Assessment of the diagnostic accuracy of automated ankle brachial pressure index devices in patients with diagnosed or suspected peripheral arterial disease: protocol for a systematic review and meta-analysis
Walshaw J,1 Lathan R,1 Huang C,2 Staniland T,3 Chetter I,1 Pymer S1
Plain English Summary
Why we are undertaking this work: Peripheral arterial disease is a common condition where narrowing of the blood vessels in the legs can reduce blood flow. This may cause symptoms that include calf, thigh and/or buttock pain when walking and this can progress to cause pain at rest and leg ulcers. The ankle brachial pressure index is a measurement that can be used to assess peripheral arterial disease. Currently, this is measured manually using a blood pressure cuff and ultrasound probe, though time constraints and staff training limit its widespread use. This causes difﬁculty in the assessment and diagnosis of peripheral arterial disease, particularly in primary care. There are automated ankle brachial pressure index devices available, which may alleviate some of this difﬁculty. However, there is limited evidence regarding their accuracy in diagnosing peripheral arterial disease.
What we aim to do: We plan to review the current evidence available for the accuracy of automated ankle brachial pressure index devices in people with known or suspected peripheral arterial disease. We will look at studies that have compared automated devices with the current methods used for diagnosing peripheral arterial disease including manual doppler ankle brachial pressure index measurements and vascular imaging. What this means: We hope the results from this review will be used to inform clinical practice and guide future clinical trials.
Review registration: Prospero ID CRD42022343920
Background: The ankle brachial pressure index (ABPI) is a common diagnostic tool used in the assessment of peripheral arterial disease (PAD). The Doppler ultrasound technique is regarded as the gold-standard method for ABPI measurement; however, time constraints and operator experience limit widespread application in clinical practice, particularly in a primary care setting. Automated ABPI devices are not currently widely used due to a lack of evidence regarding their diagnostic accuracy. The aim of this proposed systematic review and meta-analysis is to explore the current evidence for the accuracy of automated ABPI devices in people with known or suspected PAD.
Methods: Systematic searches of electronic databases and grey literature will be performed. We plan to include studies of adult patients with diagnosed or suspected PAD that have compared automated ABPI device readings with manual Doppler ABPI measurements or conﬁrmed the diagnosis of PAD using vascular imaging. Two independent reviewers will screen identiﬁed literature for inclusion and perform data extraction. Extracted data will include study and participant characteristics, a description of the index and reference tests, outcome measures and main ﬁndings. The methodological quality of selected studies will be assessed using QUADAS-2 and QUADAS-C. Meta-analysis will be performed for studies with paired designs using a bivariate random-effect model to provide pooled estimates of summary accuracy statistics. We intend to conduct subgroup analyses and meta-regression for suspected sources of heterogeneity.
Discussion: This review aims to assess the diagnostic accuracy of automated ABPI devices for detecting PAD in patients with known or suspected PAD compared with manual Doppler ABPI measurements or vascular imaging. These results will be used to inform clinical practice and guide future trials.
Target condition being diagnosed
Peripheral arterial disease (PAD) is a prevalent cardiovascular disease, estimated to affect approximately 236 million people worldwide.1 PAD is characterised by progressive narrowing of the arterial lumen, reducing blood flow to the distal extremities.2 Classic symptoms include exertional calf, thigh and/or buttock pain known as intermittent claudication, and with disease progression patients may develop ischaemic rest pain, arterial ulceration and limb loss.3 The presence of PAD is also associated with an increased risk of myocardial infarction, ischaemic stroke, and death.4,5 However, more than 50% of patients with PAD are asymptomatic and are therefore commonly underdiagnosed and undertreated.6 Detection of symptomatic or asymptomatic PAD is crucial to allow for the appropriate management to reduce disease progression and associated cardiovascular morbidities.
Index test and alternative tests
The ankle brachial pressure index (ABPI) is a non-invasive diagnostic tool widely used in the assessment of PAD and is a vital part of the clinical pathway. ABPI values of <0.9 are regarded as diagnostic for PAD, with lower values indicating increasing severity.7,8 The manual Doppler ultrasound technique is considered the gold-standard method for ABPI measurements.9 This technique uses a sphygmomanometer and Doppler ultrasound probe for accurate arterial flow readings in the brachial arteries of both arms, and usually the posterior tibial and dorsalis pedis arteries of both legs. The index is calculated for each leg by dividing the highest of the ankle pressures by the highest arm pressure.10
Imaging modalities can be used in the assessment of PAD, particularly when revascularisation procedures are being considered. These include duplex ultrasonography, contrast-enhanced magnetic resonance angiography (MRA) and computed tomography angiography (CTA). Duplex ultrasonography is the first-line imaging technique for patients being considered for revascularisation. It is easily accessible and inexpensive but is limited in the assessment of multi-level stenoses and heavily calcified vessels.11 MRA has a high diagnostic accuracy for PAD and is used in patients who require further imaging following duplex ultrasonography prior to revascularisation. CTA can also be used as an alternative imaging method when MRA is contraindicated or not tolerated.10
In the UK, an initial PAD assessment should be performed in the primary care setting. A patient who presents with features of intermittent claudication, defined as reproducible calf, thigh and/or buttock pain on exertion, or with features of critical limb-threatening ischaemia, defined as the presence of chronic rest pain, skin changes such as ulceration, non-healing wounds and/or gangrene, should be assessed for possible PAD. Such an assessment is also indicated in patients with diabetes, unexplained leg pain, those who require compression hosiery and those being considered for interventions to the leg or foot. The assessment for PAD involves a clinical history, lower limb examination and ABPI measurement.10
An ABPI value of <0.9 is regarded as confirming the presence of PAD, though a resting ABPI value of >0.9 does not necessarily exclude the diagnosis of PAD, particularly in the presence of a positive history, risk factors, or if the value is >1.4.12,13 Regardless, ABPI assessments performed in primary care facilitate earlier PAD diagnosis, therefore improving patient outcomes.14,15 In addition, most PAD management can also be executed in the primary care setting with referral to secondary care only indicated in the case of non-responding or worsening symptoms of intermittent claudication or in the case of critical limb-threatening ischaemia.
An outline of the initial assessment and management pathway for patients presenting to primary care with varying degrees of suspected PAD is summarised in Figure 1, based on current guidelines from the National Institute for Health and Care Excellence (NICE).10 To follow these guidelines on assessment and management, it is important that ABPI measurements are widely available in the primary care setting.
However, manual ABPI measurements can be time-consuming, as a period of supine rest is recommended prior to the measurement being taken and the blood pressure in each of the six arteries is measured separately.16 This, in combination with the limited expertise available in the primary care setting, means that ABPI measurements are often not performed when indicated, resulting in secondary care referrals being made earlier than necessary to diagnose or exclude PAD.17,18 These factors may also preclude the measurement of ABPI, when indicated, in other healthcare settings outside of a vascular centre.
Automated devices are becoming increasingly common for brachial blood pressure measurements in clinical practice, largely due to their simplicity and accuracy when compared to the traditional auscultation of Korotkoff sounds.19 Such devices are also available for automated ABPI measurements. However, they are not currently widely accepted by the vascular, and wider, community due to limited evidence surrounding their accuracy and diagnostic performance in PAD. It is also not clear whether the diagnostic accuracy differs between device manufacturers.
Automated ABPI devices have the potential to replace manual ABPI measurements, which may negate the need for many secondary care referrals, particularly if PAD is not present, patients are asymptomatic or symptoms are mild.10 Additionally, automated devices may improve accessibility to ABPI measurements in a variety of community and non-vascular settings. As such, these devices have the potential to improve patient care and alter the clinical pathway, better aligning it to what is recommended in the NICE guidelines (ie, diagnosis and management in primary care).
A previous systematic review was conducted in 2012, considering the reliability of automated ABPI devices. This review concluded that automated ABPI devices are valid and provide a practical alternative for the detection of PAD. However, sensitivity was low at 69%, prohibiting automated devices from replacing manual ABPI measurements due to their inferior test accuracy.20 In the 10 years following this study, new automated devices have been developed which may have improved sensitivity and specificity for PAD diagnosis. Therefore, the aim of this study is to provide an updated review of the evidence considering the role of automated ABPI devices in the detection of PAD in patients with known or suspected PAD.
Our primary objective is to determine the diagnostic accuracy of automated ABPI devices for detecting or excluding PAD in people with known or suspected PAD.
Our secondary objectives are to identify whether the accuracy of these measures is altered by differences between device manufacturers, study setting (ie, primary and secondary care) and participant characteristics.
This protocol has been developed using the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy and will be reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy (PRISMA-DTA).21,22
Types of studies
We plan to include all cross-sectional comparative studies written in the English language which evaluate the accuracy of automated ABPI devices for diagnosing or excluding PAD. Only fully paired direct comparisons will be included whereby each patient was tested using an automated ABPI device and via the manual method or another reference standard. Patients may also be randomised to receive one (of multiple) automated device or randomised to be assessed via an automated ABPI device or via the manual method. Such studies will be included if an appropriate reference standard is also used for each randomised patient. No exclusions will be made based on methodological quality or sample size.
Studies with adult participants (18 years of age and older), of any sex, in any clinical setting, who have suspected or previously diagnosed PAD will be eligible for inclusion. For those with suspected PAD, we will include all groups for whom an ABPI is indicated according to the NICE guidelines.10 This includes patients who (i) have symptoms suggestive of peripheral arterial disease; or (ii) have diabetes, non-healing wounds on the lower limbs or unexplained leg pain; or (iii) are being considered for lower limb interventions; or (iv) need to use compression hosiery.
The index test to be reviewed is automated ABPI, captured by oscillometric or plethysmographic devices. Any automated ABPI device and method will be included, regardless of whether the device has been validated for use in PAD. An ABPI value of <0.9 is widely regarded as the cut-off value for diagnosing PAD; however, studies will not be excluded if they have used different threshold values and this will be accounted for during statistical analysis.
PAD is the target condition for this systematic review, which the index and reference tests are intended to identify or exclude. Studies may categorise PAD into asymptomatic, intermittent claudication and critical limb-threatening ischaemia. No exclusions will be made based on categorisation.
Comparative test and reference standards
The comparative test considered to be the reference standard in this review will be the manual Doppler ABPI measurements. Manual Doppler ABPI measurements should be taken using a sphygmomanometer and Doppler ultrasound probe and any recognised method for calculating ABPI will be included. Additional reference standards used to confirm the presence or absence of PAD can include Doppler ultrasonography, MRA or CTA. Studies that include additional measures to assess vascular status, such as toe brachial pressure index, will be included; however, these data will not be included in the analysis.
Systematic searches will be performed using the MEDLINE, EMBASE, CENTRAL, and CINAHL databases. MeSH terms with full text synonyms will be searched and include (“peripheral arterial disease” or “peripheral arter* disease”) and (“ankle brachial index” or “ankle brachial pressure ind*”) and (“oscillometr*” or “plethysmograph*”). A draft search is shown in Appendix 1 (online at www.jvsgbi.com). Searches will be restricted to articles written in the English language; no date restrictions will be applied.
Searching other sources
The reference lists of all included studies and screened full texts will be manually reviewed for additional relevant papers. Clinical trial registries including ClinicalTrials.gov and the Clarivate Web of Science: Conference Proceedings Citation Index will be searched for ongoing studies and authors will be contacted for results where possible.
Data collection and analysis
Selection of studies
Search results will be uploaded onto the Covidence systematic review software, which automatically removes duplicated articles.23 The titles and abstracts will be screened for eligibility by two independent reviewers, and full texts of potentially relevant articles will then be independently reviewed for inclusion. Any disagreement between reviewers at either stage will be resolved by consensus or with a third reviewer. When full texts are not obtainable via conventional access methods, the authors and publishing journal will be approached to request the full article text. The number of search hits, number of duplicates removed, number of full texts reviewed, number of full texts excluded with reasons and the number of studies included will be reported using the PRISMA flow diagram.
Data extraction and management
Extraction of relevant data will be performed by two independent reviewers and recorded on two separate Microsoft Excel spreadsheets, using a bespoke data extraction form. Data extraction will be based on the Cochrane handbook.24 The extracted data will include: (i) study characteristics including year of publication, country, study design, sample size, duration, setting, and inclusion and exclusion criteria; (ii) participant characteristics including age, sex and comorbidities; (iii) description of the index test including automated device name, operator and device validation; (iv) description of the reference test(s) including equipment, operator and method for calculating ABPI if appropriate; and (v) findings related to primary and secondary outcomes, including results to recreate 2×2 diagnostic tables for estimating test accuracy. Any discrepancies in the extracted data will be resolved by reviewing the original article.
Assessment of methodological quality
Studies that meet the eligibility criteria will be appraised for risk of bias and applicability by two independent assessors using the quality assessment of diagnostic accuracy studies (QUADAS-2) tool and the QUADAS-C extension for comparative diagnostic accuracy studies.25,26 The QUADAS-C tool is shown in Appendix 2 (online at www.jvsgbi.com). Any disagreement between reviewers will be resolved by consensus or with a third reviewer. Each study will be assessed on patient selection, index test, reference standard, and flow and timing, with each domain being classified into one of three categories: (i) high risk of bias; (ii) unclear risk of bias; and (iii) low risk of bias. The effect of methodological quality will be accounted for in subgroup analyses.
Statistical analysis and data synthesis
Statistical analysis will be performed using R package mada in R language version 4.1.27 Initial data synthesis will include cross tabulation of the binary outcomes ‘PAD’ or ‘no PAD’ for automated ABPI against the reference standard, manual ABPI in diagnostic 2×2 tables (ie, true positives, true negatives, false positives and false negatives). If 2×2 tables are not provided directly, they will be back calculated from raw data where possible.24 Where data are missing to allow construction of 2×2 tables, the study authors will be contacted.
Studies with fully paired designs will be entered into a meta-analysis. The patient will be the unit of analysis. Due to expected variations in the unit of analysis used by included studies, an analysis will be performed to evaluate the impact of the unit of analysis (ie, patient vs limb). Forest plots with 95% confidence intervals (CI) and summary receiver operator characteristic (SROC) curves with 95% prediction and 95% confidence regions will be produced as part of initial exploratory analyses. Given the anticipation of a common threshold (ABPI <0.9) and for substantial study heterogeneity, as is expected in a meta-analysis of diagnostic test accuracy, we will use a bivariate random-effect model to provide pooled estimates of summary accuracy statistics.21 If there is evidence of a threshold effect, the hierarchical SROC model will be used.28 All SROC curves will be plotted with studies as weighted data points.
We plan to perform subgroup analyses and meta-regression for: (i) study characteristics (eg, study design, study setting and study quality); (ii) participant characteristics (eg, age, sex, diabetes, hypertension, smoking status and PAD severity); and (iii) comparative index test characteristics (eg, unit of analysis, ABPI calculation method, automated device type, device validation status, reference standard and threshold effect, if appropriate). This will allow us to investigate the impact of these subgroups on automated ABPI diagnostic test accuracy.
Investigations of heterogeneity
Study heterogeneity will be assessed by visual inspection of coupled forest plots and SROC plots. We expect that included studies will use a common ABPI threshold of 0.9; however, there may be slight variation in the threshold used due to equipment calibration and differences between operators.
We intend to use Spearman’s correlation coefficient to test for the presence of a threshold effect as a source of heterogeneity. For this, we will use the sensitivity and specificity of all studies and r >0.6 will indicate the presence a threshold effect.29 The aforementioned subgroup analysis will also allow us to investigate the effect of these sources of heterogeneity on automated ABPI diagnostic test accuracy.
Assessment of reporting bias
The presence of publication bias will be assessed visually using a funnel plot. If more than 10 studies are included in the analysis, funnel plot asymmetry will be examined using Deeks’ test.30
This protocol outlines a systematic review to assess the diagnostic accuracy of automated ABPI devices for detecting or excluding PAD in people with known or suspected PAD. The manual Doppler ABPI method is currently the recommended first-line investigation for PAD, though there are certain drawbacks such as the time and expertise required for measurement. These limitations also mean that ABPI measurements are rarely obtained in primary care, as is recommended in NICE guidelines. This leads to referrals to secondary care to diagnose or exclude PAD. In addition, it also means that ABPI measurements are rarely obtained in other settings including community healthcare services, prison healthcare services and non-vascular district general hospitals. Automated devices have the potential to overcome some of these drawbacks, making ABPI measurements more accessible in a variety of settings and reducing the need for some secondary care referrals. These devices are not currently widely accepted due to concerns surrounding their accuracy, particularly their sensitivity.20 However, the contemporaneous evidence for such devices is yet to be fully evaluated, an evidence gap that this review aims to fill.
An anticipated limitation of this review is considerable heterogeneity amongst study characteristics and outcomes measured, making statistical comparison challenging. Such heterogeneity has been identified in a previous review, mostly due to differences in automated devices used and methods for manual ABPI measurements.20 We plan to assess the impact of these sources of heterogeneity during our subgroup analyses.
Overall, this review aims to summarise the current evidence for the accuracy of automated ABPI devices. The results will be used to aid medical professionals in the diagnosis of PAD, altering the current clinical pathway and aligning it to what is recommended in NICE guidelines. The results may also assist in providing eligibility criteria framework for future trials designed to validate new automated ABPI devices.
November 22, 2022
1. Academic Vascular Surgical Unit, Hull York Medical School, Hull, UK
2. Institute of Clinical and Applied Health Research, University of Hull, Hull, UK
3. Hull University Teaching Hospitals NHS Trust Library Service, Hull, UK
Academic Vascular Surgery Unit, Hull University Teaching Hospitals NHS Trust, Hull HU3 2JZ, UK Email: [email protected]