Lung cancer is the second most common malignant tumor with the highest mortality rate worldwide, of which non-small-cell lung cancer (NSCLC) accounts for about 85% (1, 2). Due to the relatively insidious symptoms of early NSCLC, most diagnosis presents local spread or distant metastasis, resulting in poor prognosis. As one of the most important breakthroughs in oncology in recent years, immune checkpoint inhibitors (ICIs), characterized by programmed death 1 (PD-1), programmed death ligand 1 (PD-L1) and cytotoxic T lymphocyte associated antigen 4 inhibitors, have dramatically changed the treatment landscape, and even advanced into the first-line treatment of advanced or metastatic NSCLC (3–5). Immunotherapy can significantly improve the patient’s clinical outcome and for the first time offer hope of long-term survival for NSCLC patients. However, not all patients can benefit from immunotherapy, because only a small percentage of NSCLC patients are sensitive to ICIs (6, 7). The response rate for unscreened NSCLC patients treated with ICIs alone is usually less than 20% (3). On the other hand, unlike most traditional treatments, immunotherapy usually shows delayed response, pseudoprogression, hyperprogression and other phenomena, as well as immune related adverse events (8, 9), which brings difficulties and challenges to clinical treatment decision-making.

The complexity of immunotherapy response challenges the monitor of treatment assessment using conventional methods. Predictive biomarkers are required to determine immunotherapy achievement and survival benefit. PD-L1 expression is a typical and predictive biomarker for ICIs treatment, and patients who benefit most from immunotherapy can be selected (10). However, predictions based on PD-L1 levels are sometimes biased. Some patients without PD-L1 expression can achieve certain remission and control of lesions after immunotherapy, while some patients with PD-L1 expression cannot benefit from immunotherapy (11, 12). Tumor mutation burden is another new potential biomarker that can predict progression-free survival (PFS), but its prediction of overall survival (OS) is not reliable enough (13, 14). As the most important weapon of anti-tumor immunotherapy, tumor-infiltrating lymphocytes (TILs) CD3⁺, CD8⁺ and forkhead box protein 3 (FOXP3⁺) may be independent predictors of clinical benefit of ICIs (15–17). The most significant limitation, however, is the inability to predict efficacy from their density and location prior to treatment. Therefore, it is the focus and difficulty of ICIs treatment to find effective prognostic markers to predict immunotherapy response and screen out patients who really benefit.

With the progress of technical approaches and the wide application of ICIs, studies on dynamic monitor of treatment response and prognosis based on tissue specimen, liquid biopsy and imaging are gradually increasing. ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) is a functional noninvasive imaging modality based on glucose metabolism, which is of great value in cancer diagnosis, staging, efficacy judgment and prognosis evaluation (18). Malignant tumor cells proliferate and divide in an abnormally active manner, requiring more energy and leading to enhanced glycolysis. The high expression of glucose transporter type 1 can transport a large amount of glucose to meet its demand for rapid reproduction, resulting in the gradual aggregation of ¹⁸F-FDG in tumor cells and abnormal radioactive hyperconcentration of lesions (19). ¹⁸F-FDG PET/CT can reflect the systemic metabolism of malignant tumors at the molecular level (20, 21), so it can diagnose and evaluate the therapeutic efficacy of malignant tumors earlier and more accurately, and monitor survival prognosis. Several studies have shown that PET/CT metabolic parameters including the maximum standardized uptake values (SUV_max), the mean standardized uptake values (SUV_mean), metabolic tumor volume (MTV) and total lesion glycolysis (TLG) have important predictive value in the prognostic evaluation of malignant tumors. Some studies have found that SUV_max can be used as a potential predictor of ICIs response (22, 23), while other studies have not found their relationship (24). MTV is usually considered as a typical prognostic parameter that reflects tumor volume with metabolic activity and the metabolic volume of anatomical lesions. TLG includes tumor metabolic activity and tumor metabolic volume, which can fully reflect the overall metabolic characteristics of tumor lesions. Studies have proved that baseline MTV and TLG can better predict prognosis than SUV_max for NSCLC patients undergoing surgery, radiotherapy, chemotherapy and targeted therapy (25–28). In terms of immunotherapy, the prognostic and predictive value of these metabolic parameters in NSCLC patients treated with ICIs is still unknown. This study aimed to investigate the prognostic value of baseline ¹⁸F-FDG PET/CT metabolic parameters MTV, SUV_max, SUV_mean, and TLG for survival in advanced or metastatic NSCLC patients treated with ICIs.

PET/CT has been proved to be of great value in the diagnosis, clinical staging, radiotherapy localization, efficacy evaluation and prognosis of NSCLC (31, 43), but its prognostic status in ICIs treatment is unclear. This meta-analysis evaluated the prognostic value of baseline metabolic parameters of ¹⁸F-FDG PET/CT for advanced or metastasis NSCLC by analyzing the HR of PFS and OS in patients with high SUV_max, SUV_mean, MTV or TLG versus those with low SUV_max, SUV_mean, MTV or TLG. The pooled results showed patients with high MTV and TLG prior to ICIs treatment may had shorter OS than patients with lower metabolic rate, suggesting that MTV and TLG may be meaningful prognostic markers. High MTV may also predict worse PFS, and patients with high SUV_mean had better PFS. No correlation was found between SUV_max and OS or PFS.

Unlike traditional tumor therapy by killing cancer cells, ICIs work in an immunomodulatory way maybe by modulating tumor metabolism. The proliferation rate of tumor cells is positively related to their glucose uptake (44). Higher uptake of FDG leads to more aggressive tumors and worse prognosis (45). SUV has independent and prognostic value for OS in NSCLC patients, but most studies are based on early NSCLC (46–48). SUV_max and SUV_mean are associated with the presence of CD8⁺ TILs in patients with NSCLC (20). Takada et al. also linked high SUV_max to positivity for PD-L1 (49). These findings suggest that high FDG uptake may indicate an immunologically cold tumor. However, previous studies have also found that this may not be true for patients with advanced NSCLC. SUV_max is, to a lesser extent, considered to be an effective predictor of prognosis in patients with NSCLC, regardless of treatment or disease stage (50, 51). Our pooled results also found that baseline SUV_max did not provide any prognostic information. High SUV_mean may predict longer PFS, but had no correlation with OS. The results of subgroup analysis showed no predictive value of SUV_max and SUV_mean in first line group and undefined line group. It is worth noting that pseudoprogression can lead to possible misjudgment by PFS. Both SUV_max and SUV_mean can represent the uptake level of FDG, but cannot reflect the overall metabolism of tumor. Especially, SUV does not consider the effect of tumor volume on prognosis, and only reflects the metabolic level of mononectin in the lesions (52, 53).

When extensive metastasis occurs, the tumor load is large, which may be more valuable than the metabolic activity for primary lung cancer prognosis. Liao S et al. (54) found that MTV and TLG were better predictors of clinical outcomes in patients with stage IV NSCLC than SUV_max and SUV_mean. MTV and TLG are another type of metabolic parameters that take tumor metabolic volume and metabolic activity into comprehensive consideration. MTV can not only calculate the tumor volume, but also display the metabolic activity of the lesions. However, TLG combined with MTV and glycolysis can analyze the overall metabolic status of lesions more comprehensively. The pooled results of this study showed OS and PFS was shortened in high MTV group, indicating the larger the tumor metabolic volume and metabolic load, the shorter the possible survival, the worse the prognosis. High TLG group also showed shorter OS than low TLG group, but no significance was found for PFS. Nonetheless, significant heterogeneity was found in the analysis of MTV and TLG predicting PFS and OS. The subgroup analysis of treatment line showed that OS in high MTV group was significantly shorter than that in low MTV group in either first line group or undefined line group. However, high MTV may only predict poor PFS in first line group, but not in undefined line group. Meanwhile, high TLG was associated with poor OS in undefined line group, but not in first line group. MTV has shown good prognostic ability, but the predictive significance of TLG is unclear in the different treatment lines.

There is no theoretical explanation of the mechanism by which these metabolic parameters serve as prognostic markers for ICIs treatment. Advanced or metastatic NSCLC is often accompanied by lymph node metastasis, or liver, bone, brain and other organ metastasis. Therefore, the impact of metastasis other than the primary site of tumor should be focused on in prognosis assessment. Systemic tumor metabolic load includes the metabolic load of primary tumor, metastatic lymph node and distant metastasis, which more comprehensively reflects systemic metabolic information. MTV and TLG measure tumor volume and metabolic activity in three dimensions, providing more information about tumor aggressiveness. This study shows that NSCLC patients with high baseline MTV have a poor prognosis after ICIs treatment. MTV reflects the real volume of tumors with high metabolic activity in tumor tissues, and the high accumulation of FDG may be related to tumor necrosis or hypoxia caused by larger tumor deposition. Tumor necrosis, acidity and hypoxia can promote the establishment of immunosuppressive tumor microenvironment by recruiting a variety of immunosuppressive cells, such as myeloid-derived suppressor cells, regulatory t cells (Tregs) and tumor-associated macrophages (55, 56). These are closely associated with tumor recurrence and survival deterioration (57, 58). In this environment, FOXP3⁺, the major regulator of Tregs, has been reported to be positively correlated with MTV (59, 60). Whereas, in the anoxic environment caused by tumor necrosis and inflammation, no correlation has been found between PD-L1 expression and MTV (61). Studies have found that NSCLC patients with positive PD-L1 expression have significantly higher MTV than patients with negative PD-L1 expression (62), While other studies have shown that MTV is not significantly associated with PD-L1 expression (60, 63). Although clinical trials have confirmed that high expression of PD-L1 can make tumor cells more sensitive to immunotherapy (64, 65), and can also mediate immune escape (66, 67). PD-L1 may play a dual role, but it is vague what mechanism determines PD-L1 to activate drug resistance or response pathways. In summary, an environment with high MTV may form an immunosuppressive state, thus leading to drug resistance of ICIs, regardless of PD-L1 expression. MTV may be a better parameter to predict immunotherapy response and prognosis, but its mechanism needs to be clarified.

This is the first meta-analysis to evaluate ability of baseline metabolic parameters of ¹⁸F-FDG PET/CT for survival in advanced or metastatic NSCLC patients treated with ICIs. Our results suggest that MTV can be considered as a basis for patient stratification in the immunotherapy and prognosis of advanced or metastatic NSCLC patients. This provides some data support for clinical application of ¹⁸F-FDG PET/CT to judge the prognosis of NSCLC patients and guides the treatment of lung cancer. Although analysis software is available to measure these metabolism parameters (40–42), MTVand TLG are not typical components of standard PET/CT reports because of insufficient evidence of clinical application value and increased workload. As the clinical significance of these parameters is fully validated, radiologists will become aware of their potential applications. It is economical and feasible to help patients select the best treatment strategy by obtaining baseline MTV and TLG of NSCLC patients only once before ICIs treatment in the future. Early identification of patients who do not benefit from immunotherapy may lead to timely decisions and discontinuation of treatment, potentially reducing drug toxicity and the high cost of immunotherapy and initiating different treatment regimens earlier.

However, the study has several limitations. Firstly, heterogeneity was detected in the present meta-analysis. This may be due to bias in detection methods, differences in ICIs treatment, etc. Secondly, we cannot determine an optimal threshold to classify volume parameters as high or low. Different cut-off values and delineation strategies, as well as different histological methods, may influence the occurrence and survival of events. Thirdly, there are different types of NSCLC, such as adenocarcinoma and squamous cell carcinoma, with different prognosis. However, in the currently included studies, all of these subtypes are mixed and these studies can yield different results in response prediction and monitoring with ¹⁸F-FDG PET/CT imaging. Therefore, further large-scale prospective studies are necessary to evaluate whether MTV and TLG can be independent prognostic factors for clinical outcome in patients with different types of NSCLC.

Baseline MTV may have strong predictive power for survival, while baseline SUV_max and SUV_mean may not be appropriate prognostic markers in advanced or metastatic NSCLC patients treated with ICIs. Baseline TLG may predict OS for only a subset of patients, and its predictive value remains to be examined.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

TL and LZ conducted the literatures screening. TL and RP conducted the statistical analysis. TL, LZ and LH wrote the manuscript. LH and CY revised the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.