"Reducing Lab Stress With
Analytical Solutions To Analytical Problems"

www.accta.com

Vol. 3, No. 9. September, 2014

 In this issue...
Writing the Perfect Method
Part 9 - QC

Writing the Perfect Method!

To view previous sections, view our
Newsletter Archive.

Part 8: QC
 
I am often asked how much QC is necessary.  I usually reply, "enough" or "as much as you need to do."  Of course, this is a worthless answer!  But it does point out the fact that how much QC you do depends on a number of factors, including:

• How important is this analysis?
• How important is the answer (in terms of accuracy and precision)?
• How stable is the instrument response?
• Who will use the information and do they have expectations on the value of the number?
• What is the risk if the analysis fails?  Do you only have enough sample for one injection?  Is the sample not stable? 

Why is QC required?  Without getting into a philosophical discussion, I will say that for me, you must calibrate or check your calibration at "regular" intervals.  How often will depend on the factors listed above.  However, remember these two concepts:

• Your results are only valid if you have verified that your system is within expected calibration limits.
• If a calibration check fails, then every analysis since the most recent successful calibration is now suspect.  If you only check calibration once a week, you could lose up to an entire week of data.  Can you afford this risk?

Do:

• If this is a very important analysis, then you should probably update your calibration every day.   This may include multiple injections at a single injection or injections at multiple levels.  This procedure takes longer, but is the most rigorous, since the calibration is update for today's analysis.  Most pharmaceutical formulations are analyzed this way.
  
• If this analysis is not as critical, or calibration is a lengthy procedure, then it is usually sufficient to check/verify your calibration.  Inject a single standard, maybe in duplicate or triplicate, and verify that it analyzes at the expected amount (+/- some expected tolerance).  The tolerance should be less than the difference that you want to measure (e.g.., if a 20% difference is important (trace analysis situations), then the tolerance should be around 5-10%).  You can also set this value based on injection reproducibility.
  
• Always include an injection of blanks. Both reagent and system may be necessary.
  
• If you have many samples, or are concerned about calibration stability, or if the analysis requires much time, then insert additional check standards during the analysis. 
  
• Always end with injection of one or more calibration checks.  You must verify the calibration at the conclusion of the analysis.
  
• System suitability is an important part of validated methods, since this step verifies the performance of the instrument.  The amount of effort required will be related to the importance of the results, but at least some evaluation is required, even if it is only related to retention times.
  

Do Not:

• Ignore the need to verify system calibration with every analysis set.  It may not require more than one or two injections, but it can be a simple procedure to enhance the quality of your data.
  
• Forget to calibrate at the end of the analysis set.  You run the risk of having to lose all of your data, if the next calibration check fails.
• Do not let the analyst decide on how many replicates to inject, or how often to run check standards.  Demanding procedures are not required, but something must be specified in the method.
  

Next Month: Calculations and Reporting


Would you like us to evaluate your methods, or comment about this topic?  Contact us.

Quote for the Day

"Ascertaining if a process is capable of meeting standards is nothing more than using collected data to decide if the calculated upper and lower control limits of a process, which is stable and in control, lie within product specifications. If a process is not capable, then before SPC [Statistical Process Control] can be implemented, management, MANAGEMENT, must take action to correct the process. If management cannot or will not act to correct the process then they can expect no improvement in the quality of output from any quality improvement program, much less from SPC."

– Roger W. Berger and Thomas H. Hart,
Statistical Process Control: A Guide for Implementation,
ASQC Quality Press, Milwaukee, WI (1986)

Dowload Recent Presentations

Get your own copy of two recent presentations.  Both cover our recent work on HILIC columns.  These publications describe a simple way to evaluate multi-mode mechanisms on HILIC columns.

Pittcon, March, 2014:

Mixed Mode Mechanisms in LC: Curse or Cure?

Minnesota Chromatography Forum, May, 2014: