A systematic literature search was conducted across international and Chinese databases, including Web of Science, Embase, PubMed, Google Scholar, CNKI, Baidu Scholar, and Wanfang. The search window spanned from January 2001 to April 2024. Keywords and Medical Subject Headings (MeSH) included “kidney transplantation,” “ABOi,” “renal transplantation,” “ABO incompatible,” “ABOi-rTx,” “ABOi-KT,” and “ABOi-LDKT”.

Study eligibility was determined using the PICO framework to ensure relevance and homogeneity:

Population (P): Asian patient populations undergoing living donor kidney transplantation.

Intervention (I): ABOi-LDKT utilizing a Rituximab-based desensitization protocol.

Comparator (C): ABOc-LDKT.

Outcomes (O): Clinical outcomes including patient and graft survival with a minimum follow-up duration of 1 year, providing sufficient data for extraction.

Data were extracted independently by two investigators using a standardized form. Extracted variables included lead author, publication year, follow-up duration, dialysis history, sample sizes, and event counts (death/graft loss) at 1, 3, and 5 years. Quality assessment was performed using the Newcastle-Ottawa Scale (NOS). (http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp). Any discrepancies in data extraction or quality scoring were resolved through discussion or by consulting a third senior investigator.

Meta-analysis was performed using Stata 12.0 (StataCorp, College Station, TX, USA). Pooled Odds Ratios (ORs) with 95% Confidence Intervals (CIs) were calculated for graft and patient survival. Subgroup analyses evaluated complications including Cytomegalovirus (CMV)/BK virus infection and acute rejection. Heterogeneity was assessed via the Cochrane Q test and I² statistic. To account for potential between-study variance, a random-effects model was applied if significant heterogeneity was observed (P < 0.1 or I² > 50%); otherwise, a fixed-effects model was utilized. Publication bias was assessed using Begg’s test and funnel plots, and sensitivity analysis was conducted to verify result stability.

Recipients in both groups were required to meet the following criteria: (1) follow-up duration of >1 year in our outpatient department; (2) ≤3 human leukocyte antigen (HLA) mismatches; (3) complement-dependent cytotoxicity (CDC) <10%; (4) negative reactivity for class I and class II antibodies; (5) first kidney transplant from a direct blood relative who had provided written informed consent for altruistic donation; (6) adult recipients; and (7) implantation of the graft in the right iliac fossa. Recipients with left iliac fossa implantation (n = 15) were excluded a priori to reduce surgical and technical heterogeneity (e.g., vascular length and angle, ureteral course, and wound exposure), as these factors could influence early non-immunologic complications in our small cohort.

The experimental group received desensitization therapy based on the Technical Operating Specifications for ABO-Incompatible Living Donor Kidney Transplantation in China (2019) (7). An individualized Rituximab regimen was administered based on preoperative CD19⁺ B-cell counts: patients with >15% received 200 mg at 4 weeks, 100 mg at 2 weeks, and 100 mg at 1 day preoperatively; those with 10–15% received 100 mg at these three time points; and those with <10% received 100 mg at 4 and 2 weeks preoperatively. Isoagglutinin titers (IgM and IgG) were monitored at initial typing, before and after pharmacotherapy, before and after PE, and on postoperative days 2, 5, and 14. The target titer was <1:16. If titers exceeded 1:16, PE or Double Filtration Plasmapheresis (DFPP) was performed using albumin or fresh frozen plasma as replacement fluids until the target was met.

While the ABOc-LDKT group did not receive immunosuppressants before surgery, the ABOi-LDKT group initiated triple therapy 1–2 weeks preoperatively, consisting of Tacrolimus (0.08–0.1 mg/kg/day, target serum trough 10–12 ng/ml), Mycophenolate Mofetil (MMF, 1.0–1.5 g/day, target AUC 30–60 ng/ml), and Prednisone (20 mg/day). Induction therapy for all patients included Antithymocyte Globulin (ATG, 25 mg on days 0, 1, and 2) and methylprednisolone sodium succinate (500 mg IV on days 0, 1, and 2). Postoperatively, methylprednisolone was tapered daily to 120 mg, 80 mg, and 40 mg on days 3, 4, and 5, respectively, and switched to oral prednisone (10 mg/day) on day 6. MMF (1.0–1.5 g/day) was resumed on the evening of surgery, and oral Tacrolimus (initial dose 0.1–0.2 mg/kg/day) was started on postoperative day 3, adjusted to maintain therapeutic levels.

Routine antibacterial prophylaxis utilizing Mezlocillin sodium/Sulbactam sodium was administered from postoperative day (POD) 0 to 12 targeting Gram-negative bacteria; coverage for fungi or Gram-positive bacteria was reserved for culture-positive cases derived from donor perfusion or drainage fluid analysis. Prophylaxis against Pneumocystis jirovecii pneumonia was initiated on POD 7 with Sulfamethoxazole-Trimethoprim (SMZ-TMP) (200/40 mg daily) and continued for 6 months. Ganciclovir was administered for CMV prophylaxis from POD 7 to month 3, with dosages strictly stratified by estimated glomerular filtration rate (eGFR): (1) eGFR >70 mL/min: 1000 mg PO three times daily (tid); (2) eGFR 50–69 mL/min: 500 mg PO tid; (3) eGFR 25–49 mL/min: 500 mg PO twice daily (bid); (4) eGFR 10–24 mL/min: 500 mg PO once daily (qd); (5) eGFR <10 mL/min: 500 mg PO every other day (qod).

Comprehensive clinical data were systematically collected for recipients, including demographics, preoperative history, dialysis modality, and perioperative metrics such as hospital stay and costs. Specific surveillance outcomes included the incidence of severe pulmonary and urinary tract infection (UTI) requiring re-hospitalization within 1 year post-transplantation, while acute rejection episodes and surgery-related complications were monitored during the perioperative period. Follow-up assessments tracked patient and graft survival at 7 days, 14 days, 1 month, 6 months, and 1 year post-transplantation. Key outcomes were defined as follows: delayed graft function (DGF) was defined as the requirement for dialysis within the first week post-transplantation; Acute Rejection was suspected based on clinical indicators and confirmed by percutaneous renal biopsy according to the Banff classification; Surgical complications were categorized as ureteral stenosis, anastomotic hemorrhage, or wound infection/dehiscence; and postoperative renal function was assessed using the CKD-EPI creatinine equation (race-free) reported as eGFR(mL/min/1.73 m²).

Donor demographics and baseline characteristics were recorded following the recipient protocol. Preoperative donor glomerular filtration rate (GFR) was measured via renal dynamic scintigraphy using the Gates camera method with ^99mTc-DTPA. Net injected counts were obtained by syringe pre/post counting, and regions of interest (ROIs) were drawn over each kidney with background subtraction. Attenuation was corrected using Tonnesen depths (YL = 13.2×W/H+0.7; YR = 13.3×W/H+0.7; μ = 0.153 cm⁻¹), and the 2–3-min window provided bilateral uptake (%). Total GFR was calculated as GFR_total (mL/min) = 9.813×uptake (%) − 6.825, with single-kidney GFR = GFR_total × split function; values were BSA-standardized to 1.73 m².

Data analyses were performed using R v4.2.2 (R Foundation, Vienna, Austria) and GraphPad Prism v9.0 (GraphPad Software, San Diego, CA, USA). Continuous variables were expressed as mean ± SD or median (IQR) and tested for normality using the Kolmogorov–Smirnov test (P < 0.05 indicated non-normal distribution). For group comparisons, normally distributed continuous variables were analyzed using the unpaired Student’s t-test, while non-normally distributed variables were assessed using the Mann–Whitney U test. To evaluate longitudinal within-group changes in serum creatinine (Scr) and eGFR at different follow-up time points, paired t-tests were employed. Categorical data were compared using the Chi-square test or Fisher’s exact test as appropriate. Survival models were constructed using the “survival” and “survminer” R packages to generate Kaplan–Meier curves, with differences assessed via the log-rank test. While propensity score methods were considered to balance baseline covariates, the modest sample size of the ABOi-LDKT cohort after prespecified exclusions would have led to unstable estimates. Therefore, we focused on strict eligibility criteria and standardized perioperative management to minimize major confounding in this exploratory cohort. A two-tailed P-value < 0.05 was considered statistically significant.

The selection process used in this study is illustrated in Figure 1A. A total of 18 papers were included in this meta-analysis, consisting of 17 in English (6, 9–24) and 1 in Chinese (8). All studies were published between January 1, 2001, and April 1, 2024, involving 15,611 patients (Figure 1a). Table 1 summarizes the detailed information on the studies included in this meta-analysis. All included papers were scored 5 or higher on the NOS scale.

In the analysis of graft survival rates, 14 studies were included for the 1-year survival rate, 11 for the 3-year survival rate, and 10 for the 5-year survival rate. As shown in Table 2 and Figure 2, the 1-year graft survival rate for the ABOi-LDKT group was lower than that for the ABOc-LDKT group (Pz < 0.01, OR = 0.518, CI = 0.386 - 0.695). However, there were no significant differences in the 3-year (Pz = 0.330, OR = 1.307, CI = 0.763 - 2.240) and 5-year (Pz = 0.193, OR = 0.623, CI = 0.305 - 1.270) survival rates between the two groups. In addition, the ABOi-LDKT group had a lower patient survival rate at 3 years (Pz = 0.021, OR = 0.552, CI = 0.333 - 0.914), but not at 1 year (Pz = 0.860, OR = 0.969, CI = 0.679 - 1.381) or 5 years (Pz = 0.187, OR = 1.199, CI = 0.916 - 1.569). In the subgroup analysis of postoperative complication risks, Cytomegalovirus (CMV) infection, BK virus infection, UTI, and postoperative acute rejection were investigated. No significant differences were observed between these groups, as shown by the forest plots in Figure 3.

Sensitivity analyses were performed using a leave-one-out strategy, sequentially excluding individual studies to evaluate the robustness of the pooled results. The recalculated pooled effect sizes (HRs/ORs) and corresponding 95% confidence intervals (CIs) remained consistent, confirming that no single study disproportionately influenced the overall findings (Supplementary Table S1; Supplementary Figure S1). Furthermore, publication bias was assessed through visual inspection of funnel plots and Begg’s rank correlation test. As shown in Supplementary Figure S2 and Supplementary Table S2, the symmetrical distribution of the funnel plots, corroborated by the non-significant results of Begg’s test (P > 0.05), indicates the absence of substantial publication bias among the included studies.