Efficacy and Safety of Mizoribine in the Treatment of Lupus Nephritis: A Systematic Review and Meta-Analysis

Abstract

Introduction

This study evalöuated the clinical efficacy and safety of mizoribine (MZR) in the treatment of lupus nephritis.

Methods

We conducted a systematic review and meta-analysis in accordance with PRISMA 2020 guidelines. Literature searches were performed in PubMed, Embase, Web of Science, Cochrane Library, and ClinicalTrials.gov up to June 30, 2025. Two investigators independently screened studies, extracted data, and assessed risk of bias. Included studies enrolled patients with biopsy-confirmed lupus nephritis treated with mizoribine-containing regimens, including randomized controlled trials (RCTs), cohort studies, and single-arm designs. Primary and secondary outcomes were renal response rate (The proportion of patients achieving either complete response (CR) and partial response (PR). CR: 24-h urinary protein < 0.5 g/day with normal/stable serum creatinine ≤ 25%. PR: ≥ 50% reduction in proteinuria to < 3.5 g/day with stable serum creatinine ≤ 25%), and adverse reaction rate, 24-h urinary protein, and systemic lupus erythematosus disease activity index (SLEDAI) score, respectively. Data were analyzed in STATA 14.0 using fixed or random-effects models based on heterogeneity (random effects for I² ≥ 25% and P ≤ 0.1). Prespecified subgroup analyses were performed by treatment duration (induction period ≤ 6 months vs. maintenance period > 6 months) and control regimen (glucocorticoid group vs. glucocorticoid and immunosuppressants ).

Results

Nineteen studies (four RCTs, one cohort, 14 single-arm) with 1489 patients were included. Meta-analysis of randomized controlled trials (RCTs) showed no significant differences in renal response rates or adverse reaction rates between MZR and control groups, but a higher SLEDAI score was observed with MZR (WMD = 1.75, 95% CI 0.33–3.16, P < 0.05). Subgroup analysis indicated MZR was associated with a reduced renal response rates (RR = 0.83, 95% CI 0.71–0.97, P < 0.05), elevated proteinuria (WMD = 0.62, 95% CI 0.22–1.03, P < 0.05), and higher SLEDAI scores (WMD = 1.75, 95% CI 0.33–3.16, P < 0.05) versus controls. These negative outcomes, including increased proteinuria (WMD = 0.73, 95% CI 0.11–1.36, P < 0.05) and SLEDAI (WMD = 2.60, 95% CI 1.19–4.01, P < 0.05), were significant only in the induction period use (P < 0.05) and not sustained maintenance period (> 6 months). Single-arm studies reported high response rates (59–100%) but widely variable adverse reaction rates (5–50%). The types of adverse events included hyperuricemia, respiratory tract infection, leukopenia, etc.

Conclusions

Mizoribine demonstrates phased efficacy in lupus nephritis: it is less effective than standard regimens for the induction period (≤ 6 months) but comparable during the maintenance period (> 6 months). Therefore, it is not a first-line induction agent but serves as a practical maintenance option, particularly for patients intolerant of conventional therapies or in resource-limited settings.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40744-026-00851-4.

Key Summary Points

Why carry out this study?

Lupus nephritis is a major complication leading to end-stage renal disease, while current first-line therapies are associated with significant toxicity, highlighting the urgent need for safer and more accessible treatment options.

Although mizoribine (MZR) holds theoretical advantages, its efficacy and safety in the treatment of lupus nephritis remain controversial, necessitating systematic evidence to clarify its clinical role.

This study aims to systematically evaluate the efficacy and safety of MZR-containing regimens versus conventional therapies through meta-analysis, and to explore the influence of treatment duration and control regimen type on these outcomes.

What was learned from the study?

MZR demonstrated comparable renal response rates and safety profiles to control regimens, but it was significantly inferior in controlling proteinuria and disease activity during the induction period (≤ 6 months) compared to conventional combination therapy, with these differences resolving in the maintenance phase (> 6 months).

MZR may be more suitable as a maintenance therapy option for lupus nephritis rather than as a first-line induction treatment; however, significant heterogeneity exists in the current evidence, underscoring the need for high-quality randomized controlled trials to optimize its clinical application.

Introduction

Lupus nephritis (LN) is among the most frequent and serious clinical complications of systemic lupus erythematosus (SLE), affecting 20–60% [1] of patients with SLE. Pathologically, it is characterized by immune complex-mediated glomerular inflammation and secondary structural damage [2]. Histopathological classification of LN (particularly classes III, IV, and V) critically determines disease prognosis [3–5]. Without timely standardized treatment, LN progresses to end-stage renal disease and may cause mortality [6–9].

Current management of active patients with LN guidelines relies on glucocorticoids combined with mycophenolate mofetil or low-dose intravenous (IV) cyclophosphamide as foundational therapy, with add-on biologics (e.g., belimumab) or calcineurin inhibitors (e.g., voclosporin, tacrolimus) also recommended [6, 10–12]. However, conventional immunosuppressants carry substantial toxicities, including myelosuppression, reproductive harm, and heightened infection risk with cyclophosphamide, and frequent gastrointestinal disturbances with mycophenolate mofetil [13–16]. Novel targeted biologics represent therapeutic breakthroughs. Belimumab reduces autoantibody production by inhibiting B-lymphocyte stimulator (BLyS) [17]. Sequential therapy with rituximab and belimumab achieved renal response in 76% of patients with refractory LN (n = 25) in pilot studies [18]. Obinutuzumab or voclosporin add-on therapy substantially enhances complete response rates [19] show superior efficacy in late-stage trials [20, 21]. Nevertheless, high cost and limited access confine biologics largely to refractory cases, highlighting the need for real-world evidence to guide their clinical use.

Mizoribine (MZR), a selective inhibitor of inosine monophosphate dehydrogenase, exerts immunomodulatory effects by suppressing purine synthesis and lymphocyte proliferation [22, 23]. Initially approved for renal transplant rejection prophylaxis, it has been repurposed for LN and rheumatoid arthritis. Compared to cyclophosphamide and mycophenolate mofetil, MZR offers advantages, including oral bioavailability, reduced myelotoxicity, and lower cost [24]. Its adoption outside Japan, particularly in China and Korea, is less documented. While a pivotal multicenter phase III trial in China has demonstrated its efficacy [22], comprehensive real-world data from broader Asian populations remain limited, and the drug has not yet received approval from the U.S. Food and Drug Administration (FDA) or the European Medicines Agency (EMA). Recent clinical studies have evaluated its role in the LN induction period. A multicenter phase III RCT in China demonstrated non-inferior efficacy and superior safety of 52-week MZR plus glucocorticoids versus intravenous cyclophosphamide, with fewer severe adverse events [22]. These findings position MZR as a potential alternative for patients who are intolerant to conventional immunosuppressants or in resource-limited settings.

Nevertheless, the efficacy and safety profile of MZR in LN management remains controversial. Heterogeneity in existing clinical trial designs and patient populations impedes its broader application. This systematic review and meta-analysis comprehensively evaluates renal response rates and safety outcomes of MZR versus conventional immunosuppressants. Through subgroup analyses, we further examine the impact of therapeutic regimens and treatment duration, thereby providing evidence-based reference for clinical decision-making in appropriate patient populations.

Methods

Inclusion and Exclusion Criteria

From January 1, 1966 to June 30, 2025, a systematic literature search was performed in the PubMed, Embase, Web of Science, Cochrane Library, and ClinicalTrials.gov databases, using Medical Subject Headings (MeSH) and free-text terms with core terms “lupus nephritis” and “mizoribine”. Two investigators independently conducted study screening, quality assessment, and data extraction; disagreements were resolved through discussion or third-reviewer adjudication. Extracted data included first author, publication year, region, sample size, gender, study design, intervention/comparator details (dosing, frequency, and all immunosuppressants), treatment duration (the primary endpoint timepoint was defined per the protocol) and outcome measures.

Inclusion criteria encompassed: (1) patients with biopsy-confirmed lupus nephritis; (2) interventions involving MZR-containing regimens (mono- or combination therapy); (3) primary outcome defined as renal response rate: The proportion of patients achieving either complete response (CR) and partial response (PR). CR: 24-h urinary protein < 0.5 g/day with normal/stable serum creatinine ≤ 25%. PR: ≥ 50% reduction in proteinuria to < 3.5 g/day with stable serum creatinine ≤ 25%. with secondary outcomes including adverse reaction rate, 24-h urinary protein excretion (g/24 h), and SLE Disease Activity Index (SLEDAI) score; (4) study designs limited to randomized controlled trials (RCTs), cohort studies, and single-arm interventional studies (the largest sample size, the most complete follow-up, lowest bias risk, not included in the analysis of therapeutic efficacy comparisons). Exclusion criteria were: (1) reviews, animal studies, conference abstracts, case reports, or meta-analyses; (2) unretrievable full texts or data that could not be extracted/converted; (3) unreported outcomes; (4) studies rated as very low quality (the original Jadad scale < 2 score; the Newcastle–Ottawa Scale (NOS) evaluation criteria for cohort studies < 5 score; the revised methodological index for non-randomized studies (MINORS) sale ≤ 8 score. (5) When multiple reports originate from the same study, the outcome indicators of the report are incomplete or the shortest follow-up period.

To reduce selective reporting bias, we pre-defined an endpoint hierarchy: the protocol-defined primary endpoint was used when available; otherwise, the primary analysis timepoint or longest follow-up was selected. For continuous outcomes (proteinuria, SLEDAI), we extracted final visit values to maintain consistency. When multiple timepoints were reported, the one matching the primary endpoint definition was used; others were described narratively.

We established pre-defined criteria for assessing confounding in lupus nephritis research, including disease characteristics, renal function, treatment regimens, patient characteristics, and study design. These criteria serve as an objective benchmark for applying the ROBINS-I tool in the evaluation of non-randomized studies, such as those with historical controls.

To avoid overrepresentation from overlapping studies, when multiple reports from the same institution or recruitment period were identified, we assessed overlap by comparing author lists, recruitment periods, and baseline characteristics. If overlap was confirmed or suspected, we retained the study with the largest sample size, the longest follow-up time, or the lowest risk of bias, excluding duplicates.

Calculations and Statistical Analysis

Meta-analyses were executed in STATA 14.0 and registered on PROSPERO as CRD420251073637. Dichotomous outcomes were pooled as risk ratios (RRs) with 95% confidence intervals (CIs) [25], while continuous outcomes were used as weighted mean differences (WMDs) with 95% CIs [26]. Single-arm study data underwent descriptive synthesis without meta-analysis. Heterogeneity was quantified via I² statistics and chi-square tests (P value threshold: 0.10); fixed-effects models were applied when I² < 25% and P > 0.10, otherwise random-effects models were used [27]. Both the ROBINS-I tool [28] and the NOS [29] were used to evaluate the quality of the non-randomized studies, with the ROBINS-I tool assessing quality across seven domains. To ensure a consistent and objective assessment, we predefined the standards for evaluating the effectiveness of confounding control in non-randomized LN studies. Specifically, we predefined key prognostic factors and documented whether studies reported and adjusted for treatment changes, early events, and missing data, ensuring a transparent evaluation. The sensitivity analysis involved sequentially excluding individual studies. Subgroup analyses stratified outcomes by treatment duration (induction period ≤ 6 months vs. maintenance period > 6 months) and control regimen (glucocorticoids vs. glucocorticoids + immunosuppressants). Publication bias was assessed via Begg’s and Egger’s tests [30], and trial sequential analysis (TSA) verified evidence robustness [31]. Independent comparisons were used for studies with multiple groups to avoid unit-of-analysis error.

Ethical Considerations

This systematic review and meta-analysis were conducted using data from previously published studies. As no new human participants or animals were involved, and no primary data were collected, separate ethical approval was not required for this work.

Results

Study Cohort Characteristics

Systematic searches across PubMed, Embase, Web of Science, Cochrane Library, and ClinicalTrials.gov initially identified 407 records. After applying the inclusion and exclusion criteria, 19 studies met all criteria for inclusion: comprising four RCTs [22, 32–34], one cohort study [35], and 14 single-arm interventional studies [36–49]. One of the RCT studies [34] contributed two comparison groups. Figure 1 details the PRISMA flow diagram, while Table 1 summarizes study characteristics.

Fig. 1.

Flowchart of the literature selection process

Table 1.

Basic information on the included research literature

Author	Year	Country	Intervening vs. control measures	Research type	Intervening measures	Control measures	Follow-up timeb	Main outcome	Secondary outcome
Homma	1989	Japan	25 vs. 24	RCT	MZB 50 mg Tid + GC	Placebo + GC	6 months	①	②③
Tanaka	2010	Japan	28 vs. 29	RCT	MZB 4–5 mg/kg/day Bid + GC	GC	12 months	①	③
Feng (1)a	2014	China	30 vs. 30	RCT	MZB 300 mg/day Qod + GC	MMF 1 g Bid + GC	6 months	①	②③④
Feng (2)a	2014	China	30 vs. 30	RCT	MZB 300 mg/day Qod + GC	CTX 0.5 g 2w + GC	6 months	①	②③④
Dong	2025	China	123 vs. 120	RCT	MZB 50 mg Tid + GC	CTX 0.5–1.0 g 4w + GC	12 months	①	②③④
Aihara	2002	Japan	7 vs. 6	CS	MZB 150–200 mg/day + GC	GC	24 months	①	②
Yumura	2005	Japan	10	SAT	MZB 100–200 mg/day + GC		60 months	①	②③
Tanaka	2005	Japan	11	SAT	MZB 1.8–7.8 mg/kg 2w + GC		12 months	①	②
Yagi	2014	Japan	881	SAT	MZB + GC + other IS		36 months	-	②
Tanaka	2013	Japan	19	SAT	MZB 150 mg/day + TAC 3 mg/day + GC		36 months	①	②
Zhang	2013	China	11	SAT	MZB 150 mg/day + GC		6 months	①	②③④
Kagawa	2018	Japan	36	SAT	MZB 300 mg/day + TAC 3 mg + GC		12 months	①	-
Kuroda	2007	Japan	13	SAT	MZB 75–150 mg/day + GC		12 months	①	②③
Tanaka	2008	Japan	10	SAT	MZB 6–10 mg/kg Biw + GC		29 months	①	②
Nishi	2013	Japan	17	SAT	MZB 700 mg + MZB 350 mg Qw		24 months	①	②④
Tanaka	2014	Japan	7	SAT	TAC 3 mg + MZB 150 mg + GC		12 months	①	②
Tanaka	2006	Japan	5	SAT	MZB 250–500 mg Biw + GC		24 months	-	②③
Tanaka	2008	Japan	5	SAT	MZB 10 mg/kg Biw + GC		12 months	①	②
Tanaka	2004	Japan	6	SAT	First 6 M:MZB 400–600 mg 2w; after 6 M:MZB 200–300 mg 2w;		12 months	①	②③
Nomura	2012	Japan	6	SAT	MZB 150 mg/day + TAC 3 mg/day + GC		12 months	①	②③④

Remarks: ①Renal response rate: The proportion of patients achieving either complete response and partial response; ②Adverse reaction rate: Refers to the proportion of patients reporting any adverse event; ③24-h urinary protein; ④SLEDAI score

MZB mizoribine, TAC tacrolimus, CTX cyclophosphamide, MMF mycophenolate mofetil, GC glucocorticoid, RCT randomized controlled trial, SAT single-arm trial, CS cohort study, Bid twice daily, Tid three times daily, Qod every other day, Qw quaque week, 2w every 2 weeks, 4w every 4 weeks, Biw bis in week, 6 m 6 months

^aFeng 2014 study contributed two comparison groups (MZR vs. MMF and MZR vs. CTX)

^bFollow-up time = endpoint time, used for meta-analysis

In the cohort study by Aihara 2002 the clearly defined maintenance-phase intervention (prednisolone 15–20 mg/day plus MZB 150–200 mg/day, 1991–1995) was compared with prednisolone alone (30 mg/day, 1983–1990), with both strategies initiated at the start of the maintenance phase. The analysis was conducted per-protocol and did not adjust for potential post-baseline selection bias.

Quality Assessment

Using the Jadad scale, the four included RCTs were rated as moderate in quality, with three studies scoring 3 points and one scoring 2 points. Quality for the cohort study (NOS score = 5), the ROBINS-I tool assesses the overall risk shift as high risk. All single-arm studies exceeded the MINORS high-quality threshold (≥ 13 score), collectively supporting the evidence robustness (Table 2).

Table 2.

The original Jadad scale

Author	Year	Randomization method	Double-blind method was adopted	Dropout and loss to follow-up	Is randomization appropriate?	Is the blinding method appropriate?	Total points
Homma	1989	1	0	1	1	0	3
Tanaka	2010	1	0	0	1	0	2
Feng	2014	1	0	1	1	0	3
Dong	2025	1	0	1	1	0	3

The NOS evaluation criteria for cohort studies
Author	Year	Research on population selectivity	Inter-group comparability	Outcome measurement	Total points
Aihara	2002	3	0	2	5

The ROBINS-I tool assessment
Author	Year	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from intended interventions	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall bias risk
Aihara	2002	High risk	High risk	Low risk	Information insufficiency	Information insufficiency	Low risk	Low risk	High risk

The revised MINORS scale in 2003
Author	Year	Type	Total score/Max score	Quality level
Yumura	2005	SAT	14/16	High
Tanaka	2005	SAT	14/16	High
Yagi	2014	SAT	16/16	High
Tanaka	2013	SAT	16/16	High
Zhang	2013	SAT	15/16	High
Kagawa	2018	SAT	16/16	High
Kuroda	2007	SAT	15/16	High
Tanaka	2008	SAT	15/16	High
Nishi	2013	SAT	15/16	High
Tanaka	2014	SAT	16/16	High
Tanaka	2006	SAT	15/16	High
Tanaka	2008	SAT	15/16	High
Tanaka	2004	SAT	15/16	High
Nomura	2012	SAT	15/16	High

NOS Newcastle–Ottawa Scale, ROBINS-I Risk of bias in Non-randomized studies—of Interventions, MINORS Methodological index for non-randomized studies, SAT single-arm trial

Primary Outcome Evaluation (Controlled Studies)

For the primary outcome, meta-analysis of data from five studies (contributing six independent comparisons, including a cohort study, timepoints ranging from 6 to 24 months) reporting renal response rates demonstrated significant heterogeneity (I² = 61.0%, P = 0.03). Random-effects modeling showed no significant difference between MZR and control groups (RR = 1.01, 95% CI 0.71–1.44, P > 0.05; Fig. 2). After excluding the Aihara 2002 cohort study with high risk of bias, there was no statistically significant change in the results (RR = 0.95, 95% CI 0.7–1.29, P > 0.05; Fig. S1), The study defined 'remission' clinically, without specific lab thresholds for proteinuria or creatinine stability.

Fig. 2.

Results of the forest plot of renal remission rates

Secondary Outcome Evaluation (Controlled Studies)

The three secondary outcome indicators showed no statistically significant differences at the baseline period (P > 0.05). All continuous outcomes were analyzed using endpoint values for consistency. Secondary indicators in the analysis results at the endpoint stage: 1. Adverse reaction rates (Four studies, contributing five independent comparisons, Including a cohort study, I² = 19.3%, P = 0.29) showed no statistical significance between the experimental group and the control group in the fixed-effect model (RR = 0.95, 95% CI 0.83–1.09, P > 0.05; Fig. 3); After excluding the Aihara 2002 cohort study with high risk of bias, there was no statistically significant change in the results (RR = 0.84, 95% CI 0.45–1.56, P > 0.05; Fig S2). 2. 24-h urinary protein (five studies, I² = 47.7%, P = 0.11) showed no statistical significance between the experimental group and the control group in the random effect model. (WMD = 0.39, 95% CI − 0.01 to 0.87, P > 0.05; Fig. 4). 3. SLEDAI scores (three studies, I² = 59.7%, P = 0.08) showed a statistically significant difference between the experimental group and the control group in the random effects model (WMD = 1.75, 95% CI 0.33–3.16, P < 0.05, Fig. 5).

Fig. 3.

Adverse reaction rates forest plot results

Fig. 4.

Results of the 24-h urinary protein quantitative forest plot

Fig. 5.

SLEDAI score forest plot results

Descriptive Synthesis of Single-Arm Interventional Studies

Descriptive synthesis of 14 single-arm studies revealed substantial variability: renal response rates ranged 59–100% (12/14 studies > 86%), adverse reaction rates varied markedly (5–50%), while urinary protein trends were inconsistent (five studies reported reductions up to − 4.6 g/24 h vs. one showing increases). All three studies reporting SLEDAI scores documented reductions. The research results have significant methodological limitations, which restrict both their validity and comparability. Interpretation should be carried out with extreme caution (Table 3).

Table 3.

Reported outcomes from single-arm interventional studies

Author	Year	Overall renal response rate (%)	Adverse reaction rate (%)	24-h urinary protein			SLEDAI score
				Baseline	Post-treatment	Mean difference	Baseline	Post-treatment	Mean difference
Yumura	2005	70	10	1.54	0.29	− 1.25	–	–	–
Tanaka	2005	100	5	–	–	–	–	–	–
Yagi	2014	–	22	–	–	–	–	–	–
Tanaka	2013	89	21	–	–	–	–	–	–
Zhang	2013	82	5	3.64	1.34	− 2.31	20.60	3.80	− 16.80
Kagawa	2018	92	–	–	–	–	–	–	–
Kuroda	2007	69	8	2.70	1.80	0.90	–	–	–
Tanaka	2008	90	5	–	–	–	–	–	–
Nishi	2013	59	35	–	–	–	13.00	10.00	− 3.00
Tanaka	2014	86	7	–	–	–	–	–	–
Tanaka	2006	–	40	1.70	0.30	− 1.40	–	–	–
Tanaka	2008	80	10	–	–	–	–	–	–
Tanaka	2004	83	17	1.90	0.20	− 1.70	–	–	–
Nomura	2012	100	50	4.78	0.18	− 4.60	13.30	4.00	− 9.30

Data are reported as per original studies. Due to variations in baseline characteristics, outcome definitions, and the lack of control groups, cross-study comparisons are limited. See main text qualitative synthesis. SLEDAI Systemic lupus erythematosus disease activity index

Sensitivity Analysis

Sensitivity analysis using the leave-one-out method confirmed the robustness of the meta-analysis results. The pooled estimates for renal response, adverse reaction rates, proteinuria, and SLEDAI score remained stable, with confidence intervals consistently including the main effect estimate and none crossing the null threshold upon exclusion of any single study. This indicates that the conclusions are not driven by any individual study. The stability of the results was confirmed by sequential study exclusion, which did not cause significant shifts in the estimates, despite minor fluctuations in CI width from the large-sample studies by Feng et al. [34] and Dong et al. [22].

Subgroup Analyses

Subgroup analyses stratified by treatment duration demonstrated comparable renal response and adverse reaction rates regardless of duration. However, induction period (≤ 6 months) MZR use was associated with significantly elevated proteinuria (three studies, WMD = 0.73 95% CI 0.11–1.36, P < 0.05) and SLEDAI scores (two studies, WMD = 2.60, 95% CI 1.19–4.01, P < 0.05), effects that normalized with maintenance period (> 6 months). When stratified by control regimen, MZR showed comparable outcomes to glucocorticoid monotherapy (three studies; renal response: RR = 3.04, 95% CI 0.83–11.05, P > 0.05). In contrast, compared to glucocorticoid-immunosuppressant combinations, MZR was associated with a significantly lower renal response rates (three studies, RR = 0.83, 95% CI 0.71–0.97, P < 0.05), higher proteinuria (three studies, WMD = 0.62, 95% CI 0.22–1.03, P < 0.05), and elevated SLEDAI scores (three studies, WMD = 1.75, 95% CI 0.33–3.16, P < 0.05) (Table 4).

Table 4.

Subgroup analysis results by time of therapy and control measures

Subgroup type	Outcome indicator	Inclusion criteria for literature	Heterogeneity test		Effect model	Meta-analysis
			I2 (%)	P		RR/WMD	95% CI	P
≤ 6 months	Renal remission rate	3	72.2	0.03	Random	1.05	0.53–2.07	0.89
	Adverse reaction rate	3	33.0	0.23	Random	0.53	0.14–1.98	0.34
	24-h urinary protein	3	0	0.74	Fixed	0.73	0.11–1.36	0.02
	SLEDAI score	2	0	0.65	Fixed	2.60	1.19–4.01	0.01
> 6 months	Renal remission rate	3	63.4	0.07	Random	1.15	0.62–2.14	0.65
	Adverse reaction rate	2	0	0.73	Fixed	1.01	0.89–1.15	0.85
	24-h urinary protein	2	79.7	0.03	Random	0.14	− 0.69 ~ 0.97	0.74
	SLEDAI score	1	0	0.01	Fixed	0.80	0.01–1.59	0.06
Glucocorticoid group	Renal remission rate	3	62.2	0.07	Random	3.04	0.83–11.05	0.09
	Adverse reaction rate	2	0	0.76	Fixed	1.07	0.41–2.78	0.89
	24-h urinary protein	2	42.9	0.19	Random	− 0.01	− 0.83 to  0.8	0.97
	SLEDAI score	–	–	–	–	–	–	–
Glucocorticoid and immunosuppressants group	Renal remission rate	3	0	0.73	Fixed	0.83	0.71–0.97	0.02
	Adverse reaction rate	3	51.1	0.13	Random	0.61	0.17–2.19	0.45
	24-h urinary protein	3	0	0.67	Fixed	0.62	0.22–1.03	0.01
	SLEDAI score	3	59.7	0.08	Random	1.75	0.33–3.16	0.02

RR risk ratio, WMD weighted mean difference, CI confidence interval, SLEDAI systemic lupus erythematosus disease activity index

Publication Bias Assessment and Trial Sequential Analysis

Begg’s and Egger’s tests detected no significant publication bias (P > 0.05). Trial sequential analysis for renal response rates confirmed evidence sufficiency (cumulative n = 482 > required information size [RIS] = 405), demonstrating stable null effects (Fig. S3). For adverse reaction rates, the sample size (n = 389) reached only 84.9% of RIS (458), precluding definitive conclusions (Fig. S4). Given the limited number of studies included, the statistical test power is insufficient, and the possibility of undetected publication bias cannot be ruled out.

Discussion

This systematic review and meta-analysis evaluated the efficacy and safety of MZR in 1489 patients with LN across 19 studies. Our primary comparative conclusions are derived from controlled studies (four RCTs and one cohort study), which provide the most robust evidence for causal inference. Single-arm studies (n = 14) are presented descriptively in “Descriptive Synthesis of Single-Arm Interventional Studies” and are not pooled with controlled data due to inherent methodological limitations. Primary analyses demonstrated comparable renal response rates and adverse reaction rate profiles between MZR and controls. However, subgroup analyses revealed context-dependent efficacy: MZR showed non-inferiority versus glucocorticoid monotherapy, it was associated with significantly reduced renal response, elevated proteinuria, and higher SLEDAI scores compared to standard glucocorticoid-immunosuppressant combinations during the induction period (≤ 6 months). Notably, these inferior outcomes were transient and did not persist in the maintenance phase (> 6 months), suggesting a delayed yet ultimately comparable therapeutic effect of MZR.

The observed statistical heterogeneity (e.g., I² = 61.0% for renal response) likely stems from several clinical and methodological sources. First, the inclusion of a historically controlled cohort study (Aihara 2002), which was judged to be at high risk of bias, contributed substantially. This study showed a large effect size (RR = 12.19) but with a very wide confidence interval. Second, there were significant variations in MZR dosing regimens, concomitant therapies, and the composition of control groups across trials. Our prespecified subgroup analyses were instrumental in clarifying the sources of this heterogeneity. Notably, when MZR was compared directly with standard glucocorticoid-immunosuppressant combination regimens, the renal response rate was significantly lower (RR = 0.83, 95% CI 0.71–0.97, P = 0.02) with low heterogeneity within this subgroup (I² = 0%). This clearly identifies the strength of the control regimen as a key effect modifier and a major source of the overall heterogeneity in efficacy evaluation.

We acknowledge that despite standardization efforts, some heterogeneity in outcome definitions and follow-up durations remains across studies. To address this, we stratified analyses by treatment duration and performed sensitivity analyses, which supported the robustness of our primary conclusions. Readers should interpret the pooled estimates with the understanding that they synthesize studies with inherently variable designs. The consistency of findings across subgroups (e.g., transient inferior outcomes during induction) strengthens confidence in the observed patterns.

The observed inferiority of MZR in early disease control may be attributed to its pharmacodynamic profile. As a selective inhibitor of inosine monophosphate dehydrogenase, MZR gradually suppresses lymphocyte proliferation via purine synthesis inhibition, contrasting with the more rapid cytolytic or antiproliferative effects of agents like cyclophosphamide or mycophenolate mofetil. This slower onset of action may explain the suboptimal reduction in proteinuria and SLE disease activity within the first 6 months. Additionally, unmeasured baseline differences, potential selection bias toward refractory cases in real-world studies, and variability in concomitant steroid dosing could have contributed to these early outcome disparities.

LN manifests across a clinical spectrum ranging from asymptomatic proteinuria to nephrotic syndrome or acute kidney injury. Its heterogeneous histopathological classifications serve as major predictors of poor prognosis and increased mortality [50–53]. LN pathogenesis involves immune complex-mediated glomerular injury [54, 55], necessitating immunosuppression. Current first-line regimens (e.g., cyclophosphamide/mycophenolate mofetil + glucocorticoids) carry significant toxicities [6, 15, 16]. As an inosine-5′-monophosphate dehydrogenase inhibitor, MZR modulates purine synthesis with potentially milder myelotoxicity [23, 24, 56, 57], though its slow onset (evidenced by suboptimal induction period disease control) may limit utility in acute settings. However, its role should be contextualized within the treatment continuum: it is not recommended for rapid induction in severe LN but may be valuable for the maintenance period or in patients with intolerant to conventional agents.

The high renal response rates (59–100%) reported in single-arm studies must be interpreted with caution due to inherent design limitations, including lack of comparator groups, variability in concomitant therapies, and potential selection bias. Similarly, the wide range of adverse reaction rates (5–50%) underscores the heterogeneity in real-world tolerability and monitoring practices. Our trial sequential analysis confirmed robust evidence of comparable renal responses but indicated insufficient data to draw definitive safety conclusions, highlighting the need for larger, controlled safety studies.

MZR is contraindicated as a replacement for standard induction therapy in severe LN requiring rapid disease control. Its clinical utility should be reserved for specific scenarios: (1) patients who are intolerant to mycophenolate mofetil or cyclophosphamide; (2) resource-limited settings where its cost-effectiveness offers practical advantages; and (3) maintenance period phases leveraging its favorable maintenance period safety profile. Notably, insufficient evidence exists for pediatric, pregnant, or non-Asian populations, precluding generalized recommendations in these subgroups.

Study limitations include 1. The main limitation is the inclusion of a historically controlled cohort study, which carries a high risk of bias, confounding, and potential exposure misclassification due to incomplete treatment reporting. While its combination with RCTs may create heterogeneity, sensitivity analyses indicated that the core conclusions are robust. 2. We acknowledge heterogeneity in outcome definitions, particularly for renal response, across included studies. Original definitions of remission varied from clinical assessments to protocol-based composite endpoints, which may affect pooled estimates and interpretation of comparability. However, the use of random-effects models and sensitivity analyses supports the robustness of our primary conclusion on MZR’s overall renal efficacy. 3. The predominance of evidence from East Asia—especially Japan—limits the generalizability of our conclusions. While this provides a consistent context for evaluating MZR, variations in genetics, disease phenotypes, and clinical practice across regions may affect outcomes. Our findings are thus most applicable to East Asian populations, underscoring the need for multinational studies to confirm MZR’s efficacy and safety in diverse settings. 4. Substantial heterogeneity in MZR dosing protocols (ranging from 150 mg/day to pulse regimens) [58, 59], predominance of Asian cohort data [60], and a paucity of randomized controlled trials (n = 4) in the evidence base. 5. Future research should primarily focus on: initiating well-powered multicenter RCTs for head-to-head comparisons between MZR and first-line immunosuppressants; establishing evidence-based optimal dosing regimens (including intensity and treatment duration); validating therapeutic efficacy and safety across diverse ethnic cohorts to address current geographic evidence disparities; and investigating potential synergistic effects when combined with biologic agents through mechanistic and clinical studies.

In conclusion, MZR exhibits overall efficacy and safety comparable to those of control therapies in LN but demonstrates inferior early disease control compared to standard immunosuppressive combinations. Its clinical use should be judiciously tailored, prioritizing patients who are intolerant to conventional agents or in resource-limited settings, and should generally be reserved for maintenance rather than the induction period.

Conclusions

Mizoribine exhibits a time-dependent efficacy profile in lupus nephritis. Based on a meta-analysis of controlled studies (RCTs and cohort), while its overall efficacy and safety are comparable to those of control therapies, it demonstrates inferior induction-phase disease control compared to standard immunosuppressive regimens during the induction period (≤ 6 months). However, its efficacy becomes comparable during the maintenance period (> 6 months). Single-arm studies reported high response rates, but their wide variability and lack of comparators limit direct inference. Therefore, mizoribine is not recommended as first-line therapy for rapid remission but is a viable maintenance option for patients intolerant to conventional agents or in resource-limited settings, with its use requiring judicious tailoring to individual clinical contexts.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Xingyun Wan: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft. Xiaolong Wang: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft. Shuang Liang: Investigation, Software, Validation. Qian Wang: Investigation, Software, Resources. Chao Liu: Conceptualization, Project administration, Supervision, review & editing. Zheyi Dong: Conceptualization, Project administration, Supervision, review & editing.

Funding

No funding or sponsorship was received for this study or publication of this article. The Rapid Service Fee was funded by the authors.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Declarations

Conflict of Interest

Xingyun Wan, Xiaolong Wang, Shuang Liang, Qian Wang, Chao Liu, and Zheyi Dong declare that they have no relevant conflicts of interest to disclose. For full transparency, we disclose that author Zheyi Dong, who co-authored the meta-analysis data source (reference 22), has also previously been an author on publications sponsored by the drug’s manufacturer. The cited work was an academic undertaking, and no commercial entity has influenced the current study.

Ethical Approval

This systematic review and meta-analysis were conducted using data from previously published studies. As no new human participants or animals were involved, and no primary data were collected, separate ethical approval was not required for this work.