Persistent High Long-term Excess Mortality After Elective AAA Repair Especially in Women

Objective: The aim of this time-trend analysis is to estimate long-term excess mortality and associated cardiovascular risk for abdominal aortic aneurysm (AAA) patients after elective repair while addressing the changes in AAA management and patient selection over time. Background: Despite the intensification of endovascular aneurysm repair and cardiovascular risk management, Swedish population data suggest that AAA patients retain a persistently high long-term mortality after elective repair. The question is whether this reflects suboptimal treatment, a changing patient population over time, or a national phenomenon. Methods: Nationwide time-trend analysis including 40,730 patients (87% men) following elective AAA repair between 1995 and 2017. Three timeframes were compared, each reflecting changes in the use of endovascular aneurysm repair and intensification of cardiovascular risk management. Relative survival analyses were used to estimate disease-specific excess mortality. Competing risk of death analysis evaluated the risk of cardiovascular versus noncardiovascular death. Sensitivity analysis evaluated the impact of changes in patient selection over time. Results: Short-term excess mortality significantly improved over time. Long-term excess mortality remained high with a doubled mortality risk for women (relative excess risk=1.87, 95% CI: 1.73–2.02). Excess mortality did not differ between age categories. The risk of cardiovascular versus noncardiovascular death remained similar over time, with a higher risk of cardiovascular death for women. Changes in patient population (ie, older and more comorbid patients in the latter period) marginally impacted excess mortality (2%). Conclusions: Despite changes in AAA care, patients retain a high long-term excess mortality after elective repair with a persistent high cardiovascular mortality risk. In this, a clear sex – but no age – disparity stands out.

A meta-analysis of long-term survival following elective abdominal aortic aneurysm (AAA) repair reports a persistently high excess mortality after repair. 1 Swedish population data imply a profound sex difference and suggest that excess mortality has not improved over the last 2 decades.A notable finding considering the advances in surgical care such as the intensification of endovascular aneurysm repair (EVAR) and cardiovascular risk management (CVRM). 2lthough the persistently compromised life expectancy following elective AAA repair may imply that AAA patients are still suboptimally treated for their long-term mortality risk.It cannot be excluded that it reflects epidemiological changes in AAA care with more comorbid and/or older patients receiving repair in the current endovascular era. 3 Or that AAA disease may be associated with a residual mortality that is not fully amendable to CVRM.Finally, it cannot be excluded that the apparent lack of survival improvement observed in the Swedish population reflects a national phenomenon.The more so because a recent analysis of Danish population data implied improved survival and cardiovascular (CV) outcomes for AAA patients over time. 4,5he aim of this time-trend analysis is to evaluate longterm excess mortality and its associated CV risk for AAA patients after elective repair, while addressing the changes in AAA management (ie, the increase in EVAR and CVRM) and epidemiological shifts (ie, older and more comorbid patients) over time.Therefore, a relative survival analysis was used to estimate disease-specific excess mortality, as well as a competing risk of death analysis to evaluate the risk of CV versus non-CV death over time.

Study Population
Included patients underwent primary elective AAA repair in The Netherlands between 1995 and 2017. 6Ruptured AAAs were excluded (Supplemental Digital Content 1, Table 1, http:// links.lww.com/SLA/E782).Comorbidity burden was estimated through the Charlson Comorbidity Index (CCI) and included all diagnoses registered at the same moment as AAA repair. 7rescriptions for CVRM were extracted 1 year before repair to ensure preventive risk management (Supplemental Digital Content 1, Table 1, http://links.lww.com/SLA/E782).The date and cause of death were extracted until December 31, 2018 (Supplemental Digital Content 1, Table 1, http://links.lww.com/SLA/E782). 8Three predefined periods were created, each reflecting clear contrasts in procedural an CVRM changes.0][11] A more detailed prescription of the methods and period choice is provided in the Supplemental Material (Supplemental Digital Content 1, http://links.lww.com/SLA/E782).

Relative Survival and Competing Risk of Death Analysis
Relative survival, rather than crude survival, was applied to (1) evaluate disease-specific mortality, and (2) to adjust for sex-related, age-related, and time-related differences in life expectancy. 12,13A competing risk of death analysis was performed to estimate possible shifts in the risk of CV versus non-CV death over time. 14Sensitivity analyses were performed to estimate the impact of changes in patient age and comorbidity burden on study conclusions.Further details are provided in the Supplemental Material (Supplemental Digital Content 1, http:// links.lww.com/SLA/E782).

Statistical Analysis
All analyses were performed in SPSS, version 26 (IBM) and Stata/SE, version 12.0 (StataCorp).Comparisons between groups were analyzed using standard statistical methods.Relative survival reflects the ratio of the observed survival of the study population (ie, electively treated AAA patients), and the expected survival of the matched (sex, age, and year of operation) general Dutch population. 12,13A relative survival below 100% means that the disease-specific survival is lower than that of the reference population (ie, excess mortality).Competing risk of death analyses were performed by estimating the cumulative incidence of CV death versus non-CV death (subdistribution hazard ratios, tested by Fine and Gray models). 14A detailed description of the statistical analysis is provided in the Supplemental Material (Supplemental Digital Content 1, http://links.lww.com/SLA/E782).

Time Trends in Patient Demographics
Between 1995 and 2017, 40,730 patients (86.6% men) underwent elective AAA repair.Clear demographic changes occurred over time (Table 1), including an increase in the proportion of women (from 12.1% to 15.2%), and a ∼2-year increase in mean age at repair (P < 0.001).Moreover, the CCI increased for men in all age categories, and for women in older age categories (70-74 and > 80 years).

Time Trends in CVRM
Detailed information regarding medical CVRM was available for the years 2006 to 2017 and summarized in Figure 1.Over time, the proportion of patients with at least 1 CVRM prescription increased: statins from 24.8% to 69.6% and 26.3% to 66.9%, antihypertensive from 29.5% to 73.7% and 34.3% to 78.0%, and antiplatelets from 26.5% to 71.3% and 25.6% to 69.4%, for men and women, respectively.

Time Trends in Excess Mortality (Relative Survival)
Short-term (1 year) relative survival improved for men, but not for women, over time.Long-term relative survival (3, 5, and 10 years) remained stably impaired over time, with a clear survival disadvantage for women (Fig. 2, Supplemental Digital Content 1, Table 2, http://links.lww.com/SLA/E782).In fact, compared with men, women had a doubled mortality risk after correction for age and CCI scores (relative excess risk = 1.87, 95% CI: 1.73-2.02).There were no differences in relative survival between age categories.

Time Trends in Competing Risk Analysis of CV Versus Non-CV Death
Cause of death analysis implies a higher risk for CV death in the electively repaired AAA population compared with the general population (Supplemental Digital Content 1, Figure 1, http://links.lww.com/SLA/E782).
A competing risk analysis was performed to estimate the risk of CV versus non-CV mortality over time.This showed no change in cumulative incidence of CV versus non-CV death for all age categories and for both sexes over time.Throughout the first years after repair, the risk of CV death exceeded that of non-CV death.The dominance of CV death persisted for a longer time after surgery in women (to 6.5 years) than in men (to 3.5 years) (Fig. 3, Supplemental Digital Content 1, Table 3, http://links.lww.com/SLA/E782,Supplemental Digital Content 1, Figure 2, http://links.lww.com/SLA/E782).

The Impact of Changes in Patient Characteristics on Study Outcomes
From periods 1 to 3, the mean age at the time of repair increased to 2.7 years for men and 1.9 years for women (both P < 0.001).Comorbidity scores increased for men in all age categories and for women in age categories 70 to 74 and > 80 years.To estimate the impact of the lower repair threshold, resulting in an increase in the proportion of older and/or patients with greater comorbidity in the more recent time frame, stratified analyses by age, and sensitivity analyses addressing different degrees of comorbidity, were undertaken.Outcome comparison of age categories showed similar survival rates and equal risks of CV death (Fig. 2, Supplemental Digital Content 1, Figure 2, http://links.lww.com/SLA/E782).Sensitivity analyses addressing CCI showed a higher relative excess risk of death for patients with a higher CCI score, and a more pronounced impact of comorbidities on survival in older patients (Supplemental Digital Content 1, Table 4, http://links.lww.com/SLA/E782).Censoring all patients with a CCI score ≥ 3 receiving repair in period 3, implied a modest impact of changes in patient comorbidity on the relative survival (∼2%) for men (< 65, 65-69, 75-80 years) and for younger women (< 65 years) (Table 2).

DISCUSSION
In this 25-year nationwide study, short-term mortality after elective AAA repair was improved, as expected.Long-term mortality was stable, despite an aging patient population with increasing comorbidity.Long-term excess mortality, compared with a sex-matched and age-matched general population, however, remained persistently high.The risk of CV death did not decrease over time despite the intensification of pharmaceutical CVRM.Profound sex but no age disparity stands out with a doubled excess mortality risk for women compared with men.Despite improvements in short-term survival over time due to improvements in surgical outcomes, long-term survival remained persistently impaired.To put this in perspective, the excess mortality of elective AAA patients equals that or is higher than, reported for most malignancies. 15The absence of longterm survival benefit could be explained the gradual lowering of the threshold for repair, which resulted in older and more comorbid patients considered eligible for repair in the current era (chronological bias).In this study, this is reflected by an increase in CCI, a doubling of the proportion of octogenarians, and a two-year increase in mean age at repair over the study period.The potential impact of these changes in a patient selection over time was addressed by age-stratified and sensitivity analyses.This showed that increased age did not correspond with lower survival.Associations were found between higher CCI scores and a higher risk of excess mortality (Supplemental Digital Content 1, Table 4, http://links.lww.com/SLA/E782).To evaluate the impact of higher CCI scores on the excess long-term mortality, a sensitivity analysis censoring all with a CCI ≥ 3 in period 3 was performed.This showed a marginal increase in relative survival of 2% (predominately men), yet relative survival remained severely impaired (Table 2).Therefore, the persistently high excess mortality only marginally relates to an increase in comorbidity burden.The observation that there were no differences in excess mortality between age categories implies that the existence of an AAA reflects overall vulnerability, resulting in a univocally high mortality risk, regardless of whether a patient is 65 or 85 years old.
To gain a better understanding of possibly underlying causes for the excess long-term mortality, overall causes of death were evaluated.This implied a high CV cause of death (without AAA rupture) in the electively repaired AAA population (∼50%) compared with the general population.A result confirmed by other studies. 16To evaluate to what extent the intensification of CVRM over time influences the risk of CV mortality, a competing risk of death analysis was performed.Hypothetically, intensification of CVRM should decrease the risk for CV death, thereby exposing the patient to other competing mortality risks (eg, malignancy), and thus potentially masking a beneficial effect of CVRM on overall survival.This analysis showed an unchanged CV mortality risk over time, as well as a particularly high CV risk for AAA patients in the first years after surgery.Therefore, it is suggested that the broader implementation of CVRM has a limited impact on CV mortality risk (on population level).
The apparently limited impact of CVRM on CV mortality risk may reflect relative undertreatment of CV risk in electively repaired AAA patients.Despite consensus endorsing maximum CVRM in those patients, studies show that half of the AAA patients still do not receive optimal CVRM. 17,18Moreover, AAA patients could be relatively resistant to traditional CVRM, as risk factors for AAA disease are different from traditional (atherosclerotic) CV risk factors. 19The unchanged excess mortality could reflect the dominant role of smoking in AAA patients compared with the general population (Supplemental Digital Content 1, Table 5, http://links.lww.com/SLA/E782).Yet, prior research showed that the impact of smoking is minimal on disease-specific excess mortality for diseases highly associated with excessive smoking. 20Alternatively, although the effectiveness of CVRM on nonfatal CV events has been firmly established, an effect on survival is unclear, as literature shows  between age categories.Alternative explanations include a higher allostatic load or comorbidity profile of women. 26However, the results of this study portray equal CCI scores for men and women, with an even lower effect size of CCI scores on survival in women (Supplemental Digital Content 1, Table 4, http://links.lww.com/SLA/E782).This is supported by a metaanalysis reporting fewer baseline comorbidities for women. 27he high CV mortality of women, which persists for a longer time after surgery, points to the involvement of CV risk factors in the excess mortality for women.In this context, poorer profiles of female AAA patients are worrisome and suggest that women receive suboptimal treatment for their CV risk.This appears a general phenomenon because women, even when correctly diagnosed, are often undertreated for their CV risk factors. 28ast, in contrast to men, the absence of a decline in tobacco use in women could explain the sex-dependent survival differences. 29ur data for the Dutch population confirm and extend Swedish population data, which showed a persistent impaired survival of AAA patients who underwent elective repair (Supplemental Digital Content 1, Figure 3, http://links.lww.com/SLA/E782). 2This conclusion contrast with a recent Danish study, reporting a decrease in overall mortality and CV risk in patients diagnosed with AAA over time. 4There are several explanations for this apparent discrepancy.First, the Danish study describes a heterogenous incomparable patient population that included both patients with either intact AAA or ruptured AAA, without mentioning surgery.Moreover, the follow-up in the Danish study was limited to 2 years, and therefore the improved survival of the Danish study population presumably reflects the improvements in short-term survival observed in the current study.Given the comparable excess mortality rates observed in both the Dutch and Swedish cohorts, it is unlikely that the persistently high excess mortality reflects a national phenomenon.

Strengths and Limitations
The first strength of this study is that is relies on a large nationwide registry, with a high reported validity (84%-99%), and covers an extensive period of 25 years.A second strength is the fact that both the introduction of EVAR and CVRM occurred in welldefined timeframes.This facilitated our time-trend analysis.Availability of data from 1995 to 2017 allowed for comparison of a period with almost no EVAR and CVRM, with a period of EVARfirst strategy and extensive CVRM.Application of relative survival analyses allowed to estimate disease-specific excess mortality while accounting for the changes in patient characteristics over time (ie, more female and older patients).As a result of an aging population and lower intervention threshold, a growing number of elderly patients is, and will be, considered eligible for AAA repair. 3Age differences exert important effects on survival comparisons but are often not considered in (traditional) survival analysis.Older patients have a higher a priori risk of death.Therefore, comparing overall crude survival differences between younger and older patients introduces bias (immortal time bias). 30Relative survival analysis, which corrects for age differences in (a priori) life expectancy, overcomes this.Furthermore, to the best of our knowledge, this is the first study to apply a competing risk analysis for CV mortality in AAA patients.As opposed to traditional survival analysis, competing risk analysis addresses the competitive nature of multiple causes to the same event.
This study has some limitations inherent to the retrospective design and the use of registry data such as the quality of coding. 31,32The use of general procedural codes prohibited a distinction between EVAR and open repair.However, long-term survival rates are reported to be similar for EVAR and open repair and were not the focus of this study. 2,3Cause of death registration is prone to bias and misclassification, hence the interpretation of specific causes of death requires caution.Analyses of main categories of causes of death are less prone to misclassification.This study aimed to address time-dependent differences and it is unlikely that there are differences/bias in the cause of death registration over time.Furthermore, stratification of results inevitably leads to a restricted number of patients and therefore larger 95% CIs.This hampers the detection of more subtle differences in women.Last, while we were unable to demonstrate a benefit of CVRM on CV death, a potential benefit of CVRM on nonfatal CV events (prevention or postponement of myocardial infarction or stroke), was considered beyond the scope of this study.

CONCLUSIONS
Although improvements in surgical care resulted in significantly lower short-term mortality and broadened the spectrum of patients eligible for repair, AAA patients, especially women, retain a high long-term excess mortality risk.Despite the intensification of CVRM, the CV mortality risk remains the primary cause of death.The persistently high excess mortality appears largely independent of changes in patient selection.Future studies should focus on high sex-dependent excess mortality and strategies to reduce accompanied comorbidity risks.

FIGURE 1 .
FIGURE 1.Time trends of the implementation of pharmaceutical CVRM, stratified by sex.

FIGURE 3 .
FIGURE 3. Cumulative incidence of CV versus non-CV mortality over time stratified by sex.

TABLE 1 .
Demographics of Patients Treated With Elective AAA Repair Per Period by Sex and Age

TABLE 2 .
Sensitivity AnalysisData represent relative survival with 95% CI.Five-year relative survival of entire patient cohort versus patient cohort with exclusion of patients with CCI > 3.