Thyroid Nodules with Indeterminate Cytology: Utility of the American Thyroid Association Sonographic Patterns for Cancer Risk Stratification

Abstract

Background: The 2015 American Thyroid Association (ATA) guidelines recommend using a classification based on sonographic patterns to set the size threshold for biopsies. Each pattern is associated with a distinct estimated rate of malignancy that it was hypothesized should stratify the risk of malignancy of cytologically indeterminate thyroid nodules (ITNs).

Methods: Ultrasound images of 463 ITNs (38% atypia/follicular lesions of undetermined significance; 62% follicular neoplasms) with histological follow-up consecutively evaluated between October 2008 and June 2015 at the authors' academic cancer center were independently evaluated by three observers and classified into one of the five sonographic patterns proposed by the ATA. Nodules with sonographic patterns not defined in the classification were grouped into a non-ATA pattern category. Differences in clinical and histological findings between the sonographic patterns were assessed. The prevalence of malignancy and odds ratio for malignancy were calculated for each sonographic pattern (low and intermediate patterns were collapsed for the analysis).

Results: The distribution of size and cytological diagnosis was significantly different between sonographic patterns (p < 0.001). The overall rate of malignancy was 27%. The rate of malignancy for the very low, low/intermediate, high, and non-ATA patterns were 0%, 19%, 56%, and 36%, respectively, and were all significantly different. Compared to the low/intermediate suspicion patterns, the odds ratios for malignancy were 2.35 for the non-ATA and 5.18 for the high suspicion patterns (p < 0.001). The odds ratio of the non-ATA pattern was 0.45 over the high suspicion pattern (p = 0.04). Results were similar in both cytological categories and for each observer separately. Sonographic patterns were associated with distinct histopathological profiles (p < 0.001).

Conclusions: ATA sonographic patterns are associated with distinct clinical features and pathological outcomes, and effectively stratify the cancer risk in ITNs. Thus, the ATA sonographic patterns should be used not only to set the size threshold for biopsy, but also to personalize management after the biopsy.

Keywords: noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP); thyroid cancer; thyroid cytology; thyroid nodules; thyroid ultrasound.

FIG. 1.

Cohort selection process. AUS/FLUS, atypia/follicular lesion of undetermined significance; FN/HCN, follicular/Hürthle cell neoplasm; ITNs, cytologically indeterminate thyroid nodules (AUS/FLUS and FN/HCN); US, thyroid ultrasound.

FIG. 2.

Distribution of histological diagnoses by sonographic pattern. (A) Distribution of histological diagnosis by Bethesda category and sonographic pattern. (B) Distribution of sonographic patterns by histological diagnosis. (C) Proportion of malignancies classified as low risk, intermediate risk, or high risk by sonographic pattern (excluded two intrathyroidal parathyroid carcinomas; see Methods section for definition of groups). FTC/HCC, follicular thyroid carcinoma/Hürthle cell carcinoma; FVPTC, follicular variant of papillary thyroid carcinoma; HP/AN/CLT, hyperplastic/adenomatous nodule or chronic lymphocytic thyroiditis; NIFTP, noninvasive follicular thyroid neoplasm with papillary-like nuclear features; Other PTC, other variants of papillary thyroid carcinoma different from FVPTC. ^aIncludes one intrathyroidal parathyroid adenoma. ^bIncludes two intrathyroidal parathyroid carcinomas, three medullary thyroid carcinomas, and one poorly differentiated thyroid carcinoma. Due to small cell counts, to compare the distribution of the histological diagnoses, NIFTPs and FVPTCs were collapsed into one group, and all other invasive cancers were collapsed into another group. Differences between AUS/FLUS and FN/HCN were statistically significant (p < 0.001). The distribution of histological diagnoses between sonographic patterns was also significantly different (p < 0.001 with Fisher's exact test for count data with a simulated p-value based on 2000 replicates). The low and intermediate suspicion patterns were collapsed into a single group for this analysis.