Consensus means are tacitly assumed to provide correct target values in many external quality assessment schemes EQAS for peptide hormones and tumour markers. We suggest, however, that such targets should not be used without some evidence of their validity. Comparison of the expected and found increments in the target value on adding known quantities of International Standards to serum pools can provide confirmation of the correctness of target values or, in some cases, identify clearly incorrect targets. Validation of targets is important if EQAS are to stimulate use of correctly calibrated assays, rather than those that simply agree with the most commonly used method(s).

The quality of serum prolactin assays routinely performed by UK laboratories has been monitored in an external quality assessment scheme (EQAS) over a 10-year period, during which participation in the EQAS increased three-fold, and considerable changes in methods and standardization were introduced. The all-laboratory mean was used as the sample target value, and proved to be stable and accurate. Overall between-laboratory agreement in the clinically important range improved from a geometric coefficient of variation (GCV) of 25% to 14%. This appears to reflect the increased use of kits in place of 'in-house' assays, the more widespread availability of international standards and the absence of any marked differences in bias between the commonly used methods. Published guidelines on the clinical interpretation of prolactin values should, therefore, be widely applicable. The EQAS data indicate that, in general, the quality of performance of prolactin assays is adequate for their clinical application.

External quality assessment schemes (EQAS) have traditionally emphasised the achievement of between-laboratory consensus. Although this is important, the application of EQAS to relatively new and evolving techniques such as immunoassay calls for a wider and more searching remit if the goals of accurate assays, properly used, are to be achieved. This article outlines the principles of EQAS for peptide hormones and tumour markers, emphasising key aspects such as validation of target values, dependency of results on sample type, and assessment of method characteristics such as vulnerability to interfering factors. The latter are considered to be important as they can affect patient care more seriously than modest degrees of imprecision or inaccuracy. EQAS play a unique role in providing objective data on assays performed in many laboratories under routine conditions and the data they provide can guide improvement in diagnostic reagents and laboratory practice.

Journal Article Estimation of sensitivity of immunoassays Get access J Seth J Seth Search for other works by this author on: Oxford Academic Google Scholar Clinical Chemistry, Volume 36, Issue 1, 1 January 1990, Page 178, https://doi.org/10.1093/clinchem/36.1.178a Published: 01 January 1990

Trends in the quality of assays for serum gonadotrophins performed by laboratories in the UK EQAS during the 1980s are reviewed, with particular reference to the effects of the recent introduction of immunometric assays (IMA) as an alternative to radioimmunoassay (RIA). IMA gave results which were on average 17% higher than RIA for FSH, and 33% lower for LH. These bias characteristics were not entirely accounted for by differences in assay standardisation, but appeared to reflect the different isoforms of the hormones detected by the monoclonal antibodies used in the IMA. Between-laboratory agreement remained, consequently, unsatisfactory overall (geometric coefficient of variation, GCV, 20-30%), although good within method groups (GCV 10%). IMA were less vulnerable to non-specific background interference than many RIA, and could avoid interference from HCG. Some IMA were, however, vulnerable to interference from heterophilic antibodies in patients' sera. The differences between RIA and the various IMA in numerical values reported, and in their vulnerability to interferences underline the need for care in interpreting assay results.

A search was made for associations between poor performance in the UK External Quality Assessment Scheme (EQAS) for serum growth hormone (GH), and a range of factors including assay method, laboratory workload and staffing, and Internal Quality Control (IQC) procedures. On the basis of the factors identified as being associated with poor performance we recommend the following. 1. Laboratories using RIA for GH should routinely analyse samples at two dilutions and report a mean result. 2. The use of 125I-GH which is 5 or more weeks old should be avoided. Tracer should also be chromotographed to remove aggregate before use. 3. Laboratories using RIA should avoid using a standard curve which covers too wide a range of concentration; a curve midpoint (ie GH concentration to reduce the zero standard binding by 50%) of about 8 mU/l or less is probably acceptable. 4. It should be noted that high workloads present a risk of some loss in quality of performance. 5. An experienced member of staff other than the assayist should be responsible for checking IQC data. 6. Laboratories which do not have the resources to maintain fully their own RIA as outlined above should carefully consider use of an unbiased, precise IRMA. The UK EQAS has identified two assays (Boots-Cell-tech Sucrosep, NETRIA) that appear to meet these criteria [2]. The above observations may also be relevant to immunoassays for other peptide hormones.

Between-laboratory agreement in the UK EQAS for maternal serum alphafetoprotein has improved steadily since 1976 and the geometric coefficient of variation is now 8 to 9% at levels of 50 to 150 kU/L. The use of a common standard and commercial assay kits appear to have contributed to this trend. Within-laboratory performance is also generally good, about 50% of participants maintain a bias of less than 5%, together with a scatter of the bias of less than 10%. These data indicate that the quality of assay performance is adequate for the requirements of screening programmes for open neural tube effects. The improvement in laboratory performance is such that between-laboratory agreement is better expressed in kU/L than as multiples of the median. Errors in interpretation of clearly normal or abnormal results appear to be rare (0.4%), and contribute little to overall false positive and negative rates. However, they assume significance as most are due to avoidable errors.

The performance of laboratories in the UK External Quality Assessment Scheme for growth hormone (GH) during the years 1980 to 1987 is reviewed. The number of participating laboratories has increased steadily and is at present 67; about one half use immunoradiometric assay (IRMA) kits and the use of such kits is increasing at the expense of 'in-house' radioimmunoassays (RIAs). The consensus mean, which is used as the target value for assessing performance, has remained accurate and reproducible against this changing background. The between-laboratory geometric coefficient of variation has remained at about 18% during the period reviewed, revealing unsatisfactory between-laboratory agreement. This is in part due to poor within-laboratory performance in a small proportion of laboratories but it is also due to the negative bias of some IRMA kits. Most IRMA kits do appear, however, to provide marginally better within-laboratory precision than RIA, and are less vulnerable to non-specific interference. The laboratory interpretation of results was assessed from time to time, and was generally satisfactorily performed. In an attempt to identify the causes of poor performance, a detailed survey of assay methods and laboratory practice has been carried out; the results are described in an associated report [1].

We have compared the laboratory performance of immunoradiometric (IRMA) and radioimmunoassay (RIA) methods developed in this laboratory for measurement of serum prostatic acid phosphatase (PAP). The IRMA utilizes a radiolabelled mouse monoclonal anti-PAP and a solid phased rabbit polyclonal anti-PAP. The same rabbit antibody is used in the RIA. The IRMA shows excellent precision over a much wider working range (0.25–1000 μg/l) than the RIA (0.73–14.0 /gmg/l), and can be completed in 5 h, while the RIA requires 3 days. Levels in healthy males and in patients with benign prostatic hypertrophy are similar in both assays, upper limits of normal being 1.8 μg/l (IRMA) and 4.7 μg/l (RIA). The two assay methods correlate very well (r = 0.97) when PAP is measured in serum from prostatic cancer patients, although IRMA results are generally lower than those obtained by RIA. About 20% of patients with non-metastatic prostatic carcinoma had elevated serum PAP, whereas about 80% of those with metastatic disease had raised levels. The diagnostic efficiencies of the RIA and IRMA appeared similar. The value of the IRMA in follow-up and staging remains to be determined.

We have compared the results of serum thyrotrophin (TSH) measurements using a sensitive immunoradiometric assay (IRMA) with those of conventional thyroid function tests in 299 hospital inpatients with a range of non-thyroidal illnesses. Levels of total thyroxine (T4), free T4, total tri-iodothyronine (T3) and free T3 in the hypothyroid range were recorded in 8%, 15%, 19% and 49% of patients, respectively, whereas TSH (IRMA) was abnormally low in 1%. Furthermore, basal TSH (IRMA) accurately predicted the result of the thyrotrophin-releasing hormone test in 72 of the 74 patients in whom this test was performed and, unlike thyroid hormone measurement, identified patients with subclinical thyroid disease. It would appear that a single basal TSH (IRMA) measurement is the most appropriate screening test for thyroid dysfunction in patients with concomitant acute or chronic illness.