Postoperative biochemical remission of serum calcitonin is the best predictive factor for recurrence-free survival of medullary thyroid cancer: a large-scale retrospective analysis over 30 years

Patient selection

We retrospectively reviewed the medical records of 331 patients who were diagnosed with MTC and followed in the years 1982–2012 at Seoul National University (SNU) Hospital (n = 165) or Severance Hospitals (n = 166) in Seoul, Korea. Patients consisted of 122 males and 209 females with a mean age of 47·3 ± 14·5 years (median, 48·0 years; range, 7–81 years). The median follow-up period was 4·6 years (range, 0·2–30·8 years). Pathology data, such as tumour size, extrathyroidal extension (ETE) and lymph node (LN) metastasis, based on the World Health Organization's International Histological Classification of Tumors,15 were obtained. Tumour size was categorized into the following groups: tumours ≤1 cm, tumours 1·1–2 cm and tumours >2 cm. LNs were also assessed based on positive or negative LNs, and nonresected LNs were categorized as negative LNs. Genetic tests for RET proto-oncogene mutation (exons 8, 10, 11, 13, 14, 15, 16), which was start to implement frequently since 2008, were performed in 172 patients, and family history of MTC or the presence of phenotypes of multiple endocrine neoplasia type 2 (MEN2) was obtained by reviewing the medical records for 310 patients. To evaluate changes in pathological findings over time, the patients were classified into four groups based on the year of diagnosis: 1982–2000 (n = 71), 2001–2005 (n = 55), 2006–2010 (n = 132) and 2011–2012 (n = 73). For comparison of prognosis, we regrouped the time periods based on the similarity of clinocopathological characteristics presented (1982–2005 vs 2006–2012).

This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of SNU Hospital (IRB No. H-1301-068-459) and Severance Hospitals (IRB No. 3-2013-0306).

Follow-up protocol and definition of disease status

Measurement of preoperative serum calcitonin levels and neck ultrasound or computed tomography (CT) were carried out for most patients. Neck ultrasonography was regularly performed at the beginning of 2000. During postsurgical follow-up, the measurement of serum calcitonin levels and neck ultrasound were carried out between 2 and 6 months after initial surgery and then checked annually. For patients with calcitonin levels >10 pg/ml or those showing an increase in calcitonin levels, the presence of recurrence or distant metastasis was evaluated by imaging modalities including abdominal or chest CT or ¹⁸F FDG PET CT. Serum calcitonin measurement was started since 1994. At the SNU Hospital, it was measured using immunoradiometric assay (IRMA; CIS-Biointernational, Gif-Sur-Yvette, France) from 1994 to the present, while at the Severance Hospital, the measurement of serum calcitonin was performed using immunoradiometric assay (CT-USA-IRMA; Biosource, Nivelles, Belgium) from 1994 to 2009; then, it was replaced by a chemiluminescence DPC immunoassay (Immulite 2000; Siemens, Gwynedd, UK). A cut-off for calcitonin levels was 0–10 pg/ml for immunoradiometric assay and 0–8·4 pg/ml (male) or 0–5 pg/ml (female) for chemiluminescence DPC immunoassay. Functional sensitivity and reference interval of each kit were similar.16 Preoperative calcitonin levels were available in 254 patients, and we could not obtain them in 77 patients.

We assessed postoperative disease status within 6 months after surgery (po) and at the last follow-up. The patient was classified as being in biochemical remission (BCR) when the serum calcitonin level was <10 pg/ml and no structural disease was found by neck ultrasound or other available imaging modalities. Biochemical disease (BCD) was defined as when serum calcitonin level was ≥10 pg/ml with no evidence of structural disease, whereas clinical disease (CD) was defined as when there was structural evidence of residual disease with any imaging modalities. Serum calcitonin level was >10 pg/ml in all patients with CD.

At the last follow-up, CD included a status of either structural recurrence or persistence. The term ‘CD/structural recurrence’ was defined as newly identified structural disease in a patient previously classified as postoperative-BCR (po-BCR) or po-BCD. The other patients with CD, who showed persistent structural disease during follow-up, were classified as ‘CD/persistence’.

Statistical analysis

Data were expressed as mean ± standard deviation (SD) or n (%) for descriptive statistics. All data were analysed with IBM SPSS statistics software (version 20.0; SPSS Inc., Chicago IL, USA). A t-test or one-way anova was used for continuous variables. Pearson's chi-square test or logistic regression analysis was used for categorical variables. Kaplan–Meier survival analysis was used to estimate the overall recurrence and survival rate. The Cox proportional hazard (Cox PH) regression model was used for the identification of prognostic factors for overall recurrence and mortality. In all cases, a P value <0·05 was considered statistically significant.

Secular trends in clinicopathological characteristics of MTC

The changes in baseline clinicopathological characteristics according to the time periods are shown in Table 1. The mean age of subjects increased over time, but the percentages of males remained similar. Although the RET proto-oncogene tended to be tested more frequently with time (from 35·2 to 54·8%, P = 0·004), the frequency of positive RET mutation (from 68·0 to 27·5%, P = 0·001) or hereditary MTC (from 33·9 to 15·1%, P = 0·020) decreased. The 12 subjects, 21·1% of patients with positive RET mutation, were diagnosed initially by family screening using genetic studies, and these percentages remained constant over time (from 17·6 to 27·3%, P = 0·778). The mean tumour size (from 2·5 cm to 1·7 cm, P < 0·001) and the percentage of subjects with ETE (from 52·0 to 26·0%, P = 0·026) also decreased. However, the percentage of LN metastasis and distant metastasis remained constant over time with overall percentages of 45·8% and 8·0%, respectively. On the other hand, prophylactic central LN dissection was performed consistently over 30 years in Severance Hospital, while it began in 20038 in SNU Hospital. When we separately analysed those results in each institutes, there was significant decrease in LN metastasis as well as the proportion of positive LN in Severance Hospital, but not in SNU Hospital (Table 1).

Then, we compared the long-term outcomes using Kaplan–Meier analysis (Fig. 1). The 5-year overall recurrence rate was 14% (Fig. 1a) and significantly decreased (10% vs 18%, P = 0·031, Fig. 1c) in the 2006–2012 group than in the 1982–2005 group. However, the 5-year survival rate was 92% (Fig. 1b) and did not differ between the groups (92% vs 92%, P = 0·929, Fig. 1d). The 10-year overall recurrence rate and survival rate were 35% and 87%.

Long-term outcome of MTC and its related clinicopathological characteristics

We assessed the disease status at the last follow-up according to the postoperative disease status within 6 months after surgery (Fig. S1). Among 188 subjects with po-BCR state, 94·7% maintained BCR status. However, 2·7% have progressed to CD, in which 20·0% re-entered to BCR status after reoperation and 80·0% maintained CD status irrespective of further treatment received (operation n = 3). Among 56 patients with po-BCD, 51·8% maintained BCD status and 48·2% have progressed to CD. Among them, 7·4% achieved final BCR status, 33·3% re-entered to BCD status and maintained after operation, but 59·3% progressed CD status irrespective of further treatment received (operation n = 12). In 47 subjects with po-CD status, 2·1% or 6·4% re-entered to BCR or BCD status after operation, respectively. 91·5% maintained CD status regardless of the treatment applied (operation n = 25).

To evaluate prognostic factors for the long-term outcome of MTC, we compared the clinicopathological features of BCD, CD/structural recurrence or CD/persistence groups with the BCR group at the last follow-up (Table 2). The percentages of males and frequencies of ETE, LN metastasis and distant metastasis were higher in the other three groups than those in the BCR group. Interestingly, the po-BCR rate of the BCD or the CD/structural recurrence group was significantly lower than the BCR group (10·9% or 20·0%, respectively, vs 98·4%, both P < 0·001). No patients in the CD/persistence group achieved po-BCR. The death group showed similar clinicopathological features to the CD/persistence group; the percentages of distant metastasis, LN metastasis and ETE, and preoperative calcitonin level were significantly higher, and mean tumour size was larger than that in the BCR group (Table 2).

Prognostic factors related to long-term prognosis in MTC patients

We evaluated prognostic factors for overall recurrence and mortality using Cox PH models in subjects diagnosed after 1994, since regular measurement of serum calcitonin levels was started in 1994 (Table 3). Using univariate analysis, male gender, tumour size >2 cm, LN metastasis, ETE and failure to achieve po-BCR (no po-BCR) were poor prognostic factors for overall recurrence. However, after multivariate analysis, only no po-BCR (HR = 58·04, 95% CI 7·14–472·11; P < 0·001) remained a significant and strong variable affecting overall recurrence (Table 3, Model 1). Multivariate analysis excluding no po-BCR showed that LN metastasis (HR = 7·18, 95% CI 2·56–20·12; P < 0·001) and ETE (HR = 3·09, 95% CI 1·34–7·12; P = 0·008) were the risk factor for overall recurrence (Table 3, Model 2). When we included all subjects from 1982, the risk effects of those factors were not different (data not shown).

Most patients (95·3%) with po-BCR maintained BCR status except for 9 patients (Table 4). Nine patients with po-BCR showing a progression to BCD (n = 5) or CD (n = 4) are characterized in Table S1. Of the four patients who progressed to CD, two showed RET gene mutation with American Thyroid Association (ATA) risk B or C,17 and the other two patients had tumours larger than 2 cm.

For overall mortality, male gender, increasing age, tumour size >2 cm, LN metastasis, ETE and distant metastasis were significant poor prognostic factors using univariate analysis (Table 3). Using multivariate analysis, male gender (HR = 3·18, 95% CI 1·18–8·56; P = 0·022), tumour size >2 cm (HR = 18·33, 95% CI 2·35–143·06; P = 0·006) and distant metastasis (HR = 4·00, 95% CI 1·31–12·21; P = 0·015) remained as prognostic factors for overall mortality.

There was no disease-related mortality in subjects who achieved po-BCR, so the risk effects of no po-BCR on mortality could not be calculated. There was no mortality in all 95 subjects with tumours <1 cm, and only 2 of 108 (1·9%) patients with tumour size 1–2 cm died, while 23 of 102 (22·6%) patients with tumour size >2 cm died from MTC. Of the 25 patients showing initial distant metastasis at diagnosis, only two patients achieved BCD at last follow-up, whereas the others maintained CD/structural recurrence or CD/persistence and 14 of them were dead. The sites of distant metastasis in the two cured patients were all para-aortic LN, and both patients achieved BCD after mediastinal dissection (Table S1).

To show whether the prognostic factors have been changed over time, we did the same analysis in each period. Prognostic factors for overall recurrence or mortality in each period (1994–2005 or 2006–2012, Table S2) were similar for the entire study periods (1994–2012, Table 3). No po-BCR was associated with the strongest risk of recurrence in the 1994–2005 group; however, recurrence was not observed in patients who achieved po-BCR in 2006–2012, so no po-BCR as the risk factor for recurrence could not be evaluated in the 2006–2012 group. Other than no po-BCR, LN metastasis was the only significant factor correlated with overall recurrence (HR = 11·14, 95% CI 1·25–99·19; P = 0·031) in 2006–2012. In the same manner, multivariate analysis excluding no po-BCR in 1994–2005 also showed that LN metastasis was the risk factor for overall recurrence (data not shown; HR = 7·19, 95% CI 2·18–23·74, P = 0·001).

Failure to achieve postoperative BCR (no po-BCR) was the strongest predictive factor for recurrence

Although overall mortality was unaltered, overall recurrence improved with time, and the patients who did not achieved po-BCR showed greatest risk of recurrence. When we subanalysed the subjects according to the po-BCR status, overall recurrence was also not significantly different between the periods (Fig. 2a) as well as overall mortality (Fig. 2b). These findings differ from the results shown in Fig. 1a, suggesting the importance of the increased percent of po-BCR rate itself on the improvement of recurrence of recent period. Then, we compared the rates of po-BCR achieved according to pathological characteristics between 1994–2005 and 2006–2012 (Fig. 2c–e). The rate of po-BCR was higher in 2006–2012 in all patients, regardless of pathological aggressiveness. However, for patients with tumour size <0·5 cm or >2 cm, the rates were similar for the two time periods.

Interestingly, the long-term outcome was very different according to the postoperative status of serum calcitonin levels; patients who achieved po-BCR were very likely to remain BCR (95·2%) and showed no disease-related mortality, while patients who did not achieve po-BCR resulted CD in 57·3% and showed 21·4% of a disease-related mortality at the last follow-up (Table 4). In addition, the tumour size was larger, the preoperative calcitonin level was higher, and the percentage of LN metastasis, ETE or distant metastasis was higher in patients who did not achieve po-BCR. In multivariate analysis, tumour size >2 cm (odds ratios (OR) 3·86, P = 0·008), LN metastasis (OR 12·98, P < 0·001) and ETE (OR 8·12, P < 0·001), which showed prognostic effects on recurrence in univariated analysis (Table 3), showed significant negative predictive factors for po-BCR (data not shown).