Quality specifications for total error allowable

Analytical variations may increase the number of results in a healthy population outside a previously established reference interval A positive analytical bias increases the number of high values and a negative bias decreases the number below the lower reference limit. It is possible to link the maximum tolerable analytical bias to biological variability. The relationship below is derived on the assumption that the maximum acceptable bias is less than one quarter of the sum of the intraindividual and interindividuai variations within a population. 

By substituting different factors in the above formula different targeted analytical biases can be identified. If less analytical bias is deemed to be appropriate (e.g., one fifth of the total biological variation instead of one quarter), 0.20 would be substituted for 0.25. It has been suggested that analytical precision should be less than one half of the mean intrain-dividuai variation CVa = 0.50 CVj), Total error is derived from the smn of imprecision and bias. From the formula above 

The factor 1.65 is derived from 90% of the distribution of results excluding the highest and lowest 5% of values. Again, substituting different values for 0.50 and 0.25 above allows different goals to be established. 

Other formulas have been used to determine quality specifications for bias. Generally, analytical bias should be less than one quarter of interindividuai biological variation. 

125 for 0.250 in the above formula and minimum performance defined by substituting 0.375 for 0.250. When the factor 0.125 is used, 1.8% of values are outside one reference limit and 3.3% outside the other. When 0.375 is used 1.0% values are outside one reference limit and 5.7% outside the other. Fraser has stated that the lower number should be used when desirable performance standards are easily attained and 0.375 should be used when performance goals are not readily attainable with current procedures. Most of the literature has reported total error in terms of the sum of bias plus 1.65 times the precision (TEa Bias + 1.65 SD or CV of interindividual variation). 

---