Mario Werner

Comparative evaluation of the diagnosis of acute pancreatitis based on serum and urine enzyme assays

We compared the diagnostic sensitivities of serum amylase, lipase (assayed enzymatically and immunologically), trypsinogen and elastase-1, the 2-h-timed urine amylase excretion and the ratio of amylase and creatinine clearances in the recognition of acute pancreatitis. Serial serum and urine findings from 39 patients with acute pancreatitis, and from 42 patients with non-pancreatic causes of abdominal pain (controls), as well as findings from 24 healthy subjects (normals) were studied. Decision thresholds were established for each parameter using either the control or the normal population, and the resulting diagnostic sensitivities determined. On hospital admission, all serum assays were equally sensitive, but on subsequent days lipase, trypsinogen and elastase-1 assays all significantly surpassed the sensitivity of the serum amylase assay. On the second and subsequent hospitalization days, determination of timed urine amylase excretion offered no advantage over the serum amylase, and the ratio of amylase and creatinine clearances lacked discrimination altogether.

Appropriate utilization and cost control of the hospital laboratory: panel testing and repeat orders

Analysis of hospital laboratory utilization shows a bi-exponential relationship between the many different test combinations ordered and their frequency of use. This suggests two opposite strategies to reduce test volume. Policy could discourage use of the most popular request combinations to reduce volume, or policy could discourage use of the least popular request combinations to reduce order variety. Analysis of biochemical test orders further suggests that the largest part of test requests is associated with a small number of pathophysiological issues, namely those involving electrolytes, fluid balance, blood lipids, heart, liver, kidney or bone disease. Therefore, efficiency results when a limited number of standard test panels addressing these issues reduce order variety by discouraging infrequently used test combinations. Standard panels can be developed from an analysis of the most popular existing orders, accepting some degree of discrepancy among otherwise overlapping requests. This empiric approach avoids sterile debate about the composition of ideal panels. Since the resources of the laboratory, the mix of patients served and local medical tradition affect test orders, each institution providing health care must develop the standard panels best suited to its needs. Analysis of standing repeat orders suggests they are often used to anticipate potential pathology in stable patients. Therefore, economy results when test volume is reduced by discouraging series of futile repeat tests with unchanging normal results. Unfortunately, an empiric approach to regulate repeat orders is not available and only a more cumbersome process to build a locally agreed upon consensus between providers and users of laboratory services offers itself as a promising solution.

Linking analytic performance goals to medical outcome

As laboratorians relate analytic performance to medical goals, they face complex choices among competing subjective and objective criteria for the assessment of acceptable analytic error. Defining desirable performance as some fraction of physiologic variability provides potentially excessive benchmarks for quality. More important than the recognition of health are the clinical decisions which deal with diseases, particularly the latters degrees of severity and the different medical actions these prompt. It is equally essential to take into account the reasoning by which physicians arrive at these decisions, since their mental processes condition desirable performance goals. Considering these modalities, a universal model of analytic performance requirements uniformly applicable to all measured parameters clearly cannot be devised. Rather, tolerance limits for analytic error must be tailored to specific medical problems. To facilitate this seemingly Herculean task, this paper develops concepts and principles derived from operation research, and illustrates their application by three examples. The generic conclusion evidenced by the latter is that the linkage between analytic performance goals and medical strategies is reciprocal, namely that outcome can just as well be optimized by tailoring medical strategy to existing analytic performance as by adapting analytic performance to medical strategy.

Diagnostic discrimination and cost effective assay strategy for leukocyte alkaline phosphatase.

Decreased leukocyte alkaline phosphatase (LAP) is a first line test for chronic myelogenous leukemia (CML), generally preceding a diagnostic algorithm which also includes bone marrow biopsy, cytogenetic analysis, and molecular diagnostics. We found the analytical uncertainty of LAP assays to range from over 100% coefficient of variation at low scores to about 20% at high scores. However, the receiver-operator characteristics derived from LAP determinations in 50 consecutive cases suggest that a suitably high diagnostic decision threshold still can eliminate false negatives. As a consequence of such a strategy, as many as half the tested patients, classified unequivocally as the negatives, can avoid further invasive and costly workup. On the other hand, serial LAP determinations, whether performed to detect change to a lower or to a higher score, are unlikely to produce conclusive diagnostic signals exceeding analytical noise.

Optimization of diagnostic discrimination applied to the amniotic fluid lecithin/sphingomyelin ratio

An unrecognized disparity frequently separates the intuitive reliance placed on laboratory information by those who produce it and those who use it. To harmonize these dissimilar perceptions requires quantitative measures of the information contained in laboratory data. The assay of the lecithin/sphingomyelin ratio in amniotic fluid is used as an example to analyze the discriminatory ability of a test. This assay can be optimized from the different viewpoints of sensitivity, specificity, diagnostic effectiveness or value. In each case test results are interpreted in the light of different factors and so the optimal decision threshold to separate positive from negative results may vary. Appreciation of this concept would endow the interpretation of laboratory tests with the same flexibility traditionally available to other clinical information.

Comparative clinical evaluation of biochemical and genomic tumor markers

The clinical utilities of established biochemical tumor markers and of emerging genomic markers are compared by six formal criteria: [1] tests negative in health or benign disease, [2] produced exclusively by specific tumor cells, [3] present frequently in the targeted malignancy, [4] detectable in occult disease, [5] degree of expression reflects tumor burden and prognosis, and [6] degree of expression correlates with therapeutic result. Evaluation of eight widely accepted marker systems combining a biochemical indicator with a specific cancer, on the one hand, and five representative genomic marker-target combinations involving chromosomal translocation, gene amplification and mutation, on the other hand, produces three main conclusions: First, specified applications are sufficiently well documented for the best biochemical markers to now tailor analytical performance goals to these uses. Second, further clinical trials of genomic markers are needed to document the useful linkage of specific indicators with specific clinical problems. Third, the different profiles of marker characteristics defining the two classes of indicators suggest some mutually complementary uses.

Strategies to Integrate Laboratory Information into the Clinical Diagnosis of Hepatic and Acute Pancreatic Disease

Liver diseases provide a general model for the formal analysis of the relative diagnostic weights of information obtained from different sources. Diagnosis of liver disease has three basic purposes: (1) differentiation of hemolytic, parenchymatous, and obstructive jaundice, (2) evaluation of disease course and therapeutic success, and (3) evaluation of the nature of the liver lesion. The capabilities of diagnostic findings to attain these goals can be measured according to three criteria: (1) selectivity in differential diagnosis and etiologic selectivity, (2) sensitivity, and (3) specificity with regard to morphologic changes. Applying these yardsticks to information obtained from signs and symptoms, biophysical findings and laboratory findings establishes: (1) clinical findings and liver function tests are the underpinnings of differential diagnosis, (2) liver function tests are ideal for the evaluation of the disease course, and (3) laparoscopy and liver biopsy are ideal for the evaluation of the nature of the liver lesion, but imaging procedures and liver function tests are frequently simpler means that contribute to it.