Glomerulonephritis following COVID-19 infection or vaccination: a multicenter study in South Korea

Study design

This study addresses three main issues. We identified all de novo GN cases with suspected onset after COVID-19 infection or vaccination from multiple centers (Fig. 1A). Second, we determined the COVID-19–related histories among patients who were initially not implicated by identifying close temporal relationships in a single center (Fig. 1B). Finally, we examined the changes in biopsies during the COVID-19 vaccination or pandemic era. The histopathological review included the use of light microscopy with periodic acid-Schiff, trichrome, silver, and hematoxylin and eosin staining, immunofluorescence microscopy for immunoglobulins and complements, and electron microscopy. The biopsies for the transplanted kidney were excluded. The detailed methods are provided below.

Figure 1.

Study flow.

(A) Multicenter and (B) single-center analysis.

COVID-19, coronavirus disease 2019; GN, glomerulonephritis; HSP, Henoch-Schönlein purpura; IgAN, immunoglobulin A nephropathy.

De novo glomerulonephritis cases after COVID-19 infection or vaccination: multicenter analysis

Data were retrospectively collected from multiple centers (Severance Hospital, Yongin Severance Hospital, Wonju Severance Christian Hospital, Ilsan Paik Hospital, and Busan Paik Hospital) between April 2021 and March 2023 using a de-identified electronic documents (Fig. 1A). Patients with a clinically suspected association between COVID-19 events and GNs were included. The clinical onset of GNs in relation to COVID-19 events was evident in collected patients, most of whom had latent period of no more than 50 days after being infected or vaccinated, although the exact latent period remains unknown in some patients. Renal pathology experts thoroughly reviewed all biopsies, and relevant laboratory findings were extracted from electronic medical records. Notably, two cases demonstrating temporal associations with COVID-19 vaccines, lupus nephritis and pauci-immune crescentic GN, have been reported previously [8,17].

Investigation of COVID-19–related history in patients with biopsy-proven kidney disease in a single center

This study focused on elucidating the history of COVID-19 vaccination and infection in patients with GN. We identified COVID-19–related histories among patients who were initially not suspected of having an association with COVID-19 events. To assemble this subset of previously unsuspected biopsies, we excluded those analyzed in the multicenter series and those taken before April 2021 when the nationwide vaccination program commenced. Therefore, we meticulously collected vaccine type, vaccination date, and infection date data from patients who underwent native kidney biopsies at Severance Hospital between April 2021 and December 2022 (Fig. 1B). The investigation was conducted during outpatient visits, and comprehensive information was gathered by reviewing electronic medical records.

Concerning patients with included immunoglobulin A nephropathy (IgAN) or Henoch-Schönlein purpura nephritis (HSP), we conducted an additional investigation into the time of onset. Utilizing the same timeframe (50 days) as the multicenter case series, we identified IgAN and HSP cases that could be temporally linked to vaccination.

Changes in glomerulonephritis diagnoses during the COVID-19 pandemic

We analyzed the changes in GN diagnoses throughout the prepandemic and pandemic periods (2016–2022) at Severance Hospital. For statistical purposes, the biopsies were broadly categorized into 10 groups, including IgAN/HSP, podocytopathy (comprising minimal change disease [MCD] and focal segmental glomerulosclerosis [FSGS]), immune complex-mediated GN (encompassing lupus nephritis, C3 GN, and membranoproliferative GN), membranous nephropathy, other nephropathies (e.g., diabetes mellitus nephropathy, hypertensive nephropathy, or paraprotein-related renal disease), pauci-immune crescentic GN (associated with anti-neutrophil cytoplasm antibodies), acute interstitial nephritis (AIN), acute tubular injury (ATI), thrombotic microangiopathy (TMA), or nonspecific histology. Secondary FSGS cases attributed to clear provocative lesions (e.g., hypertensive nephropathy) were classified accordingly. Biopsies with insufficient glomeruli for a definitive diagnosis were excluded, and the nonspecific histology category was applied to cases with near-normal glomeruli, tubulointerstitium, and blood vessels.

Statistical analyses

Continuous variables are expressed as means with standard deviations (SD) for normally distributed data and medians with interquartile ranges (IQR) for non-normally distributed data, whereas categorical variables are expressed as numbers and proportions. To evaluate the causal effects of COVID-19 vaccination and infection on de novo GNs, a Bayesian structural time-series model was performed using R ‘CausalImpact’ package [18]. The causal effect was determined by assessing the disparity between predicted and observed outcomes. This evaluation helped establish the difference in the proportion of newly diagnosed GN cases at Severance Hospital before and after the commencement of the nationwide vaccination program (April 1, 2021), and before and after the peak of the COVID-19 pandemic (February 1, 2022) in South Korea. This approach considered that the nationwide vaccine program began in March 2021, and COVID-19 cases surged significantly as of February 2022 in South Korea, considering weeks of latent period [1,2,4,14]. An additional time-series analysis was conducted using an autoregressive integrated moving average (ARIMA) model. The Box-Cox transformation was performed for non-normally distributed time-series data. The auto.arima function in R ‘forecast’ package was used to automatically select an optimal model [19]. The Ljung-Box Q statistic was used to evaluate the goodness of fit of the automatically selected ARIMA models. The p-value for Ljung-Box Q statistics > 0.05 indicates that the selected ARIMA model fits well and no autocorrelation exists in the residual. Both step and slope changes following the interventions (i.e., COVID-19 vaccination or infection) were examined. All statistical analyses were performed using R (version 4.2.3; R Foundation for Statistical Computing, http://www.r-project.org) with a significance level of p < 0.05.

Clinical characteristics of patients with glomerulonephritis cases following COVID-19 infection

Clinical information from seven patients whose GN was suspected of developing following COVID-19 was collected in a multicenter study. The age at renal biopsy ranged from 24 to 76 years. The pathological diagnoses included ATI (14.3%), TMA with hypertensive nephropathy (14.3%), FSGS, collapsing variant (14.3%), MCD (28.6%), IgAN (14.3%), and HSP (14.3%). Patients with FSGS and IgAN had concurrent ATI. AKI with or without hematuria and proteinuria was the predominant clinical manifestation in three patients (42.9%). These symptoms were observed between 10 and 35 days after infection (Table 1).

Table 1.

Clinical characteristics of patients from multicenter with glomerulonephritis following COVID-19 infection

Clinical characteristics of patients with glomerulonephritis cases following COVID-19 vaccination

Nineteen patients in whom GN was suspected after COVID-19 vaccination were recruited from multiple centers. The collected GNs were composed of IgAN (n = 11, 57.9%), HSP (n = 3, 15.8%), FSGS (n = 1, 5.3%), , immunoglobulin M nephropathy (n = 1, 5.3%), lupus nephritis (n = 1, 5.3%), pauci-immune crescentic GN (n = 1, 5.3%), and AIN (n = 1, 5.3%). The age at the time of renal biopsy ranged from 18 to 70 years, and the main symptoms were proteinuria, hematuria, edema, and AKI. Both messenger RNA (mRNA)-based vaccines (BNT162b2, n = 8; mRNA-1273, n = 3) and adenovirus-based vaccines (ChAdOx1 nCoV-19, n = 2; Ad26.COV.S, n = 1) were administered. Except for one patient with HSP who developed 32 days after the first dose of ChAdOx1 nCoV-19 and another patient with AIN who developed 7 days after the first dose of BNT162b2, the initial symptoms developed after the second or third dose. The time interval between the first vaccination and symptoms varied from 1 to 50 days. Seven patients experienced symptoms within a short time, even within 1 week after vaccination, while others had a more extended interval. Some patients underwent kidney biopsy more than a year after symptom onset, whereas those who showed acute symptoms underwent kidney biopsy within 1 to 2 months. Most patients received steroids or other immunosuppressive therapies. However, progression to end-stage kidney disease (ESKD) was not observed among the patients during follow-up (Table 2).

Table 2.

Clinical characteristics of patients from multicenter with GN following COVID-19 vaccination

Association with COVID-19 or vaccination in unsuspected kidney biopsies

From April 2021 to December 2022 at Severance Hospital, 432 patients underwent native kidney biopsies, and 12 patients (2.8%) were included in the multicenter series of GNs following COVID-19 or vaccination. Among the remaining 420 patients whose connection to COVID-19–related events was initially unknown, a survey was successfully conducted to investigate COVID-19 infection or vaccination history in 143 patients.

Among 143 patients, 29 (20.3%) had preceding COVID-19 infection. Renal biopsy was conducted on an average of 99.0 days (SD, 62.1 days) after infection, and the most common histological diagnosis was IgAN/HSP. Twenty-seven patients (93.1%) also received vaccinations before tissue examination. The most commonly administered vaccine was BNT162b2. The average interval between vaccination and kidney biopsy was 347 days (Supplementary Table 1, available online).

Eighty-three patients (58.0%, 83 of 143) without a history of COVID-19 received vaccination before biopsy. The BNT162b2 vaccine was the most commonly used vaccine. The median days from the first vaccination to the kidney biopsy was 140.0 (IRQ, 77.0–272.0) (Supplementary Table 2, available online). In contrast, 24 individuals (16.8%, 24 of 143) had no history of COVID-19 infection or vaccination before biopsy. There was no significant difference in demographics or pathological diagnoses between the groups with or without a vaccination history, and IgAN/HSP was the most common diagnosis in both groups (Supplementary Table 2, available online).

Because IgAN and HSP were the leading causes of postvaccination GNs in multiple centers, we further focused on postvaccination IgAN/HSP (latent period, <50 days) among the patients with a prior vaccination history by analyzing the duration to symptomatic onset. Nine patients with IgAN/HSP (two patients with a history of IgAN) who developed symptoms within 50 days postvaccination and did not have any history of COVID-19 infection were selected. Seven (77.8%) were vaccinated with BNT162b2, and the other two (22.2%) were vaccinated with Ad26.COV.S. Five patients (55.6%) had their first vaccination before onset, whereas one (11.1%) and two (22.2%) had the second or third doses, respectively. The symptoms typically include hematuria and proteinuria. One patient presented AKI. Two patients received immunosuppressive treatment, and the remaining patients received conservative care for proteinuria. None of the patients developed ESKD (Table 3).

Table 3.

Clinical characteristics of patients from a single center with postvaccination IgAN and HSP

Biopsy diagnosis of glomerulonephritis cases during the COVID-19 pandemic era in South Korea

To examine the changes in GN diagnoses during the COVID-19 pandemic, Bayesian structural time-series models and ARIMA models were applied to analyze patients with kidney biopsies from a single institution. There was no significant difference in the frequency of kidney biopsies before and after the initiation of the nationwide vaccine program or the peak of the COVID-19 pandemic (Supplementary Fig. 1, available online). Bayesian structural time-series models showed that compared with the pre-nationwide vaccine program period, the monthly proportion of AIN, nonspecific histology, podocytopathy, and pauci-immune GN were changed as follows: the proportion of AIN, nonspecific histology and podocytopathy increased by 0.96 (95% confidence interval [CI], –0.11 to 2.00), 2.51 (95% CI, 0.37–4.60), and 5.12 (95% CI, 0.80–9.30), respectively. In contrast, the proportion of pauci-immune GN decreased by –3.71 (95% CI, –0.64 to –1.00) (Fig. 2). Based on the analysis, we found a posterior probability of a causal effect ≥ 95%, indicating a strong causal relationship in the AIN, nonspecific histology, podocytopathy, and pauci-immune GN categories (Supplementary Table 3, available online). Other diseases showed no differences in diagnosis rates before and after the nationwide vaccination program (posterior probability, <95%) (Supplementary Table 3, available online; Fig. 2). Supplementary Fig. 2 (available online) and Supplementary Table 4 (available online) present the predictions of the ARIMA model. Other nephropathy and podocytopathy entities showed significant step changes in the diagnostic rates: –1.627 (95% CI, –2.904 to –0.349) and 0.827 (95% CI, 0.168–1.485), respectively. AIN showed a significant slope change in the diagnosis rate (0.049 [95% CI, 0.013–0.084]). However, no disease was significant for either step or slope changes.

Figure 2.

The monthly trend of glomerulonephritis after the initiation of nationwide vaccine program (after April 2021) in Severance Hospital based on the Bayesian structural time-series models.

(A) Total, (B) acute tubular injury (AIN), (C) acute interstitial nephritis (ATI), (D) immunoglobulin A nephropathy (IgAN), (E) membranous nephropathy (MN), (F) nonspecific histology, (G) other nephropathies, (H) immune complex-mediated glomerulonephritis (IC-GN), (I) pauci-immune crescentic glomerulonephritis, (J) podocytopathy, and (K) thrombotic microangiopathy (TMA). The solid lines denote the observed value, the blue-dotted lines denote the predicted value, and the blue shades denote 95% confidence intervals of the predicted value.

Compared with the prepandemic period, only the monthly proportion of AIN was significantly changed in a Bayesian structural time-series model (absolute effect, 3.09 [95% CI, 1.70 to 4.50]; posterior probability of a causal effect, ≥95%) (Supplementary Table 5 and Supplementary Fig. 3; available online). In the ARIMA model, no disease showed a significant difference in either step change or slope change in diagnosis rates (Supplementary Table 6, Supplementary Fig. 4; available online).