size thresholds for repair of abdominal aortic aneurysms warrant reconsideration

Abstract

Background:

The historical size threshold for abdominal aortic aneurysm (AAA) repair is widely accepted to be 5.5 cm for men and 5.0 cm for women. However, contemporary AAA rupture risks may be lower than historical benchmarks, which has implications for when AAAs should be repaired. Our objective was to use contemporary AAA rupture rates to inform optimal size thresholds for repair.

Methods:

We used Markov-chain analysis to estimate life expectancy for AAA patients. The primary outcome was AAA-related mortality. We estimated survival using Social Security Administration life tables and published contemporary AAA rupture estimates. For those undergoing repair, we modified survival estimates using data from the Vascular Quality Initiative (VQI) and Medicare on complications, late-rupture, and open conversion. We used this model to estimate the AAA repair size threshold that minimizes AAA-related mortality for 60-year-old average health men and women. We performed a sensitivity analysis of poor-health patients, and 70- and 80-year-old base cases.

Results:

The annual risk of all-cause mortality under surveillance for a 60-year-old woman presenting with a 5.0cm AAA, using repair thresholds of 5.5cm, 6.0cm, 6.5cm, and 7.0cm, was 1.7%, 2.3%, 2.7%, and 2.8%, respectively. The corresponding risk for a man was 2.3%, 2.9%, 3.3%, and 3.4% for the same repair thresholds. For a 60-year-old average health woman, an AAA repair size of 6.1 cm was the optimal threshold to minimize AAA-related mortality. Life expectancy varied by <2 months for repair at sizes from 5.7 cm to 7.1 cm. For a 60-year-old average health man, an AAA repair size of 6.9 cm was the optimal threshold to minimize AAA-related mortality. Life expectancy varied by <2 months for repair at sizes from 6.0 cm to 7.4 cm. Women in poor-health, at various age strata, all had an optimal AAA repair size thresholds that were >6.5 cm, while poor-health men, at all ages, had an optimal repair size thresholds that were >8.0 cm.

Conclusions:

The optimal threshold for AAA repair is more nuanced than a discrete size. Specifically, there appears to be a range of AAA sizes, for which repair is reasonable, to maximize survival. Notably, they are all are greater than current guideline recommendations. These findings would suggest that contemporary AAA size thresholds for repair should be reconsidered.

Table of Contents Summary

Based on contemporary data and rupture rates, the optimal AAA repair size thresholds to minimize AAA-related mortality were 6.1cm for a 60-year-old average health woman, and 6.9 cm for a 60-year-old average health man. These size thresholds may be higher than currently recommended by societal guidelines and should be reconsidered.

Introduction:

Contemporary abdominal aortic aneurysm (AAA) practice remains predicated on the widely accepted historical 5.5cm size threshold for surgical repair.^1-6 In fact, this diameter threshold is currently endorsed by both the Society for Vascular Surgery (SVS) and European Society for Vascular Surgery (ESVS), whose guidelines cite a class 1 A recommendation.^{7, 8}

However, closer scrutiny reveals an important knowledge gap. The UK Small Aneurysm Trial (UKSAT) and the Aneurysm Detection and Management (ADAM) study, which are often cited as the origin of the 5.5 cm threshold, never actually demonstrated a benefit for repairing AAAs at 5.5 cm.^{1, 2} Rather, these landmark trials were designed to test whether repairing AAAs <5.5 cm offered a mortality benefit over surveillance.^{1, 2} In actuality, the 5.5 cm threshold remains based on historical rupture rates and autopsy studies, rather than large randomized trials.^{1, 2} To date, there surprisingly remains no level 1 evidence that documents a mortality benefit for the repair of AAAs using the size threshold ≥5.5 cm versus surveillance.⁹ Interestingly, the SVS has recently updated its endorsed estimates of AAA rupture risk to now reflect lower rates than have been historically accepted.¹⁰ Accordingly, this change has real world implications for current practice, as they may impact the perceived benefit of surgery.^{7, 8, 10} Moreover, these changes may rightfully impact the optimal size threshold at which AAA repair should be considered.

The objective of this study was to define the optimal size threshold for AAA repair in current practice, informed by SVS-endorsed rupture rates and the available literature.

Methods:

Human subjects protection

The University of Florida Institutional Review Board deemed this study exempt. All data were de-identified prior to publication, and thus the need for consent was waived.

Model construction

We created a Markov chain analysis for patients who present with an AAA and enter surveillance. A patient who enters the model can then pass through a series of health states with probabilities of each, based on their respective age, sex, health risk, and size threshold for AAA repair. Each possible outcome carries a median life expectancy of all patients who experience that outcome, which allows the calculation of average life expectancy for all patients entering the model with a given age, sex, and risk status. We can then vary the AAA repair threshold to determine the size at which survival is maximized.

We used AAA-related mortality as the primary outcome. We defined this as death due to AAA rupture, surgery, or postoperative mortality. We specifically did not use all-cause mortality as our primary outcome since this introduces an unknown amount of bias in the potential benefit of the treatment because the Markov model is unable to separate the competing risk of non-AAA related mortality from AAA-related mortality. To account for this, in the primary analysis, we assumed that patients survived until an AAA-related mortality event occurred, which should provide a conservative estimate (i.e., favoring AAA repair) of the potential benefits of surgery. We considered all-cause mortality in a sensitivity analysis, knowing that this would be limited by the competing-risk challenge listed above.

Cohorts of patients went through the Markov chain through various potential health states (Figure S1). The simulated patients progressed in 1-month cycles. During each cycle all patients were exposed to the potential risks available to them for that respective health state, including AAA rupture, surgery, and mortality. The cycles continued until all patients had reached the primary outcome. Once this had occurred, the mean survival for all patients in that simulation was calculated.

We defined male and female as sex at birth because that is how the datapoint is collected by the VQI registry. We defined poor health as a revised cardiac risk index of >0 for patients who underwent endovascular AAA repair (EVAR) and >1 for patients who underwent open AAA repair.¹¹

Model inputs and assumptions

Survival and AAA natural history

Baseline survival.

We calculated survival for AAA patients using two primary components. Baseline survival was first taken from the 2020 Social Security Administration (SSA) life tables which estimate life expectancy at birth to be 74.12 years for males and 79.78 years for females, with associated probabilities of death that advance with age (Table 1).¹² To account for the added risk of mortality for patients living with AAA disease, we modified this by adding an absolute 3% annual risk of mortality to the life tables.¹³ We further modified baseline survival for patients in poor health by increasing the annual risk of mortality by a relative 50% based on survival estimated directly from the registry data.

Table 1:

Summary of assumptions and inputs for the model

Input parameter	Base case value	Sensitivity analyses	Reference / source
AAA annual rupture rate
<5.4 cm	0.008	0.016	10
5.5-6.0 cm	0.038	0.054
6.1-7.0 cm	0.048	0.064
>7.0 cm	0.0655	0.079
Rupture mortality outside of hospital or before being offered surgery	0.55		17
AAA expansion rate	10 (10% enlargement per year)		13-16
Probability AAA not treatable with EVAR
4.0-4.9 cm	0.02		17
5.0-5.9 cm	0.04
6.0-6.9 cm	0.06
>=7.0 cm	0.2
Probability of death due to rupture	0.55
EVAR and open AAA repair in hospital mortality and complications	See table 2		VQI primary data
Probability of conversion to open
30 days
Men - elective	0.0066		17
Men - rupture	0.013
Women - elective	0.011
Women - rupture	0.021
31 days - 1 year
Men - elective	0.0088		17
Men - rupture	0.018
Women - elective	0.014
Women - rupture	0.021
Excess mortality due to AAA disease	0.003		13
Non-AAA related mortality
Men - Average health		Men - Poor health
60-69	0.016	0.024	SSA life tables
70-79	0.035	0.0525
>=80	0.095	0.1425
Women - Average health		Women - Poor health
60-69	0.0096	0.014	SSA life tables
70-79	0.024	0.036
>=80	0.073	0.1095
Post elective surgery mortality without complications	HR 1.5		VQI primary data
Post elective surgery, with complications	HR 3.7		VQI primary data
Post rupture surgery, without complications	HR 1.9		VQI primary data
Post rupture surgery, with complications	HR 3.23		VQI primary data

Legend: EVAR, endovascular aneurysm repair; AAA, abdominal aortic aneurysm; VQI, vascular quality initiative; SSA, Social Security Administration

AAA expansion.

The rate of AAA expansion over time is heterogenous.¹⁴ The UK Small Aneurysm trial estimated an approximate annual AAA expansion rate of 5% of the AAA size per year, but this is impacted by other risk factors and co-morbidities.^{15, 16} To provide a conservative model of AAA enlargement (i.e., expanding relatively fast), we estimated an annual 10% of current AAA size linear enlargement rate for all patients in the Markov model (Table 1).¹³

AAA rupture.

In 2021 the SVS published AAA implementation guidelines with revised risks for rupture based on contemporary literature.¹⁰ For our primary models, we used the midpoint of the published range for AAA rupture for each AAA size strata (Table 1). In our sensitivity analyses, we used the maximum estimated AAA rupture risk. We assumed rupture without surgery to be a mortal event. We assumed that patients who experienced rupture had a 55% chance of dying either outside of the hospital or before being offered surgery.¹⁷

AAA repair

Candidacy for endovascular versus open surgery.

We assumed an EVAR-first approach to AAA repair. We held this assumption as it minimized perioperative mortality and complications, while providing similar long-term survival.^{18, 19} We then estimated the probability that a patient presenting for AAA repair would not be an anatomic candidate for EVAR, and accordingly undergo open AAA repair, based on prior work (Table 1).¹⁷ Patients presenting with rupture were assumed to have the same likelihood of candidacy for EVAR.

Perioperative mortality and complications.

Perioperative mortality and complications were derived from primary VQI data on patients undergoing AAA repair (Table 2). We calculated the probability of in-hospital death and in-hospital permanent complications for patients presenting for elective repair and with rupture, respectively. We defined permanent complications as in-hospital stroke or renal failure requiring dialysis. We calculated these probabilities for each age, sex, and health status strata and for EVAR and open AAA repair.

Table 2:

VQI-derived probability estimates of in hospital death and complications after AAA surgery

				Elective Surgery		Surgery for rupture
Gender	Age (years)	Health class	Procedure	in-hospital death	Permanent comorbidity	In-hospital death	Permanent comorbidity
Female	60	Average	EVAR	0.001	0.002	0.13	0.07
Female	60	Poor	EVAR	0.002	0.004	0.139	0.094
Female	70	Average	EVAR	0.002	0.003	0.187	0.07
Female	70	Poor	EVAR	0.004	0.005	0.198	0.093
Female	80	Average	EVAR	0.006	0.004	0.304	0.069
Female	80	Poor	EVAR	0.011	0.008	0.319	0.092
Male	60	Average	EVAR	0.001	0.002	0.117	0.092
Male	60	Poor	EVAR	0.002	0.004	0.124	0.121
Male	70	Average	EVAR	0.002	0.003	0.169	0.091
Male	70	Poor	EVAR	0.004	0.005	0.179	0.12
Male	80	Average	EVAR	0.006	0.004	0.279	0.09
Male	80	Poor	EVAR	0.01	0.007	0.292	0.119
Female	60	Average	OAAA	0.014	0.024	0.252	0.173
Female	60	Poor	OAAA	0.024	0.04	0.285	0.242
Female	70	Average	OAAA	0.028	0.033	0.36	0.165
Female	70	Poor	OAAA	0.048	0.054	0.399	0.231
Female	80	Average	OAAA	0.077	0.051	0.548	0.153
Female	80	Poor	OAAA	0.128	0.083	0.589	0.216
Male	60	Average	OAAA	0.01	0.022	0.194	0.171
Male	60	Poor	OAAA	0.017	0.037	0.221	0.24
Male	70	Average	OAAA	0.02	0.03	0.286	0.163
Male	70	Poor	OAAA	0.034	0.05	0.321	0.229
Male	80	Average	OAAA	0.055	0.047	0.463	0.152
Male	80	Poor	OAAA	0.093	0.077	0.505	0.214

Legend: EVAR, endovascular aneurysm repair; OAAA, open abdominal aortic aneurysm repair.

Conversion to open.

We used the risk of conversion to open from EVAR based on prior work (Table 1).¹⁷ We used the probabilities of conversion at 30 days and at 1 year, for men and women, undergoing elective repair and repair for rupture, respectively. Patients who underwent conversion to open AAA were then subject to the risks associated with open AAA repair based on the respective age, sex, and risk strata (Table 2).

Postoperative survival.

We calculated postoperative survival using a combination of the SSA life tables modified for AAA disease, and the primary VQI data (Table 1). We estimated the survival of average health patients who underwent elective EVAR and did not suffer any permanent complications using the R software package “survival”. We then calculated the hazard ratio (HR) for patients who underwent elective surgery in poor health, and for those who suffered a permanent complication, respectively. We then used the HRs as multipliers to the baseline SSA mortality risks. We repeated this process for patients presenting with a ruptured AAA.

Statistical analysis

To estimate the optimal size threshold for elective AAA repair, we calculated AAA-related mortality associated with AAA repair and surveillance for every 1 mm interval of enlarging AAA diameter, beginning with a 5.0 cm aneurysm. The threshold that minimized AAA-related mortality was considered the optimal threshold. We additionally calculated the range of thresholds over which life expectancy varied by less than 2 months. We performed this for patients in average health and in poor health. We repeated this for simulated base cases that included men aged 60, 70, and 80, presenting with 5.0 cm AAAs, and women aged 60, 70, and 80, presenting with 5.0 cm AAAs. Optimal AAA repair size thresholds and the respective ranges over which life expectancy varied by less than 2 months in either direction were calculated for each respective group. All analyses were performed using the R statistical software package (V.4.1.2, The R Foundation for Statistical Computing, Vienna, Austria).

Results:

Optimal AAA size threshold for repair to minimize AAA-related mortality

The optimal size threshold to minimize AAA-related mortality for a 60-year-old average health woman presenting with a 5.0 cm AAA was 6.1 cm (Figure 1). Life expectancy varied by <2 months for repair at sizes from 5.7 cm to 7.1 cm. The optimal repair size thresholds increased in parallel with age and were 7.0 cm for a 70-year-old woman in average health, and >8.5 cm for an 80-year-old woman in average health (Figure 1, Figure S2).

Figure 1:

Life expectancy and optimal repair size thresholds for (A) 60-year-old women in average or poor health, respectively, with a 5.0 cm abdominal aortic aneurysm (B) forest plot across age strata tested (median assumed risk of rupture).

Women presenting in poor health had higher thresholds than those with average health. The optimal repair size threshold for a 60-year-old woman in poor health was 6.9 cm, and the range over which life expectancy varied by <2 months was 6.2 to 9.0 cm (Figure 1). Optimal size thresholds for 70- and 80-year-old women in poor health were all above 7.0 cm (Figure 1, Figure S2).

The optimal size threshold to minimize AAA related mortality for a 60-year-old average health man presenting with a 5.5 cm AAA was 6.9 cm (Figure 2). Life expectancy varied by <2 months for repair at sizes from 6.0 cm to 7.4 cm. The optimal repair size thresholds increased with age and were 6.9 cm for a 70-year-old man in average health, and >8.5 cm for an 80-year-old woman in average health (Figure 2, Figure S3).

Figure 2:

Life expectancy and optimal repair size thresholds for (A) 60-year-old men in average or poor health, respectively, with a 5.5 cm abdominal aortic aneurysm (B) forest plot across age strata tested (median assumed risk of rupture).

Men in poor health had increased size thresholds compared to those with average health. The optimal repair size threshold for a 60-year-old man in poor health was 8.3 cm, and the range over which life expectancy varied by <2 months was 6.9 to >8.5 cm (Figure 1). Optimal size thresholds for 70- and 80-year-old men in poor health were all above 8.5 cm (Figure 2, Figure S3).

Sensitivity analysis – maximum AAA rupture risk

Using the maximum possible AAA rupture risk, based on the SVS implementation guidelines, lowered the optimal size repair threshold for all base cases examined. Among average health women, the size threshold that minimized AAA related mortality was 5.4 cm for a 60-year-old woman, 5.8 cm for a 70-year-old woman and 7.0 cm for an 80-year-old woman, respectively (Figure S4). Optimal repair size thresholds for women in poor health were all above 6.0 cm (Figure S4).

Average health men had similar results. The repair size threshold that minimized AAA related mortality for an average health 60-year-old man was 6.0 cm. Notably, it was 6.9 cm for a 70- year-old man, and >8.5 cm for an 80-year-old man, respectively (Figure S5). Optimal repair size thresholds for men in poor health were all ≥7.0 cm (Figure S5).

Sensitivity analysis – all-cause mortality

The annual risk of all-cause mortality under surveillance for a 60-year-old woman presenting with a 5.0cm AAA using repair thresholds of 5.5cm, 6.0cm, 6.5cm, and 7.0cm, was 1.7%, 2.3%, 2.7%, and 2.8%, respectively. The corresponding risk for a man was 2.3%, 2.9%, 3.3%, and 3.4% across similar size thresholds.

The cumulative probability (not annualized) of non-AAA related mortality under surveillance for a 60-year-old woman presenting with a 5.0 cm AAA using repair thresholds of 5.5 cm, 6.0 cm, 6.5 cm, and 7.0 cm, was 1%, 1.8%, 2.6%, and 3.3% respectively (Figure 1). The optimal size threshold for repair, using all-cause mortality as the primary outcome, was 5.8 cm for a 60-year-old woman in average health, and 6.2 cm for those in poor health (Figure S6, Figure S7).

The cumulative probability (not annualized) of non-AAA related mortality under surveillance for a 60-year-old man presenting with a 5.5 cm AAA, using repair thresholds of 6.0 cm, 6.5 cm, 7.0 cm, and 7.5 cm, was 1.5%, 2.8%, 4%, and 5.1% respectively (Figure 2). The optimal size threshold for repair using all-cause mortality as the primary outcome was 6.0 cm for a 60-year-old man in average health, and 6.8 cm for those in poor health (Figure S6, Figure S7).

Discussion:

This study is among the first to use contemporary SVS-endorsed AAA rupture rates in combination with real-world data from the VQI to inform size thresholds for elective AAA repair. We found that the optimal AAA repair size threshold to minimize AAA-related mortality was 6.1 cm for a 60-year-old woman, and 6.9 cm for a 60-year-old man. Notably, these thresholds increased in parallel with advancing age. Furthermore, patients in poor health had repair size thresholds in excess of 6.5 cm for women and 7.0 cm for men. Moreover, in both the primary analysis, and the sensitivity analyses, all identified repair size thresholds were greater than the historically and widely accepted 5.5 cm diameter for men and 5.0 cm diameter for women. These findings would suggest that potentially more AAA patients are undergoing surgical repair than are necessary to minimize AAA-related mortality in current practice. This finding in particular has real world implications regarding contemporary AAA practice and may suggest that current aneurysm size thresholds for repair should be revisited or revised.

Surprisingly, there appears to be a widespread general misinterpretation of the historical data surrounding the widely held 5.5 cm size threshold. Interestingly, both the SVS, and ESVS, currently assign the 5.5 cm size threshold a Class 1 recommendation, supported by Level A evidence, the highest possible level.^{8, 10} However, this designation may in fact be unintentionally misleading. The Level A evidence (i.e., randomized trials) upon which these guidelines are based, are the ADAM study and UKSAT.^{1, 2} These landmark trials tested surveillance versus repair of small AAAs, <5.5 cm, and concluded that there was no mortality benefit for repair of small AAAs, findings which were later confirmed by testing early EVAR versus surveillance.^{1, 2, 20, 21} Based on these conclusions, there was an untested presumption that AAA >5.5 should therefore be repaired. Yet, the question of whether surgery offers a benefit over surveillance for aneurysms >5.5 cm, (or any other arbitrary threshold), has never been tested in average risk patients. Specifically, the only trial to date which has tested surveillance versus repair of aneurysms >5.5 cm was the Endovascular Repair of AAAs in Patients Physically Ineligible for Repair Trial (EVAR2), which demonstrated no benefit for surgery in this patient population.¹⁸ Therefore, it is somewhat surprising that 5.5 cm has been perpetuated as the de facto international threshold for AAA repair.

Accordingly, in the absence of high-quality evidence, prior investigators have created Markov models in an attempt to define the optimal size threshold for AAA repair.^{13, 22} The work reported herein, adds to this knowledge gap in several ways. First, our model incorporates the recently updated SVS-endorsed AAA rupture rates, which have decreased over time.¹⁰ It is important to emphasize that the assumed rate of AAA rupture is inversely correlated with the optimal size threshold for repair – when rupture risks are lower, the size when an AAA should be repaired is correspondingly increased. Moreover, advances in medical therapy, coupled with decreased observational rupture rates necessitate a new model to reflect contemporary population data. Second, prior reports have relied primarily on the older historical published literature for rates of complications, mortality, and mode of repair. By comparison, this current analysis used primary data from the VQI which permits for a more current granular patient-level estimation of the disease states included in the Markov decision model. Finally, prior investigators have used all-cause mortality as their primary endpoint.^{13, 22} The Markov model is unable to distinguish between the competing risks of AAA vs non-AAA related mortality. To address this, we used AAA-related mortality as our primary outcome (i.e., the event AAA repair is designed to prevent), and patients were assumed to live either until a given size threshold for repair was reached, or rupture had occurred. Since patients in this model were not exposed to the competing risk of non-AAA related mortality, this should amplify the utility of AAA repair. Using this methodology, our findings that the optimal size threshold for repair was more than 1 cm greater than current guidelines recommend would suggest that the benefit of repair at 5.0 cm or 5.5 cm may in fact be exaggerated.

Findings among advanced age patients or those with significant co-morbidities were more pronounced and highlight the significant implications of the study’s findings. Specifically, both men and women who presented at 70 or 80 years of age had corresponding optimal AAA size thresholds for repair of approximately 7.0 cm. Furthermore, co-morbid patients had projected optimal repair size thresholds that were above 7.0 cm for women and 8.0 cm for men. Many patients in contemporary practice with AAA disease fall into one of these 2 groups which signifies an opportunity to potentially refine AAA decision making when considering surgery. More specifically, our results would suggest that patients over 70 years of age, or in poor health, may likely derive a mortality benefit from an extended surveillance strategy compared to historical paradigms.

These results ultimately raise the question as to whether a significant cohort of patients are undergoing AAA repair prematurely, or even unnecessarily. As medical and surgical therapy has advanced, so too has its corresponding cost, particularly in the setting of AAA repair.^23-27 Safely decreasing the number of potentially unnecessary procedures could substantially mitigate this significant financial footprint. In addition, while EVAR is touted as having low perioperative rates of morbidity and mortality, it is not without risk entirely. Approximately one-third of patients can anticipate a reintervention within 10 years of their index EVAR.²⁸ Furthermore, implanted devices remain at risk of failure, and at-risk devices may not be recognized until years after reaching the healthcare market.²⁹ Given these implications for both patients and other stakeholders, additional evidence to inform optimal AAA care appears warranted. It would seem that a randomized trial to determine the true optimal repair threshold may provide current value given the noted advances in both medical and surgical therapy, as well as the fiscal sequelae imposed on our hospitals and health systems alike.

There are some important considerations that must be discussed for this Markov based analysis. Markov chain results are entirely dependent on the accuracy of the event probabilities entered into the chain and are thus subject to the inherent limitations of the studies used as sources for those probabilities. The VQI EVAR and open aneurysm registries include data on thousands of patients, and we believe they have provided us with robust age-, gender- and health-specific estimates of the probabilities of events for patients undergoing AAA surgery. However, available age- or gender-specific data on AAA rupture and expansion have low sample sizes and may be inaccurate for large groups of patients.

The Markov model is designed to test life expectancy for a population of patients, on average. This cannot capture individual nuances of each patient encounter. There are a variety of factors that may play into the decision to pursue AAA repair, that simply cannot be accounted for in this type of study. These include patient preferences, anxiety surrounding surveillance versus surgery, changes in these preferences, and patient knowledge base and education, among others.³⁰ There may be situations where repair is rightfully undertaken at lower AAA diameters, but we cannot comment on these individualized decisions with our results. Larger AAA sizes may be associated with a loss of candidacy for EVAR. While we did account for patients’ candidacy for EVAR at the time of repair or rupture, we are not able to account for the interaction between AAA growth due to surveillance, and the loss of candidacy for EVAR over time. However, because our base cases were 60-year-old average health men and women, and the similar long-term outcomes of both open and EVAR in standard risk patients, we feel that this is unlikely to affect our primary analyses described herein.^{18, 19}

Sex- or gender-specific data on AAA rupture and expansion have low sample sizes and variable values.³¹ We tried to account for this by using sex-specific VQI estimates, using a higher rate of expansion that estimated in available studies, and by performing a sensitivity analysis using the maximum estimated SVS-endorsed AAA rupture potential.^{10, 31} However, some assumptions were required for the model, which may mean that our results may be less generalizable to women. Markov models are based on the available data and are thus subject to the inherent limitations of the included studies. We did our best to address this by testing a range of AAA rupture rates and using contemporary real world VQI data where possible. In addition, there is likely a distribution of potential AAA expansion rates, which, if known, could have been used to expand our study into a Monte Carlo simulation, thus improving realism and allowing estimation of confidence intervals for estimated life expectancy under various treatment scenarios.. Based on our review of the literature, we selected a 10% annual size increase for our model. We believed this was a conservative estimate (i.e., fast expansion rate) based on the currently available data. However, we are unable to fully characterize or account for real world heterogeneity in AAA wall biology which may not adhere to this assumption. Furthermore, we are not able to account for patients who had rapid AAA expansion, which may be an indication for urgent surgery. In addition, Markov models can be limited for short time horizons, which occurred in some of our poor health and aged base case patients, and do not account for covariance of characteristics within patients. Finally, results will differ when considering AAA-related versus non-AAA related mortality. As noted previously, we believe that AAA-related mortality provides a more accurate characterization of the potential benefit derived from operative AAA repair. Nevertheless, to address this point, we conducted additional sensitivity analyses using all-cause mortality for completeness. Registry data is subject to recall bias and missing data, although the missing data for our analyses was <2%, and the VQI now includes over 1000 centers, results may not be generalized to non-VQI centers.³²

Conclusions:

In conclusion, the optimal size threshold for AAA repair is likely more nuanced than a discrete “one size fits all” approach. However, it should be noted that our Markov model would suggest that the optimal repair size to minimize AAA-related mortality is greater than 6.1 cm for women, and 6.9 cm for men. Moreover, projected optimal size thresholds for AAA repair were additionally increased among older, co-morbid patients. These contemporary findings would suggest that the surgically derived benefit for AAA repair at the historical widely held sizes of 5.0 cm or 5.5 cm may be exaggerated in current practice. This study in conjunction with a lack of level I evidence to substantiate the optimal size threshold for AAA surgical repair, highlight the potential need for a randomized trial or additional analyses leveraging the power of large datasets to further inform aneurysm care delivery.

Article Highlights.

Type of Research:

Markov-chain analysis.

Key Findings:

For a 60-year-old average health woman, an AAA repair size of 6.1 cm was the optimal threshold to minimize AAA-related mortality. For a 60-year-old average health man, the optimal AAA repair size was 6.9 cm. Optimal size thresholds increased with age and were higher for patients in poor health.

Take Home Message:

Based on contemporary data and rupture rates, the optimal AAA repair size thresholds to minimize AAA-related mortality may be higher than are currently recommended by societal guidelines and should be reconsidered.