Ankle brachial index for the diagnosis of lower limb peripheral arterial disease
Abstract
Background
Peripheral arterial disease (PAD) of the lower limb is common, with prevalence of both symptomatic and asymptomatic disease estimated at 13% in the over 50 age group. Symptomatic PAD affects about 5% of individuals in Western populations between the ages of 55 and 74 years. The most common initial symptom of PAD is muscle pain on exercise that is relieved by rest and is attributed to reduced lower limb blood flow due to atherosclerotic disease (intermittent claudication). The ankle brachial index (ABI) is widely used by a variety of healthcare professionals, including specialist nurses, physicians, surgeons and podiatrists working in primary and secondary care settings, to assess signs and symptoms of PAD. As the ABI test is non‐invasive and inexpensive and is in widespread clinical use, a systematic review of its diagnostic accuracy in people presenting with leg pain suggestive of PAD is highly relevant to routine clinical practice.
Objectives
To estimate the diagnostic accuracy of the ankle brachial index (ABI) ‐ also known as the ankle brachial pressure index (ABPI) ‐ for the diagnosis of peripheral arterial disease in people who experience leg pain on walking that is alleviated by rest.
Search methods
We carried out searches of the following databases in August 2013: MEDLINE (Ovid SP),Embase (Ovid SP), the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO), Latin American and Caribbean Health Sciences (LILACS) (Bireme), Database of Abstracts of Reviews of Effects and the Health Technology Assessment Database in The Cochrane Library, the Institute for Scientific Information (ISI) Conference Proceedings Citation Index ‐ Science, the British Library Zetoc Conference search and Medion.
Selection criteria
We included cross‐sectional studies of ABI in which duplex ultrasonography or angiography was used as the reference standard. We also included cross‐sectional or diagnostic test accuracy (DTA) cohort studies consisting of both prospective and retrospective studies.
Participants were adults presenting with leg pain on walking that was relieved by rest, who were tested in primary care settings or secondary care settings (hospital outpatients only) and who did not have signs or symptoms of critical limb ischaemia (rest pain, ischaemic ulcers or gangrene).
The index test was ABI, also called the ankle brachial pressure index (ABPI) or the Ankle Arm Index (AAI), which was performed with a hand‐held doppler or oscillometry device to detect ankle vessels. We included data collected via sphygmomanometers (both manual and aneroid) and digital equipment.
Data collection and analysis
Two review authors independently replicated data extraction by using a standard form, which included an assessment of study quality, and resolved disagreements by discussion. Two review authors extracted participant‐level data when available to populate 2×2 contingency tables (true positives, true negatives, false positives and false negatives).
After a pilot phase involving two review authors working independently, we used the methodological quality assessment tool the Quality Assessment of Diagnostic Accuracy Studies‐2 (QUADAS‐2), which incorporated our review question ‐ along with a flow diagram to aid reviewers' understanding of the conduct of the study when necessary and an assessment of risk of bias and applicability judgements.
Main results
We screened 17,055 records identified through searches of databases. We obtained 746 full‐text articles and assessed them for relevance. We scrutinised 49 studies to establish their eligibility for inclusion in the review and excluded 48, primarily because participants were not patients presenting solely with exertional leg pain, investigators used no reference standard or investigators used neither angiography nor duplex ultrasonography as the reference standard. We excluded most studies for more than one reason.
Only one study met the eligibility criteria and provided limb‐level accuracy data from just 85 participants (158 legs). This prospective study compared the manual doppler method of obtaining an ABI (performed by untrained personnel) with the automated oscillometric method. Limb‐level data, as reported by the study, indicated that the accuracy of the ABI in detecting significant arterial disease on angiography is superior when stenosis is present in the femoropopliteal vessels, with sensitivity of 97% (95% confidence interval (CI) 93% to 99%) and specificity of 89% (95% CI 67% to 95%) for oscillometric ABI, and sensitivity of 95% (95% CI 89% to 97%) and specificity of 56% (95% CI 33% to 70%) for doppler ABI. The ABI threshold was not reported. Investigators attributed the lower specificity for doppler to the fact that a tibial or dorsalis pedis pulse could not be detected by doppler in 12 of 27 legs with normal vessels or non‐significant lesions. The superiority of the oscillometric (automated) method for obtaining an ABI reading over the manual method with a doppler probe used by inexperienced operators may be a clinically important finding.
Authors' conclusions
Evidence about the accuracy of the ankle brachial index for the diagnosis of PAD in people with leg pain on exercise that is alleviated by rest is sparse. The single study included in our review provided only limb‐level data from a few participants. Well‐designed cross‐sectional studies are required to evaluate the accuracy of ABI in patients presenting with early symptoms of peripheral arterial disease in all healthcare settings. Another systematic review of existing studies assessing the use of ABI in alternative patient groups, including asymptomatic, high‐risk patients, is required.
Plain language summary
Ankle brachial index for the diagnosis of lower limb peripheral arterial disease
Peripheral arterial disease (PAD) of the legs affects 13% of people over 50 years of age. Sometimes PAD is "silent" and people are unaware they have it, but PAD can cause pain in the legs, especially with walking, and this type of symptomatic PAD affects about 5% of people in the Western world between the ages of 55 and 74 years. In PAD, fatty deposits (atherosclerosis) and blood clots cause the arteries to narrow and block. This leads to poor blood flow to the muscles during exercise, causing the classical symptom of muscle pain during walking that goes away after rest (intermittent claudication). In severe cases of PAD, symptoms of rest pain, ulceration and gangrene may develop and, if untreated, can lead to lower limb amputation. People with PAD are also at higher risk for cardiovascular disease and stroke.
The ankle brachial index (ABI) is a test that is used to facilitate diagnosis of PAD. This test uses a device for measuring blood pressure with an inflatable cuff, and blood pressure measurements are taken at the upper arm and the ankle. The equipment can be manual or digital with automatic electronic calculation of blood pressure. The ABI is widely used for assessment of PAD by specialist nurses, physicians, surgeons and podiatrists working in hospitals. Dividing blood pressure recorded at the ankle by that recorded at the arm produces a ratio. Ratios of 0.90 to 1.30 are considered normal for adults, and ratios less than 0.8 indicate that PAD is present. Lower readings (< 0.7) suggest that the disease is severe and people might develop ulcers and gangrene. People with mild to moderate PAD can arrive at a diagnosis by several routes when using the ABI: during routine diabetic foot checks in general practice, in community health clinic or hospital settings, as a screening test for PAD in people who have no symptoms and during assessment of people presenting with exertional leg pain suggestive of PAD. Once a diagnosis of PAD is established, treatment will include prescribed secondary prevention therapy and lifestyle advice (exercise, smoking cessation, diet, weight), and for those with impaired quality of life, treatment may include supervised exercise therapy, or revascularisation, which commonly involves endovascular treatment rather than surgery.
In hospitals, other tests may be used to diagnose PAD. Duplex ultrasound (DUS) shows blood flow in the arteries and is non‐invasive, but only an experienced radiologist can achieve useful images. Hospital staff can use other tests to image the blood vessels, namely, computerised tomography angiography (CTA), magnetic resonance angiography (MRA) and catheter angiography.
The ABI test is non‐invasive and inexpensive and is widely used clinically; therefore, we have reviewed all available reports obtained from a wide search of databases of medical literature to estimate its accuracy in identifying PAD in people who experience pain on walking that goes away after rest. Two review authors independently assessed studies that met inclusion criteria of the review, including use of a cross‐sectional study design; enrolment of participants with pain on walking that got better with rest; and use of duplex ultrasonography or angiography to check that results of the ABI test were accurate. One study met our criteria and provided data from 85 participants (158 limbs). Investigators compared the manual doppler method of measuring ABI with the automated method. Researchers provided only data for legs as opposed to data for patients; we were therefore unable to recalculate the analysis at the whole‐participant level.
In conclusion, we found little evidence about the accuracy of the ankle brachial index for diagnosing PAD in people presenting with exertional leg pain. The study included in our review had some flaws, and well‐designed cross‐sectional studies are needed to measure the accuracy of the ABI for diagnosing PAD in patients with early symptoms.
Authors' conclusions
Summary of findings
| Accuracy of the ankle brachial index (ABI) in diagnosing symptomatic peripheral arterial disease (PAD) | ||||
| Population: | People with intermittent claudication | |||
| Setting | Primary and secondary care settings (hospital outpatients) | |||
| Index test | Ankle brachial index | |||
| Importance | The success of management strategies for PAD depends upon the quality of the diagnostic process, which involves careful assessment of the underlying pathology with diagnostic tests that possess a high level of accuracy, to allow detection and measurement of an arterial stenosis and its distribution in the blood vessels. | |||
| Reference standard | Duplex ultrasonography or angiography | |||
| Studies | Cross‐sectional or diagnostic cohort study | |||
| Test/subgroup | Sensitivity | Specificity | No. of participants (studies) | Quality (QUADAS‐2)a and comments |
| Cut‐off ABI ratio positivity Mild PAD: 0.7 to 0.9 Moderate PAD: 0.41 to 0.69 | ||||
| Automated ABI:
Manual ABI: | 97% (95% CI 93% to 99%) 95% (95% CI 89% to 97%) | 89% (95% CI 67% to 95%) 56% (95% CI 33% to 70%) | 85 (n = 158 legs) (1 study) | Unclear risk of bias: Vega 2011 may have included patients with severe PAD (stenosis > 50%); the threshold was not reported; time between conduct of the index test and use of the reference standard is not reported. One study, no pooled analysis, sensitivity and specificity data for limb level ‐ not for participant level, as reported by study authors |
| aQUADAS‐2 is a tool used for assessment of the quality of diagnostic accuracy studies. This tool comprises four domains: patient selection, index test, reference standard and flow and timing. Each domain is assessed in terms of risk of bias; the first three domains are also assessed in terms of concerns regarding applicability. ABI: ankle brachial index. | ||||
Background
Peripheral arterial disease (PAD) of the lower limbs is common, with prevalence of both symptomatic and asymptomatic disease estimated at 13% in the over 50 age group (Hirsch 2001). Symptomatic PAD affects about 5% of individuals in Western populations between the ages of 55 and 74 years (Khan 2007). The most common initial symptom of PAD is muscle pain on exercise that is relieved by rest and is attributed to reduced lower limb blood flow due to atherosclerotic disease (intermittent claudication; IC). Patients with more severe PAD may develop rest pain, ulceration and gangrene (critical limb ischaemia; CLI), which, if untreated, can lead to lower limb amputation (Hooi 2007; Twine 2009). The presence of PAD has been shown to be a marker of underlying cardiovascular disease.
A simple, non‐invasive test known as the ankle brachial index (ABI) ‐ or the ankle brachial pressure index (ABPI) ‐ can detect PAD. Healthcare providers can use the ABI to screen asymptomatic patients at increased risk of developing PAD, for example, people with diabetes, and to assess people presenting with leg pain suggestive of PAD. Clinicians can use a low ABI, even in the absence of symptoms, to identify people who are at increased risk of cardiac and cerebrovascular disease (Ankle Brachial Index Collaboration 2008; SIGN 2007).
Dividing the highest ankle pressure (obtained in the posterior tibial, dorsalis pedis and, when required, peroneal arteries) by the highest systolic arm pressure yields the ABI ratio. Classically, healthcare providers have used a doppler probe to detect signals within the arteries, but recently designed oscillometric and photophlethysmographic devices are now available. Current guidelines do not endorse the use of these newer devices but recommend the hand‐held doppler technique (NICE 2012). Ratios of 0.90 to 1.30 are normal for adults, ratios less than 0.9 are indicative of arterial stenosis and ratios less than 0.5 are associated with CLI (Bhasin 2007; MacLeod‐Roberts 1995; NICE 2012). Individuals with aorto‐iliac disease may have normal ABI at rest and low values after exercise. An 'exercise‐ABI' test can detect this and can be performed during secondary care.
A wide variety of healthcare professionals, including specialist nurses, physicians, surgeons and podiatrists working in primary and secondary care settings, frequently use the ABI to assess PAD. These providers normally check foot and leg pulses of people presenting with leg pain suggestive of PAD before performing an ABI to determine their presence or absence. Once PAD is diagnosed, first‐line management of the condition consists of cardiac risk factor management, which includes lifestyle advice, smoking cessation, statin and antiplatelet therapy, blood pressure control and screening for and treatment of diabetes (Bhasin 2007; Heald 2006). Supervised exercise programmes can lead to symptomatic improvement, and healthcare providers can perform arterial revascularisation, in the form of angioplasty or less commonly surgery, to treat those with incapacitating disease and significantly impaired quality of life (Cassar 2003; Chang 2011; de Backer 2012; Fokkenrood 2013; Lane 2014; NICE 2012; Rutherford 1997). Physicians may prescribe naftidrofuryl oxalate for patients in whom supervised exercise therapy has not been found to be effective and who do not wish to be referred for revascularisation (NICE 2012).
In secondary care, hospital staff may use a variety of non‐invasive imaging tests for patients with suspected PAD in whom revascularisation may be considered, including non‐invasive duplex ultrasonography, computerised tomography angiography (CTA) or magnetic resonance angiography (MRA). The National Institute for Health and Care Excellence suggests duplex ultrasonography as the first‐line approach for imaging PAD, and CTA or contrast‐enhanced MRA for those who need further imaging (NICE 2012).
The ABI test is non‐invasive and inexpensive and is in widespread clinical use; a systematic review of its diagnostic accuracy in people presenting with leg pain suggestive of PAD is highly relevant to routine clinical practice.
Target condition being diagnosed
Presence or absence of peripheral arterial disease of the lower limb.
Index test(s)
Healthcare providers use the ankle brachial index (ABI) to diagnose peripheral arterial disease (PAD), by dividing highest systolic pressure measured in the arteries at the ankle (dorsalis pedis and posterior tibial arteries, or peroneal if the others are non‐detectable) by highest systolic blood pressure at the arm (brachial artery).
Physicians can calculate an ankle brachial ratio in several ways. UK clinical guidelines recommend that the patient is rested in a supine position and that blood pressure is taken by using a sphygmomanometer with an appropriately sized cuff at the brachial artery and the posterior tibial, dorsalis pedis and, when possible, peroneal arteries. A doppler probe detects audible systolic pressure (Aboyans 2012; McDermott 2000; NICE 2012).
For each leg, the healthcare professional calculates the ABI by dividing the highest ankle pressure by the highest pressure reading taken from the arm (McDermott 2000). MacLeod‐Roberts 1995 presents a classification of ABI values.
In this review, we use the threshold of less than 0.90 to distinguish between positive (< 0.90) and negative (≥ 0.90) test results. Clinicians commonly use this threshold in clinical practice, and is cited in current guidelines (NICE 2012).
The position of the patient at the time blood pressure is taken is important: For each inch that the ankle is positioned below the heart, care providers have noted a 1 mmHg increase in systolic ankle blood pressure (MacLeod‐Roberts 1995).
False negatives commonly occur in people who have calcification of the ankle artery wall, which creates incompressibility and an artificially high reading. This may occur in some patients with diabetes (Bhasin 2007; MacLeod‐Roberts 1995).
Several automated blood pressure machines are available, and all are eligible for inclusion in the review.
Clinical pathway
Healthcare providers may follow several clinical pathways to diagnose mild to moderate PAD by using the ABI: They may measure the ABI in primary care to diagnose PAD in members of the general population who report symptoms of exertional leg pain. They sometimes use ABI in addition to routine diabetic foot checks in primary care, community health settings or hospital settings as a screening test for PAD in people who have no symptoms but are at high risk. Once a diagnosis of PAD is established, healthcare staff will prescribe secondary prevention therapy and will give lifestyle advice (exercise, smoking cessation, diet, weight); for those who have impaired quality of life, they may offer supervised exercise therapy, or revascularisation, which commonly involves endovascular treatment rather than surgery.
Role of index test(s)
Practitioners use the ABI test in healthcare settings to identify PAD in people who have suggestive symptoms, and can use this test to screen those at increased risk for PAD. An ABI < 0.90 is predictive of increased risk of cardiovascular disease (Ankle Brachial Index Collaboration 2008). This review aimed to include studies evaluating the diagnostic test accuracy of the ABI used in primary and secondary (outpatient only) care settings by a range of healthcare professionals, to evaluate people presenting with leg pain on exercise that is relieved by rest, which is suggestive of underlying PAD.
Alternative test(s)
Uses of the ABI in clinical practice are diverse, and care providers do not need to consider standard alternative tests.
Rationale
The success of management strategies for PAD depends upon the quality of the diagnostic process, which involves careful assessment of underlying pathology through diagnostic tests that possess a high level of accuracy.
Objectives
To estimate the diagnostic accuracy of the ankle brachial index (ABI) ‐ also known as the ankle brachial pressure index (ABPI) ‐ for the diagnosis of peripheral arterial disease in people who experience leg pain on walking that is alleviated by rest.
Secondary objectives
We also intended to investigate the effect of sources of heterogeneity on diagnostic accuracy, specifically, study setting, previous tests, types of equipment used, types of reference standards applied, different groups of patients examined (people with type 1 or type 2 diabetes and suspected aorto‐iliac disease) and duration of symptoms, by including them as co‐variates in the meta‐analysis, if sufficient studies provided relevant data. It was our intention that we would examine graphically other potential sources of heterogeneity for signs that they were a cause of heterogeneity.
Methods
Criteria for considering studies for this review
Types of studies
We included studies of ABI that used duplex ultrasonography or angiography as the reference standard. We included cross‐sectional or diagnostic test accuracy (DTA) cohort studies examining both prospective and retrospective studies. These studies had to report that all participants received a reference standard; investigators had to present cross‐tabulated results of the index test and the reference standard (2×2 table), or had to report sufficient information to allow the 2×2 table data to be back‐calculated.
Participants
Adults with leg pain on walking relieved by rest, who are tested in primary care settings or secondary care settings (hospital outpatients only) and do not have signs or symptoms of critical limb ischaemia (rest pain, ischaemic ulcers or gangrene). We excluded from the review patients who were free of exertional leg pain, as well as those with CLI.
Index tests
Ankle brachial index (ABI), also called ankle brachial pressure index (ABPI). We included data collected by sphygmomanometers (both manual and aneroid) as well as by digital equipment that used manual or automatic inflation. We included studies that used hand‐held doppler or oscillometry to detect ankle vessels.
Target conditions
Peripheral arterial disease of the lower limbs.
Reference standards
We included studies that used duplex ultrasonography or angiography as the reference standard test, and we noted instances in which different reference standards were used to verify the presence or absence of disease in the same study population.
Search methods for identification of studies
We applied no restrictions in terms of date, language of publication or publication status of studies. We used no diagnostic method search filters.
Electronic searches
We applied no restrictions in terms of language of publication or publication status.
We searched the following databases.
-
MEDLINE Ovid (1946 to July week 5 2013).
-
Embase (Ovid SP) (1980 to 2013 week 32).
-
Cumulative Index to Nursing and Allied Health Literature (CINAHL) via EBSCO (12 August 2013).
-
Latin American and Caribbean Health Sciences (LILACS) (Bireme) (13 August 2013).
-
Database of Abstracts of Reviews of Effects (DARE) and the Health Technology Assessment Database (HTA), in The Cochrane Library (2013, Issue 7).
-
Institute for Scientific Information (ISI) Conference Proceedings Citation Index ‐ Science (14 August 2013).
-
British Library Zetoc Conference search (29 August 2013).
We used the search strategies shown in Appendix 1; Appendix 2; Appendix 3; Appendix 4; Appendix 5; Appendix 6 and Appendix 7.
We also searched MEDION (www.mediondatabase.nl/) using the 'Systematic Reviews of Diagnostic Studies' search filter (29 August 2013) (Appendix 8).
Searching other resources
We reviewed the bibliographies of review articles identified by searches for potentially relevant studies.
Data collection and analysis
Selection of studies
One review author (FC) screened titles and abstracts retrieved by the electronic searches, and a second review author (AA) checked a random sample of 10% of the studies. We obtained full papers for potentially eligible studies, including those identified by non‐electronic means. Two review authors (FC, AA) independently applied exclusion criteria to the full papers and resolved disagreements by discussion. We used a flow diagram to show results of the decision‐making process.
Data extraction and management
Two review authors (FC, AA) independently used a standard form to replicate data extraction; this form included an assessment of study quality. Review authors corroborated their data extraction and quality assessment decisions and resolved disagreements by discussion. Review authors intended to extract participant‐level data to populate 2×2 contingency tables ‐ true positives (TPs), true negatives (TNs), false positives (FPs) and false negatives (FNs)) ‐ as reported. We also extracted details of test threshold(s) used for interpretation of results.
We collected data on mortality, adverse events, the nature of the equipment used (manual or automated) and the number of technical failures.
Assessment of methodological quality
After a pilot phase involving two review authors working independently, we used the Quality Assessment of Diagnostic Accuracy Studies‐2 (QUADAS‐2) (Whiting 2011), which incorporated our review question, a flow diagram to aid reviewers' understanding of the conduct of the study when necessary and an assessment of risk of bias and applicability judgements. We presented review‐specific signalling questions and appropriate items concerning the applicability of primary studies relative to the review, together with guidance about ratings, in Appendix 9. We resolved disagreements by discussion.
Statistical analysis and data synthesis
We intended to use 2×2 contingency tables populated with participant‐level data, rather than data on limbs, to estimate sensitivity and specificity for each study. When data were adequate, we intended to perform a bivariate random‐effects meta‐analysis of sensitivity and specificity. We anticipated that we would use these estimates to create receiver operating characteristic (ROC) and forest plots. As an ABI less than 0.90 is the accepted threshold used in clinical practice, we had planned to restrict the meta‐analysis to studies that used this threshold, so that our estimates of sensitivity and specificity would be derived directly from that threshold. We intended to add items investigated for heterogeneity as co‐variates to the bivariate model.
However, available data are based on limbs as the unit of analysis. If two datapoints were obtained from the same participant (one from each leg), these datapoints will tend to be more similar to each other than datapoints from different patients, thus changing the variance in data. However, estimating within‐study variance is a key part of meta‐analysis, and current methods do not allow for studies in which a participant may contribute data from more than one potential disease site. If studies do not provide participant‐level data, we have no correct way to estimate within‐study variance, and so meta‐analysis, whether bivariate or based on hierarchical summary receiver operating characteristic (HSROC) models (Harbord 2007), or univariate meta‐analysis for sensitivity and specificity, is not an option.
We intended to perform all analyses in R 7.1 (cran.r‐project.org) and SAS 9.3 (www.sas.com).
Investigations of heterogeneity
Our planned investigations into the effect of sources of heterogeneity on diagnostic accuracy focussed on patient groups (e.g. type 1 and type 2 diabetes, suspected aorto‐iliac disease), duration of symptoms, previous tests and types of equipment (automatic or manual) by including them as co‐variates in the meta‐analyses. We intended to examine other potential sources of heterogeneity graphically for signs that they were the source of heterogeneity. We planned to group estimates in plots according to items considered potential sources of heterogeneity, as detailed above, and to present these as forest and ROC plots for visual assessment of heterogeneity.
If we found sufficient studies, we planned to investigate heterogeneity by adding items as co‐variates to the meta‐analysis model, from a bivariate or univariate analysis, depending on results of the main analysis. However, we recognised the likelihood of having too few studies to perform meta‐regression with all items listed as potential sources of heterogeneity, and under these circumstances, we planned to limit ourselves to visual inspection of ROC and forest plots. We understand that some items are investigated better with individual participant data, as they are patient‐specific, rather than study‐specific, for example, duration of symptoms, and we planned to interpret any aggregate results cautiously.
Sensitivity analyses
We intended to conduct several sensitivity analyses to compare the diagnostic accuracy of ABI in those with and without diabetes, in those with and without coronary heart disease, in smokers versus non smokers and when manual versus automated methods are used to measure the ABI.
Assessment of reporting bias
Methods for dealing with publication bias in reviews of diagnostic accuracy studies are relatively underdeveloped. Consequently, we interpreted our results cautiously, and with awareness of the likelihood of publication bias, rather than by using funnel plots, which can be challenging to interpret in this context. We planned to consider using a funnel plot of the log of the diagnostic odds ratio (lnDOR), provided we found low heterogeneity in the lnDOR (Deeks 2005).
Results
Results of the search
See Figure 1.
Study flow diagram.
We screened 17,055 records identified through searches of databases. We obtained and assessed for relevance 746 full‐text articles. A second review author (AA) checked a 10% random sample of these articles and reached 100% agreement with the first review author (FC). We scrutinised 49 studies to establish their eligibility for inclusion in the review.
We included only one study (Vega 2011) with a total of 85 participants and have described it in the Characteristics of included studies table.
We have listed the 48 studies excluded from the review, along with reasons for exclusion, in the Characteristics of excluded studies table. We excluded studies primarily because participants were not patients presenting solely with exertional leg pain, investigators used no reference standard or the reference standard used was neither angiography nor duplex ultrasonography. We have provided more than one reason for exclusion of most studies.
Methodological quality of included studies
Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.
Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.
The Characteristics of included studies table incorporates the methodological quality assessment. In Vega 2011, the risk of bias was generally 'low', but some items had 'unclear' risk of bias; we explain these below. QUADAS‐2 quality assessment items are grouped into four domains: patient selection, index test, reference standard and flow and timing.
The risk of bias arising from patient selection was unclear (Vega 2011). Although a consecutive sample of patients was reported, inappropriate exclusions may not have been avoided. Investigators included patients who had 'suspected advanced PAD'; consequently, the patient population may have been affected by disease spectrum bias.
Investigators described the execution of ABI tests well but did not state the ABI threshold used (Vega 2011); this led to a classification of "unclear risk of bias".
Vega 2011 did not report the time between ABI and angiography assessments, which might have led to misclassification due to progression of the disease to a more advanced state (disease progression bias).
We have presented in the Characteristics of included studies table details of the execution of index and reference standard tests used by Vega 2011.
Findings
The one included study (Vega 2011) did not report accuracy data at the participant level; therefore, we were unable to calculate estimates of sensitivity or specificity for individual participants. The accuracy estimates reported in the narrative synthesis below are those calculated and reported by Vega 2011. See also summary of findings Table.
This prospective study compared the manual doppler method of obtaining an ABI with the automated oscillometric method (Vega 2011). In total, researchers recruited into the study 85 patients (76 men and nine women) with a mean (standard deviation) age of 68 (11) years, who were referred for angiography with 'symptoms of intermittent claudication'. Study participants had several co‐morbidities, including diabetes (52%), hypertension (76%), hypercholesterolaemia (43.5%), ischaemic heart disease (30%), percutaneous coronary intervention (12%), coronary surgery (6%), previous stroke (22%), carotid revascularisation (2.4%) and aortic aneurysm (6%), and included current smokers (32%) and previous smokers (46%).
Doctors with no specialist training performed ABI measurements, and researchers conducted the study in a hospital catheterisation laboratory in Spain. Investigators performed manual doppler ABI by using an 8 MHz doppler probe (Dopplex II MD2/SD, ArjoHuntleigh Inc., Addison, Illinois) model and a sphygmomanometer with cuffs of appropriate size. They obtained automated ABI measurements by using automated oscillometric equipment: Omron M4‐1 (Omron Healthcare Europe BV, Hoofddorp, The Netherlands). The threshold for a positive test result was a significant lesion of > 50% occlusion detected by catheter angiography (the reference standard). Researchers defined non‐significant PAD as < 50% obstruction.
According to Vega 2011, the reported accuracy of automated oscillometric ABI was not statistically significantly different from that of the manual doppler method, with reported sensitivity of 97% (95% confidence interval (CI) 93% to 99%) and specificity of 89% (95% CI 67% to 95%) for oscillometric ABI, compared with sensitivity of 95% (95% CI 89% to 97%) and specificity of 56% (95% CI 33% to 70%) for the manual doppler ABI. The reason for the lower specificity for the doppler was that the doppler could not detect a tibial or dorsalis pedis pulse in 12 legs with normal vessels or non‐significant lesions, among a total of 27 legs. However, with the automated method, investigators could not measure blood pressure in 70 legs, 69 of which were found to have severe angiographic lesions. The superiority of the automated oscillometric method for obtaining an ABI reading over the manual method in which inexperienced operators used a doppler probe may be a clinically important finding.
Researchers reported accuracy estimates with 'limbs' as the unit of analysis, and as participant‐level data are not available, we were unable to reproduce accuracy estimates. The number of significant lesions (occlusions > 50%) detected by angiography in this study population was 131 (83%).
Discussion
The ankle brachial index (ABI) test is cheap and non‐invasive, which makes it potentially valuable in health care. Unfortunately, evidence for the accuracy of the ABI test for the detection of peripheral arterial disease (PAD) in people presenting with leg pain on exercise that is alleviated by rest is sparse. We included in this review only one study, which evaluated automated versus manual ABI equipment, and provided limb‐level data from a total of 85 participants.
The main findings of the review are the following: (1) Although both ABI tests demonstrated high levels of sensitivity, the results came from a single small study in which participants with critical limb ischaemia (CLI) may have been included, and researchers reported no threshold for the ABI; (2) investigators reported no statistically significant differences in accuracy between automated and manual ABI equipment in this small group of participants, although automated ABI was associated with a greater number of technical failures than the hand‐held doppler, and these technical failures occurred in participants with severe angiographic lesions; and (3) in light of recruitment of patients from those already referred for angiography, it seems likely that participants included in the review may have had worse PAD than those recruited directly from a primary healthcare setting, and we would not extrapolate these findings to all patients presenting in primary care.
More than half of the participants in this study had received a diagnosis of diabetes mellitus, but we were unable to produce evidence to support the advice given in clinical guidelines (NICE 2012) that the use of ABI in assessment of PAD among people with diabetes is less reliable than among those who do not have diabetes mellitus, because data were not available for such an analysis. The single included study (Vega 2011) excluded patients with an ABI > 1.4, but investigators excluded only two patients for this reason and did not reveal whether these patients had diabetes.
The review excluded 48 studies, most of which were cross‐sectional studies evaluating the accuracy of ABI or comparing different ABI techniques in the diagnosis of PAD. Unfortunately, these studies usually included patients other than those presenting with exertional leg pain, and many did not use the reference standard of duplex ultrasonography or angiography. These shortcomings are important findings of this review, and in the recommendations for research section below, we make specific suggestions to inform the design of future DTA studies of ABI for the diagnosis of PAD in people with exertional leg pain.
Summary of main results
This review found a very small amount of evidence indicating that the ABI test is accurate in the diagnosis of symptomatic PAD among people with intermittent claudication (IC). The one included study suggests that automated equipment may be more accurate than manual methods when used by individuals with no specialist training. The accuracy of manual doppler varies with operator skill, and so trained individuals may obtain more accurate results with this method. The small number of participants who took part in the study led to our cautious interpretation of the data.
Strengths and weaknesses of the review
We identified only one study for inclusion. We restricted the inclusion criteria for this review to patients with leg pain on walking relieved by rest and use of duplex ultrasonography or angiography as the reference standard; this contributed to the exclusion of a large number of studies.
Vega 2011 included some patients with 'suspected advanced PAD' and did not present data for these participants separately. This may have led to higher estimates of accuracy than would be observed in a population that was strictly recruited on the basis of leg pain alone. In addition, researchers did not report the ABI threshold.
Vega 2011 reported accuracy data at limb level only; therefore we were unable to calculate estimates of sensitivity and specificity for individual participants. We attempted to contact the study authors to obtain participant‐level data, but we received no response.
Applicability of findings to the review question
The patient population recruited to the included study suggests that the findings may not answer the review question. Investigators recruited the study population from patients referred for angiography for peripheral arterial intermittent claudication or suspected advanced PAD (Vega 2011). The percentage of people with symptoms of IC and the percentage with more advanced PAD remain unclear, but it is likely that researchers included in this population people with critical limb ischaemia. Use of angiography as the reference standard by which to verify results of the ABI (index) test may mean that the spectrum of disease is worse than in the general population seeking a diagnosis for PAD, as angiography is an invasive test conducted in hospital vascular departments. Authors of the included study themselves cautioned that their findings should not be extrapolated to the general population because of the high prevalence of PAD reported among study participants (Vega 2011).
Authors of another systematic review evaluating the accuracy of ABI for PAD suggest that accuracy is dependent on the purpose of the examination; they found ABI to be highly accurate when used to detect serious stenosis (> 50%) (Dachun 2013). The American Heart Association (AHA) Scientific Statement on measurement and interpretation of the ABI reports that areas under the receiver operating characteristic (ROC) curve are higher for ABI measured by doppler than for ABI measured by oscillometric methods (Aboyans 2012). This narrative review provides evidence on the overall diagnostic ability of the ABI in a variety of populations and settings based on four studies that did not meet the eligibility criteria for the current review (Aboyans 2012).
Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.
Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.
| Accuracy of the ankle brachial index (ABI) in diagnosing symptomatic peripheral arterial disease (PAD) | ||||
| Population: | People with intermittent claudication | |||
| Setting | Primary and secondary care settings (hospital outpatients) | |||
| Index test | Ankle brachial index | |||
| Importance | The success of management strategies for PAD depends upon the quality of the diagnostic process, which involves careful assessment of the underlying pathology with diagnostic tests that possess a high level of accuracy, to allow detection and measurement of an arterial stenosis and its distribution in the blood vessels. | |||
| Reference standard | Duplex ultrasonography or angiography | |||
| Studies | Cross‐sectional or diagnostic cohort study | |||
| Test/subgroup | Sensitivity | Specificity | No. of participants (studies) | Quality (QUADAS‐2)a and comments |
| Cut‐off ABI ratio positivity Mild PAD: 0.7 to 0.9 Moderate PAD: 0.41 to 0.69 | ||||
| Automated ABI:
Manual ABI: | 97% (95% CI 93% to 99%) 95% (95% CI 89% to 97%) | 89% (95% CI 67% to 95%) 56% (95% CI 33% to 70%) | 85 (n = 158 legs) (1 study) | Unclear risk of bias: Vega 2011 may have included patients with severe PAD (stenosis > 50%); the threshold was not reported; time between conduct of the index test and use of the reference standard is not reported. One study, no pooled analysis, sensitivity and specificity data for limb level ‐ not for participant level, as reported by study authors |
| aQUADAS‐2 is a tool used for assessment of the quality of diagnostic accuracy studies. This tool comprises four domains: patient selection, index test, reference standard and flow and timing. Each domain is assessed in terms of risk of bias; the first three domains are also assessed in terms of concerns regarding applicability. ABI: ankle brachial index. | ||||
