ARTHRITIS & RHEUMATISM Volume 37 Number 4, April 1994, pp 559-567 0 1994, American College of Rheumatology 559 KIDNEY BIOPSY IN SYSTEMIC LUPUS ERYTHEMATOSUS 111. Survival Analysis Controlling for Clinical and Laboratory Variables JOHN R. McLAUGHLIN, CLAIRE BOMBARDIER, VERNON T. FAREWELL, DAFNA D. GLADMAN, and MURRAY B. UROWITZ Objective. To examine the importance of renal biopsy as a predictor of death due to any cause in patients with systemic lupus erythematosus (SLE). Methods. The study included 123 SLE patients who had a renal biopsy between 1970 and 1984 and were followed up as part of a prospective study. Data were initially analyzed to identify clinical and laboratory features that were significantly associated with the risk of dying. Renal biopsy variables were then examined to determine whether they contributed additional information about prognosis. Results. The clinical and laboratory factors most closely associated with the risk of dying in multivariate analyses were the serum creatinine level and the SLE Disease Activity Index score. The presence of chronic renal lesions on biopsy contributed significantly to the prognostic information offered by clinical and laboratory factors in the subset of patients who had normal serum creatinine levels-the majority (85%) of patients in this study. Conclusion. These results indicate that renal biopsy serves an important role in the assessment of Supported by grants from the Arthritis Society and the Ontario Ministry of Health. Dr. McLaughlin's work was supported by a PhD Fellowship from the National Health Research and Development Program of Health and Welfare, Canada. Dr. Bombardier's work was supported by a National Health and Welfare Scholarship. John R. McLaughlin, PhD: University of Toronto, Toronto, Ontario, Canada; Claire Bombardier, MD, FRCP(C): Wellesley Hospital, Toronto, and the University of Toronto; Vernon T. Farewell, PhD: University of Waterloo, Waterloo, Ontario, Canada; Dafna D. Gladman, MD, FRCP(C): Wellesley Hospital; Murray B. Urowitz, MD, FRCP(C): Wellesley Hospital. Address reprint requests to Murray B. Urowitz, MD, FRCP(C), Rheumatic Disease Unit, The Wellesley Hospital (TW 651A), 160 Wellesley Street East, Toronto, Ontario, M4Y 153, Canada. Submitted for publication October 8, 1992; accepted in revised form August 24, 1993. prognosis in patients who do not have advanced renal disease. There is an ongoing debate over whether renal biopsy results are useful indicators of prognosis, in terms of both renal outcome and overall mortality, in systemic lupus erythematosus (SLE). In particular, there has been controversy over whether renal biopsy features add prognostic information over and above that of routine clinical and laboratory data obtained at the time of biopsy. Several previous studies examined the prognostic value of renal biopsy observations while controlling for the effects of clinical and laboratory factors by way of multivariate statistical analyses. Chronic renal lesions were associated with a poor prognosis in most (1-5) but not all (6) of these studies, which assessed prognosis in terms of renal function (2-4) or both renal function and patient survival (1,5 A). As part of an ongoing prospective study of the determinants of SLE progression, we previously examined the relationship between renal biopsy findings and clinical manifestations of SLE, showing that there was no uniform correlation between the two (7). The relationship between renal biopsy results and survival was further examined by univariate analyses, which demonstrated that the presence of chronic or proliferative lesions was associated with a higher risk of dying (8). Renal outcomes were not employed in this analysis because they occurred rarely in this series: only 6 patients developed end-stage renal disease, and of these, 5 died (8). This study examines the relative importance of renal biopsy compared with clinical and laboratory features at the time of biopsy as indicators of progno- McLAUGHLIN ET AL 560 sis in terms of mortality, among the SLE patients who underwent biopsy. This aim was met by assessing the role of clinical and laboratory measures, by examining whether there was a further contribution by renal biopsy observations, and finally, by assessing whether there was confounding by differences in treatment. PATIENTS AND METHODS Sample selection. Study subjects were selected from the University of Toronto Lupus Database. This database contains observations made prospectively during the clinical examinations of SLE patients seen at the SLE clinic of a teaching hospital since 1970. Data collection has been standardized by the use of a data collection form, which guides clinicians through the observation and definition of 207 demographic, historical, physical, laboratory, and therapeutic variables. To be included in the database, each patient must be diagnosed as having SLE, either by fulfilling 4 or more of the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) criteria for the classification of SLE (9), or by having other manifestations of SLE in addition to several of the ACR criteria. This study included the members of the SLE database who had a renal biopsy at the Wellesley Hospital from 1970 until 1984. There were 475 patients registered in the database at that time. During this period, patients were asked to have biopsies as part of the routine assessment of their SLE. This period of inclusion was chosen because the classification of renal biopsies obtained during this time was standardized by means of an expert review committee, and because it allowed a minimum of 2 years of followup for each subject (7,8). These cases were followed from the time of the patient’s first renal biopsy until January 1, 1987. Only those whose renal tissue was first examined by biopsy prior to their death were included. Thus, 123 patients were included in the present investigation. The outcome was patient death due to any cause. End-stage renal disease occurred in only 6 of the 123 patients, 5 of whom died; therefore, end-stage renal disease was not analyzed because it was too rare to be informative. Patients who were not seen at the SLE clinic during the year prior to the end-point date (i.e., 1986) were traced to verify their final outcome status. Three of these patients could not be located and were classified as lost to followup; their end-point date was set as the date of their last examination at the clinic. This report focuses on the 3 methods of classifying renal lesions that were shown in univariate analyses to be indicators of prognosis (8). These methods include classification by the World Health Organization (WHO) criteria for lupus nephritis (10,ll) and a regrouping of the WHO classes that reflects the presence of proliferative lesions (WHO class IV and subsets of classes I11 and V). Biopsies were also classified according to the presence of chronic renal lesions (glomerular sclerosis, interstitial fibrosis), as defined by Austin et a1 (3). Further details about the sample selection, data collection, and procedures employed in the review of renal biopsy samples have been previously described (7,8). In addition to the biopsy observations, prognostic importance was examined for 51 variables previously identified as possible prognostic factors. These variables were diagnostic criteria, demographic descriptors and indicators of comorbidity, therapy, and SLE status. The indicators of SLE status were based on both clinical and laboratory measures and were defined according to the criteria established by the ACR (12,13). The SLE Disease Activity Index (SLEDAI), a previously validated scale composed of 24 clinical variables (14-16), was also included in the analysis of prognosis. The question of whether treatment is associated with mortality cannot be validly assessed in an observational study such as this; however, it is still important to examine whether treatment is a possible confounder of the relationship between prognostic factors and mortality. Particular emphasis was placed on corticosteroid and immunosuppressive medications being taken at the time of biopsy, although nonsteroidal antiinflammatory drugs (NSAIDs) were also examined. Prednisone therapy at the time of biopsy was classified as none, low dose (1-39 mg/day), and high dose (240 mglday). The use of immunosuppressive medications (e.g., azathioprine, cyclophosphamide) at the time of biopsy was classified dichotomously (used or not used). Survival analysis was used to study the relationship between the possible prognostic factors and mortality. Survival time was defined as the interval from the time of biopsy until death or until January 1, 1987, at which time surviving patients were censored. Nonparametric life-table methods were used to calculate Kaplan-Meier estimates of survival probabilities (17). Proportional hazards regression models (chapter 4, ref. 18) were used to estimate relative risk (RR) and to examine the simultaneous effects of multiple prognostic factors. Relative risk was estimated as the antilogarithm of the regression coefficient for each covariate. For individual parameters in the regression models, significance tests and 95% confidence intervals (95% CI) were obtained based on the asymptotic normality of the parameter estimates. Where appropriate, the test of significance for the addition of variables to a multivariate regression model employed the likelihood ratio test, which is chi-square distributed. The best-fit multivariate model of clinical and laboratory factors was obtained in 2 stages. First, multivariate analyses were performed to identify the strongest predictors of survival among all of the manifestations that related to a single organ system (i.e., to reduce the number of highly correlated variables). Second, a stepwise multivariate procedure was employed that began with the most highly predictive variables from all of the organ systems and then selected variables that were significant at the standard probability level (a = 0.05). To determine whether renal biopsy observations provided important prognostic information beyond the clinical and laboratory variables, the biopsy variables were added to a multivariate model containing the most highly predictive clinical and laboratory variables. Treatment variables were then added to these multivariate models in order to determine whether there was confounding, as would be indicated by shifts in the regression coefficients for the main prognostic factors. 561 KIDNEY BIOPSY IN SLE Table 1. Features of the 123 SLE patients at the time of biopsy* 28 (23) 4.7 (0, 15.9) Total deaths, no. (%) Duration of followup, median years (minimum, maximum) SLE duration before biopsy, median years (minimum, maximum) Age, mean (SD) Sex ratio, fema1e:male WHO renal biopsy class, no. (%) I, normal 11, mesangial glomerulonephritis 111, segmental glomerulonephritis IV, diffuse glomerulonephritis V, membranous glomerulonephritis VI, sclerosing glomerulonephritis Serum creatinine, mean pmoleshiter (SD) Treatment, no. (%) Acetylsalicylic acid Nonsteroidal antiinflammatory agents Corticosteroids High-dose corticosteroids (240 mg/day) Immunosuppressive agents * SLE = systemic lupus erythematosus; WHO Organization. 0.8 (0, 26) 33.9 (12.6) 9: 1 8 (6.5) 52 (42.3) 20 (16.3) 32 (26.0) 7 (5.7) 4 (3.3) 109 (69) 31 (25) 10 (8) 96 (78) 28 (23) 13 (11) = World Health The appropriateness of the proportional hazards model was evaluated by testing for violations in the proportional hazards assumption using time-dependent covariates and diagnostic plots (18). There was no evidence that the proportional hazards assumption was violated, which indicated that models containing simple fixed-time covariates provided a valid depiction of the impact of the main prognostic factors. Accordingly, the results that follow were obtained from simple proportional hazards models. The SAS program (19) was used for descriptive analyses and to fit simple proportional hazards models. Regression analyses for models including time-dependent variables were performed using the Fortran program for Cox regression analysis described by Kalbfleisch and Prentice (18). RESULTS The features of the 123 study subjects are summarized in Table 1, The duration of disease before biopsy was quite short, with a median of 0.8 years, whereas the length of followup after biopsy was relatively long, with a median of 4.7 years. The age and sex distributions are similar to those usually reported for SLE patients seen in rheumatology clinics: predominantly women in their child-bearing years. Twentyeight patients (23%) died during the study period. The certified cause of death was cardiopulmonary disease in 10 patients, septicemia in 8, central nervous system disease in 4, pneumonia in 3, and unknown in 3 patients. Table 1 also shows that 49% of the patients had mild lesions on renal biopsy (WHO classes I and 11), whereas 42% had proliferative lesions (classes I11 and IV). At the time of biopsy, the number of patients receiving specific treatments was 31 (25%) taking acetylsalicylic acid, 10 (8%) taking NSAIDs, 96 (78%) taking corticosteroids, 28 (23%) taking high-dose corticosteroids (240 mg/day), and 13 (11%) taking immunosuppressive medications. During the full period of followup after biopsy, an additional 25 patients received high-dose corticosteroids and an additional 38 received immunosuppressive agents. For all patients combined, the survival rates at 5 and 10 years were 82% and 74%, respectively. In univariate analyses, whereby relative risks and significance levels were estimated using proportional hazards models, several of the 51 clinical and laboratory variables were significant prognostic factors (Table 2). Several other variables were analyzed but not listed in Table 2 because they were not associated with mortality and they occurred rarely in this sample. The variables in Table 2 with the highest relative risks were indicators of renal disease (e.g., low creatinine clearance, elevated serum creatinine, presence of urinary casts). Among the variables with relative risks that were significantly different from 1.O was a SLEDAI score 220; the cut-off value of 20 was chosen because this identified subjects in the upper quartile of the range of SLEDAI values. Variables indicating the presence of cushingoid features, infections, and muscle weakness also had relative risks that were significantly different from 1 .O. Features that achieved borderline significance (0.05 < P < 0.1) included the presence of vasculitic lesions, fever, myositis, and alopecia. For the following analyses, serum creatinine was selected as the renal manifestation of greatest prognostic importance. This was done even though creatinine clearance had a larger relative risk estimate, because creatinine clearance data were missing for a large proportion of cases (30%). Table 3 shows that 2 variables were contained in the best-fit multivariate model that was based on all of the clinical and laboratory factors: indicators of elevated SLEDAI and serum creatinine. The relative risk estimates demonstrate that with an increasing SLEDAI score, there was an increasing RR, and that there were large and statistically significant RRs among patients with SLEDAI scores >I9 (RR = 4.7) and elevated serum creatinine levels (RR = 5.9). Survival curves provide an alternative depic- 562 McLAUGHLIN ET AL Table 2. Univariate relative risk (RR) estimates and level of significance for the relationship between mortality and clinical and laboratory variables at the time of biopsy in 123 patients with systemic lupus erythematosus (SLE)* Variablet Total no. No. of deaths RR P 95% CI 14 9 8.5 <0.001 3.3,21 18 12 9 6 4.7 4.0 <0.001 0.003 2.1,11 1.6,25 48 46 29 50 5 22 25 27 5 40 34 17 9 15 4 6 43 8 8 12 18 3 10 12 9 3 13 10 7 4 6 2 2 13 15 16 1.o 1.4 3.9 2.7 4.4 2.3 2 .o 2.1 3.0 1.9 1.9 2.0 2.3 2.0 2.8 2.5 1.6 1.5 - 1 0.4 1.6 1.4 1.6 0.7 1.8 0.5,3.7 1.6,7.4 1.3J.9 1.3J3.1 1.1,5.1 0.9,4.4 0.9,4.6 0.9,lO 0.9,4.0 0.9,4.2 0.8,4.6 0.7,6.9 0.8,4.8 0.7,12 0.6,ll 0.8,3.3 0.7,3.2 0.6,3.7 0.1,2.6 0.6,4.2 0.7,3.0 0.6,4.7 0.3,1.7 0.4,7.5 0.5,4.1 0.6,3.4 0.4,6.3 0.5,2.6 0.4,3.9 Creatinine clearance <SO% of normalt Serum creatinine > 120 pmoles/liter Urinary cellular casts SLEDAI score 0-9 10-19 220 Cushingoid features Infection Muscle weakness Vasculitic lesions Fever Myositis Alopecia Hematuria (>5 RBC/hpD Age 250 Proteinuria (>3.0 gm/day) Pulmonary embolism Cardiovascular disease Thrombocytopenia (<100 x 10’iliter) Arthritis Recent period of biopsy (1978-1984) SLE duration 2 I year Lupus headache Serositis Low complement Leukopenia (<3.5 x 10’iliter) DNA binding Avascular necrosis Sex (males) Mucous membrane lesions Visual disturbance Pyuria Pleurisy 64 43 13 17 58 14 41 6 12 22 6 39 12 5 14 4 7 2 4 6 2 9 3 1.5 1.5 1.4 1.5 1.2 1.2 0.51 0.004 0.01 0.02 0.03 0.06 0.07 0.07 0.09 0.11 0.13 0.13 0.15 0.17 0.22 0.22 0.28 0.29 0.30 0.36 0.37 0.37 0.42 0.43 0.43 0.48 0.59 0.71 0.78 * Variables were excluded if they occurred in fewer than 10 patients and had P values greater than 0.80. These were: the presence of psychosis, seizures, peripheral neuropathy, pericarditis, organic brain syndrome, cerebrovascular accident, infertility, cranial neuropathy, and diabetes after steroid therapy. 95% CI = 95% confidence interval; SLEDAI = SLE Disease Activity Index; RBC/hpf = red blood cells/high power field. t Creatinine clearance data were available for 86 of the 123 patients. Cardiovascular disease was defined as a history of myocardial infarction, angina, or congestive heart failure; serositis as the presence of pleuritis or pericarditis; low complement as a CH5O value <120 hemolytic units or BlC <0.6 gdliter or C4 <0.16 gdliter. Table 3. Relative risk estimates and level of significance from the best-fit multivariate proportional hazards model containing clinical and laboratory predictors of mortality among all patients (n = 123)* 95% Variable SLEDAI score 0-9 10-19 220 Serum creatinine (>120 pmolesfliter) P RR 1.O 1.5 4.7 0.50 0.002 5.9 <0.001 * See Table 2 for definitions of abbreviations. CI 0.6,4.0 1.9,i1.9 2S713.8 tion of the effect of these prognostic factors. Figure 1 shows that the survival fraction of patients with elevated serum creatinine dropped sharply during the first year of followup relative to patients with normal creatinine levels. Figure 2 indicates that, in general, a higher SLEDAI score was associated with poorer prognosis (lower survival), and that after about 5 years, the survival fractions for patients with low (score of 0-9) and intermediate (score of 10-19) SLEDAI scores were very similar. The statistics shown in Table 4 refer to 3 563 KIDNEY BIOPSY IN SLE P P R 0 R 0 B A 0.8 B I I L 0.7. I T 0.6. Y B A B I L 0.7. I T 0.6. Y 0 0.5 Serum Creatinine 5 120 SLEDAI > 19 F 0.4. 0.4 t 1I S U S U Serum Creatinine > 120 R 0.3. R 0.3 v I 0.2 A L I 0 0.5. 3 F v I v I1 I 0.2. A L 0.1 * i 0.01 0 v - NO cases observed beyond 8.2 years l 1 2 3 4 5 8 7 .1 8 9 10 f 0.0 0 1 2 3 4 5 7 6 8 9 10 TIME AFTER BIOPSY (years) TIME AFTER BIOPSY (years) Figure 1. Survival curves for 123 patients with systemic lupus erythernatosus, according to the presence of elevated serum creatinine at the time of renal biopsy (5120 pmolesfliter in 105 patients; >I20 pmolesfliter in 18 patients). Figure 2. Survival curves for 123 patients with systemic lupus distinct models that arose by adding each of the 3 biopsy variables to the baseline model that contained only clinical and laboratory variables (as shown in Table 3). The biopsy variables did not significantly improve the predictive ability of the models, as shown by significance levels of 0.32, 0.10, and 0.18 for the erythematosus, according to the Systemic Lupus Erythernatosus Disease Activity Index (SLEDAI) class at the time of renal biopsy (score <10 for 48 patients, 10-19 for 46, and >19 for 29 patients). Table 4. Contribution of the renal biopsy variables to the best-fit multivariate model containing clinical and laboratory variables, among all patients (n = 123)* Model no. Variable added to model ~ 1. ~ 3. RR 95% CI 2 df P 1.O 0.6 0.7 1.2 1.9 4.5 0.1J.2 0.1,7.3 0.1,ll.Z 0.1,34.6 0.4,54.3 5.88 5 0.32 1.O 2.1 - 2.72 1 0.10 0.9,4.9 2.32 1 0.18 ~~ WHO class I I1 111 2. Effect of adding biopsy variable IV V VI Proliferative lesions Absent Present Chronic lesions Absent Present 1.O 2.3 0.7,7.8 * Best-fit clinical-laboratory model for all patients, as defined in Table 3. See Table 1 for WHO class definitions; df = degrees of freedom (see Table 2 for definitions of other abbreviations). McLAUGHLIN ET AL 564 Table 5. Relative risk estimates and level of significance from the best-fit multivariate proportional hazards model containing clinical and laboratory predictors of mortality in the subgroup of patients with normal serum creatinine (n = 105)* Variable SLEDAI score 0-9 10-19 220 RR P 95% CI I .o 0.76 0.02 0.4,4.0 1.2,10.4 1.2 3.5 * See Table 2 for definitions of abbreviations. WHO class, proliferative lesion, and chronic lesion classifications, respectively. Further analyses were performed on the subgroup of patients who had normal serum creatinine at the time of biopsy, because a statistical interaction was detected between the chronic lesion and the serum creatinine variables. Furthermore, it was clinically relevant to consider whether a renal biopsy contributed important prognostic information in patients with normal renal function. In this study, only 18 patients (15%) had an elevated serum creatinine value at the time of biopsy (Table 2). Among the remaining 105 patients (85%) with normal levels of serum creatinine, 19 deaths occurred. Of these patients, 74 (70%) had no indication of clinical renal disease: proteinuria, hematuria, and urinary casts were absent. The best-fit model based on all of the clinical and laboratory variables (including indicators of renal disease other than serum creatinine) in this subgroup contained only the SLEDAI variable (Table 5), for which the relative risk estimates were slightly reduced from the level for all patients shown in Table 3. Biopsy variables were then added to the baseline model for the subgroup (as in Table 5). Table 6 shows that statistical significance was achieved by both the proliferative (P = 0.03) and chronic (P = 0.004) lesion variables, but not for the WHO class variable (P = 0.12). The presence of proliferative and chronic lesions in this subset of patients was also associated with large relative risks (RR = 3.0 and 7.9, respectively). The presence of proliferative lesions did not contribute additional prognostic information to that obtained by the model containing the indicators for SLEDAI and chronic lesions (2 = 1.37, 1 degree of freedom, P = 0.24), whereas the chronic lesion variable was significant when added to the model containing the SLEDAI and proliferative lesion variables (2 = 5.08, 1 degree of freedom, P = 0.02). The role of treatment as a potential confounder of the relationship between the major predictors and mortality was then assessed for the patient subgroup by adding variables that represented treatment at the time of biopsy to the best-fit multivariate model that was identified in Table 6, namely, that which contained the SLEDAI and chronic lesion variables. Table 7 shows that there were only slight shifts in the relative risk estimates (<lo%) for the SLEDAI and chronic lesion variables, which indicated that these estimates were not confounded by the treatment variables considered. Table 6. Contribution of the renal biopsy variables to the best-fit model containing clinical and laboratory variables, among patients with normal serum creatinine (n = 105)* Model no. 1. Variable added to model WHO class I I1 2. 3. 111 IV V VI Proliferative lesions Absent Present Chronic lesions Absent Present Effect of adding biopsy variable RR 95% CI 1.o 0.7 0.6 1.8 2.1 10.6 0.1,6.0 0.1,7.9 0.3,38.8 0.1,40.8 0.9,132 1.o 3.0 2.6,3.5 1.o 7.9 2.5,25.5 2 df P 8.84 5 0.12 4.72 1 0.03 8.43 1 0.004 - - * Best-fit clinical-laboratory model for the subgroup, as defined in Table 5 ; WHO = World Health Organization (see Table 4 for other definitions). 565 KIDNEY BIOPSY IN SLE Table 7. Relative risk estimates for the SLEDAI and chronic lesion variables, among patients with normal serum creatinine (n = 105) before (baseline model) and after adding indicators of treatment at time of biopsy* Treatment variable added to baseline model ~~~~ SLEDAI score 0-9 10-19 220 Chronic lesions Absent Present Baseline model Steroids High-dose steroids Immunosuppressives 1 .o 1.4 3.6 1 .o 1.3 3.2 I .o 1.4 3.9 1.4 3.6 1.O 1.o 1.9 1.6 I .o 8.6 8.3 * Baseline model for patient subgroup, as developed in Table 6. SLEDAI Erythematosus Disease Activity Index. DISCUSSION The major findings in this study were as follows: I) information obtained by clinical and laboratory investigations, especially serum creatinine and the SLEDAI, had prognostic significance in SLE patients who had undergone renal biopsy; 2) findings on renal biopsies, especially the presence of chronic lesions, contributed additional information about prognosis in patients who had normal serum creatinine levels at the time of biopsy; and 3) the importance of the main prognostic factors was maintained even after controlling for differences in treatment. As reported previously (8), the description of the case series indicated that in relation to series reported from other centers, this group of patients had a larger proportion with normal biopsy findings (WHO class I) or mesangial lesions (class 11), a lower average value of serum creatinine, and a smaller proportion who were prescribed prednisone. In conjunction with the rarity of end-stage renal disease (6 cases), these results indicate that in this case series, there was a milder level of renal disease at the time of biopsy than in series reported from other centers. Compared with previously published studies of biopsied SLE patients, overall patient survival at 5 and 10 years was slightly higher (5,20) or similar (6,21) in this study, suggesting that patients died of conditions other than renal disease that were associated with SLE or its treatment, as we have previously reported (22). Several clinical and laboratory variables observed at the time of biopsy were found to be associated with an increased risk of death. The largest relative risks were obtained for indicators of renal 1 .o 1.o = Systemic Lupus dysfunction (low creatinine clearance, elevated serum creatinine, and urinary casts). Other clinical observations that achieved statistical significance included the presence of cushingoid features, infection, and muscle weakness. The SLEDAI, as a summary measure of multiple manifestations of active SLE, was also significantly associated with mortality. These findings are consistent with previous studies that focused on clinical and laboratory features, which have identified associations between mortality and renal disease manifestations (23) and, in particular, elevated serum creatinine (24). Only 2 of the clinical and laboratory variables were contained in the best-fit multivariate model for mortality. These were indicators of higher SLEDAI (scores of 10-19 or 220) and elevated serum creatinine (>120 pmoleslliter). The relative risk estimates were 1.5 (95% CI = 0.6,4.0) and 4.7 (95% CI = 1.9,11.9) for the 2 SLEDAI classes, and 5.9 (95% CI = 2.5,13.8) for the serum creatinine class. Each of these relative risk estimates was slightly greater than the value obtained in the univariate analyses. The ability of the SLEDAI to summarize information about a large number of variables, some of which were correlated, made this analysis of prognosis much more feasible. It should be noted, however, that SLEDAI and serum creatinine are not independent, since renal manifestations (i.e., casts, hematuria, proteinuria) are components of the SLEDAI, and serum creatinine is one of the standard measures of renal function. Even though the 2 variables were related, their correlation was not so great that the regression models failed to converge. The fact that the serum 566 creatinine variable had a significant positive coefficient in the model that also contained the SLEDAI, indicates that SLEDAI underestimated the risk associated with the presence of renal disease in this case series. In contrast to the univariate analyses (8), renal biopsy cla.sses were not significantly associated with mortality when added to the multivariate clinicallaboratory model for the complete patient sample. Relative risk estimates of 2.1 (95% CI = 0.9,4.9) and 2.3 (95% CI = 0.7,7.8) were obtained for the proliferative and chronic lesion classes, respectively. In general, these models suggested that the biopsy observations did not contribute significantly to the prognostic information provided by the SLEDAI and serum creatinine at the time of biopsy. Previous studies that reported multivariate models of the relationship between SLE prognosis and clinical and laboratory variables (2,3,25-28) each derived slightly different final models; however, there was general consistency in the finding that renal function (2,3,25-27) and SLE disease activity (2) are among the most important indicators of prognosis. The SLEDAI was not included in previous multivariate studies of prognosis; however, an activity index that included only laboratory test results was a significant predictor of renal function in the model reported by Whiting-O’Keefe et a1 (2). Of the previous studies that examined prognosis in terms of patient survival (1,5,6,25,27,29,30),3 considered the impact of a renal biopsy (1,5,6). Among previous studies that examined the importance of a renal biopsy while controlling for other factors in multivariate analyses ( M ) , 3 of which considered patient survival as an outcome (1,5,6), chronic renal lesions were associated with poor prognosis in all but one (6). A major difference from previous studies is that the present study considered prognosis in a sample of patients with less advanced renal disease and among whom end-stage renal disease occurred rarely; previous studies focused on SLE patients with lupus nephritis (1-6,28). The focus of previous studies on patients with severe renal disease would imply that they probably included a large number of study patients with elevated serum creatinine. The results of this study suggest that in such patients, prognosis in terms of mortality can be adequately assessed by clinical and laboratory measures, and therefore biopsies would provide little additional information. The present study demonstrated that biopsy variables provided important prognostic information beyond that available from clinical and laboratory measures in patients McLAUGHLIN ET AL who had normal levels of serum creatinine at the time of biopsy. In addition, it should be noted that in patients with renal dysfunction, a renal biopsy may sometimes be necessary to distinguish between active and chronic renal lesions as an aid in making therapeutic decisions. Among the strengths of this study were that data were collected prospectively using a standardized protocol, that a fixed date of entry into the study was defined, that complete and detailed clinical information was obtained on each patient, and that a concerted effort was made to obtain complete followup information. This study shares some of the usual limitations of clinic-based cohort studies. For example, it is difficult to assess how representative the patient series was of cases in the population, although it should be noted that the study included patients referred from a wide range of sources (e.g., family physicians, specialists, other patients). Whereas this study succeeded in detecting factors associated with large relative risks of dying, it did not allow variables to be ruled out as prognostic factors with certainty since confidence intervals were wide for many variables. In order to obtain more precise estimates of prognostic significance, larger studies are required; such studies could be accomplished by pooling samples of patients from different research centers. In conclusion, several factors were significantly associated with mortality in univariate analyses, and the relative importance of these factors was examined in multivariate analyses. The magnitude of the relative risk estimates demonstrated that the indicators of renal disease were strong predictors of mortality, and that the presence of active disease (e.g., high SLEDAI score) was a weaker, but still significant, indicator of risk. The result of the subgroup analysis has important clinical implications, because it suggests that the contribution of new prognostic information by renal biopsy is limited to patients who have normal renal function. In patients with renal dysfunction, as indicated by an elevated level of serum creatinine, the biopsy contributed no new information about the risk of dying. Since the majority of SLE patients have normal levels of serum creatinine (e.g., 85% in this study), these results indicate that renal biopsies do serve an important role in the assessment of prognosis in SLE. If this relationship is confirmed in other centers, further studies that examine the balance between the costs and utilities of renal biopsies in the majority of SLE patients will be required. KIDNEY BIOPSY IN SLE 567 15. Gladman D, Goldsmith C, Urowitz M, Bacon P, Bombardier C, The assistance of Dr. C. H. Chang is gratefully acknowledged. 16. REFERENCES 1. Ballou S, Chung-Park M, Waggoner DM, Kushner I: Prognostic value of clinical and renal biopsy findings in lupus nephritis (abstract). Arthritis Rheum 23:651-652, 1980 2. Whiting-O’Keefe Q, Henke JE, Shearn MA, Hopper J Jr: The information content from renal biopsy in systemic lupus erythematosus: stepwise linear regression. Ann Intern Med 96:718723, 1982 3. Austin HA, Muenz LR, Joyce KM, Antonovych lT,Kullick ME, Klippel JH, Decker JL, Balow JE: Prognostic factors in lupus nephritis: contribution of renal histologic data. Am J Med 75:382-391, 1983 4. Balow JE: Therapeutic trials in lupus nephritis. Nephron 27: 5. !\lossent HC, Henzen-Logmans SC, Vroom TM, Berden JHM, Swaak TJG: Contribution of renal biopsy data in predicting outcome in lupus nephritis: analysis of 116 patients. Arthritis Rheum 33:970-977, 1990 6. Esdaile JM, Levinton C, Federgreen W, Hayslett JP, Kashgarian M: The clinical and renal biopsy predictors of long-term outcome in lupus nephritis: a study of 87 patients and review of the literature. Q J Med 72:779-833, 1989 7. Gladman DD, Urowitz MB, Cole E, Ritchie S, Chang CH, Churg J: Kidney biopsy in SLE. I. A clinical-morphologic evaluation. Q J Med 73: 1125-1 133, 1989 8. McLaughlin JR, Gladman DD, Urowitz MB, Bombardier C, Farewell V, Cole E: Kidney biopsy in systemic lupus erythematosus. 11. Survival analyses according to biopsy results. Arthritis Rheum 34:1268-1273, 1 9 9 1 9. Tan EM, Cohen AS, Fries JF, Masi AT, McShane DJ, Rothfield NF, Schaller JG, Tala1 N, Winchester RJ: The 1982 revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum 25:1271-1277, 1982 10. Churg J, Sobin LN: Renal Disease Classification and Atlas. Vol. I: Glomerular Diseases. Tokyo, Igaku-Shoin, 1982 1 1 . Kashgarian M: New approaches to clinical pathologic correlation in lupus nephritis. Am J Kidney Dis 2:164-169, 1982 12. American Rheumatism Association Glossary Committee: Dic13. American Rheumatism Association Glossary Committee: Dictionary of the Rheumatic Diseases. Vol. 11: Diagnostic Testing. New York, Contact Associates, 1985 14. Bombardier C, Gladman DD, Urowitz MB, Caron D, Chang CH, and the Committee on Prognosis Studies in SLE: Derivation of the SLEDAI: a disease activity index for lupus patients. Arthritis Rheum 35:630640, 1992 17. 18. 19. 20. 21. 22. 23. 24. 25. Isenberg D, Kalunian K, Liang MH, Maddison P, Nived 0 , Richter M, Snaith M, Symmons D, Zoma A: Cross cultural validation and reliability of three disease activity indices in systemic lupus erythematosus. J Rheumatol 19:608-611, 1992 Petri M, Genovese M, Engle E, Hochberg M: Definition, incidence, and clinical description of flare in systemic lupus erythematosus: a prospective cohort study. Arthritis Rheum 34:937-944, 1991 Kaplan EL, Meier P: Nonparametrk estimation from incomplete observations. J Am Stat Assoc 53:457481, 1958 Kalbfleisch JD, Prentice RL: The Statistical Analysis of Failure Time Data. New York, John Wiley, 1979 SAS Institute: SAS User’s Guide: Version 5 . Cary NC, SAS Institute, 1985 Appel GB, Cohen DJ, Pirani CL, Meltzer JI, Estes D: Longterm follow-up of patients with lupus nephritis: a study based on the classification of the World Health Organization. Am J Med 83:877485, 1987 Tateno S, Kobayashi Y, Shigematsu H, Hiki Y: Study of lupus nephritis: its classification and the significance of subendothelial deposits. Q J Med 52:311-331, 1983 Rubin L, Urowitz MB, Gladman DD: The bimodal pattern revisited. Q J Med 55:87-98, 1985 Wallace DJ, Podell T, Weiner J , Klinenberg JR, Forouzesh S, Dubois EL: Systemic lupus erythematosus survival patterns: experience with 609 patients. JAMA 245:934-948, 1981 Fries JF, Weyl S, Holman HR: Estimating prognosis in systemic lupus erythematosus. Am J Med 57561-565, 1974 Ginzler EM, Diamond HS, Weiner M, Schlesinger M, Fries JF, 26. Magi1 AB, Ballou HS, Chan V, Lirenman DS, Rae A, Sutton RA: Diffuse proliferative lupus glomerulonephritis: determination of prognostic significance of clinical, laboratory and pathologic factors. Medicine (Baltimore) 63:21&220, 1984 27. Reveille JD, Bartolucci A, Alarc6n GS: Prognosis in systemic lupus erythematosus: negative impact of increasing age at onset, black race, and thrombocytopenia, as well as causes of death. Arthritis Rheum 33:3748, 1990 28. Ward MM, Studenski S: Clinical prognostic factors in lupus nephritis: the importance of hypertension and smoking. Arch Intern Med 152:2082-2088, 1992 29. Seleznick MJ, Fries JF: Variables associated with decreased survival in systemic lupus erythematosus. Semin Arthritis Rheum 21:73-80, 1991 30. Studenski S, Allen NB, Caldwell DS, Rice JR, Polisson RP: Survival in systemic lupus erythematosus: a multivariate analysis of demographic factors. Arthritis Rheum 30: 1326-1332, 1987
1/--страниц