The standard deviation will be larger if the data are spread out and smaller if the data are closely clustered about the mean. SD ranges that are either too narrow or too wide can be problematic. Control limits that are too narrow can lead to a high rate of false rejections (values being rejected when the system is in control), whereas limits that are too wide may allow control values to be accepted when the system is actually not in control.
Collecting data over too short a period of time will give too small an estimate of standard deviation. The longer data is collected, the better, because the calculations will include such variables as
- method changes;
- performance achieved before and after maintenance;
- different operators (note that it is important to periodically rotate QC testing among operators);
- changes in reagent lots; and
- changes of sample probes or pipettes.
Control limits for unassayed controls are determined by running replicate analysis in parallel with controls having previously determined ranges. The larger the number of replicates, the greater the confidence in the ranges that are established.