Assessment of the reliability of calculations of the coefficient of variation for normal and polymegethous human corneal endothelium.

In endothelial morphometry, uncertainty exists concerning how many cells should be measured. A study was undertaken to calculate mean cell area and coefficient of variation (COV) of cell areas using different numbers of cells from photo-slitlamp pictures and published micrographs. Groups of 65, 95, or 165 tesselated cells were measured and area and COV values calculated in progressive sets of 5 cells; each pair of values was compared to that obtained using all cells in each group. The results show that, for both normal (homomegethous) and irregular (polymegethous) endothelia, even cell counts as low as 50 cells can usually provide average cell area values that are within 1 to 2% of the values estimated from larger groups of cells. A similar reliability was observed for estimates of COV for normal endothelia. However, for polymegethous endothelia, even with 100 cells analyzed, the estimates of COV generally only approached a +/- 4% reliability. This uncertainty in COV estimates should be considered in both comparative studies and in regression analyses of COV changes over time or other variables.