A Commentary on Intracystic Glucose Levels Appear Useful for Diagnosis of Pancreatic Cystic Lesions: A Systematic Review and Meta-Analysis

Abstract

I read with great interest the article by Guzmán-Calderón et al. [1] about the utility of intracystic levels of glucose and carcinoembryonic antigen (CEA) to differentiate mucinous and non-Mucinous pancreatic cysts. In this review, the authors concluded that intracystic glucose appears to be better than CEA for differentiating these lesions and suggests that intracystic glucose must be measured routinely. However, some methodological concerns should be highlighted for an adequate interpretation of the results. Meta-analyses of diagnostic accuracy studies are increasingly published in the medical literature [2]. They provide pooled estimates of the sensitivity and specificity of a test. In addition, other accuracy measures such as likelihood ratios, diagnostic odds ratio, summary receiver operating characteristic (ROC) curve, and area under the curve are also frequently reported. There are several statistical packages for diagnostic test accuracy meta-analysis with diverse accessibility, usability, and methodological soundness [3]. In the article by GuzmánCalderón et al. [1], the meta-analysis was performed using the Meta-Disc version 1.4 software. This tool was developed in Spain and allows to estimate all accuracy measures previously described [4]. However, the statistical methods used to combine the diagnostic estimates have many limitations and are currently discontinued [3]. For this reason, the old version 1.4 was removed from the website (http:// www. hrc. es/ inves tigac ion/ metad isc_ en. htm) and the new version 2.0 has not yet been released. More rigorous statistical methods (such as the bivariate model and the hierarchical summary ROC model) that take into account the threshold effect and the correlation between sensitivity and specificity are currently available [2]. These methods are recommended by the Cochrane Collaboration for diagnostic test accuracy metaanalysis and can be easily fitted in the statistical package’s STATA® (StataCorp, TX, USA) and R (www.rproje ct. org) [5]. Using the mada package from R software and the bivariate model for all meta-analyses, the pooled sensitivity and specificity of intracystic glucose were 90.5% (95% CI, 86.5%–93.3%) and 81.6% (95% CI, 74.5%–87.1%), respectively. The pooled sensitivity and specificity of intracystic CEA were 61.1% (95% CI, 49%–72%) and 89.5% (95% CI, 82.2%–94%), respectively. The area under the summary ROC curve for glucose and CEA were 0.92 and 0.87, respectively. Based on these diagnostic parameters, it cannot be concluded that one test is superior to another. Therefore, the decision of which marker to use in clinical practice should be based on consideration of other test characteristics such as cost and accessibility. Finally, the methodological quality (risk of bias and applicability concerns) of each included study was not assessed by the authors. Currently, the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool is recommended for this purpose [6]. This is a key step in conducting a systematic review and must be reported [7].