Upsala J Med Sci 95: 261-264, 1990 

Quality Specifications 
P. Hyltoft Petersen' and Carl-Henric de Verdier' 

Deparlments of Clinical Chemistry at hospitals in 'Odense och 'Uppsulu 

Use of analytical components in clinical strategies are 
decided without paying much attention to the quality of 
analytical performance, e.g. from Danish recommendations: 'in 
non-insuline dependent diabetics the clinical goal is to keep 
HbA,, < 7.5 percent (1)' and 'guidelines for treatment of 
patients with P- Cholesterol (total) between 7 and 9 mmol/L ( 2 )  I .  
Influence from both biological and analytical variations seems 
to be forgotten or ignored, although the significances of these 
factors have been documented ( 3 , 4 ) .  A l s o  sampling error and 
other preanalytical factors should be considered. 
These are just examples, but they confirm the need for 

continuation of two previous NORDKEM projects on quality 
requirements (5) and quality control (6). 

Clinical strategy 

Sampling and other 
preanalytical errors -1- biological variation 

Analytical quality specifications 

The needs for analytical quality are different for various 
clinical situations as demonstrated for TSH- measurements, where 
the quality needed for basal S-TSH determinations used 'to 
predict S-TSH response to TFU-II is extremely demanding (7), 
whereas the requirements for B-TSH measurements in screening for 
congenital hypothyroidism are very loose ( 8 ) .  In these examples 

261 



the analytical procedure used for obtaining the quality required 
for 'predicting the S-TSH response' is very costly and should ' 
not be used for the screening of all newborns, where a simpler 
method is sufficient. 
The analytical quality needed should be specified for each 

clinical strategy and the most demanding specifications should 
be aimed at when new methods and equipments are introduced - 
also when control systems is designed (6). 
In practice costs, turn around-time, and other factors may 

be forgotten, interfere in obtaining the needed quality must not 
and should be the goal for later improvements. 

Analytical quality specifications 

Cost+ +Others 1 
Laboratory quality specification 

From the clinicians point of view the sources of variations 
and errors may seem less interesting, but for clinical chemists 
the splitting up of analytical variation and errors into random 
and systematic deviation is important. This will give the key 
for improvements of the quality, and by deviding the systematic 
deviation into bias (the common deviation from the 'conventional 
true value' during stable performance) and systematic error (the 
common deviation from accepted bias in the Laboratory) the 
possibilities for improving quality are reformed. 
Analytical imprecision is mainly determined by the analytical 

principle (e.g. RIA), pipetting and the equipment (e.g. 
photometer). Today many instruments meet most requirements for 
imprecision. This fact, however, does not help the problem with 
systematic error and bias. Systematic error may be intermittent 
(e.g. occational wrong performance) or persistent (e.g. batch- 
to-batch variations in kits). The bias is often caused by poor 
standardization or unspecific methods. 
The analytical quality specifications are different for 

262 



monitoring, imprecision and systematic errors are the main 
factors to be considered, whereas bias (and persistant 
systematic error) is determining in screening and diagnostics - 

recommendations are specified for monitoring 
in relation to a fixed concentration. 
Unspecific reactions and matrix effects are often ignored in 

the specifications of required quality. These factors will also 
be overlooked in the majorities of c o n t r o l  programmes, butthey 
are important and may result in considerable errors, especially 
in specimens with extreme compositions. 

and when general 

ANALYTICAL BIAS (INACCURACY) 

Calibration Matrix 
effects 

Unspecific Performance 
reactions (PRINCIPLE. EQUIPMENT) 

PROCEDURE PHYSlCAl 
(MODEL) 

UNKNOWN AQUIRED 
COMPONENTS 

The unspecific reactions are different from component to 
component so they should be specified for each analytical 
component, e.g. for S-Triiodothyronine it should be specified 
that a S-Thyroxine concentration of 200 nmol/L was not allowed 
to increase the S- Triiodothyronine result by more than 0.1 
nmol/L. 

contain a list og maximum allowable 
The analytical quality specifications should therefore, 

1) imprecision 
2) systematic error 
3 )  bias (mainly standardization) 
4 )  matrix effects (specified) 
5) unspecific reactions (specified). 

263 



Models for evaluation of imprecision, systematic error, and 
bias are available (3,4,5), but they should be reinvestigated 
in order to clearify whether they are still valid or they should 
be improved. Moreover, models for matrix effects and for 
unspecific reactions should be developed in this project on 
'medical need for quality specifications in laboratory 
medicine'. 

REFERENCE: 

1. 

2. 

3. 

4 .  

5. 

6. 

7. 

8. 

Arbejdsgruppe nedsat af Dansk Selskab for Intern Medicin: 
Ikke-insulinkrzvende diabetes mellitus, Diagnostic og 
behandling, 1988. 
Arbejdsgruppe nedsat af Dansk Selskab for Intern Medicin: 
Hyperlipidzmi, 1985. 
Lytken Larsen M. et al. A comparison of analytical goals 
for Haemaglobin Alc assays derived using different 
strategies. (Submitted for publication). 
Hyltoft Petersen P. et al. Influence of analytical quality 
and preanalytical variations on measurements of cholesterol 
in screening programmes. Scand J Clin Lab Invest, 1990;50 

H0rder M (ed.) Assessing quality requirements in clinical 
chemistry. Scand J Clin Lab Invest. 1980;40 suppl. 155. 
de Verdier C-HI Aronsson TI Nyberg A (eds.). Quality 
control in clinical chemistry - Efforts to find an 
efficient strategy. Scand J Clin Lab Invest, 1984;44 suppl. 
172. 

Wide L, Dahlberg PA. Quality requirements of basal S-TSH 
assays in predicting an S-TSH response to TRH. Scand J Clin 
Lab Invest, 1980;40 suppl 155:lOl-10. 
Hyltoft Petersen P et al. Studies on the required 
analytical quality of TSH measurements in screening for 
congenital hypothyroidism. Scand J Clin Lab Invest, 1980;40 

suppl. 198:66-72. 

suppl. 155:85-93. 

Correspondence: 
P. Hyltoft Petersen, 
Department of Clinical Chemistry, 
University Hospital, DK-5000 Odense C, Denmark 

264