Primary Stroke and Failure-to-Rescue Following Thoracic Endovascular Aortic Aneurysm Repair

Abstract

Objective:

To characterize the impact, on failure to rescue, of cerebrovascular accident as a first postoperative complication after thoracic endovascular aortic aneurysm repair (TEVAR).

Design:

A retrospective cohort study using of National Surgical Quality Improvement Program Participants User File.

Setting:

United States hospitals taking part in the National Surgical Quality Improvement Program.

Participants:

Patients >18 years, who underwent TEVAR for nonruptured thoracic aortic aneurysm between 2005 and 2018, and developed one or more major postoperative complications within 30 days after surgery.

Interventions:

None.

Measurements and Main Results:

Out of 3,937 patients who underwent TEVAR for nonruptured thoracic aneurysm, 1,256 (31.9%) developed major postoperative complications (stroke incidence: 11.4% [143/1256]). In adults <65 years old, the occurrence of stroke as the primary complication, relative to the occurrence of other complications, was associated with ten times greater risk of failure to rescue (29.4% v 4.6%; odds ratio [OR]: 10.10; 95% confidence interval [CI] 2.45–41.56; p < 0.001). The effect size was relatively lower when stroke occurred but was not the primary complication (20.0% v 4.6%; OR: 7.55; 95% CI 1.37–41.71; p = 0.020). In patients ≥65 years, the occurrence of stroke as the primary complication did not carry the similar prognostic value.

Conclusion:

Younger patients who developed stroke were up to ten times more likely to die, relative to patients who developed other major complications. Survival was substantially reduced when stroke was the primary complication. The authors’ findings imply that to maximize the survival of patients undergoing TEVAR, efforts may be needed to predict and prevent stroke occurrence as a primary postoperative morbidity event.

THORACIC ENDOVASCULAR aortic repair (TEVAR), rather than open surgical repair, has become the preferred treatment approach for diseases of the thoracic aorta. Compared with open surgery, TEVAR is less invasive and associated with fewer perioperative complications^1,2 and higher survival.^1–5 Although considered to carry a lower overall risk of surgical complications, between 2.9% and 4.3% of patients undergoing TEVAR develop stroke, of whom up to 33% may die.⁶ Besides the short-term risk of mortality, stroke after TEVAR also is associated with long-term disability and impaired quality of life.⁶

Most studies on the effects of stroke after TEVAR on mortality have not focused specifically on its timing relative to surgery and other complications. First, there is a lack of epidemiologic data regarding the proportion of stroke occurring as the first morbidity event (primary complication). Such data would improve clinicians’ knowledge about when to expect the occurrence of stroke, which could reinforce surveillance and monitoring after TEVAR.⁷ Improvement of surveillance helps with early recognition of complications, which is the gateway for effective patient rescue processes.⁸

Although it is known that stroke after TEVAR has profound implications on mortality and morbidity, it is unclear how much of this is attributable to the timing of stroke as either a primary or secondary morbidity event. An understanding of the variation of mortality risk according to whether stroke is the index complication would be critical for implementing time-specific rescue processes. Although it generally may be assumed that the occurrence of stroke as a first-ever morbidity event carries the greatest risk of mortality, this has yet to be empirically demonstrated. If proven that primary stroke portends the poorest prognosis, healthcare providers would have an empirical basis for engaging in aggressive measures to prevent the occurrence of stroke as primary morbidity event. Such data also would provide an avenue for quality control interventions because failure to rescue (FTR), defined as patient mortality after potentially treatable complication(s), is an actionable quality indicator.⁹

Therefore, the authors’ objective was to characterize the impact, on FTR, of cerebrovascular accident as the first occurring postoperative complication after TEVAR. The authors hypothesized that the occurrence of stroke as the first occurring complication after TEVAR would carry the greatest mortality risk.

Materials and Methods

Study Design and Data Sources

The authors used the National Surgical Quality Improvement Program (NSQIP) patient user file to perform a retrospective study of patients >18 years who underwent TEVAR in participating hospitals between 2005 and 2018. Briefly, the NSQIP is an ongoing data-driven quality management initiative that collects more than 300 perioperative variables, including demographic, clinical, and biological characteristics, and 30-day postoperative mortality and morbidity events, at participating hospitals across the United States. Trained surgical clinical reviewers collect these data from medical records of patients by adhering to standard definitions, resulting in a low disagreement rate (most recently 2%).¹⁰ Data collection utilizes a systematic sampling design to minimize selection bias and improve representation, thus making the NSQIP one of the most complete and reliable surgical databases. To ensure high quality and standardized data collection, clinical reviewers undergo structured continuing education and routine audits by the NSQIP committee. Further details regarding the NSQIP have been published elsewhere.^10–12 This study was approved by the authors’ center’s Institutional Review Board under a waiver of informed consent.

Study Population

The authors used the postoperative diagnosis codes 441.2 (International Classification of Diseases 9th Revision [ICD-9]) and I71.2 (10th Revision [ICD-10]) to identify patients who underwent TEVAR for nonruptured thoracic aortic aneurysm between 2005 and 2018. As in a previous study,¹³ the authors did not include patients with ruptured thoracic aneurysm because they have a disproportionately high risk of mortality. The authors kept patients in the analytical sample if they developed one or more major postoperative complications within 30 days after surgery. Major postoperative complications were considered as the occurrence of grade II, III, or IV postoperative morbidity events, according to the Clavien-Dindo classification system.¹⁴ The authors did not include cardiac arrest in their analyses (90 patients) for two main reasons. First, cardiac arrest precedes death as it represents, in most cases, the endpoint of a cascade of morbidity events. Second, the principle of FTR is to prevent the cascade of events leading to mortality after the occurrence of a sentinel complication, instead of only preventing their incidence. In line with this, the purpose of this study was to provide an epidemiologic distribution of the timing of stroke relative to other complications, and to understand the prognostic value of timing phenotypes. Unfortunately, cardiac arrest is strongly linked to mortality and provides brief opportunity for reversal.

Primary Outcome and Predictor of Interest

The primary outcome of this study was FTR, considered as the occurrence of all-cause mortality within 30 days of surgical procedure, among patients who developed major postoperative complications. The primary predictor was the occurrence of stroke, relative to other major complications. Per the study database, postoperative stroke was defined as an embolic, thrombotic, or hemorrhagic cerebrovascular event with patient motor, sensory, or cognitive dysfunction that persisted for 24 hours or more and occurred within 30 days of an surgery.¹⁵ For each postoperative complication, the NSQIP measures the number of days between the index surgery and its occurrence. The authors used this information to test the timing of stroke, relative to other complications, and divide patients who developed stroke into the following two groups: (1) patients with primary stroke and (2) patients with secondary stroke. Stroke was considered as a primary postsurgical complication if it occurred as first or among the first morbidity events after surgery (ie, occurring before or on the same day as other complications); otherwise stroke was considered as a “secondary” complication.

Statistical Analyses

All analyses were stratified by age group (<65 v ≥65 years) because the risk factors for and prognosis of stroke differ by age.¹⁶ The authors summarized categorical data by presenting their relative frequency distributions. Continuous data, with non-normal distribution, were summarized as median and interquartile range. The authors used univariate and multivariate logistic regression models to estimate the odds ratio (OR) of FTR, according to whether stroke was the first occurring postoperative complication. The authors adjusted all analyses for demographic variables and preoperative patients’ characteristics that were selected based on prior probability of confounding the association between stroke and mortality after TEVAR. These preoperative patient characteristics included gender (female v male), body mass index (continuous), current smoking status (within one year of surgery), ventilator dependence within 48 hours of surgery (yes v no), emergent case status (yes v no), American Society of Anesthesiologists (ASA) class (≥3 v <3), diabetes (noninsulin-dependent or insulin-dependent diabetes mellitus: yes v no), dialysis treatment within two weeks of surgery (yes v no), dyspnea (at rest or moderate exertion: yes v no), functional health status within 30 days of surgery (dependent v independent), history of chronic heart failure within 30 days of surgery (yes v no), history of chronic obstructive pulmonary disease within 30 days of surgery (yes v no), hypertension (yes v no), systemic sepsis (sepsis, septic shock, or systemic inflammatory response syndrome: yes v no), bleeding disorder (yes v no), steroid use for chronic condition (yes v no), and transfusion >72 hours before surgery (yes v no). The authors used Stata, version 15 (Stata-Corp), to perform statistical analysis, and considered a p value less than 0.05 to be statistically significant, unless otherwise indicated.

Sensitivity Analysis

The authors evaluated the robustness of their findings on the potential impact of unmeasured confounding by estimating the E-value. Briefly, the E-value quantifies the strength that an unmeasured confounding must have in order to nullify the observed association (ie, shifting the estimated odds ratio to 1.0).¹⁷ A larger E-value implies that the unmeasured confounding would need to be stronger for the observed association to be nullified.¹⁷

Results

Patients’ Characteristics

The authors identified 3,937 patients who underwent TEVAR for nonruptured thoracic aneurysm during the study period, of whom 1,256 (31.9%) developed major postoperative complications and were included in the analysis. The authors present their characteristics by age category and timing of postoperative stroke in Table 1. Among adults <65 years, patients who developed stroke as the primary complication were more likely to be female, to have an ASA ≥3, to have a history of COPD, to have hypertension, and to have a systemic sepsis, but were less likely to be of African-American race or to be on steroid medication. This pattern was like that observed in patients ≥65 years regarding gender. However, elderly patients who developed stroke as the primary complication were more likely to have diabetes and to be of African-American race.

Table 1.

Characteristics of the Study Population, by Age Category and Timing of Postoperative Stroke^*

Preoperative characteristies	Patients <65 Years Old 31.8% (n = 403)				Patients ≥65 Years Old 68.2% (n = 864)
	Overall	Developed other complications	Developed secondary stroke	Developed primary stroke	Overall	Developed other complications	Developed secondary stroke	Developed primary stroke
	n(%) †	n(%) †	n(%) †	n(%) †	n(%) †	n(%) †	n(%) †	n(%) †
Study population	400 (100)	368 (92.0)	15 (3.8)	17 (4.3)	855 (100)	744 (87.0)	48 (5.6)	6 (7.4)
BMI. median (interquartile range)	28.0 (24.5–32.4)	27.8 (24.6–32.4)	29.4 (28.3–31.5)	28.4 (22.1–34.4)	26.1 (22.8–30.4)	25.8 (22.7–30.4)	28.2 (24.2–31.8)	26.1 (23.0–29.8)
female sex	126 (31.5)	1 14(31.0)	4 (26.7)	8 (47.1)	412 (48.2)	354 (47.6)	23 (47.9)	35 (55.6)
African-American race	71 (17.8)	68 (18.5)	2 (13.3)	1 (5.9)	94 (11.0)	82 (11.0)	5 (10.4)	7 (11.1)
Smoker within 1 year of surgery	134 (33.5)	123 (33.4)	5 (33.3)	6 (35.3)	232 (27.1)	198 (26.6)	13 (27.1)	21 (33.3)
Ventilator dependent	7 (1.8)	4 (1.1)	2 (13.3)	1 (5.9)	17 (2.0)	13 (1.7)	3 (6.3)	1 (1.6)
Emergency case	33 (8.3)	27 (7.3)	3 (20.0)	3 (17.6)	65 (7.6)	56 (7.5)	6 (12.5)	3 (4.8)
ASA >3	264 (66.0)	243 (66.0)	11 (73.3)	10 (58.8)	509 (59.6)	452 (60.8)	30 (62.5)	27 (42.9)
Diabetes	36 (9.0)	31 (8.4)	3 (20.0)	2 (11.8)	114 (13.3)	98 (13.2)	6 (12.5)	10 (15.9)
On dialysis before surgery	15 (3.8)	13 (3.5)	0 (0.0)	2 (11.8)	31 (3.6)	27(3.6)	2 (4.2)	2 (3.2)
Dyspnea	116 (29.0)	107 (29.1)	5 (33.3)	4 (23.5)	271 (31.7)	233 (31.3)	19 (39.6)	19 (30.2)
Functional dependency	14 (3.5)	12 (3.3)	1 (6.7)	1 (5.9)	73 (8.6)	64 (8.6)	5 (10.6)	4 (6.3)
CHE 30 days before surgery	23 (5.8)	22 (6.0)	0 (0.0)	1 (5.9)	36 (4.2)	35 (4.7)	1 (2.1)	0 (0.0)
I li story of seve re COPD	34 (8.5)	29 (7.9)	1 (6.7)	4 (23.5)	208 (24.3)	185 (24.9)	8 (16.7)	15 (23.8)
Hypertension	295 (73.8)	266 (72.3)	14 (93.3)	15 (88.2)	758 (88.7)	659 (88.6)	46 (95.8)	53 (84.1)
Systemic sepsis	20 (5.0)	17 (4.6)	2 (13.3)	1 (5.9)	44 (5.2)	40 (5.4)	2 (4.2)	2 (3.2)
Bleeding disorder	18 (7.0)	16 (6.7)	1 (14.3)	1 (8.3)	69 (12.3)	58 (12.0)	6 (18.8)	5 (11.1)
Steroid use for chronic condition	12 (3.0)	12 (3.3)	0 (0.0)	0 (0.0)	46 (5.4)	37 (5.0)	5 (10.4)	4 (6.3)
Transfusion 72 hours before surgery	11 (2.8)	10 (2.7)	0 (0.0)	1 (5.9)	26 (3.0)	24 (3.2)	1 (2.1)	1 (1.6)

Abbreviations: ASA, American Society of Anesthesiology; BMI, body mass index; CHF, Congestive heart failure; COPD, chronic obstructive pulmonary disease; TEVAR, thoracic endovascular aortic aneurysm repair.

^*The authors included in their analytical sample patients who underwent TEVAR for nonruptured thoracic aortic aneurysm between 2005 and 2018 and who developed one or more major postoperative complications within the 30 days after surgery.

^†Percentages are for column.

Stroke and FTR by Age

Out of the 1,256 patients who developed major postoperative complications, 143 (11.4%) developed stroke, of whom 32 (7.1%) were adults <65 years and 111 (13.0%) were elderly patients ≥65 years (Table 2). In adults <65 years, patients with stroke were estimated to be almost nine times more likely to die, compared with patients who developed other major complications (25.0% v 4.6%; adjusted [a]OR: 8.98, 95% CI 2.84–28.44; p < 0.001). The degree to which stroke, relative to other complications, carried poorer prognosis was weaker in elderly patients ≥65 years old (23.4% v 13.2%; aOR: 2.33, 95% CI 1.40–3.88; p < 0.001) - Table 2.

Table 2.

Age-stratified Odds Ratios for Failure to Rescue Associated With Timing of Postoperative Stroke

Age	Complications	FTR,n/N (%)*	Unadjusted Analysis		Adjusted Analysis†		E-value‡
			OR (95% Cl)	p Value	OR (95% Cl)	p Value
<65 years	Other complications	17/368 (4.6)	Reference		Reference
	Secondary stroke	3/15 (20.0)	5.16(1.33–20.02)	0.018	7.55(1.37–41.71)	0.020	14.58
	Primary stroke	5/17(29.4)	8.60 (2.72–27.20)	<0.001	10.10(2.45–41.56)	0.001	19.69
≥65 years	Other complications	98/744(13.2)	Reference		Reference
	Secondary stroke	12/48 (25.0)	2.20(1.11–4.37)	0.025	2.60(1.26–5.35)	0.010	4.64
	Primary stroke	14/63 (22.2)	1.88 (1.00–3.54)	0.049	2.15(1.12–4.13)	0.022	3.72
<65 years	Other complications	17/368 (4.6)	Reference		Reference
	Stroke	8/32 (25.0)	6.88 (2.70–17.56)	<0.001	8.98(2.84–28.44)	<0.001	17.45
≥65 years	Other complications	98/744(13.2)	Reference		Reference
	Stroke	26/111 (23.4)	2.02(1.24–3.28)	0.005	2.33(1.40–3.88)	0.001	4.09

Abbreviations: CI, confidence interval; FTR, failure to rescue; OR, odds ratio.

^*Percentages are for row.

^†Adjusted for gender, body mass index, current smoking status, ventilator dependence within 48 hours of surgery, emergent case status, American Society of Anesthesiologists class, diabetes, dialysis treatment within 2 weeks of surgery, dyspnea, functional health status within 30 days of surgery, history of chronic heart failure within 30 days of surgery, history of chronic obstructive pulmonary disease within 30 days of surgery, hypertension, systemic sepsis, bleeding disorder, steroid use for chronic condition, and transfusion >72 hours before surgery.

^‡The E-value quantifies the magnitude that an unmeasured confounding must be in order to nullify the observed association (ie, causing the estimated odds ratio to 1.0), conditional on the measured covariates. A larger E-value implies that the unmeasured confounding would need to be stronger for the observed association to be nullified.

Primary Stroke and FTR by Age

Stroke was the primary complication among 4.3% of adults <65 years and 7.4% of elderly patients ≥65 years. In adults <65 years old, there appeared to be a graded relationship between FTR and the tendency of stroke to occur as the primary complication. Specifically, the occurrence of stroke as the primary complication, relative to the occurrence of other complications, was associated with ten times greater risk of FTR (29.4% v 4.6%; aOR: 10.10; 95% CI 2.45–41.56; p < 0.001). The effect size was relatively lower when stroke occurred but was not the primary complication (20.0% v 4.6%; OR: 7.55; 95% CI 1.37–41.71; p = 0.020). In elderly patients ≥65 years old, the occurrence of stroke as the primary event did not carry the highest risk for FTR. To the contrary, the adjusted odds ratio for FTR was 2.60 (95% CI 1.26–5.35, p = 0.010) when comparing patients developing secondary stroke versus patients developing other complications, and it was 2.15 (95% CI 1.12–4.13; p = 0.022) when comparing patients developing primary stroke to patients developing other complications.

Sensitivity Analyses

The E-value ranged from 3.72 to 19.69, implying that an unmeasured confounder must be associated with both the exposure and outcome measures by at least an odds ratio of 3.72-fold to nullify the observed odds ratios. Weaker confounder associations could not negate these observed ORs, implying that an unmeasured confounding must be strong to fully explain away the authors’ findings.

Discussion

In this study the authors performed a risk assessment for FTR according to the timing of stroke among patients who developed major complications after TEVAR. Overall, patients who developed stroke were up to ten times more likely to die, relative to patients who developed other major complications. In addition, the occurrence of stroke as the primary complication carried the poorest prognosis for FTR in patients <65 years. The occurrence of stroke as the primary complication did not have similar prognostic value between patients <65 years and those ≥65 years. These findings were robust for the potential impact of confounding because of factors that were not measured in this study. Of note, the incidence of stroke after TEVAR was 3.3%, which was consistent with previous estimates ranging from 2.9% to 4.3%.^18–22 In addition, the authors’ estimate of the incidence of major complications after TEVAR was consistent with a previous publication.²³

This study expanded previous research reporting on the risk of mortality associated with stroke after TEVAR,^6,24 by evaluating the temporal distribution of stroke relative to other major complications. Besides describing the relative timing of stroke, this study discriminated between its prognostic value when it occurred as the primary adverse events and afterwards. Such empirical data have two main implications for research and quality control measures, thus providing an avenue to improve the surgical care of patients undergoing TEVAR. First, knowing that most stroke events occur as the primary postoperative adverse event may be helpful in improving early identification, which is the gateway for successful rescue.²⁵ Second, the authors’ findings also may be helpful for tailoring mitigation strategies for stroke by accounting for the differential prognosis of timing phenotypes. Implementing an appropriate treatment after complication is a critical component of patient rescue and, thus, surgical quality.²⁵ Actions that can be taken in the setting of postoperative stroke include acute endovascular thrombectomy, which can be performed within six hours of stroke onset.^26,27 It is well-established that timely intervention after stroke is of the utmost importance to reduce case fatality, secondary complications, and secondary stroke.²⁸

The authors’ findings implied that to maximize the survival of patients undergoing TEVAR, efforts should go beyond just preventing the development of stroke, but also to predict and prevent its occurrence as a primary postoperative morbidity event. Such efforts may start with developing prognostic profiling for developing stroke as the primary complication, which would identify clinical and biologic features for which a clinician should look in the preoperative period. Such quantitative data would provide clinicians with the ability to make concerted efforts to systematically identify high-risk patients for primary stroke. Empirical data on the predictors of primary stroke also would be valuable for developing tailored neuroprotective actions to be taken in the preoperative settings. To date, several studies have developed prediction tools for stroke after TEVAR.^21,29–31 However, the authors are unaware of any attempt to develop a preoperative prediction model for stroke as the primary adverse event after TEVAR, which appears to be a necessary next step. Although further research is needed to reliably identify preoperative risk factors for primary stroke after TEVAR, there are existing ways clinicians can leverage to establish a prognosis. Imaging techniques, such as computed tomography, could play a role in identifying high-risk patients. The Shaggy Aorta Scoring System,³² which assesses computed tomography slices from the sinotubular junction to the aortic bifurcation for thrombi, preliminarily has been validated³¹ for use as a predictive indicator of ischemic stroke after TEVAR.

The occurrence of stroke as the primary complication carries the poorest prognosis in patients <65 years, but not in patients ≥65 years. This difference in prognostic value seems counterintuitive because the mortality risk associated with stroke increases with age.^33,34 However, age is not the principal determinant of survival in aortic aneurysm repairs.³⁵ Young adults with stroke have an increased risk of mortality compared with the general population.³⁶ It may be possible that the overall lower perioperative comorbidity burden in the younger cohort has resulted in a greater predictive value of the timing of stroke on mortality. Clearly, more research is needed to explore the mechanisms of the differential prognosis of primary stroke on younger patients, as the authors found only one study comparing aortic aneurysm between younger and older patients, which did not yield any insights relevant to the authors’ observation.³⁷

Of note, a subset (5%) of the authors’ cohort was reported to have preoperative systemic sepsis. Because sepsis is characterized by a state of systemic inflammation, it is a major determinant of both arterial and venous thromboses.³⁸ Further research may be needed to evaluate the burden of surgical mortality in patients who have sepsis and develop stroke after TEVAR. Such data would be valuable in identifying potential areas of intervention to improve the surgical risk of patients with perioperative sepsis.

Study Limitations

The interpretation of the authors’ findings had several limitations. The first limitation stemed from the absence of specific definition of stroke after TEVAR, and no recommendation on how to document these complications.³⁹ In this study, postoperative stroke was defined as an embolic, thrombotic, or hemorrhagic cerebrovascular event with patient motor, sensory, or cognitive dysfunction that persisted for 24 hours or more and occurred within 30 days of a surgery.¹⁵ This definition overlooks silent cerebral infarctions, which may be far more significant than previously realized. A recent imaging study detected cerebral infarction in 81% of TEVAR patients who underwent MRI, with 84% of those infarctions being clinically silent.⁴⁰ Despite this limitation, the NSQIP data collection follows a standard definition, reducing the potential for inter-rater disagreement.⁴¹ Second, the authors’ study database did not ascertain whether the etiology of stroke is hemorrhagic or ischemic. Additionally, the dataset did not report on other neurologic complications, such as paralysis and subdural hematoma, which are known complications of TEVAR. Furthermore, the dataset did not measure characteristics related to the aortic lesions before hospitalization, including the level of the aneurysm, the atheromatous burden of the aorta, and the need for carotid-subclavian bypass. Relatedly, the authors’ study database did not record an exhaustive list of all possible complications.⁴¹ This prevented the authors’ analyses from accounting for all possible complications, thus somewhat reducing the scope of their findings. However, the NSQIP is one of the most comprehensive surgical databases,⁴¹ which implies that this study covered a broad spectrum of possible complications. Third, the NSQIP data file does not contain identifier for hospitals, precluding any attempt to adjust the analyses for potential clustering for mortality rate at the hospital level. However, a clustering of mortality within hospitals would less likely explain away the magnitude of effect estimates, but rather their statistical significance. Fourth, the authors acknowledge two main limitations stemming from the observational nature of their study, including lack of granularity and potential for data entry errors. These limitations are inherent to database studies, and are less likely to explain the authors’ findings, which were consistent with previous literature. In addition, the authors used an unambiguous categorical outcome (mortality), which generally is helpful in reducing the occurrence of differential or nondifferential misclassification of the outcome.

Conclusion

The authors found that younger patients who developed stroke were up to ten times more likely to die, relative to patients who developed other major complications after TEVAR. Most importantly, survival was substantially reduced when stroke was the primary complication. The authors’ findings implied that to maximize the survival of patients undergoing TEVAR, efforts should go beyond just preventing the incidence of stroke, and further to establish methods to predict and prevent its occurrence as a primary postoperative morbidity event. Although the authors’ findings may require prospective validation, they may be useful for improving patient rescue through reinforcement of complication recognition and mitigation strategies specific to timing phenotypes. Further research also is needed to identify risk factors for development of primary stroke.