Demographics and baseline characteristics
1089 eligible patients with available NT-proBNP were included and divided into quartiles according to ascending order of NT-proBNP ratio (Fig. 1). Participants with the highest NT-proBNP ratio were older and more likely to have a comorbidity of hypertension or coronary artery disease or chronic kidney disease, but less likely to be female, than participants with lower NT-proBNP ratio. Higher systolic blood pressure was more prevalent among those with greatest NT-proBNP ratio, who also had a elevated level of NT-proBNP, higher white-cell count, lower percentage of lymphocyte, lower monocyte count and more death. In contrast, other characteristics like body mass index, the blood routine items (lymphocyte count, monocyte%, neutrophil%, neutrophil count), renal functions (serum urea and creatinine, eGFR), and systematic inflammatory factors (high-sensitive C-reactive protein, procalcitonin) had no significance between the four groups (Table 1, Supplyment 1).
Table 1
Demographic and clinical characteristics of the patients at baselinea.
|
Quartile 1(N = 271)
|
Quartile 2(N = 270)
|
Quartile 3(N = 275)
|
Quartile 4(N = 273)
|
P value
|
Characteristics
|
(0.003 ~ 0.053)
|
(0.054 ~ 0.131)
|
(0.132 ~ 0.371)
|
(0.376 ~ 77.778)
|
|
Age, median [IQR]-yr
|
54.00 [41.00,60.00]
|
62.00 [51.00, 69.00]
|
66.00 [56.50, 73.50]
|
67.00 [61.00, 74.00]
|
< 0.001***
|
Female sex, no. (%)
|
108 (39.90)
|
151 (55.90)
|
151 (54.90)
|
124 (45.40)
|
< 0.001***
|
Heart rate, median [IQR]
|
92.00 [80.50, 103.00]
|
90.00 [80.00, 100.00]
|
87.00 [79.00, 100.00]
|
89.00 [78.00, 104.50]
|
0.013*
|
Systolic blood pressure (mmHg), median [IQR]
|
128.00 [117.00, 139.00]
|
131.00 [119.00, 145.00]
|
129.00 [115.00, 142.75]
|
135.00 [120.25, 149.00]
|
< 0.001***
|
Diastolic blood pressure (mmHg), median [IQR]
|
82.00 [75.75, 90.00]
|
81.00 [74.00, 90.00]
|
79.00 [71.00, 87.00]
|
80.00 [71.00, 88.00]
|
0.010*
|
Mean arterial pressure (mmHg), median [IQR]
|
44.00 [37.00, 53.00]
|
50.00 [40.75, 60.00]
|
49.50 [40.00, 60.00]
|
54.00 [44.00, 66.00]
|
< 0.001***
|
Hypertension, no. (%)
|
52 (19.30)
|
78 (28.90)
|
92 (33.60)
|
109 (40.20)
|
< 0.001***
|
Coronary artery disease, no. (%)
|
5 (1.84)
|
17 (6.30)
|
28 (10.18)
|
29(10.62)
|
< 0.001***
|
Chronic kidney disease, no. (%)
|
0(0.00)
|
1 (0.37)
|
2 (0.73)
|
7 (2.56)
|
0.037*
|
White-cell count (×109/L),
median [IQR]
|
5.73 [4.64, 7.15]
|
5.40 [4.31, 7.18]
|
5.72 [4.46, 7.20]
|
7.10 [5.15,10.55]
|
< 0.001***
|
Lymphocyte (%), median [IQR]
|
21.60 [14.40, 29.10]
|
22.00 [13.60, 30.20]
|
21.20 [13.40, 29.60]
|
16.60 [8.80, 25.08]
|
< 0.001***
|
<20%, no. (%)
|
120 (44.60)
|
123 (45.90)
|
118 (43.20)
|
154 (57.00)
|
0.005**
|
Monocyte count (×109/L), median [IQR]
|
0.51 [0.38, 0.66]
|
0.45 [0.34, 0.59]
|
0.48 [0.37, 0.64]
|
0.47 [0.35, 0.58]
|
0.015*
|
NT-proBNP (pg/ml), median [IQR]
|
19.00 [9.50, 31.00]
|
70.00 [53.00, 98.00]
|
186.00 [138.00, 276.00]
|
827.00 [504.00, 1817.00]
|
< 0.001***
|
NT-proBNP ratio,
median [IQR]
|
0.03 [0.02, 0.04]
|
0.08 [0.06, 0.11]
|
0.20 [0.16, 0.27]
|
0.80 [0.53, 1.67]
|
< 0.001***
|
In-hospital Death, no. (%)
|
1 (0.40)
|
10 (3.70)
|
20 (7.30)
|
84 (30.80)
|
< 0.001***
|
aThe values shown are based on available data. IQR denotes interquartile range. *P < 0.05, **P < 0.01, ***P < 0.001. |
Receiver operator characteristic (ROC) curve of NT-proBNP ratio for prediction in-hospital death
Receiver operator characteristic (ROC) curve were showed in Fig. 2 to analyze the prognostic value of NT-proBNP and NT-proBNP ratio for prediction in-hospital death with sensitivity and specificity (NT-proBNP, sensitivity 89.47% and specificity 69.64%; NT-proBNP ratio, sensitivity 76.52% and specificity 77.93%). The area under the curve (AUC) for in-hospital death of NT-proBNP and NT-proBNP ratio were 0.868 (95% CI 0.838–0.898, P < 0.0001) (Fig. 2A) and 0.850 (95% CI 0.815–0.884, P < 0.0001) (Fig. 2B). The cutoff point of NT-proBNP and NT-proBNP ratio for prediction in-hospital death were 202.500 (Fig. 2A) and 0.316 (Fig. 2B), respectively. NT-proBNP ratio had a lower sensitivity but higher specificity than NT-proBNP.
Cumulative survival curves of NT-proBNP and NT-proBNP ratio for prediction in-hospital death
All the patients in our research were divided into quartiles according to ascending order of the NT-proBNP ratio. To make this a fair comparison, we also divided all the parcipants in this study into quartiles according to ascending order of NT-proBNP. Cumulative survival curves of NT-proBNP and NT-proBNP ratio for the prediction of in-hospital death were showed in Fig. 3A and Fig. 3B.
Results of univariate and multivariate Cox proportional hazards analyses of in-hospital death
The incidence of in-hospital death was determined for overall and the quartiles of NT-proBNP and NT-proBNP ratio in the COVID-19 patients. The univariate analysis for overall showed that both absoulte value of NT-proBNP (the unadjusted hazard ratio and 95% confidence interval: 1.000, 1.000–1.000, P < 0.0001) and NT-proBNP ratio (the unadjusted hazard ratio and 95% confidence interval: 1.074, 1.059–1.088, P < 0.0001) had a significantly increased risk of in-hospital death in the COVID-19 patients. After multivariate Cox proportional hazards analyses for overall, the mortality of NT-proBNP ratio (the adjusted hazard ratio and 95% confidence interval: 1.141, 1.048–1.242, P = 0.003) persisted while that of NT-proBNP (the adjusted hazard ratio and 95% confidence interval: 0.9999, 0.9999-1.0000, P = 0.129) did not. As compared with the first (lowest) quartile, only the quartile 3 and 4 of NT-proBNP had a significantly increased risk of death, but quartile 2, 3 and 4 of NT-proBNP ratio all had a significantly increased risk of death, in the unadjusted and adjusted analyses of COVID-19 patients (Supplyment 2).
Risk of death according to NT-proBNP and NT-proBNP ratio among the COVID-19 Participants
According to the cutoff point of NT-proBNP and NT-proBNP ratio for prediction in-hospital death, we combined the subgroups into categories designated low-risk (quartile 1 and quartile 2), intermediate-risk (quartile 3), and high-risk (quartile 4), corresponding to NT-proBNP and NT-proBNP ratio levels of the univariate and multivariate Cox proportional hazards analyses (Fig. 3C and Fig. 3D, Supplyment 3). The hazard ratios of NT-proBNP and NT-proBNP ratio in the intermediate-risk and high-risk subgroups were significantly higher than that of low-risk subgroup (Fig. 3C and Fig. 3D, Supplyment 3).