Estimating overdiagnosis in screening for abdominal aortic aneurysm: could a change in smoking habits and lowered aortic diameter tip the balance of screening towards harm?BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h825 (Published 04 March 2015) Cite this as: BMJ 2015;350:h825
- 1Department of Public Health and Community Medicine, Institute of Medicine, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden. Research Unit and Section for General Practice, FoUU-centrum Fyrbodal, Vänersborg, Sweden
- 2Research Unit for General Practice and Section of General Practice, Department of Public Health, University of Copenhagen, Copenhagen, Denmark
- Correspondence to: M Johansson
- Accepted 2 December 2014
Clinical context—Abdominal aortic aneurysms (AAAs) are often asymptomatic until they rupture, when the death rate is greater than 80%. If diagnosed before rupture, AAA can be treated with surgery, which has a mortality of 4-5%
Diagnostic change— Sweden, the UK, and the US have initiated screening programmes for AAA. There are also proposals to change the aortic diameter for diagnosis from ≥30 mm to 25 mm
Rationale for change—Early diagnosis by screening allows the opportunity of surgery to prevent ruptures
Leap of faith—Detecting asymptomatic aneurysms will reduce AAA mortality and morbidity
Impact on prevalence—Our estimates indicate that screening almost doubles AAA prevalence, but most AAAs are small and at low risk of rupture. Changing the definition of an AAA from 30 mm to 25 mm would double prevalence again
Evidence of overdiagnosis—We estimate that if 10 000 men are invited to screening, 46 AAA deaths can be prevented over 13-15 years but 176 would have an AAA ≥30 mm detected that remained asymptomatic after 13 years. A recent drop in AAA prevalence reduces the benefits of screening and worsens the benefit:harm ratio
Harms of overdiagnosis—Asymptomatic men are labelled at risk of a life threatening condition for which they will be under lifelong surveillance. Of 10 000 men invited to AAA screening, 37 (95% confidence interval 15 to 60) overdiagnosed men had unnecessary preventive surgery, of whom 1.6 (1.4 to 1.7) died
Limitations—Figures for exact calculations of overdiagnosis are not available and unlikely to emerge. The psychosocial consequences of living with a screen detected AAA are inadequately investigated. Cost effectiveness data on screening are inconclusive
Conclusion— Screening programmes have changed the meaning of an AAA diagnosis from a life threatening condition to a risk factor. AAA screening programmes should be revisited because of reduced benefits in modern populations and because data suggest considerable harm
Abdominal aortic aneurysms (AAA) are usually asymptomatic until they rupture, which is fatal in more than 80% of cases.1 Risk factors for developing AAA are smoking, male sex, advanced age, and family history.2 Screening aims to detect the aneurysm before it ruptures, enabling preventive surgery and hence reducing morbidity and mortality. However, preventive surgery has a mortality of 3.9-4.5%.3
Population level screening was introduced in Sweden,4 the United States,2 and the United Kingdom5 during the 2000s. In Sweden4 and the UK,5 all men aged 65 are offered screening with a one-off ultrasound examination.4 In the US, screening is recommended for men aged 65-75 who have ever smoked.2 Men with aneurysms of 30-54 mm are offered regular ultrasound surveillance for the rest of their lives. If the aneurysm is ≥55 mm or grows ≥10 mm annually, surgery is considered.6
Asymptomatic AAA may also be detected incidentally when a patient is receiving care for another problem. In this article we consider only aneurysms detected by screening programmes.
Evidence for current screening programmes
The decision to introduce AAA screening was based on four randomised controlled trials from the 1980s and 1990s comprising 137 214 men aged ≥65.7 8 9 10 A US Preventive Services Task Force review that included these trials estimated a 50% reduction in relative risk of AAA related mortality after 13-15 years’ follow-up (absolute risk reduction 0.46%). However, there was no significant reduction in overall mortality.2 Thus if 10 000 men are invited to screening, 46 deaths from AAA can be prevented over 13-15 years. Diagnosis could also prompt lifestyle changes to reduce risk of rupture.11
AAA screening has both benefits and unintended harms. As with many other forms of screening, overdiagnosis is the most serious harm12—that is, overdetection of aneurysms that would not have caused any symptoms during the person’s life or been the cause of their death.
Assumptions behind diagnostic change
The introduction of screening programmes has increased the frequency of testing for AAA. These programmes, based on the findings of the randomised trials, assume that earlier diagnosis will save lives without excess harm. They also assume that men who are currently invited for screening are at the same risk as those who participated in the trials of the 1980s and 1990s.
AAA is diagnosed if the aortic diameter is ≥30 mm. This definition has been broadly accepted since the 1960s.13 There has, however, been recent debate among vascular surgeons about lowering the cut-off value to 25 mm because studies show that some men with an aortic diameter of 25-29 mm at screening will later have an AAA rupture.10 14 15 16 17 18 19 It is assumed that the change would save more men from AAA death.
Effect on prevalence
There are no studies investigating how current screening programmes affect AAA prevalence. Our calculations below, based on the largest randomised trial, indicate that screening almost doubles AAA prevalence. If the definition of AAA is lowered to ≥25 mm the prevalence of AAA will double again.15 17
Evidence of overdiagnosis
Up to half of AAAs <55 mm barely grow,20 21 and it has been questioned whether these represent “true” aneurysmal disease.13 21 Most screen detected AAAs are small: 70% are <40 mm and 8.5% >54 mm.15 Average growth rates are 1.3 mm/year for 30 mm AAAs and 3.6 mm/year for 50 mm AAAs. Rupture rates are 0.05/100 person years and 0.64/100 person years for a 30 mm and 50 mm aneurysm respectively.22 Annual rupture rates for large aneurysms are 1% for AAAs of 50-59 mm and 14.1% for AAAs >60 mm.23 In summary, the smaller the aneurysm, the smaller the growth rate and the smaller the rupture risk. Therefore, by capturing mostly small aneurysms, screening can be argued to change the AAA diagnosis from a life threatening condition to a risk factor.
Proportion of men who do not benefit from diagnosis
The randomised trials did not estimate overdiagnosis7 8 9 10; neither did the US task force 2 and Cochrane6 reviews of AAA screening. Although the data are imperfect we have estimated the scale of overdiagnosis based on best available evidence. We used the largest trail (Multicentre Aneurysm Screening Study, MASS)10 to estimate the number of screened men who had an AAA diagnosed that would not have ruptured or required preventive surgery in the absence of screening over the course of the study (box 1). We calculated that 176 (95% confidence interval 150 to 202) men would be overdiagnosed for every 10 000 invited (fig 1⇓).
Box 1: Overdiagnosis estimated from a randomised controlled trial with a 13-year follow-up10
1334 of 33 883 men in the screening group had an AAA detected (absolute risk 3.94%, 95% confidence interval 3.73% to 4.14%)
739 of 33 887 men in the control group had AAA diagnosed (absolute risk 2.18%, 2.03% to 2.34%); 462 men had a rupture and 277 preventive surgery
Number of men overdiagnosed in the screening group=1334−739=595 (45%, 42% to 47%) men in the screening group were overdiagnosed
Absolute increase in risk of diagnosis=3.94-2.18=1.76
It is not only overdiagnosed men who do not benefit from being diagnosed from screening. In the MASS study, 1334 men had screen detected AAAs but only 157 (11.8%) of them avoided death as a result.10 For the rest, the early diagnosis did not improve prognosis—that is, the men died from or survived their AAA despite screening. For these men, the consequence of screening was living more years as a patient with AAA, not living longer.
To confirm these results we also estimated overdiagnosis from a cohort study of men screened in Gloucestershire (UK) from 1990-99 with 10-20 years’ follow-up.14 This study also reported data on AAA diameter, unlike the MASS study. We estimated the number of men whose care or death was not influenced by the diagnosis—that is, their AAA remained asymptomatic over the course of the study. We included all men with screen detected AAA and subtracted those who had emergency or preventive surgery and those who died from AAA rupture (table⇓). Overall 38% (311/816, 95% confidence interval 35% to 42%) of screened men who had an AAA ≥30 mm detected were overdiagnosed. [f2]
Both of our estimates of overdiagnosis have limitations that cannot be quantified. Firstly, some men would, even in the absence of screening, have had an AAA diagnosed (through incidental findings) that would have remained asymptomatic, leading to an overestimate of overdiagnosis caused by screening. Secondly, in the MASS study, some asymptomatic men in the control group might have had AAA diagnosed and treated through opportunistic screening leading to an underestimate of overdiagnosis. Thirdly, in the Gloucestershire study, some men who had preventive surgery were probably overdiagnosed (see later), thus underestimating overdiagnosis. Fourthly, new cases occurring during the long follow-up in both studies have the potential to both underestimate and overestimate overdiagnosis.24 Finally, some men in the studies may have had aneurysms that would not have ruptured until after the follow-up period, thus overestimating overdiagnosis. However, since the mortality benefit of screening was reduced after 10 years of follow-up10 this bias is likely to be small.
Changing the definition of an AAA from 30 to 25 mm is debated in the scientific community.10 14 15 16 17 18 19 We have found no evidence of benefit or analysis of the potential harms of widening of the definition. Based on the Gloucestershire study,14 10-20 years after screening, 87% (95% confidence interval 84% to 90%) of men found to have an AAA of 26-29 mm would be overdiagnosed (table⇑). This would raise the overdiagnosis rate from 38% to 58% (95% confidence interval 42% to 61%).
Drop in prevalence
Several studies have shown a drop in AAA prevalence over recent decades.14 15 25 A UK study found a 77% (4.8% to 1.1%) reduction of AAA prevalence from 1990 to 2009.14 This is probably because the prevalence of smoking has fallen25—for example, in 1974, 45% of the UK’s adult population smoked compared with 20% in 2012.26
When the incidence of the condition screened for decreases, the potential benefits also decrease. Therefore, the reduction of AAA related mortality seen in the trials is unlikely to be reproducible today. Moreover, current smoking increases AAA growth rate and doubles the risk of rupture.27 Therefore, the drop in smoking prevalence decreases rupture rates, increasing the relative rates of overdiagnosis. Consequently, the benefit:harm ratio is probably worse today than that in the trials.
Psychological burden of diagnosis
Quantitative studies on the psychosocial consequences of AAA screening are of insufficient quality to draw conclusions.2 6 28 However, qualitative interview studies show that men with screen detected AAAs report existential thoughts about frailty and mortality after diagnosis (box 2),29 30 31 with one man describing the diagnosis as “a ticking bomb inside your stomach.”29
Box 2: Quotes from an interview study with men with screen detected AAAs29
“I was very disappointed when I got the message . . . it was like to have your own death sentence.”
“We were digging . . . and then I saw on their faces that they didn´t want me to dig too much, strain myself you know, only they did it themselves, so I sat on a stool and watched.”
“But I do take it seriously, try to cut down on the smoking more and more. I’ve given up sugar, I’ve given up eating fatty food, bad cholesterol as much as possible.”
“The first thing that comes to my mind is that one might sit here and die any moment . . . of course one thinks, it comes over one that one might die at any time and the fact is that one thinks that if it´s to happen that one won´t take others along too, driving a car or something like that.”
Harms of surgery
Screening roughly doubles the rate of preventive surgery.2 8 9 10 A man who avoids death from a ruptured AAA clearly benefits from preventive surgery. In our calculations, we assume that a man who avoids a non-fatal rupture also benefits because surgery for ruptured AAAs has a higher complication rate than preventive surgery. Nevertheless, preventive surgery for AAA can result in impotence,32 myocardial infarction, respiratory failure, renal failure, ischaemic colitis, spinal cord ischaemia, and prosthetic graft infections.33 The complication rate is about 32%33 with a 30 day mortality of 4.7% for open repair.3 Endovascular techniques reduce 30 day mortality to 1.3%3 but have more long term complications; mortality is similar to that for open surgery after two years.3
In the two randomised trials that provided data on rupture rate, for every 10 000 men enrolled, 94 more men in the screened group had preventive surgery than in the control group over 13-15 years.8 10 Based on the rate of morbidity and mortality from AAA in the control group we estimate that had these 94 men not had surgery 57 would have had a ruptured aneurysm, of whom 44 would have died. Thus 37 (95% CI 15 to 60) had unnecessary preventive surgery, of whom 1.6 (1.4 to 1.7) died as a consequence.
Away from clinical trials concerns exist that AAA screening has resulted in a lowering of the threshold for surgery.34 One study found 59% of patients treated with endovascular techniques had an aortic diameter <55 mm35 despite evidence that surgery on small AAAs is not beneficial.33 This supports the view that detecting small AAAs contributes to overtreatment.34
Uncertainty over cost effectiveness
In 2009, a paper in The BMJ concluded that at a threshold of £30 000 (€41 000; $46 000) there was a less than 30% chance of AAA screening being cost effective.36 Three years later, another analysis concluded there was a 92% probability of AAA screening being cost effective at a threshold of £20 000.37 The two studies based their estimates on clinical data from different sources and neither attempted to mirror the worst and best case scenarios. They used different variables for prevalence,37 risk of rupture, annual growth rate, and costs, choosing variables that favoured their overall conclusions. Moreover, neither study considered the psychosocial costs of overdiagnosis and overtreatment.38
The proportion of men with AAA is greater among men who do not attend screening than among those who attend. Non-attendance is also correlated with lower sociodemographic characteristics.39 Therefore, AAA screening seemingly increases inequalities in health and does not reach those who would benefit the most. Additionally, 0.23% of all men with normal findings at screening experience AAA rupture within 13 years10 presenting a risk of false reassurance.
Screening men over 65 with a single ultrasound examination reduces mortality from AAA. However, according to our estimates 176 of 10 000 invited men are overdiagnosed. These men are unnecessarily turned into patients and may experience appreciable anxiety throughout their remaining lives. Moreover, 37 of these men unnecessarily have preventive surgery and 1.6 of them die as a consequence. Lowering the AAA definition from 30 to 25 mm will double AAA prevalence and substantially increase the rate of overdiagnosis.
When health authorities invite asymptomatic men to screening there should be no doubt that benefits clearly outweigh harms. We cannot judge whether this is true of AAA screening: the harms have not been adequately investigated, as is true for cancer screening.40 Furthermore, the benefit:harm ratio is likely to be worse in current screening programmes than in the trials on which they were based. We argue that AAA screening should be revisited.
Cite this as: BMJ 2015;350:h825
This article is part of a series on overdiagnosis looking at the risks and harms to patients of expanding definitions of disease and increasing use of new diagnostic technologies
Competing interests: We have read and understood BMJ policy on declaration of interests and declare MJ is supported by the Research Unit and Section for General Practice, FoUU-centrum Fyrbodal, Vänersborg, Sweden, and by Herrestads Healthcare Centre, Uddevalla, Region of Västra Götaland, Sweden. The funders had no role in the study design, data collection, analysis or interpretation.
Contributors and sources: MJ drafted the manuscript. AH and JB provided comments and contributed to revisions with important intellectual content. MJ is guarantor corresponding author.
Provenance and peer review: Not commissioned; externally peer reviewed.