Effect of analytical quality on establishing common reference intervals and their use

In the Nordic Reference Interval Project (NORIP), reference intervals were established for 25 common clinical biochemical quantities. In the project, samples from more than 3000 reference individuals collected in the 102 participating laboratories from all five Nordic countries were analysed locally. In order to maintain a high level of analytical quality and to document this quality, a common calibrator/reference preparation (CAL) and a number of control samples were analysed together with the reference samples. All these materials were serum pools of unprocessed serum from many donors in order to obtain commutable materials. The CAL was used to harmonize the many different analytical procedures and calibrations by simple recalibration by the factor T/M where T is the target value based on reference methods and M is the mean of 10 replicate measurements of CAL in each laboratory. The analytical quality specifications (analytical goals) were based on specifications created directly for the purpose of sharing common reference intervals and only the bias criteria were used because bias is the dominating problem in transfer of reference intervals. These specifications were different for the evaluation of reference values to create common reference intervals and for the laboratories to use these common reference intervals (when established). An interesting outcome was that it was only for the biologically well-regulated quantities serum-sodium and serum-calcium that the selection of the best laboratories gave considerably narrower reference intervals.

Prerequisites for establishing common reference intervals

Establishment of common reference intervals for homogeneous populations within regions is based on the same basic principles as the IFCC recommendations for individual laboratories, but a few additional prerequisites are needed. Thus, the need for common standardization and traceability during production of the reference values and with the application of the common reference intervals in the laboratories becomes crucial. Furthermore, the external control system must be geared to the purpose, using matrix-correct control materials with concentration values traceable to the same reference methods, and validation of results according to analytical quality specifications designed for the use of common reference intervals. Here, the standards may have a restrictive influence on the establishing of common reference intervals, with their demands for the use of the producers' traceability, instead of a relevant high-quality reference preparation shared by all the participants. Two main strategies for measurements are analysis immediately after the sampling, and storage of samples until analysis in one or a few analytical runs. The former strategy needs constant standardization and stability of the performance in many laboratories and in several analytical runs, resulting in between-run variation, whereas the latter precludes this between-run variation, but makes demands on the stability of the components under storage. When a considerable number of laboratories decide to establish common reference intervals, it is possible to obtain large sample sizes of reference values, which reduces the confidence intervals around the reference limits. It also makes it possible to collect samples from many subgroups, such as racial groups and groups related to different environmental conditions, as well as the traditional groupings according to age and gender, pregnancy and use of oestrogens. If all these subgroups are large, e.g. n>500, the confidence limits will be small and criteria for partitioning can be applied. Choosing reference individuals is not easy, as definitions of health, as well as rule-in and rule-out criteria vary from one investigation to the other. Therefore, the strategy and the criteria must be thoroughly described. Arguments for establishing common reference intervals are not needed. On the contrary, lack of such common reference intervals should be explained.

Reference individuals, blood collection, treatment of samples and descriptive data from the questionnaire in the Nordic Reference Interval Project 2000

The rules for recruitment of reference individuals, inclusion and preparation of individuals, blood collection, treatment of samples (and control materials) and analysis at the 102 medical laboratories attending the Nordic Reference Interval Project (NORIP) are given as well as the rules for central exclusion of reference individuals. The individuals (18-91-year-olds) should be evenly distributed on age and gender groups. The 3002 reference individuals who contributed at least one reference value to the finally suggested reference intervals were characterized using the information in the questionnaire. Gender, age and country are the main entries in the tables. Other variables in the cross-tables or figure are height, weight, body mass index, ethnic origin, heredity for diabetes, chronic disease, oestrogens or oral contraceptives, other medication, hard physical activity, previous blood donations, smoking habits, use of alcohol, hours since last meal and time of blood collection (hour, day of week, month, year). The Danes had the highest alcohol consumption and the Icelanders had the highest body mass index. The information in this article may interest potential users of the Nordic Reference Interval Project bio-bank and database (NOBIDA) in which serum, Li-heparin plasma and EDTA buffy coat from the mentioned individuals are stored below -80 degrees C.

Praecordial ECG mapping in acute anterior myocardial infarction. The evolution of ST segment and Q and R waves

Praecordial ECG mapping (42 leads) was carried out in 16 patients with anterior myocardial infarction, admitted up to 3 hours after the onset of symptoms. The mappings were performed according to a set time schedule throughout the acute phase and one year after the infarction. For each ECG map the number of leads with ST elevation greater than or equal to 1.5 mm (NST) and the sum of these (epsilon ST) were calculated. In addition, the sum of Q amplitudes (epsilon Q) and R amplitudes (epsilon R) were determined. On average, the initial epsilon ST elevation decreased from 2 to 8 hours. A secondary increase, occurring in 14 patients, was maximal during days 2-5. Based on the appearance function of creatine kinase in plasma and the QRS alterations, it is suggested that this re-elevation of the St segment was due to infarct extension (5 pats.) or expansion (9 pats.). The course of the average NST was similar to epsilon ST. epsilon Q began to develop at 3 hours, the maximum deflection being reached after about 24 hours (range 8-45). epsilon R was falling at 2 hours; the maximum rate of decrease occurred, on average, within the first 24 hours (range 12-54). One year after the infarction, both epsilon Q and epsilon R had the same level as at 8 hours. The ST elevation at 4 hours was related to the development of Q and R waves within 24 hours, but only a weak or no correlation was found. Thus, there was very little value in predicting the development of QR changes from the ST elevation at 4 hours.

Clinically based quality goals; a NORDKEM project

The three main aspects of analytical quality are 'goals for analytical quality', 'creation of analytical quality', and 'control of analytical quality'. In the NORDKEM-project 'medical need for quality specifications within laboratory medicine' the aspects of analytical quality are combined. The aims is to make an appraisal of the different approaches to goal setting, and to develop a practical procedure for assessing analytical quality requirements. The goals are used to define which demands should be met by the process of establishing and maintaining the quality, and for design of internal as well as external quality assurance procedures. The project is an umbrella project with three main subprojects and several satellite projects. In this presentation a model for evaluation of influence of analytical bias and imprecision on the outcome from a diagnostic classification based on bimodal distribution is described.

The Nordic Reference Interval Project 2000: recommended reference intervals for 25 common biochemical properties

Each of 102 Nordic routine clinical biochemistry laboratories collected blood samples from at least 25 healthy reference individuals evenly distributed for gender and age, and analysed 25 of the most commonly requested serum/plasma components from each reference individual. A reference material (control) consisting of a fresh frozen liquid pool of serum with values traceable to reference methods (used as the project "calibrator" for non-enzymes to correct reference values) was analysed together with other serum pool controls in the same series as the project samples. Analytical data, method data and data describing the reference individuals were submitted to a central database for evaluation and calculation of reference intervals intended for common use in the Nordic countries. In parallel to the main project, measurements of commonly requested haematology properties on EDTA samples were also carried out, mainly by laboratories in Finland and Sweden. Aliquots from reference samples were submitted to storage in a central bio-bank for future establishment of reference intervals for other properties. The 25 components were, in alphabetical order: alanine transaminase, albumin, alkaline phosphatase, amylase, amylase pancreatic, aspartate transaminase, bilirubins, calcium, carbamide, cholesterol, creatine kinase, creatininium, gamma-glutamyltransferase, glucose, HDL-cholesterol, iron, iron binding capacity, lactate dehydrogenase, magnesium, phosphate, potassium, protein, sodium, triglyceride and urate.

The Nordic Reference Interval Project 2000: recommended reference intervals for 25 common biochemical properties

Each of 102 Nordic routine clinical biochemistry laboratories collected blood samples from at least 25 healthy reference individuals evenly distributed for gender and age, and analysed 25 of the most commonly requested serum/plasma components from each reference individual. A reference material (control) consisting of a fresh frozen liquid pool of serum with values traceable to reference methods (used as the project "calibrator" for non-enzymes to correct reference values) was analysed together with other serum pool controls in the same series as the project samples. Analytical data, method data and data describing the reference individuals were submitted to a central database for evaluation and calculation of reference intervals intended for common use in the Nordic countries. In parallel to the main project, measurements of commonly requested haematology properties on EDTA samples were also carried out, mainly by laboratories in Finland and Sweden. Aliquots from reference samples were submitted to storage in a central bio-bank for future establishment of reference intervals for other properties. The 25 components were, in alphabetical order: alanine transaminase, albumin, alkaline phosphatase, amylase, amylase pancreatic, aspartate transaminase, bilirubins, calcium, carbamide, cholesterol, creatine kinase, creatininium, gamma-glutamyltransferase, glucose, HDL-cholesterol, iron, iron binding capacity, lactate dehydrogenase, magnesium, phosphate, potassium, protein, sodium, triglyceride and urate.

Biological variation of soluble CD163

A soluble plasma form of CD163 (sCD163) was recently identified. The protein has anti-inflammatory effects in vitro and is elevated in patients with myelo-monocytic leukaemia and infection. For rational use and evaluation of this potential new quantity it is important to have knowledge of its biological variability and to use a methodology that has a sufficiently analytical quality. The day-to-day and diurnal biological variabilities of sCD163 were studied in 12 healthy people using a sandwich ELISA. The within-run-, between run- and total analytical coefficients of variation were estimated to 3.6%, 4.8% and 6.0%, respectively. The day-to-day within-subject biological variation was estimated to 9.0%, and the between-subject biological variation to 35.9%. A diurnal variation in sCD163 concentrations with 14% lower values in the night (supine position) was observed. The ratio between total analytical variation and within-subject biological variation was 0.67. The index of individuality, calculated as the ratio between within-subject biological variation and between-subject biological variation, was low and similar to complement factors and immunoglobulins. A low index of individuality is important in a monitoring situation where small changes from the set point of the individual can be detected in serial measurements. For this purpose, the critical difference for a series of results in the same patient (significant at p < 0.05) was calculated to 30% for samples taken on different days and measured in separate runs.

Biological variation in variables associated with exercise training

To be able to identify a training induced change in a certain variable, it is necessary to know the background variation. In this study the coefficient of variation (total, between-subjects, within-subjects), the relative sources of variance (between-subjects and within-subjects), and the critical difference (within-subjects) were estimated in four categories of variables (performance and physiological variables, metabolic and hormonal variables, immunological variables, and mood state variables) in 15 moderately trained male runners measured on three different occasions over a period of 7 weeks. In the performance and physiological variables, 78.9 % of the variance was due to variation between subjects and they had the lowest critical difference (11.9 %). In contrast, the metabolic and hormonal variables had the highest critical difference (59.9 %) and 53.4 % of the variance was due to variations within subjects. The immunological and psychological variables had about two thirds of the variance arising from variation between subjects. However, the critical difference for the immunological variables was high (47.4 %), while it was relatively low for the psychological variables (26.8 %). The low critical difference and variation within subjects of the psychological and in particular the performance and physiological variables indicate that they may be beneficial as primary markers of training induced changes.

Soluble CD163: a marker molecule for monocyte/macrophage activity in disease

By immunoprecipitation we have identified a soluble plasma form of CD163 (sCD163), the IL-6 inducible macrophage-receptor for clearing haptoglobin-haemoglobin complexes. A sandwich ELISA for measuring sCD163 was established and used to determine the sCD163 levels in normal subjects and patients with inflammatory and myeloproliferative diseases. In normal subjects, the concentration of sCD163 was high (median 1.9 mg/l) with low intra-individual variation. Highly increased levels were seen in patients with sepsis, myeloid leukaemia and in patients with Gaucher disease characterized by accumulation of tissue macrophages. Although the physiological role of sCD163 remains unknown, our present data suggest that sCD163 might prove to be a valuable marker molecule in infectious and myeloproliferative diseases.

Upper reference limit, analytical quality specifications and clinical use of haemoglobin A1C

Haemoglobin A1c (HbA1c) is now the key component for monitoring glycaemic control in diabetes mellitus (DM), especially for its close relation to diabetes complications. However, treatment goals in terms of HbA1c concentrations have been difficult to define and compare because of lack of international standardization and lack of common reference values of HbA1c concentrations. The aims of our study were to document our HbA1c analysis and make it traceable to international reference laboratories with the aid of current reference preparations, to establish a reference interval based on a low-risk population, and to evaluate the analytical quality specifications, which could meet clinical needs. The s(analytical) of our method (Tosoh) was < 0.3 HbA1c%, and the mean bias as estimated from Dr Cas Weykamp's reference preparation was below 0.3 HbA1c. This was the same as that for participating Scandinavian and international reference laboratories. The concentrations were made traceable to results from the Diabetes Control and Complication Trial (DCCT) and the UK Prospective Diabetes Study (UKPDS). Risk groups for DM were ruled out from a randomly selected population in Vejle County, which isolated a "low-risk" reference population. The 97.5 reference interval in this population (N=430) was from 5.07 HbA1c% (95% CI: 5.02-5.11) to 6.24 HbA1c% (95% CI: 6.19-6.30), and the 99.9 centile was 6.62 HbA1c% (95%) CI 6.55-6.71). Body mass index, age and gender contributed marginally to the level of HbA1c concentrations. A 10% delta risk estimate of DM complications was detectable with a probability of Type I error of 40%, while adoption of a significance level of 95% and consideration to biological variation needed a risk difference of at least 33% to be detected. The critical difference was 11% for changes in either direction at s(analytical) < or = 0.2 HbA1c% and a s(biological) of 0.3 HbA1c%. Based on criteria for sharing common reference intervals and clinical utility, we accepted that the bias and s(analytical) should both be < 0.3 HbA1c% at the level of 7.0 HbA1c%.

Reproducibility of S-insulin and B-glucose responses in two identical oral glucose tolerance tests

Recently, the diagnostic criteria for type 2 diabetes mellitus have been changed, but there are disagreements about which measurements should be used. In contrast to the American Diabetes Association (ADA), The World Health Organization (WHO) still recognizes fasting and 2-h glucose concentrations measured on either plasma or whole blood as diagnostic tools. Insulin sensitivity and insulin secretion are both assumed to be involved in the pathogenesis of type 2 diabetes. The oral glucose tolerance test (OGTT) for estimating insulin sensitivity and secretion is increasingly used, e.g. in intervention trials. The objectives of this study were to estimate the coefficients of intra-individual variation (CVw) of blood glucose and serum insulin concentrations from an OGTT as well as indices of insulin sensitivity (HOMA) and insulin secretion (delta insulin30/delta glucose30) derived from this test. Following duplicate OGTTs with a median interval of 13 days (range 1-87 days), the analytical, inter-individual, and intra-individual coefficients of variation were calculated by nested ANOVA. The CVw for fasting blood glucose (7%) was considerably lower than that for 2-h post-load glucose (15%), which was again lower than for the insulin concentrations and indices of insulin sensitivity and secretion. In conclusion, the intra-individual variation is larger for 2-h post-load glucose than for fasting glucose and may question the continued use of the 2-h post-load glucose value in the diagnosis of type 2 diabetes.

Measured muscle sodium content in biopsy specimens is a reflection of true intracellular content

Determined on the basis of small skeletal muscle biopsies, muscle sodium content has a very high coefficient of variation. Furthermore, at least some of the measured sodium must originate from the extracellular space. In order to assess the applicability of the measurement of intracellular sodium on small muscle biopsy specimens, the measured sodium content was related to the content of dry solids in 25 muscle biopsy samples, and compared with the theoretical content of sodium with varying extra- and intracellular water content in biopsy samples. Four of the 25 measurements were clearly outliers. Disregarding these outliers, it was found that muscle sodium content varied with intracellular water content, whereas the theoretical effect of addition of extracellular water could not account for the observed values. The difference depended upon the specified conditions, but the slope of the theoretical regression line (-25.92 mmol x (kg dry weight)(-1) x %(-1), which was closest to the observed slope, -8.55 mmol x (kg dry weight)(-1) x %(-1), differed substantially (p < 0.0001). No association was found between the primarily intracellular ions muscle potassium and muscle magnesium on the one hand and either muscle sodium or muscle water content on the other. The measured sodium content in muscle biopsy specimens, which are freeze-dried and dissected, seems to reflect the true intracellular sodium content to some extent. The total content of sodium seems to be closely related to the content of water within the skeletal muscle cells.

Can capillary whole blood glucose and venous plasma glucose measurements be used interchangeably in diagnosis of diabetes mellitus?

According to new proposals from the American Diabetes Association (ADA) and WHO, venous peripheral plasma is the preferred system for measuring glucose for diagnosing diabetes mellitus. Owing to the instability of glucose in plasma after blood sampling, strict well-defined and standardized preanalytical conditions are essential to ensure that glucose concentration measured in plasma reflects real blood glucose in the patient. This is in contrast to the capillary whole blood measurements, which are easy to perform and well established. We investigated whether it is possible to perform analysis on capillary whole blood but express the results as plasma glucose values and hence obtain comparable results and the same predictive values for diagnosis in the individual patient? The conclusion of our investigations is that these two systems are not interchangeable and that conversion should not be done for diagnostic purposes where plasma determinations are recommended.

Strategy for determining racial and environmental similarities and differences for plasma proteins

The aim of this protocol is to establish a common basis for the production of reference values and well-defined and documented reference intervals for plasma proteins, based on common standardization, using the IFCC/BCR/CAP Certified Reference Material CRM 470. The strategy is to search for racial and environmental/geographical similarities and sources of differences in order to describe the main causes for variability among smaller or larger groups in selected societies and to estimate the sizes of differences for the different proteins according to the investigated sources. For this purpose, groups of reference individuals are selected according to race and geographical/environmental location, e.g. African Americans and Caucasians from the US. The reference individuals are groups of approximately 160 healthy male blood donors, 20 to 60 years of age. Rule-out criteria are positivity for HIV, hepatitis B and C antibodies and blood hemoglobin below the lower reference limit. Exclusion in relation to different C-reactive protein (CRP) levels will be investigated. Coagulation, storage conditions, transport, and the procedure for thawing are specified. The laboratories undertaking the measurements must have adequate analytical performance, and calibration and quality of performance are defined and documented, together with recommended control materials and procedures. Statistical models for describing distributions and for comparing groups are described. It is recommended that the data be presented as reference limits with 90% confidence intervals of those limits.

Calibration, specificity and trueness of a postheparin plasma lipoprotein lipase assay

Measurement of lipoprotein lipase activity in postheparin plasma is generally accompanied by moderate within-run variation CV(W-R) (<10%) and higher between-run variation CV(B-R) (5-25%). A calibration system was introduced in order to improve the reproducibility of measurements and to compare lipoprotein lipase activities from different days. Every day a calibration curve for lipoprotein lipase activity was constructed. Fifteen calibration curves designed over 2 years, show linearity over the whole biological spectrum and a considerable reduction of between-run variation in lipoprotein lipase activity, from 42% to 5.3% as estimated from two control postheparin plasma samples. The lipoprotein lipase calibration system is an easy and very cheap arrangement, which makes it possible to compare lipoprotein lipase activities achieved over years. When the lipoprotein lipase control values are compared with reference lipoprotein lipase samples determined in other lipase laboratories, the calibration-control system becomes an important tool for reducing analytical bias. The article reviews the original analytical criteria of catalytic measurement of lipoprotein lipase activity and describes the implementation of the calibration-control system. We describe a model for reduction of the analytical variability in the measurement of lipoprotein lipase activity. Other standardization efforts need to be made in the future, especially to define the reference material for calibration.

Serum lactate dehydrogenase isoenzyme 1 and relapse in patients with nonseminomatous testicular germ cell tumors clinical stage I

Serum lactate dehydrogenase isoenzyme 1 catalytic concentration (S-LD-1) was measured at the time of orchiectomy in 104 patients with nonseminomatous testicular germ cell tumors (NSTGCT) clinical stage I who participated in a randomized study comparing surveillance after orchiectomy (group I) and radiotherapy (group II). For 68 patients, S-LD-1 was measured in a serum sample before or on the day of the orchiectomy. Twenty-seven patients (40%) had elevated S-LD-1; median 102 U/L (range 41-335). For the remaining 36 patients. S-LD-1 was measured in a serum sample after orchiectomy: 8 of these patients (22%) had elevated S-LD-1. S-LD-1 was normalized shortly after surgery in most patients with a preorchiectomy elevated S-LD-1. Fifteen of the 68 patients relapsed: 9 out of 27 with an elevated S-LD-1 and 6 out of 41 patients with normal S-LD-1 (p = 0.13, Fisher's exact test). In group 1, those with a preoperatively elevated S-LD-1 had a lower 8-years' relapse-free survival than those with a normal S-LD-1 (40% vs. 80%, p = 0.003, log-rank test). The role of S-LD-1 in the staging, prognostication and monitoring of patients with NSGCT clinical stage I should be further explored in a large, prospective study.

[A randomized controlled trial of the use of CRP rapid test as a guide to treatment of respiratory infections in general practice]

INTRODUCTION

The aim was to assess whether the frequency of antibiotic prescriptions to patients with respiratory infections is reduced when general practitioners (GPs) use a CRP rapid test to support their clinical assessment, and to examine whether the use of the test would have any effect on the course of disease.

MATERIAL AND METHOD

A randomised controlled trial was carried out by 35 general practices in the County of Funen, Denmark, with 812 patients with respiratory infection. The main outcome measures were frequency of antibiotic prescriptions and morbidity one week after the consultation, as stated by the patients.

RESULTS

The frequency of antibiotic prescriptions was 43% (179/414) in the CRP group and 46% (184/398) in the control group (NS, OR = 0.9). At one week, increased or unchanged morbidity was stated more frequently in the CRP group (12%) than in the control group (8%) (OR = 1.6, p = 0.05). In the control group, the variable having the greatest influence on whether the GP prescribed antibiotics was the patient's general well-being (OR = 2.9, p < 0.0001), whereas in the CRP group the CRP value had the greatest influence (OR = 1.1 per unit increase [mg/l], p < 0.0001).

CONCLUSION

From on the present study, the use of a single CRP rapid test to support possible antibiotic treatment of respiratory infections in general practice cannot be recommended.

Models for combining random and systematic errors. assumptions and consequences for different models

A series of models for handling and combining systematic and random variations/errors are investigated in order to characterize the different models according to their purpose, their application, and discuss their flaws with regard to their assumptions. The following models are considered 1. linear model, where the random and systematic elements are combined according to a linear concept (TE = absolute value(bias) + z x sigma), where TE is total error, bias is the systematic error component, sigma is the random error component (standard deviation or coefficient of variation) and z is the probability factor; 2. squared model with two sub-models of which one is the classical statistical variance model and the other is the GUM (Guide to Uncertainty in Measurements) model for estimating uncertainty of a measurement; 3. combined model developed for the estimation of analytical quality specifications according to the clinical consequences (clinical outcome) of errors. The consequences of these models are investigated by calculation of the functions of transformation of bias into imprecision according to the assumptions and model calculations. As expected, the functions turn out to be rather different with considerable consequences for these types of transformations. It is concluded that there are at least three models for combining systematic and random variation/errors, each created for its own specific purpose, with its own assumptions and resulting in considerably different results. These models should be used according to their purposes.

Interpretation of sequential measurements of cancer antigen 125 (CA 125), carcinoembryonic antigen (CEA), and tissue polypeptide antigen (TPA) based on analytical imprecision and biological variation in the monitoring of ovarian cancer

The main objective with cancer antigen 125 (CA 125), carcinoembryonic antigen (CEA), and tissue polypeptide antigen (TPA) monitoring of ovarian cancer patients is to detect an early change of disease activity with high reliability. We hypothesized that a monitoring scheme for ovarian cancer patients with serological tumor markers should take into account the stochastic variation, i.e. the probability that observed increases and decreases may solely be due to analytical imprecision and normal intra-individual biological variation. The aim of this study was to provide a detailed characteristic of the within-subject mean steady state concentrations and the associated variability in healthy individuals with an age distribution representative for ovarian cancer patients. Thirty-one healthy women with a median age of 55 years comprised the study population. Sixteen blood samples were collected from each subject in four series, with four samples per series, over a period of approximately 1 year. We found that, i) natural logarithmic-transformed concentrations were more homogeneously distributed between individuals than the original concentrations, ii) the within-subject mean steady state levels, the standard deviations, and the coefficients of variation differed among subjects, and iii) the steady state variability differed among the markers. In conclusion, our data indicate that the assessment of sequential CA 125, CEA, and TPA concentrations is more complex than hitherto recognized. We suggest that it is necessary to adjust the assessment criteria to the type of marker, and that assessment may be facilitated if based on natural logarithmic transformed concentrations.

Plasma glucose reference interval in a low-risk population. 2. Impact of the new WHO and ADA recommendations on the diagnosis of diabetes mellitus

The aim of the study was to establish a reference interval of fasting venous plasma glucose (FPG) from healthy individuals. A prospective modified cross-sectional population-based study was made with random selection of 2100 persons in age-stratified groups > or = 18 years identified from the local Personal Identification Register. The invitation was accepted by 755 persons, of which 726 aged 18-92 years were eligible. They did not have a diabetes diagnosis, were non-pregnant and capable of fasting for 8 h. All participants filled in a questionnaire on medical risk factors. Blood for the FPG and haemoglobin Alc (HbAlc) measurements was drawn in accordance with a standardized procedure. A total of 302 participants carried diabetes risk indicators and were ruled out. The FPG concentrations in the remaining low-risk population (n=424) was ln Gaussian distributed. The FPG 97.50 centile in this group was 6.4 mmol/L (95% CI: 6.3-6.5 mmol/L), in contrast to the WHO and ADA theoretical limit of 6.1 mmol/L. Their diagnostic decision limit of 7.0 mmol/L FPG corresponded to the 99.86 centile of the FPG reference distribution (95% CI: 6.8-7.1). Subclassification of the reference population showed increasing FPG with increasing BMI and age and was higher in men than in women. The study determined the FPG 95% interfractile reference interval in a healthy population. The interval in glucose concentration between the 97.5 centile of the reference interval and the ADA-WHO diagnostic limit is very narrow. The clinical application of the diagnostic discriminator and the interpretation of the WHO-ADA grey zone from 6.1 to 7.0 mmol/L FPG may consequently be biased because of poorly defined limits and influence from BMI, age and gender.

Evaluation of systematic and random factors in measurements of fasting plasma glucose as the basis for analytical quality specifications in the diagnosis of diabetes. 3. Impact of the new WHO and ADA recommendations on diagnosis of diabetes mellitus

On behalf of the Danish Society of Clinical Endocrinology and the Danish Society of Clinical Chemistry we were commissioned to evaluate the influence of analytical and pre-analytical systematic and random factors on the diagnosis of diabetes, in order to provide a tool for conclusions on the analytical quality specifications needed to diagnose diabetes. A systems analysis was performed in accordance with the principles for evaluation of analytical quality specifications. The clinical setting was defined--diagnosis of diabetes in accordance with the WHO and ADA criteria with determination of fasting plasma glucose concentration (FPG) > or =7.0 mmol/L in two independent samples--with well-documented data on In (loge)-Gaussian distribution of reference values from a low-risk population and values for within-subject biological variation taken from the literature. An investigation was made of the consequences for the clinical setting of assumed errors related to the measurement of FPG. Four approaches were investigated for a single sampling and measurement and also for two independent samples: one showing the percentage of healthy individuals who had values > or = 7.0 mmol/L, one illustrating the origin of biological set-points for results > or = 7.0 mmol/L, one showing the risk of being measured > or =7.0 mmol/L when the biological set-point is known, and one showing the combined bias and imprecision for assumed percentages of false-positive (FP), defined as measurements > or = 7.0 mmol/L for the low-risk population and false-negative (FN), defined as measurements <6.4 mmol/L (the upper reference limit) for diabetics. This leaves a "grey zone" which includes the upper part of low-risk individuals, and defined by ADA and WHO as "impaired fasting glucose" (IFG). In the analysis, increasing systematic and random errors (combined analytical and pre-analytical) were assumed, and for each error condition the fractions of FP and FN were calculated. This gave plots from which the combined allowable systematic and random errors could be read off for pre-determined clinically acceptable fractions of FP and FN. The analysis does not distinguish between pre-analytical and analytical errors, as specified information on one of these is needed for specification of the other. The investigation provides a reliable basis for estimation of the needed analytical quality, and thereby for decisions about analytical quality specifications for analysis of FPG in relation to diagnosis of diabetes under optimized pre-analytical and analytical conditions. Consequences of deviations from these ideal conditions are illustrated in the figures, and should be considered for the different approaches with different performance conditions.

Optimization of preanalytical conditions and analysis of plasma glucose. 1. Impact of the new WHO and ADA recommendations on diagnosis of diabetes mellitus

The new diagnostic criteria for type 2 diabetes from the American Diabetes Association (ADA) and World Health Organization (WHO) recommend measurements on plasma and a lowering of the glucose threshold for diabetes by 0.8 mmol/L. This narrows the distance between the upper end of the reference limit and the discriminatory level to a degree where analytical quality becomes critical. The quality demands for the preanalytical and analytical phase and their consequences on diagnostic performance have to be established in the new technical system, measuring in plasma rather than in capillary whole blood. Because of the instability of glucose in blood samples it is necessary to clarify the influence of different preanalytical and analytical factors on the number of false-positive and false-negative classifications. Thus the aim of the present study was to find optimal conditions for sampling, additives, storage, transport and analysis of plasma glucose combining feasibility with an analytical bias close to zero and a within-imprecision around 1%. We have documented the analytical performance of the method itself and its traceability to an international standard. The preanalytical conditions, such as influence of antiglycolytic agent NaF, conditions for plasma separation, storage temperature and storage time before and after plasma separation were investigated. In conclusion, we recommend that blood should be drawn in tubes containing heparin and NaF and kept on ice water for not more than 1 h until centrifugation at minimum 1000 x g for 10 min. The plasma is then stable for at least 48 h at room temperature.

Influence of index of individuality on false positives in repeated sampling from healthy individuals

The index of individuality is defined as the ratio of the within-subject biological variation to the between-subject variation, i.e., the variation between the biological set-points. It has been disputed whether the index of individuality has influence on the usefulness of conventional population-based reference intervals. In this investigation we found that, as long as only a single sample is taken, for a certain change in an individual's set-point, the index of individuality has no influence on the usefulness of reference intervals. When two or more samples are taken into account, however, the outcome of the measurement is highly dependent on the index of individuality. For a low index, repeat measurement has only limited effect on the fraction of false-positive results, as the next result will be close to the first, but, when the index is high, the fraction of false-positive results will be reduced considerably through repeating the test. Moreover, the distribution of biological set-points for which the fraction of false-positive results originate is described and the influence of analytical imprecision is discussed. The calculations are performed for values of the index of individuality from 0 to 2.0 for the traditional 95% reference interval based on x +/- 2*s(total) (s(total) = total biological variation), and also for a decision limit (cut-off point) x +/- 3*s(total). The numbers are, of course, different, but the effects of the index of individuality are the same, independent of the chosen cut-off point. This concept is related to the clinical classification (diagnosis, prognosis, screening) and the difference from different principles of monitoring is discussed. Further, five examples are evaluated and aspects of index of individuality in relation to false-positive results are discussed.

Evidence for a substantial genetic influence on biochemical liver function tests: results from a population-based Danish twin study

BACKGROUND

Biochemical liver function tests are widely used in the clinic and are some of the most frequently used tests in screening for diseases in older age groups. The aim of the present study was to estimate the relative importance of genetic and environmental factors to variations in these tests among the elderly.

METHODS

We conducted a survey among Danish twins, 73-102 years of age, identified in the population-based Danish Twin Registry. Among the 2749 individuals in the study population, an interview was conducted with 79%. From these, a blood sample was collected from 290 same-sex twin pairs, total of 580 subjects, within a 6-month period and analyzed for alanine aminotransferase (ALT), lactate dehydrogenase (LDH), gamma-glutamyltransferase (GGT), bilirubin, and albumin. The interview included questions about alcohol consumption and body mass index (BMI; self-calculated height and weight). Heritability (proportion of the population variance attributable to genetic variation) was estimated using structural-equation analyses before and after correction for alcohol consumption and BMI.

RESULTS

Structural-equation analyses revealed a substantial heritability (35-61%) for the four biochemical liver function tests: ALT, GGT, LDH, and bilirubin. The remaining variation could be attributed to individuals' nonfamilial environments. Adjustment for alcohol consumption and BMI had no influence on the heritability for ALT, GGT, LDH, and bilirubin. For albumin, two models fit equally well before adjustment for alcohol and BMI: a model including additive genetic and nonshared environmental factors (AE), and a model including shared and nonshared environmental factors (CE). After adjustment, the model including shared and nonshared environment was clearly the best fitting model.

CONCLUSIONS

For both males and females, the effect of genetic factors on the biochemical liver function tests ALT, GGT, LDH, and bilirubin is substantial and accounts for one-third to two-thirds of the variation among individuals 73-102 years of age. The heritability is equal for males and females and does not change notably after controlling for alcohol consumption and BMI. For albumin, no major impact of genetic factors was found independent of BMI and alcohol consumption. An understanding of the genetic mechanisms underlying biochemical liver function tests among the very old may be of value in the interpretation of these tests in this age group.

Standardised procedures can improve the validity of susceptibility testing of uropathogenic bacteria in general practice

OBJECTIVE

To investigate whether the validity of susceptibility testing in general practice would improve when preceded by an intervention.

INTERVENTION

Instruction in standardised susceptibility testing procedures given by laboratory instructors.

METHOD

Urine specimens containing monocultures of typical uropathogenic bacteria were sent to 23 general practices before and after the intervention. Practices performed susceptibility testing by the Sensicult and the Iso-Res agar methods and the validity of the results before and after the intervention was compared. Results from susceptibility testing at the bacteriological laboratory, Odense University Hospital, were used as gold standard.

RESULTS

The median frequency of correct results increased from 82% to 98% for susceptibility testing based on Sensicult (p = 0.001) and from 90% to 96% based on Iso-Res agar (p = 0.05).

CONCLUSION

The validity of susceptibility testing in general practice improves when preceded by instruction in standardised procedures.

Effect of a new international reference preparation for proteins in human serum (certified reference material 470) on results of the College of American Pathologists Surveys for plasma proteins

OBJECTIVE

To examine the impact of a new international reference preparation for serum proteins on the among-manufacturer variance in the College of American Pathologists Surveys.

DATA SOURCE

The results of the Surveys for the years 1993-1998, supplied by the College of American Pathologists.

DATA EXTRACTION AND SYNTHESIS

Mean values for manufacturers' assays were compared for each protein in the quality control samples. Results were separated by the reported reference material from which values for calibrators had been transferred. Standard statistical parameters (means, standard deviations, and coefficients of variation) were calculated from the reported means.

CONCLUSIONS

Among-manufacturer coefficients of variation have dropped significantly for most serum proteins since the introduction of the new reference material. Possible reasons for continued differences are discussed.

[Resistance testing in general practice--is it valid?]

Resistance of uropathogenic bacteria to antibiotics is an increasing problem in primary health care. The aim of this study was to evaluate antibacterial susceptibility testing of uropathogenic when performed in general practice. Urine specimens with a known quantity of typically uropathogenic bacteria were sent to 25 general practices. The predictive values of testing a bacterial strain as susceptible ranged from 0.89 (nitrofurantoin) to 1.00 (sulphamethizole), and the predictive value of testing a bacterial strain as resistant ranged from 0.55 (trimethoprim) to 0.90 (nitrofurantoin). If susceptibility testing is to be widely implemented in general practice it would be necessary to improve the validity.

Progression criteria for cancer antigen 15.3 and carcinoembryonic antigen in metastatic breast cancer compared by computer simulation of marker data

BACKGROUND

We investigated the utility of computer simulation models for performance comparisons of different tumor marker assessment criteria to define progression or nonprogression of metastatic breast cancer.

METHODS

Clinically relevant values for progressive cancer antigen 15.3 and carcinoembryonic antigen concentrations were combined with representative values for background variations in a computer simulation model. Fifteen criteria for assessment of longitudinal tumor marker data were obtained from the literature and computerized. Altogether, 7200 different patients, each based on 50 measurements, were simulated. With a sampling interval of 4 weeks, the monitoring period for each event was approximately 3.8 years.

RESULTS

Modulation of the background variation, the starting concentrations, and the cutoffs enabled identification of criteria that were robust against false-positive signals of progression.

CONCLUSIONS

The computer simulation model is a fast, effective, and inexpensive approach for comparing the diagnostic potential of assessment criteria during clinically relevant conditions of steady-state and progressive disease. The model systems can be used to generate tumor marker assessment criteria for a variety of malignancies and to compare and optimize their diagnostic performance.

Need for revisiting the concept of reference values

The reference values concept has been adopted by health care professionals, including clinical chemists, laboratory scientists, and clinicians and simultaneously by all the official organizations in charge of the establishment of legislation. But the estimation of reference limits, and the evaluation of biological variability need to be improved at the level of the procedures, which are currently too long and too expensive and not feasible easily for all laboratories. The procedures for obtaining reference values, if we follow the original documents, are complex, and that is the main reason that clinical chemists or diagnostic kit manufacturers have not used them systematically. There is clearly a need that scientific societies and international organizations propose practical recommendations: 1) Recommendations to describe methods linked to systematic error. * How to transfer reference limits from one laboratory to another laboratory using different methods? * Should we determine reference limits for each method? * How can we differentiate bias due to the populations from these due to the method? Clear collaborations with manufacturers involved in kits and diagnostic systems are needed. 2) Practical recommendations linked to the reference population. * How to select a homogeneous population? (Careful recommendations on the choice between healthy individuals, blood donors and individuals hospitalised for other diseases should be given.) * How to estimate ethnic differences? * How to define the exclusion and inclusion criteria according to quantity? * How to deal with the question of reference limits for unstable periods, aging or old people particularly, when the difference between aging and disease is very difficult to define? 3) Practical recommendations on the statistical methods to be used. * How to make a good choice of the interquartile interval? Should we use and present only the centiles 2.5 or 97.5, or on the contrary should we give other centiles in addition, for example 5, 10, 75, 80, 85, 90? 4) Practical recommendations linked to the use of the concept of the reference values. * How to make this concept more concrete and to have official definitions which are better understandable and not only abstract? * How to demonstrate the value of using simultaneously reference limits and decision limits, and what does each of these limits bring to results interpretation? * How to improve the presentation of the results? How to give more information on biological variability in the laboratory data, taking into account the scientific validity of their determination? Should we use new information techniques and new communication systems for reaching these objectives? The responses to all these questions could only be provided if there is a concerted effort at the international level. Practical recommendations should be given, which would be very useful for a better understanding and use of reference values by laboratory scientists and clinicians.

A model for setting analytical quality specifications and design of control for measurements on the ordinal scale

A model for characterization of measurements on the ordinal scale is presented. It is based on transformation of the calculated fractions (fractiles) of positives from measurements on samples with known concentrations to a probit-natural log (probit-ln) scale. Such measurements could be made by other methods on ratio or difference scales but, for convenience (for example for speed or low cost), are measured on the ordinal scale by "simple" methods. The model is examined, and verified, using three examples from published data (haemoglobin, glucose, and leukocytes) and an external quality assessment survey on measurements of streptococcus. We show that it is possible to obtain reliable analytical quality specifications and to establish design of control systems for measurements on the ordinal scale. It is concluded that the presented probit-ln model for the ordinal scale is a tool which can improve and facilitate (i) characterizing methods with measurements on the ordinal scale, (ii) defining analytical quality specifications, (iii) designing external assessment as well as internal control schemes, (iv) validation of methods with measurements on the ordinal scale according to the analytical quality specifications, and further, (v) reduction of the number of samples required for method validation and the number of replicate measurements needed.

Assessment of CA 15.3, CEA and TPA concentrations during monitoring of breast cancer

The variability of the tumor markers cancer antigen (CA) 15.3, carcinoembryonic antigen (CEA) and tissue polypeptide antigen (TPA) during steady state concentrations and the rate of increase during progression is described. One hundred and ninety-two patients were monitored during first-line chemotherapy for metastatic breast cancer and during follow-up. Blood specimens were sampled approximately every four weeks. Steady state concentrations were registered for 77 (CA 15.3), 96 (CEA), and 127 (TPA) patients with below cutoff level values and for 28 (CA 15.3), 25 (CEA), and 11 (TPA) patients with above cutoff level values. Clinical and marker progression was registered for 75 (CA 15.3), 62 (CEA), and 57 (TPA) patients. The coefficients of total variation of steady state concentrations (comprising the intra- and interassay analytical imprecision and the within subject biological variation) were higher below (14.9% CA 15.3, 15.4% CEA, 25.9% TPA) than above cutoffs (9.6% CA 15.3,6.0% CEA, 19.9% TPA). The variability was similar for CA 15.3 and CEA but higher for TPA. During progression the rates of increase in concentrations were similar for CA 15.3 (0.0257) and CEA (0.0214) and lower than for TPA (0.0346). Our data indicate that criteria for assessment of sequential tumor marker concentrations should consider the marker in question, the steady state variability, the cutoff value, and the rate of increase during disease progression.

Randomised controlled trial of CRP rapid test as a guide to treatment of respiratory infections in general practice

OBJECTIVE

To assess whether the frequency of antibiotic prescriptions to patients with respiratory infections is reduced when general practitioners (GPs) use a C-reactive protein (CRP) rapid test in support of their clinical assessment, and to study whether using the test will have any effect on the course of disease

DESIGN

Randomised controlled trial.

SETTING

35 general practices, County of Funen, Denmark.

PATIENTS

812 patients with respiratory infection.

MAIN OUTCOME MEASURES

Frequency of antibiotic prescriptions and morbidity 1 week after the consultation, as stated by the patients.

RESULTS

In the CRP group the frequency of antibiotic prescriptions was 43% (179/414) compared with 46% (184/398) in the control group (odds ratio (OR) = 0.9, NS). After 1 week, increased or unchanged morbidity was stated more frequently in the CRP group (12%) than in the control group (8%) (OR = 1.6, p = 0.05). In the control group, the variable having the greatest influence on whether the GP prescribed antibiotics was the patients' general well-being (OR = 2.9, p < 0.0001), whereas in the CRP group the CRP value had the greatest influence (OR = 1.1 per unit increase (mg/l), p < 0.0001).

CONCLUSION

Based on the present study, the use of the CRP rapid test in support of a possible antibiotic treatment for respiratory infections in general practice cannot be recommended.

Quality specifications based on analysis of effects of performance on clinical decision-making

The graphical model for evaluation of analytical requirements for bimodal distributions, based on probit transformation and calculation of false-positive and false-negative results for assumed random and systematic analytical errors, is presented in theory. It is concluded that the bimodal model is an excellent tool for evaluation of the effect of analytical quality and, therefore, useful for estimation of analytical quality specifications for quantities used in specific clinical strategies and situations as well as based on analysis of clinical outcome.

Validity of susceptibility testing of uropathogenic bacteria in general practice

Resistance of uropathogenic bacteria to antibiotics is an increasing problem in primary health care. The aim of this study was to evaluate antibacterial susceptibility testing of uropathogenic bacteria when performed in general practice. Urine specimens with a known quantity of typically uropathogenic bacteria were sent to 25 general practices. The predictive values of testing a bacterial strain as susceptible ranged from 0.89 (nitrofurantoin) to 1.00 (sulphamethizole), and the predictive value of testing a bacterial strain as resistant ranged from 0.55 (trimethoprim) to 0.90 (nitrofurantoin). Interventions to improve the validity of susceptibility testing are desirable if the test should be incorporated in the diagnostic armamentarium in general practice.

Assessment of biological variation and analytical imprecision of CA 125, CEA, and TPA in relation to monitoring of ovarian cancer

OBJECTIVES

Changes in serial tumor marker results during monitoring of patients with ovarian cancer are due not only to deterioration or amelioration of the patient's condition, but also to preanalytical sources of variation (CPP), total random analytical error, and within-subject normal biological variation. The aim of the study was to assess (i) the analytical imprecision (CVA) and the average inherent intra- and interindividual biological variation (CVTI and CVG, respectively) for CA 125, CEA, and TPA in a group of healthy women; (ii) the significance of changes in serial results of each marker; and (iii) the index of individuality.

METHODS

The study group consisted of 31 healthy women. Sixteen blood samples from each subject were collected in four series over a period of approximately 1 year. Data analysis was based on ANOVA. The index of individuality was calculated as ((CV2A + CV2TI)/CV2G)1/2 and the critical difference for a change between two consecutive concentrations as radical2xZx(CV2P + CV2A + CV2TI)1/2 (Z = 1.65 for unidirectional and 1.96 for bidirectional changes, P </= 0.05).

RESULTS

The averages of CVTI and CVG were 35.5 and 70.6% for CA 125, 13.9 and 58.3% for CEA, and 31.4 and 62.5% for TPA, respectively. The indices of individuality were 0.5 for CA 125, 0.3 for CEA, and 0.5 for TPA. The critical differences of CA 125, CEA, and TPA were 84.6, 37.6, and 76. 3%, respectively (Z = 1.65). There were no significant differences in CA 125, CEA, and TPA concentrations between smokers and nonsmokers or in CA 125 and CEA concentrations between pre- and postmenopausal women. Postmenopausal women had higher TPA concentrations than premenopausal (P = 0.015).

CONCLUSION

CVA and CVTI contribute considerably to the variation in serial results and should, therefore, be included in the criteria for serum tumor marker assessment during monitoring of patients with ovarian cancer.

The index of individuality is often a misinterpreted quantity characteristic

The concept of the "index of individuality" was introduced by Eugene Harris in 1974. The index of individuality, calculated as (CV(A)2 + CV(I)2)(1/2)/CV(G), where CV(A), CV(I), and CV(G) are analytical, within-subject, and between-subject coefficients of variation respectively, has been used by many to investigate the utility of conventional population-based reference values. For a high index of individuality, > 1.4, it has been said that reference intervals will be more useful than for a low index, < 0.6. The validity of these concepts is investigated here and a number of our findings are at odds with the generally held opinion. The index of individuality has no impact on the fraction of individuals classified using population-based reference values, as long as the change in concentration from the usual state is of the same absolute magnitude and one sample is assayed to detect disease. However, when a measurement falling outside a reference limit is repeated in order to verify the finding, the index of individuality has considerable influence. For quantities with very low indices, the repeat test result, will be close to the first and give no new information, whereas for quantities with high indices, a repeat test will decrease the number of true positives and false positives.

Estimation of genetic and phenotypic parameters for clinical mastitis, somatic cell production deviance, and protein yield in dairy cattle using Gibbs sampling

When including clinical mastitis in the breeding goal, it is useful to know what measure of the trait is most appropriate and its relationship to the primary production traits and indicator traits in the relevant population. In this paper, genetic and phenotypic parameters for clinical mastitis, somatic cell production deviance, and protein yield were estimated for the dairy breed Danish Red. In preliminary analyses, the heritability for clinical mastitis was found to be highest in early lactation, and its genetic correlation to clinical mastitis at other stages of lactation were high. Therefore, clinical mastitis defined in early lactation was the measure of clinical mastitis used in subsequent analyses. Two bivariate analyses were performed. Each analysis fitted clinical mastitis and either somatic cell production deviance or protein yield as a continuous trait. The bivariate model was composed of a Gaussian model for the continuous trait and a threshold model for mastitis. The analyses were performed in a Bayesian setting, using the Gibbs sampler. Point estimates (mean of marginal posterior densities) of heritability for mastitis on the underlying scale were estimated to be 0.10 and 0.12 in the two analyses. The genetic correlation between mastitis and protein yield was 0.43 and between mastitis and somatic cell production deviance was 0.80. These results make clear the importance of including clinical mastitis in the breeding goal and the usefulness of somatic cell production deviance as the indicator trait for clinical mastitis. The best measure of clinical mastitis was to consider only cases in early lactation.

Analytical quality specifications for serum lactate dehydrogenase isoenzyme 1 based on clinical goals

The aim of the study was to deduce analytical quality specifications for the determination of catalytic concentration of serum lactate dehydrogenase isoenzyme 1 (S-LD-1) according to clinical goals (the clinical utility model). We defined clinical goals for false positive and false negative S-LD-1 measurements in the monitoring of patients with testicular germ cell tumors (TGCT), clinical stage I, on a surveillance only program. The absolute S-LD-1 catalytic concentrations were routinely corrected for contamination from preanalytical hemolysis. A reference group of 37 men had a near In-Gaussian distribution for the absolute S-LD-1 catalytic concentration. The geometric mean was 76 U/l and an S-LD-1 >128 U/l (99.72 percentile, the decision limit) indicated a high risk of a relapse of TGCT. We have previously shown that an S-LD-1 >160 U/l (treatment limit) was associated with a suboptimal outcome from the treatment of metastatic TGCT. The maximum allowable analytical positive bias was 5 U/l, and the maximum allowable analytical negative bias was -32 U/l. The maximum allowable analytical coefficient of variation, CV(A), was 11% (approximately 14 U/l) at a bias = -5 U/l. For S-LD-1 measurements not corrected for hemolysis, the decision limit was 145 U/l, the maximum allowable negative bias -19 U/l, and CV(A) 8%(approximately 12 U/l). A routine correction for hemolysis had a large impact on the analytical quality specifications.