Symposium on Rheumatic Diseases
Laboratory Tests in Rheumatic Diseases Wallace V. Epstein, M.D. *
The development of the modern clinical immunology laboratory is, to a significant degree, a by-product of the evolution of contemporary clinical rheumatology. No matter what the eventual utility of tests for rheumatoid factors, the discovery of such a circulating antibody with specificity for a human immunoprotein provided a powerful impetus for the growing concept of immunologically mediated chronic diseases. During the period of emphasis on humoral aberrations in rheumatic disorders, each new-found antibody associated with one or another rheumatic disorder furthered classification of disease on the basis of serologic characteristics. A polarity developed in which a nosology based on clinical and historical characteristics contrasted with rapidly evolving serologic profiles. The implications of the latter for etiology and pathogenesis were so powerful as to dominate evolving classifications of disease. Rheumatoid arthritis could be classified as either a seropositive or a seronegative disease and systemic lupus erythematosus became further defined by a wide variety of antibodies, immune complexes, and complement characteristics, each describing a disease subset with unique prognostic and therapeutic implications. The pattern of a new test showing extraordinary specificity when first described, and then gradually decreasing specificity with further clinical trial and modifications of test procedures, is well known. Major methodologic problems exist in the selection of the abnormal sera and the control sera used for evaluation of the test, and rarely have initial descriptions of the test adequately defined the accuracy and precision of the serologic technique. The population base used for estimation of the prevalence of the phenomenon is almost never broad enough to allow extrapolation of the test results from the research laboratory to the community. At present, we have the situation of changing clinical categories of rheumatic disease such as the emergence of mixed connective tissue diseases and the overlap syndromes simultaneously with the description of immunologic ';Professor of Medicine, University of California School of Medicine, San Francisco, California Preparation of this article was supported by a grant from the Robert Wood Johnson Foundation, Prince ton, New Jersey. The opinions, conclusions, and proposals in the text are those of the author and do not necessarily represent the views of the Robert Wood Johnson Foundation. Medical Clinics of North America - Vol. 61, No. 2, March 1977
377
378
WALLACE
V.
EpSTEIN
abnormalities which, in themselves, define subsets of disease. The demonstration of serum antibody to extractable nuclear antigens (ENA) strongly influenced the emergence of the entity, mixed connective tissue disease (MCTD). The laboratory de terminations used in the care of patients with rheumatic diseases share with other in vitro tests common problems related to the technique as well as to the motivation that led to ordering the test. The clinician must have well formulated reasons for ordering a single laboratory test or a group of tests and must decide, in advance, how the information will be used in diagnostic, therapeutic, and prognostic decision-making. Some relevant factors we will not discuss in this paper but which require consideration include simplicity, patient acceptability, and cost of laboratory tests. The precision of a test refers to the consistency of repeated measurements, an important characteristic if you plan to follow serum complement or DNA antibody levels. The accuracy of the test refers to the closeness of the test result to a true measure of the attribute being sought. The diagnostic sensitivity of a test expresses how frequently a test is positive when a particular disease is present, the disease, of course, not having the test itself as part of the diagnostic criteria. The diagnostic specificity of a test is an expression of how frequently a test is negative in the absence of a particular disease. We require such information concerning the tests and the laboratory performing the tests in order to determine how this information will be used, especially when tests are repeated frequently and changing results are used to guide management. As pointed out by Galen and Gambino,7 "accuracy and precision aren't good enough." It is the predictive value of a test which we frequently seek, but this depends on yet another factor-the prevalence of the disease under study. "The predictive value of a positive result defines the percentage of all positive results that are true positives. The predictive value of a test is determined by the complex interaction of three variables: sensitivity, specificity, and prevalence."7 For example, if we have a test which has a sensitivity of 95 per cent and a specificity of 95 per cent, the prevalence of the disease has a striking effect on the predictive value of our test (see Table 1). Beside the characteristics that cumulate in its predictive value we have the issue of the "normal range" of a test. The motivation for performing clinical laboratory tests ranges from screening to case finding to diagnosis, and each of these may have a normal test value. As shown in Figure 1 from Sackett,1s the value designated "A" would be an acceptable normal for someone screening for an illness, since no diseased patients would be missed. Someone doing case finding would select "B" as the limit of normal, since a positive test among disease-free persons is a false-positive for their purpose. Most interesting is the concept of "c" as a normal. This is a value selected as a guide to initiate therapy. In the usual clinical uses of laboratory test results there can therefore be at least three different "normal" limits. The applicability of this concept to a "normal" value of the serum total hemolytic complement while managing patients with systemic
379
LABORATORY TESTS IN RHEUMATIC DISEASES
Table 1.
Effect of Prevalence of Disease on Predictive Value of Tests (Sensitivity 95%; Specificity 95%) PREDICTIVE VALUE OF DISEASE PREVALENCE
A POSITIVE TEST
O.IO/C
1.9%
16.1% 27.9% 50.0% 45.0%
1.0%
2.0% 5.0';7,· 50.0%
Note: The profound effect of prevalence on predictive value is independent of both the accuracy and precision of the test. Since most of the rheumatic diseases have a prevalence in the general community in the range 0.1 to 1.0 per cent, we see that even if we had tests of 95 per cent sensitivity and specificity, the tests would still have only 1.9 to 16.1 per cent predictive value when used for screening. From Galen, R. S., and Gambino, S. R.: Beyond Normality. New York, John Wiley and Sons, 1976. Reproduced with permission.
lupus erythematosus (SLE), or of the latex F II rheumatoid factor test to diagnose rheumatoid arthritis should be immediately apparent.
Ordering Specific Laboratory Tests At present no requirements exist that the laboratory test being ordered is cost effective. Few studies explain the wide variation of test-ordering behavior among physicians; further, few studies relate the manner and frequency of repetition of tests to outcome of disease. Bombardier et al,2 have reported an 81 per cent range of variation for laboratory tests ordered by university-based physicians caring for ambulatory patients with a similar form of rheumatoid arthritis without detectable differences in short term outcomes. The use of laboratory services in the patient care setting in contrast to the clinical research setting raises issues of cost effectiveness as well as the relation to the outcome of management. Almost no information concerning these issues is available at present. The following presentation deals with issues of test sensitivity and specificity but we must recognize that in a
DISEASE - FREE
Laboratory Tests in Rheumatic Diseases Wallace V. Epstein, M.D. *
The development of the modern clinical immunology laboratory is, to a significant degree, a by-product of the evolution of contemporary clinical rheumatology. No matter what the eventual utility of tests for rheumatoid factors, the discovery of such a circulating antibody with specificity for a human immunoprotein provided a powerful impetus for the growing concept of immunologically mediated chronic diseases. During the period of emphasis on humoral aberrations in rheumatic disorders, each new-found antibody associated with one or another rheumatic disorder furthered classification of disease on the basis of serologic characteristics. A polarity developed in which a nosology based on clinical and historical characteristics contrasted with rapidly evolving serologic profiles. The implications of the latter for etiology and pathogenesis were so powerful as to dominate evolving classifications of disease. Rheumatoid arthritis could be classified as either a seropositive or a seronegative disease and systemic lupus erythematosus became further defined by a wide variety of antibodies, immune complexes, and complement characteristics, each describing a disease subset with unique prognostic and therapeutic implications. The pattern of a new test showing extraordinary specificity when first described, and then gradually decreasing specificity with further clinical trial and modifications of test procedures, is well known. Major methodologic problems exist in the selection of the abnormal sera and the control sera used for evaluation of the test, and rarely have initial descriptions of the test adequately defined the accuracy and precision of the serologic technique. The population base used for estimation of the prevalence of the phenomenon is almost never broad enough to allow extrapolation of the test results from the research laboratory to the community. At present, we have the situation of changing clinical categories of rheumatic disease such as the emergence of mixed connective tissue diseases and the overlap syndromes simultaneously with the description of immunologic ';Professor of Medicine, University of California School of Medicine, San Francisco, California Preparation of this article was supported by a grant from the Robert Wood Johnson Foundation, Prince ton, New Jersey. The opinions, conclusions, and proposals in the text are those of the author and do not necessarily represent the views of the Robert Wood Johnson Foundation. Medical Clinics of North America - Vol. 61, No. 2, March 1977
377
378
WALLACE
V.
EpSTEIN
abnormalities which, in themselves, define subsets of disease. The demonstration of serum antibody to extractable nuclear antigens (ENA) strongly influenced the emergence of the entity, mixed connective tissue disease (MCTD). The laboratory de terminations used in the care of patients with rheumatic diseases share with other in vitro tests common problems related to the technique as well as to the motivation that led to ordering the test. The clinician must have well formulated reasons for ordering a single laboratory test or a group of tests and must decide, in advance, how the information will be used in diagnostic, therapeutic, and prognostic decision-making. Some relevant factors we will not discuss in this paper but which require consideration include simplicity, patient acceptability, and cost of laboratory tests. The precision of a test refers to the consistency of repeated measurements, an important characteristic if you plan to follow serum complement or DNA antibody levels. The accuracy of the test refers to the closeness of the test result to a true measure of the attribute being sought. The diagnostic sensitivity of a test expresses how frequently a test is positive when a particular disease is present, the disease, of course, not having the test itself as part of the diagnostic criteria. The diagnostic specificity of a test is an expression of how frequently a test is negative in the absence of a particular disease. We require such information concerning the tests and the laboratory performing the tests in order to determine how this information will be used, especially when tests are repeated frequently and changing results are used to guide management. As pointed out by Galen and Gambino,7 "accuracy and precision aren't good enough." It is the predictive value of a test which we frequently seek, but this depends on yet another factor-the prevalence of the disease under study. "The predictive value of a positive result defines the percentage of all positive results that are true positives. The predictive value of a test is determined by the complex interaction of three variables: sensitivity, specificity, and prevalence."7 For example, if we have a test which has a sensitivity of 95 per cent and a specificity of 95 per cent, the prevalence of the disease has a striking effect on the predictive value of our test (see Table 1). Beside the characteristics that cumulate in its predictive value we have the issue of the "normal range" of a test. The motivation for performing clinical laboratory tests ranges from screening to case finding to diagnosis, and each of these may have a normal test value. As shown in Figure 1 from Sackett,1s the value designated "A" would be an acceptable normal for someone screening for an illness, since no diseased patients would be missed. Someone doing case finding would select "B" as the limit of normal, since a positive test among disease-free persons is a false-positive for their purpose. Most interesting is the concept of "c" as a normal. This is a value selected as a guide to initiate therapy. In the usual clinical uses of laboratory test results there can therefore be at least three different "normal" limits. The applicability of this concept to a "normal" value of the serum total hemolytic complement while managing patients with systemic
379
LABORATORY TESTS IN RHEUMATIC DISEASES
Table 1.
Effect of Prevalence of Disease on Predictive Value of Tests (Sensitivity 95%; Specificity 95%) PREDICTIVE VALUE OF DISEASE PREVALENCE
A POSITIVE TEST
O.IO/C
1.9%
16.1% 27.9% 50.0% 45.0%
1.0%
2.0% 5.0';7,· 50.0%
Note: The profound effect of prevalence on predictive value is independent of both the accuracy and precision of the test. Since most of the rheumatic diseases have a prevalence in the general community in the range 0.1 to 1.0 per cent, we see that even if we had tests of 95 per cent sensitivity and specificity, the tests would still have only 1.9 to 16.1 per cent predictive value when used for screening. From Galen, R. S., and Gambino, S. R.: Beyond Normality. New York, John Wiley and Sons, 1976. Reproduced with permission.
lupus erythematosus (SLE), or of the latex F II rheumatoid factor test to diagnose rheumatoid arthritis should be immediately apparent.
Ordering Specific Laboratory Tests At present no requirements exist that the laboratory test being ordered is cost effective. Few studies explain the wide variation of test-ordering behavior among physicians; further, few studies relate the manner and frequency of repetition of tests to outcome of disease. Bombardier et al,2 have reported an 81 per cent range of variation for laboratory tests ordered by university-based physicians caring for ambulatory patients with a similar form of rheumatoid arthritis without detectable differences in short term outcomes. The use of laboratory services in the patient care setting in contrast to the clinical research setting raises issues of cost effectiveness as well as the relation to the outcome of management. Almost no information concerning these issues is available at present. The following presentation deals with issues of test sensitivity and specificity but we must recognize that in a
DISEASE - FREE
