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).2

Although 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.4,5

The aim of this time-trend analysis is to evaluate long-term 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.

METHODS Study Population

Included patients underwent primary elective AAA repair in The Netherlands between 1995 and 2017.6 Ruptured AAAs were excluded (Supplemental Digital Content 1, Table 1, https://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.7 Prescriptions for CVRM were extracted 1 year before repair to ensure preventive risk management (Supplemental Digital Content 1, Table 1, https://links.lww.com/SLA/E782). The date and cause of death were extracted until December 31, 2018 (Supplemental Digital Content 1, Table 1, https://links.lww.com/SLA/E782).8 Three predefined periods were created, each reflecting clear contrasts in procedural an CVRM changes. In period 1 (1995–2000) open repair and rudimentary CVRM dominated; period 2 (2001–2011) was a transition period; and period 3 (2013–2017) saw an EVAR-first strategy and maximal implementation of CVRM.9–11 A more detailed prescription of the methods and period choice is provided in the Supplemental Material (Supplemental Digital Content 1, https://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.12,13 A competing risk of death analysis was performed to estimate possible shifts in the risk of CV versus non-CV death over time.14 Sensitivity 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, https://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.12,13 A 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).14 A detailed description of the statistical analysis is provided in the Supplemental Material (Supplemental Digital Content 1, https://links.lww.com/SLA/E782).

RESULTS 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).

TABLE 1 - Demographics of Patients Treated With Elective AAA Repair Per Period by Sex and Age Period 1 (1995–2000) Period 2 (2001–2011) Period 3 (2012–2017) P Elective repair, n (%) 11,096 18,731 10,903 NA  Men 9749 (87.86) 16261 (86.81) 9247 (84.81)   <65 2280 (20.55) 3117 (16.64) 1398 (12.82)   65–70 2212 (19.94) 3165 (16.90) 1826 (16.75)   71–74 2619 (23.60) 4113 (21.96) 2207 (20.24)   75–79 1846 (16.64) 3702 (19.76) 2054 (18.84)   >80 792 (7.14) 2164 (11.55) 1762 (16.16)  Women 1347 (12.14) 2470 (13.19) 1656 (15.19)   <65 230 (2.07) 331 (1.77) 197 (1.81)   65–70 240 (2.16) 388 (2.07) 280 (2.57)   71–74 358 (3.23) 641 (3.42) 362 (3.32)   75–79 332 (2.99) 659 (3.52) 431 (3.95)   >80 187 (1.69) 451(2.41) 386 (3.54) Age at time repair [mean (SD)] 70.0 (7.5) 71.4 (7.7) 72.3 (7.7) 0.000  Men 69.7 (7.4) 71.2 (7.6) 72.4 (7.7) 0.000   <65 59.7 (4.2) 59.7 (4.1) 60.0 (3.8) 0.004   65–70 67.1 (1.4) 67.2 (1.4) 67.2 (1.4) 0.263   71–74 72.0 (1.4) 72.0 (1.4) 71.9 (1.4) 0.126   75–79 76.7 (1.4) 76.8 (1.4) 77.0 (1.4) 0.000   >80 82.4 (2.3) 82.5 (2.4) 82.9 (2.6) 0.000  Women 71.7 (8.0) 72.8 (7.9) 73.6 (8.0) 0.000   < 65 58.9 (5.4) 58.9 (5.9) 59.1 (6.2) 0.944   65–70 67.2 (1.3) 67.1 (1.4) 67.3 (1.4) 0.079   71–74 72.1 (1.4) 72.1 (1.4) 72.1 (1.4) 0.987   75–79 76.9 (1.4) 76.9 (1.4) 76.9 (1.4) 0.730   >80 83.0 (2.7) 82.9 (2.8) 83.1 (2.7) 0.369 CCI score  Men, n (%)   <65 0.001 0 1274 (55.9) 1687 (54.1) 821 (58.7)    1 720 (31.6) 1061 (34.0) 379 (27.1)    2 192 (8.4) 240 (7.7) 128 (9.2)   >3 94 (4.1) 129 (4.1) 70 (5.0)   65–70 0.000    0 1400 (51.3) 2088 (52.5) 1237 (54.0)    1 923 (33.8) 1319 (33.2) 651 (28.4)    2 252 (9.2) 331 (8.3) 248 (10.8)   >3 154 (5.7) 238 (6.0) 157 (6.8)   71–74 0.000    0 1037 (49.3) 1672 (50.6) 899 (51.7)    1 698 (33.2) 1076 (32.6) 465 (26.7)    2 227 (10.8) 306 (9.3) 220 (12.6)   >3 140 (6.7) 248 (7.5) 156 (9.0)   75–79 0.001    0 903 (48.9) 1888 (51.0) 1045 (50.9)    1 597 (32.3) 1153 (31.2) 566 (27.6)    2 208 (11.3) 365 (9.8) 260 (12.7)   >3 138 (7.5) 296 (8.0) 183 (8.9)   >80 0.000    0 352 (44.4) 1128 (52.1) 905 (51.4)    1 280 (35.4) 664 (30.7) 464 (26.3)    2 97 (12.2) 187 (8.6) 226 (12.8)   >3 63 (8.0) 185 (8.6) 167 (9.5)  Women   <65 0.379    0 124 (53.9) 161 (48.6) 101 (51.3)    1 78 (33.9) 130 (39.3) 62 (31.5)    2 19 (8.3) 24 (7.3) 22 (11.2)   >3 9 (3.9) 16 (4.8) 12 (6.1)   65–70 0.142    0 153 (51.2) 239 (48.0) 169 (48.3)    1 93 (31.1) 174 (35.0) 100 (28.6)    2 30 (10.0) 56 (11.2) 45 (12.8)   >3 23 (7.7) 29 (5.8) 36 (10.3)   71–74 0.024    0 143 (47.8) 241 (45.4) 126 (43.2)    1 111 (37.1) 199 (37.5) 90 (30.8)    2 26 (8.7) 53 (10.0) 42 (14.4)   >3 19 (6.4) 38 (7.2) 34 (11.6)   75–79 0.077    0 153 (46.1) 316 (48.0) 210 (48.7)    1 119 (35.8) 228 (34.6) 133 (30.9)    2 42 (12.7) 65 (9.8) 41 (9.5)   >3 18 (5.4) 50 (7.6) 47 (10.9)   >80 0.016    0 96 (51.3) 217 (48.1) 212 (54.9)    1 58 (31.0) 154 (34.2) 86 (22.3)    2 18 (9.6) 45 (10.0) 52 (13.5)   >3 15 (8.0) 35 (7.7) 36 (9.3)

Data are presented as number (%) unless indicated otherwise.

NA indicates not available.

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.

FIGURE 1:

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

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, https://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.

FIGURE 2:

Relative survival (1, 3, 5, 10 years) for men (A) and women (B) stratified by age. Note that 10-year relative survival could only be calculated for periods 1 and 2 due to the restricted follow-up in period 3.

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, https://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, https://links.lww.com/SLA/E782, Supplemental Digital Content 1, Figure 2, https://links.lww.com/SLA/E782).

FIGURE 3:

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

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, https://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, https://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).

TABLE 2 - Sensitivity Analysis Period 1 (1995–2000) Period 3 (2012–2017) Entire cohort Entire cohort Exclusion CCI >3 Men  <65 0.88 (0.87–0.90) 0.93 (0.91–0.96) 0.94 (0.91–0.96)  65–70 0.86 (0.84–0.88) 0.89 (0.87–0.92) 0.92 (0.89–0.95)  71–74 0.88 (0.85–0.90) 0.90 (0.86–0.94) 0.93 (0.88–0.96)  75–79 0.87 (0.84–0.91) 0.93 (0.88–0.97) 0.95 (0.91–0.99)  >80 0.99 (0.91–1.06) 0.96 (0.89–1.02) 1.00 (0.92–1.07) Women  <65 0.84 (0.79–0.89) 0.94 (0.88–0.98) 0.95 (0.90–0.99)  65–70 0.80 (0.74–0.85) 0.83 (0.76–0.89) 0.86 (0.78–0.92)  71–74 0.80 (0.73–0.85) 0.77 (0.65–0.86) 0.79 (0.67–0.88)  75–79 0.74 (0.67–0.80) 0.76 (0.67–0.84) 0.79 (0.70–0.87)  >80 0.85 (0.64–1.09) 0.83 (0.71–0.95) 0.88 (0.75–1.00)

Data represent relative survival with 95% CI.

Five-year relative survival of entire patient cohort versus patient cohort with exclusion of patients with CCI>3.

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.15 The absence of long-term 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, https://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.16 To 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.17,18 Moreover, AAA patients could be relatively resistant to traditional CVRM, as risk factors for AAA disease are different from traditional (atherosclerotic) CV risk factors.19 The unchanged excess mortality could reflect the dominant role of smoking in AAA patients compared with the general population (Supplemental Digital Content 1, Table 5, https://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.20 Alternatively, although the effectiveness of CVRM on nonfatal CV events has been firmly established, an effect on survival is unclear, as literature shows heterogenous results on the effect of CV medication of life expectancy.21,22

This study showed a worrisome persistently high excess mortality for women. While inferior survival outcomes for women have been outlined previously, the basis for this sex disparity remains unclear.23–25 This study indicates that the disparity is not explained by a higher age of women at the time of repair, as there were no significant survival differences between age categories. Alternative explanations include a higher allostatic load or comorbidity profile of women.26 However, 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, https://links.lww.com/SLA/E782). This is supported by a meta-analysis reporting fewer baseline comorbidities for women.27 The 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.28 Last, in contrast to men, the absence of a decline in tobacco use in women could explain the sex-dependent survival differences.29

Our 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, https://links.lww.com/SLA/E782).2 This conclusion contrast with a recent Danish study, reporting a decrease in overall mortality and CV risk in patients diagnosed with AAA over time.4 There 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 well-defined 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 EVAR-first 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.3 Age 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).