Interlaboratory survey of Enzyme Analyses iii. Does College of American Pathologists' survey serum mimic clinical specimens?

Participants in the last College of American Pathologists Enzyme Surveys in 1978 were asked that they provide results for lactate dehydrogenase (LD), alkaline phosphatase (ALP), aspartate aminotransferase (AST), alanine aminotransferase (ALT), and creatine kinase (CK), for the six interrelated Enzyme Survey serum specimens, and were also asked to provide results for the same enzymes from six healthy persons and from 50 patients. Extensive data analysis revealed that porcine LD and CK in the Survey serum had kinetic properties similar to the native enzymes in fresh human serum. On this basis, we were able to merge the normal data from laboratories using diverse methods and to establish a "universal" reference range for these two enzymes. Survey serum may serve as a bridge for the clinical comparison of very diverse methods for determining LD and CK, possibly for AST and ALT, but not for ALP.

The 1978 College of American Pathologists survey of analyses of five serum enzymes by 450 laboratories.

Sets of survey specimens having known linear interralationships were analyzed on four occasions by approximately 450 laboratories for the five enzymes lactate dehydrogenase, aspartate aminotransferase, creatine kinase, alanine aminotransferase, and alkaline phosphatase. The results are summarized in terms of the apparent precision and relative accuracy of various analytical systems, and some apparent problems in enzyme assays are identified. The results show that interlaboratory differences in enzyme analyses are not due primarily to differences in the way laboratorians utilize their analytical systems but rather are due to fundamental differences in the instruments and reagents supplied to the laboratorians. The attainment of interlaboratory comparability of enzyme analyses is a problem that can best be addressed by the manufacturers of instruments and reagents, rather than by individual laboratorians.

An evaluation of ethylene glycol-based liquid specimens for use in quality control.

We investigated the analytical acceptability of ethylene glycol-based control sera by preparing sets of aqueous and ethylene glycol-based specimens that had added uric acid, calcium, creatinine, glucose, urea, sodium, and potassium. Ethylene glycol caused a significant systematic proportional bias in procedures involving dialysis, but had no discernible effect on methods not involving dialysis. The extent of bias was proportional to the concentration of ethylene glycol, was independent of protein concentration, and differed according to the type of dialysis system used. We conclude that ethylene glycol-based control materials can have useful applications in clinical chemistry, but caution must be exercised in their use with analytical systems that employ dialysis.

The 1977 College of American Pathologists Enzymology Survey. Principal findings.

A set of linearly related serum survey specimens was analyzed by more than 200 laboratories on four occasions in 1977 for the five enzymes lactate dehydrogenase, aspartate and alanine aminotransferases, creatine phosphokinase, and alkaline phosphatase. The survey technic, which employs specimens that are designed to have multiple informational redundancies for detecting discrepancies in the survey data, was found to be a feasible approach, as a variety of methodologic biases and analytic problems were readily detected. The specimens were shown to be stable,and the mean results obtained by a reference group of laboratories were quite reproducible. Consequently, changes observed in individual laboratories in successive surveys were attributable to intra-laboratory changes. The survey provides each participant with an objective evaluation of the linearity, precision, and calibration of his methods, in comparison with peer and reference laboratories, as well as information concerning the popularities of various methods, and the performances of various analytic systems. For each enzyme there are a few large peer groups and a large number of small peer groups. Differences among the laboratories within each peer group were due primarily to proportional biases that in some cases were clearly due to instrumental or reagent differences, rather than to methodologic differences.

An interlaboratory comparison of analyses of clinical specimens.

A pilot study was conducted to identify some of the benefits and limitations of making interlaboratory comparisons of clinical assays. Sixteen laboratories that participated in a regional quality control program and used similar instrumentation each provided analyses of specimens from three male and three female healthy subjects, 25 male and 25 female clinical subjects, and several control specimens. Analysis of the data revealed that the derived normal ranges agreed well with those provided by the instrument manufacturer, and that for some analytes the laboratories produced comparable clinical assays. The control specimens were found to be indicators of the presence of analytic bias that affected the clinical assays, but they did not always correctly indicate the kind or magnitude of bias. Using pattern recognition technics, it was shown that a laboratory's clinical assays had characteristic distributions that were apparently related to the populations served, as well as to analytic precision. The results demonstrate that useful information may be gleaned from interlaboratory surveys of clinical assays.

Use of survey-validated reference materials (survey serum) to establish target values of quality control pools.

Specimens used in the College of American Pathologists Survey Programs are sometimes manufactured in excess of Survey needs, and these excess specimens are made available to Survey participants after a survey is completed. These survey-validated reference materials (SVRM's) are a valuable resource to the clinical laboratory field. SVRM's are continually renewed, have a range of analyte concentrations, have reliable target values assigned for more than 20 analytes, and for all methods that are in common use. The materials are useful for assigning appropriate target values to other serum materials used by laboratories, for investigating the accuracy of a laboratory's analytic procedures, and as trouble-shooting aids. The wide availability of SVRM's offers a new opportunity for improving the analytic reliability of quantitative clinical assays.

A technique for determining the probability of abnormality.

Histograms of the results of quantitative clinical laboratory analyses are usually broadly skewed because some of the patients have normal values with respect to the analyte being measured while values for others are below or above normal. We describe a technique for calculating the maximum possible number of individuals in a patient population who will give normal values and, by difference, the minimum possible number of those who will give below-normal and above-normal values. The technique permits estimation of the probability that any given result represents a member of a particular subpopulation, as well as selection of decision limits for patient classification, and estimates the degree of change required for a subject with a normal value to start giving below- or above-normal values.

Performance evaluation of multi-channel analyzers by use of linearly-related survey specimens.

Sets of specimens having quantitative linear inter-relationships for 25 analytes were prepared and used in a small survey of results with multi-channel analyzers. Instrument calibration was evaluated by linear regression analysis of the analytical results, with calculation of the x- and y-intercepts and slopes. The average intercepts and slopes agreed quite well with those expected on the basis of specimen preparation, but the results from individual laboratories and from particular kinds of instruments demonstrated a variety of analytical biases. We conclude that performance of multi-channel analyzers can be evaluated effectively by such use of linearly related specimens in an inter-laboratory survey.

Design of quality-control specimens for use with a small multi-channel analyzer.

Specimens having linear relationships between concentration and instrument response for seven analytes were prepared for use with a small multi-channel analyzer. The specimens had inter-specimen and inter-constituent relationships that facilitated the performance evaluation of all channels simultaneously with each individual specimen. Techniques are described for use of the specimens in various aspects of quality control.

Interlaboratory survey of enzymatic analyses. II. Intermediate studies.

A set of six interrelated specimens was analyzed for creatinine, urea, and five enzyme constituents by about 10 laboratories on two occasions. The results from each laboratory were subjected to linear regression analysis and the correlation coefficients, X and Y intercepts, and slopes were calculated. These parameters then were used in various ways to (a) evaluate the composition, linearity and stability of the specimens, (b) assess the analytic precision of each laboratory, (c) assess the accuracy of the laboratories relative to their peer group, and (d) identify systematic differences among peer groups. The results suggest that this technic is a viable approach to the evaluation of laboratory performance in enzymatic analyses.

Interlaboratory survey of enzyme analyses. I. Preliminary.

Enzyme survey specimens were prepared by spiking portions of a normal serum pool with creatinine, urea, and five enzymes (lactate dehydrogenase, creatine phosphokinase, alkaline phosphatase, and aspartate and alanine aminotransferases), and preparing admixtures of the spiked pools with the original serum. This admixture technic established linear interspecimen relationships that could be confirmed by analyses for creatinine and urea nitrogen. Both ethylene glycol-stabilized liquid serum specimens and lyophilized specimens were prepared as sets of six to eight samples having six concentrations of each enzyme. The sets were distributed on five occasions to about ten laboratories that were widely separated geographically, and the specimens were analyzed by a variety of methods, by various instrumental systems, and in different reaction conditions, and results were reported in diverse units. This report describes how the analytic data obtained through the use of these specimens that were specifically designed for survey purposes can be analyzed statistically to provide meaningful assessments of laboratory performance in enzyme analyses.

Studies of the reliability of constituent target values established in a large inter-laboratory survey.

Lyophilized serum specimens were manufactured in such a way that fixed quantitative interspecimen relationships were established. The specimens were analyzed for about twenty constituents by several thousand laboratories in 1971, 1972, and 1973. Most of the consensus mean values derived from these analyses were found to relect accurately the relationships established in the manufacture of the specimens. In addition, analyses of some of the specimens for calcium by the National Bureau of Standards confirmed the relationships. I conclude that the consensus mean values derived for most constituents are reproducible and any biases that may be present are consistent from survey to survey.

On the design of clinical chemistry quality-control sera.

Quality-control specimens having quantitave interspecimen relationships were prepared by (a) making admixtures of two serum pools and (b) making various dilutions of the admixtures. The specimens were analyzed for many constituents to illustrate how the data obtained could be manipulated and used. The inter-related specimens permit the reduction of quality-control data to a few target values and should be particularly useful for the evaluation of multi-channel analyzers, and for use in inter-laboratory proficiency surveys. Aspects of the design of quality control and proficiency test specimens are also discussed.