Advertisement

EUS-guided FNA for GI stromal tumors: caveat lector

      To the Editor:
      GIE recently addressed EUS-guided tissue sampling in GI stromal tumors (GISTs) in 2 separate articles aimed at determining the diagnostic yield of this technique.
      • Hoda K.M.
      • Rodriguez S.A.
      • Faigel D.O.
      EUS-guided sampling of suspected GI stromal tumors.
      • Sepe P.S.
      • Moparty B.
      • Pitman M.B.
      • et al.
      EUS-guided FNA for the diagnosis of GI stromal cell tumors: sensitivity and cytologic yield.
      Although similar in their retrospective designs, the investigations differed in a few methodological aspects that deserve some comments because they could explain the significant discrepancies in the results.
      Sepe et al
      • Sepe P.S.
      • Moparty B.
      • Pitman M.B.
      • et al.
      EUS-guided FNA for the diagnosis of GI stromal cell tumors: sensitivity and cytologic yield.
      studied fewer participants, but their results were powerful enough to identify predictors of a positive EUS-guided FNA in GISTs, as previously demonstrated by others.
      • O'Neil J.
      • Al-Haddad M.
      • Leblanc J.
      • et al.
      Endoscopic ultrasound-guided fine-needle aspiration of suspected mesenchymal tumors of the gastrointestinal tract: correlation with surgical pathology [abstract].
      Other major strengths of their study included the availability of a criterion standard (surgical pathology) that allowed them to calculate the test properties of EUS-FNA with a high degree of confidence: the endosonographic appearance of lesions, the size of the needle used, and the inclusion of patients with tumors in the third layer. In contrast, Hoda et al
      • Hoda K.M.
      • Rodriguez S.A.
      • Faigel D.O.
      EUS-guided sampling of suspected GI stromal tumors.
      included only tumors in the fourth layer because all third-layer lesions, irrespective of size and location, were removed by endoscopic mucosal resection. Nonetheless, the exclusion of this subgroup of patients is a potential source of selection bias because third- and fourth-layer GISTs might not exhibit a similar exposure (EUS-FNA) to the outcome (diagnostic yield) relationship. In fact, Sepe et al
      • Sepe P.S.
      • Moparty B.
      • Pitman M.B.
      • et al.
      EUS-guided FNA for the diagnosis of GI stromal cell tumors: sensitivity and cytologic yield.
      showed that the origin of a GIST within a specific wall layer seemed to be the best predictor of cytological yield. The report of patient outcomes could have strengthened the investigation because the clinical significance of a “suspicious FNA” (positive for spindle cells but insufficient material for diagnostic stains) and its impact on patient management was not presented.
      As addressed in the accompanying editorial,
      • Al-Haddad M.
      • Dewitt J.
      EUS-guided sampling of suspected GI mesenchymal tumors: cells, cores, or a combination?.
      the lack of a reliable criterion standard was a major limitation and indeed a threat to the validity of the study by Hoda et al. Its absence limited the main objective of the study (evaluation of the test performance characteristics of EUS-FNA), albeit feasible if the authors had used the appropriate statistical methods.
      • Joseph L.
      • Gyorkos T.W.
      • Coupal L.
      Bayesian estimation of disease prevalence and the parameters of diagnostic tests in the absence of a gold standard.
      The deleterious effect of the lack of a criterion standard became even more obvious in their logistic regression analysis, bearing its own methodological problems. Five of the 7 independent variables selected for the multivariate analysis were automatically excluded from the model (reported as not available) by the statistical software. There is a significant difference between variables “not statistically significant” and “not available.” The former consist of independent variables that did not reach a β coefficient level (slope of the curve) of statistical significance. In contrast, the latter represent an automatic deletion of variables by the statistical software, which unavoidably occurs when variables have a high degree of colinearity, suggesting a suboptimal model selection. This could explain some of the unexpected findings and limitations annotated by the excellent editorial review provided by Al-Haddad and DeWitt
      • Al-Haddad M.
      • Dewitt J.
      EUS-guided sampling of suspected GI mesenchymal tumors: cells, cores, or a combination?.
      as well as the significant differences in results if compared with the work of Sepe et al.
      • Sepe P.S.
      • Moparty B.
      • Pitman M.B.
      • et al.
      EUS-guided FNA for the diagnosis of GI stromal cell tumors: sensitivity and cytologic yield.
      Although both studies provide substantial information on a similar topic and their differences are relevant to clinical practice, good-quality data can still provide biased information if not properly analyzed. As it stands, the significant discrepancy in results raises concern about potentially overlooked methodological errors. Caveat Lector!

      References

        • Hoda K.M.
        • Rodriguez S.A.
        • Faigel D.O.
        EUS-guided sampling of suspected GI stromal tumors.
        Gastrointest Endosc. 2009; 69: 1218-1223
        • Sepe P.S.
        • Moparty B.
        • Pitman M.B.
        • et al.
        EUS-guided FNA for the diagnosis of GI stromal cell tumors: sensitivity and cytologic yield.
        Gastrointest Endosc. 2009; (70:254-61)
        • O'Neil J.
        • Al-Haddad M.
        • Leblanc J.
        • et al.
        Endoscopic ultrasound-guided fine-needle aspiration of suspected mesenchymal tumors of the gastrointestinal tract: correlation with surgical pathology [abstract].
        Gastrointest Endosc. 2008; 67: AB207-AB208
        • Al-Haddad M.
        • Dewitt J.
        EUS-guided sampling of suspected GI mesenchymal tumors: cells, cores, or a combination?.
        Gastrointest Endosc. 2009; 69: 1224-1227
        • Joseph L.
        • Gyorkos T.W.
        • Coupal L.
        Bayesian estimation of disease prevalence and the parameters of diagnostic tests in the absence of a gold standard.
        Am J Epidemiol. 1995; 141: 263-272

      Linked Article

      • Response
        Gastrointestinal EndoscopyVol. 71Issue 3
        • Preview
          In his letter entitled “Caveat lector,” Dr da Silveira states that there are significant discrepancies between our paper,1 a recent publication by Sepe et al,2 and an abstract by O'Neil et al,3 which is commented on by two of the abstract's authors (Al-Haddad and Dewitt) in an editorial4 regarding our article.
        • Full-Text
        • PDF