The efficacy of hormone assays in the differential diagnosis of amenorrhea and menopause.

The differential diagnosis of menopause and amenorrhea is currently based on the assay of circulating follicle stimulating hormone, luteinising hormone, and estradiol. The diagnostic performance of the three hormone assays, both as single and combined tests, was evaluated considering as reference data the results from 300 subjects for either condition, and assuming menopause-amenorrhea prevalence ratios corresponding to 1 and 10. In the calculation an "allocation" scheme was adopted, and the uncertainty associated with the diagnostic performance parameters was accounted for. The results obtained clearly demonstrate that the addition of a test for luteinising hormone or estradiol (or both) to the testing for follicle stimulating hormone is not justified in terms of improvement of the diagnostic information.

The uncertainty associated with the predictive value of test results.

The uncertainty associated with predictive value of test results was taken into consideration, as concerns both sampling error (related to the size of the statistical reference samples) and analytical imprecision (unavoidably involved by the measurement itself). A software package, developed for the statistical calculations, was used for the treatment of the results obtained for serum free thyroxin in euthyroid and dysthyroid subjects, assumed as an experimental model. Examples are shown for the obtainable functions predictive value vs. estimate and the related uncertainty regions. These data could help in comparing test results and, in particular, in preparing fully informative laboratory reports. The difficulties involved by an extensive application of the procedure are discussed.

Error models and quality control performance.

As an alternative to the oversimplified error schemes currently adopted in establishing quality control (QC) strategies, a complex model was assumed implying (a) the distribution of errors (critical error is regarded as a value discriminating between "effective errors" to be detected and "subcritical errors" which do not interfere with the medical decision whose detection is considered as a false-reject signal), and (b) the possibility of simultaneous losses of precision and accuracy. The control data recorded for digoxin radioimmunoassay over a one-year period were used for (1) deriving the probability density functions of random and systematic errors, through a within-run across-level normalisation procedure; (2) obtaining the functional relationships between the critical random or systematic error and the QC performance statistics (sensitivity, specificity, predictive value), weighted for the error prevalences, through integration of the probability density functions and the power functions associated with an exemplifying control rule; and (3) describing the functions which correlate the corrected performance statistics with the allowable error (whose individual values account for all possible combinations of critical random errors and critical systematic errors), by extending to the tridimensional space the above procedures. Analysis of the resulting data shows that it is necessary to revise the criteria for the choice and optimisation of QC schemes.

Italian external quality assessment scheme in immunoassay.

This paper deals with the organization, the data processing and some of the results obtained in Italian external quality assessment (EQA) schemes for hormones, tumor markers and hepatitis B markers. The EQA for hormones and tumor markers includes up to sixteen analytes together with the participation, in 1990, of about 250 laboratories. Laboratory results were used to prepare periodic and end-of-period reports. The former includes the results (with the related statistical parameters) obtained by all participants and by laboratories using the same method, as well as the histogram of the data. The end-of-period report contains estimates of imprecision and average bias for all laboratories, for each laboratory and for the more widely employed kits. From 1980 to 1988, laboratory variability improved significantly for TSH, progesterone, estradiol, testosterone, CEA and ferritin, slightly for cortisol, FSH, prolactin and AFP, while there was no improvement for both total T3 and T4. For LH we found an unusually high variability mainly due to systematic differences between kits based on different monoclonal antibodies. About 200 laboratories participated in the EQA for hepatitis B markers (HBsAg and anti-HBs) organized in 1990. For these analytes the periodic reports show the percentage of negative and positive results and the histogram of the responses (absorbance or counts) normalized with respect to the cut off.

Performance assessment of coupled tests: the effects of statistical non-independence.

Limited to two-test associations (series and parallel schemes), the effects of statistical non-independence were studied through a mathematical approach and an experimentally-based evaluation. Both procedures were applied to results for total hormones and free fractions in euthyroid and dysthyroid subjects. Assuming independence, the sensitivity of combined tests was found to increase in parallel coupling, and to decrease, symmetrically, in series coupling, depending critically on the degree of between-test correlation and on the value of single test sensitivity (the opposite modifications obviously occur for specificity). A more complicated situation resulted for the predictive value of test associations, where a prediction based on a mathematical model was found not to be generally valid; in this case, calculations using the correct values of conditional probabilities of coupled tests seemingly remain the safest procedure.

Interassay variability of immunometric methods for thyrotropin in an external quality assessment survey: evidence that functional sensitivity is not always adequate for clinical decisions.

We investigated the ability of current immunometric methods for thyrotropin (TSH; thyroid-stimulating hormone) to distinguish between low-normal and subnormal hormone concentrations by using the data from an external quality assessment (EQA) survey in 1990. We computed the interassay (between-run) precision profiles from results from 101 laboratories, which used the five most popular kits in the survey; during the control period (one year) each laboratory assayed 4 EQA pools distributed (as hidden replicates) in five occasions. The interassay CV was relatively low (9-13%) for three pools in the normal TSH range (greater than 0.8 milli-int. unit/L) but markedly higher (30-40%, except for one more precise kit) in the subnormal range (0.2 milli-int. unit/L). We calculated the effect of the between-run variability on the diagnostic accuracy (discrimination between normal and subnormal values) for three representative TSH concentrations: 0.2, 0.4, and 0.5 milli-int. unit/L (0.3 milli-int. unit/L was considered the lower normal limit). The three concentrations were reasonably discriminated (P less than or equal to 5%), and only one kit showed a between-run CV less than 18% at 0.2 milli-int. unit/L. For the other four less-precise kits, only the higher TSH value (0.5 milli-int. unit/L) could be classified with an acceptable diagnostic reliability. With the most precise kit, one can distinguish two TSH concentrations in the 0.3-0.5 milli-int. unit/L range that differ by at least 30%; with the other kits, differences greater than 50-60% are needed for reliable discrimination. Thus many laboratories fail to achieve the functional sensitivity of a second-generation assay, even if they use immunometric methods. TSH assays with a better interassay precision in the low concentration range are needed.

The imprecision profile in immunoassay. A study using a resampling technique.

A resampling ('bootstrap') technique was applied to assess the reliability of the calculated imprecision profile (IP), as obtained from the dose/response curve and the response/error relationship (RER) using the cumulative data relative to two assays, i.e. a T4 radioimmunoassay (RIA) and a TSH immunofluorometric assay (IFMA), both run in duplicate. Mean values and the related uncertainty of the estimated dose errors were compared for different RER fitting conditions and different sizes of the duplicate response sets. The following observations were made: (a) compared to the maximum-likelihood procedure, the least-square fit proved to be unsuitable for estimating the parameters in the general RER equation variance(R) = aRb (where R indicates the response), (b) the simplifying assumption of a within-method constancy of the exponent in the RER equation, while acceptable for the T4 RIA, did not hold in the case of the TSH IFMA implying a much wider response range, (c) for both assays, response sets of ca. 100 duplicates were apparently compatible with an acceptable definition of the IP (+/- 10 to +/- 20% uncertainty).

Analytical and clinical assessment of a commercial kit for the simultaneous determination of free thyroxine and thyrotropin.

The SimulTRAC FT4/TSH kit (Becton Dickinson), allowing a simultaneous determination of free thyroxine (FT4) and thyrotropin (TSH), was assessed in terms of analytical quality and clinical performance. The results of a multicenter trial were included in this study to obtain a more complete and reliable information. The current validation procedures (ie evaluation of analytical imprecision and sensitivity, between-kit comparison of estimates and--limited to TSH--check of response linearity on dilution) demonstrated quite acceptable analytical characteristics for both the FT4 and the TSH tests. The diagnostic sensitivity and predictive value were derived for separate and combined tests from a relatively large number of cases (ie 401 euthyroids, 127 overt and 48 subclinical hypothyroids, 205 overt and 80 compensated hyperthyroids). The TSH test proved more effective than FT4 in discriminating both overt and subclinical dysfunctions, and, in this respect, the test association was found to add a little advantage--if any. However, the extension of the concepts of diagnostic efficiency to any combination of test results--favoured by their production in a single assay--provides a basis to establish diagnostic protocols and to predict costs and benefits of medical actions.

Assessment of qualitative tests.

For qualitative (2-class) tests, which provide binary (yes/no) information, the correctness of specimen classification remains the most important criterion for performance evaluation. However, a more informative picture emerges from the relationship between percentage of positive results and analyte concentration, which allows some inherent test characteristics to be derived (the positive/negative discrimination concentration and the "grey zone" around it). The information content of evaluation approaches is decidedly improved by the availability of numerical results (counts per minute, absorbance) in most situations of clinical interest. The concentration/response functions underlying the quality of individual tests may thus be derived and compared, and laboratory staff given a more objective criterion to judge individual performance. Examples are drawn from the authors' experience in running external quality assessment programs for tests for infectivity markers.

External quality assessment of assays for hormones and tumor markers in Italian laboratories.

External quality assessment (EQA) programs run by CNR/Tecnostandard for immunoassays of hormones and tumor markers, started in 1980, presently include as many as 20 analytes; about 300 laboratories are involved in these programs. For all immunoassays submitted to the EQA, the inspection of cumulative results allows the current situation to be documented for total variability and its within-kit and between-kit components (the former accounting for the reproducibility and robustness of the kits and the latter for their systematic differences of estimation). For 13 assays subjected to EQA for longer, the variability trends over time are depicted, and single factors affecting the overall quality of particular assays are identified. Among these, experimental simplification of kit structure, alignment of calibrators with an acknowledged reference material, and adoption of monoclonal-antibody based two-sites assays can be mentioned. On the contrary, neither automation of the procedures nor (more expectedly) increasing use of nonisotopic techniques has proved effective in significantly improving the analytical quality.

Immunoassay external quality assessment in Italy: HBsAg and anti-HBs tests.

The mode of operation of the CNR/Tecnostandard external quality assessment scheme for HBsAg and anti-HBs assays is outlined, and the relevant results reported. Emphasis is given to the retrospective evaluation of data in an attempt to derive a picture of the state of the art in Italian laboratories. The approaches followed for this evaluation include analysis of the rate of correct results and inspection of the relationships between analyte concentration and either the percentage of positive classifications of samples or the numerical test responses. Information is given on the "average" performance of individual kits as actually used in participants' laboratories.

Assessment of response/error relationship and imprecision profile in immunoassay using computer simulation procedures.

Simulation procedures were applied to assess the response/error relationship (RER) and the imprecision profile (IP) for two model assays, a T4 RIA and a TSH IFMA both using duplicate samples. In order to define the reference functions, the mean data obtained in 10 successive experiments for dose/response curve (DR), RER and IP were employed. The following conclusions emerged from the study: (a) run sizes of ca. 100 duplicates can acceptably describe within-assay IPs, irrespective of the data distribution through the dose range; (b) the contribution of DR fitting error to the total variability of estimate can be disregarded in the case of small series but not for the larger ones; (c) the variability components related to the response error can be efficiently controlled by applying criteria based on RER parameters.

ELISA for specific anti-toxoplasma IgM antibodies: aspects related to serum interference.

Two different methods were used to prepare solid-phase antigen (Ag) from soluble extracts of tachyzoites of Toxoplasma gondii: (A) physical adsorption on polystyrene beads; and (B) formaldehyde fixation of Ag previously dried in microtitration wells. In both cases a horseradish peroxidase conjugate with anti-IgM IgG was used as tracer. The assay scheme consisted of sequential incubations of diluted serum samples and tracer solution (1 or 2 h, 37 degrees C), colour development in the presence of substrate (10 min at room temperature), addition of H2SO4, and absorbance reading at 492 nm. In procedure A no cut-off value for positives could be determined owing to a large overlap between positive and negative sera. The extent of overlap directly correlated with the total IgM content of samples. With negative sera similar values were obtained with sensitized and untreated beads: thus a correction could be made by directly subtracting absorbance values determined in parallel runs with uncoated beads. Results with negative sera correlated with total IgM concentration in procedure B also, but much less variability of blank values allowed negative and positive sera to be effectively discriminated. A series of reference positive and negative sera was correctly classified by both procedures A and B. However, the latter appeared preferable, as not requiring blank correction.