I f you need to quantitate in the LOQ range, you need a calibration curve from LOQ to ~10LOQ. The RSD is normally higher than at higher concentrations. You need to determine the real LOQ (not from LOD) by preparing several concentrations of LOQ solutions, do 6 replicate injections and the concentration of LOQ that gives you

Chaojun,

There is no need to obtain the linearity starting from LOD.

What you can expect is to have linearity starting from LOQ, i.e. 2 or 3 x LOD.

How you determine the LOD ? on the basis of signal to noise ratio? or using a standard deviation of you calibration curve?

regards,

G

I f you need to quantitate in the LOQ range, you need a calibration curve from LOQ to ~10LOQ. The RSD is normally higher than at higher concentrations. You need to determine the real LOQ (not from LOD) by preparing several concentrations of LOQ solutions, do 6 replicate injections and the concentration of LOQ that gives you

@ Dr Andrei,

"......gives you

The method suggested by Andrei is ok - but the LOQ is determined on the basis of assumed/accepted RSD limit. It can't be 20%!

For some applications it can be

Dear Chaojun Wu,

I agreed that the calibration curve start from LOQ. But Looking at your range, if you start with 0.07 ng/mL or even 1 ng/mL to 1000 ng/mL is very wide range. I'm not really sure your range of sample concentration. But if the maximum sample concentration is 500 ng/mL, I don't think you need calibration range until 1000 ng/mL. If you have wide sample range, I would suggest you have two different calibration curve which with lower range calibration curve (1 ng/mL - 100 ng/mL) and higher range calibration curve (50 ng/mL - 1000 ng/mL).

LOD conceptually is not equal to LLOQ. Your curve range is much higher that LOQ you can hit, it does not mean it is not properly set. The curve range is set based on the real need your case, based on the dosing formulation.

There is no need to give any explanation in this regards.

The use of LOQ as a start of the calibration curve is much more logical and more practical than LOD. you can calculate LOQ as the ratio of ten times of standard deviation of blank readings to the slope of the calibration curve after the extraction of the analyte.

You have gotten 0.07 ng/mL limit of detection which can be the best result depending on the type of Instrument detection limit you used and your method. but the range from 15-100 ng/mL is your working calibration linear range that is depending on your analyte abundance so which is depending on R ( R2).

You have to check instrument response of analyte and other wise pre concentrate your sample .

Once again, I want to thank everyone here for helping me. You are very kind!

Dear Chaojun Wu,

When our detection limit is below 1 ng/ml then the probability of both random as well as systematic error increases with higher altitude. In this case we can minimize the random error by increasing the number of replication but we have no control on the systematic error and this error causes large deviation from your true value. So, I would suggest you that draw a different calibration curve taking the same results below 10 ng/ml and you definitely get good linearity. Because at lower rang systematic errors of each measurement are nearly equal which neutralize the whole problem.

Dear Chaojun Wu,

When our detection limit is below 1 ng/ml then the probability of both random as well as systematic error increases with higher altitude. In this case we can minimize the random error by increasing the number of replication but we have no control on the systematic error . Draw a different calibration curve taking the same results below 10 ng/ml and you definitely get good linearity.

When our detection limit is below 1 ng/ml then the probability of both random as well as systematic error increases with higher altitude. In this case we can minimize the random error by increasing the number of replication but we have no control on the systematic error . Draw a different calibration curve taking the same results below 10 ng/ml and you definitely get good linearity.