Use of the recently updated AJCC Cancer Staging Manual, 8th edition, produced about 10% better accuracy when interpreting T1a invasive melanomas in a national study.
For stage T1b or greater cases, using the AJCC 8 instead of the AJCC 7 yielded about 6% higher concordance between interpretations and consensus reference diagnoses, said Joann G. Elmore, MD, MPH, of the David Geffen School of Medicine at the University of California, Los Angeles, and her associates. “Intraobserver reproducibility of diagnoses also improved,” but “concordance and reproducibility remain low,” the researchers wrote. The report was published in.
For the study, 187 pathologists each interpreted one of five glass slide sets of melanocytic skin biopsies. At least 8 months later, they interpreted the same set again, both times recording their interpretations on the online Pathology Assessment Tool and Hierarchy for Diagnosis (MPATH-Dx). Researchers then used AJCC 7 and AJCC 8 criteria to group these interpretations into one of five MPATH-Dx classes. The final analysis included only the 116 cases of invasive melanoma, because changes in the AJCC 8 criteria affected only MPATH-Dx classes IV (T1a) and V (T1b and higher), the investigators explained.
For T1a invasive melanomas, overall concordance between interpretations and consensus diagnoses rose from 44% (95% confidence interval, 41%-48%) under the AJCC 7 criteria to 54% (95% CI, 51%-57%) under the AJCC 8 criteria. For stage T1b or higher lesions, concordance rose from 72% to 78%. Intrapathologist reproducibility improved from 59% to 64% for T1a cases and from 74% to 77% for T1b or higher cases. Thus, the AJCC 8 produced “modest” improvements in concordance and reproducibility over the AJCC 7, the researchers said.
One of the most likely explanations is that the AJCC 8 eliminated dermal mitotic activity as a criterion for T1b melanomas, which is “potentially unreliable” in thin lesions, they added. Persistently low concordance and reproducibility “suggest that conventional histopathology has been parsed to a degree that falls below the limits of reliability,” they concluded.
The National Cancer Institute provided funding. Dr. Elmore disclosed ties to Myriad Genetics, SciBase, and the National Institutes of Health.
SOURCE: Elmore JG et al. JAMA Network Open. 2018 May 18. .