Another question from the Cut Your QC webinar in September:
"How do we account for issues with the controls themselves (i.e. real life stability is shorter than manufacturer's recommendation)?"
Of course, this is the nightmare of the QC professional: if the controls introduce more variability, the signals from QC are not completely reflective of the method performance.
The classic terminology is "matrix effect" and "commutability." Commutability is good, it's the ability of a control to replicate the behavior of a real patient sample. Anytime there is a difference between how a control behaves and a patient sample would behave, that's the matrix effect. Not the cool matrix with Keanu Reeve and slow-motion bullet ballets, either.
When we use the imprecision data from our controls to calculate a Sigma metric, then we build our assessment assuming (hoping) that the control is as commutable as possible. If there is a matrix effect, our Sigma metric is impacted. Since the typical matrix effect is negative, our Sigma metric might be lower than usual based on the heightened imprecision of an imperfect control. If our bias is also based on controls (control mean vs peer group mean), then our dependency is doubled.
If we know that our controls are seriously compromised by matrix effects, calculating a Sigma metric off of them is not a great idea. A control with behavior greatly distorted from patient sample behavior may not even be a good choice for regular QC.
If you have matrix-distorted controls but still want to calculate a Sigma metric, you can look into using pooled patient samples as a way to determine imprecision. Bias, you'll need to find another method of determining, perhaps through EQA/PT, reference method, or reference materials.
For more on bias, check out this free article: What About Bias in Lab Measurements? https://www.westgard.com/lessons/quality-management/what-about-bias.html
For more on QC challenges: Is QC Quality Compromised? https://www.westgard.com/essays/basic-qc-practices/qc-quality-compromised.html
P.S. To the specifics of the question itself. If the manufacturer of the control says the control has a stable lifespan of X months, but in practice you see that it starts degrading at X-Y months, then that is the true effective stability period. You can complain to the manufacturer, perhaps there is a better control lot that has a better lifespan, but if not, they should provide you with lots at the right frequency. Don't use a control lot once it's begun to deteriorate - you'll be chasing out of control events that are being created by the control, not the method. If the manufacturer is wrong about the effective stability of the control, I would try to get them to provide discounts, refunds, etc.