Diagnostic Performance of MRI, SPECT, and PET in Detecting Renal Cell Carcinoma: A Meta-Analysis


 Background: Renal cell carcinoma (RCC) is one of the most common malignancies worldwide. Noninvasive imaging techniques, such as magnetic resonance imaging (MRI), single photon emission computed tomography (SPECT), and positron emission tomography (PET), have been involved in increasing evolution to detect RCC. This meta-analysis aims to compare to compare the value of MRI, SPECT, and PET in the diagnosis of RCC, and to provide evidence for decision-making in terms of further research and clinical settings.Methods: Electronic databases including PubMed, Web of Science, Embase, and Cochrane Library were systemically searched. Studies concerning MRI, SPECT, and PET for the detection of RCC were included. Pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), with their respective 95% confidence interval (CIs) and the area under the summary receiver operating characteristic (SROC) curve (AUC) were calculated.Results: A total of 44 articles were finally detected for inclusion in this meta-analysis. The pooled sensitivities of MRI, SPECT, and PET were 0.80, 0.81, and 0.88, respectively. Their respective overall specificities were 0.90, 0.54, and 0.87. Results in the subgroup analysis of the performance of SPECT that the pooled sensitivity, specificity, and AUC of the prospective SPECT studies included were 0.80, 0.42, 0.80, respectively. In the analysis of 18F-FDG PET, the pooled sensitivity, specificity, and AUC were 0.88, 0.86, and 0.92, respectively. PET studies showed a pooled sensitivity, specificity, and AUC of 0.80, 0.85, and 0.85, respectively in the diagnosis of primary RCC. The pooled sensitivity, specificity, and AUC of PET studies in detecting recurrent or metastatic RCC were 0.93, 0.88, and 0.94.Conclusion: Our meta-analysis manifests that MRI and PET present better diagnostic value for the detection of RCC in comparison with SPECT. PET is superior in the diagnosis of recurrent or metastatic RCC.


Introduction
Renal cancer is one of the most frequently diagnosed cancers worldwide, which ranks the 6th most frequently con rmed malignant tumor in men and the 8th in women [1]. 90% of all renal malignant tumors tend to be renal cell carcinoma (RCC) [2]. There are three major histological subtypes of renal cell carcinoma: clear cell RCC, papillary RCC, and chromophobe RCC [3]. It is manifested that over one-half of patients with renal cell carcinoma are asymptomatic [4].
Unfortunately, approximately 33-50% of suspected patients are diagnosed with metastatic diseases at the time of initial detection, furthermore, 20-40% of patients with con rmed RCC progress to metastatic diseases even after surgical resection [5,6]. Consequently, timely and accurate detection of the early stage and advanced stage of the disease is of great signi cance.
Biopsy diagnosis is still the gold standard for con rmation of RCC although it is an invasive modality that may result in unnecessary adverse outcomes [7]. Various noninvasive imaging approaches are commonly employed in the detection of RCC [8]. For decades, ultrasound has been used as one of the rst-line modalities for diagnostic imaging of patients with renal lesions due to its cost-effective nature, however, the e cacy of renal imaging is not satisfactory especially in patients with suspected malignancies [8]. Although computed tomography (CT) has been utilized as the con rmative standard for RCC imaging for decades, it manifested poor performance in differentiation among solid masses, fat-poor AML, and oncocytoma [9,10]. Compared to CT, magnetic resonance imaging (MRI) plays an increasingly important role in the diagnosis and restaging of RCC, particularly in patients with unclear results, allergic reactions, pregnancies, as it has no ionizing radiation exposure, superior soft tissue resolution [11,12]. In recent years, targeted imaging approaches have made great progression in the diagnosis of RCC. Single photon emission computed tomography-computed tomography (SPECT) imaging is used to differentiate RCC and detect metastases in renal cancer [13,14]. Furthermore, positron emission tomography (PET) imaging utilizing 18 F-uoro-deoxy-glucose (FDG) and other tracers ( 124 I-girentuximab, 68 Ga-DOTATOC, 11 C-acetate, 18 F-Fluoride) has been studied as diagnostic biomarkers in RCC [15][16][17][18][19][20]. Especially, PET plays an important role in the detection of recurrent or metastatic RCC [20,21].
As far as we are concerned, a large number of studies have assessed the diagnostic performance of non-invasive approaches in terms of RCC, nevertheless, the results are heterogeneous [13,16,[20][21][22][23][24]. This study aimed to generate a more comprehensive comparison of the diagnostic performance of MRI, SPECT, and PET in the detection of RCC by conducting a meta-analysis, and subsequently to guide the diagnosis and differentiation of RCC in the eld of scienti c research and clinical application.

Materials And Methods
The entire process of this meta-analysis was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) [25].

Search strategy and selection criteria
The electronic databases of PubMed, Web of Science, Embase, and Cochrane Library were comprehensively searched with a publication date from inception to January 31, 2021. Articles in the English language were considered. The following key terms were used for the database research: "magnetic resonance imaging", "MRI", "single-photon emission computed tomography", "SPECT", "positron emission tomography", "PET", "renal cell carcinoma". Besides, we manually screened the references of the articles included for more potentially eligible studies. The inclusion criteria of studies were as follows: 1) MRI, SPECT, and/or PET were used for the detection of RCC in patients with suspected or con rmed RCC; 2) a reference standard was utilized to assess diagnostic performance; 3) absolute numbers of patients with true positive (TP), false positive (FP), true negative (TN) and false negative (FN) results can be retrieved in the published articles or recalculated based on other parameters (accuracy rate, sensitivity, speci city, positive predictive value (PPV), negative predictive value (NPV), number of all participants) presented in the manuscripts. In case that the studies were undertaken by the same research group, those with the largest sample size or the most complete information were included to avoid duplicates. Articles were excluded if they were case reports, reviews, letters, news, conference abstracts, animal studies, or studies with insu cient data.
Two independent investigators (QY and HX) conducted the process of literature search and study inclusion. Discrepancies were resolved by discussion. If no consensus was reached, a third author (JN) was involved.
Data extraction and quality assessments Two researchers (QY and YZ) independently performed the title and abstract screening according to the inclusion criteria. A full-text reading of the literature was conducted for the nal inclusion. The following information was extracted from each study: rst author's name, year of publication, study design, type of RCC (primary or recurrent/metastatic), number of patients analyzed, percentage of the male, age of the participants, reference standard, imaging modality and type of radiotracers used in the study, absolute numbers of patients with TP, TN, FP, and FN numbers.
To evaluate the methodological quality of the enrolled studies, we used the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool. This method contains four main components in terms of participant selection, index test, reference standard, as well as ow and timing, all the components are assessed in terms of risk of bias, and the rst three components are also evaluated the concerns of applicability [26].

Statistical analysis
We calculated pooled sensitivity, speci city, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and the 95% con dence intervals (CIs) and the area under the summary receiver operating characteristic (SROC) curve (AUC). A Cochran Q value and the I 2 statistic were used to detect the heterogeneity of studies included. I 2 statistics in the range of 0-25%, 25-50%, 50-75%, and 75-100% were considered to be of insigni cant, low, moderate, and high heterogeneity between studies, respectively [27]. Meta-regression was performed to investigate the possible source of heterogeneity between the included studies. A Deeks' method was introduced to statistically test the asymmetry of the funnel plot and detect publication bias. We conducted sensitivity analysis to evaluate the impacts of one single study on the overall outcomes. All statistical analyses were processed on the study basis using the Stata 15.0 software and Review Manager 5.3 software. A p value <0.05 was considered to be statistically signi cant.

Study selection and characteristics
A total number of 896 articles were identi ed from the online databases. Among them, we excluded 135 duplicates and 640 irrespective studies based on an initial screening of titles and abstracts. After the full text con rmation for eligibility of the remaining 121 articles, 44 articles with 50 studies and 2545 patients were identi ed for nal inclusion in this meta-analysis. No additional studies were found through reference screening of the included papers. Figure 1 shows the ow of the database search and literature selection process. Detailed characteristics of studies included were shown in Table 1. The results of the quality evaluation of the included studies manifested that the high quality of the included studies ( Figure 2).

Diagnostic performance of imaging modalities
We performed the sensitivity analysis to assess the impacts of single study on the overall outcomes. Two SPECT studies and 4 PET studies was identi ed as outliers.   Table 2.

Heterogeneity and publication bias
Deek's tests for publication bias yielded p values of 0.94, 0.69, and 0.02 for MRI, SPECT, and PET, which revealed that there was a possible publication bias in the pooled analysis of PET studies.

Discussion
Renal cell carcinoma is the most commonly diagnosed subtype of kidney cancers and accounts for approximately 2-3 % of all malignancies [19]. The research of Motzer et al. demonstrated that the average 5-year survival rates for patients with RCC decreased with the disease stages (I to IV), from 96-23% [28]. Moreover, the early signs and symptoms of RRC are not speci c which introduces di culties for the early detection of this disease in primary or metastatic sites [29]. Renal biopsy is an accurate method to establish a histological diagnosis for RRC, however, it is may induce a risk of procedural adverse events [30]. Noninvasive approaches namely MRI, SPECT, and PET have been in evolution during the past decades [13,31,32]. Based on various studies of the diagnostic value of noninvasive modalities in the detection of RCC, we carry out a meta-analysis to compare the diagnostic e cacy of these approaches.
The meta-analysis was processed on the basis of study design, type of imaging modalities, type of radiotracers, type of RCC. To our knowledge, some of these dimensions have not been discussed in relevant meta-analyses [33][34][35]. This is one of the strengths of this study. Results revealed that the pooled sensitivity of PET (0.88 [0.77, 0.94]) was the highest. MRI demonstrated the highest overall speci city (0.90 [0.84,0.94]). MRI and PET showed high diagnostic performance in detecting RCC. Results of subgroup analysis manifested that PET imaging had better performance than PET combined with CT. This result was opposite to Ma's study. The possible reasons may be that Ma's study only included studies using 18 F-FDG while studies utilizing 18 F-FDG, 18 F-Fluoride, 124 I-girentuximab, and 11 C-acetate were enrolled [34]. Interestingly, results of the analysis of 18 F-FDG PET demonstrated similar diagnostic parameters to the meta-analysis of Ma et al. [34]. Furthermore, our research indicated that PET imaging revealed better performance in detecting recurrent or metastatic RCC than primary RCC. The underlying reason may be renal excretion of the tracers and low uptake by the primary RCC [20,36,37]. The current study con rmed this nding with a superior level of evidence.
In this meta-analysis, we conducted a detailed literature search to improve the probability of searching as many related studies as possible. Two independent investigators completed the whole process of data extraction using standardized electronic forms. Furthermore, we evaluated the heterogeneity between the studies included. There were signi cant heterogeneities among studies. Distinctions in the year of publication, study methodology, patient characteristics, reference standard, and radiotracers may be the source of heterogeneity. Unfortunately, meta regression was not able to be performed to investigate the likely cause of heterogeneity due to limited number of covariates extracted from the enrolled studies. Subgroup analysis was undertaken to explore the possible source of heterogeneity. For the analysis of PET, the source of heterogeneity may be attributed to the type of radiotracers, type of RCC, and study design. However, not all potential source of heterogeneity was analyzed because of the insu cient number of studies in different subgroups. On account of this limitation, the e cacy of heterogeneity assessment in the study may be biased. Besides, publication bias was detected through the Deeks' funnel plot asymmetry test in the analysis of PET studies. The publication bias may be attributed to the strict exclusion criteria of this meta-analysis. Although there is heterogeneity among studies included and publication bias, the ndings of this analysis may introduce evidence and assistances concerning scienti c research and clinical practice in the detection of RRC. In regard to further research, novel radiotracers with higher uptake ratios between tumor tissues to normal tissues and lower levels of renal excretion need to be further investigated on account of the results of this meta-analysis. In terms of application in the clinical setting, MRI is recommended as the favorable imaging method to help detect RCC due to lack of radiation exposure and high soft-tissue resolution. PET is more suitable for the diagnosis of recurrent or metastatic RCC than primary tumors under the current development of functional imaging modalities. Of note, combined employment of various detection techniques is may be of assistance to increase the overall diagnostic accuracy.

Availability of data and materials
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Authors' Contributions
QY conceived and designed this study. HX and YZ were responsible for the collection, extraction, and analysis of the data. QY was responsible for data analysis and writing the paper. JN and SH performed the quality evaluation of the writing and polished the English language. All authors reviewed the paper and reached an agreement to approve the nal manuscript.  Methodological assessment of studies included on the QUADAS-2 tool.   Forest plot for the detection performance of PET