Kidney outcome after mild to moderate COVID-19

ABSTRACT

Background

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has a remarkable kidney tropism. While kidney effects are common in severe coronavirus disease 2019 (COVID-19), data on non-severe courses are limited. Here we provide a multilevel analysis of kidney outcomes after non-severe COVID-19 to test for eventual kidney sequela.

Methods

This cross-sectional study investigates individuals after COVID-19 and matched controls recruited from the Hamburg City Health Study (HCHS) and its COVID-19 program. The HCHS is a prospective population-based cohort study within the city of Hamburg, Germany. During the COVID-19 pandemic the study additionally recruited subjects after polymerase chain reaction–confirmed SARS-CoV-2 infections. Matching was performed by age, sex and education. Main outcomes were estimated glomerular filtration rate (eGFR), albuminuria, Dickkopf3, haematuria and pyuria.

Results

A total of 443 subjects in a median of 9 months after non-severe COVID-19 were compared with 1328 non-COVID-19 subjects. The mean eGFR was mildly lower in post-COVID-19 than non-COVID-19 subjects, even after adjusting for known risk factors {β = −1.84 [95% confidence interval (CI) −3.16 to −0.52]}. However, chronic kidney disease [odds ratio (OR) 0.90 (95% CI 0.48–1.66)] or severely increased albuminuria [OR 0.76 (95% CI 0.49–1.09)] equally occurred in post-COVID-19 and non-COVID-19 subjects. Haematuria, pyuria and proteinuria were also similar between the two cohorts, suggesting no ongoing kidney injury after non-severe COVID-19. Further, Dickkopf3 was not increased in the post-COVID-19 cohort, indicating no systematic risk for ongoing GFR decline [β = −72.19 (95% CI −130.0 to −14.4)].

Conclusion

While mean eGFR was slightly lower in subjects after non-severe COVID-19, there was no evidence for ongoing or progressive kidney sequela.

Graphical Abstract

Open in new tabDownload slide

INTRODUCTION

Coronavirus disease 2019 (COVID-19) is associated with a wide range of symptoms and extrapulmonary organ dysfunctions [1]. Post-mortem studies have demonstrated that the underlying severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) not only invades the lung, but also possesses a remarkable tropism to kidney, liver and other organs [2]. This organ tropism is associated with infection-like cellular damage and a profibrotic molecular signature [3, 4]. The clinical relevance of such organ effects are illustrated by autopsy studies on patients with severe COVID-19 showing that the detection of virus particles in the kidney correlates with acute kidney injury (AKI) and patient outcome [5]. Indeed, kidney injury is common in the acute phase of severe COVID-19 and the high incidence exceeds that predicted by common risk factors [6].

Relevant organ dysfunctions are not restricted to the acute phase of severe COVID-19 but can also be evident after recovery [7]. Such sequelae have been observed for cardiovascular diseases as well as for chronic kidney disease (CKD) [8, 9]. However,s severe COVID-19 only accounts for a minority of SARS-CoV-2 infections, as most infections take mild or moderate courses [10]. Relevant organ impairment after non-severe COVID-19 could affect a large population with a tremendous impact on general and specialized healthcare. Thus a key question in the pandemic is whether there are sequelae in important organs such as the kidney after mild and moderate COVID-19.

Recently, we and others observed a minor decline in the estimated glomerular filtration rate (eGFR) after non-severe COVID-19 [9, 11]. These data raise the possibility of relevant kidney sequelae and require further deep analysis of kidney functions after non-severe COVID-19. To investigate whether there is are relevant kidney sequelae on a cross-sectional, population level after non-severe COVID-19 and to test the hypothesis of ongoing kidney injury, we report a multilevel analysis including kidney function, urine abnormalities correlating with COVID-19-related kidney effects and a high-sensitivity biomarker indicating progressive kidney disease in the recently described post-COVID-19 cohort of the Hamburg City Health Study (HCHS) [11].

MATERIALS AND METHODS

Study design and population

This is a cross-sectional analysis of kidney outcomes within the recently described HCHS post-COVID-19 program [11]. Briefly, the HCHS is a prospective population-based cohort study on randomly selected residents of the city of Hamburg, Germany, ages 45–74 years [12]. During the COVID-19 pandemic the study opened an additional program recruiting individuals of similar age with prior SARS-CoV-2 infections via public newspaper announcements as well as the COVID-19 test centre of the University Medical Center Hamburg-Eppendorf. SARS-CoV-2 infections had to be verified by polymerase chain reaction (PCR) at least 4 months before the study visit, which took place between March and December 2020. Individuals were included if SARS-CoV-2 infections were reported as mild (symptoms of a common cold or asymptomatic courses) or moderate (no need for intensive or intermediate care unit treatment). This post-COVID-19 cohort was matched by age, sex and education to historic controls from the main HCHS program. After identification from the official local inhabitant file, control subjects completed the study visit at a randomly chosen date. Subjects were eligible as controls for this analysis if the study visit occurred between the initiation of the study in 2016 and before 1 January 2020, as well as during the months March–December. This period was chosen to rule out prior COVID-19 or SARS-CoV-2 vaccination, as there was no COVID-19 case in Germany before January 2020 and SARS-CoV-2 vaccines were not available at that time [13]. The restriction to study visits performed between March and December was used to match the post-COVID-19 cohort and exclude seasonal effects. All participants provided informed consent and the study was performed in adherence to the Declaration of Helsinki. The HCHS, registered at ClinicalTrials.gov (NCT03934957), as well as its COVID-19 program were approved by the local ethics committee (PV5131).

Data collection

All participants underwent a detailed assessment of various organ systems and functions, including a general laboratory and urine workup, as well as establishment of a biobank as published previously [12]. Pre-existing comorbidities as presented in the baseline characteristics as well as information on the severity of SARS-CoV-2 infections were assessed by patient interviews. While creatinine, cystatin C and urine dipstick analysis were performed directly during the study visit, all other urinalyses were performed on samples frozen at −80°C. Urinary Dickkopf3 (DKK3) was measured at the Saarland University Medical Center with a commercially available DKK3 enzyme-linked immunosorbent assay (ELISA) (Refine; DiaRen UG, Homburg/Saar, Germany) according to the manufacturer's protocol. All other parameters were measured according to the routine clinical standard at the Institute of Clinical Chemistry and Laboratory Medicine of the University Medical Center Hamburg-Eppendorf, Germany, using Atellica Solution analysers and tests (Siemens Healthineers, Erlangen, Germany). The laboratory personnel performing the measurements were blinded with respect to the study results. Limits of detection for urinary albumin, protein and DKK3 were 3 mg/l, 60 mg/l and 30 pg/ml, respectively. The interassay coefficient of variation of the DKK3 ELISA is <4% for repeated measurements, as described previously [14]. eGFR was calculated with the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula [15].

Outcome measures

We defined five main renal outcomes: eGFR calculated with the combined CKD-EPI formula for creatinine and cystatin C [15], albuminuria quantified by immunonephelometry and presented as the urine albumin:creatinine ratio (UACR), the urinary biomarker DKK3 indexed to the urinary creatinine concentration (mg DKK3/g creatinine) as well as haematuria and pyuria assessed by urine dipstick. eGFR was chosen to study long-term effects on general kidney function and albuminuria as an important and independent risk factor for progressive kidney disease [16]. Haematuria and pyuria were analysed in a semiquantitative manner and analysed to test the hypothesis of persistent kidney involvement, as these parameters were found to be commonly affected during COVID-19-related kidney injury [17–19]. Urinary DKK3 was used to investigate the risk of ongoing decline in kidney function [20]. DKK3 is a profibrotic glycoprotein expressed in kidney tubular cells upon stress [21]. The urinary concentration of DKK3 correlates with and predicts future GFR decline independent of the degree and cause of pre-existing impairment in kidney function [14].

To test the hypothesis of relevant pathological conditions after non-severe COVID-19, the main outcomes were categorized into predefined categories. For eGFR and albuminuria, these categories correspond to the CKD stages defined by the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines [22]. The cut-off for urinary DKK3 was set to values shown to indicate future and progressive eGFR loss [20]. Creatinine and cystatin C with their corresponding sole eGFRs (eGFRcrea and eGFRcys), proteinuria quantified by the protein:creatinine ratio as well as urine pH and specific gravity detected semi-quantitively by urine dipstick were defined as secondary outcomes.

Statistical analysis

Continuous variables are presented as means and standard deviations (SDs), categorical variables as numbers and percentages. Baseline characteristics are presented as crude values and corrected for the weighting and clustering of the flexible matching design. Outcomes were investigated with mixed linear or logistic regression models as indicated. Results of regression analyses are reported as β or the odds ratio (OR) with 95% confidence intervals (95% CIs) for continuous and categorical variables, respectively. All regression models were controlled for a clustering of the matching parameters age, sex and education as random effects. Other covariates were included in the regression models as fixed effects as indicated. The chi-squared test was used to test for the trend of categorical variables between the cohorts in matching weighted analyses. The association of outcome parameters with SARS-CoV-2 antibody titres was investigated with linear regression models for continuous and boxplot analyses with chi-squared test for categorical variables. Missing values (see Supplementary Table 1 for frequencies of missing values per variable) were imputed using random forest-based imputations with missforest (version 1.4) [23]. Values below the limit of detection were set to half the limit of detection. To account for multiple testing, P-values were adjusted using the Bonferroni method by correcting for the five main outcomes. Statistical analyses were performed using R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Matching and baseline characteristics

The 443 subjects of the HCHS post-COVID-19 program were matched to 1328 non-COVID-19 controls, resulting in a well-balanced distribution of the matching parameters age, sex and education between the two cohorts (see Supplementary Table 2 for matching characteristics). Besides a crude description, baseline characteristics are presented after correction for the matching clustering to permit a valid comparison (Table 1). The two cohorts were similar in most anthropometric and social data as well as vital signs and comorbidities. However, hypertension was more common in post-COVID-19 (60.7%) compared with non-COVID-19 subjects (52.6% after correcting for the matching cluster). The same trend was observed for CKD (11.1% versus 9.3%) and diabetes (6.1% versus 4.3%). Current smoking was more common in the non-COVID-19 cohort (20.4 versus 7.4%). Despite the higher prevalence of hypertension in the post-COVID-19 cohort, this group was less likely to receive angiotensin-converting enzyme inhibitors (ACEIs) than the non-COVID-19 cohort (5.2% versus 9.0%). For the post-COVID-19 cohort, infections occurred in a mean 252 days before the study visit (SD 84.1, median 286 days). A total of 61.6% (253 individuals) had experienced a mild COVID-19 course and 38.4% had a moderate COVID-19 course. Only 7% (31 individuals) were hospitalized due to COVID-19, but no subject had been admitted to the intensive care unit. The mean antibody titres were 749.6 (SD 2668.9) binding antibody units/ml for anti-spike and 72.2 (SD 100) times the cut-off value for anti-nucleocapsid antibodies (see Petersen et al. [11] for detailed baseline characteristics of the post-COVID-19 cohort). SARS-CoV-2 vaccination was not comprehensively available at the time of the study and no individual was vaccinated before the visit.

Risk of relevant kidney pathologies after non-severe COVID-19

The main outcomes by predefined relevant disease conditions were compared with mixed logistic regression models between the two cohorts to investigate the hypothesis of relevant kidney changes following non-severe COVID-19 (Table 2). Even after controlling for the matching cluster and relevant confounding risk factors such as body mass index, pre-existing CKD, hypertension and diabetes as well as the use of angiotensin receptor blockers (ARBs) or ACEIs (Model 3), we found no increase for CKD, defined by an eGFR <60 ml/min/1.73 m², in individuals after non-severe COVID-19 [OR 0.90 (95% CI 0.48–1.66), adjusted P = 1.000] over individuals without prior COVID-19. Also, there was no change in severely increased albuminuria [UACR >300 mg/g; OR 0.76 (95% CI 0.49–1.09), adjusted P = .893] or urinary DKK3:creatinine ratio >470 mg/g_creatinine [OR 1.31 (95% CI 0.24–6.97), adjusted P = 1.000] between the two cohorts. Pyuria [OR 1.21 (95% CI 0.90–1.63), adjusted P = 1.000] and haematuria rates [OR 0.78 (95% CI 0.62–0.99), adjusted P = .217] were equally similar in post-COVID-19 and non-COVID-19 subjects. These effects were consistent in subgroup analyses of individuals with and without pre-existing CKD (Supplementary Table 3).

Effects on CKD stages and predefined DKK3 categories

We further divided the continuous main outcomes eGFR and albuminuria into CKD stages as defined by the KDIGO guidelines [22] to investigate the effects on mildly reduced kidney function and changes in the severity distribution of CKD (Table 3). A matching weighted comparison indicated a non-significant trend towards differences in the distribution of CKD stages between the two cohorts (adjusted P = .054), which were due to an increase in individuals with an eGFR of 60–90 ml/min/1.73 m² (43.8% versus 35.3% in post-COVID-19 and non-COVID-19 subjects, respectively). This was confirmed by regression analyses adjusting for additional risk factors (Supplementary Table 4). There was no difference in the incidence of individuals with advanced CKD stages between the post-COVID-19 and non-COVID-19 cohort.

Mildly increased UACR occurred in 4.5% in the post-COVID-19 cohort and 4.2% in the non-COVID-19 cohort, but severely increased UACR and levels >1000 mg/g_creatinine were not observed in any subjects of the two cohorts. Importantly, there was no difference in the distribution of UACR stages between post-COVID-19 and non-COVID-19 subjects (adjusted P = 1.000). Urinary DKK3:creatinine ratios were classified into predefined categories, which were previously shown to correlate with future eGFR loss (Table 3) [20]. Again there was no difference in the distribution over the categories of DKK3:creatinine ratios (adjusted P = 1.000) and only a few individuals had very high ratios >1000 mg/g_creatinine (2.7% in the post-COVID-19 cohort and 3.1% in the non-COVID-19 cohort). Regression analyses with adjustment for additional risk factors also revealed similar distributions of UACR and DKK3:creatinine ratio categories between the two cohorts (Supplementary Table 4).

Mean difference in renal function, albuminuria and DKK3 levels

Regarding general kidney function, we found a slightly lower mean eGFR calculated by a combination of creatinine and cystatin C in the post-COVID-19 cohort compared with the non-COVID-19 cohort (Table 4). This difference persisted after adjustment for the matching cluster and known CKD risk factors [β = −1.84 (95% CI −3.16 to −0.52), adjusted P = .032]. In contrast, we found a trend towards even lower urinary DKK3 concentrations indexed to urinary creatinine in post-COVID-19 subjects [β = −72.19 (95% CI −130 to −14.4), adjusted P = .072]. Levels for UACR were equal between the two groups [β = −0.95 (95% CI −5.32–3.43), adjusted P = 1.000].

These findings also applied to subgroup analyses of individuals without pre-existing CKD (Supplementary Table 5). Subgroup analyses in individuals with pre-existing CKD were limited by the small number of individuals and showed no significant differences. None of the outcome parameters showed a significant correlation with SARS-CoV-2 anti-spike or anti-nucleocapsid antibodies (Supplementary Figs. S1 and S2).

Regarding the secondary outcomes (Table 4), creatinine [β = 0.02 (95% CI 0.01–0.03), adjusted P = .007] and its derived eGFR [β = −2.02 (95% CI −3.39 to −0.65), adjusted P = .020] also showed mild kidney impairment in the post-COVID-19 cohort. Although not statistically significant, a similar trend was observed for cystatin C [β = 0.01 (95% CI 0–0.03), adjusted P = .325] and its corresponding eGFR [β = −1.32 (95% CI −2.76–0.12), adjusted P = .367]. Proteinuria was equal between cohorts [β = −4.14 (95% CI −12.75–4.47), adjusted P = 1.000].

Categorized analysis of urinary dipstick findings

To further analyse the severity in the primary outcomes haematuria and pyuria, the urinary dipstick findings were analysed in semi-quantitative categories between the post- and non-COVID-19 cohorts (Table 5). While low-grade haematuria (up to 10 cells/µl) was common in both cohorts (20.1% in post-COVID-19 and 23.3% in the matching weighted analysis of non-COVID-19 subjects), marked stages occurred rarely. There was no obvious difference in the severity of haematuria between groups (adjusted P = 1.000). The distribution of pyuria categories was also similar between the two cohorts (adjusted P = .385). Regression analyses adjusting for additional risk factors equally found no difference in haematuria and pyuria stages (Supplementary Table 4). Neither the severity of haematuria nor pyuria correlated with SARS-CoV-2 anti-spike or anti-nucleocapsid antibodies in post-COVID subjects (Supplementary Figs. S1 and S2).

Regarding the secondary outcomes (Table 5), there was a difference in the specific gravity, indicating more concentrated urine in the post-COVID-19 cohort (adjusted P < .001). No difference between post-COVID-19 and non-COVID-19 subjects was found for glycosuria (adjusted P = .279) or urinary pH (adjusted P = .143).

DISCUSSION

This study provides a multilevel analysis of kidney outcomes in individuals 9 months after non-severe COVID-19 compared with a population-based cohort of subjects without prior COVID-19. Despite a slight reduction in eGFR with a corresponding increase in individuals with mildly reduced eGFR, our analysis found no evidence for progressive or ongoing kidney disease after non-severe COVID-19.

The hypothesis of a relevant kidney sequela after non-severe COVID-19 arises from the kidney tropism and molecular effects of SARS-CoV-2 on the kidneys as well as the excess incidence of kidney effects in severe COVID-19 [2, 4, 6, 7]. Recently we and others observed slight changes in the eGFR after mild and moderate SARS-CoV-2 infections, supporting this hypothesis [9, 11]. Although similar, the effects in both studies were marginal and restricted to one parameter (eGFR). Importantly, such a slight decrease in eGFR after a potential injury does not necessarily reflect progressing or persisting kidney disease [24] and calls for a multilevel analysis of various kidney parameters to better rate the possibility and relevance of a distinct and progressive kidney sequela. Also, a careful assessment of confounders is important when looking at such small effects, especially as we still rely on retrospective and cross-sectional data.

Analogous to previous analyses [9, 11], our study reveals a slight reduction in eGFR in post-COVID-19 compared with non-COVID-19 subjects even after controlling for known CKD risk factors. While this implied an increase in individuals with mildly reduced kidney function, this reduction did not affect the CKD prevalence according to the KDIGO definition [22]. The latter finding differs from the previous study by Bowe et al. [9], who similarly observed minor effects on eGFR but also an increase in severe kidney events within 1 year after mild COVID-19. However, it must be noted that the cohort investigated by Bowe et al. [9] was significantly sicker and experienced a high rate of late AKI. This could indicate that very frail individuals or those with pre-existing advanced kidney diseases are still at an increased risk after COVID-19, albeit probably due to partial resolution of mild AKI during the acute infection or secondary effects than a distinct COVID-19-induced CKD.

Although the observed COVID-1-attributed eGFR change of ≈2 ml/min/1.73 m² is low, the effect is about twice that of sodium–glucose cotransporter 2 inhibitors (SGLT2-Is) and could become highly relevant over years in case of an ongoing effect [25]. However, we found no alterations in commonly observed abnormalities during COVID-19-related kidney involvement such as haematuria, pyuria and proteinuria [17–19], indicating no ongoing kidney effects.

Progressive decline in kidney function might be due to other effects than persisting infection and recent data suggest a direct pro-fibrotic effect of SARS-CoV-2 on the kidney [4]. Thus we also investigated the pro-fibrotic biomarker DKK3, known to indicate progressive eGFR decline independently of other risk factors or the underlying cause of kidney injury [14]. Further, DKK3 was recently shown to correlate with the transition from AKI to CKD in patients with severe COVID-19 and might thus be a good biomarker to investigate COVID-19 kidney sequelae [26]. In our cohort, DKK3 even tended to be lower in post-COVID-19 than non-COVID-19 subjects. This observation might be due to the increased rate of active smoking in the non-COVID-19 cohort, given that smoking can increase tubular DKK3 expression even without other evidence of kidney injury [27]. Still, our data indicate no increased risk for a progressive eGFR decline after non-severe COVID-19. We also found no change in albuminuria after non-severe COVID-19 as another important and independent risk factor for progressive CKD [16]. In the light of no suspicion in any other investigated renal parameter, the minor eGFR reduction appears not indicative for a distinct or progressive kidney disease, but rather suggests an aging of the kidneys during non-severe SARS-CoV-2 infection by about 2 years without further damage.

To the best of our knowledge this is the first study providing a deep and multilevel investigation of kidney outcomes after non-severe COVID-19 in a very well-characterized cohort including parameters of kidney function, kidney involvement during SARS-CoV-2 infection and a highly sensitive marker for progressive kidney disease. Despite its strengths, our study has several limitations. Importantly, data derived from a cross-sectional analysis without assessment of longitudinal data or availability of baseline eGFR values and controls were investigated before the COVID-19 pandemic. Despite a thorough matching of the cohorts and adjustment for potential confounders, our approach bares a remaining risk of selection bias as well as residual confounding. Also, our analysis is restricted to infections that occurred in 2020 and its validity for novel SARS-CoV-2 mutants still needs to be investigated. As SARS-CoV-2 vaccination was not comprehensively available at the time of the study, all subjects were unvaccinated against SARS-CoV-2 when the study visits occurred. The median duration after COVID-19 in our study was 9 months. Studies with a longer latency after COVID-19 as well as prospective studies are needed to validate our findings. Finally, a comparison with patients with other viral infections might help to investigate whether the minor changes are specific for SARS-CoV-2.

In conclusion, despite a slight eGFR reduction, this study shows no distinct or progressive kidney sequela 9 months after non-severe COVID-19. This has important implications, as progressing kidney disease after non-severe COVID-19 would have resulted in a severe burden of general and kidney healthcare.

ACKNOWLEDGEMENTS

The technical assistance of Silvia Chilla and Martina Wagner is highly acknowledged. We are thankful to Danny Schreier for support with Atellica Solution analyses. Also, we appreciate the statistical advice from Alina Goßling.

FUNDING

C.S.-L. and T.B.H. were supported by the City of Hamburg, Germany (LFF-OS 95-2021). T.B.H. was supported by the Deutsche Forschungsgemeinschaft (CRC1192, HU 1016/8-2, HU 1016/11-1, HU 1016/12-1), the Bundesministerium für Bildung und Forschung (STOP-FSGS-01GM2202A, NephrESA-031L0191E), the Else-Kröner Fresenius Foundation (Else Kröner-Promotionskolleg–iPRIME) and the H2020-IMI2 consortium BEAt-DKD (115974); this joint undertaking receives support from the European Union's Horizon 2020 research and innovation program and European Federation of Pharmaceutical Industries and Associations and Juvenile Diabetes Research Foundation. T.R. acknowledges Deutsche Forschungsgemeinschaft grants (25440785-SFB877, P6-KFO306, 80750187-SFB841).

AUTHORS’ CONTRIBUTIONS

C.S.-L., E.L.P., R.T., S.B. and T.B.H. conceptualized and designed the analysis. S.S., A.A., M.L., P.K., T.R., T.Z. and D.F. were involved in data acquisition. Statistical analysis was performed by S.H., E.L.P. and F.H.. C.S.L., S.H., E.L.P., F.H., D.F. and T.B.H. were responsible for analysis and data interpretation. Manuscript was drafted by C.S.-L. and T.B.H. and revised by S.H., E.L.P., M.L., P.K., T.R., R.T.. All authors read and approved the final version of the manuscript.

DATA AVAILABILITY STATEMENT

The data will be shared upon reasonable request to the corresponding author.

CONFLICT OF INTEREST STATEMENT

The authors of have no conflicts of interest to disclose.

REFERENCES