ABI determination is useful for the assessment of peripheral vessel status and cardiovascular risk in epidemiologic studies. New automated tools based on photo-plethysmography were established which can simultaneously measure ABI and other vascular parameters such as pulse wave velocity. These devices promise an easier handling compared to Doppler sonography which could improve the standardization of these assessments in large epidemiologic trials. In the present study, we aimed at introducing plethysmography-based ABI assessments in an epidemiologic context by assessing feasibility, agreement with Doppler sonography, intra- and inter-observer concordances. We compared the results between the two devices Vicorder and Vascular Explorer whose measurements are based on the deflation and inflation method respectively.
We also analysed the performance of a number of correction methods of the raw values of Vicorder and Vascular Explorer, namely observer or examiner revised values for both methods and the deflation method of Vascular Explorer. The latter one showed better agreement with Doppler especially for patients with peripheral artery disease .
Acceptance scores of both devices are similarly high. Moreover, we documented a comparably short learning and training phase for both devices and observers, including the training for manual ABI corrections. Therefore, we conclude that both assessments are feasible in large population based studies. However, measurements with Vascular Explorer were more time-consuming mainly due to multiple inflations of the ankle cuff in case of low signal quality. This applies in particular for subjects with high blood pressure and might be preventable by appropriate software modifications. Considering the economic aspect, the costs for both PPG-devices in our hands were comparable (Vicorder: about 13,000€, Vascular Explorer: about 11,000€) but an order of magnitude higher than for hand-held Doppler (below 1,000€).
While measurement triplicates of Doppler express high concordance, it was considerably lower for the automated measurements. The high concordance of the repeated Doppler measurement could be attributed to the subjective evaluation of the acoustic Doppler signal. Accordingly, after visual correction of the raw ABI values, both PPG-based devices improved their within-set concordances. Deflation modus in Vascular Explorer did not improve the within-set concordance indicating that the differences between measurement replicates are not modus-specific. We conclude that at least three repeated measurements are required in an epidemiologic study to control the variability of automated ABI measurements.
In the present study, we observed a strong bias of automated methods towards higher ABI values compared to Doppler. The bias is especially pronounced for Vascular Explorer. The observed bias is caused by either a negative bias of brachial pressures (Vicorder) or a positive bias for ankle pressures (Vascular Explorer). The negative bias for brachial pressures is in accordance to findings reported by Jobbagy et al.  who demonstrated a substantial delay in PPG detection of artery opening during the cuff deflation at the upper arm when compared with deflationary oscillometry performed by OMRON M4. Higher ankle pressures during automated PPG-based measurement can be attributed to the assessment of overall limb circulation, whereas Doppler detects only one particular artery of interest. Accordingly, a possible sustained leg perfusion through collaterals might be a source of higher ABI-values. Moreover, the higher biases observed for the Vascular Explorer could be a result of the prolonged tibial artery occlusion times due to frequently longer or repeated ankle cuff inflations mentioned above. The temporary increase of systolic and diastolic blood pressure as reaction on the cuff inflation is well documented especially in subjects with hypertension . Similarly, we observed reactive pressure increase by Vicorder, however, almost exclusively at the first measurement. Thus, further improvement of the current sensors and detection algorithms might avoid the redundant cuff inflations which could improve the measurements for the inflationary methods in the future.
The deflationary method of Vascular Explorer ameliorated the observed bias. However, it was still significantly larger than for Vicorder. We showed in  that the deflation method of Vascular explorer works especially well for improving the concordance of low ABI values which are less common in our study population. Vicorder and Vascular Explorer showed good intra- and inter-observer concordances. Intra-observer concordances appeared to be virtually identical. In contrast, the inter-observer concordance was (non-significantly) better for Vicorder.
For both devices, observer or examiner based corrections of raw values failed to provide clearly better results with respect to agreement with Doppler or intra- and inter-observer concordances. Therefore, we recommend using the raw values. In our study, and on the basis of our SOPs and Software Versions, Vicorder performed slightly better with respect to time requirements, agreement with Doppler, variance of repeated measurements and inter-observer agreement. Therefore, we decided to use Vicorder in our population based cohort aiming at recruiting 10,000 individuals (LIFE study). However, since both, Software Version and SOPs can be subject to changes, our finding should not be over-interpreted in the sense that Vascular Explorer is generally inferior.
Compared to Doppler both PPG-based methods do not constitute an advantage regarding time requirement and accuracy of ABI assessment. However, a major potential of PPG-based methods is the availability of other traits of vessel status such as pulse-wave parameters. Hence, the additional time requirement might be justified by a deeper characterization of individual’s vessel status. Since PPG-based ABI values are inflated compared to Doppler, it will be necessary to adjust Doppler-based cut-offs for risk stratification. We could show that proper adjustments of corresponding cut-offs could result in the same diagnostic power to detect PAD as Doppler based ABI. But definite recommendations for adjusted cut-offs would require larger sample sizes of both, PAD patients and controls .
There are several limitations of our study: We analysed a convenience sample which could imply that we over-estimated the acceptance but under-estimated the time requirements compared to a population-based sample. We used Doppler sonography as gold-standard since invasive blood pressure measurements and angiography are ethically not justified for our putatively healthy individuals. The sample size of our study is limited in view of the extensive assessment programme required to answer our questions. This did not allow us to estimate concordances and biases with high accuracy. Additionally, due to the same reason, we neither performed any sub-group analyses nor analyses of the effects of covariates on the performance of methods. Finally, we like to emphasize again that our results depend on the actual hard- and software versions of the devices and our SOPs. Since there are ongoing activities regarding improvements, it will be necessary to repeat our study if new versions are available. However, we believe that our study provides a first rationale for the application of automated PPG-based methods for vascular assessments in large epidemiologic studies.