Prognostic Significance of Circulating RET M918T Mutated Tumor DNA in Patients With Advanced Medullary Thyroid Carcinoma

Abstract

Context:

Interpretation of calcitonin measurement to predict the prognosis of medullary thyroid carcinoma (MTC) requires multiple measurements over an extended time period, making it an imperfect biomarker for evaluating prognosis or disease behavior. Single circulating cell-free DNA (cfDNA) values have been shown to be a valuable prognostic marker for several solid tumors.

Objective:

We tested the hypothesis that cfDNA containing the RET M918T mutation could be detected in the blood of patients with advanced MTC whose tumor harbored an M918T mutation and would be able to predict overall survival more reliably than calcitonin.

Design:

The level of cfDNA containing RET M918T mutation was measured in the plasma of patients with MTC via droplet digital polymerase chain reaction.

Patients:

Patients had a confirmed sporadic MTC diagnosis, a serum calcitonin measurement >100 pg/mL, and tumor tissue biopsy results providing RET M918T mutation status. There were 75 patients included in this study, 50 of whom harbored an RET M918T mutation by tissue biopsy.

Results:

RET M918T cfDNA was detected in 16 of 50 patients (32%) with a positive tissue biopsy. The detection of RET M918T cfDNA strongly correlated with worse overall survival and more accurately predicted a worse outcome than calcitonin doubling time.

Conclusions:

Liquid biopsy is able to detect RET M918T mutations in patient plasma with high specificity but low sensitivity. In patients with established somatic RET M918T mutations, the allelic fraction of circulating tumor DNA is prognostic for overall survival and may play a role in monitoring response to treatment.

We found that detection of circulating tumor DNA strongly correlates with a worse survival outcome in patients with MTC.

Medullary thyroid carcinoma (MTC) is an aggressive form of thyroid cancer caused by germline mutations of the RET receptor tyrosine kinase (multiple endocrine neoplasia, type 2) or in the majority of sporadic MTC activating mutations of RET or RAS (1). The screening for germline RET mutations is a well-established paradigm that allows for prophylactic surgery in asymptomatic gene carriers (2). For patients with advanced MTC, the introduction of molecular targeted therapies has provided new treatment approaches. In sporadic MTC, somatic RET mutations are found in ∼40% of tumors, with a single mutation, M918T, accounting for nearly 70% of RET mutations (3). The presence of somatic RET mutation, particularly M918T, has been associated with worse prognosis in some studies (4, 5). Inhibiting the action of this key pathogenic tyrosine kinase may be one of the mechanisms of action that led to the US Food and Drug Administration approval of vandetanib and cabozantinib for the treatment of MTC (6, 7). Recent studies indicate that patients with an RET M918T somatic mutation respond better to cabozantinib treatment (8). However, for many patients somatic mutation testing of the primary tumor is not performed because of access to tumor tissue, or issues related to DNA amount or quality. Additionally, there is the risk that the mutation status of the metastatic lesions may have diverged from that of the primary counterpart (9, 10).

Liquid biopsy, the detection of mutations in circulating plasma cell-free DNA (cfDNA), can serve as a potential surrogate for analysis of existing tumor DNA. Liquid biopsy provides a potential method for evaluating existing tumor mutation status, and samples can be easily obtained at any time, such as before, during, and after treatments. In addition, measurement of the fraction of circulating tumor cfDNA relative to circulating normal cfDNA has been shown to be prognostic and predictive of response to treatment, recurrence, or progression in some cancers (11, 12). Several studies have shown measurement of circulating tumor cfDNA to be a sensitive and specific tumor biomarker in colorectal, ovarian, lung, and breast cancers (13–15). The clinical use of circulating tumor cfDNA analysis is gaining acceptance as a valuable tool for treatment decision-making (11, 16–18). The usefulness of cfDNA testing in the management of thyroid cancer is less clear. A single-center study using a targeted Taqman® polymerase chain reaction (PCR) approach found that cfDNA could be useful for detection of BRAF V600E mutation in plasma samples from patients with papillary thyroid carcinoma (PTC) and also demonstrated that cfDNA levels decreased after surgical removal of the tumor (19). A separate study using a digital sequencing approach suggested that tumor cfDNA shedding is less common in patients with PTC compared with other cancer types, but it included only four patients (13). A larger study performed in patients with the more aggressive anaplastic thyroid carcinoma (ATC) found a high concordance between liquid and tissue biopsy results when untreated patients were tested (20). Notably, there have been no reports describing the detection of circulating tumor cfDNA in MTC, which like ATC has a higher incidence of aggressive disease relative to PTC (21).

Based on the success of detecting cfDNA mutations in other cancer types, it is reasonable to hypothesize the presence of circulating tumor DNA in MTC. Furthermore, a noninvasive approach to identify tumoral mutation status is necessary when there is an absence of available tumor tissue or when metastatic tumor tissue is difficult to sample. Therefore, we examined the sensitivity and specificity of RET M918T cfDNA detection by droplet digital PCR (ddPCR) compared with conventional tissue genetic testing in a large, clinically relevant cohort for MTC.

Patients and Methods

Patients

All cases were derived from patients who were treated at the University of Texas MD Anderson Cancer Center for sporadic MTC. Patients gave written informed consent before collection of blood and tumor samples under institutional review board–approved protocols. Participants had evidence of biochemical persistent or recurrent disease and MTC tumor somatic mutation data or tumor tissue available to perform the analysis. TNM staging was based on seventh edition American Joint Committee on Cancer (AJCC) criteria (22). Date of diagnosis was defined as the date of pathologically confirmed MTC, either biopsy or initial surgery when the biopsy was indeterminate. Blood collection for cfDNA analysis was performed concurrently with routine standard of care measurements of blood calcitonin and carcinoembryonic antigen (CEA) concentrations. Vital status was censored at last contact or time of death. Overall survival was measured from the date of diagnosis until death from any cause. Calcitonin and CEA doubling times were determined by using the value obtained at the time of cfDNA sampling and three additional tests performed in the previous 2 years as recommended by American Thyroid Association guidelines (23).

RET M918T gene mutation analysis in MTC tumor samples

When possible, mutation detection was performed on the most recent tumor tissue available, with metastatic tissue used for 51 of the 75 patients. The median time between tissue and blood sampling was 26.1 months. Fifty-six patients had mutation analysis performed as part of the clinical care in a Clinical Laboratory Improvement Amendments–certified laboratory, 39 by next-generation sequencing, and 17 by pyrosequencing. For both methods mutation detection >10% was considered positive. Formalin-fixed, paraffin-embedded tissue was obtained for the remaining 19 patients to determine tumor mutational status by Sanger sequencing of RET exon 16 PCR products as previously described (24). The assignment of RET M918T as a somatic mutation was based on the lack of mutation in paired peripheral blood cells or normal thyroid DNA.

cfDNA isolation

Plasma was prepared from EDTA-treated blood within 4 hours of blood sample collection and frozen at −20°C to allow batch processing of DNA. Processing involved consecutive centrifugations of 2000 RPM for 10 minutes at 4°C to ensure plasma was free of contaminating cells. Before cfDNA isolation, frozen plasma was thawed on ice and then centrifuged to remove particulate. Cell-free DNA was isolated from plasma with the QIAamp Circulating Nucleic Acid Kit (Qiagen, Venlo, The Netherlands). A standardized plasma volume of 3 mL was adapted for each sample, with a 50 µL elution volume. DNA amount was quantified by Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA).

Droplet digital polymerase chain reaction

The detection of the RET M918T mutation was performed with the BioRad QX200 ddPCR system and the validated PrimePCR™ ddPCR™ RET M918T Mutation Assay (assay ID: dHsaCP2506896), which was multiplexed with an assay for detection of RET unmutated codon 918 (wild-type) (assay ID: dHsaCP2506897). Each reaction contained the following: 11 µL of 2X Supermix, 1.1 µL of 20X Assay 1 HEX (Wild-Type Allele), and 1.1 µL of 20X Assay 2 FAM (Mutant Allele) per reaction, for a total of 13.2 µL per reaction. The remaining 8.8 µL contained cfDNA sample. The cfDNA yields using the aforementioned protocol ranged from 0.2 to 6.4 ng/µL when measured by Qubit, allowing a median estimated input of 5 ng per assay. A nontemplate control of water was run for each assay, as well as positive (MZ-CRC-1 cell line) and negative (TT cell line) sheared DNA controls. Droplets were counted with the QX200 Droplet Reader and data analyzed in the BioRad QuantaSoft software package, with thresholds for FAM and HEX detection manually set based on results from the no-template control. We required a minimum of 8000 counted droplets. The allelic fraction (AF) was calculated with the ratio of channel 1 counts (M918T cfDNA) over channels 1 and 2 (M918T and M918WT) and the value standardized as a percentage.

Statistical analysis

The associations of clinicopathologic variables were evaluated with an unpaired Student t test with Welch correction for continuous variables and trending χ² or Fisher’s exact tests for categorical data. Overall survival was defined as the elapsed time from blood collection to death. Patients were censored at last follow-up if death had not occurred. Univariate regression survival analysis, hazard ratio determination, and Kaplan–Meier plots were generated with GraphPad Prism version 6.0 software (GraphPad Software, La Jolla, CA). Log rank Mantel–Cox tests were used to detect differences between groups. P values <0.05 were considered significant.

Results

Patient characteristics

Seventy-five patients with sporadic MTC with elevated serum calcitonin level (>100 pg/mL) and with somatic RET M918T mutation status determined through solid tissue biopsy testing (50 positive, 25 negative) were included in the analysis. Patient characteristics of subjects enrolled in this study are summarized in Table 1. The entire cohort of patients with MTC were analyzed based on the presence or absence of a somatic RET M918T mutation determined through tumor tissue biopsy. No significant differences were observed between these two groups related to age, primary tumor characteristics, stage at diagnosis, tumor marker levels at sampling, or whether they were undergoing therapy with vandetanib or cabozantinib at the time of cfDNA sampling (Table 1). We viewed the somatic RET M918T negative group as a control population for cfDNA assay validation, with the recognition that the rare acquisition of new RET mutations in metastatic disease has been previously reported (9, 10).

Table 1.

Demographic and Clinical Characteristics of Patients With MTC

	RET M918T (+) by Tissue Biopsy	RET M918T (−) by Tissue Biopsy	P
n (male/female) (%)	50 (24/26) (52)	25 (11/14) (44)	0.809
Median age at diagnosis, y (range)	47.8 (16.5–70.3)	49.7 (15.2–84.2)	0.136
Median age at sampling, y (range)	53.4 (19.0–73.0)	56.9 (22.2–84.9)	0.083
Primary tumor characteristics
Median size, mm (range)a	22 (4–80)	27 (8–83)	0.360
Extrathyroidal extension, yes/no (%)	31/12 (76)	13/9 (59)	0.401
AJCC stage, n (%)
I	0 (0)	1 (4)	0.743
II	2 (4)	0 (0)
III	2 (4)	1 (4)
IVa, IVb, IVc	46 (92)	23 (92)
Tumor markers at sampling
Calcitonin, pg/mL median (range)	1190 (105–66,017)	1123 (283–66,400)	0.581
CEA, ng/mL median (range)	23.3 (2–1274)	51 (4.1–2632)	0.310
Therapy at sampling, yes/no (%)	15/35 (30)	5/20 (20)	0.417

^aTumor size was unavailable for five positive and three negative patients.

Detection of circulating RET M918T cfDNA in patients with MTC

Patient plasma DNA was examined for the presence or absence of the RET M918T mutation by ddPCR. The assay of both RET mutant M918T and total RET cfDNA populations allows calculation of the AF of mutant cfDNA to total cfDNA present in circulation. This approach seeks to minimize quantification errors associated with measurement of plasma volume, efficiency of cfDNA isolation, and elution volumes. It also addresses DNA quantification errors associated with spectrophotometry or fluorescence because of the small size of cfDNA (25). A disadvantage of using AF is that the correlation with tumor burden is affected by conditions altering contribution of normal cfDNA to the circulation, such as chemotherapy (11). For this reason, we also calculated RET M918T cfDNA copy number relative to the input patient plasma volume. A comparison of RET M918T cfDNA AF and copies per milliliter of plasma is shown in Fig. 1. In the control patient group (RET wild-type by tumor tissue biopsy), a single sample tested positive for RET M918T cfDNA with an AF of 0.21% and estimated 1.36 mutant cfDNA copies per milliliter of plasma. Because we could not distinguish between acquisition of a new RET M918T mutation and a false positive, we used these values to establish thresholds. For both quantification methods the concordance of RET M918T mutation detection between liquid and tissue biopsy results was low: 32% (16/50) for AF and 40% (20/50) for copies per milliliter of plasma.

Figure 1.

Validation of RET M918T cfDNA detection by liquid biopsy. The study cohort included 75 patients with MTC with tumor tissue biopsy performed to determine the status of somatic RET M918T mutation. Somatic mutation was absent in tumor tissue of 25 patients (RET WT) and present in 50 patients (RET M918T). The detection of RET M918T in plasma cfDNA from these patients is displayed as either the proportion of mutated to total DNA (AF) or as number of mutant DNA copies per milliliter of plasma. WT, wild type.

Clinical correlations with AF of circulating RET M918T cfDNA

The concordance of mutation detection observed by liquid and tissue biopsy in MTC is lower than has been reported for other solid tumors, such as lung, breast, and colorectal cancers and melanoma (13, 26). Given that all patients enrolled in this study had biochemical evidence of tumor burden, we sought to identify clinical findings associated with the presence of circulating RET M918T cfDNA. The 50 patients who were RET M918T positive by tissue biopsy were divided into two groups based on an AF cutoff of 0.25%. Table 2 summarizes the baseline clinical correlates observed for these two groups. No significant differences were observed for baseline clinical parameters, including sex, age at diagnosis, tumor size, extrathyroidal extension, nodal metastasis, and length of time between diagnosis and plasma sampling. Patients with stage IVc distant metastatic disease at the time of diagnosis were more likely to have a positive RET M918T cfDNA assay at the time of sampling (P = 0.010). At the time of liquid biopsy sampling, significant associations were found with CEA level, number of metastatic sites, and whether the patient was being treated with vandetanib or cabozantinib.

Table 2.

Comparison of Demographic and Clinical Characteristics Between Patients With MTC With RET M918T cfDNA AF <0.25% and >0.25%

	AF <0.25%	AF >0.25%	P
n (male/female) (%)	34 (14/20) (41)	16 (10/6) (62)	0.227
Median age at diagnosis, y (range)	46.4 (16.5–67.3)	48.6 (21.6–70.3)	0.356
Median age at sampling, y (range)	52.7 (19.0–73.1)	54.5 (25.7–71.1)	0.981
Primary tumor characteristics
Median size, mm (range)a	20 (4–80)	31.5 (7–70)	0.997
Extrathyroidal extension, yes/no (%)	23/5 (82)	8/5 (62)	0.241
T status, n (%)
T1	7 (21)	3 (21)	0.590
T2	6 (18)	0 (0)
T3	15 (46)	9 (64)
T4	5 (15)	2 (14)
N status, n (%)
N0	2 (6)	2 (12.5)	0.263
N1a	2 (6)	2 (12.5)
N1b	30 (88)	12 (75)
M status, n (%)
M0 or Mx	28 (82)	6 (37.5)	0.003
M1	6 (18)	10 (62.5)
AJCC stage, n (%)
II	2 (6)	0 (0)	0.010
III	0 (0)	2 (12.5)
IVa	25 (73)	4 (25)
IVb	1 (3)	0 (0)
IVc	6 (18)	10 (62.5)
Tumor markers at cfDNA sampling
Calcitonin, pg/mL median (range)	786 (105–66017)	1604 (204–17707)	0.459
CEA, ng/mL median (range)	15.1 (2–283.5)	161.6 (2.1–1186)	0.023
Metastatic sites at cfDNA sampling, n (%)
Neck	14 (41)	11 (69)	0.128
Mediastinum	6 (18)	3 (19)	1.000
Lung	13 (38)	9 (56)	0.360
Liver	13 (38)	10 (63)	0.136
Bone	7 (21)	5 (31)	0.486
Other	1 (3)	1 (6)	0.542
Number of sites at cfDNA sampling, n (%)			0.033
Zero	7 (21)	1 (6)
One	11 (32)	2 (12)
Two	9 (26)	6 (38)
Three	4 (12)	4 (25)
Four	2 (6)	2 (12)
Five	1 (3)	1 (6)
Therapy at cfDNA sampling, yes/no (%)	7/27 (32)	8/8 (50)	0.049

^aTumor size was unavailable for three <0.25% and two >0.25% patients.

RET M918T cfDNA level and overall survival in patients with MTC

To evaluate the prognostic significance of RET M918T cfDNA level relative to established parameters, we performed a univariate Mantel–Cox regression survival analysis for the RET M918T–mutated MTC patient cohort. At the time of censoring, 10 of 50 patients were deceased, with an overall median follow-up of 15.7 months (1.2 to 29.7 months) from cfDNA sampling. Metastasis at diagnosis, therapy at sampling, calcitonin doubling time (<24 months), and RET M918T cfDNA AF (both the AF cutoff of 0.25% and a higher 1% cutoff) were found to be significant indicators of poor survival (Table 3). Unlike in previous reports, calcitonin and CEA doubling times <6 months were not found to be significantly prognostic for survival in our cohort (27–29). The AF cutoff measure of >0.25% best predicted outcome (hazard ratio 25.31, P < 0.0001) and as observed by Kaplan–Meier analysis (Fig. 2). Overall survival was compared for 48 patients with both calcitonin doubling time and RET M918T cfDNA AF data. For patients with calcitonin doubling time <24 months, survival was significantly worse compared with those with doubling times >24 months, although the survival curves eventually cross (Fig. 2). The trend is consistent with the established prognostic role of calcitonin doubling time. In this analysis, five of the 30 patients with calcitonin doubling time >24 months were found to have RET M918T cfDNA AF values >0.25%, three of whom died during our study. Among the 18 patients with calcitonin doubling time <24 months, eight had RET M918T cfDNA AF values <0.25%. This redistribution of 13 patients contributed to the more robust separation (P < 0.0001) observed for the Kaplan–Meier analysis based on an RET M918T cfDNA AF cutoff of 0.25% (Fig. 2).

Table 3.

Clinical and Biological Predictors of Survival in the RET M918T MTC Patient Cohort

	Survival Hazard Ratio (95% Confidence Interval)	P
Age at diagnosis, <45 y/>45 ya	0.703 (0.193–2.484)	0.591
Age at sampling, <45 y/>45 y	1.092 (0.266 –4.487)	0.903
Sex, male/female	1.477 (0.426–5.084)	0.542
M1 status at diagnosis, yes/no (n = 39)	7.607 (2.091–28.38)	0.002
Metastatic sites at sampling, >2/≤2	2.665 (0.826–13.36)	0.099
Therapy at sampling, yes/no	3.369 (1.036–15.19)	0.045
Calcitonin doubling time (n = 48)
<6 mo/>6 mo or <0	3.389 (0.993–50.36)	0.055
<24 mo/>24 mo or <0	3.328 (1.032–15.03)	0.046
CEA doubling time (n = 47)
<6 mo/>6 mo or <0	3.247 (0.914–43.50)	0.066
<24 mo/>24 mo or <0	2.498 (0.732–11.37)	0.132
RET M918T cfDNA AF
>1%/<1%	15.52 (9.514–205.1)	<0.0001
>0.25%/<0.25%	25.31 (6.920–115.9)	<0.0001

^aThe patient cohort size is 50 except where noted.

Figure 2.

Overall patient survival was examined for the same 48 patients by calcitonin doubling time <24 months (top plot) or with an RET M918T cfDNA AF cutoff of 0.25% (bottom plot). For both plots follow-up was from time of first cfDNA sampling, with calcitonin doubling time based on retrospective data. Two RET M918T–positive patients were excluded because of the lack of three calcitonin measurements in the previous 2 years. Patients A and B are described further Fig. 3.

Despite the significant prognostic value of RET M918T cfDNA single time point measurement, clear outliers were observed. The prognostic strength of calcitonin and CEA measurement is based on comparison of expression levels over time. Therefore, we examined the longitudinal changes in calcitonin and RET M918T cfDNA AF for two patient outliers present in Fig. 3. Patient A was diagnosed with stage IVa, T3N1bM0 disease 58.8 months before the first cfDNA sampling (Fig. 3A, time = 0). A second surgery for recurrence was performed ~5 months before sampling, followed by external beam radiation to the bilateral neck and superior mediastinum. At sampling the patient had stable disease with a calcitonin doubling time of <0 months and RET M918T cfDNA AF of 0. Progression was noted 15 months later, with increasing calcitonin and borderline detection of RET M918T cfDNA (AF = 0.25%). At the subsequent visit, a 2.4-fold increase in calcitonin and 8.0-fold increase in RET M918T AF level was observed. Both calcitonin and RET M918T AF levels continued to rise. Systemic treatment was never initiated because of poor Eastern Cooperative Oncology Group status, and the patient died within 10 months of the initial detection of RET M918T cfDNA. In contrast, patient B presented to our hospital 19 months before initial cfDNA sampling for evaluation of recurrence with liver metastasis. Subsequently, progressive disease was noted on imaging, with rising calcitonin levels. At initial sampling the patient had a 4.8-month calcitonin doubling time, with an RET M918T cfDNA AF of 12.8% (Fig. 3B). Vandetanib treatment was initiated 1 month later, with notable responses observed by imaging, calcitonin (13-fold reduction), and RET M918T cfDNA (16-fold reduction). Disease progression on vandetanib was observed on imaging at 20 months, resulting in discontinuation of vandetanib and initiation of cabozantinib. Progression of liver metastases continued on cabozantinib, with coincident increases in calcitonin and RET M918T cfDNA leading to discontinuation of the treatment and an acute rise in circulating tumor DNA.

Figure 3.

Longitudinal examination of calcitonin and RET M918T cfDNA levels. The biomarker levels and interventions for (A) patient A and (B) patient B from Fig. 2 are shown. Time shown is relative to the first cfDNA sampling, which was 0 for patient A, who died 25.3 months later.

Discussion

A growing literature has demonstrated the increasing clinical utility of liquid biopsy to examine circulating cfDNA for the presence of tumor-derived DNA. Liquid biopsy provides a noninvasive approach to identifying prominent driver mutations, and the monitoring of circulating tumor cfDNA level has potential as cancer biomarker (11, 16, 18, 30–33). Clear associations of circulating tumor cfDNA level with tumor burden, prognosis, and response to treatment are emerging for several cancer types. Additionally, repeat blood samples provide the opportunity to uncover the presence of new driver mutations when resistance develops as an aid for alternative treatment choices (34).

In this study we examined whether circulating tumor cfDNA containing the most common MTC driver mutation, RET M918T, could be detected by liquid biopsy of patient plasma samples. Because surgery is the primary treatment approach for this disorder, tissue is typically available for mutation analysis. However, in some cases the length of time from surgery to recurrence can lead to loss of sample or degradation of DNA in the original tumor sample. Thus our first goal was to determine the potential of liquid biopsy as a replacement for tissue biopsy. As expected, we were able to detect RET M918T circulating tumor cfDNA in patients with tissue biopsy proven mutations. However, with the AF tumor cfDNA cutoff of >0.25%, only 32% of patients were positive despite clinical evidence of disease (Fig. 1). This is much lower than the >75% detectability reported for most major cancer types (13, 26) and the 70% BRAF V600E detection rate observed for a cohort of 19 patients with ATC (20). Our findings may be a conservative estimate resulting from a combination of tumor dynamics and assay sensitivity. Samples tested in this study produced a median RET cfDNA copy number ∼1200 per assay, which limits sensitivity, as does the false positive detection of a single RET M918T droplet in a patient without a mutation present in tumor tissue. Although we cannot rule out acquisition of an RET M918T mutation in this patient sample, single-droplet false positives were also observed infrequently for control cell line DNA (data not shown). The RET M918T mutation detection rate increased to 40% when a cutoff of two mutant copies per milliliter of plasma was used. This result is similar to that observed by Bettegowda et al. (13) for non–small cell lung, prostate, renal cell, and pancreatic ductal carcinomas when the same cutoff is applied, and therefore low concordance between liquid and tissue biopsy is not unique to MTC. Thus, although it is clear that ddPCR liquid biopsy can be used for detection of somatic driver mutations in MTC, our data suggest that a negative result is not informative in the absence of tumor tissue data.

Similar to other cancer types, the tumor stage at diagnosis is highly prognostic for patients with MTC (35). The 5- and 10-year survival for patients with MTC with M1 disease at presentation has been reported to be ~66% and 39%, respectively, in a large single-center study (36). The availability of calcitonin and CEA as tumor biomarkers led to the establishment of measurement of their doubling time as being more prognostic than measures of tumor staging (27–29, 37). We recently reported circulating tumor cell number to be a better indicator of prognosis than calcitonin or CEA doubling time (38). Therefore, it is not surprising that detection of RET M918T cfDNA is also predictive of worse overall survival. The survival curve analysis demonstrated that detection of RET M918T cfDNA at a single time point accurately identified patients with poor prognosis despite favorable calcitonin doubling time (a calculation that requires at least four measurements over a 2-year period). Furthermore, for the two most prominent patient outliers, RET M918T cfDNA increased before patient A’s death, and initiation of systemic treatment was associated with reduced levels for patient B (Fig. 3). Therefore, in patients with an established driver mutation, such as RET M918T, measurement of circulating tumor cfDNA appears to be highly predictive of tumor progression and may have the potential to alert the clinician to start systemic therapy earlier. Additionally, where cfDNA is present, changes in level could guide decisions as to whether continued treatment is beneficial.

Our study is not without limitations. The rarity of MTC combined with the referral pattern to our institute and the need for known RET M918T tumor status resulted in a small study cohort with predominantly advanced disease. Furthermore, because of the prospective nature of this study, our median follow-up from cfDNA sampling was only 15.7 months, which may account for the limited number of deaths observed. We also recognize that our findings may be unique to patients with MTC with a somatic RET M918T mutation. Patients with the RET M918T mutation are reported to have worse prognosis (4, 5) and were found to be specifically enriched in the vandetanib (64%) and cabozantinib (55%) phase III trials, supporting a role for this mutation in more aggressive disease (6, 7).

Conclusions

To our knowledge, this is the largest study to evaluate the use of cfDNA as a liquid biopsy for somatic tumor mutation detection in MTC. Our findings indicate that a positive test can be used as a surrogate for tissue biopsy in patients with advanced MTC. However, a negative test does not mean absence of a mutation in the tumor, and follow-up tissue mutation testing should be performed when clinically relevant. In patients with RET M918T–positive tumor tissue, a negative liquid biopsy result, similar to stable calcitonin levels, appeared to be associated with stable or slowly progressing disease, whereas positive circulating levels correlated with aggressive disease. In this setting, the detection of circulating RET M918T cfDNA may be advantageous compared with other biomarkers given the predictive value of a single time point measurement. Particularly, RET M918T cfDNA could be used as a more precise biomarker of early response, or lack thereof, to therapy but should be studied further in prospective clinical trials. Finally, in the absence of tumor tissue data, liquid biopsy is a reasonable alternative for patients with large disease burden being considered for systemic therapy.

The positive detection of RET M918T cfDNA in patients with advanced MTC serves a role in prognosis, would allow targeting of therapy, and establish a baseline for temporal studies.