Author + information
- Received August 6, 2018
- Revision received October 22, 2018
- Accepted October 30, 2018
- Published online February 11, 2019.
- Peter Chiu, MDa,b,
- Andrew B. Goldstone, MD, PhDa,b,
- Justin M. Schaffer, MDc,
- Bharathi Lingala, PhDa,
- D. Craig Miller, MDa,
- R. Scott Mitchell, MDa,
- Y. Joseph Woo, MDa,
- Michael P. Fischbein, MD, PhDa and
- Michael D. Dake, MDa,∗ (, )@StanfordMed
- aDepartment of Cardiothoracic Surgery, Stanford University, School of Medicine, Stanford, California
- bDepartment of Health and Research Policy, Stanford University, School of Medicine, Stanford, California
- cThe Heart Hospital Baylor Plano, Plano, Texas
- ↵∗Address for correspondence:
Dr. Michael D. Dake, Stanford University Medical School, Falk CV Research Center, 300 Pasteur Drive, Stanford, California 94305.
Background For the management of descending thoracic aortic aneurysms, recent evidence has suggested that outcomes of open surgical repair may surpass thoracic endovascular aortic repair (TEVAR) in as early as 2 years.
Objectives The purpose of this study was to evaluate the comparative effectiveness of TEVAR and open surgical repair in the treatment of intact descending thoracic aortic aneurysms.
Methods Using the Medicare database, a retrospective study using regression discontinuity design and propensity score matching was performed on patients with intact descending thoracic aortic aneurysms who underwent TEVAR or open surgical repair between 1999 and 2010 with follow-up through 2014. Survival was assessed with restricted mean survival time. Perioperative mortality was assessed with logistic regression. Reintervention was evaluated as a secondary outcome.
Results Matching created comparable groups with 1,235 open surgical repair patients matched to 2,470 TEVAR patients. The odds of perioperative mortality were greater for open surgical repair: high-volume center, odds ratio (OR): 1.97 (95% confidence interval [CI]: 1.53 to 2.61); low-volume center, OR: 3.62 (95% CI: 2.88 to 4.51). The restricted mean survival time difference favored TEVAR at 9 years, −209.2 days (95% CI: −298.7 to −119.7 days; p < 0.001) for open surgical repair. Risk of reintervention was lower for open surgical repair, hazard ratio: 0.40 (95% CI: 0.34 to 0.60; p < 0.001).
Conclusions Open surgical repair was associated with increased odds of early postoperative mortality but reduced late hazard of death. Despite the late advantage of open repair, mean survival was superior for TEVAR. TEVAR should be considered the first line for repair of intact descending thoracic aortic aneurysms in Medicare beneficiaries.
In 1994, thoracic endovascular aortic repair (TEVAR) was introduced as an alternative to open surgical repair for treatment of descending thoracic aortic aneurysms (1). Following U.S. Food and Drug Administration (FDA) approval in 2005, the use of TEVAR has been increasing (2,3). This shift in practice has come as a result of excellent perioperative outcomes with TEVAR reported in small prospective nonrandomized trials (4,5). However, larger studies using either the Medicare database or meta-analytic methods have suggested that the survival advantage of TEVAR may be lost by 2 years with open surgical repair potentially having superior midterm outcomes (6,7). Reflecting the uncertainty of the comparative effectiveness, professional guidelines fail to offer guidance outside of technical infeasibility for TEVAR or poor candidacy for open surgical repair (8–10).
Efforts to elucidate either short- or long-term results using data from the Nationwide Inpatient Sample and the Medicare database have been limited due to the reliance on imprecise diagnostic and procedural codes from the International Classification of Diseases-Ninth Revision (ICD-9) (2,3,6,11). ICD-9 codes fail to distinguish among ascending, arch, and descending thoracic aortic aneurysms. Aneurysms in each location are treated with a different approach, and expected survival varies (12–16). As such, the inability to distinguish among these disease processes may introduce substantial bias. We undertook the current study using Current Procedural Terminology (CPT) codes, which are able to distinguish among operations on different segments of the thoracic aorta, to compare midterm outcomes of TEVAR and open surgical repair in Medicare beneficiaries with intact undissected descending thoracic aortic aneurysm.
Data collection and study population
We retrospectively reviewed data from the Centers for Medicare & Medicaid Services administrative database from 1999 to 2010 with follow-up through 2014. Patient demographics and survival data were obtained from the Beneficiary Summary file; ICD-9 diagnosis codes pertaining to the descending thoracic aorta were obtained from the MedPAR file; comorbidity data were obtained from the Chronic Conditions Summary file; and surgeon-billed CPT codes pertaining to prior and current surgical procedures were obtained from the Carrier file. We limited our analysis to patients with a CPT code specific to open surgical repair of the descending thoracic aorta (33875) and CPT (33880 and 33881) or ICD-9 (39.73) codes for endovascular repair of the thoracic aorta (17,18).
Patients who underwent concomitant procedures (e.g., ascending aorta or transverse arch repairs, repairs of the thoracoabdominal aorta, endovascular or open repair of the abdominal aorta, and cardiac surgical procedures) were excluded. Patients with aortic dissection, ruptured thoracic aortic aneurysm, trauma, or aortoenteric fistulae were also excluded (Online Figure 1). We determined each patient's underlying aortic pathology with an algorithm (described previously ) that incorporated ICD-9 diagnosis codes recorded during prior and current hospitalizations.
Regression discontinuity design
An analysis using propensity score alone was not appropriate due to the lack of anatomic information—a factor influencing treatment selection (19)—in this administrative database. As such, we used the abrupt change in practice pattern following the introduction of TEVAR in a regression discontinuity design (20). The introduction of TEVAR as a commercially viable alternative to open surgical repair occurred with FDA approval in late 2005. Patients undergoing operations prior to FDA approval were encouraged towards open surgical repair with 100% compliance. Following FDA approval of TEVAR, virtually all patients were compliant with treatment encouragement toward TEVAR, with 94.4% compliance (Online Figure 2). Our analysis then proceeded as an “intent to treat”; patients in the second one-half of the study period undergoing open surgical repair were treated similarly to noncompliers in a randomized trial and analyzed as part of the TEVAR group. This approach was used due to the inability to identify patients who were anatomically ineligible for TEVAR in the early phase of the study.
The introduction of TEVAR as an alternative to open surgical repair increased the pool of patients clinically eligible for aortic repair resulting in an imbalance in baseline covariates (Table 1). Observed differences in baseline covariates between patients encouraged toward open surgical repair (prior to FDA approval of TEVAR) and patients encouraged toward TEVAR (following FDA approval of TEVAR) were balanced using propensity score matching (21,22). A nonparsimonious logistic regression model was used to estimate the probability of encouragement to open surgical repair. Optimal matching was then performed with 2 TEVAR patients matched to each open surgical repair patient to estimate the average treatment effect for open surgical repair (22). Balance between covariates was assessed between the 2 groups before and after matching using the standardized differences approach; a standardized mean difference <0.1 was considered to be ideal and <0.2 was considered to be acceptable (23,24).
Center volume was defined as the volume of open surgical repairs performed prior to the introduction of TEVAR. This allowed for an estimate of operative experience that was less biased by the rapid expansion of treating centers resulting from the introduction of TEVAR. Centers in the 90th percentile of operative volume during this time (>10 open surgical repairs among Medicare beneficiaries in the early period under study) were defined a priori to be high-volume centers.
The primary endpoint was all-cause mortality. Vital status and date of death were validated with National Death Index data from 1999 to 2008 or, if these data were unavailable, an internal Medicare determination of death; agreement between these sources exceeded 99% (25). The secondary endpoint was reintervention (either open or endovascular) on the descending thoracic aorta using the following CPT (open: 33875; endovascular: 33880, 33881, 33883, 33884, and 33886) and ICD-9 codes (endovascular: 39.73). Follow up was available through 2014.
Survival was evaluated using the Kaplan-Meier method and restricted mean survival time (RMST); RMST is the population average for survival during the specified follow-up period (26). To account for nonproportionality of hazards, we used a time-partitioning technique separating perioperative mortality (≤180 days) from the late hazard (>180 days); early mortality was assessed using logistic regression with a robust variance estimator. Hospital volume was tested as an effect modifier for open surgical repair, and 95% confidence intervals (CIs) were estimated with 500 bootstrap replicates (14). Late mortality was then assessed with Cox proportional hazards regression in a landmark analysis (i.e., contingent on 180-day survival). Aortic reintervention was evaluated with death as a competing risk using the Fine-Gray method (27). A 2-sided p value <0.05 was considered to be statistically significant. Analyses were conducted with R version 3.4.1 (R Foundation, Vienna, Austria).
To control for institution-level variation, we performed a sensitivity analysis restricted to the 349 hospitals present in both the early and late periods. The effect of a secular trend was evaluated by comparing open repair in the early and late periods using inverse probability weighting (28). To account for surgical volume as an effect modifier for open surgical repair, additional analyses were performed comparing open surgical repair at a high-volume center to TEVAR and separately open surgical repair at a low-volume center to TEVAR. Finally, the population was restricted to the time period from 2004 to 2006 (i.e., the year before and the year after FDA approval) to examine the effect of TEVAR in closer detail around the time of the discontinuity (20). For additional details on our statistical analyses, please refer to the Methods section in the Online Appendix.
We identified 16,955 Medicare patients with specific codes for endovascular (n = 11,411) or open surgical repair (n = 5,544) of the descending thoracic aorta. Patients who underwent concomitant cardiac surgical procedures; underwent branch vessel revascularization other than carotid-to-subclavian bypass; and had aortic dissection, trauma, or aorto-enteric fistula were excluded (Online Figure 1). Prior to matching, the cohort consisted of 1,235 open surgical repair and 4,580 TEVAR patients. There was a decrease in the frequency of open surgical repair following the introduction of TEVAR and a substantial increase in the overall frequency of descending thoracic aortic aneurysms treated in Medicare beneficiaries (Online Figure 2).
Prior to matching, TEVAR patients tended to be older and have a greater burden of chronic diseases than open surgical repair patients; TEVAR patients were also less likely to be treated at a high-volume open surgical center (Table 1). Our matching algorithm successfully matched each open surgical repair patient with 2 TEVAR patients; balance was achieved across all available covariates (Table 1, Online Table 1, Online Figure 3A). There were 183 patients noncompliant with treatment encouragement to TEVAR; noncompliance was not observed among open surgical repair patients. Unmatched patients in the TEVAR group (n = 2,110) were older and had more chronic comorbidities than patients who were successfully matched (Online Table 2, Online Figure 3B).
The median duration of follow-up in the matched group was 5.6 years (interquartile range [IQR]: 0.7 to 10.0 years) for open surgical repair patients and 4.7 years (IQR: 2.5 to 6.4 years) for TEVAR patients. Mortality was significantly lower among TEVAR patients than open surgical repair patients, log-rank test p < 0.001 (Figure 1). There was clear evidence that the hazard of open surgical repair varied over time, with an early phase of increased risk and a later phase with lower risk of death compared with TEVAR. Mortality at 180 days was greater among open surgical repair patients, 23.8% (95% CI: 21.4% to 26.1%), compared with TEVAR, 10.2% (95% CI: 9.0% to 11.4%), and the interaction between open surgical repair and hospital volume was significant: odds ratio (OR) for high-volume open surgical centers with respect to TEVAR: 1.97 (95% CI: 1.53 to 2.61) and for low-volume open surgical centers: 3.62 (95% CI: 2.88 to 4.51), p value for interaction = 0.002 (Table 2). Conversely, late hazard of death was reduced in the open surgical repair group, hazard ratio (HR) 0.86 (95% CI: 0.77 to 0.95; p = 0.004) (Online Figure 4). The difference in RMST was −209.2 days (95% CI: −298.7 to −119.7 days; p < 0.001) revealing a substantial survival disadvantage with open surgical repair compared with TEVAR at 9 years (Table 2, Online Figure 5).
Among those patients treated after the introduction of TEVAR, patients who were matched had significantly lower comorbidity burden (Online Table 2) and lower mortality during follow-up than patients who went unmatched, HR: 0.70 (95% CI: 0.65 to 0.76; p < 0.001) (Online Figure 6). Sensitivity analyses demonstrated that institutional differences (Online Tables 3 and 4, Online Figure 7), time period (Online Tables 5 and 6, Online Figure 8), adjustments for center volume (Online Tables 7 to 10, Online Figures 9 and 10), and limiting the study period to the time immediately surrounding the discontinuity (Online Tables 11 and 12, Online Figure 11) did not affect the inference from the main analysis. Although there was a reduction in perioperative mortality for open surgical repair in the second one-half of the study period (OR: 0.70; 95% CI: 0.49 to 0.99; p = 0.04), there was no difference in risk of midterm death (HR: 1.21; 95% CI: 0.97 to 1.51; p = 0.1). For additional details on the sensitivity analyses, please see the Online Appendix.
Within the matched regression discontinuity analysis, there were 293 first-time reinterventions on the descending thoracic aorta, and 90.7% (266 of 293) of these reinterventions were performed endovascularly. Moreover, there was substantial risk associated with reintervention. Among open surgical repair patients, reintervention (28.4% open, 71.6% TEVAR) was associated with a 23.5% risk of 180-day mortality. Among TEVAR patients, reintervention (1.9% open, 98.1% TEVAR) was associated with a 19.3% risk of 180-day mortality. Open surgical repair patients experienced a significantly lower likelihood of reintervention at 9 years, 5.3% (95% CI: 3.9% to 6.6%), compared with TEVAR, 10.1% (95% CI: 8.8% to 11.5%), HR: 0.45 (95% CI: 0.34 to 0.60; p < 0.001) (Figure 1).
The use of open surgical repair for intact undissected descending thoracic aortic aneurysms declined precipitously after the introduction of TEVAR despite an increase in the overall frequency of aneurysms repaired. To determine the comparative effectiveness of these competing strategies, we performed a propensity-matched analysis within a regression discontinuity design. Open surgical repair was associated with higher early mortality than TEVAR; however, the late hazard of death and risk of reintervention were lower among patients who underwent open surgical repair (Central Illustration). Despite the potential improved durability of open surgical repair, the initial mortality advantage of TEVAR over open surgical repair persisted until 9 years post-operatively, resulting in a significant survival benefit associated with TEVAR.
Differences in survival between TEVAR and open surgical repair
Hospital volume significantly affected perioperative outcome, one of the major driving forces for the difference between TEVAR and open surgical repair. Regionalization of care by limiting open surgical repair to a group of high-volume and high-performing centers should be considered given the superior outcomes seen at centers of excellence. Referral to an aortic center may further improve the likelihood of appropriate treatment selection after weighing the risks and benefits of each approach. Furthermore, the declining use of open surgical repair for patients with descending thoracic aortic aneurysms suggests that concentrating care at a select group of regional centers with adequate volume to maintain an active program will become increasingly more important as the number of centers performing enough operations to be considered high-volume dwindles.
Beyond the perioperative period, the hazard of death associated with TEVAR was higher than after open surgical repair, and this may have been due to a difference in durability of repair. In the Medtronic Thoracic Endovascular Registry and the 5-year follow-up of the VALOR trial, there was ongoing risk for aortic death beyond the perioperative period in patients treated with TEVAR; midterm reintervention due to endoleak, aneurysm growth, or migration occurred in 14% to 16% of patients with descending thoracic aortic aneurysms (29,30). Reintervention likely underestimates the risk of graft or endograft failure and disease progression, as some patients may die with occult failure while others may be turned down for reintervention on account of technical or clinical considerations. These findings point to the absolute necessity of close follow-up for TEVAR patients to detect late complications. The increased hazard for reoperation is an intrinsic limitation of TEVAR given the inability to occlude feeding intercostal vessels, ensure adequate distal and proximal seal in the face of disease progression within adjacent segments, and avoid leaks between stent grafts with or without migration of components.
Despite potential differences in durability with a concomitant increase in late hazard of death for TEVAR, the mean survival of patients out to 9 years was >6 months greater for patients receiving TEVAR than for open surgical repair. Most of this survival difference was accumulated in the early phase; as such, whether a survival difference would exist favoring open surgical repair in a younger cohort of patients with lower perioperative risk and greater potential long-term survival is unknown. Inclusion of younger patients—mean age 51 years for open surgical repair—in the meta-analysis by Cheng et al. (7) may explain the lower perioperative mortality and the earlier convergence of survival curves. An important area for future investigation, using a clinical database, may be to develop a predictive algorithm for risk stratifying patients undergoing open surgical repair, with the lowest-risk patients potentially benefiting the most from open surgical repair. Unfortunately, a comparable study in younger patients would not be possible using the Nationwide Inpatient Sample or the Society of Thoracic Surgeons-Adult Cardiac Surgery Database due to the absence of long-term follow-up in those databases. Additionally, state-level databases, which rely on ICD-9 codes, would also be inadequate for such an analysis.
Differences between TEVAR and endovascular aortic repair
The existing published data on TEVAR versus open surgical repair suggests that the survival benefit of TEVAR is lost as early as 2 years but possibly as late as 5 years (6,7,31). These results appeared to parallel the modest overall survival benefit attributable to endovascular aortic repair for abdominal aortic aneurysms, which achieved parity at approximately 3 to 4 years with an RMST difference reported by Schermerhorn et al. (32) of only 5.6 days at 8 years (32–34). This is in stark contrast with the substantial survival advantage discovered in our study, with an RMST difference of 209.2 days and convergence of survival curves at 9 years. Despite the reduced durability compared with open surgical repair seen in our analysis, endovascular repair may be better suited to the descending thoracic aorta than the abdominal aorta, as there more frequently exist long segments of aorta that may act as appropriate landing zones capable of resisting disease progression for a longer time post-procedure. This may explain the difference in comparative effectiveness of endovascular and open repair between the descending thoracic aorta and the abdominal aorta.
A substantial minority of patients treated in the latter one-half of the study period were excluded as a result of our matching algorithm. Reflective of the greater age and comorbidity burden, the survival of these patients was significantly worse after TEVAR than those patients who were able to be matched mirroring results observed by our group previously (35). Due to the reduced risk of perioperative morbidity and mortality, endovascular treatment has been offered to patients who were traditionally considered to be poor surgical candidates contributing—in part—to the greater frequency of descending thoracic aortic aneurysms treated. Excluding these patients improves the internal validity of our study due to the lack of eligibility for open surgical repair in this subset of patients. Given the high risk and poor survival of this cohort, TEVAR appears to be appropriate if any intervention is to be undertaken at all.
This investigation was limited by its retrospective design in an administrative database, and although causal inference techniques were used to evaluate the comparative effectiveness of the 2 competing treatments, our study falls short of a randomized trial in its ability to posit a causal effect. Moreover, unmeasured confounding in the form of unobserved clinical variables, such as frailty or anatomic features, may be present. Regression discontinuity with an “intention-to-treat” design was used to limit many of these potential sources of bias, which have not been addressed in prior investigations. Propensity score matching was used to address imbalances in baseline covariates given the increase in the chronic comorbidities observed in the TEVAR period. Following matching, all covariates were balanced adequately. With respect to outcome measures, our use of an administrative database left us unable to determine whether endoleaks were present following TEVAR (i.e., that the thoracic endograft was functioning appropriately during follow-up). Finally, the perioperative outcome for open surgical repair at high-volume aortic centers in a contemporary cohort may be better than we observed in the Medicare database, thus altering the risk-benefit tradeoff to be less weighted toward TEVAR. Although we performed sensitivity analyses to evaluate the potential effect of hospital-level characteristics and secular trends on our inference, the true effect in a contemporary cohort could not be estimated.
Among Medicare beneficiaries, patients undergoing open surgical repair for intact undissected descending thoracic aortic aneurysm had greater odds of early death than TEVAR patients. Despite the lower risk of reintervention and lower late hazard of death, open surgical repair only achieved parity with TEVAR after 9 years, resulting in a substantial survival benefit associated with TEVAR. The superior survival observed in patients undergoing TEVAR compared with open surgical repair suggests that TEVAR ought to be considered first-line among Medicare beneficiaries with open surgical repair restricted to high-volume centers and patients with low risk of perioperative mortality.
COMPETENCY IN MEDICAL KNOWLEDGE: Among Medicare beneficiaries with intact descending thoracic aortic aneurysms, endovascular stent-grafting by an experienced operator is the preferred method of thoracic aortic repair.
TRANSLATIONAL OUTLOOK: Further studies are needed to define the optimum timing of intervention for patients with thoracic aortic aneurysms undergoing elective endovascular therapy.
This work was conducted with support from a KL2 (to Dr. Chiu) and TL1 (to Dr. Goldstone) Mentored Career Development Award of the Stanford Clinical and Translational Science Award to Spectrum (National Institutes of Health [NIH] KL2 TR 001083, NIH TL1 TR 001084, NIH TR 001085). The funder had no involvement in the design or conduct of the study; in the collection, analysis, or interpretation of the data; or in the preparation, review, or approval of the manuscript. Dr. Miller has served as co-Principal Investigator for the Abbott Vascular COAPT MitraClip Trial, Stanford Principal Investigator for the COMMENCE Trial and PARTNER I, II, and III Trials, Edwards Lifesciences, and Stanford Principal Investigator for the Medtronic SURTAVI Trial; has served as a consultant for Medtronic; and has served on the executive committee of Edwards Lifesciences. Dr. Mitchell has received consulting fees from W.L. Gore. Dr. Dake has received consulting fees from W.L. Gore, Abbott Vascular, and Medtronic; and has received consulting and lecture fees from Cook Medical. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Listen to this manuscript's audio summary by Editor-in-Chief Dr. Valentin Fuster on JACC.org.
- Abbreviations and Acronyms
- confidence interval
- current procedural terminology
- hazard ratio
- International Classification of Diseases-Ninth Revision
- odds ratio
- restricted mean survival time
- thoracic endovascular aortic repair
- Received August 6, 2018.
- Revision received October 22, 2018.
- Accepted October 30, 2018.
- 2019 American College of Cardiology Foundation
- Goodney P.P.,
- Travis L.,
- Lucas F.L.,
- et al.
- Cheng D.,
- Martin J.,
- Shennib H.,
- et al.
- Hiratzka L.F.,
- Bakris G.L.,
- Beckman J.A.,
- et al.
- Erbel R.,
- Aboyans V.,
- Boileau C.,
- et al.
- Williams J.B.,
- Peterson E.D.,
- Zhao Y.,
- et al.
- Vapnik J.S.,
- Kim J.B.,
- Isselbacher E.M.,
- et al.
- Seabrook G.R.
- Chiu P.,
- Sailer A.,
- Baiocchi M.,
- et al.
- O'Keeffe A.G.,
- Geneletti S.,
- Baio G.,
- Sharples L.D.,
- Nazareth I.,
- Petersen I.
- Morales D.L.,
- McClellan A.J.,
- Jacobs J.P.
- Uno H.,
- Claggett B.,
- Tian L.,
- et al.
- Patterson B.,
- Holt P.,
- Nienaber C.,
- Cambria R.,
- Fairman R.,
- Thompson M.
- Patel R.,
- Sweeting M.J.,
- Powell J.T.,
- Greenhalgh R.M.,
- for the EVAR Trial Investigators