2 Reliability. The performance check can be done using the same buffers or reagents that will be dispensed by the pipettor, thus eliminating possible discrepancies in between pipettor performance with calibrator solution and with normal dispense reagent.
In the example presented above, the mean dispense volumes for 2 of the 8 channels fell outside the acceptable range (198 - 202 L). Both mean values are significantly different from the nominal value of 200 L (P<.05, students t-test), however, neither is significantly different from the lower limit (198 L). Thus the channels are probably out of specification, but the ultimate Pass/Fail decision depends upon the laboratorys approach to taking measurement uncertainty into account. In a recent publication,14 Ellison et al nicely summarize and interpret the International Standards Organizations recommended method for quantitative estimation of uncertainty.15 The paper also stresses the importance of correct interpretation of accuracy results, so that they are judged neither overly optimistically, nor unduly pessimistically. In practice, such statistical analyses can be avoided by merely repeating the calibration measurements to ensure that the calibration is within or outside the acceptable range.
The terms accurate and precise are relative. A 5% error may be considered insignificant in one method and totally unacceptable in another. Pipettor manufacturers specifications for dispense volumes of 100 - 200 L are typically less than 1% for both accuracy (+ %) and imprecision (CV%), if obtained under nearly ideal conditions. For clinical analyses, medically-allowable errors (of which pipetting error would be only one of several potential contributors) can be 5%, 10% or 20%.16 Thus many clinical laboratories accept an accuracy requirement of + 3% and a precision requirement of + 1.5% for pipet dispense volum