Simultaneous inter-arm and inter-leg systolic blood pressure differences to diagnose peripheral artery disease: a diagnostic accuracy study

Introduction

Henni et al. have commented our recently published paper about diagnostic performance of inter-leg systolic blood pressure difference (ILSBPD) to detect peripheral arterial disease (PAD) [1].

Firstly, Henni et al. noted that encoding absence of results on the automatic sphygmomanometer (i.e. “oscillometric errors”) as a "zero mmHg" pressure induces a bias in the results, since this finding is not inherently associated to a low or absent value of pressure. Certainly, there are other factors that can lead to oscillometric errors, such as arrhythmia, movement during measurements, inaccurate wrapping of the cuff or calcification. However, in our study, patients with arrhythmia (specifically atrial fibrillation) were excluded from the analysis, so arrhythmia cannot be responsible for such oscillometric errors. Moreover, our automatic blood monitor displays specific icons when arrhythmia, movement during measurement or inaccurate wrapping of the cuff is detected, thus excluding these factors easily as the cause of oscillometric errors. Therefore, low pressure (and much less commonly calcification) is the main factor leading to oscillometric errors in our work.

With regard to calcification, Aboyans et al. reported that PAD coexists with calcified limbs in up to 80% of cases, suggesting that calcified limbs should be considered as PAD equivalents [2]. Thus, both in low ankle pressure and calcification, the inability of the oscillometer to record ankle pressure should be considered as a qualitative message of pathology, and encoding this as “zero mmHg” may be a preferable approach.

In the literature, there is evidence that the consideration of oscillometric errors as “zero mmHg” in the oscillometric ABI to detect PAD is a preferable approach. As Henni et al. commented, we proved, in a previous meta-analysis, that considering oscillometric errors as PAD equivalents (“zero mmHg approach”) resulted in higher diagnostic accuracy at the expense of sensitivity of the oscillometric ABI to detect PAD [3]. In this sense, we also found that all but one of the 23 subjects with oscillometric errors that we analyzed in a previous work had either low ABI or calcification [4]. Similarly, Gómez-Huelgas et al. found that in all their 17 cases with oscillometric errors, the magnetic resonance angiography confirmed the presence of significant arterial stenosis (>50%) in the inferior members [5].

Begg et al. studied the influence of uninterpretability in diagnostic tests and reported that a relevant factor is the potential repeatability of the test, i.e. whether the cause of uninterpretability is a transient phenomenon or an inherent property of the subject [6]. In this sense, Henni et al. asked about the number of missing oscillometric results observed on one of the two recordings. Regarding PAD patients, 12 legs exhibited oscillometric errors in the two consecutive measurements and only 3 showed oscillometric errors in one measurement alone. Thus, double consecutive errors are far more common, indicating a condition which is usually linked to an inherent property of the subject (habitually low pressure), rather than an arbitrary finding. Therefore, oscillometric errors are (if not inherently, at least often) associated to a characteristic of the patient, and not inclusion (more than inclusion) of the oscillometric errors in the analysis would result in an uninterpretable results exclusion bias which would potentially bias sensitivity down and specificity up [7]. Specifically, in our work, if oscillometric errors are not considered, sensitivity would lower from 69.6% to 54.5% with no change in specificity.

Although we agree with Henni et al. that the consideration of oscillometric errors as “zero mmHg” infraestimates the real ankle pressure, thus overestimating ILSBPD, our study is designed as a diagnostic accuracy study, with no interest in the calibration or the level of agreement between oscillometric and Doppler ILSBPD. In response to Henni et al., when only one of the two consecutive measurements had a missing value, the other value (and not the mean) was considered as the final value.

Second, as Henni et al. commented, predictive values (PV) are strongly dependent on the prevalence of the disease. The lower prevalences of PAD found in the two large studies commented by Henni et al. (4.5% and 7.6% vs. 13.5%) [8,9] can be explained by the lower mean age (56.0 and 64.7 vs. 70.5 years) and the lower prevalence of diabetes (15.0% and 15.7% vs. 31.6%), two factors closely related with PAD. First of all, it should be noted that the application of the sensitivity and the specificity of our test in populations with different characteristics may not be straightforward, if spectrum effect is present [10]. Henni et al. commented that as prevalence decreases, PV + would decrease with low capacity of the test to ascertain PAD. However, the inherent consequence would be that, as prevalence decreases, PV– would increase with high capacity of the test to rule out PAD. Considering the great feasibility of the test, it may be better used as a “screening test” (to rule out the disease), rather than a “diagnostic test” (to rule in the disease). In this sense, ILSBPD would yield PV– of 98.3% and 97.1% in the abovementioned populations, with excellent capacity to rule out the disease.

Third, as Henni et al. commented, reference number 6 is incomplete. We apologize for the mistake. We agree with Henni et al. that PAD is a bilateral and systemic disease. However, in our work it can be read that “…atherosclerosis is habitually not symmetrically distributed between right and left limbs”. We have never questioned PAD bilateralism, but its perfect symmetry.

Disclosure statement

No potential conflict of interest was reported by the author(s).