CLINICAL PROFICIENCY TESTING ESTABLISHES THAT THE
DIAGNOSIS OF DIABETES MELLITUS CAN BE INFLUENCED
BY DIFFERENT METHODOLOGY
G. S. Lima-Oliveira1, C. Albuquerque2, R. Doellinger2, V. Biasoli2, A. Duarte3.1Federal University of Parana, Curitiba-PR, Brazil, 2ControlLab, Rio de Janeiro, Brazil, 3University of São Paulo LIM56, São Paulo, Brazil
ABSTRACT
Background: From 2003 to 2007, the ADA introduced several changes inthe diagnostic criteria of diabetes mellitus. However, none of the publishedguidelines specified which glucose measurement methodology should beused in the clinical laboratory. Latin American laboratories utilize severalmethods, especially oxidase and hexokinase assays.
Objective: To evaluate if different methodology of glucose determinationwould change the diagnosis of diabetes mellitus in clinical laboratories.
Methods: This study received IRB approval and identification of alllaboratories were preserved. To compare different methodology among thedifferent laboratories the authors used data from the clinical proficiencytesting of ControlLab, supported by the Brazilian Society of ClinicalPathology/ Laboratory Medicine (SBPC/ML). For the different methods theControlLab calculates the robust mean, standard deviation and coefficientof variation (CV), employing the ISO 13528 (algorithm A). To calculate theacceptance range, ControlLab applies to the robust mean the limitsrecommended by Brazilian regulatory agencies. The rate of success is
calculated by the formula: results in the acceptance range divided by allresults. We evaluated glucose results reported by 1154 laboratories toControlLab between July, 2003 and July, 2007, obtained by differentmethodology in Latin American laboratories (oxidase, colorimetric and UVhexokinase in several types of clinical analyzer systems). To evaluate ifdifferent methods of glucose determination would change the diagnosis weused ANOVA and Dunnett´s T3 tests.
Results: The rate of success is significantly higher using the UVhexokinase glucose determination method (Dunnett´s test p< 0,0001).
Discussion and conclusion: It is crucial to establish guidelinesstandardizing the methodology for the diagnosis and therapeuticmonitoring of diabetes in clinical laboratories because several methodscan be influenced by methodology used in open or closed clinical analyzerautomation. This work demonstrates that oxidase and colorimetrichexokinase assays have higher coefficients of variation, and this can leadto diagnostic mistakes.
BACKGROUND
In 2003 the Expert Committee on theDiagnosis and Classification of DiabetesMellitus of American Diabetes Association(ADA) introduced several changes to thediagnostic criteria for diabetes and forlesser degrees of impaired glucoseregulation (IFG/IGT) established in 1997by International Expert Committee. In2004 the ADA published a new statementabout diagnosis and classification ofdiabetes mellitus. (Table 1) However,there was no recommendations aboutspecifics glucose measurement methodsin clinical laboratory. Currently, Brazilianlaboratories use on several methods,mainly, glucose oxidase (oxidase) andhexokinase. Taken together, one canargue if diagnosis of diabetes mellituscould be influenced by different methodsand methodologies assay in Brazilian andLatin American clinical laboratories.
OBJECTIVE
Based on those statements, we proposedto evaluate the influence of the applicationof different methods in Brazilian and LatinAmerican laboratories.
METHODS AND PROCEDURES
From July 2003 to July 2007, 45 differentserum samples for glucose proficiencytesting were sent by ControlLab (4 timesper year) to 1,154 laboratories, to testdifferent methods and methodologies inLatin American laboratories.
All laboratories returned the result of theglucose proficiency testing with theinformation about method (oxidase,
colorimetric hexokinase or UVhexokinase), reagent and equipment usedto determination the concentration ofglucose.
To compare different methodologiesamong the different laboratories theauthors used data from the clinicalproficiency testing of ControlLab,supported by the Brazilian Society ofClinical Pathology/Laboratory Medicine(SBPC/ML).
This study was approved by our InternalReview Board (IRB) and identification ofall laboratories was preserved.
Statistical methods applied byControlLab´s Clinical ProficiencyTesting Program
For the each different method, it wasdetermined the robust mean, standarddeviation and coefficient of variation (CV),employing the ISO 13528 (algorithm A).To calculate the acceptance range, thelimit of 13% recommended by Brazilianregulatory agencies (ANVISA/REBLAS)was applied to the robust mean.
The rate of success was calculated by thefollowing formula: results in theacceptance range divided by all results.
Statistical methods applied by our workgroup.
To evaluate if different methods of glucosedetermination (oxidase, colorimetrichexokinase or UV hexokinase) wouldchange the diagnosis of diabetes mellituswe used ANOVA and Dunnett´s T3 tests.
To evaluate if different methodologies(opened or closed system) would changethe diagnosis of diabetes mellitus wereformed six groups and Mann-Whitney´stest was used. The groups formed are: (1)Oxidase – opened system; (2) Oxidase –closed system; (3) Colorimetrichexokinase – opened system; (4)Colorimetric hexokinase – closed system;(5) Ultraviolet hexokinase – openedsystem; (6) Ultraviolet hexokinase –closed system.
P values < 0.05 were consideredstatistically significant.
RESULTS
The rate of success was significantly higher using theultraviolet hexokinase glucose determination method(Dunnett´s test p< 0,0001), when not considering themethodology (Figure 1).
It was observed a difference between opened and closedsystem methodologies for each method (oxidase,colorimetric and ultraviolet hexokinase). No difference inperformance was observed when we compared oxidaseversus ultraviolet hexokinase by closed system andoxidase versus colorimetric hexokinase by openedsystem (Table 2 and Figure 2).
From 1,154 laboratories evaluated, 81% perfomed thelaboratorial diagnosis of diabetes by opened systems(71% by oxidase method). Only 18% used closedsystems (Table 3).
DISCUSSION AND CONCLUSION
It is important to establish guidelines standardizing themethodology for the diagnosis and therapeuticmonitoring of diabetes mellitus in clinical laboratories.Several methods can be influenced by methodology,such as opened or closed clinical analyzer automation.Currently, the most frequent methodology used in LatinAmerica to glucose determination is opened systems.
In the present study, we sought to compare the results ofa proficiency testing of glucose of 1,154 laboratories andverify if different methods and methodologies couldinterfere in the diagnosis of diabetes mellitus.
The present analysis showed that theperformance of a method is influencedby the methodologies application(opened or closed system). Forexample, the isolated analysis of figure 1could lead to an erroneouslyinterpretation of a superiority of only onemethod. A more careful analysis ofTable 2 and Figure 2 lead us concludethat closed system yields to higher rateof success.
Taken together, our data providesinteresting insights about differentperformance obtained with methods andmethodology. It highlights concernsabout methodology standardization.
REFERENCES
National Diabetes Data Group: Classification anddiagnosis of diabetes mellitus and other categories ofglucose intolerance. Diabetes 28:1039-57,1979.
The Expert Committee on the Diagnosis andClassification of Diabetes Mellitus: Report of theexpert committee on the diagnosis and classificationof diabetes mellitus. Diabetes Care 20:1183-97, 1997.
The Expert Committee on the Diagnosis andClassification of Diabetes Mellitus: Follow-up report onthe diagnosis of diabetes mellitus. Diabetes Care26:3160-67, 2003.
American Diabetes Association: Diagnosis andClassification of Diabetes Mellitus. 27:S5-10, 2004.
ANVISA – Procedimento GGLAS nº. 02/43: Critériospara a Habilitação de Provedores de Ensaios deProficiência. 2ª.ed., Brasília, 2002.
ISO 13528:2005(E) – Statistical Methods for use inProficiency Testing by Interlaboratory Comparisons.Annex C (normative) – Robust Analysis.
________
[email protected]@controllab.com.b
