Incidence among men of asymptomatic abdominal aortic aneurysms: estimates from 500 screen detected cases

Transcription

1 5 J Med Screen 1999;6:5 54 Incidence among men of asymptomatic abdominal aortic aneurysms: estimates from 5 screen detected cases K A Vardulaki, T C Prevost, N M Walker, N E Day, A B M Wilmink, C R G Quick, H A Ashton, R A P Scott Institute of Public Health, University of Cambridge, Forvie Site, Robinson Way, Cambridge CB2 2SR, UK K A Vardulaki, epidemiologist T C Prevost, statistician N M Walker, computer scientist NEDay,professor of epidemiology A B M Wilmink, specialist registrar, vascular surgery Department of Surgery, Hinchingbrooke Hospital, Huntingdon, Cambridge PE18 8NT, UK A B M Wilmink C R G Quick, consultant vascular surgeon Scott Research Unit, CMEC, St Richard s Hospital, Chichester, West Sussex PO19 4SE, UK H A Ashton, research associate R A P Scott, consultant vascular surgeon Correspondence to: Ms K A Vardulaki. Accepted for publication 1 February 1999 Abstract Objectives To determine the incidence of asymptomatic abdominal aortic aneurysms and the implications for an ultrasound screening programme in England and Wales. Methods First screen data were obtained from the Chichester and Huntingdon screening studies and used to estimate the prevalence of abdominal aortic aneurysms. The incidence of new, asymptomatic aneurysms was estimated from the prevalence rates observed in the Huntingdon screening study. Setting Screening programmes in Huntingdon and Chichester using ultrasound to screen all men over the age of 5 and men over age 65 respectively. Results The prevalence of abdominal aortic aneurysms ranged between 5.32% and 8.2% and between 6.18% and 9.88% of men aged between 65 and 79 in Chichester and Huntingdon respectively. Annual incidence rates, estimated by age, rose steadily reaching a peak of.67% of the Huntingdon population per year at age 65. Thereafter incidence falls. Estimates of the incidence of new asymptomatic abdominal aortic aneurysms, based on the observed prevalence data, were calculated and showed a peak at age 65. Conclusions Hypotheses are overed to explain this unexpected early peak in incidence. This information should allow the definition of the optimum age for screening, and the relative benefits of screening at diverent intervals if widespread screening is adopted in the future. (J Med Screen 1999;6:5 54) Keywords: prevalence; incidence; abdominal aortic aneurysm Ruptured abdominal aortic aneurysms cause approximately 2.1% of all deaths in men and.75% of all deaths in women over the age of 65 in England and Wales. 1 The benefit of screening by ultrasound is currently being evaluated in a randomised controlled trial, involving a single screening test for men aged and annual or three-monthly recalls for those with an enlarged aorta. If this trial provides substantial evidence of a reduction in mortality from ruptured aortic aneurysms, discussion on the introduction of screening as a public health measure would have to consider the age at which screening should be overed, whether screening should be repeated, and, if so, at what interval. Two important variables which determine the optimum screening strategy are the age-specific incidence rates of asymptomatic abdominal aortic aneurysms and the distribution of growth rates of aneurysms within the population. A pilot screening programme for abdominal aortic aneurysms has been in operation for several years in Chichester and Huntingdon. The results of these two programmes provide estimates of age-specific prevalence rates of abdominal aortic aneurysms, and the follow up of detected aneurysms provides estimates of their rate of growth. A separate paper on the distribution of aneurysmal growth by initial diameter and by age is published elsewhere. 2 Direct estimation of the incidence of new abdominal aortic aneurysms would require repeat screening of people initially without an aneurysm. Neither the Chichester nor the Huntingdon programmes rescreened normal subjects to any appreciable extent and to our knowledge there are no reported estimates of incidence based on rescreening normal subjects. One can, however, estimate age-specific incidence rates indirectly from age-specific prevalence rates, the approach taken in this paper. Our estimates are based on the fact that the number of prevalent aneurysms in a population at age i +1, is given by the number of prevalent aneurysms at i, plus the newly incident aneurysms, minus the aneurysms that have been removed from the population by death or repair. Strictly, longitudinal data are required so that the prevalence at diverent ages refers to the same people. If, however, one assumes that the changes in underlying risk of abdominal aortic aneurysms by year are small over a five year period, then one can use the age-specific prevalence rates determined cross sectionally at a single screening test. The incidence rate estimates presented in this paper refer to men only and are based on the results of the Huntingdon screening programme, which covered a wide age range (5 and older). The Huntingdon prevalence rates are similar to those for Chichester in the age group and consistent with those reported from other studies. 3 Estimates are based on systematic case ascertainment in a representative sample of the general population.

2 Asymptomatic abdominal aortic aneurysms 51 Table 1 Centre Prevalence of abdominal aortic aneurysms by age, sex, and centre Number of subjects with an abdominal aortic aneurysm Total population Prevalence (%) Increase over five years (%) Chichester Huntingdon Methods STUDY DESIGN AND POPULATIONS The Huntingdon Aneurysms Screening Project began in November 1991 with the intention to screen all men over 5 in the Huntingdon district. A detailed description of the protocol used in this study has been published elsewhere. 4 A total of 11 men aged over 5 were screened and 515 abdominal aortic aneurysms were detected. The analysis used first scan data for all men aged between 5 and 89, screened between 21 November 1991 and 6 January A second dataset came from a randomised controlled trial carried out in Chichester, which began screening over 5 men and women in A total of 182 abdominal aortic aneurysms were detected at first scan (ages 65 79).The methods and five year results of this study were published in Anteroposterior measurements of aortic diameter were used in both centres, though in Chichester the maximum of the anteroposterior and transverse measurements was used as the defining diameter in the primary analysis. 5 DATA ANALYSIS Table 1 shows the observed prevalence of abdominal aortic aneurysms, defined as an aortic diameter >3 mm, by five year age band, for both centres. A three parameter linear logistic model, which allowed prevalence to level ov in older age groups, was fitted to the prevalence data from the Huntingdon study (residual sum of squares (RSS) = 229.1). where a, b, and c are constants. This model fitted the data better than linear or exponential patterns of growth. From the raw data (table 1) it can be seen that the prevalence of abdominal aortic aneurysms rises steadily from.37% at age 5 54 to 9.4% at age The fit of a linear model to the data was substantially worse (RSS = 667.8) than that of the logistic model, reflecting that the rate of increase is not constant over time. The rate, the diverence in prevalence rates in five year bands, increases linearly from 1.37% between ages 5 54 and 55 59, to 3.21% between and In both studies, while prevalence still increases, the rate of increase falls after An exponential model is inappropriate because it would predict a very high prevalence in the oldest age group, much higher than the observed 11.19% in the 8+ group. Our three parameter linear logistic model included prevalence data for all men aged between 5 and 84. Prevalence by age was used to calculate an estimate of the incidence of asymptomatic abdominal aortic aneurysms in men. Incidence was defined as the percentage of the population expected to become aneurysmal over one year. It is assumed that the progression of an aneurysmal patient can be followed over one year by comparing the prevalence of abdominal aortic aneurysms in two consecutive age groups (for example, 69 and 7). By taking account of the number of deaths expected to occur in one age group, the expected prevalence estimate after one year can be directly compared with the observed prevalence estimate in the next age group. The method used is described in the appendix. Death rates for abdominal aortic aneurysms are presented by the OYce of National Statistics 1 in five year age bands and so a logistic regression model was fitted to the data to estimate death rates for aneurysms in yearly age bands. It has been suggested 6 that there is a degree of underreporting of deaths from abdominal aortic aneurysms as diagnosis may only follow a postmortem examination, and not all deaths result in autopsy. To allow for this and the successful repair of aneurysms, we looked at various models of the degree of underreporting, by age. Values for the expected magnitude of underreporting of deaths from abdominal aortic aneurysms, by age, are diycult to estimate. In this paper we assume that there is no underreporting under the age of 65. After age 65 we explore a range of diverent age dependent levels of underreporting. It has also been suggested that patients with abdominal aortic aneurysms have associated cardiovascular disease, which may result in higher all-cause death rates than the general population. 7 A recent paper, however, presents follow up results demonstrating high relative survival ratios for patients who had elective or emergency aortic repair, 8 in the period from four weeks to five years after surgery. These results show that apart from surgery related mortality within four weeks of surgery, excess mortality among those who have developed an abdominal aortic aneurysm, and gone on to have surgery, is small and these patients can be expected to have normal life expectancy after recovery. The proportion of patients with an abdominal aortic aneurysm that would be found unfit for surgery, on reaching the criteria, is unknown and it is among these patients that higher death rates might be seen. We considered the evect of known risk factors on incidence and the possibility of a cohort evect, but risk factors, including smoking, hypertension, and vascular disease, are unlikely to vary substantially between cohorts of men born a few years apart.

3 52 Vardulaki, Prevost, Walker, et al Prevalence of abdominal aortic aneurysms (% of population) Observed prevalence (5 year age bands) 95% CI for observed prevalence Estimated prevalence 95% CI for estimated prevalence Figure 1 Observed and estimated prevalence of abdominal aortic aneurysms. Results Table 1 shows the prevalence of abdominal aortic aneurysms by five year age band and sex. Prevalence ranged between 5.32% and 8.2% among the Chichester male population aged between 65 and 79. In Huntingdon, prevalence in the same age groups ranged between 6.18% and 9.88%. Figure 1 shows the three parameter linear logistic curve that best fits the prevalence data, with confidence intervals around each observed prevalence (in five year age bands) and around the regression line. The increase in prevalence with age can be seen clearly in fig 1 for all ages between 5 and 74. Prevalence rises steadily, with the steepest increase between ages 59 and 74. The rate of increase then begins to slow down, levelling ov after age 74. Figure 2 shows the incidence, by age, estimated from prevalence using the Huntingdon dataset. The percentage of the population expected to join the aneurysmal group over one year ranged from.8% (95% confidence interval (CI).58 to.14) of men aged 5 to.285% (95% CI.237 to.323) of 84 year old men. Incidence peaked at age 65 with.67% (95% CI.565 to.781) of the 65 year old male population becoming aneurysmal Incidence of abdominal aortic aneurysms (% of population) Estimated incidence 95% Confidence Interval Figure 2 Estimated incidence of abdominal aortic aneurysms. % Underreporting Incidence.6 A B Figure 3 Minimum level of underreporting required for incidence not to decrease with age. over one year. Incidence declined after age 65. The confidence intervals shown incorporate the uncertainty of having estimated the prevalence of abdominal aortic aneurysms and the mortality attributable to aneurysms in yearly age bands. The sensitivity of these incidence estimates to underreporting of deaths due to abdominal aortic aneurysms is explored, by assuming age dependent underreporting of death rates for aneurysms over age 65. To generate agespecific incidence rates which continue to increase after age 65, it is necessary to assume rapidly increasing levels of age-specific underreporting, from levels close to zero at age 65 to levels at age 7 and over, in which more than 75% of deaths due to abdominal aortic aneurysms are misclassified (see figs 3A and 3B). We regard such levels as implausible. Our sensitivity analysis shows that at plausible levels of age dependent underreporting, the general shape of the age-specific incidence curve seen in fig 2 is maintained. Discussion The prevalence of abdominal aortic aneurysms increases between the ages of 5 and 74. At this point the prevalence reaches around 1% of the population. A logistic curve gave a good fit to the Huntingdon data, in which prevalence increases with age and begins to level ov after age 74 (see fig 1). Both linear and exponential models do not fit the data as well as this three parameter linear logistic model. Other families of curves for example, the Gompertz, could have been used; they give essentially the same results. We rejected a linear model as the rate of increase is not constant over time (see table 1), it does not allow for a tail in younger age groups, and does not allow prevalence to level ov in older age groups. An exponential model was also rejected as it would predict a prevalence of around 2% in the oldest age group. The aspect of the results presented in this paper which requires most discussion is the strong indication that the incidence of abdominal aortic aneurysms stops increasing with age at around age 65. After age 65, the incidence may even fall. The critical feature of the Huntingdon age-specific prevalence rates which induces this evect on the estimates of age-specific incidence is the change in the pattern of increasing 8

4 Asymptomatic abdominal aortic aneurysms 53 Rate per 1 population per year Figure 4 Death rates from abdominal aortic aneurysms in England and Wales, prevalence with age. Between ages 5 and 7 the increase of prevalence with age is accelerating (see the last column in table 1).The diverence between successive five year prevalence rates increases smoothly with age from 1.37% to 3.21%. This increase reflects continuously increasing age-specific incidences. After the 7 74 age group, however, there is a sharp fall in the rate of increase in the five year prevalence rate from 3.21 to.49, then As can be seen from the relation between prevalence and incidence discussed in the appendix, this fall in the rate of increase in the age-specific prevalence would follow from a fall in the age-specific incidence rates. It would not occur if the age-specific incidence continued to rise. Two questions should be asked, however. Firstly, is this fall real? and, secondly, if real, are there explanations which would allow a continuously increasing age-specific incidence? That the fall in the rate of increase in age-specific prevalence is real, and not a statistical artefact, is supported by the same finding in the Chichester data, with a drop from 2.18 to.47 (see table 1). The age-specific prevalence rates are slightly lower in Chichester than in Huntingdon, but the pattern of the change with age is very similar. In terms of other explanations, the observed pattern of age-specific prevalence may have arisen from a more rapid depletion with age of the prevalent pool of men with aneurysms than is allowed for. This more rapid depletion might occur either because the death rate from abdominal aortic aneurysms increases more rapidly with age than described by the OYce of National Statistics (ONS) mortality statistics, or because men with aneurysms have a substantially higher death rate from other, probably vascular, causes. However, recent results from Australia 9 indicate strongly that there is little overall excess mortality among men with abdominal aortic aneurysms who have undergone elective repair (that is, deaths due to causes other than rupture of an aneurysm). A variety of levels of age dependent underreporting of deaths attributed to abdominal aortic aneurysms have been explored. The proportion of deaths wrongly certified might be expected to increase with age as a sudden death at age 85 is less likely to result in an autopsy than one at age 65. However, allowing for diverent levels of underreporting by age does not make a substantial diverence to the pattern of age-specific incidence rates. The degree of underreporting required to prevent a decline in incidence after age 65 would require in excess of 6% of deaths from abdominal aortic aneurysms, over age 65, to be wrongly certified, which we regard as largely implausible. A further cause of depletion of the prevalence pool would be people leaving the pool after emergency or elective surgery. In the study presented here, however, the number of such people is relatively small. It might also be argued that as death rates from abdominal aortic aneurysms increase with age, at least up to ages (fig 4), one might expect agespecific incidence to increase in parallel. We note, however, that in the case of abdominal aortic aneurysms, death is related to the prevalence of larger aneurysms and not the incidence of small ones. The data and the modelling presented in this paper indicate strongly that the age-specific incidence of new abdominal aortic aneurysms reaches a peak in the mid-6s age range, and then declines at least until age 8. Further support for this pattern of age-specific incidence comes from two sources. Firstly, in Huntingdon the overall distribution of abdominal aortic diameter, not just the proportion with a diameter greater than or equal to 3 mm, was examined by age. 9 Surprisingly, although the prevalence of abdominal aortic aneurysms rises to over 11% over age 8, the 75th centile of aortic diameter does not change with age. This indicates that for 75% or more of the population the aorta is not expanding with age, unlike the increase in blood pressure with age that avects most of the population. Thus only a relatively small proportion of the population are at risk of an expanding aorta, and hence of an aneurysm. The age-specific incidence pattern presented in this paper is consistent with the existence of a relatively small pool of men susceptible to the development of an abdominal aortic aneurysm. Also, when age-specific mortality from ruptured abdominal aortic aneurysms is considered as a proportion of total age-specific mortality (fig 5), the pattern seen is similar (at least up to age 8) to that proposed for incidence in figs 2, 3A, and 3B, at least in general shape. This pattern of age-specific proportional mortality suggests that mortality from abdominal aortic aneurysms does not behave with age in a similar fashion to mortality from other vascular conditions. Our estimates of incidence of asymptomatic abdominal aortic aneurysms use data from over 1 subjects screened in Huntingdon. The incidence of abdominal aortic aneurysms could be estimated by rescreening a population found to have a normal aortic diameter at the prevalence screen, but such a study would need many years of follow up. An incidence of

5 54 Vardulaki, Prevost, Walker, et al % Of all deaths due to abdominal aortic aneurysms Figure 5 Proportional mortality (deaths from abdominal aortic aneurysms as a proportion of all deaths, 1996) % among a population of 1 observed over five years would generate only 25 cases. Estimates would be dependent on a test able to distinguish between a 29 mm aortic diameter, deemed as normal, and a 3 mm aneurysmal diameter. Owing to the length of time required to collect these data, the inevitable loss to follow up would influence any estimate. Estimates of the incidence of asymptomatic abdominal aortic aneurysms provide an indication of the degree of morbidity we would expect to detect at diverent time intervals after a prevalence screen. It is essential to know the incidence in order to assess the most appropriate age for screening and the relative benefits of screening at diverent intervals. The rate of growth of aneurysms of divering diameters and their corresponding risk of rupture has been examined in a previous paper. 2 In combination with these estimates of incidence, such information should permit an accurate evaluation of the potential benefits derived from different screening strategies and allow for the design and adequate provision of local screening facilities and surgical services. We intend to explore the implication for screening policy of the incidence results presented here in a later analysis. We thank Shaun Seaman at the Institute of Public Health for his help with Splus. Appendix 1 With data from one prevalence scan, assumptions as to the progress of abdominal aortic aneurysms over time in diverent age groups can be made. The assumption is made that the people expected to die from an aneurysm will be those described as aneurysmal at the prevalence scan. The data from Chichester suggest that this is a valid assumption as no study subject died of an aneurysm within five years after a normal aortic scan. The calculation is as follows: the yearly incidence in year i to i +1is given by: where P i is the prevalence rate of abdominal aortic aneurysms at age i, r i is the annual rate of deaths attributable to aneurysms at age i, and a i is the annual rate of deaths attributable to all causes at age i. All death rates were obtained from ONS mortality statistics. 1 The estimate of incidence obtained from the formula outlined above will be a minimum estimate as it is likely that a small percentage of incident cases will die from another cause, rather than from an abdominal aortic aneurysms, during this one year period. As indicated in the text, the number of people leaving the prevalence pool in this series owing to elective surgery is small, and therefore ignored. 1 OYce of National Statistics. Mortality statistics, cause, England and Wales London: TSO, (Series DH2, No 22.) 2 Vardulaki KA, Prevost TC, Walker NM, et al. Growth rates and risk of rupture of abdominal aortic aneurysms. Br J Surg 1998;85: Wilmink ABM, Quick CRG. Epidemiology and potential for prevention of abdominal aortic aneurysm. Br J Surg 1998;85: Morris GE, Hubbard CS, Quick CRG. An abdominal aortic aneurysm screening programme for all males over the age of 5 years. Eur J Vasc Surg 1994;8: Scott RAP, Wilson NM, Ashton HA, et al. Influence of screening on the incidence of ruptured abdominal aortic aneurysm: 5 year results of a randomised controlled study. Br J Surg 1995;82: Lederle F. Screening for snipers: the burden of proof. J Clin Epidemiol 199;43: Aune S, Amundsen SR, Evjensvold J, et al. Operative mortality and long term relative survival of patients operated on for asymptomatic abdominal aortic aneurysm. Eur J Vasc Endovasc Surg 1995;9: Norman PE, Semmens JB, Lawrence-Brown MMD, et al. Long term survival after surgery for abdominal aortic aneurysm in Western Australia: population based study. BMJ 1998;317: Wilmink ABM, Pleumeekers HJCM, Hoes AW, et al. The normal infrarenal aortic diameter in relation to age: only a part of the population shows an increase in older age groups. Eur J Vasc Endovasc Surg 1998;16: