Does Immobilization Post Injection Reduce Contrast Extravasation in MR Arthrography of the Shoulder?

We have read with interest the manuscript entitled ‘‘Does Immobilization Post Injection Reduce Contrast Extravasation in MR Arthrography of the Shoulder?’’ by Muylaert et al. which is one of few articles assessing measures to reduce iatrogenic extravasation from theglenohumeral joint following arthrography prior to MR. The authors should be commended for contributing to this currently sparse area of research, particularly in their efforts to test for a correlation between extravasation and image quality. However, one could legitimately consider whether the potential impact of extravasation on image interpretability is substantial enough to warrant investigation and procedural changes. Multiple systematic reviews and meta-analyses have demonstrated that MR arthrography has a high baseline sensitivity and specificity in the diagnosis of several lesion types, most recently labral lesions, regardless of the presence of extravasation. Furthermore, the risk of iatrogenic extravasation resulting in the misdiagnosis of lesions is reduced by current knowledge of imaging features that distinguish between clinically significant (e.g. the J sign associated with HAGL lesions) and insignificant leakage. This prospective study by Muylaert et al. has several design strengths. The subjects came directly from a relevant patient population and were randomized to either the intervention or control group, and the design controlled for age and gender. The 3 T MRI platform used is state of the art. The authors minimized confounding by assessing for the J sign and immediate post-fluoroscopy extravasation prior to MRA imaging. All participants received the intervention they were assigned to and no dropouts occurred during the study. During initial fluoroscopy, contrast was injected into participants’ anteroinferior glenohumeral quadrant. The article referenced evidence in support of rotator interval and posterior injection approaches but argued that these methods also had high rates of leakage, in their clinical experience. Current evidence does indeed suggest a high frequency of extravasation with these approaches, though there is some disagreement in the literature on which is comparatively greater. In the absence of a gold standard for grading the extent of extravasation, the authors created an ordinal five-point scale (1: none, 2: less than 2 cm, 3: 2-5 cm, 4: 5-10 cm, 5: more than 10 cm) based on their clinical observations. The validity of this scale was not quantified, but it demonstrated a strong interrater reliability (0.81). Image quality was also graded on an ordinal five-point scale (1: very poor, 5: very good) and demonstrated only a moderate interrater reliability (0.49). Although not mentioned, it is assumed that the assessors were blinded when grading images, so the presence of observation bias is unlikely. Muylaert et al. discovered no significant difference in extravasation or image quality between the interventional and control groups, though this may have been due to limitations in the data analysis. A Chi-square test was used to measure between group differences, which was the appropriate choice for the variables measured. However, given that extravasation was originally measured in centimeters, another option would have been to present the data as a set of continuous variables which provide more complete information than ordinal or dichotomous ones. Therefore, a t-test or Wilcoxon rank sum test would have higher power than a Chi-square test when assessing the same sample and may reveal a statistically significant difference that would otherwise go unnoticed. Likewise, a valid Chi-square test should ideally have a sample size >60, whereas the sample size in Muylaert et al. is only 50. The discussion section of the paper thoroughly assessed the limitations in its methodology and provided hypotheses for the mechanisms of iatrogenic extravasation. One paragraph mentioned that extravasation may be related to the fluid pressure surpassing the physiologic joint capacity, as seen in large elbow or knee effusions, which has been corroborated by others. Of interest, the current authors stated that their team recently transitioned to using a rotator interval approach as they found it ‘‘faster, less painful, and overall easier’’ with which some authors agree. Again, we thank the authors for their insightful contribution to the MR arthrography literature.