In the univariate analysis, age, BMI, lymphocytes, neutrophils, LDH, CRP, ferritin, D-dimer, albumin, AST, ALT, and creatinine were significantly associated with severe COVID-19. These results show a similar trend to previous studies [2, 6]. The percentage of GGO and consolidation within total lung on chest CT on COVID-19 pneumonia at admission was quantified using an imaging tool. A multiple regression analysis of the factors associated with that proportions showed that the associated factors were sex, BMI, LDH, CRP, and albumin. The association of an exacerbation of COVID-19 pneumonia with these five factors was analyzed by a multivariate Cox proportional hazards analysis using a sample adjusted for background factors by propensity score matching. According to the results, only LDH showed a significant difference, indicating that for every 1 increase in LDH from the normal value, the risk of severity increased 1.003-fold. In other words, the risk increased exponentially by 2-, 3-, 6-, 10-, and 18-fold as the LDH increased by 200, 400, 600, 800, and 1000 above the normal value.
The search for clinical factors associated with the severity, i.e. application of mechanical ventilation and mortality, of COVID-19 has been the subject of many previous studies [1–5]. However, many of those studies were based on univariate analyses and did not adequately adjust for background factors. Some studies using large cohorts, in which background factors were adequately adjusted in multivariate analyses, have been reported [6, 7]. However, inflammatory markers, such as CRP and LDH, which have been frequently reported to be associated with the severity of COVID-19 [18, 19], were not included in those analyses. In addition, there have been reported that chest CT is useful for assessing the severity of COVID-19 [20, 21]. However, few studies have evaluated the association between COVID-19 severity and chest CT by a multivariate analysis with adequate adjustment for background factors. Few studies have employed a method adequately adjusting for background factors to identify clinical factors associated with the COVID-19 pneumonia lesion extent within the total lung on chest CT at admission and predicted severe disease after admission.
There were two novel points associated with the present study. First, we identified the details of clinical factors associated with the progression of COVID-19 pneumonia on chest CT at admission using a multiple regression analysis. We automatically quantified the extent of COVID-19 pneumonia on chest CT using an image analysis system. The fact that we used that numerical values as the outcome of the multivariate analysis is very different from conventional studies using CT severity scores as the outcome, which rely on a subjective semiquantitative evaluation by radiologists . Using our objective quantitative evaluation, we found that the five factors of sex, BMI, LDH, CRP, and albumin were significantly associated with progression of COVID-19 pneumonia on chest CT at admission.
Women reportedly have a higher incidence of ARDS than men, suggesting an association between disease severity and female hormones . A similar mechanism may have led to the significant sex-based difference in severity of COVID-19 pneumonia in the present study.However, a study on the association between COVID-19 pneumonia and CT severity score by a univariate analysis reported no significant difference in sex . In our multiple regression analysis, the partial regression coefficient was as low as 1.11, even though the difference was statistically significant, so the difference in COVID-19 pneumonia severity based on sex on chest CT may not be a problem in real-world clinical practice.
The four other factors of BMI, LDH, CRP, and albumin, have also been reported to be associated with ARDS. Obesity is known to promote an inflammatory response and the endothelial changes seen in ARDS . Since there are reports suggesting an association between COVID-19 pneumonia and the BMI , obese patients may be prone to elevated inflammatory adipokines induced by COVID-19 infection. Inflammatory markers, such as LDH and CRP, have been reported to be useful in predicting the early onset of ARDS and its prognosis [26, 27]. Similarly, many studies have suggested that LDH and CRP may be prognostic factors for severity of COVID-19 [28, 29]. Hypoalbuminemia has been suggested to be associated with ARDS, as it causes increased alveolar capillary permeability and promotes edema formation . Our multiple regression analysis showed a significant inverse correlation between albumin levels and COVID-19 pneumonia, suggesting that edema related to vascular permeability associated with hypoalbuminemia may contribute to the lesion extent of COVID-19 pneumonia on chest CT.
A comparison of the standardized partial regression coefficients of these four factors obtained by a multiple regression analysis revealed that LDH, BMI, CRP, and albumin, in that order, were strongly associated with disease progression of COVID-19 pneumonia. Although these factors have been analyzed and compared as categorical variables in previous studies, we treated them as continuous variables and directly compared the strength of the association. As a result, we revealed for the first time that LDH is the most suitable factor for assessing the COVID-19 pneumonia lesion extent within total lung on chest CT at admission. Since LDH is an enzyme contained in cells that catalyzes the conversion of pyruvate to lactate—the final step of anaerobic glycolysis— and elevated LDH indicates the degree of cell damage associated with tissue hypoperfusion, LDH may most directly reflect the extent of lung damage.
The second novel point of our study is that we found that LDH is a potential predictor of severe disease after admission. In our univariate analysis without considering the effect of confounders, the factors LDH, CRP, and albumin were significantly associated with an exacerbation of COVID-19 pneumonia. In contrast, our multivariate analysis using a cohort adjusted for background factors, including those factors suggested to be associated with COVID-19 severity in a previous study , by propensity score matching, showed that only LDH was significantly associated with a subsequent exacerbation of COVID-19 pneumonia in the clinical course. Although CRP, albumin, and BMI can be used to assess the disease progression of COVID-19 pneumonia on chest CT at admission, they may not be predictors of severe disease after admission.
Many previous studies have reported the association between the severity of COVID-19 including mortality and LDH . However, LDH was treated as a categorical variable, and the cut-off value was set independently by each study, so the results varied among studies. Therefore, it would be difficult to use the results as a specific indicator of risk of COVID-19 severity in real-world clinical practice. We overcame this problem by treating LDH as a continuous variable. We showed that the hazard ratio of LDH for an exacerbation of COVID-19 pneumonia was 1.006 in a univariate analysis and 1.003 in a multivariate analysis using a propensity score-matched cohort. This is a small difference, but it indicated that the larger the increase in LDH from the normal value, the larger this difference becomes, indicating that overestimation was controlled by adjustment for background factors. No study has ever evaluated LDH as a continuous variable and specifically reported the risk of COVID-19 severity according to individual LDH datapoints. It may be possible to predict the risk of COVID-19 exacerbation after admission simply by measuring the LDH value at admission.
We showed that LDH represents the COVID-19 pneumonia lesion extent within total lung at admission and may be a predictor of severe disease after admission. This mechanism may have a biological explanation. Although the details of COVID-19 infection and cytotoxicity are not known and are the subject of active research, recent experiments using lung organoids have reported that cells die after infection with COVID-19 and that the virus may induce cytokine storm and cause cytotoxicity [32, 33]. In other words, the cell death caused by COVID-19 infection of lung cells may lead to extracellular release of LDH and induce a direct increase in serum LDH. However, studies on the route of entry of COVID-19 have suggested that COVID-19 may enter the lungs via the respiratory tract and then spread throughout the body via vascular endothelial cells, entering other organs . In other words, the virus spreads from the lungs to systemic organs and induces a cytokine storm, which is thought to cause multi-organ damage. Through this indirect cellular damage, extracellular release of LDH is also triggered, which is expected to increase the level. CRP is a protein produced in the liver  and adipocytes  in response to inflammatory cytokines such as IL-6, and is thought to reflect the intensity of inflammation or the degree of cytokine storm. It may be an indicator of indirect cytotoxicity like LDH, but it does not reflect the degree of direct cytotoxicity like LDH, so it may be less effective than LDH for evaluating the degree and prognosis of COVID-19 pneumonia. Similarly, the BMI and albumin level are indirectly related to inflammation [37, 38], but they do not reflect direct cellular damage, so they may be less effective as assessment factors than LDH.
Several limitations associated with the present study warrant mention. First, this was a single-center study, which limits the sample size. Since a multivariate analysis is dependent on the sample size and the sample size of the adjusted cohort will be reduced by propensity score matching, further validation in a larger cohort may be necessary. Second, all cases used in this study were patients with mild or moderate symptoms at admission, and patients with severe symptoms were not included. It is therefore necessary to evaluate whether or not similar results can be obtained in patients with severe symptoms. Third, no distinction was made between the mutant strains of COVID-19. Judging from the timing of our study and the prevalent variants of COVID-19 in Japan, it is likely that majority of patients were infected with B.1.1.7 (Alpha strain according to the World Health Organization classification)  or B.1.617.2 (Delta strain) . There have been reports of differences in viral load depending on the strain and consequent differences in severity depending on the viral load ; it may therefore be necessary to include the type of mutant strains as a variable in multivariate analyses. Fourth, the pneumonia imaging findings in the automated image analysis of chest CT may have been overanalyzed as COVID-19 pneumonia, since bacterial and viral pneumonia cannot be distinguished in this manner. Although the settings for the automatic image analysis were based on those used in several previous studies, the accumulation of COVID-19 pneumonia imaging may require more COVID-19-specific image analysis settings. However, in our multivariate analysis, we included factors that are commonly elevated in bacterial pneumonia, i.e. white blood cell count, neutrophil count, and CRP, so we believe that we were able to reduce this effect statistically.