Non-Iodine-Avid Disease Is Highly Prevalent in Distant Metastatic Differentiated Thyroid Cancer With Papillary Histology

Abstract

Context

Patients with radioactive iodine (RAI) refractory metastatic differentiated thyroid cancer (DTC) have poor prognosis. Early identification of RAI refractoriness may improve care.

Objective

This work aimed to characterize DTC patients with distant metastases (DM) at diagnosis who presented with non–iodine-avid disease.

Methods

Retrospective analyses of DTC patients with DM at diagnosis who presented between 2012 and 2020 were performed. Iodine uptake in DM was correlated with tumor histology and mutational profile. The difference in uptake between BRAF^V600E-like (BVL) and RAS-like (RL) cancers based on insights from The Cancer Genome Atlas was evaluated.

Results

Among 78 patients, 48.7% had negative uptake in DM on the first posttherapy scan. Negative scans were highly prevalent in papillary thyroid carcinoma (PTC) with papillary architecture, PTC with BRAF^V600E mutation, and PTC with both BRAF^V600E and TERT promoter mutations (71.1%, 80.9%, and 100%, respectively). BVL and RL tumors exhibited distinct uptake patterns with negative scan prevalence of 76.9% and 14.3% (P = .005). Multivariate logistical regression confirmed high odds of negative uptake in BVL tumors with either BVL mutations or papillary architecture, 19.8 (95% CI, 2.72-144), and low odds of negative uptake in RL tumors with either RL mutations or follicular architecture, 0.048 (95% CI, 0.006-0.344), after adjusting for age, sex, race, RAI preparation method, bone metastases, and RAI dose. Patients with negative scans were significantly older (62.4 vs 47.0 years, P = .03).

Conclusion

Among DTC patients with DM at diagnosis, non–iodine-avid disease is highly prevalent in patients with BVL cancers, particularly with BRAF^V600E and TERT promoter mutations, and is associated with an older age. Better strategies are needed to improve RAI treatment response for these patients.

Differentiated thyroid cancers (DTCs) are cancers arising from thyroid follicular epithelial cells including papillary thyroid carcinoma (PTC), follicular thyroid carcinoma (FTC), and Hürthle cell carcinoma (HCC), accounting for about 84%, 2%, and 2% of all thyroid cancers, respectively (1-3). The prevalence of distant metastases (DM) at diagnosis in DTC is 1.2% to 2.2% (4, 5). Radioactive iodine treatment (RAI-T) is the mainstay of therapy in patients with DM, and prognosis depends on the response to RAI-T. Durante et al (6) demonstrated an estimated 10-year survival rate of 92% in metastatic thyroid cancer patients who achieved remission with RAI-T compared to an estimated 10-year survival rate of 10% in those without iodine uptake in the DM, comprising about 32% of their cohort. The authors found that 23% patients with PTC, 8% of patients with well-differentiated FTC, and 47% of patients with poorly differentiated FTC lacked iodine uptake after RAI-T. Since then, there have been a paucity of data examining the prevalence of non–iodine-avid disease in patients with DM from DTC (7).

Advancement in understanding the molecular pathogenesis of DTC has led to the identification of oncogenic drivers for most DTC and oncogenic events associated with dedifferentiation and iodine refractoriness (3, 8-12). Redifferentiation studies using targeted therapy drugs with the goal of increasing iodine avidity have shown promising results in patients with iodine-refractory DM (13-18). However, these patients frequently had undergone several rounds of failed RAI-T before redifferentiation therapy. Early identification of these patients may allow for more effective and individualized treatment strategies to optimize care.

The principal oncogenic drivers of DTC include mutually exclusive BRAF^V600E mutation, activating mutations in RAS, and receptor tyrosine kinase gene fusions resulting in activation of the mitogen-activated protein kinase (MAPK) signaling pathway (3, 19). The Cancer Genome Atlas (TCGA) classified PTC into 2 distinct tumor types based on multidimensional genomic data: BRAF^V600E-like (BVL) and RAS-like (RL) tumors. In general, BVL tumors exhibit papillary architecture, enriched with BRAF^V600E mutation and RET fusions that result in higher MAPK signaling leading to dedifferentiation, whereas RL tumors tend to be the follicular variant of PTC (FV-PTC) displaying follicular architecture, enriched with RAS mutations that result in lower MAPK signaling leading to less dedifferentiation (8). The findings from TCGA raised the question whether FV-PTC should be classified together with FTC; both have follicular architecture on histology and share similar mutational profiles (20-24). Although RL tumors are more differentiated than BVL tumors, no clinical studies have compared the iodine avidity of these 2 tumor types.

We retrospectively reviewed DTC patients with DM at diagnosis at our institution. Our main aim was to identify patients with radioactive iodine–refractory (RAI-R) disease at diagnosis in a well-differentiated DTC (wDTC) cohort by evaluating iodine uptake in the DM on the posttherapy scan after first RAI-T. Even with well-differentiated histology, with RAI-R disease, these patients are likely to have a worse prognosis with increased side effects and minimal benefits from repeated RAI-T and may benefit from individualized treatment strategies (6, 25, 26). We aimed to correlate iodine uptake with histology and molecular data. Furthermore, deriving insights from TCGA, we classified our cohort into BVL and RL tumors and explored the difference in iodine uptake between the 2 tumor types. Given that age is a recognized poor prognostic factor (27, 28, 29), we also evaluated whether age played a role in terms of iodine avidity.

Materials and Methods

Patient Selection

This retrospective study was approved by the institutional review board at the University of California, San Francisco (UCSF); informed patient consent was waived. Adult DTC patients (age ≥ 18 years) with DM seen at UCSF between June 2012 to June 2020 were identified through International Classification of Disease (ICD) - 9 and ICD-10 codes. Electronic medical records were reviewed, and data were extracted from clinical documentation, surgical pathology reports, radiology reports, and molecular testing results if performed. Patients with thyroid cancer other than wDTC were excluded (poorly differentiated thyroid carcinoma [n = 24], anaplastic thyroid carcinoma [n = 35], medullary thyroid carcinoma [n = 63], and thyroid carcinoma showing thymic-like differentiation [n = 1]). Only patients who had total thyroidectomy followed by RAI-T were included. Patients with wDTC and DM at diagnosis (n = 82) were identified by (1) biopsy-proven DM at the time of diagnosis (n = 18); (2), iodine uptake in the DM on the first posttherapy scans corroborated by subsequent clinical courses and cross-sectional imaging studies (computed tomography [CT], magnetic resonance imaging) or positron emission tomography CT scans) (n = 28): (3), persistent thyroglobulin (Tg) elevation greater than or equal to 10 ng/mL and rising Tg consistent with significant biochemical disease (30) after RAI-T without significant locoregional disease, with subsequent clinical courses and cross-sectional imaging consistent with DM (n = 11); and (4), cross-sectional imaging demonstrating structural disease at distant sites within 1 year of initial cancer diagnosis corroborated by subsequent clinical courses (n = 25) (30, 31).

Definition of Positive and Negative Iodine Scans

Iodine scan reports and clinic notes were reviewed to obtain the percentage uptake on the pretherapy scan, preparation method (recombinant human thyrotropin [rhTSH] or thyroid hormone withdrawal [THW]), therapeutic dose administered, and sites of uptake on the posttherapy scan. We categorized the first posttherapy scan results into negative or positive scans. A negative scan was defined as uptake limited to the neck only without uptake in the DM. A positive scan was defined as uptake in any DM. Diffuse liver uptake was not considered positive uptake; only focal liver uptake corroborated by cross-sectional imaging was considered positive uptake. Scans with faint and minimal uptake in the lungs were classified as negative scans as reviewed by an independent radiologist and nuclear medicine physician (R.R.F.) blinded to other clinical data. Uptake without correlating cross-sectional imaging findings and not consistent with a subsequent clinical course demonstrating presence of DM was also classified as negative.

Histology

DTC subtypes and PTC subtypes were recorded based on surgical pathology reports of the primary tumors or clinic notes when pathology reports were not available. PTC subtypes included classic variant (CV, n = 28), follicular variant (FV, n = 16), tall cell variant (TCV, n = 6), diffuse sclerosing variant (DSV, n = 4), hobnail variant (HV, n = 1), and columnar cell variant (CCV, n = 1). If the subtype of PTC was not reported, the histology was listed as an unknown variant (UV, n = 9). For FV-PTC, 2 subtypes were classified depending on the report: infiltrative FV (inFV, n = 5) and encapsulated FV (eFV, n = 4). When the subtype of FV-PTC was not specified, the pathology was listed as unknown FV (unFV, n = 7). Original surgical pathology slides of 4 patients with UV-PTC were available and reviewed by a pathologist (E.K.) blinded to other clinical data to render the histological diagnoses.

Molecular Testing and Genomic Profiling

Since 2009, it has been a routine practice to test for BRAF^V600E for PTC 0.4 cm or greater operated at UCSF, using polymerase chain reaction with allele-specific fluorescent probe melting curve analysis (until 2019) or next-generation sequencing of exon 15 of the BRAF gene (October 2019 onward) (32). Comprehensive genomic profiling was performed using commercially available cancer gene–targeted, capture-based, next-generation sequencing assays including UCSF500, Tempus xT, or Foundation One. These assays cover more than 300 genes and are designed to detect substitutions, insertions, deletions, and select gene rearrangements using DNA isolated from formalin-fixed, paraffin-embedded tumor tissue (33, 34).

Classification of BRAF^V600E-like and RAS-like Tumors and Iodine Uptake Between the 2 Groups

To compare iodine uptake status between BVL and RL tumors, PTC subtypes and FTC were divided into these 2 groups according to the mutational data from TCGA on PTC (model 1) (8). For additional analyses, according to the multidimensional genomic data from TCGA demonstrating phenotype and genotype correlation, that is, papillary architecture in BVL tumors and follicular architecture in RL tumors (8), we then added the tumors without mutational data and classified them into these 2 groups based on histology. Tumors with papillary architecture were added to the BVL tumors and tumors with follicular architecture were added to the RL tumors (model 2). Two CV-PTCs with NTRK fusions were excluded from this classification because these fusions were found neutral in the spectrum between BVL and RL tumors in terms of genomic signaling in TCGA (8, 35); 2 HCCs were also excluded from this analysis because of their unique genomic signature different from those of PTC and FTC (36, 37). There were 9 tumors with unknown PTC subtypes (UV-PTC), of which 3 had mutational data and 6 had no mutational data. These 6 UV-PTCs without mutational data were excluded from this classification. One TCV-PTC with the genetic duet of BRAF^V600E and TERT promoter mutations, 2 FTCs, and 1 CV-PTC were also excluded because of missing scan results.

Model 1

BVL tumors (n = 26) included 21 tumors harboring BRAF^V600E mutation and 5 tumors harboring RET fusions (3 NCOA4-RET and 2 CCDC6-RET). RL tumors (n = 7) included 3 PTCs s with NRAS mutations (2 FV-PTC and 1 CCV-PTC), 1 inFV-PTC with a FGFR2 fusion, and 3 FTCs with NRAS mutations.

Model 2

Tumors with papillary architecture but without mutational data included 8 CV-PTCs without any molecular or genomic testing, 4 CV-PTCs and 1 TCV-PTC negative for BRAF^V600E when only BRAF^V600E testing was performed, and 1 HV-PTC without testing, adding an additional 14 BVL tumors with papillary architecture to the 26 tumors with mutational data, leading to a total of 40 BVL tumors. Tumors with follicular architecture without known mutations or without genomic testing included 4 inFV-PTCs, 4 eFV-PTCs, 4 unFV-PTCs, 8 FTCs, and 1 FTC without a commonly known driver mutation (with multiple other mutations) and were classified RL, adding additional 21 tumors to the 7 tumors with mutational data, leading to a total of 28 RL tumors.

Numbers of negative scans between BVL and RL tumors both in model 1 and model 2 were compared; the association of having a negative iodine scan with each group in both models was assessed by logistical regression as described in “Statistical Analysis.” Because 10% to 30% of FV-PTCs may harbor BRAF^V600E, particularly inFV-PTC (8, 38), a sensitivity analysis was also performed for model 2 by reclassifying the 4 inFV-PTC as BVL and excluding 4 unFV-PTC, leading to 44 BVL tumors and 20 RL tumors.

Age and Iodine Uptake

Since CV-PTC and TCV-PTC represent the majority of PTC with papillary architecture, the mean age of CV-PTC and TCV-PTC patients with negative scans was compared with the mean age of those with positive scans. Because age did not meet the normality assumption owing to smaller number (n = 32), Kruskal-Wallis test was used for comparison of means. We also performed logistical regression adjusting for sex, race, method of RAI preparation, presence of bone metastases, and RAI dose.

Statistical Analysis

Cohort characteristics were described using mean with SD or median (25th-75th percentiles) as appropriate. We compared the percentage of negative scans and characteristics of those with BVL and RL tumors (model 1 and model 2) using chi-square, Fisher exact, t test, analysis of variance, or Kruskal-Wallis tests as appropriate.

To assess the association of BVL and RL tumors and iodine avidity, we used univariate and multivariate logistical regression to study the relationship between our predictors: BVL or RL tumors and negative scans in the DM both for model 1 and model 2. Our multivariate logistical regression model was adjusted for potential confounders including age, sex, race, method of RAI preparation (rhTSH vs THW), presence of bone metastases, and RAI dose. We interpreted the results of all logistical regression analysis using odds ratio (OR). A 2-tailed nominal P value of .05 was considered to indicate statistical significance. All statistical analyses were performed using STATA, version 14.1 (STATA Corp).

Results

Demographics and Clinical Characteristics of the Cohort

We identified 82 wDTC patients with DM diagnosed at the time or within 1 year of cancer diagnosis. The mean age was 56.2 ± 17.3 years and 39% were men. PTC constituted 80.5% (n = 66) of the cohort whereas FTC and HCC represented 17.1% (n = 14) and 2.4% (n = 2), respectively. PTC subtypes in the cohort are shown in Fig. 1; the most common subtype was CV-PTC (42%, n = 28), followed by FV-PTC (24%, n = 16), TCV-PTC (9%, n = 6), and DSV-PTC (6%, n = 4). Nine PTCs had no subtype documentation; they were listed as UV-PTC. Of the 16 FV-PTCs, there were 5 inFV-PTCs, 4 eFV-PTCs, and 7 unFV-PTCs. The most common sites of metastases are lung only (73%), bone only (11%) or lung and bone (13%). Of the 82 patients, 30 patients had pretherapy scan neck uptake data available before the first RAI-T. The median neck uptake was 1.89% (interquartile range [IQR], 1.1%-4.55%). In the negative scan group (n = 14), the median neck uptake was 1.7% (IQR, 0.89%-2.61%). In the positive scan group (n = 16), the median uptake was 2.4% (IQR, 1.1%-6.9%). The mean RAI dose was 140.4 ± 42.4 mCi. RAI-T was prepared by THW, rhTSH and unknown method in 42.7%, 47.6%, and 9.8% of the cohort, respectively. Detailed clinical characteristics of the cohort are shown in Table 1.

Figure 1.

Papillary thyroid carcinoma (PTC) subtypes.

Table 1.

Demographics and clinical characteristics of the study cohort

Characteristics	DTC with DM at diagnosis (n = 82)	Model 1b			Model 2d
		BVL (n = 26)	RL (n = 7)	P	BVL (n = 40)	RL (n = 28)	P
Age, y (mean ± SD)	56.2 ± 17.3	55.8 ± 18.2	61.9 ± 14.1	0.43	53.9 ± 18.8	56.9 ± 14.9	0.49
Male sex, n (%)	32 (39.0)	12 (46.2)	2 (28.6)	0.67	17 (42.5)	8 (28.6)	0.31
Histology, n (%)
Papillary	66 (80.5)	26 (100)	4 (57.1)	.006	40 (100)	16 (57.1)	< .001
Follicular	14 (17.1)	0 (0)	7 (100)		0 (0)	12 (42.9)
Hürthle cell	2 (2.4)	N/A			N/A
Site of metastasis, n (%)
Lung only	60 (73%)	23 (88.5)	2 (28.6)	.004	36 (90.0)	12 (42.9)	< .001
Bone only	9 (11%)	0 (0)	2 (28.6)	.04	0 (0)	8 (28.6)	< .001
Lung and bone only	11 (13%)	3 (11.5)	2 (28.6)	.28	4 (10.0)	6 (21.4)	.30
Other sitesa	2 (3%)	0 (0)	1 (14.3)	.21	0 (0)	2 (7.1)	.17
RAI dose, mCi (mean ± SD)	140.4 ± 42.4	130.4 ± 33.2	160.6 ± 31.9	.04	130.5 ± 39.4	163.5 ± 37.3	.001
Neck uptake on pretherapy scan median (IQR) (n = 30)	1.89% (1.1-4.55)	2.61% (0.87-5.9)	1.1% (0.87-3.4)	.08	2.51% (0.87-49)	1.50% (1.10-2.58)	.46
Negative scan group median (IQR) (n = 14)	1.70% [0.885, 2.61]
Positive scan group median (IQR) (n = 16)	2.37% [1.10, 6.90]
RAI-T preparation, n (%)
Hormone Withdrawal	35 (42.7%)	9 (34.6)	3 (42.9)	.99	18 (45.0)	9 (32.1)	.51
Recombinant human TSH	39 (47.62%)	15 (57.7)	4 (57.1)		18 (45.0)	17 (60.7)
Unknown	8 (9.8.%)	2 (7.7)	0 (0)		4 (10.0)	2 (7.1)
Negative scans after RAI-T, n (%)c	38 (48.7%)	20 (76.9)	1 (14.3)	.005	28 (70.0)	4 (14.3)	< .001

Abbreviations: BVL, BRAF^V600E like; DM, distant metastases; DTC, differentiated thyroid cancer, IQR, interquartile range; RAI, radioactive iodine; RAI-T, radioactive iodine treatment; RL, RAS like; TSH, thyrotropin.

^a Other sites of metastasis included liver, brain, and adrenal glands in addition to lungs and bones.

^b Four individuals with missing scan results were excluded.

^c Model 1: BVL and RL according to mutational data.

^d Model 2: BVL and RL according to available mutational data combined with histology data.

Correlation of Tumor Histology and Mutational Data

In this cohort of 82 patients, 57.3% (n = 47) had some genomic testing: either BRAF^V600E testing only or more detailed genomic profiling. Of the 66 PTCs, 41 tumors had either BRAF^V600E testing only or more detailed genomic profiling. The breakdown of the mutational results correlating with the histology subtypes is detailed in Table 2. The overall BRAF^V600E prevalence in PTC was 53.7% (n = 22). Eight PTCs were tested for BRAF^V600E only and were negative for the mutation, observed in 4 CV-PTCs, 1 enFV-PTC, 2 unFV-PTCs, and 1 TCV-PTC. Eight BRAF^V600E–positive tumors had additional genomic testing and 7 tumors were found to harbor TERT promoter mutations, observed in 3 TCV-PTCs, 2 CV-PTCs, and 2 UV-PTCs. Fourteen BRAF^V600E–positive tumors were tested for BRAF^V600E only, observed in 9 CV-PTCs, 2 TCV-PTCs, 1 DSV-PTC, 1 UV-PTC, and 1 unFV-PTC. Other oncogenic drivers were (a), RET fusions (n = 5, 3 NCOA4-RET and 2 CCDC6-RET) seen in 3 DSV-PTCs, 1 CV-PTC, and 1 PTC with focal follicular and solid features; (b), NTRK fusions (n = 2, 1 TPM3-NTRK1 and 1 ETV6-NTRK3) in 2 CV-PTCs; and (c), an FGFR2 fusion (n = 1, FGFR2-KIAA1598) in 1 inFV-PTC. NRAS mutations (n = 3) were seen in 2 unFV-PTCs and 1 CCV-PTC. TERT promoter mutations also coexisted with 1 inFV-PTC harboring FGFR2-KIAA1598 and 1 CV-PTC harboring ETV6-NTRK3.

Table 2.

Histology and mutational data of the study cohort

Histology	PTC (n = 66)										FTC (n = 14)	HCC (n = 2)
Mutations	CV (n = 28)	TCV (n = 6)	DSV (n = 4)	HV (n = 1)	CCV (n = 1)	UV (n = 9)	enFV (n = 4)	inFV (n = 5)	unFV (n = 7)	Follicular/Solid features (n = 1)
BRAF V600E+ (C)	1
BRAF V600E+ TERT+ (C)	2	3				2
BRAF V600E+ (S)	9	2	1			1			1
BRAF V600E– (S)	4	1					1		2
RET fusionsa	1		3							1
NTRK fusionsb	2
NRAS+ only					1				2		2
NRAS+ TERT											1	1
FGRR2 KIAA598 fusion								1
No driver mutation											1c	1d
No testing	9	0	0	1	0	6	3	4	2	0	10	0

Abbreviations: C, comprehensive genomic profiling by UCSF 500 gene panel, Tempus xT or Foundation One; CCV, columnar cell variant; CV, classic variant; DSV, diffuse sclerosing variant; enFV, encapsulated follicular variant; FTC, follicular thyroid carcinoma; HCC, Hürthle cell carcinoma; HV, hobnail variant; inFV, infiltrative follicular variant; PTC, papillary thyroid carcinoma; S, selective BRAF^V600E testing only; TCV, tall cell variant; unFV, unknown follicular variant; UV, unknown PTC variant.

^a RET fusions: 3 tumors with NCOA4-RET fusion and 2 tumors with CCDC6-RET fusion.

^b NTRK fusions: 1 tumor with NTRK3-ETV6 fusion and 1 tumor with NTRK1-TPM3 fusion.

^c No commonly known driver mutation was found; AR, FANCA, Jak1, NF1, and TSC2 mutations were identified.

^d No reportable alterations based on Foundation One testing were found; VUS in BCOR, ERBB2, and NOTCH were identified.

Of the 14 FTCs, 4 tumors had genomic profiling. NRAS mutations alone were observed in 2 tumors, NRAS and TERT promoter mutations in 1 tumor, and multiple genetic alterations (AR, FANCA, JAK1, NF1, and TSC2) in 1 tumor. One of the 2 HCCs had an NRAS mutation with a TERT promoter mutation and DNMT3A R736H. The other HCC was found to have no reportable genomic alterations based on Foundation One testing.

Tumor Histology and Negative Scans

Of 82 patients, 4 patients with missing scan results were excluded from this analysis (2 FTCs, 1 CV-PTC, and 1 TCV-PTC). We investigated the association of uptake scan result with specific tumor histology in 78 wDTC patients with known scan results. The percentages of negative scans based on histology are shown in Fig. 2. The overall prevalence of negative scans in this cohort of wDTC was 48.7% (38 in 78 scans). Of the 64 PTC scans, 57.8% (37 in 64 scans) were negative whereas none of the 12 FTC scans and 1 of the 2 HCC scans were negative. Negative scans were more commonly observed in PTC subtypes with papillary architecture, 77.8% in CV-PTC (21 in 27 scans), 75% in DS-PTC (3 in 4 scans), and 60% in TCV-PTC (3 in 5 scans). The only CCV-PTC had a negative scan. For UV-PTC, negative scans were 66.6% (6 in 9 scans). Of the PTC cohort with follicular architecture, there were 16 FV-PTCs and 18.8% (3 in 16 scans) had negative scans. The prevalence of negative scans in eFV-PTC, inFV-PTC, and unFV-PTC were 25% (1 in 4 scans), 20% (1 in 5 scans), and 14.3% (1 in 7 scans), respectively. After excluding FV-PTC (n = 16), CCV-PTC (n = 1), and UV-PTC (n = 9), the prevalence of negative scans in PTC with papillary architecture increased to 71.1% (27 in 38 scans) vs 57.8% in the whole PTC cohort.

Figure 2.

Prevalence of negative scans by differentiated thyroid cancer (DTC) histology.

Mutational Data and Negative Scans

Of the 47 patients who had tumor molecular testing or genomic profiling, 1 CV-PTC harboring BRAF^V600E and TERT promoter mutations was excluded from the analysis because of missing scan results. Eight PTCs had BRAF^V600E testing only and were negative for BRAF^V600E; these tumors were also excluded. The percentages of negative scans based on mutational data are shown in Fig. 3. Notably, all tumors with the duet of BRAF^V600E and TERT promoter mutations (n = 6) had negative scans in the DM. The prevalence of negative scans in tumors with BRAF^V600E mutation regardless of TERT promoter mutation status and in tumors with RET fusions was 80.9% (17 in 21 scans) and 60% (3 in 5 scans), respectively. In comparison, only 1 of the 7 NRAS-mutated tumors (14.2%) observed in CCV-PTC had a negative scan. This tumor also had MEN1 copy number loss. Unlike the duet of BRAF^V600E and TERT promoter mutations, both tumors with the duet of RAS and TERT promoter mutations observed in 2 FTCs had positive scans. Of the 2 tumors with NTRK fusions, 1 tumor with TPM3-NTRK1 had a positive scan whereas the other tumor harboring both ETV6-NTRK3 and TERT promoter mutations had a negative scan. The tumor with FGFR2-KIAA1598 and TERT promoter mutations had a positive scan.

Figure 3.

Prevalence of negative scans by mutational data. BVL, BRAF^V600E like; RL, RAS like. **BRAF V600E mutation regardless of TERT mutation status. BVL tumors included BRAF V600E mutation and RET fusions with or without TERT promoter mutations. RL tumors included NRAS mutations and FGFR2 fusion. Tumors with NTRK fusions were categorized as neutral because these fusions were found neutral in the spectrum between BVL and RL tumors in terms of genomic signaling in The Cancer Genome Atlas (8, 48).

BRAF ^V600E-like vs RAS-like Tumors and Negative Scans

Using available mutational data, in model 1, we identified 26 BVL tumors (21 BRAF^V600E and 5 RET fusions) and 7 RL tumors (6 NRAS mutations and 1 FGFR2 fusion). The prevalence of negative scans in the BVL tumors was 76.9% (20 in 26 scans), significantly higher than 14.3% (1 in 7 scans) in the RL tumors (P = .005) (see Table 1).

Given that close to 50% of our cohort did not have molecular testing or genomic profiling, in model 2, we further categorized tumors into BVL tumors and RL tumors based on histology when mutational data were not available (see “Materials and Methods”). We identified an additional 14 BVL tumors (12 CV-PTCs, 1 TCV-PTC, and 1 HV-PTC) and an additional 21 RL tumors (4 inFV-PTCs, 4 eFV-PTCs, 4 unFV-PTCs, and 9 FTCs), leading to a total of 40 BVL tumors and 28 RL tumors. The prevalence of negative scans in BVL and RL tumors was 70.0% (28 in 40 scans) and 14.3% (4 in 28 scans), respectively (P ≤ .001) (see Table 1).

Association of BRAF^V600E-like vs RAS-like Tumors With Negative Scans

We used multivariate logistical regression to examine the association between BVL vs RL tumors and having a negative scan in the DM (Table 3). In model 1, our univariate analysis showed that BVL and RL tumors had an OR of 11.4 (95% CI, 1.94-67.3) and 0.046 (95% CI, 0.004-0.479) of having a negative scan, respectively. In multivariate logistic regression, we adjusted for confounders including age, sex, race, method of preparation for RAI (rhTSH vs THW), presence of bone metastases, and RAI dose. BVL tumors trended toward higher odds of having a negative scan (OR 19.1; 95%, CI 0.891-410) and RL tumors also trended toward lower odds of having a negative scan (OR 0.056; 95% CI, 0.001-1.92). To account for power because of small numbers with mutational data, we also completed a second analysis for model 2. In the univariate analysis, BVL tumors had higher odds of having a negative scan (OR 12.1; 95% CI, 3.76-39.2), and RL tumors had lower odds of having a negative scan (OR 0.074; 95% CI, 0.022-0.259). When adjusted for the aforementioned variables, BVL tumors remained significantly associated with higher odds of having a negative scan (OR 19.8; 95% CI, 2.72-144), and RL tumors remained significantly associated with lower odds of having a negative scan (OR 0.048; 95% CI, 0.006-0.344). To account for the possibility of misclassifying 10% to 30% of FV-PTC especially inFV-PTC harboring the BRAF^V600E mutation (8, 38), we performed a sensitivity analysis. Instead of classifying the 4 inFV-PTCs and the 4 unFV-PTCs without mutational data as RL, we reclassified the 4 inFV-PTCs as BVL and excluded the 4 unFV-PTCs. The associations between BVL and RL tumors with negative scans were not changed significantly.

Table 3.

Multivariate logistical regression analysis on the association between negative scans and BRAF^V600E-like and RAS-like tumors

Model 1b
	BVL tumors (n = 26)		RL tumors (n = 7)
	Odds ratio (95% CI)	P	Odds ratio (95% CI)	P
Univariate analysis	11.4 (1.94-67.3)	.007	0.046 (0.004-0.479)	.01
Multivariate analysisa	19.1 (0.891-410)	.06	0.056 (0.001-1.92)	.11
Model 2 c
	BVL tumors (n = 40)		RL tumors (n = 28)
	Odds ratio (95% CI)	P	Odds ratio (95% CI)	P
Univariate analysis	12.1 (3.76-39.2)	< .001	0.074 (0.022-0.259)	< .001
Multivariate analysisa	19.8 (2.72-144)	.003	0.048 (0.006-0.344)	.004

Abbreviations: BVL, BRAF^V600E like; RL, RAS like.

^a Adjusted for age, sex, race, type of preparation for RAI, presence of bone metastasis, and RAI dose.

^b Model 1: classification of BVL and RL tumors according to mutational data.

^c Model 2: classification of BVL and RL tumors according to available mutational data combined with histology data.

Age and Iodine Avidity in Classic Variant Papillary Thyroid Carcinoma and Tall Cell Variant PTC

We evaluated if there was a significant difference in age in those with positive vs negative iodine scans in the DM of patients with papillary architecture, specifically those with CV-PTC (n = 27) and TCV-PTC (n = 5). Of these 32 patients, the 24 patients with negative scans were significantly older with a mean age of 62.4 ± 3.2 years, compared to a mean age of 47.0 ± 7.2 years in those with positive scans (P = .03). In a logistical regression, the association between age and uptake was not significant, likely due to a lack of power with 32 patients (data not shown).

Discussion

Although RAI-T is the primary therapeutic modality in patients with distant metastatic DTC and is associated with an improved outcome, prognosis is poor when the metastases are RAI-R (6, 39, 40). Usually, metastatic disease is labeled as RAI-R after failure of one or more RAI-Ts, not before the first RAI-T. These patients have usually undergone one or more rounds of RAI-T with increased risks of adverse effects from repeated RAI-T without clinical benefits (25, 26). Since the publication by Ho et al in 2012, several studies have demonstrated that either MEK or BRAF inhibitors or both, depending on the underlying mutation, can restore iodine avidity in about 40% to 60% of RAI-R patients (13-18). Early identification of tumor characteristics predictive of iodine resistance may guide a tailored individualized treatment approach including early tumor redifferentiation therapy before the first RAI-T, although studies are needed to demonstrate improved outcomes in this strategy.

In our study, we found a high prevalence (48.7%) of negative scans following the first RAI-T in metastatic wDTC. This is higher than previously reported by Durante and colleagues (6) (32%), which included poorly differentiated FTC. The higher prevalence of negative scans seen in our study might be due a higher prevalence of PTC, especially CV-PTC, in our cohort, whereas FTC was the predominant tumor type in the study by Durante et al (6). Simões-Pereira et al (41) demonstrated 21.4% of metastatic CV-PTCs were iodine avid compared to 76.5% metastatic FTC.

Our patients with negative scans likely had non–iodine-avid or RAI-R disease at presentation although the retrospective nature of this study did not allow us to determine if all of these patients were prepared by a low-iodine diet, without iodine contamination such as preceding use of CT iodinated contrast imaging, and did not have false-negative uptake in the DM if a large thyroid remnant was present. A large remnant can potentially cause negative iodine uptake in the DM because of its ability to trap iodine more effectively. Of the 30 patients who had pretherapy neck uptake data, the median uptake in the negative group was 1.7% compared to the 2.4% in the positive scan group. Our limited data suggest remnant uptake likely did not affect our iodine uptake results.

RAI-R disease is defined by the 2015 American Thyroid Association guidelines as having no, partial, or progressive loss of uptake or disease progression within 12 to 18 months despite positive uptake (30). Therefore, the actual percentage of RAI-R disease in our cohort was likely higher since our study defined positive scans as uptake in any of the DM and negative scans as no uptake in all the DM without considering partial uptake and response. Furthermore, 46% of our patients with positive scans in the DM had progressive disease (data not shown), supporting an underestimation of RAI-R disease. Using the criteria from the 2015 American Thyroid Association guidelines (30) to define RAI-R disease, Shobab et al (7) identified 76 patients with synchronous DM in a 4-year period and found 71% of their cohort had RAI-R disease; their study included patients with poorly differentiated thyroid carcinoma.

In our study, negative scans were highly prevalent in patients with papillary architecture (71.1%), in patients with BRAF^V600E mutation (80.9%), and in patients with the duet of BRAF^V600E and TERT promoter mutations (100%). In the study by Durante et al (6), among the 187 patients with PTC, 23% had negative scans but no subtype analyses were performed. The high prevalence of negative scans in our patients with papillary architecture is likely reflected by the predominate BRAF^V600E mutation observed in our PTC cohort. In a previous study, 73% of all PTC patients operated at our institution were found to have BRAF^V600E mutation (42). Compared to other oncogenic drivers, BRAF^V600E generates the highest MAPK output, leading to dedifferentiation involving normal thyroid hormone synthesis genes, and this mutation has been associated with a worse prognosis and RAI-R disease (3, 8, 9, 12, 43-46).

Not all patients with BRAF^V600E mutation had TERT promoter mutation testing in our cohort. Only 8 patients had comprehensive genomic testing after BRAF^V600E testing and 7 were found to have additional TERT promoter mutations. The one patient with BRAF^V600E mutation only had a positive scan. Our result of 100% negative scans in all these 7 patients with BRAF^V600E and TERT promoter mutations is not surprising. The addition of TERT promoter mutations to BRAF^V600E has been demonstrated to have a synergistic effect on PTC-related mortality and is associated with the worst clinical outcomes (40, 47, 48). Tan et al (49) showed that MAPK activation driven by BRAF mutation can robustly activate mutant TERT promoter, which in turn renders BRAF-mutated tumors antiapoptotic and less differentiated. Liu et al (10) demonstrated loss of iodine avidity in all 32 patients with recurrent metastatic CV-PTC harboring this duet, and this loss of iodine avidity was associated with the lowest normal thyroid gene expression. In another study by Yang and colleagues (11) evaluating the influence of TERT promoter mutations on iodine uptake in a Chinese cohort specifically examining the tumor to background iodine uptake, both TERT promoter and BRAF^V600E mutations were found to be associated with a non–iodine-avid status. The duet was found to have markedly reduced uptake to near background level after initial RAI-T in the 10 patients harboring both mutations. Although limited, our result adds to support the finding by the study of the Chinese cohort (11) that this duet can predict non–iodine-avid disease and likely RAI-R status at diagnosis. With current molecular and clinical data, RAI-T is likely futile in this subset of patients. Given the poor prognosis of RAI-R status, better strategies are needed for these DTC patients with DM early on rather than repeated ineffective RAI-T.

In contrast, most of the FV-PTCs of various subtypes including eFV-PTC, inFV-PTC, and unFV-PTC and all FTCs had positive iodine scans. Although inFV-PTC was found more likely to have lymph node metastases, akin to tumors with papillary histology such as CV-PTC (50), the iodine uptake pattern was more similar to the rest of the tumors with follicular architecture, including eFV-PTC, unFV-PTC, and FTC. The effect of TERT promoter mutations on iodine uptake in tumors with NRAS mutations could not be determined because the small number in our study (n = 2) limited any definitive conclusion (see Fig. 2).

Age is an important prognostic factor in DTC. Similar to other studies (6, 7, 51), we found that a negative scan was associated with an older age (62.4 ± 3.2 years vs 47.0 ± 7.2 years), specifically in CV-PTC and TCV-PTC. Although the association disappeared after multivariate logistical regression adjusting for age, sex, race, RAI preparation method, bone metastases, and RAI dose, this is likely due to a small number of patients in this cohort. Not all our CV-PTC patients underwent genomic evaluation, and data were not sufficient for us to evaluate the association of mutational status and age in terms of iodine uptake. Clinically, older patients may also benefit from a different strategy including mutational testing or genomic profiling and consideration of early redifferentiation to optimize their care, especially if the tumors are found to harbor BRAF^V600E with or without TERT promoter mutations.

The overarching conclusion and insights from TCGA is that PTC with papillary architecture such as CV-PTC and TCV-PTC are distinctly different from PTC with follicular architecture as in FV-PTC, raising the question whether FV-PTC should be reclassified with FTC (8, 35). Both FV-PTC and FTC share similar mutational profiles and clinical behaviors with a propensity to metastasize distantly rather than locally, the latter more commonly observed in CV-PTC and TCV-PTC (20-24, 52, 53). We found that BVL tumors and RL tumors exhibited very distinct iodine uptake patterns mirroring the multidimensional genomic data from TCGA; fewer than 25% of patients with BVL tumors had positive uptake in the DM compared to more than 85% of patients who had RL tumors.

Because not all our patients had mutational data, in model 2, we also modeled the comparison in iodine uptake between BVL tumors and RL tumors by incorporating tumor histology, based on insights derived from TCGA demonstrating phenotype and genotype correlation. With increased power, we were able to demonstrate consistent and distinct iodine uptake patterns between the 2 tumor types even after multivariate analysis adjusting for age, sex, dose of RAI, method of preparation, and absence or presence of bone metastases. Although we made assumptions classifying tumors without mutational data to BVL and RL based on histology, the number of misclassified tumors was likely low. BVL mutations such as BRAF^V600E and RET fusions are in general uncommon in tumors with follicular architecture, and RL mutations such as RAS mutations are uncommon in tumors with papillary architecture (3, 8, 54, 55). We performed a sensitivity analysis to account for the possibility of 10% to 30% of FV-PTC harboring BRAF^V600E (8, 38); the results demonstrated the same conclusion.

Our iodine scan results reflected the real-world clinical experience correlating with the multidimensional genomic data from TCGA (8) showing that these 2 tumor types are distinct; RL tumors are more differentiated than BVL tumors in terms of iodine uptake. This may have an important clinical implication in deciding when molecular testing or genomic profiling should be performed. When the cost is prohibitive, priority should be given to patients with BVL tumors with papillary architecture.

While RAI-T is usually the mainstay of therapy after thyroidectomy for patients with DM arising from DTC, our results demonstrated that RAI-T can be futile in patients with DTC with papillary architecture or BVL mutations. Although we do not have outcome data, lack of iodine uptake has been associated with a poor prognosis (6). Strategies to optimize care include molecular testing for patients with DTC with papillary architecture, especially in older patients. If the duet of BRAF^V600E and TERT promoter mutations is found, early redifferentiation before RAI-T should be considered although the effectiveness of such strategy is yet to be studied. In our study, a small subset of patients with BRAF^V600E mutation and RET fusions had iodine-avid DM on the posttherapy scans. Patients, especially younger patients with BRAF^V600E mutation only or RET fusions, may be treated with RAI-T empirically; small-volume pulmonary metastases may be treated successfully (56). However, if no uptake is noted on the posttherapy scan in the setting of structurally apparent DM, the presentation would be considered RAI-R. Obtaining a diagnostic whole-body scan is another strategy for these patients. If uptake is absent in the structurally apparent DM, early redifferentiation therapy should also be considered before RAI-T. The benefit of empiric RAI-T may be questionable when a diagnostic whole-body scan shows no uptake in the structural disease even if uptake is noted on the posttherapy scan (57). At the genomic level, tumors with BRAF^V600E mutation are heterogeneous with a range of dedifferentiation (8). Furthermore, mutations in the subunits of the SWI/SNF chromatin remodeling complexes have been implicated in iodine refractories and resistance to redifferentiation therapy targeting the MAPK pathway (58). While the best treatment strategy is yet to be determined, empiric RAI-T may not be the one-size-fits-all solution as our understanding of molecular pathogenesis of iodine refractoriness continues to advance.

Limitations and Strengths

This study has several limitations. Our study is retrospective; therefore, cause and effect relationships could not be established. Because our cohort was selected from a single tertiary hospital institution, there could be selection bias and generalizability would be limited. The cohort was selected relying on ICD-9 and ICD-10 codes; patients without proper coding might be missed. Clinical characteristics were extracted from chart review of clinical notes and reports without re-review of all radiology images and all surgical pathologies. These could result in the misclassification of patients. Because of the retrospective nature of this study, we were unable to determine if all our patients were properly prepared for RAI-T, hence potentially limiting the certainty of our iodine uptake results. Another limitation is that we might have missed patients with indolent DM at diagnosis that could take years to become apparent. But this would only strengthen our finding that non–iodine-avid disease was prevalent in distant metastatic DTC. On the other hand, we might have included patients who did not have DM at diagnosis but developed DM later because of how we identified some of our patients with DM, specifically using elevated Tg and documented structural disease within 1 year of cancer diagnosis. This could potentially falsely elevate our prevalence of negative iodine scans. However, given that Tg is the most sensitive marker for persistent or recurrent disease (30), when significantly elevated early in the course despite of RAI-T, the elevation likely represents microscopic DM at diagnosis. Structural disease documented within 1 year of diagnosis also likely represents persistent disease rather than recurrent disease (30, 31). Last, molecular testing or genomic profiling was not uniformly performed, lowering the power of our multivariate analysis using mutational data only.

Our strength is that we were able to demonstrate for the first time the distinct iodine uptake patterns between BVL and RL tumors mirroring the genomic data from TCGA. Our findings of negative iodine scans and therefore non–iodine-avid status in PTC with papillary architecture and in BVL tumors were consistent with the underlying molecular tumorigenesis of DTC in which activation of the MAPK signaling pathway leads to tumor dedifferentiation, and therefore decreased iodine avidity (3, 8).

Conclusions

In this cohort of wDTC patients with DM at diagnosis, we found a high prevalence of non–iodine-avid disease, especially in patients with papillary architecture and with BRAF^V600E mutation. The duet of BRAF^V600E and TERT promoter mutations conferred 100% non–iodine avid-disease. Older patients with PTC with papillary architecture were also likely to have non–iodine-avid disease. Better strategies or clinical trials are needed for these patients to optimize their care.

Glossary

Abbreviations

BVL

BRAF ^V600E like

CCV

columnar cell variant

CT

computed tomography

CV

classic variant

DM

distant metastases

DSV

diffuse sclerosing variant

DTC

differentiated thyroid cancer

eFV

encapsulated follicular variant

FTC

follicular thyroid carcinoma

FV

follicular variant

HCC

Hürthle cell carcinoma

HV

hobnail variant

ICD

International Classification of Diseases

inFV

infiltrative follicular variant

IQR

interquartile range

MAPK

mitogen-activated protein kinase

OR

odds ratio

PTC

papillary thyroid carcinoma

RAI

radioactive iodine

RAI-R

radioactive iodine refractory

RAI-T

radioactive iodine treatment

rhTSH

recombinant human thyrotropin;

RL

RAS like

TCGA

The Cancer Genome Atlas

TCV

tall cell variant

Tg

thyroglobulin

THW

thyroid hormone withdrawal

TSH

thyrotropin

UCSF

University of California, San Francisco

UV

unknown variant

wDTC

well-differentiated differentiated thyroid cancer

Financial Support

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) T32 (grant No. 5T32DK007418 to M.H.S.) and the National Center for Advancing Translational Sciences, National Institute of Health, through UCSF-CTSI (grant No. UL1 TR001872). Its contents are solely the responsibility of the authors and do not necessarily represent the official reviews of the National Institutes of Health.

Disclosures

The authors have nothing to disclose.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.