manual integration

[JI2002]

We have an expert here told us in the training class that no manual integration is allowed by FDA(GC or HPLC), is that true?

[diefenwald]

really ? very interesting....

[Dan]

I have never heard that from the FDA; nor have I heard any one say that.

Of course, manual integration is to be a last resort and any peak that is manually integrated should be easily noted as such (The CDS will (or better) allow for this).

It all depends on what you put in your SOPs. That is probably what the FDA would want to see, the proper documented procedures.

Regards,
Dan

[Bruce Hamilton]

We have an expert here told us in the training class that no manual integration is allowed by FDA(GC or HPLC), is that true?

When an "expert" cites something that surprises, always ask for a reference to the policy/document/source. You shouldn't change your processes without verifying their perception.

As Dan notes, it's all about your SOPs. If the method has been validated and permits manual integration, you only have to comply the method.

Manual integration is typically a last resort option but, unless the FDA has recently changed policy, it's nor outlawed - provided the CDS retains both the original and processed data using 21CFR-compliant protocols, and identifies what was changed, and by whom. What you can't do is start using manual integration when the method only permits defined parameter ( automatic ) integration.

It's all about risk assessment and effective uses of resources. Spending two months producing an automatic, validated, method for some samples may not be justified.

Bruce Hamilton

[JI2002]

Thank you for your replies.

The "expert" was one of the few people outside HP to use 5890 GC before it came out 30 something years ago, and he is one of the few people outside Agilent to know about 7890 GC. When he talked about data system, he said he strongly opposes manual integration and everything should be done by automatic integration by manipulating the integration parameters. His points was that human eyes can't see the peak start point and end point. I said it's impossible to avoid manual integration, you can minimize using manual integration but you can't eliminate it no matter how fancy your software is. You can have the best peak shape, resolution and baseline in the standards, but it's impossible to get rid of all interferences in the samples. If chromtographers can't see the start point and end point of the peak, why all the automatic integration need to be cheched by chemists?(I recently read an article about integration, it's mentioned that chromatographer's eyes excell in identifying the start point and end point of the peak, but we just don't know the peak area) He got tipped off, and said if you are auditted by FDA and they see a manual integration, you are done. I guess he just made that up to intimidate us.

[DR]

He got tipped off, and said if you are auditted by FDA and they see a manual integration, you are done. I guess he just made that up to intimidate us.

In all matters FDA related - where the FDA has not issued a policy statement or a 483 addressing a specific topic like this - experts are routinely kicked to the curb when the FDA does issue a statement or 483 based on a specific topic like this. Avoid experts who speak on behalf of the FDA without 1) citations of policy and/or 483s to back up their claims and/or 2) a résumé entry indicating that they were employed by the FDA.

[Bruce Hamilton]

The 5890A came out in 1984 - hardly 30 something years ago.

I worked with one of the first, if not the first, in NZ, and was working with a 5830 and 5792A ( my favourite GC - with the column around the fan blades - excellent retention repeatability ) before the 5890A appeared. The fused silica columns were just starting to appear as well.

In those days, there was no Chemstation ( it appeared as HP3365 about 10 years later running on Windows/386, IIRC ), and data storage/reprocessing was extremely limited ( tapes on 3388A, various 3390 integrators ), so getting the chromatography correct first time was important.

Your expert may have been friendly with HP, and know all about the 5890, but if he had access to it before release, he should have pointed out the silly faults, like the easily broken side door latch, the stupidity of having the detector gas pressure control valves acting also as shut off valves, and the fracturing of brazed gas lines on the split injector top as it rotated when unscrewing the septum nut,. They were fixed on the Series II, but all happened to us within weeks of receiving it.

Being around a long time means younger readers should consider the potential for Altzeimers, as well the experience we oldies have.

Bruce Hamilton

[MikeD]

HP had some kind of graphical interface from at least 1976 with a 264x terminal on a 3352 memory-based system powered by a HP1000 mini. It was very expensive at the time - I think £12,000 for the terminal alone and equivalent to over $100,000 ! today. Hard to understand how we could afford any of this stuff. I used a graphical GC interface from 1983 on the HP3357 data system with CPLOT. It wasn't mouse driven but you could expand and select start/stop points on screen with great precision . This type of manual integration was and is the reference by which to judge the automatic result. Perhaps the "expert" was thinking of something cruder. Nevertheless commercial automatic integration programs are very clever indeed, as anyone would know if you they ever tried to write their own.

I think the PC-based 3365 (ALICE) was 1989 and heralded the final days of HP's 3350 series, although ours was finally put to sleep in 1996, the HP1000 mini running nearly 24/7 for an astonishing 15 years.

[Peter Apps]

If there is an issue with manual integration in a regulated environment I would guess that it is the temptation to massage peak areas to conform to the required results. How this is different from setting the autointegration parameters to achieve the same result is beyond a simple analyst like me

Peter