Systematic review and meta-analysis of the growth and rupture rates of small abdominal aortic aneurysms: implications for surveillance intervals and their cost-effectiveness

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 08/30/02. The contractual start date was in November 2009. The draft report began editorial review in July 2012 and was accepted for publication in November 2012. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors' report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

© Queen's Printer and Controller of HMSO 2013. This work was produced by Thompson et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Abdominal aortic aneurysm screening

Abdominal aortic aneurysm (AAA) is an important cause of mortality, with around 4500 deaths each year in England and Wales. 1 Death due to an AAA usually occurs as a result of rupture, the overall survival rate from which is only about 20%. 2,3 AAAs are usually asymptomatic until rupture occurs, but can be detected by ultrasound (US) screening. Following detection of an AAA (aortic diameter of at least 3.0 cm), either through screening or incidentally, the principal surgical strategy is to repair the AAA before rupture occurs. However, elective open AAA repair is associated with a mortality rate in the region of 5%. 4 Although minimally invasive endovascular techniques can reduce this operative risk to < 2%,5 late complications make the long-term outcomes similar whatever operative method is chosen. 6

The evidence in favour of screening older men for an AAA by US has come from four population-based randomised trials in different countries, with the evidence being summarised in a Cochrane review. 7 In men, offering screening reduces aneurysm-related mortality by almost half [odds ratio (OR) 0.60, 95% confidence interval (CI) 0.47 to 0.78]. The data in women are too few to draw conclusions. Longer-term (10-year) follow-up of men randomised in the large UK Multicentre Aneurysm Screening Study (MASS) confirms a continued reduction in aneurysm-related mortality and indicates that offering screening is cost-effective within a 10-year time horizon. 4 Long-term modelling based on the MASS trial suggests that the lifetime cost-effectiveness of a policy of offering screening to men aged 65 years is very favourable, with an estimated incremental cost-effectiveness ratio (ICER) of £2970 (95% CI £2030 to £5430) per quality-adjusted life-year (QALY) saved. 8 Following this evidence, a national AAA screening programme for men aged 65 years was initiated in the UK. 9 However, any national AAA screening programme must have clear policies for intervention and surveillance protocols. Such policies will depend, in a large part, on the growth rate and rupture rate of small aneurysms.

Surveillance for small abdominal aortic aneurysms

The two UK AAA screening trials used diameters of either 5.5 cm or 6.0 cm, as measured by US, as the threshold for considering elective surgery, whereas the Danish screening trial used a threshold of 5.0 cm and in the Australian screening trial the intervention threshold was decided by local vascular surgeons. 7 A 5.5-cm threshold is being used in the UK NHS National Abdominal Aortic Aneurysm Screening Programme (NAAASP). 9

Screening programmes identify AAAs with a wide range of diameters, but the majority (about 90%) have diameters < 5.5 cm. 10 The evidence about managing these ‘small’ aneurysms has come from four randomised trials of intervention by open or endovascular surgical repair compared with surveillance for aneurysms 4.0–5.4 cm in diameter. 11 These trials did not show a survival benefit associated with a policy of early surgery below the 5.5-cm-diameter threshold, even after prolonged follow-up. 12 One reason that these trials have not shown a survival benefit is because the risk of rupture is very low while the aneurysm is still small.

The majority of small AAAs grow relatively slowly, but there is substantial variation in growth rates between different individuals. 13 To prevent undetected progression of an AAA to a size where there is a high risk of rupture, repeated radiological scanning is necessary. Most patients with small AAAs require surveillance for many years and undergo multiple AAA measurements.

There is no randomised trial comparing different surveillance protocols, and no recent systematic review of the growth and rupture rates of small AAAs. As a result, screening and surveillance programmes in different countries have adopted different intervals (Table 1), without a good evidence base for their choice. NAAASP employs surveillance intervals of 1 year for 3.0–4.4-cm AAAs and of 3 months for 4.5–5.4-cm AAAs. This choice corresponds to the intervals used in the large UK MASS trial, which has provided the majority of the randomised evidence worldwide (in terms of participants, person-years of follow-up and number of events) on the benefit of AAA screening. 4

Growth rates of small abdominal aortic aneurysms

Apart from the screening trials, many other observational studies have reported on growth rates of small aneurysms and recommended different frequencies of follow-up examinations. Published studies report estimates of AAA growth in different ways, making them difficult to summarise. In addition, they only rarely report the variability in growth rates between patients, which is crucial for determining appropriate surveillance intervals. 17 In many cases simple linear regression has been used to estimate growth rates. This does not account for the tendency for aneurysm growth to accelerate with increasing aneurysm diameter, nor for the bias introduced by termination of the available measurements after the threshold for surgical intervention is reached. Unbiased estimation of growth rates requires more complex statistical analysis using likelihood-based methods, for example based on multilevel modelling. 13 Application of such analytical techniques shows that aneurysm growth rates vary considerably between people. Although most aneurysms show a pattern of accelerating growth with increasing aneurysm diameter, some show a slowing of aneurysm growth, some linear growth, and some aneurysms even decrease in diameter over time.

Some studies report faster aneurysm growth in smokers and slower aneurysm growth in patients with diabetes, but evidence is inconsistent about sex differences. 13 Two recent systematic reviews have addressed the efficacy of pharmacotherapy for decreasing aneurysm growth rates. 18,19 These reviews concluded that, although beta blockers do not reduce growth rates, there is some promise that statins and other anti-inflammatory agents may reduce aneurysm growth rates. There is inconsistent evidence about the effect of blood pressure and other variables.

Rupture rates of small abdominal aortic aneurysms

The risk of rupture increases with AAA size. In patients with small aneurysms (< 5.5 cm in diameter), the risk of surgery is considered greater than the risk of rupture and a policy of surveillance is usually adopted. The low risk of rupture makes it difficult either to obtain reliable estimates of rupture rates of small aneurysms or to investigate how these rupture rates vary according to aneurysm size, sex and other patient characteristics. Reliable estimates of rupture rates are essential for planning the management of small aneurysms. Although some synthesis of data for rupture in large aneurysms has been reported previously,20 systematic reviews and meta-analyses of studies focused on the rupture rate of small aneurysms (< 5.5 cm) have not been reported. Randomised trials have reported aneurysm rupture rates of 1% per annum or less for aneurysms of 4.0–5.5 cm in diameter. 21,22 The rupture rate of smaller aneurysms appears to be even lower. 23

Need for systematic review and individual patient data analysis

To bring together the evidence on growth and rupture rates for small AAAs, a systematic review is required. However, individual studies publish their results in many different ways, making a quantitative synthesis problematic. This is compounded by the misuse of simple linear regression as an analysis tool, and by the selective (or lack of) information presented on how growth and rupture rates vary with patient characteristics. Thus, analysis of individual patient data (IPD) is likely to be necessary to obtain reliable estimates of small aneurysm growth and rupture rates, and for modelling the impact of different surveillance intervals.

Costs and cost-effectiveness

Different surveillance intervals lead to different associated costs, in terms of both the number of screens and the number of elective and emergency operations carried out. They also have different benefits in terms of life expectancy. Costs and benefits therefore have to be formally traded against each other by estimating the cost-effectiveness of different screening intervals, and expressing this as the incremental cost per QALY gained.

Overall aim of the project

This project has therefore sought to evaluate the available evidence pertinent to designing an evidence-based surveillance programme, by means of systematic reviews of the literature and analysis of IPD from surveillance studies. The choice of surveillance intervals should balance the frequency of rescreening intervals against their cost, and will depend on both aneurysm growth rates and aneurysm rupture rates. Although we focus on the UK context, and the NAAASP, our work clearly also has implications for other population-based screening and surveillance programmes.

Aims and objectives

The overall aim is to synthesise the available evidence on the growth and rupture rates of small AAAs, in such a way that can inform the appropriate choice of surveillance intervals.

The following questions are addressed:

• What are the growth rates of small AAAs?

• What are the rupture rates of small AAAs?

• How does aneurysm size influence growth and rupture rates?

• How do other individual characteristics affect growth and rupture rates?

• What are the clinical implications of varying the surveillance intervals for small AAAs?

• What are the cost implications of varying the surveillance intervals for small AAAs?

• What is the incremental cost-effectiveness of changing the surveillance intervals from those currently implemented in the NAAASP?

To answer these questions we have:

1. conducted a systematic literature review of studies using US to monitor the growth of small AAAs

2. conducted a systematic literature review of the rates of rupture of small AAAs.

As this published information was not sufficient for detailed modelling of AAA growth and rupture rates, we then:

1. 3. identified studies from these systematic reviews that could provide IPD to investigate AAA growth and rupture rates

2. 4. collated and harmonised individual data from 18 surveillance studies

3. 5. developed an extensive statistical modelling framework to analyse the growth rates of small AAAs by size and sex using likelihood-based multilevel modelling

4. 6. undertook IPD meta-analyses of growth rates and of rupture rates

5. 7. identified important patient characteristics associated with aneurysm growth and with aneurysm rupture

6. 8. modelled surveillance intervals for persons identified with small AAAs, according to aneurysm size and sex.

We used bespoke statistical analyses from the IPD meta-analyses to provide inputs for health economic modelling, and:

1. 9. developed an extended health economic modelling framework to estimate the incremental costs and cost-effectiveness of a range of surveillance strategies.

Structure of the report

Two systematic literature reviews were conducted: one of AAA growth rates and one of AAA rupture rates. The methods of these reviews are presented in Chapter 3, with results in Chapter 4. The protocols, search strategies and lists of excluded studies for these reviews are given in Appendices 1 and 2.

Under the name of the RESCAN Collaboration, IPD on small AAA surveillance were obtained from collaborating study groups. We collated data on both AAA diameters over time, to estimate growth rates, and ruptures. The collating and analysis of these data are described in Chapter 5 (with detailed statistical methods in Appendix 3). The results, given in Chapter 6, are expressed in a way that is directly relevant to the choice of surveillance intervals for different sized aneurysms.

The cost implications of different surveillance strategies, and the comparative cost-effectiveness of using them in an AAA screening programme, are then studied. The methods used for this are given in Chapter 7 (and additional details in Appendix 4), with results in Chapter 8.

Chapter 9 provides a discussion of all these component studies and conclusions are given in Chapter 10. A list of published papers arising from this work is provided in Acknowledgements.

Systematic literature review of small abdominal aortic aneurysm growth rates

Systematic literature review of small abdominal aortic aneurysm rupture rates

Systematic review and meta-analysis of small abdominal aortic aneurysm growth rates

Comparison of small aneurysm growth rates

Overall average growth rate estimates are shown in Figure 2, with separate estimates for the Huntingdon and Chichester studies. 34 There was very substantial heterogeneity between study estimates (I² = 98%), with reported growth rates ranging from – 0.33 mm/year to + 3.95 mm/year. The Huntingdon study34 was reported as having a negative growth rate and the largest growth rate was reported from a study which only included 40- to 49-mm-diameter aneurysms. 33 One study could not be included because no SE could be calculated from the information reported. 30 The random-effects meta-analysis estimate of the average aneurysm growth rate was 2.32 mm/year (95% CI 1.95 mm/year to 2.70 mm/year). Restriction of the meta-analysis to those studies that only used ultrasonography imaging after 1990 (reflecting potentially better imaging quality), the growth rate reduced slightly to 2.05 mm/year (95% CI 1.64 mm/year to 2.45 mm/year), although the heterogeneity remained high (I² = 97%).

Rates are summarised in Figure 3 for those studies that reported mean growth rates by 5-mm-diameter bands. 29,32,35–38 Estimates within these size bands generally were more homogeneous than the overall estimates, although heterogeneity still remained (I² ranged from 11% to 91%). Growth rates increased steadily with aneurysm diameter band, with a pooled mean growth rate of 1.81 mm/year (95% CI 1.55 mm/year to 2.07 mm/year) for individuals whose baseline diameter measured 3.0–3.4 cm compared with 4.96 mm/year (95% CI 4.25 mm/year to 5.66 mm/year) for individuals whose baseline diameter measured 4.5–4.9 cm. The greater number of patients in the smaller diameter bands is one reason why the widths of the CIs increase with aneurysm diameter.

The statistical method used to calculate growth rates and the study design (retrospective or prospective) were significantly associated with growth rates. Studies that used hierarchical modelling or linear regression had, on average, growth rates that were respectively 1.52 mm/year (SE 0.51 mm/year) and 0.40 mm/year (SE 0.40 mm/year) smaller than studies that used (last – first diameter)/(time elapsed) to report growth rate. Prospective studies had, on average, growth rates that were 1.11 mm/year (SE 0.39 mm/year) smaller than either retrospective or partially retrospective designs, although this effect was no longer significant after adjusting for the method used to calculate growth rate.

A meta-regression with average aneurysm diameter as an explanatory variable resulted in a strong trend between diameter and growth rate (Figure 4). After adjusting for average diameter, between-study heterogeneity was estimated to be about twice that of the heterogeneity between estimates within a study. A 1-cm increase in AAA diameter was found to be associated with a 1.62 mm/year (SE 0.20 mm/year) increase in growth rate. The age and sex of included patients was available for nearly all studies and were assessed as explanatory variables but found to be not statistically significant; a 10% increase in the percentage of women in a study increased the growth rate by a modest 0.13 mm/year (SE 0.26 mm/year), whereas a 5-year increase in mean study participant age decreased the growth rate by 0.01 mm/year (SE 0.56 mm/year).

Based on the linear relation between growth rate and aneurysm diameter (see Figure 4), the size of aneurysms is expected to grow exponentially with time. Given the meta-regression relationship shown, aneurysms with diameters of 3.0, 3.5, 4.0, 4.5 and 5.0 cm would be expected, on average, to take 9.6, 6.2, 4.0, 2.3 and 1.1 years, respectively, to reach 5.5 cm.

However, there was insufficient published data for the detailed modelling of AAA growth rates necessary to better inform surveillance intervals. For this, IPD are necessary, as described in Chapters 5 and 6.

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram – small AAA growth studies.

FIGURE 2.

Overall AAA growth rate estimates, one for each study, by baseline AAA diameter size range. Estimates are pooled using a random-effects meta-analysis, due to considerable heterogeneity (I² = 98%). An estimate from Schewe et al. 30 could not be included due to no SE being available.

FIGURE 3.

Aneurysm growth rate by 5-mm size ranges of baseline aneurysm diameter: random-effect meta-analyses conducted within subgroups.

FIGURE 4.

Meta-regression of AAA growth rate estimates by AAA diameter. The overall regression line is shown by the solid bold line. The mid-point of the reported size range is used to calculate AAA diameter for range-specific estimates, whereas the mean baseline diameter is used for overall estimates. Estimates from the same study are joined by a dotted line. The circles have an area in proportion to the amount of information (inverse variance of the estimate).

TABLE 2 - Published growth rates of small AAAs (initial diameter of 3.0 to 5.5 cm)

A, anterior; AC, author communication; AP, anterior–posterior; P, posterior.

Patient group extended in Vega de Céniga et al. , 2008. 33

Studies listed by author in alphabetical order.

Figures shown in bold have been subsequently calculated from the published data using the methods described in the text.

Author, year	Country, enrolment dates	Number of patients (women)	Aortic diameter measured [mean baseline diameter (mm)]	US/CT	Frequency of follow-up (intervention threshold, if specified)	Length of follow-up	Method of growth rate estimation and reporting	Baseline AAA size range(s) (mm)	Growth rates reported in mm/year (SE)
Brady et al., 200413	UK, 1991–5	1743 (387)	Maximum external AP [43]	US	≥ 50 mm: every 3 months; < 50 mm: every 6 months (5.5 cm)	3312 person-years	Method: random-effects models (linear, quadratic)	Overall: 28–85 (n = 1743)	2.6 (0.03)
							Mean growth rates reported
Brown et al., 200329	Canada, 1976–2001	895 (207)	Not reported [40]	US (1976–82); CT (1983 onwards)	6-monthly	3088 person-years	Method: not reported	Overall: 30–49 (n = 895)	3.8 (0.2)
							Mean growth rates reported	30–34 (n = 191)	2.8 (0.3)
								35–39 (n = 204)	3.4 (0.3)
								40–44 (n = 306)	4.5 (0.3)
								45–49 (n = 194)	5.2 (0.4)
Karlsson et al., 200939	Sweden, 2002–5	211 (50)	Wider of the external AP and transverse [40 in both groups]	US	6-monthly	18–36 months	Method: linear regression	Overall: 35–49 (n = 211)	2.2 (0.1)
							Median growth rates reported
Propranolol Aneurysm Trial Investigators, 200240	Canada, 1993–8	548 (87)	External AP [38 in both groups]	US	6 months (5.0 or 5.5 cm by centre)	Mean 2.5 ± 1.1 years (range 0–4.5 years)	Method: unclear (last – first diameter)/time or linear regression	Overall: 30–50 (n = 531) [Growth rate only calculated for 531 patients]	2.4 (0.1)
							Mean growth rates reported
Lederle et al., 200222	USA, 1992–6	567 (2)	Maximal external cross-sectional in any plane [47]	US; CT (baseline and ≥ 53 mm)	Every 6 months	Mean 3.0 years	Method: (last – first diameter)/time	Overall: 40–54 (n = 567)	3.2 (0.08)
							Median growth rate reported
Lindholt et al., 200035	Denmark, 1994–8	151 (0)	Maximum external AP [mean diameter not reported]	US	Annually (5.0 cm)	1–5 years	Method: (last – first diameter)/time	Overall: 30–49 (n = 151)	2.8 (0.20)
							Mean growth rate reported
McCarthy et al., 200337	UK, 1990–2002	496 (0)	Maximum AP	US	26–39 mm: 12 months; > 40 mm: 6 months (5.5 cm)	Mean 4.2 (SD 2.4) years	Method: (last – first diameter)/time	Overall: 30–39 (n = 496)	1.9 (0.1)
			External A to internal P [34]				Median growth rate reported
Santilli et al., 200236	USA, 1990–6	790 (0)	Wider of external AP and lateral [33]	US	At least one further scan > 90 days after initial scan (5.5 cm)	3071 person-years	Method: not reported	Overall: 30–39 (n = 790)	1.6 (0.07)
							Mean growth rates reported	30–34 (n = 578)	1.5 (0.08)
								35–39 (n = 212)	2.0 (0.13)
Schewe et al., 199430	Germany, 1976–93	134 (not reported)	Mean of AP and transverse [37]	US	6 months	536 person-years (5.0 cm)	Method: (last – first diameter)/time	30–39 (n = 69)	2.3 (no SE)
							Mean growth rates reported	40–49 (n = 30)	3.1 (no SE)
								> 50 (n = 11)	3.9 (no SE)
Schlosser et al., 200841	Holland, 1996–2005	147 (16)	AP diameter [39]	US	30–39 mm: 12 months; 40–55 mm: 6 months (5.5 cm)	Median 3.3 years (range 0.5–11.1 years)	Method: linear regression	Overall: 30–55 (n = 147)	2.5 (0.26)
						599 person-years	Mean growth rate reported
Schouten et al., 200631	Holland, 2002–5	150 (24)	Maximum external AP [38]	US	6–12 months (5.0 cm)	Median 3.1 years	Method: linear regression	Overall: 25–53 (n = 150)	2.95 (0.23)
							Mean growth rate reported
Solberg et al., 200538	Norway, 1994–5	234 (49)	External transverse or AP [35]	US	Every 3 or 6 months	1212 person-years	Method: linear regression	Overall: 30 to > 49 (n = 234)	1.82 (0.14)
							Mean growth rates reported	30–34 (n = 87)	1.80 (0.25)
								35–39 (n = 58)	1.75 (0.14)
								40–44 (n = 23)	2.31 (0.48)
								45–49 (n = 11)	3.36 (0.95)
								> 49 (n = 9)	2.63 (1.57)
Vardulaki et al., 199834	UK, Chichester 1983–96; Huntingdon 1991–6	Chichester: 519 men and women; Huntingdon: 498 men (0 women)	In Chichester the wider of AP and transverse measurements	US	3–12 months (Chichester: 6.0 cm initially; Huntingdon: 4.5 cm)	Chichester: mean 3.4 years	Method: random-effects model based on exponential growth	Chichester:	0.9 (0.1)
			External AP in Huntingdon [mean diameters not reported]			Huntingdon: mean 2.3 years	Percent change in growth rates reported	Overall: 30 to ≥ 50 (n = 519)
								30–39	0.7 (0.2)
								40–49	1.5 (0.4)
								≥ 50	3.5 (0.7)
								Huntingdon:	− 0.3 (0.2)
								Overall: 30–49 (n = 498)
								30–39	– 0.6 (0.3)
								40–49	0.9 (0.6)
aVega de Céniga et al., 200632	Spain, 1988–2004	352 (19)	Wider of external AP and transverse [36]	US (for 30–39 mm); CT (for 40–49 mm)	Annually (group 30–39)	Overall mean 55.2 ± 37.4 months (range 6.3–200 months)	Method: (last – first diameter)/time	Overall: 30–49 (n = 352)	2.87 (0.23)
					6-monthly (group 4.0–4.9) (5.0 cm)		Mean growth rates reported
Vega de Céniga et al., 200833	Spain 1998–2004	195 (12)	Wider of maximum external AP or transverse [42.5]	CT	6 months (5.0 cm)	Mean 50 ± 36.4 months (range 6.5–194 months)	Method: (last – first diameter)/time	Overall: 40–49 (n = 195)	3.95 (0.43)
							Mean growth rate reported

TABLE 3 - Quality assessment of studies included in the systematic review

C, Chichester; H, Huntingdon; N, no; obs, observational study; P, prospective; PR, both prospective and retrospective; R; retrospective; RCT, randomised controlled trial; U, unclear; Y, yes.

To observation or surgery.

Quality aspect	Brady et al., 200413	Brown et al., 200329	Karlsson et al., 200939	Propranolol Aneurysm Trial Investigators, 200240	Lederle et al., 200222	Lindholt et al., 200035	McCarthy et al., 200337	Santilli et al., 200236	Schewe et al., 199430	Schlosser et al., 200841	Schouten et al., 200631	Solberg et al., 200538	Vardulaki et al., 199834		aVega de Céniga et al., 200632	Vega de Céniga et al., 200833
													C	H
After 1990 (image quality)	Y	N	Y	Y	Y	Y	Y	Y	N	Y	Y	Y	N	Y	N	Y
Standardisation of imaging	Y	N	Y	Y	N	Y	Y	N	N	N	N	N	N	N	N	N
Variability of diameter given	Y	N	N	Y	For CT only	Y	N	N	N	N	Y	N	N	N	N	N
Study type (P, R, PR)	P	PR	P	P	P	P	P	P	PR	P	R	P	P	P	PR	PR
Study type (RCT/obs)	RCT + obs	Obs	RCT	RCT	RCT	RCT + obs	Obs	Obs	Obs	Obs	Obs	Obs	Obs	Obs	Obs	Obs
Inclusion to obs comprehensive	N	Y				N	Y	N	N	Y	N	N	U	U	Y	Y
Intervention policy stated at start (diameter, cm)	Y (randomised at 4.0 or 5.5a)	N	N	Y (either 5 or 5.5)	Y (randomised at 4.0 or 5.5a)	Y (5.0)	Y (5.5)	Y (randomised at 4.0 or 5.5a)	Y (5.0)	Y (5.5)	Y (5.0)	N	Y (6.0, later 5.5)	Y (4.5)	Y (5.0)	Y (5.0)
Patient censorship defined	Y	N	Y	Y	Y	Y	Y	Y	N	Y	U	Y	Y	N	U	Y

Systematic review and meta-analysis of small abdominal aortic aneurysm rupture rates

Comparison of small aneurysm rupture rates

The seven studies reporting on aneurysms in the baseline diameter range of 3.0 to 5.5 cm, and with follow-up restricted to this range (conditional follow-up to 5.5 cm in diameter, larger aneurysms excluded), provided rupture rates estimated as varying from 0 to 1.61 ruptures per 100 person-years (Figure 6). The studies in the figure are sorted according to the mid-point of the diameter range and exhibit a slight trend towards higher rupture rates for larger diameters (for example, no ruptures were reported within 12 months of a measurement of < 4.0 cm). Nevertheless, the size range only partly explains the considerable heterogeneity between the rupture rate estimates, demonstrated by an I² value of 89%. Two studies estimate the rupture rate to be greater than 1.0 rupture per 100 person-years,23,43 whereas the point estimates from the remaining five studies22,29,44,45,47 are all below 0.5 ruptures per 100 person-years. Owing to the large statistical and clinical heterogeneity, a formal synthesis of the results (meta-analysis) was not carried out.

Rupture rates estimated from total reported ruptures and follow-up (including aneurysms > 5.5 cm) from all 14 studies22,23,29,32,36,37,40–47 (Figure 7) ranged from 0 to 2.51 ruptures per 100 person-years. In this figure, the studies are sorted according to the mid-point of the baseline diameter range. The between-study variation is even more extreme, with an I² value of 96%.

The published information on rupture rates is thus inadequate for the detailed modelling needed to inform surveillance intervals. For this, IPD are necessary, as described in Chapters 5 and 6.

FIGURE 5.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram – small AAA rupture studies.

FIGURE 6.

Rupture rates (per 100 person-years) for small AAAs in each study, reporting conditional follow-up to 5.5-cm threshold, sorted by reported size range (total 5934 patients). Aneurysms reaching > 5.5 cm have been excluded. These studies are depicted in order of increasing aneurysm size range, but the distribution of diameters within these size ranges is not available.

FIGURE 7.

Total rupture rates (per 100 person-years) for small AAAs in each study, sorted by baseline diameter range (total 9779 patients). Including follow-up > 5.5 cm. These studies are depicted in order of increasing aneurysm baseline size range, but the distribution of diameters within these size ranges and over time is not available.

TABLE 4 - Studies reporting on the rupture rate of small AAAs (3.0–5.5 cm in diameter)

NR, not reported.

For many studies the conditional follow-up for an aneurysm diameter of 3.0 to 5.5 cm, to aneurysm rupture or death could not be established.

Author, year and country	Total number of patients (women) [baseline AAA diameter]	Total ruptures per total person-years reported	Total small aneurysm ruptures per small aneurysm person-years [diameter range for subgroups] (fatal ruptures by 30 days)	Rate of small aneurysm ruptures per 100 person-years (95% CI)	Comments
Brown, 199923	2257 (465) [3.0–9.7 cm]	103/4102	67/3215 [3.0–5.9 cm]	2.08 (1.62 to 2.65)	Person-years for the 3.0–4.9 cm size range calculated from number of ruptures per reported rupture rate
UK			27/2600 [3.0–4.9 cm] (fatal NR)	1.04 (0.68 to 1.51)
			40/615 [5.0–5.9 cm] (fatal NR)	6.50 (4.65 to 8.86)
Buckenham, 200742	198 (50) [2.8–7.8 cm]	5/325.05	3/unknown (3 fatal)	Unknown	Threshold diameter for surgery changed from 5.5 cm to 5.0 cm in June 2004 Person-years estimated from median follow-up
New Zealand
Brown, 200343	372 (99) [5.0–5.4 cm]	9/560	9/560 [5.0–5.4 cm] (fatal NR)	1.61 (0.73 to 3.05)
Canada
Scott, 200544	1333 (0) [3.0–5.4 cm]	36/5465.3	12/3110.8 [3.0–5.4 cm] (fatal NR)	0.39 (0.20 to 0.67)
England
Brown, 200329	895 (207) [3.0–4.9 cm]	0/3088	0/3088	0.00 (0.00 to 0.12)	Some patients also may be reported in study Brown 200343
Canada
Santilli, 200236	790 (0) [3.0–3.9 cm]	0/3071.32	0/unknown	Unknown	Cause of death only available in 43% of patients, so some deaths from rupture may have been excluded Only included patients with more than one scan
USA
Reed, 199745	176 (64–69) [< 3 to > 6 cm]	11/862	1/333 [3.0–5.0 cm] (fatal NR)	0.30 (0.01 to 1.67)	Rupture rates calculated from last US diameter Only included patients with more than one scan
USA
Guirgius, 199146	300 (89) [2.5–5.0 cm]	14/850	2/unknown (fatal NR)	Unknown
Canada
Armstrong, 200747	334 (3) [4.0–5.4 cm]	2/946	2/946 [4.0–5.4 cm] (1 fatal)	0.21 (0.03 to 0.76)	Surveillance group only
USA
Vega de Céniga, 200632	352 (19) [3.0–5.0 cm]	2/1619	0/unknown	Unknown	Only includes patients with more than one scan
Spain
Propranolol Aneurysm Trial Investigators, 200240	552 (88) [3.0–5.0 cm]	At least 3/1380	At least 2/unknown (< 5.5 cm fatal NR 2/3 total ruptures)	Unknown	Excluded patients whose AAA had grown by < 2 mm in previous year No sudden deaths reported as rupture after mortality committee review
Canada
McCarthy, 200337	1423 (0) [2.5–4.0 cm]	2/5045	2/unknown (2 fatal)	Unknown	Unclear how many person-years are included beyond 5.5 cm
England
Lederle, 200222	567 (5) [4.0–5.4 cm] surveillance group only	11/1833	9/1833 [4.0–5.4 cm] (8 fatal)	0.49 (0.22 to 0.93)	Two covered ‘hole in wall’ at elective repair cases excluded: not true ruptures
USA
Schlosser, 200841	230 (23) [3.0–5.5 cm]	7/755	6/unknown (< 5.5 cm fatal NR; 6/7 total ruptures)	Unknown
Netherlands

TABLE 5 - Further characteristics of included studies reporting on the rupture rate of small AAAs (3.0–5.5 cm in diameter)

MRI, magnetic resonance imaging; NR, not relevant; P, prospective study; PR, prospective and retrospective contributions; R, retrospective study; PY person-year.

Elective repair is contingent on patient fitness and consent; in addition, most studies reported being willing to repair the aneurysm for aneurysm-related symptoms and rapid growth (> 1 cm/year).

Author, year and type of study (R, P, PR)	Inclusion criteria (aortic diameter ranges)	Age (years) of participants (mean)	Surveillance policy	Intervention policya	Ascertainment of rupture described?	Time between last diameter and rupture reported?	Rupture rates as published
Brown and Powell, 199923	AAA >3 cm (3.0–3.9; 4.0–5.5; 5.6–9.7 cm)	70	US	> 5.5 cm	Yes	Yes	Annual rupture rate: 2.2% (95% CI 1.7% to 2.8%); large and small aneurysms
P			4–4.9 cm: 6 months
			5–5.5 cm: 3 months
Buckenham et al., 200742	AAA 3–5.5 cm; fit for surgery (2.8–7.8 cm)	73 (median)	US, varied with sex and size	Men > 5.5 cm	No	No	NR
PR				Women > 5.0 cm
Brown et al., 200343	AAA 5–5.9 cm; unfit for surgery	73	US, 6-monthly	All patients unfit for surgery	No	No	1% per year (men); 3.9% per year (women); large and small aneurysms
P
Scott et al., 200544	AAA 3–5.5 cm; men aged 65–74 years	NR	US	5.5 cm	No	No	36 per 1000 person-years (95% CI 15 to 75 per 1000 person-years) for those with an AAA > 5.5 cm at recall
P			3–4.4 cm: yearly
			4.5–5.4 cm: 3-monthly
Brown et al., 200329	AAA 3.0–5.0 cm	69	US, CT, 6-monthly	NR	No	NR	No ruptures
P
Santilli et al., 200236	AAA 3–3.9 cm	69	US, variable	NR	No	NR	No ruptures
P
Reed et al., 199745	AAA not for surgery (< 3 to > 8 cm)	74	US, NR	NR	Yes	Yes	By last diameter: 3–3.99 cm: 0 ruptures per PY; 4–4.99 cm: 0.007 ruptures per PY; 5–5.99 cm: 0.11 ruptures per PY
P
Guirgius and Barber, 199146	AAA initially managed non-operatively (2.5–9.3 cm)	70	US, CT, 6-monthly	NR	No	No	NR
PR
Armstrong et al., 200747	AAA 4–5.4 cm	NR	US, CT, MRI	> 5.4 cm	No	No	Cumulative incidence only reported
PR			3–4 cm: yearly
			4–5.4 cm: biannually
Vega de Céniga et al., 200632	AAA 3–4.9 cm	71	US: yearly if < 4 cm	> 5.0 cm	No	No	NR
P			CT: every 6 months
Propranolol Aneurysm Trial Investigators, 200240	AAA 3–5.0 cm; no contraindications to propranolol	69	US, 6-monthly	Variable, > 4.5 cm or > 5.0 cm by centre	No	No	NR
P
MaCarthy et al., 200337	AAA 2.6–3.9 cm	65	US	> 5.5 cm	No	Yes	2-year rupture rate: 1.4% (3.5–3.9-cm group), 0% at 3 years for smaller aortas
P			2.6–3.9 cm: yearly
			4.0 cm: 6-monthly
Lederle et al., 200222	AAA 4–5.4 cm; randomised to surveillance	68	US, 6-monthly	> 5.4 cm	Yes	No	< 0.6% per year of follow-up of unrepaired aneurysms
P
Schlosser et al., 200841	AAA 3–5.5 cm	66	US	> 5.5 cm	No	No	0.9% per PY, including aneurysms > 5.5 cm
P			3–3.9 cm: yearly
			4–5.5 cm: 6-monthly

Data collection

Under the name of the RESCAN Collaboration, requests for IPD were addressed to the corresponding authors of 25 studies, each including more than 100 patients with small AAAs (3.0–5.5 cm in diameter) in a surveillance programme. Data sets for potential inclusion had been identified through a systematic literature search up to the end of 2009, as described in Chapters 3 and 4, and manual searching of abstracts from vascular surgical meetings during the interval 2007–10. Additional studies published up to September 2012 are discussed in Chapter 9.

The items requested included patient age, sex, sequential aneurysm diameters, ethnicity, smoking history, body mass index (BMI), presence of diabetes, dates of aneurysm repair, aneurysm rupture or death. The method used for US AAA diameter assessment was recorded for each study as internal (inner to inner luminal surface) or external (outer to outer aortic wall). A few studies used CT scans for some AAA measurements. Supplementary information for baseline histories of medication use, coronary and peripheral arterial disease as well as blood pressure was requested, if available. To allow for changing environmental factors (e.g. smoking), improvements in imaging quality as well as improving cardioprotective drug therapy, we also included year of patient enrolment as a variable. Additional information regarding AAA rupture (including death due to rupture), operative AAA repair and non-AAA death was collected for each patient where this was available. The definition of aneurysm rupture was pragmatic, based on locally used definition and reporting.

The RESCAN collaborators provided data for 18 studies, which were then subject to cleaning through an iterative process of correspondence with the study investigators to improve data quality, prior to the harmonisation of the data sets. All data sets were anonymised and, therefore, the study was subject to an ethical waiver.

For each individual, the baseline measurement was defined as the first measurement recorded between 3.0 and 5.4 cm. Measurements taken prior to a patient reaching a threshold of 3.0 cm were removed from the data sets. All data following baseline measurement were used until either a rupture event or censoring occurred. Censoring was defined as the first occurrence of: (1) a diameter measurement ≥ 5.5 cm (with this measurement included in the data set to properly account for informative dropout);17 (2) non-rupture-related death; (3) an elective AAA operation; or (4) the administrative censoring date of the data set. A patient's series of measurements was not censored if the aneurysm diameter dropped to < 3.0 cm, thus allowing some patients to have negative growth rates.

Focus group

In order to determine the level of risk of rupture or progression to 5.5 cm that would be acceptable to patients with small AAA, we conducted a patient focus group meeting. Men with small AAAs under surveillance in the Leicestershire AAA Surveillance Programme were invited to attend a meeting with a consultant vascular surgeon (MJ Bown) after verbal and written information about the RESCAN project was provided. A further explanation of the aims of the RESCAN project and its potential outcomes was given to the patients who attended the meeting prior to discussing the specific points below. Although the main aim of this focus group was to determine what risk of rupture and what risk of progression to a size threshold of 5.5 cm was acceptable to patients, a number of other issues relating to the surveillance of small AAA were discussed and these results are included below. The issues directly discussed were the acceptability of different intervals between surveillance scans, whether or not surveillance should be personalised, what factors patients think would improve surveillance, and what risks of rupture/progression were acceptable.

Statistical methods

Full details of the statistical methods are provided in Appendix 3. A summary is given here.

Study characteristics

Individual patient data were obtained from 18 different studies from Europe, Canada, the USA and Australia, with patients being admitted to these studies between 1983 and 2008; a summary of these studies is given in Table 6. A total of 15,475 patients are included, with an average of 4617 AAA measurements and 4.0 years of follow-up per study (Table 7). Not all patient demographics were available from all studies, and in some studies the data were incomplete for some individuals. Ethnicity was poorly reported. Baseline drug history was fully available in seven studies (see Table 7). Prevalence of drug use varied considerably between these studies (ACE inhibitors 9–41%, beta blockers 14–52%, calcium channel blockers 12–28%, statins/lipid-lowering drugs 0–78%, antiplatelet agents 27–94%). Data were not available from two randomised controlled trials (RCTs) and five other studies with a total of 2487 persons (see Appendix 3).

Study-specific thresholds for surgical intervention varied from 4.5 cm up to 6.0 cm (see Table 6). The majority of studies measured external (outer-to-outer) wall diameters, although three studies10,37,53 measured internal diameters. Rupture and censoring events were recorded in 14 out of the 18 studies allowing rupture rates to be investigated. The remaining four studies were used in the growth analysis only (see Table 6).

Four studies, Leeds (Professor D Julian A Scott, University of Leeds, 2010–12, personal communication), PIVOTAL,51 Galdakao32 and Stirling (Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication), measured diameters using both US and CT scans. CT measurements were on average significantly larger than US measurements [Leeds 3.91 mm (SE 0.33 mm), PIVOTAL 1.75 mm (SE 0.21 mm), Galdakao 1.77 mm (SE 0.10 mm), Stirling 2.46 mm (SE 0.27 mm)]. However, there was no evidence of an effect of modality on estimated growth rates (i.e. a CT measurement was found to be larger than an US measurement throughout follow-up by a constant amount). Therefore, within each study an additive adjustment could be applied so that all measurements were included in the analysis of growth rates.

Small abdominal aortic aneurysm growth rates in men

The overall pooled estimates of AAA growth in men increased by about 0.5 mm/year for each 0.5-cm increase in baseline AAA diameter (Figures 8 and 9). The average estimated growth rate in men was 1.3 mm/year (95% CI 1.0 mm/year to 1.5 mm/year) for a 3.0-cm aneurysm and 3.6 mm/year (95% CI 3.3 mm/year to 3.9 mm/year) for a 5.0-cm aneurysm (Table 8). However, there was also substantial heterogeneity in growth rates between studies, so the 95% prediction intervals were wide. Much of this heterogeneity is unexplained. Although smokers and non-diabetics have higher rates of growth (see Predictors of abdominal aortic aneurysm growth rates), these factors did not explain the heterogeneity between studies: adjusting for baseline AAA diameter, smoking and diabetes only reduced the overall percentage of variation due to heterogeneity from 97.5% to 94.1%. Excluding the three studies (Chichester,53 Gloucestershire37 and MASS10) that measured internal diameters, the pooled mean growth rates decreased by between 0.10 mm/year and 0.13 mm/year, depending on baseline diameter, but the percentage of variation due to heterogeneity only decreased by between 1.5% and 3.3%.

The length of time taken for men under surveillance to have a 10% chance of reaching the threshold size of 5.5 cm decreased substantially with baseline AAA diameter (Figure 10). For men with a 3.0-cm aneurysm, the estimated mean time taken to have a 10% chance of reaching 5.5 cm was 7.4 years (95% CI 6.7 years to 8.1 years). The corresponding mean times for 4.0-cm and 5.0-cm aneurysms were 3.2 years (95% CI 3.0 years to 3.4 years) and 0.7 years (95% CI 0.6 years to 0.8 years), respectively (see Table 8). Again, the prediction intervals were substantially wider than the CIs because of the heterogeneity between studies.

Small abdominal aortic aneurysm rupture rates in men

Overall the number of ruptures was small, with 178 reported events among 11,262 men under surveillance and 50 among 1314 women. Crude rupture rates in men in each study varied between 0 and 0.77 ruptures per 100 person-years (see Table 6). There were 12 studies with at least two ruptures available to assess the relationship between rupture rates and AAA diameter in men, and five studies in women (see Table 6). Rupture rates in men increased with baseline AAA diameter, approximately doubling for every 0.5-cm increase (Figures 11 and 12). The average rupture rates were 0.05 per 100 person-years (95% CI 0.03 to 0.07 per 100 person-years) for a 3.0-cm AAA and 0.64 (95% CI 0.43 to 0.95 per 100 person-years) for a 5.0-cm AAA. The degree of heterogeneity between studies also increased with baseline AAA diameter; I² ranged from 0% to 82% for diameters 3.0–5.4 cm. It was not possible to investigate the cause of this heterogeneity, due primarily to the small number of ruptures observed in each study making multivariate modelling infeasible. The time taken for men under surveillance to reach a rupture risk of 1% decreased with baseline AAA diameter (Figure 13). For AAA diameters in the range 3.0–4.5 cm the average time to reach a rupture risk of 1% was more than 2 years (see Table 8). For a 5.0-cm AAA, a 1% risk of rupture was reached on average in 1.4 years (95% CI 1.2 years to 1.8 years).

Small abdominal aortic aneurysm growth and rupture rates in women

The dependence of growth and rupture rates on AAA diameter was very similar in women and men (Figures 14 and 15 compared with Figures 8 and 11). Heterogeneity between studies was slightly less for women. Although absolute growth rates were similar for women and men (particularly for larger baseline AAA diameters), there were marked differences in the absolute risks of rupture. Women had about a fourfold greater rupture risk for all AAA sizes (see Predictors of abdominal aortic aneurysm growth rates), and reached a rupture risk of > 1% in a much shorter time than men (see Table 8).

Predictors of small abdominal aortic aneurysm growth rates

Unadjusted analysis of each demographic variable on aneurysm growth revealed significant effects among a number of individual study estimates. However, there was considerable heterogeneity between study estimates for many of the demographics considered (as measured using I²). When estimates were pooled using random-effects meta-analysis (with adjustment for initial aneurysm diameter), current smoking was significantly associated with faster aneurysm growth rates, whereas increased BMI, diabetes, increased pulse pressure, and history of cardiovascular disease were associated with slower growth rates (Table 9). After adjustment for all demographics, medical history and drug history, only current smoking and diabetes remained significant predictors of aneurysm growth (see Table 9). The growth rate in current smokers was on average 0.35 mm/year (SE 0.07 mm/year; p < 0.001) faster than in ex or never smokers, whereas diabetics had growth rates that were on average 0.51 mm/year (SE 0.10 mm/year; p < 0.001) slower than non-diabetics. This evidence is consistent across the studies as demonstrated in Figure 16. Men and women had similar growth rates.

Associations between baseline use of cardiovascular drugs and aneurysm growth rate are shown in Table 10. Univariate analyses suggest that many of the drugs could be protective against aneurysm growth, with the largest effect coming from the use of statins/lipid-lowering drugs. However, the magnitude of these effects was relatively small and could be confounded with their prescription in specific risk populations. Indeed, after adjustment for other demographics, the pooled meta-analysis estimates were no longer statistically significant for any class of drug investigated (see Table 10). Although the use of many of these drugs increased over time, calendar year of enrolment did not appear to influence aneurysm growth rate.

Predictors of small abdominal aortic aneurysm rupture rates

The differences in the crude rupture rates between the studies were large (from 0.07 up to 1.1 per 100 person-years). However, some of these differences could be explained by different baseline diameter ranges, different lengths of follow-up after last diameter measurement and proportions of women in each study. After adjustment for aneurysm diameter, the influence of demographics and medical history on rupture rates revealed a strong association between smoking and rupture [pooled hazard ratio (HR) 2.02, 95% CI 1.33 to 3.06; p = 0.001], and a far higher risk of rupture for women than men (pooled HR 3.76, 95% CI 2.58 to 5.47; p < 0.001) (Figure 17). The risk of rupture was also found to increase significantly for older patients, patients enrolled earlier in calendar time, those with lower BMI, and with higher mean arterial blood pressure or pulse pressure (Table 11). It was not possible to conduct fully adjusted analyses of these effects because of the small number of ruptures observed. However, effects did remain consistent after adjustment for age and calendar time.

Any association between cardiovascular drugs and aneurysm rupture was difficult to investigate with the present data because of the low prevalence of rupture and insufficient reporting of the drugs. Only six studies13,39,40,51,52,55 reported both rupture events and use of drugs. Of these, only two estimates, on average, could be obtained for each class of drug, none of which showed statistical significance (Table 12).

Inputs to health economic modelling

These IPD from 18 surveillance studies (see Table 6) were next used to provide inputs relating to growth and rupture rates for the health economic modelling, as described in Chapters 7 and 8.

FIGURE 8.

Forest plots of mean AAA growth rate (estimate and 95% CI) in men according to baseline AAA diameter. (a) Mr Simon Parvin, Royal Bournemouth Hospital, 2010–12, personal communication; (b) Professor D Julian A Scott, University of Leeds, 2010–12, personal communication; (c) Mr Matthew J Bown, University of Leicester and the NIHR Cardiovascular Biomedical Research Unit, 2010–12, personal communication; (d) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (e) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 9.

Estimated mean growth rate given baseline AAA diameter, in men, by study and overall (with 95% CIs and prediction intervals). The area of each symbol is inversely proportional to its SE. (a) Mr Simon Parvin, Royal Bournemouth Hospital, 2010–12, personal communication; (b) Professor D Julian A Scott, University of Leeds, 2010–12, personal communication; (c) Mr Matthew J Bown, University of Leicester and the NIHR Cardiovascular Biomedical Research Unit, 2010–12, personal communication; (d) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (e) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 10.

Estimated time given baseline AAA diameter for which there is a 10% chance that the threshold diameter for surgery (5.5 cm) has been reached, in men, by study and overall (black diamonds represent 95% CIs, error bars are 95% prediction intervals). (a) Mr Simon Parvin, Royal Bournemouth Hospital, 2010–12, personal communication; (b) Professor D Julian A Scott, University of Leeds, 2010–12, personal communication; (c) Mr Matthew J Bown, University of Leicester and the NIHR Cardiovascular Biomedical Research Unit, 2010–12, personal communication; (d) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and(e) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 11.

Forest plots of rupture risk (estimate and 95% CI, log-scale) in men according to underlying AAA diameter. (a) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (b) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 12.

Rupture risk according to underlying AAA diameter, in men, by study and overall (log-scale, with 95% CIs and prediction intervals). The area of each symbol is inversely proportional to its SE. (a) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (b) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 13.

Estimated time given underlying AAA diameter for which there is a 1% chance of rupture, in men, by study and overall (black diamonds represent 95% CIs, error bars are 95% prediction intervals). (a) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (b) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 14.

Forest plots of mean AAA growth rate (estimate and 95% CIs) in women according to baseline AAA diameter. (a) Professor D Julian A Scott, University of Leeds, 2010–12, personal communication; (b) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (c) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 15.

Forest plots of rupture risk (estimate and 95% CIs, log-scale) in women according to underlying AAA diameter. a, Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; b, Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication.

FIGURE 16.

Effect of (a) smoking status (current vs. ex/never) and (b) diabetes on aneurysm growth rates (mm/year) in individual studies and meta-analysis: adjusted estimates with 95% CI. (a) Professor D Julian A Scott, University of Leeds, 2010–12, personal communication; and (b) Mr Matthew J Bown, University of Leicester and the NIHR Cardiovascular Biomedical Research Unit, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

FIGURE 17.

Effect of (a) sex (female vs. male) and (b) smoking (current vs. ex/never) on aneurysm rupture rates in individual studies and meta-analysis: HRs (95% CI) adjusted for aneurysm diameter. (a) Professor Charles N McCollum, University Hospital of South Manchester, 2010–12, personal communication; and (b) Mr Richard Holdsworth, Stirling Royal Infirmary, 2010–12, personal communication. UKSAT, United Kingdom Small Aneurysm Trial.

Introduction

Previous studies have estimated the cost-effectiveness of the screening strategy that was evaluated in the MASS trial. Estimates of the cost-effectiveness of this strategy have been adjusted over time, principally to reflect the emerging long-term data from the trial. The original cost-effectiveness estimates were based simply on 4 years of follow-up and estimated a mean cost per life-year gained within that truncated period of £28,400 (95% CI £15,000 to £146,000)56 at 2000–1 prices. Using the observed data at 10 years of follow-up, the estimate had fallen to £7600 (95% CI £5100 to £13,000) despite revaluing costs to 2008–9 prices. 4 Based on informal modelling, the original cost-effectiveness paper had suggested that the cost-effectiveness over 10 years would indeed be around £8000 per life-year saved. 56

However, it is clearly recognised that an investment in a screening programme needs to be assessed over a longer period that does not cut short the benefits from avoided aneurysm-related mortality and will require a formal model. Such a model, taking a 30-year perspective but initially based on the 4-year MASS results, was developed and this estimated the cost per life-year gained over a 30-year period as £2320 (with a 95% uncertainty interval of £1600 to £4240). 8 As the period of follow-up of the MASS trial has extended, the uncertainty associated with the longer-term effects of screening has been reduced, and scope now exists to compare the model results at 10 years and as necessary to recalibrate the model to more accurately reflect the longer-term observed data.

All these estimates have assumed a screening programme in elderly men with a surveillance pattern of: no recall for patients with an aortic diameter of < 3.0 cm; yearly rescanning of patients with an aortic diameter of 3.0–4.4 cm; 3-monthly rescans for patients with an aortic diameter of 4.5–5.4 cm; and consideration for surgery if an aortic diameter of ≥ 5.5 cm. This surveillance strategy adopted in MASS was based on the data available and expert clinical opinion at the time of the planning of the trial in 1995–6. The same pattern was subsequently adopted by the NAAASP. The analysis of growth and rupture rates in this RESCAN project provides an evidence base and opportunity to investigate whether or not a different surveillance strategy might be better than that used in MASS and subsequently by NAAASP.

Other economic models of AAA screening have of course been published. A systematic review of the models to 2006 emphasised the variability in their estimates of cost-effectiveness. 57 Since that review, a number of further modelling studies have been published drawing on data from a variety of sources: for example studies relating to Italy,15 to the Netherlands and Norway,58 to Canada59,60 and to Denmark. 61 These have all concluded that screening is acceptably cost-effective, with the exception of a study using data from Denmark and other sources, which concluded that screening did not seem to be cost-effective. 62 A recent modelling study, again using Danish data, contradicts that previous conclusion, suggesting that screening men at age 65 years is highly cost-effective compared with no screening and, additionally, that rescreening after 5 years may be a cost-effective extension to the programme. 63 However, no models of cost-effectiveness of AAA screening have been published that have specific relevance to NAAASP, other than those already cited relating to and derived from the MASS study, which is the largest randomised trial of AAA screening and contributes most to the international evidence. 7

The focus of the cost-effectiveness element of the current study is to analyse the implications of the RESCAN analyses, relating to the international evidence on growth rates and rupture rates, for the most cost-effective surveillance strategy following AAA screening. Cost-effectiveness is of course a function not only of these clinical parameters but also the behavioural, resource-use, and cost data relevant to a particular screening programme. Hence, we have chosen to use such country-specific data relevant to the current UK programme.

The work specifically undertaken on the model for this study consisted of the following elements:

1. Validation and recalibration of the previously published model8 to best reflect the accumulated 10-year follow-up data from MASS, including incorporation of time-dependent parameter values.

2. Adaption of the model structure to fully incorporate unobserved ‘tunnel’ states to reflect the growth and rupture probabilities for all aneurysms, whether or not reobserved and remeasured at a recall scan.

3. Re-estimation of current unit costs for screening interventions and elective and emergency surgery.

4. Incorporation of data from NAAASP, including attendance rates, prevalence of identified aneurysms, and distribution of aneurysm sizes at initial screening.

5. Incorporation of size-specific growth and rupture rates from the analysis of the IPD surveillance data sets reported in Chapters 5 and 6.

Each of these aspects of our development of the model is to enable us to address the issue of the cost-effectiveness of alternative surveillance strategies, in the context of the NAAASP, and each aspect is addressed in turn in this chapter.

The initial model

The starting point for the economic analysis in this project was the economic model that had previously been developed by members of the research team, to estimate the long-term cost-effectiveness of a screening programme using the MASS trial screening and surveillance protocol. Details of this model have previously been published. 8 In summary, it used a Markov model, with 3-monthly cycles, to compare the introduction of a formal screening programme of an invitation for a one-off US scan for men aged 65 years, and the MASS surveillance strategy described above if an aneurysm is identified, with the policy of no systematic screening. The original model is represented in Figure 18.

The model used a 30-year time horizon (so approximating lifetime results). The methods were consistent with those recommended for the National Institute for Health and Care Excellence (NICE)'s Technology Appraisal Programme. 64 The cost perspective was that of the NHS and the model estimated life-years gained and adjusted these for age-specific utility values relating to the UK population to provide ICERs in terms of QALYs.

For this model, parameter values were estimated from patient-level data from the 4-year analysis of the MASS study and the model structure was internally validated by comparing its results against observed event and cost-effectiveness data over the 4-year time period. In addition, it was externally validated against key outcomes from the Cochrane review of AAA screening and surveillance. 7

Validation and recalibration of the previously published model to better reflect the 10-year follow-up data from the Multi-centre Aneurysm Screening Study

The first step in the current study was to revalidate the model comparing the 10-year model outcomes against the 10-year observed data from the follow-up of MASS, and, where necessary, to adjust or ‘recalibrate’ the model parameters to better reflect that observed data. We used the 8.9–11.2 years of observed follow-up (mean 10.1 years) to better estimate the values for the parameterisation of the economic model. Outputs of the model were compared against those observed in the trial, by considering the numbers of key observed events and cost-effectiveness results at the end of this follow-up period.

Given the staggered recruitment of participants into the trial and the range of follow-up periods for individual patients, for the purpose of comparing numbers of key events the model replicated censoring patterns by removing equal proportions as observed in the trial of numbers in each state for every 3-month cycle, so that the total person-years of follow-up were comparable. The model was then run over 11.25 years to provide comparability. To produce estimates of costs and life-years at 10 years comparable with those observed in MASS, the model was run for 10 years with costs based on 2008–9 prices and survival based on all-cause mortality. Given the characteristics of the MASS trial population, it might be expected that mortality rates would not be comparable with the national statistics, and so for the internal validation exercise MASS-specific mortality data were used to estimate the probability of mortality in each 3-month cycle, rather than using national mortality statistics that were incorporated into the model to give it greater external validity or generalisability.

Initially the model did not appear to perform particularly well against observed data in terms of key events or cost-effectiveness results over the longer follow-up period. As the length of follow-up had increased, trends may have emerged in key parameter values and differences between modelling and observed outputs could be partly attributed to the use of time-constant parameters. Time-dependent transition probabilities were therefore estimated using ‘logistic’ and ‘Poisson’ regressions, to improve the fit of the model. The differences between the observed and modelled data may also have been in part due to parameters used to inform the modelling that could not be observed in MASS. In particular, two parameters, the probability of opportunistic detection among those not in active screening and the rupture rate of undetected large aneurysms, were seen as potentially unreliable. The first had been based on calculations utilising data from the control arm of the trial, to give a crude estimate of the detection rate. The latter had been estimated to fit the 4-year data, assuming it lay between the rupture rate in detected large aneurysms and the rupture rate among aneurysms contraindicated for elective surgery. A recalibration exercise was conducted using a range of figures for both these parameters, to obtain model results that gave more similar numbers of key events (i.e. elective operations, emergency operations and AAA deaths) to those observed in MASS.

It was not easy to achieve complete consistency between the observed and modelled clinical outcomes. Given that the long-term ICER was the primary outcome for the cost-effectiveness modelling, the recalibration focused on eliminating disparities in the modelled and observed differences between the arms.

Table 13 shows key event rates as observed in MASS, as estimated by the original model and as estimated by the model following recalibration. The pattern of events over time from the observed data, the initial model estimates and the recalibrated estimates are shown in Figure 19a–h.

This recalibration process achieved similarity between the modelled and observed differences in key events between the arms, as well as in the resultant ICERs (Table 14). The ICER was £7600 per life-year based on observed 10-year data: the original model estimated an ICER at 10 years of £18,000, the recalibrated model estimated the ICER at £8900, closer to that from the observed data. This suggested that, though imperfect, the model would be suitable for extrapolation of cost-effectiveness results over the long term.

Adapting the model structure to fully incorporate unobserved ‘tunnel’ states

The previously constructed Markov model8 needed adaption to account more explicitly for the incidence of rescans. The original model, with 3-monthly cycles, was built with a view to considering one surveillance policy (that pertaining to the MASS data of yearly recall for AAAs measuring 3.0–4.4 cm and 3-monthly recall for medium AAAs measuring 4.5–5.4 cm). The model averaged out the surveillance for the smaller aneurysms, effectively assuming that surveillance scans were being conducted every cycle but assigning only one-quarter of the rescan costs to each cycle. This meant that individuals were able to transition to larger aneurysmal states in each 3-month cycle and on reaching the large state they could be considered for elective surgery or returned to screening. However, in reality, those in a small AAA state (3.0–4.4 cm) are not rescanned every cycle and cannot move into the surveillance pattern for a larger aneurysm until the increased size of their aneurysm has been identified through a scan. In the original model, costs associated with rescanning were averaged across the cycles, 0.25 of a rescan cost per cycle for the small aneurysmal state and assigned as one per cycle for the medium-sized AAA. Although this approximation was adequate when considering the one surveillance strategy, the model needed to be adapted for the purposes of answering questions around different surveillance policies.

This adaption was achieved through the inclusion of ‘tunnel’ states for aneurysm growth, accounting for patients in whom aneurysm growth occurred but was not observed or acted on. The structure of the extended model is represented diagrammatically in Figure 20. These tunnel states allow aneurysms to grow in between scans, where the individual is subject to the relevant rupture rate. An individual can stay in the tunnel state that they have entered in subsequent cycles, or transition to larger AAA tunnel states. Only when a scheduled surveillance scan occurs are individuals then able to be move out of these unobserved tunnel states into observed states. On this they can then be subject to more frequent scanning, or events that occur as a result of entering the large AAA state (i.e. consideration for elective surgery). This provides a more accurate representation of reality and most importantly can allow for the effects of differing policies by adjusting the number of cycles that individuals spend in these states before a rescan and a return to observed states. Aneurysms that are detected opportunistically are assumed to enter an observed state before moving into tunnel states. However, given that this opportunistic detection can occur in every cycle, an approximation in the cycles immediately following opportunistic detection is used to bring these individuals into the same rescanning schedule as the rest of the detected aneurysm cohort. This retains programming efficiency and has the effect that some opportunistically detected patients will be assumed to receive their first 1-year scans after a shorter period than a full year.

It may be useful to consider the current screening policy of 1 year and 3 months for small and medium AAAs, respectively, as an example of the way the new structure works. An individual invited to and having attended screening, with a small aneurysm observed at baseline, would be invited to return for another scan in 1 year, equivalent to four cycles in the model. Assuming the AAA had not grown, nor ruptured and the individual had not dropped out of surveillance or died, they would move into the small (unobserved) tunnel state in cycle 2. If no transition had occurred in cycles 3 and 4, the individual would then move out of the small unobserved ‘tunnel’ state back into the small observed state where a rescan takes place. They would then move into tunnel states again following this scan. Alternatively, the aneurysm could have grown and the patient transitions into the tunnel state for medium aneurysms in cycle 2 and possibly into the tunnel state for large aneurysms in cycle 3. The tunnel states mean that this individual will not be treated as having a large aneurysm as regards the surveillance policy until this growth has been observed at the next scan in cycle 4.

These techniques for modelling growth also mean that the accrual of costs due to rescanning should be more accurate. Costs will accrue as rescans occur over time, as opposed to averaging costs over a number of cycles. By implementing these changes the model can be used to investigate a range of potential new surveillance policies and compare their cost-effectiveness.

Re-estimation of current unit costs for screening interventions, and elective and emergency surgery

The various estimates of cost-effectiveness based on MASS, referred to earlier, have so far all used original unit cost estimates at 2000–1 prices, updated as necessary only for general health service inflation. There was a clear need for a more thorough update of costs to reflect a range of changes in practice. All costs were re-estimated at 2010–11 price levels.

Incorporating data from the NHS Abdominal Aortic Aneurysm Screening Programme

To reflect current AAA screening in the UK, data from the NAAASP were utilised in the economic modelling. The attendance rate from those invited to screening in the NAAASP has been lower than that of the MASS trial (which was 83%); a figure of 73% from the NAAASP has been used in the modelling. 71 A lower prevalence of AAAs has also been noted in data from the NAAASP, which might have some effect on cost-effectiveness results. In the MASS trial the prevalence was 4.9%, but the current prevalence observed in the NAAASP is 1.6% and this rate was used in the modelling. Data were provided to us from the NAAASP on the distribution of aneurysm sizes detected in the screening programme, to reflect possible changes in aneurysm sizes observed at baseline compared with the MASS data used in the original model.

In addition, non-AAA mortality rates were estimated using data from Hospital Episodes Statistics72 and the Office for National Statistics (ONS) for 2010,73 incorporated into the modelling as age-specific 3-month probabilities.

Incorporating size-specific growth and rupture rates from the RESCAN analyses

Analysis of cost-effectiveness using the adapted and updated model

These developments enable us to use the model to estimate the life-years, QALYs, costs and net monetary benefits for a series of strategies with different recall frequencies, compared initially with the base case of the existing strategy, and by looking at differences in net benefit to extend to comparisons between any two strategies.

In this particular context, a probabilistic sensitivity analysis (PSA), as might typically be provided as a representation of the overall uncertainty in the estimates of cost-effectiveness, would not adequately characterise that overall uncertainty. In particular, it is not readily feasible to estimate the correlated uncertainty around the very many (480) age- and size-related aneurysm growth and rupture rates derived from the IPD meta-analysis incorporated into the model, or to provide evidence-based estimates of uncertainty distributions around other parameters, or to characterise the underlying structural uncertainty in this complex model. Therefore, rather than provide a potentially misleading PSA, we have chosen to employ simple one-way sensitivity analyses to characterise the effects on net benefit of uncertainty around growth rate and rupture rate estimates, and to check whether or not the conclusions regarding the preferred strategy are sensitive to this uncertainty. For comparison we illustrate the effects of uncertainty around other important parameters in the model.

FIGURE 18.

Original health economic model structure. Reproduced with permission from figure 1 of Kim LG, Thompson SG, Briggs AH, Buxton MJ, Campbell HE. How cost-effective is screening for abdominal aortic aneurysms? J Med Screen 2007;14:46–52. http://dx.doi.org/10.1258/096914107780154477. 8

FIGURE 19a.

Number of elective operations in control group over 10 years' (mean) follow-up.

FIGURE 19b.

Number of emergency operations in control group over 10 years' (mean) follow-up.

FIGURE 19c.

Number of AAA-related deaths in control group over 10 years' (mean) follow-up.

FIGURE 19d.

Number of non-AAA deaths in control group over 10 years' (mean) follow-up.

FIGURE 19e.

Number of elective operations in invited group over 10 years' (mean) follow-up.

FIGURE 19f.

Number of emergency operations in invited group over 10 years' (mean) follow-up.

FIGURE 19g.

Number of AAA-related deaths in invited group over 10 years' (mean) follow-up.

FIGURE 19h.

Number of non-AAA deaths in control group over 10 years' (mean) follow-up.

FIGURE 20.

Extension of original model structure to include unobserved ‘tunnel’ states.

FIGURE 21.

Distribution of AAA diameter at first screen from NAAASP data and Chichester Screening Programme data.

Alternative surveillance strategies

We used the adapted and updated model described in Chapter 7 to examine the health benefits (in terms of life-years) and overall costs, modelled over a 30-year period, for a range of alternative surveillance strategies as compared with the current strategy which we refer to as strategy A. Table 23 summarises the values and sources of clinical and cost parameters used. The alternative strategies included both lengthening and shortening the current time intervals between rescans for those identified with aneurysms between 3.0 and 4.4 cm (base case 1 year) and 4.5 and 5.4 cm (base case 3 months). We assumed that a recall more frequent than 3 months would be unworkable, so did not consider any such options. In analysing any of the different surveillance strategies we assume that the same strategy would also apply to all opportunistically identified aneurysms in both arms. This assumption explains the very small difference in the control arm life-years and cost between strategies.

Table 24 sets out the key comparison, which is illustrated on the cost-effectiveness plane in Figure 22. Strategy A represents the base case of the current screening strategy, with an ICER against no formal screening of £5572 per life-year gained and £7143 per QALY gained. To identify whether or not an alternative strategy was an improvement, and which was best, we focused on the ICERs (for both life-years and QALYS) and the net monetary benefit calculated as (Net QALYs × £20,000)–Net costs. We used the threshold value of £20,000 per QALY to be consistent with the opportunity cost of interventions within the NHS as articulated by NICE. 75

Each of the alternative strategies demonstrated appropriate directional changes in the life-years gained from screening and the QALYs gained from screening. The strategies, which extend one or more intervals, slightly reduce the QALY gain (by missing a small proportion of aneurysms that would go on to rupture) and slightly reduce the cost (by avoiding additional scans), whereas strategies that decrease intervals increase the QALY gain and increase the costs. This effect is shown in Figure 22, where the strategies that increase intervals (strategies B, C, D and G) fall in the south-west quadrant of the cost-effectiveness plane (where both costs and QALYs are reduced), whereas a strategy that decrease intervals (strategy E) falls in the north-east quadrant (where both QALYs and costs increase). Strategy F, in which all aneurysms between 3.0 and 5.5 cm are recalled at 6-monthly intervals, can be eliminated from further consideration in that it reduces QALYs and increases costs, so is dominated by existing strategy A which is clinically more effective and cheaper.

The economic question for strategy E is whether or not the increase in QALYs is sufficient to justify the increase in costs, while for each of strategies B, C, D and G it is whether or not the reductions in cost are sufficient to compensate for the QALY losses. This is assessed against a view of the acceptable cost per QALY or threshold. Using the threshold of £20,000 per QALY, we can see that strategy E far exceeds the acceptable threshold and is not cost-effective. Given the only other practical strategy, to increase frequency of scans, strategy F, was dominated by the present screening strategy, we can conclude that decreasing recall intervals is unlikely to be cost-effective. Looking at the strategies (B, C, D and G) that increase screening intervals, we are seeking a strategy that releases at least £20,000 in cost savings per QALY lost. All options meet that test, and of these we need to choose the one with the highest net benefit (highest excess in value of cost savings minus value of lost QALYs). Strategy C, with a recall pattern of 2 years for aneurysms between 3.0 and 4.4 cm and 3 months for aneurysms between 4.5 and 5.4 cm, has the highest net benefit, and is the most cost-effective strategy. Increasing intervals beyond this does not provide sufficient additional cost savings to justify the additional QALY losses.

To summarise these figures as clearly as possible, our analysis suggests that the effect of changing from the current surveillance strategy to surveillance strategy C, the best of the options considered, would mean that the loss of value from the QALY gain per man invited would be equivalent to a loss of £1.24, but that the cost would be reduced by £2.57 per man invited, giving a net monetary benefit gain of £1.33.

Uncertainty and sensitivity

As was indicated in Chapter 7, in this particular case it is not feasible, given available data and analytical possibilities, to provide a robust, meaningful PSA that encompasses all the parameter and structural uncertainty. We have therefore focused on a more appropriate series of one-way sensitivity analyses. These are summarised in Table 25.

The first of these applies growth and rupture rates drawn from the three UK population-based screening studies that used internal aortic diameter measurements (as does the NAAASP), namely MASS, Gloucester and Chichester (see Chapter 6), and which could therefore be seen as most applicable to the UK programme. Strategy C clearly remains the preferred option and, although it is slightly less cost-effective than in the base case, the net benefit is still positive and substantial. As an alternative way to address the uncertainty around growth and rupture rates, the next sensitivity analyses reduce or increase growth rates by 10% and rupture rates by 30%. These ranges approximate one SD in the estimates of the value of the relevant parameter. Again, in each case, the preference for strategy C, over all other strategies, remains unchanged. Using an estimate of operative mortality rates from the NAAASP, similarly does not change the choice of strategy. Further sensitivity analyses consider the possibility that dropout rates (from recall) might be affected by different recall strategies but again the preference for strategy C is unaltered. Finally, we consider the effect of (arbitrarily) different relative costs for elective and emergency surgery and for the cost of rescanning. In no case does the choice of strategy change, although, not surprisingly, the magnitude of the cost savings from less frequent recall in strategy C (and hence its net benefit) are somewhat affected by the unit cost of rescans.

Discussion and conclusions

Long-term modelling of the implications of screening programmes is always difficult and typically involves unverifiable assumptions and estimates that have long-term implications. Screening for AAA is an unusual case in that we have been able to compare results from a long-term model initially based on 4-year data with 10 years of observation. The resultant problem was to identify a set of model parameters that would replicate the observed results at 10 years. We were able to do this imperfectly, but nevertheless can have more confidence in the recalibrated model than would otherwise have been appropriate with the original model. Then, with the systematic reviews and the RESCAN analyses of growth and rupture rates, we were able to go further in estimating the cost-effectiveness implications of alternative surveillance strategies.

These show that from a cost-effectiveness perspective lengthening the surveillance interval for aneurysms of 4.5–5.4 cm reduces net monetary benefit and we have argued that decreasing that interval would not be practical. However, increasing the interval for recall of men with aneurysms between 3.0 and 4.4 cm from 1 year to 2 years improves cost-effectiveness, but increasing it further to 3 years worsens cost-effectiveness compared with 2 years and (marginally) compared with the current 1-year interval.

It is important to recognise that the absolute differences in outcomes and costs between the surveillance options we have considered are small, as are the absolute values of the net monetary benefit per man invited. They can be put into perspective by multiplying them up to reflect the impact on a large-scale screening programme. With a programme inviting around 260,000 men per year as will broadly be the situation when the NHS programme covers the whole of England,73 the (present value of the) cost difference for the preferred option would be of the order of £660,000 per year, but the QALY loss would be equivalent to around 16 QALYs. Given the remaining uncertainties there has to be a question of whether or not the differences involved justify a change from the existing surveillance programme particularly as it would have to be explicit that this strategy was expected to be slightly inferior in terms of clinical effectiveness.

FIGURE 22.

Cost-effectiveness plane for new screening interval strategies.

Systematic review of small abdominal aortic aneurysm growth rates

The methods and results of our systematic review are presented in Chapters 3 and 4. This is the first systematic review of small aneurysm growth rates since their inclusion in the earlier surgical management of aneurysms review of Hallin et al. ,76 which covered the literature from 1990 to 1997. Unlike this earlier study, which included only 816 patients from 13 different studies, we did not have to relax the eligibility criteria for studies (50 or more patients) and this allowed the identification of 15 suitable studies with 7630 persons for inclusion in a meta-analysis (see Table 2).

There are many reasons which might explain the large heterogeneity in growth rates between studies, particularly methodological and patient population issues. The methodological issues include the fact that patients have been entered into these studies since 1976, a 35-year time window during which the quality of aortic imaging has improved dramatically. Second, some recent studies have used imaging modalities other than US (such as CT) and it is known that there is poor agreement between CT and US aortic diameter measurements. 77,78 Some studies have used both methods of imaging. 22,29,32 However, a sensitivity analysis of studies that used only US imaging after 1990 was not associated with a reduction in heterogeneity. Third, aortic diameter has been measured in a variety of different planes, measurement cursors have been positioned on both the external and internal walls of the aorta (see Table 2) and only a minority of studies specified the standardisation of imaging protocol between observers (see Table 3). Fourth, growth rates have been estimated in a variety of different ways using (last – first diameter)/(time elapsed), linear regression or complex or multilevel modelling (see Table 2), and it has been reported that linear regression can overestimate aneurysm growth by up to 40% compared with multilevel modelling. 13 Hierarchical modelling also resolves the problem of a small number of measurements possibly leading to exaggerated growth rates. 13 There was evidence that some of the heterogeneity is explained by the use of linear regression and, in particular, (last − first diameter)/(time elapsed) to report aneurysm growth rates.

The patient population issues include ethnicity (which was reported rarely), the intervention policy of each study (which varied from aneurysm repair at 4.5 cm in the Huntingdon study to 6 cm in the Chichester study34), the variable proportion of women in each study (which ranged from 0% in some studies to 24% in one Canadian study29) and whether the aneurysm was identified by population screening or in hospital. During the period of patient recruitment to these studies, 1976–2005, there have been important improvements in medical therapy, with some studies reporting that use of statins slows the growth of aneurysms while ACE inhibitors may increase the growth of small aneurysms. 31,79 Medical therapies, smoking and comorbidities were reported inconsistently across the 15 studies included in our review and frequently only at study entry; therefore, their influence on growth rates could not be evaluated (see Table 2).

Although this meta-analysis, based on the published literature, provides very strong evidence that growth rates increase markedly with aneurysm size, it only provides partial information about appropriate surveillance intervals for small aneurysms. The growth rate for a 3.5-cm aneurysm is estimated at 1.9 mm/year, whereas that for a 4.5-cm aneurysm is 3.5 mm/year, both rates being somewhat lower than those presented in the earlier review. 76 Given an exponentially increasing aneurysm diameter, as implied from the meta-regression in Figure 4, it would take on average 6.2 years for a 3.5-cm aneurysm to grow to 5.5 cm, whereas a 4.5-cm aneurysm would take only 2.3 years. So surveillance intervals for a 4.5-cm aneurysm should be of the order of one-third of those for a 3.5-cm aneurysm. However, the meta-regression does not provide information about the variability of growth rates between individuals necessary to inform the choice of appropriate surveillance intervals.

Systematic review of small abdominal aortic aneurysm rupture rates

The methods and results of our systematic review were presented in Chapters 3 and 4. Rupture of an AAA is defined as blood leaking through the aortic wall into the peritoneal or retroperitoneal spaces; the diagnosis of rupture therefore depends on imaging (usually CT), findings at laparotomy or autopsy. A particular constellation of clinical symptoms including circulatory collapse, in the presence of a known aneurysm, also will give rise to the reporting of death from aneurysm rupture without autopsy. Contained or intramural rupture also may occur, when either the leak has sealed spontaneously or the bleeding remained intramural (the intramural blood from these ruptures usually can be detected by CT imaging). In contrast, abdominal or back pain and aortic tenderness in the presence of a known aneurysm does not necessarily imply rupture, although such aneurysms may be repaired urgently and listed as an emergency hospital admission. Very surprisingly, none of the eligible studies for this review reported their diagnostic criteria for aneurysm rupture and only 322,23,41 of 14 studies reported the evidence from which a diagnosis of rupture was made. This lack of clarity about the diagnosis of rupture is a major limitation of these studies. In the European Society for Vascular Surgery Guidelines for the management of AAAs,80 a clear definition of ruptured aneurysm is provided and hopefully this will be used in future studies: ‘AAA rupture is defined as bleeding outside the adventitia of a dilated aortic wall. Rupture is further classified into free rupture in the peritoneal cavity and retroperitoneal rupture where the retroperitoneal tissues provide tamponade and reduce temporarily the volume of blood loss’.

Although many studies report the number of aneurysm ruptures, few studies have length of follow-up information to permit the estimation of rupture rates. Nevertheless, there are many more studies, with almost 10,000 patients, describing rupture rates of small aneurysms than were available for a recent systematic review of rupture rates in large aneurysms. 20 The latter, including only 533 patients in total, reported a rupture rate of 10.3 (95% CI 7.5 to 14.3) per 100 person-years for aneurysms 5.0–5.9 cm in diameter in those considered unfit for aneurysm repair. 20 The rupture rates reported here, for aneurysms of 3.0–5.5 cm in diameter, are much lower at 0–1.61 per 100 person-years, with a trend for rupture rates to increase with increasing diameter.

As no study reported the diagnostic criteria for rupture, the evidence from post-mortem studies was examined. There are limited data from post-mortem studies and because the post-mortem diameter is lower than any in vivo measurement, post-mortem data may be unreliable. A prospective autopsy series of 78 aneurysms where the autopsy aorta was subject to pressure inflation at 80–100 mmHg did not show any ruptures in specimens of ≤ 5.0 cm. 81 This same study reported that without pressure inflation post-mortem diameters were usually smaller than any in vivo measurement, probably by 0.4–0.5 cm or in some cases by almost 50%. Therefore, data from the only large-scale autopsy series, which reported retrospectively on the rupture of aortic aneurysms in almost 24,000 consecutive autopsies at the Massachusetts General Hospital in the period 1952–75, are unreliable. 82

Both the PIVOTAL and CAESAR trials of endovascular repair compared with surveillance for small aneurysms reported in 2010 after the search for this review closed. 51,83 In the PIVOTAL trial51 the rupture rate in the surveillance was 0.17 per 100 patient-years (although this includes some follow-up after aneurysm repair, which occurred in 31%). In the CAESAR trial,83 although mean conditional follow-up was not reported, the rupture rate appears to be < 1 per 100 person-years. These are in keeping with the results reported here.

A new large prospective study based on population screening would be the most robust method of assessing the rupture rates of small aneurysms in the twenty-first century. Such a study would take time to set up and several years of follow-up before it could report. For now, an alternative way to assess the rupture rates of small aneurysms is through synthesis of a large amount of IPD from relevant studies. This is the strategy we pursued.

Analysis of individual patient data surveillance data sets

We presented our analysis of overall growth and rupture rates in men and women from 18 surveillance data sets in Chapters 5 and 6. Key results are shown in Table 8. This research has definitively quantified the extent to which AAA growth rates and risk of rupture increase with AAA diameter, to provide a scientific basis for selecting surveillance intervals for patients with small AAAs. For example, we have shown that the time taken for an AAA of ≤ 4.0 cm to have a 10% chance of growing to 5.5 cm is at least 3 years. For men with an AAA of ≤ 4.0 cm, it takes over 3.5 years to have a risk of rupture > 1%.

These results are based on the largest collection of IPD made to date, and consistent methods of statistical analysis across multiple studies. Although three additional studies84–86 have been published up to September 2012, since the literature search that identified the studies for this meta-analysis was completed, one84 had fewer than 100 patients, and another85 appears to have considerable overlap with data already included in our analysis. The third study86 includes 453 patients with 12 months of follow-up and would be unlikely to alter our findings which are based on 15,475 patients with longer follow-up.

Our findings have important implications for AAA surveillance programmes. As screening for AAA is becoming established in many countries there will soon be large cohorts of individuals under surveillance for AAAs. For example, the NAAASP provides yearly scans for 3.0–4.4-cm diameter AAAs and 3-monthly scans for 4.5–5.4-cm AAAs. 9 The 95% confidence estimates for rupture risk from our study suggest that these surveillance intervals could safely be extended to 3 years for AAAs < 4.5 cm and yearly for 4.5–5.4-cm AAAs. The risk of rupture would be maintained at < 1%, a figure that appears to be acceptable to patients with AAAs. This would reduce the average number of surveillance scans from approximately 22 to only 7 for a patient with a 3.0-cm AAA detected by screening. Even if the lower 95% prediction limits of our estimates are applied, to acknowledge that the populations in each study may have different growth and rupture rates, these surveillance intervals could be reduced to 2-yearly for 3.0–3.9-cm AAAs, yearly for 4.0–4.9-cm AAAs and 6-monthly for 5.0–5.4-cm AAAs. Using the same example as above, this would reduce the average number of surveillance scans required for a 3.0-cm AAA from approximately 22 to 10.

Most population-based screening programmes for AAA are focused on the elderly male population, aged ≥ 65 years. This study, which included 1743 women (11%), showed that overall there was no difference in aneurysm growth rates between men and women. Interestingly, when aneurysm growth rate was estimated using hierarchical modelling, allowing for accelerating growth rate with increasing aneurysm diameter, data from Tromsø no longer showed that growth rates were faster in women. 38 Only Edinburgh54 and Leeds (Professor D Julian A Scott, personal communication) showed a significantly different growth rate in women. This underscores the heterogeneity observed both in this meta-analysis of the influence of sex and many other covariates and in previously published meta-analyses based on varied methods for estimating growth rate. 19,87

The fourfold higher rupture rate found for women compared with men is intriguing, particularly given that these data demonstrate that both sexes have similar growth rates. The sex imbalance in the risk of rupture has been identified in previous studies;48 differences in anatomy, structure, sex steroids and smoking habits have all been suggested to play a role. 88 The clinical implication is that a lower AAA diameter threshold for surgery should be adopted for women, a recommendation already made by the joint council of the American Association for Vascular Surgery and Society for Vascular Surgery,89 but one not yet supported by randomised trial evidence. As the rupture rate for women with a 4.5-cm AAA is about the same as that for a man with a 5.5-cm AAA, a threshold for surgery of 4.5 cm might seem appropriate in women. However, the operative mortality and discharge outcomes appear to be worse in women than men. For instance, the US Medicare database shows that operative mortality after endovascular repair (now the majority procedure) is 3.2% in women compared with 1.8% in men, and similar findings have been reported in other studies. 90–92 Therefore, competing risks must be carefully considered before amending guidelines or recommendations for intervention in women.

Despite the strong conclusions that can be drawn from this study, there remain some areas of uncertainty where future research efforts could be focused. The most obvious need is for further evidence for women with aneurysms in the diameter range 4.5–5.4 cm. Second, the substantial heterogeneity between studies we identified in growth and rupture rates could not be explained by adjustment for individual factors or study-level characteristics. If the underlying causes for these differences are biological rather than simply methodological, these may yield insights for developing interventions to lower growth rates in patients with small AAA; such treatments have been sadly lacking up to the present time. 18,19 Such advances will only be possible through the analysis of large data sets, most likely to become available from surveillance within the national AAA screening programmes that are currently being set up. Third, recommended surveillance intervals may be lengthened in the future given the generally falling national rates of AAA rupture currently being experienced. 48,93 Last, more research is needed on the acceptable risk to patients of exceeding 5.5 cm or rupture before the next scan. An alternative approach is to consider the cost-effectiveness of different surveillance policies. Decreasing surveillance frequency would reduce surveillance costs. However, it may also very slightly increase rupture rates, thereby decreasing overall life expectancy in AAA patients under surveillance and increasing the burden and cost of emergency surgery. Hence, we conducted a formal evaluation to identify which surveillance intervals for different sized AAAs would be most cost-effective, as reported in Chapters 7 and 8.

Factors influencing growth and rupture rates

Factors apart from sex that influence growth and rupture rates were systematically examined in the analysis of the IPD surveillance data sets (see Chapters 5 and 6). Aneurysm growth rates were consistently increased by about one-sixth in current smokers and decreased by about one-quarter in patients with diabetes. There was no evidence of any change in aneurysm growth rates over time and neither age, mean arterial pressure nor cardioprotective drugs, had a significant effect on growth rates. In contrast, small aneurysm rupture rates were doubled in current smokers, with age, mean arterial pressure and calendar year of enrolment being additional factors influencing rupture rates. These data suggest that the follow-up and management of the patient with small AAA could potentially be tailored to factors such as current smoking, diabetes and blood pressure, in addition to aneurysm diameter.

In every study, the growth rate of aneurysms in patients with diabetes was on average lower than in patients without diabetes, with no heterogeneity observed in the meta-analysis. Furthermore, the magnitude of the change in growth rate (approximately 25%) associated with diabetes was far greater than for any other covariate tested. This suggests that routine surveillance of small aneurysms might safely be less frequent in patients with diabetes. The biology of the aortic wall may be altered in diabetes with increased glycosylation, which together with other changes might underlie the reduced aneurysm growth rate in patients with diabetes. Most of these studies did not discriminate between type I and type II diabetes, although unpublished data from the United Kingdom Small Aneurysm Trial (UKSAT) suggest that most of the patients with diabetes (78%) were not taking insulin and in the Leeds study only type II diabetes was reported (Professor D Julian A Scott, personal communication). This might explain the trend for increased BMI to be associated with a reduced aneurysm growth rate in the unadjusted analysis. However, overall the patients with diabetes were few, concordant with the observation that diabetes is negatively associated with AAA. 94 As there were only four ruptures reported in patients with diabetes, it was not possible to reliably evaluate the influence of diabetes on aneurysm rupture.

Smoking was second only to diabetes in the magnitude of its effect on aneurysm growth but while the effect of current smoking on aneurysm growth was modest (∼ 16%), current smoking doubled aneurysm rupture rates. The direction of the effect of smoking on aneurysm growth was consistent among all studies, except Leeds (Professor D Julian A Scott, personal communication). A limitation of our study, impacting on heterogeneity, is that non-smoking or ex-smoking status might have been defined differently in the different studies, particularly with respect to time from smoking cessation or very light tobacco usage.

The studies included in this meta-analysis span patients studied over a > 25-year period, since 1983. The prevalence of smoking has been decreasing since the mid-1970s and at the same time awareness of cardiovascular risk prevention in patients with small aneurysms has been increasing and the quality of US has improved dramatically. For these reasons, we explored whether or not the recruitment date of patients within each study (as a proxy for changing environmental, ultrasonographic and drug factors) might influence either aneurysm growth or rupture rates. Interestingly, although aneurysm growth rates appeared to be unchanged over time, rupture rates appeared to decrease. This is consistent with better control of hypertension (higher mean arterial pressure increases rupture rates) and lower prevalence of smoking as well as with the decreasing incidence of aneurysm rupture, which is being reported from several different countries. 95–97 The contrasting effects of enrolment date on growth and rupture rates also may indicate that although cardioprotective drugs might influence rupture rates they are unlikely to have any strong influence on aneurysm growth.

Individually, none of the cardioprotective drugs (lipid-lowering drugs, aspirin, beta blockers, calcium channel blockers or ACE inhibitors) showed significant associations with aneurysm growth, and the magnitude of any individual effect was very small (about 0.2 mm/year for statins and 0.1 mm/year for other drugs). One of the included studies was a formal evaluation of the effect of propranolol on aneurysm growth rate in about 500 patients. 40 Based on the findings reported here, this trial was very underpowered and trials of approximately 3700 patients would be necessary to show a reduction in aneurysm growth rates of 0.2 mm/year at 80% power, assuming growth rates have a SD 2.2 mm/year. The findings reported here for lipid-lowering drugs are consistent with other meta-analyses showing a possible association between statin usage and slower aneurysm growth rate. 18,19,98 The finding that ACE inhibitors increased growth rates in the UKSAT79 was not substantiated by other studies. Particularly in the older studies, few of the patients included in the current meta-analysis were taking cardioprotective medications and even if patients were taking these drugs at baseline, there is no evidence about continued patient compliance with medication. This is pertinent since both of the randomised trials assessing the effect of beta blockers on aneurysm growth illustrated the poor compliance with these drugs in patients with small aneurysms. 40,99

This individual patient meta-analysis of the influence of demographic factors and drugs on small aneurysm growth and rupture rates overcomes a number of limitations found in meta-analyses of published data. 18,19,87,100 First, data are strictly limited to the aneurysm diameter size range of 3.0–5.5 cm, the range which is relevant to most population screening programmes. Second, the study uses a single statistical method of estimating aneurysm growth rates, to overcome some of the heterogeneity associated with using different methods of growth rate estimation, particularly the use of (last – first diameter)/(time elapsed). Third, through application of an adjustment for overestimation of diameters using CT scans compared with ultrasonography, studies using both CT and US data could be included. Fourth, as rupture is a rare event in small aneurysms, there is an opportunity to study important factors, other than aortic diameter, which may influence rupture rates and alter patient management.

Clinical and policy conclusions

In our systematic reviews of the growth and rupture rates of small AAA we primarily identified that both of these outcomes were related to AAA size, with larger AAA having higher rates of growth and higher rates of rupture. We estimated that the average time for a 3-cm AAA to grow to 5.5 cm is 9.6 years and for a 5-cm AAA is 1.1 years. We were unable to provide the same size-based estimates for risk of rupture from our systematic review since the data informing such an analysis were insufficient. We can only conclude that the rupture rate of small AAAs (3.0–5.5 cm in diameter) appears to lie between 0 and 1.61 per 100 person-years.

In order to address the deficiencies of the systematic reviews we went on to obtain IPD for 15,475 patients in AAA surveillance programmes. We harmonised and combined the data from these individual patients to obtain estimates of AAA growth and rupture rates. As expected, we identified that AAA growth and rupture rates were related to AAA size. In this analysis we determined outcomes for men and women separately. For men with a 3.0-cm AAA, the time taken to have a 10% chance of exceeding 5.5 cm was 7.4 years, and for a 5.0-cm AAA the time was 0.7 years. For women these times were 6.9 years and 0.7 years. We also assessed the time for small AAAs to exceed a 1% per annum rupture risk. In men with 3.0-cm and 5.0-cm AAAs these times were 9.3 years and 1.4 years, respectively. The times for women were 3.5 years and 0.7 years. These data highlight the sex difference in AAA rupture rates, with women more likely to rupture at lower AAA sizes than men, suggesting that the threshold for surgical intervention in women should be lower than in men.

From our analysis of IPD we identified that smoking was independently associated with both accelerated AAA growth and the risk of AAA rupture. Although all patients with AAA are routinely offered smoking cessation therapy in the majority of clinical practices, this finding underlines the importance of instituting measures to reduce tobacco consumption among patients with AAA and provides the most robust evidence to date with which to counsel patients. This information is also of importance to the non-specialist NHS practitioner, as the majority of patients with small AAAs will be managed in surveillance programmes with little or no physician contact.

Although we identified that higher mean arterial pressure was associated with a higher risk of AAA rupture, we were unable to demonstrate an effect of antihypertensive drugs. The appropriate use of antihypertensive medications, with targets of 130/80 mmHg, is nevertheless important in AAA patients. Although this strategy will require further investigation to determine any effect on rupture rates, the expected benefits in terms of overall cardiovascular mortality reduction justify its recommendation.

Our study has implications for surveillance strategies. From our IPD meta-analysis, we have demonstrated that most of the smallest AAAs will remain quiescent over many years. This suggests that current surveillance strategies for small AAAs could be refined to reduce the number and frequency of surveillance scans required and therefore the costs associated with AAA surveillance.

In theory the individual characteristics that influence growth or rupture rates, such as smoking and diabetes, could be taken into account to tailor surveillance intervals for each individual. However, the effects of such variables are not substantial enough to justify the organisational difficulty that such personalised surveillance strategies would involve, despite the desire for this from patients. Similarly, AAA measurements before the current one are informative about a person's rate of AAA growth17 and could, in principle, also be taken into account.

Our cost-effectiveness investigations showed that lengthening the surveillance intervals from 1 year for 3.0–4.4-cm AAAs was cost-effective, whereas lengthening the 3-month interval for 4.5–5.4-cm AAAs was neutral with respect to cost-effectiveness. However, the differences in costs were small, up to about £4 within an overall cost of £300 per man invited to AAA screening. Also our results may be susceptible to the detailed modelling assumptions made and parameter estimates used, and the fact that our updated health economic model did not validate very well on the 10-year MASS trial data. Nevertheless, the cost-effectiveness analyses support the lengthening of surveillance intervals for the smallest aneurysms, in line with our clinical conclusions.

Research implications

From the work we have undertaken, we draw a number of conclusions about future research priorities, as follows.

We identified that the definition and diagnostic criteria of AAA rupture were poorly reported in published studies. There is a need for standardised definitions and reporting, so that rupture rates can be more reliably estimated. One major hurdle to overcome is the determination of whether or not AAA rupture is reported accurately in mortality statistics. Many sudden deaths, particularly in elderly populations who are at the highest risk of AAA rupture, are likely to be recorded on death certificates as due to cardiac or cerebrovascular causes, missing many cases of AAA rupture. Declining post-mortem examination rates will contribute to the potential under-reporting of AAA-related mortality.

We demonstrated that surveillance intervals for most small aneurysms can be extended safely, with such extension being cost-effective for the smallest aneurysms. These recommendations are based on opinions from a single patient focus group rather than a more widespread patient preference study. Given the current knowledge, there is a need to investigate patient preferences concerning extending surveillance intervals, especially in the upper half of the small aneurysm diameter range where such extension could be clinically acceptable but without cost-effectiveness benefit.

We have determined that hierarchical modelling provides a methodologically appropriate framework for analysing data on AAA growth rates and recommend that this method be used in future studies.

Based on the observation of growth rates in patients with 3.0-cm diameter aneurysms, modelling could be conducted to indicate appropriate recall intervals for men identified as having an aortic diameter of 2.5–2.9 cm at initial aneurysm screening visit.

We observed large variation in growth rates between studies, which was largely unexplained. The identification of both methodological reasons, for example how AAA diameters are measured, and biological causes for this variation are important. The latter could lead to the identification of factors and biological pathways related to increased growth rates or rupture rates that were amenable to intervention. Such interventions might then be assessed in randomised trials.

We have shown that the clinical progression of an AAA is different in women compared with men, with women at substantially higher risk of rupture. Currently women are not screened for AAAs in the UK, and those with opportunistically detected AAAs are usually entered into generic rather than sex-specific surveillance programmes. In addition to obtaining evidence to investigate the efficacy and cost-effectiveness of AAA screening in women, surveillance programmes and criteria for considering surgery need to be tailored for women. Formal evaluation of a lower intervention threshold for women, either via a randomised trial or modelling studies, is a research priority.

We have shown that increasing mean arterial blood pressure increases the risk of aneurysm rupture. There is a need to formally evaluate whether or not reducing blood pressure to ≤ 130/80 mmHg would reduce the risk of rupture, in both small aneurysms managed in surveillance programmes and in high-risk patients with larger aneurysms who, as part of the National Quality Improvement Programme, are being turned down for elective aneurysm repair in increasing numbers.

The findings that age, sex and blood pressure appear to have strong associations with aneurysm rupture but not aneurysm growth are intriguing and hint at the presence of differential mechanisms promoting aneurysm growth and aneurysm rupture. The association of diabetes with retarded AAA growth rates reiterates the intriguing negative association between diabetes and AAA prevalence and further exploration of the mechanistic pathways leading to diabetes may provide insights into the pathogenesis of AAA.

Although we have provided evidence about appropriate surveillance intervals for small AAAs, we have always used the currently accepted diameter of 5.5 cm for considering elective surgery. Related research could aim to determine whether it is possible to develop personalised criteria or patient-specific thresholds for elective surgery.

Comparable studies with substantial patient numbers are required for the kind of research undertaken here. Future research should focus on the collection and use of large, well-structured and internationally compatible data sets, along with statistical and epidemiological methods needed for their harmonisation and analysis. For example, work is required to determine how to combine data based on different methods of AAA size measurement. The recent inception of national AAA screening programmes in several countries may facilitate such research.

Contributions of authors

SG Thompson was principal investigator of the project and the RESCAN Collaboration, provided overall direction and overview, supervised the statistical analyses, and collated and revised material for this report.

LC Brown undertook data extraction and evaluation of published studies in the systematic reviews, collated and harmonised the IPD surveillance data sets acquired from the RESCAN collaborators, and edited this report.

MJ Sweeting undertook statistical analyses for the systematic reviews, developed the statistical modelling framework for growth and rupture in the IPD data sets to inform surveillance intervals, and provided information from these analyses in a form suitable for the health economic analyses.

MJ Bown organised the patient focus group, collaborated on the analyses of the IPD surveillance data sets, and provided clinical input into the RESCAN Collaboration.

LG Kim undertook reanalyses of the MASS trial data and provided input into developing the health economic modelling framework.

MJ Glover developed the health economic modelling framework and undertook the cost-effectiveness analyses.

MJ Buxton supervised the health economic modelling work and drafted the conclusions from the cost-effectiveness analyses.

JT Powell directed the systematic reviews, invited investigators to participate in the RESCAN Collaboration, provided epidemiological and biological input for the analyses of growth and rupture rates, and led the drafting of a number of the resulting published papers.

All authors provided comments on a draft version of this report.

Other contributions

Systematic reviews, acquisition and management of data sets: Dr Susan Gotensparre (Imperial College London, London, UK).

Clinical input: Professor Roger Greenhalgh (Imperial College, London); and Professor Ross Naylor (University Hospitals of Leicester NHS Trust, Leicester, UK).

Imaging input: Mr Tim Hartshorne (University Hospitals of Leicester NHS Trust Leicester, Leicester, UK).

Epidemiology and review/approval of systematic review protocols: Professor F Gerald R Fowkes (University of Edinburgh, Edinburgh, UK).

Other RESCAN collaborators

Professor PE Norman (University of Western Australia, Crawley, WA, Australia).

Mr S Parvin (Royal Bournemouth Hospital, Bournemouth, UK).

Mrs Hilary Ashton (St Richard's Hospital, Chichester, UK).

Mr RTA Chalmers (Edinburgh Royal Infirmary, Edinburgh, UK).

Mr JJ Earnshaw (Gloucestershire Royal Infirmary, Gloucester, UK).

Mr ABM Wilmink (Heart of England NHS Trust, Birmingham, UK).

Professor DJA Scott (University of Leeds, Leeds, UK).

Professor CN McCollum (University Hospital of South Manchester, Manchester, UK).

Dr S Solberg (Rikshospitalet, Oslo, Norway).

Dr K Ouriel (Syntactx, New York, NY, USA).

Professor A Laupacis (St Michael's Hospital, Toronto, ON, Canada).

Dr M Vega de Céniga (Hospital de Galdakao, Bilbao, Spain).

Mr R Holdsworth (Stirling Royal Infirmary, Stirling, UK).

Dr L Karlsson (Gävle Hospital, Gävle, Sweden).

Professor JS Lindholt (Viborg Hospital, Viborg, Denmark).

Other acknowledgements

The authors would like to thank Dr Karen Welch (Cochrane Collaborative Review Group on Peripheral Vascular Diseases, University of Edinburgh Medical School, Edinburgh, UK) for conducting the search strategy for the systematic reviews. We also thank Dr Angela Talbot (Medical Research Council Biostatistics Unit, Cambridge, UK) for help with submission of this report to the Health Technology Assessment journal.

This project was supported by the UK National Institute for Health Research Health Technology Assessment Programme (project 08/30/02). The views expressed in this report are not necessarily those of the UK NHS.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Publications

1. RESCAN collaborators, Bown MJ, Sweeting MJ, Brown LC, Powell JT, Thompson SG. Surveillance intervals for small abdominal aortic aneurysms: a meta-analysis. JAMA 2013;309:806–13.

2. RESCAN collaborators, Sweeting MJ, Thompson SG, Brown LC, Powell JT. Meta-analysis of individual patient data to examine factors affecting growth and rupture of small abdominal aortic aneurysms. Br J Surg 2012;99:655–65.

3. Sweeting MJ, Thompson SG. Making predictions from complex longitudinal data, with application to planning monitoring intervals in a national screening programme. J Roy Statist Soc Ser A 2012;175:569–86.

4. Sweeting MJ, Thompson SG. Joint modelling of longitudinal and time-to-event data with application to abdominal aortic aneurysm growth and rupture. Biom J 2011;53:750–63.

5. Powell JT, Sweeting MJ, Brown LC, Gotensparre SM, Fowkes FGR, Thompson SG. Systematic review and meta-analysis of growth rates of small abdominal aortic aneurysms. Br J Surg 2011;98:609–18.

6. Powell JT, Gotensparre SM, Sweeting MJ, Brown LC, Fowkes FGR, Thompson SG. Rupture rates of small abdominal aortic aneurysms: a systematic review of the literature. Eur J Vasc Endovasc Surg 2011;41:2–10.

7. Sweeting MJ, Thompson SG, Brown LC, Greenhalgh RM, Powell JT. Use of angiotensin converting enzyme inhibitors is associated with increased growth rate of abdominal aortic aneurysms. J Vasc Surg 2010;52:1–4.

1.0 Protocol

1.1 Introduction

From spring 2009, the UK National aneurysm-screening programme has been slowly launched across England. The programme uses ultrasonography to detect and measure the diameter of abdominal aortic aneurysms (AAAs) to reduce the number of deaths. However, the programme still needs clear policies for their intervention and surveillance protocols. The policies, which are currently under development, will in large depend on the growth rate (GR) and rupture rate of aneurysms. At present, large aneurysms (over 5.5 cm in diameter) are referred to elective surgery (open surgery or endovascular aneurysm repair, EVAR), while optimal management of small aneurysms (less than 5.5 cm in diameter), which represents ca. 90 per cent of all screen-detected AAAs,1 seems to be in favour of surveillance. 2–5

A number of studies have reported on growth rates of small aneurysms and recommended the following frequencies of follow-up examinations (Table 1). The reported growth rates of small aneurysms (3.0 cm to 3.9 cm in aortic diameter) ranged from 0.2 to 0.41 cm per year,6,7 while Santilli et al. reported a lower rate of 0.11 cm per year. 8 In many cases, simple linear regression has been used to estimate growth rates. However, this does not take into account that aneurysmal growth rate may accelerate with an increased diameter. 9 In addition, it has been reported that aneurysm growth is faster in smokers and slower in patients with diabetes, while differences between genders are still unclear. 9–13 There have been several trials, using antibiotics and beta blockers, to reduce AAA growth (Table 2). None of these trials provide persuasive evidence that a treatment to slow aneurysm growth has been identified, but they should provide good quality data for aneurysm growth.

Most of these studies of aneurysm diameter surveillance have used ultrasonography for monitoring. A few studies have used CT scanning3 or a mixture of CT scanning and ultrasonography. In some comparative studies, the measurements recorded by CT scanning were different to those recorded by ultrasonography. 14–16 This observation will have to be accounted for in the systematic review and meta-analyses proposed. Therefore, it will be useful to assess the mean difference between ultrasonographic and CT diameter measurements to inform the analysis of this systematic review.

This systematic review will be made up of three elements, i.e. literature searching, summary of data, and meta-analysis (Figure 1). Studies will also be identified for an individual patient data (IPD) meta-analysis. The published data on GR for small AAA has not yet been synthesized nor has safe surveillance intervals been determined, especially in women. Therefore, further to the systematic review we will aim to re-analyse all the available individual patient data to achieve more precise estimates of growth rates and determine probabilities of growth within different ranges of aortic diameter. This, in turn, will have an impact on the re-screening intervals for patients with small AAAs.

TABLE 1 - Summary of follow-up data of small abdominal aortic aneurysms

Author	Aortic diameter	Follow-up	Additional info
Santilli et al., 20028	3.0 cm to 3.9 cm	After 3 years	Male patients
McCarthy et al., 200317	3.0 cm to 3.4 cm	At 3 years	n/a
Lindholt et al., 20006	3.0 cm to 3.5 cm	At 3 years	n/a
Lindholt et al., 20006	3.5 cm to 3.9 cm	Every other year	n/a
McCarthy et al., 200317	3.5 cm to 3.9 cm	At 1 year	n/a
Lindholt et al., 20006	4.0 cm	Every year	n/a

TABLE 2 - Selection of trials to reduce AAA growth (antibiotics and beta blockers)

Author	Agent	Aortic expansion rate
Mosorin et al., 200118	Doxycycline	Doxycycline may positively change the outcome of patients with small AAA.
Vammen et al., 200119	Roxithromycin	Roxithromycin reduced the expansion rate of AAAs, in comparison to placebo.
Baxter et al., 200220	Doxycycline	Doxycycline is linked with a gradual reduction in plasma MMP-9 levels. Additional studies are needed to determine effects of doxycycline on aneurysm growth.
Karlsson et al., 200921	Azithromycin	Azithromycin had no effect on AAA expansion.
Propranolol Aneurysm Trial Investigators, 200222	Propranolol	The agent did not significantly affect growth rate of small AAAs.
Lindholt et al., 199923	Propranolol	Only 22% of small AAA was treatable with propranolol (40 mg twice a day) for two years, in comparison to placebo. Need large-scale studies.

FIGURE 1.

An outline of the strategy behind the systematic review.

References

1.0 Protocol

1.1 Introduction

From spring 2009, the UK National aneurysm screening programme has been slowly launched across England. The programme uses ultrasonography to detect and measure the diameter of abdominal aortic aneurysms (AAAs) to reduce the number of deaths. However, the programme still needs clear policies for their intervention and surveillance protocols. The policies, which are currently under development, will in large depend on the rupture rate (RR) and growth rate of aneurysms. At present, large aneurysms (over 5.5 cm in diameter) are referred to elective surgery (open surgery or endovascular aneurysm repair, EVAR), while optimal management of small aneurysms (less than 5.5 cm in diameter), which represents ca. 90 per cent of all screen-detected AAAs,1 seems to be in favour of surveillance. Three trials showed that survival in patients with small AAAs is not significantly improved by elective surgery,2–4 while other trials focused on whether early endovascular repair offer a survival benefit to patients with small AAAs. 5,6 Thus, there are a number of published trials that investigated rupture rates and growth rates of both large and small aneurysms. The consensus is that aneurysms, which are 4.0–5.5 cm in diameter, have a RR of 1% per annum or less,2,7 while smaller aneurysms have an even lower RR. 8

This systematic review will be made up of three elements, i.e. literature searching, summary of data, and meta-analysis (Figure 1). Studies will also be identified for an IPD meta-analysis. The published data on RR has not yet been synthesised nor has safe surveillance intervals been determined. Therefore, further to the systematic review we will aim to re-analyse all the available individual patient data to achieve more precise estimates of rupture rates and determine probabilities of rupture within different ranges of aortic diameter. This, in turn, will have an impact on the re-screening intervals for patients with small AAAs.

FIGURE 1.

An outline of the strategy behind the systematic review.

References

Statistical methods

Excluded studies

Calculation of growth transition probabilities for health economic models

We wish to estimate 3-month transition probabilities between 5-mm-wide size states as shown in Figure 23. The state occupancy prevalence for each of the states at any time t following screening given baseline diameter y(0) < 5.5 can be calculated using the fitted mixed-effects models described in Appendix 3.

FIGURE 23.

Five-mm-wide Markov transition model.

Specifically, assume a discrete Markov chain with cycles of 3 months with states 0–6 as shown in Figure 23. This model is a simplified version used in the health economic analyses and does not consider any competing risks, such as surgery or death. Let S(t) represent the state occupied in this model at time t, t = 0,1,2, . . . ,T.

The probability of making a transition from state 0 to state 1 in any 3-month period is assumed to remain constant, and is given by q₀₁ = 0.00207. It is also assumed that the probability of being screened normal (i.e. in state 0 at time t = 0) is π₀ = 0.983, and the probability of having a small aneurysm of size u at screening is denoted by π(u), where π(u) is taken from the Chichester study baseline distribution of AAA diameters. For these calculations, no AAAs are assumed to be large at screening.

Then the (occupancy) probability of remaining in state 0 (< 3.0 cm) after t time periods is given by:

The occupancy probabilities of being in states s (s = 1, . . . , 6) at time t are expressed as the weighted sums of probabilities conditional on screened diameter:

where P_s(t|y(0) < 3.0) and P_s(t|y(0) = u) are the occupancy probabilities of being in state s at time t given the screened diameter is normal or a small aneurysm of size u, respectively.

Given an AAA is detected at screening (≥ 3.0 cm), the occupancy probability of being in state s at time t is calculated using the tail areas of the predictive distribution from each fitted mixed-effects model separately as follows:

where U_s and L_s are the upper and lower cut-points of state s, respectively (e.g. 3.5 cm and 3.0 cm for state 1), and μ_{y|y(0) = u}(t) and v_{y|y(0) = u}(t) are the predicted mean and variance as defined in Appendix 3. Note that U₆ = ∞. Meanwhile, the occupancy probability of being in state s at time t given being screened normal, assuming that all newly developed aneurysms are initially measured at 3.0 cm, is:

where the second line follows if we further assume that once screened normals develop an aneurysm, they follow the same growth trajectory as the screen detected patients when they were screened, that is we assume a time-stationary Markov chain, i.e. P_s(t|y(l) = 30) = P_s(t – l|y(0) = 30).

The occupancy probabilities can now be related to the one-step (3-month) transition probabilities via the Chapman–Kolmogorov equations, which state that:

where q_s,s + 1(t) is the 3-month transition probability at time t from state s to state s + 1, recalling that q₀₁ is independent of time. Rearranging the equation above gives an expression for q_s,s + 1(t) in terms of q_{s – 1,s}(t). Hence, starting with calculating q₁₂(t) in terms of q₀₁, we can recursively estimate all the 3-month transition probabilities for any time t.

Calculation of ‘equivalence’ probabilities for different health economic models

We wish all the Markov models considered in the health economic analyses to be comparable in terms of the probability of developing large AAA, in the absence of any interventions. To achieve this, we recognise that all Markov models being considered will be collapsed versions of the general 5-mm size model.

Consider a collapsed Markov model with S* states and a known mapping from these states to those from the general 5-mm state model. For example, a two-state Markov model with states 3.0–4.4 cm and 4.5–5.4 cm has the mappings {{1} → {1,2,3},{2} → {4,5},{3} → {6}} from this model to the general 5-mm state model. The occupancy probabilities in each size state at time t for the collapsed model can be written in terms of the general 5-mm model and, hence, the new ‘equivalence’ transition probabilities can be derived. For example, in the two-state model the occupancy probability in state 1 at time t is as follows and is set equal to the occupancy probability in the first three states of the general 5-mm state model.

Rearranging and solving the recursive equation we can find values for q_1,2^*(t), noticing that P₀^*(t) = P₀(t), q_0,1^* = q_0,1 and P₁^*(0) = ∑_s = 1³P₁(0). A similar strategy is then used to find values for q_2,3^*(t). This approach can be extended and used for any collapsed model of interest.