Population-based study of incidence and outcome of acute aortic dissection and pre-morbid risk-factor control: 10-year results from the Oxford Vascular Study

Abstract

Background

Acute aortic dissection is a preventable life-threatening condition. However, there have been no prospective population-based studies of incidence or outcome to inform understanding of risk factors, strategies for prevention, or projections for future clinical service provision.

Methods and Results

We prospectively determined incidence and outcomes of all acute aortic dissections in a population of 92,728 in Oxfordshire, UK, during 2002-2012. Among 154 patients with 173 acute aortic events, 54 patients had 59 thoraco-abdominal aortic dissections (52 incident events - 6/100,000, 95%CI 4-7; 37 Stanford type-A, 15 type-B; 31 men, mean age =72.0yrs). Among patients with type-A incident events, 18 (48.6%) died prior to hospital assessment (61.1% women). The 30-day fatality rate was 47.4% for type-A cases who survived to hospital admission and 13.3% for type-B cases, although subsequent 5-year survival rates were high (85.7% for type-A; 83.3% for type-B). Even though 67.3% of patients were on antihypertensive drugs, 46.0% of all patients had at least one systolic-BP ≥180 mmHg in their primary-care records over the preceding 5 years, and the proportion of BPs in the hypertensive range (>140/90 mmHg) averaged 56.0%. Premorbid BP was higher in patients with Type-A dissections that were immediately fatal than in those who survived to admission (mean/SD pre-event SBP = 151.2/19.3 vs 137.9/17.9; p<0.001).

Conclusions

Uncontrolled hypertension remains the most significant treatable risk factor for acute aortic dissection. Prospective population-based ascertainment showed that hospital-based registries will underestimate not only incidence and case-fatality but also the association with pre-morbid hypertension.

Introduction

Acute manifestations of aortic pathology, including symptomatic and ruptured thoracic and abdominal aneurysms and acute dissections, represent some of the most serious vascular emergencies. Acute aortic dissection has a particularly high case-fatality,^1–2 despite well-established treatment guidelines.^3–5 However, unlike coronary and cerebral arterial disease,^6–10 data on risk factors, incidence and outcome of acute aortic disease are limited and there have been no prospective population-based studies.^11–18 There is some evidence that prevalence of abdominal aortic aneurysm and incidence of rupture are declining,^19–21 or at least following the broader shift in incidence of vascular events to older age groups,²² but trends in acute dissection are uncertain.²³

Existing hospital-based studies, often from specialist centres, or studies of retrospective registry data,^11–18 such as the International Registry of Acute Aortic Dissection (IRAD),^11,24–26 may underestimate incidence and case-fatality by incomplete inclusion of deaths prior to hospital admission, which might also bias assessment of risk factors and predictors of outcome. Of the two studies of the epidemiology of aortic dissection commenced since 1980,^1,12 both were retrospective and used only routinely collected diagnostic or mortality coding data to ascertain subjects and neither assessed pre-morbid risk factors or functional outcome. We therefore prospectively studied event rates, incidence, risk factors, early case fatality, and long-term outcome of all acute aortic events occurring in a population of 92,728 in Oxfordshire, UK, during 2002-2012, as part of the Oxford Vascular Study (OXVASC).

Methods

The OXVASC study population comprises all individuals (10-year average = 92,728), irrespective of age, registered with 63 family physicians in nine general practices in Oxfordshire, UK. In the UK, the vast majority of individuals register with a general practice, which provides their primary health care and holds a lifelong record of all medical consultations, and details of medications, blood pressure measurements, and investigations. All participating practices held accurate age-sex patient registers, and allowed regular searches of their computerised diagnostic coding systems. All practices refer patients to only one secondary care centre.

Case ascertainment was by prospective daily searches for acute events (“hot pursuit”) and retrospective searches of hospital and primary care administrative and diagnostic coding data (“cold pursuit”). Hot pursuit was based on the daily assessment of all patients with a possible vascular event identified by: 1) Daily searches of Emergency Department admission and symptom/diagnosis registers; 2) Daily listing from the central admissions department of all patients from our general practices admitted to hospital, and assessment of these patients in hospital; 3) Daily visits to the cardiac surgery and vascular surgery wards and review of daily lists of all patients referred to vascular surgery; 4) Daily identification via Bereavement Officers of patients dead on arrival at hospital or who died soon after; 5) Daily assessment of all patients undergoing diagnostic angiographic, angioplasty/stenting or arterial surgical procedures in any territory.

The methods of cold pursuit were: 1) Weekly review of all listed surgical procedures undertaken by vascular and cardiovascular surgery; 2) Frequent contact with general practices and monthly searches of computerised practice diagnostic codes; 3) Monthly practice-specific list of all patients with relevant diagnostic codes from the coding departments covering all acute and community hospitals; 4) Monthly visits to the Coroner’s Office to review out-of-hospital deaths; 5) Review of all death certificates and relevant clinical details in the study general practices; 6) Practice-specific listings of all ICD-10 death codes from the local Department of Public Health; 7) Review of vascular surgery outpatient clinic letters to identify all patients attending vascular clinics who were not admitted to hospital.

A study clinician assessed patients as soon as possible after the event. Informed consent was sought, where possible, or assent was obtained from a relative. Standardised clinical history and cardiovascular examination were recorded. We also recorded from the patient, their hospital records and their general practice records, details of the clinical event, medication, past medical history, all investigations relevant to their admission and all interventions occurring subsequent to the event.

If a patient died prior to assessment, we obtained eyewitness accounts and reviewed any relevant records. If death occurred outside hospital or prior to investigation, the autopsy result was reviewed. Clinical details were sought from primary care physicians or other clinicians on all deaths of possible vascular aetiology. Initial clinical assessments were made by study clinical research fellows alongside the clinical teams. All diagnoses were subsequently reviewed by a vascular surgeon.

All patients were followed-up by a research nurse or via their family doctor, with recurrent events also identified by the on-going study surveillance. If a recurrent vascular event was suspected, the patient was assessed by a study physician. Disability was assessed with the modified Rankin Scale (mRS).²⁷

All patients with acute vascular events affecting the aorta from the 1^st April 2002 to the 31^st March 2012 were included. Acute aortic dissection was classified according to the Stanford system²⁸ into type-A (proximal to left subclavian artery origin) or type-B (distal to left subclavian artery origin). Event rates were defined as the total number of vascular events that led to separate clinical presentations during the study period. All events were categorised as first-ever incident or recurrent and specific to territory. An incident event implied the first ever aortic dissection, irrespective of type. Second events were always recurrent, irrespective of type (i.e. a first type-A event would be classed as recurrent in a patients with only a previous type-b event). An extension of a previous dissection was not considered to be a recurrent event if it occurred within 6-months of the first event.

Patients who had an event whilst temporarily away from Oxfordshire were included, but visitors to Oxfordshire who were not registered with one of the study practices were excluded. The population structure derived from the general practice age/sex registers was stable over the study period. For the purpose of analysis we used the mean population derived from the ten mid-year population age-sex structures.

Statistical analysis

Analyses were done with SPSS (version 19.0; SPSS Inc, Chicago, Ill). Age-specific and sex-specific rates (per 100,000 population per year) were calculated for incidence and mortality. Since all comparisons were within-population, rates were not standardised to any external populations. Numeric data were expressed as means (standard deviation) or proportions, as appropriate. Group differences in continuous variables were examined with Student t test or Mann-Whitney U test for parametric and non-parametric variables, respectively. Group differences in categorical variables were examined with Fisher Exact test or χ2 test, as appropriate. Gender-differences in outcome were assessed for using multivariable binary logistic regression analysis with adjustment for age. Survival rates were derived by Kaplan-Meier analysis.

The Oxford Vascular Study has been approved by our local research ethics committee and all eligible patients gave informed consent. Study assent was obtained from next of kin if the patient was unable to consent.

Results

A total of 173 events occurred in 155 patients during the study period. 54 patients had 59 thoraco-abdominal aortic dissections (52 incident events - 6/100,000, 95% Confidence Intervals (CI) 4-7) and 101 patients had 114 ruptured/symptomatic aortic aneurysms (101 incident events - 11/100,000, 95%CI 9-13). Of the 52 incident dissections, 37 (71.2%) were Stanford type-A and 15 (28.8%) were type-B. Two patients had a type-A dissection during the study period, but had previous type-A dissections prior to the study period (both surgically repaired) and were therefore classed as recurrent events. Five patients had two separate dissections (excluding acute extensions) during the study period (2 recurrent type-A dissections after an incident type-A dissection, 1 type-B dissection after an incident type-B dissection, 1 type-B dissection after an incident type-A, and 1 type-A dissection after an incident type-B).

Table 1 shows the demographics and risk factors for incident acute aortic dissections by gender and by Stanford classification. Mean (SD) age for acute aortic dissection was 72.0 (14.5) years and the male:female ratio was 1.5:1. Women suffered from acute dissections at a significantly higher mean age than men overall (79.3 versus 67.1 years; p=0.002) and in those cases not associated with Marfan’s (79.3 versus 69.6 years; p=0.003). Of type-B aortic dissection patients, 53.3% had prior history of stroke compared with only 13.5% of type-A dissections (p=0.005). Of the 8 patients with type-B dissection who had history of stroke, 7 had ischaemic stroke and one was a subarachnoid haemorrhage. 3 patients had known Marfan’s syndrome.

Table 1.

Demographics and risk factors for incident aortic dissection by gender and type

	Total	Total (excl. Marfan’s)	Male	Female	P	Type-A	Type-B	P
	n=52	n=49	n=31	n=21		n=37	n=15
Mean (SD) age, years	72.0 (14.5)	73.8 (11.5)	67.1 (14.9)	79.3 (10.4)	0.002	72.5 (15.4)	70.8 (12.3)	0.70
Male	31 (59.6%)	27 (55.1%)				20 (54.1%)	11 (73.3%)	0.23
Prior Vascular disease
Angina	5 (9.6%)	5 (10.2%)	2 (6.5%)	3 (14.3%)	0.38	4 (10.8%)	1 (6.7%)	1.00
Acute coronary syndrome	5 (9.6%)	5 (10.2%)	3 (9.7%)	2 (9.5%)	1.00	3 (8.1%)	2 (13.3%)	0.62
TIA	4 (7.7%)	3 (6.1%)	2 (6.5%)	2 (9.5%)	1.00	1 (2.7%)	2 (13.3%)	0.07
Stroke	13 (25.0%)	12 (24.5%)	6 (19.4%)	7 (33.3%)	0.25	5 (13.5%)	8 (53.3%)	0.005
Peripheral arterial disease	3 (5.8%)	2 (4.1%)	2 (6.5%)	1 (4.8%)	1.00	2 (5.4%)	1 (6.7%)	1.00
Risk factors
Current smoker	9 (17.3%)	9 (18.4%)	6 (19.4%)	3 (14.3%)	0.72	6 (16.2%)	3 (20.0%)	1.00
Ever smoked	32 (61.5%)	31 (63.3%)	21 (67.7%)	11 (52.4%)	0.48	20 (57.1%)	12 (80.0%)	0.20
Marfans syndrome	3 (5.8%)	0 (0.0%)	3 (9.7%)	0 (0.0%)	0.26	1 (2.7%)	2 (13.3%)	0.20
Hypertension	35 (67.3%)	34 (69.4%)	18 (58.1%)	17 (81.0%)	0.08	25 (67.6%)	10 (66.7%)	0.95
Diabetes mellitus	6 (11.5%)	5 (10.2%)	4 (12.9%)	2 (9.5%)	1.00	3 (8.1%)	3 (20.0%)	0.34
Cardiac failure	3 (5.8%)	3 (6.1%)	0 (0.0%)	3 (14.3%)	0.06	0 (0.0%)	3 (20.0%)	0.02
Atrial fibrillation	8 (15.4%)	8 (16.3%)	5 (16.1%)	3(14.3%)	1.00	4 (10.8)	4 (26.7%)	0.21
Medications
Statin	15 (28.8%)	14 (28.6%)	9 (29.0%)	6 (28.6%)	0.97	8 (21.6%)	7 (46.7%)	0.07
Aspirin	19 (36.5%)	19 (38.8%)	10 (32.3%)	9 (42.9%)	0.44	11 (29.7%)	8 (53.3%)	0.11
Other antiplatelet agents	2 (3.8%)	2 (4.1%)	1 (3.2%)	1 (4.8%)	1.00	1 (2.7%)	1 (6.7%)	0.50
Warfarin	3 (5.8%)	1 (2.0%)	3 (9.7%)	0 (0.0%)	0.26	1 (2.7%)	2 (13.3%)	0.20
Antihypertensives
0	17 (32.7%)	16 (32.7%)	13 (41.9%)	4 (19.0%)		13 (35.1%)	4 (26.7%)
1	11 (21.2%)	11 (22.4%)	4 (12.9%)	7 (33.3%)		9 (24.3%)	2 (13.3%)
2	14 (26.9%)	14 (28.6%)	9 (29.0%)	5 (23.8%)		10 (27.0%)	4 (26.7%)
≥ 3	10 (19.2%)	8 (16.3%)	5 (16.1%)	5 (23.8%)	0.17	5 (13.5%)	5 (33.3%)	0.39

Categorical data is expressed as 'n (%)' throughout

The most prevalent vascular risk factors were known hypertension (67.3%) and having ever smoked (61.5%). For hypertension, the recorded prevalence increased significantly over the 10-years study period from 42.9% in years 1-3 to 85.7% in years 8-10 (trend over 10 years - p=0.008), reflecting increased routine screening in primary care. With regards to the 92,728 study population, 23,808 subjects were aged ≥55, of which 8,954 (37.6%) had diagnosed hypertension and 11,480 (48.2%) had ever-smoked. There were no statistically significant sex differences in risk factors for aortic dissection, although hypertension tended to be more common in women, who were also more likely to be on antihypertensive medication.

Pre-morbid blood pressure (BP) measurements were available for 50/52 (96.2%) patients with a mean of 24.7 (+/- 19.5) recordings per patient. Pre-morbid control of BP was poor despite 67.3% of patients being on antihypertensive medication. Over the 5 years prior to the incident dissection, the proportion of all BPs recorded in primary care in individual patients that were greater than 140/90 mmHg averaged 56.0% and was similar for type-A (55.4%) and type-B (57.1%) dissections (p=0.84). Results were similar in analyses of measurements over the 15 years prior to the incident dissection, with an average of 58.3% >140/90 mmHg. Maximum prior recorded SBP was ≥180 mmHg in 46.0% of patients and was similar for men (41.4%) and women (52.4%), but tendered to be more common in those suffering type-B dissections (60.0%) compared to type-A (40.0%) (p=0.19). Among those patients who had been started on antihypertensive treatment, 61.9% of subsequent BP readings still exceeded 140/90 despite the majority of patients being on combination therapy (21.2% monotherapy; 26.9% dual, 19.2% ≥ triple therapy.) In patients with diagnosed hypertension, the proportion of patients on two or more antihypertensive drugs did increase during the study period from 28.6% in years 1-3 to 66.7% in years 8-10 (trend over 10 years - p=0.02).

Figure 1 (and Supplemental Table 1) shows numbers and age/sex-specific incidence rates for incident dissection events. Age of onset was approximately 10 years later in women versus men, but rates were similar at age ≥75 years.

Figure 1.

Age- and sex-specific rates per hundred thousand population for incident acute aortic dissections (2002-2012)

A total of 19 interventional procedures were performed following acute dissection during the study period. 15 patients (11 with type-A dissection) underwent emergency surgical treatment; 12 open aortic surgical repairs, 2 endovascular aortic stenting procedures (EVAR), 1 renal artery stenting procedure and 2 open limb procedures (amputations). In two patients aortic intervention was delayed > 30 days from event following initial conservative treatment (1 open repair and 1 EVAR). For patients surviving until hospital discharge (40.7% of total), median length of stay was 11.5 days (interquartile range 6 - 22.5). Two patients required community hospital rehabilitation. For incident cases, 80.8% were fully independent (modified Rankin Scale (mRS) 0-2) prior to event (mean mRS 1.5 (1.1 SD)). For those discharged from hospital alive, mean 30-day mRS was 2.6 (1.0) and mean 6-month mRS was 2.1 (1.3), with 72.7% of these patients independent at 6 months post-event.

Of the 52 patients with incident events, 33 died during the study period. The 30-day and 5-year fatality rates (figure 2) were 55.8% and 64.5%, but were higher for type-A dissections (73.0% and 76.8% respectively) than for type-B dissections (13.3% and 33.3% respectively). The incident dissection was immediately fatal in 18 cases (13 were certified as dead at home and 5 were dead on arrival at hospital), all with type-A dissections. In patients with incident type-A dissection who survived to hospital admission, 30-day mortality was 47.4%. Among those who survived to hospital discharge, subsequent 5-year survival rates were high (85.7% for type-A; 83.3% for type-B). After adjustment for age, immediate death, 30-day, and 5-year mortality rates were higher in women versus men (adjusted OR: 6.4, 95% CI 1.5-26.5, p=0.01 and 3.6, 0.95-13.6, p=0.06; and adjusted HR: 2.0, 0.95-4.3, p=0.07 respectively). During the 5 years prior to the event, BP was higher in patients with Type-A dissections that were immediately fatal than in those cases who survived to admission (Table 2): mean (SD) pre-event SBP = 151.2 (19.3) versus 137.9 (17.9); p<0.001.

Figure 2.

Mortality of acute aortic dissections with numbers at risk tabled below. Top: 30-day; Middle: 5-year; Bottom: 5-year in those surviving to hospital discharge.

Table 2.

Demographics and risk factors of patients with incident type-A dissection events who died immediately versus those who survived to hospital admission

	Immediate Death	Survived to Admission	P
	n=18	n=19
Mean (SD) age, years	70.8 (18.7)	74.2 (11.6)	0.51
Male	7 (38.9%)	13 (68.4%)	0.07
Prior Vascular disease
Angina	1 (5.6%)	3 (15.8%)	0.60
Acute coronary syndrome	1 (5.6%)	2 (10.5%)	1.00
TIA	0 (0.0%)	1 (5.3%)	1.00
Stroke	4 (22.2%)	1 (5.3%)	0.18
Peripheral arterial disease	1 (5.6%)	1 (5.3%)	1.00
Risk factors
Current smoker	4 (22.2%)	2 (10.5%)	0.39
Ever smoked	9 (50.0%)	11 (57.9%)	0.92
Marfans syndrome	1 (5.6%)	0 (0.0%)	0.49
Hypertension	12 (66.7%)	13 (68.4%)	1.00
Diabetes mellitus	1 (5.6%)	2 (10.5%)	1.00
Cardiac failure	0 (0.0%)	0 (0.0%)	1.00
Atrial fibrillation	1 (5.6%)	3 (15.8%)	0.60
Medications
Statin	5 (27.8%)	3 (15.8%)	0.45
Aspirin	5 (27.8%)	6 (31.6%)	0.80
Other antiplatelet agents	0 (0.0%)	1 (5.3%)	1.00
Warfarin	0 (0.0%)	1 (5.3%)	1.00
Pre-Morbid Blood Pressure Control
Mean (SD) Systolic BP in last 5yrs	151.2 (19.3)	137.9 (17.9)	<0.001
Mean % of BPs ≥140/90 mmHg in last 5yrs	71.9 (28.7)	39.1 (21.8)	0.01
Mean % of BPs ≥140/90 mmHg ever	59.9 (29.2)	49.7 (23.9)	0.26

Categorical data is expressed as 'n (%)' throughout

Discussion

Previous studies of aortic dissection have been hospital-based, restricted by age or vascular territory, and have had incomplete ascertainment due to exclusion of out-of-hospital cases.^1,4,12,26 Previous epidemiological studies of aortic dissection also predated widespread use of modern diagnostic technology such as CT angiography, and so diagnostic accuracy was potentially questionable.^1,2,4,12 Our study is the first prospective population-based study with comprehensive case-ascertainment and high levels of patient participation.²⁹

We have reported four main findings. First, although the age and sex distributions of our cases were similar to those in comparable studies,^2,11,12,30 the incidence of acute dissection in our study was higher than previously estimated, probably due to more complete inclusion of deaths prior to hospital admission. It is also possible that improvements in vascular imaging might have contributed and that the incidence of acute dissection may have increased since previous studies. Second, we showed that incidence of acute aortic dissection is now approximately half that of ruptured/symptomatic aortic aneurysms and therefore represents a significant clinical burden. Third, we showed that hospital-based databases, such as the International Registry of Acute Aortic Dissection (IRAD), which include only patients who reach designated tertiary referral centres alive, will miss a substantial proportion (~50%) of patients with acute type-A dissections who die at home or before hospital admission, thereby underestimating overall disease incidence and both short and long-term case fatality. This selection bias is also likely to account for the lower in-hospital mortality rates found in IRAD compared to other studies where a significant transfer time to a specialist centre is not required.

Finally, although patients with dissections had low rates of prior vascular disease, including ischaemic heart disease, diabetes, and peripheral arterial disease, we confirmed previous observations that hypertension is very strongly associated with aortic dissection.^2,30 Moreover, in the first detailed analysis of pre-morbid blood pressure control in patients presenting with acute dissection, we showed that prior control of BP was poor, despite the majority of patients being on combination antihypertensive therapy, particularly in immediately fatal cases. This high rate of treatment-resistant hypertension may reflect increased aortic stiffness or other pathophysiological mechanisms responsible for the development of acute aortic syndromes. However, better control of BP would nevertheless be likely to reduce incidence and case fatality. Trials of BP-lowering drugs should include aortic dissection in composite outcomes intended to determine the overall impact of treatment.

Our findings also have implications for future clinical service provision. Over the next 40 years the UK population aged 75 or above will double (from 8% to 15%).³¹ Our data suggest that the proportion of incident dissection events that occur over age 75 will reach 49.7% in 2030 and 57.3% by 2050 (see Supplemental Projection Analysis). Improved primary prevention, in particular more aggressive management of hypertension and smoking cessation, may reduce future incidence rates, but treatment-resistant hypertension is likely to remain a challenge.

Although we consider our findings to be valid, our study has a number of potential shortcomings. First, the OXVASC population is 94% white and the proportions of other racial groups are too small (3.1% Asian, 1.5% Chinese and 1.4% Afro-Caribbean) to determine separate event rates and outcomes.³² However, the proportion of whites is similar in the UK as a whole (88% white) and in many other western countries (Australia - 90%; France - 91%; Germany - 93.9%), although our results will not necessarily apply in the United States (72.4% White, 12.6% African American, 4.8% Asian, 0.9% American Indian, 6.2% Chinese/Other). However, there are few differences in the demographics and outcome of aortic dissections between the IRAD centres in Europe (99% white) and those in North America (91% white).³³ Second, based on the Index of Multiple Deprivation (IMD),³⁴ the electoral districts covering our population are less deprived than the rest of England (mean IMD score: 8.69 versus 16.98, t-test - p<0.001). However, our population still covers a broad range of deprivation, with 22% of our districts ranking in the lower third nationally. Third, it is unlikely that we ascertained every single acute aortic event in our population as some patients suffering acute type-B aortic dissection events can be asymptomatic or choose not to present to medical services. Moreover, although we identified many out-of-hospital deaths, not all patients who die in the community undergo post-mortem examination, and so it is feasible that some aortic dissection sudden deaths may have been incorrectly certified.

In conclusion, in the first ever prospective population-based study of acute aortic disease, we have shown that incidence of acute aortic dissection is higher than previous estimates, and have demonstrated the importance of inclusion of out-of-hospital deaths if incidence and outcome are to be reliably determined. We have also highlighted the critical association between acute aortic dissection and hypertension, particularly hypertension resistant to medication. Future demographic changes are likely to lead to increased numbers of events over the next few decades, particularly in the elderly, with implications for health service provision and future research, especially since surgical outcome has not improved substantially in recent years despite significant progress in diagnosis and endovascular treatment.^23,28,35

Short commentary.

In the first prospective population-based study of acute aortic dissection, all aortic events over a 10 year period (2002-2012) were ascertained in a population of 92,728 in Oxfordshire, UK. Incidence of acute aortic dissection was higher than previous estimates, due at least partly to inclusion of out-of-hospital deaths, which accounted for a third of all cases. The most prevalent pre-morbid risk factor for acute aortic dissection was hypertension, particularly hypertension resistant to medication. Even though 67% of patients were on antihypertensive drugs, 46% of all patients had at least one systolic-BP ≥180 mmHg in their primary-care records over the preceding 5 years, and the proportion of BPs in the hypertensive range (>140/90 mmHg) averaged 56%. Hospital-based registries will underestimate not only incidence and case-fatality of acute aortic dissection but also the association with pre-morbid hypertension, which remains the most significant treatable risk factor. Future demographic changes are likely to lead to increasing numbers of aortic dissections over the next few decades, particularly at older ages, with implications for health service provision and future research.