Abstract
Objective
To characterize the longitudinal trajectory of estimated glomerular filtration rate (eGFR) in patients with systemic lupus erythematosus (SLE) and identify predictors of the change in eGFR trajectory.
Methods
The longitudinal eGFR levels of patients in the Hopkins Lupus Cohort were modelled by piecewise linear regression to evaluate the slope of different line segments. The slopes were classified into declining (≤ −4 mL/min/1.73 m2 per year), stable (−4 to 4 mL/min/1.73 m2 per year), and increasing (≥4 mL/min/1.73 m2 per year) states. The transition rate between states and the impact of clinical parameters were estimated by a Markov model.
Results
The analysis was based on 494 SLE patients. At a mean follow-up of 8.8 years, 347 (70.2%), 107 (21.7%), 33 (6.7%), and 7 (1.4%) patients had zero, one, two, and three state transitions, respectively. In patients with no transition, 37 (10.7%), 308 (88.8%), and 2 (0.6%) were in declining, stable, and increasing state, respectively. In patients with one transition, 43 (40.2%) changed from declining to stable state while 29 (27.1%) changed from stable to declining state. When patients were in a non-declining GFR state, those who were younger and African Americans were more likely to transition to a declining GFR state. In adjusted analyses, high blood pressure, C4 and low hematocrit were associated with change from non-declining to declining state. High urine protein-to-creatinine ratio also tended to be associated with change from non-declining to declining state. African American patients were less likely to move from declining to non-declining state. Use of prednisone was associated with change from declining to non-declining state.
Conclusions
Patients with high blood pressure, low complement C4, low haematocrit, and high urine protein-to-creatinine ratio are more likely to have a declining eGFR trajectory, while the use of prednisone stabilizes the declining eGFR trajectory.
INTRODUCTION
Systemic lupus erythematosus (SLE) is a chronic multisystem autoimmune disease that can affect the kidneys in approximately 50% of the patients.[1] Lupus nephritis (LN) is a strong risk factor of adverse clinical outcomes including end-stage renal disease (ESRD) and mortality. Around 10% of patients with LN can progress to ESRD.[2] The prognosis after ESRD is poor despite the improvements in treatment of LN and ESRD in recent years.[3] The prevalence of LN varies among different races and ethnicities. African American patients are known to have a higher prevalence of LN and a poorer prognosis as compared to Caucasian patients.[1, 4, 5] On the other hand, a rising prevalence of high body mass index, diabetes mellitus and hypertension is observed among patients with LN.[3]
The renal function of SLE patients changes dynamically over time.[6] This can depend on their underlying SLE disease activity, received treatment, and other related comorbidities. Reduced urinary and renal function may not be observed in some SLE patients early in their illness, yet over half of the patients can develop renal manifestations ranging from mild proteinuria to rapidly declining renal function later in the course.[7, 8] As a result, some patients can have a relatively stable renal function trajectory earlier in their clinical course, which is followed by a declining renal function trajectory after renal manifestations. In contrast, treatment of LN can ameliorate the inflammation and may restore a stable renal function trajectory from a declining trajectory. While previous studies focused on the clinical outcomes including ESRD, survival and remission in patients with LN, the data remain scarce on the longitudinal renal function trajectory of SLE patients.
We postulated that patients with SLE can be characterized as being in one of three states with respect to kidney function, as measured by glomerular filtration rate (GFR), i.e. stable GFR, declining GFR, or improving GFR. Using data from a large American clinical cohort, we characterized the states that patients were in over time, and identified variables that predicted transitions from one state to another.
METHODS
Study Design and Data Source.
We performed a retrospective cohort study using data from the Hopkins Lupus Cohort. The Hopkins Lupus Cohort is a longitudinal study of patients with SLE enrolled and followed at John Hopkins University since 1987. The cohort was approved by the Johns Hopkins University School of Medicine Institutional Review Board. All patients gave informed, written consent. According to the protocol, patients are prospectively followed every 3 months, or more frequently depended on clinical indication.
Subjects.
All patients with SLE who participated in the Hopkins Lupus Cohort in or after year 2006 were identified. Patients with at least 17 clinical visits were included. Patients aged below 18 years old were excluded. Patient records with missing creatinine measurement (2% of visits) or before the date of SLE diagnosis were excluded. All patients met the revised American College of Rheumatology or Systemic Lupus International Collaborating Clinics classification criteria for SLE [9, 10]. Patient records after a gap in cohort visits of one year or more were excluded. Patients were followed until death, diagnosis of ESRD or renal failure, last follow-up date, or last date of data collection (January 31, 2018), whichever came first.
Data collection.
Data were retrieved from the Hopkins Lupus Cohort database in February 2018. The baseline date was defined as the date of the first visit after year 2006. Demographic data including age, gender and ethnicity were collected and recorded at baseline. Data on serum creatinine, complement component 3 (C3) and component 4 (C4) level, anti-double stranded DNA, SELENA revision of the SLE Disease Activity Index (SLEDAI) [11], urine protein-to-creatinine ratio, hematocrit and systolic blood pressure were prospectively collected at each quarterly clinic visit. Treatment for SLE including the use of hydroxychloroquine, prednisone, immunosuppressants, and antihypertensive medications was also reported at each visit.
Renal function assessment.
Estimated glomerular filtration rate (eGFR) was determined using the Chronic Kidney Disease Epidemiology Collaboration equation expressed as a single equation: GFR (in mL/min/1.73 m2) = 141 × min (Scr/κ, 1)α × max(Scr/κ, 1)−1.209 × 0.993Age × 1.018 [if female] × 1.159 [if black]; where Scr is serum creatinine in mg/dL, κ is 0.7 for females and 0.9 for males, α is −0.329 for females and −0.411 for males, min indicates the minimum of Scr/κ or 1, and max indicates the maximum of Scr/κ or 1.[12]
Definitions.
High systolic blood pressure was defined as systolic blood pressure ≥140 mmHg. Urine protein-to-creatinine ratio >0.5 g/day was considered as high. Hematocrit <41% for males and <36% for females was defined as low.
Statistical analysis.
Data were analyzed using Statistical Product and Service Solutions (SPSS) version 25.0 (SPSS, Inc., Chicago, Illinois), SAS (9.4; SAS Institute Inc., Cary, NC), R software (3.5.1; R Foundation for Statistical Computing, Vienna, Austria) and Joinpoint Regression Program (Version 4.6.0.0. - April 2018; Statistical Methodology and Applications Branch, Surveillance Research Program, National Cancer Institute). Continuous variables were expressed in mean ± standard deviation or median (interquartile range [IQR]), as appropriate, while categorical variables were presented as frequency (percentage).
Trajectory analysis was used to subtype patients.[13], The trajectory of eGFR was modeled separately for each patient by piecewise linear regression which consisted of line segments connected at changepoints. The location of changepoints and the corresponding regression coefficients of each line segments were estimated by the grid search method; the least-square estimate with the lowest residual sum of square among all possible locations for changepoints was identified.[14] The number of changepoints was chosen by Bayesian information criterion; the model with the minimum Bayesian information criterion was selected as the final model. A minimum of eight data points excluding the changepoints was included in each line segment to ensure sufficient data points on measurement of mid to long-term trend, and prevent short-term fluctuation. As a result, a minimum of seventeen data points was required for one changepoint to occur. The slope of each line segment represented the annual change in eGFR. The change in eGFR trajectory was indicated by the change of slope of the line segments. The location of changepoints indicated the time of the change in eGFR trajectory.
At each point in time, based on their estimated eGFR trajectories, to summarize the salient features of the trajectories, patients were defined to be in one of three possible states of eGFR trajectory, namely declining state, stable state, and increasing state. Each of the three states ---declining, stable, and increasing state ---corresponded to a different range of slope of the line segment. For the declining state, stable state, and increasing state, the range of slope was ≤ −4, −4 to 4, and ≥4 mL/min/1.73 m2 per year, respectively. We analysed these states assuming a Markov model where time in each state followed an exponential distribution that could depend on covariates. The transition rate and transition probabilities with 95% confidence interval (CI) between the three states were evaluated. Due to the limited number of transitions between stable state and increasing state, we combined stable state and increasing state into one state, namely a “non-declining” state to examine the impact of covariates on the transition rate. Transition from the non-declining state to declining state was considered as deterioration in renal function trajectory, while transition from declining state to non-declining state were considered as improvement in renal function trajectory.
Transition rates were modelled by a proportional hazards model.[15] Hazard ratios and adjusted hazard ratios (aHR) with 95% CI for the covariates on state transition were reported. We included covariates (percentage of missing data): patients’ age (0%), eGFR level (0%), high systolic blood pressure (0.1%), high urine protein-to-creatinine ratio (8.0%), detectable anti-double stranded DNA (2.2%), SLEDAI without renal variables (0.2%), low complement C3 level (1.3%), low complement C4 level (1.6%), low hematocrit level (0.3%), and use of prednisone (0.1%), hydroxychloroquine (0%), and antihypertensive drugs (0.05%) in each clinical visit as time-dependent covariates. At each point in time, the rate of transition from one state to another was modelled as a function of the current value of these time-varying predictors. Patients’ gender and race were included as time-independent covariates. Missing data were excluded from the analysis.
Results
Baseline Characteristics.
We identified 2,515 SLE patients with 63,890 clinical visits in the Hopkins Lupus Cohort database. We excluded 48,561 records of clinical visit of 2,021 patients according to the exclusion criteria. Finally, 494 SLE patients with 15,329 clinical visits were included and analyzed. Two hundred and sixty-one (52.8%) patients were between 40 and 59 years old at the first clinical visit; 183 (37.0%) were younger than 40 years old; 50 (10.1%) were at or above 60 years old at the first clinical visit; 455 (92.1%) patients were female; 267 (54.0%) patients were Caucasian; 185 (37.4%) patients were African American (Table 1). The median time between SLE diagnosis and baseline visit of the patients was 6 years. Their mean SLEDAI was 2.7 ± 3.4 at baseline. Three hundred and eight (62.3%) patients had an eGFR ≥90 mL/min/1.73 m2 at baseline; 152 (30.8%) had an eGFR between 60 and 89 mL/min/1.73 m2 at baseline; 34 (6.9%) had a baseline eGFR <60 mL/min/1.73 m2 (Table 1). Patients were followed at a mean follow-up of 8.8 ± 2.6 years. Eight patients developed renal failure or ESRD while 13 patients died.
Table 1.
Clinical characteristics of patients with systemic lupus erythematosus (SLE) at the baseline clinic visit.
| Clinical characteristics | All patients N=494 |
|---|---|
| Age at the first clinical visit (years) | |
| <40 | 183 (37.0) |
| 40-59 | 261 (52.8) |
| ≥60 | 50 (10.1) |
| Year since SLE diagnosis at first clinical visit (years) | |
| <2 | 116 (23.5) |
| 2-5 | 128 (25.9) |
| 6-9 | 99 (20.0) |
| ≥10 | 151 (30.6) |
| Gender (n, %) | |
| Female | 455 (92.1) |
| Male | 39 (7.9) |
| Race/ethnicity (n, %) | |
| Caucasian | 267 (54.0) |
| African American | 185 (37.4) |
| Other | 42 (8.5) |
| Estimated glomerular filtration rate (mL/min/1.73 m2) | |
| ≥90 | 308 (62.3) |
| 60-89 | 152 (30.8) |
| <60 | 34 (6.9) |
| SLEDAI | 2.7 ± 3.4 |
| SLEDAI without renal variables | 2.1 ± 2.6 |
| Detectable anti-double stranded DNA (n, %) | |
| Yes | 116 (23.5) |
| No | 378 (76.5) |
| Low complement component 3 (n, %) | |
| Yes | 79 (16.0) |
| No | 415 (84.0) |
| Low complement component 4 (n, %) | |
| Yes | 75 (15.2) |
| No | 419 (84.8) |
| High urine protein-to-creatinine ratio a | |
| Yes | 47 (9.5) |
| No | 447 (90.5) |
| Low hematocrit (n, %) b | |
| Yes | 169 (34.2) |
| No | 325 (65.8) |
| High systolic blood pressure c | |
| Yes | 100 (20.2) |
| No | 394 (79.8) |
| Use of oral prednisone (n, %) | |
| Yes | 208 (42.1) |
| No | 286 (57.9) |
| Use of immunosuppressants (n, %) | |
| Yes | 159 (32.2) |
| No | 335 (67.8) |
| Use of hydroxychloroquine (n, %) | |
| Yes | 368 (74.5) |
| No | 126 (25.5) |
| Use of antihypertensive drugs (n, %) | |
| Yes | 226 (45.7) |
| No | 268 (54.3) |
| Follow-up duration (years) | 8.8 ± 2.6 |
Continuous variables were expressed in mean ± standard deviation.
High urine protein-to-creatinine ratio was defined as urine protein-to-creatinine ratio >0.5.
Low hematocrit was defined as <41% for males and <36% for females.
High systolic blood pressure was defined as ≥140 mmHg.
SLE = systemic lupus erythematosus, SLEDAI = systemic lupus erythematosus disease activity index.
The eGFR trajectories.
The eGFR trajectory of nine selected patients was shown (Figure 1). The median (IQR) number of clinical visits of patients was 32 (24 – 38) during follow-up. Among 494 eGFR trajectories, 347 (70.2%) had no state transition; 107 (21.7%) had one state transition; 33 (6.7%) had two state transitions and 7 (1.4%) had three state transitions. Among 347 eGFR trajectories with no state transition, 37 (10.7%), 308 (88.8%), and 2 (0.6%) were in declining, stable, and increasing state, respectively. Among 107 eGFR trajectories with one state transition, 43 (40.2%) changed from declining to stable state; 29 (27.1%) changed from stable to declining state; 11 (10.3%) changed from declining to increasing state; 11 (10.3%) changed from increasing to declining state; 10 (9.3%) changed from increasing to stable state and 3 (2.8%) changed from stable to increasing state. Among 33 eGFR trajectories with two state transition, 8 (24.2%) were initially in a stable state, followed by a declining state and then a stable state; 7 (21.2%) were initially in a declining state, followed by an increasing state and then a stable state; 6 (18.2%) were initially in a declining state, followed by an increasing state and then a declining state; 12 (36.4%) had other types of eGFR trajectories.
Figure 1.
eGFR trajectory of selected patients A to I. Gray line represented the estimated trajectory by piecewise linear regression. Black dot represented the eGFR of the patient in each visit. eGFR= estimated glomerular filtration rate.
The frequency of state transition.
One hundred and ninety-four (1.3%) state transitions occurred among 14,835 possible transitions of 494 patients (Table 2). Deterioration of eGFR trajectory occurred more frequently than improvement in eGFR trajectory; 96 (3.5%) transitioned from a declining state to an increasing or stable state, i.e. non-declining state, among 2,778 possible transitions from declining state, while 74 (0.6%) transitioned from a non-declining state to a declining state among 12,057 possible transitions from non-declining state.
Table 2.
Estimated glomerular filtration rate (eGFR) state at each visit, given state at the previous visit.
| eGFR state at the previous visit | eGFR state at current visit (N, %) |
||
|---|---|---|---|
| Declining state | Stable state | Increasing state | |
| Declining state | 2,682 (96.5) | 62 (2.2) | 34 (1.2) |
| Stable state | 44 (0.4) | 11,265 (99.6) | 6 (0.05) |
| Increasing state | 30 (4.0) | 18 (2.4) | 694 (93.5) |
One hundred and forty-seven (29.8%) patients had at least one state transition at a mean follow-up of 8.8 ± 2.6 years. Given that a patient was in stable state, the probability that the patient would still be in stable state after one, five, and ten years was 99%, 94%, and 91% respectively. Given that a patient was in declining state, the probability that the patient would still be in declining state after one, five, and ten years was 88%, 56%, and 37% respectively. Given that a patient was in increasing state, the probability that the patient would still be in increasing state after one, five, and ten years was 76%, 28%, and 11% respectively (Table 3).
Table 3.
Estimated probability and expected time of being in each state of estimated glomerular filtration rate (eGFR) trajectory after 1, 5, and 10 years given one’s initial state.
| Initial state | Estimated probability of being in each status (95% CI) | Expected time spent in each trajectory (years) | ||||
|---|---|---|---|---|---|---|
| Declining | Stable | Increasing | Declining | Stable | Increasing | |
| 1 year | ||||||
| Declining | 0.88 (0.85-0.90) | 0.08 (0.06-0.11) | 0.04 (0.03-0.06) | 0.937 | 0.042 | 0.022 |
| Stable | 0.01 (0.009-0.02) | 0.99 (0.98-0.99) | 0.002 (0.0008-0.004) | 0.006 | 0.993 | 0.0009 |
| Increasing | 0.15 (0.10-0.22) | 0.10 (0.06-0.16) | 0.76 (0.69-0.81) | 0.079 | 0.050 | 0.871 |
| 5 years | ||||||
| Declining | 0.56 (0.46-0.71) | 0.34 (0.23-0.53) | 0.09 (0.05-0.17) | 3.748 | 0.924 | 0.328 |
| Stable | 0.05 (0.03-0.08) | 0.94 (0.91-0.98) | 0.009 (0.003-0.02) | 0.135 | 4.842 | 0.022 |
| Increasing | 0.35 (0.18-0.66) | 0.37 (0.18-0.75) | 0.28 (0.16-0.45) | 1.210 | 1.042 | 2.748 |
| 10 years | ||||||
| Declining | 0.37 (0.23-0.66) | 0.55 (0.33-1.00) | 0.08 (0.03-0.21) | 6.022 | 3.201 | 0.777 |
| Stable | 0.08 (0.05-0.15) | 0.91 (0.84-1.00) | 0.02 (0.005-0.05) | 0.459 | 9.457 | 0.084 |
| Increasing | 0.31 (0.12-0.82) | 0.58 (0.24-1.00) | 0.11 (0.04-0.33) | 2.899 | 3.467 | 3.634 |
CI = confidence interval.
The impact of covariates.
Table 4 shows the association between clinical characteristics and transition from a non-declining eGFR trajectory to a declining trajectory. Based on the univariate analyses, we found that young age, female gender, and African-American race were associated with a higher risk of transition from non-declining state to declining state. However, these differences are generally diminished after adjustment for differences in clinical covariates in the multivariable analysis. The most important independent predictors in the multivariable analyses were high systolic blood pressure (aHR 1.77, 95% CI 1.01 – 3.12, P=0.047), low complement C4 (aHR 1.96, 95% CI 1.02 – 3.77, P=0.044), and low hematocrit (aHR 1.71, 95% CI 1.02 – 2.86, P=0.040).
Table 4.
Univariate and multivariable analysis on the identified predictiors of transition from non-declining state to declining state.
| Parameters | Univariate analysis | Multivariable analysis | ||
|---|---|---|---|---|
| Hazard ratio (95% CI) | P value | Adjusted hazard ratio (95% CI) | P value | |
| Age (years) | ||||
| <40 | Referent | Referent | ||
| 40-59 | 0.51 (0.31 – 0.84) | 0.008 | 0.84 (0.48 – 1.46) | 0.536 |
| ≥60 | 0.53 (0.28 – 1.03) | 0.061 | 1.27 (0.58 – 2.76) | 0.555 |
| Male gender | 0.26 (0.07 – 1.08) | 0.064 | 0.28 (0.07 – 1.14) | 0.075 |
| Race | ||||
| Caucasian | Referent | Referent | ||
| African American | 1.63 (1.01 – 2.62) | 0.045 | 0.99 (0.57 – 1.72) | 0.974 |
| Other | 1.19 (0.50 – 2.83) | 0.697 | 0.88 (0.36 – 2.16) | 0.787 |
| eGFR (mL/min/1.73 m2) | ||||
| ≥90 | Referent | Referent | ||
| 60-89 | 0.36 (0.21 – 0.65) | <0.001 | 0.43 (0.23 – 0.78) | 0.005 |
| <60 | 0.27 (0.10 – 0.73) | 0.010 | 0.22 (0.07 – 0.66) | 0.007 |
| High systolic blood pressure a | 1.82 (1.07 – 3.09) | 0.028 | 1.77 (1.01 – 3.12) | 0.047 |
| High urine protein-to-creatinine ratio b | 2.74 (1.40 – 5.33) | 0.003 | 2.07 (1.00 – 4.29) | 0.051 |
| Detectable anti-dsDNA | 1.84 (1.10 – 3.07) | 0.020 | 0.98 (0.52 – 1.86) | 0.963 |
| SLEDAI without renal variables | 1.13 (1.05 – 1.23) | 0.002 | 1.02 (0.92 – 1.14) | 0.699 |
| Low complement component 3 | 1.56 (0.88 – 2.75) | 0.126 | 0.83 (0.42 – 1.66) | 0.602 |
| Low complement component 4 | 2.53 (1.49 – 4.30) | <0.001 | 1.96 (1.02 – 3.77) | 0.044 |
| Low hematocritc | 2.12 (1.34 – 3.35) | 0.001 | 1.71 (1.02 – 2.86) | 0.040 |
| Use of oral prednisone | 1.93 (1.22 – 3.05) | 0.005 | 1.47 (0.90 – 2.40) | 0.125 |
| Use of hydroxychloroquine | 0.95 (0.51 – 1.76) | 0.873 | 1.09 (0.56 – 2.13) | 0.808 |
| Use of antihypertensive drugs | 0.81 (0.52 – 1.28) | 0.360 | 0.86 (0.52 – 1.43) | 0.555 |
| Use of Immunosuppressant drugs | 2.01 (1.27, 3.17) | 0.0003 | 1.61 (0.95, 2.71) | 0.077 |
High systolic blood pressure was defined as systolic blood pressure ≥140 mmHg.
High urine protein-to-creatinine ratio was defined as urine protein-to-creatinine ratio >0.5.
Low hematocrit was defined as <41% for males and <36% for females.
anti-dsDNA = anti-double stranded DNA, CI = confidence interval, eGFR=estimated glomerular filtration rate, SLEDAI=systemic lupus erythematosus disease activity index.
Table 5 shows the variables related to transition from a declining eGFR trajectory to a non-declining trajectory. African American patients (aHR 0.50, 95% CI 0.31 – 0.81, P=0.004) had a lower rate of transitioning from declining state to non-declining state than Caucasian patients. The use of prednisone (aHR 1.61, 95% CI 1.04 – 2.49, P=0.032) was associated with an improvement of eGFR from declining state to non-declining state. Compared to patients with eGFR ≥90 mL/min/1.73 m2, patients with eGFR between 60 and 89 mL/min/1.73 m2 (aHR 1.87, 95% CI 1.09 – 3.20, P=0.023) and eGFR below 60 mL/min/1.73 m2 (aHR 2.72, 95% CI 1.52 – 4.87, P<0.001) had a higher transition rate from declining state to non-declining state (Table 5). The use of hydroxychloroquine was not associated with a lower transition rate from non-declining state to declining state (aHR 1.09, 95% CI 0.56 – 2.13, P=0.808), or a higher transition rate from declining state to non-declining state (aHR 0.85, 95% CI 0.49 – 1.49, P=0.575) (Tables 4 and 5).
Table 5.
Univariate and multivariable analysis on the identified predictiors of transition from declining to non-declining state.
| Parameters | Univariate analysis | Multivariable analysis | ||
|---|---|---|---|---|
| Hazard ratio (95% CI) | P value | Adjusted hazard ratio (95% CI) | P value | |
| Age (years) | ||||
| <40 | Referent | Referent | ||
| 40-59 | 1.39 (0.85 – 2.28) | 0.192 | 1.19 (0.63 – 2.25) | 0.520 |
| ≥60 | 1.80 (1.03 – 3.13) | 0.038 | 1.19 (0.63 – 2.25) | 0.591 |
| Male gender | 0.40 (0.06 – 2.67) | 0.345 | 0.34 (0.04 – 2.59) | 0.298 |
| Race | ||||
| Caucasian | Referent | Referent | ||
| African American | 0.55 (0.36 – 0.83) | 0.004 | 0.50 (0.31 – 0.81) | 0.004 |
| Other | 0.88 (0.40 – 1.94) | 0.745 | 0.82 (0.34 – 1.99) | 0.667 |
| eGFR (mL/min/1.73 m2) | ||||
| ≥90 | Referent | Referent | ||
| 60-89 | 2.05 (1.23 – 3.43) | 0.006 | 1.87 (1.09 – 3.20) | 0.023 |
| <60 | 3.36 (1.98 – 5.73) | <0.001 | 2.72 (1.52 – 4.87) | <0.001 |
| High systolic blood pressure a | 0.72 (0.43 – 1.22) | 0.221 | 0.68 (0.40 – 1.17) | 0.167 |
| High urine protein-to-creatinine ratio b | 0.85 (0.43 – 1.69) | 0.647 | 0.70 (0.33 – 1.52) | 0.368 |
| Detectable anti-dsDNA | 0.92 (0.57 – 1.47) | 0.717 | 1.21 (0.71 – 2.07) | 0.486 |
| SLEDAI without renal variables | 0.95 (0.86 – 1.04) | 0.246 | 0.97 (0.86 – 1.10) | 0.672 |
| Low complement component 3 | 0.95 (0.57 – 1.58) | 0.836 | 0.94 (0.51 – 1.75) | 0.851 |
| Low complement component 4 | 0.93 (0.52 – 1.67) | 0.810 | 1.11 (0.55 – 2.23) | 0.772 |
| Low hematocrit c | 1.30 (0.87 – 1.93) | 0.206 | 1.54 (0.98 – 2.41) | 0.061 |
| Use of prednisone | 1.55 (1.04 – 2.32) | 0.032 | 1.61 (1.04 – 2.49) | 0.032 |
| Use of hydroxychloroquine | 0.88 (0.51 – 1.50) | 0.631 | 0.85 (0.49 – 1.49) | 0.575 |
| Use of antihypertensive drugs | 1.82 (1.10 – 3.00) | 0.020 | 1.37 (0.81 – 2.32) | 0.234 |
| Use of immunosuppressants | 1.02 (0.68 – 1.52) | 0.958 | 0.75 (0.48 – 1.17) | 0.203 |
High systolic blood pressure was defined as systolic blood pressure ≥140 mmHg.
High urine protein-to-creatinine ratio was defined as urine protein-to-creatinine ratio >0.5.
Low hematocrit was defined as <41% for males and <36% for females.
anti-dsDNA = anti-double stranded DNA, CI = confidence interval, eGFR=estimated glomerular filtration rate, SLEDAI=systemic lupus erythematosus disease activity index.
During cohort participation, 141 of the patients had one or more renal biopsies of whom 129 were diagnosed with lupus nephritis. Among those with lupus nephritis 58 (45%) experienced at least one declining state, compared to only 112/365 (31%) among those without lupus nephritis (p=0.0033). Lupus nephritis class was defined based on a patient’s first biopsy that was positive for lupus nephritis and some were had more than one class. Among those with Class II lupus nephritis, 10/34 (29%) experienced a declining state during follow-up. In contrast, the proportion with a declining state was 21/38 (55%), 16/42 (38%), 27/50 (54%), and 0/1 (0%) for classes III, IV, V, and VI respectively.
DISCUSSION
This cohort study examined the long-term eGFR trajectory in patients with SLE, evaluated the time spent in each types of trajectory and the frequency of changes of trajectory, and identified potential predictors on the change of eGFR trajectory. The majority of patients showed a relatively stable renal function over time without rapid decline of renal function, whereas a significant proportion of patients experienced a change of eGFR trajectory. In those changes, deterioration of renal function, i.e. from non-declining to declining trajectory, was more common than improvement in renal function. If patients started at a declining trajectory, they spent on average 6.0, 3.2, and 0.8 years in declining, stable, and increasing state, respectively, in ten years. High blood pressure, low complement C4 and low hematocrit were correlated with a deterioration in eGFR trajectory. High urine protein-to-creatinine ratio also tended to be associated with a deterioration in eGFR trajectory. Moreover, African American patients were less likely to have an improvement in eGFR trajectory, while the use of prednisone was associated with an improvement in eGFR trajectory.
Unlike previous studies [27], our analysis describes declines in GFR of all patients, not just those diagnosed with lupus nephritis. We found that even among those without recognized nephritis, 31% of the patients experienced a period of declining GFR. This finding, and predictors of that experience would be lost if we based the analysis on simply those with recognized nephritis.
Proteinuria is one of the common features in renal manifestations of SLE.[8] Urine protein-to-creatinine ratio is currently a recommended measurement for proteinuria,[16] and has been validated in patients with LN.[17, 18] Our findings showed that SLE patients with high urine protein-to-creatinine ratio tended to move from a non-declining eGFR trajectory to a declining eGFR trajectory in the future. Thus, urine protein-to-creatinine ratio can be a potential predictor of deterioration of renal function and ESRD on top of a measurement for proteinuria in patients with SLE.
Previous studies found that African American patients with SLE typically have a poorer renal function when compared to Caucasian patients. LN is more prevalent among African American patients than among Caucasian patients.[4] African American patients with LN are also more likely to progress to end-stage renal disease.[19] Similarly, our findings showed that African-American patients were less likely to stabilize their renal function and thus more likely to remain a declining eGFR trajectory.
Low serum complement C3 and C4 levels is one of the hallmarks of SLE patients, particularly in those with LN.[20] Previous studies suggested that a decrease in complement levels is associated with an increase in renal SLE activity.[21] A reduced level of serum component C3 also correlates with renal histologic deterioration and is hence a good biomarker for the severity of LN, whereas serum complement C4 level did not correlate with renal histology.[22, 23] Although low C3 level is suggested to be a more important biomarker for actual renal damage, a low C4 level may predict a subsequent lupus renal flare which defined by a significant increase in serum creatinine and urine protein-to-creatinine ratio.[23] Our results showed that a low C4 level instead of C3 level was associated with a higher rate of change from non-declining trajectory to declining trajectory. This may suggest a relationship between low C4 level and a future reduction in renal function.
Prednisone is currently one of the first-line treatments for different organ system manifestations of SLE including LN.[24] It exerts both anti-inflammatory and immunosuppressive actions to SLE. Our results showed that the use of prednisone was associated with an improvement of patients from a declining eGFR trajectory to a non-declining trajectory. However, long-term treatment of prednisone may be associated with multiple irreversible side effects, particularly in patients with high-dose prednisone exposure.[25, 26]
In an earlier analysis, Hanly et al. examined the changes in renal function including eGFR and proteinuria in patients with LN using a multistate model approach.[27] They focused on patients with LN and modelled their eGFR levels as categories. They also identified predictors that were associated with moving from a category of higher eGFR level to a category of lower eGFR level and vice versa. Unlike their approach, we characterized the common longitudinal eGFR trajectory of patients with SLE by piecewise linear regression with changepoints and modelled the change in eGFR trajectory instead of eGFR level itself. We also identified predictors that were correlated with change in trajectory of eGFR instead of the change in value of eGFR. Our work can contribute on top of the work by Hanly et al. to visualize the longitudinal trajectory of renal function in SLE patients and examine the association between predictors and the change of trajectory over time.
Regarding the predictors of transitions between different eGFR levels, Hanly et al. included the patient demographics and the use of lupus treatment in their main analysis; steroid use was not included in the analysis.[27] They suggested that male gender predicts improvement in eGFR, whereas older age and African American as compare to Caucasian are associated with deterioration. Interestingly, they showed that Asians had both a higher rate of improvement and deterioration of eGFR. Age at diagnosis also showed contradictory result before and after standardization. In a smaller subgroup of patients with complete data, Hanly et al. also examined the impact of proteinuria, baseline anticardiolipin antibody and renal biopsy chronicity scores on the change of state of eGFR level. They showed that higher estimated proteinuria state was associated with deterioration and less improvement in eGFR level. In contrast, we included different clinical parameters in our study. Besides patients’ demographics, we also included high blood pressure and the use of antihypertensive medication, low hematocrit, low serum complement C3 and C4 levels, SLEDAI without renal variables, the presence of anti-dsDNA and the use of prednisone. We showed a similar result in African American patients and proteinuria as reflected by high urine protein-to-creatinine ratio. Moreover, we showed the impact of other important covariates including low serum complement C4 levels, low hematocrit as well as the use of prednisone on the change in eGFR trajectory. This may facilitate the management on the renal function of SLE patients by monitoring their serum and urinary markers.
The strength of our study includes a large sample size and a regular three-month interval of measurement of clinical parameters and drug information. We also examined the impact of time-varying clinical parameters and time-varying use of medications on the change of eGFR trajectory. Data from real-life cohorts represent a wider spectrum of patients than those in randomized controlled trials, in which patients at multiple co-morbidities are often excluded. Findings from real-life cohorts are thus more readily applicable to routine clinical practice. Also, stringent exclusion criteria are adopted to minimize bias. Nonetheless, our study has a few limitations. First, missing data and drop-out may lead to biases as in other retrospective studies. Patients who leave the cohort might be either sicker (and less able to come to cohort visits), or healthier (and less interested in coming to cohort visits). Second, the definition of state of eGFR trajectory depends on the cut-off of the slope of line segments. Third, the selection of treatment of SLE depends on clinician’s recommendation instead of study protocol. The impact of medications may be confounded by the rationale of clinician’s recommendation.
In conclusion, this cohort study characterizes the common trajectory of renal function over time in patients with SLE and estimate the frequency of change in trajectory. Patients with high blood pressure, low complement C4, low haematocrit, and probably high urine protein-to-creatinine ratio are more likely to have a declining eGFR trajectory. African American patients were less likely to improve from a declining eGFR trajectory than Caucasian patients, while the use of prednisone stabilizes the declining eGFR trajectory.
Acknowledgments
Grant support: This work was supported by NIH RO1 AR069572, and by NSF award #1418590.
Disclosures:
Michelle Petri
Terry Yip has served as a speaker for Gilead Sciences.
Dr. Saria has grants from: Gordon and Betty Moore Foundation, the National Science Foundation, the National Institutes of Health, the Defense Advanced Research Projects Agency (DARPA), and the American Heart Association. Dr. Saria is a founder of and holds equity in Bayesian Health. This arrangement has been reviewed and approved by the Johns Hopkins University in accordance with its conflict of interest policies. She is the scientific advisory board member for PatientPing. She has received honoraria for talks from a number of biotechnology, research, and health tech companies.
Abbreviations:
- aHR
adjusted hazard ratio
- C3
component 3
- C4
component 4
- CI
confidence interval
- eGFR
estimated glomerular filtration rate
- ESRD
end-stage renal disease
- IQR
interquartile range
- LN
lupus nephritis
- SLE
systemic lupus erythematosus
- SLEDAI
systemic lupus erythematosus disease activity index
REFERENCES
- 1.Almaani S, Meara A, Rovin BH. Update on Lupus Nephritis. Clin J Am Soc Nephrol 2017;12(5):825–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Alarcon GS. Multiethnic lupus cohorts: what have they taught us? Reumatol Clin 2011;7(1):3–6. [DOI] [PubMed] [Google Scholar]
- 3.Costenbader KH, Desai A, Alarcon GS, et al. Trends in the incidence, demographics, and outcomes of end-stage renal disease due to lupus nephritis in the US from 1995 to 2006. Arthritis Rheum 2011;63(6):1681–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Richman IB, Taylor KE, Chung SA, et al. European genetic ancestry is associated with a decreased risk of lupus nephritis. Arthritis Rheum 2012;64(10):3374–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Freedman BI, Wilson CH, Spray BJ, Tuttle AB, Olorenshaw IM, Kammer GM. Familial clustering of end-stage renal disease in blacks with lupus nephritis. Am J Kidney Dis 1997;29(5):729–32. [DOI] [PubMed] [Google Scholar]
- 6.Steinberg AD, Steinberg SC. Long-term preservation of renal function in patients with lupus nephritis receiving treatment that includes cyclophosphamide versus those treated with prednisone only. Arthritis Rheum 1991;34(8):945–50. [DOI] [PubMed] [Google Scholar]
- 7.Austin HA. Clinical evaluation and monitoring of lupus kidney disease. Lupus 1998;7(9):618–21. [DOI] [PubMed] [Google Scholar]
- 8.Cameron JS. Lupus nephritis. J Am Soc Nephrol 1999;10(2):413–24. [DOI] [PubMed] [Google Scholar]
- 9.Hochberg MC. Updating the American College of Rheumatology revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum 1997;40(9):1725. [DOI] [PubMed] [Google Scholar]
- 10.Petri M, Orbai AM, Alarcon GS, et al. Derivation and validation of the Systemic Lupus International Collaborating Clinics classification criteria for systemic lupus erythematosus. Arthritis Rheum 2012;64(8):2677–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Petri M, Kim MY, Kalunian KC, et al. Combined oral contraceptives in women with systemic lupus erythematosus. N Engl J Med 2005;353(24):2550–8. [DOI] [PubMed] [Google Scholar]
- 12.Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009;150(9):604–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Schulam P, Wigley F, Saria S. Clustering Longitudinal Clinical Marker Trajectories from Electronic Health Data: Applications to Phenotyping and Endotype Discovery. Proceedings of the Twenty-Ninth Aaai Conference on Artificial Intelligence 2015:2956–64. [Google Scholar]
- 14.Lerman PM. Fitting Segmented Regression Models by Grid Search. Appl. Statist 1980;29:77–84. [Google Scholar]
- 15.Marshall G, Jones RH. Multi-state Markov models and diabetic retinopathy. Stat Med 1995;14(18):1975–83. [DOI] [PubMed] [Google Scholar]
- 16.Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group. KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl. 2013;3(1): 1–150. [Google Scholar]
- 17.Christopher-Stine L, Petri M, Astor BC, Fine D. Urine protein-to-creatinine ratio is a reliable measure of proteinuria in lupus nephritis. J Rheumatol 2004;31(8):1557–9. [PubMed] [Google Scholar]
- 18.Leung YY, Szeto CC, Tam LS, et al. Urine protein-to-creatinine ratio in an untimed urine collection is a reliable measure of proteinuria in lupus nephritis. Rheumatology (Oxford) 2007;46(4):649–52. [DOI] [PubMed] [Google Scholar]
- 19.Freedman BI, Langefeld CD, Andringa KK, et al. End-stage renal disease in African Americans with lupus nephritis is associated with APOL1. Arthritis Rheumatol 2014;66(2):390–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Tsokos GC. Exploring complement activation to develop biomarkers for systemic lupus erythematosus. Arthritis Rheum 2004;50(11):3404–7. [DOI] [PubMed] [Google Scholar]
- 21.Ho A, Barr SG, Magder LS, Petri M. A decrease in complement is associated with increased renal and hematologic activity in patients with systemic lupus erythematosus. Arthritis Rheum 2001;44(10):2350–7. [DOI] [PubMed] [Google Scholar]
- 22.Garin EH, Donnelly WH, Shulman ST, et al. The significance of serial measurements of serum complement C3 and C4 components and DNA binding capacity in patients with lupus nephritis. Clin Nephrol 1979;12(4):148–55. [PubMed] [Google Scholar]
- 23.Birmingham DJ, Irshaid F, Nagaraja HN, et al. The complex nature of serum C3 and C4 as biomarkers of lupus renal flare. Lupus 2010;19(11):1272–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hahn BH, McMahon MA, Wilkinson A, et al. American College of Rheumatology guidelines for screening, treatment, and management of lupus nephritis. Arthritis Care Res (Hoboken) 2012;64(6):797–808. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zonana-Nacach A, Barr SG, Magder LS, Petri M. Damage in systemic lupus erythematosus and its association with corticosteroids. Arthritis Rheum 2000;43(8): 1801–8. [DOI] [PubMed] [Google Scholar]
- 26.Thamer M, Hernan MA, Zhang Y, Cotter D, Petri M. Prednisone, lupus activity, and permanent organ damage. J Rheumatol 2009;36(3):560–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Hanly JG, Su L, Urowitz MB, et al. A Longitudinal Analysis of Outcomes of Lupus Nephritis in an International Inception Cohort Using a Multistate Model Approach. Arthritis Rheumatol 2016;68(8):1932–44. [DOI] [PMC free article] [PubMed] [Google Scholar]