A retrospective study using serum samples from a prospective biobank of kidney transplant recipients at Seoul National University Hospital from August 2016 to December 2019 was undertaken. Routine serum samples were collected preoperatively (T0) and at 2 weeks postoperatively (T1), 3 months postoperatively (T2), and 1 year postoperatively (T3). Data on clinical variables, transplantation admission and operation history, cause of end-stage renal disease, lab findings, immunosuppressive therapies, and rejection and infection history were extracted and categorized based on this time interval (T0–T3).

The occurrence of specific episodes, such as infection and rejection, was defined according to the intervals between the serum samples collected: P0 if it occurred between T0 and T1, P1 if it occurred between T1 and T2, P2 if it occurred between T2 and T3, and P3 if it occurred within 3 months after T3 sampling. Infections that occurred in P0 were excluded from the analysis as they overlapped with admission for transplantation and might not reflect true opportunistic infection.

During the study period, a total of 569 kidney transplantations were performed. Excluding patients with missing samples from any of the four time periods, a total of 192 patients had all four samples and were included for the final analysis.

The study was designed to evaluate (1) the serial changes in serum levels of immunoglobulins (IgG, IgA, and IgM) and complement components (C3 and C4) across four key postoperative intervals (T0, T1, T2, and T3) to assess the temporal changes in immune function, (2) the prevalence of hypogammaglobulinemia (HGG), hypocomplementemia (HCC), and infection, and (3) the association between HGG and infection to identify significant risk factors for infection among transplant recipients.

The serum samples from the biobank were preserved in liquid nitrogen tanks at -195°C and thawed before use. The IgG, IgA, IgM, C3, and C4 levels collected at each time period (T0–T3) were measured using an automated chemical analyzer (TBA-Fx8, Toshiba, Tokyo, Japan) and immunoassay reagents supplied by the manufacturer (Abott, Tokyo, Japan).

Based on the measured serum levels, hypogammaglobulinemia (HGG) was defined as a serum IgG level below the normal range or the presence of more than one class of immunoglobulins below the normal range. The normal ranges were defined as follows: IgG, 700–1,600 mg/dL; IgA, 70–400 mg/dL; and IgM, 40–230 mg/dL (15). Hypocomplementemia (HCC) was defined as serum C3 (hypoC3) or C4 (hypoC4) levels below the normal range. The normal ranges were defined as follows: C3, 83–171 mg/dL and C4, 14–38 mg/dL (11).

The infection episodes that were analyzed were severe infection, defined as any infectious episode requiring hospitalization and treatment, and graft-related viral infections, defined as infection by any of the following pathogens: polyoma BK virus (BKV), cytomegalovirus (CMV), and Epstein–Barr virus (EBV). CMV, EBV, and BKV are routinely screened at our center through serum quantitative polymerase chain reaction (qPCR) preoperatively and at 2 weeks postoperatively, 3 months postoperatively, 6 months postoperatively, and 1 year postoperatively. BKV infection was defined as viremia with a viral load over 1,000 copies/mL detected through qPCR or biopsy-proven BKV-associated nephropathy. EBV infection was defined as viremia with a viral load over 4,000 copies/mL detected through qPCR. CMV infection was defined as viremia with a viral load over 1,000 copies/mL detected through qPCR with or without signs and symptoms of disease or the presence of specific symptoms in a target organ with positive CMV in tissue culture.

Patients who underwent preoperative desensitization and rejection treatment received Pneumocystis jirovecii prophylaxis of 400 mg trimethoprim/80 mg sulfamethoxazole daily for 6 months. CMV prophylaxis for 6 months with valacyclovir was administered to patients who underwent induction treatment with anti-thymoglobulin or was a CMV-seronegative recipient with a CMV-seropositive donor (CMV D+/R-).

All patients were prescribed a triple immunosuppressant regimen comprised of tacrolimus, mycophenolate mofetil, and steroid. For the first 3 months post-transplantation, the tacrolimus trough level (FK C0) was targeted to 8–12 ng/mL. The target levels were then lowered to 6–10 ng/mL until 1 year after transplantation and to 4–6 ng/mL thereafter. Mycophenolate mofetil was administered at a fixed dose of 500 mg twice a day. IV steroids of 500 mg were used on the day of operation, followed by rapid tapering to achieve a daily dose of oral prednisolone at 5 mg within the first month after transplantation.

Incompatible living donor transplantation (iLDKT) was defined as recipients who received kidneys from ABO-incompatible donors or had pre-formed donor-specific antibodies (DSA). The desensitization protocol for such iLDKT recipients was as follows: rituximab administration 2–4 weeks prior to transplantation, followed with pretransplant plasmapheresis (PPH) with intravenous immunoglobulin (IVIG) of 0.2 g/kg prior to transplantation until the ABO antibody titer levels dropped below 1:32 and the DSA MFI levels below 1,000.

For induction immunosuppression, basiliximab was administered (20 mg) at days 0 and 4 after transplantation. In high-risk patients with high panel-reactive antibody titers (PRA) (>80%) or who were initially DSA-positive, anti-thymoglobulin (ATG) (1.5 mg/kg/day) was administered instead of basiliximab and was maintained until postoperative day 3 or 4.

A rejection event was defined as any biopsy-proven rejection, including borderline rejection, defined in accordance with the Banff ‘17 classification by an experienced renal pathologist. Biopsy was undertaken as per protocol at reperfusion and at 2 weeks and 1 year postoperatively or performed when indicated due to worsening of the graft, defined as greater than 20% increase in baseline serum creatinine or an appearance of de novo DSA to human leukocyte antigen.

If clinically suspected, rejection treatment with steroid pulse therapy was undertaken immediately after biopsy without waiting for histologic confirmation. All other forms of rejection treatment were started after a histologic diagnosis was made by the pathologist.

Quantitative data were shown as mean ± standard deviation (SD), and qualitative data were presented as relative frequencies (%). The Wilcoxon signed-rank test with Bonferroni correction was used to evaluate the statistical significance of the temporal changes in serum IgG, IgA, IgM, C3, and C4 levels at the four different time points (T0–T3).

Given the potential of immune modulating covariates in transplant recipients to act as risk factors for hypogammaglobulinemia, we initially conducted logistic regression using generalized estimating equations (GEE). As the outcome variable HGG was measured repeatedly, GEE was used to adjust for the clustering effects within one patient. Variables known to impact and cause HGG were selected as variables for univariate analysis. Baseline factors such as age, BMI, dialysis state, iLDKT, underlying diseases (DM, pulmonary disease, and liver disease), CMV serology and treatment, immunosuppressant use (preoperative desensitization and induction and postoperative rejection and treatment), and low immune function such as baseline HGG or HCC were selected for analysis. Then, a stepwise backward variable selection method was performed using significant variables from the univariate analysis.

A time-dependent Cox proportional hazards model was then employed to explore the relationship between hypogammaglobulinemia and infection. To further account for time-varying exposure of HGG on the occurrence of severe infection and graft-related viral infection, we utilized a marginal structural Cox proportional hazards model with stabilized weights derived from inverse probability treatment weighting (IPTW) (16, 17). The marginal structural model statistically adjusted for the confounding effect of the imbalanced variables on both the probability that an individual will be allocated to two HGG states (having HGG or not having HGG) as well as the probability of the occurrence of infection at a certain time point. The stabilized weight at time t consisted of the product of the treatment weight and censoring weight and was as follows, where A(k) was the HGG state at time k, and A ¯ (k-1) was the HGG state prior to time k:

Similarly, C(k) was the censoring state at time k, C ¯ (k -1) was the history, L ¯ (k) (was the covariate history before time k and including time k, and V was the time-independent covariate observed on the index date.

Treatment weights were calculated based on the inverse of each individual’s probability of belonging to one of the two HGG states at each observed time point. Both time-dependent and time-independent covariates were considered in this calculation. Similarly, censoring weights were calculated based on the probability of being censored at each time point and considering both the time-dependent and time-independent covariates. The covariates that were included for calculating stabilized weights were immune modulatory factors. The time-dependent covariates were rejection event occurrence, rejection treatment with IV steroid, rejection treatment with plasmapheresis and immunoglobulin, rejection treatment with ATG, and rejection treatment with other antibodies. Time-independent covariates were preoperative desensitization including plasmapheresis, immunoglobulin usage and antibody treatment, and induction with anti-thymoglobulin. This approach aimed to provide an unbiased estimation of the causal effect of HGG levels on infection risk.

All P-values were two tailed, and a P-value of<0.05 was considered statistically significant. Statistical analysis was performed using SPSS version 27.0 (SPSS Inc., Chicago, IL, USA) and SAS 9.4 (SAS institute Inc., Cary, NC, USA) for Windows.

The study was approved by the Institutional Review Board at Seoul National University Hospital (IRB no. 2106-094-1228). The need for informed consent was waived due to the retrospective nature of the study.

During the study period, a total of 569 kidney transplantations were performed. Among them, 192 patients had serum samples taken at all four time periods and were included in the study. None of these patients experienced graft failure or mortality within the study period. Most of the transplantations were living donor transplantations (81.8%, n = 157), of which 27.6% had incompatible donors. Preoperative desensitization was performed in a total of 50 patients (26%). Rituximab was administered to 48 patients (25%), and plasmapheresis with IVIG was performed in 32 patients (16.7%). Induction with ATG was undertaken in 25 patients (13%). Infection prophylaxis was undertaken for CMV in 35 patients (18.2%) and for Pneumocystis pneumonia (PCP) in 17 patients (8.9%). A total of 84 patients (43.8%) had at least one rejection event within the study period. Of these patients, 60 (71.4%) were pathologically borderline T-cell-mediated rejections. There were 54 rejection episodes in P0, 21 rejection episodes in P1, eight in P2, and 29 in P3. The clinical characteristics, including baseline characteristics and transplantation-related characteristics, are summarized in Table 1 .

The serum IgG, IgA, and C3 levels showed statistically significant changes over time with a p-value of less than 0.005. The mean values of serum IgG and IgA significantly decreased in P0 (T0 IgG 1,238.57 ± 336.05 mg/dL to T1 IgG 904.8 ± 243.8 mg/dL and T0 IgA 235.22 ± 114.38 mg/dL to T1 IgA 185.2 ± 78.8 mg/dL). The mean C3 levels significantly increased in P0 and P1 and then decreased in P2 (T0 = 82.3 ± 18.0 mg/dL, T1 = 91.6 ± 18.2 mg/dL, T2 = 108.3 ± 20.6 mg/dL, and T3 = 101.5 ± 19.3mg/dL). The mean levels of IgM and C4 did not change significantly (p = 0.473 and p = 0.327). The changes in the mean levels of immunoglobulin and complement levels are shown in Figure 1 .

A total of 54 patients had HGG during the first year after transplantation. Among the 54 patients, eight (14.8%) had HGG at baseline, and half of these patients had persistent HGG throughout all four periods of detection. Excluding baseline HGG patients, a total of 46 (24.9%) of 192 patients had newly developed HGG in the post-transplant period (T1, n = 22; T2, n = 17; and T3, n = 7). HypoC3 was found in 118 patients, and among these patients, 62 also had hypoC4. A total of 98 (83.0%) of 118 patients had hypoC3 at T0, and only six of them had persistent hypoC3 throughout all four periods. Excluding baseline hypoC3 patients, only 20 (16.9%) patients had newly developed hypoC3 in the post-transplant period (T1, n = 18; T2, n = 2; and T3, n = 0). HypoC4 was found in 74 patients during the study period, and 37 (50.0%) of them had hypoC4 at T0. Excluding baseline hypoC4 patients, 37 (50.0%) patients had newly developed hypoC4 in the post-transplant period (T1, n = 21; T2, n = 6; and T3, n = 10). The total number of patients with both HGG and HCC was 15 (T0, n = 4; T1, n = 11, T2, n = 2; and T3, n = 2) and were too few for any further analysis.

The total number of infectious episodes that occurred was 158 in 127 patients. Table 2 shows the summary of types of infection by pathogen and clinical syndromes. Out of these cases, 36 (22.8%) were severe infections that required hospitalization, and 34 (21.5%) were graft-related viral infections. The primary pathogen for severe infection was bacterial in 75% (n = 27/36). In graft-related viral infection, the incidence of each viral infection was CMV 73.5% (n = 25/34), BKV 23.5% (n = 8/34), and EBV 2.9% (n = 1/34). Severe infection requiring hospitalization occurred most frequently in the P1 period (n = 19, 52.8%) and decreased with time (P2, n = 14 and P3, n = 3). Graft-related viral infection occurred most frequently in the P1 period (n = 19, 55.9%) and decreased with time (P2, n = 9 and P3, n = 6).

Figure 2 shows the prevalence of HGG and HCC along with infection and rejection events at each period post-transplantation. In both univariate and multivariate analysis for factors associated with HGG, immune-modulating factors such as rejection treatment with IVIG and baseline HGG were significantly associated with HGG ( Table 3 ). The single most associated factor related to HGG was presence of baseline HGG (OR 23.16, 95% CI: 7.508–71.461).

In the time-dependent Cox model analysis, the unadjusted analysis showed no significant association between HGG and infection. However, in the adjusted model with stabilized weights related to immunosuppression, HGG was significantly associated with infection. If HGG was present, the hazard ratio (HR) for hospitalization due to infection and graft-related viral infection was 1.895 (95% CI: 1.871–1.920; p< 0.001) and 1.152 (95% CI: 1.144–1.160; p< 0.001), respectively ( Table 4 ).