Reply

Authors


To the Editor:

We thank Filippou and colleagues for their comments. We agree that the level of concordance we achieved in our study was strikingly high, and that the reasons for this might provide insight into how best to approach the use of musculoskeletal US for the assessment of gout.

The authors correctly note that our image interpretations were performed on static images and not at the time of acquisition. Separating the activities of scanning and image interpretation may have a beneficial effect on the ability of the reader to apply an objective protocol to interpretation. We agree that standardizing and evaluating the interreader and intrareader consistency of image acquisition are important, but we also believe these would be better served by a separate study. Moreover, the fact that all images were acquired by a single ultrasonographer using uniform patient and transducer positioning likely resulted in consistent views and image quality that allowed for a more homogenous interpretation of images. Our investigators scored all of the images without any knowledge of the corresponding patients' crystalline status (gout/hyperuricemia/healthy) by assigning each patient a number and shuffling the order of images before evaluation. While the ultrasonographer in our study was not blinded to the diagnosis, the standardized positioning of the patients and the probe likely was sufficient in preventing bias in image acquisition that could have altered the results of the study.

The authors also correctly note that our exclusion of patients with other forms of arthritis from the control group may have made it easier for the readers to identify gout, since the task at hand was largely one of determining gout versus no gout (or really, crystal deposition versus no deposition). As this suggests, US may be more useful in gout assessment when the clinical judgment has already been reduced to a simple yes/no determination, rather than asking the ultrasonographer/reader to identify an unknown diagnosis based strictly on clinical appearance. Such an interpretation essentially proves the importance of understanding pretest probability, and in this regard US may be no different than any other modality. One lesson from our study may therefore be applying clinical information to the fullest extent possible in order to increase the power of the US study by narrowing the scope of the question. Conversely, our assessments in this study were related only to urate deposition, not to the presence of synovitis or other features of gout that may readily be found in other diseases. In this regard, the main confounding diagnosis would have been chondrocalcinosis, which also presents as linear hyperechoic (bright) signals along (but below the surface of) the cartilage. Patients with known chondrocalcinosis were excluded from our study; nevertheless, chondrocalcinosis was detected incidentally, and by both readers, with no confusion as to its distinction from gout.

Finally, Filippou et al inquire about the background training of our US readers and suggest that readers with common training and/or experience may be more likely to produce concordant readings; we agree with this suggestion completely. In our study, the 2 readers had both similarities and differences in their skills and experience. One reader had 4 years of experience with musculoskeletal US and has served as an instructor at the American College of Rheumatology biannual US courses designed specifically for rheumatologists. The other reader had <1 year of US experience, but spent time with an expert who has published studies on US in crystal deposition disease in order to learn US techniques and how to review images prior to initiating the study. Our 2 readers prepared for the study by reviewing images both separately and together prior to the study to establish common guidelines for interpretation. The study images were then read separately and in a blinded manner. Therefore, although the 2 readers had very different levels of experience, we made efforts to homogenize their approach prior to the actual study. While such efforts may not always be possible in clinical practice, they testify to the importance of common standards, and the importance of training according to such standards, to maximize the interpretability of this modality.

Rennie G. Howard MD*, Michael H. Pillinger MD*, Soterios Gyftopoulos MD*, Jonathan Samuels MD*, Ralf G. Thiele MD†, Christopher J. Swearingen PhD‡, * New York University Hospital for Joint Diseases, New York, NY, † University of Rochester, Rochester, NY, ‡ University of Arkansas, Little Rock, AK.

Ancillary