1. Validation, Verification and Quality Control
Thomas Novicki Ph.D. D(ABMM)
Marshfield Clinic
Marshfield WI

2. Today’s Topics Validation and Verification of ID and AST systems
Quality Control (QC) of Identification (ID) and Antimicrobial susceptibility test (AST) systems
(We will not be discussing proficiency testing due to time constraints)

3. Disclosures
I have nothing relevant to this presentation to disclose

4. Validation and VErification

5. Validation and Verification
Validation: Documenting in vitro diagnostic device performance in your lab
Verification: An ongoing demonstration of the vendor’s performance claims using a variety of tools

6. Validation and Verification (or is it Verification and Validation…?)
Note that some entities (e.g. CLSI, CLIA) reverse the definitions
Initially Verify
Then periodically Validate
The definitions used don’t matter, but their consistent use in your lab is important.
What does your accrediting body use?

7. Who Guides Us? Vendors of commercial systems
The Clinical and Laboratory Standards Institute (CLSI)
The Feds
CLIA – Clinical labs
FDA – Instruments
Your accrediting body (COLA, CAP, JC)
(Did I miss anything…?)

8. Validation and Verification
The Centers for Medicare and Medicaid Services (CMS), (a part of the US Department of Health and Human Services) regulates clinical laboratory testing through the Clinical Laboratory Improvements Amendments (CLIA) that became effective in (Recall that it all began in 1988)

9. Validation
For FDA cleared and approved tests, the lab demonstrates performance claims by the device manufacturer, typically found in the product label (‘package insert’)
For lab-developed tests (LDTs), a more rigorous set of studies demonstrating performance is necessary

10. Validation of LDTs
Will not discuss further except to say that any modification of an FDA IVD makes it an LDT
Be wary of modifying the more critical FDA-approved test (e.g. TB, HIV IVD devices )
Qualitative tests easier to modify than quantitative
Simple cleared tests (e.g. MRSA chromo-agars for additional anatomical sites) are reasonably modified

11. iClick? 1
CLIA requires a lab to validate which of the following for every new diagnostic test
Accuracy
Precision
Reportable and Reference ranges
All of the above
1 and 2

12. iClick? 1 - Discussion
CLIA specifies that the user of an FDA IVD device of Moderate/High complexity will
Validate accuracy,
Precision,
Reportable range, and
Reference range
You must satisfactorily complete validation of the device before beginning its use.
You must document all validation study data and maintain it for at least 2 years after test retirement.

13. Validation – AST Two methods for an unmodified FDA-cleared instrument
Compare to a reference method
Compare to an existing system
Method B probably the best fit for most labs with an existing system For method A, see Cumitech 31A
(Cumitech 31A, Verification and validation of procedures in the clinical microbiology laboratory, ASM Press 2009)

14. Disclaimer What I’m about to say on validation…
‘… is more what you’d call guidelines than actual rules’

15. Validation – AST
Source of samples/isolates to test
> 30 well-characterized clinical isolates per panel that reflects the mix common to your population and also some with unusual AST patterns
Can include proficiency or commercial validation panels, but be wary:
Reflective of your population?
Incorrect sample matrix?
If from the device manufacturer, a self-fulfilling prophecy?
Note that with care a single panel can be used for both AST and ID validations

16. Validation – AST
How to deal with discrepancies when two non-reference systems are compared?
 No reference method, so grade results, then apply predetermined limits on the number and types of discrepancies

17. Validation – AST For each bug/drug result, calculate %s for…
Categorical (S/I/R) Errors
Minor Error I vs S or I vs R
Major Error S vs R or R vs S
No Very Major Error category: ‘True’ result is not known
Essential Errors (> + 1 MIC dil. difference)
Can only score when both methods generate discrete numbers (i.e. do not score < or > values)

18. Validation – AST
Using the total number of evaluable bug/drug data points, an acceptable validation has:
< 5% Major Errors AND
< 10% Major + Minor Errors AND
< 10% Categorical Errors AND
< 10% Essential Errors
Also look for trends in a particular drug or bug which may point towards a problem area
Consult your vendor rep with any validation problems

19. Validation – AST Precision (reproducibility)
Test five isolates X3 for three to five days
Calculate intra-and inter-run categorical and essential % agreements
Isolates should have known susceptibility profiles, such as ATCC strains
Use both Gram positive and Gram negative strains if validating GP and GN panels
Acceptable precision > 95%agreement

20. Validation – ID
Test a range of Gram positive and Gram negative organisms ID’d by reference or current method(s)
Minimum of 20 patient isolates; larger labs test more. Additionally, test QC strains
< 10% discrepancies between reference and new IDs

21. Validation – ID Level of agreement may need to be adjusted e.g.
Old ID of ‘Coag-negative Staphylococcus sp’
New instrument ID’s to species level
Precision
Test 2-4 isolates X3 for three to five days
Calculate % intra- and inter-run agreements
Accept > 95% agreements

22. Validation
What to do if validation fails? Do not use the new test and either…
Withdraw test from further consideration OR
Develop and implement a corrective plan with the manufacturer. Then, re-validate

23. Verification – Component of the QA Process
An ongoing, multi-faceted process that assures the validated test continues to perform at the expected level
Personnel competency
Proficiency challenges
QC organisms or analytes
Comparison with other labs
Instrument PM/calibration
Trend analysis
Discrepant resolution
Pre- and post-analytic controls
Written procedures documenting the process

24. I think, therefore I control quality
Quality Control
I think, therefore I control quality

25. iClick? 2
You have an instrument that uses micro-broth dilution MIC plates in your lab. This system incubates and reads plates automatically but also has a manual plate reader. Your tech rep tells you to read a plate manually if QC fails on the instrument. If QC then passes, your QC is acceptable for the entire system and no follow-up action is needed.
True
False

26. iClick? 2: discussion
QC is performed to assure the accuracy of a test result by assessing all analytical components
Reagents
Instrument (where applicable)
Test precision (i.e. repeatability)
Operator technique
Data interpretation

27. CAP on AST QC
‘For AST of either disk or dilution type, control organisms are tested with each new lot or batch of antimicrobials or media and each day the test is performed thereafter.’ (Note that >3 failures per month per drug/bug combo is unacceptable)

28. CAP on AST QC
‘However, the frequency of test monitoring may be reduced to weekly …if the laboratory can document satisfactory performance with daily control tests as suggested by CLSI guidelines.’ The CAP checklist then goes on to briefly describe how to do so

29. AST QC Resources
COLA, JC have similar verbiage, but is the checklist guidance enough?
IMHO*, No!
You need CLSI AST documents in your lab.
(*In my humble opinion.)

30. CLSI AST Documents
M2 (disk diffusion) and M7 (MIC), 3 year cycle
Test performance
 QC (including daily-to-weekly QC) 
M100, annual cycle
Recommended drugs to test and report, interpretive breakpoints
 QC strains, QC troubleshooting, when to re-validate QC 
Others documents for other classes of microorganisms
(Explore these and more at choose Microbiology from the Shop drop down menu)

31. iClick? 3 How many of you have these CLSI documents in your lab:
2012 M2 (disk diffusion), M7 (MIC) and M100 (Breakpoints, QC)
2012 M7 and M100
2012 M2 and M100
2012 M2 and M7
Any older versions of these documents
None of the above

32. Conversion to Weekly AST QC
When day-of-use QC is being successfully performed at least once a week, a lab may consider weekly AST QC by a validation process:
Test QC strains 20 or 30 consecutive days
If 0-1 failures per drug-bug combo in 20 days OR 2-3 failures in 30 days
THEN Weekly QC may be implemented

33. Weekly AST QC
Perform QC testing once per week and with any new reagent
If weekly AST QC fails
If readily identifiable error (e.g. wrong QC strain; see M2 or M7 for more examples) correct error and retest once. If QC passes, document findings and continue weekly QC
If no error is found…

34. Weekly AST QC Re-validation
Perform 5 consecutive days of QC on the failed drug-bug combo
(We automatically take out a fresh QC strain from the freezer at this point)
If ALL 5 days pass, document and resume weekly testing
If a failure occurs again, STOP TESTING and thoroughly evaluate the process for error
New QC strain
New lot of reagent
Correct process errors
Consult the manufacturer

35. Weekly AST QC Re-validation
Once a likely error has found, re-validate
Do not include data from the prior 5-day sequence
Must again pass these criteria
0-1 failures per drug-bug combo/20 days OR
2-3 failures per drug-bug combo/30 days
Remember that, during re-validation the lab is on a daily QC schedule-a failure means that that drug’s results are not released

36. AST QC References Disk diffusion MIC M02-A11 Section 15
M100-S22 Tables 3C, 3D
MIC
M07-A9 Section 16
M100-S22 Tables 4F, 4G

37. AST QC Special Considerations
M100-S22 Tables 3C & 4F, ‘Reference Guides to QC Frequency’
When and how much re-validation to perform when a system change occurs (e.g. AST instrument repair, Abx formula change & Etc)
Release of patient results when QC fails?
CLSI says OK with many qualifiers, but our lab generally just says No

38. iClick? 4
You are performing a weekly AST QC re-validation for drug X on your favorite automated ID/AST instrument when you incur a failure on the same AST panel for drug Y. What do you do?
Ignore the failure of drug Y and press on with the 5-day testing of drug X
Now treat the failure of drug Y as a second failure and start a 5 day test of drug Y
Stop testing, evaluate the whole process, then perform a new 20/30 day validation on the whole AST panel

39. iClick? 4 discussion
The correct answer isn’t clear, but Marshfield Labs treats the failure of the second drug as a failure and starts a 5 day test on that drug also Why?  Subsequent failures may reflect a larger problem

40. ID QC
CLIA requires a lab to show positive and negative reactions for each new lot and/or shipment of biochemical.  This includes kit and automated instrument ID methods!

42. ID QC
There is a QC solution for automated ID instrument users:
Streamlined QC
Stay Tuned for the Next Talk!

43. The End
Thank you…
Questions?