Ki67 proliferation indices in advanced jejunoileal well‐differentiated neuroendocrine tumours are most prognostic when evaluated in mesenteric tumour deposits

Background Ki67 proliferation indices are used for grading well‐differentiated neuroendocrine tumours (NETs). Selection of tissue for staining can be challenging, especially in jejunoileal NETs (JI‐NETs), which can be multifocal and produce multiple metastases. This study aimed to determine parameters for selecting the most prognostically appropriate tissue for Ki67 staining in JI‐NETs. Materials and Methods We identified JI‐NETs with at least one nodal metastasis and/or mesenteric tumour deposit. For each, we recorded patient age and sex; size of largest primary tumour, nodal metastasis and/or deposit; AJCC staging; and patient outcome (recurrence and disease‐specific death). We stained the largest primary, node and deposit in each case for Ki67 (clone SP6, Biocare Medical CRM325, 1:50) and also assessed mitotic rates. We photographed hotspot areas on each Ki67 slide. Ki67 index was calculated for each photo using QuPath digital image analysis software after training using a random forest supervised classifier. Twenty cases underwent manual Ki67 counting to validate QuPath results. All variables were compared to patient outcomes using standard univariate and multivariate Cox proportional hazard regression ( α = 0.05). Results Among 75 JI‐NETs from 42 female and 33 male patients (mean age 61 years), 71 had positive nodes and 47 had deposits. Mean sizes of the largest primary tumours, nodal metastases and deposits were 1.97, 1.28 and 2.0 cm. Mean Ki67 was 1.82 (largest primaries), 2.24 (largest nodes) and 2.08 (largest deposits). In 51 cases (68%), the Ki67 of a node or deposit was higher than the primary tumour. Twenty‐one tumours (28%) recurred, with median projected time to recurrence time of 108 months. Over a median overall follow‐up of 72 months, six patients (8%) died of disease. Tumour deposit Ki67 was the only factor associated with increased risk of progression on multivariate analysis (hazard ratio [HR] 1.22 for each 1.0% increase, 95% confidence interval [CI] 1.02–1.46, P = 0.032); primary and nodal Ki67 were not, nor was the highest Ki67 or the Ki67 of the largest focus within each case. No Ki67 indices were significantly relevant to overall survival on multivariate analysis, though largest tumour deposit size was (HR 1.51 for each 1.0 cm increase, 95% CI 1.08–2.11, P = 0.015). Conclusions Tumour deposits appear to play a key role in the outcome of patients with JI‐NETs. For meaningful Ki67 evaluation of JI‐NETs, the largest tumour deposit should be stained, if available.