Chromatography Forum

if it is impossible to integrate the peak by system then try force peak instead of manual integration that is much better option.........

I know that at least one regulatory inspector I have met doesn't treat manual integration as an immediate bad thing, but that it has 'a bad smell'.

There have been cases in the past whith (naughty) analysts who have manually adjusted integrations to either get bads runs to pass, or particular samples to pass whatever limits they have to (see McDowall, R. D. 2006, "Quality assurance implications for computerized systems following the Able Laboratories FDA Inspection", The Quality Assurance Journal, vol. 10, no. 1, pp. 15-20.). This 'Testing into Compliance' is fraudulent, and mustn't be allowed.

In modern chromatography systems, all this will be audit trailed, some also update the result for a peak in real time, which does vaguely encourage integrating into compliance, however, there is a need for manual integration. Not every sample chromatograms like a standard and not every run happens under exactly the same conditions....

This means that manual integration will be seen as a 'bad smell' that might indicate something nasty lurking under the surface. To combat this, make sure you have a procedure in place that describes how staff will do automatic and manual integration, that all staff are aware of it, and that you have audit trails and adequate records to show no fraud has occured.


Regards,

Paul.

...doesn't treat manual integration as an immediate bad thing, but that it has 'a bad smell'

Perhaps the best explanation I have seen. Well said!

I've not used manual integrations in many years.
Gone through HP (Agilent) 3352,3357, LAS, Chemserver and noe Varian Galaxie.

We finally removed recorders 10yrs or more ago.
However I can still remember scissors & weights or rulers and heights.
OK I'm old.

However back to the question.
A scenario, a basic 1 GC environment with only one or two peaks from a packed column with good seperation.
A recorder, or a data system.........

For fast GC, capillary etc, I guess it's a none starter but in the correct circumstances cheap and cheerful is sometimes adequate and OK.

The inspectors do not have time to question everything and everyone - if the paperwork looks convincing, they are unlikely to ask questions unless there are problems. And they will miss things, from time to time, that should capture their attention.

If you can get better results one way or another, this would not concern an inspector as long as the procedure is properly validated, and the operator trained to realise if there are problems - such training being recorded.

Unfortunately, too many labs are managed by people who have "trouble doing sums" and low mechanical aptitude - they have found the safest way is to do everything by the book, and they are in the job because higher management finds them an acceptable interface.

Life has told me that such people do not "think outside the box", so that when they design procedures, those procedures may not be adequate - but the inadequacy does not show when everything is running to plan.

The safe answer: If you are changing something within a method, you must validate it against the existing procedure, demonstrate that what you want to do is a definite improvement, and demonstrate that the improvement is consistent between operators.

If that contravenes someone else's idea of "what is an adequate audit trail", then it is up to them to devise a way to bring it into compliance - even if it requires a witness to do that. To do otherwise, they must demonstrate that the improvement is not significant to the properties of the product.

Manual stuff is not forbidden, as long as the procedure contains adequate safeguards.

Having said that, I've mostly had the trump card of being a QP when I needed to force a change. Life can be hard when you know more than the boss ...

Check this out. FDA states manual integration is acceptable.

http://www.gmp-compliance.org/eca_news_84.html

true, but a phenomenally naive document. For example:

Q: On PC's you can easily change date and time. How can I control this id somebody changes the date and/or the time.

A: If the system allows to change time and date without leaving an audit trail, you can develop a procedural control which states that date and time must not be changed. And you must check and document if the procedures are followed.

[saracasm] Right, so if we can't stop people from cheating, we'll just ask them nicely not to cheat, and that will sort it all out then? [/sarcasm]

This thread has been an interesting re-read for me, as a lot has changed in 2 years, and perhaps I wouldn't write exactly the same today as I did above. But I still won't use manual integration except in extremis, with enormous warnings attached, and I still will use smoothing, but carefully documented and applied to everything equally.

OK, so here is my view:
Deciding where the integration line" should" be is based on experience of chromatography and company consistancy. We use the software to do it right automatically as many times as possible to save me time and the opportunity to play with the numbers.

So if it looks wrong to me, it probably is wrong. Should I leave it wrong? Obviously not. Should I to get the software to do it better? Obviously Yes.

But how much time should I spend making a complex processing set of parameters, just to make it do what my chromatographic knowledge tells me is correct? Using events like FORCE PEAK overrule the algorithm just the same as manually integrating. I have seen people spend hours creating processing methods with 30 or 40 events in them to " get the integration right" for all samples. Even more 'dodgy' is where different methods are used for Standards and Samples... even if both are automatic.

I personally fail to see how doggedly sticking to algorithms to place baselines can be more compliant than drawing the baselines by hand, on the odd occasion it is needed. Providing there is a recording of the original automatic version ( or versions where you tried to get it right), a good audit trail with accountability and a proper reason, and a detailed and documented peer review process.

H
PS Did I read someone was still printing chromatograms?
http://www.fda.gov/Drugs/GuidanceCompli ... 124787.htm

We had a huge investigation at my last company site by the MHRA (due to various reasons I can't disclose!) but one of which was some idiot manually integrating chromatograms in a way which was intentionally misleading for study data. Our software package was so badly in need of an update he was able to do this and pass results which clearly should have failed - this is why the regulators want to move away from manual integration to minimise (stopping is fairly impossible!) mis-use.

I myself used to be a QA auditor and I in practice see no problem with manual integration so long as it is documented why it is being done. The MHRA after this issue at my previous work asked all other departments using chromatography systems to print off an "unmodified" original chromatogram as well as the "modified" chromatogram if you were using manual integration so that no-one could save over the original integration with the manual and lose the data etc. You also had to fill out a standard form detailing which injection numbers required manual integration and why. So it is accepted by the MHRA (not sure about FDA) however it has to be fully traceable and explainable and its a lot of hassle so you can see why people live with automatic integration rathe than have to go through all that rigmarole!

lynz x