Background
In recent years the clinical application of novel immune checkpoint inhibitors (ICIs) has constituted a major breakthrough in the management of advanced melanoma, leading to unprecedented response and survival rates [
1,
2]. ICIs are monoclonal antibodies that promote tumoricidal effects by targeting regulatory pathways in T-cells. The two most effective classes of ICIs are directed towards the cytotoxic T-lymphocyte-associated protein 4 (CTLA-4; ipilimumab) or the programmed cell death protein 1 (PD-1; nivolumab, pembrolizumab) [
3,
4]. The usage of ICIs is considered nowadays the standard practice for the treatment of metastatic melanoma [
5]. Despite these improvements, however, a significant amount of patients—approximately 40–45%—show no response to immunotherapy [
6].
Immunotherapeutic agents act markedly different than usual cytotoxic approaches, notably by generating inflammations rather than direct lysis. This unique mechanism of action can lead to novel response patterns, which pose relevant challenges in the interpretation of treatment response by conventional imaging approaches [
7]. Pseudoprogression, defined as an initial increase in tumor burden followed by tumor regression, represents a distinct, atypical response pattern, initially described in melanoma patients undergoing ipilimumab therapy [
8,
9]. Pseudoprogression is considered the result of a transient immune cell infiltration of the tumor [
9,
10]. Another biologic explanation for the phenomenon could be a continued tumor cell growth until a sufficient response to immunotherapy takes place [
11]. Regardless of etiology, since pseudoprogression may be misclassified as progressive disease, its reliable and early identification would offer significant therapeutic implications in patient management.
In order to capture the atypical patterns of tumor response described with ICIs, several modified radiologic response criteria have been proposed, including the immune-related response criteria (irRC) [
9], the immune-related response evaluation criteria in solid tumors (irRECIST) [
12], the immune response evaluation criteria in solid tumors (iRECIST) [
13], and the immune-modified response evaluation criteria in solid tumors (imRECIST) [
14]. Despite their differences, all these criteria require or at least recommend a confirmation of progressive disease in follow-up CT or MRI scans [
15]. Similar attempts have been made with 2-deoxy-2-(
18F)fluoro-
D-glucose (
18F-FDG) positron emission tomography/computed tomography (PET/CT), leading to the proposal of the respective, metabolic criteria, namely the PET/CT Criteria for Early Prediction of Response to Immune Checkpoint Inhibitor Therapy (PECRIT) [
16], the PET Response Evaluation Criteria for Immunotherapy (PERCIMT) [
17], the immune PET Response Criteria in Solid Tumors (iPERCIST) [
18] and the immunotherapy-modified PET Response Criteria in Solid Tumors (imPERCIST) [
19]. However, the PET-based approaches have included smaller patient cohorts than the radiologic ones.
Considering the potential benefit of reliably identifying the non-negligible number of non-responders early during immunotherapy, we aimed to investigate the phenomenon of early metabolic disease progression in 18F-FDG PET/CT in metastatic melanoma patients under ICIs.
Discussion
Through the restoration of T cell function, ICIs activate the adaptive immune system to produce enhanced anti-tumor responses [
26]. The activation, differentiation and functions of T-cells are regulated by glucose metabolism. In particular, the “Warburg effect”, originally used to describe the shift of cancer cells from an aerobic mitochondrial oxidative metabolism to aerobic glycolysis to cover their demands, is also a key process for the sustainability of activated lymphocyte metabolism [
27‐
29]. At the same time, the Warburg effect constitutes the fundamental of molecular imaging with
18F-FDG PET/CT in oncology [
30]. In this context, the aforementioned similarity of cancer cell and activated lymphocyte metabolism—perfectly suited to match their functional needs [
31]—leads to manifestation of both cell types by means of
18F-FDG PET/CT, inevitably, raising the issue of specificity.
The reliable differentiation of true disease progression from pseudoprogression is clinically relevant, since it can identify non-responders who will have a shorter duration of benefit from the treatment [
32,
33]. However, the early and valid stratification of response to ICIs remains yet an insufficiently addressed issue [
26,
34,
35]. We have previously highlighted the contribution of PET/CT in the reliable differentiation of metabolic responders from non-responders early during immunotherapy, stressing, however, the emergence of signs of pseudoprogression, which should be taken into consideration during PET/CT interpretation [
20]. In the present study we tried to address this diagnostic challenge, focusing our investigation exclusively on melanoma patients exhibiting early signs of progression (uPMD) on
18F-FDG PET/CT under ICIs. Following the recommendations of the novel response criteria to immunotherapy [
13,
18], we studied these patients with at least one additional PET/CT after the demonstration of uPMD in order to confirm or exclude the diagnosis of disease progression. A strength of the study is that serial PET/CT monitoring included at least 3 strictly defined and clinically relevant time-points during the course of immunotherapy: (a) shortly before the start of treatment, (b) early during therapy (after the 2 first cycles), and (c) soon after administration of 4 cycles of ICIs.
There are three major findings from our analysis. Firstly, the majority (83.9%) of patients with signs of early metabolic progression on PET/CT—already after administration of two ICIs cycles—eventually show confirmed progressive disease. On the other hand, approximately every sixth patient with initial signs of metabolic progression has eventually pseudoprogression. Importantly, survival analysis revealed a longer OS for patients with pseudoprogression compared to those with cPMD; this difference in survival may not be statistically significant, which is very likely attributed to the small number of patients in the pseudoprogression group, but a clear trend was recognized in the respective Kaplan–Meier curves. Secondly, we demonstrate that the incidence of pseudoprogression can be markedly decreased after application of novel response criteria and identification of signs of irAEs. Thirdly, patients eventually responding to ICIs with cPMD exhibit a higher SLRmean after the first 2 cycles of treatment than those showing pseudoprogression.
The detection of non-responders already after administration of the first ICIs cycles carries significance, since it can limit the treatment-associated toxicity and financial burden in patients that are unlikely to profit from ICIs [
36]. In our cohort, 83.9% of patients with early uPMD (according to EORTC) had a confirmed metabolic progression when scanned at a later time point. These patients would potentially benefit from an early cessation of the non-effective, potentially toxic treatment and a change in therapeutic management at the appropriate time.
On the other hand, the identification of uPMD at an early time point comes at a cost: the misdiagnosis due to the phenomenon of pseudoprogression of some late responders as PMD would deprive these patients of the beneficial effect of immunotherapy. Indeed, a non-negligible number of patients (16.1%) eventually showed a subsequent remission of the initial uPMD. This finding suggests that pseudoprogression is not uncommon, which is in line with the results of a recent study by Pires da Silva et al. In that study, involving 140 melanoma patients treated with combined immunotherapy, one-third of all patients with progressive disease -according to RECIST 1.1—had eventually pseudoprogression and exhibited similar survival compared with non-progressors [
33].
These challenging results call for a more detailed analysis of the specific PET/CT findings of the 5 patients with pseudoprogression, in order to possibly identify imaging characteristics suggestive of this phenomenon and develop approaches that could address it. Specifically, in 2 patients of the pseudoprogression group—both of which showed a very good clinical response—misdiagnosis could be tackled after application of the recently proposed PERCIMT, instead of the EORTC criteria. The cornerstone of PERCIMT is the finding that the absolute number of newly emerged
18F-FDG-avid lesions is more predictive of clinical outcome than SUV changes during melanoma immunotherapy [
17]. In particular, neither a mere increase (> 25%) in tumor SUV nor the development of one new hypermetabolic lesion in follow-up PET/CT scan mean disease progression per se, as defined by the EORTC criteria. Instead, PERCIMT suggest the application of a threshold of four newly emerged lesions—with a decreasing cutoff of lesion number as the functional diameter of the lesions increases—for patient classification to progressive disease (Table
1) [
37]. The hitherto preliminary application of PERCIMT has shown promising results in patient stratification [
21,
38‐
41].
Besides PERCIMT, various novel metabolic response criteria have been recently proposed as alternative approaches to the conventional PET criteria—mainly the PET Response Criteria in Solid Tumors (PERCIST)—which were based on data derived from cohorts under cytotoxic therapies [
42]. These novel criteria seem to outperform PERCIST regarding their prognostic value in immunotherapy. In particular, the iPERCIST criteria use the “wait and see” approach, requiring a dual time-point evaluation for confirmation of the initial signs of PMD on PET/CT. iPERCIST introduce two new categories of response, derived from iRECIST: the unconfirmed progressive metabolic disease (UPMD), defined at 2 months after start of treatment (equivalent to 4 cycles of therapy), and the confirmed progressive metabolic disease (CPMD), requiring another evaluation 4 weeks after manifestation of UPMD [
18]. Another set of promising, novel criteria developed for immunotherapy evaluation are the imPERCIST, which follow the changes between and after the end of ipilimumab administration of the peak SUV of
18F-FDG in up to 5 measurable tumor/target lesions corrected for lean body mass (SULpeak), as suggested by PERCIST. However, imPERCIST have two major differences in comparison with PERCIST: firstly, PMD is not defined by the appearance of new lesions but exclusively by the increase of SULpeak. Secondly, the selection of target lesions at follow-up scan(s) is based on the 5 hottest lesions among all lesions in each scan—including newly emerging lesions -, irrespective of the distribution of lesions at baseline PET/CT imaging [
19]. Indicatively, the application of iPERCIST and imPERCIST in the herein presented group of 5 patients with pseudoprogression after the initial 2 ICIs cycles would have led to correct classification of one patient as SMD, thus, reducing the pseudoprogression cohort to 4 patients. Respectively, after the administration of 4 cycles of therapy, 2 more patients would show metabolic benefit to treatment (SMD) according to these criteria.
With further consideration to the pseudoprogression group, one patient developed a transient sarcoid-like lymphadenopathy of the mediastinal and hilar nodes. He eventually showed a very good clinical response. In clinical practice, the emergence of this response pattern during immunotherapy can create a diagnostic dilemma for imaging specialists and clinicians, since it may mimic disease progression. Based on the present findings and the steadily growing literature in the field [
43,
44], sarcoid-like lymphadenopathy should be not be interpreted as disease progression, but rather acknowledged as a possible irAE, which may be, moreover, associated with improved tumor response and disease control [
33,
45,
46]. Therefore, in such cases we recommend the performance of at least one follow-up PET/CT scan or histopathological examination of the respective imaging findings.
Finally, the remaining 2 cases were difficult to identify as potential pseudoprogression patterns. One patient showed a transient, generalized lymphadenopathy, which constitutes a rather sporadic response pattern under immunotherapy, previously described in some case reports [
47]. The second patient demonstrated a marked advance in both the number and metabolism of the baseline lesions, which subsided after 6 months, representing an example of the more delayed but durable responses, sometimes observed under ICIs [
48].
Aside from applying various response criteria, our analysis also involved the semi-quantitative assessment of immune activation signs on PET as a supplementary approach for prediction of the immunotherapy effect. This was attempted through the calculation of the ratio of
18F-FDG uptake (SUV) in the spleen and the liver (SLR). Our results showed that SLR
mean after the first 2 ICIs’ cycles was significantly higher for patients eventually showing cPMD compared to those with pseudoprogression, suggesting that increased spleen glucose metabolism under ICIs may correlate with negative clinical outcomes. This finding is consistent with recently published literature investigating spleen metabolism by means of
18F-FDG PET, which highlighted the significant association between a high SLR and a poor outcome to immunotherapy [
23,
49]. At the same time, these results should be interpreted with caution, since other studies in the field—although not explicitly involving SLR calculations—have shown a less contributory role of spleen metabolism in differentiating responders from non-responders to ICIs [
46,
50,
51], supporting the opinion that spleen metabolism is a “double-edged” biomarker with debatable results [
52]. On the whole, the investigation by means of functional imaging not only of the tumor but also of the patient’s host immune system gradually gains significance as potential surrogate marker of treatment response. Future prospective studies should determine whether this approach could indeed serve as prognosticator of immunotherapy.
Taken together, our findings show that the majority of non-responders can be identified early during immunotherapy, which may carry serious clinical implications. At the same time, approximately 16% of patients with uPMD eventually responds to treatment with remission of signs of pseudoprogression. These results support the use of early PET/CT scanning during immunotherapy, i.e. after application of two cycles of ICIs, but at the same time call for careful handling of the findings of early metabolic progression. In this context, the performance of another follow-up PET/CT at a second time-point for confirmation of early PMD is recommended. Given that the median time required for remission of pseudoprogression was 2.3 months in our cohort, and the fact that a long delay could risk disease decompensation to the point of rendering the patient incapable of receiving salvage chemotherapy, the late follow-up PET/CT examination may be performed approximately 8–10 weeks after initial, early progression. Besides that, we suggest the application of some practical tools for tackling the phenomenon of pseudoprogression, such as the usage of novel metabolic response criteria as well as the identification of sarcoid-like lymphadenopathy as a possible irAE rather than as a manifestation of disease progression. Finally, our results support the supplementary role of SLRmean as potential prognosticator in melanoma patients under immunotherapy.
Our study has some limitations. Firstly, this is a retrospective analysis of prospectively acquired data with a relatively limited number of patients due to the strict inclusion criteria applied. A validation of these findings in larger patient cohorts would be therefore required. Moreover, several patients were not further examined with PET/CT—in terms of the present analysis—after the first 4 cycles of ICIs, since they were either clinically characterized as progressors and subsequently changed/stopped therapy, or they soon thereafter died. Theoretically, a confirmation of the cPMD findings with another PET/CT scan later on (after the 4 cycles), may have revealed remission of signs of pseudoprogression in some of these patients. However, the clinical definition of progressive disease was always predominant, leading to respective, early management changes. Another limitation lies in the fact that most of the patients (84%) described in the present series were treated with ipilimumab monotherapy, which is no longer the standard of care in melanoma; instead, melanoma treatment nowadays involves anti-PD-1 monoclonal antibodies, applied both as single agents and in combination with ipilimumab. Further, the vast majority of the PET/CT-positive findings were not histopathologically confirmed, which is, however, not usually possible in the clinical setting. Finally, this study focused on two sets of criteria, one conventional (EORTC) and one novel (PERCIMT). The other recently proposed, immunotherapy-modified PET response criteria, such as iPERCIST and imPERCIST, were applied only in a subcohort of the studied patient population (pseudoprogression group). However, we are in the ongoing process of evaluating the performance of all available PET criteria—conventional and modified—in an extended patient cohort, including not only patients with PMD but all classes of response to ICIs; these evaluations will be the topic of a future work of our group.