Log in
Register|
konkous ☆ Greece, 2011-06-28 15:52 (5090 d 14:54 ago) Posting: # 7176 Views: 4,460 |
|
|
Dear forum members During pre-study validation one has to run 30 runs to get only 8 valid. All reasons for the rest of the runs failing are documented and explained. I wonder what would this look like from an auditor's point of view? |
|
Ohlbe ★★★ France, 2011-06-28 16:04 (5090 d 14:42 ago) @ konkous Posting: # 7177 Views: 3,843 |
|
|
Dear Konkous, ❝ I wonder what would this look like from an auditor's point of view ? At first sight, simple: garbage. And there is a chance there will not be a second sight. Unless you can demonstrate that this was just bad luck, and an unfortunate accumulation of various different events totally unrelated to the method itself (in 12 of the 20 runs ?), there is a very large chance that an auditor will consider your method as unreliable. Rather go back to method development. Regards Ohlbe — Regards Ohlbe |
|
Helmut ★★★   Vienna, Austria, 2011-06-29 05:21 (5090 d 01:25 ago) @ Ohlbe Posting: # 7179 Views: 3,809 |
|
|
Dear Ohlbe! ❝ […] (in 12 of the 20 runs ?) Cough. Failure rate was even 73% (22/30) – not just 60% (12/20).  I’m wondering why an analyst would wait that long for a miracle to happen and not go back to method development earlier.— Dif-tor heh smusma 🖖🏼 Довге життя Україна! ![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png) Helmut Schütz ![[image]](https://static.bebac.at/img/CC by.png) The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
Ohlbe ★★★ France, 2011-06-29 11:59 (5089 d 18:47 ago) @ Helmut Posting: # 7185 Views: 3,811 |
|
|
Dear Helmut, ❝ Cough. Failure rate was even 73% (22/30) – not just 60% (12/20). That's what happens when deleting the text to avoid TOFU ! Looks like I couldn't believe the figure myself...Regards Ohlbe — Regards Ohlbe |
|
konkous ☆ Greece, 2011-06-29 12:46 (5089 d 18:00 ago) @ Ohlbe Posting: # 7189 Views: 3,812 |
|
|
Thank you for your answers, i totally agree. It seems that the problems causing this extremely high failure rate were the following: 1. An instument parameter causing high variability in response especially at concentrations near the LLOQ 2. The calibration curve fit which appears to be problematic regarding the ULOQ. The back calculated value is always between 83-87% of the nominal value. All weghting factors (1/x, 1/x2, 1/y, 1/y2) applied seem to have more or less the same problem. Assuming that someone has 2 or 3 consecutive runs establishing accepted linearity, accuracy and precision would it be acceptable to start analyzing study samples? Should the complete set of validation runs (consisting of about 8 or 9 runs)be completed prior to main study? Is it acceptable to perform the remaining validation runs after the main study? |
|
Ohlbe ★★★ France, 2011-06-29 13:11 (5089 d 17:35 ago) @ konkous Posting: # 7190 Views: 3,814 |
|
|
Dear Konkous, ❝ [...] The back calculated value is always between 83-87% of the nominal value. All weighting factors (1/x, 1/x2, 1/y, 1/y2) applied seem to have more or less the same problem. Then the problem is not with weighting but with the model itself. A linear model is not always suitable in MS/MS, and a quadratic model can give a better fit and is well accepted. This being said, if you have real problems with the linearity of the detector and the response gets saturated you will have to decrease the ULOQ and dilute study samples. ❝ Assuming that someone has 2 or 3 consecutive runs establishing accepted linearity, accuracy and precision would it be acceptable to start analyzing study samples? Should the complete set of validation runs (consisting of about 8 or 9 runs) be completed prior to main study? Is it acceptable to perform the remaining validation runs after the main study? You can start analysing your samples before you have the results of the long-term stability. You can also skip dilution integrity, if you think you won't need to dilute any samples in your study. All other parameters should be validated before you start analysing your study samples. This was discussed at length last year in Brussels at the EBF/EUFEPS meeting with the members of the Writing Committee of EMA's bioanalytical guideline. Industry representatives argued that starting analysing the samples before all validation results are available would be their own risk. Regulators answered that it was not just industry's risk, but there were also some ethical considerations: if you realise after analysing your samples that the method is flawed, and you have no sample volume left to re-analyse, you will have to repeat the study and expose again subjects to the drug, or expose and kill more animals. Regards Ohlbe — Regards Ohlbe |
Ing. Helmut Schütz