Association between hypogammaglobulinaemia and severe infections during induction therapy in ANCA-associated vasculitis: from J-CANVAS study

Abstract

Introduction

Antineutrophil cytoplasmic antibody-associated vasculitis (AAV), including microscopic polyangiitis (MPA), granulomatosis with polyangiitis (GPA) and eosinophilic granulomatosis with polyangiitis (EGPA), is a systemic vasculitis with elevated ANCA that causes inflammation primarily in small and medium blood vessels with the development of various organ manifestations [1]. Although the prognosis has improved dramatically with the establishment of remission-induction therapy with CYC and rituximab (RTX), the mortality rate remains high compared with that of healthy controls [2]. The 1-year mortality rate is 11–12%, and infection is the main cause of death (48–50% of deaths). [3, 4]. Reported risk factors for severe infections include age, abnormal kidney function and high disease activity [5–9].

Serum IgG is an indicator of humoral immunity for protection against infection and is known to be decreased during the treatment of autoimmune rheumatic diseases (AIRDs), including AAV, especially when RTX is used [10, 11]. Particularly in AAV, the frequency of low IgG after RTX is high compared with that in RA and systemic lupus erythematosus [11–13]. This could be attributed to differences in overall treatment intensity, including glucocorticoids and RTX re-treatment strategies [14, 15], as well as differences in disease-specific patient background and immunopathology [16]. Previous studies on AIRDs reported that low IgG before or after RTX use increased the risk of developing severe infections [15, 17]. Therefore, the EULAR and European Renal Association-European Dialysis and Transplant Association (EULAR/ERA-EDTA) recommend serum immunoglobulin measurement prior to each course of RTX and in patients with recurrent infection [18]. In addition, the ACR/Vasculitis (ACR) recommends immunoglobulin supplementation for MPA/GPA patients receiving remission maintenance therapy with RTX who have hypogammaglobulinaemia (serum IgG <300 mg/dl) and recurrent severe infections [19]. However, RTX is not the only agent that causes a decrease in serum IgG levels. The use of glucocorticoids, CYC and other immunosuppressive drugs causes secondary hypogammaglobulinaemia [14, 15, 20]. Nevertheless, it is not clear whether a decrease in serum IgG affects the development of severe infections in patients using drugs other than RTX.

In this study, we hypothesized that a decrease in serum IgG levels during the remission-induction period in patients with AAV is associated with the development of severe infections, regardless of treatment regimens. To address this, we conducted a study using data from a nationwide registry in Japan.

Materials and methods

Study design and patients

This multicentre retrospective cohort study was conducted using data from the Japan Collaborative Registry of ANCA-Associated Vasculitis (J-CANVAS), a nationwide registry established by 24 referral sites in Japan. The registry enrolled adult (≥20 years) patients with AAV who were newly diagnosed or who had severe relapse from January 2017 to June 2020. All patients were classified as MPA, GPA or EGPA based on the definitions of the 2012 International Chapel Hill Consensus Conference [1] and the European Medicines Agency algorithm [21]. The follow-up period for individual patients was from disease onset to death, loss to follow-up, or June 2021. Patients who dropped out or developed a severe infection on the first day of enrolment in the registry were excluded from the study.

Data collection

All clinical information was collected retrospectively at each medical site by referring to clinical records. The baseline characteristics prior to treatment initiation or enhancement, including patient background [age, sex, disease type (MPA/GPA/EGPA), relapse case (vs newly diagnosed), body weight, height, smoking and comorbidity], laboratory tests [serum albumin, serum creatinine, blood count (neutrophil, lymphocyte, haemoglobin), serum C-reactive protein, serum IgG, ANCA serotype (MPO-ANCA/PR3-ANCA)] and organ involvement [Birmingham vasculitis activity score (BVAS) 3.0 [22], interstitial lung disease] were collected. The treatment details [glucocorticoid dose (in prednisone equivalents), methylprednisolone pulse, RTX, CYC, plasma exchange, high-dose intravenous immunoglobulin, mycophenolate, methotrexate, azathioprine, mizoribine, trimethoprim-sulfamethoxazole prophylaxis] used during the 24 weeks after treatment (remission-induction period) and the minimum serum IgG levels during the same period were recorded. Severe infections and deaths during the 48 weeks after treatment were recorded as outcomes. In this registry, no information was collected on the site of infection and the pathogen. All collected data were recorded and integrated using an electronic data capture system, Viedoc (PCG solutions, Uppsala, Sweden).

Exposure

The main exposure was hypogammaglobulinaemia (low IgG) during remission-induction period defined as a minimum serum IgG levels of <500 mg/dl up to 24 weeks after treatment initiation, according to previous studies in primary immunodeficiency [23, 24]. Serum IgG concentrations were measured in the laboratories of each institution by turbidimetric immunoassay.

Outcomes

The primary outcomes were the incidences of first severe infections defined as ‘infections requiring hospitalization or prolongation of existing hospitalization’ [25] up to 48 weeks after treatment initiation.

Statistical analysis

All data were summarized as the median [interquartile range (IQR)] or absolute numbers (percentage). The cumulative incidence function (CIF) was used to describe the incidence of severe infections considering death as a competing risk. Cases that were lost to follow-up prior to the incidence of an outcome within 48 weeks were defined as censored, and cases with no outcome at 48 weeks were also censored (right censoring) at that time. The Fine–Gray proportional subdistribution hazard model [26] was used to estimate the association between low IgG and severe infection, adjusted for potential confounding factors. In selecting these confounders, we created a directed acyclic graph (DAG) to structure the relationships between each variable (Supplementary Fig. S1, available at Rheumatology online). These variables were selected based on those previously reported to be involved in severe infections [5–9], those included in typical treatment regimens, and those considered part of the general patient background. Based on the DAG, multivariate analyses were performed by several different models that considered different scenarios for bias mechanisms {adjusted variables for each model: model 1 [age, serum creatinine, BVAS 3.0]; model 2 [female, body mass index (BMI), disease type (GPA/MPA/EGPA), relapse case, current smoking in addition to the variables in model 1]; model 3 [diabetes mellitus, chronic bronchitis or bronchiectasis, serum albumin, interstitial lung disease in addition to the variables in model 2]; and model 4 [intravenous methylprednisolone pulse therapy, prednisone initial dose, CYC, RTX in addition to the variables in model 3]}. From these several models, we selected ‘model 4’ as our main analysis, which adjusts for maximum confounding. The magnitude of an association was described using point estimates of subdistribution hazard ratios (HRs) and 95% CIs. In addition, to check for non-linear associations between severe infections and serum IgG, confounding-adjusted HRs were estimated by the restricted cubic splines (RCS) functions using the Fine–Gray model [27]. The knot for analysing the association was set to 3 (placed on the 10th, 50th and 90th percentile of the predictor value range). Finally, subgroup analyses were performed to examine the interaction between treatment (intravenous methylprednisolone pulse therapy, RTX and CYC) and low IgG, and the interaction was evaluated by the Wald test. Because several variables, including minimum serum IgG, were found to be missing, multiple imputations by chained equation [28] were used to estimate the data for these missing values under the assumption of missing at random. Multiple imputations were performed for all explanatory variables used in the multivariate analysis, and 100 multiply imputed datasets were created [29]. The above-mentioned analyses based on the Fine–Gray model (estimation of HRs, RCS analysis and subgroup analysis) were performed on imputed data and also on complete case data as a sensitivity analysis. Patient characteristics, the univariate CIF, and prednisone dosage over time were described only for complete cases. All these analyses were performed using STATA version 17 (Stata Corp, College Station, TX, USA) and R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria). For the creation of the DAG, we used the DAGitty web application (version 3.0) [30].

Ethics

This study complied with the Helsinki Declaration. Because of the retrospective study design, written informed consent for each patient was omitted. We used only anonymized data and provided an opportunity for patients to opt out. The ethics committees of the Kyoto Prefectural University of Medicine and the other participating institutions approved the protocol (approval no. ERB-C-1928).

Results

Of the 662 Japanese patients with AAV, one patient who was censored at day 0 after treatment and four patients who developed severe infections at day 0 after treatment were excluded. Finally, 657 AAV patients were included in the study (392 with MPA, 139 with GPA and 126 with EGPA; 554 were new cases and 103 were relapsing cases) (Supplementary Fig. S2, available at Rheumatology online). Patient characteristics are presented in Table 1. Of the 657 patients enrolled, the minimum IgG levels up to 24 weeks after treatment initiation was measured in 510 patients, of whom 77 (15.1%) had low IgG (<500 mg/dl). While in the remaining 147 patients (22.4%), the minimum IgG levels were missing. The missing numbers of other variables are presented in Supplementary Table S1, available at Rheumatology online. The distribution of the minimum serum IgG levels up to 24 weeks after treatment initiation is shown in Supplementary Fig. S3, available at Rheumatology online, and the proportions of the minimum serum IgG levels in each treatment subgroup are shown in Supplementary Fig. S4, available at Rheumatology online. The differences in characteristics between patients with and without low IgG are described in Table 1. By 48 weeks after treatment, 657 patients had been followed for a median (IQR) of 336 (125–336) days, and 111 (16.9%) had developed severe infections. The median time to onset of first severe infection was78.5 (31–126) days with low IgG and 90.5 (27.5–163) days without low IgG. All-cause death occurred in 37 (5.6%) patients up to 48 weeks, of which 16 (43.2%) were related to infection.

Association between low IgG and severe infections

The cumulative incidence function is shown in Fig. 1; patients with low IgG had a shorter time to severe infections compared with those without low IgG.

Figure 1.

Incidence of severe infections with and without low IgG. The cumulative incidence function of severe infections was compared between patients with and without low IgG (<500 mg/dL) for complete cases (n = 510). ^aLow IgG: minimum serum IgG less than 500 mg/dL up to 24 weeks after treatment

Open in new tabDownload slide

The failure time analyses using the Fine–Gray model were performed after multiple imputations for missing data. The HR of low IgG to severe infections was 2.22 (95% CI: 1.42, 3.49) in the univariate analysis. In the multivariate models, the HR was 1.75 (95% CI: 1.03, 3.00) for model 4. The results of other models are also included in Table 2, and the full version of the results for variables other than serum IgG is shown in Supplementary Table S2, available at Rheumatology online. For the sensitivity analysis, the same analysis was performed for complete cases and the results were consistent with these (Supplementary Table S3, available at Rheumatology online).

Non-linear association between minimum serum IgG and severe infections

The results of the RCS analysis are shown in Fig. 2. The HRs tended to be higher as the IgG levels decreased towards the lower range, whereas the HRs were higher as the IgG levels increased above a certain value, indicating a U-shaped association. The lowest HR was observed at a minimum serum IgG of 946 mg/dl.

Figure 2.

Non-linear association of minimum serum IgG and the risk of severe infections. This figure shows the non-linear association between minimum serum IgG levels up to 24 weeks after treatment and severe infection (adjusted hazard ratios and 95% CI) by restricted cubic spline model (three knots). The adjusted hazard ratios and 95% CIs were estimated by the Fine–Gray model after multiple imputations using 17 variables as explanatory variables as shown for model 4 in Table 2. The serum IgG reference value was set at 500 mg/dL. The adjusted hazard ratios are shown as a solid line and the 95% CIs are shown as dashed lines

Open in new tabDownload slide

Subgroup analyses by treatment

Supplementary Fig. S4, available at Rheumatology online shows the proportions of low IgG in each treatment subgroup. The proportions were higher in the groups that received RTX or CYC compared with the proportion in the group that was neither RTX nor CYC. The same multivariate analysis (model 4 performed for each treatment group) had an HR of 2.61 (95% CI: 1.08, 6.31) for RTX users and 1.27 (95% CI: 0.59, 2.70) for RTX non-users. The P-value for the interaction between RTX and low IgG was 0.199. Similar analyses performed in subgroups with and without CYC, with and without methylprednisolone pulse therapy, and with and without RTX or CYC, showed no apparent evidence of the interactions (Fig. 3).

Figure 3.

Subgroup analysis in each treatment on the association between low IgG and severe infections. RTX: rituximab; mPSL: methylprednisolone. In treatment-specific subgroups, the adjusted hazard ratios of low IgG for severe infections were estimated by the Fine–Gray model after multiple imputations using 17 variables as explanatory variables as shown for model 4 in Table 2. ^aOne case with missing RTX and CYC data was excluded from the subgrouping

Open in new tabDownload slide

Prednisone dosage at different time points with and without low IgG

The prednisone dosage in patients with and without low IgG at each time point from 0 to 24 weeks is described in Fig. 4. Patients with low IgG tended to receive higher prednisone doses than those without low IgG.

Figure 4.

Prednisone dosage at different time points with and without low IgG. Prednisone dosages over time were compared between patients with and without low IgG (<500 mg/dL) for complete cases. Vertical boxes indicate the quartile range (25th—75th percentile), and vertical lines indicate the 1.5 quartile range. Outliers are shown as circles. ^aLow IgG: minimum serum IgG less than 500 mg/dL up to 24 weeks after treatment

Open in new tabDownload slide

Discussion

In this study, we studied the association between low IgG levels (<500 mg/dl) up to 24 weeks after remission-induction therapy and severe infections up to 48 weeks after treatment initiation using data from the J-CANVAS registry. The results showed that low IgG was independently associated with the incidence of severe infections. In addition, a non-linear association between the minimum serum IgG levels and severe infections was expressed by the RCS functions, which indicated a U-shaped association. Moreover, there was no apparent heterogeneity in the effects of low IgG on severe infections between treatments including RTX, CYC and methylprednisolone pulse therapy.

The association between low IgG and the incidence of severe infections demonstrated by this study has also been studied with patients receiving RTX in AAV [31] and other AIRDs [17]. Although these studies focused on the baseline values of IgG, Md Yusof et al. reported that low IgG levels (<600 mg/dl) before and after RTX administration were associated with the development of severe infections in AIRDs [15]. In addition, Podestà et al. reported that patients who required intravenous antibiotic therapy by 6 months after treatment initiation had significantly lower serum IgG levels at that 6-month time point [32]. These previous studies only included RTX-treated patients; however, other drugs (e.g. glucocorticoids and CYC) also cause a decrease in B-cell counts [33] and serum IgG levels [34]. In the present study, low IgG occurred to the same extent in patients receiving CYC or RTX, which was consistent with previous clinical trials [34], and moreover, low IgG also occurred in patients who did not receive RTX and CYC. We found that decreased serum IgG after remission-induction therapy was associated with severe infections, regardless of whether the treatment regimen included RTX. Overall, our findings confirm and generalize those of previous studies to all AAV patients in the remission-induction period.

The following mechanisms may be involved in the association between low IgG and the development of severe infections: immunoglobulins, glycoproteins produced as soluble molecules by plasma cells, have a major role in host defence against extracellular infections through neutralization, complement activation, opsonization and antibody-dependent-cell-mediated cytotoxicity [35], and hypogammaglobulinaemia is thought to result in the inadequacy of these functions leading to the increased risk of severe infection. It is also possible to consider that low IgG may not only be directly related to susceptibility to infection but may also be an indirect indicator reflecting the degree of immunosuppression induced by the sequence of remission induction therapy.

Using RCS in this study, we found a U-shaped association between IgG and severe infections with an increased risk trend also observed at higher IgG levels above 946 mg/dl. One explanation for this trend may be due to the influence of residual disease activity. In AAV, disease activity has been reported to be associated with the development of severe infections [6], and in RA, hypogammaglobulinaemia after RTX administration is reported to be an indicator to good treatment response [12]. Accordingly, we speculated that patients with higher minimal serum IgG levels had residual disease activity due to inadequate treatment intensity, which may have led to an association with severe infections.

To determine whether there was any treatment-specific influence on the association between low IgG and severe infections, we examined the effects and interactions in subgroups of each treatment. However, no obvious heterogeneity in the effect of low IgG was observed between any of the treatments. Taken together with the results of the RCS, serum IgG after remission-induction therapy may reflect the degree of overall immunosuppressive status brought about by the balance between the intensity of remission-induction therapy and AAV disease activity. Notably, our patients with low IgG tended to receive a higher prednisone dosage than those without low IgG. Recent clinical trials reported that high-dose glucocorticoid treatment leads to a greater degree of IgG reduction than low-dose treatment [36]. Tieu et al. reported that glucocorticoid use at 12 months after treatment initiation was associated with moderate/severe hypogammaglobulinaemia in a study of patients with AIRDs including AAV [37]. Reducing glucocorticoid doses as early as possible in combination with RTX or CYC may help avoid low IgG and thereby reduce the incidence of severe infections.

The first strengths of this study were the treatment-independent association between low IgG and severe infections in patients with AAV based on a nationwide large-scale cohort. Second, we considered competing risks for severe infections and visualized the non-linear association between serum IgG levels and severe infections using RCS. This is because no studies have used the RCS in secondary hypogammaglobulinaemia, and there have been no large-scale reports analysing factors involved in the development of severe infections while considering competing risks.

This study had several limitations. First, because not all IgG measurements were collected, it is possible that in some cases the timing of the onset of severe infection (up to 48 weeks after treatment initiation) could preceded the date of low IgG (<500 mg/dl up to 24 weeks after treatment). However, because serum IgG levels are generally not decreased by infection, we think that reverse causality is unlikely. Further studies are needed to determine whether low IgG is a predictor of severe infections. Second, the number of cases in the subgroup analysis might be insufficient. In this subgroup analysis, we concluded that there was no obvious interaction between treatment and hypogammaglobulinaemia, although some uncertainty remains. Considering these limitations, further studies to evaluate the association between IgG and severe infections in more detail, including changes in IgG over time after treatment initiation, would be beneficial.

In conclusion, this study suggests that a decrease in serum IgG is a useful indicator of the immunosuppressive status of patients during the remission-induction period regardless of the treatment regimen. Therefore, the routine monitoring of serum IgG can provide an opportunity to optimize treatments (e.g. glucocorticoid reduction) to reduce severe infections in all AAV patients.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

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

Funding

This work was supported by the Japan Society for the Promotion of Science KAKENHI (Grant No. JP22K16348).

Disclosure statement: S.O. has received honoraria for speakers bureaus from Asahi Kasei Pharma and Chugai. T.Ki. has received honoraria for speakers bureaus from Asahi Kasei Pharma and Chugai. H.N. has received a consulting fee from Kyowa Kirin. D.N. has received scholarship grants from Asahi Kasei Pharma, Chugai and Kyowa Kirin. N.T. has received honoraria for speakers bureaus from Asahi Kasei Pharma, Chugai, Kyowa Kirin, Mylan, Pfizer and Towa Pharmaceutical. A.N. has received honoraria for speakers bureaus from Asahi Kasei Pharma and Chugai. S.H. has received research grants from Asahi Kasei Pharma, Chugai, Pfizer and Shionogi, and honoraria for speakers bureaus from Asahi Kasei Pharma, Chugai, Kyorin and Pfizer. K.M. has received research grants from Asahi Kasei Pharma and Chugai. T.Take. has received research grants from Asahi Kasei Pharma and Chugai, and honoraria for speakers bureaus from Asahi Kasei Pharma, Chugai, Kyowa Kirin, Pfizer and Takeda. M.Kat. has received honoraria for speakers bureaus from Asahi Kasei Pharma, Chugai, Kyowa Kirin and Pfizer. R.O. has received a research grant from Asahi Kasei Pharma. T.Taka. has received scholarship grants from Chugai and Kyowa Kirin, and honoraria for speakers bureaus from Chugai and Kyowa Kirin. T.Ka. has received honoraria for speakers bureaus from Chugai. W.F. has received honoraria for speakers bureaus from Asahi Kasei Pharma, Chugai and Pfizer. Y.Ka. has received research grants from Asahi Kasei Pharma and Chugai, and honoraria for speakers bureaus from Asahi Kasei Pharma, Chugai Pharmaceutical and Mylan. The other authors declare that they have no conflicts of interest to disclose.

Acknowledgements

We thank J. Ludovic Croxford, PhD, from Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript. This work was supported by J-STAR-CR, the Committee on Clinical Research, Japan College of Rheumatology. This work was presented as a workshop at the 66th Annual General Assembly and Scientific Meeting of the Japan College of Rheumatology.

References