Abstract
Objective. The frequency of endstage renal disease (ESRD) from systemic lupus erythematosus (SLE) in the United States has not improved over the last few decades in large population datasets. Understanding the risk factors for renal failure in SLE could lead to earlier detection of lupus nephritis and potentially more effective treatments in those with markers of poor prognosis.
Methods. The Hopkins Lupus Cohort, comprising 2528 patients was used. One hundred fifty-one patients experienced renal failure after SLE diagnosis, defined as dialysis or renal transplant. We estimated the risk of renal failure in subgroups defined by demographics, laboratory tests, and the American College of Rheumatology/Systemic Lupus International Collaborating Clinics (ACR/SLICC) classification criteria satisfied within 1 year of SLE diagnosis.
Results. The overall incidence of renal failure within 20 years of SLE diagnosis was 8.4%. The risk was much higher (20.0%) among those who experienced proteinuria within the first year of diagnosis. Demographic predictors included African American ethnicity [rate ratio (RR) 1.82, P = 0.0012] and age ≥ 40 years at SLE diagnosis (RR 0.51 vs those with diagnosis at < 30 yrs of age, P = 0.019). Among immunologic markers, low C3 was a strong predictor of renal failure (RR 2.00, P = 0.0011).
Conclusion. Proteinuria within the first year of diagnosis of SLE is one of the most important predictors of ESRD. Our data also confirm African American ethnicity, younger age at SLE diagnosis, and low C3 as strong predictors of renal failure.
Models to estimate the risk of endstage renal disease (ESRD) after diagnosis of chronic kidney disease stages 3 to 5 have been developed for the general population1, but comparable studies have not been performed for patients with systemic lupus erythematosus (SLE). ESRD remains a major challenge in SLE management. Large population datasets have shown no improvement in SLE renal outcomes over the last few decades2,3. Groups particularly at risk include African Americans and, to a lesser extent, Hispanic Americans4. Patient-specific factors matter as well. Our group previously showed that predictors of poor renal outcomes included nonadherence with clinic visits and uncontrolled hypertension5.
Advances in the treatment of lupus nephritis (LN) have not yet translated into improvement in long-term renal outcomes. Clinical trials have shown the equivalence (and, in non-Whites, the superiority) of mycophenolate mofetil (MMF) over oral or intravenous cyclophosphamide6,7 and the superiority8 or equivalence9 of MMF over azathioprine. However, on average, only 50% of patients with LN have a complete renal response to MMF at 1 year. The recent phase 1 results of the US National Institutes of Health (NIH) Foundation Amplified Medicines Partnership suggest that nonresponders may have activated other renal pathways (such as interferon-α and fibrotic pathways) that are important10.
Several demographic and immunologic factors have been shown to predict ESRD in those already diagnosed with LN. The most important demographic factor is ethnicity, with worse outcomes in non-Whites. LN among non-White patients was a predictor of renal failure in the Netherlands11,12. African Americans with LN had poorer renal outcomes in the NIH cohort13. Both African Americans and Hispanic Americans with LN had worse outcomes in a study in Miami14. African American ethnicity increases mortality after LN diagnosis15.
The important prognostic role of immunologic markers in LN (e.g., better prognosis with early normalization of low complement during treatment) has been identified in both cohort studies and analyses of randomized clinical trials (RCT) of LN16. In these studies, low complement, particularly C3, appeared to be an important marker of renal outcomes13,17,18,19,20. However, not all lupus nephritis is associated with low complement.
Most previous studies have examined outcomes only after the onset or identification of LN. Past studies have examined predictors of renal outcomes at 5 years21 or 10 years11 but not beyond. This is the first cohort study of predictors of ESRD in an SLE population with both White and African American ethnicity followed for almost 13 years on average, in which detailed information on clinical and immunologic predictors are available since the time of SLE diagnosis. In addition, we have developed a formula to estimate the risk of renal failure within 10 and 20 years of SLE diagnosis.
MATERIALS AND METHODS
Patients. The Hopkins Lupus Cohort is an ongoing longitudinal cohort of patients with SLE presenting from the community and counties surrounding Johns Hopkins University School of Medicine and which was begun in 1986. Patients who met classification criteria for SLE [either the revised American College of Rheumatology (ACR) criteria22,23 or Systemic Lupus International Collaborating Clinics (SLICC) classification criteria24] are enrolled. All patients give written, informed consent. The study has been approved on a yearly basis by the Johns Hopkins University School of Medicine Institutional Review Board (Study number NA_00039294).
At enrollment into the cohort, a detailed patient history of SLE is collected based on patient report and medical records. Subsequently, patients are seen quarterly (or more often if medically necessary). At each visit, disease activity indices [including the physician global assessment as part of the Lupus Activity Index25 and the Safety of Estrogens in Lupus Erythematosus National Assessment (SELENA) revision of the SLE Disease Activity Index26] are completed by 1 rheumatologist (MP). Laboratory tests were performed at the Johns Hopkins Clinical Laboratories as well, to complete the indices (complete blood count, serum creatinine, urinalysis, urine protein /creatinine, C3, C4, anti-dsDNA by Crithidia). Renal failure was defined as the need for dialysis or renal transplant.
This analysis is based on the Hopkins Lupus Cohort database as of August 2019. At that time, there were 2577 patients in the cohort who had follow-up since SLE diagnosis. Of these, 13 were excluded because they reported renal failure before the date of SLE diagnosis, and 30 were excluded because the date of renal failure was equivalent to the date of SLE diagnosis. Another 6 were removed due to unknown date of renal failure. The analysis was based on the remaining 2528 patients who were followed for a mean of 12.8 years. Baseline information on patients who were diagnosed with SLE before cohort entry was obtained from the review of all medical record data from onset of SLE. The patient characteristics are shown in Table 1.
Characteristics of patients included in the analysis.
Statistical analyses. The survival analyses were based on the time from SLE diagnosis until the end of follow-up or renal failure/transplant for patients in our cohort. For patients diagnosed with SLE before entering our cohort, the date of diagnosis and clinical manifestations at the time of diagnosis were based on the comprehensive review of all medical records and patient history collected at cohort entry.
We estimated the risk of experiencing renal failure over time among all patients and within subgroups of patients using the Kaplan-Meier approach. Cox regression models were used to estimate the joint association between multiple patient characteristics and renal failure. These models were built in a stepwise fashion, including as candidates those variables found significantly associated in univariate models and excluding those which were no longer statistically significant (P > 0.05) after adjustment for other variables.
RESULTS
A total of 151 patients developed renal failure after the diagnosis of SLE. Figure 1 shows the estimated cumulative incidence of renal failure in our cohort. The risk of renal failure was estimated to be 2.7% (95% CI 2.1–3.5%) within 5 years, 4.8% (95% CI 4.0–5.9%) within 10 years, and 8.4% (95% CI 7.0–10.0%) within 20 years of SLE diagnosis.
Incidence of renal failure by time since diagnosis based on Kaplan-Meier estimation. The dotted lines constitute 95% CI for the incidence at each timepoint. DX: diagnosis; SLE: systemic lupus erythematosus.
The 10- and 20-year risk of renal failure, by patient subgroup, is shown in Table 2. In univariate subgroup analyses, we found a statistically significant elevated risk of renal failure among those who were diagnosed with SLE at a younger age, in African Americans, and those with lower education (high school and below). There was no statistically significant change in risk based on calendar time of diagnosis. Those with proteinuria at the time of SLE diagnosis were at the highest risk. The 10- and 20-year risks of renal failure in patient subgroups defined by the presence of immunologic markers are shown in Table 3. Patients who had low complement, anti-dsDNA, anti-Sm, or lupus anticoagulant (LAC; positive dilute Russell viper venom test) at any time during their disease course were at higher risk of renal failure.
10-year and 20-year risk of renal failure in patient subgroups.
10- and 20-year risk of renal failure in patient subgroups defined by immunologic markersa.
We used multiple Cox regression to derive a formula for risk prediction based on multiple variables. The higher risk of renal failure among African Americans persisted after adjustment for socioeconomic status (as measured by education and income). Education was no longer statistically significantly associated with renal failure after adjustment for race and ethnicity. Although income < US$50,000 was still a statistically significant predictor of higher risk after adjustment for other variables, we did not include it in the final risk formula because (1) it was missing for 10% of the observations, and (2) we chose to confine the risk model to biologic variables. The final multiple regression model is shown in Table 4. Demographic factors that remained in this final model were male sex [rate ratio (RR) 1.69, P = 0.062], age ≥ 40 years at SLE diagnosis (RR 0.51 compared to those with age of diagnosis < 30 yrs, P = 0.019), and African American ethnicity (RR 1.82, P = 0.0012). Those with proteinuria at the time of SLE diagnosis were at significantly increased risk (RR 2.75, P < 0.0001). Low C3 was the immunologic marker with the strongest association with renal failure (RR 2.00, P = 0.0011). The other immunologic markers were not statistically significantly associated with renal failure after adjustment for low C3. The model had good discrimination (c-statistic = 0.75) and good calibration (Figure 2).
Predictors of renal failure based on a multivariable Cox proportional hazards modela.
Calibration plots for the estimation of the 20-year risk of endstage renal disease with the Cox models. Dark bars indicate the average 10-year risk in each group based on the model. Gray bars indicate the Kaplan-Meier estimates of the 10-year risk in each group.
The RR in Table 4 can be used to calculate the 20-year risk for a patient with any risk factor profile using the formula:
where RR is the RR for that patient relative to a female who was diagnosed with SLE before the age of 30 and who had none of the other risk factors. For example, for a male diagnosed with SLE before the age of 30 with proteinuria in the first year after diagnosis, and no other risk factors, the hazard would be increased by a factor of 1.69 (because the patient is male) and by a factor of 2.75 (because of the proteinuria). Thus, the overall RR for such a patient would be (1.69 * 2.75), which equals 4.65. The 20-year risk is therefore estimated to equal 1 – 0.971(4.65), which computes to 12.8%. The 10-year risk can be calculated in the same way after substituting 0.984 for 0.971 in the above formula. Using this approach, we estimated that for females with all other risk factors (low age at diagnosis, African American, proteinuria, low complement, and lupus anticoagulant), the 10-year risk was equal to 21% (95% CI 13–28%) and the 20-year risk was estimated to be 35% (95% CI 23–45%).
DISCUSSION
Based on our cohort, among those without renal failure at the time of SLE diagnosis, the risk of renal failure 5, 10, and 20 years after SLE diagnosis was estimated to be 2.7%, 4.8%, and 8.4%, respectively. Note that this did not include 1.7% of the patients who had renal failure before or at the time of SLE diagnosis, so the percentage of SLE patients with renal failure within 10 or 20 years of SLE diagnosis is somewhat higher than our estimates of incidence after diagnosis. Our 5-year risk estimates are consistent with those of Plantinga, et al who reported 5-year risks of 2.5% and 6.4% in White and Black patients, respectively, in Atlanta, Georgia27. Our findings are also consistent with the incidence rates reported in Taiwan (2.5% in 6 years)28, and somewhat below those reported in Okinawa (9.4% in 10 years)29. The risk of renal failure did not decline over calendar time.
There are 3 major findings. First, while the 20-year risk of renal failure was 8.4% for the overall cohort, for a patient with a history of proteinuria at the time of SLE diagnosis, the risk of renal failure within 20 years was 20%. This large difference in outcome, not previously appreciated, can allow consideration of more aggressive therapies in this subgroup. Examples include multitarget therapy with MMF and tacrolimus30, newer calcineurin inhibitors31, experimental sequential regimens such as rituximab followed by belimumab32 on a background of MMF, and newer anti-CD20 biologics (ClinicalTrials.gov: NCT02550652).
Second, confirming previous studies, certain demographic factors, including male sex, African American ethnicity, and SLE diagnosis before age 40 years were statistically significant predictors of renal failure. The higher risk among African Americans persisted after adjustment for socioeconomic status as measured by education, although it is possible that socioeconomic status may not be on par with educational attainment in certain ethnic minorities. There was no difference in risk between SLE diagnosis at < 30 years and 30–39 years after adjustment for other predictors.
Third, consistent with the findings of other investigators13, we found that the immunologic marker that was the strongest predictor of renal failure was low C3. Low C4 and anti-dsDNA (which are highly correlated with low C3) were not statistically significant predictors of renal failure after adjustment for low C3 (although they were significantly associated in univariate analyses).
To our knowledge, this is the first 30-year cohort study to assess demographic, clinical, and immunologic predictors of later renal failure in SLE. It is the largest study to include a balance of White and African American patients, and the only one to have complete data on risk factors (due to the protocolized nature of follow-up).
Our study has several limitations. As this is a single-center study of patients who agreed to participate in the research cohort, drawing inferences about general SLE patients should be done with caution. Like most survival analyses, our estimates are based on the assumption that at any given timepoint, the risk of ESRD among those censored before that timepoint is similar to the risk among those observed at that specific timpoint. To the extent that this assumption is violated, our estimates would be biased. Because we did not have complete information on all deaths in the cohort, we could not account for the competing risk of death in our analysis, resulting in a slight overestimation of the 10- and 20-year cumulative risk. Our multivariable model could also have been biased due to missing covariate information, or unmeasured or residual confounding. Our analyses for immunological markers were based on whether they were ever manifested during cohort follow-up rather than at SLE diagnosis, which could limit their utility for renal outcome prediction early in the disease course. A final limitation is that, even in this large cohort, the number of events was limited to 151, limiting the power to identify subtle associations and the precision of our estimates.
We think these predictors of renal failure in SLE will be immediately useful in clinical practice to identify patients who might benefit from more aggressive treatment. In particular, those with LN in the first year after diagnosis have a greatly increased risk of later renal failure that would justify more aggressive induction therapy. These predictors will also be important in either stratifications for RCT or for adjustment in outcome analyses.
Footnotes
The Hopkins Lupus Cohort was funded by grants NIH AR043727 and NIH AR069572.
- Accepted for publication April 7, 2020.
- Copyright © 2021 by the Journal of Rheumatology