Surgeon disagreement on abdominal tumors hampers treatment planning

<p>Background The Peritoneal Cancer Index recorded at laparotomy is based on visual and palpable inspection of peritoneal surfaces for disease. This study aimed to analyze interobserver variation in assigning the lesion score (LS) and morphologic term (MT) to characterize peritoneal lesions (PL) and predict the probability of malignancy (POM) among surgeons with expertise in cytoreductive surgery (CRS) for peritoneal malignancies.</p><p>Methods The study selected 80 intraoperative images of PLs depicting different morphologic appearances of PLs arising from different primary tumors in various peritoneal regions. In the study, 50 expert peritoneal malignancy surgeons were asked to assign an LS to the region in question, select the MT or MTs to describe the PL, and predict the POM. Information on the presence of disease on histopathology was not provided at this point. Inter-observer reliability was evaluated using Krippendorff's alpha (alpha). A consensus was reached if any option received more than 75% of the votes.</p><p>Results The study participants comprised 41 (82%) of 50 experts. Consensus on LS was achieved for 18 images (22.5%), with low agreement (alpha = 0.174). For MTs, a consensus was reached for 21 images (26.5%; alpha = 0.0902, denoting low and unreliable agreement. Of these 21 images, the MT used was "tumor nodule" for 90.4% of the images (p &lt; 0.001). The POM was accurately predicted in 52.5% of the cases (alpha = 0.155). Administration of neoadjuvant chemotherapy had no impact on the surgeons' assessment of the three parameters.</p><p>Conclusions Even the most experienced CRS surgeons showed high interobserver variation and unreliable agreement in the description and accurate characterization of PLs on pictorial records. A Delphi consensus to standardize the MT used and scoring of PL could reduce discordance.</p>