of procedure”) representing the time period during this learning curve (prior to January 1st, 2004), and following this learning curve (after January 1st, 2004).
Second, the Food and Drug Administration specifically approved a device for use in TEVAR in 200519. Therefore, to examine any effect secondary to changes in patient selection following FDA approval, we examined results before and after January 1st, 2005, for both open and endovascular repair.
Once we established a cohort of patients with diagnosis codes and procedural codes for TAA repair, we followed them over time to establish our two main outcome measures: peri-operative and long-term survival, using one and five-year survival rates. First, to define peri-operative mortality, we sought to capture all deaths occurring in the period following surgery. We defined peri-operative mortality as death occurring within the index hospitalization (regardless of post-operative day), as well as any death within thirty days (irrespective of inpatient or outpatient status). The outcome of peri-operative death was a binary categorical variable, and was analyzed using chi-squared tests.
Second, survival at one and five years was established using the Medicare denominator file to establish the date of death. We censored those patients who survived until the end of our analysis (at December 31st, 2007). Survival curves were estimated using Kaplan-Meier analysis, and life table analysis was used to establish rates of five year survival with surrounding 95% confidence intervals. Log-rank tests were used to determine significant differences in survival between groups, and p values <0.05 were considered significant.
To account for differences in patient characteristics between the open repair and TEVAR cohorts, we performed two additional analyses. First, we used a survivor function wherein age, gender, race, era of procedure, and Charlson score16 was adjusted using a Cox proportional hazards model to estimate survival. This allowed comparison of survival estimates, adjusted to reflect survival within the strata of minimal patient-level risk within each group20. The lowest-risk group from these analyses, stratified by rupture status and repair type, was used to demonstrate these adjusted results using Kaplan Meier plots21, 22. In patients undergoing repair for rupture, the survival curves by repair type crossed one another, and therefore the assumption of proportional hazards was not satisfied, precluding calculation of a adjusted hazard ratio for endovascular repair in ruptured patients. Accordingly, for ruptured patients, we stratified our Cox models by repair type, and calculated hazard ratios for the remaining covariates, as reported in Supplemental Table 2.
Second, we used propensity matching methods to create similar cohorts for survival analysis 23. First, we generated a propensity score for the likelihood of undergoing TEVAR, based on a multivariable logistic model that described the association between preoperative patient characteristics and the choice to perform TEVAR. Because the range of propensity score was broad, we utilized a stratified propensity analysis24, 25. In other words, adequate propensity matches were limited to the lowest-risk quartile of propensity score, because higher risk patients undergoing TEVAR were not able to be adequately “matched” within the open surgical cohort. Within a sample of patients between age 65 and 75 who underwent surgery during the latter 4 years of our study period, we propensity matched patients by age and comorbidity. This allowed us to generate two cohorts that were matched in terms of age, gender, race, era of procedure, comorbidities that constitute the Charlson score, as well as the Charlson score itself. (Table 2). Chi-squared tests and t-tests were used to ensure that there were no significant differences in preoperative characteristics between the open repair and TEVAR cohorts. We randomly selected 550 patients each from these two matched cohorts, and we compared four-year survival between these groups using Kaplan Meier analysis. Given the small sample size in the TEVAR rupture cohort, only intact TAA was studied in this portion of the analysis.
Further details of the models used in the adjusted survival analysis and propensity matched analysis are available in Supplementary Table 2. All analyses were performed using SAS (SAS Institute, Cary, NC), and STATA 10 (STATA, College Station, TX).
Overall, we studied 12,573 Medicare patients who underwent open procedures, and 2,732 patients who underwent TEVAR (Figure 1). By presentation status, 13,998 patients presented for surgery with intact TAA (11,565 open repair, 2,433 TEVAR), while 1,307 patients underwent surgery for ruptured TAA (1,008 open repair, 299 TEVAR). Patient characteristics in the cohort are shown in Table 2.
Several demographic differences existed between open repair and TEVAR patients with intact TAA. First, patients undergoing TEVAR were significantly older than patients undergoing open repair (75.9 years versus 73.8 years, p<0.0001). Second, the proportion of male patients was slightly higher among patients undergoing TEVAR than open repair (58.7 versus 55.4%, p<0.0001). Other differences were also apparent across the open repair and TEVAR cohorts, including higher rates of diabetes, myocardial infarction, COPD, chronic renal failure, as well as a higher proportion of black patients (7.5% TEVAR, 3.8% open, p=0.001).
Differences also existed between patients presenting with intact and ruptured TAA. First, patients with ruptured TAA were older, on average, than patients with intact TAA, by over two years (76.4 years versus 74.1years, p=0.0001). Second, the proportion of black patients was significantly higher in the ruptured cohort as compared to the intact cohort (7.3% versus 4.4%, p=0.0001). Third, the incidence of comorbidities such as diabetes, cerebrovascular disease, COPD, and renal failure were slightly higher in patients undergoing repair for rupture as compared to patients presenting for elective repair.
Lastly, within the group of patients undergoing repair for ruptured TAA, patients undergoing TEVAR were significantly older than patients undergoing open repair (77.9 years versus 76.4 years, p=0.001). Further, patients selected for TEVAR for ruptured TAA had a higher incidence of diabetes, COPD, black race, renal failure, and malignancy than patients selected for open repair.
As shown in Figure 2, the lowest peri-operative mortality rate occurred in patients undergoing repair of intact TAA using TEVAR (6.1% (95% CI 5.1–7.0%)). While the peri-operative mortality rate for open repair was slightly higher (7.1% (95% CI 6.7–7.6%)), the clinical magnitude of this difference was small, and borderline in terms of statistical significance (p=0.07). Among patients presenting with ruptured thoracic aneurysms, peri-operative mortality was 28.4% (95% CI 23.2–33.5%) for TEVAR, and 45.6% (95% CI 42.5–48.7%) for open repair (p=0.0001).
Unadjusted long-term survival varied by presentation (intact versus ruptured) as well as repair type (open repair versus TEVAR) (Figure 3a). Even though patients with intact TAA selected for TEVAR had lower peri-operative mortality, patients selected for open repair reclaimed the survival advantage within the first year after surgery (1 year survival by life table analysis: 87% open repair (95% CI 86–88%), 82% TEVAR (95% CI 80–83%), log rank p=0.001). This survival advantage continued to accumulate over time, as seen in our five-year survival data (5-year survival by life table analysis: 72% (95% CI 71–73%) open repair, 62% TEVAR (95% CI 60–65%, log rank p=0.001). As seen in Figure 3a, the steeper slope in the survival curve for TEVAR patients undergoing repair of intact TAA demonstrates the poorer long-term survival in patients selected for TEVAR as compared to those selected for open repair.
Similarly, patients with ruptured TAA repaired with TEVAR had significantly better short-term survival. However, similar to intact repair, this survival advantage disappeared by 1.5 years post-operatively. Thereafter, survival rates remained similar between repair type. After five years, by life table analysis, fewer than 30% of patients were alive after repair of their ruptured TAA, irrespective of the type of repair (26% (95% CI 23–30%) open repair, 23% (95%CI 16–32%) TEVAR, log rank p=0.37).
In survival analyses adjusted for age, sex, race, era of procedure, and Charlson comorbidity score, our results were similar to our unadjusted analyses. In patients with intact TAA, the peri-operative survival advantage incurred by TEVAR disappeared within the first year, and five year survival was significantly worse in patients undergoing TEVAR (89% versus 79%, log rank p<0.0001) (Figure 3b). In patients with ruptured TAA, the survival advantage incurred by TEVAR disappeared within the first 90 days following surgery, and five year survival was significantly worse in patients undergoing TEVAR compared to open repair (62% versus 45%, log rank p=0.0001).
In this analysis, we compared patients who were propensity-matched (within the lowest-risk quartile of propensity score) by age, sex, race, year of repair, and comorbidity score, thereby generating two distinct cohorts of patients. These cohorts of patients were similar in terms of patient characteristics available for measurement in administrative claims, as shown in Table 2, and underwent either open repair or TEVAR. Within this propensity-matched open repair and TEVAR cohort, there were no significant statistical differences in patient characteristics.
In the propensity-matched cohorts, peri-operative mortality was statistically similar in patients undergoing TEVAR compared to open repair (4.5% (95% CI 2.8–6.2%) open repair versus 4.2% (95% CI 2.5–5.8) TEVAR, p=0.78). Any difference in peri-operative mortality incurred by TEVAR disappeared within the first year post-operatively, and late survival was significantly worse in patients undergoing TEVAR (Figure 3c). By life table analysis at five years’ post-operatively, late survival was significantly worse in patients undergoing TEVAR (73% (95% CI 68–76%)) than open repair (81% (95% CI 77–85%)) within the propensity-matched cohort (log rank p=0.007).
Finally, to examine any effect of change in practice pattern following the FDA approval of commercial devices specifically designed for TEVAR, we examined in-hospital/30 day mortality as well as 3-year survival before and after FDA approval of a commercial device specifically designed for TEVAR in 2005. Before FDA approval in 2005, 631 patients underwent TEVAR of intact TAA, and 30-day/in hospital mortality was 7.1%. Following FDA approval, the number of patients in our cohort undergoing TEVAR was much larger (n=1,802), and represented a larger proportion of all patients undergoing TAA repair (TEVAR = 8.5% of all TAA prior to FDA approval, TEVAR = 27.2% of all TAA following FDA approval, p<0.0001). Patient characteristics of those undergoing TEVAR were largely similar before and after FDA approval (Table 3), except patients undergoing TEVAR following FDA approval were slightly older. While peri-operative mortality following TEVAR was lower following FDA approval (5.8%), this difference was not significantly lower than peri-operative mortality before FDA approval. Similarly, 3-year survival was similar following TEVAR before and after FDA approval (Table 4).
However, for patients undergoing open repair, patient characteristics, peri-operative mortality, as well as 3-year survival were all significantly different before and after FDA approval. Following FDA approval, patients selected for open repair were less likely to have a history of heart disease, cerebrovascular disease, or COPD. Patients selected for open repair following FDA approval also had a lower average Charlson comorbidity score. Finally, peri-operative mortality decreased from 8.4% to 5.4% (p=0.0001) following FDA approval, and 3 year survival was higher in patients undergoing open repair following FDA approval (Table 4).
While the competing mortalities of observation and open surgical repair for TAA have been investigated for several decades, the effect of TEVAR on survival remains less well studied. Most reports regarding survival after TEVAR are single center series, or the results of highly selected industry-sponsored registries4, 6, 26, 27–29. Our study examined survival following open repair and TEVAR in national, real-world practice, and found that while peri-operative mortality is lower in TEVAR, patients selected for TEVAR have worse long-term survival than patients selected for open repair. These results suggest that higher risk patients are being offered TEVAR, and that some do not benefit based on long-term survival.
Understanding the effect of a less invasive endovascular option for aneurysm repair on patient selection and survival following aortic aneurysm surgery has been a topic of extensive study, but primarily in patients with infrarenal abdominal aortic aneurysm (AAA)3, 30. Several randomized trials31–33 have demonstrated lower peri-operative mortality with endovascular techniques. However, in these trials, as well as large observational analyses30, the survival advantage gained by an endovascular approach consistently disappears within two years after surgery, and little difference in long-term survival is evident thereafter across procedures. Patients who experience late death following infrarenal AAA repair most commonly die from cardiopulmonary comorbidities unrelated to their aneurysm, and relatively few experience aneurysm-related death, in either open or endovascular repair34. In other words, in both randomized trials and in real-world practice, while EVAR is as effective in preventing aneurysm-related death as open repair, it does not result in prolonged improvement or detriment in survival. Rather, it has decreased peri-operative morbidity and mortality, and expanded the pool of patients undergoing elective repair35.
Our results demonstrate that the treatment of TAA follows a similar course to infrarenal AAA, with one important exception. As with infrarenal AAA, we found a survival advantage in short-term mortality for patients who undergo TEVAR as compared to open repair, especially in patients presenting with ruptured TAA, and this finding has been reported in similar analyses in other national datasets36. Further, as with infrarenal AAA, any survival advantage gained in the peri-operative period following endovascular repair was lost within two years after surgery. However, unlike infrarenal AAA, wherein long-term survival is similar across procedure type, adjusted survival at five years was significantly worse for patients selected for TEVAR as compared to open repair. Therefore, the widespread application of TEVAR has resulted in a cohort of patients who may have previously not undergone surgery, but now undergo TEVAR. Patients selected for TEVAR achieve worse survival than patients undergoing open repair, and many of these deaths occur within the first two years after TEVAR. These deaths could be due to the selection of “sicker” patients for TEVAR, although our finding of poorer survival following TEVAR persists, even in propensity-matched analyses that account for differences in patient risk measurable using administrative claims. Alternatively, these differences in survival could be explained by device-related complications occurring within the first five years following surgery.
As with patients with infrarenal AAA, we suspect that the loss of survival advantage is secondary to patient-level comorbidities. For example, in the EVAR-2 trial37, survival was similar among patients treated with endovascular repair and patients who did not undergo repair. All patients in EVAR-2 were deemed “unacceptable for open repair”, as they often had comorbidities that limited their survival, but not their ability to undergo endovascular repair. The data available from our study supports this presumption, as TEVAR patients tend to be older and have higher comorbidity scores than the patients selected for open repair. However, it is important to acknowledge that our findings are based solely on administrative claims, and our analysis is therefore limited in terms of clinical detail.
Given the lack of anatomic and procedural detail in our dataset, it remains uncertain if the lack of survival advantage seen in patients selected for TEVAR could be related to late-occurring, device-related complications. To best determine whether or not late device-related complications are contributing to the poorer “real-world” survival in patients undergoing TEVAR, post-implantation follow-up using device-specific registries will be necessary. Efforts in this regard have already been discussed and implemented by specialty societies interested in outcomes of endovascular procedures, and these registries will also provide more robust clinical detail for risk adjustment as compared to the administrative data used in our current work38.
Finally, following FDA approval and widespread implementation of TEVAR, the peri-operative mortality associated with open repair declined significantly, and a small but significant survival benefit was evident 3 years following surgery. While indirect, this evidence suggests that after FDA approval, higher risk patients were being offered TEVAR rather than open repair, and patients selected for open repair were lower risk than those selected for TEVAR. However, more definitive characterization of these changes is necessary, and will require registry-based data with more detailed covariates for risk adjustment than currently available from administrative data.
Our findings add important context to the studies used to demonstrate the efficacy of endovascular repair of TAA12, 13, 39–41. First, when compared to data from TEVAR clinical trials (Table 5), it is evident that peri-operative mortality is higher in “real-world” practice than in the centers of excellence where the clinical trials were performed. Second, when examining the relative effectiveness of TEVAR in comparison to open surgical repair at five years, we see that it is important where the “bar” is set, in terms of open surgical repair. In real-world practice, patients selected for open repair had better survival the surgical controls used in clinical trials (72% five-year survival in Medicare, 67% five-year survival in the single trial that reported this measure). Further, patients selected for TEVAR in real-world practice performed slightly worse than patients studied in the clinical trial (62% five-year survival in Medicare versus 68% five-year survival in the clinical trial). Collectively, these two differences resulted in the disparity in conclusions between our study, wherein TEVAR patients fared significantly worse at five years, and the clinical trial, wherein outcomes were similar at five years. Which rate is right? Certainly, future trials and analyses will address this question. While a formal meta-analysis comparing these results is beyond the scope of this manuscript, a recent meta-analysis addressed this question, and included the clinical trials shown in Table 5, as well as several single-center studies reporting 2 and 3-year outcomes. These investigators found little long-term survival benefit for patients undergoing TEVAR compared to open surgical repair 42. Therefore, given the findings in our study and others, there is little evidence to suggest that long-term survival is better in patients selected for TEVAR as compared to open surgical repair. Further, dependent upon the surgical controls selected for comparison, survival after TEVAR may be worse. 13, 39–41
Our study has several limitations. First, as mentioned previously, the limitations of administrative data in describing clinical details of aortic aneurysm repair and providing clinical-level covariates for risk-adjustment have been well described43. Therefore, we sought to be highly specific in our attempt to create a cohort of descending thoracic aneurysms, and searched for both procedural and diagnostic codes commonly used for branched, fenestrated, ascending, and transverse arch thoracic aortic repair, and eliminated these patients from our cohort of isolated descending TAA to ensure uniformity in our cohort. Second, our comparison of patients undergoing open repair and TEVAR procedures is risk-adjusted using demographic data, ICD-9 derived diagnoses, and a propensity-matched model studying low-risk patients undergoing TEVAR and open surgical repair. Further attempts to develop comparative populations, either via instrumental variable analysis44 or other statistical techniques designed to account for selection bias in observational studies, would be limited by the absence of important clinical variables such as aneurysm extent, size, and prior aneurysm surgery. Third, FDA approval was granted in 2005, and our dataset extends to 2007, limiting our insight into patients who underwent surgery in the post-FDA approval era. However, because patients with TAA experience limited survival regardless of procedure type, the overall size of our cohort combined with the frequency of these events allowed insight into significant survival differences even in this limited time period. And lastly, our dataset does not allow us to discern causes of death34. It is possible that differences in aneurysm-related deaths might better inform decision-making about which patients should undergo open repair or TEVAR, but these data would be unlikely to alter the primary conclusion of our study.
In summary, our study demonstrates that short and long-term survival following surgery for TAA varies by presentation and repair type. The consistent short-term survival advantage offered by TEVAR disappears within the first two years after surgery, and patients currently selected for TEVAR have worse long-term survival than patients selected for open repair. These results suggest that higher risk patients are being offered TEVAR, and that some do not benefit based on long-term survival. Future work is needed to identify TEVAR candidates unlikely to benefit from repair.
Using Medicare claims from 1998–2007, we examined short and long-term survival of patients with descending thoracic aortic aneurysms (TAA) following open surgical repair and endovascular repair (TEVAR). Overall, we studied 12,573 patients who underwent open repair, and 2,732 patients who underwent TEVAR. Peri-operative mortality was lower in patients undergoing TEVAR as compared to open repair for both intact (6.1% versus 7.1%, p=0.07) and ruptured TAA (28% versus 46%, p=0.0001). However, patients with intact TAA selected for TEVAR had significantly worse survival than open patients at one year (87% open, 82% TEVAR, p=0.001) and five years (72% open, 62% TEVAR, p= 0.001), and analyses adjusting for patient-level comorbidities produced similar results. Therefore, while peri-operative mortality is lower with TEVAR, patients selected for TEVAR have worse long-term survival than patients selected for open repair in our observational analysis of Medicare patients. Future work is needed to determine if these deaths are due to the selection of high-risk patients for TEVAR, or due to late device-related complications from the TEVAR itself.
P.P.G and D.H.S received a grant from the Hitchcock Foundation at Dartmouth Medical School to fund the research described herein. P.P.G. received a Career Development Award from the National Heart, Lung, and Blood Institute (1K08HL05676-01).
Presented at the Society For Vascular Surgery’s Vascular Annual Meeting, June 17th, 2011, Chicago IL.
Conflict of Interest Disclosures:
M.F.F received research and grant support as well as consultative fees from W.L. Gore, Cook Medical, and Medtronic, although none of this funding supported the research described herein. No other conflicts of interest are disclosed.