VICTORIA, B.C. – Ultrasound risk criteria for adults are no match for the training, skill, and gut instinct of an experienced radiologist when evaluating pediatric thyroid nodules, results of a retrospective cohort study reported at the annual meeting of the American Thyroid Association suggest.
“In 2015, the ATA commissioned a pediatric task force that developed valuable guidelines specific to our pediatric patients. These guidelines recommend performing an FNA [fine-needle aspiration biopsy] in any nodule with a concerning clinical history or a concerning ultrasound feature,” commented first author Ana L. Creo, MD, a pediatric endocrinology fellow at the Mayo Clinic in Rochester, Minn.
She and her colleagues analyzed findings from diagnostic ultrasound in 112 patients aged under 21 years who had 145 thyroid nodules that were ultimately assessed histologically or cytologically.
Results showed that the radiologists’ overall impression and the ATA risk-stratification system for adults had the same high sensitivity, picking up 9 out of 10 malignant cases, she reported. But the radiologists’ overall impression had much higher specificity, correctly classifying 8 out of 10 benign cases, versus about 5 out of 10 for the risk stratification.
“These findings may have implications in trying to avoid unnecessary FNAs, particularly in our population,” Dr. Creo summarized. “Our million dollar question is trying to get in the heads of the radiologists to figure out what really goes into that overall impression. And if we can apply a specific score to that, I think that would be most clinically useful.”
“Based upon these results, further work is needed to determine the usefulness of the adult ATA ultrasound risk stratification in children, moving towards an ultrasound-based stratification system specific to our pediatric patients,” she concluded.
One session cochair, Catherine A. Dinauer, MD, a pediatric endocrinologist and clinician at the Yale Pediatric Thyroid Center, New Haven, Conn., commented, “It seems as though we’re pretty good at picking up which nodules are malignant, but I still feel like so many of the nodules that are benign are suspicious by ultrasound. Trying to tease out what is it about those benign ones may allow us to figure out in which ones we could avoid biopsy. That’s where we see we are not that good at it.”
Nodule attributes that might help in this regard include subtypes of microcalcifications, irregular margins, and position of the nodule in the gland relative to the skin, she proposed.
The other session cochair, Yaron Tomer, MD, chair of the department of medicine and the Anita and Jack Saltz Chair in Diabetes Research at the Montefiore Medical Center, New York, stressed knowing one’s radiologist and questioned the generalizability of the findings.
“You have to know your own radiologist well and how well you can trust them. Probably, the investigators chose some of the top radiologists in their institution and maybe even in the nation, so we have to be careful as to whether this applies to places that don’t have access to such great radiologists,” he commented. “But I think even if the guidelines are not perfect, they are the best we have right now.”
The investigators studied pediatric patients (mean age, 15.5 years) with thyroid nodules who underwent initial ultrasound at the Mayo Clinic during 1996-2015, had at least a year of follow-up, and for whom histology or cytology results were available. Those with a known genetic tumor syndrome or a history of radiation exposure were excluded.
Two blinded radiologists assessed nodule ultrasound features using the Thyroid Imaging and Reporting Data System (TIRADS) () and then rendered their overall impression: malignant, indeterminate, or benign.
Next, an independent reviewer assigned each nodule an ATA adult risk category (Thyroid. 2016;26:1-133): high, intermediate, low, or very low suspicion.
Finally, both measures were compared against the reference standard of the nodule’s histology or cytology results.
Ultimately, 34% of the patients had malignant nodules, Dr. Creo reported. “This is likely quite a bit elevated from the true prevalence due to our intentional study design requiring follow-up, likely excluding some patients with benign nodules,” she commented.
Patients with benign and malignant nodules did not differ significantly on any of a variety of sociodemographic and clinical factors, such as family history and mode of detection.
For comparison of sensitivity, the investigators combined the ATA risk categories of high and intermediate suspicion and combined the overall radiologists’ impression of malignant and indeterminate. “We felt this best answered the practical clinical question of how many malignant nodules would be missed if FNA was not performed, assuming FNA would typically be performed if the ATA risk stratification was high or intermediate or if the radiologist’s overall impression was malignant or indeterminate,” Dr. Creo explained.
Results here showed that the ATA risk stratification and the radiologists’ overall impression had the same high sensitivity of 90%.
For comparison of specificity, the investigators compared the ATA risk category of high suspicion with the radiologists’ overall impression of malignant.
Results showed that the overall impression had specificity of 80%, whereas the risk category had a specificity of only about 52%. Findings were similar when analyses instead used ATA high suspicion and intermediate suspicion combined.
“The key ultrasound characteristics that drove the diagnosis included having a solid component, calcifications, irregular margins, and hypoechogenicity – all similar to those seen in pediatric studies and similar to those in adult studies as well,” Dr. Creo noted.
Compared with benign nodules, malignant nodules significantly more often had a greater than 75% solid component (84% vs. 64%; P = .01), contained calcifications (60% vs. 18%; P less than .0001), had irregular margins (70% vs. 46%; P = .0073), and were hypoechogenic (74% vs. 51%; P = .0073). Notably, size and presence of halo did not differ significantly.
“Our study adds to previous work in that it had a relatively large pediatric sample size, used strict inclusion criteria with at least a year of follow-up to increase the validity of the diagnosis, and had precise definitions of the ultrasound features,” concluded Dr. Creo, who disclosed that she had no relevant conflicts of interest.
At the same time, the study had limitations, such as its use of a referral population, likely loss to follow-up of some patients with benign nodules, and possible clustering effect. “Lastly, we had the benefit of extremely experienced pediatric radiologists, and their overall diagnostic accuracy may not universally apply across all radiologists,” she said.