The first PET-based response criteria embedding metabolic response to treatment in solid tumours were proposed by the European Organisation for Research and Treatment of Cancer (EORTC) in 1999 [
19]. Based on these criteria, a complete metabolic response (CMR) is reached when all tumour lesions are no longer detectable against adjacent background activity, whereas progressive metabolic disease (PMD) is defined as an increase in SUVmax of ≥25% from baseline imaging or the appearance of new metastatic lesions. The EORTC criteria do not specify the number of lesions to be measured or the minimum measurable lesion SUVmax, but rather refer to the background activity for the definition of CMR. A partial metabolic response (PMR) is defined as a reduction in SUVmax of between 15% and 25% or >25% after one or more cycles of chemotherapy. Stable metabolic disease (SMD) is considered a response not classifiable in any of the other categories. The EORTC criteria were the first to be applied for the assessment of response of solid tumours to immunotherapy [
20]. A summary of the available and/or proposed response criteria with for use with FDG PET is presented in Table
1.
Table 1
Available and/or proposed response criteria for use with FDG PET
Complete response (CR) | Complete resolution of FDG uptake | Disappearance of all metabolically active tumours | RECIST 1.1 (disappearance of all target lesions; reduction in short axis of target lymph nodes to <1 cm; no new lesions) | Clinical benefit | Complete resolution of all preexisting 18F-FDG-avid lesions; no new 18F-FDG-avid lesions | Clinical benefit |
Partial response (PR) | Minimum reduction of ±15–25% in tumour SUV after one cycle of chemotherapy, and >25% after more than one treatment cycle | Decline in SULpeak by 0.8 unit (>30%) between the most intense lesion before treatment and the most intense lesion after treatment | RECIST 1.1 (decrease in target lesion diameter sum >30%) | Clinical benefit | Complete resolution of some preexisting 18F-FDG-avid lesions. No new, 18F-FDG avid lesions. | Clinical benefit |
Stable disease (SD) | increase in SUV of less than 25% or a decrease of less than 15% | Does not meet other criteria | Does not meet other criteria | Change in SULpeak of the hottest lesion of >15% | Clinical benefit | Neither PD nor PR/CR | Clinical benefit |
Change in SULpeak of the hottest lesion of ≤15% | No clinical benefit |
Progressive disease (PD) | Increase in tumour FDG uptake of >25%; increase in maximum tumour of >20%; new metastases | Increase in SULpeak of >30% or the appearance of a new metabolically active lesion | RECIST 1.1 (increase in target lesion diameter sum of >20% and at least 5 mm or new lesions) | No clinical benefit | Four or more new lesions of <1 cm in functional diameter or three or more new lesions of >1.0 cm in functional diameter or two or more new lesions of more than 1.5 cm in functional diameter | No clinical benefit |
In 2009, ten years after the introduction of the first PET-based criteria, Wahl et al. proposed the PET Response Criteria in Solid Tumors (PERCIST) [
21]. They are rather similar to the EORTC criteria, and therefore the response assessment provided by PERCIST tend to give very similar results, but with some differences in terms of response classification (Table
1). The major innovations of PERCIST were the use of SUV lean (SUV normalized by lean body mass, or SUL) for the assessment of tumour response and the identification of a minimum tumour SUL equivalent to 1.5 times the mean SUL of the liver for a lesion to be evaluable. PERCIST also show some similarities with the morphological criteria (i.e. RECIST), by recommending the measurement of SUL in up to five tumours (up to two per organ) corresponding to the target lesions. These criteria were also the first to introduce the concept of SULpeak within the area of highest uptake in the tumour, which can be measured within a spherical region of interest of diameter 1.2 cm (1 cm
3 volume). The use of PERCIST criteria with respect to response to immunotherapy has been described only rather recently [
22].
In an attempt to find the perfect fit between morphological and metabolic responses, Cho et al. [
23] evaluated different criteria (i.e. RECIST 1.1, irRC, PERCIST and EORTC) in a small cohort of 20 patients with advanced melanoma treated with either ipilimumab (
n = 17) or nivolumab (
n = 3). This imbalance in type of treatment agent somewhat limits the generalizability of these criteria since the authors found pseudoprogression, which is implied by these criteria, mainly with ipilumumab and seldom with anti-PD1/PD-L1 agents. Notwithstanding this observation, the cohort was prospectively investigated after days 21–18 and 4 months following the start of therapy with the aim of defining the best combination for response assessment in immunotherapy. In particular, the best combination of parameters, which were termed by the authors as PECRIT (PET/CT Criteria for Early Prediction of Response to Immune Checkpoint Inhibitor Therapy), included either a change in the sum of RECIST 1.1-based target lesion diameters (method 1), and a change in SULpeak of >15.5% of the hottest lesion (method 3) [
23]. Combining morphological and metabolic criteria led to an accuracy of 95% (sensitivity 100%, specificity 93%). One of the most interesting aspects of this study is the introduction of clinical benefit (CB) into the definition of response (Table
2). In particular, this classification applies to patients with a complete response (CR) or partial response (PR) according to morphological criteria plus all patients with stable disease (SD) with a decrease in SULpeak greater than the cut-off value of 15.5%.
Table 2
Principal studies investigating the role of FDG PET/CT in the evaluation of response of solid tumours to immunotherapy
| Prospective | 22 | Melanoma | Ipilimumab | EORTC after two cycles of treatment (early) and at the end of treatment after four cycles (late) | Early response evaluation after (two cycles) is predictive of final treatment outcome in patients with PMD and SMD |
| Prospective | 27 | Melanoma | 20 pembrolizumab, 7 nivolumab | Visual analysis (qualitative visual inspection, positive when FDG uptake greater than background activity or hepatic uptake; Deauville score) | 43% of patients who had residual disease by CT criteria, either PR or SD, were FDG-negative |
| Prospective | 31 | Melanoma | Ipilimumab | Fractal and multifractal analysis before and after two and after four cycles of treatment | Operator-independent method with a correct classification rate of 83.3% |
| Prospective | 20 | Melanoma | 16 Ipilimumab, 1 nivolumab, 3 BMS-936559 | RECIST 1.1 and PERCIST at early (4 weeks) and late assessment (4 months) | Combined anatomical and functional data at 21–28 days (PECRIT) criteria predicted response with 100% sensitivity, 93% specificity and 95% accuracy. Introduction of clinical benefit in response criteria |
| Prospective | 24 | NSCLC | Nivolumab | RECIST 1.1 versus PERCIST; additional semiquantitative analyses (SUVmax, MTV, TLG) | Metabolic response on PET (especially TLG) associated with therapeutic response and survival at 1 month after nivolumab |
| Prospective | 27 | NSCLC | 23 nivolumab, 4 pembrolizumab | Baseline semiquantitative analysis | SUVmax ≤17.1 (sensitivity 88.9%) or a SUVmean ≤8.3 (sensitivity 100%) identified fast progression after 8 weeks of therapy |
| Prospective enrolment, retrospective PET analysis | 41 | Melanoma | Ipilimumab | RECIST and appearance of new FDG-avid lesions (PERCIMT); patients were dichotomized into those with and those without clinical benefit | A cut-off of four newly emerged FDG-avid lesions on posttreatment PET/CT gave reliable indication of treatment failure |
| Prospective | 41 | Melanoma | Ipilimumab | EORTC and PERCIMT after two cycles of immunotherapy | PERCIMT to interim PET/CT provides a more sensitive predictor of final response than EORTC criteria |
CB also appears to be the main goal for the criteria recently proposed by the group from Heidelberg for FDG PET evaluation of response to immunotherapy in patients with melanoma [
24,
25]. The PERCIMT (PET Response Evaluation Criteria for Immunotherapy) classification takes into consideration the observed relevance of the absolute number of new lesions on FDG PET scan and its more robust predictive role compared to pure SUV changes during the course of treatment with ipilimumab. In particular, the authors dichotomized patients according to CB from the treatment (CR/PR and SD) or no CB from the treatment, i.e. progressive disease determined as the appearance of: (a) four or more new lesions <1 cm in functional diameter, (b) three or more new lesions >1.0 cm in functional diameter, or ©) two or more new lesions >1.5 cm in functional diameter. In all cases, the functional diameter is considered the lesion diameter measured in centimetres based on the fused PET/CT images.
In a cohort of 20 patients with advanced melanoma treated with either ipilimumab or nivolumab [
23] with response to treatment assessed early (days 21–18) and late (4 months) after the start of therapy, separate early assessment with PERCIST and EORTC criteria demonstrated suboptimal accuracies of 70% and 65%, respectively, for the prediction of best overall response at 4 months. Therefore, the authors proposed PECRIT (the combined criteria) that had an accuracy of 95%, as being better associated with CB in melanoma patients treated with immunotherapy.
FDG PET for immunotherapy response assessment
The principal studies investigating the role of FDG PET/CT in the evaluation of response of solid tumours to immunotherapy are summarized in Table
2. Given the major impact that immunotherapy with checkpoint inhibitors has had on the treatment of metastatic melanoma, it is not surprising that the first report on metabolic response during the course of immunotherapy related to FDG PET evaluation after two cycles of ipilimumab and at the end of treatment in 22 patients with melanoma [
20]. In this initial analysis, the EORTC criteria were used for response assessment and showed that early PET (after two cycles) was predictive of outcome (late response) in patients with PMD and SMD. Already in this first report, the authors recognized the appearance of new lesions, conventionally defining disease progression, as being a potential cause of response misclassification.
One year later, another group [
26] investigated the role of residual metabolic activity on FDG PET in patients with metastatic melanoma presenting with a prolonged response to immunotherapy with anti-PD1 agents (i.e. pembrolizumab and nivolumab). No defined metabolic response criteria were used for the definition of response on PET, and the authors relied mostly on visual assessment and qualitative analysis based on background tissue comparisons in a manner similar to the Deauville score. Overall, 27 patients were analysed, of whom 15 (56%) had a positive FDG PET scan with a biopsy-proven melanoma residue in eight (62%). Of the remaining 12 patients with a negative PET scan, six presented with a residual lesion on CT and five had ceased treatment, but none of these patients showed recurrence during 6–10 months of follow up. In summary, 43% of patients with residual disease based on CT criteria (either PR or SD) were negative on FDG PET. Occasionally, metabolically active lesions in patients with CB from immunotherapy in the long term may show positive findings on PET that can be considered to be a result of immune cell infiltrates rather than melanoma localizations.
A more recent study including 20 patients with advanced melanoma treated with either ipilimumab or nivolumab [
23] assessed responses to treatment early (days 21–18) and late (4 months) after starting therapy. Various morphological criteria (RECIST 1.1, irRC) and metabolic criteria (PERCIST and EORTC) were directly compared to define the best combination for the assessment of response to immunotherapy with checkpoint inhibitors. Interestingly, the authors found low inter-criteria agreement (kappa = 0.48–0.7) among RECIST, PERCIST and EORTC in the early assessment, whereas there was good to excellent agreement between CT modalities and PET in the late evaluation.
Another recent study prospectively enrolled 41 patients with metastatic melanoma treated with ipilimumab [
24] and evaluated at baseline and 3 months later. Changes in SUVmax and SUVmean during the course of immunotherapy were not correlated with clinical response (
t test;
p = 0.06 and 0.05, respectively), whereas the number of new lesions was able to define disease progression (Tables
1 and
2), improving the identification of patients who will show CB (Wilcoxon test,
p < 0.0001). Moreover, optimal cut-off points for the total number of new lesions on the basis of their functional diameter (measured diameter on fused PET/CT images) were defined: (a) four new lesions led to an observed sensitivity of 84% and a specificity of 100%, (b) a functional size >1.0 cm for a cut-off of three new lesions led to a sensitivity of 90% and a specificity of 90%, and ©) a functional size >1.5 cm for a cut-off of two new lesions led to a sensitivity of 94% and a specificity of 90%. By combining all the available data, the authors proposed the PERCIMT criteria.
The group from Heidelberg in the same cohort of patients also investigated the applicability of PERCIMT at interim evaluation after two cycles of immunotherapy (Table
2) and compared the results with those using the EORTC criteria [
25]. Patients were divided into two groups, those showing metabolic benefit, including SMD, PMR and CMR, and those showing no metabolic benefit (PMD). Overall, agreement between the two sets of response criteria was poor (kappa = 0.46; McNemar test
p = 0.001). The PERCIMT showed a significantly higher sensitivity than the EORTC criteria in predicting CB (93.6% versus 64.5%, respectively;
p = 0.004), but did not show a significantly higher specificity (70.0% versus 90.0%, respectively;
p = 0.5) in predicting no CB. The superiority of the new proposed response criteria is therefore questionable, first because of the limited number of patients on which the PERCIMT were developed (
n = 41), and second because the EORTC criteria appear to be better at identifying patients who will not respond to ipilimumab than the PERCIMT, although the difference in specificity did not reach significance.
Only a few studies have investigated the response of tumour types other than melanoma to immunotherapy, especially non-small cell lung cancer (NSCLC), and some are case reports [
22,
27‐
29]. In a recent study assessing response of NSCLC to immunotherapy [
22], 24 patients treated with nivolumab were investigated at baseline and 1 month after the start of treatment. Response was determined using either morphological (RECIST 1.1) or PERCIST criteria, along with SUVmax, metabolic tumour volume (MTV) and total lesion glycolysis (TLG). The value of PET in predicting PR and progressive disease was significantly higher than that of CT. This was also shown in a multivariate analysis that confirmed FDG uptake (i.e. TLG) after administration of nivolumab as an independent factor predicting PFS (HR 3.624;
p < 0.001) and OS (HR 2.461;
p = 0.012).
In a recent study of the use of an anti-PD-L1 agent, atezolizumab, in the treatment of NSCLC, the potential of FDG PET/CT for assessing response was evaluated. FDG PET scans at baseline and 6 weeks were evaluable in 103 patients. Patients with an early FDG response at 6 weeks according to the EORTC criteria achieved a higher objective response rate on subsequent CT than metabolic nonresponders (17/ 23, 73.9% versus 5/80, 6.3%). Possible pseudoprogression was identified in only two patients [
30].
Additional considerations with respect to tumour metabolism
Another important aspect to be considered, particularly during baseline evaluation, is that FDG PET can provide useful information on the metabolic state of the tumour microenvironment and on the expression of checkpoint inhibitors. Indeed, in patients with NSCLC, there is a statistically significant association between tumour metabolic parameters on PET and PD1/PD-L1 expression, along with the presence of CD8+ tumour infiltrating lymphocytes (TILs), in resected tumour specimens [
31,
32]. The presence of immune infiltrate is already known as a good predictor of response to immunotherapy. Several studies have shown that responding patients have a significantly higher expression of CD8+ TILs, and PD1 and PD-L1 cells before treatment than patients with progression [
33,
34]. In addition, Mazzaschi et al. [
35] found that patients with CB and longer progression-free survival following treatment with nivolumab showed CD8+ lymphocytes with low expression of PD1, while the PD1-to-CD8 ratio was a prognostic factor in univariate and multivariate analyses.
It is not surprising that some initial evidence, although limited to a cohort of 27 patients with NSCLC [
28], has shown the value of FDG PET in predicting response to immunotherapy with checkpoint inhibitors. Grizzi et al. [
28] found that almost all patients classified as fast progressors after 8 weeks of immunotherapy showed SUVmax ≤17.1 or SUVmean ≤8.3 on baseline PET. The apparently low specificity of these cut-off values, which conversely maintain high sensitivity, is attributable to the fact that response to immunotherapy depends on multiple factors. Imaging and metabolic data, analysed visually, semiquantitatively or with dedicated algorithms [
36], are pieces of the “puzzle”. As a consequence, the metabolic characteristics of the tumour and its environment at baseline may be part of a larger panel of predictive factors of response to immunotherapy [
37].