The Accuracy of MRI for Esophageal Cancer Staging

HomeRadiologyVol. 299, No. 3 PreviousNext Reviews and CommentaryFree AccessEditorialThe Accuracy of MRI for Esophageal Cancer StagingMariska M. G. Leeflang Mariska M. G. Leeflang Author AffiliationsFrom the Department of Epidemiology and Data Science, Amsterdam Public Health, Amsterdam University Medical Centers, University of Amsterdam, Meibergdreef 9, PO Box 22700, 1105 AZ Amsterdam, the Netherlands.Address correspondence to the author (e-mail: [email protected]).Mariska M. G. Leeflang Published Online:Mar 30 2021https://doi.org/10.1148/radiol.2021210374MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In See also the article by Lee et al in this issue.Dr Leeflang is an associate professor in clinical epidemiology in the Department of Epidemiology and Data Science at the Amsterdam University Medical Centers. Her research interests focus on methods development for the evaluation of medical tests and diagnostic test accuracy systematic reviews. She is also co-convenor of the Cochrane Screening and Diagnostic Test Methods Group.Download as PowerPointOpen in Image Viewer Staging of esophageal cancer is important because it informs the prognosis of the patient as well as decisions about the appropriate treatment. The earliest tumor stage is high-grade dysplasia, and it is denoted as stage T0. It then ranges from tumor stage 1 (T1) to T4, with declining survival rates (1). After the patient has been treated with neoadjuvant chemotherapy, monitoring of the patient may detect early recurrence or aggravation of the tumor. Staging may be performed with a combination of imaging modalities, although each of those modalities (PET, CT, endoscopic US, and endoscopic US–guided biopsy) may have adverse effects and may have limited sensitivity (2–4). Because of the recent developments in MRI technology, Lee and colleagues (5) synthesized the sensitivity and specificity of MRI for the staged diagnosis of esophageal cancer.The systematic review addresses the following: (a) differentiation between tumor stage 0 (T0) and stage T1 or higher, (b) differentiation between tumor stage up to stage T2 and T3 or higher, and (c) differentiation between nodal stage 0 (N0, no spread to lymph nodes) and stage N1 or higher. The authors retrieved 20 studies of the diagnostic accuracy of MRI for determining the stage of esophageal carcinoma before esophagectomy and pathologic staging. The quality of the included studies was assessed using a standard checklist for the Quality Assessment of Diagnostic Accuracy Studies (version 2), and the meta-analysis they performed took into account the bivariate nature of the data at hand.Eleven studies analyzed the sensitivity and specificity of MRI to differentiate between stage T0 and stage T1 or higher. In two of those, the participants without esophageal cancer were healthy volunteers. This may have caused both a high risk of bias and concerns regarding applicability, as healthy volunteers are usually not among the patient population who would undergo MRI. The results were variable, especially for specificity. This may have been caused by the low numbers of participants without esophageal cancer, leading to chance variation and broad CIs. The overall sensitivity was 92% (95% CI: 82, 96), and the overall specificity was 67% (95% CI: 51, 81).The median prevalence among the studies without healthy control participants was 80% (calculated from Table E3 [online]), which means that of every 100 patients assessed for stage T0 disease versus stage T1 disease or higher, 80 persons had stage T1 disease or higher. One could hypothesize a population of 100 patients who were all examined for esophageal cancer with MRI. Of the 80 patients who in reality had stage T1 disease or higher, 74 (92%) were also found to have stage T1 disease or higher according to their MRI results. Six of the 80 patients with stage T1 disease or higher were indicated to have stage T0 disease according to their MRI results. Of the 20 patients with stage T0 disease, 13 will be told that they are in stage T0, whereas seven will be told that they have at least stage T1 disease. In summary, this means that 81 patients will be told that they have stage T1 disease or higher (positive MRI results), whereas seven of these will actually have a false-positive result. Whether this is a worrying result depends on the next steps to be taken. It also means that 19 patients will be told that they have stage T0 disease, of whom six will actually have a higher stage (and thus have false-negative results). Here as well, the next steps determine whether this number is acceptable or not. These numbers may of course vary depending on the rounding factors and CIs.Unfortunately, the review does not explain what these next steps are, and whether the 8% of the patients with higher disease stages will be understaged is problematic. Also, Lee et al (5) claim that the summary estimates for sensitivity and specificity are similar in the overall set of studies and in those focusing on the use of MRI after neoadjuvant treatment. Although the CIs overlap and sensitivity indeed seems to be similar, the summary estimate for specificity is considerably lower in the treated population (55%, 95% CI: 37, 71) than in the population before treatment (70%, 95% CI: 46, 86). The lower specificity may be a problem in monitoring because specificity declines with multiple testing rounds. If a patient has positive test results (above a certain threshold), then the patient will be referred for further treatment, but the patient will not be set back to having a negative result. If a patient has negative test results, then this patient will be tested again next time, with another chance to receive a potentially false-positive result. Hence, monitoring may improve sensitivity, but it will also increase the risk of false-positive findings.The second question was about the distinction between stage T2 disease or lower and stage T3 disease or higher. The overall sensitivity for the 10 included studies was 86% (95% CI: 76, 92), and the specificity was 86% (95% CI: 75, 93). The median prevalence among the studies was 46% (calculated from Table E3 [online]), which means that of every 100 patients undergoing MRI, 46 persons had stage T3 disease or higher. Of these, about six patients will be regarded to have stage T2 disease or lower (false-negative findings), according to the MRI results. Of the 54 patients with stage T2 disease or lower, eight will be considered to have stage T3 or higher disease (false-positive findings). Again, whether these numbers are problematic depends on the consequences of being staged as stage T2 or lower or stage T3 or higher.Ten included studies addressed the third question about nodal involvement. The more regional lymph nodes were involved, the higher the N staging. The overall sensitivity for detecting patients with at least one regional lymph node involved was 71% (95% CI: 60, 80), and the specificity was 72% (95% CI: 64, 79). The median prevalence among the studies was 48% (calculated from Table E3 [online]), which means that of every 100 patients undergoing MRI, a median of 48 persons had at least one regional lymph node affected.The authors conclude that MRI has a sensitivity superior to that of CT and PET. Although this indeed seems to be the case, this conclusion is based on an indirect comparison. In the ideal world, one would include patients suspected of having esophageal cancer and have them all undergo CT, PET, and MRI to draw firm conclusions about the relative sensitivity between the three modalities. A systematic review addressing such a comparative question should ideally limit the inclusion to studies directly comparing two or more different modalities against the reference standard. However, in practice, this is difficult if not impossible. It may be too burdensome or too costly to apply all relevant modalities to all patients, and the patient groups may be too small to randomize between modalities. Still, this does not justify indirect comparisons without any caveats. It is possible that the studies evaluating PET and CT included different patients so that the difference between the modalities is not because of the techniques, but because of the included patients.In conclusion, MRI has a high sensitivity and relatively low specificity for helping diagnose esophageal cancer. It remains to be seen whether the sensitivity and specificity are sufficiently high to replace other modalities. This will depend on the consequences of false-negative findings or false-positive findings in patients. It will also depend on the true difference in accuracy between the modalities, in addition to cost, burden, and other factors that may influence the use of a test in practice.Disclosure of Conflicts of Interest: M.M.G.L. disclosed no relevant relationships.References1. Berry MF. Esophageal cancer: staging system and guidelines for staging and treatment. J Thorac Dis 2014;6(Suppl 3):S289–S297. Medline, Google Scholar2. Leong TL, Loveland PM, Gorelik A, Irving L, Steinfort DP. Preoperative Staging by EBUS in cN0/N1 Lung Cancer: Systematic Review and Meta-Analysis. J Bronchology Interv Pulmonol 2019;26(3):155–165. Crossref, Medline, Google Scholar3. Sun F, Chen T, Han J, Ye P, Hu J. Staging accuracy of endoscopic ultrasound for esophageal cancer after neoadjuvant chemotherapy: a meta-analysis and systematic review. Dis Esophagus 2015;28(8):757–771. Crossref, Medline, Google Scholar4. Crabtree TD, Kosinski AS, Puri V, et al. Evaluation of the reliability of clinical staging of T2 N0 esophageal cancer: a review of the Society of Thoracic Surgeons database. Ann Thorac Surg 2013;96(2):382–390. Crossref, Medline, Google Scholar5. Lee SL, Yadav P, Starekova J, et al. Diagnostic Performance of MRI for Esophageal Carcinoma: A Systematic Review and Meta-Analysis. Radiology 2021. 10.1148/radiol.2021202857. Published online March 30, 2021. Google ScholarArticle HistoryReceived: Feb 9 2021Revision requested: Feb 17 2021Revision received: Feb 18 2021Accepted: Feb 22 2021Published online: Mar 30 2021Published in print: June 2021 FiguresReferencesRelatedDetailsAccompanying This ArticleDiagnostic Performance of MRI for Esophageal Carcinoma: A Systematic Review and Meta-AnalysisMar 30 2021RadiologyRecommended Articles Diagnostic Performance of MRI for Esophageal Carcinoma: A Systematic Review and Meta-AnalysisRadiology2021Volume: 299Issue: 3pp. 583-594Advanced-Stage Nasopharyngeal Carcinoma: Restaging System after Neoadjuvant Chemotherapy on the Basis of MR Imaging Determines SurvivalRadiology2016Volume: 282Issue: 1pp. 171-181Application of N Descriptors Proposed by the International Association for the Study of Lung Cancer in Clinical StagingRadiology2021Volume: 300Issue: 2pp. 450-457Breast MRI during Neoadjuvant Chemotherapy: Lack of Background Parenchymal Enhancement Suppression and Inferior Treatment ResponseRadiology2021Volume: 301Issue: 2pp. 295-308Effect of Longitudinal Variation in Tumor Volume Estimation for MRI-guided Personalization of Breast Cancer Neoadjuvant TreatmentRadiology: Imaging Cancer2023Volume: 5Issue: 4See More RSNA Education Exhibits High-Resolution MRI Pretherapeutic Evaluation for Esophageal Cancer? What the Radiologists Need to Know?Digital Posters2020How To Interpret Esophageal Cancer MR For Practical Use In Oncologic Decision MakingDigital Posters2021Gastroenteropancreatic Neuroendocrine Tumors (GEP-NETs): Case-based Pictorial Review Of The AJCC 8th Edition Staging SystemDigital Posters2021 RSNA Case Collection Locally advanced, metastatic prostate adenocarcinomaRSNA Case Collection2020Recurrent breast cancer on CEUSRSNA Case Collection2021Synovial sarcoma of neckRSNA Case Collection2020 Vol. 299, No. 3 Metrics Altmetric Score PDF download