|Home | About | Journals | Submit | Contact Us | Français|
The published criteria for the proteinuria increase that constitutes a proteinuric flare in lupus glomerulonephritis (SLE GN) vary widely, likely because they are largely based on expert opinion. Ideally, the threshold for proteinuric flare should be set sufficiently high so that spontaneous variation in proteinuria does not likely explain the increase, but not so high that the patient is needlessly exposed to prolonged heavy proteinuria before a flare is declared and therapy is increased. Here we describe an evidence-based approach to setting the threshold for proteinuric flare based on quantifying the spontaneous variation in urine protein/creatinine (P/C) ratio of SLE GN patients who are not experiencing SLE flare.
SLE GN patients (N = 71) followed in the Ohio SLE Study (OSS) were tested at pre-specified bimonthly intervals within windows of ± 1 week, median follow-up > 44 mo, visit compliance > 90%. To assess spontaneous P/C ratio variation under no-flare conditions, we excluded P/C ratios measured within ± 4 month of renal flare.
For those with mean no-flare P/C ratios ≤ 0.5, the published flare thresholds are set well above the 99% confidence interval (CI) of the no-flare P/C ratios. The opposite is seen in those with patients whose mean no-flare P/C ratios ≥ 1.0.
Current thresholds for proteinuric flare appear to be set either too high or too low. A randomized trial would be needed to test whether re-setting the thresholds would result in faster remission, less therapy, and less chronic kidney disease.
It is generally agreed that an increase in proteinuria is the most common manifestation of a moderate or severe flare of SLE glomerulonephritis (GN). However, there is not general agreement regarding the amount of proteinuria increase that should constitute a proteinuric flare. As discussed in Methods, the criteria that have been proposed can be categorized as having low, intermediate, or high thresholds, depending upon the minimum increase in proteinuria that is deemed to constitute a proteinuric flare . To the best of our knowledge, the published criteria are based largely on expert opinion. Likely this accounts for the wide variation in the recommended thresholds for proteinuric flare.
It is also generally agreed that the proteinuria increase that defines a proteinuric flare should be set sufficiently high so that it is statistically unlikely that the proteinuria increase is attributable to spontaneous variation in proteinuria. On the other hand, the increase should not be so high that the patient is needlessly exposed to heavy proteinuria before a flare is declared and therapy is increased.
We suggest that the ideal approach to a rigorous definition of proteinuric flare would be to first determine the spontaneous variation in proteinuria in SLE GN patients who are clinically stable and not experiencing SLE flare. Flare thresholds would then be set above the upper boundary of spontaneous variation in proteinuria. This type of analysis requires a database in which SLE GN patients are tested at regular, pre-specified intervals so that the results are not confounded by ascertainment bias (e.g., sick SLE patients may be more likely to appear for testing than those who are well). The database also needs to include a large number of proteinuric flares so that the expected differences in proteinuria between flare and no-flare conditions can be defined. In addition, proteinuria needs to be measured accurately. This is best accomplished by measuring the protein/creatinine (P/C) ratio of intended 24-hr urine collections [2–4]. The Ohio SLE Study (OSS) database meets each of these conditions. The present study is based on analysis of the OSS database.
The Ohio SLE Study [3–13] has enrolled and followed 106 patients with recurrently active SLE, 71of whom (the renal cohort) have or had major kidney manifestations documented by kidney biopsy (ISN/RPS Class III, IV, or 5) . They are the object of the present study. To date, median OSS follow-up is greater than 44 months, involves more than 2400 visits, 90% of which are bimonthly and within pre-specified windows of ± 1 week. The renal SLE patients have provided intended 24-hr urine collections at 84% of their OSS visits. Random spot urine collections were provided at the rest of their OSS visits. The OSS patients received standard of care as previously described [14, 15].
To assess spontaneous variation in urine P/C ratios under no-flare conditions, the following protocol was used.
Using the approach described above, 58 of the 71 SLE GN patients (82%) provided 894 individual consecutive urine P/C ratios, which were assembled into 110 P/C ratio data sets. Of these 110 data sets, 9 (8%) were of 4-months’ duration (3 P/C ratios over 4 months) and 101 data sets (92%) were of 6 months or greater. Of the 58 patients in this study, 28 patients contributed P/C ratio data sets to 2 or 3 of the 5 P/C ratio Groups. The rest (30 patients) contributed only one P/C ratio data set to one of the 5 P/C ratio Groups. The P/C ratio Groups to which the 110 P/C data sets were stratified are as follows:
All clinical analyses were performed in the laboratories of The Ohio State University Medical Center. Urine protein was measured by an automated pyrogallol red method, coefficient of variation (CV) 3.3% at control level 71.6 mg/dl. Urine creatinine was measured by an automatic picric acid method: CV 2.8% at control level 79.0 mg/dl .
All mean values shown are ± 1 SD. To calculate the CI of the P/C ratios, bootstrapping techniques were used  because the P/C ratios are not normally distributed. One thousand replicates of the P/C ratios of each group were used. The normal based 95% and 99% CI intervals were very similar to the corresponding percentile based on CI (data not shown).
Table 2 shows the baseline clinical characteristics of the OSS patients described in the present study. The OSS is a typical US SLE cohort in which about 30 to 40% are AA. AA produce about 21% more creatinine than EA . Thus AA will tend to have higher serum creatinine levels, but lower urine protein/creatine ratios than EA. The influence of race on P/C ratios was not taken into consideration in the analyses provided below.
Table 3 shows the key statistics describing the P/C ratio data sets of P/C ratio Groups 1–5. As shown, P/C ratio variability increases according to the mean value of the P/C ratio Group. The individual P/C ratio values according to P/C ratio Group are displayed graphically in the next 3 figures.
Figure 1 shows the individual P/C ratios measured under no-flare conditions for the P/C ratio Groups 1 and 2. The individual P/C ratios are shown in relationship to the mean value and 99% CI for each of the P/C ratio groups. Also shown are the P/C ratio thresholds for representative low, intermediate, and high-threshold criteria for proteinuric flare (see Table 1). To convert the low, intermediate, and high proteinuria thresholds to P/C ratios, it was arbitrarily assumed that the mean 24-hr urine creatinine excretion was 1.3 g. As shown in Figure 1, the upper boundary of the 99% CI for the individual P/C ratios of Group 1 and Group 2 are well below even the low threshold criteria for proteinuric flare.
Figure 2 uses the same format as Figure 1 except that the individual P/C ratios of P/C ratio Groups 3 and 4 are shown. As can be seen, the upper boundary of the 99% CI for the individual P/C ratios of Groups 3 and 4 are well below even the low threshold criteria for proteinuric flare. The exception is the LTC-3 (BILAG B) where the threshold for proteinuric flare is at about the 95% CI of the P/C ratios that occur spontaneously under no-flare conditions. This suggests that this threshold for proteinuric flare is set at or near the appropriate level.
Figure 3, uses the same format as Figures 1 and and22 except the individual P/C ratios of P/C ratio Group 5 are shown. As shown, a pattern emerges that is different from that of P/C ratio Groups 1–3 but resembles that of P/C ratio Group 4 in that for P/C ratio Group 5 some of the flare thresholds are at P/C ratios that commonly occur spontaneously under no-flare conditions. This suggests that the threshold for proteinuric flare for Group 5 may be set too low.
The present work used the database of the OSS to assess the spontaneous variation in urine P/C ratio in SLE GN patients under no-flare conditions with the goal of developing an approach to validating SLE GN proteinuria flare criteria. The OSS is uniquely well suited for the present study. It is the largest prospective SLE database in which the patients are tested bimonthly and without bias regarding the patient’s condition (the bimonthly visits were pre-specified and within windows of ± 1 week). Proteinuria is assessed from the protein/creatinine (P/C) ratio of urine collections, the great majority of which were intended 24-hr urine collections. These provide the most reliable estimate of 24-hr proteinuria [2–4]. After each OSS visit it was determined, using pre-specified criteria , whether an SLE flare had occurred since the previous OSS visit. On this basis each OSS visit was designated a flare, or no-flare visit. All patients were receiving standard of care management of their SLE and their renal manifestations [14, 15]. To focus the analysis on the P/C ratios obtained under no-flare conditions, P/C ratio testing within ± 4 months of renal flares were excluded. Also excluded were P/C ratio data sets with missing values (see Methods). From these data the spontaneous variation in urine P/C ratio in the SLE GN cohort under no-flare conditions was estimated.
The no-flare P/C ratios were stratified into 5 separate P/C ratio groups based on the mean P/C ratio of the individual P/C ratio data sets. The rationale is that P/C ratio variability is strongly determined by the mean P/C ratio of the data set (see Table 2). Also, the five P/C ratio Groups were chosen to correspond to the baseline levels of proteinuria used in the published flare criterion.
We suggest that our estimate of P/C ratio variability in SLE GN patients under no-flare conditions is reliable, and generalizable. This interpretation is based on the following: a) the OSS is a typical U.S. SLE GN cohort; b) they are receiving standard of care; c) the bimonthly testing is highly consistent and without ascertainment bias; d) The OSS SLE GN patients have a demonstrated propensity to experience proteinuric flare. Specifically, each had experienced proteinuric flare in the past, and of the 58 patients who contributed P/C ratio data sets to this study, 42 of the 58 (72%) experienced flare during OSS follow-up; e) each P/C ratio Group had broad representation from the 58 patients of this study.
To the best of our knowledge, the present work is the first to determine what should constitute a proteinuric flare by first examining the expected variation in P/C ratio in SLE GN patients who are not experiencing SLE flare (no-flare status).
Our analysis shows that for SLE GN patients whose P/C ratio data sets have mean values ≤ of 0.5 (P/C ratio Groups 1–3) the upper boundary of the 99% CI of the spontaneous variation in P/C ratios under no-flare conditions is well below even the low threshold criteria for proteinuric flare. For example, if the mean baseline (pre-flare) 24-hr urine P/C ratio is 0.15 (this corresponds to a 24-hr proteinuria of about 200 mg, P/C ratio Group 1, see Methods), the upper boundary of the 99% CI is a P/C ratio of 0.34. A BILAG A proteinuria flare, however, would not be declared until the P/C ratio is nearly 0.8 (see Figure 1). For P/C ratio Group 3 (mean baseline P/C ratio ≥ 0.38 to ≤ 0.77), the P/C ratio would have to increase to > 1.5 or greater before a flare is declared (see Figure 2). However, the upper boundary (see Table 2) of the 99% CI of the P/C ratios in this P/C ratio group is 1.1 (see Table 2). Thus, for Group 3 patients, the best case scenario for declaring proteinuric flare would be when P/C ratio increases from 0.77 to ≥ 1.5. The worse case scenario would be when P/C ratio needed to increase from < 0.38 to ≥ 1.5 before a flare is declared.
For P/C ratio Group 4 a somewhat different picture emerges in that the low threshold criterion is well beneath the upper boundary of the 99% CI. The high threshold criteria, however, remain well above the 99% CI.
For P/C ratio Group 5, the trends noted for P/C ratio Group 4 are accentuated: the threshold criteria are now much below the 99% CI for spontaneous variation in P/C ratio.
Taken together, our analysis suggests that the most common types of proteinuric flare (P/C ratio Groups 1–3) would have delayed treatment of proteinuric flares, even if the low threshold criteria are used. On the other hand, for those with relatively high baseline proteinuria (P/C ratio Groups 4 and 5), therapy might be introduced before there is decisive evidence that proteinuria has changed.
To rigorously test whether a re-setting of the proteinuria flare thresholds would result in better outcomes would require a prospective trial in which SLE GN patients would be randomly assigned to management by either traditional flare criteria (for P/C ratio Groups 1–3 the current low threshold criteria are recommended) or to more stringent flare criteria (the upper boundary of the 99.9% CI seems appropriate for P/C ratio Group 1–3). For P/C ratio Groups 4 and 5 an absolute rather than a relative increase in proteinuria might be appropriate (see Table 2). Ideally the study physicians would be blinded as to whether the patient is being managed by “traditional” or “stringent” criteria.
The primary outcome of the suggested randomized trial would be 24-hr urine P/C ratio and serum creatinine level after management of the patients according to traditional or stringent flare criteria for a prolonged period (e.g., 3 years). The hypothesis is that the stringent criteria would result in lower serum creatinine levels and lower levels of proteinuria because flares would be treated earlier and remission would be achieved more quickly. This interpretation is based on the widely held belief that severe SLE flare requires more aggressive treatment than less severe SLE flares. Secondary outcome measures would be the number of relapses during follow-up and the amount of therapy (prednisone and immunosuppressives) needed to achieve the outcomes. The hypothesis is that the stringent flare criteria would result in fewer relapses and lesser use of steroids and immunosuppressives, for the reasons discussed above.
The issue of early treatment of kidney disease in order to prevent chronic kidney disease (CKD) may be particularly important for SLE GN patients because their kidney disease usually comes on at an earlier age than other common forms of glomerulopathy. For example, idiopathic membranous nephropathy has a peak incidence around the fifth decade of life whereas lupus nephritis has a peak incidence in the second or third decade of life . Thus exposure to conditions that can lead to chronic kidney disease (CKD) may be 20 to 30 years longer in SLE patients compared to those with idiopathic membranous nephropathy. Furthermore, GFR normally is lost at the rate of about 1 ml/min/year beginning at about age 40 [28, 29]. Also, “natural progression” of kidney disease emerges after > 50% of GFR is lost [28, 29]. Thus, the combination of these mechanisms of GFR decline could lead to end-stage renal disease later in life, even though the SLE nephritis itself was no longer active.
A further incentive to optimize preservation of kidney function in the SLE patient is the newly recognized independent association between CKD and cardiovascular disease [28, 29]. Thus, the threat to the SLE GN patient over time is not only kidney failure but also cardiovascular disease fostered by CKD.
The need to suppress proteinuria is particularly important in the SLE GN patients with heavy proteinuria. As we have previously discussed, trials in chronic proteinuric kidney disease show that for each 1 g of proteinuria reduction, GFR decline is slowed by about 1 to 2 ml/min/yr [24, 28, 29]. However, once 24-hr proteinuria has been reduced to < 500 ml/24 hr (P/C ratio of about 0.4), GFR loss attributable to proteinuria likely is minimal [28, 29]. Thus, the goal in SLE GN management is to reduce proteinuria to this level or lower [14, 15]. Also to be taken into account in assessing the risk of progression to CKD is whether the SLE GN flare is accompanied by hematuria or an increase in serum creatinine. These findings increase the risk of progression as described by Moroni et al 
In summary, we suggest the present work provides rigorous evidence that the currently used proteinuria flare criteria may not be optimal. This work is hypothesis generating. Suggestions are provided as to how this work might proceed in order to test this hypothesis.
Supported in part by NIH grants UO1 DK48621, PO1 DK 55546, the James D. Casto Research Fund and by Award Number UL1RR025755 from the National Center For Research Resources. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center For Research Resources or the National Institutes of Health.