Molecular Testing Predicts Incomplete Response to Initial Therapy in Differentiated Thyroid Carcinoma Without Lateral Neck or Distant Metastasis at Presentation: Retrospective Cohort Study

Introduction

For patients with differentiated thyroid carcinoma (DTC) who present with lateral neck disease or distant metastasis, total thyroidectomy (with or without therapeutic neck dissection or adjuvant radioactive iodine [RAI]) is uniformly recommended because of the intrinsically high risk for recurrence. Fortunately, most patients present with less extensive disease and can thus be considered for less extensive treatment, including thyroid lobectomy alone. ^1,2

In cases that do not clearly fall within the bookends of the recurrence risk continuum, treatment decisions are more nuanced and use risk stratification schemas, such as the 2015 ATA Risk Stratification System (RSS), to predict recurrence and inform management. While surgical pathology results represent a key snapshot of cancer biology and thus should factor heavily into risk prediction schemas, such reliance on surgical pathology can theoretically subject nearly half of patients to two operations based on newly identified pathological information not known preoperatively. ^{3 –5}

For example, a patient with a 2-cm indeterminate nodule on cytology found after diagnostic lobectomy to be a follicular thyroid carcinoma with microscopic extrathyroidal extension and extensive angioinvasion after diagnostic lobectomy would likely require completion thyroidectomy to facilitate adjuvant RAI and/or surveillance—compared with if the nodule were a 2 cm completely encapsulated follicular variant papillary thyroid carcinoma. Understanding which patients are more likely to recur before any treatment begins may help to expedite care, ensure that the first surgery is the last one, and minimize the risk for recurrence.

Molecular testing (MT) of fine needle aspirates (FNA) improves the diagnostic yield of detecting cancer in indeterminate thyroid nodules. ^{6 –8} Recent studies have also demonstrated links between MT results and surgical pathology, ^9,10 and early recurrence. ^{11 –16} Among patients eligible for either lobectomy or total thyroidectomy as their initial therapy, our study aims were to assess (1) if conventionally available preoperative factors are associated with incomplete response to initial therapy, and (2) if MT results can be a surrogate for the RSS to estimate risk of recurrence. In this way, we eventually anticipate being able to move the ability to risk stratify to a timepoint before any treatment begins.

Materials and Methods

Consecutive patients who underwent index thyroidectomy (lobectomy or total) for any clinical indication and had pathologically identified primary DTC (papillary, follicular, oncocytic/Hürthle cell) between November 1, 2017 and October 31, 2021 were abstracted from the electronic health record to conduct this retrospective cohort study. Patients with medullary thyroid cancer or noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) were not abstracted. Patients who had initial thyroid surgery earlier or elsewhere for benign disease and then completion thyroidectomy for thyroid carcinoma during the study period were not abstracted. Patients who had previous thyroid surgery earlier or elsewhere for DTC and then required reoperation during the study period were also not abstracted. In cases of multifocal cancer, the thyroid cancer type with the most aggressive histology was recorded, as this dictated clinical care. The University of Pittsburgh Institutional Review Board approved this study (QIIRB No. 1055).

Patients who had lateral neck nodal disease or distant metastasis at presentation were excluded as they represented the highest risk patients (i.e., total thyroidectomy necessary). Patients who had carcinomas smaller than 1.0 cm and low risk according to the RSS were also excluded as they represented the lowest risk patients (i.e., lobectomy alone appropriate; Fig. 1).

OPEN IN VIEWER

FIG. 1.

Participant flow diagram to create the study cohort. ATA, American Thyroid Association; DTC, differentiated thyroid carcinoma.

Indications, preoperative evaluation, extent of surgery, and postoperative cancer management followed the 2015 ATA and 2020 AAES Thyroidectomy Guidelines. ^1,2 The dates of surgery and last clinical encounter were used to calculate follow-up.

Results

Recurrence

Over median follow-up of 18 months (IQR 9–30), 50 (5.3%) patients developed a biochemical (n = 18) or structural (n = 32) recurrence. Patients with recurrences had longer median follow-up (22.5 vs. 17 months, p = 0.006). Patients who recurred were more likely to be men (p = 0.02) with larger tumors (p = 0.03). Age (p = 0.89), preoperative levothyroxine use (p = 0.19), and Bethesda cytology category (p = 0.57) were not associated with recurrence (Table 2). As expected, patients who had more extensive operations were more likely to recur, as were those who received RAI (72.0% vs. 37.4%, p < 0.001). Aggressive histological features, such as extrathyroidal extension (p < 0.001) and angioinvasion (p < 0.001), were more often found in cancers that recurred (Table 3). Structural recurrences were identified in the central compartment (n = 13), lateral compartment (n = 12), and distant sites (n = 7).

There were 36 (72.0%) patients who recurred after total thyroidectomy, of which 21 (58.3%) were structural and 15 (41.7%) biochemical. There were 12 (24.0%) patients who recurred after completion thyroidectomy, of which 9 (75.0%) were structural and 3 (25.0%) biochemical. There were two patients who developed structural recurrences after lobectomy. No patients were identified to have a biochemical recurrence after lobectomy alone.

Recurrences were detected in 37 (4.4%), 12 (12.0%), and 1 (20.0%) patient(s) with AJCC stage I, II, and III disease, respectively (Fig. 2A). In the patient with stage III disease, recurrence was structural. Recurrences were detected in 17 (2.9%), 20 (6.7%), and 13 (22.8%) patients with RSS-low, -intermediate, and -high cancers, respectively (Fig. 2B). RSS-high patients recurred as structural in 10 (76.9%) and biochemical in 3 (23.1%). Both AJCC staging and RSS categories were significantly correlated with recurrence in an incremental stepwise manner (p < 0.001 for both).

OPEN IN VIEWER

FIG. 2.

(A–C) Observed recurrence rates across AJCC stage (A), ATA Risk Stratification System (B), and MRGs (C). AJCC, American Joint Committee on Cancer; MRGs, Molecular Risk Groups.

Among patients who had MT (n = 440), recurrences were detected in 4 (1.8%), 8 (4.7%), and 11 (23.4%) patients with MRG-low, -intermediate, and -high cancers, respectively. MRG categories significantly correlated with recurrence in an incremental stepwise manner (p < 0.001; Fig. 2C). All MRG-high cancers had structural recurrence, whereas recurrences within the MRG-low category were predominantly biochemical (Supplementary Fig. S1).

RFS was similar between RSS (Fig. 3A) and MRG (Fig. 3B) categories, suggesting both have similar ability to risk stratify. The 36-month RFS for RSS-high patients was lower than RSS-intermediate patients (47.5% vs. 92.2%, p = 0.004), while the 36-month RFS for RSS-intermediate patients was no different than RSS-low patients (92.2% vs. 94.9%, p = 0.07). The 36-month RFS for MRG-high patients was lower than MRG-intermediate patients (53.4% vs. 91.1%, p < 0.001), while the 36-month RFS for MRG-intermediate patients was no different than MRG-low patients (91.1% vs. 96.4%, p = 0.37).

OPEN IN VIEWER

FIG. 3.

(A, B) Recurrence-free survival among patients who had molecular testing (n = 440) stratified by ATA RSS categories (A) and by MRGs (B). ATA RSS, American Thyroid Association Risk Stratification System.

Discussion

Given that ∼20–30% of thyroid nodules are indeterminate on cytology and that it is currently considered appropriate to perform either a lobectomy or total thyroidectomy for DTCs between 1 and 4 cm, completion thyroidectomy may be necessary based on pathological information. ^{1,3 –5} With >100,000 thyroidectomies performed annually in the United States, ²⁵ this problem is significant. Knowing which patients in the preoperative setting can be safely triaged to less extensive treatment commensurate with the intrinsic risk for disease recurrence would greatly facilitate high-quality personalized patient-centered care. ²⁶

In this study of 945 patients with histological DTC where we focused on those without lateral neck or distant metastasis based on routine preoperative staging imaging who could be managed with either lobectomy or total thyroidectomy, two-thirds were classified by the existing ATA system as RSS-low risk for recurrence. Even still, incomplete responses were observed across all RSS categories in short-term follow-up. Tumor size was the only conventional preoperative factor associated with recurrence and was associated with a 30% increased risk per centimeter increase in tumor size. After MT results were incorporated in the form of MRGs, model discrimination improved by 27%. This improvement in estimating recurrence risk was not significantly different than when the RSS categories were incorporated, suggesting that the MRGs, which can be known preoperatively, carry predictive capability similar to disease-specific staging characteristics that are typically obtainable only after initial operation.

Numerous prior studies have attempted to identify factors predictive of thyroid cancer recurrence. ^27,28 But they all rely on histopathological factors to achieve enough predictive power to be clinically applicable since conventional preoperative factors are limited. We studied only five factors in the preoperative setting as other factors, such as race/ethnicity, body mass index, smoking status, Hashimoto's thyroiditis, personal history of radiation therapy in childhood, or family history of thyroid cancer, have been studied in detail and have not been associated with thyroid cancer recurrence, statistically or conceptually. ^29,30 We found that although male sex and tumor size were significantly associated with recurrence only tumor size remained significant in multivariable analyses.

MT results were then examined because it is another preoperative variable that is emerging as a useful prognosticator. Previous studies have reported that specific molecular profiles are associated with key histopathological features linked to disease aggressiveness and recurrence, and, separately, independently with recurrence and distant metastasis. ^{10,14 –16,22,31} In our prior study examining a cohort inclusive of all DTC with MT information available, we detected an incremental increase in recurrence risk with each MRG category. ²² To expand on this study, we sought to directly compare the ability of the MRGs with the ATA RSS in predicting recurrence among patients eligible for either lobectomy or total thyroidectomy. If the MRGs could be a surrogate for pathological information, knowing the thyroid cancer molecular profile before initiating treatment may help to guide the extent of surgery in particular, among other benefits of this added information.

We observed that the MRGs remained a significant and independent risk factor for recurrence on multivariable analysis even in this more limited patient sample and improved predictive ability beyond tumor size alone. Because the MRGs can be readily known preoperatively, this finding suggests that tumor size and MRGs are two factors that could clarify recurrence risk before any treatment begins, including potentially the extent of surgery. MT thus offers an avenue to inform shared decision-making without relying solely on postoperative data. Similar to earlier studies, ^32,33 we confirmed the ability of the RSS categories to predict recurrence, which is the current gold standard method by which thyroid cancer care is delivered.

Moreover, our findings demonstrate that the MRGs are statistically equivalent to the RSS categories in their ability to predict recurrence risk. When we tested this finding by taking another approach to examine the interplay between RSS categories and MRGs by constructing a model that included both RSS categories and MRGs simultaneously (i.e., Model 4), only the MRGs remained significantly associated with recurrence. Taking the findings of Models 2, 3, and 4 together, the MRGs are equally accurate as the RSS categories at predicting recurrence, if not better, within these data. The statistical findings underscore clinical observations that interaction between the MRGs and RSS categories likely exists.

However, until additional validation data are available, the MRGs should not completely replace the gold standard method in routine thyroid cancer care. It is important to stress that testing for a limited number of molecular markers (e.g., BRAF^V600E or RAS mutations only) is insufficient for cancer prognostication as it requires knowing the mutational status of multiple genes, including TERT and TP53. However, even when large genetic panels are used, a combination of MT and clinical findings using the armamentarium of available tools (e.g., preoperative imaging and pathological findings) may provide the most comprehensive prediction of recurrence to improve care. As another tool, comprehensive MT results may potentially move the ability to risk stratify to a timepoint before any treatment begins.

We observed a 13–14% rate of requiring completion thyroidectomy in this cohort. While this is lower than the theoretical rate of 40–50% assessed in retrospective studies, this rate is consistent with the reported completion thyroidectomy rate of 12–25% after publication of the 2015 ATA Guidelines. ^{3 –5,34,35} Reasons to proceed with a completion thyroidectomy are not always apparent in review of medical records, which can include patient preference, clinician preference, or oncological considerations. Further prospective studies are needed to better understand the molecular, patient-specific, and cancer-related variables that would possibly allow further reduction in the need for completion thyroidectomy.

This investigation has several additional limitations. First, as a retrospective study, selection bias is probable and additional unknown preoperative factors predictive of recurrence may exist. For example, ultrasound characteristics were excluded because results were not uniformly reported during the study period and the ability of ultrasonography to provide reliable prognosticative discrimination between low- and intermediate-risk disease is variable. ^{17 –19} In addition, we did not include socioeconomic variables in the assessment although this would be important to consider for future analyses.

Second, this study was conducted at a single high-volume institution in the United States thus possibly limiting generalizability. A split-sample internal validation approach was not taken due to the small sample size of our study and would not have improved our ability to answer the study question. Indeed, it is important to recognize that the statistical ability of the RSS categories and MRGs to correctly categorize patients with and without recurrence is limited to this patient cohort. Larger sample sizes, other settings, or longer follow-up may alter these findings and explain why there is not a greater separation in RFS between the intermediate- and low-risk groups within both the RSS categories and MRGs. Future multi-institutional studies are certainly needed to externally validate findings.

Third, pathologists were not blinded to MT results, when present. Nevertheless, pathologists followed CAP standards when evaluating surgical specimens. ²⁴ Fourth, we could not ascertain whether MT results factored into the decision-making for completion thyroidectomy in these patients or how MT informed the initial extent of surgery. Fifth, patients who recurred were followed for longer and thus the true recurrence rate may be underestimated. Patients with low-risk tumor features were possibly monitored less intensely resulting in incomplete capture of follow-up.

Last, overall follow-up was short and thus biochemical recurrences were included, which accounted for a significant proportion (36%) of the overall recurrences in this study. Biochemical recurrences may never result in actionable recurrences and thus recurrence may be overestimated. Longer follow-up of this cohort is needed.

Conclusions

In this retrospective cohort study, MT results improved preoperative risk prediction beyond the single conventional preoperative factor of tumor size. Furthermore, the predictive ability of the MRGs was similar to the RSS categories, suggesting that the MRGs could be a valuable surrogate for histology to accurately inform prognosis.