Diagnostic and Predictive Value of Using RGD PET/CT in Patients with Cancer: A Systematic Review and Meta-Analysis

The purpose of this study was to assess the diagnostic value of arginine-glycine-aspartic acid (RGD) PET/CT for tumor detection in patients with suspected malignant lesions and to determine the predictive performance of RGD PET/CT in identifying responders. Methods. The PubMed (Medline), EMBASE, Cochrane Library, and Web of Science databases were systematically searched for potentially relevant publications (last updated on July 28th, 2018) reporting the performance of RGD PET in the field of oncology. Pooled sensitivities, specificities, and diagnostic odds ratios (DORs) were calculated for parameters. The areas under the curve (AUCs) and Q⁎ index scores were determined from the constructed summary receiver operating characteristic (SROC) curve. We explored heterogeneity by metaregression. Results. Nine studies, five including 216 patients that determined diagnostic performance and three including 75 patients that determined the predictive value of parameters, met our inclusion criteria. The pooled sensitivity, pooled specificity, DOR, AUC, and Q⁎ index score of RGD PET/CT for the detection of underlying malignancy were 0.85 (0.79-0.89), 0.93 (0.90-0.96), 48.35 (18.95-123.33), 0.9262 (standard error=0.0216), and 0.8606 for SUVmax and 0.86 (0.80-0.91), 0.92 (0.88-0.94), 40.49 (14.16-115.77), 0.9312 (SE=0.0177), and 0.8665 for SUVmean, respectively. The pooled sensitivity, pooled specificity, DOR, AUC, and Q⁎ index score of RGD PET/CT for identifying responders were 0.80 (0.59-0.93), 0.74 (0.60-0.85), 15.76 (4.33-57.32), 0.8682 (0.0539), and 0.7988, respectively, for SUVmax at baseline. Conclusion. The interesting but preliminary data in this meta-analysis demonstrate that RGD PET/CT may be an ideal diagnostic tool for detecting underlying malignancies in patients suspected of having tumors and may be able to efficiently predict short-term outcomes.


Introduction
Angiogenesis, the process of new blood vessel formation from preexisting vasculature, is recognized as a key mechanism involved in tumor growth, invasion, and metastasis [1]. In 1971, Judah Folkman first stressed that a tumor could not grow larger than a few millimeters in diameter without angiogenesis occurring to transport nutrients and oxygen [2,3]. The vascular effects of antitumor therapy on microvessel density may precede the impact on tumor size by a long interval, especially when the tumor volume cannot change significantly [4].
Integrin-mediated cell adhesion plays an important role in many essential normal cellular and pathological functions [5]. Integrin v 3, which has been widely studied, is significantly upregulated in tumor cells and activated endothelial cells but not in resting vessel cells in normal regions [6]. Therefore, imaging integrin v 3 expression may be valuable for diagnosis and assessing suitable patients for particular treatments. Imaging of various tripeptide RGD sequencecontaining integrins has been extensively evaluated because of their high affinity and specificity with regard to integrin v 3. Compared to other methods of imaging v 3, the PET/CT approach is likely to be widely applied in tumor patients because of its high sensitivity to low amounts of tracer and its unlimited depth penetration [7]. Indeed, there have been great efforts to develop radiolabeled RGD peptides, and 18 F, as a radioisotope for labeling peptides, is commonly used because its half-life is suitable for routine clinical use [8]. In addition, 18 F-Galacto-RGD [9][10][11][12][13][14], 18 F-Fluciclatide-RGD [15,16], 68 Ga-NOTA-PRGD2 [17], 18 F-Alfatide [18], 18 F-FPPRGD2 [19], and 18 F-Alfatide II [20,21] have been under clinical investigation for their relevance to the diagnostic ability of RGD PET/CT, but only sensitivity has been studied to date. Among them, 18 F-Alfatide II and 68 Ga-NOTA-PRGD2 show advantages over the others in terms of easy preparation, fast labeling, and in vivo pharmacokinetics [22]. The results of the majority of the studies have demonstrated positive potential, though some studies have had small sample sizes. To further evaluate the diagnostic ability of RGD PET/CT, we conducted a meta-analysis of the clinical literature to obtain data regarding the sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR).
Even when treated by the standard therapy of concurrent chemoradiotherapy (CCRT), one-third of patients with advanced non-small cell lung cancer (NSCLC) still experience local treatment failure [23]. Thus, finding an effective prediction tool to select patients likely to respond to therapy would help guide individualized treatment. Although morphologic treatment effects cannot be detected by conventional imaging techniques at early time points but only after several weeks or months [24,25], imaging tumor angiogenesis may be used for patient risk stratification before starting therapy [22]. Therefore, we performed a meta-analysis to evaluate the role of RGD PET/CT in predicting the shortterm outcomes of therapy among NSCLC patients.

Methods
. . Literature Search Strategy. All relevant articles were retrieved from the PubMed (Medline), EMBASE, Cochrane Library, and Web of Science databases. The databases were searched based on the following keywords and text words: ("RGD" OR "alfatide") and ("PET" OR "positron emission tomography") and ("neoplasms" OR "tumor" OR "cancer" OR "neoplasm" OR "tumour"). The search was last updated on July 28th, 2018. There were no beginning date restrictions.
. . Inclusion and Exclusion Criteria. Published articles were selected according to the following inclusion criteria: (1) original articles revealing the performance of RGD PET or PET/CT for the diagnosis of neoplasms or predicting the efficacy of treatment; (2) studies in which the final diagnoses of patients were confirmed by histopathology or comprehensive assessment containing clinical and radiologic follow-up; (3) studies in which RGD PET or PET/CT was used as the single reference standard for neoplasm diagnosis; (4) studies in which the short-term efficacy of cancer therapy was based on tumor regression grading (TRG), response evaluation criteria in solid tumors (RECIST), or changes in the volumes of residual lesions on MRI; (5) studies in which RGD PET or PET/CT was performed at baseline or during therapy for the prediction of a curative effect; (6) articles with sufficient data to acquire true positive (TP), false positive (FP), false negative (FN), and true negative (TN) rates; (7) articles published in English. The exclusion criteria included the following: (1) duplicate studies; (2) letters, case reports, reviews, comments, and meeting abstracts; (3) laboratory studies, animal studies, or studies unrelated to the diagnosis or prediction of cancer; (4) studies with patients who had been treated before for diagnostic purposes; (5) studies that were not related to the prediction of short-term outcomes due to inconsistent long-term outcomes and too few studies.
. . Data Extraction. The following data were extracted from the studies: basic information of the studies (names of the first authors, country of origin, year of publication, and study design), population characteristics (number of subjects or lesions, sex distribution, and age distribution), technical aspects (imaging methods, parameters, or PET technique), the effect index (the TP, FP, FN, and TN rates for the PET imaging), and methods of tumor determination or response criteria.
. . Quality Assessment of Included Studies. Two reviewers independently assessed the quality of the included studies. Studies testing the diagnostic value of RDG PET or PET/CT were assessed by QUADAS-2. The scale consists of the following four domains: patient selection, index test, reference standard and flow, and timing. Each section includes assessment of the risk of bias ("low", "high", or "unclear") and the applicability of diagnostic accuracy [29]. Studies investigating the predictive value of RDG PET or PET/CT were evaluated using the Newcastle-Ottawa Scale (NOS). The scale involves the following three items: subject selection criteria, comparability of subjects, and outcomes. Highquality articles have total scores of more than 5 points, with a maximum total score of 9 points [30].
. . Statistical Analysis. Pooled estimates of the sensitivity, specificity, and DOR with 95% confidence intervals (CIs) were determined based on bivariate analysis of patients or lesions. We evaluated heterogeneity among the studies through the likelihood ratio I 2 index. We assigned categories of low, moderate, and high heterogeneity to I 2 values of 25%, 50%, and 75% [31]. The DerSimonian-Laird method (random effect model) was applied for the meta-analysis if heterogeneity existed; otherwise, the Mantel-Haenszel method (fixed effects model) was employed. Summary receiver operating characteristic (SROC) curves with the AUC and the Q * estimate were obtained. We analyzed the diagnostic abilities of the four parameters and heterogeneity by metaregression. Publication biases were assessed by Deek's funnel plots. The main analyses were performed in Meta-Disc 1.4.

Results
. . Search Results. Using a previously established retrieval strategy, we initially identified 1310 relevant papers from the databases, as follows: 461 in EMBASE, 299 in PubMed, 3 in the Cochrane Library, and 547 in Web of Science. In total, 1216 manuscripts were excluded for the following reasons: 346  were duplicates; 269 were reviews, case reports, or meeting abstracts; and 601 were about basic experiments or animal experiments. The remaining 94 articles were subjected to further full-text assessment. After careful reading, 85 of the papers were excluded for the following reasons: 73 were unrelated to the diagnostic or predictive value of RGD PET/CT; 10 lacked sufficient data to acquire or calculate TP, FP, FN, and TN rates; and 2 had endpoints that were not short-term responses. Ultimately, 9 studies were included. Six articles with 9 sets of data and a total of 216 patients were eligible for inclusion to assess the diagnostic value of RGD PET/CT, and 3 articles with 75 patients were suitable for evaluating the predictive value of RGD PET/CT ( Figure 1) [8,[26][27][28][32][33][34][35].
. . Characteristics of the Included Studies and Quality Assessment. The main characteristics of the studies are shown in Table 1. Three of the selected articles each diagnosed both the primary tumor and the lymph node metastasis status of patients with lung cancer, one of which also examined the diagnostic potential of 68 Ga-Alfatide II PET/CT in differentiating between NSCLC and tuberculosis patients. Six studies used different RGD radioligands. We selected the number of lesions or metastases for analysis rather than the number of patients when both were counted in the studies. Several parameters (SUV max , SUV mean , tumorto-normal tissue ratios, and visual analysis) were used as diagnostic parameters. Eight sets of data were selected for SUV max , five sets of data for SUV mean , three sets of data for visual analysis, and only one set of data for tumor/blood ratios (T/N). Tumor lesions or metastases were all finally diagnosed by histopathology. The patient-based diagnostic parameters of RGD PET/CT for the studies are shown in Table 2. The diagnostic sensitivity ranged from 75.00% to 100%, and the diagnostic specificity ranged from 44.44% to 100%.
The main characteristics of the three studies that investigated the prediction of short-term effects are presented in Table 3. These studies enrolled patients with locally advanced rectal cancer, GBM after surgical resection, and advanced NSCLC. The parameter was SUVmax before CCRT. Xiaohui Luan et al. investigated the predictive value of multiple parameters, including SUVpeak and T/N. Overall, shortterm response criteria differed according to tumor type. The predictive parameters of RGD PET/CT from the studies are listed in Table 4. The prediction sensitivity ranged from 73% to 100%, and the prediction specificity ranged from 66.7% to 88.9%.
The quality of the studies that investigated diagnostic potential was assessed by QUADAS-2, and these studies were determined to have a low risk of bias ( Figure 2). The quality of the studies that investigated the prediction of short-term outcomes was assessed by the NOS, with scores ranging from 7 to 8; this result indicates that the studies were of high quality (Table 5). There was no overlap among the cohorts enrolled in the articles. In addition, all the included studies were prospective.
. . Heterogeneity and reshold Effect Assessment. Heterogeneity among the studies was determined using the Chisquare test. There was clear heterogeneity in the sensitivity and specificity of the parameters for diagnosis, but there was no significant heterogeneity for the predictive value. Furthermore, there was no threshold effect among the studies (Table 6). Thus, the DerSimonian-Laird method (random effects model) was adopted to estimate pooled data for the diagnostic value, and the Mantel-Haenszel method (fixed effects model) was used for the predictive value.
. . Diagnostic Performance and Heterogeneity Analysis. The pooled sensitivity, specificity, DOR, and SROC for SUVmax           . . Subgroup Analysis of Diagnostic Performance. To obtain further insight, we analyzed heterogeneity with regard to lesion location, RGD radioligand, and tumor type for SUVmax and for lesion location and RGD radioligand for SUVmean. However, no factors were found to clearly affect heterogeneity (Table 7).
. . Publication Bias. Due to the small number of studies that investigated other parameters, only studies involving SUVmax and SUVmean were tested for publication bias.
Deek's funnel plots indicated no significant publication bias (P=0.474 for SUVmax and P=0.603 for SUVmean). We did not analyze publication bias for predictive studies because the number of included studies was relatively small.

Discussion
This is the first meta-analysis to evaluate the diagnostic performance of RGD PET/CT in the detection of underlying malignancy and to assess the ability of RGD PET/CT to predict the short-term effects of treatment. The tumor diagnostic ability for SUVmax was slightly superior to that of SUVmean, though there was no significant difference. There was considerable heterogeneity in SUVmax and SUVmean in the literature, and we have not yet reached a consensus regarding the factors influencing the diagnostic value of RGD PET/CT because of the limited number of studies included in the meta-analysis. Overall, the different receptor binding affinities and receptor-binding kinetics of different radiotracers may be an important factor. The inclusion of different tumor types in this meta-analysis may be another important reason for the observed inter-tumor heterogeneity; v 3 expression level and tumor angiogenesis vary among tumor types. Meta-analysis of the use of the same RGD PET for the diagnosis of the same type of tumor is more persuasive [36].
As the receptor binding affinity and tumor retention of RGD radioligands improve, multiple types of RGD peptides are being used in clinical studies; 18 F-Galacto-RGD and 18 F-Fluciclatide were the first two such compounds. These two compounds have similar distribution characteristics, including heterogeneous tumor uptake in monitoring sensitivity for primary and metastatic foci. The sensitivity for evaluating primary lesions ranges from 83% to 100%, though the sensitivity of 18 F-Galacto-RGD is only 33-54% for metastatic lymph nodes (LNs) and 46-78% for distant metastases [9-11, 13, 14, 16, 37]. Although the sensitivity of 18 F-Fluciclatide was relatively better, reaching 88-94% for all lesions and 71-88% for metastases [15,16], the moderate level of sensitivity for metastases is insufficient for tumor staging. As a result, to enhance the performance of integrin imaging, multimeric RGD peptides with increased receptor binding affinity and tumor retention have been used in clinical studies [38,39]. The results showed that 18 F-Alfatide PET/CT and 18 F-FPPRGD2 PET/CT are able to clearly identify all primary lesions (100% sensitivity) [18,19]; however, fewer primary lesions (83.8%) can be distinguished with 68 Ga-NOTA-PRGD2 PET/CT [17]. We comprehensively compared the diagnostic ability of 18 F-Alfatide II RGD PET/CT and 68 Ga-NOTA-PRGD2 PET/CT using data from studies by Fei Kang and Kun Zheng. Although the detection rates for primary lesions and metastatic LNs were similar (76.92% versus 75% for primary lesions and metastatic LNs, respectively, in the study by Fei Kang; 80.88% versus 77.14% in the study by Kun Zheng), the specificity for metastatic LNs was slightly inferior to that for primary lesions (92.31% versus 100% in the study by Fei Kang; 82.61% versus 97.58% in the study by Kun Zheng, P>0.05) [33,34]. 18 F-Alfatide II PET/CT has been performed for diagnosing bone metastases [20] and brain metastases [21], and among different types of metastases, 18 F-Alfatide II PET/CT demonstrates excellent diagnostic sensitivity for osteolytic metastases (100%), mixed bone metastases (100%), and mixed bone metastases (98%) but moderate sensitivity for osteoblastic metastases (70%). Regarding brain metastases, all 20 lesions from patients were identified by 18 F-Alfatide II PET/CT. Yue Zhou et al. demonstrated a relatively higher sensitivity of 18 F-Alfatide PET/CT for detecting NSCLC (90.0%) and squamous cell carcinoma (SCC) (100%) than adenocarcinoma (AC) (83.3%). This may be due to the low affinity of 18 F-Alfatide for AC [26]. The much easier radiosynthetic procedure for 18 F-Alfatide and 18 F-Alfatide II compared to others would facilitate large-scale clinical trials [22]. Nonetheless, the diagnostic ability of RGD PET/CT still requires larger sample-size clinical trials for validation.
Overall, 18 F-FDG PET/CT cannot completely replace invasive staging methods because of its relatively low specificity and high uptake by inflammatory LNs [40]. Compared with the 30.21% positive predictive value of 18 F-FDG PET/CT, 68 Ga-NOTA-PRGD2 PET/CT had a value of 90% [34]. FDG, as an analog of glucose, is transported into cells by glucose transporters (GLUTs) and phosphorylated by hexokinase. Similar to malignant cells, inflammatory cells exhibit increased expression of GLUT and increased affinity of GLUT toward deoxyglucose, leading to high uptake of FDG but not necessarily an increase in integrin v 3 [34,41,42]. 68 Ga-NOTA-PRGD2 PET/CT may be complementary to 18 F-FDG PET/CT because of its lower sensitivity but higher specificity in the diagnosis of tumors. However, in the study by Song Gao et al., uptake by benign lesions in nine patients was heterogeneous [8]. Jiang Wu et al. also found lower specificity for 18 F-Alfatide II PET/CT (54.5% for SUVmax, 63.6% for SUVmean) than 18 F-FDG (81.8% for SUVmax, 81.8% for SUVmean) in differentiating between breast cancer and benign breast lesions [43]. Thus, it remains difficult to clearly distinguish inflammatory pseudotumors from malignant lesions. Angiogenesis is an early pathological change in chronic inflammation, which may be the reason for the observed low specificity [8,44], and neovascularization may vary depending on different stages of inflammation. Although we demonstrated an excellent specificity for tumor diagnosis in our meta-analysis, we cannot rule out the possibility of a lower specificity when using RGD PET/CT, especially in the early pathological stage of chronic inflammation.
In our meta-analysis, SUVmax from RGD PET/CT before CCRT may be able to predict short-term outcomes of treatment. In a study by Xiaohui Luan et al. addressing the ability of 18 F-alfatide PET/CT to identify responders, the AUC of T/N in the lung (AUC=0.944) at baseline was higher than that of SUVmax, SUVpeak, T/N in the blood, and T/N in the muscle (AUC=0.815, 0.864, 0.889, and 0.901, respectively). In another study, SUVmax (AUC=0.737) before treatment and SUVmax (AUC=0.846), T/N (AUC=0.785) during treatment were also able to predict short-term outcomes, and SUVmax during treatment had superior predictive value compared to the volumetric parameters (AUC=0.786) of MRI. As a vital process in the growth and progression of tumors, reversal of angiogenesis may occur earlier than tumor cell death; however, FDG PET cannot detect it earlier. These findings may explain why RGDPET/CT can predict tumor treatment response [28,45,46], though neither Nadia Withofs et al. nor Xiaohui Luan et al. were able to identify responders using 18 F-FDG PET/CT [27,35,46].
In mouse experiments, tumors with medium and high uptake of 99 m-Tc-3P-RGD2 SPECT/CT before treatment responded well, with a greater degree of tumor response compared to tumors with low uptake levels [47]. In a pilot study with only four patients treated with bevacizumab-containing drugs (one patient with disease progression, one with a partial clinical response, and two with complete response), the patients exhibited different degrees of SUVmean decreases using 18 F-FPPRGD2 PET/CT (1.6%, 7.9%, 25.2%, and 25.0%, respectively) [48]. In a study by Andrei Iagaru, patients with a 59.8% decrease in 18 F-FPPRGD2 uptake had no recurrent GBM, whereas patients with a 4.8% decrease had recurrent disease [19]. RGD PET/CT may have potential for early prediction of the response to antiangiogenesis therapy, though these preliminary findings should be confirmed in larger studies.
There are limitations in this meta-analysis. First, expression of v 3 varies in different types of tumors, which may affect the capability of RGD PET diagnostic and prediction. Second, the receptor binding affinity and receptor-binding kinetics of different radiotracers vary, which may also affect results. These limitations are all due to the limited number of studies. Therefore, more research with large sample sizes is urgently needed.

Conclusion
The interesting but preliminary data of this meta-analysis demonstrate that RGD PET/CT may be an ideal diagnostic method for detecting underlying malignancies in patients suspected of having tumors and may be able to predict short-term treatment outcomes. It is necessary to conduct large-scale clinical trials for RGD PET/CT to further study its diagnostic ability and predictive value for short-term outcomes.

Ethical Approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Consent
Informed consent does not apply as this is a review manuscript.

Conflicts of Interest
The authors declare no conflicts of interest.