There were no cases of surgery with survival for ruptured AAAs.
eMethods. Sample Size and Target Population
eFigure 1. CONSORT Flow Diagram for Men Aged 65 to 74 Years
eFigure 2. The Timing and Number of Men Aged 64 to 83 Years Undergoing Open Surgery and Endovascular Procedures (EVAR) for AAAs
eFigure 3. Cumulative Deaths From AAAs in Men Aged 65 to 74 Years
eFigure 4. AAA-Related Events in Men Aged 65 to 74 Years With Aortic Diameters <30 mm at Baseline
eTable 1. Comparison of Men Who Did and Did Not Attend Screening
eTable 2. Comparison of Screened Men Who Smoked vs Men Who Never Smoked
eTable 3. Comparison of Crude Numbers and Percentages (Not Rates/Person-years) in the 3 Largest Trials of Screening for AAAs in Men Aged 65 to 74 Years (65-73 Years for the Danish Trial)
McCaul KA, Lawrence-Brown M, Dickinson JA, Norman PE. Long-term Outcomes of the Western Australian Trial of Screening for Abdominal Aortic AneurysmsSecondary Analysis of a Randomized Clinical Trial. JAMA Intern Med. 2016;176(12):1761–1767. doi:10.1001/jamainternmed.2016.6633
Copyright 2016 American Medical Association. All Rights Reserved.
Does screening older men for abdominal aortic aneurysms (AAAs) reduce mortality from AAAs in the long term?
In this randomized clinical trial of 38 480 men (aged 64-83 years) in Western Australia, use of administrative databases, such as the electoral roll, to identify and invite the men for AAA screening increased the detection rate and number of elective operations in the screened group, but mortality was not significantly reduced.
It is unlikely that a national AAA screening program will be effective in an Australian health care setting.
Mortality from ruptured abdominal aortic aneurysms (AAAs) remains high. The benefit of screening older men for AAAs needs to be assessed in a range of health care settings.
To assess the influence of screening for AAAs in men aged 64 to 83 years on mortality from AAAs.
Design, Setting, and Participants
This randomized clinical trial performed from April 1, 1996, through March 31, 1999, with a mean of 12.8 years of follow-up (range, 11.6-14.2 years) included a population-based sample from a single metropolitan region in Western Australia identified via the electoral roll. Data analysis was performed from June 1, 2015, to June 1, 2016.
Randomization to an invitation to undergo ultrasonography of the abdominal aorta or a control group without invitation.
Main Outcomes and Measures
Surgery for and mortality from AAA.
A total of 49 801 men aged 64 to 83 years were identified for the study. Men living too far from screening centers (n = 8671) or who died before invitation (n = 2650) were excluded, resulting in 19 249 men in the invited group and 19 231 controls (mean [SD] age, 72.5 [4.6] years; 95% white). Of 19 249 men invited for screening, 12 203 (63.4%) attended. There were more elective operations (536 vs 414, P < .001) and fewer ruptured AAAs (72 vs 99, P = .04) in the invited group compared with the control group. Overall, there were 90 deaths from AAAs in the invited group (mortality rate, 47.86 per 100 000 person-years; 95% CI, 38.93-58.84) and 98 in the control group (52.53 per 100 000 person-years; 95% CI, 43.09-64.03) for a rate ratio of 0.91 (95% CI, 0.68-1.21). For men aged 65 to 74 years, the AAA mortality rate in the invited group was 34.52 per 100 000 person-years (95% CI, 26.02-45.81) compared with 37.67 per 100 000 person-years (95% CI, 28.71-49.44) in the control group for a rate ratio of 0.92 (95% CI, 0.62-1.36). The number needed to invite for screening to prevent 1 death from an AAA in 5 years was 4784 for men aged 64 to 83 years and 3290 for men aged 65 to 74 years. There were no meaningful differences in all-cause, cardiovascular, and other mortality risks.
Conclusions and Relevance
Use of the electoral roll to identify and invite men aged 64 to 83 years for screening for AAAs had no significant effect on the overall mortality from AAAs.
isrctn.org Identifier: ISRCTN16171472
The initial reports1- 4 of 4 large randomized clinical trials of screening for abdominal aortic aneurysm (AAA) in men 65 years or older indicated that screening reduces mortality from AAAs. The long-term results of the Multicentre Aneurysm Screening Study (MASS) in the United Kingdom revealed that the benefit was sustained at 13 years with a 42% (95% CI, 31%-51%) relative reduction in mortality from AAAs.5 In the Danish trial of screening, there was a 66% (95% CI, 43%-80%) reduction at 14 years of follow-up.5 Screening for AAAs also appeared to be cost-effective in these 2 studies.6,7 In MASS, the absolute risk reduction in mortality was only 0.46% after approximately 14 years. The early results of the Western Australian trial of screening for AAAs were reported in 2004.3 This trial included men aged 65 to 83 years, which was older than the men in MASS (65-74 years old) and the Danish trial (65-73 years old). Although results after 43 months of follow-up demonstrated a relative risk reduction of 39%, this finding was not statistically significant, and long-term results to enable comparison with the MASS and Danish trials have not been reported.
The results of the trials have been considered encouraging enough to recommend screening programs in the United States and a number of other countries.8- 11Quiz Ref ID In the past 5 years, epidemiologic studies12- 15 have suggested that the incidence of AAAs may be decreasing. Consistent with this trend is the finding that the prevalence of AAAs in the national screening programs in the United Kingdom and Sweden is approximately half that seen in the screening trials.10,16 This finding has generated uncertainty about the value of introducing new national programs.15,17 To add to the available evidence, this article presents the long-term outcomes for the Western Australian trial of screening for AAAs.
The trial design has been reported in detail elsewhere.3,18 Briefly, this was a population-based randomized clinical trial to be analyzed on an intention-to-treat basis, with 90% power to detect a relative reduction of 50% in the mortality from AAAs in men invited for screening during 5 years. Because the available population of men aged 65 to 74 years was smaller than that required by our power calculation, we also planned to include men aged 75 to 79 years. In January 1996, men were identified from an electronic copy of the electoral roll and randomized into control and invited groups in strata defined by 5-year age group and postcode of residence. Because of imprecision in the dates of birth available from the Electoral Commission, some men aged 64 years and 80 to 83 years were also included (eMethods in the Supplement). There was an a priori intention to analyze outcomes separately in men aged 65 to 74 years, and the results for these men are reported to facilitate comparison with the other trials. The study was approved by the Human Research Ethics Committees of the University of Western Australia and the Health Department of Western Australia. All men attending for screening provided written informed consent to participation in the study.
Quiz Ref IDFrom April 1, 1996, through March 31, 1999, men were screened for AAAs at 5 community-based clinics and were given a letter that contained the results of the scan, with a copy for the general practitioner. Any follow-up investigations or referral to a surgeon were to be arranged by the general practitioner. No attempt was made by the investigators to influence clinical management with regard to threshold for intervention or method of repair. Data analysis was performed from June 1, 2015, to June 1, 2016.
For the analysis undertaken in the earlier report of midterm results, men in the control group were assigned a virtual date of screening, which was the median scheduled date of examination for men from the same postcode area randomized to the invited group. In retrospect, this approach was problematic. For the present analysis, we defined the date of entry into the trial as the date of invitation for screening in the invited group, with matched men in the control group being assigned a similar invitation date.
Follow-up was from the invitation date until death or June 30, 2010, whichever came first. Electronic record linkage to population-based, name-identified records was used to identify men undergoing surgery for AAAs and deaths from AAAs and all other causes (eMethods in the Supplement). Any death within 30 days of surgery was considered a death caused by an AAA. Deaths caused by specific complications of AAA surgery, such as aortoenteric fistula and graft sepsis, occurring after 30 days were also included as late AAA deaths.
The deaths are presented as crude proportions, rates, and rate ratios. All percentages are reported with 97.5% CIs calculated using the exact method. Differences in percentages between the control and invited groups were tested using χ2 tests. Mortality rates were calculated by dividing the number of deaths by the total accumulated person-time and are reported per 100 000 person-years. Rate ratios with 95% CIs were used to compare mortality between the control and invited groups. We used competing risks regression to estimate cumulative incidence functions of mortality from AAAs during the trial.19 All analyses were conducted using STATA statistical software, version 14.1 (StataCorp).
A total of 49 801 men aged 64 to 83 years were identified for the study. Quiz Ref IDMen living too far from screening centers (n = 8671) or who died before invitation (n = 2650) were excluded, resulting in 19 249 men in the invited group and 19 231 controls (mean [SD] age, 72.5 [4.6] years; 95% white). The trial numbers are summarized in Figure 1 and eFigure 1 in the Supplement. Differences in some numbers compared with an earlier report3 are discussed in the eMethods in the Supplement. Of 19 249 men invited for screening, 12 203 (63.4%) attended, and for men aged 65 to 74 years, 8950 of 13 266 (67.5%) attended. The overall prevalence of an AAA 30 mm or larger in diameter in screened men was 7.2% and of an AAA 55 mm or larger was 0.5%. For men aged 65 to 74 years, the prevalence of an AAA 30 mm or larger in diameter was 6.6% (4.5% in those aged 65 years) and of an AAA 55 mm or larger was 0.4%. The median follow-up duration was 12.8 years (range, 11.6-14.2 years).
Quiz Ref IDSignificantly more elective operations (536 vs 414, P < .001) were performed in the invited group compared with the control group (Table 1), an increase seen mainly in the first year after screening (eFigure 2 in the Supplement). The 30-day mortality after elective surgery was lower in the invited group compared with the control group (3.4% vs 4.1%, P = .54) and was even lower in men who attended for screening (2.7%) (eTable 1 in the Supplement). The overall proportion of endovascular repairs was 59.4%, increasing from 32.7% in 1996 to 1999 to 82.2% in 2005 to 2010; this number did not differ substantially between the invited and control groups (eFigure 2 in the Supplement). There were significantly fewer ruptured AAAs in the invited group compared with the control group (72 vs 99, P = .04). However, the 30-day mortality after surgery for ruptured AAAs was higher in the invited group compared with the control group (61.5% vs 43.2%). This pattern of results was similar for men aged 65 to 74 years (Table 1).
The cumulative deaths from AAAs in men aged 64 to 83 years are shown in Figure 2 (see eFigure 3 in the Supplement for men aged 65-74 years). For the first 4 years after screening, mortality was slightly higher for the invited group, then the 2 groups became similar until slightly favoring the invited group. Deaths from AAAs and other causes are detailed in Table 2. Overall, there were 98 deaths from AAAs in the control group (52.53 per 100 000 person-years; 95% CI, 43.09-64.03) compared with 90 in the invited group (47.86 per 100 000 person-years; 95% CI, 38.93-58.84) for a rate ratio of 0.91 (95% CI, 0.68-1.21). For men aged 65 to 74 years, the mortality rate in the invited group was 34.52 per 100 000 person-years (95% CI, 26.02-45.81) compared with 37.67 per 100 000 person-years (95% CI, 28.71-49.44) in the control group for a rate ratio of 0.92 (95% CI, 0.62-1.36). Quiz Ref IDThe number needed to invite for screening to prevent 1 death from an AAA in 5 years was 4784 for men aged 64 to 83 years and 3290 for men aged 65 to 74 years. There were no meaningful differences in all-cause, cardiovascular, and other mortality risks (Table 2). The differences between men attending and not attending screening within the invited group are given in eTable 1 in the Supplement. Despite screening, men who had ever smoked were more likely to experience a ruptured AAA and die of an AAA than those who had never smoked (eTable 2 in the Supplement). The numbers of AAA events (death, rupture, or elective operation) in men with aortic diameters smaller than 3 cm at baseline appeared to increase from approximately 8 years after screening (Figure 3 and eFigure 4 in the Supplement).
Use of the electoral roll to identify and invite men aged 64 to 83 years for screening for AAA resulted in a small but not significant reduction in mortality from AAAs after nearly 13 years of follow-up. The findings were similar in the subgroup of men aged 65 to 74 years. This lack of benefit was primarily attributable to a relatively low rate of rupture and death from AAAs, along with a high rate of elective surgery for AAAs, in the control group.
The strength of this study is that it was truly population based and was undertaken in a geographically discrete and isolated city with access to a reliable mechanism for follow-up during an extended period. General practitioner lists were not used for recruitment because they are neither mandatory nor comprehensive in Australia. Because voting is compulsory, the electoral roll was considered to be a comprehensive and reasonably representative method to contact the target population. This approach minimized any sociodemographic or clinical bias but entailed the weakness of being unable to exclude individuals who are not appropriate for screening because of prior AAA repair, major comorbidities, or advanced malignant tumors, as was done in the other trials. Outcomes were assessed using linkage to administrative data, although validation of individual medical records could not be undertaken because of ethical constraints. Some miscoding of causes of death is inevitable, although there is no reason for any systematic bias in these errors. Postmortem rates are low in Australia,20 and some cases of sudden death may have been incorrectly attributed to AAAs. This type of error may have been more frequent in screened men known to have an AAA, thereby reducing the apparent benefit of screening. From a public health policy perspective, this factor makes our conclusions conservative. The need to include men older than 75 years to achieve our required power was a recognized limitation. Although these older men had a higher prevalence of AAAs, they were less likely to attend and probably less likely to have been suitable candidates for elective AAA repair. However, both the absolute and relative risk reductions were similar in the men aged 75 to 83 years compared with those aged 65 to 74 years, suggesting that the inclusion of older men did not reduce the efficacy of screening.
The main outcome of interest in our trial was mortality from AAAs. As previously reported,21 the effect of screening on quality of life was modest, but we have not assessed nonfatal outcomes after surgery for AAAs. The early morbidity of elective open surgery and the cumulative long-term complications of endovascular surgery remain substantial.22 Given that screening inevitably results in a large increase in elective surgery, this unmeasured morbidity has raised questions about the underestimation of the harms of screening.23,24
For men aged 65 to 74 years, the 8% reduction in mortality from AAAs seen in our trial is considerably less than that seen in MASS (42% reduction) and the Danish trial (66% reduction) at a similar length of follow-up.5,6 It is likely that a number of factors contributed to these differences (eTable 3 in the Supplement). The response fraction (attendance) in our trial was only 68% compared with 80% in MASS and 77% in the Danish trial. This difference may relate to the use of the electoral roll as a means of recruitment. If our response fraction had been 80%, as seen in MASS, the estimated relative risk reduction could have been a more comparable 29% (37 deaths in the invited group vs 48 in the controls).
Another reason for the lack of benefit in the present trial appears to be the lower than expected mortality from AAAs in the control group. Men aged 65 to 74 years in the Australian control group were less likely to experience a ruptured AAA (0.4% vs 1.4% in MASS), and as a result, the absolute risk of death from AAAs was lower (0.39% vs 1.12% in MASS). The reason for this difference is unclear but could be attributable to a relatively high proportion of men in the control group undergoing elective surgery for AAAs during the trial period (2.1% vs 0.82% in MASS). This finding may reflect a higher incidental diagnostic rate attributable to the long-standing availability of open access imaging (ultrasonography and computed tomography) for general practitioners in Australia. In addition, the design of the trial, with results sent to general practitioners, compared with the other trials that referred directly for surgical assessment may have sensitized practitioners to the possibility of AAAs and led to more testing and referrals of men in the control group. Finally, the mortality rate after elective surgery in the control group was slightly lower in our trial compared with MASS (4% vs 5.1%). This finding may be attributable to the much higher proportion of endovascular repairs (60% vs 13%) and possibly a tendency for Australian vascular surgeons to operate at a threshold diameter of 5 cm rather than 5.5 cm (although there are limited data to assess this possibility).
A final factor to consider is that AAAs are becoming less common (at least in men). The incidence of and mortality from AAAs are clearly decreasing in various countries, including Australia and the United States.12,15 The prevalence of AAAs in men aged 65 to 74 years was 4.5% to 6.0% at the time of the main trials, yet screening studies9,10,16 in the United Kingdom and Sweden indicate that the prevalence in men approximately 65 years of age is now less than 2%. The reason for the decrease in incidence and prevalence is multifactorial but is probably driven by differences in rates of smoking and cessation13,15 because the relative risk for AAA events is 3- to 6-fold higher in smokers compared with nonsmokers.25 In Australia, the smoking rate in men 65 years or older decreased from 15% in 1995 to 8% in 2011.26 Although this trend also occurred in other countries, the overall smoking rate in Australia (15%) was lower than in the United States and Europe (approximately 20%-27%) over many years.27,28 This difference may have affected the prevalence of AAAs and incidence of rupture in these countries differentially.
There has been considerable interest in the long-term treatment of individuals with nonaneurysmal aortic diameters at screening. In MASS, the rate of ruptured AAAs appeared to start increasing at approximately 8 years after baseline screening.29 We also observed an increase in the numbers of interventions for AAAs after approximately 8 years. Most, but not all, of these interventions were in men with baseline aortic diameters in the 25- to 29-cm range. A multicenter study30 found that 26% of men and women with subaneurysmal aortic diameters of 2.5 to 2.9 cm developed an AAA larger than 5.4 cm in diameter within 10 years. Together, these studies29,30 suggest that men with aortic diameters in the 2.5- to 2.9-cm range are at risk of incident AAAs, although the benefit of surveillance of these men remains to be established. Interestingly, aortic diameters larger than 2.5 cm in men are also associated with future (nonaneurysmal) cardiovascular events.31- 33
Our results suggest that a national screening program using administrative databases, such as the electoral roll, to identify men aged 64 to 83 years or 65 to 74 years is unlikely to be effective. Differences in health care systems, including imaging practices, and the effectiveness of health promotion interventions, such smoking cessation, control of hypertension, and use of statins, may influence the benefit of screening for AAAs and should be considered in individual countries before the introduction of screening. Selective screening of smokers and ex-smokers34 or use of a risk scoring system35 might be more effective, but an earlier analysis36 of this cohort found that doing so would fail to identify approximately 25% of AAAs. The small overall benefit of population-wide screening does not mean that finding AAAs in suitable older men is not worthwhile because deaths from AAAs in men who actually attended for screening were halved by early detection and successful treatment.
Corresponding Author: Paul E. Norman, MB, DS, School of Surgery, University of Western Australia and Harry Perkins Institute of Medical Research, Fiona Stanley Hospital, PO Box 480, Murdoch 6151, Australia (firstname.lastname@example.org).
Accepted for Publication: August 31, 2016.
Published Online: October 31, 2016. doi:10.1001/jamainternmed.2016.6633
Author Contributions: Drs McCaul and Norman had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Lawrence-Brown, Dickinson, Norman.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: McCaul, Norman.
Critical revision of the manuscript for important intellectual content: Lawrence-Brown, Dickinson, Norman.
Statistical analysis: McCaul.
Administrative, technical, or material support: Norman.
Study supervision: Dickinson, Norman.
Conflict of Interest Disclosures: None reported.
Funding/Support: The study was supported by grants 964145 (Drs Norman, Lawrence-Brown, and Dickinson), 403963 (Dr Norman), and 1006266 (Drs Norman and McCaul) from the National Health and Medical Research Council Project.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Additional Contributions: The late Konrad Jamrozik, DPhil, was instrumental in establishing the study, and the late Max Le, BSc, made a major contribution to running the trial. We thank all the research staff and all the participants involved in the study and earlier piloting of the trial. We gratefully acknowledge the use of information provided by the Department of Health of Western Australia and the use of linkages provided by Western Australian Data Linkage Branch.