Chromatography Forum

I got calibration data from LC-MSMS detection in plasma sample.
The data is not fit in the linear model y=ax+b in range of 0.1 - 30 ng/ml.
The coefficient of determination is higher than 0.99 and I accept.
But the accuracy is bad (~50-60%).

1. I found some paper weighted data with 1/x2. How could
they weighted the data? Could I used Excel to do this or any program.

2. I try the power model Y = aX^b. The accuracy at LQC is accepted in some curve.

3. What are reasons that some lc-msms data is not fit to simple linear y=ax+b model or it sould be weigted.

Standard curves over a 300-fold range often give excellent coefficients of determination, even when the curve is not actually linear! I suggest you visually inspect your calibration curve. Get out graph paper if necessary. Does it look like a straight line? Or does it drop off at higher concentrations? (not at all uncommon) If the latter, you need to reduce the magnitude of your standard curve and, if necessary, dilute samples so you are working in the linear range of the assay.

If the curve LOOKS linear, especially if the 50% errors are mostly confined to the lowest concentrations, weighting usually works quite well to bring the residuals down. Most MS and chromatographic software programs have this option. You may be able to program Excel to do it, but I don't think it's part of the default package.

Thank you for your answers.

It drop at higher concentration.
I must determine wide range ~0.5-24 ng/ml plasma level in pharmacokinetic study.

1. Could two range of calibration curves are appropiate to use?
2. Can I use other equation for calibration curve?