Female sex independently predicts mortality following TEVAR for intact descending thoracic aortic aneurysms

Abstract

Objective

While sex differences in the pathogenesis, presentation, and outcomes of repair for abdominal aortic aneurysms are well studied, less is known about sex differences following thoracic endovascular aneurysm repair (TEVAR). The goal of this study was to evaluate the association between sex and morbidity and mortality following TEVAR.

Methods

A retrospective review of all TEVAR in the Society for Vascular Surgery Vascular Quality Initiative (SVS-VQI) registry from 2011–2015 was conducted, excluding those with dissection, trauma, and rupture. Statistical analysis was performed using the Fisher’s exact test and the Mann-Whitney U test for categorical and continuous variables. Multivariable logistic regression and Cox hazards modeling were used to account for differences in demographics, comorbidities, and aneurysm characteristics in 30-day mortality and long-term survival.

Results

We identified 2,574 patients (40% women) who underwent TEVAR. Women were older, less likely white, and had smaller aortic diameters but larger aortic size indices (aortic diameter/body surface area). Women also had more chronic obstructive pulmonary disease, but less coronary artery disease and fewer coronary interventions. Women were more likely to be symptomatic at presentation and subsequently have a non-elective procedure. Women had higher estimated blood loss (EBL > 500cc: 20% vs. 17%, P = .04), were more likely to be transfused (29% vs. 21%, P < .001), and more frequently underwent iliac access procedures (4.3% vs. 2.1%, P < .01). Operative time and left subclavian intervention were similar. Postoperatively, women had increased median hospital (5 vs. 4 days, P < .001) and intensive care unit lengths of stay (2.5 vs. 2, P < .001) and were less likely discharged home (75% vs. 86%, P < .001). Mortality was higher for women at 30 days (5.4% vs. 3.3%, P < .01) and one year (9.8% vs. 6.3%, P < .01). After adjusting for age, aortic size index, symptoms, and comorbidities, female sex remained independently predictive of 30-day (OR 1.5, 95% CI: 1.1–2.1, P < .01) and long-term mortality (HR 1.3, 95% CI: 1.03–1.6, P = .02).

Conclusions

Even after adjusting for differences in age and comorbidities, female patients have higher perioperative mortality and lower long-term survival following TEVAR. These findings, along with the rupture risk by sex, should be considered by clinicians when determining the timing of intervention.

Introduction

Sex differences in outcomes following vascular surgery exist and are well described for abdominal aortic aneurysms (AAA), with female patients having higher rates of rupture and worse outcomes following both open and endovascular repair.^1–5 However, much less is known about sex differences in outcomes following repair of descending thoracic aortic aneurysms.^{6, 7} Although several prospective trials have suggested sex differences, especially in vascular complications, after TEVAR, these have been secondary observations and not part of the primary study aim.^8–10 Institutional studies assessing sex differences following TEVAR have suggested no difference in mortality, although these were limited by small sample sizes.^{11, 12} One recent, large retrospective study of TEVAR using NSQIP data showed female sex to be associated with increased 30-day mortality.¹³ However, the association lost statistical significance after adjusting for intraoperative variables, including operative time and use of iliac artery exposure, suggesting the mortality difference may be, in part, explained by varying aortoiliac and femoral arterial disease burden between sexes. Further investigation with aneurysm- and vascular-specific data is needed to better elucidate this relationship. Therefore, the goal of this study was to evaluate the relationship between sex and outcomes following TEVAR in a contemporary series, with focus on 30-day and long-term mortality.

Methods

The Beth Israel Deaconess Medical Center Institutional Review Board approved this study and waived informed consent due to the use of de-identified data.

Population

This is a retrospective cohort study using the Society for Vascular Surgery Vascular Quality Initiative (SVS-VQI), which is a national clinical registry established as a collaboration between regional quality groups in an effort to improve patient care through the prospective collection of clinical data. At the time of this analysis, the VQI included 16 regions and over 300 participating hospitals. Within participating hospitals, complete capture of all included procedures is expected, with regular performance reviews to ensure compliance. More information about the VQI can be found at www.vascularqualityinitiative.org. We identified all patients undergoing TEVAR for descending thoracic aortic aneurysm between 2011 and 2015, and excluded those with trauma (n=262), dissection (n=873), or rupture (n=305). As the TEVAR database in the VQI contains TEVAR and complex EVAR patients, all complex EVAR patients were also excluded (n=699).

Variables

Demographics, comorbid conditions, operative details, and in-hospital postoperative outcomes were identified for all patients. Aortic diameter was defined by the VQI as the maximum total aortic diameter within the diseased segment being treated. Body mass index (BMI) and body surface area (BSA) were calculated for each patient using height and weight information. We used the standard formula for BMI: BMI = weight (kg)/height (m²), and the Dubois and Dubois formula¹⁴ for BSA: BSA = [weight (kg)^0.425 × height (m)^0.725] × 0.20247. Aortic size index (ASI) was defined as aortic diameter/BSA.^{3, 15} Smoking history was defined as current or former smoking. A single preoperative creatinine value was used to estimate the glomerular filtration rate (GFR) for each patient using the Modification of Diet in Renal Disease Study equation,¹⁶ and renal insufficiency was considered present for estimated GFR < 30 mL/min/1.73m². Preoperative anemia was defined as hemoglobin < 10 g/dL. Preoperative functional status was defined as independent if the patient could perform light work and self-care and was not under assistive care or bed-bound. The presence or absence of other comorbidities were recorded in a binary fashion. Symptomatic patients were those presenting with symptoms but without rupture, as defined by the VQI. Intraoperative iliac access procedures included all staged or concomitant iliac angioplasty, stent, bypass, or temporary conduit used to facilitate delivery of the endovascular device, as defined by the VQI. Thirty-day and long-term mortality were deduced using linkage between the VQI and the Social Security Death Index. In-hospital postoperative complications were recorded per the VQI registry and included new stroke, myocardial infarction, pulmonary complication including pneumonia or reintubation, temporary or permanent renal replacement therapy, intestinal ischemia, any postoperative blood transfusion, and spinal cord ischemia. A composite variable for major adverse events was created which includes 30-day death, stroke, temporary or permanent new renal replacement therapy, and spinal cord ischemia. Of note, as there were missing data for the spinal cord ischemia variables, patients with missing data were treated as if they did not have the complication for the composite endpoint.

Statistical Analysis

Categorical variables were presented as percentages. Continuous, non-normally distributed variables were presented as median (interquartile range [IQR]). Differences between cohorts were assessed using the Fisher’s exact test for categorical variables and the Mann-Whitney U test for continuous variables, where appropriate. Logistic and Cox hazards regression modeling were utilized to assess the independent association between female sex and 30-day and long-term mortality, respectively. Purposeful selection was used to identify covariates for inclusion in our multivariable models. This method includes variables identified on univariate analysis with P < .1 for each endpoint of interest and clinically relevant factors shown to be predictive of adverse events in previous studies.¹⁷ Kaplan-Meier survival estimates were stratified by sex and compared using the Log Rank test. All tests were conducted two-sided, and a P-value of less than 0.05 was considered significant. Statistical analysis was conducted using STATA version 14.1 (StataCorp LP, College Station, TX).

Results

Demographics

We identified 2,574 patients who underwent TEVAR, of which 40% (1,038) were female. Women were slightly older than men (73 vs. 72 years, P = .03) and were less often white (75% vs. 81%, P < .001) (Table I). Aortic diameter was smaller in female patients (5.8 cm vs. 6.0 cm, P = .02), but after accounting for body size, females had larger aortic size indices (3.2 cm/m² vs. 2.8 cm/m², P < .001). Women were less likely to have known coronary artery disease (19% vs. 26%, P < .001) or to have had prior coronary artery bypass grafting or percutaneous coronary intervention (15% vs. 31%, P < .001). However, female patients were more likely to have chronic obstructive pulmonary disease (33% vs. 28%, P = .01) and preoperative anemia (20% vs. 12%, P < .001). Women were less likely to have had any prior aortic intervention (23% vs. 30%, P < .001). Female patients were also less likely to be fully independent at baseline (88% vs. 93%, P < .001). There was no difference in rate of connective tissue disease, renal insufficiency, or diabetes (Table I).

Table I.

Demographics and preoperative comorbidities of patients undergoing TEVAR % or median (IQR)

	Women	Men	P-value
	N = 1,038	N = 1,536
Age, years	73 (67–79)	72 (64–78)	.03
White Race	75%	81%	< .001
Body Mass Index, kg/m2	26.4 (22.8–30.5)	26.8 (23.9–30.4)	.02
Maximum Aortic Diameter, cm	5.8 (5.0–6.6)	6.0 (5.1–6.8)	.02
Aortic Size Index, cm/m2	3.2 (2.6–3.9)	2.8 (2.2–3.3)	< .001
Connective Tissue Disorder	1.3%	1.4%	1.0
Prior Aortic Surgery	23%	30%	< .001
Fully Independent	88%	93%	< .001
Prior Stroke	11%	11%	.80
Coronary Artery Disease	19%	26%	< .001
Prior Coronary Intervention	15%	31%	< .001
Congestive Heart Failure	11%	13%	.11
Hypertension	89%	88%	.20
COPD	33%	28%	.01
Ever Smoker	74%	81%	< .001
Renal Insufficiency	40%	37%	.15
Anemia	20%	12%	< .001
Diabetes	17%	18%	.49

IQR = Interquartile Range; COPD = Chronic Obstructive Pulmonary Disease

Operative Details

Female patients were more likely to present symptomatically (16% vs. 10%, P < .001), and subsequently were less likely to have an elective procedure (79% vs. 85%, P < .001) (Table II). Intraoperatively, female patients were more likely to undergo iliac access procedures (4.3% vs. 2.1%, P < .01). Of the types of iliac access performed, female patients were slightly more likely to undergo aorto- or ilio-femoral bypass (1.2% vs. 0.5%, P = .06) and temporary conduit placement (1.0% vs. 0.3%, P = .03), but rates of percutaneous transluminal angioplasty/stent and stentgraft placement were similar. Arm/neck access procedures and left subclavian interventions were performed at similar rates. The proximal and distal extent of disease were both slightly more distal in female patients (Table II; Proximal: P = .10; Distal: P = .04), but this analysis was limited by 70% missing data. Female patients were more likely to have an estimated blood loss > 500 mL (20% vs. 17%, P = .04) and were also more likely to undergo intraoperative blood transfusion (29% vs. 21%, P < .001). While there was a trend towards reduced operative time in female patients (150 vs. 161 minutes, P = .08), this was not statistically significant. However, in the subset of patients who did not undergo iliac access procedures, females had significantly shorter operative times (146 vs. 160 minutes, P = .03). Female patients were slightly more likely to undergo conversion to open thoracic aortic repair (0.6% vs. 0.2%, P = .054), although this was not quite statistically significant and the rates in both groups were low.

Table II.

Operative characteristics and concomitant procedures by sex % or median (IQR)

	Women	Men	P-value
	N = 1,038	N = 1,536
Symptomatic	16%	10%	< .001
Elective Procedure	79%	85%	< .001
Operative Time, minutes	150 (93–240)	161 (99–245)	.08
If No Iliac Access Procedure
General Anesthesia	95%	95%	.72
EBL > 1L	20%	17%	.04
Any Intraoperative Blood Transfusion	29%	21%	< .001
Contrast Used, mL	100 (60–150)	94 (60–145)	.40
Proximal Extent of Diseasea			.10
Zone 0–1	6.2%	9.1%
Zone 2	16%	16%
Zone 3	38%	36%
Zone 4+	30%	25%
Distal Extent of Diseasea			.04
Zone 4	28%	34%
Zone 5+	72%	66%
Concomitant Procedures
Left Subclavian Intervention	9.9%	9.7%	.89
Iliac Access Procedure	4.3%	2.1%	< .01
PTA/Stent	1.3%	0.9%	.32
Stentgraft	1.1%	0.7%	.39
Aorto- or iliofemoral bypass	1.2%	0.5%	.06
Temporary conduit	1.0%	0.3%	.03
Arm/Neck Access	5.9%	5.1%	.43
Conversion to Open	0.6%	0.2%	.054

IQR = Interquartile Range; EBL = Estimated Blood Loss; PTA = Percutaneous Transluminal Angioplasty

^aLimited by 70% missing data

Outcomes

Female patients were more likely than male patients to have a major adverse event (7.6% vs. 4.3%, P < .001). Male and female patients had similar in-hospital postoperative rates of stroke, myocardial infarction, dysrhythmia, and pneumonia (Table III). However, female patients were more likely to be reintubated (6.6% vs. 4.7%, P = .04), to be treated with new postoperative renal replacement therapy (0.8% vs. 0.2%, P = .03), and to be transfused postoperatively (29% vs. 19%, P < .001). Female patients were much more likely to have spinal cord ischemia (4.0% vs. 1.0%, P < .01), although there was 65% missing data with this variable. Female patients were also more likely to have an access artery injury (2.9% vs. 1.2%, P < .01). Female patients had longer lengths of stay in both the intensive care unit (ICU) (2.5 vs. 2 days, P < .001) and the hospital (5 vs. 4 days, P < .001). Even amongst patients with no complications, female patients had higher lengths of stay (4 vs. 3 days, P < .001). Women were also more likely to be discharged to a skilled nursing facility (25% vs. 14%, P < .001).

Table III.

Unadjusted postoperative mortality and complications % or median (IQR)

	Women	Men	P-value
	N = 1,038	N = 1,536
Mortality
Thirty Day	5.4%	3.3%	<
One Year	12%	8.1%	<
In-Hospital Morbidity
Major Adverse Event	7.6%	4.3%	< .001
Stroke	0.9%	1.5%	.25
Myocardial Infarction	2.3%	2.0%	.68
Dysrhythmia	8.4%	7.6%	.5
Pneumonia	2.3%	1.9%	.48
Reintubation	6.6%	4.7%	.04
Intestinal Ischemiaa	1.3%	0.6%	.29
Temporary or Permanent Dialysis	0.8%	0.2%	.03
Postoperative Blood Transfusion (any)	29%	19%	< .001
Units Blood Transfused	0 (0–1)	0 (0–0)	< .001
Spinal Cord Ischemiaa	4.0%	1.0%	< .01
Access Artery Injury	2.9%	1.2%	< .01
ICU Length of Stay, days	2.5 (1–4)	2 (1–3)	< .001
Hospital Length of Stay, days	5 (3–8)	4 (2–7)	< .001
LOS in Patients with No Complications	4 (2–6)	3 (2–5)	< .001
Discharged to SNF	25%	14%	< .001

IQR = Interquartile Range; ICU = Intensive Care Unit; LOS = Length of Stay; SNF = Skilled Nursing Facility

^aLimited by 65% missing data

Unadjusted mortality was higher in female patients at 30 days (5.4% vs. 3.3%, P < .01) and one year (12% vs. 8.1%, Log-Rank P < .01) (Table III). After adjusting for age, aortic size index, symptom status, and comorbidities including statin use, female sex remained associated with 30-day mortality (Odds Ratio [OR] 1.8, 95% Confidence Interval [CI] 1.1 – 2.9, P = .01) (Table IV). Adding iliac access procedure to the model does not affect the relationship between sex and mortality (OR 1.8, 95% CI 1.1 – 2.8, P = .02). Kaplan-Meier estimates showed reduced survival in female patients compared to male patients (Figure 1), with an 88% survival in females compared to 92% survival in males at a median follow-up time of 21 months. Additionally, after Cox hazards modeling was used to adjust for age, aortic size index, symptom status, and comorbidities, female sex was still independently associated with mortality during follow-up (Hazard Ratio [HR] 1.3, 95% CI 1.01 – 1.8, P = .046) (Table V).

Table IV.

Logistic regression model for 30-day mortality

	Odds Ratio	95% Confidence Interval	P-Value
Female Sex	1.8	1.1 – 2.9	.01
Age, by decade	1.3	1.001 – 1.7	.049
Aortic Size Index	1.2	0.99 – 1.5	.06
Renal Insufficiency	1.8	1.1 – 2.8	.01
Anemia	1.6	0.9 – 2.9	.08

Other covariates include symptom status, coronary artery disease, chronic obstructive pulmonary disease, and statin use, which were not significantly associated with 30-day mortality.

Figure 1.

Kaplan-Meier survival estimates show reduced freedom from death in female compared to male patients (Log Rank P < .01).

Table V.

Cox regression for long-term mortality

	Hazard Ratio	95% Confidence Interval	P-Value
Female Sex	1.3	1.01 – 1.8	.046
Age, by decade	1.5	1.3– 1.8	< .001
Aortic Size Index	1.04	1.03 – 1.05	< .001
Coronary Artery Disease	1.5	1.1 – 2.0	.02
COPD	1.4	1.1 – 1.9	.02
Anemia	2.3	1.6 – 3.2	< .001

COPD = Chronic Obstructive Pulmonary Disease

Other covariates include symptom status, renal insufficiency, and statin use, which were not significantly associated with long-term mortality.

Discussion

We found that female patients have higher 30-day and long-term mortality compared to male patients following TEVAR for intact descending thoracic aortic aneurysms. These associations persisted after adjusting for differences in age, comorbidities, and symptom status, female patients. The reason for this is likely multifactorial, but possible explanations include more complicated aortic and/or access vessel anatomy, higher rates of symptomatic aneurysms, worse baseline health, or other social impacts on health status.

Women have been shown in multiple series to experience worse outcomes than men following open and endovascular repair of abdominal aortic aneurysms.^18–21 However, few studies directly evaluated sex differences following repair of thoracic aortic aneurysms. Sub-analyses from the prospective VALOR⁸ and Gore TAG⁹ device trials showed no difference in perioperative mortality between sexes. While the VALOR trial suggested a sex difference (2.5% vs. 1.7%),⁸ the Gore TAG trial showed no trend,⁹ although mortality in the latter population was very low at 1.2%, which has not been duplicated in subsequent studies. The population in the VQI, which consists of large academic medical centers, as well as smaller community hospitals, is likely more generalizable to the real world. The sex differences we witnessed may be due to the fact that our cohort included patients with more medical and anatomic complexity than those seen in clinical trials, where inclusion and exclusion criteria are stricter, and procedures are limited to centers with demonstrated clinical expertise.

Arnaoutakis et al. used NSQIP data from 2005–2011 to evaluate sex differences following TEVAR, and similar to our findings, found that female patients had higher mortality. In contrast, however, they did not find female sex to be independently predictive of mortality after adjusting for differences in age, comorbidities, use of iliac access procedures, and operative time. The prospective clinical trials of the early Gore and Medtronic devices had similar findings, noting both increased use of iliac conduits and higher rates of access-vessel injuries in female patients.^{8, 9} These findings have also been demonstrated in the abdominal aortic aneurysm literature, in which a national study of over 15,000 EVAR patients showed that females were significantly more likely to have an iliac conduit or direct iliac access (26% vs. 18%, P = .02), and iliac access, in turn, was associated with a nearly three-fold increase in perioperative mortality.²² Based on these results, we hypothesized that at least some of the sex differences could be explained by smaller, more diseased access vessels in female patients. While we also found higher rates of iliac conduit use (4.3% vs. 2.1%) and access artery injury (2.9% vs. 1.2%), adjusting for conduit use did not completely account for the mortality difference in our cohort. Notably, rates of conduit use and access artery injury were much lower in both male and female patients than in prior studies, which may represent the use of contemporary, lower-profile endografts as well as surgeons’ experience. However, we do not have information in our dataset regarding access vessel size or the type of endograft used to evaluate this further. We did find that female patients had higher rates of spinal cord ischemia, although conclusions from this are limited given the high percentage of missing data. If all patients with missing data were assumed to have no spinal cord ischemia, female patients were still more likely to develop this complication (1.5% vs. 0.3%, P < .01), but further study of this should be done to determine if this discrepancy does in fact exist, and if it is related to increased degree of aortic coverage.

In addition to discrepancies in anatomy, female patients in this series were more likely to be symptomatic at presentation. This has been replicated in patients with abdominal aortic aneurysms, as women are more likely to present with symptoms and/or rupture.²⁰ Furthermore, we have seen in patients with AAA that patients with intact but symptomatic aneurysms have worse outcomes than patients undergoing elective repair.²³ However, in our analysis, symptom status did not predict perioperative or long-term mortality, and even after adjusting for this, female sex was still predictive of mortality. Therefore, severity at presentation alone is not the only explanation for the mortality differences we identified.

An alternative explanation for the mortality difference we identified may be the older age and increased medical complexity of the female patients undergoing TEVAR. We found that women were older and more often had chronic obstructive pulmonary disease. Additionally, as noted in the thoracic and abdominal aortic aneurysm literature,^{9, 24} female patients were less likely to have coronary artery disease or prior coronary intervention. This may represent less coronary disease burden, but could also mean that female patients have more undiagnosed disease, leading to increased perioperative cardiac complications. Notably, however, there were no differences in the postoperative rates of myocardial infarction or dysrhythmia in our series. Despite the worse 30-day and long-term mortality in women even after adjusting for age and comorbidity differences, there is likely some degree of unaccounted-for medical complexity or frailty that is contributing to this disparity.

We also found that female patients were less likely to be functionally independent preoperatively and had longer hospital and ICU lengths of stay, with increased discharge to a skilled nursing facility. Other studies of TEVAR showed similarly prolonged hospital lengths of stay in female patients.^{8, 9, 13} While the increased time in hospital could be related to higher rates of complications, length of stay was longer even in patients with no postoperative complications. This, in conjunction with the higher rates of discharge to skilled nursing facilities, may be related to disparities in social factors, with more females perhaps playing the role of caregiver within their family, without a caregiver able to provide for them in return. These social factors, which are not included in the recorded variables, may also contribute to the mortality disparities seen in this study.

The decision to repair an aortic aneurysm weighs the rupture risk, life expectancy, and the morbidity and mortality risk associated with operative repair. Therefore, if female patients experience higher perioperative mortality, this should factor into the operative decision making. Different aneurysm size thresholds by sex have been suggested in abdominal aortic aneurysms, given the higher rupture risk at smaller diameters in female patients.^{25, 26} Additionally, female patients in our study also had higher rates of COPD, which is also known to correlate with rupture risk.²⁷ We have no data from this study regarding risk of rupture, but the operative mortality in female patients remained higher even after adjusting for aortic diameter and aortic size index. Further research is needed in order to determine the ideal threshold for repair, either by diameter or aortic size index, weighing the reduced life expectancy and increased operative burden with the potential rupture risk.

These data must be interpreted in the context of the study design. Clinical registries often have incomplete data and limited variable definitions. In particular, long-term follow-up information in the VQI is limited, so we are unable to adequately account for late endoleak and reintervention or determine if sex differences in late reinterventions contribute to the higher late mortality in female patients. Additionally, only in-hospital perioperative outcomes are recorded, so readmissions are not captured and other postoperative complications may be underreported, especially following an operation with a median hospital stay less than one week. While the VQI does provide more granular data than other non-vascular-specific registries, it does not include information regarding presence of aortic thrombus, access vessel size, and certain concomitant procedures, such as multi-stage hybrid repairs. However, many of the variables we can measure can be used as surrogate markers for these findings, including the use of iliac access procedures for access vessel size. Additionally, to fully understand disparities, social factors must be considered, and these variables are difficult to quantify or even capture in any registry. Again, patient functional status and discharge status can be used as markers for these findings. Further study of sex differences following TEVAR are warranted to mitigate these disparities as best as possible.

Conclusion

Female patients have higher mortality than male patients following TEVAR in a real8 world population, even after adjusting for their older age and increased medical comorbidities. Further research into the etiology of these sex differences will allow us to determine the optimal threshold for repair of thoracic aortic aneurysms in female patients.