The role of the size in thyroid cancer risk stratification

Only a minority of cases of differentiated thyroid carcinoma (DTC) have a poor clinical outcome. Clinical outcomes and molecular aspects were assessed in: 144 DTC ≤ 40 mm without distant metastases (group 1); 50 DTC > 40 mm without distant metastases (group 2); and 46 DTC with distant metastases (group 3). Group 3 had a worse outcome than the other two groups: during the follow-up, patients more frequently had persistent disease, died, or underwent further treatment. The outcomes did not differ between groups 1 and 2. Group 3 had a higher prevalence of TERT promoter mutations than group 2 (32.6% vs 14%). Group 1 had a higher frequency of BRAF mutations than groups 2 or 3 (61.1% vs 16.0% and 26.1%, respectively), while RAS mutations were more common in group 2 than in groups 1 and 3 (16.0% vs 2.1% and 6.5%, respectively). Groups 1 and 2 shared the same outcome, but were genetically distinct. Only lymph node involvement, distant metastases, older age and (among the molecular markers) TERT promoter mutations were independent predictors of a worse outcome. Metastatic DTC had the worst outcome, while the outcome was identical for large and small non-metastatic DTC, although they showed different molecular patterns. TERT promoter mutations emerged as an independent factor pointing to a poor prognosis.

The patients in group 3 were more frequently males (p < 0.01), and were older than those in the other two groups (p < 0.01). Group 3 included a higher proportion of cases of widely invasive FTC (WI-FTC), with 72.7% in group 3 and 37.5% in group 2 (there were no cases of FTC in group 1). Patients in group 3 were also more likely to undergo a second treatment (80.0% versus 9.0% in group 2, and 5.6% in group 1, p < 0.01), and to have a worse outcome, with higher rates of persistence/recurrence or death (78.2% versus 6.0% in group 2, and 4.9% in group 1, p < 0.01). Patients with metastatic tumors (group 3) had a shorter DFS (p < 0.01) (Fig. 1), and a worse CSS (p < 0.01) (Fig. 2) then the other two groups. No differences in patient outcomes emerged between groups 1 and 2.
Molecular characteristics. The molecular characteristics of the three groups are summarized in Table 2. BRAF mutations. BRAF mutations were detected in 88/144 (61.1%) of patients in group 1, 8/50 (16.0%) in group 2, and 12/46 (26.1%) in group 3 (p < 0.01). The BRAF mutation was significantly more frequent in group 1 than in the other two groups (p < 0.01). The type of BRAF mutation was related to the histotype: BRAF V600E was detected in all but one PTC, while BRAF K601E was the only BRAF mutation found in cases of FTC. RAS mutations. RAS mutations were detected in 3/144 (2.1%) cases in group 1, 8/50 (16.0%) in group 2, and 3/46 (6.5%) in group 3. They were more common in large, non-metastatic tumors than in the other two groups (p = 0.006).
Among the 27 patients with TERT mutations, 8 (29.6%) also had BRAF mutations, and 4/27 (14.8%) also had RAS mutations. In one case (of WI-FTC), we found a homozygous TERT promoter mutation together with BRAFK601E and PIK3CA mutations (this case had been reported in 2011 18 , but TERT analysis had not been done at the time). A combination of TERT and PTEN mutations was also found in one case of large FTC without distant metastases. TP53, PTEN and PIK3CA mutations. A TP53 mutation was found on exon 6 in only 1/133 patients (0.8%) in group 1 (G293R), and in 1/44 (2.3%) in group 2 (D208N). PTEN mutations were detected on exon 5 in 1/50 patients (2%) in group 2 (D93N). PIK3CA mutations were detected in 1/43 patients (2.3%) in group 2 (R516K) who had PTC, and in 2/40 (5.0%) of patients in group 3 (one carrying an E545A mutation, the other carrying a H1047R mutation), both cases of WI-FTC.
Clinical and molecular correlations. In the series as a whole, and within the three groups, RAS, PTEN, TP53 and PIK3CA mutations were not associated with patient outcomes.
BRAF mutations were also unassociated with outcomes in the series as a whole, but in group 1 they were weakly associated with a worse prognosis. Indeed, of the 137 patients with a remission or indeterminate response at the end of the follow-up, 81 (59.1%) were BRAF-mutated, as opposed to 7/7 (100%) with persistence of disease (p = 0.03).
BRAF was not associated with outcome in groups 2 or 3. www.nature.com/scientificreports/ Whether alone or in association with other molecular events, TERT promoter mutations correlated with the worst outcomes: they were detectable in 18/46 patients (39.1%) with persistent disease, and only in 9/183 patients (4.9%) with an excellent response (p < 0.01).
At univariate analysis, male sex, older age at diagnosis, type of cancer, TNM parameters, tumor stage, need for a second treatment, and TERT mutations were all associated with a worse outcome (p < 0.01).
TERT promoter mutations also influenced outcomes for patients in groups 2 and 3. In group 2, patients with persistent disease or disease-related death were more frequently TERT-mutated than patients in remission or an indeterminate status: 2/3 (66.7%) versus 5/47 (10.6%), respectively (p = 0.0073). In group 3, none of the patients with a biochemical remission or indeterminate response (0/10) were TERT-mutated, while 41.7% of the patients with TERT promoter mutations were among those with a worse prognosis (15/36) (p = 0.01).  www.nature.com/scientificreports/

Discussion
Although the mortality rate for DTC has remained stable, its incidence has been rising rapidly in recent decades due to the increasing use of neck ultrasound for thyroid diseases and other, unrelated conditions 7,19-23 . Many authors have focused on defining clinical, pathological and molecular characteristics of DTC useful for the purposes of identifying the minority of aggressive tumors associated with a higher likelihood of progression or death. This would enable customized treatment decisions and follow-up protocols, possibly right from the time Table 2. Molecular characteristics of the three groups (also by histology).  www.nature.com/scientificreports/ of diagnosis. With this in mind, various classification systems have been proposed, such as the TNM classification (useful for predicting the risk of death) and the ATA grouping score (suitable for establishing the risk of persistence/recurrence) 6,8 . From a clinical standpoint, the presence of distant metastases and a tumor size greater than 40 mm have conventionally been associated with a poor prognosis, warranting radical surgery associated to radioactive iodine (RAI) therapy 6,7 . The impact of tumor size per se on patient mortality is still debated, however. Given these premises, we examined a group of cases with distant metastases (regardless of primary tumor diameter), and a group with non-metastatic DTC larger than 40 mm in size, seeking to establish whether these forms of DTC, which are known to be more aggressive, share similar molecular features and clinical courses. We compared the characteristics of these "high-risk" groups with those of a group of cases of DTC not exhibiting these two features of aggressiveness. All patients had been diagnosed, treated and followed up according to the same standards of care at the same institution.
As expected, our findings confirm the impact of distant metastases on a patient's clinical outcome 24 , as metastatic disease is associated with a significantly worse DFS and CSS. None of the patients without metastases died of their disease regardless of the size of their DTC. The presence of distant metastases was also the most influential independent prognostic factor in terms of poor prognosis, with an OR of around 40.0. Patients with distant metastases, were more often male, older at the time of their diagnosis, with lymph node involvement. They were more likely to need of a second treatment, and they had the worst outcome. Patients with large, but non-metastatic DTC shared the same outcome as those with small non-metastatic lesions. In other words, a large tumor size was not a risk factor in our series when considered alone. A shorter median follow-up for the group with large but not metastatic tumors may have influenced this result, although the 4.3-year median follow-up for this group should have been long enough for most relapses to become apparent 25 .
In recent years, several studies have suggested that certain molecular events-mainly BRAF, RAS, TP53 and TERT promoter mutations, alone or in combination-may be useful for stratifying the behavior of DTC, and predicting which tumors are more likely to recur and/or be fatal [26][27][28] .
The BRAF V600E mutation is the most common molecular event in DTC (found in 40-60% of cases). Its presence has been historically associated with less-differentiated phenotypes carrying a poor prognosis, but its role is still debated [29][30][31][32][33] . The most important multicenter retrospective studies on the role of BRAF in DTC found it associated with key negative prognostic factors. Xing et al. reported, for instance, that it was not independently associated with cancer-related death, but it was independently associated with disease persistence/recurrence 34,35 . Some monocentric retrospective studies confirmed the BRAF mutation as an independent negative prognostic factor related to persistent disease 36 , and to aggressive characteristics such as loss of the ability to concentrate radioiodine, or acquisition of a capacity for glucose uptake 37 . On the other hand, prospective studies on patients who underwent fine needle aspiration biopsy for molecular definition and total thyroidectomy failed to confirm an role of BRAF mutations on patient outcomes 38,39 .
These different findings may stem from the fact that, particularly in the setting of prospective studies, the early detection of a BRAF mutation (particularly at cytology) prompts early surgery, so this genetic alteration may not have the time to confer the aggressive tumor characteristics found in retrospective studies. We confirmed the published data on the prevalence of BRAF mutations 34 , but only in our group 1 (61.1%), while it was markedly lower in groups 2 and 3 (16.0% and 26.1%, respectively). This is in line with reports on other metastatic DTC series in which this genetic event was not associated with the tumor's potential to spread to distant sites 34,40 . In short, the mutation does not give the cancer cells a metastatic advantage. Our study has a possible bias, however, regarding the lack of FTCs among the consecutively-selected cases in group 1. This could have increased the frequency of BRAF mutations in group 1 by comparison with the other groups, since this mutation is uncommon in FTCs 41 .
Our data revealed a higher prevalence of the RAS mutation than reported elsewhere in the literature 34,42,43 , particularly in our group 2 (16.0% of cases revealed RAS mutations). We found no association between these mutations and the patients' clinical features, probably because of the small size of this sample.
TERT promoter mutations in DTC were first described in 2013 44,45 . The literature has since consistently demonstrated an association between TERT promoter mutations and poor outcomes. This mutation in DTC patients is associated with older age, larger tumors, distant metastases, and advanced stage at diagnosis 13,16,17 . In our series too, the frequency of TERT promoter mutations rose with DTC aggressiveness and risk of progression: it was 3.8% group 1, 14.0% in group 2, and 32.6% in group 3.
In line with previous reports, TERT promoter mutations in our series as a whole were associated with older age at diagnosis, more advanced tumor stage, more frequent need for a second treatment, and worse outcomes (around 67% of patient had persistent disease or died of their disease). At multivariate analysis, TERT promoter mutations was the only variable independently associated with a negative prognosis.
There are reports in the literature on associations between TERT mutations and other oncogenic molecular events. This condition was found related to an aggressive DTC behavior, including distant metastases and a lack of radioiodine uptake capability 13,37,45 . We found no such negative impact of simultaneous mutations on the prognosis for patients with DTC, by comparison with those harboring a single TERT promoter mutation. It could be useful to consider larger series in order to clarify the real impact of multiple mutations. The other molecular events considered here-PIK3CA, PTEN and TP53 mutations-seem to be rare, and almost exclusive to high-risk DTC, as previously reported 17 . It is worth noting that the presence of a TERT promoter mutation was able to influence the prognosis even in our group 3, with metastatic disease.
It has been reported that age, tumor size, extrathyroidal extension, and nodal and distant metastases are predictors of a patient's mortality risk [46][47][48][49][50][51] . Our data, focusing on tumor size and distant metastases, only confirm the latter as a predictor of mortality.
It is noteworthy that patients with larger DTCs shared much the same outcome as those with smaller lesions, though they differed considerably from a molecular standpoint. RAS and TERT mutations were more common in group 2, while BRAF mutations were more prevalent in group 1. Group 2 had a molecular profile more similar www.nature.com/scientificreports/ to that of group 3, except for a higher frequency of RAS mutations in the former. TERT promoter mutations were more frequent than in group 1, but still lower than in metastatic patients. Based on our results and the conflicting data in the literature regarding its prognosis, we wonder whether large DTCs may be a sort of intermediate entity, in between small indolent DTCs and those with a poor outcome. Our data suggest a shift in the molecular profile from DTC < 4 cm in size to metastatic tumors, with large DTCs seeming to come somewhere in between, in terms of their molecular features, although they have much the same outcomes as the smaller, non-metastatic tumors. It may be that many other molecular events need to accumulate in large tumors to have any metastatic potential and affect the prognosis. The clinical importance of cancer size is still debated. Most authors have observed an indolent behavior of microcarcinomas 52 , but some have reported distant metastases already at diagnosis or during the follow-up, and a poor prognosis even for tumors less than 10-15 mm in size [53][54][55] . Little is known as yet about the molecular patterns of such small but aggressive DTCs. In our series, the median size of tumors in group 3 (with metastatic disease) was smaller than in group 2, and ranged very widely, from 2 up to 90 mm, although their molecular pattern and other characteristics associated with poor outcome (lymph node involvement) were very different. These data suggest that some DTCs are inherently aggressive, already developing aggressive mutations even while they remain small in size. A better knowledge of the molecular pattern of advanced thyroid disease could help us to identify promising therapeutic targets for advanced DTC. The treatment of choice for metastatic thyroid cancer is RAI therapy, but as DTC progresses, it can lose its capacity for iodine uptake 56 , becoming refractory to RAI. Other therapies involving tyrosine kinase inhibitors can be attempted in such cases. The loss of iodine uptake capacity in some thyroid cancers has been associated with involving BRAF V600E 57 or TERT promoter mutations, or both 58 . Phase II trials with the BRAF V600E inhibitors vemurafenib 59 and dabrafenib 60 (two FDA-approved drugs for treating BRAF V600E-mutated melanoma) obtained promising results in thyroid cancers with this mutation. Vemurafenib also proved capable of restoring radio-iodine uptake, to some degree at least 61 . As discussed earlier, however, few cases of metastatic thyroid disease are associated with BRAF V600E mutations, so other therapeutic targets are needed. Given the frequency of its mutations in metastatic cancers, and their prognostic role in thyroid cancer, the TERT promoter could be a future therapeutic target. Telomerase inhibitors (e.g. Imetelstat 62 ), drugs inducing telomere dysfunction (e.g. 6-thio-2′-deoxyguanosine 63 ), and adenoviral gene therapies that induce telomerase promoter-driven oncolytic activity 64 are currently under investigation as potential anticancer agents. Based on the latest literature (as confirmed by the present study), telomerase could be a promising therapeutic target for RAI-refractory thyroid cancer, but no data are available as yet.
Our study confirms the prognostic impact of distant metastases, but not primary cancer size on patient outcomes. From the available literature, it is still hard to say whether large tumor size has an impact on patients' prognosis. Our data suggest that larger DTCs differ in their molecular features from smaller ones, but are not more aggressive. We found a higher prevalence of TERT promoter mutations in patients with metastatic disease. This mutation confirmed its association with tumor aggressiveness, and was the only molecular event capable of significantly and independently influencing DTC outcome.

Methods
Patients. We selected 240 consecutive patients with a histological diagnosis of DTC who underwent total thyroidectomy from 2007 to 2016, and were followed up by the Endocrinology and Radiotherapy Units in Padua. We first ensured that adequate frozen material was available in the Tissue Bank after first surgery for all cases. Then patients were grouped as follows: group 1, DTC with the largest tumor focus ≤ 40 mm, without distant metastases; group 2, DTC with the largest tumor focus > 40 mm, without distant metastases; and group 3, metastatic DTC at diagnosis or detected during the follow-up, regardless of tumor size.
Clinical data were obtained from the electronic medical records. Surgical pathology specimens were analyzed by two expert pathologists (FG and GP). All pathological samples were reviewed and a histological diagnosis was established following the 4th edition of the WHO classification (WHO-2017) 65 . Pathological staging was done according to the 8th edition of the TNM staging system 8 . If there were multiple foci, the largest tumor dimension was considered.
All patients underwent total thyroidectomy and RAI treatment (median dose 200 mCi). When further treatments were indicated, these involved surgery, RAI treatments, external beam radiation, bisphosphonates (in cases of multiple bone lesions), and tyrosine kinase inhibitors (in cases of progressive metastatic RAI-refractory disease).
Following recent ATA guidelines 6 , we identified four possible outcomes, defining response as: excellent; biochemical incomplete; structural incomplete; or indeterminate. The median patient follow-up was 79.8 months (12.9-237 months).
This study was conducted according to the guidelines laid down in the Declaration of Helsinki. All patients participating in the study gave their written informed consent. The Ethical Committee for Clinical Experimentation at Padua Hospital approved the study protocol (Ref. www.nature.com/scientificreports/ 2 and 3, by direct sequencing (ABI PRISM 3130, Applied Biosystems, Foster City, CA), as previously described elsewhere 66,67 . Statistical analysis. Categorical data were summarized using frequencies and percentages. Distributions of the continuous variables were assessed, and data were summarized accordingly. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.