Title page Manuscript type: Meta-analysis. Title: Screening for abdominal aortic aneurysm reduces overall mortality in men. A meta-analysis of the mid- and long- term effects of screening for abdominal aortic aneurysms. Authors: Jes S. Lindholt (1), Paul Norman (2) Institutions: (1). Vascular Research Unit, Viborg Hospital, Denmark (2). School of Surgery, University of Western Australia Words: 3009 including Abstract, tables, legends and references Tables: 1 Figures: 2 References: 15 Responsible for correspondence concerning the manuscript: Ass. Prof., Ph.D. Jes S. Lindholt Vascular Research Unit, Viborg Hospital, Postbox 130, 8800 Viborg, Denmark E-mail: Jes.S.Lindholt@Sygehusviborg.dk Tlf : +45 8927 2447 1
Abstract. Background Four randomised controlled trials of screening older men for abdominal aortic aneurysms (AAA) have been completed. Consequently, a meta-analysis was performed to examine the updated pooled effects of screening on both mid- and long-term AAA-related and total mortality, and operations for AAA. Methods Pooled mid-term (3½-5 years) and long term (7-15 years) effects were calculated as odds-ratios (ORs) with 95% confidence intervals in fixed effect models. Long-term data from the West Australian trial were limited to all-caurse deaths. Heterogeneity between the studies was assessed by the χ 2 -test. In cases of heterogeneity, random effect models were used. Results The pooled mid-term analysis showed the offer of screening caused a significant reduction in AAArelated mortality (OR= 0.56, 95% CI: 0.44,0.72),overall mortality (OR= 0.94, 95% C.I.: 0.86; 1.02), and emergency operations (OR=0.55, 95% C.I.: 0.39; 0.76), while the number of elective operations increased significantly (OR= 3.27, 95% C.I.: 2.14; 5.00). The long-term results also showed a significant reduction in AAA-related mortality (OR= 0.47, 95% C.I.: 0.25; 0.90), overall mortality (OR= 0.94, 95% C.I.: 0.92; 0.97) and emergency operations (OR=0.48, 95% C.I.: 0.28; 0.83), while the number of elective operations increased significantly (OR=2.81, 95% C.I.: 2.40; 3.30). Conclusion Population screening for AAA reduces AAA-related and overall mortality, however differences may exits which could influence local cost effectiveness of screening 2
Introduction In spite of an increasing elective surgery for asymptomatic abdominal aortic aneurysm (AAA), the sex- and age-standardized mortality from ruptured AAA continues to increase(1;2). The overall mortality of ruptured AAA is about 80-90 %(1) compared to a 30 day postoperative mortality of 3.3% after elective AAA repair in Denmark in 2006 (www.karbase.dk). The presence of an asymptomatic phase with the opportunity of a relatively low risk treatment compared to the symptomatic phase raised the question of whether screening for AAA would be effective. If seriously considered, all the criteria for screening formulated by WHO and the Council of Europe would need to be fulfilled(1). Firstly, ultrasound scanning is a valid, suitable and acceptable method of screening. It is fast and safe with an estimated sensitivity and specificity of 98 and 99 %, respectively(3). Attendance rates of 53-79 % have been consistently achieved, and about 95% of individuals with small AAAs attends surveillance programs(1;2). Secondly, the indications for treatment of asymptomatic AAA are clear. The diameter of an AAA remains the most useful risk factor for rupture, and based upon two high quality randomised controlled trials (4;5), the threshold for elective surgery is a diameter of 55 mm or more. For screening to be effective, the treatment must be acceptable to patients. Those surviving surgery have the same quality of life as the matched background population; only 2-5% of the patients refuse surgery, although contraindications to surgery may be present in 15% of cases. As 85-90% of screen-detected cases are initially too small to warrant surgery, they are kept under surveillance, and such surveillance reduces quality of life(6). However, these issues seem minor, if the reduction in mortality (at acceptable cost) is substantial (1). Randomised controlled trials of screening older men aged have been running since the late 1980s in order to assess the role of screening for AAA. 3
Aim The aims of the analysis were to examine the updated pooled mid- and long-term effects of screening on AAA-related and total mortality, and operations. Search strategy Medline was searched with the words Screening AND aortic aneurysms with limitation to randomised controlled trials reported in English. 72 papers were retrieved. Relevant data from 4 trials were identified in 9 papers: The Chichester study (UK) with reports after 5(7), 10 and 15(8) years of follow up, the Multicentre Aneurysm Screening Study - MASS (UK) with reports after 4 (9)and 7 years(10), the West Australian Aneurysm Screening Study (AUS) (11) with results after 3.6 years, and the Viborg Study (DK) with reports of results after 5(12), 7 (AAA-related deaths) (13) and 10 years (all deaths and AAA operations) (14). Analyses Pooled mid-term effects were analysed at 3-5 years, and long-term effects at 7-15 years, of follow up. Due to lack of long-term results from the West Australian trial, data from this trial were not used in the long-term analysis, with the exception of all-cause mortality where data were available after eleven years of follow up. The analyses were performed using the intention to treat principle with calculation of pooled oddsratios (ORs) and 95% confidence intervals (CI) in fixed effect models. Heterogeneity between the studies was assessed by the χ 2 -test. In such cases, random effect models were used. Reviewer Manager 4.2 was used as software. 4
Description of the studies All four trials described the method of randomisation. None were blinded of obvious reasons. The participants were mainly men aged 64-83 years. The Chichester study also included 65-80 year old women(7). These were excluded from the meta-analysis in order to maximise study homogeneity. The age groups varied between the studies; 64-73 years in Viborg, 65-74 years in MASS, 65-80 years in Chichester, and 65-83 years in the Western Australian study. The participants were identified with various methods: In the British studies by patient registers of General Practioners and Family Health Service lists allowing selection of men suitable for potential AAA repair, while the Australian Study used the electoral roll. In Denmark, non-selective digitalised personal identification was used. All studies used computerised randomisation with the participants (1:1) for being offered screening or control status. Three of the studies established screening sessions outside hospital by a mobile team, frequently at General Practioners, while the Viborg Study performed the screening sessions in the County hospitals by a mobile screening team. All studies used abdominal ultrasonography for screening, and defined an AAA as an abdominal aorta of 3 cm or more in maximal diameter. The indication for surgery varied between the studies: the British studies if AAA-related symptoms were present, if the annual AAA growth rate was 1 cm or more, or if the maximal aortic diameter exceeded 5.5 cm (MASS) or 6 cm (Chichester study). In the Viborg study, if AAA-related symptoms were present or if the maximal aortic diameter exceeded 5 to 5.5 cm. In the Western Australian Study, the indications for surgery were left to the individual surgeons. Similarly, the intervals between surveillance scans varied between the studies; from every third month in the British trials in cases with AAAs of 4.5cm and over to annual scan in the Viborg Study. 5
Results The number of participants, age groups, attendance and prevalence of AAA are summarised in Table 1. The trials included a total of 125,576 men. Attendance rate ranged from 70% in WesternAustralia to 80% in MASS trial. The prevalence of AAA ranged from 4.0% in the Viborg Study to 7.7% in Chichester. The analyses of the mid-term results showed the offer of screening caused a significant reduction in AAA-related mortality (OR= 0.56, 95%.C.I.:0.44;0.72, Figure 1). However, the inclusion of men above 80 must be questioned, since many may not be offered elective surgery for AAA. If these are excluded from the meta-analysis, screening significantly reduced AAA-related mortality (OR=0.51, 95% C.I.: 0.40; 0.67). In addition, there was a significant reduction in overall mortality (OR=0.93, 95% C.I.: 0.90; 0.96), however, heterogeneity between the studies was present and statistical significance was lost in a random effect model (OR= 0.94, 95% C.I.: 0.86; 1.02). A significantly higher number of planned operations (OR=3.27, 95% C.I.: 2.14; 5.00), and significantly fewer emergency operations were observed (OR= 0.55, 95% C.I.:0.39; 0.76, Figure 1). The analyses of the long-term results showed the offer of screening caused a significant reduction in AAA-related mortality (OR= 0.47, 95% C.I.:0.25; 0.90, Figure 2). A significant reduction of overall mortality was also noticed (OR= 0.94, 95% C.I.:0.92; 0.97) - without significant heterogeneity between the studies. If a random effect model had been used, the reduction would still be statistically significant (OR= 0.95, 95% C.I.:0.90; 0.99). A significantly greater number of elective operations (OR= 3.27, 95% C.I.:2.14; 5.00), and significantly fewer emergency operations was also noticed (OR= 0.48, 95% C.I.:0.28; 0.83, figure 2). Overall, there were significantly more operations in the invited group compared to the controls (OR=1.75, 95% C.I.:1.54; 1.99). 6
Discussion A meta-analysis should only be performed, if it includes new data. Recently, a Cochrane review(15) was published, but it did not include data published after 2004-5, thus missing important data from three of the four randomised trials. In addition, it did not exclude unspecified aortic ruptures, which tends to underestimate the benefit in the MASS trial. The results of this updated meta-analysis appear to identify evidence of significant benefit in men; pooling data from the trials showed a significant mid-term reduction in mortality from AAA related and overall mortality after 3-5 years, which were sustained in the long-term analysis. Also a significant increase in rates of elective surgery of asymptomatic AAA resulting from screening was noticed, together with a significant decrease in rates of emergency surgery. However, there are some uncertainties in this review. Odds ratios are not the most suitable and precise test to use Cox proportional hazards ratio analysis would be more appropriete, but would require access to a merged database containing raw data from all studies. However, significant heterogeneity was seen in the mid-term analysis; although overall mortality tended to decrease in all studies, only the Western Australia study showed a significant decrease in all-cause mortality, and the pooled result was not significant using a random effect model. Consequently, the novel long-term results from the Western Australian study is very interesting, and together with the other trials long term results, it shows screening for AAA significantly decreases overall long-term mortality. Heterogeneity was noticed for elective surgery. There are obvious explanations for this, including differences in the prevalence of AAAs and indications for surgery. Heterogeneity was also noticed in the longterm results of AAA-related mortality. This may be due to the observed differences between the British studies and the Viborg Study; the British studies included aortic rupture at unspecified site. If these cases were removed from the analysis, the OR in the MASS trial falls from 0.58 to 0.44 compared to 0.23 (95% C.I.: 0.11; 0.47) in the Viborg Study. Although not a 7
necessarily a significant factor, AAA-related deaths may have been over-estimated in the screen group of MASS as cause of death simply relied on the death certificate, while the two other trials had independent reviewers classify the cause of death in cases suspected to be due to AAA. This may result in bias against screening for AAA; those with a known (screen-detected) AAA with coexisting ischemic heart disease dying suddenly, are more likely to have the cause of death attributed to AAA, than a patient without an AAA. Finally heterogeneity was noticed for emergency operations, where the Chichester study and the Western Australian study did not observe any reduction with screening, while significant reductions were noticed in the MASS and Viborg Studies. This may be a due to younger populations in the MASS and Viborg Study. In conclusion, screening for AAA reduces AAA-related deaths and overall mortality, however, local differences in health care settings could influence the local cost effectiveness of screening. 8
Table 1. Characteristics of the four randomised screening trials for abdominal aortic aneurysms. Viborg Study Western Australia MASS Chichester, Men Combined Age (y) 64 73 65 83 65 74 65 80 Participants (No) 12 639 38 704 67 800 6040 125 576 Max. Follow-up (y) 9.6 3.6 7 15 Attenders (%) 77 63 80 73 74 AAA-Prevalence 4.0% 7.2% 4.9% 7.7% 5.5% 9
Figure 1. Meta-analysis of the mid-term effects of screening 64-83 year old men : AAA-related mortality, total mortality and operations for AAA. 10
Figure 2. Meta-analysis of the long-term effects of screening 64-83 year old: - AAA-related mortality, total mortality and operations for AAA. 11
Reference List (1) Lindholt JS. Considerations and experiences of screening for abdominal aortic aneurysms. Copenhagen: Fadl s Forlag; 1998. (2) Wilmink AB, Quick CR. Epidemiology and potential for prevention of abdominal aortic aneurysm. Br J Surg 1998 February;85(2):155-62. (3) Lindholt JS, Vammen S, Juul S, Henneberg EW, Fasting H. The validity of ultrasonographic scanning as screening method for abdominal aortic aneurysm. Eur J Vasc Endovasc Surg 1999 June;17(6):472-5. (4) Long-term outcomes of immediate repair compared with surveillance of small abdominal aortic aneurysms. N Engl J Med 2002 May 9;346(19):1445-52. (5) Lederle FA, Wilson SE, Johnson GR, Reinke DB, Littooy FN, Acher CW et al. Immediate repair compared with surveillance of small abdominal aortic aneurysms. N Engl J Med 2002 May 9;346(19):1437-44. (6) Lindholt JS, Vammen S, Fasting H, Henneberg EW. Psychological consequences of screening for abdominal aortic aneurysm and conservative treatment of small abdominal aortic aneurysms. Eur J Vasc Endovasc Surg 2000 July;20(1):79-83. (7) Scott RA, Wilson NM, Ashton HA, Kay DN. Influence of screening on the incidence of ruptured abdominal aortic aneurysm: 5-year results of a randomized controlled study. Br J Surg 1995 August;82(8):1066-70. (8) Ashton HA, Gao L, Kim LG, Druce PS, Thompson SG, Scott RA. Fifteen-year follow-up of a randomized clinical trial of ultrasonographic screening for abdominal aortic aneurysms 1. Br J Surg 2007 June;94(6):696-701. (9) Ashton HA, Buxton MJ, Day NE, Kim LG, Marteau TM, Scott RA et al. The Multicentre Aneurysm Screening Study (MASS) into the effect of abdominal aortic aneurysm screening on mortality in men: a randomised controlled trial. Lancet 2002 November 16;360(9345):1531-9. (10) Kim LG, RA PS, Ashton HA, Thompson SG. A sustained mortality benefit from screening for abdominal aortic aneurysm. Ann Intern Med 2007 May 15;146(10):699-706. (11) Norman PE, Jamrozik K, Lawrence-Brown MM, Le MT, Spencer CA, Tuohy RJ et al. Population based randomised controlled trial on impact of screening on mortality from abdominal aortic aneurysm. BMJ 2004 November 27;329(7477):1259. (12) Lindholt JS, Juul S, Fasting H, Henneberg EW. Screening for abdominal aortic aneurysms: single centre randomised controlled trial. BMJ 2005 April 2;330(7494):750. (13) Lindholt JS, Juul S, Henneberg EW. High-risk and Low-risk Screening for Abdominal Aortic Aneurysm Both Reduce Aneurysm-related Mortality. A Stratified Analysis from a Single-centre Randomised Screening Trial. Eur J Vasc Endovasc Surg 2007 February 27;34:53-8. 12
(14) Lindholt JS, Juul S, Fasting H, Henneberg EW. Preliminary ten year results from a randomised single centre mass screening trial for abdominal aortic aneurysm. Eur J Vasc Endovasc Surg 2006 December;32(6):608-14. (15) Cosford PA, Leng GC. Screening for abdominal aortic aneurysms. Cochrane Database Syst Rev 2007;No.: CD002945(Issue 2). 13