In this study we evaluated statistical reasons to support application of specific conditions to FMD testing on the brachial artery. While assessment of FMD of the brachial artery either with proximal (upper arm), or distal (forearm) occlusion has been found to be correlated with coronary endothelial function[12, 13], we evaluated whether other test characteristics may provide a basis for selecting one approach over the other.
We have confirmed the well established finding that the maximal FMD is greater and delayed with upper arm occlusion[6, 14–20]. We showed a significantly greater and delayed FMD following proximal occlusion (16.2% and 107.2 seconds, respectively) compared to distal occlusion (7.3% and 67.8 seconds). The difference in the magnitude of FMD between occlusion sites (8.9%) is greater than the difference reported in a recent meta-analysis (2.4%). This discrepancy may be due to three major reasons. First, the measures were made in populations with different baseline endothelial function. Second, differences in baseline diameters in each of the two occlusion sites, may have affected the shear stress, the stimulus for endothelium-dependent dilation. In this study, shear stress was not analyzed. Finally, our use of an automated method of analysis with all available frames. Others studies that demonstrated greater FMD in response to proximal occlusion rather than distal occlusion [14–20] captured brachial artery images continuously, but then examined frames from specific time-points to define the artery's diameter and the frequency of these intervals was not consistent between studies.
The time to maximal FMD may vary not only by occlusion site, but also by the method used to detect the diameter from captured frames. For example, adhering to the recommended time interval to maximal FMD at 60 seconds after the release of the occlusive cuff  would have resulted in missing the true peak-FMD of 40% of the measurements we obtained with forearm occlusion (figure 3b), and about 95% of the cases of upper arm occlusion (figure 3a). Random measurement errors and underestimation of the actual FMD are less likely when examination of brachial artery dilation is done continuously over the entire course of the session and analyzed from all available images with automated computer based edge detection.
The assessment of FMD of the brachial artery is a functional bioassay for in vivo endothelial function in humans. It has been stated that the primary mechanism responsible for vessel dilation with forearm occlusion is endothelial release of nitric oxide(NO). Doshi et al. demonstrated that most of the dilatory response to proximal occlusion is NO-independent. Additional mechanisms, such as direct effect of hypoxia on the smooth muscle, altered myogenic response, or altered release of other mediators involved in vessel reactivity, may be involved in the dilation of the brachial artery following upper arm occlusion. NO-mediated vasodilation occurs ~60–80 seconds following reperfusion[24, 25]. Therefore, in the absence of a pharmacologic intervention, the increased time to maximal FMD and the prolonged dilatory response (Dend 11.3% vs. 3.3% p <0.0001) for proximal occlusion relative to distal occlusion may reflect other underlying mechanisms. NO is not the only mediator involved in FMD following proximal and distal occlusions; other endothelium-derived factors are probably involved differently with the two occlusion locations. It is essential to conduct validation studies in order to understand the role of different components of endothelial function in one approach or the other of the test.
Many studies have assessed the variability of the brachial artery diameter measurements, both at baseline and during the dilatory response. While the baseline diameter was shown to have relatively small variability (CV range 1–13%), the maximal FMD was associated with a much greater range of variability (CV range 1–84%). Various factors may contribute to the variability of brachial artery diameter, and these can be largely divided into technical and physiological factors. Ultrasound of the brachial artery is technically challenging, and measurement errors – even in magnitude of fraction of millimeters – especially when added to effect of intrinsic factors affecting FMD, may affect the accuracy of the test. Therefore, it is essential to apply a rigorous methodology to the test. In doing so, we utilized a single ultrasound machine, all studies were performed by one technician, and all images were read by one off-line reader. We found relatively low within-session variability and day-to-day variability of baseline diameter with no difference between proximal and distal placements. Similar day-to-day variability was found by Uehata et al. Greater variability of peak FMD is shown by the relatively high fraction of participants (more than a third, in both occlusion sites) that did not reach their maximal FMD by the average time-to-peak FMD (figure 3). The small impact of technical factors on baseline brachial artery diameter variability may indicate that physiological factors contribute most of peak FMD variability as suggested previously.
In this study we did not quantify the shear stress. It has been observed that the shear stress computed as area under the curve, is the major contributor to the magnitude of FMD. Therefore it is plausible that variability in shear stress may affect variability and repeatability of FMD, and understanding this interplay requires additional study.
We also assessed the repeatability of FMD, an important quality of this outcome as a potential tool in clinical studies where repeated measures are needed, and as a potential indicator for the risk of future coronary disease. Previous studies have shown inconsistent results of brachial artery FMD repeatability. Malik at al reported poor repeatability of FMD with an intraclass correlation coefficient of 0.1. In contrast, Welsch et al reported an intraclass correlation coefficient of 0.92 comparing day-to-day FMD. We demonstrated a reasonable repeatability of FMD measurements. Repeatability of FMD was better for the distal cuff location as shown by the smaller coefficient of repeatability (CR = 10.1% vs. 14.1%) and the greater difference on average of two repeated measurements with upper arm cuff placement (2.4%). Even with accurate methodology, our repeatability was only modest; suggesting that some degree of measurement error is inevitable in any setting. However, it is reasonable to assume that studies that use more than one sonographer or are performed in more than one center may be subject to a greater degree of measurement error than observed in this study. Hence, such studies must be extremely well coordinated, using highly trained technical staff, and probably employ a larger sample size than single site studies. The use of duplicate measurements may be important in reducing measurement error, as evidence show that repetitive reactive hyperemia does not effect FMD measurements , though the minimum time between measurements has not been clearly identified.