Impact of Center Experience on Patient Radiation Exposure During Transradial Coronary Angiography and Percutaneous Intervention: A Patient‐Level, International, Collaborative, Multi‐Center Analysis

Abstract

BackgroundThe adoption of the transradial (TR) approach over the traditional transfemoral (TF) approach has been hampered by concerns of increased radiation exposure—a subject of considerable debate within the field. We performed a patient‐level, multi‐center analysis to definitively address the impact of TR access on radiation exposure.Methods and ResultsOverall, 10 centers were included from 6 countries—Canada (2 centers), United Kingdom (2), Germany (2), Sweden (2), Hungary (1), and The Netherlands (1). We compared the radiation exposure of TR versus TF access using measured dose‐area product (DAP). To account for local variations in equipment and exposure, standardized TR:TF DAP ratios were constructed per center with procedures separated by coronary angiography (CA) and percutaneous coronary intervention (PCI). Among 57 326 procedures, we demonstrated increased radiation exposure with the TR versus TF approach, particularly in the CA cohort across all centers (weighted‐average ratios: CA, 1.15; PCI, 1.05). However, this was mitigated by increasing TR experience in the PCI cohort across all centers (r=−0.8; P=0.005). Over time, as a center transitioned to increasing TR experience (r=0.9; P=0.001), a concomitant decrease in radiation exposure occurred (r=−0.8; P=0.006). Ultimately, when a center's balance of TR to TF procedures approaches 50%, the resultant radiation exposure was equivalent.ConclusionsThe TR approach is associated with a modest increase in patient radiation exposure. However, this increase is eliminated when the TR and TF approaches are used with equal frequency—a guiding principle for centers adopting the TR approach.

Hand dysfunction after transradial artery catheterization for coronary procedures

“…Nevertheless, despite of its clear advantages over TFA, TRA is not without limitations and is associated with longer operator learning curve[ 15 , 16 ], increased radiation exposure in individual operators at the start of their learning curves[ 17 , 18 ] and higher case radial proportion to translate the better results of randomized trials into clinical practice[ 11 , 19 , 20 ]. Moreover, vascular complications such as RAO[ 21 ] and radial artery spasm[ 22 ] are not uncommon and very recently concerns have been raised that patients undergoing TRA PCI may encounter hand dysfunction[ 23 ].…”

What's in a Day? A Guide to Decomposing the Variance in Intensive Longitudinal Data

“…This small but statistically significant negative inertia indicates negative carry-over of affect from moment to moment, which is theoretically not very plausible. A possible explanation is that, as we mentioned above, inertia estimates are negatively biased in a centered AR model (Hamaker and Grasman, 2014 ). In any case, the estimate is at least closer to the true value (zero) than the estimate from the two-level AR(1) model.…”

“…Values of ϕ i closer to one indicate higher inertia, which implies strong carry-over of affect, whereas ϕ i values close to zero indicate weak carry-over. Values between zero and minus one are possible, but not expected when studying affective inertia (Hamaker and Grasman, 2014 ), and values larger than one would imply that the process is not stationary.…”

“…We prefer the centered model notation as it allows us to put a normal distribution on the person's trait level of affect, and, optionally, to predict this trait level using other person-level variables. This makes substantive sense, whereas the uncentered model formulation would put the normal distribution on the substantively uninteresting intercept a i , which seems rather arbitrary and which has the undesirable implication that the trait level itself is assumed to be non-normally distributed in the population (Hamaker and Grasman, 2014 ). For these reasons, the centered model notation is preferable, but it has the downsides that we need to have an estimate of each person's mean μ i beforehand, and that the estimate of the average (fixed) inertia will have a negative bias, as demonstrated by Hamaker and Grasman ( 2014 ).…”

“…This involves using some estimate for each person's trait level μ i around which their predictor is centered. One option is to use the sample mean (i.e., the observed mean of each person's time series) but here we use the empirical Bayes estimate, which results in slightly less bias and higher coverage rates for the fixed inertia parameter (Hamaker and Grasman, 2014 ).…”