Tools, Technologies and Training for Healthcare Laboratories

Abbott ARCHITECT c8000

This is a new one for Westgard web: we're going to provide Sigma analysis for a study where Sten Westgard was in fact a co-author on the paper. This poster was presented at the AACC 2008 conference and analyzed the Abbott ARCHITECT c8000

October 2008

[Note: This QC application is an extension of the lesson From Method Validation to Six Sigma: Translating Method Performance Claims into Sigma Metrics. This article assumes that you have read that lesson first, and that you are also familiar with the concepts of QC Design, Method Validation, and Six Sigma. If you aren't, follow the link provided.]

This is a new one for Westgard web: we're going to provide Sigma analysis for a study where a Westgard (me) was in fact a co-author on the paper. This poster was presented at the AACC 2008 conference:

ARCHITECT® c8000® Instrument Enzyme Comparability at MultiCare Health System

J. Baker,1 S. Westgard,2 C.W. Wilson,3 D. Hunter,3 S. Price,3 A. DeFrance,3 G.W. Osikowicz3
1MultiCare Health System, Tacoma, Washington; 2Westgard QC, Madison, Wisconsin; 3Abbott Diagnostics Division, Dallas, Texas
View the entire poster here

This study was a little different than the usual method validation studies we see. This study involved multiple instruments at multiple sites. These were instruments in the field, not just limited studies being performed. This type of data is nice to see because it's more "real world" than the typical evaluation study.

Six CC analyzers were examined across multiple sites. Enzyme assays evaluated ALT, AST, alkaline phosphatase (ALP), creatine kinase (CK), amylase, gamma GT (GGT), and lactate dehydrogenase (LD).)

The Precision and Comparison data

Two daily quality controls (BioRad Liquichek) were run over 30 days, (n = 30 per inst, 6 inst = 180), wit the data transferred via the Web with AbbottLink. Total percent CV was calculated from routine daily QC. Between-instrument bias percent was also calculated from daily QC.

Sigma metric and QC recommendation derived from and EZ Rules® 3.

Imprecision Estimates:

Bias estimates (added to the right):

Ideally, the bias is calculated at the critical decision levels for test. Here, we are assuming that the control levels (the levels where controls are run) are the critical levels.

Usually, the studies make use of the regression equation of the comparison of methods study. Here, the instrument values were compared directly at the control levels and those differences were calculated as the bias. Thus, we won't need to calculate bias at the critical decision level. We've already gotten it.

Determine the quality requirements at the critical decision level

Now that we have both bias and CV estimates, we are almost ready to calculate the Sigma metrics for these analytes. The last thing we need is the quality requirement for each method. CLIA provides the quality requirements we need and we don't even have to calculate the requirement at any particular level.

But, again, this study is unique. The labs were in Canada, which is not governed by CLIA. Instead, Abbott used the German rules as the source for quality requirements. Thus, Total allowable error (TEa%) was derived from RiliBäk (Deutsches Arztebkatt|Jg. 100|Heft 50|12.Dezember 2003) except amylase, where the quality requirement was obtained from the biodatabase from Ricós (Scand J Clin Lab Invest.1999; 59: 491-500).

Calculate Sigma metrics

Now we have all the pieces in place.

Remember the equation for Sigma metric is (TEa - bias) / CV:

For ALT level 1, (23 - 4.5) / 4.3 = 4.3

The metrics are displayed along the right columns.

A better representation of the data can be seen here:

A lot of good news! Only one of the levels is below Six Sigma (that ALT level 1). Everything else is world class.

Summary of Performance by Normalized Sigma-metrics chart and Normalized OPSpecs chart

Here's a graphic depiction of the Sigma performance of these analytes:


Again, nearly all of the analytes are in the Six Sigma "bulls eye."

What does that mean for QC? Let's take another look, this time using a Normalized OPSpecs chart:


Nearly all of these analytes could be controlled using 3s limits with 1 or 2 controls or even 4s limits with 2 controls.


Again, if this were our lab, it would be time to celebrate. Moreover, the good news is that the data here shows performance across instruments and across sites. So even more variability is incorporated into the calculations than the usual method validation study. This analysis is more "real world" than many posters get - so having excellent news here is a good sign that the instrument will work well in many laboratories.