Development and validation of an ankle brachial index risk model for the prediction of cardiovascular events

Abstract

Background

The ankle brachial index (ABI) is related to risk of cardiovascular events independent of the Framingham risk score (FRS). The aim of this study was to develop and evaluate a risk model for cardiovascular events incorporating the ABI and FRS.

Design

An analysis of participant data from 18 cohorts in which 24,375 men and 20,377 women free of coronary heart disease had ABI measured and were followed up for events.

Methods

Subjects were divided into a development and internal validation dataset and an external validation dataset. Two models, comprising FRS and FRS + ABI, were fitted for the primary outcome of major coronary events.

Results

In predicting events in the external validation dataset, C-index for the FRS was 0.672 (95% CI 0.599 to 0.737) in men and 0.578 (95% CI 0.492 to 0.661) in women. The FRS + ABI led to a small increase in C-index in men to 0.685 (95% CI 0.612 to 0.749) and large increase in women to 0.690 (95% CI 0.605 to 0.764) with net reclassification improvement (NRI) of 4.3% (95% CI 0.0 to 7.6%, p = 0.050) and 9.6% (95% CI 6.1 to 16.4%, p < 0.001), respectively. Restricting the FRS + ABI model to those with FRS intermediate 10-year risk of 10 to 19% resulted in higher NRI of 15.9% (95% CI 6.1 to 20.6%, p < 0.001) in men and 23.3% (95% CI 13.8 to 62.5%, p = 0.002) in women. However, incorporating ABI in an improved newly fitted risk factor model had a nonsignificant effect: NRI 2.0% (95% CI 2.3 to 4.2%, p = 0.567) in men and 1.1% (95% CI 1.9 to 4.0%, p = 0.483) in women.

Conclusions

An ABI risk model may improve prediction especially in individuals at intermediate risk and when performance of the base risk factor model is modest.

Keywords

Ankle brachial index cardiovascular diseases risk assessment

Introduction

The accurate prediction of major cardiovascular events in individuals without vascular disease is an important public health goal, allowing targeting of preventive measures to those at increased risk. The Framingham risk score (FRS), which combines classic cardiovascular risk factors such as cigarette smoking and total cholesterol, is the most well-known method of risk prediction. However, the FRS has limited accuracy,¹ and attempts have been made to improve prediction by incorporating additional risk factors, novel biomarkers, and measures of subclinical atherosclerosis.²

The ankle brachial index (ABI), which is the ratio of ankle:arm systolic pressure, is a measure of atherosclerosis in symptomatic and asymptomatic individuals. Originally developed as a diagnostic tool for leg artery disease, the ABI is also an indicator of generalized atherosclerosis and of future cardiovascular events, independent of established risk factors.³ Since the ABI can be measured using only a handheld Doppler probe and sphygmomanometer in many healthcare settings, the possibility of using the ABI to enhance risk prediction is of interest.

The ABI Collaboration was formed of investigators performing longitudinal studies of the ABI and incident cardiovascular events. In a meta-analysis comprising 48,294 subjects, a low ABI (≤0.90) compared to a normal ABI (1.11–1.40) was related to a 2–3-fold increase in both 10-year major coronary events and cardiovascular mortality independent of the FRS.³ Furthermore, in cardiovascular risk stratification using the FRS, subsequent inclusion of the ABI resulted in reclassification of risk (low, intermediate, high) in 1 in 5 men and 1 in 3 women. However, the impact of reclassification in improving risk prediction was not assessed.

In this present study, our aim was to develop an ABI prediction model incorporating FRS and ABI into a single equation. The ABI model was then evaluated in a different population and the effect of reclassification of 10-year risk assessed. Since risk prediction varies according to gender and ethnic group,⁴ the model was developed separately in men and women and only in whites who comprised the largest ethnic group in the ABI Collaboration studies.

Methods

Study design

The study design was an analysis of individual participant data from 18 prospective cohort studies. Risk prediction models were fitted in a development dataset and evaluated in internal and external validation datasets.

Inclusion and exclusion of studies

The study was based on 20 cohort datasets in the ABI Collaboration. The literature search and collation of data from 16 of these studies^5–20 were described previously.³ Since then, experts informed us of four further studies,^21–24 from which corresponding data were acquired. Two studies included only non-White ethnic groups^7,8 and were excluded, leaving 18 studies in the final analysis. Individual participant data were extracted and analysed using version 9.2 of SAS System for Windows (2002–2008; SAS Institute, Cary, NC, USA).

Subjects, variables, and missing data

Subjects with nonvalid ABI, with prevalent coronary heart disease as defined in each study at baseline, and without follow up for vital status were excluded. Framingham covariates extracted were age, gender, systolic and diastolic blood pressures, total and high-density lipoprotein cholesterol, and smoking and diabetes indicators. ABI, study location, and ethnic group were obtained and only subjects included who were classified as ‘white’ using individual study classifications. Analyses were not performed on non-White ethnic groups because of small sample sizes. The numbers of men and women in the final analysis were 24,375 and 20,377, respectively. Only 7.4% of data in men and 2.9% in women were missing, predominantly for total and high-density lipoprotein cholesterol. Imputation was performed separately by gender using the SAS procedure PROC MI with the MCMC full-data imputation method. FRS was calculated for all individuals using the methods of Wilson et al.²⁵ In keeping with our previous investigation,³ ABI was categorized into four groups: ≤0.90, 0.91–1.10, 1.11–1.40, >1.40.

Development, internal validation, and external validation datasets

The studies were divided into two groups separately by gender to create, first, a model development and internal validation dataset and, secondly, an external validation dataset. Studies were allocated to these datasets based on study location (USA or not), earlier vs. later start date, larger vs. smaller study, age range of participants, and median duration of follow up in order to achieve a balance across datasets. Studies with one or more wholly imputed covariates were constrained to be in the external validation dataset. The characteristics of the studies are shown in Table 1. The first dataset was then divided in two by randomly selecting half the participants in each study. Prediction models were fitted in the first half (development dataset) and validated in the other half (internal validation dataset) and finally in the external validation dataset.

Table 1.

Participants and characteristics of studies in the development/internal validation and external validation datasets for men and women

Study sample	Location	Age (range years)	Follow up (median, IQR) years)	Cardiovascular deaths	Major coronary events	Study start
Men
Development/internal validation studies
Belgian Physical Fitness Study¹³	2020	EU	40–57	10.0 (10.0–10.0)	22	32	1976
Edinburgh Artery Study⁹	690	EU	55–76	15.5 (9.0–15.9)	84	113	1987
MONICA Augsburg Survey²¹	654	EU	25–74	12.8 (12.6–12.9)	54	51	1989
InCHIANTI Study¹⁷	507	EU	23–94	6.0 (5.9–6.1)	34	–	1998
Hoorn Study¹⁶	271	EU	51–77	12.5 (9.8–13.1)	27	–	1989
ARIC Study⁵	4748	USA	45–64	13.2 (12.4–13.9)	111	452	1987
Cardiovascular Health Study⁶	1555	USA	65–96	11.2 (7.4–11.6)	223	363	1989
San Luis Valley Diabetes Study¹⁹	675	USA	24–74	15.6 (14.4–16.9)	53	98	1984
Health ABC Study²²	663	USA	68–80	9.1 (8.5–9.4)	43	74	1997
San Diego Study¹⁸	243	USA	41–81	16.6 (10.3–22.3)	77	80	1978
All studies	12,026	728	1263
External validation studies
Limburg PAOD Study¹⁰	1037	EU	41–78	7.1 (6.6–7.7)	35	83	1987
Men Born in 1914 Study¹¹	392	EU	68–69	13.2 (8.1–13.9)	85	92	1982
Health in Men Study¹⁵	3217	Australia	65–83	6.3 (6.0–6.6)	142	–	1997
Mr OS Study²⁴	4167	USA	64–100	6.9 (6.4–7.4)	215	–	2003
Framingham Offspring Study¹⁴	1423	USA	29–86	7.4 (6.6–8.2)	20	56	1995
Rotterdam Study¹²	2113	EU	55–98	10.9 (8.3–11.9)	216	230	1990
All studies	12,349	713	461
Women
Development/internal validation studies
Edinburgh Artery Study⁹	702	EU	54–75	15.8 (14.2–16.1)	41	57	1987
MONICA Augsburg Survey²¹	629	EU	25–74	12.8 (12.7–12.9)	20	13	1989
Hoorn Study¹⁶	286	EU	50–76	12.6 (10.5–13.2)	25	–	1989
InCHIANTI Study¹⁷	654	EU	21–95	6.1 (5.9–6.2)	34	–	1998
San Diego Study¹⁸	313	USA	38–81	19.6 (13.0–22.6)	76	65	1978
San Luis Valley Diabetes Study¹⁹	838	USA	21–71	15.8 (14.6–17.1)	56	70	1984
ARIC Study⁵	5719	USA	45–64	13.2 (12.5–14.0)	73	235	1987
Cardiovascular Health Study⁶	2322	USA	64–100	11.3 (10.0–11.7)	224	302	1989
Women’s Health and Ageing Study²⁰	476	USA	65–101	5.0 (3.8–5.1)	63	115	1992
Health ABC Study²³	742	USA	68–80	9.2 (8.8–9.4)	41	52	1997
All studies	12,681	653	909
External validation studies
Limburg PAOD Study¹⁰	1324	EU	40–78	7.1 (6.7–7.6)	27	54	1987
Study of Osteoporotic Fractures²³	1233	USA	65–89	17.1 (11.5–20.4)	244	–	1986
Framingham Offspring Study¹⁴	1703	USA	29–86	7.4 (6.6–8.3)	5	24	1995
Rotterdam Study¹²	3436	EU	55–106	11.1 (9.6–12.2)	334	239	1990
All studies	7696	610	317

Subjects in each of the development/internal validation studies were allocated randomly to either the development or internal validation datasets.

–

, not available.

Model fitting

Two models were fitted each for the primary outcome of major coronary events (myocardial infarction or death due to coronary heart disease) and the secondary outcome of cardiovascular mortality (death due to coronary heart disease or stroke) using Cox’s proportional hazards model, as follows: model 1: Framingham risk score²⁵ fitted as a continuous variable (FRS); model 2: as per model 1 with addition of ABI group (FRS + ABI).

These models were first fitted for men and women separately in the development datasets. Details of the fitted FRS + ABI for major coronary events are shown in Supplementary Table 1 and measures of model performance^26–28 in Supplementary Table 2 (available online).

For comparison with the FRS, an additional risk factor model was fitted in the development dataset in which, instead of using the overall single FRS value, each individual Framingham risk covariate, such as smoking, was used separately. And then a further model was fitted with addition of ABI group (Supplementary Table 1). Assumptions for Cox’s analyses were checked formally.

Outcome measures

To quantify the effect of inclusion of the ABI in risk prediction models, the C-index and net reclassification improvement (NRI) were derived.

The C-index is analogous to C-statistic, which is the area under the receiver operating characteristic curve and is a measure of the model’s discrimination of events and non-events. The C-index is for survival data, being the fraction of occasions where the predictor score correctly predicts the earlier event for pairs of individuals in the sample. This was achieved using the methods of Harrell et al.²⁹ and Pencina and D’Agostino³⁰ adopting the simplified method for confidence intervals.

For the calculation of the NRI, reclassification tables comparing predicted risk categories for models with and without ABI were constructed for the external validation dataset. Ten-year risk categories of <10%, 10–19%, and ≥20% were selected for major coronary events and <2%, 2–4%, and ≥5% for cardiovascular mortality. NRIs were calculated taking account of censored data.³¹ Confidence intervals and p-values were derived using methods for the standard NRI.

For major coronary events, analysis was repeated with a wider intermediate group of 5–19%, akin to some recent studies of risk prediction.^32,33

Results

Predicting events in the whole study population

Table 2 shows that C-indices in the internal validation dataset were mostly similar to those in the development dataset but in the external validation dataset were mostly lower. In the latter, C-indices for the FRS in men were 0.672 (95% CI 0.599 to 0.737) for major coronary events and 0.684 (95% CI 0.625 to 0.738) for cardiovascular mortality, and in women were particularly low: 0.578 (95% CI 0.492 to 0.661) and 0.449 (95% CI 0.377 to 0.523), respectively. Incorporation of the ABI led to a modest increase in the C-index for men to 0.685 (95% CI 0.605 to 0.764) for major coronary events and 0.710 (95% CI 0.652 to 0.762) for cardiovascular mortality. However, in women, the increase was much larger, to 0.690 (95% CI 0.605 to 0.764) for major coronary events and 0.652 (95% CI 0.579 to 0.719) for cardiovascular mortality so that the C-indices approximated more closely to those of men.

Table 2.

C-indices for Framingham risk score predictions of major coronary events and cardiovascular mortality in men and women with and without the ankle brachial index for the development and internal and external validation datasets

Prediction model	Development	Internal validation	External validation
Major coronary events
Men	(n = 5632)	(n = 5638)	(n = 4962)
FRS alone	0.715 (0.655–0.768)	0.721 (0.664–0.722)	0.672 (0.599–0.737)
FRS + ABI	0.721 (0.661–0.773)	0.721 (0.664–0.722)	0.685 (0.612–0.749)
Women	(n = 5869)	(n = 5872)	(n = 6459)
FRS alone	0.661 (0.587–0.728)	0.676 (0.599–0.745)	0.578 (0.492–0.661)
FRS + ABI	0.681 (0.607–0.746)	0.710 (0.633–0.775)	0.690 (0.605–0.764)
Cardiovascular mortality
Men	(n = 6010)	(n = 6016)	(n = 12,349)
FRS alone	0.809 (0.737–0.865)	0.794 (0.719–0.852)	0.684 (0.625–0.738)
FRS + ABI	0.817 (0.746–0.872)	0.797 (0.723–0.855)	0.710 (0.652–0.762)
Women	(n = 6339)	(n = 6342)	(n = 7696)
FRS alone	0.568 (0.484–0.647)	0.604 (0.514–0.688)	0.449 (0.377–0.523)
FRS + ABI	0.667 (0.585–0.740)	0.689 (0.600–0.765)	0.652 (0.579–0.719)

Values are C-index (95% CI)

The number of subjects in major coronary events data sets are lower than in cardiovascular mortality datasets because of nonfatal events not ascertained in four studies in men (Health in Men, Mr Os, Hoorn, INCHIANTI) and three studies in women (Hoorn, InCHIANTI, Study of Osteoporotic Fractures)

ABI, ankle brachial index; cardiovascular mortality, death due to coronary heart disease or stroke; FRS, Framingham risk score; major coronary event, myocardial infarction or death due to coronary heart disease.

The effect of the FRS + ABI model on prediction, in comparison to the FRS, is shown in Figure 1 according to low, intermediate, and high 10-year risk for a major coronary event. Within each FRS risk category, the FRS + ABI model refined the degree of risk corresponding to trends in event rates (except for a low event rate in women classified at intermediate risk of 10–19% by both models).

The impact of reclassification of risk using the FRS + ABI model compared to the FRS is shown in Table 3. For major coronary events, the NRI was 4.3% (95% CI 0.0 to 7.6%, p = 0.050) in men and 9.6% (95% CI 6.1 to 16.4%, p < 0.001) in women and included a net increase in risk category in those having an event. For cardiovascular mortality, the NRI was 5.7% (95% CI 2.7 to 7.9%, p < 0.001) in men and 15.7% (95% CI 11.3 to 20.2%, p < 0.001) in women in whom improved classification occurred in those having and not having a cardiovascular death. Detailed reclassification data for the primary outcome of major coronary events are shown in Supplementary Tables 3 and 4.

Table 3.

Reclassification in predicting major coronary events and cardiovascular mortality for the Framingham risk score with ankle brachial index compared to Framingham risk score alone in men and women

	Persons having event		Persons not having event		NRI	p-value
	Up	Down	Up	Down	NRI	p-value
Major coronary events
Men	9.1	6.7	6.8	8.3	4.3 (0.0–7.6,)	0.050
Women	18.9	2.5	8.0	2.9	9.6 (6.1–16.4)	<0.001
Cardiovascular mortality
Men	5.3	6.2	4.5	10.6	5.7 (2.7–7.9)	<0.001
Women	19.2	10.7	6.2	13.4	15.7 (11.3–20.2)^a	<0.001

Values are % (95% CI). Results derived from external validation dataset. Data for primary outcome of major coronary events shown in Supplementary Tables 3 and 4. Reclassification up and down shows the percentage of persons in whom the 10-year risk category increased or decreased, respectively, between <10%, 10–19%, and ≥20% for major coronary events and <2%, 2–4%, and ≥5% for cardiovascular mortality

Based on crude event rates and not survival method because lack of subjects in low risk group did not permit accurate estimation

Cardiovascular mortality, death due to coronary heart disease or stroke; major coronary event, myocardial infarction or death due to coronary heart disease; NRI, net reclassification improvement: NRI = {(P(event|up) – P(event)) * P(up) + (P(event) – P(event|down)) * P(down)}/{P(event) * (1 – P(event)}, with P(event), P(event|up), and P(event|down) all estimated using the Kaplan–Meier approach.

Predicting events in subjects at intermediate risk

Restricting use of the ABI model to only those at intermediate 10-year FRS risk had a greater effect (Table 4) than in all subjects. In those with a 10–19% risk for a major coronary event, incorporation of the ABI resulted in a NRI of 15.9% (95% CI 6.1 to 20.6%, p < 0.001) in men and 23.3% (95% CI 13.8 to 62.5%, p = 0.002) in women. This was due to a net increase in subjects having an event reclassified as higher risk and in those not having an event reclassified as lower risk. In restricting use of the ABI to those at intermediate 10-year risk of 2–4% for cardiovascular death, NRIs were likewise higher than in the whole population but were similar in men and women: 20.2% (95% CI 11.5 to 29.1%, p < 0.001) and 18.0% (95% CI 13.1 to 22.9%, p < 0.001), respectively.

Table 4.

Reclassification in predicting major coronary events and cardiovascular mortality for the Framingham risk score with ankle brachial index compared to Framingham risk score alone in men and women at intermediate risk

	Persons having event		Persons not having event		NRI	p-value
	Up	Down	Up	Down
Major coronary events
Men	14.7	8.9	8.1	15.7	15.9 (6.1–20.6)	<0.001
Women	38.9	13.9	13.2	26.4	23.3 (13.8–62.5)	0.002
Cardiovascular mortality
Men	24.7	4.1	9.1	8.8	20.2 (11.5–29.1)	<0.001
Women	26.5	0.0	8.5	0.0	18.0 (13.1–22.9)^a	<0.001

Values are % (95% CI). Results derived from external validation dataset. Data for primary outcome of major coronary events shown in Supplementary Tables 3 and 4. Study subjects are only those in FRS 10-year intermediate risk categories of 10–19% for major coronary events and 2–4% for cardiovascular mortality. Reclassification up and down shows the percentage of persons in whom 10-year risk either increased above or decreased below intermediate risk category, respectively

Based on crude event rates and not survival method because probability estimates not feasible when no person reclassified down

The impact of reclassification on major coronary events using the FRS + ABI model was analysed using a wider FRS intermediate 10-year risk category of 5–19% (Supplementary Tables 5 and 6). This categorization resulted in very few numbers in the <5% risk group. In the whole population, the NRI for men was modest (3.1% (95% CI 0.6 to 6.4%, p = 0.018)) but for women was considerable (20.4% (95% CI 11.6 to 22.5%, p < 0.001)), with improved net reclassification for those having and not having an event. Restricting the FRS + ABI model to the 5–19% intermediate group led to a higher NRI in men (7.9% (95% CI 3.7 to 11.5%, p < 0.001)) but a lower NRI in women (13.0% (95% CI 7.3 to 17.9%, p < 0.001)).

Predicting events using cardiovascular risk covariate model

C-indices for the newly developed risk factor model in predicting major coronary events in the external validation dataset were 0.683 (95% CI 0.611 to 0.748) in men and 0.788 (95% CI 0.709 to 0.850) in women, which were slightly higher in men and considerably higher in women than the corresponding FRS C-indices in Table 2. Incorporation of the ABI resulted in only a slight improvement, increasing C-indices to 0.690 (95% CI 0.618 to 0.754) in men and 0.791(95% CI 0.712 to 0.852) in women, with nonsignificant NRIs of 2.0% (95% CI –2.3 to 4.2%, p = 0.567) and 1.1% (95% CI –1.9 to 4.0%, p = 0.483), respectively. In only those at intermediate 10–19% risk, NRIs were 7.7% (95% CI 0.0 to 13.0%, p = 0.049) in men and 2.4% (95% CI –3.0 to 10.5%, p = 0.275) in women.

Discussion

Main findings

In this analysis combining data from 18 population-based studies, a new ABI risk model incorporating the FRS + ABI was developed and then validated in an external population. In comparison to the FRS, as measured by C-index and NRI, the ABI model led to an improvement in performance, but mostly in women. Restricting use of the ABI model to those at intermediate 10-year risk of 10–19% for major coronary events resulted in higher NRIs than in the whole population. In a wider intermediate risk group of 5–19%, the NRIs were lower than in the 10–19% risk group.

The more impressive results in women than in men might be related to the particularly poor performance of the FRS in women in the external validation dataset in which C-indices were much lower than in the development and internal validation datasets (Table 2). This may have been due to chance or unexpected population differences. The main effect of incorporating the ABI may have been to compensate for this poor performance by bringing the C-indices up towards those of men and not that the ABI model was inherently superior in women than in men. When a better performing model based on fitting individual risk covariates was used instead of the FRS, incorporation of the ABI in the prediction of major coronary events led to no significant improvement. These results suggest that the impact of the ABI is not a fixed phenomenon but is influenced by how well the base risk factor model performs.

Other studies

The Multi-Ethnic Study of Atherosclerosis (MESA) found that the ABI contributed to an improvement in classification of cardiovascular events and non-events compared to traditional risk factors as measured by integrated discrimination improvement.³⁴ In the Atherosclerosis Risk in Communities (ARIC) Study and Rotterdam Study, each components of the ABI Collaboration, the effect on prediction of adding the ABI to a cardiovascular risk equation was examined^32,35 and found, in the whole study populations, to have minimal impact. In keeping with the present study, both the ARIC and Rotterdam studies found improved performance if the ABI was used only in patients at intermediate risk. In all men and women combined, the NRI was 0.8% in the ARIC Study and 0.6% in the Rotterdam Study and, in those at intermediate risk, was 8.3% and 7.3%, respectively. A similar result (NRI 7.0%) was found in the elderly aged 70–79 years at intermediate risk in the Health, Aging and Body Composition Study.³⁶ However, it has been pointed out recently that findings in intermediate groups may be overly optimistic due to possible bias in their estimation.³⁷

Comparison with coronary artery calcium

In the Rotterdam Study, coronary artery calcium (CAC) showed better prediction than the ABI when added to a locally developed risk factor model.³⁵ The NRIs for models in the whole population were 19.3% for CAC and 0.6% for ABI and, in those at intermediate risk, were 39.3% for CAC and 7.3% for ABI. Likewise in intermediate risk individuals in MESA, the NRI was much higher for CAC than for other measures of subclinical atherosclerosis including ABI.³⁶ However, measurement of CAC is technically sophisticated, expensive, and involves radiation exposure,³⁸ and so is less suited than the ABI to measurement in primary care or in population screening. The ABI can be measured easily with minimal training, using simple inexpensive equipment and in less than 15 min.³⁹ A risk prediction programme incorporating measures of subclinical disease might benefit from tiered assessment with more complex and costly tests targeted on fewer individuals. Research is required on the costs and effectiveness of such an approach.

Limitations

A principal strength of this study was that, due to large numbers of subjects, models could be tested in both internal and external validation datasets derived from 18 population studies from several countries. A limitation of using multiple studies is that measurement of variables, including ABI, and the ascertainment and definition of endpoints were not identical. However, studies were only included where consistent and valid methods were used. Furthermore, using two study outcomes with hard endpoints (major coronary events and cardiovascular mortality) and finding broadly similar results, provided some reassurance on validity. Nevertheless, very large simulated datasets have been used as an alternative approach to overcoming some of these difficulties.⁴⁰ A further limitation of the study was that the exclusion criteria at baseline of only subjects with coronary heart disease meant that some subjects with other cardiovascular disease would be included, but were likely to make up a small proportion of the total population. Insufficient data were available to study non-White populations.

The FRS applied in this study²⁵ has been adopted in many practice settings but, more recently, the Framingham group produced a general cardiovascular disease risk factor model to be used in primary care for predicting a wider range of events.⁴¹ Due to limitations in our dataset, assessment of these outcomes was not feasible but, if predicting a wide range of events was desirable, evaluation of ABI risk models in this context would be useful, as would the contribution of the ABI to other risk models such as SCORE.⁴²

Clinical implications

The ABI has been recognized for some time as potentially useful in screening healthy individuals for cardiovascular risk.⁴³ However, guidelines have varied recommendations on use of the ABI. Recently, the US Preventive Services Task Force, while recognizing that use of the ABI would reclassify some men and women in the FRS intermediate risk category, considered the evidence insufficient to assess the benefits and harms, and concluded that clinical discretion was required.⁴⁴ On the other hand, an American College of Cardiology Foundation/American Heart Association guideline included a recommendation, based on level of evidence B (nonrandomized studies), that ‘Measurement of ABI is reasonable for cardiovascular risk assessment in asymptomatic adults at intermediate risk’.⁴⁵ Recent multi-Society European guidelines had a similar recommendation.⁴⁶ The results of our study provide some support for the use of an ABI risk model especially in individuals at intermediate risk and when performance of the base risk factor model is modest. Furthermore, if physicians are uncertain about how well the FRS performs in their practice, using the ABI model is likely to compensate for any deficiencies in the FRS.

Footnotes

Acknowledgements

The Atherosclerosis Risk in Communities study is carried out as a collaborative study. The authors thank the staff and participants of the ARIC study for their important contributions.

References

Brindle

Beswick

Fahey

. Accuracy and impact of risk assessment in the primary prevention of cardiovascular disease: a systematic review. Heart 2006; 92: 1752–1759.

Greenland

Smith

SC Jr

Grundy

. Improving coronary heart disease risk assessment in asymptomatic people: role of traditional risk factors and non-invasive cardiovascular tests. Circulation 2001; 104: 1863–1867.

Ankle Brachial Index Collaboration. Ankle brachial index combined with Framingham Risk Score to predict cardiovascular events and mortality. A meta-analysis. JAMA 2008; 300: 197–208.

D’Agostino

RB Sr

Grundy

Sullivan

for the

CHD

Risk Prediction Group . Validation of the Framingham coronary heart disease prediction scores. Results of a multiple ethnic groups investigation. JAMA 2001; 286: 180–187.

Weatherley

Nelson

Heiss

. The association of the ankle-brachial index with incident coronary heart disease: the Atherosclerosis Risk in Communities (ARIC) study, 1987–2001. BMC Cardiovasc Disord 2007; 7: 3–3.

Newman

Shemanski

Manolio

. Ankle-arm index as a predictor of cardiovascular disease and mortality in the Cardiovascular Health Study. The Cardiovascular Health Study Group. Arterioscler Thromb Vasc Biol 1999; 19: 538–545.

Abbott

Petrovitch

Rodriguez

. Ankle/brachial blood pressure in men >70 years of age and the risk of coronary heart disease. Am J Cardiol 2000; 86: 280–284.

Resnick

Lindsay

McDermott

. Relationship of high and low ankle brachial index to all-cause and cardiovascular disease mortality: the Strong Heart Study. Circulation 2004; 109: 733–739.

Leng

Fowkes

Lee

. Use of ankle brachial pressure index to predict cardiovascular events and death: a cohort study. BMJ 1996; 313: 1440–1444.

10.

Hooi

Kester

Stoffers

. Asymptomatic peripheral arterial occlusive disease predicted cardiovascular morbidity and mortality in a 7-year follow-up study. J Clin Epidemiol 2004; 57: 294–300.

11.

Ogren

Hedblad

Isacsson

. Non-invasively detected carotid stenosis and ischaemic heart disease in men with leg arteriosclerosis. Lancet 1993; 342: 1138–1141.

12.

van der Meer

Bots

Hofman

. Predictive value of noninvasive measures of atherosclerosis for incident myocardial infarction: the Rotterdam Study. Circulation 2004; 109: 1089–1094.

13.

Kornitzer

Dramaix

Sobolski

. Ankle/arm pressure index in asymptomatic middle-aged males: an independent predictor of ten-year coronary heart disease mortality. Angiology 1995; 46: 211–219.

14.

Murabito

Evans

Nieto

. Prevalence and clinical correlates of peripheral arterial disease in the Framingham Offspring Study. Am Heart J 2002; 143: 961–965.

15.

Fowler

Jamrozik

Norman

. Prevalence of peripheral arterial disease: persistence of excess risk in former smokers. Aust N Z J Public Health 2002; 26: 219–224.

16.

Jager

Kostense

Ruhe

. Microalbuminuria and peripheral arterial disease are independent predictors of cardiovascular and all-cause mortality, especially among hypertensive subjects: five-year follow-up of the Hoorn Study. Arterioscler Thromb Vasc Biol 1999; 19: 617–624.

17.

McDermott

Guralnik

Albay

. Impairments of muscles and nerves associated with peripheral arterial disease and their relationship with lower extremity functioning: the InCHIANTI Study. J Am Geriatr Soc 2004; 52: 405–410.

18.

Criqui

Langer

Fronek

. Mortality over a period of 10 years in patients with peripheral arterial disease. N Engl J Med 1992; 326: 381–386.

19.

Hiatt

Hoag

Hamman

. Effect of diagnostic criteria on the prevalence of peripheral arterial disease. The San Luis Valley Diabetes Study. Circulation 1995; 91: 1472–1479.

20.

McDermott

Fried

Simonsick

. Asymptomatic peripheral arterial disease is independently associated with impaired lower extremity functioning: the women’s health and aging study. Circulation 2000; 101: 1007–1012.

21.

Lamina

Meisinger

Heid

Kora Study Group . Association of ankle-brachial index and plaques in the carotid and femoral arteries with cardiovascular events and total mortality in a population-based study with 13 years of follow up. Eur Heart J 2006; 27: 2580–2587.

22.

Sutton-Tyrrell

Venkitachalam

Kanaya

. Relationship of ankle blood pressures to cardiovascular events in older adults. Stroke 2008; 39: 863–869.

23.

Vogt

Cauley

Kuller

. Prevalence and correlates of lower extremity arterial disease in elderly women. Am J Epidemiol 1993; 137: 559–568.

24.

Collins

Ewing

Diem

Osteoporotic Fractures in Men (MrOS) Study Group . Peripheral arterial disease is associated with higher rates if hip bone loss and increased fracture risk in older men. Circulation 2009; 119: 2305–2312.

25.

Wilson

D’Agostino

Levy

. Prediction of coronary heart disease using risk factor categories. Circulation 1998; 97: 1837–1847.

26.

Akaike

. A new look at the statistical model identification. IEEE Trans Automatic Control 1974; 19: 716–723.

27.

Schwarz

. Estimating the dimension of a model. Ann Stat 1978; 6: 461–464.

28.

Nagelkerke

NJD

. A note on a general definition of the coefficient of determination. Biometrika 1991; 78: 691–692.

29.

Harrel

FE Jr

Lee

Mark

. Multivariate prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996; 15: 361–387.

30.

Pencina

D’Agostino

. Overall C as a measure of discrimination in survival analysis: model specific population value and confidence interval estimation. Stat Med 2004; 23: 2109–2123.

31.

Pencina

D’Agostino

RB Sr

Steyerberg

. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med 2011; 30: 11–21.

32.

Murphy

Dhangana

Pencina

. Ankle-brachial index and cardiovascular risk prediction: an analysis of 11,594 individuals with 10-year follow up. Atherosclerosis 2012; 220: 160–167.

33.

Yeboah

McClelland

Polonsky

. Comparison of novel risk markers for improvement in cardiovascular risk assessment in intermediate-risk individuals. JAMA 2012; 308: 788–795.

34.

Criqui

McClelland

McDermott

. The ankle-brachial index and incident cardiovascular events in MESA (Multi-Ethnic Study of Atherosclerosis). J Am Coll Cardiol 2010; 56: 1506–1512.

35.

Kavousi

Elias-Smale

Rutten

JHW

. Evaluation of newer risk markers for coronary heart disease risk classification. A cohort study. Ann Intern Med 2012; 156: 438–444.

36.

Rodondi

Marques-Vidal

Butler

. Health, Aging, and Body Composition Study. Markers of atherosclerosis and inflammation for prediction of coronary heart disease in older adults. Am J Epidemiol 2010; 171: 540–549.

37.

Paynter

Cook

. A bias-corrected net reclassification improvement for clinical subgroups. Med Decis Making 2013; 33: 154–162.

38.

Kim

Einstein

Berrington de Gonzalez

. Coronary artery calcification screening: estimated radiation dose and cancer risk. Arch Intern Med 2009; 169: 1188–1194.

39.

Mohler

ER III

Treat-Jacobson

Reilly

. Utility and barriers to performance of the ankle-brachial index in primary care practice. Vasc Med 2004; 9: 253–260.

40.

Wald

Simmonds

Morris

. Screening for future cardiovascular disease using age alone compared with multiple risk factors and age. PLoS ONE 2011; 6: e18742–e18742.

41.

D’Agostino

Vasan

Pencina

. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 2008; 117: 743–753.

42.

Conroy

Pyorala

Fitzgerald

. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J 2003; 24: 987–1003.

43.

Hirsch

Haskal

Hertzer

. American Association for Vascular Surgery; Society for Vascular Surgery; Society for Cardiovascular Angiography and Interventions; Society for Vascular Medicine and Biology; Society of Interventional Radiology; ACC/AHA Task Force on Practice Guidelines Writing Committee to Develop Guidelines for the Management of Patients With Peripheral Arterial Disease; American Association of Cardiovascular and Pulmonary Rehabilitation; National Heart, Lung, and Blood Institute; Society for Vascular Nursing; TransAtlantic Inter-Society Consensus; Vascular Disease Foundation. ACC/AHA 2005 practice guidelines for the management of patients with peripheral arterial disease (lower extremity, renal, mesenteric, and abdominal aortic). Circulation 2006; 113: e463–e654.

44.

Moyer

. Preventive Services Task Force. Screening for peripheral artery disease and cardiovascular disease risk assessment with the ankle-brachial index in adults: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2013; 159: 342–348.

45.

Greenland

Alpert

Beller

. American College of Cardiology Foundation; American Heart Association. 2010 ACCF/AHA guideline for assessment of cardiovascular risk in asymptomatic adults: a report of the American College of Cardiology Foundation/American Heart Association task force on practices guidelines. J Am Coll Cardiol 2010; 56: e50–e103.

46.

Fifth Joint Task Force of the European Society of Cardiology, European Association of Echocardiography, European Association of Percutaneous Cardiovascular Interventions, et al. European guidelines on cardiovascular disease prevention in clinical practice (version 2012). Eur J Prev Cardiol 2012; 19: 585–667.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.19 MB