We recently had a consultant come in to our QA lab to conduct gap analysis in respect to ISO 17025 (and other A2LA requirements). We currently follow the FDA guidelines for our industry but are seeking ISO 17025 accreditation since we are an international company (wouldn't be a bad idea to hold ourselves to a higher level, right?). We currently don't collect any QC data or run QC samples alongside our unknowns -- we were under the impression that our method validation/verification data was sufficient to prove our methods are good for the matrices we're running. Well, the consultant is telling us that for ISO 17025 accreditation we need to start doing QC activities for everything we want to be accredited for (One of his statements was, "How do you know a method you validated 5 years ago is still operating properly?").

:bonk:<--- This is how we feel. QC? Seriously? Do we really need to add this much work to our already stressed system???

I have two questions:
1) Is the consultant making an extremely conservative interpretation of the standard and A2LA requirements? (or is he spot-on and we should listen ... should've seen the faces of everyone in the room when he started to speak about QC - went from active-listening to "I think I'm going to be sick. :jawdrop:)

2) How are other companies/Quality Assurance labs handling this?

Please help!!! :frust:

Your lab currently does zero activity to assure that the data it produces is "good"? How does your lab know that its data is accurate, repeatable, and defensible (if need be, in court)?

I see that you do perform method validation/verification activity, so this may be your QC, but it sounds like you are not verifying on a regular basis, just when you implement a new method. What types of analysis and instrumentation are you running? I'm not familiar at all with FDA guidelines, but for what I've seen on EPA and other standardized testing, method validation/verification is more than just as a new method is implemented/one-time action. First thing I think of for instrumentation is instrument drift. Even with non-spectrometric analysis, say a titration procedure, one would want to implement some type of regular QC. We use a titration method here for Si determination that is probably 70-80 years in use, but even though the chemistry behind the analysis is rock solid, there can be error in weighing sample, sample digestion, and titration. Some type of regular method QC to assure data quality still needs to be followed with this test once it was formally implemented in my industry. Frequency of QC depends on testing type, standard used, etc.

How does your lab know that its data is accurate, repeatable, and defensible (if need be, in court)?

Great point - defensible in a court of law or in the court of customer opinion! Most of the time data is met with little "opposition" - but when it occurs, you better be ready to offer up strong justification of the techniques, and everything that supports them (environment, calibration, training, etc.) You will not want to be on the short end of such a situation. Forget the standard - look at how it can impact your organization!

First, :thanx: I appreciate all insight into this matter - this is new for us.

What types of analysis and instrumentation are you running?

We run various technologies in chromatography, spectroscopy, wet chemistry, and physical testing. Because of our wide-range of testing capabilities I think one step we're going to take in the accreditation process is to scale-back our scope a bit (initially, we had everything under the sun - that we are capable to do - on our draft scope). It was suggested by the assessor to start out with the 20% of tests that produce 80% of our results, which is something we probably should've considered before we began.

Your lab currently does zero activity to assure that the data it produces is "good"? How does your lab know that its data is accurate, repeatable, and defensible (if need be, in court)?

Correct - we do zero. We've always relied on method validation/ verification and instrument maintenance (PM, IQ/OQ/PQ, calibrations, etc) activities. Any time our data is questioned we pull out that data (along with training/analyst-qualification records and any additional supporting info) and it has been sufficient for whoever requested it - thus no reason for us to suspect on-going QC was necessary.

Frequency of QC depends on testing type, standard used, etc.

Our assessor is saying we need to run a QC sample with every analysis! Could we, instead, have a policy that says we run QC with every 10 analyses? It would be great if you could expand on this, Joe ...

In any case, I'm getting a sense that QC is a standard practice across "lab-land" and it's "mind-boggling" (for lack of a better term) that a QA lab in the pharma industry does not do it. Thus, we should stop our crying and pull up the boot-straps and get a QC system in place, especially if we want 17025 accreditation.

But I would like to put this question out there: How to other GMP/ FDA-regulated labs deal with this issue? Do you think the FDA would tell me their process??

I remain ... :confused:

Voodoo,

if FDA methods do not specify QC activity, then you might be able to define your own activity schedules. I'm not familiar with 17025, so I can't say if it also has a preferred schedule of QC activity, or just wants you to establish and maintain one.

Here is what I see, coming from the metals industry and working in a QC lab under a 9001:2008 cert: ASTM 305 for spectroscopic analysis calibration curves has section 8.2.3 on establishing frequency of verification of the calibration curves with a reference material. ASTM 882 addresses QC and accountability in the chem lab, noting the use of QC samples. CH 1 in the EPA methods is devoted to QC of analysis. This includes the method validation/verification upon instituting any method in a lab, but also covers ongoing verification activity of the method (from sampling to sample prep to actual instrument/wet analysis) as it is used down the weeks, months, and years. This includes use of matrix blank samples to watch for contamination, spiked analyte samples, and reference materials and/or matrix-matched QC samples built from individual RM's. Just some examples.

Frequency of QC: ASTM has you look at repeatability on reference materials to determine frequency of QC checks. I believe EPA has the user doing it with each batch of samples run with a method. I do metals analysis with WD-XRF and ICP-OES, and have different strategies for each. With XRF, precision is very tight and instrumental drift is not normally a big problem, so daily QC is fine for my situation. With ICP, QC depends on batch size, for me. I tend to run a QC sample, then run 5 unknowns, then run the QC again. With wet analysis for metals, I typically have a QC sample run in duplicate alongside each batch of duplicate unknowns. For physical tests, like particle size in sieves, or Ash content in coals, QC is mainly a physical calibration check of the apparatus, with maybe a yearly QC sample run to verify. It really depends on the test method and apparatus.

I'm not sure, under maintenance activity, what "IQ/OQ/PQ" meant in your post. What is calibration activity mean for you, with wet and spectrochemical analysis? Regular drift correction, or calibration by checking with a reference material?

Again, I don't know 17025, so I don't know of any constraints it imposes on frequency of QC. I AM surprised FDA methods don't go beyond initial method validation/verification specifications though. But I have no experience at all with pharma or FDA. In absence of any FDA or 17025 mandates on QC frequency, I would experiment with each piece of analytical equipment and method, to see what frequency is justified. If your methods on the spectrometers and the chromo are similar enough to each other, for each piece of equipment, then look solely at the instruments, and not by each method. Pick one method on one instrument and start with more frequent QC and watch your repeatability and accuracy. Then back off frequency until your repeatability suffers enough that you would not be comfortable with the results; say 2 sigma of the "true" value of the selected analyte in a reference material. If FDA or 17025 does not specify, you don't want to QC too much and waste time and resources running QC with each sample, when running a QC per batch of samples would suffice to show stable, repeatable analysis that one would have confidence in.

All the things you do now ARE part of QC in a lab. But in many situations, I would want to see historical evidence (trend analysis) that lab testing is able to put out quality data that is defensible. If you have to perform major maintenance on your ICP, say a new detector, how do you know the method will still be valid?

I agree with your assessor, to start in light with your scope and not try to have everything you do put under the scope of your cert, at first. One of the 3rd party labs I use did the same thing, over 10 years ago. Like you, they did myriads of different testing types, working with metals, coal, and brick. It worked out well for them.

I like the reply Joe gave.
To add :2cents: I am currently responsible for setting up 17025 in our lab. We have implemented QC on all the methods regardless of if we are going for accreditation for the method or not. We will be seeking accreditation on release parameters, regulatory parameters and food safety.

Earlier this year I posted a comment regarding calibration frequency. In essence everything you do for 17025 is aimed at proving that you have control over your technology and processes. We may have overdone it a bit/ or have we? ( No cost can be placed on your reputation - Poor results can ultimately kill customers).

17025 speaks of competency and not simple compliance.

We are in the fortunate position of having masses of certified reference materials. This allows us to run a reference sample with each batch. Is it necessary? - If you can prove that your processes are stable - reduce your frequency to once per week or every two weeks. ( you have proof that things are under control). Remember that running a reference sample without evaluating the data statistically - repeatability and Reproducibility - has no meaning. I can provide you with a statistical spreadsheet in excel that was put together for this purpose. (You may have to validate the calculations in the spreadsheet to satisfy your inspection authority)

Something else we have implemented is what we term an "internal control plan" (ICP). For each piece of equipment we have constructed a set of worksheets describing what needs to be monitored, when, what is the specification, what must be done in case of deviations etc. A record is then kept of all parameters monitored - Proof that we have control over the equipment and we can id possible problems before they can affect our results.

Ultimately the decision lies with you on how much or how little. Can you PROVE (as in court) that you have everything under control?

Useful answers from Joe Cruse and Zamclachia.

17025 cannot specify QC frequency in detail. "Appropriate for the type and volume of work" is about as far as it goes.

The requirements are in clause 5.9 which I have paraphrased.

The mandatory shalls are:


have procedures capable of detecting trends
document a plan and review it
define acceptance criteria
analyze data and take corrective action if outside limits


You may (optional):


use CRMs if possible and/or secondary reference material
participate in external QA/proficiency testing (but if not then need documented reason)
repeat tests
retest old samples

Voodoo Queen,

There are two things to look at here.

First, the FDA requirements, whether that be under Part 11 or some other area. There, others have replied.

Second, the requirements for accreditation to ISO/IEC 17025:2005, and the specific requirements of your chosen AB (seems to be A2LA).

For ISO/IEC 17025, Clause 5.9 is the Assurance of Quality requirements specifically. Under that, International Laboratory Accreditation Cooperation (ILAC) and Asia Pacific Laboratory Accreditation Cooperation (APLAC) requirements that where proficiency tests or inter-laboratory comparisons (PT/ILC) are available, those must be used to provide the evidence of compliance with Clause 5.9.1; however, the key there is where available as they are not always.

If PT/ILC is notavailable, then you must provide evidence of acceptable alternatives. The Clause provides four more examples, but others may exist also.

Remember, this is a SYSTEMIC approach, not just a specific test, although specific tests are where the PT/ILC or other methods will concentrate in order to develop the evidence. The Assurance of Quality however is meant to cover the ENTIRE system.

Remember also, the results of assurance of Quality MUST be reviewed in Management Review, to include past efforts and planning for future and on-going efforts. Additionally, other Clauses may be involved such as handling and preparation of sample, control of records, control of non-complying work, and corrective and preventive action Clauses.

So, the short answer is yes you must do this; the longer answer is to have a look at everything that may be involved to get to the higher standard.

Hope this helps.