Summary of results
This rigorous systematic review and meta-analysis has identified that in patients with sMTC who had an excellent response to their initial treatment, those with T3/4 disease, the presence of nodal disease and AJCC stage III/IV disease are at a higher risk of recurrence. We did not find evidence that either somatic RET mutations or the specific RET M918T were associated with recurrent disease.
This is broadly consistent with other literature, including another recently published systematic review by Vuong et al. [
7]. However, there are some differences, particularly with respect to
RET and
RAS mutations. One key finding from this review is that without a standard definition of disease recurrence in sMTC, the datasets available on this topic are heterogeneous. Much of the literature fails to differentiate recurrent from persistent disease which are two very different subsets of groups with different management strategies. This significantly contaminates the literature on disease recurrence. Furthermore, as can be seen in Fig.
1, a large number of studies do not report recurrence with respect to clinico-pathological characteristics in a way that allows for the pooling of data and only one author contacted provided data in this manner.
Clinicopathological characteristics
Whilst it is not surprising that patients with later stage disease are more likely to experience adverse outcomes in comparison to earlier stage disease, this is the first study to our knowledge to show that this is the case in sMTC in the context of recurrence. The majority of sMTCs present with locoregional and metastatic disease [
2] and therefore many patients are already in higher risk categories for recurrence. Our findings are consistent with a large-scale analysis of the American Surveillance, Epidemiology and End Results (SEER) Program dataset. This has previously shown that survival correlates with local, regional and distant disease [
18] and therefore it is reasonable to expect AJCC disease stage correlates with recurrent disease. What is still unanswered is which aspect of staging may be the most important factor. In our meta-analysis, stage III and IV AJCC staging had the highest odds ratio for recurrence rather than T3 or 4 disease or the presence of nodal disease alone. However, we cannot comment on degree to which each contributes to the risk of recurrence.
Mutational analyses
We did not demonstrate that patients with somatic
RET mutations or the specific mutation
RET M918T predicted a higher risk of recurrent disease. This contradicts the findings of a recent meta-analysis by Vuong et al
. [
7] which found that any
RET mutation including M918T is associated with recurrent disease. There are a number of possible explanations for this discrepancy. Firstly, their definition of recurrence differed from our own. We specifically investigated recurrence as defined in our methods, whereas Vuong et al
. included persistent disease in their analysis of recurrence. We elected not to class ‘persistent’ disease as recurrence, as the presence of persistent disease itself selects a group of patients who will already be managed differently once persistent disease is established. Secondly, the only two studies with clear positive results in their analysis were from Zedenius et al. [
19] and from Elisei et al. [
20]. Neither of these studies investigated recurrent disease but censored their outcomes if patients were ‘not cured’ or had ‘persistent disease’. This also affected other studies in their analysis by Cho et al. [
21] and Moura et al
. [
22] and Cavedon et al. [
23]. Finally, Chuang et al. [
24] which had less than 10 patients was included in their study but was excluded in ours.
Two studies reported only
RET M918T mutational status. These were included in the pooled analyses for both the overall
RET and
RET M918T mutation status [
10,
14]. Their removal from the overall
RET mutation analyses did not affect the significance of the result, although it did reduce I
2 to 0%.
We attempted to pool data on
RAS mutations. However, no study was eligible for inclusion as none reported on recurrent disease as we defined it. Of the three studies included in the Vuong et al
. meta-analysis, two [
25,
26] were contaminated with cases of persistent disease and therefore had a different case definition to the current study, and the other did not measure recurrent disease [
23].
Impact on initial surgical management and adjuvant therapy
The British Thyroid Association recommends that patients with sMTC should be offered a total thyroidectomy and central nodal clearance with lateral neck clearance if there is evidence of central nodal disease or ipsilateral lateral neck involvement [
5]. Management of the contralateral neck is more controversial. The American Thyroid Association management guidelines suggest offering a contralateral neck dissection if basal calcitonin levels are > 200 pg/mL [
1]. However, the British Thyroid Association do not make such an assertion noting that contralateral neck dissection may increase the chance of biochemical cure but have a limited impact on survival [
5]. We have found that patients with T3/T4, N1 and stage III/IV disease are at increased risk of recurrence in those who have had an excellent response to treatment. Whilst decision-making in locally advanced disease can be straightforward with the need for radical resection of loco-regional disease, it remains to be seen if risk factors can be used to stratify patients into groups that may benefit from more aggressive surgical management.
Limitations
Firstly, despite a large number of studies potentially eligible for inclusion on full text review, we were only able to extract data from 10 manuscripts. We found that whilst studies often included a number of outcome variables and may have reported recurrence, more often than not, they failed to report results in a way amenable to pooling for meta-analysis. For example, some studies provided unseparated data on cases of sporadic and familial MTC, whilst others did not report recurrence. Where recurrence was reported, it was often not reported in a way which allowed data to be pooled with respect to individual clinicopathological or mutational characteristics, or was confounded by persistent disease. We attempted to mitigate this by contacting authors for further data. We contacted 34 groups (totalling 39 manuscripts) offering the opportunity for inclusion of their work in this analysis. We received three replies with only one group able to provide data. This explains the absence of large datasets, including recently published studies [
27] from groups who have contributed valuable work in this field. As a result, there may be some degree of publication bias given how few studies were included. However, this could not be formally assessed as there were inadequate numbers of included studies to accurately assess funnel plots or to perform regression-based assessments. Therefore, we are unable to comment on when patients recur, and patterns of recurrence. What is clear is that if a more in-depth risk profiling of recurrence in this rare disease is going to be established, it will require a multi-centre collaborative approach with open data sharing.
Secondly, there are large confidence intervals for the clinico-pathological characteristics we have explored. The exception for this is the mutational analyses where these intervals are smaller. Whilst we can be confident in the data presented, we have refrained from making firm assertions regarding the strength of these associations. We certainly cannot, from this data, explore the relationship between these characteristics and disease recurrence.
Thirdly, with recent advancements in imaging, more modern studies may be more likely to identify recurrence. Therefore, there may be an underestimation of recurrence in the earlier literature. However, only three of the average follow-up times recorded was less than five years and all were more than 2.5 years. Therefore, many patients with recurrent disease would have had time to develop identifiable disease to be included in this study.
Additionally, it was unclear how some cases were collected, such as whether they were consecutive. This could be particularly important, for example, in data looking retrospectively at RET mutations that rely on the availability of stored tissue. It may be that more tissue is available in larger tumour specimens that themselves may indicate more aggressive disease and therefore be subject to selection bias.
Finally, one potential criticism of this review is the omission of calcitonin from analyses. Large series suggest that the failure to achieve biochemical remission is an important predictor of recurrent disease in familial and non-familial disease [
28]. However, throughout this review, we have been concerned regarding the definitions of recurrence throughout the literature. A failure to achieve biochemical remission within 6 months suggests persistent disease, as opposed to recurrence following biochemical cure. Therefore, this group of patients are a subgroup in their own right that merit investigation regarding factors contributing to disease recurrence. Chen et al.have recently suggested that the early ratio of pre and post-operative calcitonin can predict which patients develop structurally identifiable disease following a biochemically incomplete response to treatment and this requires validation in other cohorts [
29]. With respect to allcomers with sMTC, we suggest that in addition to overall and disease-free survival, progression-free survival would be a particularly important outcome measure to consider in any future trials. This will at least partly account for stable persistent vs progressive persistent disease in patients with a biochemically incomplete response to treatment.