Reply to: “Histopathologic features of melanoma in difficult-to-diagnose lesions: A case-control study; methodological issues”

Abstract

To the Editor: I thank Drs Ayubi and Safiri1Ayubi E. Safiri S. Histopathologic features of melanoma in difficult-to-diagnose lesions: a case-control study; methodological issues.J Am Acad Dermatol. 2017; 77: e149Abstract Full Text Full Text PDF PubMed Scopus (1) Google Scholar for their comments regarding our recent article on the accuracy of different histopathologic features of melanoma in difficult-to-diagnose melanocytic neoplasms.2Gonzalez M.L. Young E.D. Bush J. et al.Histopathologic features of melanoma in difficult-to-diagnose lesions: a case-control study.J Am Acad Dermatol. 2017; 77: 543-548.e1Abstract Full Text Full Text PDF PubMed Scopus (8) Google Scholar I would like to emphasize that this was an exploratory hypothesis-generating research project designed to identify which histopathologic attributes were most likely to be of value in designing a diagnostic algorithm for the difficult melanocytic neoplasm. Ayubi and Safiri recommend (1) additional validation studies and (2) better demonstration of temporality between independent and dependent variables. I agree with Ayubi and Safiri that follow-up confirmatory research studies are needed to help validate our findings. A cross-validation study using a new data set based on molecular anomalies associated with melanoma is in progress. Although additional statistical analysis of our original data set could be performed, this was not part of our original study design. The infrequency of some of the attributes in our study does not lend itself to resampling methods such as bootstrapping. Retroactive data analysis outside the original experimental design can lead to erroneous conclusions through “p-hacking” and “HARKing” (hypothesizing after the results are known).3Forstmeier W, Wagenmakers EJ, Parker TH. Detecting and avoiding likely false-positive findings—a practical guide. Biol Rev Camb Philos Soc. 2016; http://dx.doi.org/10.1111/brv.12315 [Epub ahead of print].Google Scholar I believe that cross-validation based on a novel data set will be a more valid approach. The concerns raised by Ayubi and Safiri regarding temporality are a function of the case-control study design. Because all case-control studies are retrospective, they will never demonstrate temporality as conclusively as cohort studies or randomized trials. However, cohort studies and randomized trials to assess the prospective accuracy of histopathologic features of malignancy are rare. It would be unethical to assign patients who might have a malignancy to an observation arm if the standard of care requires surgical excision. Case-control study designs such as ours provide useful information when observation without intervention is impractical or unethical.4Gordis L. Epidemiology.5th ed. Elsevier Saunders, Philadelphia, PA2014Google Scholar Histopathologic features of melanoma in difficult-to-diagnose lesions: A case-control study; methodological issuesJournal of the American Academy of DermatologyVol. 77Issue 5PreviewTo the Editor: We read with great interest the article authored by Gonzalez et al that was published in the Journal of the American Academy of Dermatology in 2017.1 The authors aimed to measure the accuracy of histopathologic features in difficult-to diagnose melanocytic tumors and the interobserver agreement of those features. They constructed a prediction model and found that asymmetry, single-cell melanocytosis, solar elastosis, pagetoid melanocytosis, and broad surface diameter were the most predictive factors of outcome studied. Full-Text PDF