The Application of Lactate Dehydrogenase in Coronavirus Disease 2019 as the Best Indicator for the Progression and Clinical Status: A Case-Control Study

Introduction Coronavirus disease 2019 (COVID-19) is now ocially a pandemic. Current studies observed extensive abnormal indexes in COVID-19 patients and signicant differences between mild and severe patients. However, which index would perform better as the indicator of disease progression merits further investigation. Methods We enrolled COVID-19 patients who were admitted to Shanghai Public Health Clinical center. We closely monitored the following candidate indictors: white blood cell, lymphocyte, platelet, CD4 T cell, CD8 T cell, alanine aminotransferase, estimated glomerular ltration rate (eGFR), brin degradation products (FDP), D-dimer, creatine kinase, myoglobin, troponin T (TnT), N-terminal pro-B-type natriuretic peptide (NT-proBNP), lactase dehydrogenase (LDH), C-reactive protein (CRP), and procalcitonin. The correlation with illness severity were assessed by Spearman analysis and the ability of differentiating the clinical statuses was quantied as the AUC value of the ROC curve. NT-proBNP, AUC had the ROC AUC of over 0.90 (0.927). Conclusions This study found LDH to be a superior indicator for COVID-19 status and had the potential to optimize the clinical management strategy.


Introduction
Coronavirus disease 2019 (COVID-19) is now o cially a pandemic. Critically ill patients with  are marked by the high mortality and catastrophic expenditure, despite the respiratory support and comprehensive treatment that are usually administered to these groups. Therefore, suitable indicators would be of much signi cance in the early treatment of COVID-19 patients and may play a vital role in the prevention and blockage of the disease course.
Although multiple studies have reported laboratory or clinical markers relating to progression to severe or critical illness, it is still di cult to predict which type of patients may progress or relieve. Current researches generally adopted the laboratory ndings on hospital admission as their index variable.
However, if some of the patients had already developed severe illness prior to admission, this would create a bias in which analyzing laboratory ndings on admission would reveal only the differences between mild and severe patients, rather than risk factors of disease progression.
What kind of indicators are good predictors? We propose the "CDEF" principle for an ideal indicator, namely correlation, differentiation, early-warning, and feasibility. "Correlation" refers that indicators should clearly parallel with the clinical status. For COVID-19 patients, the indicators are supposed to re ect dynamic uctuations of the respiratory function. "Differentiation" refers the performances in differentiating between mild and severe illness. "Early-warning" refers that a perfect indicator is expected to show signi cant abnormalities prior to the clinical status progresses. Oxygen saturation is a satisfactory variable in accordance with the patient's current respiratory status, but it cannot re ect the role of predictive warning. Regarding to the feasibility pro le, to make sure an indicator can be wildly used in the clinical routine practice, the abnormalities of indicators in severe cases should be consistent with the established clinical signi cance, like beyond the normal range.
In order to discover indicators that can re ect early disease progression and issue early warning of possible advances to critically ill status, we followed up 326 cases of laboratory con rmed COVID-19 patients in Shanghai and tried to nd the best indicator in line with the "CDEF" rule.

Study population and de nition
This is a retrospective case-control study. From January 20, 2020 and March 15th, 2020, we enrolled all patients with COVID-19 according to World Health Organization (WHO) interim guidance in Shanghai Public Health Clinical Center (SPHCC). The study was approved by the Ethics Committees of Huashan Hospital, Fudan University. All patients who participated in the study gave informed consent.
The severity or clinical condition of COVID-19 patients was classi ed into mild, ordinary, severe and critical illness according to the Chinese Clinical Guideline for COVID-19 pneumonia diagnosis and treatment (6th edition) [ 1 ]. Because the severity of a patient 's disease can change dynamically, we viewed the severity of COVID-19 as a stage of disease development. We de ned the mild, and ordinary illness as mild-ordinary stage, severe illness as the severe stage and critical illness as critical stage. Symptoms, vital signs, laboratory values and chest CT scan or X-ray were monitored daily for patients in the severe or critical stage and every 2-3 days for patients in mild-ordinary stage.
We grouped the patients according to their disease progression and their severity on admission. Patients who continued to be in the mild-ordinary stage during hospitalization were de ned as the persistent mildordinary group (PM Group), and patients who had experienced the severe or critical stage during the hospitalization were de ned as the severe/critical group (S/C Group). Patients in the S/C Group were further divided into a mild-ordinary-at-admission group (MA Group) and a severe-at-admission group (SA Group) according to the clinical stage at admission (Fig. 1).

Data Collection
We obtained epidemiological, demographic, clinical, laboratory and radiology data from patients' medical records. The data were reviewed by a trained team of physicians.
We collected variables with clinical signi cance based on our clinical experience and previous studies about COVID-19 [2][3][4]. In SPHCC, the following candidate parameters were monitored closely, including white blood cell (WBC), lymphocyte, platelet, CD4 T cell, CD8 T cell, alanine aminotransferase (ALT), estimated glomerular ltration rate (eGFR), brin degradation products (FDP), D-dimer, creatine kinase (CK), myoglobin, troponin T (TnT), N-terminal pro-B-type natriuretic peptide (NT-proBNP), lactase dehydrogenase (LDH), C-reactive protein (CRP), and procalcitonin (PCT). We collected the values of these parameters measured at admission, the rst day of each clinical stage, the midpoint date of each clinical stage and the day of death.

Data management and statistical analysis
Correlation between the laboratory parameters and disease severity (mild-ordinary, severe, critical, and death) were analyzed using Spearman's correlation. To analyze the performance of candidate parameters for indicators of disease severity, we created receiver operating characteristics (ROC) curves which were summarized by area under the curve (AUC) estimates. ROC curve analysis was also performed to determine the cut-off value, sensitivity and speci city of indicators in differentiating the cases in mildordinary stage from those in severe or critical stage.
All statistical analysis was conducted using IBM SPSS software (Version 22.0) and R (Version 3.6.1).
Basic characteristics were sorted by median (interquartile range, IQR) or mean (standard deviation, SD) in consecutive variables, count (percentage) in categorical variables. Means for continuous variables were compared using one-way Anova analysis. Categorical variables were compared with the use of Chisquare test. Dynamics of laboratory ndings along with disease progression were plotted and Spearman's correlation were used to evaluate the correlation. All tests were two-sided and P value < 0·05 was considered as statistically signi cance.
As of 15th April 2020, 299 patients continued to be in the mild-ordinary stage since admission (PM Group), the other 27 patients experienced severe or critical stage (S/C Group), including 20 patients who as in mild-ordinary stage at admission but deteriorated during hospitalization (MA Group) and 7 patients initially severe at admission (SA Group) (Fig. 1).
There were more men and elders in the S/C Group (p < 0.05) ( Table 1). 109 (36.5%) in the PM Group and 17 (63.0%) in S/C Group reported comorbidities (p = 0.002). Overall symptoms patterns were similar between 3 groups, with fever (80.9%, 264/326) and cough (49.7%, 162/326) being the most common symptoms. Dyspnea was observed in 5 (18.5%) in the S/C Group while 2 (0.67%) in the PM Group (p < 0.001). 315 (96.6%) cases had abnormalities consistent with viral pneumonia on chest radiology or lung CT.     The level of D-dimer, CRP at admission were above normal range both in the PM group and the MA group; the levels of LDH, NT-proBNP, myoglobin, CD4 T cell, eGFR, FDP and PCT were beyond normal range only in the MA group but remained normal in the PM group (Table 2, Fig. 2D, 2G, 2J, 2L, 2M), indicating a nature warning effect of the upper limit of normal of these parameters.
In order to evaluate the diagnostic performance of the above 5 parameters in differentiating the severity of COVID-19, we conducted the ROC analysis (Fig. 4). Comparing between mild-ordinary stage and severe/critical stage (Fig. 4A), LDH had the best performance, with the highest ROC AUC of 0.951. With a cutoff value of 385.5 U/L, the sensitivity of LDH in differentiating severe COVID-19 was 49.7% and the speci city was 95.3%. PCT ranked second, with the ROC AUC of 0.905. With a cutoff value of 0.055 ng/ml, PCT had a sensitivity of 74.1% and a speci city of 79.2%. ROC AUC of D-dimer, NT-proBNP and FDP was less than 0.90. Comparing between mild-ordinary and severe stages (Fig. 4B), only LDH had the ROC AUC of over 0.90 (0.927). With a cutoff value of 345 U/L, the sensitivity of predicting severe cases was 48.2% and the speci city was 92.3%.

Discussion
This study is the rst clinical study to describe the correlation and lymphocytes stayed relatively at in rst 7 days since the onset-of symptoms and began to rise signi cantly after that. This is consistent with Zhou's nding that non-survivors of COVID-19 progressed to sepsis on the average of seven days. In another words, the non-survivors experienced several stages of the disease, i.e. mild, severe and even critical ill, in the rst seven days and stayed in critical stages in the following 14 days. Therefore, the signi cantly elevated part of the curve described the patients in critical stage and these indicators were related to death, but not early predictors for disease progression. Third, predictors should show abnormalities earlier than the deterioration of the clinical condition manifested by the patient's symptoms and signs. The right way to nd those indicators is to compare the parameters of severe cases in their mild stage to the persistent mild patients [ 4 ]. It is common to compare admission ndings of the severe cases to the mild ones. However, as previously mentioned, many patients were seriously ill when they were admitted to the hospital. This kind of comparison would show the difference between severe and mild patients but not risk factors for developing severe cases. Fourth, for indicators that appear abnormal in most persistent mild cases, even if they also showed statistical differences to severe ones, it may have limited clinical application value for clinicians. These indicators need to establish a cut-off speci c for COVID-19, and it is also affected by technical factors such as laboratory examinations, resulting a higher threshold for acceptance and application in clinical practice.
We summarized a "CDEF" rule for indicators, namely correlation, differentiation, early-warning, and feasibility. We measured the correlation by Spearman analysis and found that LDH, PCT, NT-proBNP, MYO and D-dimer correlated well to the severity of COVID-19. The ability of differentiating the clinical conditions was quanti ed as the AUC value of the ROC curve, and we noted the superior performance of LDH in differentiating between mild and severe patients. Early-warning was to show abnormalities even in the mild stage of the disease, therefor helping clinicians to nd high-risk patients who might deteriorate. One way to achieve feasibility was to warn the clinicians when the indicators become abnormal, i.e. beyond the normal range, rather than another unestablished cutoff. In this study, CRP and D-dimer levels were above the upper limit of normal both in mild and severe cases although there were signi cant differences between mild and severe cases. Thus, above-normal CRP or D-dimer had the di culty to indicate the disease progression.
Lactate dehydrogenase is a cytoplasmic glycolytic enzyme found in almost every tissue. Its elevation generally indicates tissue damage. Raised LDH is a common nding in patients infected with MERS-CoV[ 8,9 ], H7N9[ 10, 11 ] and H5N1 [ 12 ]. It is reported to be independent factors of mortality for patients with severe acute respiratory syndrome [ 13 ] and H1N1 infection [ 14 ]. It is also one of the biomarkers most strongly associated with ARDS mortality [ 15 , 16 ]. Our research did not combine LDH with other indicators.
The rst reason is that LDH's ROC AUC for predicting severity of COVID-19 is more than 0.95. The combination of other indicators that are inferior to LDH is of limited signi cance for improving prediction performance. The main signi cance of the early predictors is to identify high-risk patients in order to allocate medical resources more rationally and improve the prognosis, but not predicting the prognosis itself. Therefore, we believe that even if LDH alone may slightly inferior to indicator combination in the predicting accuracy, it can greatly improve the convenience in clinical practice. What's more, the present indictor combination or work ow of COVID-19 [ 17,18 ] still lack large-scale clinical veri cation, while LDH has been widely proved to be an important marker to indicate the progress of the disease [3,4,5,6,17,18]. Our research further shows that LDH has outstanding practical predictive performance for disease progression from many aspects.
This study has some limitations. First, we did not measure viral load and some patients lacked cytokine testing, which could be factors related to the severity of the disease. Second, we did not test the LDH isoenzymes due to limited resources. LDH isoenzyme analysis in the future may help to identify the source of increased LDH.
In conclusion, LDH was found to be a superior indicator of disease status among COVID-19 patients and had the potential to optimize the clinical management strategy.