Immunotherapy Monitoring with Immune Checkpoint Inhibitors Based on [18F]FDG PET/CT in Metastatic Melanomas and Lung Cancer

Immunotherapy with checkpoint inhibitors has prompted a major change not only in cancer treatment but also in medical imaging. In parallel with the implementation of new drugs modulating the immune system, new response criteria have been developed, aiming to overcome clinical drawbacks related to the new, unusual, patterns of response characterizing both solid tumors and lymphoma during the course of immunotherapy. The acknowledgement of pseudo-progression, hyper-progression, immune-dissociated response and so forth, has become mandatory for all imagers dealing with this clinical scenario. A long list of acronyms, i.e., irRC, iRECIST, irRECIST, imRECIST, PECRIT, PERCIMT, imPERCIST, iPERCIST, depicts the enormous effort made by radiology and nuclear medicine physicians in the last decade to optimize imaging parameters for better prediction of clinical benefit in immunotherapy regimens. Quite frequently, a combination of clinical-laboratory data with imaging findings has been tested, proving the ability to stratify patients into various risk groups. The next steps necessarily require a large scale validation of the most robust criteria, as well as the clinical implementation of immune-targeting tracers for immuno-PET or the exploitation of radiomics and artificial intelligence as complementary tools during the course of immunotherapy administration. For the present review article, a summary of PET/CT role for immunotherapy monitoring will be provided. By scrolling into various cancer types and applied response criteria, the reader will obtain necessary information for better understanding the potentials and limitations of the modality in the clinical setting.


Introduction
Starting with the first outstanding results of the anti-cytotoxic T lymphocyte antigen-4 (CTLA-4) antibody, Ipilimumab, in melanoma [1] and following use of antibodies against programmed cell death protein 1 (PD-1) and its ligand PD-L1 (nivolumab, pembrolizumab, atezolizumab) in non-small cell lung cancer (NSCLC) [2][3][4][5], immunotherapy with checkpoint inhibitors has gradually changed the management of malignant tumors by improving the long term benefit and survival. Clinicians have become acquainted along the way with new ways of considering clinical benefit, meaning to recognize objective progression not necessarily as an upfront sign of treatment failure. From an imaging point of view, new semantic artifices have been implemented to help handle the variegated patterns of response that accompany treatment with immune checkpoint inhibitors (ICI). It is therefore not surprising that the number of response criteria has consequently increased, both for morphological and metabolic imaging (Table 1). For the present review article, a summary of the role of PET/CT for immunotherapy monitoring will be provided. By scrolling into various cancer types and applied response criteria, the reader will obtain necessary information for better understanding the potentials and limitations of the modality in the clinical setting.  [11] complete resolution of [ 18 F]FDG uptake reduction of a minimum of 15% ± 25% in tumor SUV after 1 cycle of chemotherapy, and >25% after more than one treatment cycle increase in SUV of less than 25% or a decrease of less than 15% increase in tumor FDG uptake > 25%, increase of the maximum tumor > 20%, new metastases as progressive disease PERCIST (2009) [12] disappearance of all metabolically active lesions SULpeak reduction ≥ 30% in the hottest target lesions neither PMD nor PMR/CMR SULpeak increase ≥ 30% in the hottest target lesion as progressive disease PERCIMT (2018) [13] disappearance of all metabolically active lesions

New Concepts in Tumor Response during Immunotherapy
Born to overcome the limitations of conventional criteria, and driven by the need to avoid unnecessary treatment withdrawal, immunotherapy-derived response criteria have embraced concepts such as pseudo-progression, hyper-progression or dissociated progression to move beyond the immunotherapy era. Although previously described as an unconventional response pattern in gliomas treated with chemoradiotherapy [17], pseudoprogression is now more broadly associated with ICI and corresponds to the appearance of new lesions or the occurrence of tumor enlargement during therapy, followed by disease regression or stabilization at subsequent imaging [18]. The phenomenon is more frequent during anti-CTLA-4 therapy and tends to affect fewer cancer patients treated with anti-PD-1/L1 agents. Nevertheless, the rate of pseudo-progression in general does not exceed 10% [19,20].
Hyper-progression, on the other hand, refers to a very peculiar pattern of response to ICI, and was firstly described in 2016 by Champiat et al. [21]. Its occurrence ranges from 4% to 29%, proving a large variability of cases according to the casuistics [20,22]. Substantially, hyperprogressive disease (HPD) corresponds to a massive increase of tumor burden, over twice the amount compared to (prior to) treatment start. Notwithstanding, controversies exist on the exact way HPD is defined in clinical practice. While Champiat et al. [21] defined HPD as a twofold or greater increase of tumor growth rate (TGR) during immunotherapy [20], other authors used different descriptions. For instance, Kato et al. defined HPD as a time to treatment failure (TTF) < 2 months, a 50% increase in tumor burden compared to pre-immunotherapy imaging obtained within 2 months of the treatment initiation, and > 2-fold increase in progression pace [20,23]. In other cases, like for Saâda-Bouzid et al., HPD could be computed based on tumor growth kinetic ratio (TGKR), where TGK is defined as the difference of the sum of the largest diameters of target lesions per unit of time, which in the case of HPD has to be ≥ 2 when compared to baseline [20,24,25]. More simply, Matos et al. [26] used as parameter for HPD a 40% increase of the sum of the target lesions from baseline to the first evaluation and/or an increase of 20% plus the appearance of new lesions in two different organs [27]. Although comparison only to baseline imaging, without utilization of data before treatment start, has made some authors define as "fast progression" rather than "hyperprogression" the cases reported by later authors [24][25][26], strictly speaking the occurrence of this "non-response", is in any of the cases, a dramatic failure. In fact, patients with this type of progression during ICI (call it "hyper-" or "fast") have a worse outcome with a significantly shorter survival rate [20][21][22][23][24][25][26]28].
To add further confusion to the already intricate situation, recently a new pattern of tumor behavior during ICI has been described in advanced lung cancer [29,30]; this consists of a "dissociated response", i.e., a contemporary shrinkage of some tumor lesions along with the increase of others in various organs [18], which occurs in around 10% of patients [31]. Given the potential benefit still obtainable for patients showing an immune dissociated response (iDR), some authors [30] have suggested iDR as a surrogate marker of favorable outcome and treatment efficiency [31].
Along with the abovementioned new patterns of response, immunotherapy with ICI can determine several immunologically mediated alterations of healthy tissues and organs, also known as immune-related adverse events (irAEs) [18]. The incidence of these events is higher for anti-CTLA-4 antibodies (80%) and during combination therapy, while it reaches in general 27% for anti-PD-1 and 17% for anti-PD-L1 regimens [18,32]. The occurrence of irAEs, based on the severity of the event, might require immediate ICI discontinuation [33,34]. This will not necessarily prevent fatality, which is surprisingly related to colitis in 70% of the cases treated with anti-CTLA-4, followed by pneumonitis (35%), hepatitis (22%) and neurotoxicities (15%) for anti-PD-1/anti-PD-L1 antibodies [33,34]. From an imaging point of view, irAE interpretation can sometimes be as challenging as other unconventional patterns of response described during ICI. Given the potentially fatal events related to their occurrence, it is fundamental to be aware of their appearance and describe them promptly in the report and to the clinician treating the patient (Figure 1) [35]. Notwithstanding, there is also a positive aspect with irAEs, which is their potential predictive role for treatment benefit. Indeed, being an expression of immune system response, although abnormal and undesirable in most cases, irAEs represent a precognitive sign of longer progression-free (PFS) and overall survival (OS) [36]. From first reports to later meta-analyses, irAE development seems to be positively associated with overall response rate (ORR), PFS, and OS in patients treated with immunotherapy, regardless of lesion site, type of ICI and irAE [36,37], although, grade 3 or higher toxicities have resulted prognostically in worse OS [37].

Response Assessment in Solid Tumors Treated with Checkpoint Inhibitors
Keeping in mind the abovementioned peculiarities of imagine interpretation during ICI, imagers require adequate instruments to assess immunotherapy benefit, which from a metabolic point of view consists mainly in the use [ 18 F]FDG PET/CT for response assessment (Figures 2 and 3). As previously anticipated, quite an extensive number of response criteria have been proposed for this purpose in recent years (Table 1). During initial studies, consolidated criteria, such as EORTC (European Organization for Research and Treatment of Cancer) [11] and PERCIST (PET Response Criteria in Solid Tumors) [12], have represented the simplest way to assess tumor response, followed later by subsequent adaptations to ICI. This is the case in the instance of PECRIT criteria (PET/CT Criteria for early prediction of Response to Immune checkpoint inhibitor Therapy), introduced by Cho et al. [16], which combine both morphologic (contemplating a change in the sum of diameters of target lesions according to RECIST 1.1) and metabolic response (i.e., a reduction in the SULpeak > 15.5% for the hottest lesion on PET) to assess clinical benefit of ICI. Other authors have introduced PERCIMT (PET Response Evaluation Criteria for IMmunoTherapy), firstly described in melanoma patients [13]. Herein, the appearance of up to four new lesions, depending on their size (Table 1), can be tolerated to obtain clinical benefit (CB) and support treatment continuation [13,38]. More recently, other alternative approaches to PERCIST have been used, including iPERCIST [15] and immunotherapy-modified PER-CIST5 (imPERCIST) [14]. For the latter, the definition of a progressive metabolic disease (PMD) becomes less stringent, requiring in fact an increase in the sum of SULpeaks of 30%, with new lesions being eventually included in the sum of SULpeak [14,18]. The principle behind all these new adaptations is substantially the same: to avoid unnecessary and premature treatment withdrawal during immunotherapy. but can we depict one of them as the best response criteria for response assessment during ICI? Actually, not. Some reports have attempted to compare various methods, particularly in melanoma and NSCLC patients [14,[38][39][40][41][42][43], proving the superiority of some of the utilized criteria over others (Table 2). Ultimately, all available response criteria, metabolic or morphological, retain the capability to predict response and outcome. What makes one criteria better than the other is most likely to be the interpretation ability of the imager and the correct contextualization of the results into clinical practice. This should not limit, in any case, the continuous research in the field, since robust data must be produced to optimize response criteria for response assessment during ICI, not forgetting the absolute necessity to ascertain the perfect timing for treatment discontinuation for patients to receive long-term clinical benefit.        Notes: PubMed database was searched from 2010 until September 2021 for the terms: ("fluorodeoxyglucose f18" OR ("fluorodeoxyglucose" AND "f18") OR "fluorodeoxyglucose f18" OR ("18f"AND "fdg") OR "18f fdg") AND "pet" AND ("immunotherapy" OR "immunotherapies" OR "immunotherapy s") AND ("cancer s" OR "cancerated" OR "canceration" OR "cancerization" OR "cancerized" OR "cancerous" OR "neoplasms"OR "cancer" OR "cancers").
Of special interest also is the risk stratification of patients based on volumetric parameters already obtained at baseline, with patients having a higher MTV and TLG being at higher risk of poor outcome or HPD compared to others [53,55,71,73,74,76,80]. In this context, to further improve the predictive role, a combination of metabolic tumor burden (MTV and TLG) with other clinical parameters has been performed. In particular, circulating inflammatory markers, such as neutrophyl-to-lymphocyte ratio (NLR) and its derived value (dNLR) have proved to better stratify patients undergoing immunotherapy with ICI into risk groups (i.e., higher values predicting poor outcome), both at baseline and after treatment start [71,72,76,80]. Similarly, the combination of volumetric parameters on PET with circulating tumor cells (CTC) count and soluble PD-L1 [72,75,83], or lactate dehydrogenase (LDH) [88] has been reported to be as useful for risk stratification. Thanks to the capability of [ 18 F]FDG PET/CT to depict underlying immunological status, expressed as bone marrow or lymphatic organ activation (i.e., bone marrow-to-liver ratio, spleen-toliver ratio) or by the development of irAEs, it is also possible to combine metabolic and immunological parameters to improve response prediction and outcome [48][49][50][51][52][53]56,57,81].
The downside of the previously mentioned findings, despite being fascinating and promising, is that most of the original data derive from retrospective analyses or from limited, single centered, prospective cohorts (Table 2). Consequently, their clinical relevance remains circumscribed to theory, until large prospective multicentric imaging trials are properly conducted.

Next Generation Imaging for Immunotherapy in Cancer
Radiomics and artificial intelligence (AI) have become a constant mantra in applied sciences, and this includes, necessarily, medical imaging. Automated machine or deep learning algorithms also represent the next frontier of imaging for immunotherapy in cancer, since they might be able to extract precious information, invisible to the naked eye or to conventional measurements. We have known for some years that image heterogeneity is a marker of underlying histological and genetic complexity; but which features could be better associated with specific tumor aspects still requires thorough investigation. What emerges from initial reports published so far on radiomics and AI in the context of immunotherapy setting is that no unique parameter or feature can be defined as superior (Table 2). While features like "skewness" and "kurtosis", well known from other types of treatment, might represent a marker of treatment failure during ICI in lung cancer [90], for other authors either Small Run Emphasis (SRE), multiparametric radiomics signature (mpRS), cytolytic activity score (CytAct), deeply learned score (DLS), or long zone emphasis (LZE) [89,[91][92][93][94] can be as effective. What is missing in this clinical scenario is a solid ground truth, which can only be obtainable from preliminary reports validating imaging parameters with targets specifically relevant for immunotherapy, as in the case of PD-L1 expression. Unfortunately, evidence in this regard is extremely limited, particularly when concerning metabolic imaging [94,101].
On the other hand, PET imaging during immunotherapy implies another frontier of development, with radiolabeled immune-based tracers, also known as Immuno-PET. This includes the targeting with radiolabeled antibodies, antibody fragments, or small proteins of checkpoints (i.e., CTLA-4, PD-1, PD-L1) [102][103][104][105][106], tumor infiltrating lymphocytes (ex. CD3, CD4, CD8) [107][108][109][110], cytokines (ex. IL-2) [111], enzymes (ex. Granzyme B, dCK deoxycytidine kinase, dGK deoxyguanosine kinase) [112][113][114][115], and potentially any other element involved in immune system response [116]. The possibility of detecting non-invasively checkpoint expression prior to the administration of ICI, as well as the identification on the entire tumor mass of the amount and pattern of distribution of immune cells, can have priceless clinical implications [106,110]. The same compound used for treatment, ex. ipilimumab, nivolumab, pembrolizumab, atezolizumab, and so forth, [105,106,[117][118][119], would be labeled and imaged with PET to detect the actual targeting of tumor sites ( Figure  4). Similarly, it would be able to detect the status of lymphocyte activation, exhaustion or cytotoxic capacity by simply injecting radiolabeled molecules targeting enzymes like Granzyme B, a downstream effector of tumoral cytotoxic T cells [113,115,120], or by checking the deoxyribonucleotide kinase activity [112,114]. The majority of data belong mostly to the preclinical setting, with ongoing research aiming to translate the results from bench to clinical practice [106,119,121]. The hope is that in the near future the data will be mature enough to implement immuno-PET into the diagnostic pathway for cancer patient candidates to undergo immunotherapy with checkpoint inhibitors.

Endnote Remarks
The introduction of immunotherapy in cancer treatment has represented a turning point in medical oncology, but also a new challenge for diagnostic imaging. The multitude of adapted response criteria and the numerous research studies published within a relatively short period of time demonstrate the capability of our community to face challenges and find solutions. From a nuclear medicine point of view, practical directives/guidelines are in the pipeline, along with previously published position papers or comments [122,123] on how to deal with the assessment of tumor response in the era of checkpoint inhibitors. The battlefield should, anyhow, move to clinical validation and recognition by the medical oncology community, which remains skeptical and firmly anchored to morphological criteria. Superior data are required in this regard, since non-inferiority would not be sufficient, given the larger availability of radiological devices (i.e., CT) and the reduced costs of the procedures compared to PET imaging. The astonishing technological leap of the last decade might be the game changer (immune-PET, Radiomics, AI), along with the improved awareness among nuclear medicine physicians of the clinical trial requirements in case of imaging studies, which should represent the backbone of any novel clinical indication or new tracer development.
Funding: This paper received no external funding.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on motivated request to the corresponding author.
Conflicts of Interest: E.L. reports receiving grants from AIRC (Associazione Italiana per la Ricerca sul Cancro) and from the Italian Ministry of Health, and faculty remuneration from ESMIT (European School of Multimodality Imaging and Therapy) and MI&T congressi. No other potential conflicts of interest relevant to this article exist.