1. Introduction
Over the last few years, the development and introduction in the clinic of immune checkpoint inhibitors (ICIs) have radically changed the landscape in the management of metastatic melanoma. In particular, systemic treatment with the monoclonal antibodies ipilimumab, which targets the T-lymphocyte-associated protein 4 (CTLA-4), as well as nivolumab and pembrolizumab, which target the programmed cell death protein 1 (PD-1), has led to unprecedented response and survival rates in advanced melanoma and is nowadays considered the first-line therapy for unresectable stage III and IV melanoma patients [
1,
2,
3,
4,
5,
6,
7].
At the same time, the widespread usage of these powerful therapeutic tools has raised some challenges with regard to the reliable assessment of response to them. Being a completely novel treatment, immunotherapy with ICIs is associated with responses, which can be different from conventional cytotoxic approaches, notably by generating inflammations rather than direct lysis [
8]. A distinct feature of immunotherapy, compared to chemotherapy, is the more delayed response to agent administration with durable responses that may persist even after treatment cessation [
9]. Other novel patterns associated with ICIs include the phenomenon of pseudoprogression, defined as an initial increase in tumor burden followed by tumor regression, the hyperprogressive disease, which is an aggressive pattern of rapid, marked disease progression, as well as dissociated responses, characterized by the regression of some lesions and the concurrent growth of other lesions or the appearance of new ones [
10,
11,
12].
To make things more complicated, immunotherapeutic agents are also linked with the development of a “new class” of side effects that resemble autoimmune responses, the immune-related adverse events (irEs), whose spectrum is wide, affecting almost every organ of the body and can occur at any point in the treatment course [
13]. These adverse events can also pose relevant interpretation challenges to standard imaging methods.
Positron emission tomography/computed tomography (PET/CT) represents a hybrid imaging modality, which combines the molecular-metabolic detail of PET—mostly utilizing the glucose analog
18F-fluorodeoxyglucose (
18F-FDG) as a radiotracer, with the anatomic precision of CT.
18F-FDG PET/CT is the elective imaging technique in detecting metastatic disease in advanced melanoma [
14,
15,
16,
17]. It is, moreover, considered a relatively reliable tool for the monitoring of immunotherapy, although certain limitations are encountered with concern to response assessment [
18,
19,
20,
21,
22,
23].
In clinical routine, the evaluation of response to immunotherapy by means of PET/CT is primarily visual and subjective in nature, with quantitative—thus more objective—assessments being mainly restricted in the calculation of the semiquantitative parameter, standardized uptake value (SUV). SUV represents tissue activity within a volume of interest (VOI) corrected for injected activity and body weight, and is calculated when the tracer has reached equilibrium, usually at 60 min post-injection (p.i.). However, the generally accepted method for accurate analysis of
18F-FDG metabolism and pharmacokinetics is a two-tissue compartment model [
24]. A prerequisite for this is the performance of full dynamic PET (dPET) studies for at least 60 min.
Data on the application of dynamic PET/CT in immunotherapy monitoring are limited [
25]. The aim of the present prospective study is to investigate the predictive role of quantitative, dynamic
18F-FDG PET/CT performed early during immunotherapy in metastatic melanoma patients undergoing treatment with PD-1 inhibitors.
4. Discussion
The vast majority of oncological treatment response evaluations are based on changes in tumor size [
34,
35], which represent, however, the last step in a series of metabolic and functional processes and are, thus, considered late-occurring events. Moreover, due to their unique mechanism of action, ICIs can generate novel, previously unreported response patterns, whose reliable interpretation poses challenges in conventional imaging modalities. In this framework, the role of
18F-FDG PET/CT is increasingly investigated since it can detect therapy effects on a metabolic level before morphological changes take place [
36]. Tumor imaging by means of
18F-FDG PET/CT is based on the metabolic programming of cancer cells, in particular on their preference to metabolize glucose by aerobic glycolysis, rather than oxidative phosphorylation, for ATP generation so-called Warburg effect [
37]. Apart from its well-documented affinity for tumor tissue,
18F-FDG also accumulates at sites of infiltration by cytotoxic T-cells, the main immune cells involved in the antitumor mechanism of ICIs, allowing their visualization by means of PET/CT [
23,
38].
These features render imaging with
18F-FDG PET/CT attractive for immunotherapy monitoring, but on the other hand, they may partly lead to interpretation pitfalls, including false-positive findings, such as the phenomenon of pseudoprogression [
9,
39] or the emergence of irEs at non-tumor tissue. One strategy to tackle this limitation is the application of more specific imaging biomarkers than
18F-FDG, currently under investigation [
40,
41,
42,
43]. In the absence, however, of such tracers widely available for clinical use, objective quantification of tracer uptake is gaining significance.
SUV is a widely used semiquantitative PET metric because of its simplicity and reproducibility [
44]. These attributes have led to its broad investigation as a potential biomarker of immunotherapy response assessment [
19,
20,
22,
45,
46]. Another quantification approach involves the employment of dynamic PET/CT protocols, which offer the possibility of more detailed calculations of tracer kinetics compared to standard imaging. In the case of
18F-FDG, the performance of early, dynamic PET/CT studies and the subsequent application of compartment and fractal analysis allow the investigation of dedicated parameters of the tracer’s metabolism [
47]. However, data on dynamic, quantitative
18F-FDG PET/CT in ICIs’ monitoring remain scarce [
25]. This is mainly attributed to the difficulty of implementation of dynamic PET/CT in routine clinical practice: first, it is more time-consuming than conventional, static PET/CT scanning since the acquisition of a full dynamic PET scan is a 60-min process. Moreover, sophisticated software tools are required for the performance of the complicated compartment and fractal analysis in order to evaluate the dynamic data. Despite these obstacles, in the present study, both approaches, the semiquantitative and the quantitative, were utilized in an attempt to investigate the predictive role of “full” quantitative
18F-FDG PET/CT analysis on the survival of metastatic melanoma patients treated with PD-1 blockade.
Semiquantitative analysis of
18F-FDG uptake revealed that median SUV values of melanoma lesions decreased by 4% to 30% as a response to anti-PD-1 treatment. Moreover, SUV values derived from interim PET/CT adversely affected PFS. These findings are in line with the results by Ito et al. in 60 melanoma patients treated with ipilimumab and examined with PET/CT before and after treatment. The authors of that study observed a significant association between changes in tumor SUV and overall survival (OS) [
22]. Moreover, Nobashi et al. showed in a heterogeneous cohort of 40 patients undergoing immunotherapy that clinical responders demonstrated a significant decrease of SUV
max between baseline and restaging PET/CT [
47]. On the other hand, our group had previously failed to show a significant correlation between SUV changes and clinical benefit to ipilimumab [
25,
45]; instead, the number of a newly emerging,
18F-FDG-avid lesions could more reliably serve patient classification to progressive disease and, thus, predict response to immunotherapy (sensitivity = 84%, specificity = 100%). Based on this, a new set of response criteria was suggested, the PET response evaluation criteria for immunotherapy (PERCIMT) [
45]. The PERCIMT-based dichotomization of the herein studied cohort into metabolic responders and nonresponders showed that the first ones had a significantly longer PFS than the latter [
26]. Interestingly, Cho et al. found in a group of 20 melanoma patients treated with different ICIs that an increase in
18F-FDG tumor uptake in PET/CT performed at 3–4 weeks into therapy may correlate with clinical benefit, reaching a 100% sensitivity, a 93% specificity, and a 95% accuracy. This observation led to the introduction of the PET/CT Criteria for early prediction of Response to Immune checkpoint inhibitor Therapy (PECRIT), which combines anatomic and functional imaging data collected at follow-up PET/CT [
20]. Based on these mixed—and partly contradictory—results, we can assume that SUV changes in tumor lesions may have a predictive value in immunotherapy outcome, but SUV is far from being considered a perfect prognostic indicator. SUV changes may be helpful in assessing partial or complete response but do not enable a reliable distinction between borderline cases of stable and progressive disease. Furthermore, several issues still remain to be clarified, mostly deriving from the nonspecific nature of
18F-FDG, which, as mentioned above, can accumulate in both tumor lesions and ICIs’-induced sites of inflammation.
Quantitative analysis of the dynamic PET/CT data of melanoma lesions after application of compartment modeling revealed a decreasing, not statistically significant, pattern of all
18F-FDG kinetic parameters regarding their interval changes as a response to therapy. Nevertheless, survival analysis did not reveal any effect of these parameters on PFS. This finding is in accordance with previously published results by our group on 25 metastatic melanoma patients undergoing ipilimumab monotherapy. In that study, due to the lack of survival data at the time point of evaluation, the patients’ best clinical response—based on a combination of clinical, biochemical and imaging data—was used as a reference. Similar to the present study, no significant differences between responders and nonresponders to ipilimumab were observed [
25].
On the other hand, a quantitative tumor parameter found to negatively influence survival was the degree of
18F-FDG heterogeneity, reflected by the index FD. FD is calculated with the box-counting method and provides an estimate of the complexity of a dimensional structure [
33]. In the present analysis, higher FD values and subsequently a more heterogeneous distribution of the glucose analog in melanoma lesions early during anti-PD-1 treatment had an adverse effect on PFS. The potential of fractal mathematics in oncological research and practice remains relatively unknown to most clinicians. Nevertheless, the hitherto rather limited but steadily growing literature in the field implies a possible contribution of the investigation of tumor heterogeneity, utilizing fractal principles on PET images, in tumor characterization and patient prognosis [
48,
49,
50,
51].
Aside from melanoma lesions, we also assessed the thyroid gland, whose dysfunction is a frequent endocrine irAE induced by PD-1 inhibitors [
52,
53], as well as primary and secondary lymphoid organs, given the critical impact of systemic immune responses for effective cancer immunotherapy [
54].
The quantitative analysis of the thyroid showed a non-negligible SUV increase in the gland as a response to treatment, which is in line with previously published results [
46]. Interestingly, patients with higher FD on interim PET/CT showed significantly longer PFS, suggesting a positive survival effect of the heterogeneity of
18F-FDG distribution in the thyroid during PD-1 blockade. Although the pathophysiological basis for these findings cannot be easily clarified, we note the documented association between the development of thyroid dysfunction and an improvement in survival of patients treated by PD-1 blockade [
55]. Importantly, in a recently published meta-analysis, a significant association between the development of endocrine irAEs, among which thyroid dysfunction, and a favorable benefit on survival was highlighted [
56].
The bone marrow represents the main hematopoietic organ in adults and a supportive organ to immune cell function [
57]. In this context, the assessment of bone marrow glucose metabolism during cancer immunotherapy is of interest. Nobashi et al. reported on a decrease of bone marrow SUV
max after immunotherapy in patients with different tumors, in both clinical responders and nonresponders to treatment [
46]. On the other hand, Schwenk et al. found a significant increase of glucose metabolism in the bone marrow of responders, while a decreased
18F-FDG uptake was observed in nonresponders [
58]. In our analysis, the parameters SUV
max and k
3 derived from interim PET/CT positively affected the patient suggesting that the degree of activation of the bone marrow during PD-1 blockade, as reflected by respective changes in
18F-FDG uptake and kinetics, may play a predictive role in treatment outcome of metastatic melanoma. Of course, this needs to be validated by further studies, including larger cohorts.
Finally, we evaluated the glucose metabolism in the spleen, the largest secondary lymphoid organ and blood-filtering unit in the body [
59,
60,
61]. The spleen is involved in response to systemic inflammatory stimuli, which, in terms of PET imaging, can be observable by increased
18F-FDG uptake in the organ [
58,
62,
63,
64]. In the present analysis, only minor SUV changes were observed as a response to treatment. Moreover, no parameter both on the baseline and interim PET/CT had any effect on patient survival. These findings are supportive of previously published results of our group and others, suggesting a rather poor contribution of spleen metabolism, studied by
18F-FDG PET/CT, in monitoring and prediction of outcome in melanoma patients under immunotherapy [
46,
58,
65].
Taken together, there are two major findings from our study. First, the SUV values, as well as FD of metastatic melanoma lesions, seem to have an adverse effect on PFS already after the application of two cycles of PD-1 blockade. Second, we show for the first time that quantitative parameters derived from specific reference tissues, namely the thyroid and the bone marrow, can also be predictive of PFS, but in this case with a positive effect. Although these findings could suggest the wider usage of dynamic PET/CT in oncological clinical practice, we are aware of the practical considerations that accompany this approach, rendering its routine application burdensome. In view of this, we can recommend for the time being the use of dynamic PET/CT—performed in combination with conventional, static, whole-body PET/CT—for selected oncological cases, most likely in terms of investigating treatment response to specific therapeutic regimens. By complementing the information offered by conventional imaging with the multiparametric, pharmacokinetic data extracted by dynamic PET/CT, the diagnostic certainty of the reading physician could be enhanced, while our understanding of the pathophysiology involved in the natural history of the tumor and its response to treatment would be improved. Moreover, the future perspective of quantitative, dynamic PET/CT seems prosperous: the recent advent of new PET/CT scanners, which allow dynamic studies over several bed positions by using a continuous bed movement, as well as the introduction of new PET/CT scanners with an extended field of view (>1 m) will facilitate the use of dynamic PET protocols and reduce the whole acquisition time, making dynamic PET/CT an attractive and cost-effective approach in oncological imaging [
66].
Our study has some limitations. Foremost, the relatively small number of studied patients does not allow the drawing of more firm conclusions. This is, however, mainly attributed to the strict inclusion criteria applied in the study. The lack of histological validation of the vast majority of the PET/CT positive findings constitutes another limitation of our analysis. However, this is usually not possible in the clinical setting. Moreover, despite the fact that a two-bed position protocol, which allows the study of a relatively large field of view of 44 cm, was used, the dynamic PET/CT acquisition was confined only in the anatomic area of the thorax/upper abdomen, not allowing for whole-body dynamic studies to be performed.