Serum C3 Levels Are Diagnostically More Sensitive and Specific for Systemic Lupus Erythematosus Activity Than Are Serum C4 Levels

Serum C3 Levels Are Diagnostically More Sensitive and Specific for Systemic Lupus Erythematosus Activity Than Are Serum C4 Levels Denise M. Ricker, MD, Lee A. Hebert, MD, Richard Rohde, MS, Daniel D. Sedmak, MD, Edmund J. Lewis, MD, John D. Clough, MD, and The Lupus Nephritis Collaborative Study Group* • To determine whether serum C3 or C4 is more likely to be normal during systemic lupus erythematosus (SLE) remission and abnormal during SLE relapse we studied twelve SLE patients who presented with severe nephritis. The patients were followed long term (12 to 77 months) through multiple relapses (N = 41) and remissions (N = 13) defined by protocol. A total of 471 serum samples were obtained at defined intervals during these relapses and remissions and were analyzed for C3 and C4 levels by two different methods: nephelometry (N) and radial immunodiffusion (R). During SLE remission (defined by protocol and without reference to serum complement levels), C3 measured by Nassay (C3N) and by R-assay (C3R) tended to be normal (specificity of 93% and 71%, respectively). By contrast, C4 measured by N-assay (C4N) and by R-assay (C4R) showed no such tendency (specificity of 50% for both C4N and C4R). During SLE relapse (defined by protocol and without reference to serum complement levels), C3N and C3R were more likely to be abnormal (sensitivity 95% and 85%, respectively) compared with C4N and C4R (sensitivity 56% and 54%, respectively, P < 0.001 compared with corresponding values for the C3 assay). Analysis by receiver-operator characteristic (ROC) curves demonstrated that the reduced diagnostic sensitivity of C4 versus C3 is not explained by use of an inappropriate lower limits of normal (LLN) for C4. Analysis of the regression curves of C4 on C3 in'individual patients demonstrated that the reduced diagnostic sensitivity of C4 relative to that of C3 also is not explained by imprecision of the C4 assay, or by relatively smaller changes in C4 compared with C3 as SLE activity changes. Rather, the reduced diagnostic sensitivity of C4 (failure to be normal during SLE relapse) and reduced diagnostic specificity of C4 (failure to be normal during SLE remission) appears to be explained by the much broader range of normal of C4, relative to C3, and the increased prevalence of C4 null genes in the SLE population. Analysis of the slopes of the individual regressions of C4 on C3 also demonstrates that net molar consumption (moles consumed - moles produced) of C3 is approximately 6 times greater than that of C4, a value similar to that which can be predicted from previous in vitro studies of classical pathway activation. If serum complement levels (C3, C4) are used to monitor SLE activity in patients with a history of renal manifestations, it is sufficient to measure only C3 levels. This should simplify and reduce the cost of monitoring SLE patients.

© 1991 by the National Kidney Foundation, Inc. INDEX WORDS: Complement C3 and C4; systemic lupus erythematosus.

I

T IS GENERALLY accepted that the hypocomplementemia noted in active systemic lupus erythematosus (SLE) is largely the result of activation of the classical complement pathway.l Thus, in active SLE both C3 and C4 are consumed. The extent to which plasma C3 and C4 levels change in active SLE is determined by the rate of consumption relative to the rate of synthesis of the respective complement component. Classical pathway activation should result in a From the Department ofMedicine and Pathology, The Ohio State University, Columbus, OH; the Department ofMedicine, Rush Presbyterian College ofMedicine, Chicago, IL; and the Department ofRheumatic and Immunologic Diseases, Cleveland Clinic Foundation, Cleveland, OH. * See Appendix for a list of members of The Lupus Collaborative Study Group. Supported in part by National Institutes of Health Grants No. HL25404, AM27770, and RR00034. Address reprint requests to Lee A. Hebert, MD, Ohio State University Medical Center, 1654 Upham Dr, Room N21O, Columbus, OH 43210-1228. © 1991 by the National Kidney Foundation, Inc. 0272-6386/91/1806-0009$3.00/0 678

greater consumption of C3 than C4 because, based on in vitro studies, incorporation of one C4b molecule into the classical pathway C3 convertase, C4b2b, requires the activation of several hundred C4 molecules, but one C4b2b molecule can result in the activation of2,000 C3 molecules before the C4b2b molecule decays.2 Thus, based on these estimates, one C4 molecule results in the consumption of about six C3 molecules. In addition, C3b generated by the classical pathway can result in the formation of the alternative pathway C3 convertase C3bBb. 1 This would lead to further consumption of C3 relative to C4. On the other hand, C3 is present in plasma in greater molar quantity than C4 l and has a synthesis rate that may be greater than that of C4. 3,4 In addition, SLE plasma often contains an inhibitor of the amplification convertase C3bBb. 5 These conditions should help maintain C3 levels, relative to C4 levels, during classical pathway activation. Because of these differences in the production and use of C3 versus C4, it is difficult to predict whether C3 or C4 levels will be more responsive

American Journal of Kidney Diseases, Vol XVIII, No 6 (December), 1991: pp 678-685

679

C3 AND SLE ACTIVITY

to changes in SLE activity. Clinical studies that have compared the diagnostic sensitivity and specificity ofC3 and C4 have been contradictory. Some studies suggest that, compared with serum C4 levels, serum C3 levels are a better reflection of the degree of SLE activity.6-9 Other studies suggest just the opposite.I,ID,11 In the present study, the diagnostic sensitivity of serum C3 versus C4 was determined in SLE patients being closely followed under a study protocol 12 through multiple relapses and remissions. In addition, two different methods were used to measure C3 and C4 levels in each of the patients serum specimens. METHODS

Patients Seventeen consecutive SLE patients with glomerulonephritis followed by one Lupus Nephritis Collaborative Study Group (LNCSG) Center (The Ohio State University, Columbus, OH) in the National Institutes of Health (NIH) multicenter study of plasmapheresis in severe lupus nephritis l2 were evaluated for the present study. Twelve patients were selected because they were followed long term (12 to 77 months) and had multiple sera analyzed (total 471) for C3 and C4 by both nephelometry (N) (performed in the clinical laboratories of The Ohio State University Hospitals) and radial immunodiffusion (R) (performed in the Core Laboratories of the LNCSG at Rush-Presbyterian Hospitals, Chicago, IL). All patients met the American Rheumatologic Association (ARA) revised criteria for classification as SLE and, on entry to the LNCSG Study, had a renal biopsy showing evidence of "severe" glomerulonephritis (World Health Organization [WHO] class IV, or class III or class V with >50% of glomeruli involved with "active" lesions). The mean patient age was 26 years ± 6 years. Eleven patients were female; one was male. Eleven patients were white; one was black.

Definition of SLE Relapse The definitions given below are those used for the LCNSG study. Only those types of relapses observed in the patients in the present study are described. The relapses described in this study occurred after the patients initial presentation with severe glomerulonephritis, and at least 3 months after completing their initial course of treatment, which consisted of prednisone, cyclophosphamide, and, in some patients, a 4week course of plasmapheresis. 13 The relapses identified in the present study are as follows. Minor renal relapse. Reappearance of red blood cell casts in the urinary sediment. Major renal relapse. (1) An increase in serum creatinine of greater than 22 ILmoijL (0.3 mg/dL) if previous value was less than 152 ILmoijL (2.0 mg/dL); greater than 30 ILmoijL (0.4 mg/dL) if previous value was 152 ILmoijL to 381 ILmoijL (2.0 to 5.0 mg/dL); or greater than 45 ILmoijL(0.6 mg/dL) if previous value was greater than 381 ILmoijL (5.0 mg/dL), or (2) significant increase in proteinuria to greater than 1 g/d from previous value ofless than 200 mg/d; or to greater than 2 g/d from previous

value of 200 to 1,000 mg/d; or more than double the previous value if that value was greater than 2 g/d. Minor non renal relapse. (1) SLE skin rash, (2) mild serositis, (3) myalgias or arthralgias, or (4) oral temperature greater than 101°F in the absence of infection. Major nonrenal relapse. Central nervous system involvement (seizures) or retinitis (cytoid bodies). The date of each relapse was taken as the date when the patient was examined and was found to fulfill the criteria of SLE relapse. The serum C3 and C4 levels measured on that occasion were not used to determine if the patient was in relapse or remission.

Definition of Relapse C3 and C4 Values The relapse C3 value was taken as the nadir C3R (C3 measured by radiai immunodiffusion) measured within ±4 weeks of the date of the relapse, as defined previously. The relapse C3N (C3 measured by nephelometry) was taken as the C3N value measured on the same serum specimen as the relapse C3R value. The relapse C4 was taken as the nadir C4R (C4 measUred by radial immunodiffusion) measured within ±4 weeks of the date of the relapse, as defined previously. The relapse C4N (C4 measured by nephelometry) was the value measured on the same serum specimen as the relapse C4R value.

Definition of SLE Remission A remission period was defined as a 4 to 12 month period of no SLE activity, ignoring C3 and C4levels, while the patient was receiving prednisone 20 mg orally every other day or less. In addition, the remission period had to be separated from the nearest relapse by at least 4 months. Thus, a minimum of 12 months of follow-up was required to define a single remission period of 4 months in duration. The rationale of this rigorous definition of remission is that it enhances the likelihood that remission C3 and C4 levels reflect periods of little or no SLE activity. The remission C3N, C3R, C4N, and C4R values were taken as the mean of all of the respective values measured during the remission period. To minimize the disparity among patients in the number of data points averaged for a given period of remission, it was arbitrarily determined that a given remission period could not exceed 12 months. Thus, consecutive 12-month periods of remission were arbitrarily taken as separate remission periods.

Management of SLE Relapse The details of the management protocols have been published elsewhere. 13 In brief, for minor renal relapse the prednisone dose was increased from "maintenance" levels (20 mg every other day) to 30 mg daily. If the prednisone dose at time of relapse was not yet at maintenance level, the dose was increased to the level that was "two steps" higher in the prednisone taper schedule. This usually resulted in a 20 mg/d increase in prednisone dose. During relapse, patients were monitored at weekly or biweekly visits. If the patient's condition stabilized or improved, the prednisone dose was tapered to maintenance levels over the next 10 weeks. If the patient was judged not to have improved, therapy for a major renal relapse was begun. For major renal relapse the prednisone dose was increased to 60 mg daily. If at the end of 2 weeks the patient was judged to have improved and the improvement was sustained, the prednisone dose was tapered over the next 20 weeks to a

680 maintenance dose of 20 mg every other day. If the patient did not improve, the patient was removed from the protocol and treated according to the physician's best judgment. For minor nonrenal relapse the prednisone dose was increased to 30 mg daily. If the patient had not yet reached this point in the prednisone taper schedule, the dosage was increased to that which was two steps higher in the taper schedule. During relapse the patient was then seen every 2 weeks for the next 6 weeks. If improvement was noted after I week, the prednisone was tapered according to the schedule. If no improvement occurred, the therapy was continued for I more week. If the patient improved, the tapering schedule was begun. If there was no improvement and the physician believed that more treatment was warranted, the patient was considered to have a major extrarenal manifestation and was treated according to that protocol. For major nomenal relapse prednisone dose was increased to 60 mg each day. During relapse, the patient was seen weekly. If there was improvement by 3 weeks (I week if CNS involvement was present), the prednisone dose was tapered to maintenance levels over the next 20 weeks. If there was no improvement or worsening, the patient was removed from protocol and treated according to the physician's best judgment.

Frequency of Testing For patients judged to be in remission and on maintenance prednisone therapy (20 mg every other day), C3, C4, and routine blood and urine studies were performed at 2-month intervals. After the onset of relapse, C3, C4, and appropriate routine laboratory studies were obtained at I or 2 week intervals until it was clear that a remission was underway. The frequency of testing was then decreased until the patients were being tested each 2 months.

Assay for C3 and C4 The blood samples for Rand N assay were collected at the same time into separate glass tubes without anticoagulant. For the specimens collected for R assay, the blood was allowed to clot for 60 minutes in a 37°C water bath. Serum was then separated from cells by centrifugation and stored at -70°C until mailed on dry ice to the Core Laboratory where they were stored at -70°C until assayed. For the specimens collected for N assay, the blood was allowed to clot at room temperature for 2 hours. Serum was then separated from cells by centrifugation. The serum was stored at -20°C until assayed. In all instances the storage of these samples before assay did not exceed I week. For N assay, serum C3 and C41evels were quantitated using an automated Beckman Rate Nephelometer (Beckman Instruments, Inc., Brea, CA). This rate nephelometer measures light scatter, resulting from antigen-antibody interactions, during the ascending phase of the precipitation reaction. The instrument is single-point calibrated as per instrument protocol. Each sample run was controlled by the inclusion of C3 and C4 normal control specimens (pooled normal human serum) before and following patient samples. For R assay, commercial plates (Calbiochem, San Diego, CA) were prepared with C3 or C4 antisera. A standard serum solution with known C3 and C4 concentration was used for construction of a standard curve. Patient serum (5 ILL/well) was applied to plates that were then incubated at room temperature for 48 hours

RICKER ET AL to endpoint. After incubation, the diameters of the precipitin rings were measured to the nearest 0.1 mm. These values were then compared with the standard curve determined by the reference sera. All samples were done in duplicate.

Range of Normal for Complement Assays For N assay, 137 healthy second-year medical students (30% female, 70% male) were used as control subjects. The mean value ± 2 standard deviations was taken as the range of normal (C3N, 97 to 155 mg!dL; C4N, II to 44 mg!dL). For R assay, the normal range for C4R was taken as that provided by the manufacturer (20 to 50 mg!dL). However, for C3R the range of normal provided by the manufacturer (55 to 120 mg!dL) was not used because analysis of our data (discussed in the next section) indicated that a lower limit of normal (LLN) of 55 was inappropriately low for our patient population. For the present study a LLN for C3R of 66 mg! dL was used because receiver-operator characteristic (ROC) curve analysis (discussed in the next section) showed this to be the optimum value for diagnostic sensitivity and specificity. Adjustment of C4R LLN was not done because ROC curve analysis (discussed in the next section) showed that there was no LLN ofC4R that would importantly improve its diagnostic sensitivity and specificity.

Data Analysis All mean values are shown ±I standard error of the mean (SEM). Other statistical tests used are discussed in relationship to the data presented. Sensitivity for a given assay was defined as the ratio true positives/(true positives + false negatives). Specificity for a given assay was defined as the ratio true negatives/(true negatives + false positives). In our study, a true-positive result is a complement level that is below the LLN in a patient in SLE relapse. A truenegative result is a complement level within normal limits in a patient who is in SLE remission. A false-positive result is a complement level that is below the LLN in a patient who is in SLE remission. A false-negative result is a complement Table 1. Number of Relapses and Remission Periods for Each Study Patient Relapses Patient

Renal

902 903 904 905 906 907 908 912 914 915 916 917

7 0 5 2 0 2 2 3 0

Totals

Nonrenal

2 0 2 4

Remission Periods

1 2

1 2 3 1 0 0 0

0 4 0 3 0 3 1 0 2 1 0 0

25

16

14

C3 AND SLE ACTIVITY level within normal limits in a patient who is in SLE relapse. Thus, the sensitivity of a given complement assay is the proportion of patients with SLE relapse that have an abnormally low value for that assay. The specificity for a given complement assay is the proportion of patients in SLE remission that have a normal value for that assay. For each C3 and C4 assay an ROC curve was constructed. An ROC curve can be used to determine the diagnostic usefulness of a test and to compare its diagnostic usefulness with that of other tests. I'. IS In ROC curve analysis the "receiver" is an observer or instrument (in the present case, the measured C3 and C4 values), and the "operating characteristics" are the variation in diagnostic sensitivity and specificity for the test at various arbitrarily determined LLN for the test. The ROC curves for the present study were generated as follows: the true-positive rate (sensitivity) for a given complement assay was plotted on the y-axis and the false-positive rate (1 - specificity) was plotted on the x-axis at different "cut points" (arbitrarily chosen LLN for the given complement assay, at which the diagnostic sensitivity and specificity of the test was evaluated). Thus, the whole ROC curve for a given test shows how the true-positive rate varied as a function of the falsepositive rate over a range of arbitrarily chosen LLN that spans all relevant LLN for that test. The ROC curve enables one to choose the optimum LLN for a diagnostic test; that is, the optimum LLN for a test that maximizes the true-positive rate and minimizes the false-positive rate. For an ideal diagnostic test, the optimum LLN is that which achieves 100% true-positive results with 0% false-positive results. Thus, the ROC curve for an ideal diagnostic test would increase vertically sharply at a 0% false-positive rate and then plateau at the 100% true-positive rate level. Conversely, a ROC curve for a useless test would approach a diagonal line. That is, at all LLN for that test the true-positive rate approximates the false positive rate. Such a test would be no better than a "coin toss" in determining whether the patient is normal or abnormal. The diagnostic value of one test compared with another can also be determined by comparing their respective ROC curves. The better test will have a significantly higher true-positive rate at all useful falsepositive rates.

681

that in almost one third of the relapses, the nadir C3 or C4 levels were noted after the diagnosis of SLE relapse was made and treatment was instituted. Thus, it was common for complement levels to decrease in the first few weeks after starting steroid therapy, even though the patients were begun on therapy that ultimately proved to be successful (data not shown). Table 3 shows the mean ± SEM for the serum complement levels measured by N assay and by R assay during remissions and relapses in the SLE patients of this study. As can be noted, during relapse the mean nadir C3N and C3R were significantly less than their respective LLN. By contrast, during relapse mean nadir C4N and C4R values were not below their respective LLN. This is not explained by an inappropriately low LLN for the C4 assay, as discussed below. Table 3 also shows that during SLE remission, the mean values for C3 and C4 were within the normal range for the assay. Table 4 shows, for each of the complement assays, the diagnostic sensitivity (likelihood that the test will be abnormal during SLE relapse) and diagnostic specificity (likelihood that the test will be normal in SLE remission). As can be observed, only the C3 assays are both sensitive and specific. To assess whether the increased diagnostic sensitivity of C3 versus C4 is simply the result of a difference in the lower limits of normal selected for the assays, ROC curves were developed for each assay. These are shown in Figs 1 and 2. As can be noted, the true-positive rate for C3 exceeds that of C4 at all useful false-positive rates for both the N assay and the R assay. To test statistically the dif-

RESULTS

Table I shows the number of relapses and remission periods for each study patient. As can be noted, 14 remission periods occurred in 6 patients and 41 relapses (25 renal and 16 nonrenal) occurred in 11 patients. No distinction was made between renal versus nonrenal relapses because the frequency of hypocomplementemia was not different between renal versus nonrenal relapse. In addition, the ratio ofC4levei to C3level during relapse was the same for renal and nonrenal relapses (data not shown). Table 2 shows when nadir complement levels were first observed in relationship to the relapse. As can be noted, in the majority of SLE relapses the nadir complement level was observed at the time of diagnosis of SLE relapse. Of interest is

Table 2. Relationship Between Onset of SLE Relapse (Designated Week 0, Determined By Criteria That Ignored Serum Complement Levels) and the Week at Which the Nadir Complement Level for That Relapse was First Measured % of Relapses That Show Nadir C3/C4 Values Before, At, or After Onset of the Relapse

C3N C4N C3R C4R

Week -1 to -4

Week 0 (relapse)

Week +1 to +4

0 3 8 8

74 68 58 60

26 29 34 32

NOTE. This analYSis includes only those relapses (N = 38) in which serum complement levels were measured both 1 to 4 weeks before and 1 to 4 weeks after the relapse.

RICKER ET AL

682 Table 3. C3N Versus C4N and C3R Versus C4R During Relapses (N and Remissions (N = 13) in the SLE Patients of This Study

Relapse' Mean ± SEM % of LLNt, mean ± SEM Remission§ Mean ± SEM % of LLNt, mean ± SEM

=

41)

C3N

C4N

C3R

C4R

70 ± 2.6 73 ± 2.7:1:

13 ± 0.9 122 ± 8.6

52 ± 2 .5 79 ± 3.8:1:

20 ± 2.0 103 ± 9.9

70 ± 2 .3 127 ± 4 .1

25± 4.5 127 ± 22.6

116 ± 4.9 119 ± 5.1

18 ± 2 .1 167 ± 20

, Values shown are the mean ± SEM of the nadir value in mg/dL for each relapse. t Mean ± SEM for each nadir value when expressed as a percent of the LLN for the given assay. :I: P < 0.001 compared with the corresponding values for C4N (in the case of C3N) or C4R (in the case of C3R), paired ttest. § Values shown are the mean ± SEM of all values in mg/dL during the remission period.

ferences between the ROC curves ofC3 versus C4 the following analysis was performed: for each complement assay, the number of correct diagnoses (defined as the given complement level was less than the LLN for that assay when the patient had active SLE, or the given complement level was equal to or greater than the LLN of normal for that assay when the patient had inactive SLE) and the number of incorrect diagnoses (the obverse of the definition of correct diagnoses) was determined. The LLN used for C3N (97 mgjdL) and C4N (11 mg,ldL) were those determined from a study of a normal population (see the Methods section). The LLN used for C3R (66 mgjdL) was chosen because as can be noted from Fig 2 this value achieved the highest true-positive rate for false-positive rates less than 20%. The LLN used for C4R (20 mg,ldL) was that provided by the manufacturer. The LLN for Table 4. The Diagnostic Sensitivity and Specificity of C3 and C4 Measured by Both Nephelometry (N) and Radial Immunodiffusion (R)

Number of relapses'
Sensitivity (%) Number of remissions' ~ LLN


C3N

C3R

C4N

C4R

39 2 95t

34 7 83:1:

23 18 56

22 19 54

13 1 93t

10 4 71:1:

7 7 50

7 7 50

, The LLN for the respective assays are as follows: C3N 97 mg/dL, C3R 66 mg/dL, C4N 11 mg/dL, C4R 20 mg/dL. t P < 0.001 comparing total correct to total incorrect diagnoses for the C3N assay versus the C4N assay by x 2 . :I: P < 0.001 comparing total correct to total incorrect diagnoses for the C3R assay versus the C4R assay by x 2 •

C4N and C4R were not adjusted, as was ~one for C3R, because as can be noted from Fig 1 and 2 there is no cutpoint that importantly improves the sensitivity and specificity of the C4 assays. That is, at all cutpoints the true-positive rate and the falsepositive rate are nearly the same. Comparison of the C3 and C4 assays showed that the C3N assay made significantly more correct diagnoses (for the C3N assay 52 correct, 3 incorrect diagnoses; for the C4N assay 30 correct, 25 incorrect diagnoses, P < 0.001, X2 analysis). The same analysis was performed comparing C3R to C4R. That analysis 100

~

+

• •

+ 75

... III III

•

a:

III

> :,:::; ·iii

50

•

0

Q. QJ

:J

t!=

25

0 0

25

50

75

100

False Positive Rate (%)

Fig 1. The ROC curve for C3N (+) and for C4N (.6.). The true-positive rate is shown in relationship to the false-positive rate when C3N was assigned the following LLN (in mg/dL): 51, 56, 61, 66, 71, 76, 81, 86, 91, 92, 93, 94,95,96,97,98,99,1oo,101,106,111,116,121,and 122. The lower limits assigned to C4N (in mg/dL) were 9, 12, 15, 16, 17, 19, 20, 21, 24, 27, 30, 33, 36, and 39. The number of points on the ROC curves are less than the number of LLN tested because points with overlapping results are shown as a single point.

683

C3 AND SLE ACTIVITY

100

~

75

+

Q) +-'

•

III

rr

Q)

-~

•

50

+-'

-iii 0

0.. QJ

::J

~

25

0

~-.---,

0

25

50

75

100

False Positive Rate (%)

Fig 2. The ROC curve for C3R (+) and for C4R (A). The true-positive rate is shown in relationship to the false-positive rate when C3R was assigned the following lower limits of normal (in mg/dL): 32, 35, 38, 41, 44, 47, 50,53,56,59,62,65,66,67,68,69,70,71,74,77,80, 83, 86, 89, 92, and 95. The lower limits assigned to C4R (in mg/dL) were 3, 6, 9, 12, 15, 18, 19, 20, 21, 22, 23, 24, 27, 30, 33, 36, 39, 42, 45, 48, 51, 54, 57, and 60. The number of points on the ROC curves are less than the number of LLN tested because points with overlapping results are shown as a single point.

showed that the C3R assay made significantly more correct diagnoses than the C4R assay (for C3R 47 correct and 8 incorrect diagnoses; for C4R 29 correct and 26 incorrect diagnoses, P < 0.00 1, X2 analysis). Thus, we found that by either N assay or R assay C3 was a more accurate assessment of SLE activity than was C4. To assess whether the increased diagnostic sensitivity ofC3 versus C4 was the result oflarger relative changes in C3 versus C4 as SLE activity changes, the following analysis was performed. The regression of C4N on C3N was calculated using all values on each patient (mean 60 ± 7 values per patient). In all but two patients the regression of C4 on C3 was well described by a straight-line relationship. The correlation coefficients of C4 on C3 for the individual patients, ranked in ascending order, are 0.33, 0.36, 0.67, 0.69,0.73,0.78,0.78,0.84,0.88,0.90, 0.91 , 0.92 (mean 0.73 ± 0.06). For each patient, the slope of the regression (LlC4/ LlC3) was determined and divided by the ratio mean C4/mean C3 for that patient. Thus, the ratio (LlC4/mean C4)/(LlC3/ mean C3) was determined. This ratio is the ratio of the fractional change in C4 to the fractional change in C3 for that patient as SLE activity

changes. The individual ratios were 0.37, 0.46, 0.73, 0.73, 0.76, 0.87,0.91,1.0, 1.0, 1.0, 1.2, 1.3. The mean ratio was 0.87 ± 0.08, a value not significantly different from 1.00. Thus, on average, the fractional change in C4 is not different from that of C3 as SLE activity changes. Analysis of the slopes of the regression of C4 on C3 also permitted an assessment of the net molar consumption (moles consumed - moles produced) of C4 compared with that of C3 as SLE activity changes. That is, the molecular weight of C4 (approximately 200 kd) is comparable with that of C3 (approximately 190 kd).2 Thus, the slope of the regression of C4 on C3 (LlC4/ LlC3), expressed in mg/dL for each component, is an approximation of the ratio of the net molar consumption of C4 versus C3 as SLE activity changes. The slopes of the individual patients, ranked in ascending order, are 0.042, 0.045, 0.101 , 0.118, 0.136, 0.168, 0.169, 0.182, 0.217, 0.275, 0.282, and 0.284 (mean 0.17 ± 0.02). Thus, on average, the net molar consumption of C3 is approximately 6 times greater than that of C4. DISCUSSION

The present study was undertaken to compare the diagnostic sensitivity and specificity ofC3 versus C4 with regard to identifying patients with active versus inactive SLE. Two different, standard methods of C3 and C4 measurement (N assay and R assay) were used. The study was performed in 12 SLE patients with severe glomerulonephritis (GN) on initial presentation, and who were subsequently followed through multiple relapses and remissions. The relapses and remissions were identified using criteria that did not include interpretation of the serum C3 and C4 levels. The present study demonstrates that, irrespective of the assay used, serum C3 levels are more sensitive and specific indicators of SLE activity than are serum C4 levels. That is, during SLE relapse, C3 levels were abnormally low in 83% (R assay) to 95% (N assay) of instances, but C4 levels (either method) were abnormally low in only approximately 55% of instances of active SLE. During SLE remission, as defined in the present study, C3levels were normal in 71 % (R assay) to 93% (N assay) of instances, but C4 levels (either method) were normal in only 50% of instances of SLE remission. Thus, measurement ofC3levels, but not C4levels,

684

reliably separated active from inactive SLE particularly when C3 was measured by N assay. To assess whether the relative lack of diagnostic sensitivity and specificity of C4 levels compared with C3 levels could be corrected by choosing an optimum LLN for C4, ROC curves were constructed. This analysis showed that the low diagnostic sensitivity and specificity of C4, relative to C3, could not be explained by failure to use an optimum LLN for C4. That is, the ROC curves demonstrated a higher true-positive rate (sensitivity) for C3 compared with C4 at all relevant false-positive rates (1 - specificity). The greater diagnostic sensitivity and specificity of C3 versus C4 also could not be attributed to greater precision in the measurement of C3 versus C4. Indeed, in almost all patients C3 levels were highly correlated with C4 levels. Thus, changes in C3 levels accurately predict changes in C4 levels. We also found that the ratio of (Ll C4/C4)/(Ll C3/C3), as SLE activity changed, was not significantly different from 1.00. Thus, reduced diagnostic sensitivity ofC4 versus C3 also cannot be explained by relatively smaller changes in C4 compared with C3 as SLE activity changed. Our data indicate that the reduced diagnostic sensitivity ofC4 compared with C3 is in part related to the broader range of normal of C4, relative to that of C3. For example, for the 14 centers involved in the LNCSG the mean of the individual ratios of the upper limit of normal to the LLN for the C3 and C4 assays was 2.0 ± 0.1 for C3 versus 3.1 ± 0.2 for C4 (P < 0.001, paired t-test). The relatively broader range of normal for C4 appears to be the result of the high prevalence (up to 40%) of one or two C4 null genes (genes that do not produce C4) in the normal population. 16,17 Another factor that reduces the diagnostic sensitivity of C4 levels in SLE is that SLE patients have an increased (up to 80%) prevalence of one or two C4 null genes. 16 Thus, SLE patients with C4 null genes will tend to have low C4 levels even when there is little or no SLE activity. By contrast, SLE patients with a full set ofC4 genes can maintain C4levels in the normal range, even though their SLE is active, causing a substantial reduction in C4 from baseline levels. These properties of C4 in the SLE population also reduces the diagnostic sensitivity and specificity of the C4 assay compared with the C3 assay. The disagreement between the present study and the previous studies that suggested that the C4 level is more likely to be abnormal during SLE activity

RICKER ET AL

than the C3 level may be explained by the facts that the previous reports studied patients who had high rates (50% to 71 %) of abnormally low C4levels even when in remission, 8, to the reports did not provide data to support their contention that C4 was more useful diagnostically than C3,1,11 or their conclusions were based on a single C4 value in a single patient. 18 Thus, although it is possible that in an individual SLE patient C4 levels might be diagnostically more useful than C3 levels, there is no convincing evidence for that in our patient population or in the current literature. Although the present study demonstrates that measurement of C4 is not useful for monitoring activity of SLE patients, the present study does not establish the precise role for measuremeD;t of C3 levels in the monitoring of SLE patients. That would require a much larger study and would need to include SLE patients without renal involvement because of the evidence that hypocomplementemia is less likely to occur in nonrenal relapses. 7,tO,19 We did not find that hypocomplementemia is less common in nonrenal relapse. However, all of our patients had as a presenting SLE manifestation severe renal involvement. It is possible that such patients are more likely to develop hypocomplementemia with SLE relapse, renal or nonrenal, than are SLE patients who have never had renal involvement. Perhaps SLE patients who have never had renal involvement form a lesser amount of immune complex or form immune complexes that are less pathogenic because they do not activate complement as well. The present study also demonstrates that during active SLE the net molar consumption of C3 is approximately 6 times greater than that ofC4. This net molar consumption of C3 relative to that of C4 approximates that which can be predicted from the in vitro studies showing that only about 1 in 400 activated C4 molecules is incorporated into the C3 convertase C4b2b but that one of these convertases can consume about 2,000 C3 molecules. 2 Our analysis also assumes that if changes in C3 and C4 production occur, the changes are proportional to their respective baseline production rate. In summary, if complement levels are used to monitor SLE activity in patients with a history of renal manifestations, it is sufficient to measure only C3 levels. This should simplify and reduce the cost of the serologic assessment ofSLE activity.

685

C3 AND SLE ACTIVITY

APPENDIX The Lupus Nephritis Collaborative Study Group: Edmund J. Lewis, MD, Principal Investigator, B. S. Kasinath, MD, Coprincipal Investigator, Anil K. Bidani, MD, Director, Central Immunology Laboratory, Rush-Presbyterian-St Luke's Medical Center, Chicago. Biostatistics Coordinating Center: John Lachin, ScD, Shu-ping Lan, MA, MPH, Patricia Cleary, MS, George Washington University, Bethesda, MD. Pathology Reading Committee: Melvin M. Schwartz, MD, Chairman and Director of the Central Pathology Laboratory, RushPresbyterian-St Luke's Medical Center, Chicago; Jay Bernstein, MD, William Beaumont Hospital, Royal Oak, MI; Gary S. Hill, MD, Francis Scott Key Medical Center, Baltimore; Keith Holley, MD, Mayo Clinic, Rochester, MN. Clinics: Marc A. Pohl, MD, John Clough, MD, Gordon Gephardt, MD, Cleveland Clinic, Cleveland; Tomas Berl, MD, University of Colorado, Denver; Nathan Levin, MD, Henry Ford Hospital, Detroit; Lawrence G . Hunsicker, Stephen Bonsib, MD, University of Iowa, Iowa City; Norman Simon, MD, Hartmann F:riederici, MD, Evanston Hospital, Evanston, IL; Francesco del Greco, MD, Frank A. Carone, MD, Northwestern Uni-

versity, Chicago; Lee Hebert, MD, Hari M. Sharma, MD, Ohio State University, Columbus; Eric Neilson, MD, John Tomazewski, MD, University of Pennsylvania, Philadelphia; Howard L. Corwin, MD, Melvin M. Schwartz, MD, RushPresbyterian-St Luke's Medical Center, Chicago; Andrew Levey, MD, Angelo Ucci, MD, Tufts-New England Medical Center, Boston; Howard Shapiro, MD, Barbara F. Rosenberg, MD, William Beaumont Hospital, Royal Oak, MI; Jacob Lemann, MD, John Garancis, MD, Medical College ofWisconsin, Milwaukee; Kenneth Shapiro, MD, Praveen Chander, MD, New York Medical College, Valhalla, NY; Fred Whittier, MD. John W. Graves, MD, Joan Bathon, MD, Roger Riley, MD, West Virginia University, Morgantown.

ACKNOWLEDGMENTS The authors gratefully acknowledge the generous gift of Mr William Miller that helped support this work. The authors also thank Ms Carmela Price for her secretarial assistance, Mrs Maddie Hebert for her assistance in data cGllection, and Dr Melvin Moeschberger of the Biostatistics Laboratory for his help in statistical analysis of the data.

REFERENCES I. West CD: The complement profile in clinical medicine. Complement Inflamm 6:49-64, 1989 2. Hughes-Jones NC: The classical pathway, in Ross GD (ed): Immunobiology of the Complement System. Philadelphia, PA, Saunders, 1986, pp 21-44 3. Carpenter CB, Ruddy S, Shehadeh IH, et al: Complement metabolism in man: Hypercatabolism of the fourth (C4) and third (C3) components in patients with renal allograft rejection and hereditary angioedema (HAE). J Clin Invest 48: 1495-1505, 1969 4. Ruddy S, Carpenter CB, Chin KW, et al: Human complement metabolism: An analysis of 144 studies. Medicine 54: 165-178, 1975 5. Waldo FB, Forristal J, Beischel L, et a1: A circulating inhibitor of fluid-phase amplification. C3 convertase formation in systemic lupus erythematosus. J C1in Invest 75: 1786-1795, 1985 6. Morrow WJW, Isenberg DA, Todd-Pokropek A, et al: Useful laboratory measurements in the management of systemic lupus erythematosus. Q J Med LI: 125-138, 1982 7. Valentijn RM, van Overhagen H, Hazevoet HM, et al: The value of complement and immune complex determinations in monitoring disease activity in patients with systemic lupus erythematosus. Arthritis Rheum 28:904-913, 1985 8. Swaak AJG, Groenwold J, Bronsveld W: Predictive value of complement profiles and anti-dsDNA in systemic lupus erythematosus. Ann Rheum Dis 45:359-366, 1986 9. Esdaile JM, Levinton C, Federgreen W, et a1: The clinical and renal biopsy predictors of long-term outcome in lupus nephritis: A study of 87 patients and review of the literature. Q 1 Med 72:779-833, 1989

10. Lloyd W, Schur PH: Immune complexes, complement, and anti-DNA in exacerbations of systemic lupus erythematosus (SLE). Medicine 60:208-217, 1981 II. Frank MM: Complement in the pathophysiology of human disease. N Engl 1 Med 316:1525-1530, 1987 12. Hebert L, Nielsen E, PohI M, et a1: Clinical course of severe lupus nephritis during the controlled clinical trial of plasmapheresis therapy (PPT). American Society of Nephrology, Annual Meeting, Washington, DC, December 7-10,1986 (abstr) 13. Clough 1D, Lewis EJ, Lachin 1M, et al: Treatment protocols of the lupus nephritis collaborative study of plasmapheresis in severe lupus nephritis. Apheresis 301-307, 1990 14. Robertson EA, Zweig MH: Use of receiver operating characteristic curves to evaluate the clinical performance of analytical systems. Clin Chern 27:1969-1574, 1981 15. Beck JR, Shultz EK: The use of relative operating characteristic (ROC) curves in test performance evaluation. Arch Pathol Lab Med 110:13-20, 1986 16. Atkinson JP: Complement deficiency. Predisposing factor to autoimmune syndromes. Am J Med 85:45-47, 1988 17. Braun L, Schneider PM, Giles CM, et al: Null alleles of human complement C4. Evidence for pseudogenes at the C4A locus and for gene conversion at the C4B locus. J Exp Med 171 :129-140, 1990 18. Lewis EJ, Carpenter CB, Schur PH: Serum complement component levels in human glomerulonephritis. Ann Intern Med 75:555-560, 1971 19. Schur PH, Sandson J: Immunologic factors and clinical activity in systemic lupus erythematosus. N Engl J Med 278: 533-538, 1968