Extending the Clinician's Hand Through thyroid Ultrasound - Do it Yourself-for the Uninitiated Busts Expertise?

The last decade has seen significant advances in the decision-making processes that guide the management of nodular thyroid disease.[1] A significant advance has been the availability of high-resolution ultrasound[2] with standardized reporting using either the Thyroid Imaging Reporting and Data System (TIRADS)/American Thyroid Association risk category.[3] Guided aspiration and the use of Bethesda system of reporting have further enhanced our ability to guide shared decision-making.

The ultrasound may be considered as an extension of the history and physical examination. An ultrasound performed in the office as part of the clinical examination is now part of specialized practices with some evidence to support its credibility.[4] Besides the time saving and cost-effectiveness that appear to be obvious benefits, a single operator performed ultrasound reduces interobserver variability that appears to plague ultrasonography done in the community often not by the same radiologist and increasingly by radiology technicians and reported by radiologists. Except for size and vascularity, interobserver agreement appears to be slight to fair for most malignancy-associated features. On the other hand in competent hands, intraobserver agreement is significant for most features[56] making a compelling argument for ultrasound performed by the clinician in the office.

The converse is the argument that a trained radiologist has expertise that do it yourself-for the uninitiated clinicians can’t match. The burden of proof is to demonstrate that clinician-performed ultrasounds are noninferior to radiologist-performed ultrasounds (RPUSs) in detecting malignancy. In this issue of the journal, Mohanapriya and Chandrasekaran[7] present data supportive of the use of the ultrasound in the surgeon's office to evaluate and provide decision support for the management of thyroid nodules. In this large series with surgical correlation, surgeon-performed ultrasound (SPUS) had greater negative predictive value (NPV) (98 V 83.8) than RPUS. Interestingly, it outperformed fine-needle aspiration cytology (FNAC) (NPV of 81.08) in this regard. Impressive indeed!.

Despite this, the cautious optimist in me will hasten to point out that the proverbial deck was loaded in favor of the surgeon. The SPUS was performed by a single surgical resident while the RPUS was performed by rotating radiology residents; the FNAC was done by pathology residents. Clearly, an unequal comparison given that competency of trainees is significantly different from an experienced radiologist. One might argue – that SPUS does exactly this – it mitigates the variability in competencies in the community; many of us do not have the luxury of having a radiologist who specializes in thyroid ultrasound at our beck and call.

This study is consistent, and in some ways, an improvement on other published studies of SPUS by single operators.[8] It must be pointed out that performing and interpreting thyroid ultrasound is not a “conference acquired skill.” It requires training validation and repeated performance to maintain expertise; the learning curve is fairly steep. Legal restrictions on owning an ultrasound and operating one in India do not advance the cause of enthusiasts who dream of tab-connected ultrasound probes in their offices. Having said that it is a skill worth carrying in one's quiver.

Mohanpriya et al. and others also provide the scientific rationale to incorporate training in thyroid ultrasound and FNAC in the postgraduate curriculum. The value of acquiring these skills by the endocrinologist in training cannot be overemphasized.

The game changer of-course will be the incorporation of machine learning algorithms/artificial intelligence (AI) to disrupt thyroid sonology. Initial results sound like the first chess matches between Kasparov and Deep Blue-sensitivity for the AI system similar but lower specificity.[9] This will only get better if we go by the experiences in other areas. To paraphrase Christensen, thyroid ultrasound is ripe for disruptive innovation by machine learning; the benefits – not missing malignancy and avoiding unnecessary surgery – are too important to pass on.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.