Dynamic risk stratification in the follow-up of thyroid cancer: what is still to be discovered in 2017?

Dynamic risk stratification in DTC patients treated with total thyroidectomy and RAI ablation

The idea of dynamic risk stratification, proposed in Tuttle et al. (2010) has been confirmed by a few retrospective analyses, published recently (Table 3). The first data come from an Italian study (Castagna et al. 2011). In fact the authors called this new approach ‘delayed risk stratification’ (DRS) but it unequivocally reflected Tuttle’s idea. Castagna et al. (2011) retrospectively reviewed a group of 512 DTC patients, who were stratified soon after initial therapy (total thyroidectomy and RAI ablation) on the basis of ATA 2009 (Cooper et al. 2009) and ETA 2006 (Pacini et al. 2006) guidelines. Next, 8–12 months after primary treatment, they were subjected to DRS involving their clinical status. Initially, at the time of treatment, 45.1% of patients were classified as ETA low-risk and 54.9% of remaining subjects as ETA high-risk. Simultaneously, 47.6% and 52.4% of these patients fulfilled ATA low-risk criteria and intermediate/high-risk criteria, respectively. At the subsequent control, 8–12 months later, complete remission was stated in 87.8% of ETA and 87.2% of ATA low-risk patients and in 53.3% and 52.2% of ETA and ATA high-risk patients, respectively. At this time point, patients showing complete remission were restratified as DRS low-risk category, whereas subjects who had a persistent DTC were involved into DRS high-risk category. Thereby, nearly 50% of patients at first involved to ATA and ETA high-risk groups moved to a low-risk category and nearly 12% of patients from ATA and ETA low-risk groups were reclassified as a high-risk category. In the next step of the analysis, to demonstrate the validity of such re-classification, the authors attempted to correlate ATA, ETA and DRS with final outcomes. Complete remission was achieved in comparable rates of patients from ATA and ETA low-risk groups: 91.4% and 90.8%, respectively, whereas this value was significantly higher (96.6%) in DRS low-risk category. Considering ATA intermediate/high-risk and ETA high-risk groups, complete remission was noticed in 60.8% and 61.6% of patients, respectively. The percentage of patients with complete remission in DRS high-risk category was only in 27.1%. Conversely, 33.9% of patients from ATA intermediate/high-risk group and 32.8% from ETA high-risk group demonstrated persistent DTC. These rates were significantly lower than this one, noticed in DRS high-risk category (66.6%). Interestingly, no significant differences in recurrence rate after a period of complete remission between low- and high-risk groups were observed, regardless of the stratification system used. The authors believed that patients in clinical remission had a good prognosis irrespective of initial risk stage. Finally, to assess the ability to predict the outcomes (remission/persistent DTC), the authors calculated a positive predictive value (PPV), negative predictive value (NPV) and proportion of variance (PVE) for each particular stratification system. The values of all above-mentioned factors achieved for DRS (PPV 72.8%, NPV 96.3% and PVE 62.1%) were significantly better than those related to ATA and ETA initial risk classification (Castagna et al. 2011). Thus, the authors unequivocally concluded about the superiority of DRS over other stratification systems performed at DTC diagnosis.

The superiority of DRS (also these authors used the term ‘delayed’ proposed by the Italian group) was also demonstrated by other analysis of 398 DTC patients, who underwent total thyroidectomy and RAI ablation (Hong et al. 2014). This group had an initial risk evaluation, performed after RAI treatment, on the basis of TNM and ATA 2009 (Cooper et al. 2009). After 8–15 months, it was subjected to DRS and classified as having an excellent, acceptable, biochemical incomplete or structural incomplete response. Using ATA criteria, 20.4% of patients were evaluated as low risk, 65.1% as intermediate risk and 14.6% as high risk. At the time of DRS, 57.5% of patients demonstrated an excellent response, 19.6% acceptable response, 15.6% incomplete biochemical and 7.3% incomplete structural response. Considering ATA low-risk group, an excellent response was stated in 75.3%, acceptable in 16%, incomplete biochemical in 7.4%, and incomplete structural in 1.2% of patients. In ATA intermediate-risk group, 60.2% of patients showed an excellent response, 20.1% acceptable response, 16.2% biochemical incomplete and 22.4% structural incomplete response, whereas in ATA high-risk class, these rates were respectively 20.7, 22.4, 24.1 and 24.1%. At the end of this analysis, the authors compared the association of ATA 2009 risk system and DRS with progression-free survival (PFS). A multivariate analysis, which included DRS and the factors used for ATA risk stratification (extrathyroidal extension, regional lymph node metastases, histologic type) demonstrated a statistically significant association only between DRS and PFS. In another multivariate analysis of ATA risk stratification and the factors involved in DRS (stimulated serum thyroglobulin (Tg) level ≥1 ng/mL and TSH suppressed Tg level ≥1 ng/mL), ATA risk stratification did not show any significant association with PFS either but both stimulated and suppressed Tg levels ≥1 ng/mL did (Hong et al. 2014). In conclusion, according to the data given by ATA, Tg evaluation is crucial for risk assessment.

Korean authors evaluated dynamic risk stratification in a group of 715 DTC patients after total thyroidectomy and RAI ablation (Jeon et al. 2014). The authors called their approach a ‘modified dynamic risk stratification’ because they focused also on the status of antiTg antibodies in their risk assessment. Like in other papers, at DTC diagnosis, the patients were categorized according to ATA 2009 criteria (Cooper et al. 2009) as low risk (12%), intermediate risk (81%) and high risk (7%). Next, 6–24 months after an initial therapy, they were reclassified to one of the following category: excellent response, acceptable response, biochemical incomplete response and structural incomplete response: 61, 20, 7 and 12% of patients, with disease-free survival (DFS) rates of 98.3, 71.6, 54.2 and 6.8%, respectively. PVE value of this modified dynamic risk stratification was significantly higher than ATA and TNM risk stratification systems: 44.59% vs 12.14% and 8.68% respectively (Jeon et al. 2014).

Another paper, confirming the relevance of DRS, involved into a retrospective analysis a group of 916 DTC patients treated with total thyroidectomy and RAI ablation in a single institution between the years 2000 and 2013 (Kowalska et al. 2016). Likewise, the patients underwent an initial staging according to ATA 2009 (Cooper et al. 2009) and ETA 2006 (Pacini et al. 2006) guidelines and were restratified 9–12 months after primary treatment (DRS) on the basis of response criteria proposed in Momesso & Tuttle (2014) as having an excellent response, biochemical incomplete response, structural incomplete response or indeterminate response. Patients, in whom an excellent response was demonstrated, were restratified by DRS as low risk, whereas remaining subjects as high risk. The majority of patients (62%) presented stage I at DTC diagnosis. Using ATA 2009 criteria 59.6, 34.6 and 5.8% of the patients were initially stratified as low, intermediate and high risk, respectively. While according to ETA classification, 20.9% were considered as very low risk, 37.3% as low risk and 41.8% as high risk. Nine to twelve months later, 79.8% of patients achieved an excellent treatment response (DRS low risk), whereas remaining 20.2% of patients were restratified at as DRS high-risk (among them 3.3% with biochemical incomplete response, 6.4% with incomplete structural response and 10.5% with indeterminate response). At the final follow-up, 83.6% of patients were free of disease, 12% had persistent DTC and 4.4% died, among them 1.9% due to DTC. However, there is no information to which risk-group the patients with cancer-related death were classified. Similar to Castagna study, Kowalska et al. (2016) calculated PPV, NPV and PVE. Their results were concordant with the data reported in Castagna et al. (2011), where DRS showed the highest ability in a prediction of DTC outcomes, with PPV, NPV and PEV 56.8, 98.5 and 56.7%, respectively (Kowalska et al. 2016).

Recently, a dynamic prediction of the risk of recurrence has been applied in a group of older DTC patients above the age of 60 years (Morosán et al. 2016). Ninety patients at median age of 65 years (range 62–82), treated with total thyroidectomy and RAI ablation between 2000 and 2013, and with follow-up period longer than 2 years, were included into a retrospective analysis. These patients were initially stratified using ATA 2009 criteria (Cooper et al. 2009), whereas their response to treatment was evaluated 2 years later. According to ATA 2009 risk classification 47% of patients were initially allocated to low-risk, 43% to an intermediate-risk and 10% to high-risk groups. After 2 years, an excellent response was stated in 74, 36 and 0%, an indeterminate response in 24, 51 and 100% and an incomplete response in 2, 13 and 0% of patients from ATA low-, intermediate- and high-risk groups, respectively. At the end of follow-up, 95% of patients from ATA low-risk group demonstrated NED, whereas in the ATA high-risk group, 100% of subjects showed persistent DTC. Considering ATA intermediate-risk group 61% of patients were free of disease and 39% had persistent disease. When this group was restratified 2 years later using response criteria, 100% of patients with an excellent response and only 61% of subjects with indeterminate/incomplete responses were disease free at final follow-up (Morosán et al. 2016).

The most recent study analyzed the impact of DRS on prognosis in intermediate-risk PTC (Tavarelli et al. 2017). This retrospective analysis included 560 pT3 PTC patients treated with total or near-total thyroidectomy and RAI ablation in a single institution. The study population was split into three groups according to tumor size and the presence of extrathyroidal invasion (ETI): group 1 (n = 160) pT3 ≤ 10 mm with ETI; group 2 (n = 328) pT3 > 10 mm with ETI and group 3 (n = 78) pT3 due to tumor size >4 cm without ETI. At initial presentation, group 2 demonstrated more often lymph node metastases and extranodal extension. Lymph node metastases and multifocality were more frequent in group 1. The differences between the groups were significant. Moreover, patients from group 3 received a significantly higher RAI activity. Ten-year DFS was 89, 67, and 82% in groups 1, 2 and 3, respectively. Lymph node involvement, especially in the lateral compartment (N1b), male sex and group 2 had a negative significant impact on DFS. The patients were reevaluated 6–8 months after an initial treatment. Clinical remission was obtained in 78.4% of patients, whereas 21.6% had persistent disease. In patients with NED at the six- to eight-month check-up, 10-year DFS significantly improved in all groups: from 89% to 98% in group 1, from 67% to 96% in group 2 and from 82% to 90.5% in group 3. Considering patients with persistent PTC at check-up none of the patients from group 1 were disease free. DFS decreased to 19% in group 2 and to 57% in group 3 (Tavarelli et al. 2017).

Dynamic risk stratification in DTC patients treated with total thyroidectomy or lobectomy, without RAI ablation

To date, the results of two studies, evaluating the utility of dynamic risk stratification in patients treated only with surgery (total thyroidectomy or lobectomy) without RAI ablation, have been published. Criteria on how to classify treatment responses in these patients, proposed in Momesso et al. (2016), are given in Table 2.

The first one included 507 DTC patients, among whom, 187 underwent lobectomy (L) and 320 underwent total thyroidectomy (TT) (Momesso et al. 2016). First, after the operation, patients were categorized according to ATA 2015 modified criteria (Table 1) as low, intermediate and high risk and next, during the first 2 years of follow-up, they were restratified according to treatment response (excellent, incomplete biochemical, incomplete structural or indeterminate). Considering L and TT patients alone, there were 81.9% and 91.4% of ATA low-risk patients and 18.1% and 8.6% of ATA intermediate-risk patients, respectively. There were no ATA high-risk patients in both groups, what is in accordance to indications for lobectomy. An excellent response was obtained in 64.2%, incomplete biochemical response in 9.1%, incomplete structural response in 2.1% and indeterminate response in 24.6% of L patients. These values were 16.5, 2.2, 1.9 and 79.4% for TT patients, respectively. Such high percentage of patients demonstrating an indeterminate response is surprising. The authors explained it by the presence of nonspecific US findings, nonstimulated Tg value between 0.2 and 5.0 ng/mL, and positive stable or declining antiTg antibodies. As it could be expected the risk of DTC recurrence was low in patients with an excellent and indeterminate response but very high in patients demonstrating an incomplete structural response: 100% in both groups. Eight TT patients and 21 L patients required additional treatment. All but one showed NED at final follow-up (Momesso et al. 2016).

The results of the second study have just been published (Park et al. 2017). The authors retrospectively analyzed a group of 357 DTC patients after lobectomy or TT. As in the above-mentioned study, patients were initially evaluated according to ATA 2015 criteria (Table 1) as low or intermediate risk, 52.4% and 47.6%, respectively. DFS and recurrence risk did not differ significantly between ATA low and intermediate-risk groups. DRS, carried out during the first 2 years of follow-up classified 71.7% of patients as having an excellent response, 18.5% with an indeterminate response, 8.4% with a biochemical incomplete response and 1.4% of remaining patients with an incomplete structural response. The differences in DFS regarding treatment responses were statistically significant. The relative risk (RR) of DTC relapse among patients with an indeterminate response was nearly 2 times higher, with a biochemical incomplete response nearly 21 times higher and with a structural incomplete response nearly 243 times higher comparing to patients showing and an excellent response (Park et al. 2017).

Dynamic risk stratification in the light of the newest ATA guidelines

All retrospective studies discussed above clearly demonstrated that the use of dynamic risk assessment allowed for a better prediction of DTC recurrence risk. This idea has been reflected in the newest, updated 2015 ATA GLs (Haugen et al. 2016). The 2015 ATA GLs clearly emphasize that, although this initial staging system allows for a valuable assessment of DTC recurrence risk, it provides static information based only on data available at DTC onset and therefore it is not sufficient for a life-long follow-up. All retrospective analyses discussed above, as well as our personal clinical experience, prove that substantial proportion of patients initially classified as an intermediate and high risk achieve complete remission (or using current term: an excellent response) and become low risk of DTC recurrence. Thus, there is a necessity to restratify them appropriately taking into consideration their response to treatment applied. It has been proposed to carry out such restratification during the first 2 years of follow-up and at any point later. As it has been confirmed by the papers, already cited by us, the risk of DTC recurrence in a particular patient varies according to the achieved best response. Noteworthy, criteria for treatment response published in the current 2015 ATA GLs (Table 2) are in fact related only to patients who underwent TT and RAI ablation (Haugen et al. 2016). To restratify the subjects treated with TT or lobectomy alone, one should base on the criteria proposed in Momesso et al. (2016) (Table 2). Likewise, this evaluation considers nonstimulated serum Tg level, TgAb level and imaging studies (Momesso et al. 2016). These Tg cut-off values are different than these proposed for patients after TT and RAI ablation and have not been validated by an independent study yet.

There are still some unanswered questions that in our opinion need to be resolved in the nearest future.

Histopathological controversies regarding 2015 ATA initial risk stratification

First important information, required for an adequate risk stratification, should be provided by a postoperative histopathological report that has to include basic tumor features necessary for TNM staging and additional data regarding vascular invasion, the number of invaded vessels, the number of examined lymph nodes, the number metastatic lymph nodes and their size of as well as presence/absence of extranodal extension. Histopathologic variants of PTC or FTC should also be identified.

Surprisingly, 2015 ATA GLs did not specify poorly differentiated thyroid cancer (PDTC), although this type of thyroid cancer is obviously characterized by a high risk of relapse and a bad prognosis. Moreover, Hurthle cell carcinoma (HCC, oncocytic variant of follicular carcinoma) was mentioned in the description of recommendation 46 concerning the basic principles of histopathological evaluation, but it was not present in the graph of ‘Risk of Structural Disease Recurrence’. Encapsulated HCC with an extensive vascular invasion (EVI) implies a high risk of recurrence observed in 64% of such tumors (Xu et al. 2015). This paper also reported a high risk of recurrence in patients without distant metastases at the time of diagnosis. HCCs with EVI relapsed even vascular invasion counted 4 or 7 foci and capsular invasion was unifocal (Xu et al. 2015).

AJCC/UICC stratification is strongly recommended for each newly diagnosed DTC patient to predict DTC mortality. Simultaneously with TNM classification, the patients should be subjected to an initial stratification to evaluate the risk of persistent or recurrent DTC. The updated guidelines still recommend the ATA 2009 initial clinic-pathologic risk stratification system (Cooper et al. 2009), currently completed with additional variables such as the extent of lymph node involvement, mutational status and/or vascular invasion and focus on few significant factors such as clinical and pathological metastatic lymph node involvement, extension of vascular invasion, multifocality of PTC, encapsulated follicular variant of papillary thyroid carcinoma and, considering FTC, minimally invasive follicular thyroid carcinoma (Table 1).

The 2015 ATA GL still classify patients as a low, intermediate and high risk of recurrence. Patients having intrathyroidal tumor, ≤5 lymph node micrometastases, below <0.2 cm in diameter fulfill a low-risk criteria only when they are diagnosed with classical of follicular PTC variant or minimal invasive FTC. The intermediate-risk group involve PTC tumors with aggressive histology, RAI-avid DTC cervical foci outside thyroid bed, microscopic extrathyroidal extension, vascular invasion or >5 metastatic lymph nodes 0.2–3 cm in size. Patients showing aggressive PTC variants also belong to this group. Unfortunately, ATA do not specify the risk for PTC diffuse sclerosing variant, because its prognostic implication is controversial. This variant is characterized by a higher rate of local and distant metastases at presentation and lower DFS than classical PTC. Nevertheless, the overall mortality is rather low, with 10-year disease-specific survival of nearly 93%. However, for patients’ safety, we propose to classify this variant to a high-risk category. Finally, the patients with gross extrathyroidal extension, lymph node metastases >3 cm in diameter, distant metastases (M1) or after incomplete tumor resection constitute a high-risk group. The risk of recurrence varies according to these groups. One could guess, because ATA does not define it, that it is lower than 10% for low-risk subjects, between 10 and 30% for intermediate-risk patients and above 30% for the high-risk group. We believe that the risk has to be unequivocally established by prospective studies. Interestingly, the risk of structural disease recurrence is depicted rather as a continuum than clearly separated 3 groups.

It has to be stressed that both clinical and pathological features listed in 2015 ATA GL risk stratifications systems may be subjective and vary in evaluation from institution to institution. Accumulation of pathological features, which are subjective in interpretation without established standardized criteria, may impact on clinical management.

Minimal extrathyroidal extension (mETE) is an important factor present in the ATA Risk Stratification System and Risk of Structural disease recurrence. It points pT3 tumors to an intermediate-risk group. ETE is characterized by carcinomatous involvement of the sternothyroidal muscle or perithyroid soft tissue. This factor is evaluated and diagnosed by a microscopic examination. Individual pathologists’ interpretation and disagreement on histological criteria for mETE influences inter-observer variation in identification of mETE. The study performed among 11 thyroid experienced pathologists, based on analysis of 69 scanned slides selected from potentially controversial cases with mETE, presented that the overall strength of agreement in identifying mETE was slight (kappa coefficient factor 0.14). Most pathologist interpreted as mETE perithyroidal fat, skeletal muscle, nerve and thick-walled vessel involvement. Less than half of pathologists also reported using desmoplasia as a criterion for diagnosis of mETE. Inter-observer concordance of the identification of mETE between expert pathologists was moderate to poor (Su et al. 2016). Thus, one has to be aware that mETE is a controversial prognostic factor in PTC and in ATA risk stratifications schemes.

Clinical N0 or N1 evidence is identified by imaging or either palpation preoperatively or intraoperatively. Sonographic examination of lymph nodes among others depends on clinician’s skills, sonographic equipment and patient’s body. These factors are difficult to standardize. Reproducibility of palpation performed by surgeon preoperatively or intraoperatively is related to different factors, which made this examination subjective.

Surgical completeness of lymph node dissection, macroscopic examination and careful microscopic assessment influence the number of positive lymph nodes. On the other hand, micrometastases can be missed by a pathologist. On each stage, both surgical and pathological examinations errors can be made and preclude adequate lymph node assessment (Urken et al. 2016).

Another important issue is the lack of pathological criteria for evaluation of extranodal extension (ENE). ENE is a factor, which increases the risk of recurrences, therefore, the standardization of pathological designation of ENE in the future studies is particularly important (Urken et al. 2016). Ito et al. (2009) analyzed ENE in 2 age groups of patients >55 years old and <55 years old. In the first group, only large lymph node metastases in the lateral cervical compartment affected DFS, whereas in the younger one the number of metastatic nodes, ENE and size of metastatic nodes influenced DFS (Ito et al. 2009). According to the College of American Pathologists Protocol for the Examination of Specimens from Patients with carcinomas of the Thyroid Gland from 2017 (Seethala et al. 2017) the presence or absence of ENE should be noted in a pathology report, but the details of pathological criteria required for diagnosis are not explained. The agreement between 11 experts in thyroid pathology in histological diagnosis of ENE was only fair with kappa coefficient factor of 0.35. The study showed that pathologists utilized different histologic criteria for ENE diagnosis. While most pathologists used a criterion ‘tumor beyond nodal capsule’, some of them needed desmoplastic stromal reaction outside node to diagnose ENE. The full agreement was not reached in identification of perinodal skeletal muscle, nerve or thick-walled vessel (Du et al. 2016). The lack of concordance in identifying ENE could impact on the disease prognosis, patient care and interpretation of clinical studies that evaluating the influence of ENE on the disease prognosis. Noteworthy, the level of identification of ENE in a group of general pathologists could be lower than one in the cited paper.

Taking into a consideration the recent data regarding a possible impact of BRAF (Xing et al. 2013, 2015) and TERT mutations (Melo et al. 2014, Xing et al. 2014) on DFS and PTC disease-specific mortality, the authors of the guidelines believe that mutational status may improve risk estimation but when interpreted in the context of other clinical and histopathological risk factors. However, this statement requires a prospective validation.

The lack of prospective studies

According to the 2015 ATA GL, an initial recurrence risk estimates should be continually modified during follow-up, because the risk of recurrence and disease-specific mortality can change over the time as a function of the clinical course of the disease and the response to the therapy. This is a strong recommendation. However, the quality of evidence is low because to date there is no prospective study, which validates the data coming from retrospective analyses discussed above. The question arising is whether the use of dynamic risk stratification is safe in patients initially categorized as high risk? The diagnosis of an excellent response in these patients, demonstrating according to TNM classification a higher risk of cancer-related death, may influence their further follow-up to be less strict. Such situation may result in a delay in the diagnosis of DTC recurrence. Therefore, the necessity to carry out prospective studies is our important task for the nearest future.

Dynamic or delayed risk stratification

The idea proposed by Tuttle et al. (2010) and currently recommended by the newest 2015 ATA GL concerned a continuous = dynamic risk stratification to evaluate recurrence risk and adjust patient’s management to it. However, some papers, analyzed in this review, proposed the term ‘DRS’ (Castagna et al. 2011, Hong et al. 2014, Kowalska et al. 2016) probably in opposite to an ‘initial risk stratification’. One should ask the question whether ‘DRS’ means exactly the same as ‘dynamic risk stratification’. DRS in fact reflects a single-point risk assessment, whereas dynamic risk stratification is a real-time evaluation during the whole follow-up that incorporates an individual response to therapy, what is necessary for the personalized approach to ongoing management. Tuttle et al. (2010) proposed to base the assessment of treatment response on clinical data obtained during first 2 years of follow-up but 2015 ATA GL suggest to describe the clinical status at any point during follow-up.

Differences between published studies

There are some issues concerning retrospective studies, described in this review that has to be discussed.

All are single institution studies so a selection bias has to be considered. The time, when the treatment response was evaluated varies across the studies (Table 3). In some papers the authors analyzed all clinical data obtained during first 2 years of follow-up to assess treatment response (Tuttle et al. 2010, Momesso et al. 2016, Park et al. 2017), whereas the others carried out risk restratification in different single time points between 8 and 24 months after initial therapy (Castagna et al. 2011, Hong et al. 2014, Jeon et al. 2014, Kowalska et al. 2016, Morosán et al. 2016). One may ask the question which data are more accurate for risk restratification: these collected during the first 2 years of follow-up or those obtained during single assessment between 8 and 24 months after initial therapy?

Another important issue is related to different follow-up protocols used in distinct centers and outcome variables of reported studies. The number of follow-up visits and diagnostic examinations may influence the diagnosis of cancer recurrence. Considering outcome variables, some of papers analyzed the percentage of patients with NED or persistent disease, recurrence rates, and the number of cancer-related deaths (Tuttle et al. 2010, Castagna et al. 2011, Kowalska et al. 2016, Morosán et al. 2016, Momesso et al. 2016, Park et al. 2017), whereas other studies performed PFS or DFS analyses (Hong et al. 2014, Jeon et al. 2014, Tavarelli et al. 2017) to compare the patients with different stages of initial stratification or treatment response. We have to be aware that important limitation of PFS or DFS is how closely patients are followed. If visits are rare the diagnosis of cancer recurrence may be delayed.