Optimal histopathological analysis of biopsies from metastases of neuroendocrine tumor (NET) of the lung requires more than morphology only. Additional parameters such as Ki-67 labeling index are required for adequate diagnosis, but few studies have compared reproducibility of different counting protocols and modalities of reporting on biopsies of lung NET. We compared the results of four different manual counting techniques to establish Ki-67 LI. On 47 paired biopsies and surgical specimens from 22 typical carcinoids (TCs), 14 atypical carcinoids (ACs), six large cell neuroendocrine carcinomas (LCNECs), and five small cell carcinomas (SCCs) immunohistochemical staining of Ki-67 antigen was performed. We counted, in regions of highest nuclear staining (HSR), a full ×40-high-power field (diameter = 0.55 mm), 500 or 2000 cells, or 2 mm2 surface area, including the HSR or the entire biopsy fragment(s). Mitoses and necrosis were evaluated in an area of 2 mm2 or the entire biopsy fragment(s). Between the four counting methods, no differences in Ki-67 LI were observed. However, a Ki-67 LI higher than 5% was found in only four cases when in an HSR, 500 cells were counted (18%), five (23%) when in an HSR 2000 cells were counted, four (18%) when 2 mm2 were counted, and one (5%) TC case when the entire biopsy was counted. A 20% cutoff distinguished TC and AC from LCNEC and SCC with 100% specificity and sensitivity, while mitoses and necrosis failed to a large extent. Ki-67 LI in biopsy samples was concordant with that in resection specimens when 2000 cells, 2 mm2, or the entire biopsy fragment(s) were counted. Our results are important for clinical management of patients with metastases of a lung NET.