Background
Melanoma is an aggressive form of skin cancer that is increasing in prevalence worldwide [
1]. Metastatic melanoma is a highly aggressive and difficult to treat cancer, particularly when patients present with advanced-stage disease that is unresectable [
2]. Recent advances in understanding the molecular mechanisms of melanoma oncogenesis and immune evasion have resulted in the introduction of BRAF and immune checkpoint inhibiting agents. These new treatments have improved the melanoma survival [
3‐
6], with the greatest treatment benefit observed when treatment is initiated at a lower disease burden [
2,
7‐
9].
In patients with advanced stage melanoma, treatment decisions are based upon clinical and imaging findings. In recent years, positron emission tomography with 2-deoxy-2[fluorine-18] fluoro-D-glucose integrated with computed tomography (FDG PET/CT) has emerged as a powerful imaging tool for initial staging and evaluating treatment response in metastatic melanoma [
10,
11].
18F–FDG is a radio labelled glucose analogue which reflects tumour metabolic activity. Commonly, FDG PET/CT is used to determine tumour burden as it provides a high tumour-to-background intensity ratio which facilitates computer generated measurements of total body metabolic tumour volume (MTV) and total lesion glycolysis (TLG) from which metabolic tumour burden (MTB) can be quantitatively calculated [
12,
13].
As a blood-based biomarker, circulating tumour DNA (ctDNA) offers a non-invasive and easily accessible method of providing a real-time “snap shot” of tumour burden. The level of ctDNA sensitivity however differs between tumour types, AJCC stages, mutant forms and between patients [
14]. In AJCC stage IV melanoma patients, ctDNA has been recognised as a valuable biomarker for tumour genetic profiling, monitoring disease progression, response to therapy and as a predictor of clinical outcome [
15‐
23]. Whilst, ctDNA has been detected in 73–89% of patients prior to therapy initiation [
15,
16,
19,
24] and the absence of ctDNA in these patients has been suggested as a prognostic marker for a better disease outcome [
19], the level of overall tumour burden at which ctDNA can be detected has not yet been quantified.
Numerous platforms are available for the detection of ctDNA, however the droplet digital PCR (ddPCR) platform has been shown to be the most sensitive, capable of detecting mutant DNA (such as
BRAF V600E) at 0.001% frequency abundance [
25,
26]. The lower limit in tumour size that shed detectable amounts of ctDNA into the blood is however unclear, and it may vary between cancer types. This information is critical for clinical validation of the efficacy ctDNA as a non-invasive complimentary method to functional imaging for monitoring of tumour burden in melanoma patients.
Here we determined the levels of ctDNA measured by ddPCR targeting tumour specific mutations in a cohort of 32 treatment naive stage IV melanoma patients. We evaluated whether the presence of ctDNA was associated with progression free survival and whether ctDNA levels correlated with MTB measured by FDG PET/CT.
Discussion
Here we demonstrated a significant correlation between the ctDNA levels in plasma and MTB measured by PET/CT in metastatic melanoma patients. Moreover, we observed a threshold for detecting ctDNA at an MTB score of ≤10. Finally, we confirm previous reports showing that undetectable ctDNA is associated with longer PFS.
With regards to patients with undetectable ctDNA, several cases are worth highlighting. In our study, one case (patient 4) with subcutaneous metastases only and noted to have low disease burden, was found negative for plasma ctDNA. Previously, subcutaneous metastases have been shown to be associated with low levels of ctDNA, despite extensive disease [
22]. Patients 16, 31 and 32 had nodal metastases only and ctDNA was not detectable. Whilst Wong et al. [
22] have shown that patients with nodal involvement often display high levels of ctDNA, we did not confirm this finding. It is worth noting that no other patients in our cohort had nodal disease alone, thus extrapolation of our results suggesting that ctDNA is undetectable in node only disease should be considered with caution. Moreover, this may be confounded by the fact that in two cases (patients 16 and 32) the MTB was below our threshold of 10. The third patient (patient 31) had an MTB of 33.49 but the lack of ctDNA could be attributed to the fact that cfDNA was extracted from only 1 ml of plasma.
Undetectable ctDNA in patients 9, 10 and 30 may be explained by the sites of their metastatic disease, nodal and brain (patient 9) and brain only (patient 10 and 30). Previously, low or undetectable ctDNA levels have been observed in patients with brain metastases [
22,
29,
30].
One significant outlier is patient 29, where despite having bone metastases and a significant MTB, we were unable to detect ctDNA. Bone metastases have previously been associated with high levels of ctDNA [
22]. In this case, cfDNA was also extracted from only 1 mL of plasma which is likely to have a significant impact on detection levels of ctDNA [
31].
Remarkably patient 28 (with nodal and liver metastases) had undetectable ctDNA however, MTB was ≤10. No obvious limitations were evident in this case, particularly with regards to the volume of plasma collected. Thus, we are confident that with our current assay sensitivities, we are not able to detect ctDNA in patients with a MTB value of ≤10, regardless of disease site or mutation. Nevertheless, this observation needs to be corroborated in larger studies, taking in consideration diverse mutated genes and the site of metastases.
To our knowledge, this is the first study in melanoma that has directly compared the level of ctDNA with the exact MTB calculated from the sum of TLG for all evaluable lesions. Recently, Wong et al. [
22] showed that ctDNA levels correlate with qualitative analysis of whole body MTV in metastatic melanoma patients. Similar to our findings, a strong correlation was observed between ctDNA levels and MTV (
r = 0.61;
P < 0.001). Whilst it is difficult to make a direct comparison between this study and our own due to the different methodologies employed to detect ctDNA (digital PCR and targeted sequencing) and different reporting mechanisms of MTV (Wong reported results in mL, whilst we reported results in cm
3), it is interesting to note there is a remarkable difference in the median levels of plasma ctDNA reported by both studies. Wong et al. reported a median ctDNA concentration of 1112 copies/mL of plasma (range 63–97,000) which is considerably higher than the copies/mL of plasma that we reported, indicating an enrichment of patients with lower disease burden in our study. Finally, our measurements incorporated PET parameter TLG to define all measureable lesions combined with the metabolic activity of each tumour, thus measuring both tumour volume and aggressiveness. This approach provides a comprehensive score of not only of tumour burden but also of tumour activity and therefore, an overall perspective of the disease status of patient.
In line with our study, the presence of ctDNA has been directly correlated with patient survival and MTB in advanced stage non-small cell lung cancer (NSCLC) [
12]. This study assessed allele frequency measured by NGS in 24 NSCLC patients and MTB (calculated from the sum of TLG for all evaluable lesions) and found a significant correlation (
P = 0.001). The authors also reported a significantly shorter median overall survival (OS) in patients with detectable ctDNA compared to those with no detectable ctDNA. In our study, we have specifically chosen not to assess OS as this endpoint includes death from any cause, and thus is influenced by co-morbidities and access to systemic therapies over different periods of time. For this reason, our analysis focused on PFS, which was significantly shortened in patients with detectable ctDNA. Finally, multivariate Cox regression analyses demonstrated that ctDNA is not an independent variable but rather a reflection of other disease burden measurements such as MTB and disease stage.
There are a number of noteworthy limitations to our study, largely associated with its retrospective nature. Firstly, the timing of blood collections in accordance with PET/CT imaging was varied. In the Winther-Larsen and colleagues study [
12], blood samples were collected at the time of inclusion into the study, and PET/CT imaging. This resulted in a median interval of 2 days between imaging and blood draw, which is considerably shorter than our median of 1.9 weeks. However, for both studies the timing between blood collection and imaging may be an important factor to consider. Given that ctDNA has a half-life of less than 2 h and have been shown to increase as new lesions become apparent [
32], ideally blood draw should be conducted immediately after imaging to ensure that ctDNA detected is a true reflection of lesions identified in the image. Importantly, in all our cases, plasma was collected after imaging and prior to treatment. Secondly, PET/CT imaging was conducted at different institutions, which may have resulted in inter-institutional differences in quality control and scanning [
33]; the use of multiple scanner models has been associated with variability in the standard uptake of FDG readings [
34]. We did not however observe any substantial deviation in the correlation between ctDNA and MTB due to scanner models (see Additional file
4). Finally, we acknowledge that this cohort is heavily biased for
BRAF V600 mutated cases, and future studies should address this across multiple mutations.