Level of interobserver variability estimation as a valuable tool: assessment of tumour budding in colon cancer.

Aims: Tumour budding is considered to be a good marker for progression and prognosis in colorectal carcinomas. A uniform classification system has been established recently. The natural element of uncertainty in the practice of human medicine is also exhibited in the assessment of tumour budding. We tested the hypothesis that interobserver variability can be estimated during the assessment process and investigated its potential clinical implication.
Methods and Results: Six investigators with different levels of experience could perceive different levels of difficulty (LOD) and estimated different levels of interobserver variability (LOIV) (Li1, lower than average; Li2, average; Li3, higher than average) during the assessment of tumour budding in 244 cases of colon cancer (pT3/4). In total, the LOIV showed following distribution: Li1: 36.1%, Li2: 43.9% and Li3: 20.0%. The LOIV was correlated significantly with the LOD given by the investigator. In total, the agreement rates with the final consensus classification were: Li1: 93.4%, Li2: 78.5% and Li3: 58.4%. The relative risk of disagreement with the final consensus classification was more than six times higher when a case was estimated to have a high rather than a low interobserver variability.
Conclusion: Our data show that the investigator can estimate the interobserver variability during the ongoing rating process in pT3/4 colon cancer. The LOIV/LOD seems to be a valuable parameter of the assessment quality. For Li3 cases further measures seem mandatory.
(© 2018 John Wiley & Sons Ltd.)