The accuracy of an oscillometric ankle-brachial index in the diagnosis of lower limb peripheral arterial disease: A systematic review and meta-analysis
2017; Wiley; Volume: 71; Issue: 9 Linguagem: Inglês
10.1111/ijcp.12994
ISSN1742-1241
AutoresÁngel Herráiz-Adillo, Iván Cavero‐Redondo, Celia Álvarez‐Bueno, Vicente Martínez‐Vizcaíno, Diana P. Pozuelo‐Carrascosa, Blanca Notario‐Pacheco,
Tópico(s)Cerebrovascular and Carotid Artery Diseases
ResumoInternational Journal of Clinical PracticeVolume 71, Issue 9 e12994 META-ANALYSISFree Access The accuracy of an oscillometric ankle-brachial index in the diagnosis of lower limb peripheral arterial disease: A systematic review and meta-analysis Ángel Herráiz-Adillo, Ángel Herráiz-Adillo orcid.org/0000-0002-2691-0315 Department of Primary Care, Health Service of Castilla-La Mancha (SESCAM), Tragacete, SpainSearch for more papers by this authorIván Cavero-Redondo, Corresponding Author Iván Cavero-Redondo ivan.cavero@uclm.es Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, Spain Correspondence Iván Cavero-Redondo, Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, Spain. Email: ivan.cavero@uclm.esSearch for more papers by this authorCelia Álvarez-Bueno, Celia Álvarez-Bueno Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, SpainSearch for more papers by this authorVicente Martínez-Vizcaíno, Vicente Martínez-Vizcaíno Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, Spain Universidad Autónoma de Chile, Facultad de Ciencias de la Salud, Talca, ChileSearch for more papers by this authorDiana P. Pozuelo-Carrascosa, Diana P. Pozuelo-Carrascosa Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, SpainSearch for more papers by this authorBlanca Notario-Pacheco, Blanca Notario-Pacheco Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, SpainSearch for more papers by this author Ángel Herráiz-Adillo, Ángel Herráiz-Adillo orcid.org/0000-0002-2691-0315 Department of Primary Care, Health Service of Castilla-La Mancha (SESCAM), Tragacete, SpainSearch for more papers by this authorIván Cavero-Redondo, Corresponding Author Iván Cavero-Redondo ivan.cavero@uclm.es Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, Spain Correspondence Iván Cavero-Redondo, Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, Spain. Email: ivan.cavero@uclm.esSearch for more papers by this authorCelia Álvarez-Bueno, Celia Álvarez-Bueno Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, SpainSearch for more papers by this authorVicente Martínez-Vizcaíno, Vicente Martínez-Vizcaíno Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, Spain Universidad Autónoma de Chile, Facultad de Ciencias de la Salud, Talca, ChileSearch for more papers by this authorDiana P. Pozuelo-Carrascosa, Diana P. Pozuelo-Carrascosa Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, SpainSearch for more papers by this authorBlanca Notario-Pacheco, Blanca Notario-Pacheco Universidad de Castilla-La Mancha, Health and Social Research Center, Cuenca, SpainSearch for more papers by this author First published: 29 August 2017 https://doi.org/10.1111/ijcp.12994Citations: 19AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Summary Introduction Peripheral arterial disease (PAD) remains underdiagnosed and undertreated, partly because of limitations in the Doppler ankle-brachial index (ABI), the non-invasive gold standard. Objective This systematic review and meta-analysis aims to compare the diagnostic accuracy of the oscillometric ABI and the Doppler ABI, and to examine the influence of two approaches to analysis: legs vs subjects and inclusion of oscillometric errors as PAD equivalents vs exclusion. Methods Systematic searches in EMBASE, MEDLINE, Web of Science and the Cochrane Library databases were performed, from inception to February 2017. Random-effects models were computed with the Moses-Littenberg constant. Hierarchical summary receiver operating characteristic curves (HSROC) were used to summarise the overall test performance. Results Twenty studies (1263 subjects and 3695 legs) were included in the meta-analysis. The pooled diagnostic odds ratio (dOR) for the oscillometric ABI was 32.49 (95% CI: 19.6-53.8), with 65% sensitivity (95% CI: 57-74) and 96% specificity (95%CI: 93-99). In the subgroup analysis, the “per subjects” group showed a better performance than the “per legs” group (dOR 36.44 vs 29.03). Similarly, an analysis considering oscillometric errors as PAD equivalents improved diagnostic performance (dOR 31.48 vs 28.29). The time needed for the oscillometric ABI was significantly shorter than that required for the Doppler ABI (5.90 vs 10.06 minutes, respectively). Conclusions and relevance The oscillometric ABI showed an acceptable diagnostic accuracy and feasibility, potentially making it a useful tool for PAD diagnosis. We recommend considering oscillometric errors as PAD equivalents, and a “per subject” instead of a “per leg” approach, in order to improve sensitivity. Borderline oscillometric ABI values in diabetic population should raise concern of PAD. Review criteria Systematic searches in EMBASE, MEDLINE, Web of Science and the Cochrane Library databases were performed through predefined search criteria. Studies reporting a 2 × 2 contingency table comparing Doppler ABI (reference test) and oscillometric ABI (index test) were included. Message for the clinic The oscillometric ankle-brachial index (ABI) has proven good diagnostic performance and excellent feasibility; thus, it might be a useful tool for diagnosing peripheral arterial disease (PAD). To detect individuals at high cardiovascular risk, we suggest considering oscillometric errors as PAD equivalents and a “per subject” instead of a “per leg” approach as the unit of analysis. Borderline oscillometric ABI values in diabetic population should raise concern of PAD. 1 INTRODUCTION Peripheral arterial disease (PAD) is an age-dependent manifestation of atherosclerosis, which is highly prevalent in Western countries. Uncommon before the age of 50, its rates increase to about 20% by the age of 80.1 Moreover, PAD has proved to be an independent risk factor for coronary artery and cerebrovascular disease, and all-cause mortality.2 However, this condition remains both underdiagnosed and undertreated, with no consensus regarding on whom and when screening should be performed.3-5 Underdiagnosis can be attributed to the fact that only one out of three patients suffering from PAD are symptomatic,6 and because invasive catheter digital subtraction angiography, which is considered the gold standard for PAD diagnosis, is an invasive test that requires both iodinated contrast and ionising radiation. Nevertheless, patients with PAD but without claudication are also at increased risk of cardiovascular disease and mortality.7 Thus, in an attempt to overcome angiography limitations, the Doppler ankle-brachial index (ABI), because of its simplicity and availability, is considered the non-invasive gold standard for PAD. However, there is a lack of standardisation in ABI measurements. While the American Heart Association suggests using the higher Doppler value between posterior tibial or dorsalis pedis arteries, others recommend the lower value in an attempt to improve sensitivity in PAD diagnosis8, 9 and cardiovascular risk prediction.10 In addition, although PAD is classically defined as an ABI ≤0.9, the ideal cut-off may be influenced by clinical setting variables such as population characteristics or disease prevalence.11 ABI measured by oscillometry is a simple, reproducible and automatic method that is becoming popular, since it surpasses the limitations of the Doppler with regards to equipment, training and time constraints. Both the oscillometric and the Doppler ABI techniques are not fully standardised, in such a way that several procedures have been suggested: simultaneous vs sequential and unique vs multiple measurements. In addition, studies comparing the oscillometric ABI with the Doppler ABI differ in whether they consider calcified members and oscillometric errors as PAD equivalents or not. Moreover, two units of analysis are equally used yielding potentially different results: those analysing legs as independent measurements and those analysing subjects (defining as PAD subjects those with one or two pathological legs). A previous meta-analysis reported that the oscillometric ABI is a reliable and practical alternative to the conventional Doppler ABI, with 69% sensitivity and 96% specificity.12 However, although it has been reported that some statistical methods for meta-analyses of diagnostic accuracy might result in misleading summary estimates of sensitivity and specificity, no previous study has comprehensively reviewed and compared the accuracy of both the oscillometric and the Doppler method using Hierarchical Summary Receiver Operating Characteristic (HSROC), which is currently considered the most rigorous multivariate meta-analysis approach.13 Thus, the present study aims to identify and evaluate evidence regarding the diagnostic performance of the oscillometric ABI to detect PAD as compared with the Doppler ABI using HSROC meta-analysis procedures, and to examine the influence of two strategies of analysis: (i) subjects vs legs, and (ii) oscillometric errors analysed as PAD equivalents vs exclusion of oscillometric errors. 2 METHODS 2.1 Protocol and registration The protocol of this study was included in PROSPERO as “The accuracy of oscillometric ankle-brachial index in the diagnosis of lower limb peripheral arterial disease. The influence of two units of analysis and oscillometric errors: a systematic review and meta-analysis” with the registration number: CRD42016051120. 2.2 Literature search We systematically searched MEDLINE (via PubMed), EMBASE, the Cochrane Central Register of Controlled Trials, the Cochrane Database of Systematic Reviews and the Web of Science databases from their inception to February 2017. The search strategy comprises three comprehensive search terms combined with Boolean operators: (“ankle brachial index” OR “ankle brachial indices” OR “ankle-brachial” OR “ankle-arm”) AND (oscillomet* OR automat*) AND (usefulness OR accuracy OR sensitivity OR specificity OR comparison OR diagnosis OR diagnostic). The literature search was complemented by reviewing citations of the articles considered eligible for the systematic review. These steps were performed independently by two reviewers (AH and CA) and disagreements were solved by consensus or involving a third researcher (IC). 2.3 Selection criteria We aimed to identify original articles analysing the diagnostic performance of the oscillometric ABI (index test) compared with the Doppler ABI (reference standard) used to diagnose PAD. The following inclusion criteria were used: (i) study participants: individuals aged ≥18 years; (ii) the oscillometric ABI as the index test; (iii) the hand-held continuous wave Doppler ABI as the reference standard test; (iv) outcome: PAD diagnosis; and (v) study design: cross-sectional and comparative studies with either prospective or retrospective data collection. The exclusion criteria were: (i) insufficient data to calculate diagnostic odds ratio (dOR); (ii) studies conducted only on patients diagnosed with PAD; and (iii) studies written in a language other than English or Spanish. When the same study reported ABI measurements using two different oscillometers14 or observers,15 those maximising dOR were chosen for the meta-analysis. Studies in which a double analysis was possible,16, 17 “per subjects” and “per legs” analysis, an analysis “per legs” was computed for the global meta-analysis because it yielded narrower confident intervals. 2.4 Data extraction and quality assessment After analysing original reports, the following data were extracted: (i) author identification, (ii) year of publication, (iii) Doppler ABI calculation, (iv) oscillometric ABI calculation, (v) oscillometric device, (vi) Doppler probe, (vii) average time to perform the Doppler ABI and the oscillometric ABI techniques, (viii) setting, (ix) age, gender and number of participants, (x) prevalence of diabetes mellitus, (xi) prevalence of PAD, (xii) whether or not calcified limbs and oscillometric errors were excluded from analysis, (xiii) unit of analysis (subjects vs legs), (xiv) parameters summarising the accuracy of the test: cut-off, area under the curve (AUC), and a 2 × 2 contingency table (true positives, true negatives, false positives and false negatives) to calculate dOR, sensitivity and specificity. When necessary, we directly contacted the authors for additional data. Studies from which it was not possible to collect a 2 × 2 contingency table were excluded from the meta-analysis. Quality assessment of studies was performed using the Quality Assessment of Diagnostic Accuracy Studies-2 tool (QUADAS-2) to evaluate four domains in each study: (i) patient selection, (ii) index test, (iii) reference standard and (iv) flow of patients and timing of the tests. All four domains were evaluated regarding the risk of bias and the first three domains were also evaluated in terms of concerns regarding the applicability of results.18 Two investigators (AH and CA) assessed each study's methodological quality independently and disagreements were resolved by consensus or with a third investigator (IC). 2.5 Statistical analysis and data synthesis This study is reported according to the PRISMA statement19 and it fulfils the Cochrane Collaboration Handbook recommendations.20 The dOR, sensitivity, specificity, positive likelihood ratio (PLR) and negative likelihood ratio (NLR), as well as their corresponding 95% confidence intervals (CIs), were calculated globally and by subgroups. A continuity correction was made by adding 0.5 to all cell counts of the 2 × 2 tables to avoid indeterminate values of dOR, PLR and NLR.21 PLR and NLR were directly meta-analysed after excluding a significant threshold effect, which was studied through correlation between sensitivity and specificity, and a “shoulder-like” appearance of the HSROC curve.22 The dOR is a measure of the effectiveness of a diagnostic test that combines sensitivity and specificity into a single number, which could take values from 0 to infinity.23 A value of 1 indicates null diagnostic ability of the test, while higher values represent better discriminatory test performance. Moses’ constant of linear model was used to compute the dOR. This approach is based on the regression line using the logit of the dOR of each study as a dependent variable and an expression of the positivity threshold of the study as an independent variable.24 HSROC curves were used to summarise the overall test performance. They were also used to evaluate the magnitude of heterogeneity, in such a way that wider prediction regions suggest larger heterogeneity.25, 26 Additionally, the I2 statistic was used to evaluate heterogeneity across studies, with values of 50% corresponding to small, medium and large heterogeneity, respectively.27 Because of large heterogeneity in most cases, dOR estimates were pooled using a random-effects model with the Der Simonian and Laird method. Subgroup analyses were conducted according to factors potentially causing heterogeneity, such as unit of analysis (“per subjects” vs “per legs”), oscillometric error consideration (inclusion vs exclusion) and the nature of the populations (Primary care, intermediate cardiovascular risk clinics and Vascular services). The “per legs” analysis considered each leg as an independent unit of analysis for comparing the oscillometric and the Doppler measurements. Conversely, in the “per subjects” analysis, individuals rather than legs were the unit of analysis, considering as PAD subjects those with at least one leg with an ABI ≤0.9. In the subgroup analysis, oscillometric errors are defined as the incapacity of the oscillometer to report a value of ankle blood pressure. When oscillometric errors were included into the analysis, they were considered as PAD equivalents. Random-effects univariate and multivariate meta-regressions were used to separately evaluate the effects of potential covariates in dOR, sensitivity and specificity: (i) unit of analysis (subjects vs legs); (ii) oscillometric errors (inclusion vs exclusion); (iii) calcified legs (inclusion vs exclusion); (iv) timing of oscillometric measurements (simultaneous vs sequential); (v) validation of oscillometric devices (yes vs no); (vi) oscillometric devices specifically designed for ABI (yes vs no); (vii) standard oscillometric and Doppler calculation (yes vs no); (viii) Doppler test blinded to the oscillometric test results (yes vs no); (ix) population recruitment (consecutive vs not) and (x) patients’ characteristics: age, gender, sample size, prevalence of diabetes and prevalence of PAD. Sensitivity analyses were conducted by removing studies one by one in order to assess the robustness of the summary estimates and to detect whether any particular study accounted for a large proportion of heterogeneity. Finally, publication bias was assessed using both Deeks’ statistical test and a funnel plot.28 Publication bias is suspected when a non-vertical line for the slope of the coefficient is present (P < .10), thus proving asymmetry. Statistical analyses were performed using StataSE software, version 14 (Stata Corp, College Station, TX, USA). 3 RESULTS 3.1 Baseline characteristics The search retrieved a total of 472 articles, of which 209 were duplicates. After screening the titles and abstracts of the remaining 263 studies, 155 were excluded on the basis of the previously described criteria, leaving 108 full-text articles to be reviewed. Of those, 77 were excluded, leaving 31 articles for qualitative synthesis and 20 for the final meta-analysis, shown in Figure 1.19 Figure 1Open in figure viewerPowerPoint Literature search PRISMA flow diagram The 31 studies comprising this review included 5527 participants: 11 studies (n = 1760) used “per subjects” analysis, 11 studies (n = 1947) used “per legs” analysis and 11 studies (n = 2125) did not clearly describe the strategy of analysis, shown in Table 1. After exclusions, 1538 subjects (11 studies) and 3695 legs (11 studies) were analysed. Reasons for such exclusions were: (i) limb calcification,16, 17, 29-34 (ii) oscillometric errors14, 31, 33, 35-39 and (iii) not all participants had their limbs measured using both the oscillometric and the Doppler.40 In two studies,16, 17 a double analysis (“per subjects” and “per legs”) was performed. Table 1. Characteristics of the populations, devices and accuracy outcomes of the studies included in the meta-analysis that conducted: “per subjects” (not legs) analysis, “per legs” (not subjects) analysis and with not well defined unit of analysis Reference Study population n Age (years) Gender (% men) PAD prevalence Doppler ABI Oscillometric ABI Oscillometric device Sensitivity (%) Specificity (%) dOR AUC “Per subjects” analysisd Benchimol29 Cardiology consultants 219a 212b 55.0 62 29.7 ↑DPA or PTA/↑BA Ankle/↑BA OMRON M4g 76 95 52.1 Benchimol40 Preventive medicine consultants 354a 196b 50.5 74 13.3 PTA/↑BA Ankle/↑BAj OMRON HM 722g 92 98 362.6 Ena31 Diabetic patients from Internal Medicine office 110a 93b 71.0 65 32.3 ↑DPA or PTA/↑ BA Ankle/↑BAj OMRON M6g 67 87 12.7 0.870 Novo-García16 Primary care subjects with type 2 diabetics 215a 193b 69.5 56 19.7 DPA or PTA/BA same side Ankle/BA same side OMRON M6g 37 92 6.8 0.764 Arévalo-Manso41 Ictus or transient isquemic attack subjects 30a 67.8 66 26.7 ↑DPA or PTA/↑ BA Ankle (PTA)/↑BA OMRON HEM-907g 100 100 765.0 1.000 Nelson39 High cardiovascular risk patients 250a 242b 71.2 69 21.5 DPA or PTA/↑BA AUSC NR OMRON HEM-907g 62 92 16.8 0.870 Takahashi32 Atomic bomb survivors 115a 113b 72.3 41 8.8 PTA/R BA Ankle/↑BAi OMRON VP2000f, g 50 100 207.0 Umuerri42 Hypertension clinic consultants 153a 57.5 67 41.8 Louder DPA or PTA/↑BA Ankle/↑BA OMRON M3g 61 90 13.1 0.787 Forés, 201414 Primary care subjects 88a 72.7 48 13.6 NR Ankle/↑BAj OMRON M6g 33 97 15.8 Forés, 2014 Primary care subjects 88a 72.7 48 13.6 NR Ankle/↑BAi Microlife Watch BP Officef,g 8 97 3.9 Špan, 201633 General practitioner's office 136a 68.2 NR 10.3 ↑DPA or PTA/↑ BA NRi MESI ABPI MDf,h 57 99 105.9 Herráiz-Adillo, 201617 Primary care and vascular consultants 90a 82b 70.4 56 30.5 ↑DPA or PTA/↑ BA Ankle/↑BAi OMRON M3g 84 100 549.4 0.970 “Per legs” analysise Beckman, 200630 Vascular clinic consultants 201a 345c 66.0 47 27.8 ↑DPA or PTA/↑BA Ankle/↑BAj CasMed740h R: 73 L: 88 R: 95 L: 85 32.9 Vinyoles35 Hypertensive primary care subjects 100a 157c 66.4 39 11.5 PTA or DPA if PTA undetected/↑BA Ankle/↑BA OMRON HEM 705CPg R: 37 L: 20 R: 93 L: 97 7.2 MacDougall36 Vascular clinic consultants 57a 62c 65.0 84 32.3 ↑DPA or PTA/average (L+R) BA or ↑BA Mid-calf/average (L+R) BA OMRON HEM 711Cg 71 89 19.1 Aboyans15 Vascular clinic consultants 54a 108c 58.2 52 23.1 ↑DPA or PTA/average (L+R) BA or ↑BA Ankle/average (L+R) BAj Spengler ProMg 76 96 69.0 Korno34 Vascular clinic consultants 61a 120c 67.0 NR 56.7 ↑DPA or PTA/↑BA same side Ankle/↑BAj CasMed 740h 71 92 27.4 0.920 Wohlfahrt44 General population 839a 1678c 54.3 47 1.1 ↑DPA or PTA/R BA NRi Boso ABI system 100f,h 78 98 147.5 Kollias37 Cardiovascular clinic consultants 93a 183c 62.5 62 13.1 ↑DPA or PTA/↑BA NRi Microlife Watch BP Officef,g 83 97 128.0 0.981 Rosenbaum43 Cardiovascular risk factors (≥2) 157a 314c 59.1 51 17.8 OSC ↓DPA or PTA/OSC R BA Ankle/↑BAi SCVLf,g 57 89 10.3 Novo-García16 Type 2 diabetics 215a 403c 69.5 56 14.4 DPA or PTA/BA same side Ankle/BA same side OMRON M6g 24 94 4.9 0.724 Sinski38 Coronary artery disease patients 80a 158c 70.1 66 35.4 PTA/↑ BA NRi Microlife Watch BP Officef,g 46 98 34.9 Herráiz-Adillo17 Primary care and vascular consultants 90a 167c 70.4 56 25.1 ↑DPA or PTA/↑ BA Ankle/↑BAi OMRON M3g 79 96 77.3 0.958 Not well defined unit of analysis Di Yacovo51 Internal medicine department 145 70.0 47 NR NR NR NR 66 87 13.0 Fröhlich52 Outpatient clinic 100 NR NR 40.0 NR NR Angio Experienceh 56 92 14.6 0.740 Fröhlich52 Outpatient clinic 100 NR NR 40.0 NR NR Boso ABI system 100f,h 86 87 41.1 0.920 Hamel53 Hospitalised patients ≥ 65 years 287 79.6 48 13.3 ↑DPA or PTA/↑BA Ankle/↑BA OMRON M6g 34 96 12.4 Khukhua54 No PAD symptoms 104 57.9 57 9.6 NR NR OMRON 705 ITg 90 100 – Bakalakou55 NR 130 67.0 66 30.4 NR NR OMRON M4g 80 95 76.0 0.900 Campens56 Cardiology department 99 NR NR 22.0 NR NR Datascope Acutorr Plusg 80 97 129.3 Rantamaula57 Diabetic patients 100 NR NR 18.5 NR NR OMRON M3g 43 94 11.8 Balkanay58 Cardiovascular clinic 53 46.9 66 17.0 NR NR Microlife Watch BP Officef,g 6 97 2.1 Laroche59 >55 years old with any ischaemic event 505 69.0 63 NR NR NR SCVLf,g 64 67 3.6 Gerald Seinost60 NR 372 NR NR NR NR NR NR R: 81 L: 74 R: 83 L: 84 R: 20.8 L: 14.9 Liu61 Diabetic patients 230 NR NR NR NR NR OMRON VP1000f,g R: 94 L: 96 R: 95 L: 97 R: 297.7 L: 776.0 ABI, ankle brachial index; AUC, area under curve; AUSC, auscultatory technique; BA, brachial artery; DOR, diagnostic odds ratio; DPA, dorsalis pedis artery; L, left; NR, not reported; OSC, oscillometric; PAD, peripheral arterial disease; PTA, posterior tibial artery; R, right. n, sample size: a: subjects; b: number of subjects really analysed, after excluding those with calcified limbs, oscillometric errors and those without both Doppler and oscillometric measurements; c: number of legs really analysed, after excluding calcified limbs and oscillometric errors. PAD prevalence (according to Doppler ABI): d: considering subjects (1 or 2 pathological legs); e: considering legs. Ankle brachial index formulas: ↑: highest value is considered; ↓: lowest value is considered. Characteristics of oscillometers: f: specifically designed for ABI measurements; g: validated for arm BP measurement; h: not validated; i: simultaneous oscillometric measurements; j: not simultaneous oscillometric measurements. The studies were conducted in 18 countries, with participants ranging in age from 46.9 to 79.6 years. The prevalence of PAD across studies considering subjects (one or two pathological legs) and legs varied from 8.9% to 41.8% and from 1.1% to 56.7%, respectively. Studies which used “per legs” analyses as compared with those using “per subjects” analyses involved younger participants (60.5 vs 64.5 years old), more women (49.1% vs 38%), less prevalence of diabetes (29.8% vs 37.9%), less cardiovascular events (16.5% vs 24.4%), similar mean oscillometric ABI (1.063 vs 1.062) and higher mean Doppler ABI (1.101 vs 1.038). 3.2 Study quality Quality assessment of the included studies was performed using the QUADAS-2 tool. Most studies had bias in patient selection (domain 1) and in the reference test (domain 3), see Figure S1. Considering patient selection, six studies (30%) had exclusions that were a potential risk of bias (PAD subjects)15, 29, 33, 35, 40, 41 and in two studies (10%),15, 17 there was concern about a case–control design. In eight studies (40%), the reference standard did not fulfil the standard ABI calculation16, 35, 38-40, 32, 42, 43 and in four studies (20%),16, 31, 39, 41 the Doppler test was not blinded from the oscillometric test results. One study (5%)40 had partial verification bias. Table S1 provides detailed data on the QUADAS-2 assessment of the studies and the rules used to score each domain. 3.3 Meta-analysis Figure 2 depicts the dOR forest plot of the included studies. Heterogeneity across studies comparing oscillometric and Doppler ABI measurements was high in dOR (I2 = 75.6%), moderate in sensitivity (I2 = 46.1%) and absent in specificity (I2 = 0.0%). The pooled estimates for the diagnosis of PAD were 32.49 for dOR, 65% for sensitivity, 96% for specificity, 15.33 for PLR and 0.30 for NLR. Table 2 depicts the global estimates of accuracy in the diagnosis of PAD. Figure 3 shows the global HSROC curve estimating the discriminating accuracy of the oscillometric ABI for identifying PAD. Figure 2Open in figure viewerPowerPoint Forest plot of the diagnostic odds ratio of the oscillometric ankle brachial index in comparison to the Doppler ankle brachial index to detect peripheral arterial disease Table 2. Pooled estimations of accuracy parameters in the diagnosis of peripheral arterial disease: global, by unit of analysis (“per subjects” vs “per legs”) and regarding oscillometric errors (included vs excluded) Type of analysis No. of studies Sensitivity (%) Specificity (%) PLR NLR dOR Global 20 65 (57-74) 96 (93-99) 15.33 (8.8-26.8) 0.30 (0.18-0.50) 32.49 (19.6-53.8) “Per subjects” 11 67 (57-78) 95 (90-100) 21.79 (10.3-46.0) 0.27 (0.13-0.54) 36.44 (16.7-79.3) “Per legs” 11 62 (51-76) 96 (92-99) 12.50 (5.8-26.8) 0.33 (0.16-0.67) 29.03 (14.6-57.9) OSC errors included as PAD equivalents 11 63 (50-78) 94 (89-99) 15.25 (7.2-32.3) 0.26 (0.13-0.51) 31.48 (13.6-72.9) OSC errors not included 11 58 (46-74) 95 (90-100) 15.57 (7.2-33.8) 0.31 (0.15-0.62) 28.29 (13.2-60.6) dOR, diagnostic Odds Ratio; OSC, oscillometric; NLR, negative likelihood ratio; PAD, peripheral arterial disease; PLR, positive likelihood ratio. Values in parentheses are 95% confidence intervals. Figure 3Open in figure viewerPowerPoint Hierarchical summary receiver operating characteristic curve summarising the ability of the oscillometric ankle brachial index to detect peripheral arterial disease in comparison to the Doppler ankle brachial index Figures 4 and 5 depict the global forest plots of sensitivity and specificity in the meta-analysis. Figure 4Open in figure viewerPowerPoint Forest plot of the sensitivity of the oscillometric ankle brachial index in comparison to the Doppler ankle brachial index to detect peripheral arterial disease Figure 5Open in figure viewerPowerPoint Forest plot of the specificity of the oscillometric ankle brachial index in comparison to the Doppler ankle brachial index to detect peripheral arterial disease 3.4 Time of measurements in Doppler ABI and oscillometric ABI Six and seven studies reported time of measurements in the Doppler ABI and the oscillometric ABI, respectively. The Doppler ABI time measurements ranged from 6.65 to 14.00 minutes, while those of the oscillometric ABI ranged from 2.0 to 8.1 minutes. The time needed for the Doppler ABI was significantly longer (10.06 minutes, 95% CI: 6.76-13.35) than that required for the oscillometric ABI (5.90 minutes, 95% CI: 5.08-6.73), also showing higher intra and inter study variability, see Figure S2. 3.5 Subgroup analysis 3.5.1 Unit of analysis (“per subjects” vs “per legs”) “Per subjects” analyses showed higher dOR than “per legs” analyses: 36.4 (I2 = 73.5%) vs 29.0 (I2 = 80.7%), see Figure S3. Pooled estimates of accuracy parameters in this subgroup analysis (sensitivity, specificity, PLR and NLR) are depicted in Table 2. Figures S4 and S5 show the HSROC curves by unit of analysis. 3.5.2 Inclusion or not of oscillometric errors When oscillometric errors were analysed as PAD equivalents, dOR and sensitivity increased from 28.29 to 31.48 and from 58% to 63%, respectively. Specificity did not change substantially (95% vs 94%), see Table 2. 3.5.3 Nature of
Referência(s)