Abstract
The diagnosis of early mycosis fungoides (MF) is one of the greatest challenges in dermatopathology. Additional diagnostic modalities, such as immunohistochemistry and molecular testing, like T-cell receptor (TCR) gene rearrangement testing, are widely implemented. However, no standalone method to diagnose early-stage MF with high sensitivity and specificity is established. Several algorithms for early-MF diagnosis have been proposed: Guitart et al (2001) proposed a 10-point scoring system to evaluate the likelihood of early MF based purely on histology. The International Society of Cutaneous Lymphoma (ISCL) in 2005 proposed an algorithm integrating clinical findings as well as immunopathologic and molecular biological evaluation into morphological assessment. Although clinicopathological correlation included in the latter is intuitively expected to improve a diagnosis of early MF, no study to date has directly compared these two diagnostic algorithms. We conducted a head-to-head comparison of the Guitart algorithm and a modified ISCL algorithm to determine relative accuracy rates and variability in diagnoses of early MF.
Clinicopathologically well-characterized cases of MF (n=23) and non-MF “distractors” (n=22) from the archives of the University of Washington that had corresponding clinical photographs and available TCR gene rearrangement testing results were included. In part 1, participants were presented with H&E WSIs and asked to score each of the 45 cases based on the Guitart algorithm. In part 2, participants were presented with H&E WSIs, corresponding clinical photographs, summarized results of immunohistochemical studies, and TCR gene rearrangement results, and asked to score each of the 45 cases based on a simplified ISCL algorithm. McNemar Chi-squared test was used to test if the proportions of correct responses are significantly different between the algorithms.
Six raters provided responses; five were board-certified in dermatopathology and two in hematopathology (one in both), with the mean years of practice being approximately 7 years (range: less than one to 18). The simplified/modified ISCL algorithm showed a significantly higher accuracy rate as well as a lower average variance compared to the Guitart algorithm, with general preference of raters for the ISCL algorithm.
This is the first study directly comparing two existing algorithms for the diagnosis of early MF. A modified ISCL algorithm showed a higher accuracy rate in differentiating MF/non-MF compared to the Guitart Algorithm. These findings support the importance of an integrative approach in diagnosis of early MF.