Fifteen-year follow-up of a randomized clinical trial of ultrasonographic screening for abdominal aortic aneurysms

Abstract

Background

Long-term benefits of screening for abdominal aortic aneurysm (AAA) are uncertain. These are the final results of a randomized controlled screening trial for AAA in men, updating those reported previously. Benefit and compliance over a median 15-year interval were examined.

Methods

One group of men were invited for ultrasonographic AAA screening, and another group, who received standard care, acted as controls. A total of 6040 men aged 65–80 years were randomized to one of the two groups. Outcome was monitored in terms of AAA-related events (surgery or death).

Results

In the group invited for screening, AAA-related mortality was reduced by 11 per cent (from 1·8 to 1·6 per cent, hazard ratio 0·89) over the follow-up interval. Screening detected an AAA in 170 patients; 17 of these died from an AAA-related cause, seven of which might have been preventable. The incidence of AAA rupture after an initially normal scan increased after 10 years of follow-up, but was still low overall (0·56 per 1000 person-years).

Conclusion

Screening with a single ultrasonography scan still conferred a benefit at 15 years, although the results were not significant for this population size. Fewer than half of the AAA-related deaths in those screened positive could be prevented. Registration number: ISRCTN 00079388 (http://www.controlled-trials.com).

The 41st Annual Scientific Meeting of the Vascular Society of Great Britain and Ireland was held in Edinburgh, UK on 22–24 November 2006 under the presidency of Mr John Wolfe. The abstracts of original work presented at the meeting have been published in a BJS supplement online (www.bjs.co.uk; Br J Surg94(S1):1–13), and the prize abstracts were published in the February 2007 issue of BJS (Br J Surg94:254–256), The following five articles are based on presentations from the Edinburgh meeting and have been accepted for publication after peer review by the Journal's referees. Other papers from the meeting will appear in subsequent issues.

Introduction

This is the final report of the first randomized controlled trial of abdominal aortic aneurysm (AAA) screening, set up in 1988 and carried out in a primary care setting. The results of the effectiveness of screening on AAA-related mortality were reported after 5 years¹ and 10 years². The relative risk reduction in AAA-related mortality in those invited for screening was 42 per cent at 5 years and 21 per cent at 10 years. The study's initial findings were used to plan the Multicentre Aneurysm Screening Study. This larger trial, involving 67 800 men aged 65–74 years, has so far confirmed the advantage of a screening programme in terms of both AAA-related mortality³ and health economic criteria⁴ after 4 years of follow-up.

The benefit of screening and early treatment for AAA is reduced over time by additional AAA-related deaths in the group invited for screening. This study focused on the long-term benefit of screening and compliance. It assessed why some patients remain at risk of rupture despite the screening programme and considered the potential for reducing this at-risk group.

Methods

Details of this randomized controlled study have been published previously^1,2,5. In summary, 6431 men aged 65–80 years from nine general practices within the catchment area of St Richard's Hospital, Chichester, were identified between 1988 and 1991, and selection was based on date of birth only. At the start of the study two decades ago, most general practitioners had only paper patient records, and there were few electronic links with the health authority records. The availability of computerized systems has allowed the information to be updated; this information was more recently cross-checked with electronic downloads from the health authority and through patient flagging with the Office of National Statistics mortality surveillance system. This highlighted anomalies (for example, incorrect sex assigned, duplicate registration, previously unknown deaths occurring out of area), allowing the information held on the database to be corrected. As a result of these rigorous checks, 391 men were excluded from the original data. The authors are confident that the data used for this analysis are now as accurate as is possible using all available sources. No formal stopping rules were put in place for this trial, but the 30-day elective surgical mortality rate was monitored throughout to ensure that it remained within acceptable levels.

An external research unit randomized the patients within each practice separately. In all, 6040 men were eligible for the study, of whom 2995 were randomized to the group invited for screening. The remaining 3045 were allocated to the control group, who received standard care. The date of randomization was defined as the first ultrasonography scan date in each practice, and was used as the relevant date for both controls and those invited for screening in each practice when assessing inclusion in the study and survival. Women were included in the original screening programme, but are not in this analysis. The low incidence of AAA in women suggests that this group should be analysed separately, although the small numbers make interpretation of the results difficult. Results on women have been published elsewhere⁶.

Screening clinics were held at a general hospital, a cottage hospital and a general practice. In the screening clinic, the aortic diameter was measured in two planes (transverse and longitudinal) and the maximum aortic diameter was recorded. An aortic diameter of 3 cm or more was considered aneurysmal, and these patients were offered surveillance at intervals related to the maximum aortic diameter: 3 monthly if the AAA diameter was 4·5–5·9 cm, or annually if the AAA diameter was 3·0–4·4 cm. Any patient who reached certain criteria (an AAA diameter of 6·0 cm or more, rapid aortic expansion of at least 1·0 cm in a year or developing symptoms attributable to the aneurysm) was referred to a vascular surgeon for consideration of elective AAA surgery.

Data were monitored over the 15-year surveillance interval to 31 March 2005. An AAA event was defined as an elective or emergency AAA operation, or death where AAA was the specified cause. Cause of death was established from death certificates, the Office of National Statistics flagging system and reports by the local registrar of deaths to the health authority. The primary outcome measure was death from an AAA (International Classification of Diseases 9, codes 441·3–6). Analysis of the two groups was performed using Cox regression, censoring other causes of death. The secondary outcome measure was the incidence of AAA rupture.

Results

The median age of the men at randomization was 72 years. The attendance rate was 74 per cent at the initial scan, and an AAA was detected in 170 men (7·7 per cent) (Figure 1). Table 1 shows the final status of all subjects at 31 March 2005; no patient remains under surveillance.

Fig. 1

Flow diagram of the randomized trial. *Includes deaths to 31 March 2005. AAA, abdominal aortic aneurysm

Table 1

Final outcome for all subjects at 31 March 2005

No. of patients	Controls (n = 3045)	Refused scan (n = 779)	Scanned normal (n = 2046)	Scanned AAA (n = 170)
Died from AAA (including AAA-related postoperative deaths)
No AAA surgery	41	16	8	13
Died ≤ 30 days after AAA surgery	12	3	2	3
Died > 30 days after AAA surgery	1	1	0	1
Total	54	20	10	17
Surviving emergency AAA surgery
Died from other causes	7	0	0	5
Alive	2	0	1	1
Total	9	0	1	6
Surviving routine AAA surgery
Died from other causes	9	0	0	24
Alive	9	2	0	14
Total	18	2	0	38
No AAA surgery
Died from other causes	1997	531	1334	95
Alive	967	226	701	14*
Total	2964	757	2035	109

*

Additional information on 14 surviving patients: two declined further follow-up; seven were discharged as unfit; three were discharged as borderline normal; one had a revised diagnosis; one moved away. AAA, abdominal aortic aneurysm.

Surgery

In the control group, 40 men had AAA surgery (21 as an emergency) compared with 57 in the group invited for screening (16 as an emergency) (Table 2). There were no 30-day postoperative deaths after 60 routine operations. This contrasts with the high mortality rate for emergency procedures: 20 of 37.

Table 2

Operations for abdominal aortic aneurysm and subsequent 30-day mortality

			Invited for screening (n = 2995)
					Attended screening (n = 2216)
Operation type	Controls (n = 3045)		Refused initial scan (n = 779)		Normal aorta (n = 2046)		AAA (n = 170)
	Operations	Deaths	Operations	Deaths	Operations	Deaths	Operations	Deaths
Routine	19	0	2	0	0	0	39	0
Emergency	21	12	4	3	3	2	9	3
Total	40	12	6	3	3	2	48	3

AAA, abdominal aortic aneurysm.

Of the 57 men who had surgery in the group invited for screening, the AAA had been detected as a result of the screening programme in 48. Four of these patients had surgery based on clinical judgement, without reaching the trial criteria. Of 170 men with an AAA detected, 86 reached the trial criteria for surgery. However, 42 did not have surgery during the follow-up interval (Table 3).

Table 3

Outcome for patients with an abdominal aortic aneurysm who fulfilled criteria for elective surgery in the screened group

Outcome	No. of patients	AAA operation	No AAA operation
			Died from AAA	Died from other causes	Alive
Deemed unfit for elective surgery	20	2	6	11	1
Declined outpatient appointment	2	0	1	1	0
Declined elective surgery	7	2	2	3	0
Returned to screening group	12	0	0	12	0
Planned surgery	40	40	0	0	0
Not referred to GP by screening team	1	0	0	1	0
Clinical decision pending	4	0	0	4	0
Total	86	44*	9	32	1

*

Four further patients had surgery without fulfilling the study criteria for elective surgery and are not included here. Three died within 30 days of their operation. AAA, abdominal aortic aneurysm; GP, general practitioner.

Twelve patients returned to screening after clinical review. They had aortic expansion of 1 cm or more in 1 year, but none had an aortic diameter of 6 cm or more. The aorta remained under 4·5 cm in nine patients, and the three others were assessed as unfit, although further surveillance was considered appropriate until the aorta reached 6 cm. These unfit patients died from causes unrelated to the aneurysm.

Of 39 patients with a screen-detected aneurysm who had an elective operation, all survived at least 30 days afterwards (Table 2). Two died within a year, but 31 survived 5 or more years after surgery.

Deaths from abdominal aortic aneurysms

There were 63 AAA ruptures in the control group, compared with 54 in the group invited for screening (Table 4). The number of AAA-related deaths was 54 (1·8 per cent) of 3045 in the control group and 47 (1·6 per cent) of 2995 in those invited for screening (Fig. 1). This is an 11 per cent reduction in the AAA mortality rate as a result of the screening programme, although this was not a significant difference (hazard ratio 0·89, 95 per cent confidence interval (c.i.) 0·60 to 1·32). There was no evidence that this hazard ratio varied with age at randomization. Fig. 2 shows cumulative AAA-related mortality over 15 years since randomization. Any difference between the AAA-related mortality in the two groups did not emerge until 4 years after randomization.

Fig. 2

Cumulative mortality from abdominal aortic aneurysm (AAA) over 15 years of follow-up by randomization group

Table 4

Mortality and incidence of ruptured abdominal aortic aneurysm

	Controls (n = 3045)	Invited for screening (n = 2995)
Person-year of observation	30·03	29·45
(1000s)
Total AAA-related deaths	54	47
Rate per 1000 person- years	1·80 (1·38, 2·35)	1·60 (1·20, 2·12)
Hazard ratio	1·00	0·89 (0·60, 1·32)
Non-fatal ruptured AAA	9	7
Total incidence of ruptured	63	54
AAA
Rate per 1000 person- years	2·10 (1·64, 2·69)	1·83 (1·40, 2·40)
Hazard ratio	1·00	0·88 (0·61, 1·26)
Total deaths from all causes	2067	2036
Rate per 1000 person-	68·83	69·14
years	(65·92, 71·86)	(66·20, 72·21)
Hazard ratio	1·00	1·01 (0·95, 1·07)

Values in parentheses are 95 per cent confidence intervals. AAA, abdominal aortic aneurysm.

Within the group invited for screening, 20 AAA-related deaths occurred in those who did not have a scan (Fig. 1). A further ten AAA-related deaths occurred in men with a maximum aortic diameter within normal limits at the initial scan: their mean initial maximum aortic diameter was 2·2 (range 2·0–2·9) cm. The mean time between the normal scan and their AAA-related death was 10·8 (range 3·8–15·0) years. In these ten patients, the mean age at the initial scan was 69·6 (range 65–76) years and the mean age at death was 80·4 (range 74–90) years.

An additional 17 AAA-related deaths occurred in men with a screen-detected AAA: five failed to comply with follow-up either by not attending for a follow-up scan (two patients) or by declining elective surgery (three), nine were classed as unfit for elective surgery on clinical grounds, two died during the surveillance cycle, one of whom had a 3-cm AAA detected 12 months earlier at his last scan and was undergoing bowel surgery, and one died from ruptured AAA 12 years after elective AAA surgery.

Deaths from all causes

As expected, the overall cumulative all-cause mortality did not show any differences between the two groups in this elderly population (Table 4), although this trial was not sufficiently powered to detect a difference in all-cause mortality. However, when comparing the all-cause mortality in the different groups, the rate was higher in the patients with a screen-detected AAA (141 of 170, 82·9 per cent) than in the controls (2067 of 3045, 67·9 per cent).

Discussion

In this randomized study of AAA screening, some reduction in AAA-related mortality was evident even 15 years after a scan. Although interpretation of the data should be cautious owing to the small sample size, the results offer an indication of possible lasting benefits after a single scan.

The Chichester study is the only randomized study of AAA screening with follow-up extending beyond 7 years^3,7,8. The initial trends suggested that AAA-related mortality could be reduced¹, and its procedures and early data were used as a pilot to plan and set up the Multicentre Aneurysm Screening Study. The 4-year report from this much larger trial has confirmed the benefit predicted 10 years ago³. The present data show an increasing number of AAA ruptures or AAA-related deaths, with diminishing benefits of screening over 15 years compared with the published results at 5 years¹ and 10 years². This decline was expected, as even the youngest surviving patients in this trial are now aged over 80 years, and levels of fitness for surgical treatment generally decrease with age.

The AAA-related deaths were higher in the non-attendees, with 2·6 per cent (20 of 779) dying from an AAA compared with 1·2 per cent (27 of 2216) who attended screening. Increasing attendance rates has the greatest potential to improve the benefit of screening, which could be achieved by offering screening to men of about 65 years, who are more likely to attend their appointment^1,9 and to be fit enough for treatment. The use of publicity, which was not possible during a randomized trial, could further enhance attendance rates. A national UK screening strategy is likely to screen men once at 65 years¹⁰, which should maximize both the uptake of screening and the potential for surgical treatment.

Advances in AAA treatment may make surgery safer for high-risk patients¹¹, although endovascular procedures were not available routinely at the trial hospital. The entry age range for this 15-year study was 65–80 years, and, as the prevalence of AAAs increases with age, these patients are likely to be less fit for elective surgery, which may account for the high number of patients returned to screening and the additional aneurysm-related deaths in the screened group. However, although the initial results of endovascular surgery were encouraging¹², some patients are unsuitable or do not benefit from this type of treatment, particularly those who are unfit^13,14. Of the 86 patients who satisfied the criteria for elective surgery, 32 died from other causes with their aneurysm intact. However, the proportion of all-cause deaths in the patients with AAA was higher than that of the controls, possibly because of associated atherosclerotic disease.

Ten men died from an AAA despite having a normal initial ultrasonography scan, and further investigations have shown that five had a maximum aortic diameter of 2·5–2·9 cm. A study in Gloucester demonstrated that 2·1 per cent of men have an aortic diameter of 2·6–2·9 cm¹⁵ and recommended that this subgroup should be rescanned at 5-year intervals. A Danish study screening men of 65–73 years has recommended that some patients with an ectatic aortic¹⁶ (infrarenal aortic diameter of 2·5–2·9 cm or distal: renal aortic diameter ratio of 1·2), with a prevalence of 7·5 per cent, should be rescanned after 5 years. Screening these additional patients with aortic measurements at the upper end of normal would have significant cost implications for the screening strategy. Based on the present data, surveillance of these additional patients with an aortic diameter of 2·5–2·9 cm could potentially have saved five AAA-related deaths over the 15 years. The overall rate of AAA rupture in men with an initially normal scan (AAA diameter less than 3·0 cm) was 0·56 (95 per cent c.i. 0·32 to 0·99) per 1000 person-years, and the overall rate of AAA-related death in those screened normal was 0·47 (95 per cent c.i. 0·25 to 0·88) per 1000 person-years. As these rates are low, the cost-effectiveness of repeat screening for this group needs formal evaluation. The number of men aged 65 years in this study was small, and any analysis of the 10–15-year risk of rupture in those with borderline normal aortic diameters would not be meaningful. Future meta-analyses using data from existing screening programmes may be able to offer clear guidance on this issue.

The overall 30-day elective postoperative mortality rate in this study was abnormally low compared with other publications¹⁴. This may reflect the experience and expertise of a well established vascular team who undertake about 80 AAA operations each year. It is surprising that there were so many elective procedures in the control patients and screen non-attendees, whose AAAs were identified incidentally. This may be because of the high profile that clinicians in the area give to AAAs. Inevitably, in a community-run randomized trial, there was some contamination of the control group.

There are some minor differences in the reported numbers of subjects in this study and previous publications. However, because of improved data checks and registration with the Office of National Statistics mortality surveillance system, the data used for the present analysis were as accurate and complete as they could be. No patient with a screen-detected AAA remains under surveillance, and this is therefore the last report updating the trial.

Acknowledgements

Sources of financial support were the NHS Research and Development Programme and the Medical Research Council. L.K. receives a Raymond and Beverly Sackler Studentship award.