Primary membranous nephropathy (pMN) is the leading cause of nephrotic syndrome in non-diabetic adults.1 Rituximab (RTX) is now established as a first-line immunosuppressive therapy for patients at moderate or high risk of progressive disease.2 Approximately 30–40% of patients, however, fail to achieve remission.3 Interindividual variability in RTX pharmacokinetics, is a major contributor to treatment failure, driven largely by the degree of proteinuria4; hypoalbuminaemia, body weight, and anti-drug antibody formation are additional established covariates.5 The mean serum half-life of RTX in MN is approximately 11.5 days—roughly half that observed in non-proteinuric autoimmune conditions—and month-3 residual drug levels are dramatically lower than in patients without proteinuria.6 Urinary loss of intact RTX through the damaged glomerular filtration barrier has been identified as the principal mechanism of this accelerated clearance.7

Serum RTX concentration has accordingly been investigated as a predictor of treatment outcome. Teisseyre and colleagues reported that undetectable serum rituximab (<2μg/mL) at month 3 independently predicted failure to achieve remission in a 68-patient prospective cohort.8 However, a three-month delay in identifying treatment failure prolongs exposure to the thromboembolic, infectious, and metabolic complications of persistent nephrotic syndrome, and postpones any treatment modification. The predictive value of serum RTX at earlier timepoints in pMN has not been studied.

In this prospective cohort study, our primary objective was to determine whether serum rituximab measured on day 15, immediately before the second infusion, predicts composite clinical response (complete or partial remission) at month 6 in moderate- and high-risk pMN. Secondary objectives were to assess the association of day-15 levels with response at months 1 and 3, the relationship of day-105 levels and CD19+ B-cell depletion with month-6 response, and to compare rituximab exposure between monotherapy and combination (calcineurin inhibitor plus corticosteroid) regimens.

We hypothesised that serum rituximab measured on day 15, immediately before the second infusion, predicts composite clinical response at month 6 in moderate- and high-risk pMN—a hypothesis this study supports. Building on the observation that undetectable month-3 serum RTX independently predicts treatment failure, we advance this monitoring window by approximately 10 weeks—to our knowledge the earliest timepoint at which serum RTX has been shown to carry prognostic value in pMN.8

Bootstrap resampling supported the 56μg/mL operating threshold: independent derivation rules converged on 52–58μg/mL, and diagnostic performance at 56μg/mL was maintained across resamples, while the continuous exposure–response relationship was strong and internally validated (optimism-corrected area under the curve 0.82). Defining the precise decision limit within the 50–60μg/mL range, and confirming its transferability across assay platforms, is the task of larger externally validated cohorts using standardised assays; our data indicate that a clinically actionable pharmacokinetic signal is measurable as early as day 15.

The biological plausibility of day 15 as a predictive timepoint is supported by the pharmacokinetic profile of RTX in pMN. Fogueri et al. reported a mean serum half-life of approximately 11.5 days—roughly half that observed in non-proteinuric conditions.6 Larrosa-García and colleagues confirmed a similar half-life and identified proteinuria>2.4g/day as a threshold for significantly accelerated clearance,11 while Boyer-Suavet and colleagues provided direct clinical evidence that month-3 residual RTX levels in pMN are approximately sixfold lower than in non-proteinuric patients receiving an identical regimen.12 Day 15 therefore corresponds to approximately one elimination half-life after the first infusion, making it a physiologically rational timepoint at which underexposure first becomes detectable.

The mechanism driving this underexposure was rigorously characterised by Allinovi and colleagues in a prospective Florentine cohort, who demonstrated that urinary loss of intact RTX through non-selective proteinuria directly reduces serum peak levels and predicts treatment failure—a commendable methodological contribution that nonetheless requires laboratory expertise not readily available outside specialist centres.7 Broader implementation of RTX monitoring faces additional practical barriers: LISA-TRACKER, employed by both Teisseyre8 and Destere,13 requires dedicated automated instrumentation and carries higher per-test costs.14 Inter-assay variability compounds this—Truffot and colleagues showed that the Promonitor kit underestimates RTX concentrations by a mean of 69% relative to LC-MS/MS, whereas LISA-TRACKER shows only 4% bias.15 The IDKmonitor kit employed in the present study has not been evaluated in such head-to-head analyses, and our 56μg/mL threshold may not be directly transferable across platforms. Nevertheless, measuring serum RTX at day 15 with a widely available ELISA on a conventional microplate reader represents a pragmatic alternative that could extend pharmacokinetic monitoring beyond the handful of centres currently equipped for it.

Despite more severe disease at baseline—lower albumin, higher proteinuria, and a greater proportion of high-risk classification—patients receiving combination therapy achieved day-15 and day-105 rituximab concentrations comparable to those on monotherapy. This was unexpected, as hypoalbuminaemia and heavy proteinuria are established drivers of accelerated RTX clearance.6,16 A pharmacokinetically plausible explanation is that CNI-mediated podocyte stabilisation17 may attenuate urinary rituximab loss and thereby help preserve systemic exposure, although causal inference is not possible in this observational dataset. This would be consistent with the striking remission rates reported by Waldman and colleagues with concurrent RTX and cyclosporine—92% CR+PR at 9 months18—and with subsequent data showing that CNI addition to RTX in refractory disease improves outcomes.19,20 To our knowledge, no prior study has directly compared serum RTX levels between monotherapy and combination therapy in pMN; this finding is hypothesis-generating and requires confirmation in a prospective, adequately powered cohort.

It should be noted that proteinuria is not the sole determinant of rituximab pharmacokinetics. In addition to several measurable clinical factors, urinary rituximab loss has been documented even in patients with selective proteinuria, and FcRn polymorphisms represent a further source of inter-individual variability.7,8,12 Hartinger and colleagues, through population pharmacokinetic modelling, identified body weight, serum albumin, disease type, treatment duration, and anti-drug antibody formation as independent covariates of rituximab exposure, and argued that dosing strategies should differ across glomerular diseases including FSGS, SLE, and MN.5 In the present cohort, the major measurable clinical determinants of rituximab clearance were largely comparable between responders and non-responders, yet day-15 serum RTX concentrations differed substantially—further supporting the value of direct pharmacokinetic measurement over any single clinical surrogate.

Established prediction models in pMN rely on parameters assessed at month 3 or later. Beck et al. showed that anti-PLA2R antibody kinetics precede clinical remission by months.21 In a prospective cohort of 132 patients, Ruggenenti and colleagues showed that a 50% fall in anti-PLA2R antibody levels preceded an equivalent reduction in proteinuria by approximately 10 months.22 A secondary analysis of the MENTOR trial reported a C-statistic of 0.93 when combining 3-month anti-PLA2R and albumin changes.23 Day-15 RTX measurement may offer a prognostic signal at a timepoint when none of these parameters have yet had time to evolve—though validation in larger cohorts is needed before firm conclusions can be drawn.

Our day-105 data are consistent with previously published month-3 findings. Detectable RTX at day 105 was associated with significantly higher response rates compared with undetectable levels (Supplementary Table 4), consistent with earlier month-3 immunomonitoring findings.8 Day-15 measurement, however, provides a prognostic signal at a timepoint when therapeutic intervention remains most feasible.

CD19+ B-cell depletion was achieved in 95.6% of patients at day 15 but did not predict clinical response (Supplementary Table 5), consistent with the GEMRITUX trial,24 a retrospective cohort evaluating B-cell depletion thresholds in MN,25 and a prospective study of post-RTX CD19 kinetics demonstrating that depletion was near-universal within weeks regardless of subsequent clinical outcome.26 Peripheral CD19 depletion reflects circulating B-cell elimination but does not capture tissue-resident B cells, long-lived plasma cells, or other effector mechanisms relevant to pMN pathogenesis.27 KDIGO 2021 does not endorse CD19 as a primary monitoring tool, reserving it as one factor in the evaluation of treatment failure.2

We propose a hypothetical, stepwise algorithm for the management of day-15 underexposure, summarised in Fig. 3 and outlined below. A day-15 RTX level below the exploratory threshold identifies insufficient drug exposure at a timepoint when therapeutic modification remains feasible. Additional RTX dosing may suffice in patients whose low levels reflect urinary loss—an approach supported by Allinovi and colleagues, in whom re-treatment of underexposed non-responders yielded clinical response in five of eight patients.7 CNI addition could simultaneously reduce proteinuria and preserve drug exposure, as our combination group data suggest and as the high remission rates with concurrent RTX-cyclosporine reported by Waldman and colleagues support.18 For patients in whom anti-drug immunity rather than urinary loss underlies treatment failure, agent substitution is preferable to dose escalation. Obinutuzumab has demonstrated markedly higher response rates than rituximab in refractory pMN,28–30 and its humanised structure may additionally overcome pre-existing anti-RTX immunity, which accumulates progressively with repeated courses.29,31–33 Obinutuzumab was not evaluated in the present study; the cited data support this management rationale rather than representing outcomes of our cohort. Because anti-drug antibodies were not measured, this algorithm cannot yet distinguish urinary loss from immunogenicity as the mechanism underlying day-15 underexposure. Accordingly, it cannot be used to infer the cause of underexposure in individual non-responders. Clinical implementation would require external validation in independent cohorts, together with cross-platform assay comparison. At present, the data support risk stratification only and do not justify protocolised dose adjustments.

Fig. 3.

Proposed stepwise treatment algorithm for moderate- or high-risk primary membranous nephropathy based on day-15 rituximab levels. ADA: anti-drug antibody; CNI: calcineurin inhibitor; pMN: primary membranous nephropathy; RTX: rituximab.

This study has several limitations. No formal a priori sample-size or power calculation was performed; this is an exploratory single-centre cohort of 46 patients with 14 non-responders, which constrains multivariable modelling and external validation of the 56μg/mL threshold, derived from tertile distribution and regarded as exploratory, and residual confounding cannot be excluded. RTX was measured at only two timepoints (days 15 and 105), leaving intermediate kinetics undefined; tacrolimus and cyclosporine were not analysed separately, and the IDKmonitor ELISA has not been evaluated against LISA-TRACKER or LC-MS/MS, so the threshold may not transfer directly across platforms. Anti-drug antibodies were not measured, although 31 of 46 patients (67%) were rituximab-naïve, in whom pre-existing anti-rituximab antibodies would not be expected. Nevertheless, de novo anti-drug antibodies may have developed during follow-up, but this was not characterised in the present cohort. Complete and partial remission were analysed as a composite endpoint owing to limited event count, and follow-up was limited to 6 months, leaving the predictive value of day-15 levels for longer-term outcomes unknown.

In conclusion, day-15 serum RTX concentration predicts clinical response at month 6 in moderate- and high-risk pMN, offering the earliest pharmacokinetic signal yet described in this disease. The 56μg/mL threshold identified in this study should be considered an exploratory operating value for hypothesis generation and internal risk stratification. It should not be interpreted as a decision threshold ready for clinical application. If confirmed in multicentre cohorts, this measurement could enable a precision medicine approach to pMN, individualising treatment intensity from the outset rather than correcting course only after failure.

The study was approved by the Ankara Etlik City Hospital Ethics Committee (Decision No: AEÅžH-EK1-2024-0050, May 5, 2024) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.

During the preparation of this work, the authors used Claude (Anthropic) to assist with language editing and figure preparation. The authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

This work was supported by the Gazi University Scientific Research Projects Coordination Unit (BAP) [project code TTU-2025-9663].

Authors declare no conflicts of interest.

Data are available upon reasonable request from the corresponding author.