The effect of using different blood lactate sampling sites in combination with different exercise test designs on the validity and accuracy for prediction of maximal endurance running velocity was investigated. Ten aerobically all-round trained firemen and nine aerobically endurance trained long-distance runners performed six differently designed treadmill running blood lactate accumulation tests. Each test consisted of five consecutive running periods on a treadmill of either 4, 6 or 8 min duration, with a mean increase in running velocity between each period of either 0.25 or 0.5 m.s-1. The corresponding treadmill running velocity to a lactate concentration of 4.0 mmol.l-1 in capillary and venous hemolysed blood and plasma for each running velocity. The mean running velocity from a maximal 12 km run for the firemen and a maximal 21 km run for the runners served as the reference of maximal endurance running velocity. There were both significant (p < 0.001) and similar relationships (r = 0.86-0.94) and no difference in mean prediction error between the predicted and measured maximal endurance running velocities with all tested protocols. However, there was a high risk of making both over- and underestimations (5% to -4%). The lowest risk of making an inaccurate prediction was found when a running duration of 8 min for each running period was used in combination with an increase in running velocity of 0.25 m.s-1, and the lactate measurements were performed in hemolysed capillary blood.