Association of D-dimer elevation with inflammation and organ dysfunction in ICU patients with COVID-19 in Wuhan, China: a retrospective observational study

Demographic and clinical characteristics

In total, 158 patients with severe or critical COVID-19 were included in this study. The median age of the patients was 62 years (interquartile range: 56–70 years) and 91 patients (57.6%) were men. Seventy-seven patients (48.7%) died during treatment in the ICU. Forty-nine patients (31.0%) received anticoagulant therapy and 102 (64.5%) received mechanical ventilation [68 (43.0%) received noninvasive ventilation and 81 (51.3%) received invasive ventilation] (Table 1).

D-dimer level on ICU admission was correlated with laboratory test results

The 158 patients were divided into four groups based on their D-dimer level at the time of ICU admission. The D0 group comprised patients with a normal D-dimer level (D-dimer<1.5μg/mL). The D1–D3 groups comprised patients with an abnormal D-dimer level, as follows: D1 (1.5≤D-dimer<10μg/mL), D2 (10≤D-dimer<40μg/mL) and D3 (D-dimer≥ 40μg/mL). The laboratory test results were divided into three categories (abbreviations and ranges of normal values are shown in Supplementary Table 3): high (above normal range), normal (within normal range), and low (below normal range). Heat maps were used to show the numbers of patients in the various laboratory test result categories by D-dimer level (Figure 1). At the time of ICU admission, the number of patients with abnormalities in 46 parameters varied by D-dimer level (Figure 1A). During ICU treatment, the number of patients with abnormalities in 52 parameters varied by D-dimer level (Figure 1B). These parameters reflected coagulation function, inflammation level, liver function, renal function, lipid metabolism and myocardial injury. Furthermore, the values of these parameters significantly differed among the four D-dimer groups (Table 1). We evaluated the correlations between D-dimer level and other laboratory test results. The D-dimer level was positively correlated with the white blood cell count (WBC, r = 0.523, P < 0.001), neutrophil count (r = 0.557, P < 0.001), CRP level (r = 0.441, P < 0.001), procalcitonin level (r = 0.357, P < 0.001), hypersensitive sensitive troponin I level (r = 0.562, P < 0.001), lactate dehydrogenase level (r = 0.633, P < 0.001) and α-Hydroxybutyrate dehydrogenase level (r = 0.654, P < 0.001) for all 158 patients on the first day after ICU admission. On that day, the D-dimer level was negatively correlated with the lymphocyte count (r = -0.419, P < 0.001) and albumin level (r = -0.524, P < 0.001) (Supplementary Table 1). The results of multivariate Cox regression analysis showed increasing odds of in-ICU death with older age, a higher SOFA score, and higher D-dimer, lactate dehydrogenase, apolipoprotein A, and apolipoprotein B levels on ICU admission (Supplementary Table 2).

D-dimer level on ICU admission was associated with 7- and 14-day mortality rates

There were no significant differences in the 7-day mortality rate among the D1, D2, and D3 groups. Within 14 days after ICU admission, the death rates were as follows: D0 group, 5 of 57 patients, 13.66%; D1 group, 14 of 38 patients, 43.57%; D2 group, 21 of 31 patients, 76.12%; and D3 group, 23 of 32 patients,74.84% (Figure 2, Supplementary Figure 2). The 14-day mortality rate was significantly lower in the D0 group than the D1, D2, and D3 groups (all P<0.01), and significantly lower in the D1 group than in the D2 and D3 groups (all P<0.05). However, there was no significant difference in the 14 day mortality rate between the D2 and D3 groups (p=0.963) (Figure 2). In addition, there was no significant difference in the ICU mortality rate between male and female patients (Supplementary Figure 1).

D-dimer level on ICU admission was associated with dynamic changes in organ function during ICU treatment

Organ dysfunction was related to the D-dimer level at the time of ICU admission: in the D3 group, 16 of 32 patients (50%) had abnormal liver function, 5 of 32 (15.63%) had abnormal renal function, 29 of 32 (90.63%) had myocardial injury, 11 of 32(34.38%) had thrombocytopenia. 26 of 32 (81.25%) had severe ARDS and 9 of 32 (28.13%) had DIC. There was no significant difference in organ function status between the D1 and D2 groups at the time of ICU admission (Figure 3). However, patients in the D2 group experienced rapid deterioration of organ function after ICU admission (Figure 3C–3F). In particular, significantly more patients in the D2 group had thrombocytopenia and DIC, compared with the D1 group, within 4-7 days after ICU admission (Figure 3E, 3F). Additionally, significantly more patients in the D2 group had severe ARDS, compared with the D1 group, within 8-14 days after ICU admission (Figure 3D). Notably, by day 14 after ICU admission, the numbers of patients with liver injury, kidney injury, severe ARDS and DIC were similar between the D2 and D3 groups (Figure 3A, 3B, 3D–3F).

Further analysis of the numbers of patients with organ dysfunction over time showed that the growth rates of the numbers of patients with liver injury, myocardial injury, severe ARDS and DIC peaked at 1-3 days after ICU admission (Figure 4A, 4C–4F), while the growth rates of the numbers of patients with kidney injury and thrombocytopenia peaked at 4-7 days after ICU admission (Figure 4B, 4E). Importantly, within 7 days after ICU admission, the growth rates of the numbers of patients with organ dysfunction was higher in the D2 group than in the other three groups (Figure 4).

Dynamic relationships of inflammation and lipid metabolism markers with D-dimer level

To investigate the causes of rapid progression of organ dysfunction in the patients in the D2 group, we compared the dynamic changes in SOFA score, inflammation status and whole blood cells between the D1 and D2 group patients within 7 days after ICU admission. The results showed that the SOFA score was significantly higher in patients in the D2 group than in those in the D1 group, both 1-3 and 4-7days after ICU admission (Figure 5A). With the exception of neutrophil count (P=0.003), there was no significant difference in any other parameter between the two groups, at the time of ICU admission (Figure 5B–5H). At 1-3 days after ICU admission, the levels of inflammation-related markers such as CRP (P=0.018), procalcitonin (P=0.004) and lactate dehydrogenase (P=0.026), were significantly different between the two groups (Figure 5B–5H). At 4-7 days after ICU admission, there were significant differences between the two groups in the levels of inflammation-related markers such as lymphocyte count (P=0.029), IL-6 (P=0.008) and ferritin (P=0.026) (Figure 5D–5G), as well as in the levels of lipid metabolism-related markers such as lipoprotein (low-density lipoprotein, P=0.031; high-density lipoprotein, P=0.001), cholinesterase (P=0.007), cholesterol (P=0.005), apolipoprotein (apolipoprotein A, P=0.001; apolipoprotein B, P=0.035) (Figure 6).

Study design and data collection

This study included 158 patients (≥18 years of age) with confirmed COVID-19 who were admitted to ICUs in Jinyintan Hospital (a designated hospital for patients with COVID-19 in Wuhan, China) between January 20, 2020, and February 26, 2020. All patients were diagnosed with COVID-19 (severe and critical type) in accordance with the World Health Organization interim guidance and Chinese guidelines for COVID-19 management (version 7.0) [27, 28]. Epidemiological data, clinical findings, laboratory test results, and clinical records were obtained from the electronic medical record system. All data were collected by two researchers and then reviewed and confirmed by the authors of this study. The primary clinical outcome, all-cause death was monitored until March 4, 2020 (i.e., the final follow-up). This study was approved by the Research Ethics Commission of Jinyintan Hospital (KY-2020–56.01).

Laboratory procedures

The clinical laboratory of Jinyintan Hospital detected severe acute respiratory syndrome coronavirus 2 nucleic acid using the real-time reverse transcriptase polymerase chain reaction assay as described previously [29]. All laboratory data collected within 14 days after ICU admission were included in the analysis. Laboratory analyses included complete blood count, coagulation function parameters (e.g., D-dimer, fibrinogen, prothrombin time, prothrombin activity, thromboplastin time, Activated partial thromboplastin time, fibrin degradation products, and antithrombin III, inflammation markers (e.g., interleukin [IL]-6, serum ferritin, C-reactive protein [CRP], procalcitonin, and lactate dehydrogenase), and various other assessments (liver function, renal function, electrolytes and myocardial enzymes). All clinical laboratory data were generated by the clinical laboratory of Jinyintan Hospital.

Definition

COVID-19 severity was defined on the basis of the Chinese guidelines for COVID-19 management (version 7.0) [28]. Disseminated intravascular coagulation (DIC) was defined in accordance with the criteria established by the International Society of Thrombosis and Hemostasis [30]. The Berlin definition of acute respiratory distress syndrome (ARDS) was used [31]. Acute kidney injury, liver injury and cardiac injury were defined in accordance with a previous study [32].

Statistical analysis

Continuous data were compared using the Mann–Whitney U-test or Kruskal–Wallis test and are presented as the median (interquartile range). Categorical data were compared using the chi-square test or Fisher’s exact test, and are presented as counts. Univariate Cox regression analysis was used to screen potential predictors. Considering collinearity between variables, multivariate regression ultimately included age, gender, the Sequential Organ Failure Assessment (SOFA) score, D-dimer level, hypersensitive sensitive troponin I, white blood cell count, and levels of lymphocytes, C-reactive protein, lactate dehydrogenase, apolipoprotein A, and apolipoprotein B. Spearman rank correlation analysis was used to explore the association between D-dimer subgroup and clinical characteristics. Kaplan–Meier analysis was used to evaluate the cumulative mortality of patients with different D-dimer levels in the ICU. SPSS 23.0 and GraphPad Prism 7.0 were used for the analyses. P<0.05 was considered statistically significant.