European Review for Medical and Pharmacological Sciences 2020; 24: 7240-7252 Whole-body magnetic resonance imaging is superior to skeletal scintigraphy for the detection of bone metastatic tumors: a meta-analysis G. SUN1, Y.-X. ZHANG2, F. LIU3, N. TU4 1Department of Radiology, Zaozhuang Hospital of Traditional Chinese Medicine, Zaozhuang, Shandong, China 2No. 2 Department of Orthopedics, Zaozhuang Hospital of Traditional Chinese Medicine, Zaozhuang, Shandong, China 3Zaozhuang Hospital of Traditional Chinese Medicine, Zaozhuang, Shandong, China 4Department of Magnetic Resonance, Zaozhuang Municipal Hospital, Zaozhuang, Shandong, China Gang Sun and Yongxiang Zhang contribute equally to this work Abstract. – OBJECTIVE: The meta-analysis Introduction aims to compare the diagnostic performance of whole-body magnetic resonance imaging (MRI) Any type of metastatic cancer that spreads via and skeletal scintigraphy (SS) for the detection the bloodstream can infiltrate the bone marrow of skeletal metastases. 1,2 MATERIALS AND METHODS: We searched and give rise to bone metastases . Certain types Medline, Scopus, Embase and Cochrane library of cancers, such as prostate or breast cancer, are databases for identifying fifteen eligible stud- known for their ability to cause skeletal metas- ies with a total of 1939 participants, and the tases, with a prevalence of up to 70%3,4. There is quality of these studies was assessed accord- high variability in the metabolic activity of dif- ing to Quality Assessment of Diagnostic Accu- ferent types of bone metastases. It is therefore racy Studies (QUADAS-2) guidelines. Sensitivi- important to choose the suitable diagnostic meth- ties, specificities, diagnostic odds ratios (DOR), positive likelihood ratios (PLR), and negative od among a variety of radiological and nuclear likelihood ratios (NLR) were calculated. Sum- medical imaging techniques for the detection of mary receiver operating characteristic curves bone metastases in patients with certain types of (sROC) were generated using bivariate models cancers. for whole-body MRI and skeletal scintigraphy. Skeletal scintigraphy (SS) with labeled phos- RESULTS: Whole-body MRI had higher but phonates (99MTC-phosphonates) is routinely used comparable patient-based higher specificity for the detection of local bone metabolism in an compared to SS (99% vs. 95%). However, it had 5 markedly higher sensitivity (94% vs. 80% re- early phase of some types of cancers . Therefore, spectively), DOR (966 vs. 82), and LPR (54.4 vs SS is most effective in the visualization of me- 17.1). LNR of whole-body MRI was <0.1 (0.06), tastases that are associated with reactive hyper- while LNR of SS was >0.1 (0.22). The area under metabolism of bone. This includes metastases of curves (AUC) for whole-body MRI and SS were prostate and breast cancer, neuroendocrine tu- 0.99 and 0.95 respectively. mors, and osteosarcomas5-8. However, SS is rel- CONCLUSIONS: We demonstrate that both whole-body MRI and SS have good diagnostic atively insensitive for tumors that are not hyper- performance. However, MRI is superior for di- metabolic and can lead to false-positive results in agnostics of bone metastases, as it has higher cases of post-treatment bone matrix regeneration sensitivity, higher diagnostic accuracy, and can (flare phenomenon)9. be used for both confirmation and exclusion of Magnetic resonance imaging (MRI) is quickly metastatic bone disease. becoming a method of choice for detecting bone Key Words: metastases due to its high soft-tissue contrast, Magnetic resonance imaging, Skeletal scintigraphy, high spatial resolution, and no requirements for Bone metastatic tumors, Meta-analysis. intravenous contrast medium10. In the 2011 me- 7240 Corresponding Author: Nan Tu, MD; e-mail: [email protected] Whole-body MRI and detection of bone metastatic tumors: a meta-analysis ta-analysis, Yang et al11 showed that MRI had 91% tigraphy”, “Skeletal Scintigraphy”, “Magnetic sensitivity and 95% specificity, being superior to Resonance Imaging”, “Sensitivity”, “Specificity”, planar skeletal scintigraphy. These findings were “Diagnosis”, and “Diagnostic Accuracy Studies”. further confirmed in subsequent studies involv- Similar terms were also used in Cochrane library, ing, among others, breast, prostate, lung and renal Scopus, Embase for literature search of published cancers13-23. However, these studies have certain studies. The time period was from inception to limitations, such as a lack of pathological verifica- December 2019 without any language restrictions. tion of skeletal metastases, selection bias, and not Bibliographies of retrieved studies were searched, taking into consideration an improved SS with and eligible studies were included. the use of more advanced single-photon-emission computerized tomography/combined with CT Study Selection (SPECT/SPECT-CT) apparatus17,24,25. Two authors independently screened the ti- The purpose of this meta-analysis is to com- tle and abstract of the records identified during pare the efficiency of whole-body MRI and skele- the literature search. The full-text article was re- tal scintigraphy in the detection and characteriza- trieved for studies deemed relevant. The further tion of skeletal metastases. full-text screening was again done by the two au- thors independently and those studies matching the inclusion criteria were finally included in our Materials and Methods review. Disagreements between the two authors during this process were solved via consultation Eligibility Criteria with a third investigator. The inclusion criteria were as follows: studies comparing the diagnostic performance of skele- Data Extraction Process tal scintigraphy and whole-body MRI irrespec- Data extraction for the required characteristics tive of study design employed; studies that report from the included studies was done by the primary required statistics of the above-mentioned tech- investigator. The data extracted were study design, niques or provide data to calculate these rates; setting, index test, reference standards (gold stan- full-text studies or published as conference ab- dard/comparator), comorbidities, the total number stracts were included while unpublished data, of participants, average age, inclusion, exclusion case reports, and studies with smaller sample size criteria, sensitivity, and specificity. The extracted (fixed at 10 for the current review) were excluded. data were transferred into STATA software. Participants: patients with a primary malignant tumor in sites other than skeletal sites. Risk of Bias Assessment In Included Index test: studies that used skeletal scintigra- Studies phy and whole-body MRI for the identification of The risk of bias was assessed by two authors bone metastasis. independently using the Quality Assessment of Reference standards: studies where the diag- Diagnostic Accuracy Studies-2 (QUADAS-2) nostic accuracy is compared with a definitive di- tool. Domains used for assessing the risk of bias agnosis of bone metastasis by histopathological or were the selection of patients, characteristics of biopsy findings. index test and reference standard, timing, and flow of assessments. The risk of bias was finally Type of Outcome Measure interpreted as low, high or unclear. Pooled sensitivity, specificity, positive like- lihood ratio (PLR) and negative likelihood ratio Statistical Analysis (NLR) and diagnostic odds ratio (DOR) Meta-analysis was done with the selected stud- ies using STATA 14.2 software (StataCorp, Col- Search Strategy lege Station, TX, USA). Bivariate meta-analysis For this meta-analysis, we identified relevant was done to obtain the pooled estimates of the studies by searching the following Medline, Sco- diagnostic accuracy estimates like sensitivity and pus, Embase and Cochrane library databases. The specificity, PLR and NLR, and DOR for MRI and following medical subject headings (MeSH terms) Skeletal Scintigraphy. Summary Receiver Opera- and free-text terms were used in PubMed in sev- tor Characteristic curves (sROC) was made. The eral combinations: “Validation Studies”, “Bone summary estimate obtained in the sROC was the Metastasis”, “Skeletal Metastasis”, “Bone Scin- area under the curve (AUC) with 95% confidence 7241 G. Sun, Y.-X. Zhang, F. Liu, N. Tu interval (CI). AUC value closer to 1 indicates the Results higher diagnostic performance of the imaging techniques. Search Results Forest plot was used to graphically represent We have conducted a systematic search to find the study level and overall pooled diagnostic studies reporting the diagnostic performance of measures. Likelihood ratio (LR) scattergram was MRI and SS for the diagnosis of bone metasta- made to find the clinical significance of these im- ses. Totally, 2834 records found, out of which aging techniques. Fagan plot was constructed to 1189 studies from Medline, 835 from Scopus, demonstrate how much the result of MRI or Skel- 598 from Embase, and 212 from the Cochrane etal Scintigraphy changes the probability that a library (Figure 1). After the removal of dupli- patient has a diagnosis of bone metastases. cates, the remaining 2178 studies were subjected Heterogeneity was evaluated via the following to the title, abstract and keywords screening. At methods: graphical representation through bivari- that stage, 1981 studies were eliminated, due to ate box plot, Chi-square test for heterogeneity and different outcomes (n=1480) and irrelevant diag- I2 statistics to quantify the inconsistency. Potential noses (n=501). Full text of 197 relevant