BRAF^V600E Mutation is Associated with an Increased Risk of Papillary Thyroid Cancer Recurrence

Abstract

Background

Papillary thyroid carcinoma is the most common endocrine malignancy and one of the most common cancers worldwide. However, the optimal timing and frequency of surveillance to assess for recurrence remain undetermined. As the incidence of thyroid cancer continues to rise worldwide, identifying risk factors for recurrence and investigating intervals and durations of surveillance are paramount to adapt treatment and follow-up plans to high-risk individuals and to reduce interventions for low-risk patients.

Methods

Our dataset included an unselected cohort of papillary thyroid carcinoma (PTC) patients who underwent a total thyroidectomy (or unilateral then completion thyroidectomy) at a single institution from 2000 to 2007. BRAF genotyping was performed on available specimens by a validated PCR-based assay. Pathologic structural recurrence was the primary outcome. We performed univariate and multivariable analyses to identify predictors of cancer recurrence.

Results

In total, 599 patients underwent complete resection of the thyroid gland for PTC. The cohort was young (mean age 45.0 years), predominately female (n = 462, 76.9%), and median follow-up was 10.3 years (IQR 5.4–12.2). Recurrence occurred more commonly in the BRAF^V600E group (18.6 vs. 9.9%, p = 0.02). BRAF independently predicted PTC recurrence (HR 2.81, p = 0.006).

Conclusions

BRAF mutation is an independent predictor of papillary thyroid carcinoma long-term recurrence. Understanding molecular characteristics of individual thyroid cancers may help risk-stratify patients and direct them toward more appropriate initial care and long-term surveillance strategies.

Introduction

Papillary thyroid carcinoma (PTC) is the most common endocrine malignancy and one of the most common cancers worldwide [1, 2]. Papillary thyroid carcinoma is often treated effectively through surgical resection, but recurrence is common, ranging from 20 to 30%, and may occur years after index diagnosis and treatment [3]. The long-term morbidity and mortality associated with PTC remain poorly understood, and the optimal timing and frequency of surveillance to assess for recurrence remain undetermined. As the incidence of PTC continues to rise worldwide with a continued overall excellent prognosis [1], there has been a shift toward management that limits morbidity of overtreatment [4]. Identifying risk factors for recurrence and investigating intervals and durations of surveillance are paramount to adapt treatment and follow-up plans to high-risk individuals and to reduce interventions for low-risk patients.

Over the past decade, BRAF^V600E and other molecular markers have emerged as potential risk factors for PTC-related death and PTC recurrence. BRAF^V600E is a protooncogene mutation that encodes an upregulated active form of the B-Raf protein leading to an overactive MAP kinase signaling pathway and contributing to tumorigenesis [5, 6]. The association between BRAF^V600E mutation and mortality remains unclear with some research, suggesting that BRAF mutation positivity is associated with increased PTC-related mortality [7], while other evidence suggests the relationship between BRAF mutation positivity and mortality is not as strong when the appropriate covariates are included [8].

Some evidence exists to support an association between BRAF^V600E mutational status and PTC recurrence [9]. Other studies have indicated that the BRAF mutation is not associated with an increased risk of recurrence [10, 11] or with a more aggressive clinical course [12, 13]. As hospitals and health systems work to deliver high quality and affordable health care, it is critically important to develop appropriate risk stratification to understand who is at risk of recurrence and warrants continued surveillance. Multiple scoring systems are available to offer prognosis about disease-specific survival for PTC [14-16]; however, given that few patients die of PTC, the use of these systems for risk stratification is limited. The ATA recurrence risk is the only formal scoring system specifically created to estimate the likelihood of recurrence for papillary thyroid cancer [17]. The 2015 ATA revised guidelines for the management of thyroid cancer underscore the need for improved stratification for the intensity and frequency of surveillance. BRAF testing is feasible and inexpensive and may be performed on both preoperative FNA specimens and surgical pathologic specimens.

The purpose of this study is to investigate the association between BRAF and PTC long-term recurrence and to assess the additional utility of BRAF mutational status in improving existing PTC risk stratification systems. Our previous work suggested that there may be an improvement in risk stratification when combining BRAF mutation status with other clinicopathologic information [18]. Herein, we present an extended follow-up of our unselected retrospective cohort study with a median follow-up over 10 years. This study is an updated cohort that includes a larger patient population and a longer duration of follow-up.

Material and methods

Population

Our dataset included an unselected cohort of patients with PTC who underwent total thyroidectomy (or unilateral thyroidectomy with subsequent completion thyroidectomy) at the Massachusetts General Hospital from 2000 to 2007. Patients’ demographics, clinical histories, and pathology results were retrospectively abstracted from the hospital electronic medical record. BRAF genotyping was performed on available specimens by a validated PCR-based assay with paraffin-embedded primary tumor specimens. A single-nucleotide primer extension PCR-based assay was used by the Massachusetts General Hospital Pathology Service.

Variables

Age was classified as both a continuous variable and also a categorical variable (<55 vs. ≥55 years of age). Tumor size was categorized according to AJCC-TNM 8^th edition tumor staging system [4]. Pathologic characteristics including lymphovascular invasion, extrathyroidal extension, lymph node positivity (lateral levels II–V and central level VI), and tumor multifocality were analyzed as dichotomous variables (either present or absent) and only included when specifically described in the pathology report. Tumor subtypes were categorized according to their respective associated risk of recurrence: low risk (follicular variant of PTC, cribriform-morular, warthin-like), normal risk (classical, solid, oncocytic), and high risk (diffuse sclerosing, tall cell). Radioactive iodine treatment was analyzed as a dichotomous variable. All missing demographic, clinical, pathologic, and mutational variables were excluded from the final analysis. Recurrence was defined as a new pathologically confirmed (fine needle aspiration or formal surgical pathology) papillary thyroid carcinoma or distant metastases more than 90 days from the index operation in a previously treated patient with a clinical disease-free interval. Serum thyroglobulin levels alone were not used to characterize recurrence. Person-time was defined as the interval from the operation until the time of first confirmed recurrence or until the last physician follow-up visit.

Analysis

Descriptive statistics were reported as means with a standard deviation for continuous variables and as proportions for categorical variables. Predictors of PTC recurrence were chosen based on known risk factors. We calculated the number of person-years at risk for each patient from the date of their operation until the date of recurrence or death or until their last physician follow-up. For the univariate analysis, we performed tests of equality across strata to determine whether or not to include the predictor in the final multivariate model. For categorical variables, we used the log-rank test of equality. For continuous variables, we used a univariate Cox proportional hazard regression and included the predictor in the multivariable model if the test has a p value of 0.25 or less. We subsequently tested each variable within the analysis for interaction with the other remaining variables and included statistically significant interaction terms within our final multivariable model.

We used Kaplan–Meier analysis to estimate the risk of recurrence for papillary thyroid cancer. The proportional hazard assumption was validated through the Schoenfeld residuals. Two-sided p values of 0.05 were considered statistically significant. We subsequently calculated the receiver–operator curve (c-index), a measure of model discrimination, for different classification systems with and without BRAF^V600E status. For the c-index, a value of 0.5 represents no predictive ability and 1.0 represents perfect discrimination between non-recurrent and recurrent cases. Data were analyzed using Microsoft Excel and Stata version 15 (StataCorp). The study was approved by the Internal Review Board of Massachusetts General Hospital.

Results

An unselected cohort of patients (n = 599) who underwent complete resection of the thyroid gland for PTC at Massachusetts General Hospital between 2000 and 2007 were included in the study. Mean age was 45.0 years (SD 14.5). The cohort was predominately female (n = 462, 76.8%). Mean tumor size was 1.79 cm (SD 1.22). Multifocality was present in 294 patients (49.0%). Median follow-up was 10.3 years [IQR 5.4–12.2]. Lymphovascular invasion was present in 36 (6.0%) tumor samples, and extrathyroidal extension was present in 88 (14.6%) tumor samples. A majority of the cohort received postoperative radioactive iodine treatment (n = 482, 81.5%). Normal risk was the most common subtype (85.3%). The American Joint Committee on Cancer, 8^th Edition (AJCC) tumor stage distribution was as follows: T1 71.1%, T2 19.1%, T3 6.8%, and T4 3.0%. Recurrence occurred in 78 patients (13.0%) patients during the study period. Death occurred in 24 (4.3%) of patients during the study period, six of which were attributable to thyroid cancer.

Of the 599 patients, tissue was available for BRAF testing for 504 samples (84%). Within our study population, 411 (69%) of samples successfully underwent BRAF genotyping. The BRAF^V600E mutation was present in 260 tumor samples. Recurrence occurred more commonly in the BRAF^V600E group (18.5 vs. 9.9%, p = 0.02, Table 1). There was no statistically significant difference in mortality between groups with and without the BRAF mutation (p = 0.21).

Table 1.

Demographic and clinical characteristics of patient cohort

Variable	All patients (n = 599)	Wild type (n = 151)	BRAFV600E (n = 260)	p value*
Age in years (SD)	45.0 (14.5)	43.4 (14.5)	45.8 (14.7)	0.14
Female, n (%)	462 (76.8)	121 (80.1)	194 (74.6)	0.20
Tumor size in mm (SD)	17.9 (12.2)	20.0 (13.9)	17.6 (12.0)	0.12
Multifocal, n (%)	294 (49.0)	72 (48.0)	123 (47.3)	0.89
Family Hx PTC, n (%)	28 (4.7)	7 (4.6)	16 (6.1)	0.52
Follow-up (years), median (IQR)	10.3 (5.4–12.2)	10.6 (6.2–13.1)	9.7 (3.9–11.2)	0.09
Lymph node metastases	198 (33.1)	45 (29.8)	97 (37.3)	0.12
Lymphovascular invasion, n (%)	36 (6.0)	13 (8.7)	18 (6.9)	0.52
Extrathyroidal extension (%)	88 (14.6)	15 (9.9)	49 (18.9)	0.02
Radioactive iodine	480 (81.5)	124 (83.2)	201 (79.1)	0.32
Subtype, n (%)				0.001
Normal risk (ref.)	509 (85.3)	112 (74.7)	247 (95.0)
Low risk	79 (13.2)	35 (23.3)	10 (3.9)
High risk	9 (1.5)	3 (2.0)	3 (1.1)
T stage (AJCC 8th), n (%)				0.97
T1	427 (71.1)	104 (68.9)	180 (69.2)
T2	115 (19.1)	31 (20.5)	50 (19.2)
T3	41 (6.8)	13 (8.6)	18 (6.9)
T4	18 (3.0)	3 (2.0)	12 (4.6)
Recurrence, n (%)	78 (13.0)	15 (9.9)	48 (18.5)	0.02
Time to recurrence in years, mean (SD)	3.0 (3.2)	2.85 (2.62)	3.33 (3.55)	0.91

Low risk = follicular variant of papillary thyroid cancer (PTC), warthin-like, cribriform-morular variant; normal risk = classical PTC, mixed follicular variant PTC, solid, oncocytic; high risk = tall cell variant, diffuse sclerosing variant

^*Represents comparison of wild type versus BRAFV600E

Multivariable Cox regression was performed, and variables associated with recurrence included BRAF^V600E mutation (hazard ratio [HR] 2.81 [1.34–5.89]), central or lateral lymph node metastases (HR 2.22 [1.31–3.77], and radioactive iodine treatment (HR 3.38 [1.03–11.1], Table 2). BRAF^V600E mutation status was independently associated with a shorter time to recurrence (Fig. 1). Each prediction model displayed some ability to discriminate the outcome of recurrence with c-indices ranging from 0.541 to 0.649. When BRAF mutation status was added to these different models, BRAF^V600E improved model discrimination (Table 3).

Table 2.

Factors associated with papillary thyroid cancer recurrence

Variable	Hazard ratio	p value	Confidence interval
BRAFV600E mutation	2.81	0.006	1.34–5.89
Age ≥ 55 years AJCCa T stage (T1 is ref.)	1.88	0.337	0.52–6.79
T2	0.81	0.595	0.38–1.74
T3	1.88	0.177	0.75–4.72
T4	1.22	0.846	0.16–9.12
Lymph node metastases	2.22	0.003	1.31–3.77
Radioactive iodine	3.38	0.045	1.03–11.1
Extrathyroidal extension	1.31	0.411	0.69–2.47

^aAmerican Joint Committee on Cancer, Tumor Stage

Fig. 1.

Recurrence for PTC patients by BRAF mutation status. Data represent Kaplan Meier curve for the proportion of PTC patients without recurrence by wild type (straight line) versus BRAFV600E mutation (dotted line). A higher proportion of BRAF BRAF V600E patients experience recurrence as compared to BRAF wild type patients

Table 3.

BRAF mutation improves model discrimination for predicting recurrence

Model	c-index (95% confidence intervals)	c-index with BRAF (95% confidence intervals)
AMES	0.541 (0.482–0.600)	0.617 (0.545–0.689)
AJCC-TNM 7th edition	0.606 (0.545–0.667)	0.657 (0.585–0.728)
AJCC-TNM 8th edition	0.604 (0.550–0.659)	0.671 (0.603–0.738)
ATA recurrence risk	0.649 (0.593–0.705)	0.682 (0.615–0.749)
MACIS	0.581 (0.504–0.657)	0.655 (0.579–0.731)

Discussion

Our study offers further evidence that the BRAF^V600E mutation is associated with an increased risk of recurrence for patients with papillary thyroid carcinoma. Our study provides one of the largest unselected cohorts (i.e. BRAF status was unknown at the time of surgery) of PTC patients (removing selection bias as a factor), and the longitudinal median follow-up extends beyond ten years. The BRAF^V600E mutation occurs in roughly 40–60% of papillary thyroid carcinomas and remains one of the most common genetic mutations [19]. Given the high survival associated with PTC and need for continued surveillance, our study highlights the critical importance of developing improved risk prediction models of recurrence and for exploring potential predictors of recurrence—including BRAF—as these efforts will help guide future management practice. Additionally, for those patients with recurrence, predicting which of those patients are truly at risk of mortality rather than local recurrence would surely add value to current prediction models.

Previous work has been controversial regarding the prognostic significance of the BRAF mutation. In a multi-institutional retrospective study of 2099 patients with PTC who underwent total thyroidectomy, PTC recurrence rates were higher in BRAF mutation-positive versus BRAF mutation-negative patients (47.7 vs. 26.0 per 1000 person-years) and remained significantly higher after adjustment for clinicopathologic features [9]. Further work demonstrated that BRAF was associated with decreased survival and other clinicopathologic characteristics including lymphovascular invasion and extrathyroidal extension [8]. Nevertheless, subsequent studies provided evidence to suggest that BRAF was not an independent predictor for worse outcomes for PTC patients [11, 20]. Many of these studies were retrospective and restricted to single institutions. Subsequent meta-analyses and multi-institutional studies were performed which demonstrated that BRAF mutation positivity predicts PTC recurrence. A meta-analysis of 2247 PTC patients from four published studies and two institutional cohorts demonstrated a higher likelihood of recurrence (odds ratio 2.09, p = 0.002) for patients with BRAF mutation [21]. Our finding that BRAF^V600E independently predicts recurrence is consistent with these previous studies.

Our finding that BRAF assessment adds value to risk prediction models for PTC is consistent with our previous study that explored the potential prognostic value of molecular markers in PTC patients [18]. In contrast to our prior work, our model did not find multifocality to be a predictor of recurrence. Additionally, new AJCC staging criteria (8^th edition) have been created since the previous cohort data was published, and the new AJCC model demonstrated a similar c-statistic at baseline and larger value with the addition of the BRAF mutation status, which suggests improved model discrimination. Interestingly, in a recent single-institution study of 231 patients with PTC tested for molecular markers, BRAF^V600E was not associated with recurrence or death [10]. Their study created a molecular risk score that included additional molecular markers (RET, MAPK, and pAKT), and recurrence was independently associated with the ATA high-risk category (HR 2.8 [1.3–6.0]) and the molecular high-risk signature (HR 5.4 [2.5–12.0]) in their multivariate model. While their study conclusion differs from our finding that BRAF^V600E is associated with recurrence, the value of molecular marker analysis in predicting recurrence is evident. Of the included models in our study, only the ATA was specifically designed to prognosticate about recurrence risk. Thus, it is not surprising that it is the ATA model that had the highest baseline c-statistic, indicating the model’s superiority in predicting recurrence.

In our study, radioactive iodine therapy was associated with an increase in the risk of recurrence. Nearly 80% of our patients received postoperative radioactive iodine treatment, and 94% of recurrences occurred within this group. While our data do suggest that receipt of RAI treatment was a predictor of recurrence, BRAF mutational status was unknown at the time of surgery (i.e. was done post hoc), and decisions regarding RAI treatment were made with clinicopathologic information. Therefore, the RAI-treated cohort likely represents a selected group of more aggressive patients with a higher likelihood of recurrence than the non-treated group. A previous cohort study did not identify an association between radioactive iodine and recurrence in their PTC cohort during a 27-year median follow-up [3].

There are limitations to our study. Our study was an unselected historical cohort of patients and was not explicitly randomized. Data were collected retrospectively and remain open to unknown selection bias. In the 504 patients with available tissue, BRAF genotyping was only successful in 411 tumor specimens. Therefore, assessment of the incremental value of BRAF genotyping was only available for 81% of persons in our cohort. BRAF testing is now available using IHC and is likely to be successful in essentially all cases provided sufficient viable tissue is present for assessment, at a reduced cost, and within 24 h of test initiation [22]. Moreover, telomerase reverse transcriptase (TERT) mutation analysis was not performed, and evidence suggests that the combination of TERT and BRAF mutations increased the aggressiveness of PTC [23]. Therefore, we do not know if additional mutations other than BRAF are driving the relationship between BRAF and PTC recurrence. While our focus was on structural recurrence, we acknowledge that having an actionable thyroglobulin level could have been useful to validate our marker of recurrence. Additionally, the extent of additional surgical exploration beyond a total thyroidectomy was not uniform among the cohort, and we were unable to validate retrospectively if lymph node sampling was done in a prophylactic or therapeutic manner. Additionally, we acknowledge that the heterogeneity of subtypes between BRAF and wild-type groups may have contributed to the differences in recurrence. We performed an analysis of our cohort after excluding the follicular variants (n = 72), and BRAF remained an independent predictor of PTC recurrence.

In summary, our data offer evidence that BRAF mutation status is an independent predictor of recurrence for over a decade. Similar to previous studies, recurrences commonly occurred in the first five years [3, 24, 25]. Additionally, the BRAF mutation status provides additional discrimination to existing PTC risk stratification models. This raises the question of preoperative genotyping. As we continue to work to identify patients who may undergo observation versus surgical intervention, additional data including molecular markers may be useful in risk stratification. Further investigation into additional markers and risk stratification models is warranted.