Mid-spectrum melanoma diagnoses have low reproducibility, accuracy
medwireNews: A study of US pathologists highlights the difficulties of diagnosing melanocytic skin lesions, finding low levels of intra- and interobserver reproducibility and accuracy in diagnoses ranging from moderately dysplastic nevi to early-stage invasive melanoma.
There was less variability when diagnosing benign nevi and high-stage invasive disease, which “are at the polar ends of the histopathologic spectrum,” say the researchers.
But they add that the low level of diagnostic precision seen for cases in the middle of the spectrum is of clinical concern as “[a] pathologist’s visual interpretation is the cornerstone for diagnosing melanocytic lesions.”
To evaluate the current quality of diagnostic practices, the team recruited 187 practicing pathologists from 10 US states, all of whom assessed one of five sets of 48 skin biopsy specimens, while 118 evaluated the same set again at least 8 months later.
Specimens classed as benign nevi (class I) or invasive melanoma of at least stage pT1b (class V) in the first phase were likely to be given the same class when assessed by the same pathologist , with corresponding intraobserver concordance rates of 77% and 83%.
However, a diagnosis of moderate atypia (class II), severe atypia or melanoma in situ (class III), or pT1a invasive melanoma (class IV) in the first phase was confirmed by the same pathologist in the second phase in only 35%, 60%, and 63% of cases, respectively.
The pattern was similar when considering interobserver reproducibility, with the least variability seen for class I and V diagnoses, but the interobserver concordance rates were lower than the intraobserver rates. For instance, a pairwise comparison showed that a class II diagnosis made by one pathologist concurred with that of another pathologist in only 25% of instances, which was lower than the intraobserver concordance rate of 35%.
Accuracy, which was assessed by comparison with a consensus reference diagnosis, was similarly variable, being as high as 92% for class I specimens and as low as 25% for class II specimens. The accuracy for the other classes was 72%, 43%, and 40% for class V, IV, and III specimens, respectively.
In light of their findings, Joann Elmore (University of Washington School of Medicine, Seattle, USA) and team estimate that at a population level, only a quarter (26%) of cases diagnosed as class II would be confirmed by a consensus reference panel comprising experienced pathologists, as would 35% of class III and 51% of class IV diagnoses. By contrast, the majority of class I and V diagnoses would be confirmed by the expert consensus panel, with estimated validation rates of 92% and 78%, respectively.
They write in The BMJ that “[e]fforts to improve clinical practice should include simplification of terminology by use of a standardized classification system, acknowledgment of the extant uncertainty of specific diagnoses in pathology reports, and development of more sophisticated diagnostic tools to support pathologists.”
medwireNews is an independent medical news service provided by Springer Healthcare. © 2017 Springer Healthcare part of the Springer Nature group