Comparison of Antigens Conjugates and Cutoff Values

By far the most important parameter in an ELISA, or in any test that measures a specific autoantibody, is the antigen. Each antigen has to be optimized to detect the clinically important autoantibody it was designed to measure. Very often a native antigen produced in a human or closely related mammal works better than a cloned antigen or an antigen from a source far removed from humans on the evolutionary tree. The interpretation generally given for this finding is that most autoantibodies are produced by an immune response against the antigen in the host. So even though the immune response is abnormal in the sense that the person is reacting against something in their own body, it is a typical antibody response because the autoantibodies are exquisitely tuned to recognize the antigen that stimulated them in the first place.

Native antigens may work best for a number of reasons. Some autoantibodies recognize epitopes that are expressed in a macromolecular complex comprised of two or more separate macromolecules. There are diagnostically important antibodies in SLE and drug-induced lupus patients that recognize chromatin but not isolated DNA-free histones or histone-free DNA [55, 56]. Similarly, antibodies recognizing the native RNP particle, but not the individual proteins or RNA moiety, are found in patients with SLE and mixed connective tissue disease [57]. In some cases autoantibodies recognize parts of proteins that are changed by post-translational modifications. The best example of this is the recently discovered autoantibody reactivity in RA patients that recognize only proteins whose arginines have been changed to citrulline [5]. Finally, there are numerous examples of autoantibodies that recognize conformational epitopes that are present on the native form of the protein but not on the denatured protein [58].

For the above reasons, cloned proteins expressed in bacteria or insect cells, and synthesized peptides, rarely work as well as native antigen to detect autoan-tibodies. Some exceptions to the above statement are autoantibodies against ri-bosomal P proteins that react with a 23-amino-acid peptide [59] and autoantibodies to SS-A/Ro-52 that react with the denatured protein better than the native [60].

The most important feature in the conjugate is the class of immunoglobulin that it recognizes. The main choices are the class-specific antibodies that recognize the heavy chains of IgG, IgM, or IgA and the polyspecific conjugates that react with all the heavy chains. Antibodies that recognize the kappa and lambda light chains are also polyspecific because all classes of immunoglobulins have the same light chains. For some tests, such as IgM RF, IgA anti-tTG, and IgG anti-CCP, there is no debate about the most important class of immunoglobulin to measure. However, for many other very common tests there is no consensus. This is particularly true for IIF on HEp-2 cells, anti-DNA by both ELISA and IIF, anti-histone, and anti-MPO and anti-PR3. Some manufacturers make kits with IgG-specific conjugates, while others are made with polyspecific conjugates. Any attempt to make international standards for these antibodies needs to take into account the different specificities of the conjugates that are used in different laboratories.

The cutoff value between negative and positive in a test that measures autoan-tibodies is extremely important. In subjective tests such as IIF, the technician examining the slide under the microscope has to decide what strength of im-munofluorescence and what pattern should be considered positive. Different people and microscopes may give different results. The results with Ouchter-lony immunodiffusion are less subjective than IIF but still depend on the sharpness of the eyes, the indirect light source, and the attention to detail of the viewer.

One of the main advantages of ELISA over many other techniques to measure autoantibodies is that it is objective. The optical density (O.D.) of the patient is compared to that of a calibrator or standard curve and given a value. The biggest challenge to anyone developing an ELISA is setting the value that divides negative or indeterminate from positive. Kits made by various in-house methods and by different manufacturers may yield opposite results on a given sample simply because the cutoffs are quite different. This dichotomy arises because different statistical methods and different control groups are used to determine the cutoff between negative and positive.

The most egregious approach occurs when normal blood donors are used as the control group and a value of two or three standard deviations above the mean is chosen as the cutoff. In this case, both the control group and the statistical analysis are inappropriate. Very rarely is the blood from a healthy person, such as the average blood donor, sent to a clinical laboratory for autoantibody tests. Usually the person is sick. Thus, the correct control group should consist of people with autoimmune diseases who are expected to be negative for the au-toantibody in question and people with infectious diseases with symptoms similar to those of an autoimmune disease. Very often these latter groups have higher antibody levels, and higher binding on ELISA, than normal blood donors. The average and standard deviation are useful statistical tools only when the distribution of values in a population yields a normal, or bell-shaped, curve. In an ELISA, the distribution of values in the negative population is not bell shaped but usually resembles the right half of a bell. That is, a large percentage of the population will yield the lowest results, with smaller and smaller percentages yielding higher and higher results.

The best way to determine the correct cutoff is to perform a simple nonpara-metric statistical analysis of the expected negative and positive groups [61]. All the samples should be put in rank order from highest O.D. to lowest O.D. At least 60, but preferably about 200, patients who are expected to be negative for the autoantibody but have symptoms related to the disease in question should be tested. In addition, more than 20, or as many as possible, people with the disease or known to be positive for the autoantibody should also be tested. After all the patients and controls are put in rank order, one can examine the results and determine the best cutoff to yield the optimum sensitivity and specificity for that ELISA.

A recent paper showed great differences in the specificity, but only moderate changes in the sensitivity, of the anti-chromatin ELISA depending on the cutoff that was chosen. When a cutoff was used that yielded 98% specificity (i.e., 2% positive) in blood donors, the sensitivity was 86% in SLE patients, the group who were expected to be positive [62]. However, the specificity in other groups that were expected to be negative was poor. People with infectious diseases also showed 2% false-positive results, while 13% of SSc patients were positive. With this cutoff this test has a high sensitivity but poor specificity, and it may not help doctors in their diagnosis. When the value for the cutoff was raised so that no blood donors were positive, then no one with infectious diseases or SSc was positive, while the sensitivity in the SLE patients only dropped to 71% (A. Doria, personal communication). With this cutoff the test is still quite sensitive, but is now very specific and has significant clinical utility.

0 0

Post a comment