Membranous Nephropathy: Quantifying Remission Duration on Outcome

Abstract

Although change in proteinuria has been proposed as a surrogate for long-term prognosis in membranous nephropathy (MGN), variability in proteinuria levels and lag between these changes and acceptable end points, such as ESRD, has limited its utility. This cohort study examined the prognostic significance of remission duration in 376 patients with biopsy–proven idiopathic/primary MGN who achieved a remission after a period of nephrotic-range proteinuria. We defined complete remission (CR), partial remission (PR), and relapse as proteinuria ≤0.3, 0.4–3.4, and ≥3.5 g/d after CR or PR, respectively. The exposure variable was the remission status of patients at fixed landmarks (3, 6, 12, 24, and 36 months) after the date of first remission. The primary outcome was ESRD or 50% reduction in eGFR. We fitted Cox proportional hazards models to examine the association of remission status at each landmark and the primary end point. Persistent remission associated with unadjusted hazard ratios for the primary outcome that ranged by landmark from 0.35 (95% confidence interval, 0.20 to 0.61) to 0.56 (95% confidence interval, 0.31 to 1.04). Separate analyses for PR and CR yielded similar results. After adjustment, maintaining remission associated with significantly reduced risk of the primary outcome at all landmarks. Durable remissions associated with improved renal survival. Although the longer the remission, the greater the improvement, patients with remission durations as short as 3 months had improved renal prognosis compared with patients who relapsed. This study validates and quantifies PR and CR as surrogates for long-term outcome in MGN.

Membranous nephropathy (MGN) remains one of the most common causes of adult–onset nephrotic syndrome.^1–3 The natural history of MGN has been described on multiple occasions over several decades^4–10 but is still roughly divided into thirds: one third spontaneous remission, one third persistent proteinuria, and one third progressing to kidney failure.^9–11 The prognosis is also classified by proteinuria (i.e., non-nephrotic versus nephrotic) and changes in proteinuria (i.e., complete remission [CR], partial remission [PR], and no remission).^4–11 A relapse after remission is common in MGN and has complicated the ability to interpret the long-term benefits of remission.¹²

CR is associated with excellent long–term kidney survival.^13,14 A recent literature review of surrogate end points in MGN concluded that CR, but not PR, could potentially be used as an outcome measure for clinical trials.¹⁵ There is also evidence to support the long-term benefit of persistent non–nephrotic proteinuria with kidney survival of >80%–90% at 10 years. Treatment guidelines for MGN recommend only conservative treatment for these patients.^16,17

In contrast, patients with persistent high–grade nephrotic–range proteinuria have a rate of deterioration to ESRD that is proportional to proteinuria severity and duration.^17–19 The cumulative incidence of ESRD/CKD in this latter cohort is between 50% and 60% within 10 years of presentation.^{7,9,17,19–21}

Major controversies in the management of MGN include the timing of immunosuppressive (IS) therapy, what to use, how long to treat, and the response targets of treatment.^17,21 Recent MGN guidelines focused on treatment for this clinical phenotype with the referenced end points including reduction in proteinuria, modification of the ESRD rate, a doubling of serum creatinine, or 50% decline in eGFR. When only the hard end points are used, evidence to support IS therapy is largely lacking.²² This has produced a major divide between the surrogate outcome measures of change in proteinuria (i.e., CR and PR) used by the medical community, and the hard outcomes required by health regulatory agencies.^23,24 The validity of proteinuria reduction is further complicated by how it is induced (spontaneously versus treatment), follow-up duration, and relapse given that the latter can be as high as 50%.^{11,13,17,25–29} These limitations all contribute to the reluctance to use proteinuria as a bona fide surrogate end point.

Studies have shown that benefits accrue in kidney survival proportionate to the amount of time spent with lower levels of proteinuria. However, these studies are retrospective and descriptive and cannot be used to predict outcome in any individual patient.^{17–20,30,31} Quantifying the effect of duration of CR and PR on kidney survival remains unclear, particularly in PR. The complexity is accentuated by the temporal lag, often measured in years, between presentation, treatment, remission, relapse, and progression to ESRD.^{4,5,8,11–13,17–22,25–31} This time lag remains a major concern of physicians, patients, and health regulatory agencies.^15,22–24

Our study objectives are (1) to examine the durability of proteinuria reduction in relationship to hard outcome measures and (2) to determine if a quantitative relationship can be established at different time points along the remission/relapse trajectory that could provide temporal guideposts for MGN management.

Results

After applying the exclusion criteria, the study cohort was comprised of 376 patients with MGN who achieved either CR or PR after a documented period of nephrotic-range proteinuria (Figure 1). Table 1 summarizes the baseline characteristics of the study cohort. The total follow-up time was 1786 person-years (median follow-up: 3.4 [1.2–6.7] years; maximum: 27 years; mean age was 49.9±15.3 years old, 63.6% were men, and 62.5% were of white ancestry). The median duration of nephrotic-range proteinuria before remission was 362 (176–781) days, with a mean peak proteinuria of 10.3±6.2 g/24 h and accompanying hypoalbuminemia in 82% of patients. At remission, the median eGFR was 69 (49–90) ml/min per 1.73 m², and the median proteinuria was 2.1±0.9 g/24 h. Spontaneous remission, as defined in Concise Methods, occurred in 196 of 376 (52.1%), including 56 of 196 (28.6%) on renin-angiotensin system blockade (RASB) therapy. Before remission, 47.9% of patients had received prednisone and/or IS medication. After remission, there were 62 patients who had a 50% reduction in kidney function, including 34 who progressed to ESRD. The cumulative probability of a CR over the first 36 months was 15%. Cumulative probabilities of relapse at 1, 2, and 5 years were 36.1%, 43.8%, and 56.5%, respectively.

Figure 1.

Composition of the final study cohort. The final study cohort was composed of all idiopathic/primary membranous nephropathy patients in the GNR excluding those that never had nephrotic range proteinuria or never remitted from nephrotic range proteinuria. GNR, Glomerulonephritis Registry.

Table 1.

Baseline patient characteristics at remission

Variables	n (%), Mean (±SD), or Median (IQR)
Demographics
Age at remission, yr	49.9 (±15.3)
Sex
Men	239 (63.6%)
Women	137 (36.4%)
Ethnicity
White	235 (62.5%)
Black	15 (4.0%)
Asian	16 (4.3%)
Other	110 (29.3%)
Laboratory measurements
Duration of nephrotic syndrome, d	362 (175.5–780.5)
GFR at remission, ml/min per 1.73 m2	69.0 (48.5–89.5)
Proteinuria at remission, g/24 h	2.1 (±0.9)
Serum albumin, g/L	34.1(±6.7)
Mean BP at remission, mmHg
Systolic	132 (±20)
Diastolic	81 (±11)
Medications
Spontaneous remissiona	196 (52.1%)
Treatment medication at remissionb	180 (47.9%)
Prednisonec or ISd (nonspontaneous)	180 (100%)
Prednisonec alone (without IS)	52 (29%)
ISd alone (without prednisone)	59 (33%)
Prednisonec and ISd	69 (38%)
Azathioprined	49 (27.2%)
Cyclophosphamided	82 (45.6%)
Cyclosporind	83 (46.1%)
MMFd	6 (3.3%)
Other ISd	3 (1.7%)

IQR, interquartile range; MMF, mycophenolate mofetil.

^aDid not receive prednisone in the 3 months before remission and no IS medication in the 9 months before remission. Of the 196 patients, 56 (28.5%) received angiotensin–converting enzyme inhibitor (ACEi) and/or angiotensin II receptor blocker (ARB) during 3 months before remission.

^bOf 180 patients, 124 (68.9%) received ACEi and ARB during 3 months before remission.

^cDuring 3 months before remission.

^dDuring 9 months before remission. This includes patients who received more than one of the IS agents during this timeframe. At the time of first remission, 135 received prednisone and/or IS medication (these patients could have received more than one IS medication): nine received azathioprine, 35 received cyclophosphamide, 42 received cyclosporin, six received MMF, and two received other IS medication. Ignoring the 3- or 9-month defined time for spontaneous remission, 278 patients had at some time received prednisone and/or one or more IS medication before their first remission: 148 had received azathioprine, 186 had received cyclophosphamide, 182 had received cyclosporin, eight had received MMF, and six had received other IS medication.

Figure 2 illustrates the risk of ESRD or 50% reduction in kidney function using Kaplan–Meier curves among patients who relapsed versus stayed in remission by the 3-, 6-, 12-, and 24-month landmarks (Figure 2). At each landmark, the cumulative probability of the primary outcome was lower in patients who remained in remission, beginning as early as the 3-month landmark and sustained to the 24-month landmark. Although the advantage of remaining in remission was maintained at each landmark, the cumulative probability of progressing to an end point declined at later landmarks in both groups. Specifically, the cumulative probabilities of the primary outcome in patients after first remission in relapse versus who stayed in remission by the 3-, 6-, 12-, and 24-month landmarks were 26.8% versus 15.2% (P=0.06), 30.4% versus 11.6% (P<0.001), 20.2% versus 8.8% (P=0.002), and 16.8% versus 5.1% (P=0.01), respectively. Similar patterns were observed when the risks of ESRD or 50% reduction in kidney function were assessed separately (Supplemental Figures 1 and 2) at all landmarks up to 24 months (Supplemental Table 1). At the 36-month landmark, there was no statistically significant difference in outcome over the 5-year follow-up between patients remaining in remission versus relapsing (P=0.30) (Supplemental Figure 3).

Figure 2.

Hard endpoints of either end-stage renal disease or 50% reduction in kidney function reached in the remission and relapse cohorts following each landmark. Using the status of each of the patient at these time points, the subsequent risk of ESRD or 50% reduction in kidney function was calculated by Kaplan–Meier methods. Final outcomes in each of these cohorts refer to their status at 5 years following each of the landmark analysis at (A) 3 months, (B) 6 months, (C) 12 months, and (D) 24 months after first remission.

The results of the univariable and multivariable Cox proportional hazard models at each landmark are displayed in Table 2. The unadjusted and adjusted models indicate that staying in remission is associated with a persistent reduction in the risk of the primary outcome. In the final adjusted model, patients who stayed in remission for at least 3 months had a hazard ratio (HR) of 0.48 (95% confidence interval, 0.25 to 0.93; P=0.03) for the primary outcome versus patients who relapsed by 3 months. Similar reductions in HRs were seen for the 6-, 12-, and 24-month landmarks The longer the duration of remission time (per 30 days) before each landmark, the less likely one is to progress to the primary outcome (Supplemental Table 2).

Table 2.

Landmark analysis of the association of relapse status with primary end point (ESRD or 50% reduction in kidney function) using Cox proportional hazard model

Relapse Status, moa	n	Model 1 HR (95% CI)b	P Value	Model 2 HR (95% CI)c	P Value	Model 3 HR (95% CI)d	P Value
3	347	0.56 (0.31 to 1.04)	0.07	0.54 (0.29 to 1.00)	0.05	0.48 (0.25 to 0.93)	0.03
6	339	0.35 (0.21 to 0.58)	<0.001	0.36 (0.21 to 0.61)	<0.001	0.35 (0.20 to 0.61)	<0.001
12	302	0.44 (0.26 to 0.75)	0.002	0.43 (0.25 to 0.73)	0.002	0.37 (0.21 to 0.67)	0.001
24	255	0.55 (0.61 to 0.99)	0.05	0.52 (0.29 to 0.94)	0.03	0.40 (0.22 to 0.77)	0.01

95% CI, 95% confidence interval.

^aCompares patients who stayed in remission with patients who relapsed.

^bModel 1 adjusted for relapse status.

^cModel 2 adjusted for model 1, age, sex, and ethnicity.

^dModel 3 adjusted for model 2, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, and GFR at remission.

Figure 3 depicts the time to the primary outcome with further subclassification of remission sustained to each landmark (i.e., relapse, CR, or PR). Table 3 shows the results of multivariable Cox proportional hazards models on the basis of the type of remission sustained to each landmark (CR versus PR). Reaching a CR usually occurs slowly and had a variable time period of PR first. This is illustrated by the increasing number of patients with CR at the various landmarks from 14 at the 3-month landmark to 39 at the 24-month landmark. The cumulative probability of patients with CR reaching the primary outcome was significantly lower than in either the PR or relapse cohort from the 3-month to the 24-month landmarks, although the reductions in the relative hazards of the primary outcome in the models still confirm the independent value of PR (versus relapse) on kidney survival (Table 3).

Figure 3.

Time to the primary outcome of ESRD or 50% reduction in kidney function with sub-classification of remission to PR or CR sustained to each landmark. Using the status of each of the patient at these time points, the subsequent risk of ESRD or 50% reduction in kidney function stratified by PR and CR status was calculated using Kaplan–Meier methods. Final outcomes in each of these cohorts refer to their status at 5 years following each of the landmarks at (A) 3 months, (B) 6 months, (C) 12 months, and (D) 24 months after remission.

Table 3.

Landmark analysis of the association of PR and CR status with ESRD or 50% reduction in kidney function using Cox proportional hazard model

Relapse Statusa	n	Multivariable HR (95% CI)b	P Value
6 Mo
Stayed in remission (PR)	339	0.34 (0.20 to 0.60)	<0.001
Stayed in remission (CR)		0.19 (0.04 to 0.83)	0.03
12 Mo
Stayed in remission (PR)	302	0.40 (0.22 to 0.73)	0.003
Stayed in remission (CR)		0.19 (0.04 to 0.85)	0.03
24 Mo
Stayed in remission (PR)	255	0.46 (0.24 to 0.88)	0.02
Stayed in remission (CR)		0.24 (0.07 to 0.83)	0.03

95% CI, 95% confidence interval.

^aReference group: stayed in relapse.

^bMultivariable model adjusted for relapse status, age, sex, ethnicity, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, and eGFR at remission.

Figure 4 illustrates the subgroup analysis using the 6-month landmark to fit the multivariable Cox proportional hazard models. There were no statistically significant interactions between relapse status and age, sex, spontaneous versus drug-induced remission, or level of kidney function (on the basis of mean eGFR during the 6 months after remission onset). Similarly, using the same method, no significant interactions were shown at other landmarks (Supplemental Table 3).

Figure 4.

Interactions between relapse status and age, sex, spontaneous versus drug-induced remission, and level of kidney function. Subgroup analysis of the association of relapse status with ESRD or 50% reduction in kidney function at 6 months after first remission using Cox proportional hazard model and landmark analysis. ACEi, angiotensin–converting enzyme inhibitor; ARB, angiotensin II receptor blocker; 95% CI, 95% confidence interval.

Table 4 displays results from the sensitivity analyses using the Cox proportional hazard model at each landmark. When the final model was further adjusted for the duration of nephrotic syndrome before remission (sensitivity analysis 1), peak GFR during remission (sensitivity analysis 2), mean diastolic BP during remission (sensitivity analysis 3), spontaneous versus drug-induced remission (sensitivity analysis 4), any history of prednisone and/or IS before remission (sensitivity analysis 5), or never or ever hypoalbuminemic (Supplemental Table 4), patients who stayed in remission had a decreased relative hazard for the primary outcome at all landmarks. When death was included as part of the composite outcome, the HRs at each landmark remained unchanged. We repeated this analysis using the dual definition of remission used in our original paper on PR in MGN (i.e., both a 50% reduction in peak proteinuria and reaching subnephrotic level).³⁰ This changed the total number to 307 patients but neither the HRs nor the significance of the benefit of staying in remission at all landmarks (Table 5).

Table 4.

Sensitivity analysis of the association of relapse status with the primary end point (ESRD and 50% reduction in kidney function) using Cox proportional hazard model and landmark analysis

Relapse Status, moa	n	Sensitivity Analysis 1 HR (95% CI)b	P Value	Sensitivity Analysis 2 HR (95% CI)c	P Value	Sensitivity Analysis 3 HR (95% CI)d	P Value	Sensitivity Analysis 4 HR (95% CI)e	P Value	Sensitivity Analysis 5 HR (95% CI)f	P Value
3	347	0.48 (0.25 to 0.92)	0.03	0.44 (0.23 to 0.85)	0.02	0.49 (0.26 to 0.95)	0.04	0.45 (0.23 to 0.90)	0.02	0.53 (0.28 to 1.01)	0.06
6	339	0.36 (0.21 to 0.62)	<0.001	0.33 (0.19 to 0.57)	<0.001	0.35 (0.95 to 0.99)	<0.001	0.38 (0.22 to 0.68)	0.001	0.34 (0.20 to 0.59)	<0.001
12	302	0.38 (0.21 to 0.68)	0.001	0.35 (0.19 to 0.62)	<0.001	0.38 (0.21 to 0.68)	0.001	0.40 (0.22 to 0.72)	0.002	0.37 (0.21 to 0.65)	0.001
24	255	0.39 (0.20 to 0.73)	0.004	0.35 (0.18 to 0.66)	0.001	0.41 (0.21 to 0.79)	0.01	0.42 (0.22 to 0.82)	0.01	0.46 (0.25 to 0.86)	0.02

95% CI, 95% confidence interval.

^aCompares patients who stayed in remission with patients who relapsed.

^bSensitivity analysis 1 included relapse status, age, sex, ethnicity, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, mean GFR during remission, and duration of nephrotic syndrome.

^cSensitivity analysis 2 included relapse status, age, sex, ethnicity, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, and peak GFR during remission.

^dSensitivity analysis 3 included relapse status, age, sex, ethnicity, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, mean GFR during remission, and mean diastolic BP during remission.

^eSensitivity analysis 4 included relapse status, age, sex, ethnicity, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, mean GFR during remission, and spontaneous remission.

^fSensivity analysis 5 included relapse status, age, sex, ethnicity, any history of prednisone before remission, any history of IS medications before remission, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, and mean GFR during remission.

Table 5.

Landmark analysis of the association of relapse status with primary end point (ESRD or 50% reduction in kidney function) using Cox proportional hazard model in patients who also had 50% reduction from peak proteinuria

Relapse Status, moa	n	Model 1 HR (95% CI)b	P Value	Model 2 HR (95% CI)c	P Value	Model 3 HR (95% CI)d	P Value
3	307	0.48 (0.25 to 0.92)	0.03	0.45 (0.24 to 0.87)	0.02	0.41 (0.20 to 0.82)	0.01
6	300	0.31 (0.18 to 0.55)	<0.001	0.32 (0.18 to 0.56)	<0.001	0.31 (0.17 to 0.56)	<0.001
12	267	0.42 (0.24 to 0.74)	0.003	0.39 (0.22 to 0.70)	0.002	0.39 (0.20 to 0.73)	0.003
24	224	0.47 (0.25 to 0.89)	0.02	0.42 (0.22 to 0.80)	0.01	0.33 (0.16 to 0.67)	0.002

95% CI, 95% confidence interval.

^aCompares patients who stayed in remission with patients who relapsed.

^bModel 1 adjusted for relapse status.

^cModel 2 adjusted for model 1, age, sex, and ethnicity.

^dModel 3 adjusted for model 2, prednisone, IS medications, antihypertensives, angiotensin–converting enzyme inhibitor, angiotensin II receptor blocker, and GFR at remission.

Discussion

Our findings suggest that, regardless of age, baseline eGFR, spontaneous or drug-induced remission at each landmark, and whether patients with MGN remain in PR or CR, their long–term kidney outcome steadily improves as a consequence of maintaining a remission of proteinuria.

The management of the patient with MGN has evolved over the last two decades. The recent Kidney Disease Improving Global Outcomes (KDIGO) GN guidelines summarize the current evidence and strength of the recommendation for treatment.¹⁷ These guidelines used all hard outcomes but also, included changes in proteinuria. In contrast, a recent Cochrane systematic review focused only on acceptable hard end points defined by health agencies given the uncertainty surrounding proteinuria changes.²² This variance in end points results in major differences in the treatment recommendations.

Although there is literature indicating that long-term benefit accrues after remission of proteinuria in MGN, these data have been derived only at the end of the observation period. Moreover, quantitating the value of remission prospectively, the prognostic relevance of how remission is achieved, the effect of the starting GFR, and the importance of CR or PR remain controversial or unknown. These issues have made the use of proteinuria as a surrogate outcome unacceptable to health regulatory agencies, although regularly used by physicians. This dichotomy creates a therapeutic conundrum on how to balance risk versus benefits of different treatments. Although the KDIGO GN guidelines do recommend that, under certain conditions, patients not be exposed to the risks of potent IS drugs, they are recommended in the persistent higher–grade proteinuria group, despite weak evidence by Cochrane standards. This is because the definitive hard outcomes require years to reach, well beyond the limits of standard clinical trials and a difficult time horizon for nephrologists and their patients to interpret. This means that even the randomized, controlled trials that have been done are of insufficient duration to directly assess these benefits.

Even the current literature supporting a benefit to proteinuria remission is limited by problems with data completeness/quality, crossover of patients to alternative therapies, and the inherent difficulty of continuing to monitor patients who have had a remission of proteinuria.^{13–15,23,24} We examined this gap in evidence by looking at long-term outcomes of patients with nephrotic–range idiopathic/primary MGN who achieved at least a PR. We used, as a starting point, the time of remission to the subnephrotic state and measured its durability on outcome using the landmark analysis method.³² This method is commonly used in cancer studies to assess drug efficacy, and we chose this analytic approach given the comparable remitting/relapsing course of MGN. Furthermore, the landmark analysis provides a more accessible result when the duration (e.g., remission time) of a time-dependent exposure and its relationship to a specific outcome are of primary interest.

We postulated that the association of remission status with the primary outcome of ESRD or a 50% reduction in baseline eGFR varies with remission duration and may be modified by age, sex, drug-induced versus spontaneous remission, and kidney function at the time of remission. We found a strong association between remission duration and the primary outcome beginning as early as 3 months after remission onset. This improvement was maintained even when the secondary end points, ESRD or a 50% reduction in eGFR, were examined separately. In addition, as time in remission increased, both the separation between the relapse and remission groups for primary and secondary end points narrowed and the absolute proportion of patients progressing to the primary end point was reduced, suggesting improved kidney survival, despite a subsequent relapse (Figure 2).

Examining CR versus only a PR, an advantage (especially early postremission) was seen in the CR group with regards to the primary outcome (Figure 3, Table 3). This supports the recent review that CR could be an acceptable indication for clinical trials.¹⁵ When remission duration was examined per 30-day increments, a continuous reduction in the primary end point was observed and remained essentially unchanged at each of the landmarks (Supplemental Table 1). Sensitivity analyses showed that the duration of the nephrotic syndrome before remission, peak GFR during the remission, the mean diastolic BP during remission, spontaneous versus drug induced defined as per methods, or ever receiving IS and/or prednisone therapy did not significantly modify the HR for relapse status (Table 4). The addition of death to the primary end point did not change the results. We repeated this analysis using the dual definition of remission used in our original paper (i.e., both a 50% reduction in peak proteinuria and reaching subnephrotic level). This changed neither the HRs nor the significance of the benefit of staying in remission at all landmarks (Table 5). When we broadened the definition of treatment to include all patients ever exposed to IS or prednisone, again, we found no change to either the HR or significance of staying in remission (Table 4, model 5).

There are several limitations to our study that deserve mention. First, we cannot fully account for differences in conservative management or new treatment options that have evolved over the time of the study in our remission cohort. The data collection, although done prospectively at least yearly, is only periodic and limited by the frequency of the patients’ scheduled visits to their primary physician.³³ Second, as in any clinical cohort study,³⁴ there is variability in follow-up and loss of patients over time and a certain amount of missing data. The former was handled with standard time to event methods, and the latter was addressed using multiple imputation.³⁵ Third, we excluded patients who had reached an end point or were lost to follow-up before a given landmark. This may reduce precision and generalizability. However, the numbers reaching the hard outcome were reasonable, and the total observation period of up to 8 years postremission was an appropriate timeframe to assess long-term outcomes in patients with nephrotic MGN. Fourth, we did not account for the variable definitions of PR in the literature^15,30 but used the clinically applicable end point of reaching a non-nephrotic level of proteinuria, because our main objective was to determine the benefits of the durability of this clinical state. We did include the more recent dual definition of remission,³⁰ and this did not significantly alter the results (Table 5). We believe that this strengthens the real world applicability of our results. Fifth, we did not have specific data on the dose or duration of RASB before remission. However, we did include exposure to RASB at the time of remission in our statistical models and could not show a significant difference in outcome between this exposure and either the spontaneous or drug–induced remission subcohort. Sixth, as in all observational studies, there may be some residual confounding due to unmeasured factors that may influence both the duration of remission and the likelihood of adverse renal outcomes. Seventh, the generalizability of our results may be limited to populations with characteristics that are similar to the study cohort. However, the significant number of patients and the consistent results of a steadily improving kidney survival benefit at the different landmarks strongly suggest that these data are applicable to the general MGN population.

Our findings do suggest that, regardless of age, eGFR, and spontaneous or drug-induced remission, for each month that the patient remains in PR or CR, their long–term kidney outcome steadily improves. This improvement also applies when the definition of remission is broadened to include a 50% reduction in peak proteinuria. Whether a PR or CR somehow modifies or resets the disease process, perhaps through modification of the phospholipase A2 receptor (PLA2R) autoantibody or other autoimmune phenomena,³⁶ is of interest but cannot be answered by these data. Similarly, the important question of whether the remission state is due to immunologic remission or residual kidney damage cannot be answered by these data. However, we know that up to one third of patients during their most active phase and an even higher percentage with low-level proteinuria are not associated with PLA2R, and therefore, in this sizable cohort, this information would not be definitive in determining whether remission was related to a modified immunologic status or residual injury.^36,37 It is likely, in those associated with PLA2R positivity, that antibody status (and/or change in titer) will improve the granularity of risk prediction in this cohort, but our data support, regardless of this, that remission durability will be important in assessing the patients’ long-term prognoses.

Our results suggest that the relative benefit of staying in remission is remarkably stable, even out to the 24-month landmark (i.e., HR, 0.50 or two times less likely to progress to the outcome), and that the likelihood of progressing to an end point declines in both groups. This does not mean that relapses should be ignored but does mean that the decision to start or reintroduce IS therapy need not be immediate, and this does add a new dimension to the question of when to initiate treatment in those who relapse postremission. This should allow the risks of de novo or reintroduction of IS therapy to be more precisely assessed by accounting for factors that have developed during the remission time, such as increased age, comorbid conditions, and complications related to previous treatment. These data should also help regulatory agencies to set indications and encourage the pharmaceutical industry to invest in efficacy trials in MGN given the reduction in sample size and the shortened timeframe to PR or CR compared with ESRD. These data significantly strengthen the evidence supporting the utilization of remission, both PR and CR, as bona fide surrogate outcomes for patients with MGN.

Concise Methods

This is a cohort study using data from the Toronto Glomerulonephritis Registry (TGNR).^30,33 All patients with biopsy–proven idiopathic or primary MGN from inception of the registry in the early 1970s to November of 2011 who were enrolled in the TGNR were eligible for study inclusion and followed up as described previously. Patients entered this study cohort if they achieved CR or PR (i.e., the date of study entry was the documented date of first CR or PR). Secondary causes of MGN, such as lupus nephritis, were excluded, whether they were detected at presentation or over follow-up.¹⁷ Additional exclusion criteria included patients who never had nephrotic-range proteinuria and patients who never experienced PR or CR at any time during follow-up.

We defined the nephrotic state as proteinuria ≥3.5 g/24 h as measured by 24-hour urine protein collections. CR was defined as proteinuria ≤0.3 g/24 h and PR was defined as proteinuria 0.4–3.4 g/24 h after a period of nephrotic-range proteinuria. Relapse was defined as proteinuria ≥3.5 g/24 h after a period of CR or PR.

The exposure variable of interest was the duration of CR and/or PR. The exposure was examined as a continuous (i.e., per 30 days) or binary (i.e., relapsed or stayed in remission) variable at specific time points after first remission (i.e., 3, 6, 12, 24, and 36 months). Other baseline patient and clinical characteristics (i.e., age, sex, ethnicity, duration of nephrotic syndrome before remission, BP, proteinuria, eGFR, spontaneous remission, and medications) were collected at the time of first remission. Those characteristics that varied over time were subsequently updated at 3, 6, 12, 24, and 36 months after first remission. The Modification of Diet in Renal Disease Study equation was used to calculate eGFR.³⁸ Medications recorded in the TGNR included corticosteroids (>90% as prednisone), angiotensin–converting enzyme inhibitor, angiotensin receptor blocker, other antihypertensives, diuretics, and IS medications (including azathioprine, cyclophosphamide, chlorambucil, cyclosporin, tacrolimus, mycophenolate mofetil, and rituximab). Spontaneous remission was defined as achieving remission with no exposure to corticosteroids within 3 months and no IS medications within 9 months of the first remission date.

The primary end point was the composite of ESRD (defined by persistent eGFR <15 ml/min per 1.73 m², start of chronic dialysis, or preemptive kidney transplantation) or a 50% reduction in eGFR over follow-up relative to the landmark. All patients were assessed for end points up to 5 years after each landmark. Secondary outcomes included the individual components of the primary outcome (i.e., ESRD or a 50% reduction in eGFR).

For normally distributed continuous variables, means (±SDs) and the t test were used to describe and compare differences across groups respectively. Non–normally distributed data were described using medians (interquartile ranges), and differences across groups were compared using the Wilcoxon rank sum test. For all categorical variables, frequencies/percentages and the chi-squared test were used to describe and compare differences across groups, respectively.

To ensure that the analysis properly accounted for the timing of first remission in relation to the relapse event, a landmark analysis was conducted.³² In this approach, fixed time points over follow-up are used as the starting point to estimate the survival probabilities of each group on the basis of group membership (relapse versus no relapse) at each landmark. The landmarks were 3, 6, 12, 24, and 36 months after first remission. At each landmark, only patients who continue to be at risk for the outcome (i.e., are alive and have not experienced the outcome before the landmark) are included. The Kaplan–Meier product limit method was used to assess the time to the primary end point from each landmark. Log-rank tests were used to compare the cumulative event probabilities between patients who were in remission versus relapsed by each landmark. Multivariable Cox proportional hazard models were fitted at each landmark to adjust for potential confounders. The fully adjusted models included age, sex, ethnicity, drug exposure on the basis of our definition or ever used (including prednisone and/or other IS medications), RASB exposure (angiotensin–converting enzyme inhibitor or angiotensin receptor blocker), other antihypertensive medications, systolic/diastolic BP, and the mean eGFR over the remission period. The method of multiple imputation by chained equations was used to impute missing covariates.³⁵

Subgroup analyses were performed to assess the risk of the primary outcome by relapse status across strata of age (<50 or ≥50 years old), sex (man or woman), type of remission (spontaneous or drug induced), and kidney function at remission (≤60 or >60 ml/min per 1.73 m²). Sensitivity analyses were also conducted that adjusted the statistical models for additional covariates and included death as part of the primary outcome. This analysis was repeated, separating the remission cohort into those achieving CR versus only PR. A two–tailed P value of <0.05 was considered statistically significant. Stata/SE 12.0 (StataCorp, College Station, TX) was used to perform all analyses. The research ethics board at the University Health Network approved the study.

Disclosures

None.