NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Guirguis-Blake JM, Beil TL, Sun X, et al. Primary Care Screening for Abdominal Aortic Aneurysm: A Systematic Evidence Review for the U.S. Preventive Services Task Force [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2014 Jan. (Evidence Syntheses, No. 109.)

Cover of Primary Care Screening for Abdominal Aortic Aneurysm

Primary Care Screening for Abdominal Aortic Aneurysm: A Systematic Evidence Review for the U.S. Preventive Services Task Force [Internet].

Show details

3Results

Literature Search

Our literature search yielded 2,723 unique citations. From these, we provisionally accepted 204 articles for review based on titles and abstracts (Appendix B). After screening the full-text articles, 51 studies (68 articles) were judged to have met the inclusion criteria (Appendix A). The remaining 151 full-text articles were excluded (Appendix D).

Overview of Included Studies

Twenty-four studies, including 13 RCTs, eight cohort studies, and three case-control studies that were reported in 44 published papers, were included in our systematic review.13-16,25,41,67,89-124 Of those studies, four major RCTs—two good- and two fair-quality—investigated the benefits of one-time screening for AAA in general asymptomatic populations;13-16 these four studies and two additional fair-quality cohort studies also assessed harms associated with one-time screening for AAA.104,105 No RCTs were available for assessing the effect of rescreening, but seven observational studies (six cohort studies and one case-control study) examined the benefits of rescreening,96-102 and one of these also reported harms.99 Additionally, two large, good-quality RCTs reported benefits and harms of early open surgery for small AAA,106,108 and two moderate-sized, fair-quality RCTs assessed the benefits of early EVAR for small AAA.113,115 Those two RCTs and two additional fair-quality registry studies also reported harms associated with early EVAR for small AAA.122,123 Four small- to moderate-sized RCTs investigated the benefits and harms of beta-blockers and antibiotics for small asymptomatic AAA,116,118-120 and one additional RCT reported harms associated with the use of beta-blockers.117 Because of the complexity of the evidence body, we report information regarding study design, patient and intervention characteristics, and study outcomes about screening for AAA and treatment of small AAA, respectively.

KQ 1. What Is the Effect of One-Time AAA Screening on Health Outcomes in an Asymptomatic Population Age 50 Years and Older?

Summary of Results

Four large population-based screening RCTs of men age 65 years and older examined the effectiveness of one-time AAA screening, showing that AAA prevalence varies from 4 to 7.7 percent and the majority of screen-detected AAAs are small, measuring smaller than 4 to 4.5 cm. Invitation for screening in men age 65 years and older was associated with reduced AAA-related mortality, AAA rupture rates, and number of emergent surgeries but not all-cause mortality.

Study Details

Two fair-quality and two good-quality population-based screening RCTs from the United Kingdom, Denmark, and Australia assessed the efficacy of AAA screening in population-based settings: the Multicentre Aneurysm Screening Study (MASS) (n=67,800);13,89,90,129 the Chichester, United Kingdom screening trial (n=15,775);14,25,91,124 the Viborg County, Denmark screening trial (n=12,639);15,67,92,93,125 and the Western Australian screening trial (n=41,000) (Appendix E Table 1).16 All trials identified potential participants age 64 or 65 years and older from population registries or regional health directories. MASS identified participants from four centers in the United Kingdom, the Chichester trial included nine general practices in Chichester, the Viborg trial included the population from Viborg County, and the Western Australian trial included participants from a capital city and satellite towns. Reported mean (or median) ages ranged from 67.7 to 72.6 years, and the oldest study participants were age 80 years. One study, the Chichester trial,14 included women,25 while the other three recruited only men. Other than age and sex, no studies reported outcomes by any other demographic information. The Viborg trial reported AAA-related comorbidity risk factor information from hospital discharge data, indicating that 26.5% of all participants had at least one cardiovascular risk factor or COPD.93,125 The Western Australian trial reported cardiovascular comorbidity and risk factor information for the screened group and analyzed the association between the risk factor and AAA diagnosis, but these risk factors were not collected for the control group, nor were they linked to mortality outcomes.8 Three studies had no trial exclusions; only MASS excluded patients who 1) were identified by their primary care physician as too high risk to be screened, 2) were terminally ill, or 3) had other serious health problems or prior AAA repair.

All trials randomized participants to two groups: the control group received usual care, while the invited group received a letter invitation for one-time ultrasound screening (Appendix E Table 1). All trials considered normal aortic diameter to be smaller than 3 cm and defined AAA as 3.0 cm or larger. Three of the RCTs (MASS, Viborg, Chichester) further prescribed specific postscreening surveillance protocols for AAAs measuring 3.0 cm or larger with repeat ultrasounds,13,14,67 while one trial (Western Australian) sent initial ultrasound results to primary care physicians for management.16 In MASS, those with aortic diameters measuring 3.0 to 4.4 cm were rescanned yearly, those measuring 4.5 to 5.4 cm were rescanned at 3-month intervals, and those measuring 5.5 cm or larger were urgently referred to a vascular surgeon.13 In the Viborg trial, individuals with ectatic aortic size of 2.5 to 2.9 cm were offered a repeat scan at 5 years, those measuring 3.0 to 4.9 cm were offered annual scans, and those measuring 5.0 cm or larger were referred to vascular surgery.67 In the Chichester trial, patients with AAAs measuring 3.0 to 4.4 cm were rescanned annually, those measuring 4.5 to 5.9 cm were rescanned every 3 months, and those measuring 6 cm or larger were referred to a vascular surgeon, as were those with an increase in diameter of 1 cm or more per year.14

The primary outcome reported in trials was AAA-specific mortality (defined as all AAA deaths plus all deaths within 30 days of AAA surgical repair); all four trials also reported AAA rupture rate and all-cause mortality as benefit outcomes (Appendix E Table 1). Mortality data and causes of death were ascertained from death certificates in all studies, and three of the RCTs additionally involved an independent blinded review of autopsy reports and/or hospital records for all AAA-related deaths. Mean followup in these trials ranged from 3.6 to 15 years. Local and national health departments, research councils, and heart foundations funded these studies. MASS13 and the Viborg trial67 were assigned a good-quality rating based on USPSTF criteria.86 MASS had the greatest number of participants and the highest adherence to screening, with clear reporting of randomization, allocation, blinding of outcome assessors, and confirmation of equal followup in the invited and control groups.13 The Viborg trial, while the smallest, adequately reported randomization and blinding of outcome assessors. There was, however, a difference between attendees and nonattendees, with nonattendees being significantly older than attendees (one third of those age 73 years vs. <20% of those ages 65 to 67 years did not attend).67 The Chichester and Western Australian trials were assigned fair-quality ratings due to inadequate description of blinding of outcome assessors and lack of reporting of loss to followup (Chichester) or lack of detail regarding randomization method (Western Australian).14,16 All trials appeared to use intention-to-treat analysis; adherence to screening varied from the lowest adherence in the Western Australian trial (62.5% of those invited attended screening) to the highest adherence in MASS (80.2% adherence). All studies reported outcomes for attendees and nonattendees in the invited group separately. Three studies reported low loss-to-followup rates in the participants with AAA: MASS (72% at 10-year followup),13 Viborg trial (75.1% retention rate in invited group; 58.0% in control group at 52-month followup),67 and Western Australian trial (87.1% retention in invited group; 84.9% in the control group at 3.6-year followup).16

AAA Prevalence in the Screened Population

AAA prevalence on the initial screening for male attendees varied from 4.0 and 4.9 percent in the Viborg trial and MASS to 7.6 and 7.7 percent in the Chichester and Western Australian trials, respectively (Table 1). The two latter trials with higher AAA prevalence rates recruited older participants (Chichester median age, 72 years; Western Australian mean age, 72.7 years; compared with mean ages of 67.7 and 69.2 years in the Viborg trial and MASS, respectively). Three of the four trials (MASS, Chichester, and Western Australian) reported prevalence of AAA by size at initial screening. MASS and the Western Australian trial reported that the majority of AAAs (71% to 80%) were small, measuring 3.0 to 4.4 cm. In Chichester, approximately 60 percent of the detected AAAs were 3.0 to 3.9 cm. The prevalence of larger AAAs (≥5 cm or ≥5.5 cm) in the screened population was consistent across studies and was reported as 0.4 to 0.6 percent (Appendix F).

Table 1. AAA Prevalence, Rupture, and Surgery Data for One-Time Screening Trials (KQs 1 and 3).

Table 1

AAA Prevalence, Rupture, and Surgery Data for One-Time Screening Trials (KQs 1 and 3).

Effect of Population Screening on AAA-Related Mortality

Table 2 presents the mortality results of the four population-based screening RCTs for men. MASS and the Viborg trial found statistically significant AAA-related mortality benefit in the invited group compared with the control group at each of the followup time points, while the Western Australian and Chichester studies had ORs/HRs of less than 1, but were not statistically significant. In these four trials, 26.3 to 77.4 percent of AAA-related deaths in the invited group occurred in nonattendees; however, all results used intention-to-treat analysis.

Table 2. All-Cause and AAA-Related Mortality Data for One-Time Screening Trials (KQs 1 and 3).

Table 2

All-Cause and AAA-Related Mortality Data for One-Time Screening Trials (KQs 1 and 3).

Meta-analysis of the four trials, using a random-effects model (Figure 2), produced a summary risk ratio (RR) of 0.57 (95% CI, 0.44 to 0.72 ), 0.38 (95% CI, 0.17 to 0.86), and 0.50 (95% CI, 0.31 to 0.79) in favor of screening at 3 to 5 years, 6 to 7 years, and 10 to 11 years, respectively. Pooled analysis of the three longest studies at 13 to 15 years (Chichester, MASS, and Viborg; n=86,446) likewise showed a statistically significant benefit with AAA screening (RR, 0.58 [95% CI, 0.39 to 0.88]). High heterogeneity was detected at all time points after the 3- to 5-year time period, making the point estimate less reliable. Nonetheless, qualitative synthesis shows AAA-related mortality benefit appears as a consistent and persistent result over time when examining the individual trial results. Sensitivity analysis using HRs versus RRs and Peto ORs versus RRs did not alter conclusions (additional meta-analyses are shown in Appendix K). No publication bias was identified (Appendix K).

Figure 2 is a forest plot depicting the relative risk for AAA-related mortality in one-time screening trials. This figure is discussed in the Results section.

Figure 2

Pooled Analysis of AAA-Related Mortality in One-Time Screening Trials (Random-Effects Model).

Effect of Population Screening on All-Cause Mortality

Each trial, except Viborg at 5.9 years, demonstrated no statistically significant benefit in all-cause mortality for men in the invited versus control group at each followup time point (Table 2).

Meta-analysis performed at four different time points ranging from 3 to 5 years up to 15 years of followup demonstrate no all-cause mortality benefit with an invitation for AAA screening compared with control (Figure 3). At the 15-year followup, pooled analysis from the longest trials, Chichester, MASS, and Viborg (n=86,446), produced a summary RR of 0.98 (95% CI, 0.97 to 1.00) using a random-effects model. Sensitivity analyses, performed using a fixed-effects model, changed the 6- to 7-year followup results to be statistically significant (OR, 0.96 [95% CI, 0.93 to 0.99]), and sensitivity analysis using studies with HRs as the measure of effect found a statistically significant reduction in overall mortality at 10 to 11 years (HR, 0.97 [95% CI, 0.95 to 0.99]) and at 13 to 15 years (HR, 0.97 [95% CI, 0.96 to 0.99]) (Appendix K). No publication bias was detected (Appendix K).

Figure 3 is a forest plot depicting the relative risk for all-cause mortality in one-time screening trials. This figure is discussed in the Results section.

Figure 3

Pooled Analysis of All-Cause Mortality in Screening Trials (Random-Effects Model).

Effect of Population Screening on AAA Rupture

Individual study results for AAA rupture rate in men at different followup endpoints are presented in Table 1 and Figure 4. Only MASS shows fewer AAA ruptures in the invited group at 4.1-year (RR, 0.51 [95% CI, 0.38 to 0.69]), 7-year (RR, 0.53 [95% CI, 0.43 to 0.65]), 10.1-year (RR, 0.52 [95% CI, 0.43 to 0.62]), and 13.1-year followup (RR, 0.57 [95% CI, 0.49 to 0.67]). At 4.3 years of followup, one of three trials (Viborg) showed a statistically significant reduction in AAA rupture. The Chichester trial showed fewer AAA ruptures at 5 and 10 years but not at 15 years. At 10 years, the Viborg and Chichester trials and MASS showed a benefit from screening. Using a random-effects model, meta-analysis demonstrated that screening was associated with a lower AAA rupture rate at all except the 15-year followup: 3 to 5 years (RR, 0.52 [95% CI, 0.35 to 0.79]), 7 years (RR, 0.53 [95% CI, 0.43 to 0.65]), and 10 to 11 years (RR, 0.27 [95% CI, 0.11 to 0.65]). Of note, despite MASS results included in the 13- to 15-year pooled analysis, there was still no statistically significant benefit in rupture at the 15-year followup. Heterogeneity was high at all time points; therefore, the magnitude of screening benefit on rupture is less certain. Sensitivity analyses using a fixed-effect model did alter the 13- to 15-year point estimate (RR, 0.61 [95% CI, 0.53 to 0.70]); sensitivity analysis using HRs did not alter conclusions (Appendix K). No apparent publication bias was found (Appendix K).

Figure 4 is a forest plot depicting the relative risk for rupture in one-time screening trials. This figure is discussed in the Results section.

Figure 4

Pooled Analysis of Rupture in One-Time Screening Trials (Random-Effects Model).

Emergent Repairs for AAA Rupture

Three trials reported rates of emergent AAA repairs for rupture in men in the invited and control groups (Table 1). Two of the three trials (Viborg and MASS) showed fewer emergency surgeries in the invited group at all measured time points. Meta-analysis using a random-effects model showed fewer emergent surgeries for AAA rupture at 3 to 5 years (RR, 0.42 [95% CI, 0.28 to 0.64]) (Figure 5). Pooled point estimates at the 13- to 15-year followup of the Viborg trial and MASS showed a reduction in emergent repairs in the invited group (RR, 0.42 [95% CI, 0.32 to 0.54]). Sensitivity analysis using fixed- versus random-effects models did not substantially alter results.

Figure 5 is a forest plot depicting the relative risk for emergent repairs of ruptures in one-time screening trials. This figure is discussed in the Results section.

Figure 5

Pooled Analysis of Emergent Repairs for Ruptures in One-Time Screening Trials (Random-Effects Model).

KQ 1a. Does the Effect of One-Time Screening Vary Between Men and Women, Smokers and Nonsmokers, Older and Younger Patients, Patients With and Without a Family History of AAA, and Patients of Different Races/Ethnicities?

Summary of Results

Only one population screening RCT examined AAA screening in women, showing that there is a low prevalence of AAA in women and that most AAAs detected at screening are small. There was no difference in AAA rupture rates at 5- and 10-year followup or in all-cause mortality at 5 years between the invited and control groups. Rupture rate is low, and most AAA ruptures occur in women age 80 years and older.14,25

Subanalyses from one RCT addressing ages younger and older than 65 years showed similar AAA-related or all-cause mortality benefit from screening in the older and younger screened age groups.93 Another subanalysis from one RCT showed no AAA-related mortality benefit from screening in patients age 75 years and older or in those ages 65 to 74 years.16

Study Details

Women

Only the Chichester study recruited female participants ages 65 to 80 years (59% of participants were women; n=9,342 women) (Appendix E Table 2). Sex-specific results at 5-year followup are reported with the larger trial results;14 5- and 10-year sex-specific results are published separately.25 Compared to men in every age cohort, more invited women refused screening. For example, in the 65-year-old cohort, 27.3 percent of invited women refused screening compared with 19.5 percent of men; in the 76- to 80-year-old cohort, 41.7 percent of invited women refused screening, while 33.8 percent of men refused. Prevalence of AAA in screened women was six times lower than that reported in Chichester men (1.3% vs. 7.6%, respectively) (Table 1). In the screened group, no women were diagnosed with AAA at the age of 65 years, 1 percent were diagnosed at ages 66 to 70 years, 1.8 percent at ages 71 to 75 years, and 1.6 percent at ages 76 to 80 years. The majority of AAAs (30/40) were small, measuring 3 to 3.9 cm; six AAAs measured 5.0 cm or larger. There was no difference in AAA rupture rates between the invited and control groups at 5- or 10-year followup (Table 1). At 5 years, three ruptures occurred in the invited group (0.06%) and three ruptures occurred in the control group (0.06%). At 10 years, 10 ruptures occurred in the invited group (0.21%) and nine ruptures occurred in the control group (0.19%). AAA-specific mortality was low in both groups (four deaths [0.08%] in the invited group and nine deaths [0.19%] in the control group; no statistical analysis) (Table 2). AAA-related mortality was reported in the entire unscreened population in Chichester, and while more than half of the AAA deaths in men occurred before age 80 years, the majority (70%) of AAA deaths in women occurred at age 80 years and older. All-cause mortality at 5 years was similar in both groups: 10.7 percent in the invited group and 10.2 percent in the control group. All-cause mortality was not reported for women at 10-year followup.

Older Adults

The oldest participants in the four major screening trials ranged in age from 73 to 80 years (Appendix E Table 2). Two of the population-based screening trials reported AAA-related mortality outcomes stratified by age or provided a subgroup analysis by age.16,93 The 13-year followup of the Viborg trial performed a subgroup analysis of 5,429 men ages 64 to 65 years and showed similar AAA-related mortality benefit in the 64- to 65-year-old group (HR, 0.36 [95% CI, 0.14 to 0.93]) compared with the 66- to 73-year-old group (HR, 0.33 [95% CI, 0.18 to 0.62]).93 Likewise, the lack of all-cause mortality benefit from screening was similar in the younger and older age groups (HR, 0.98 [95% CI, 0.89 to 1.07] in 64- to 65-year-olds; HR, 0.98 [95% CI, 0.93 to 1.03] in 64- to 73-year-olds). Of note, these subanalyses were not sufficiently powered. The Western Australian trial showed no AAA-related mortality benefit in the invited group age 75 years and older (OR, 1.13 [95% CI, 0.56 to 2.29]) or in those ages 65 to 74 years (OR, 0.82 [95% CI, 0.37 to 1.84]).16

None of the population-based screening RCTs reported smoking history, AAA family history, or race/ethnicity descriptive data for participants (Appendix E Table 2). All studies were conducted in majority Caucasian populations.

KQ1b. Does the Effect of One-Time Screening Vary Between Different Screening Approaches?

Summary of Results

One of the population-based RCTs collected risk factor information after randomization and analyzed AAA prevalence, AAA-related mortality, and overall mortality comparing low- and high-risk screening strategies. This simulation showed that there is a tradeoff: high-risk screening reduces the number of patients screened but prevents only half of AAA deaths.93,125 This simulation used study methods that could lead to underascertainment of high-risk status, which would bias the estimated impact of high-risk screening toward a lesser effect.

Study Details

High-Risk Versus Low-Risk Screening Strategy

The Viborg trial collected risk factor information on all participants after trial randomization and analyzed the yield of high- and low-risk screening approaches based on comorbidity with 5- and 13-year followup in men ages 64 to 73 years.93,125 The high-risk group included participants with one or more of the following conditions: hypertension, myocardial infarction (MI), COPD, ischemic heart disease, peripheral occlusive arterial disease, stroke, or transient ischemic attack. The low-risk group did not have any of the above comorbidities. The presence of comorbidities was extracted from hospital discharge diagnoses. Attendance in the high-risk group was significantly higher than in the low-risk group (78.8% vs. 75.8%, respectively; p<0.001).

Forty-six percent of the AAAs diagnosed in all participants (invited plus control) were in the high-risk group. In the screened group, prevalence of AAA in the high-risk group was more than double that of the low-risk group (6.7% vs. 2.9%). AAA-related mortality and total deaths were reported for each high-risk group, showing that the highest AAA-related mortality rate in the control group was in subgroups with peripheral artery disease (PAD) and COPD (AAA-related mortality, 4.62 vs. 3.42 per 1,000 years, respectively). Out of 339 patients with PAD, 157 died in the mean followup period of 5.1 years; one patient died of AAA-related causes in the invited group and four AAA deaths occurred in the control group. Out of 860 patients in the COPD group, 372 died at 5.2-year mean followup; one patient died of AAA-related causes in the invited group and seven AAA deaths occurred in the control group.

Kaplan Meier estimates over 8 years of followup showed that invitation for screening similarly reduced AAA-specific mortality in both low- and high-risk groups (HR, 0.24 [95% CI, 0.09 to 0.63] vs. 0.22 [95% CI, 0.08 to 0.65], respectively).

At 5-year followup, of the 30 AAA-related deaths that were prevented in the mass screening trial, selective high-risk screening would have prevented nearly half (14/30) of those deaths, and high-risk screening would have required 72.9 percent fewer screening invitations compared with mass screening. On the other hand, low-risk screening (only screening those without any of the risk factors) would have prevented about half (16/30) of the AAA deaths in the whole mass screening.

At 13-year followup, however, the high-risk AAA population realized less of an AAA-related mortality benefit from screening compared with the low-risk group (HR, 0.42 [95% CI, 0.2 to 0.87] vs. 0.29 [95% CI, 0.14 to 0.6], respectively). Results showing diminished benefit in the high-risk group at 13 years should be interpreted cautiously, as the Viborg trial's independent end point committee found classification bias, in which deaths of participants with known heart disease and AAA were more likely to be attributed to heart disease than to AAA. The effects of screening were therefore diminished, especially in the high-risk population with a high prevalence of coronary disease. At 13-year followup, there was no overall mortality benefit from invitation for screening (HR, 1.04 [95% CI, 0.95 to 1.13]).

Care should be taken in drawing definitive conclusions about high-risk screening yield based on this study. Because comorbidities were extracted from hospital discharge data, the high-risk cohort was likely sicker (i.e., required hospitalization) than a population defined as having one or more risk factors, thereby potentially biasing results against targeted, high-risk screening. For example, many participants in the Viborg trial with hypertension, PAD, or COPD would never have been hospitalized and therefore would never have been defined as high risk. Also, it is important to note that this was a simulation study; participants were randomized in the trial prior to risk factor collection or analysis, thereby making this study similar to a cohort study rather than an RCT.

There are no population-based screening trials comparing low- and high-risk screening approaches.

KQ 2. What Is the Effect of Rescreening for AAA on Health Outcomes or AAA Incidence in a Previously Screened, Asymptomatic Population Without AAA?

Summary of Results

Five fair-quality and one good-quality prospective cohort studies and one fair-quality case-control study with various rescreening protocols showed that AAA-related mortality over 5 to 12 years is rare (<3%) among those with normal aortas (<3 cm) on the initial scan.96-102 Over 5 to 12 years, very few of those with aortas measuring smaller than 3 cm, particularly smaller than 2.5 cm, developed any AAA. As one would expect, some (19% to 88%) of those with initial diameters of 2.5 to 2.9 cm progressed to a diameter larger than 3.0 cm after 5 to 6 years. A smaller percentage grew to larger than 5 cm (0% to 2.4%) at 5 years,98,100,126 and 15% had progressed at 10 years.17 One fair-quality individual patient data meta-analysis reported data on time-to-event for AAA incidence and rupture for a subgroup of patients with subaneurysmal aortic dilatation (2.5 to 2.9 cm), also called ectatic aorta.127 Mean time to rupture (after initial subaneurysmal detection) was 18.7 years (95% CI, 18.3 to 19.1) and was quite rare (<1%), although 8.3 percent developed a large aneurysm (≥5.4 cm) over a mean of 13.2 years.

A small number of participants with normal aortas were included in these studies, there are no matched controls in most studies, and most studies measured expansion rates rather than health outcomes.

Study Details

Five fair-quality and one good-quality prospective cohort studies96-99,101,102 and one fair-quality case-control study100 (two in the United States, four in the United Kingdom, and one in Denmark), with the number of total participants ranging from 223 to 15,098 and with mean followup ranging from 4 to 12 years, examined the yield of rescreening participants with normal or ectatic aortas (Appendix G Table 1).96-103 Four of these trials recruited patients from subsets of larger population-based screening and treatment RCTs described in KQs 1 and 4.96,99-101 The largest two studies were subsets of a screening RCT (n analyzed=4,308; subset from the Chichester trial)99 and a treatment trial (n analyzed=5,151; subset from the ADAM trial).101 Definitions of normal and ectatic aortas differed; inclusion criteria based on aortic diameter were defined as follows: 2.5 to 2.9 cm,96,100 2.6 to 2.9 cm,98 smaller than 2.6 cm,102,103 smaller than 3 cm,97,99 or 3 cm and larger.101 Ultrasound measurement techniques varied, with some measurements obtained using inner-to-inner wall measurements,17,99,126 outer-to-outer wall measurements,96,100,101 or by unspecified measurements.98 Repeat screening occurred at various intervals after the initial normal scan, as follows: 3 months to annually,98 annually,96 at 5 years and 12 years,102,103 every 2 years for up to 10 years or once after 5 years,99 once at 4 years,101 once at 3 to 5 years,100 and every 2 years.97 While three of these trials also analyzed the yield of various surveillance strategies for small AAA (≥3 cm), these data are not reviewed here.97,99,100 One individual patient data meta-analysis examined the subgroup with ectatic or subaneurysmal aortic diameters (2.5 to 2.9 cm) using time-to-event data for 1,696 individuals (66 women) from eight European centers.127 Median age at first scan was 66 years (range, 56 to 71 years); period of followup varied (median, 4 years [range, 0.1 to 19.0]), as did mean number of scans per individual (2.0 to 6.7). Benefit outcomes reported in these trials included all-cause mortality,17,96 AAA-related mortality,17 AAA rupture (fatal and nonfatal),17,96,97,101-103,126 and AAA growth rate.98-100,102,103,126 Four of these studies reported use of procedures.17,96,100,101

Patient characteristics are shown in Appendix G Table 2. Studies included men age 65 years and older, with only one study including men as young as age 50 years.101,127 Mean age was reported in four of the studies and ranged from 65.6 to 74.8 years,96,98,99 with most studies averaging 65 to 68 years and one study with an older population mean age of 74.8 years.98 Only one study reported inclusion of female participants (2.4%), and a few women (N=66) were included in the patient-level meta-analysis.101 While one other study followed a subset of ADAM patients, there is no mention of the sex of participants; however, there cannot be a large number of women, given the low number of women in the original cohort.96 Only the two studies which followed subgroups from the ADAM trial included risk factor information, including smoking history, family history, diabetes, COPD, hypertension, peripheral vascular disease, and CVD risk factors.96,101

AAA Incidence (New Cases)

Each study reported the percentage of participants with initially normal scans who eventually developed an AAA measuring 3.0 cm or larger during the 5- to 12-year followup period (Table 3). This AAA incidence varied considerably from 1 to 88 percent,96-103 with the highest incidence of 88 percent found in the study with older participants (mean age, 74.8 years) who had an aortic measurement of 2.6 to 2.9 cm on their original scan.98 Half or more of the incident AAAs were small, measuring 3 to 3.9 cm; in almost all studies, only 0 to 1.3 percent of normal AAAs (<2.5 or <3.0 cm) expanded to larger than 5 cm during the followup period. In a study of 547 men with an initial measurement of 2.6 to 2.9 cm, 15 percent developed an AAA larger than 5.4 cm over 10 years. 17 Based on individual patient data meta-analysis of 1,696 individuals (6 women) with subaneurysmal aortic diameters, 1,011 (59.6%) developed an AAA (≥3 cm) after a mean followup of 4.7 years (median followup, 4.0 years [range, 0.1 to 16.3 years]), while 8.3 percent developed a large AAA (>5.4 cm) after a mean followup of 13.2 years (median, 12.6 years [range, 1.2 to 19.5 years]).127

Table 3. AAA Prevalence, Rupture, and Surgery Data for Rescreening Trials (KQ 2).

Table 3

AAA Prevalence, Rupture, and Surgery Data for Rescreening Trials (KQ 2).

AAA Rupture

Reported AAA rupture rates were low in participants with ectatic or normal aortas (Table 3). Studies that included an initial diameter smaller than 2.6 or 3.0 cm reported no ruptures at 4 to 12 years.101-103 A small study of 223 patients with an initial aortic diameter of 2.5 to 2.9 cm reported no AAA ruptures at 5.9-year mean followup,96 but a study of 547 patients with initial aortic diameter of 2.6 to 2.9 cm reported that 2.4% had experienced a rupture at 10 years.17 An individual patient data meta-analysis found AAA rupture to be rare (14/1,631).127 A study of 2,691 men with an initial normal aorta reported a 0.07 percent rupture rate (95% CI, 0.02 to 0.30) at 10-year mean followup. The denominator for this rate included a population of patients who were screened once plus a population rescreened every 2 years for 10 years, so it is unclear whether these two ruptures occurred in the rescreened population.97

AAA-Related Mortality

AAA-related mortality rates ranged from 0 to 2.4 percent (Table 4). One study reported no AAA-related deaths at 5- and 12-year followup for those with an initial diameter of smaller than 2.6 cm.102,103 Two studies followed a population of patients with an initial diameter of 2.6 to 2.9 cm and although there were no AAA-related deaths at 5 years, 2.4% had died of AAA-related causes at 10 years.

Table 4. All-Cause and AAA-Related Mortality Data for Rescreening Trials (KQ 2).

Table 4

All-Cause and AAA-Related Mortality Data for Rescreening Trials (KQ 2).

All-Cause Mortality

All-cause mortality was not well reported in these studies, but one study of individuals with an initial aortic diameter of 2.6 to 2.9 cm reported that 34% had died at 10 years. 17

AAA Procedures

Several studies reported no AAA procedures at 4 to 5.9 years,96,100,101 but one study of 547 participants with an initial aortic diameter of 2.6 to 2.9 cm reported that 11.5% had undergone a procedure by 10 years.17 Most of these procedures were elective (9.7%).

Operative Mortality

Only one trial reported 30-day operative mortality (11.1%), and most deaths occurred in those with ruptured aortas (Table 4).17

KQ 2a. Does the Effect of Rescreening Vary Between Men and Women, Sizes of AAA, Smokers and Nonsmokers, Older and Younger Patients, Patients With and Without a Family History of AAA, and Patients of Different Races/Ethnicities?

Summary of Results

Only two prospective cohort studies (one fair-quality, one good-quality) analyzed AAA detection and AAA-related mortality rates with rescreening using uni- or multilogistic regression models by risk factor; both are subsets of the ADAM trial.96,101 There were no AAA ruptures or AAA-related deaths in participants with normal aortas on initial screening in either of these studies. For detection rates with rescreening at 4 and 5.9 years, results are conflicting: one smaller cohort study (N=223) showed no association between risk factors and subsequent AAA detection, while the other, much larger, cohort (N=2,622) showed three risk factors (current smoker, coronary artery disease, and any atherosclerosis) associated with AAA detection at rescreening. One fair-quality cohort study examined the association between age of participants with aortic size smaller than 3.0 cm at initial scan and later development of AAA and found no association between age group and AAA-related mortality. Conclusions about the yield of rescreening in subgroups is limited by very few nonwhite or female participants and by the use of national death certificate information only for mortality data.

Study Details

A good-quality prospective cohort study involved a random subset of 5,151 veterans with normal aortas at initial screening who were offered rescreening at 4 years (subset of the ADAM trial) and given questionnaires about risk factor information.101 Of those invited, 2,622 were rescreened. Participant information on 17 characteristics was collected: age, sex, race, height, weight, waist circumference, family history, smoking history, hypertension, hypercholesterolemia, coronary artery disease, claudication, cerebrovascular disease, deep venous thrombosis, diabetes, COPD, and nonskin cancer. This cohort study compared characteristics of the 58 subjects in whom new AAAs (≥3 cm) were detected with the 2,564 participants who did not have AAAs diagnosed at the 4-year followup; risk factor information on nonattendees, including the deceased, was also reported. Most of the detected AAAs were 3 to 3.4 cm (n=45), with a few in the 4- to 4.9-cm range (n=3). Nearly 3 percent (2.7%) of patients with AAA detected at the second screening had a positive family history of AAA compared with 6.0 percent of those without AAA. Less than 4 percent (3.4%) of participants with AAA detected at the second screening were black compared with 7.0 percent without AAA. Eighty-six percent of those with AAA detected were ever smokers (>100 cigarettes over a lifetime) compared with 74 percent in the group without AAA detected. Thirty-six percent of those with AAA detected were current smokers. Fifty-eight percent of those with AAA detected had any atherosclerosis compared with 42 percent without AAA detected. None of the 58 patients with AAA detected at rescreening were women. Mean age of those with AAA detected was similar to the mean age of those without AAA (67.3 and 66.0 years, respectively).

A univariate logistic model for predicting new AAA found the following characteristics to be significant: ever smoked (OR, 2.20 [95% CI, 1.04 to 4.66]), number of years smoked (OR, 1.26 [95% CI, 1.08 to 1.47]), current smoker at the time of initial screening (OR, 3.31 [95% CI, 1.92 to 5.72]), coronary artery disease (OR, 1.73 [95% CI, 1.03 to 2.91]), and any atherosclerosis, defined as coronary disease, cerebral vascular disease, or claudication (OR, 1.93 [95% CI, 1.14 to 3.28]; this composite factor resulted from subsequent reanalysis). A step-wise multivariate logistic model using significant factors from the univariate analysis resulted in three remaining significant factors: current smoker (OR, 3.09 [95% CI, 1.74 to 5.5]), coronary artery disease (OR, 1.81 [95% CI, 1.07 to 3.07]), and any atherosclerosis (OR, 1.97 [95% CI, 1.16 to 3.35]). There was one AAA repair (likely from a false-negative result on initial screening) and no AAA deaths reported in any of the participants with a normal aorta, although this outcome came from national death records.

The smaller, fair-quality cohort study followed 223 patients selected from one of the Department of Veterans Affairs (VA) sites of the ADAM trial with normal aortas (<3 cm) on initial screening with yearly scans. The study found that 114 (63%) of these patients developed an aneurysm measuring larger than 3 cm at the mean 5.9-year followup, with the vast majority of detected AAAs being small (106/114 were 3.9 cm; 3/114 were >5 cm).96 Multivariate logistic regression analysis did not identify any risk factors associated with the development of AAA. No AAA repairs, ruptures, or deaths were reported in those with initially normal aortas, although cause of death was obtained from death certificates only.

One fair-quality cohort study examined risk of AAA-related death in the 166 patients who developed AAA with rescreening (out of 4,308 with an initial normal scan who were rescreened) following an initial normal scan (<3.0 cm) and found no association between age group in this group and AAA death.99

KQ 2b. Does the Effect of Rescreening Vary Between Different Time Intervals?

In the five fair-quality and one good-quality prospective cohort studies96-99,101,102 and one fair-quality case-control study,100 screening occurred at various intervals after the initial normal scan, as follows: 3 months to annually,98 annually,96 at 5 years and 12 years,102,103 every 2 years for up to 10 years or once after 5 years,99 once at 4 years,101 once at 3 to 5 years,100 and every 2 years (Appendix G Table 1).97 It is not possible to draw conclusions about the effect of screening frequency on health outcomes based on this small collection of heterogeneous studies with few, if any, reported AAA-related deaths.

KQ3. What Are the Harms Associated With One-Time and Repeated AAA Screening?

Summary of Results

All four large population-based screening RCTs (two fair-quality, two good-quality) provide information on operative mortality and number of AAA surgeries (Table 1).13,14,94,95 Meta-analysis of available data at each time point (3 to 5 years, 6 to 7 years, 10 to 11 years, and 13 to 15 years) showed up to a doubling of risk for any AAA-related operation in the invited group, driven by more elective surgeries. The risk of emergency surgery was halved in the invited group compared with the control group. Thirty-day postoperative mortality after elective surgery was similar in screened and control groups, but was significantly reduced after emergency surgery in the screened group compared with the control group, for all time periods up to 10-year followup.

Five small observational studies (two cohort, one case-control, two cross-sectional) report conflicting results on the influence of AAA screening on quality of life and anxiety/depression measures. One study reported short-term decreases in quality of life at 12 months104 in those who screened positive for AAA, while four studies showed no clinically important decline in quality of life measures in those who screened positive compared with unscreened controls.13,95,105,128

Three cohort studies94,96,99 reported number of procedures, finding that relatively few (0.5%) of those with initially normal aortas will require elective or emergency surgery over 5 years (Table 1).

Study Details

One-Time Screening Harms

30-day postoperative mortality: all AAA surgeries, 3- to 5-year followup. Two of the four major screening RCTs (MASS, Chichester)13,14 showed a statistically significant benefit in the invited group compared with the control group, while two trials showed no statistically significant difference at the 3- to 5-year followup period. Pooled data from the four trials using a random-effects model was performed, showing a point estimate of RR of 0.32 (95% CI, 0.21 to 0.48) (Figure 6) for 30-day mortality at 3 to 5 years. The fixed-effects model produced similar findings.

Figure 6 is a forest plot depicting the relative risk for 30-day mortality in one-time screening trials. This figure is discussed in the Results section.

Figure 6

Pooled Analysis of 30-Day Mortality in One-Time Screening Trials (Random-Effects Model).

30-day postoperative mortality: all AAA surgeries, 7- to 15-year followup. MASS showed a similar decrease in 30-day mortality in the invited group at 7-, 10.1-, and 13.1-year followup (7-year RR, 0.32 [95% CI, 0.21 to 0.48], 10.1-year RR, 0.37 [95% CI, 0.25 to 0.54], 13.1-year RR, 0.46 [95% CI, 0.33 to 0.65]) (Figure 6).90 Pooled results at 13 to 15 years, weighted mostly by the MASS results, showed a reduced 30-day postoperative mortality in the invited group (RR, 0.46 [95% CI, 0.34 to 0.63]).

30-day postoperative mortality: elective and emergency surgeries. Pooled data from MASS and the Western Australian trial showed no difference in 30-day mortality from elective surgery between the invited and control groups at 3- to 5-year followup (RR, 0.70 [95% CI, 0.35 to 1.41]) (Figure 7).13 Results from the fixed-effects model yielded identical results. MASS reported 7- and 10-year outcomes, showing no difference in 30-day mortality from elective surgery between the invited and control groups.89,90

Figure 7 is a forest plot depicting the relative risk for 30-day mortality due to elective surgery in one-time screening trials. This figure is discussed in the Results section.

Figure 7

Pooled Analysis of 30-Day Mortality Due to Elective Surgery in One-Time Screening Trials (Random-Effects Model).

Pooled data from MASS and the Western Australian trial at 3- to 5-year followup showed reduced 30-day mortality from emergency surgery in the invited group (RR, 0.15 [95% CI, 0.07 to 0.32]) (Figure 8). Results from the fixed-effects model yielded similar results. MASS reported reduced 30-day mortality from emergency surgery at 7- and 10.1-year followup (7-year RR, 0.17 [95% CI, 0.09 to 0.31]; 10.1-year RR, 0.22 [95% CI, 0.13 to 0.36]).89,90 MASS showed no difference in 30-day postoperative mortality between the groups at the 13.1-year followup.129

Figure 8 is a forest plot depicting the relative risk for 30-day mortality due to emergency surgery in one-time screening trials. This figure is discussed in the Results section.

Figure 8

Pooled Analysis of 30-Day Mortality Due to Emergency Surgery in One-Time Screening Trials (Random-Effects Model).

Number of AAA operations. In all trials, there were consistently more AAA-related operations in the invited group compared with the control group (Table 1). At 3 to 5 years, pooled analysis using a random-effects model from the four trials showed a doubling of any AAA-related operations in the screened group (RR, 2.16 [95% CI, 1.84 to 2.53]) (Figure 9). For every 1,000 individuals invited to be screened, five more underwent any AAA-related operation over the next 3 to 5 years (data not shown). Based on pooled data from three trials (MASS, Viborg, and Chichester), increased risk of AAA-related surgery in the screened group remained after 10 to 11 years, although the effect was somewhat diminished (RR, 1.57 [95% CI, 1.35 to 1.83) (Figure 9). At 13 to 15 years, pooled data from three trials (Chichester, MASS, and Viborg) were similar to the 10- to 11-year time period (RR, 1.54 [95% CI, 1.38 to 1.72]). Sensitivity analysis using the fixed-effects model yielded similar results.

Figure 9 is a forest plot depicting the relative risk for AAA operations in one-time screening trials. This figure is discussed in the Results section.

Figure 9

Pooled Analysis of AAA Operations in One-Time Screening Trials (Random-Effects Model).

Number of elective operations. Elective operations were more common in the screened group in every trial at each of the followup time points (Table 1, Figure 10), with pooled relative risks for the various time points ranging from 3.25 (95% CI, 2.13 to 4.96) at 3 to 5 years to 2.07 (95% CI, 1.53 to 2.79) at 15 years. Heterogeneity was low at the 10- to 11-year and 13- to 15-year time periods. Based on pooled data at 3 to 5 years, 92 percent of all AAA operations in the screened group were elective compared with 64% of AAA-related operations in the control group (data not shown).

Figure 10 is a forest plot depicting the relative risk for elective operations in one-time screening trials. This figure is discussed in the Results section.

Figure 10

Pooled Analysis of Elective Operations in One-Time Screening Trials (Random-Effects Model).

Number of emergency operations. Individual trial results varied, with most suggesting reduced emergency AAA operations in those invited for screening compared with controls over 5 to 15 years of followup (Table 1, Figure 11). Pooled results suggested that screening halved the risk of emergency surgery at 3 to 5 years (RR, 0.50 [95% CI, 0.29 to 0.86]), with some heterogeneity in individual RCT results (I2=39%). The Western Australian trial alone reported a nonsignificant increased risk of emergency operations in the screened group at 3.6 years; this study also had a much higher proportion of elective surgeries (87%) in its control group compared with the other three trials (pooled percentage, 57%) (data not shown), which changed the relative effect of emergency operations between arms. By 13 to 15 years of followup, differences in emergency operations remained significant, and heterogeneity was low at this time point. Sensitivity analysis using the fixed-effects model yielded similar results.

Figure 11 is a forest plot depicting the relative risk for emergency operations in one-time screening trials. This figure is discussed in the Results section.

Figure 11

Pooled Analysis of Emergency Operations in One-Time Screening Trials (Random-Effects Model).

Quality of life. One cohort, one augmented case-control, and one subsampling study were constructed from MASS,13 Western Australian,95 and Viborg94 trial participants to address quality of life (data not shown). Two other observational comparisons were constructed from the Gloucestershire Aneurysm Screening program,105 a regional Swedish screening program,104 and one small screening study set in rural Australia.128 The Viborg study was excluded, as it did not report baseline quality of life to allow adjusted comparisons among subgroups with and without AAA and with and without screening attendance.94 The studies reported quality of life outcomes using different questionnaires, including the Short-form 36-item Health Survey (SF-36), Screen Quality of Life, European Quality of Life-Five Dimension (EuroQOL-5D), and the General Health Questionnaire (GHQ).13,94,95,104,105,128 The SF-36 (range, 0 to 100) is a self-administered questionnaire evaluating eight domains: vitality, physical functioning, bodily pain, general health perceptions, physical role functioning, emotional role functioning, social role functioning, and mental health. The EuroQOL-5D (range, 0 to 100) is a generic measure of health status that provides a descriptive profile and a single index value that can be used in the clinical and economic evaluation of health care and in population health surveys. This self-administered questionnaire asks individuals to evaluate five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. A lower number denotes poorer quality of life. Questionnaires were administered at various time points, including prior to screening, after screening, or in selected subgroups of those with screen-detected AAA undergoing surgery or surveillance. Timing of questionnaires was no longer than 12 months after a specific event.

Prescreening quality of life. Baseline quality of life was gathered in 2009 from participants recruited to the Western Australian study during a specified period. Even prior to screening, those eventually found to have a small AAA had lower mean age-adjusted self-perceived general health on the EuroQOL-5D compared with those with normal aortas.95 In the Gloucestershire Screening program, the GHQ was completed before screening and again after 1 month by 61 men with screen-detected AAA and 100 consecutively screened men with normal aortas.105 No difference in overall score was seen in those with and without AAA, before or after screening, although anxiety decreased significantly in both groups 1 month after screening. Prior to undergoing ultrasound screening in a regional Swedish screening program, men and women completed the SF-36.104 No baseline differences were seen in any SF-36 scale scores between men and women with AAA (n=27) and screen-negative age- and sex-matched controls. In the Australian study, the baseline SF-36 scores were not different between the men who subsequently screened positive or negative.128

Postscreening quality of life. Six weeks after screening, MASS compared depression, state anxiety, and quality of life using the SF-36, EQ-5D, Hospital Anxiety and Depression Scale, and Spielberger state/trait anxiety scales in a subset of those who screened positive for AAA, who screened negative, and who were not invited to screening. Those who screened negative showed no worse scores than unscreened controls, although those with screen-detected AAA had significantly worse anxiety, physical health, mental health, and self-rated health and health index quality of life scores than screen-negative participants. Even among those screening positive, anxiety and depression scores were well under clinical cutoffs, and quality of life scores were similar to those of unscreened controls. Longer-term data on the impact of screening on quality of life were not available, although physical and mental health scores and weighted health index scores in screen-positive patients undergoing surveillance or surgery tended to rebound after 3 months. In the Gloucestershire Screening program, no differences in overall GHQ scores were seen in those with and without AAA 1 month after screening, and anxiety was significantly reduced from baseline in both groups.105 In the Western Australian trial, those with and without AAA had similarly sized, nonsignificant improvements in quality of life from baseline to 12 months after screening. In Swedish men and women, SF-36 scores were not different 12 months after screening in those with and without AAA; however, those with AAA showed significant decreases from baseline in physical functioning, social functioning, and mental health.104 In the rural Australian study, SF-36 scores were not different 6 months after screening in those with and without AAA, but only the screen-negative group had a statistically significant improvement in the SF-36 dimensions of general health, social function, and freedom from bodily pain.128 Small numbers make these results imprecise.

Rescreening Harms

No comparative data consider the effect of rescreening versus no rescreening on those without AAA on initial screening. Small cohort studies suggest that relatively few (0.5%) of those with initially normal aortas will require elective or emergency surgery over 5 years (Table 3). No studies examined quality of life outcomes for rescreening.

AAA surgeries. Three fair-quality cohort studies examined procedure rates in rescreened cohorts, showing that this rate varied from 0 to 0.5 percent (Table 3).94,96,99 The largest of these three cohorts, with a 5-year followup, reported 17 elective and six emergency surgeries (out of 4,308 rescreened). 99

30-day mortality. In the Hafez cohort study, six participants (out of 23 undergoing AAA operation [26.1%]) died within 30 days of surgery (Table 4).99

KQ 4. What Is the Effect of Pharmacotherapy Versus Placebo or Surgery (Open and EVAR) Versus Surveillance on Treatment-Relevant Intermediate Health Outcomes in an Asymptomatic Population With Small AAA Identified by Screening?

Summary of Results

Eight RCTs assessed the effects of early surgery versus surveillance (k=4)106,108,113,115 and pharmacotherapy versus placebo (k=4)116,118-120 in the treatment of patients with small AAA. In the two good-quality surgical trials comparing early open surgery with surveillance,106,108 all-cause and AAA-related mortality at 5 years did not differ between the two approaches, although early open surgery significantly reduced the 5-year rupture rate (18 fewer AAA ruptures per 1,000 individuals treated with open surgery rather than surveillance). From available data, mortality effects after 8 and 12 years were unchanged;41,109 very limited data evaluated subgroup differences, with no evidence of treatment differences in age- or aneurysmal diameter-specific subgroups. In the two fair-quality trials comparing early EVAR with surveillance,113,115 shorter-term followup (at around 2 years) suggested no mortality or rupture benefit from early EVAR (leading to early cessation of both trials). A single good-quality drug trial found no significant effect on mortality or AAA growth rate after 2 years of beta-blocker use;116 one good-quality and two fair-quality trials of different antibiotics used for 4 to 15 weeks also found no significant effect on mortality and found small, somewhat mixed effects on AAA growth rates, which are difficult to interpret.118-120

Study Details

Early Open Surgery Versus Surveillance

Two good-quality RCTs of early open surgery (UKSAT and ADAM),106,108 conducted in the United Kingdom and the United States, enrolled a large number of patients (1,090 in UKSAT; 1,136 in ADAM) with small AAA (4.0 to 5.4 cm) and randomized them to early open surgery or surveillance (Appendix H Table 1). Both studies actively managed patients for a mean of approximately 5 years (4.6 years in UKSAT; 4.9 years in ADAM). In addition to 5-year followup at the end of active management, UKSAT reported results at 8 and 12 years.109,130 Over 99 percent of patients were followed after 12 years in UKSAT, and approximately 86 percent in the ADAM trial after 5 years. The primary outcome for both trials was all-cause mortality, and secondary outcomes were cost (UKSAT) or AAA-related mortality reported by an independent outcomes committee (ADAM).

Important patient characteristics, such as age and smoking history, were comparable between the two studies, although UKSAT included a higher proportion of female patients (17.5% vs. 0.8%), lower rate of hypertension (39% vs. 56.4%) and ischemic heart disease (14% probable ischemic heart disease vs. 41.9% coronary disease), and fewer patients with diabetes (2.5% vs. 9.8%) (Appendix H Table 2). Mean AAA diameter at baseline was similar for the two studies (4.6 and 4.7 cm). For patients randomized to the early surgery group, procedures were undertaken 6 to 12 weeks after randomization, with 520 patients (92.4%) in UKSAT and 527 (92.6%) in the ADAM trial receiving procedures after a mean followup period of approximately 5 years. In the surveillance group, patients received an ultrasound every 3 to 6 months and were referred to surgery when the AAA diameter reached 5.5 cm, when the growth rate exceeded 0.7 cm in 6 months or 1 cm per year, or when they developed symptoms. By the end of 5-year followup, 321 (60.9%) patients in the UKSAT surveillance group and 349 (61.6%) in the ADAM surveillance group had undergone open surgical repair (Table 5).

Table 5. AAA Growth Rate, Rupture, and Surgery Data for Open Surgery vs. Surveillance Trials for Small AAA (KQs 4 and 5).

Table 5

AAA Growth Rate, Rupture, and Surgery Data for Open Surgery vs. Surveillance Trials for Small AAA (KQs 4 and 5).

The two RCTs of early open surgery versus surveillance (UKSAT, ADAM) reported outcome data at 5-year followup.106,108 UKSAT, however, did not directly report AAA-related mortality. To determine this, we combined deaths from ruptured AAA data and 30-day operative mortality data. Although time-to-event data analyses were conducted in both trials, the effect estimates (i.e., HR) were available for pooling only for all-cause mortality.

The effects on all-cause mortality, AAA-related mortality, and rupture were consistent between the two trials (Figure 12, Table 6). At 5 years of followup, the ADAM trial reported similar findings between treatment groups in all-cause mortality (25.1% vs. 21.5%; HR, 1.21 [95% CI, 0.95 to 1.54]) and AAA-related mortality (3.0% vs. 2.6%; HR, 1.15 [95% CI, 0.56 to 2.31]).106 UKSAT found an increase in all-cause mortality in those undergoing surveillance, though the difference was not significant (30.6% vs. 46.7%; HR, 0.91 [95% CI, 0.72 to 1.16]).108 Pooling the two trials suggested no statistically significant difference in all-cause mortality (RR, 1.07 [95% CI, 0.91 to 1.25]) or AAA-related mortality (RR, 0.93 [95% CI, 0.64 to 1.37]) (Figure 12) between early open surgery and surveillance. Early open surgery did, however, significantly lower rupture risk (RR, 0.28 [95% CI, 0.13 to 0.62]), with 18 fewer AAA ruptures per 1,000 individuals treated with open surgery rather than surveillance after 5 years.

Figure 12 is a forest plot depicting the relative risk for all-cause mortality, AAA-related mortality, and rupture in early open surgery versus surveillance trials at 5-year followup. This figure is discussed in the Results section.

Figure 12

Pooled Analysis of All-Cause Mortality, AAA-Related Mortality, and Rupture in Open Surgery vs. Surveillance at 5-Year Followup (Random-Effects Model).

Table 6. All-Cause and AAA-Related Mortality Data for Open Surgery vs. Surveillance Trials for Small AAA (KQ 4).

Table 6

All-Cause and AAA-Related Mortality Data for Open Surgery vs. Surveillance Trials for Small AAA (KQ 4).

UKSAT further reported outcome data at 8 and 12 years of followup.41,109 The results consistently showed no statistically significant difference in all-cause mortality (43.0% vs. 48.2% at 8 years; 64.3% vs. 66.8% at 12 years; RR, 0.96 [95% CI, 0.88 to 1.05]) and AAA-related mortality (9.6% vs. 7.0% at 8 years; 9.6% vs. 9.5% at 12 years; RR, 0.73 [95% CI, 0.49 to 1.09]) (Table 6). Early open surgery was reported to reduce the risk of rupture (1.8% vs. 4.3% at 8 years; 2.3% vs. 4.5% at 12 years). However, the relative magnitude of effect decreased over time due to a very similar, proportional increase in ruptured aneurysms in both arms, suggesting the reduced rupture benefit with open surgery occurs within the first 5 years (Figure 12).

The data for exploring heterogeneity across studies were very limited. Sensitivity analyses using reported HRs confirmed no statistically significant difference in all-cause mortality in early open surgery versus surveillance. The use of HRs, however, suggested higher heterogeneity (I2=63.1%) than the use of risk ratios (I2=21.4%), which might suggest a difference in the timing of deaths between the two trials (Appendix K). The use of alternative statistical models (random- vs. fixed-effects) did not show any significant change in 5-year effect estimates (Appendix K).

Early EVAR Versus Surveillance

Two medium-size, fair-quality, industry-funded RCTs (Comparison of Surveillance Versus Aortic Endografting for Small Aneurysm Repair [CAESAR],113 Positive Impact of Endovascular Options for Treating Aneurysm Early [PIVOTAL]115) were undertaken to evaluate the effect of early EVAR in patients with small aneurysms. CAESAR (N=360) measured the primary outcome of all-cause mortality and secondary outcomes of AAA-related mortality, rupture, growth, perioperative mortality, and conversion to open repair, while PIVOTAL (N=728) measured the primary outcome of AAA-related mortality and rupture and secondary outcomes of all-cause mortality and AAA-related mortality in smokers versus nonsmokers and conversion to open repair. Trials were conducted in the United States and Italy and randomized patients with small AAA (CAESAR, 4.1 to 5.4 cm; PIVOTAL, 4.0 to 5.0 cm) to undergo early EVAR versus surveillance (Appendix H Table 2). The CAESAR trial reported results at a median followup of 2.7 years, and the PIVOTAL trial reported results at a mean followup of 1.7 years. Notably, both RCTs conducted interim analyses and found that detection of meaningful difference in primary outcomes between EVAR and surveillance was unlikely if patient enrollment were to continue (i.e., futility).113,115 Thus, both trials subsequently stopped recruiting patients early, but they completed scheduled followup in those who had already been enrolled. Likely due to early stopping of enrollment, the two studies did not adequately achieve balance between randomized arms in important prognostic factors, such as family history, sex, and diabetes.

While the mean age (70.5 vs. 68.9 years) and AAA diameter (4.45 vs. 4.72 cm) were similar between the trials, there was a higher proportion of smoking patients and patients with coronary artery disease in the PIVOTAL trial (91.0% vs. 55.3% and 55.4% vs. 39.2%, respectively) (Appendix H Table 2).115 Both studies randomized patients to undergo EVAR within 30 days or to surveillance (Appendix H Table 1). Three hundred and twenty-two patients (88.9%) allocated to the early EVAR group underwent EVAR procedures in the PIVOTAL trial, and 171 (94.0%) in CAESAR (Table 7). In each of the studies, four patients received open surgery instead of EVAR in the early EVAR group. Patients in the surveillance group underwent assessment every 6 months and were offered EVAR when AAA diameter reached 5.5 cm, the growth rate exceeded 0.5 cm in 6 months or 1 cm per year, or they developed symptoms. Among patients randomized to surveillance, 71 (39.9%) in the CAESAR trial and 108 (30.1%) in the PIVOTAL trial received EVAR by the end of followup.

Table 7. AAA Growth Rate, Rupture, and Surgery Data for EVAR vs. Surveillance Trials for Small AAA (KQs 4 and 5).

Table 7

AAA Growth Rate, Rupture, and Surgery Data for EVAR vs. Surveillance Trials for Small AAA (KQs 4 and 5).

Both trials of early EVAR versus surveillance (CAESAR, PIVOTAL) reported the effect on all-cause mortality, AAA-related mortality, and rupture (Tables 7 and 8).113,115 At the end of followup, early EVAR did not achieve a statistically significant reduction in the risk of all-cause mortality, AAA-related mortality, or rupture in either of the two RCTs compared with surveillance, but the number of events was generally small, with only zero to two events (AAA-related death or rupture) in each group reported in both studies (Figure 13, Tables 7 and 8). Both trials reported an all-cause mortality rate of approximately 4 percent, with similar results seen across treatment interventions (Table 8). Similarly, AAA-related mortality was found to be very similar across treatment arms and ranged from 0.3 to 0.6 percent (Table 8). Pooling data on all-cause and AAA-related mortality confirmed no difference in mortality between strategies, but pooled estimates were both in the direction of greater mortality with early EVAR (RR, 1.07 [95% CI, 0.62 to 1.86] for all-cause mortality; RR, 1.46 [95% CI, 0.24 to 8.94] for AAA-related mortality) (Figure 13). Both trials reported zero ruptures in the early EVAR group and low numbers in those receiving surveillance (0.3% and 1.1%) (Table 8). Pooling data on ruptures confirmed no difference in rupture rates between strategies (RR, 0.25 [95% CI, 0.03 to 2.26]) (Figure 13).

Table 8. All-Cause and AAA-Related Mortality Data for EVAR vs. Surveillance Trials for Small AAA (KQ 4).

Table 8

All-Cause and AAA-Related Mortality Data for EVAR vs. Surveillance Trials for Small AAA (KQ 4).

Figure 13 is a forest plot depicting the relative risk for all-cause mortality, AAA-related mortality, and rupture in EVAR versus surveillance trials at 5-year followup. This figure is discussed in the Results section.

Figure 13

Pooled Analysis of All-Cause Mortality, AAA-Related Mortality, and Rupture in EVAR vs. Surveillance Trials (Random-Effects Model).

There was no statistically significant heterogeneity across the two trials (I2=0% for all outcomes). Sensitivity analyses conducted with statistical models (random- vs. fixed-effects) did not show any significant change of the estimates (Appendix K).

Pharmacotherapy Versus Placebo

One good-quality and two fair-quality placebo-controlled parallel RCTs investigated the effectiveness of antibiotics (doxycycline, azithromycin, roxithromycin),118-120 and one good-quality RCT investigated the effectiveness of the beta-blocker propranolol116 compared with placebo for small AAA for the following outcomes: delay of AAA growth (primary outcome), all-cause mortality, AAA rupture, and surgery (Appendix H Table 1). These trials, conducted in Finland, Demark, Sweden, and Canada, recruited participants from vascular referral centers as well as community/population screening programs, had varying sample sizes with small antibiotic trials (34 to 211) and a single larger beta-blocker trial (N=552), and followed patients for 1.5 to 2.5 years. One study additionally reported results at 5 years of followup.121 Only the doxycycline trial reported blinding outcome assessors.118

Important differences existed in patient characteristics across the four RCTs (Appendix H Table 2). Although the inclusion criteria were generally consistent, two trials included patients with AAA diameters of 3.0 to 5.0 cm,116,120 one trial included AAA diameters of 3.0 to smaller than 5.5 cm,118 and the other one included AAA diameters of 3.5 to 5.0 cm.119 The four trials included patients of similar age (mean age, 68.4 to 72.5 years), but the proportion of female patients differed considerably between three trials (0% to 18.5%), and one trial exclusively enrolled men.120 In three trials, about one third (34% to 40.0%) of patients had a smoking history, whereas the other trial included 59.5 percent of patients who were current or ever smokers.120 The distribution of cardiovascular risk factors (e.g., the proportion of patients with hypertension, MI, or stroke) also differed across studies. Ultrasound measurements were performed using aortic anterior-posterior diameters in all trials, with three of the four trials using the larger of the axial or transverse measurement planes.118-120 Only one trial116 reported using outer-to-outer wall measurements, while the other three trials did not specify which wall measurements were used.

While three trials compared antibiotics with placebo, those antibiotics—including doxycycline, roxithromycin, and azithromycin—have different mechanisms (Appendix H Table 1).118-120 The treatment duration ranged from 4 to 15 weeks. Patients were offered surgery when meeting surgical criteria, but these criteria were inadequately described in two trials.118,119 The beta-blocker trial randomized patients to receive propranolol or placebo, with a target dose of between 80 and 120 mg twice daily, for a mean of 2.5 years.116

All three antibiotic trials reported all-cause mortality, the use of surgical procedures, and AAA growth rate (Tables 9 and 10).118-120 We pooled the three trials in order to assess the overall effect of relatively short-term antibiotic use on all-cause mortality data. The pooled estimate suggested no effect of antibiotics above placebo on reducing all-cause mortality (RR, 0.92 [95% CI, 0.43 to 1.96]) (Figure 14). Despite the differences in patient characteristics and interventions, there was no statistically significant heterogeneity (I2=0%). Pooling these trials suggested no statistically significant reduction in the use of surgical procedures for AAA in those taking antibiotics for 4 to 15 weeks compared with placebo (RR, 0.89 [95% CI, 0.51 to 1.55]) (Appendix K). The CIs for each trial were very wide, and no meaningful heterogeneity was found across studies (I2=8.4%).

Table 9. AAA Growth Rate, Rupture, and Surgery Data for Pharmacotherapy vs. Placebo Trials for Small AAA (KQs 4 and 5).

Table 9

AAA Growth Rate, Rupture, and Surgery Data for Pharmacotherapy vs. Placebo Trials for Small AAA (KQs 4 and 5).

Table 10. All-Cause and AAA-Related Mortality Data for Pharmacotherapy vs. Placebo Trials for Small AAA (KQ 4).

Table 10

All-Cause and AAA-Related Mortality Data for Pharmacotherapy vs. Placebo Trials for Small AAA (KQ 4).

Figure 14 is a forest plot depicting the relative risk for all-cause mortality in trials of antibiotics versus placebo. This figure is discussed in the Results section.

Figure 14

Pooled Analysis of All-Cause Mortality in Trials of Antibiotics vs. Placebo (Random-Effects Model).

The reporting of AAA growth rate varied across the three trials (Table 9). Two trials reported median and interquartile range of annual growth rate,118,119 whereas the other reported mean annual growth rate only.120 The results were inconsistent across trials. One study showed no improvement with azithromycin over placebo (median, 2.2 vs. 2.2 mm per year; p=0.85).119 Another study suggested a possible reduction in growth rate in patients using doxycycline versus placebo (median, 1.5 vs. 3.0 mm per year; no statistical testing reported).118 Additionally, this study found that only one (7%) patient in the doxycycline group compared with five (41%) in the placebo group had an annual growth rate of 5 mm per year or more during 1.5 years of followup (p=0.06). In the third study, roxithromycin significantly reduced the growth rate compared with placebo at both 2 and 5 years of followup (mean, 1.56 vs. 2.75 mm per year at 2 years; p=0.02; 1.16 vs. 2.52 mm per year at 5 years; p=0.06);120 nevertheless, the difference was too small to be clinically important, and care providers who measured AAA diameter were not blinded in those trials, which is a potential threat to the accuracy of measurement.

In the trial of propranolol versus placebo, there was no statistically significant reduction in all-cause mortality (12% vs. 9.6%; p=0.36), AAA-related mortality (0.7% vs. 0.7%; p=1.0), rupture (0.4% vs. 0.7%; p=0.25), or AAA growth rate (mean, 0.22 vs. 0.26 cm per year; p=0.11) (Table 10).116

Sensitivity analyses that compared random- versus fixed-effects models did not show any significant difference in the effect estimates (Appendix K).

KQ 4a. Does the Effect of Pharmacotherapy, Surgery, and Surveillance Differ Between Men and Women, Smaller and Larger Aneurysms, Smokers and Nonsmokers, Older and Younger Patients, Patients With and Without a Family History of AAA, Patients With and Without Diabetes, Patients With and Without COPD, and Patients of Different Races/Ethnicities?

Summary of Results

At 5 years, two good-quality RCTs reported all-cause mortality by subgroups of age and AAA diameter, showing no significant differences.106,108 One good-quality RCT reported no sex-specific subgroup difference in all-cause mortality.108,103

Study Details

Open Surgery Versus Surveillance: Subgroups by Sex, AAA Diameter, and Age

Sex. Only one trial of early open surgery versus surveillance reported all-cause mortality data by sex and found no significant sex-specific subgroup differences (adjusted HR for men, 0.9 [95% CI, 0.76 to 1.06]; adjusted HR for women, 0.89 [95% CI, 0.62 to 1.28]; p=0.76).108,103 Through 12 years of followup, UKSAT found similar numbers of deaths in male and female participants in the early surgery group (8.0% in men vs. 8.4% in women) and slightly more deaths among women than men in the surveillance group (8.5% in men vs. 10.0% in women).

AAA diameter. When considering mortality by AAA diameter through 5 years of followup, both trials reported an increase in mortality as AAA diameter increased, with no significant differences between treatment groups. ADAM reported that 21.3 percent of those in the early surgery group and 16.2 percent of those in the surveillance group with an AAA diameter of 4.0 to 4.4 cm died (RR, 1.48 [95% CI, 0.92 to 2.38]).106 Similarly, UKSAT reported more deaths in the early surgery group than in the surveillance group in those with AAAs measuring 4.0 to 4.4 cm (63 vs. 53 per 1,000 person-years; HR, 1.14; p=0.26).108 In those with AAAs measuring 4.5 to 4.8 cm, there was no difference between treatment groups in UKSAT, and there were slightly more reported deaths in those receiving surgery in the ADAM trial (RR, 1.27 [95% CI, 0.81 to 1.99]). In both trials, there were no differences between treatment groups in those with AAAs measuring 5.0 to 5.4 cm (ADAM RR, 1.02 [95% CI, 0.71 to 1.47]; UKSAT HR, 0.79; p=0.26). We were able to conduct subgroup analyses across the two trials to examine whether the effect on all-cause mortality differed by AAA diameter. Our results confirmed what was reported above and showed no statistically significant difference in effects across the three AAA diameter subgroups (p=0.92 for interaction test) (Appendix K).

Age. In the ADAM trial, across all age groups, more deaths were seen in those receiving early open surgery, with an unsurprising increase in all-cause mortality as age increased (data not shown). This trend was slightly different in UKSAT, with more deaths in the surveillance group in those ages 60 to 71 years and more deaths in the early surgery group in those ages 72 to 76 years (data not shown). The subgroup data in UKSAT could imply a possible benefit of surgery in younger patients and a possible benefit of surveillance in older patients. These differences, however, were found to be nonsignificant, so the data should be interpreted with caution. In those ages 50 to 59 years in ADAM, there were no differences in mortality between those receiving early surgery and those undergoing surveillance (RR, 1.02 [95% CI, 0.38 to 2.73]).106 Participants ages 60 to 69 years had 61 deaths in the early surgery group and 55 deaths in the surveillance group (RR, 1.34 [95% CI, 0.93 to 1.93]). In UKSAT, those ages 60 to 66 years had more deaths in the surveillance group than in the early surgery group (42 vs. 36 per 1,000 person-years; HR, 0.76; p=0.10). This trend was the same in those ages 67 to 71 years, with more deaths in those undergoing surveillance than early surgery (60 vs. 51 per 1,000 person-years; HR, 0.80; p=0.10).108 In ADAM, at ages 70 to 79 years there were 74 deaths (27.3%) in those receiving surgery and 59 deaths (24.9%) in those receiving surveillance (RR, 1.10 [95% CI, 0.78 to 1.55]). In contrast, UKSAT reported more deaths in the early surgery group in those ages 72 to 76 years than in the surveillance group (72 vs. 48; HR, 1.25; p=0.10). Given that there was no overall effect on all-cause mortality, subgroup-specific effect modification is unlikely.131

As UKSAT and ADAM did not report results by comorbidity, family history, or race, no subgroup analyses are possible.

EVAR Versus Surveillance: Subgroups

Neither of the two trials comparing early EVAR surgery with surveillance reported data on subgroup effects.113,115

Pharmacotherapy Versus Surveillance: Subgroups

Of the three trials analyzing the effectiveness of antibiotics, only one reported data on subgroup effects by treatment arm (data not shown).120 The roxithromycin trial (n=92) reported aneurysm expansion rates in each treatment group by AAA size through 2 years of followup. The results showed that roxithromycin was more effective at slowing aneurysm growth in both smaller and larger aneurysms, but not at significant levels. In aneurysms that were smaller than 3.65 cm at baseline, roxithromycin reduced aneurysm expansion compared with placebo, though not significantly (1.34 vs. 2.28 mm per year; p=0.17). This trend was similar to what was seen in aneurysms that were 3.65 cm or larger at baseline, with the difference again found to be not significant (1.76 vs. 3.27 mm per year; p=0.08).

The trial investigating the effectiveness of the beta-blocker propranolol did not report subgroup effects by treatment arm.116

KQ 5. What Harms Are Associated With Pharmacotherapy, EVAR and Early Surgery, and Surveillance in an Asymptomatic Population With Small AAA Identified by Screening?

Summary of Results

Both RCTs of early open surgery found a 50 percent increased risk of AAA-related surgical procedures in those randomized to early surgery instead of surveillance; for every 1,000 persons with small AAA managed with surveillance, 313 persons would avoid AAA-related surgery over 5 years.106,108 For those requiring surgery in the surveillance arm, there was minimal difference in surgical complications after delayed open surgical repair compared with more immediate surgery (Table 11). No difference in 30-day operative or postoperative mortality was seen between early open surgery and delayed surgery; however, compared with immediate open surgery, 45 more individuals per 1,000 with small AAA undergoing delayed open surgery after surveillance would experience any complications, and two more would experience a major complication, particularly surgery-related MI. At least over the first 1 to 2 years after small AAA discovery, health perception/overall health was improved in those undergoing early open surgery, although there was no difference in overall quality of life (Appendix I Table 1).107,132

Table 11. Harms Data in Studies of Treatment for Small AAA (KQ 5).

Table 11

Harms Data in Studies of Treatment for Small AAA (KQ 5).

As expected, those with small AAA randomized to early EVAR more than doubled their risk of undergoing AAA-related surgery over the next several (1.5 to 2.5) years compared with those undergoing aneurysm surveillance (Table 7).113,115 Between 484 and 582 out of every 1,000 persons with small AAA undergoing surveillance rather than early EVAR would be expected to avoid any AAA-related surgery during that time period. Delaying surgery until indicated after surveillance did not result in increased surgery-related harms, with some data suggesting better results in those undergoing delayed (as opposed to early) EVAR. Endoleaks were the most common complications after EVAR, occurring in 6 to 15 percent of participants in trials and one registry study. Differences in the risk of endoleaks between early and delayed EVAR were minimal, but favored delayed EVAR in one trial. Reinterventions over an unspecified time period were similar in early versus delayed EVAR in PIVOTAL, but were significantly increased after early EVAR in CAESAR, again favoring delayed EVAR (Table 11). As with early open surgery, short-term quality of life was significantly improved at 6 months in those undergoing early EVAR, but quality of life differences between approaches were not maintained after about 3 years of followup (Appendix I Table 1).114

Propanolol use more than doubled medication discontinuation or patient dropout compared with placebo (37.7% vs. 21.3%; p <0.0001 in the Propranolol Aneurysm Trial [PAT]; 60% vs. 40%; p-value not reported in the Danish trial), while relatively short-term antibiotic use had few reported harms or patient dropouts (0% to 2.4%) (Table 11).116,117

Study Details

Harms Associated With Early Open Surgery Versus Surveillance

Receipt of surgical procedures. At 5-year followup, approximately 93 percent of patients in the early open surgery group had received AAA repair in both trials, as opposed to approximately 61 percent in the surveillance group (Appendix K, Table 5).106,108 The effect estimates were nearly identical between the two studies, and the pooled estimates (RR, 1.51 [95% CI, 1.44 to 1.59]) suggested that, at 5 years after randomization, 313 per 1,000 persons with small AAA managed with surveillance rather than open surgery would avoid any surgical procedure for AAA repair. By 12-year followup, UKSAT reported an additional 14 percent of the control group receiving AAA surgery. Nearly all surgeries (98% to 99%) were elective in both groups.

Operative mortality. At 5-year followup, the 30-day operative mortality in the early open surgery group of UKSAT was 5.8 percent (n=520) compared with 7.1 percent (n=321) in the surveillance group (Table 11).108 In ADAM, 30-day operative mortality at 5 years was 2.0 percent (n=526) in the surgery group compared with 1.8 percent (n=340) in the surveillance group.106 Pooling the data suggested no statistically significant difference in 30-day operative mortality between the two strategies (RR, 0.86 [95% CI, 0.54 to 1.36]) (Figure 15). In UKSAT, through 8 years of followup, each group had similar 30-day postoperative mortality rates compared with their respective 4.6-year followup rates. This suggests that delayed surgery did not alter 30-day postoperative mortality.

Figure 15 is a forest plot depicting the relative risk for 30-day and postoperative mortality in early open surgery versus surveillance trials. This figure is discussed in the Results section.

Figure 15

30-Day and Postoperative (30-Day Plus In-Hospital) Mortality in Early Open Surgery vs. Surveillance Trials (Random-Effects Model). * The number of events in UKSAT was estimated based on the reported proportion that was adjusted for age and sex.

The results for postoperative mortality (adding in-hospital deaths after 30 days to 30-day operative mortality) were similar to 30-day mortality, with pooled results suggesting no statistically significant difference between the two strategies (RR, 0.86 [95% CI, 0.54 to 1.37]) (Appendix K).

Surgical complications. The ADAM trial, but not UKSAT, reported 30-day readmissions and (nonfatal) complications associated with AAA repair in both groups (Table 11).106 Findings were limited by relatively few participants and low individual event rates. Patients in the early open surgery group tended toward a slightly higher rate of 30-day readmissions (20.5% vs. 16.5%), but had a significantly lower risk of any surgical complications (52.3% vs. 56.8%; p=0.026). Nonetheless, the event rate for total major complications was higher in the surveillance group (4.4% to 7.9%), with a significantly higher risk of surgery-related MI reported (1.0% vs. 3.8%; p=0.0051).

Quality of life. Both UKSAT and ADAM reported quality of life, although only UKSAT reported numerical data (Appendix I Table 1). UKSAT reported change in quality of life, measured by the Medical Outcomes Study 20-item Short-form Health Survey, 1 year after randomization.132 This questionnaire measures multiple domains of patient health, including physical functioning, role functioning, social functioning, metal health, current health perception, and bodily pain. The total score ranges from 0 to 100, with a higher score indicating better health. In patients of both groups, quality of life appeared to decrease 1 year after randomization across all domains (1- to 6-point decrease), except that health perception was improved in the early surgery group (approximate 6-point increase). There was no statistically significant difference in the change of quality of life from baseline between early open surgery and surveillance in most domains, although patients in the early surgery group had better health perception after 1 year. The difference in mean change of health perception was 6.7 points (95% CI, 3.41 to 9.99) between groups, suggesting a clinically meaningful improvement.133

The ADAM trial measured patients' quality of life using the SF-36, and collected data every 6 months until the end of followup.107 Patients were followed for 3.5 to 8 years (mean, 4.9 years), and approximately 86 percent of patients completed followup. The study analyzed change in quality of life over time using a repeated-measures model and adjusting for baseline measurements. The authors reported a statistically significant decrease in all SF-36 subscales over time for the entire population (p<0.001). Although numerical values are not reported, they are presented graphically. No difference, however, was found between the early repair group and the surveillance group in all SF-36 subscales, except that the early repair group had statistically higher general health scores (p<0.001). This difference was due mainly to significantly higher scores during time points between 6 months and 2 years (p<0.05).

The ADAM trial also reported rates of impotence among participants.107 In a repeated-measures analysis, there was a statistically higher risk of developing impotence over time in patients with early repair than those undergoing surveillance (p<0.03). The statistically significant difference occurred between 18 months and 4 years. No numerical values were reported.

Harms Associated With Early EVAR Versus Surveillance

Two registry studies assessing EVAR reported harms data,122,123 in addition to the two RCTs of early EVAR versus surveillance described in KQ 4 (Appendix H Table 1).113,115 The two fair-quality registry studies, conducted in Australia (Australian Safety and Efficacy Register of New Interventional Procedures-Surgical [ASERNIP-S])122 and Europe (European Collaborators on Stent/Graft Techniques for Aortic Aneurysm Repair [EUROSTAR]),123 prospectively collected data from patients with AAA who underwent EVAR. The ASERNIP-S study collected perioperative and intermediate outcome data associated with EVAR from Australia's national audit and included more than 90 percent of procedures performed. The EUROSTAR study documented all patients undergoing EVAR with commercially available devices approved in continental Europe. Both studies reported results according to the size of AAA at baseline, thus allowing the assessment of harms associated with EVAR in patients with small AAA. The ASERNIP-S study enrolled 478 patients with small AAA (≤5.5 cm), and the EUROSTAR study included 4,392 patients, of which 1,962 had small AAA (4.0 to 5.4 cm). The median length of followup was 3.2 years (interquartile range, 2.4 to 3.7 years) in ASERNIP-S and approximately 1.7 years (range, 1 month to 8 years) in EUROSTAR.

Use of surgical procedures. After 1.7 years of followup, endovascular procedures were undertaken in more than 89 percent of patients in the early EVAR group in the PIVOTAL trial and 96 percent at 2.5 years in the CAESAR trial (Table 7).113,115 In the surveillance group, EVAR was undertaken in 31 percent of patients in PIVOTAL and 48 percent in CAESAR. Both studies found a statistically significant increase in the receipt of surgical procedures in the early EVAR group, with a greater increase suggested in PIVOTAL than in CAESAR (pooled RR, 2.41 [95% CI, 1.68 to 3.45]) (Appendix K). Using the pooled estimate, 549 of every 1,000 individuals with small AAA managed with surveillance rather than early EVAR would avoid any surgical procedure for AAA in 1.5 to 2.5 years. Given that the two trials had quite different estimates (reflected in high heterogeneity when pooled), the estimate for the number avoiding surgery could range from a low of 484 per 1,000 (as seen in CAESAR) to a high of 582 per 1,000 (as seen in PIVOTAL).

Operative mortality. In both trials, 30-day operative mortality after EVAR was rare; only one patient died in the early EVAR group in each trial (0.3% in PIVOTAL; 0.6% in CAESAR), while one patient undergoing repair in the surveillance group of PIVOTAL died and none died in CAESAR (Table 11).113,115 Because of this low event rate, the CI of the pooled estimates was very wide (RR, 0.63 [95% CI, 0.08 to 5.12]) (Figure 16), but these results do not suggest increased operative mortality from delaying surgery through surveillance. In the registry studies, the 30-day mortality among those receiving EVAR was slightly higher (1.1% in ASERNIP-S; 1.6% in EUROSTAR).122,123 The difference was likely due to the fact that registry studies included higher-risk patients or to other differences in community practice.

Figure 16 is a forest plot depicting the relative risk for 30-day operative mortality in early EVAR versus surveillance trials. This figure is discussed in the Results section.

Figure 16

30-Day Operative Mortality in Early EVAR vs. Surveillance Trials. * Both random-effect (D-L) and fixed-effect (M-H) models were included due to the very low event rates.

Complications. The reporting of complications and the time periods assessed differed considerably across the four studies (two RCTs and two registry studies) (Table 11).113,115,122,123 In PIVOTAL, approximately 4 percent of patients required reinterventions over an unspecified time period in both groups, suggesting no difference in reinterventions for those undergoing early versus delayed surgery after surveillance.115 Endoleaks were the most frequent 30-day complication, occurring in 10 to 12 percent of patients receiving EVAR, but were not different between those undergoing early versus delayed EVAR (after surveillance). Other 30-day complications, including endograft or peripheral thromboses, wound infections, and systemic complications, occurred in about 15 percent of EVAR recipients, with no difference in early versus delayed surgery.

In the CAESAR study, the percentage of patients with any morbidity was significantly higher in the early EVAR group than the surveillance group at 30 days (18% vs. 6%; p<0.01) and after 30 days to 2.5 years (19% vs. 5%; p<0.01) (Table 11).113 Although the groups did not differ in the number of major adverse events early or late in the study, the early EVAR group had significantly more endoleaks after 30 days to 1 year (12% vs. 2.8%; p=0.028) and significantly more secondary procedures than the surveillance group (5.7% vs. 0%; p=0.03).

Both the ASERNIP-S and EUROSTAR registry studies reported mortality and complications after EVAR in participants with small AAA (30-day postoperative mortality in ASERNIP-S, timing not reported in EUROSTAR), with much less detail in EUROSTAR (Table 11).122,123 The rate of systemic complication (defined as cardiac, pulmonary, renal, cerebral, or gastrointestinal complications) was similar: 13.4 percent in ASERNIP-S and 12.0 percent in EUROSTAR. The rate of device and procedural complications, however, appeared much higher in ASERNIP-S (10.7% within 30 days of surgery) than EUROSTAR (2.9%, unclear timeframe). The two registry studies additionally reported complications at longer followup (median of 3.2 years in ASERNIP-S and 1.7 years in EUROSTAR). In the ASERNIP-S study, 97 (20.3%) patients had endoleaks during followup (28% requiring reintervention). EUROSTAR compared the cumulative probability of various types of endoleaks (i.e., type I proximal, type I distal, type II, and type III), but incomplete reporting did not allow us to determine the cumulative probability of having an endoleak. The cumulative probability of conversion to open surgery was 6.6 percent.123

Quality of life. Data from one trial (CAESAR) compared short-term (6 months) and longer-term (mean, 3 years [SD, 1.2 years]) quality of life in those with small AAA randomized to early EVAR versus surveillance114 (Appendix I Table 1). Compared with baseline, overall quality of life improved at 6 months in patients receiving EVAR (Figure 16), with larger benefits in the mental health summary score than the physical health summary score. In contrast, quality of life decreased slightly from baseline to 6 months in the surveillance group, thereby favoring early EVAR for all of these measures. By the end of followup, both groups showed decreases in quality of life from baseline, and none of these quality of life summary scores differed between groups.

Harms Associated With Pharmacotherapy

In addition to the four RCTs described in KQ 4,106,108,113,115 one additional RCT of propranolol versus placebo provided harms data only.117 This RCT randomized 54 patients with small AAA to receive either 40 mg of propranolol twice daily (n=30) or placebo (n=24) for 2 years (Appendix H Table 1). A large proportion of patients (60% in propranolol group vs. 29% in placebo group) dropped out of the study, mainly because of adverse events. Thus, while the trial was ineligible for assessment of benefits because of significant loss to followup, it provided useful data about harms associated with propranolol in the treatment of AAA.

Two trials assessing propranolol reported data about adverse effects that led to medication discontinuation116 or patient dropout.117 Both studies suggested that propranolol at least doubled patient discontinuation of medications or dropout from the study (37.7% vs. 21.3%; p <0.0001 in PAT;116 60% vs. 40%; p-value not reported in the Danish trial117) (Table 11). In PAT, propranolol was associated with a higher rate of fatigue, shortness of breath, and bradycardia or atrioventricular block, and was possibly associated with higher risk of heart failure.116 Because of the small sample size in the Danish trial, adverse events were generally very low. More adverse events, however, occurred in patients receiving propranolol across all adverse event categories.117

The three antibiotic trials reported medication discontinuation and associated side effects (Table 11).118-120 Generally, results suggested that these antibiotics used over 4 to 15 weeks were not associated with a significant increase of harms. In the trial of doxycycline, two patients (one with doxycycline, one with placebo) out of 32 total participants discontinued medication because of allergic reactions.118 The trial of roxithromycin reported no medication discontinuation.120 Two patients in the treatment group and two in the placebo group, out of a total of 247 participants, discontinued medication in the trial of azithromycin due to gastrointestinal symptoms, diarrhea, and arthralgia. Additionally, one patient in the treatment group discontinued medication due to allergic reaction, but this was found to be caused by an antihypertension medication. The trial of azithromycin also reported that 21 patients (13 in the treatment group and eight in the control group; p=0.37) had side effects. Details about the side effects, however, were unknown. Two of the three antibiotic trials119,120 reported a need for surgery in each group; however, due to a small number of events overall, conclusions are limited.

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.6M)

In this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...