The sampling efficiency of the unbiased stereological vertical-section method was analyzed in five hydroxyapatite-coated implants. They were inserted into humans and harvested after 1 year. To find an optimal sampling design for histomorphometric analyses, sampling efficiency was estimated by variance analyses at different sampling levels (humans, sections, fields of view, and number of counting items) and intensities. Only minor changes in variance were observed when the initial scheme was reduced to include just one of the two possible implant sides, every third field of view, and half the density of the probe; this reduced the total workload at the microscope to less than 10% for all sections. In addition, the number of sections for analysis could be reduced to every fourth section per implant (three to four sections for evaluation) without significantly increasing variance. The study demonstrated that biological variation contributed to the majority of the total observed variance. Optimizing the sampling design could significantly reduce the workload at the hard-tissue microtome and the microscope without reducing the quality of the data that were unbiased and that had low sampling variance as compared with the true biological variation.