Date Published: March 14, 2019
Publisher: Public Library of Science
Author(s): Luca Valerio, Ron J. Peters, Aeilko H. Zwinderman, Sara-Joan Pinto-Sietsma, Ivana Isgum.
Changes in the microcirculation may be used as a surrogate outcome in studies on cardiovascular disease. We assessed the reliability characteristics of the sublingual microcirculation parameters Vascular Density (VD), Red Blood Cell Filling (RBCF), and Perfused Boundary Region (PBR) as obtained by sidestream darkfield imaging.
For each of the three parameters, the variance components of measurement, the Intraclass Correlation Coefficient (ICC), the Standard Error of Measurement, and the limits of agreement were estimated for the intra-rater setting (N = 50) and the inter-rater setting (N = 48). Subsequently, as a proof of concept, the reliability measures were used for a power analysis to design studies to evaluate the effect of acute stimuli–i.e. having a meal (N = 50) and cigarette smoking (N = 21) on the three parameters.
Reproducibility was poor for all three parameters. The intra-rater ICC for 2 measurements was 0.28 (95% CI: 0.04, 0.53) for the VD, 0.51 (95% CI: 0.27, 0.69) for the RBCF, and 0.33 (95% CI: 0.08–0.56) for the PBR. The standard errors of measurement and the limits of agreement for all three parameters were larger than most statistically significant intra-individual or inter-individual differences reported in previous studies. The proofs of concept showed that sample sizes in excess of 600 subjects are necessary to reach statistical significance for the observed effects of having a meal or smoking on VD and PBR.
The reliability of the three sublingual microcirculation parameters in their current form appears to be low and a large sample size is advisable for their use in conditions similar to those we describe.
Longitudinal investigation of cardiovascular disease is time-expensive and requires large groups of participants. In the search for reliable surrogate markers of disease, microcirculation has been proposed as a candidate target. For this purpose, the Sidestream Dark Field  video recordings are increasingly used to measure the sublingual microcirculation in the general population, after measurements conducted using its precursor, Orthogonal Polarization Spectral imaging, had been suggested to be of prognostic relevance in critically ill patient populations . The parameters used include two markers of microvascular perfusion, the Vascular Density (VD) [3–5] and the Red Blood Cell Filling percentage (RBCF) , and an indicator of the endothelial surface layer usually referred to as the glycocalyx, the Perfused Boundary Region (PBR) . Since they can be obtained from a non-invasive, fast, and semi-automated procedure, these parameters have been considered markers of vascular damage with potential diagnostic, prognostic, and therapeutic values[4, 7, 8]. However, subsequent studies did not yield consistent results. In particular, the association of these parameters with the risk factors for atherosclerosis has been confirmed by some studies[4, 9] and disputed by others. In order to better determine whether the correlations reported for a biomedical parameter are true or spurious, knowing how reliable they are is of paramount importance. In particular, it allows researchers to calculate an adequate sample size for future studies which is necessary to apply this method in clinical practice. We approached this matter in two ways: First, we calculated the contributions of different sources of variation to the total variability of the measurements; second, we calculated parameters of measurement error, which helps to determine the clinical applicability of research findings obtained under comparable conditions . We then performed two experimental proofs of concept to use the reliability parameters for sample size calculation.
The results of the reproducibility analyses are shown in Table 2. The Bland-Altman plots for the best researched parameter, the PBR, are shown in Fig 2; those for the RBCF and the VD are shown in S2 Fig.
Under our testing conditions, the sublingual microcirculation parameters VD, RBCF, and PBR display poor reproducibility, that seems to originate from poor correlation between measurements on the same subject and high residual variance over the whole range of values rather than association between the true value and either variance or the difference between repeated measurements (S1 Fig).