The predictive value of serum level of cystatin C for COVID-19 severity

To investigate the potential prognostic value of Serum cystatin C (sCys C) in patients with COVID-19 and determine the association of sCys C with severe COVID-19 illness. We performed a retrospective review of medical records of 162 (61.7 ± 13.5 years) patients with COVID-19. We assessed the predictive accuracy of sCys C for COVID-19 severity by the receiver operating characteristic (ROC) curve analysis. The participants were divided into two groups based on the sCys C cut-off value. We evaluated the association between high sCys C level and the development of severe COVID-19 disease, using a COX proportional hazards regression model. The area under the ROC curve was 0.708 (95% CI 0.594–0.822), the cut-off value was 1.245 (mg/L), and the sensitivity and specificity was 79.1% and 60.7%, respectively. A multivariable Cox analysis showed that a higher level of sCys C (adjusted HR 2.78 95% CI 1.25–6.18, p = 0.012) was significantly associated with an increased risk of developing a severe COVID-19 illness. Patients with a higher sCys C level have an increased risk of severe COVID-19 disease. Our findings suggest that early assessing sCys C could help to identify potential severe COVID-19 patients.

Hospital approved the study protocol (NO.2020GFKY005). All clinical investigations were conducted based on the principles expressed in the declaration of Helsinki. Written informed consent was waived by the ethical committee due to the retrospective nature of our study of routine clinical data.

Study design and participants. This is a single-center, retrospective, observational study done at Tumor
Center of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology (Wuhan, China), a designated hospital to treat patients with COVID-19. We analyzed consecutive patients admitted between 15 February and 14 March 2020 because of a COVID-19 disease based on World Health Organization interim guidance 11 . Laboratory confirmation of COVID-19 infection was performed using Reverse Transcription-Polymerase Chain Reaction (RT-PCR) detection by the local health authority as previously described 12 .
Data collection and outcome measures. We retrospectively reviewed the electronic medical records of 168 consecutive eligible patients with COVID-19 using a digital database. We extracted the epidemiological, demographic, clinical, laboratory data on admission, the chest computed tomography (CT) image, and outcome data using a standardized data collection form. In cases of disagreement, we reached a consensus after team discussion. In case of missing or uncertain data, we obtained and clarified data by direct consulting with attending doctors and other healthcare providers. We constructed a vascular risk factor score based on the following well-documented vascular risk factors in each patient: hypertension, diabetes, dyslipidemia, atrial fibrillation, current smoker, overweight and physical inactivity [13][14][15][16][17] . Quick Sequential Organ Failure Assessment (qSOFA) was calculated based on data regarding systolic blood pressure, respiratory rate, and mental status within 24 h after admission 18 Our primary outcome was severe COVID-19 disease defined as fever or suspected respiratory infection, plus one of: respiratory rate > 30 breaths/min; severe respiratory distress; or SPO 2 ≤ 93% on room air based on the interim guidance of the World Health Organization 11 . Our secondary outcome was death, which was limited by the duration of our observation period. All the authors agreed on the study protocol and reviewed the manuscript.
Measurements of cystatin C, creatinine and Bun. Serum cystatin C was measured by a turbidimetric immunoinhibition assay using the cystatin kit (Tina-quant cystatin C Gen. 2, Roche) with an automatic biochemical analyzer (Cobas c-system, Roche, Switzerland). This method is traceable to a primary reference material with values assigned by the international cystatin C reference material (ERM-DA471/IFCC) as a calibrator 19 . This procedure had a total coefficient of variation of 2.2% at a cystatin C level of 1.0 mg/L and of 1.4% at a level of 4.0 mg/L. Reference range 0.55-1.09 mg/L. Serum creatinine was determined by a traceable method (enzymatic assay calibrated against the National Institute of Standards and Technology standard reference material [SRM 967]) with an automatic biochemical analyzer (Cobas c-system, Roche, Switzerland) 20 . The total coefficient of variation is < 4.0%, and the reference range is 44-133 μmol/L. Serum Bun was determined by an enzymatic method with an automatic biochemical analyzer (Cobas c-system, Roche, Switzerland) according to the kit protocol. This method is traceable to a primary reference material (SRM 909b). This procedure had a total coefficient of variation of 1.2% at a level of 7.2 mmol/L and of 0.7% at a level of 35.1 mmol/L. The reference range is 2.9-8.2 mmol/L.

Statistical analysis.
We summarized continuous and categorical variables as median (interquartile range, IQR) and numbers (percentages), respectively. We used the t-test or Mann-Whitney test to compare differences in continuous variables, and the chi-square test or Fisher's exact test to compare differences in categorical variables where appropriate. The classification performance of sCys C, Bun and sCr to discriminate between severe and non-severe cases was evaluated by calculating the area under the ROC curve (AUROC) and its 95% confidence intervals (95% CI). We defined the score with the largest Youden Index as the optimal cut-off value for predicting COVID-19 severity. Based on the optimal cutoff value, we calculated the sensitivity and specificity. The differences between AUROC were evaluated by using a method as previously described 21 . We divided patients into higher and lower group based on sCysC levels above and below the cut off value. We calculated the absolute event rate per 1000 patient-days for severe COVID-19 illness and death during our observation period. Kaplan-Meier curves depicted the risk for outcome events stratified by sCys C levels (higher vs lower). We implemented univariable and multivariable COX proportional hazards regression models to examine the association between sCys C level and the risk of developing a severe COVID-19. For multivariate analysis, to avoid overfitting in the model, we chose the variables based on previous findings and clinical constraints. Previous studies have shown older age, qSOFA, vascular risk factor burden, and several laboratory abnormal findings to be associated with an increased risk of severe COVID-19 13,18 . Therefore, we chose qSOFA, vascular risk factor burden, decreased lymphocyte, and increased lactate dehydrogenase (LDH) in addition to age and sex for our multivariable regression model. All analyses were performed using STATA 12.0 (StataCorp LP, College Station, TX) and SPSS for Windows (SPSS 25.0, IBM, Inc., Chicago, IL, USA).

Results
Baseline difference between patients with severe and non-severe COVID-19. We admitted  Fig. 1a). Figure 1b,c show the serum levels of creatinine and Bun were higher   Fig. 3). In univariable analysis, a higher sCys  www.nature.com/scientificreports/ C level was associated with an increased risk of severe COVID-19 (unadjusted HR 4.95, 95% CI 2.31-10.57). After adjustment for age and sex, a higher sCys C level remained significantly associated with severe COVID-19 (adjusted HR 3.04 95% CI 1.30-7.13). This relationship did not change after additional adjustment for decreased lymphocyte, increased LDH, qSOFA or vascular risk factor burden along with age and sex (Table 3). A sensitivity analysis excluding three patients who reported previous renal disease resulted in similar findings (Table 4).

Discussion
The early and fast evaluation of severely ill COVID-19 patients is paramount to ensure early medical monitoring and interventions for these patients due to the medical resource constraints during the COVID-19 pandemic.
Our most important finding is that in COVID-19 patients a higher sCys C level is associated with an increased risk of experiencing severe COVID-19 illness. Our results highlighted the considerable predictive performance of sCys C for severe COVID-19 disease on ROC curve analysis. At the time of admission, a sCys C value over 1.245 (mg/L) was highly predictive of developing a severe COVID-19 disease. Our findings were in line with a previous hospital-based study that showed the highest baseline sCys C level was associated with more severe inflammatory status and unfavorable outcomes among COVID-19 patients 22 . These findings suggest that sCys C could serve as a potential inflammatory target for preventing COVID-19 from the likely progression of critical illness and mortality, in addition to representing early renal insufficiency. Since the sCys C is generally readily available at hospital admission, our findings may contribute to identify COVID-19 patients with poor prognosis at an early stage.   www.nature.com/scientificreports/ Our ROC analysis showed that sCys C had a better predictive performance for severe COVID-19 than two conventional renal function indicators (Bun and sCr). Accumulating evidence has demonstrated that Bun and sCr levels might not be good indicators for the early detection of renal injury 23,24 . sCys C is emerging as a novel index with a high sensitivity and specificity for evaluating renal function, independently of age, sex, weight, inflammation, and other factors 25,26 . The renal clearance of 5-40 kDa molecules like cystatin C (13.3 KDa) is decreased more than that of low molecules like creatinine (0.133 kDa) 27 . Since cystatin C is mainly excreted via glomerular transport, a reduction in their glomerular filtration rate would result in a simultaneous increase of sCys C levels. Previous studies suggest that a decrease in the pore diameters of the functional pores might account for the increased sCys C levels 28,29 . Recently, researchers from Lund University (Sweden) proposed the designation 'Shrunken pore syndrome' to reflect a reduced pore size of the glomerular membranes, which might help explain the superiority of cystatin C as a predictor of mortality 30,31 . In line with a previous study 32 , our findings that sCys C was significantly increased in patients with severe COVID-19 compared to those with nonsevere COVID-19 illness (1.28 [1.05-1.83] vs 1.07 [0.92-1.22]) indicate that severe SARS-CoV-2 infection might damage the kidney. Our findings were supported by a previous hospital-based observational study that showed COVID-19 patients with renal involvement had a mortality of 11.2%, compared to 1.2% in those without renal involvement 33 . These results suggest that renal complications in COVID-19 were associated with higher mortality.
Our findings have important implications for early COVID-19 prevention and treatment. For example, hydroxychloroquine is a safe and highly tolerable approach with minimum side effects. A retrospective longitudinal cohort study showed that hydroxychloroquine treatment (OR 3.891, 95% CI 1.196-12.653) was associated with renal function recovery in patients with lupus nephritis 34 . A large-sampled multicentered observational study has shown that the exposure of hydroxychloroquine is associated with a decreased risk of hospitalization from COVID-19 (OR 0.53; 95% CI 0.29-0.95) 35 . Additional, findings from some recent random-controlled studies raise the concern about whether dexamethasone treatment could improve the outcome of COVID-19 36,37 Whether patients with higher sCys C levels might benefit form early hospital treatment measures such as hydroxychloroquine or dexamethasone need to be addressed in future large-scale studies.
The underlying mechanisms of the association between sCys C level and COVID-19 severity remain unclear. SARS-CoV-2 infection may directly cause endothelial damage, infects renal tubular epithelium and podocytes, causing mitochondrial dysfunction and acute tubular necrosis. This process is mediated through the angiotensinconverting-enzyme 2 (ACE2) dependent pathway 38,39 Moreover, available evidence suopport that kidney impairment in COVID-19 patients may be caused by an interplay of virus-mediated injury, a dysregulated inflammatory response, hypercoagulation, and microangiopathy 40 . A severe inflammatory process during SARS-CoV-2 infection might also probably induce increased levels of cystatin C similar in size to cytokines like interleukin-6. This hypothesis is supported by a previous study showing that sCys C level was positively correlated with inflammatory indicators such as IL-6, tumor necrosis factor-α, and hypersensitive C-reactive protein 41 . These aforementioned inflammatory indicators have been reported to be related to the COVID-19 severity 42,43 . The underlying interaction between circulating cystatin C and viral infection may provide insight into our better understanding of pathophysiological events in COVID- 19.
The results of the present study should be interpreted within the constraints of its limitations. First, we conducted this retrospective study at a single-centered hospital with limited sample size. Moreover, we retrospectively collected data from medical records and laboratory data regarding prior renal function in some patients was not accessible. Therefore, we may underestimate the true overall incidence of COVID-19-associated elevated sCys C. However, a sensitivity analysis excluding those patients who reported previous renal disease did not alter the association between a higher sCys C level and a higher COVID-19 risk. Notably, we did not include data regarding urine output, a defining characteristic of acute kidney insufficiency. Therefore, more studies with a broad geographic scope are needed to get a more comprehensive understanding of role of renal impairment in COVID-19. Our strength include the association of sCys C with the COVID-19 severity was verified using previous well-validated cofounders in consecutive laboratory-confirmed participants.

Conclusions
Our findings that COVID-19 patients with a higher sCys C level were at an increased risk suggest that this population needs early prevention and treatment.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request (Email: houweidu@fjmu.edu.cn). Table 4. Association between higher sCys C level and severe COVID-19 illness in a sensitivity analysis. Abbreviations: sCys C = serum cystatin C; COVID-19 = coronavirus disease 2019; HR = Hazards ratio; Multivariable Model 1: Adjusted for age, sex, increased lactate dehydrogenase, decreased lymphocyte, and qSOFA (Quick Sequential Organ Failure Assessment). Multivariable Model 2: Adjusted for age, sex, increased lactate dehydrogenase, decreased lymphocyte, and vascular risk factor burden.

Unadjusted
Adjusted for age and sex