Background
Lung cancer is one of the most common malignant tumors and represents the leading cause of death by cancer in Europe and in the USA [
1,
2]. Non-small cell lung carcinoma (NSCLC) accounts for 85% of lung cancers [
3], and close to 70% of patients present with locally advanced or metastatic disease at the time of diagnosis. Recently, immunotherapy has become a standard treatment after chemotherapy in NSCLC, whether locally advanced or metastatic, offering the potential for prolonged response and survival time [
4]. Two phase III studies have demonstrated higher overall response and survival rates in patients with NSCLC treated with nivolumab versus docetaxel [
5,
6].
Positron emission tomography/computed tomography (PET) with fluorodeoxyglucose (FDG) is widely used to assess the response to chemotherapy. However, the mechanisms of action of checkpoint inhibitors differ from those of chemotherapies. In some cases, the initial response to nivolumab may be an increase in the size and metabolism of tumor lesions due to lymphocyte infiltration, leading to false interpretations [
7]. Several response patterns have been described with computed tomography (CT), but data with FDG PET monitoring are scarce. Because of the new mechanisms of action of therapies and the heterogeneity of malignant disease, methods of assessing response are complex, and the role of PET is not well established.
Recently, several attempts to establish adapted imaging criteria as follow-up to immunotherapy treatment have been proposed, notably the immune Response Evaluation Criteria in Solid Tumors (iRECIST) [
8]. Regarding metabolic imaging, PET Response Criteria in Solid Tumors (PERCIST) have been proposed for solid tumor monitoring but have not been validated with new immunotherapies. Consequently, the authors proposed other response criteria using a single-time-point PET analysis for immunotherapy response assessment, especially in melanoma patients. Cho et al. studied patients with advanced melanoma under immunotherapy using an early PET evaluation method combined with a CT scan (3–4 weeks) to predict the best overall response at 4 months [
9]. The conclusion of the work was that the combination of functional and anatomical imaging parameters obtained early after the start of treatment appears predictive for eventual response at 4 months. Anwar et al. proposed using an absolute number of 4 newly emerged FDG-avid lesions on post-therapy PET scans in melanoma patients under ipilimumab to provide reliable information for treatment response [
10,
11]. However, as well-written in the paper from Pinker et al., a new response category, including an indeterminate response, based on a dual-time-point evaluation could be relevant during the metabolic evaluation of immunotherapy, as already shown with iRECIST [
12]. This indeterminate response or “unconfirmed” status recognizes that delayed responses or immune-mediated flares can both occur in the first months of immunotherapy treatment [
12]. This additional category provides the clinician flexibility to continue treatment in good general status patients and to perform a subsequent evaluation to confirm or deny a truly progressive disease [
12].
Based on this assumption, we chose to introduce and evaluate metabolic criteria based on a dual-time-point evaluation, as proposed with iRECIST. We designed a retrospective study to evaluate the prognostic value of FDG PET using modified PERCIST criteria adapted from iRECIST, which we called Immune PET Response Criteria in Solid Tumors (iPERCIST), in monitoring patients with NSCLC treated with an anti-programmed death-1 checkpoint inhibitor.
Discussion
In this study, FDG PET monitoring with iPERCIST was an effective tool for discerning NSCLC patients who could benefit from treatment with nivolumab. For responders to nivolumab according to iPERCIST (CMR, PMR, or SMD at SCAN-2, or with pseudo-progression confirmed by SCAN-3), the 1-year survival rate was greater than 90%, against 11% for non-responders. Additionally, OS was better for responders than non-responders at 19.9 vs. 3.6 months, p = 0.0003. Therefore, the prognostic value of iPERCIST might help physicians monitor immunotherapy in NSCLC patients.
FDG PET is currently the most widely used molecular imaging modality in clinical practice for staging and restaging NSCLC. However, few data are available for evaluating immunotherapy with FDG PET, especially in lung cancer, for which anti-PD-1 or anti-PDL-1-based immunotherapies are taking a crucial role in treating locally advanced or metastatic tumors [
15]. Moreover, because the antineoplastic activity of immunotherapy is related to the activation of T cells against tumor cells, FDG accumulation might cause false-positive findings, as was underlined in RECIST 1.1, evolving the criteria toward iRECIST [
8]. Consequently, implementing and evaluating PET-based criteria for immunomodulatory therapy [
16] is needed.
We proposed to use iPERCIST derived from PERCIST, which was introduced by R. Wahl in 2009 [
13,
14]. We modified iPERCIST by introducing two new categories of response derived from iRECIST: UPMD and CPMD, indicating that all metabolic progression observed at 8 weeks should be confirmed by another PET study 4 weeks later. However, the decision to continue immunotherapy treatment after the first evaluation is based on both clinical and imaging data. As recommended and discussed in the paper from Seymour et al. [
8], the continuation of treatment beyond imaging progression (UPMD in iPERCIST) is permitted in patients who are clinically stable until the next assessment. Pseudo-progression is a rare but clearly described condition under PD1 inhibitor treatment in lung cancer [
17]. At the time of the study, most of the described cases of pseudo-progression in lung cancer occurred in patients with clinical improvement or stabilization. The available data suggested that the UPMD patients with clinical deterioration had a progressive disease. Although these patients had no SCAN-3, they were followed after stopping immunotherapy; 6/9 UPMD patients showed tumoral progression on their CT scan approximately 2 months after the start of salvage chemotherapy, 2/9 patients died shortly (approximately 1 month) after the stopping nivolumab due to sepsis, and one patient had passed away at follow-up. Nevertheless, since the end of our study, a few case reports of patients with initial clinical worsening followed by a durable response were reported [
18].
One important point that should be highlighted is that SMD patients after UPMD are considered metabolic responders with iPERCIST in our study. Indeed, we observed that most SMD patients evaluated by PET after 4–6 cycles of nivolumab treatment had a possible sustained metabolic response. As illustrated in the survival curve in Additional file
3: Figure S2, OS did not significantly differ between SMD patients and CMR + PMR patients when evaluated by iPERCIST.
The comparison of the results obtained with iRECIST and iPERCIST suggests that iPERCIST might be more relevant than iRECIST in terms of prognostic information. Indeed, 11/28 patients (39%) were classified differently with iPERCIST. In a review of the 2016 literature (6 series, 268 patients), discrepancies between RECIST and PERCIST were observed in approximately 38% of tumor responses, in line with our rate of disagreement [
19]. In our study, PERCIST allowed five morphologically stable patients to be reclassified as PMR. The OS of these five patients was longer than the OS of the whole study population (520 vs. 459 days, respectively). Similarly, 2/11 patients were reclassified from progressive disease to stable disease, with their OS longer than the OS of the whole study population (573 vs. 459 days, respectively). Two other morphologically stable patients were reclassified as PMD, with their OS shorter than the OS of the whole study population (360 vs. 459 days). These data suggest that reclassification using PERCIST provides additional prognostic information compared to iRECIST in this setting.
Several studies have documented the value of FDG PET in patients receiving nivolumab, whether with large-cell B lymphomas [
20] or melanomas [
9]. In NSCLC, Kaira et al. recently investigated the role of FDG PET as a biomarker of therapeutic efficacy in a prospective study of 24 patients [
21]. The authors showed that changes in FDG uptake evaluated by total lesion glycolysis 1 month after the start of nivolumab had significant prognostic value. Of note, SUVmax or metabolic tumor volumes failed to have prognostic value in multivariate analyses in their study. In addition, PERCIST using SULpeak was not evaluated in their study, although we considered PERCIST to be the most reliable criteria available. Indeed, the superiority of PERCIST over SUVmax has been demonstrated, especially in NSCLC treated with radiotherapy [
22] or chemotherapy [
23].
However, other authors showed that increased FDG uptake during immunotherapy predicted immune activation and could predict the best overall response [
9]. Cho et al. studied the prognostic value of FDG PET 1 month after the start of treatment in patients with melanoma. In patients with stable disease at month 1 according to RECIST 1.1, PET could differentiate between two populations where an increase of 15.5% in SULpeak of the lesion with the highest uptake was associated with eventual clinical benefit (i.e., partial or complete response at 4 months or stable disease ≥ 6 months). The authors interpreted their findings by suggesting that an early inflammatory response at the site of the tumor, induced by boosting immunity, is a marker of a sustained response. This study lacked additional PET imaging after SCAN-2 to confirm the presence of an immune reaction rather than tumoral progression, which could explain the low accuracy and negative predictive value of PERCIST (70% and 42%, respectively) for predicting the best overall response after 4 months, because of the possible misclassification of pseudo-progression in PMD.
Recent studies of melanoma patients receiving ipilimumab and those evaluated by FDG PET suggested the use of new criteria for imaging evaluation [
10,
11]. The PET Response Evaluation Criteria for Immunotherapy (PERCIMT) added new criteria defining a threshold of two to four new lesions in an evaluation scan to classify patients under immunotherapy as PMD. Again, a single-time-point evaluation here does not seem reliable enough in the setting of immunotherapy and can lead to a risk of misclassifying pseudo-progression as disease progression. This issue was observed in initial observations, which found that 10% of patients who received ipilimumab for melanoma showed a clinical response (including partial response and stable disease) that would have been misclassified as disease progression by World Health Organization criteria [
24]. Using FDG PET, we assumed that the risk of misclassification might be even higher because of the better sensitivity of FDG PET, particularly for detecting small lymph nodes or bone metastasis. Notably, we identified 1 patient (1/28, 4%) with a pseudo-progression, which is in accordance with the rate from the literature that ranges from 0 to 6% [
25].
Our study is mainly limited by its retrospective design and potential recruitment bias. Of note, the response rate in our study was higher than that usually reported with studies of nivolumab at 40% compared with 20% reported in a recent publication [
4]. The choice of the most relevant imaging test evaluation in the context of immunotherapy treatment is an important issue, particularly considering that CT scans are less expensive and more widely available compared to PET scans. Prospective studies will need to validate the cost-effectiveness of a whole-body metabolic evaluation with FDG PET compared to conventional anatomical assessment. Finally, we did not compare FDG PET with predictive biomarkers, such as PD-L1 expression in tissue sampling [
16], because of its unavailability at the time of the study.