-
- admin
- Site Admin
-
- Posts: 647
- Joined: Wed Aug 11, 2004 12:19 am
By Anonymous on Thursday, December 4, 2003 - 01:41 am:
Can anyone say the differnce between the terms accuracy and recovery in method validation.generally we use the term recovery while performing validation(ie. spiking known amounts of impurities and calculating the percentage of recovery.
-------------------------------------------------------------------------------------------------------
By Gert on Friday, December 5, 2003 - 01:34 am:
Hi,
The accuracy is the used as a messure of how close you can predict the real concentration in a certain sample. In this case you have to spike blanc matrix samples with your compound to be analysed at a certain concentration. Then you analyse this sample as if it was a real sample any you compare the measured concentration with the concentration you spiked.
In bioanalysis we use the recovery as a meassure of the sample preparation efficienty.
In this case you prepare a sample with a known concentration. You analyse this sample and compare the response of the peak(ie height or area) with the peak response of an academic solution with the same concentration as the prepared sample, with the concentration factor takeninto respect as needed.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Tuesday, February 10, 2004 - 02:51 am:
Gert presented what seemed a nice version of what GLP, etc. etc. ruled. But to us old guys who worked on common sense these terms had different practical meaning:
Accuracy should be determined, for instance, with at least two different methods, until chemically possible doubts are removed. An example shows how this differs from current "rules". One gets a peak at rt of the analyte, it may be, lets say, composed of 50% analyte and 50% "dirt". Now you spike such a sample with a known amount of analyte and the peak area is increased as expected. Your "accurracy" determined this way seems perfect, but in actuallity your analysis is 100% off.
On recovery: Again, rules seem to require only relative recoveries. An absolute recovery would require to use, as an example, a radioactively labeled analyte. With that you can determine were all of the analyte went. Or another possibility I can think of: You do a flow injection of a known analyte to get at the area (careful, this is very tricky).
It all depends on what you want to do, satisfy rules or to know the composition of your analyte.
-------------------------------------------------------------------------------------------------------
By Tom Mizukami on Tuesday, February 10, 2004 - 09:47 am:
Hi Hans,
I didn't follow your standard addition example. If your interference is from the matrix or external to the system then your area won't double as expected.
In a pharmaceutical setting, I think it would be rare to be performing analytical method validation so early in the process that the compound primary characterization and reference standards were not available.
I understand your point about orthogonal quantitation. This may be necessary if pure reference standards are not available. However, if you have pure reference standards and complete compound characterization, then I think demonstrating column recovery via flow injection analysis and relative recovery and accuracy should be sufficient.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Tuesday, February 10, 2004 - 11:24 pm:
Tom,
in your pharmaceutical setting you are usually sure of what´s in your unknown, so you can be reasonably sure (not necessarily certain, even here!) that relative values are close to the absolute value (the real thing). My example is quite usual for a sample with unknown matrix. Here a slightly different, and hopefully more straight forward, hypothetical example: Your unknown produces a peak at the retention time of your analyte of an area = 100. This area is in reality composed of 50 analyte + 50 dirt. Now you spike this unknown with analyte that you know should give a peak of 150. Your spiked sample then produces a peak of 250. You think everything is fine, in actuality your analysis of analyte in the unknown is off by a factor of 2.
Or simply, spiking does not give you any information on homogeneity of your peak.
Darn, my English is really getting rusty.
-------------------------------------------------------------------------------------------------------
By Tom Mizukami on Wednesday, February 11, 2004 - 12:29 pm:
Hans,
Of course your example is correct, but who would do this. If you have a known good reference standard and analyte in an unknown matrix and decided to quantitate via standard addition you would perform multiple spikings and do the regression to determine the matrix component of your sample area. Anyway, you know all of this and I understand you. Your English is much better than my German.
Getting back to anonymous, accuracy and recovery can often be combined in the pharmaceutical setting because as Hans said we have a nice clean simple matrix and normally have a sample of placebo available to test for interference. This plus the availability of pure reference standards makes our job easier than some of our colleagues.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Thursday, February 12, 2004 - 01:55 am:
Tom,
hopefully I am wrong on this, but the impression is that it´s widespread to do just that: One spiking and then be perfectly happy when an expected peak increase is obtained.
Of course, with biological matricees one gets into the situation often that the spread of the data is such that an ANOVA, and such, doesn´t help much.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Friday, February 13, 2004 - 02:46 am:
Gert, and Tom, if you are still looking in:
I have succombed to reading too carelessly and to the power of suggestion. I didn´t register the "blanc" in the "spike blanc matrix" phrase of Gert´s contribution. Thus my examples refer to the very common cases where you can´t get a blanc matrix. Therefore, I am back at my first statement: There is no way to get at accuracy without different analyses. Spiking can only yield precision in such cases.
I didn´t see the "blanc", because I see red when I see "spiking" and "matrix" in connection with all those analyses without a blanc matrix.
Tom,
I am now a bit confused about your suggestion of multiple spiking and a regression. With a blanc matrix analysis you get the amt. of possible overlapping substance.
-------------------------------------------------------------------------------------------------------
By Lime on Friday, February 13, 2004 - 06:24 am:
Hi ,
to me,accuracy is a validation parameter. and you can evaluate it by studying recovery.and they show you that how close your experimental value to true value.
hope helps.
-------------------------------------------------------------------------------------------------------
By Tom on Friday, February 13, 2004 - 11:27 am:
Hans,
Don't read too much into it. I think we are doing more to confuse than clarify. I often get questions like the original from anonymous from our chemsits. Many questions arrise around accuracy, recovery, and precision in HPLC without a true mass balance to pin it all down.
Technically you are correct and the regulatory guidance acknowledges that we are inferring accuracy from precision, linearity, and specificity. I always add column recovery using FIA to ensure there is no gross mass defect. Given precision, linearity, specificity, and column recovery I don't think it is unreasonable to infer accuracy. Within this context accuracy is often reported as percent recovery.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Monday, February 16, 2004 - 02:45 am:
OK, Tom.
Now immunassays are a multibillion dollar business based only on precision, as are most of the quality controls in med. lab. tests. In this environment HPLC is treated the same way by most "analysts" even though one seldomly knows anything about the matrix (one does not know whether the analyte peak is homogeneous). Just that much: It´s a hell of a problem.
One more thing on the original question: The reason for having those two words is, of course, that theoretically one can have a lousy recovery, but still have an excellent accuracy if you know what your recovery is and if you know that it stays constant. The lower the recovery the lower is the robustness, usually.
-------------------------------------------------------------------------------------------------------
By Anonymous on Monday, February 16, 2004 - 04:11 am:
"ACCURACY AS RECOVERY " IS WHAT IS ACCEPTABLE.
-------------------------------------------------------------------------------------------------------
By B,Buglio on Monday, February 16, 2004 - 06:18 pm:
Accuracy, as I understand it, depends on whether
youre dealing w the API or the drug product. In
the former case you would want to compare the
method against other methods(eg titration or UV
assay). In the latter case one is looking for the
effect of the matrix on the assay so one would
assay the drug product then spike the product and
determine the recovery.
-------------------------------------------------------------------------------------------------------
By Anonymous on Friday, June 18, 2004 - 12:03 pm:
How do you go about validating a standard addition method? I want to validate a method for OVI. I know it is linear based on std add. How would you go about validating for FDA regs?
Can anyone say the differnce between the terms accuracy and recovery in method validation.generally we use the term recovery while performing validation(ie. spiking known amounts of impurities and calculating the percentage of recovery.
-------------------------------------------------------------------------------------------------------
By Gert on Friday, December 5, 2003 - 01:34 am:
Hi,
The accuracy is the used as a messure of how close you can predict the real concentration in a certain sample. In this case you have to spike blanc matrix samples with your compound to be analysed at a certain concentration. Then you analyse this sample as if it was a real sample any you compare the measured concentration with the concentration you spiked.
In bioanalysis we use the recovery as a meassure of the sample preparation efficienty.
In this case you prepare a sample with a known concentration. You analyse this sample and compare the response of the peak(ie height or area) with the peak response of an academic solution with the same concentration as the prepared sample, with the concentration factor takeninto respect as needed.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Tuesday, February 10, 2004 - 02:51 am:
Gert presented what seemed a nice version of what GLP, etc. etc. ruled. But to us old guys who worked on common sense these terms had different practical meaning:
Accuracy should be determined, for instance, with at least two different methods, until chemically possible doubts are removed. An example shows how this differs from current "rules". One gets a peak at rt of the analyte, it may be, lets say, composed of 50% analyte and 50% "dirt". Now you spike such a sample with a known amount of analyte and the peak area is increased as expected. Your "accurracy" determined this way seems perfect, but in actuallity your analysis is 100% off.
On recovery: Again, rules seem to require only relative recoveries. An absolute recovery would require to use, as an example, a radioactively labeled analyte. With that you can determine were all of the analyte went. Or another possibility I can think of: You do a flow injection of a known analyte to get at the area (careful, this is very tricky).
It all depends on what you want to do, satisfy rules or to know the composition of your analyte.
-------------------------------------------------------------------------------------------------------
By Tom Mizukami on Tuesday, February 10, 2004 - 09:47 am:
Hi Hans,
I didn't follow your standard addition example. If your interference is from the matrix or external to the system then your area won't double as expected.
In a pharmaceutical setting, I think it would be rare to be performing analytical method validation so early in the process that the compound primary characterization and reference standards were not available.
I understand your point about orthogonal quantitation. This may be necessary if pure reference standards are not available. However, if you have pure reference standards and complete compound characterization, then I think demonstrating column recovery via flow injection analysis and relative recovery and accuracy should be sufficient.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Tuesday, February 10, 2004 - 11:24 pm:
Tom,
in your pharmaceutical setting you are usually sure of what´s in your unknown, so you can be reasonably sure (not necessarily certain, even here!) that relative values are close to the absolute value (the real thing). My example is quite usual for a sample with unknown matrix. Here a slightly different, and hopefully more straight forward, hypothetical example: Your unknown produces a peak at the retention time of your analyte of an area = 100. This area is in reality composed of 50 analyte + 50 dirt. Now you spike this unknown with analyte that you know should give a peak of 150. Your spiked sample then produces a peak of 250. You think everything is fine, in actuality your analysis of analyte in the unknown is off by a factor of 2.
Or simply, spiking does not give you any information on homogeneity of your peak.
Darn, my English is really getting rusty.
-------------------------------------------------------------------------------------------------------
By Tom Mizukami on Wednesday, February 11, 2004 - 12:29 pm:
Hans,
Of course your example is correct, but who would do this. If you have a known good reference standard and analyte in an unknown matrix and decided to quantitate via standard addition you would perform multiple spikings and do the regression to determine the matrix component of your sample area. Anyway, you know all of this and I understand you. Your English is much better than my German.
Getting back to anonymous, accuracy and recovery can often be combined in the pharmaceutical setting because as Hans said we have a nice clean simple matrix and normally have a sample of placebo available to test for interference. This plus the availability of pure reference standards makes our job easier than some of our colleagues.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Thursday, February 12, 2004 - 01:55 am:
Tom,
hopefully I am wrong on this, but the impression is that it´s widespread to do just that: One spiking and then be perfectly happy when an expected peak increase is obtained.
Of course, with biological matricees one gets into the situation often that the spread of the data is such that an ANOVA, and such, doesn´t help much.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Friday, February 13, 2004 - 02:46 am:
Gert, and Tom, if you are still looking in:
I have succombed to reading too carelessly and to the power of suggestion. I didn´t register the "blanc" in the "spike blanc matrix" phrase of Gert´s contribution. Thus my examples refer to the very common cases where you can´t get a blanc matrix. Therefore, I am back at my first statement: There is no way to get at accuracy without different analyses. Spiking can only yield precision in such cases.
I didn´t see the "blanc", because I see red when I see "spiking" and "matrix" in connection with all those analyses without a blanc matrix.
Tom,
I am now a bit confused about your suggestion of multiple spiking and a regression. With a blanc matrix analysis you get the amt. of possible overlapping substance.
-------------------------------------------------------------------------------------------------------
By Lime on Friday, February 13, 2004 - 06:24 am:
Hi ,
to me,accuracy is a validation parameter. and you can evaluate it by studying recovery.and they show you that how close your experimental value to true value.
hope helps.
-------------------------------------------------------------------------------------------------------
By Tom on Friday, February 13, 2004 - 11:27 am:
Hans,
Don't read too much into it. I think we are doing more to confuse than clarify. I often get questions like the original from anonymous from our chemsits. Many questions arrise around accuracy, recovery, and precision in HPLC without a true mass balance to pin it all down.
Technically you are correct and the regulatory guidance acknowledges that we are inferring accuracy from precision, linearity, and specificity. I always add column recovery using FIA to ensure there is no gross mass defect. Given precision, linearity, specificity, and column recovery I don't think it is unreasonable to infer accuracy. Within this context accuracy is often reported as percent recovery.
-------------------------------------------------------------------------------------------------------
By HW Mueller on Monday, February 16, 2004 - 02:45 am:
OK, Tom.
Now immunassays are a multibillion dollar business based only on precision, as are most of the quality controls in med. lab. tests. In this environment HPLC is treated the same way by most "analysts" even though one seldomly knows anything about the matrix (one does not know whether the analyte peak is homogeneous). Just that much: It´s a hell of a problem.
One more thing on the original question: The reason for having those two words is, of course, that theoretically one can have a lousy recovery, but still have an excellent accuracy if you know what your recovery is and if you know that it stays constant. The lower the recovery the lower is the robustness, usually.
-------------------------------------------------------------------------------------------------------
By Anonymous on Monday, February 16, 2004 - 04:11 am:
"ACCURACY AS RECOVERY " IS WHAT IS ACCEPTABLE.
-------------------------------------------------------------------------------------------------------
By B,Buglio on Monday, February 16, 2004 - 06:18 pm:
Accuracy, as I understand it, depends on whether
youre dealing w the API or the drug product. In
the former case you would want to compare the
method against other methods(eg titration or UV
assay). In the latter case one is looking for the
effect of the matrix on the assay so one would
assay the drug product then spike the product and
determine the recovery.
-------------------------------------------------------------------------------------------------------
By Anonymous on Friday, June 18, 2004 - 12:03 pm:
How do you go about validating a standard addition method? I want to validate a method for OVI. I know it is linear based on std add. How would you go about validating for FDA regs?
