Diagnostic Performance of Noninvasive Imaging Modalities for Localization of Insulinoma: A Meta-Analysis

Insulinomas is the most common functional neuroendocrine tumor found only in the pancreas. The early detection of insulinoma is of importance. Studies comparing the performance of noninvasive modalities were limited by sample size and heterogeneity between studies. The aim of this meta-analysis was to evaluate the diagnostic performance of PET/CT, SPECT/CT, CT and MRI for the localization of insulinoma, and to provide evidence for clinical practice. Methods


Background
Insulinomas is also called insulin β cell tumor, which is a relatively rare tumor found only in the pancreatic tissue [1,2].
The incidence of sporadic insulinoma is 1-4 cases/per million/per year [3,4]. It is a major cause of endogenous hyperinsulinemic hypoglycemia (EHH) in nondiabetic individuals [5]. Approximately 90% of these tumors are solitary and benign, 10% of them are malignant [6]. In the presence of metastases, insulinoma is often considered malignant [7], accurate diagnosis is essential due to the severe hypoglycemia that may lead to life-threatening consequences [8].
Previous studies with small sample-sizes have investigated the e cacy of PET/CT and SPECT/CT over conventional imaging techniques for insulinoma localization. The results of these studies were heterogeneous. Therefore, the purpose of this study was to conduct a meta-analysis by synthesizing the published evidence to generate an accurate comparison of the diagnostic performance of PET/CT, SPECT/CT, CT and MRI for the localization of insulinoma, and secondly to provide evidence and hints for clinical decision-making and implement. articles. If studies recruited participants over the same period of time or from the same study center, only the researches with the most pertinent outcomes was included to avoid duplications. The types of research included prevalence surveys, case-control studies, longitudinal studies and randomized controlled trials (RCTs). Case reports, conference abstracts, and successive cases seen in a unit were excluded. Two independent investigators conducted the process of literature search and study inclusion. When disagreement occurred, they discussed their arguments, and a third reviewer was involved while no consensus was achieved.

Data Extraction And Quality Assessments
Data were extracted by two independent investigators from each included article. Name of the rst author, year of publication, total number of participants, duration of insulinoma, age, gender of patients, type of insulinoma, absolute numbers of participants with TP, FP, TN, and FN outcomes were collected. The QUADAS-2, a revised tool for the quality assessment of diagnostic accuracy studies, was used for the assessment of methodological quality, risk of bias and applicability concerns of the included studies [33,34].

Statistical analysis
The Stata 16.0 software and Review Manager 5.3 software were utilized for statistical analysis in this study. A P value less than 0.05 was considered to be statistically signi cant. We calculated pooled sensitivity, speci city, positive Likelihood Ratio (+ LR) and negative Likelihood Ratio (-LR), diagnostic odds ratio (DOR), concordance rate (number of patients detected by imaging modalities compared reference standard) and the 95% con dence intervals (CIs) using maximum likelihood estimates. Summarized receiver operating characteristic (SROC) curves were used for comparison of diagnostic performance of different modalities. The Cochran Q and the I 2 statistics were used to assessed the heterogeneity of results between studies included. I 2 values in the ranges of 0-25%, > 25-50%, > 50-75%, and > 75% reveal insigni cant, low, medium, and high heterogeneity, respectively [35]. Moreover, funnel plots were created to assess publication and related bias. We used a Deeks' method to statistically examine the asymmetry of the funnel plot so as to detect potential publication bias.

Study selection and characteristics
A total of 482 articles were identi ed from the databases searched. One hundred and seventeen duplicates were removed and 264 studies were excluded through an initial screening. After a full text assessment for eligibility of the remaining 101 articles, 19 studies were identi ed for inclusion in this meta-analysis. No additional studies were found through reference screening of the included papers. Figure 1 shows the ow of the literature search and study selection process. The included 19 studies containing a total of 708 patients with diagnosed or suspected insulinoma, provided relevant detection results of PET/CT, SPECT/CT, CT and MRI in those participants. These articles were published from 1996 to 2020. Table 1 showed details of the studies included.

Quality Of Studies
Quality assessment by QUADAS-2 scale showed that 13 studies had low risk of bias for patient selection, 1 study had high risk of bias and 5 studies had unclear risk. 13 studies had low risk of bias for index test, risk of bias for others were unclear. Twelve studies had low risk of bias for reference standard, 1 study had high risk of bias, and 6 had unclear risk of bias. In the part of ow and timing, 3 studies had unclear risk of bias, 2 had high risk and 14 had low risk. Clinical applicability concerns of each study included were also evaluated (Fig. 2).  (Fig. 4).

Subgroup Analysis Of Molecular Imaging Modalities
The  Table 2).

Discussion
The application of cancer imaging has been involved in the entire process of the management of patients with con rmed insulinoma for decades [36]. Different imaging modalities (MRI, CT and ultrasonography) have been used for insulinoma localization, but they showed low accuracy and sensitivity in previous studies [37][38][39], the main reason may be that the size of insulinomas are evenly distributed throughout the pancreas and less than 2 cm in size in most cases [40]. With the development of SPECT and PET imaging, this issue can be addressed owing to improved spatial resolution and more accurate quanti cation in the localization of insulinoma [25,41]. In this meta-analysis, we evaluated the diagnostic performance of PET/CT, SPECT/CT, CT and MRI for the localization of insulinoma, which were not researched in previous meta-analyses. In addition, we compared the diagnostic performance of GLP-1R based PET/CT and SSTR based PET/CT imaging modalities in this meta-analysis. These are the We had planned to conduct a meta-analysis based on individual data, unfortunately, these data were not easy to obtain, thus all statistical analyses were conducted on study level. The number of studies on PET or SPECT were small according our preliminary research, so we included studies on PET/CT, SPECT/CT, CT and MRI which are used in the detection of insulinoma. Nineteen studies with a total of 708 patients with diagnosed insulinoma were enrolled in the metaanalysis. Results of pooled analyses showed that PET/CT exhibited higher pooled sensitivity over conventional noninvasive imaging techniques (CT and MRI). The AUC value of PET/CT imaging superior to other modalities.
PET/CT showed the highest concordance rate compared to gold standard. GLP-1R based PET/CT manifested better diagnostic performance in comparison with SSTR based PET/CT imaging modality.
In this meta-analysis, we did a detailed literature search in both Chinese and English language to enhance the probability to retrieve all relevant studies as we can. Data extraction was conducted by two independent investigators using a pre-designed form. Besides, we assessed the heterogeneity between studies included, quality of each study along with publication bias. Heterogeneity for between studies included was detected, the results of studies can be consolidated with caution. The potential source of heterogeneity between studies may be the demographic characteristic of participants, study design, duration of insulinoma, type of radiotracers, and type of imaging equipment. We intended to conduct the meta-regression to statistically investigate the reason for the heterogeneity between studies included, unfortunately, the numbers of studies in each subgroup were inadequate for metaregression due to the inclusion criteria of this study. It is hoped that with the increased of the number of studies or the adjustment of inclusion criteria, this analysis can be carried out in the future. Furthermore, Deeks' funnel plot asymmetry tests indicated that publication bias may not affect the pooled results between studies.
Based on the outcomes of this meta-analysis, we may conclude that GLP-1R based PET/CT imaging for localization of insulinoma demonstrates favorable imaging results and diagnostic accuracy. Availability of data and materials The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.

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

Funding
This work was independently supported by grants from the National Natural Science Foundation of China to Jianbing Zhu (No.6192780137). The funding body had no in uences on the design of the study and collection, data analysis, interpretation of results, and writing of the manuscript. Methodological quality assessment of studies included