Intrinsic and routine quality of serum total potassium measurement as investigated by split-sample measurement with an ion chromatography candidate reference method

^a Address correspondence to this author at: Fax 32-9-264 81 98; e-mail Linda.Thienpont{at}rug.ac.be.

	Abstract

Top Abstract Introduction Materials and Methods Results and Discussion Summary and Conclusions References

 
We evaluated the intrinsic quality of eight routine test systems for the measurement of serum total potassium (K⁺), as well as the routine quality of four of these systems, using a group of 60 single-donation serum samples that had been certified with an ion chromatography reference method. The intrinsic quality of the tests was evaluated by analysis of the sera in the manufacturers' application laboratories under strict internal quality control. The routine quality was evaluated by analysis of the same sera in five (per system) routine laboratories under daily working conditions. The results of the study were interpreted in light of the most stringent specifications derived from the biological variation of K⁺, which require limits of 6.3% for total error and 1.6% for systematic error. The study revealed that the intrinsic quality of all systems was excellent. None of the test systems yielded a substantial number of results outside the 6.3% total error limit, and only one test system exceeded the 1.6% systematic error limit. The majority of the routine laboratories reproduced the manufacturers' intrinsic quality. In particular, most laboratories satisfied the 6.3% total error limit. However, several laboratories exceeded the 1.6% systematic error limit. Generally, there was a considerable difference in quality between the participating laboratories. This showed that the major problems for serum K⁺ analysis (for samples with no unusual matrices and with concentrations within the reference interval) are at the routine laboratory.

	Introduction

Top Abstract Introduction Materials and Methods Results and Discussion Summary and Conclusions References

 
In clinical chemistry, the importance of accurate (in the analytical sense) routine test systems has long been recognized (1)(2)(3). Among the major tools for the assessment of accuracy are method comparisons between routine and reference methods, using a group of patient specimens and the split-sample design. In recent years, the latter approach has received support through legislation, through international standardization, and from the International Federation for Clinical Chemistry (4)(5)(6)(7). However, it should be realized that the accuracy of a result reported by a routine laboratory depends on the intrinsic quality of the test system used and the analytical quality provided by the laboratory. Unfortunately, the distinction between intrinsic and routine quality often is not clear-cut when evaluating test systems. Therefore, for the current study of test systems for serum K analysis, we decided to strictly distinguish between the quality that is intrinsically available at the manufacturers' facilities and the quality that is realized when the tests are performed under routine laboratory conditions.

The specimens that were used for the evaluation consisted of a group of 60 single-donation serum samples that had been certified with a candidate reference method using ion chromatography (IC)¹ (8)(9). The eight systems being evaluated for intrinsic quality used ion selective electrodes (ISEs) to measure K. The systems that used direct ISE were manufactured by AVL, Chiron, Dade (two systems), and Johnson & Johnson (J&J); the systems that used indirect ISE were manufactured by Boehringer Mannheim, Beckman, and Roche. For the Chiron, J&J, Boehringer, and Beckman systems, routine quality was evaluated in five clinical laboratories, using the same panel of samples. The results of the study are interpreted on the basis of the strictest limits for total error (TE, 6.3%) (10) and systematic error (SE, 1.6%) (11) of serum K analysis, derived from the biological variation of the analyte.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results and Discussion Summary and Conclusions References

 
IC REFERENCE METHOD
The IC reference method was used as described previously (8)(9), with the following modifications. The eluent was a 7 mmol/L aqueous H₂SO₄ solution, and front cut was applied (12). The IC measurements were performed as two independent duplicates on two different days, performed over the course of 9 days and on two different instruments (DX100 and DX500, Dionex). The internal quality control (IQC) samples were standard reference materials (SRMs) 909b1 and 909b2 from NIST (Gaithersburg, MD). The SRMs were measured at the beginning, in the middle, and at the end of each assay. The quality-control criteria to satisfy were a maximum overall CV of 1.5%, a maximum SE of 0.65%, and a maximum TE of 1.6% (13)(14).

test systems performed by or under the responsibility of the manufacturers
The analyses in the application laboratories were performed in duplicate in one assay under strict IQC. The manufacturers used their own IQC samples at two or more concentrations. The IQC samples were placed at the beginning of the assay (n = 6), after each tenth duplicate (n = 3), and at the end of the assay (n = 6). Five test systems that used direct ISE were investigated. The measurements with the AVL system (AVL Medical Instruments) were performed at the site of the Belgian distributor, Merck-Belgolabo, using an Omni system and the Combitrol 1–3 (AVL) IQC samples. The measurements with the Chiron system (Chiron Diagnostics) were done in the Chiron application laboratory at Zaventem (Belgium) with a Chiron 654 analyzer and the Certain Lyte Level 1–3 (Chiron Diagnostic) IQC samples. The measurements with the two Dade systems (Dade International) were performed in the Dade application laboratory at Munich (Germany), using a Dimension AR system (direct ISE) and a Dimension RXL instrument [direct integrated multisensor technology (IMT)] analyzer. For IQC on both instruments, Monitrol I and Monitrol II (Dade) samples were used. The measurements with the J&J system (Johnson & Johnson Clinical Diagnostics) were done in the J&J application laboratory at Illkirch (France) using a Vitros 250 analyzer and the X1395 and Y1397 (J&J) IQC samples.

The test systems below used indirect ISE. The measurements with the Boehringer system (Boehringer Mannheim) were performed in the manufacturer's application laboratory using a Hitachi 917 instrument and the PNU 186720, the PNU 189638, and the PPU 188184 IQC samples. The analyses with the Beckman system were done at the site of the Belgian distributor, Analis, using a Synchron CX7 analyzer and the Decision Level 1–3 (Beckman) IQC samples. The analyses with the Roche system (Roche Diagnostics) were performed by the clinical routine laboratory of the hospital, Maria Middelares, in Ghent (Belgium), using the Integra instrument and the Roc SN and SP (Roche Diagnostics) IQC samples.

routine analyses performed in clinical laboratories
The routine analyses with the Beckman, Boehringer, Chiron, and J&J test systems were performed in Belgian clinical laboratories (five laboratories for each system). The measurements were done under conditions strictly identical to those for measurement of patient samples, i.e., in singlets and without any additional calibration or IQC precautions. Measuring the samples in a single assay was not mandatory. We guaranteed the anonymity of the participating routine laboratories.

The instruments and the reference intervals they applied are listed in Table 1 (this table also includes the data from the manufacturers).

serum samples
The native samples were single donations from 60 healthy persons (30 males and 30 females). They were purchased from WBAG Resources. Fifteen donors were younger than 20 years, 5 were older than 50, and the other age groups were 20–30 years of age (n = 15), 30–40 years of age (n = 15), and 40–50 years of age (n = 10). The blood was allowed to coagulate for 1 h after sampling, after which the serum was isolated by centrifugation. The resulting fresh sera were stored at 4 °C for 3 days. They were then sterile-filtered, fractionated into 500-µL portions in Eppendorf vials, and frozen. The sera were kept at -20 °C for 2 weeks and sent on dry ice to the IC laboratory in Ghent. The samples were treated according to the instructions of the Ethical Commission of the University of Ghent. All samples were sent on dry ice from the IC laboratory to the laboratories participating in the study. When the samples arrived, the person designated as responsible for the study checked whether the samples were still frozen, which was always the case. The samples were kept at -20 °C until analysis. Values for total protein, cholesterol, bilirubin, and triglycerides were determined by WBAG Resources on a Hitachi 717 instrument; almost all were within the reference range. Sera with suspected interferences were not selected.

calculations
Outlier removal.
To prevent large errors from influencing the method comparisons, we treated values outside the 2.5-SD limit as outliers. The statistical chance to observe values outside this limit is 1%; in our case, this corresponded to ~1 of 60 values; in practice we rejected a maximum of three values to avoid "overfitting" the data.

Imprecision.
For IC, we calculated the within-day imprecision (CV_wd) from the deviations of the within-day duplicates and the mean K concentration according to the formula below:

We calculated the between-day imprecision (CV_bd) with the same formula but using the deviations of the between-day duplicates. In addition, the total imprecision (CV_tot,REF) of the reference method values was calculated with the following formula:

The imprecision of the analyses in the application laboratories was estimated using the differences between duplicates (see above) and the results of the IQC measurements (CV_IQC). The total analytical imprecision for the method comparisons (single measurements for the routine method vs quadruplicate measurements for the reference method; CV_tot,COMP) was calculated with the following formula:

This formula was used to calculate the 95% prediction limits of variation. The CV_IQC was used to account for the total imprecision during the method comparison, including eventual drifts and/or shifts. In contrast, the imprecision estimated from the differences between duplicates (CV_wd) did not reflect shifts or drifts because they were performed immediately after each other.

For the method comparisons for the routine laboratories, no imprecision data were available. Therefore, the corresponding CV_tot,COMP calculated for the application laboratories was used.

SE.
To estimate the SE of the test systems performed at the manufacturer and at the routine laboratories, we calculated the mean deviation of the results from IC and the 95% confidence interval. Because the routine laboratories only measured the sera in singlets, we also used the singlet results from the application laboratories. In addition, we used linear regression to calculate the deviation for the sample with the lowest (3.56 mmol/L) and the highest (5.42 mol/L) K concentration. These deviations were then evaluated against the SE limit proposed by Fraser et al. (11) for routine test systems for K (1.6%).

data presentation
The results of the method comparisons are presented in the form of bias plots. As indicated above, we used only single measurements (even for the application laboratories, where the samples had been measured in duplicate) for the method comparisons. This allowed us to evaluate the performance of the test systems against the limits for TE, which also referred to a single measurement. Therefore, in all data presentations, the y-axis represents the percentage of the deviations of the test system singlets from the mean of the IC quadruplicates. The respective values (mmol/L) measured for the 60 samples by IC are plotted on the x-axis. The plots also contain the regression lines for the deviations, together with the 95% prediction limits calculated as 1.96 x the CV_tot,COMP. Finally, the strictest TE limit for serum K analysis (TE = 6.3%), as proposed by Ricós et al. (10), was included.

	Results and Discussion

Top Abstract Introduction Materials and Methods Results and Discussion Summary and Conclusions References

 
IC REFERENCE METHOD
All analyses were performed according to recently proposed specifications for serum K reference measurement, namely, a maximum overall CV of 1.5% (here for the quadruplicates), a maximum SE of 0.65%, and a maximum TE of 1.6% (13)(14). The limit for TE, however, was specified for 12 measurements. Because only four measurements were performed in our study, the TE limit was changed to 3.0%.

The SE and TE were controlled by use of the NIST 909b reference materials. The observed deviations were 0.20 ± 0.25% for the SRM 909b1 (certified value, 3.424 mmol/L) and 0.27 ± 0.18% for the SRM 909b2 (certified value, 6.278 mmol/L). Both values satisfied the 0.65% limit for SE cited above. Furthermore, neither the 1.5% CV limit (this also holds true for the 60 serum samples) nor the 3% TE limit for the quadruplicates were exceeded. The total imprecision, CV_tot,REF, was 0.6%. From these results, we can conclude that the reference method was under control during the study.

intrinsic quality
Internal quality control.
During the study, a strict IQC protocol was imposed on the respective application laboratories. At least two IQC samples (with low and high K concentrations) were measured in blocks at the beginning of the assay (n = 6), after each tenth duplicate (n = 3), and at the end of the assay. All manufacturers used their own IQC samples, each of which had a method-dependent target value. The mean deviations from the respective target values for the IQC samples with low, medium (if applicable), and high concentration are represented in Table 2 . For our study, we aimed for a maximum deviation from the target (2% of the mean) of all IQC results per sample. However, this limit could not always be realized in practice. The low-concentration IQC samples of three of the manufacturers (AVL, Beckman, and Chiron) had mean deviations >2.0%. These deviations had previously been observed during experiments preliminary to the study. However, the manufacturers claimed that this would not negatively influence the results of the analyses of these samples. In the calculation of the mean AVL deviations, the first IQC block was not taken into account because the results for all samples showed a pronounced negative bias (-5.56%, -3.23%, and -2.28%, respectively). The analyses were not repeated because the system was stable after this first IQC block, and the deviation was not reflected in the sample measurements.

Imprecision.
Imprecision was calculated using the IQC results (CV_IQC) and using the differences between duplicates (CV_wd); these values are shown in Table 3 . Interestingly, for Boehringer and J&J, the CV_IQC values were substantially higher than the CV_wd values. This was due to a system drift over the total run in the order of 2% for Boehringer and to a continuous fluctuation of the IQC values in the same magnitude for J&J. The influence of the system drift on the duplicate measurements was negligible because both samples immediately followed each other. This observation led us to calculate the CV_tot,COMP values (see above) from the CV_IQC and not from the CV_wd data, because the total imprecision of the routine results was involved. For the Dade ISE test, the CV_IQC and CV_wd values were similar and extremely low, indicating a high repeatability and a negligible drift during the study. Imprecision was also very low for the Chiron, Roche, and Beckman systems (the CV_IQC and CV_wd for all three systems were 0.5%).

Method comparison: systematic error.
For each test system evaluated in the application laboratories, the deviation (%) of the singlet results for the group of sera from the results (mean of quadruplicates) obtained by the IC reference method was calculated. From these data, we derived the mean deviation to estimate the systematic error of the test systems in comparison with IC (Table 4 ). Because of a concentration dependency of these deviations, the deviations of the samples with the lowest (3.56 mmol/L) and the highest (5.42 mmol/L) K concentrations were also calculated. When comparing these deviations with the most stringent limit for SE of 1.6% (11), one has to consider that the reference method was performed under the condition of a maximum SE of 0.65%. Therefore, the 1.6% limit has to be extended in our case to 2.25% (1.6 0.65%). Only the Dade ISE system exceeded this limit and showed a constant bias of ~3%. However, the AVL, Beckman, and Chiron systems showed a slight undercalibration in the low concentration range, and the Dade IMT system showed a slight undercalibration in the high concentration range. Interestingly, this fact is also visible in the respective IQC data (Table 2 ), which stresses the importance of adequate IQC.

In summary, with the exception of the Dade ISE system, all of the investigated test systems were well calibrated in the reference range.

Method comparison: total error.
The method comparisons are shown in the form of a bias plot (see Materials and Methods) in Fig. 1 . As explained above, these bias plots were constructed from singlet results. This was purposely done because in this way, the method comparisons directly give an impression about the magnitude of the TE. Therefore, Fig. 1 also reflects the high intrinsic quality of the investigated systems. None of them gave a substantial number of results outside the strictest TE limit of 6.3%. Nevertheless, the slightly poorer performance of the Dade IMT system (compared with the other systems) can also be observed in the bias plots. In particular, the variation around the regression line is considerably greater in the Dade IMT system than in the other test systems. This variation can originate from two sources, namely, from the total analytical variation of the reference and routine methods (CV_tot,COMP) or from sample-related effects (15). To distinguish between the two sources, we included the 95% prediction limit for variation caused by the CV_tot,COMPin Fig. 1 . For the same purpose, we compared the predicted variation; with the observed variation, the latter was obtained from regression analysis (S_yx; Table 5 ). To make the data more comparable, we expressed the S_yx values in percentages. For the Dade IMT system, S_yx is considerably higher than CV_tot,COMP. Similarly, it can be seen in the bias plot (Fig. 1 ) that the number of results outside the 95% prediction limit is considerably greater than the statistically tolerable number of three. Both of these problems might indicate some sensitivity of the Dade IMT system to sample-related matrix effects because no drifts or shifts were observed during analysis (shifts or drifts would also produce higher values for S_yx). For the Dade ISE, the Beckman, and the Chiron systems, the values for S_yx were greater than CV_{tot, COMP}. Similarly, the number of outliers exceeded the statistically expected number (Fig. 1 ). For these systems, however, the respective CV_tot,COMP values were very low. Therefore, it seems likely that the imprecision of those test systems has been underestimated.

View larger version (43K): [in this window] [in a new window]   Figure 1. Percentages of deviation of singlets measured in the respective application laboratories from the IC results (quadruplicates). The line calculated with linear regression analysis is represented as a dashed line (- - - - - -); the 95% prediction limits are dotted lines (... ... ... . . ). The broader, gray lines are the 6.3% TE limits.

View this table: [in this window] [in a new window]   Table 5. Calculated total imprecision (CVtot,COMP) and observed variation (Syx) in the method comparisons performed in the application laboratories.

routine quality
The values for bias and S_yx observed in the study for the test systems in the routine laboratories are shown in Table 6 . For comparison, the corresponding values from the manufacturers are included. In the bias plots (Fig. 2 ), only the results of the laboratories with the best and the worst performance are shown (the prediction intervals are those of the respective manufacturers).

View this table: [in this window] [in a new window]   Table 6. Deviations from IC and Syxvalues for each application laboratory and the corresponding routine laboratories.

View larger version (44K): [in this window] [in a new window]   Figure 2. Percentages of deviation of singlets measured in clinical laboratories from the IC results (per test system only the results of the routine laboratories with the best and worst performance are shown).

For Beckman laboratories I and IV, the S_yx values considerably exceeded the values of the application laboratory. For laboratory IV, this was partly because it only could report results with two, instead of three, significant digits. For laboratory I, the greater variation was due to within-run recalibration, which produced blocks of successive samples that were biased when compared with the mean deviation. With the exception of laboratory I, all laboratories using the Beckman system produced results well within the 6.3% TE limit. The 2.25% SE limit, however, was exceeded in several laboratories, especially in the lower concentration range. The latter was caused to a great extent by the undercalibration of the Beckman system itself (Table 4 and Fig. 1 ). It should be mentioned in this connection that such low bias values as those for serum K are generally difficult to realize in practice (16). Therefore, these findings should not be overinterpreted.

Most of the laboratories that used the Boehringer system could generally reproduce the intrinsic quality of the test. The slightly higher imprecision in laboratories I and II was because they reported results with only two significant digits. However, laboratories I and II showed a bias of -4% and 3%, respectively, in the lower concentration range and thus exceeded the 2.25% limit for SE. With the exception of one result in laboratory II, all deviations were very well within the 6.3% TE limit.

For the laboratories that used the Chiron system, only laboratories I and III were able to reproduce the intrinsic quality of the test. Laboratory IV showed a mean bias of -4.8% and an increased imprecision. The latter was observed in particular among the first 15 samples. The laboratory found no reason for this, but problems with electrode adaptation can occur. The combination of bias and increased imprecision produced a high number of results outside the 6.3% TE limit. Laboratories II and V showed greatly increased values of S_yx and a pronounced negative bias in the low (-7%; laboratory II) and high (-11.5%; laboratory V) concentration ranges. The effect of the increased S_yx value and bias was a high number of results outside the 6.3% TE limit. Laboratory II, however, used a different type of electrode than the one that was used by the application laboratory (Table 1 ).

For the J&J system, the majority of the routine laboratories could reproduce the intrinsic quality of the applications laboratory. Only laboratory III showed a slightly increased value of S_yx, with the consequence that results were outside the 6.3% limit for TE. However, system calibration was a slight problem. Laboratories I and IV showed a bias of approximately -3% and thus exceeded the 2.25% SE limit.

Reference intervals.
To investigate whether the observed biases would be reflected in laboratory reference intervals, we asked the routine laboratories which reference intervals they used. Most laboratories followed the recommendations of the manufacturers and used a reference interval of 3.5–5.1 mmol/L. J&J used a reference interval of 3.6–5.0 mmol/L; however, the mean of 4.3 mmol/L was the same as the other reference intervals (Table 1 ). Four laboratories (Beckman laboratory II, Boehringer laboratory V, and Chiron laboratories III and IV) reported considerably different reference intervals. Most strikingly, their mean value was 7% higher than those of the others. However, this was not reflected in their calibration. Perhaps that they accounted in this way for higher preanalytical errors that caused increased K values through hemolysis.

	Summary and Conclusions

Top Abstract Introduction Materials and Methods Results and Discussion Summary and Conclusions References

 
The study demonstrated a high intrinsic quality in K measurement systems for serum samples with normal matrices and K concentrations. This holds true for direct as well as indirect ISE. The majority of the routine clinical laboratories could reproduce the intrinsic quality under daily routine conditions. However, the very strict limits for bias calculated from the biological variation of K are a challenge for the routine laboratories. Nevertheless, some laboratories could not reproduce the intrinsic quality offered by the manufacturer. In consequence, improvement in the quality of K analysis seems to be most needed in routine laboratories. The means to achieve this improvement are adequate system installation and maintenance and proper internal quality control.

	Acknowledgments

 
We thank the directors of the following Belgian laboratories for their participation in the study: the laboratory of the H. Hartkliniek (Roeselare), Henry Serruys Ziekenhuis (Oostende), Mariaziekenhuis (Poperinge), St.-Niklaasziekenhuis (Kortrijk), St.-Andrieskliniek (Tielt), AZ Menen Campus (Menen), O.L.-Vrouw van Lourdes Kliniek (Waregem), Kliniek Zwarte Zusters (Ieper), St.-Rembertziekenhuis (Torhout), Bijlokehospitaal (Ghent), Universitair Ziekenhuis (Ghent), St.-Jozefkliniek (Ghent), O.L.V. Ziekenhuis (Ieper), Stadskliniek (Lokeren), Algemeen Ziekenhuis Ten Bosch (Willebroek), Medilab (Ghent), and the laboratories Nuytinck (Evergem), Van Poucke (Kortrijk), and Bruyland (Kortrijk). The study was supported by the Research Fund of the University of Ghent, grant 01116692.

	Footnotes

 
Laboratorium voor Analytische Chemie, Faculteit Farmaceutische Wetenschappen, Universiteit Gent, Harelbekestraat 72, B-9000 Gent, Belgium.

¹ Nonstandard abbreviations: IC, ion chromatography; ISE, ion selective electrode; TE, total error; SE, systematic error; IQC, internal quality control; SRM, standard reference material; and IMT, integrated multisensor technology.

	References

Top Abstract Introduction Materials and Methods Results and Discussion Summary and Conclusions References

 

1. Cali JP. An idea whose time has come [Editorial]. Clin Chem 1973;19:291-293. [Web of Science][Medline] [Order article via Infotrieve]
2. Cali JP. Rationale for reference methods in clinical chemistry. Pure Appl Chem 1976;45:63-68.
3. Tietz NW. A model for a comprehensive measurement system in clinical chemistry. Clin Chem 1979;25:833-839. [Free Full Text]
4. Amended proposal for a European Parliament and Council directive (In vitro diagnostic medical devices). COM (96) 643 final/20.12.1996. Brussels: Commission of the European Union, Directorate-General Industry, 1996..
5. European Committee for Standardization (CEN), TC 140 "In vitro diagnostic medical devices", WG4. Measurement of quantities in samples of biological origin. Metrological traceability of values assigned to calibrators and control materials. PrEN first draft. Brussels: CEN, 1997..
6. International Organization for Standardization, Committee on clinical laboratory testing and in vitro diagnostic test systems. ISO/TC 212. WG 2 Reference systems. Second draft. Geneva: ISO, 1997..
7. IFCC. Working Group on Standardization of Cortisol Measurements. Preparation and certification by ID-GC/MS of a panel of serum samples suitable for use in the validation of cortisol measurement procedures [Abstract]. Medlab 97, 12th IFCC European Congress of Clinical Chemistry. Basel: IFCC, 1997..
8. Thienpont LM, Van Nuwenborg J, Stöckl D. Ion chromatography as potential reference methodology for the determination of total sodium and potassium in human serum. J Chromatogr A 1995;706:443-450. [Web of Science][Medline] [Order article via Infotrieve]
9. Thienpont LM, Van Nuwenborg J, Reinauer H, Stöckl D. Validation of candidate reference methods based on ion chromatography for determination of total sodium, potassium, calcium and magnesium in serum through comparison with flame atomic emission and absorption spectrometry. Clin Biochem 1996;29:501-508. [Web of Science][Medline] [Order article via Infotrieve]
10. Ricós C, Baadenhuijsen H, Libeer J-C, Hyltoft Petersen P, Stöckl D, Thienpont LM, Fraser CG. External quality assessment: currently used criteria for evaluating performance in European countries, and criteria for future harmonization. Eur J Clin Chem Clin Biochem 1996;34:159-165. [Web of Science][Medline] [Order article via Infotrieve]
11. Fraser CG, Hyltoft Petersen P, Ricós C, Haeckel R. Proposed quality specifications for the imprecision and inaccuracy of analytical systems for clinical chemistry. Eur J Clin Chem Clin Biochem 1992;30:311-317. [Web of Science][Medline] [Order article via Infotrieve]
12. Van Nuwenborg JE, Stöckl D, Thienpont LM. Practical aspects of suppressed ion chromatography for cations in human serum. J Chromatogr A 1997;770:137-141. [Web of Science][Medline] [Order article via Infotrieve]
13. Stöckl D, Reinauer H, Appel W, von Klein-Wisenberg A. Diskussionsvorschlag für ein einheitliches Referenzmethodenkonzept auf der Grundlage der "Richtlinien der Bundesärztekammer zur Qualitätssicherung im medizinischen Laboratorien". Lab Med 1991;15:336-339.
14. Stöckl D, Reinauer H. Development of criteria for the evaluation of reference method values. Scand J Clin Lab Invest 1993;53(Suppl 212):16-18.
15. Hyltoft Petersen P, Stöckl D, Blaabjerg O, Pedersen B, Birkemose E, Thienpont L, et al. Graphical interpretation of analytical data from comparison of a field method with a reference method by use of difference plots [Opinion]. Clin Chem 1997;43:2039-2046. [Abstract/Free Full Text]
16. Stöckl D. Desirable performance criteria for quantitative measurements in medical laboratories based on biological analyte variation–hindrances to reaching some and reasons to surpass some. Clin Chem 1993;39:913-914. [Free Full Text]