SARS-CoV-2 infection in chronic kidney disease patients with pre-existing dialysis: description across different pandemic intervals and effect on disease course (mortality)

We performed our analyses of patients suffering from dialysis-dependent CKD retrieving data from LEOSS (https://​leoss.​net/​) (Fig. 1) [13]. In LEOSS, clinical data is reported anonymously and retrospectively in an electronic case report form using the online platform ClinicalSurveys.net of the University Hospital of Cologne [14]. The anonymization procedure has been published previously [15].

The transmitted data set consisted of 1277 patients, which were documented in 80 different study sites and diagnosed between January 2020 and May 2021. Exclusion criteria are illustrated in Fig. 1. We excluded in total 8.3% (106/1277) patients resulting in a data set of 207 patients suffering from CKD5D (cases) and 964 patients suffering from CKD not requiring dialysis (potential controls).

Chosen parameters included sociodemographics, comorbidities, details on CKD, clinical and diagnostic parameters, as well as administered therapies. Symptoms and diagnostic parameters were determined within 48 h after first SARS-CoV-2 positive testing. Pre-existing comorbidities and clinical events were documented by investigators according to clinical definitions using anamnestic information and medical records. Prior immunosuppressive medication at baseline was defined as administered within an interval of 3 months before SARS-CoV-2 infection. Advanced respiratory support was defined as mechanical ventilation or extracorporeal membrane oxygenation (ECMO). The month of first diagnosis was assigned to one of three intervals of pandemic based on infection rates and evolving virus variants in Germany [16]: January 2020–June 2020, July 2020–January 2021, February 2021–May 2021. Death within the observational period was used as end-point in the regression analyses.

Data management and analyses were performed using R, version 4.1.0 [17]. Figure 1 illustrates the workflow.

We described patients’ characteristics as absolute numbers and percentages. Group differences between the three different intervals of pandemic were determined using Chi-squared test or when applicable Fisher’s exact test. We controlled for multiple comparison using the Bonferroni correction.

Variables relevant for the regression analyses were analysed for missingness (supplementary (S) Table S1) and imputed iteratively via fully conditional specification (FCS) with proportional odds model or polytomous logistic regression depending on the nature of the respective variable using the R package MICE (https://​cran.​rproject.​org/​web/​packages/​mice/​mice.​pdf). This resulted in a total of 5 imputed data sets.

For the propensity-score matched pair analysis, each case was matched to one control in each imputed data set using the R package MatchIt (https://​cran.​rproject.​org/​web/​packages/​MatchIt/​index.​html). Exact matching was performed on age, gender and phase (according to LEOSS criteria, see Figure S1) at first SARS-CoV-2 detection; propensity-score matching (nearest neighbour) on hypertension, chronic heart failure, coronary artery disease, diabetes mellitus type 2, chronic obstructive pulmonary disease (COPD), active oncological disease, obesity, prior immunosuppressive medication and therapy limitations. Matching quality was assessed via standardized mean differences.

LEOSS was approved by the applicable local ethics committees of all participating centers and registered at the German Clinical Trials Register (DRKS, No. S00021145).

The cohort consisted of 207 patients suffering from CKD5D recruited in LEOSS and diagnosed between January 2020 and May 2021. All patients underwent hemodialysis. Vascular hypertensive (46.2%, 61/132), secondary (22.0%, 29/132) and primary glomerular disease (9.1%, 12/132) were the leading etiologies of CKD. Most CKD5D patients also suffered from hypertension (79.6%, 164/206). Other frequent comorbidities included diabetes mellitus type 2 (44.2%, 88/199), coronary artery disease (36.3%, 74/204) and obesity (31.4%, 44/140). Prior immunosuppressive medication was present in 17.6% (34/193). History of transplantation (51.5%, 17/33), rheumatological disease (15.2%, 5/33) and other reasons (33.3%, 11/33) were given as indication. Details are depicted in Table 1.

At first detection of SARS-CoV-2, 46.2% (84/180) of our patients reported fever, 36.0% (64/178) dyspnea, 28.5% (49/172) dry cough, 7.1% (12/170) myalgia, 4.1% (7/171) headache and 2.4% (4/167) hypogeusia and/or hyposmia. At this time point, most patients (64.3%, 133/207) were assigned to the uncomplicated phase according to LEOSS criteria (see Figure S1). During the further course of disease, 19.8% (41/207) underwent critical phase, 14.3% (29/209) required advanced respiratory support and 26.6% (55/207) patients deceased.

The pandemic was divided into three intervals based on infection rates and evolving virus variants in Germany: January 2020–June 2020, July 2020–January 2021, February 2021–May 2021. When comparing across the pandemic intervals, patients’ characteristics did not differ except for gender (Table 1). The percentage of recruited female patients increased over time: in the first interval, 34.5% (20/58) of the patients, in the second 40.5% (47/116) and in the third interval 60.6% (20/33).

There was no significant difference between the pandemic intervals regarding existing symptoms and laboratory parameters at first SARS-CoV-2 detection. The latter is illustrated in Fig. 2. CRP was generally elevated ≥ 30 mg/dl in 63.4% (90/142), D-dimers > 2 × upper limit of normal (ULN) in 61.3% (46/75) and lymphocytes were below 800/μl in 58.2% (52/91).

Therapy limitations were reported in 23.4% (36/154) of the patients. Throughout the pandemic, administered treatment changed: Steroids > 0.5 mg/kg prednisolone equivalents were used in 7.7% (4/52) of the patients during the first interval, in 43.2% (48/111) during the second one and in 30.3% (10/33) during the third interval. Use of chloroquine was more frequent in the first interval (15.7%, 8/51 versus 1.9%, 2/106 in the second interval versus 3.0%, 1/33 in the third interval). Remdesivir was administered in 5.8% (11/191) of the patients with no differences in frequency of use throughout the pandemic.

We performed a multivariable logistic regression on fatal outcome using the whole data set of 1171 patients suffering from CKD in LEOSS after multiple imputation. Pooled results are shown in Table 2. Increasing age was identified as risk factor with the greatest adjusted odds ratio (aOR) in the category > 85 years (aOR 7.34, 95% CI 2.45–21.99, p < 0.001). Chronic heart failure (aOR 1.67, 95% CI 1.25–2.23, p < 0.001), coronary artery disease (aOR 1.41, 95% CI 1.05–1.89, p = 0.021) and active oncological disease (aOR 1.73, 95% CI 1.07–2.80, p = 0.027) were further predictors for fatal outcome. Dialysis dependency did not show a significant association with mortality in CKD patients (aOR 1.08, 95% CI 0.75–1.54, p = 0.692).

Patients suffering from CKD5D were matched via propensity score to controls suffering from CKD not requiring dialysis. Controls were predominantly assigned to CKD stage 3 (52.1%, 395/758), followed by stage 4 (15.8%, 120/758) and stage 2 (15.0%, 114/758), according to the definition of the international guideline group Kidney Disease Improving Global Outcomes (KDIGO). A detailed description of controls suffering from CKD not requiring dialysis is given in Table S2. In the univariate and multivariable conditional regression stratified by dialysis, dialysis-dependency was not significantly associated with fatal outcome (aOR 1.34, 95% CI 0.70–2.59, p = 0.375). Pooled results are shown in Table 3. We performed sensitivity analyses using the unimputed data set (Table S3) that confirmed our results with dialysis-dependency not being significantly associated with fatal outcome but exhibiting a risk tendency (aOR 1.40, 95% CI 0.73–2.69, p = 0.31). Univariate and multivariable results of the matched-pair analyses using the imputed and unimputed data set are illustrated in Fig. 3.

Interestingly—despite of changing patients’ management and testing strategies, evolving virus variants and applied vaccinations—sociodemographics, clinical characteristics and laboratory findings at first diagnosis did not significantly differ in dialysis-dependent CKD patients over time. However, the changing landscape of COVID-19 therapeutics and recommendations for specific strategies, as well as the overall limited options for CKD patients is reflected in our data. While hydroxychloroquine has been more frequently used at the very beginning of the pandemic, it declined after randomized controlled trials failed to detect a benefit [19]. The increasing use of steroids follows recommendations by (inter)national medical societies and the WHO that evolved after the first interval of the pandemic [20‐22]. Remdesivir, as the first antiviral drug approved but currently without clear recommendation for use [20, 22] and in particular with precaution for patients with reduced GFR [11], has interestingly been administered throughout the whole pandemic in some patients undergoing dialysis for chronic dialysis dependency.

Our multivariable analysis confirmed already known risk factors also for patients suffering from CKD, such as age, chronic heart failure, coronary artery disease and an active oncological disease [23‐25], which was described for the whole LEOSS cohort [18, 26‐28] and which we published in a smaller CKD cohort previously [8]. In contrast, broadly accepted risk factors, such as male sex, hypertension or diabetes mellitus failed to present as additional risk factors in our model, which might be due to the overall high prevalence in our cohort. It might also be important to note that hypertension and diabetes mellitus often have been identified without being adjusted for CKD [23, 29] or in a cohort where prevalence of CKD was low [30, 31].

Pre-existing need for dialysis by itself was neither in our multivariable regression analysis nor in our propensity-score matched-pair analyses significantly associated with fatal outcome in SARS-CoV-2-infected patients, thus confirming our previously published results [8]. The OpenSAFELY project with 17,278,392 individuals similarly identified CKD as one of the highest risk factors for death but, in contrast, identifies a history of dialysis as an additional factor in a secondary analysis [24]. The slight difference in a history of dialysis and dialysis-dependency might account for these discrepancies. Flythe et al. addressed in a retrospective cohort study (STOP-COVID) in 4264 critical ill patients with COVID-19 (143 patients with preexisting kidney failure receiving maintenance dialysis) a similar question as the present study [5]. They demonstrated that dialysis-dependent CKD patients had a shorter interval from symptom onset to intensive care treatment than non-dialysis-dependent CKD patients and detected higher mortality rates for both—dialysis-dependent and -independent CKD patients. In line, they showed that hazards of in-hospital death is higher in patients with dialysis-dependent kidney failure compared to patients without pre-existing CKD [5]. Further studies report high mortality within the range of 20–30% among SARS-CoV-2-infected patients suffering from dialysis-dependent CKD [10, 25, 32‐34] which is comparable to our results (26.6%, 55/207). A UK registry study stressed kidney replacement therapy as crucial risk factor. In particular in center hemodialysis patients had a high mortality with a peak in April 2020 [35]. A more recent prospective observational study demonstrated an increased mortality (35.7%) of hemodialysis patients within the first year after infection [36]. Remarkably, these patients died also after discharge of the hospital. Moreover, anti SARS-CoV2 antibodies decrease with time indicating that humoral responses were low after infection. A similar response has been described after vaccination in this vulnerable cohort [37]. However, in these studies, a direct comparison to dialysis-independent CKD patients is lacking. Thus, the studies analyzed different dialysis cohorts in different countries at different time point during the pandemic. Subsequently the results might differ. Our study highlights CKD and decreased glomerular filtration rate (GFR) independent of dialysis as relevant risk factors for severe COVID-19.

As 100% of our dialysis patients were on hemodialysis and none on peritoneal dialysis, we are unable to generate insights in potential benefits of hemodialysis, i.e., the intermittent anticoagulation wit heparin, usually three times a week, or the regular health care utilization that would allow swifter diagnosis and therapy. Previous reports have, however, not shown any difference in the disease course between peritoneal dialysis and hemodialysis [25].

One of the strengths of our study lies in being based on data of LEOSS, which has uniformly and standardized collected data since the beginning of the pandemic. Thus, all three intervals of the pandemic, as well as cases and controls derive from one data source operating on a transregional level with more than 131 sites. Nevertheless, there are still several limitations as outpatient sites are underrepresented in LEOSS, study sites have changed over time and important confounders (e.g., socioeconomic background, COVID-19 vaccination status, frailty) might not have sufficiently been considered in the matching or regression analysis. Dialysis could also have an impact on other endpoints, such as ICU admission or thromboembolic complications which should be addressed in further analyses.

In conclusion, our results indicate that not chronic dialysis dependency itself but rather the associated age, co-morbidities and underlying diseases are important modifiers of disease severity and death. However, the high mortality in both, cases and controls, should raise awareness for SARS-CoV-2-infected patients suffering from CKD, and should be considered when discussing about recommendations for vaccine booster shots.