Keywords
COVID-19 - platelet count - thrombosis - low-molecular-weight heparin
Introduction
In the early phase of the coronavirus disease 2019 (COVID-19) pandemic, reports revealed
that D-dimer levels in plasma were greatly elevated among the severely ill patients,
and that anticoagulation seemed to improve survival.[1]
[2] The virus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was found
to trigger thromboinflammation and endothelial damage, leading to an increase in thrombin
generation and altered fibrinolysis.[3]
[4]
[5]
[6] Platelets appear to be hyperactive and play a pivotal role, but how SARS-CoV-2 interacts
with parent cells (megakaryocytes) and/or platelets is under intense research.[7]
[8]
Mild thrombocytopenia at admission is associated with severe disease and mortality.[9]
[10]
[11] However, the association between the trends of platelet counts over time and unfavorable
outcomes is not fully elucidated even though a recent publication has reported that
deceased count is associated with mortality.[9]
[10]
[12]
In hospitalized COVID-19 patients, venous thromboembolism (VTE) is the most common
thrombotic manifestation, with a reported prevalence in the intensive care unit (ICU)
population of around 30% in the early phase of the pandemic, dominated by pulmonary
embolism (PE).[13]
[14]
[15]
[16] The incidence of VTE was then three times higher in COVID-19, compared with other
severe respiratory infections.[17] Notably, a high incidence of VTE has been described despite prophylactic anticoagulation
treatment in COVID-19.[18] Antithrombotic prophylaxis with low-molecular-weight heparin (LMWH) seems to improve
outcome, although final results from large randomized clinical trials to provide guidance
on optimal LMWH dosage are still lacking.[19]
[20]
[21]
[22]
[23]
[24] Observational studies have so far reported different results regarding if higher
doses of LMWH are efficient to reduce mortality and thrombosis in severe COVID-19,
this depending on cohort composition, study design, and outcome, as well as differences
in antithrombotic guidelines between institutions.[12]
[21]
[22]
[23]
[25] At Karolinska University Hospital, reports of thrombosis despite antithrombotic
prophylaxis resulted in hospital-wide recommendations for increased treatment with
LMWH in severe COVID-19 (April 9, 2020).[26] These hospital-wide recommendations were consistent with International Society on
Thrombosis and Haemostasis (ISTH) guidelines published at the end of May.[27]
Our aim was to see if the dynamics of D-dimer levels and platelet counts during the
first week in hospital were associated with death and thrombosis. Further, we investigated
if intensified anticoagulation in severe COVID-19 had any effect on D-dimer levels
and platelet counts, and if they were related to survival and thrombosis.
Methods
This is a cohort study based on retrospectively collected data from patients with
verified COVID-19 at Karolinska University Hospital (Karolinska), Stockholm, Sweden,
a tertiary referral hospital. The project was given ethics approval (D-nr 2020–01752)
by the Swedish Ethics Review Authority.
The cohort comprised patients admitted to Karolinska's two sites: Huddinge and Solna,
between March 5 and April 22, 2020. They were followed from admission until discharge
or death, last day of data collection being June 30, 2020. All patients were identified
as having COVID-19 by way of at least one interleukin 6 (IL-6) test result recorded
in the Karolinska University Laboratory (KUL) database, and SARS-CoV-2 was confirmed
with a positive reverse transcription polymerase chain reaction test. IL-6 was included
in the centralized COVID biobank and supposed to be assayed in all in-hospital COVID-19
patients. All included patients were older than 18 years of age at admission and assessed
as having severe disease defined by SaO2 <94% in room air at sea level.[28]
[29] Exclusion occurred if follow-up was unknown or if the patient contracted the disease
while hospitalized. Patients could have been transferred from another hospital in
Stockholm county, but follow-up was exclusive to Karolinska. The referring hospitals
are connected to the KUL enabling collection of laboratory information from admittance
and through follow-up. Patients admitted to the ICU, a unit where intubation and mechanical
ventilation is possible, had respiratory failure and were all assessed by intensivists
prior to ICU admission. Ward patients were not transferred to the ICU at any time
during their hospital stay.
According to hospital-wide recommendations, first published online on March 27, COVID-19
patients were recommended to receive antithrombotic prophylaxis with LMWH in-hospital.
Revised hospital-wide recommendations on April 9 (presented in [Table 1]) included modified recommendations with instructions on who should be considered
for enhanced prophylactic treatment (i.e., ICU ward or equivalent, D-dimer levels
over 3 mg/L FEU [fibrinogen-equivalent units], fibrinogen over 8 mg/L, coexisting
cancer, previous thrombosis, or known thrombophilia), and treatment duration after
discharge from hospital.[26] In ICU-specific recommendations (April 3), allowance was given for higher doses
of LMWH than for regular prophylaxis, but these were not specified. Patients with
diagnosed thrombosis or a strong clinical suspicion of PE received treatment doses
of LMWH. Very few patients received unfractionated heparin.
Table 1
Hospital recommendations on thromboprophylaxis and antithrombotic therapy with LMWH
to patients with COVID-19, implemented March 27 and updated April 9, 2020
Regular thromboprophylaxis[a]: all hospitalized patients
|
Adm.
|
Body weight
|
<50 kg
|
50–90 kg
|
<90 kg
|
Inj., sc
|
Dalteparin (sc)
|
2,500 IE qd
|
5,000 IE qd
|
75 IE/kg qd
|
Inj., sc
|
Tinzaparin (sc)
|
2,500 IE qd
|
4,500 IE qd
|
50 IE/kg qd
|
Enhanced thromboprophylaxis added after update on April 9[a]
[b]: to patients with symptoms corresponding to need of ICU care, D-dimer over 3 mg/L
FEU, fibrinogen over 8 g/L, cancer, previous thrombosis, or known severe thrombophilia
|
Adm.
|
Body weight
|
<50 kg
|
50–90 kg
|
<90 kg
|
Inj., sc
|
Dalteparin
|
2,500 IE bin
|
5,000 IE bin
|
75 IE/kg bin
|
Inj., sc
|
Tinzaparin
|
2,500 IE bin
|
4,500 IE bin
|
50 IE/kg bin
|
Treatment dose[a]
[b]: to patients with confirmed thrombosis/pulmonary embolism or strong suspicion of
thromboembolism
|
Adm.
|
|
|
|
|
Inj., sc
|
Dalteparin
|
200 IE/kg per day, qd/bin
|
|
|
Inj., sc
|
Tinzaparin
|
175 IE/kg per day, qd/bin
|
|
|
Infusion
|
Heparin
|
400–500 IE/24 h
|
|
|
Abbreviations: bin, twice daily; FEU, fibrinogen-equivalent unit; ICU, intensive care
unit; inj., injection; qd, once daily; sc, subcutaneous.
a If platelet count is over 30 × 109/L and no bleeding symptoms.
b Dose reduction should be considered if reduced kidney function (eGFR < 30 mL/min).
Outcomes
In-hospital death covered patients who died from COVID-19 in hospital. Thrombosis
covered venous, arterial, and catheter-related thromboses. VTE included symptomatic
PE and/or deep vein thrombosis, diagnosed by either computed tomography or duplex
ultrasonography. Strongly suspected clinical PE was defined by stated in free text
in the medical records if the patient received full-dose anticoagulation and/or systemic
thrombolysis and was clinically too unstable to undergo confirmative examination.
Arterial thrombosis was acute ischemic stroke defined as altered focal neurological
status and confirmed by typical findings in computed tomography. Acute myocardial
infarction was noted if found in the patient medical records as text or as ICD-10
code included in I.21. Catheter-related thromboses were both arterial and venous and
verified by duplex ultrasonography.
Data Collection
Descriptive data of the patients were extracted retrospectively from medical records
and included information on age, sex, body mass index (BMI), hypertension, diabetes
mellitus type 2, chronic obstructive pulmonary disease (COPD), heart disease (i.e.,
ischemic heart disease, arrhythmia, and cardiomyopathy), continuous treatment with
anticoagulants, and time of onset of COVID-19 symptoms. During the study period, we
collected information of hospital admission, start of LMWH (prophylactic or treatment
dose), maximum dose of LMWH during hospitalization, anticoagulant-related bleeds,
thrombotic events, need of mechanical ventilation, in-hospital death, and length of
hospital stay. Anticoagulant-related bleeds were noted if found in the medical records
and divided into levels of severity of bleeding according to ISTH criteria.[30] Discharge from hospital was either to home or an inpatient rehabilitation facility.
From the KUL database, data of D-dimer levels in plasma and platelet counts in blood
were collected continuously during the study period or for a maximum of 21 days. Coagulation
laboratory parameters were measured at KUL, and automatically analyzed ([Supplementary Methods]: Laboratory Analysis, available in the online version).
Statistical Methods
Levels
Descriptive data for continuous and categorical variables were summarized as medians
with interquartile range (IQR) and counts with percentages (%), respectively.
First levels of D-dimer and platelet counts were only used if they were taken within
3 days from hospital admission. All analysis concerning first levels was adjusted
for if LMWH was given the same day as the first level was measured, or after. To analyze
speed and direction of change of D-dimer levels and platelet counts, person-specific
linear regression models were fitted to D-dimer levels and platelet counts occurring
within the first 7 days in hospital. The model used was univariate ordinary least-squares
regression.
where yi
is the outcome, β
0 is the intercept, β
1 the mean increase in y per day (slope), and εi
the error term. The person-specific slopes and intercepts were used as summary measures
to describe the starting levels and speed and direction of change in further analysis.
The slopes can be interpreted as the mean change per day for each patient within the
first days in hospital. For thrombosis, only information prior to the occurrence was
included in the analysis.
Analysis
The relationship of D-dimer levels and platelet count dynamics (first level and estimated
slopes) to death and thrombosis was estimated using multivariable logistic regression
with robust standard errors. We adjusted for the admission date of each patient to
account for the constant improvement in the treatment of COVID-19 patients during
the study period.
To continue we investigate the effect of protocol change, i.e., the modified antithrombotic
hospital-wide recommendations, published on April 9, 2020 ([Supplementary Methods], available in the online version). A random intercept linear mixed effects regression
model was applied to calculate the daily means of D-dimer levels and platelet counts
among patients during their first 7 days at hospital. Person-specific random intercepts
were used to take into account the differences in disease severity at hospital admission
and to control for the intra-person correlation. Probability of death and thrombosis
before versus after protocol change was estimated using logistic regression.
Interrupted time series regression was also used to further investigate the effect
of protocol change. The model was mixed effects linear regression with random intercept
and the parameter of interest was the interaction of time and time period after protocol
change, which can be interpreted as the difference in speed of change in the outcome
variable before and after protocol change. For the whole cohort, all models were adjusted
for established risk factors associated with the severity of COVID-19, i.e., age,
sex, diabetes mellitus type 2, heart disease, and COPD. BMI was not used because of
too much missing data. We also adjusted for continuous treatment with DOACs or warfarin,
hypothesizing a protecting effect on the outcomes.[31]
Graphs
Graphs were produced using predictions from unadjusted linear regression ([Fig. 1]) and logistic regression models ([Fig. 2]). The dynamics of change in D-dimer levels and platelet counts were allowed to be
nonlinear using restricted cubic splines with three knots. For the linear regression
model, clustered robust standard errors were used to account for the intra-person
correlation.
Fig. 1 The predicted probability of death and thrombosis regarding the different slopes
(rate of change in levels) of D-dimer levels and platelet counts during the first
week in hospital in patients with COVID-19. The blue line is the predicted probability of death and the red line is for thrombosis in connection with the different slopes of biomarker levels during
the first week in hospital (rate of change in levels; β-coefficients estimated by
linear regression); D-dimer level (A) and platelet count (B). The y-axis is the predicted probability of death and thrombosis, the x-axis in (A) is the D-dimer slope (β-coefficients) and in (B) the change in mean daily platelet count. Modeling was performed by applying logistic
regression and restricted cubic splines with three knots.
Fig. 2 Predicted daily mean D-dimer level and platelet count from March 5 to April 4, 2020
on patients with COVID-19. Prediction was based on measurements every given day on
patients treated within the first 7 days at hospital; the red line represents D-dimer level and the blue line platelet count. The shaded areas are 95% confidence intervals. The green line is the lowest smoothed day-specific mean percentage of patients admitted that day,
receiving either intensified prophylactic treatment or full-dose with LMWH (enhanced
dose) at any time during their hospital stay. Modeling was performed by using linear
regression with clustered robust standard errors. Restricted cubic splines with three
knots were used to allow flexibility over time. LMWH, low-molecular-weight heparin.
The analysis was performed using Stata, version 15 (StataCorp, 2017, Stata Statistical
Software: Release 15; College Station, Texas, StataCorp LLC). p-Values smaller than 0.05 were considered statistically significant.
Results
During the study period, March 5 to April 22, 429 patients with severe COVID-19 were
admitted to Karolinska and included in the final cohort ([Supplementary Fig. S1], available in the online version). The median age was 59 years (IQR: 50–66). Thirty-seven
percent were women in the total cohort, the figure being 21% in the ICU group ([Table 2]).
Table 2
Admission characteristics, subgrouped to level of care
Variables
|
All
N = 429
|
ICU
N = 246
|
Ward
N = 183[a]
|
Median age, y (IQR)
|
59 (50–66)
|
59 (51–64)
|
60 (46–69)
|
19–39 years of age
|
41 (9.6)
|
19 (7.7)
|
22 (12)
|
40–69 years of age
|
316 (74)
|
198 (80)
|
118 (65)
|
70–93 years of age
|
72 (17)
|
29 (12)
|
43 (23)
|
Women
|
159 (37)
|
50 (21)
|
109 (60)
|
Median BMI, kg/m2 (IQR)[b]
|
29.0 (26.1–32.5)
|
29.0 (26.4–31.9)
|
28.4 (26.0–32.9)
|
Severe obesity (BMI 35 or more)
|
48 (14)
|
24 (12)
|
24 (13)
|
Hypertension
|
163 (38)
|
95 (38)
|
68 (37)
|
Diabetes mellitus type 2
|
117 (27)
|
68 (28)
|
58 (32)
|
Heart disease[c]
|
68 (16)
|
25 (10)
|
43 (23)
|
Chronic obstructive pulmonary disease
|
73 (17)
|
43 (17)
|
30 (16)
|
Anticoagulant treatment
|
19 (4.4)
|
5 (2.0)
|
17 (9)
|
In age group 70 years and older
|
11 (2.6)
|
2 (0.8)
|
9 (5)
|
Median days with Covid-19 symptoms (IQR)
|
7 (5–10)
|
7 (6–11)
|
7 (5–10)
|
Abbreviations: BMI, body mass index; ICU, intensive care unit; IQR, interquartile
range.
Note: All data are presented as count and percentage (%) unless otherwise noted.
a 35 patients restricted to ward.
b Missing results in 22% of cases.
c Includes ischemic heart disease, arrhythmia, and cardiomyopathy.
Days in Hospital, Death, and Thrombosis
COVID-19 patients were hospitalized for a median period of 15 days (IQR: 8–27). Ninety-four
patients died from COVID-19 at hospital, with a 19% case-fatality rate in 30 days
and a cumulative incidence of 22%. Thrombotic events had a 30-day incidence of 18%
and were six times more frequent in patients at the ICU compared with the ward (cumulative
incidence of 30% (83/429) and 5% (9/183), respectively). The most common thrombotic
event was PE, found in 65 patients (15 ICU patients diagnosed by high clinical likelihood
of PE, but judged too unstable to undergo further examination). Arterial thrombosis
occurred in five patients ([Table 3]).
Table 3
Outcomes within 7 and 30 days, antithrombotic treatment, and bleeding
Outcomes, subgrouped to level of care
|
All (N = 429)
|
ICU (N = 246)
|
Ward (N = 183[a])
|
Median days of in-hospital stay (IQR)
|
15 (8–27)
|
23 (16–37)
|
8 (5–13)
|
Patients discharged within 7 days
|
93 (22)
|
13 (5.3)
|
80 (44)
|
Patients discharged within 30 days
|
337 (79)
|
158 (64)
|
179 (98)
|
In-hospital death
|
94 (22)
|
73 (30)
|
21 (11)
|
Within 7 days
|
23 (5.4)
|
12 (4.9)
|
11 (6.0)
|
Within 30 days
|
80 (19)
|
60 (24)
|
20 (11)
|
Thrombosis
|
83 (19)
|
74 (30)
|
9 (4.8)
|
Within 7 days
|
31 (7.2)
|
23 (9.3)
|
8 (4.4)
|
Within 30 days
|
77 (18)
|
68 (28)
|
9 (4.9)
|
Venous thromboembolism
|
72 (17)
|
65 (26)
|
7 (3.8)
|
Pulmonary embolism[b]
|
65 (15)
|
58 (24)
|
7 (3.8)
|
Arterial thrombosis[c]
|
5 (1.1)
|
4 (1.6)
|
1 (0.5)
|
Catheter-related thrombosis
|
6 (1.4)
|
5 (2.0)
|
1 (0.5)
|
Mechanical ventilation
|
212 (49)
|
212 (86)
|
0
|
Within 7 days
|
202 (95)
|
202 (82)
|
0
|
Within 30 days
|
212 (100)
|
212 (100)
|
0
|
Antithrombotic treatment with LMWH
|
367 (87)
|
240 (100)
|
127 (69)
|
Intermediate or full dose
|
210 (57)
|
183 (76)[d]
|
27(15)[e]
|
Standard prophylactic dose only
|
156 (43)
|
57 (24)
|
99 (54)
|
Bleeding on antithrombotic treatment[f]
|
27 (6.3)
|
27 (11)
|
0
|
Fatal
|
1
|
1
|
0
|
Major[g]
|
10
|
10
|
0
|
Clinically relevant minor
|
13
|
13
|
0
|
All data are presented as count and percentage (%) unless otherwise noted. IQR is
inter-quartile range
a Thirty-five patients restricted to ward care level.
b Fifteen were clinically strongly suspected but not confirmed at ICU, given full-dose
anticoagulant treatment, and four patients received thrombolysis
c Four patients had ischemic stroke and one critical ischemia in the left leg.
d Fifteen patients treated with unfractionated heparin.
e Three remained on DOAC.
f Four ICU patients received blood transfusion.
g Four patients had intracranial bleeding after ischemic stroke.
Antithrombotic Treatment with Low-Molecular-Weight Heparin and Bleeding during Treatment
Three hundred and sixty-seven patients were treated with LMWH, with more than half
receiving an intermediate or full dose, escalating from the initial prophylactic dose
([Table 3] and [Fig. 2]). At the ICU, almost all (240/246) patients received antithrombotic treatment, in
contrast to those outside the ICU, where only 69% received such treatment. Bleeding
after antithrombotic treatment occurred only in ICU patients (cumulative incidence:
11%). According to ISTH classification, 24 of 27 bleeds were categorized as clinically
relevant and 22 of them occurred during intermediate- or full-dose regimens with LMWH.[30] Prior to the bleeds, eight patients received infusion with heparin, two patients
had been given central thrombolytic treatment and seven, platelet inhibitors. Fifteen
of the patients suffered from a thrombotic event prior to their bleed. Four bleeds
were intracranial after ischemic stroke, classified as major ([Table 3]).
Associations between D-Dimer Levels, Platelet Counts, Death, and Thrombosis
D-Dimer levels and platelet counts (first levels at hospital, peak levels, and day
of peak) are presented in [Table 4]. We performed logistic regression analysis to assess if the risk of death and thrombosis
was associated with the first D-dimer level and platelet count in hospital. Whereas
the first D-dimer level at hospital was significantly associated with thrombosis (odds
ratio [OR]: 1.10; 95% confidence interval [CI]: 1.04–1.16), it was not associated
with death ([Table 4]). If categorized, the OR for thrombosis would increase with each category compared
with the lowest D-dimer level percentile (≤0.96 mg/L FEU), first being significant
at levels three times higher than normal (<0.5 mg/L FEU). Regarding death, the OR
increased compared with the lowest D-dimer level percentile up to 4.1 mg/L FEU, only
significantly associated in the stratum 1.69–4.1 mg/L FEU (OR: 2.79; 95% CI: 1.30–5.93).
Both death and thrombosis were associated with peak D-dimer levels (OR: 1.07; 95%
CI: 1.05–1.10 and 1.09; 95% CI: 1.06–1.11, respectively). We found no significant
associations with either the first or the peak platelet count in relation to death
and thrombosis ([Table 4]).
Table 4
D-Dimer and platelet count levels in association with death and thrombosis
|
All
|
Death
|
Thrombosis[a]
|
|
N = 429
|
N = 94
|
OR(a)
|
[95% CI]
|
p
|
N = 83
|
OR(a)
|
[95% CI]
|
p
|
D-Dimer levels in plasma (mg/L FEU)
|
First level at hospital[b], median (IQR)
|
0.96 (1.70–0.59)
|
1.27 (0.77–2.80)
|
0.99
|
[0.95–1.03]
|
ns
|
1.29 (0.76–3.95)
|
1.10
|
[1.04–1.16]
|
***
|
0th–50th percentile (0–0.95 mg/L FEU)
|
179 (42)
|
24 (26)
|
ref.
|
|
|
25 (30)
|
ref.
|
|
|
50th-75th percentile (0.96–1.68 mg/L FEU)
|
90 (21)
|
21 (22)
|
1.52
|
[0.73–3.16]
|
ns
|
15 (18)
|
1.15
|
[0.55–2.40]
|
ns
|
75th-90th percentile (1.69–4.1 mg/L FEU)
|
53 (12)
|
20 (21)
|
2.79
|
[1.30–5.93]
|
**
|
15 (18)
|
2.58
|
[1.20–5.55]
|
**
|
90th-100th percentile (4.2–35 mg/L FEU)
|
37 (8)
|
13 (14)
|
1.95
|
[0.79–4.83]
|
ns
|
17 (20)
|
4.80
|
[2.09–11.0]
|
***
|
Missing levels
|
70 (16)
|
16 (17)
|
1.33
|
[0.66–2.92]
|
ns
|
11 (13)
|
1.07
|
[0.45–2.55]
|
ns
|
Peak level, median (IQR)
|
3.85 (1.25–10.6)
|
11.8 (3.70–34)
|
1.07
|
[1.05–1.10]
|
***
|
13.0 (5.45–35)
|
1.09
|
[1.06–1.11]
|
***
|
Day at peak level
|
6 (2–11)
|
5 (3–10)
|
|
|
|
5 (2–7.5)
|
|
|
|
Slope (β) from admission to 7 days
|
No change
|
70 (16)
|
12 (13)
|
ref.
|
|
|
10 (12)
|
ref.
|
|
|
Decrease in slope
|
65 (15)
|
4 (4)
|
2.36
|
[0.73–7.69]
|
ns
|
6 (7)
|
1.64
|
[0.54–4.92]
|
ns
|
Increase in slope
|
190 (44)
|
56 (60)
|
6.06
|
[2.10–17.5]
|
***
|
50 (60)
|
3.05
|
[1.17–7.91]
|
*
|
Missing
|
104 (24)
|
22 (23)
|
2.38
|
[0.72–7.87]
|
ns
|
17 (21)
|
1.18
|
[0.64–5.54]
|
ns
|
Platelet counts in blood (×109/L)
|
First count at hospital[a], median (IQR)
|
201 (159–260)
|
191 (151–255)
|
|
|
|
220 (164–279)
|
|
|
|
Less than 150 ×109/L
|
80 (19)
|
22 (24)
|
1.44
|
[0.77–2.70]
|
ns
|
14 (17)
|
0.91
|
[0.46–1.80]
|
ns
|
150–350 × 109/L
|
320 (75)
|
68 (72)
|
ref.
|
|
|
59 (71)
|
ref.
|
|
|
Over 350 × 109/L
|
27 (6)
|
3 (4)
|
0.37
|
[0.77–2.70]
|
ns
|
9 (11)
|
2.30
|
[0.90–5.84]
|
ns
|
peak count, median (IQR)
|
460 (334–577)
|
408 (288–506)
|
1.00
|
[1.00–1.00]
|
***
|
420 (317–528)
|
1.0
|
[1.00–1.00]
|
ns
|
Day at peak count
|
9 (5–13)
|
8.5 (4–13)
|
|
|
|
7 (2–12)
|
|
|
|
Slope (β) from admission to 7 days
|
No change or decrease in slope
|
110 (26)
|
43 (46)
|
ref.
|
|
|
31 (37)
|
ref.
|
|
|
Mean increase on 10–35 × 109/L platelets per day
|
219 (51)
|
44 (47)
|
0.48
|
[0.27–0.85]
|
*
|
36 (43)
|
0.47
|
[0.26–0.85]
|
*
|
Mean increase more than 35 × 109/L platelets per day
|
97 (23)
|
6 (6)
|
0.16
|
[0.06–0.41]
|
***
|
13 (16)
|
0.36
|
[0.17–0.80]
|
*
|
Abbreviations: CI, confidence interval; FEU, fibrinogen-equivalent unit; IQR, interquartile
range; OR; odd ratio, (a): adjusted for age, sex, heart disease, diabetes mellitus
type 2, chronic obstructive pulmonary disease, anticoagulant treatment, date on admission
and when LMWH was given in relation to either first D-dimer level measurements or
platelet count.
Note: All data are presented as count and percentage (%) unless otherwise noted. p-Values are marked with * for <0.05, ** for <0.01, and *** for <0.005.
a Only information prior to thrombosis was included in the analysis.
b Within first 3 days.
We conclude that the higher the D-dimer level was, the higher the risk was for thrombosis,
in contrast to the risk of death where higher levels were not associated with increased
risk. First and peak levels of platelet count were not good markers. This is well
illustrated in [Supplementary Fig. S2] (available in the online version) showing the probability for death and thrombosis
in relation to D-dimer level and platelet count.
Trend Analysis of D-Dimer Levels and Platelet Counts in the First Week at Hospital,
and Their Association with Death and Thrombosis
To understand the direction and velocity of change, dynamics, and their implications
as regards D-dimer levels and platelet counts during the first 7 days at hospital,
we calculated a slope based on the rate of change in levels. This was followed by
predicting the probabilities of death and thrombosis in relation to the rate of change
in D-dimer levels and platelet counts, graphically illustrated in [Fig. 1]. The probabilities of death and thrombotic events increased with a faster increase
of D-dimer levels in plasma. The probability of death seemed to increase even faster
than that of thrombosis. To further investigate this, we categorized patients into:
no or minor change, increase or decrease in slope. We could confirm that a mean increase
in slope of more than 0.05 mg/L FEU per day was significantly associated with both
death and thrombosis (OR: 6.06; 95% CI: 2.10–17.5 and 3.12; 95% CI: 1.12–8.11) compared
with no change ([Table 4]). This corresponded to an estimated cumulative incidence of death of 29.5% if the
slope increased compared with 11.5% if there was no change in D-dimer levels. For
thrombosis, the cumulative incidence rates were 26.5 and 9.4%, respectively. In sensitivity
analysis, we divided the intercept in the regression models at the 75th percentile.
We observed that an even decrease in slope was associated with death and thrombosis
when the intercept was high compared with the reference with minor change or decrease
at a lower intercept. However, this group with decreasing D-dimer levels from a high
starting point was very small, which resulted in uncertain estimates and wide CIs
(data not presented). Still, this was visualized in [Fig. 1], where a decrease in D-dimer levels also seemed to affect the probabilities of thrombosis
and death.
Rate of change in platelet counts during the first week at hospital was striking.
The faster the platelet counts increased, the lower the probability of dying or having
a thrombotic event, as graphically presented in [Fig. 1]. This was well supported by the results of regression analysis; a mean increase
of more than 35 × 109/L platelets per day decreased the risks of death (OR: 0.16; 95% CI: 0.06–0.41) and
thrombosis (OR: 0.36; 95% CI: 0.17–0.80) compared with no change or a decrease in
slope. For platelet counts with a daily mean increase of 10–35 × 109/L, the ORs were 0.48 (95% CI: 0.27–0.85) and 0.47 (95% CI: 0.26–0.85) for death and
thrombosis, respectively ([Table 4]). The importance of increasing platelets was further indicated by the calculated
cumulative incidence of death of 15.8% for patients with a daily mean increase in
platelets of 10 × 109/L or more compared with 39.1% for patients with no change or decrease. Corresponding
cumulative incidence rates of thrombosis were 15.6 versus 28.4%. There was no significant
interaction between the first platelet count at hospital and how fast the platelet
count increased per day (data not presented), this might indicate that the dynamics
were independent of the first level. To summarize, the rate of increase in platelet
count demonstrated an impressive association with risk reduction in both thrombosis
and death. The pattern of D-dimer levels was less robust due to a lower number of
observations compared with platelet counts ([Fig. 1] and [Table 4]).
Comparison between the Time Periods Before and After the Modification of Antithrombotic
Hospital-Wide Recommendations
The daily mean levels of D-dimer and platelet counts were estimated, illustrated graphically
in [Fig. 2]. Added to the figure is the lowest smoothed percentage of patients admitted daily
who were given high antithrombotic prophylactic doses or a treatment dose of LMWH
that could have been increased during their hospital stay. Modified hospital-wide
recommendations on anticoagulation with added enhanced prophylactic doses were published
on April 9, 2020 ([Table 1]). We investigate if this modification had any effect on average D-dimer levels and
platelet counts among patients. For each patient, the levels from the first 7 days
in hospital were included in the analysis. When comparing the time periods before
and after the modification, the daily mean levels changed significantly after April
9; D-dimer levels decreased in plasma by 1.93 mg/L FEU (1.00–2.87) mg/L FEU and platelet
counts increased by 65 (54–76) × 109/L on average. Interrupted time series regression analysis estimated the difference
in the rate of change (β-coefficients) before and after the protocol change to be
−0.13 for D-dimer (95% CI: −0.32–0.06) and 5.95 for platelet counts (95% CI: 4.03–7.87).
When comparing the cumulative incidence of death and thrombosis before and after the
shift in the hospital's anticoagulation recommendation on April 9, 24.5% died before
compared with 13.5% after the shift. For thrombosis, the cumulative incidence rates
were 21.6 versus 11.5%. This corresponded to a significant risk reduction for death
and thrombosis if the patient was admitted after the modification of hospital-wide
recommendations (OR: 0.48; 95% CI: 0.25–0.92 and 0.35; 95% CI: 0.17–0.72).
Discussion
We found that the dynamics of platelet counts in combination with D-dimer levels were
associated with death and thrombosis in our cohort of 429 severely ill patients with
COVID-19. As faster the platelet counts increased, the more the risk declined. A trend
analysis of daily mean levels graphically shows a gradually decreasing D-dimer levels
and increasing platelet counts among patients over the course of the study period.
These changes in levels significantly coincided with the implementation of hospital-wide
recommendations of intensified anticoagulation and decreased relative risks of mortality
and thrombosis.
Our study population is similar in characteristics to previous studies of patients
with severe COVID-19 infections during the first phase of the pandemic in Europe.[14]
[18]
[32]
[33] Similarities included thrombotic events were six times more frequent in ICU patients
versus patients treated at the general ward and the most common thrombotic event,
by far, was PE, found in 25% of the ICU patients.[12]
[14]
[18]
[23]
[33] In contrast to previous studies, we have completed follow-up of all patients until
hospital discharge or death. This may explain why we at the ICU experienced a higher
cumulative death rate of 30%, of whom 84% died within 30 days, compared with rates
of 13 to 24% in previous studies, all with shorter follow-up times.[14]
[34]
Bleeding has been reported as a spontaneous event or an unwanted complication of anticoagulation
treatment in COVID-19 cases, with a prevalence rate of 2 to 21%. The wide range has
been explained by the dosage of anticoagulation and the combination with platelet
inhibitors or fibrinolytic therapy, but also by more frequently reported bleeds among
mechanically ventilated ICU patients.[2]
[19]
[35] This was also seen in the current cohort; bleeds occurred only in the ICU-treated
patients, with a cumulative incidence of 11%, and the majority had experienced a thrombotic
event before the bleed ([Table 3]). Our results support theories of COVID-19 coagulopathy without prominent bleeding.[36]
[37] Taken together, bleeding manifestations seem to be a later complication of severely
ill ICU-treated COVID-19 patients, typically in combination with antithrombotic treatment
other than LMWH alone.
In our cohort, we observed the same pattern with elevated first D-dimer levels and
peak as in previous studies ([Table 4]).[32]
[38]
[39] The higher the first D-dimer level was in hospital, the higher the risk was for
thrombosis, in contrast to the risk of death where levels over 4.1 mg/L FEU did not
associate with increased risk ([Table 4]). More information was provided when studying the dynamics in D-dimer levels over
the first week. Even if the patient had a low first D-dimer level, a fast increase
would indicate a higher risk of thrombosis and death than unchanged levels. We could
also note that if a patient's D-dimer level dropped from a higher starting point,
the increased risk may remain, but this needs to be confirmed in other studies. This
highlights that first D-dimer levels are far better assessed in regard to the following
levels to truly estimated risk of death and thrombosis. The slope might indicate an
ongoing thromboinflammatory process that gets out of control in patients with potential
risk factors. In other studies, with thromboelastography, delayed or absent fibrinolysis
was found as a sign of this disturbance of normal anticoagulation.[37]
[40]
[41]
In contrast to previous studies, first platelet count was not a good risk marker of
death and thrombosis in our cohort as we found no association with thrombocytopenia.[11]
[42] Again, studying the dynamics in the first week in hospital provided much more captivating
results. A fast increase of more than 35 × 109/L platelets per day demonstrated a robust association with a large risk reduction
for both thrombosis and death ([Table 4] and [Fig. 1]). If used as a dynamic risk marker, the platelet count may remain within the reference
range but still be of high clinical value and should not be disregarded. Platelets
appear to be hyperactive and play a key role, but how SARS-CoV-2 interacts with the
megakaryocytes and/or platelets is still unknown.[8] The thrombi seen at autopsy are platelet-rich, indicating that consumption could
be a plausible explanation, but other reasons such as increased platelet clearance
and/or an impaired immune response in host-defense also contribute to an inability
to increase platelet counts.[9]
[11]
[43]
[44] On the whole, dynamics of D-dimer levels and platelet counts have not previously
been assessed in depth in this context and our findings emphasize to use them both
as dynamic markers in severe COVID-19 and to not only focus on first levels.
To continue, there was a considerable difference in patients' daily mean of D-dimer
levels and platelet count in the first week in hospital after the hospital-wide recommendations
of intensified anticoagulation was implemented in April 9. In patients admitted after
April 9, the risks of thrombosis and death were significantly reduced in comparison
with patients admitted earlier and this decrease was confirmed in a large Swedish
study.[45] As visualized in the graph in [Fig. 2], the changes in D-dimer levels and platelet counts began before the hospital-wide
recommendations of intensified anticoagulation was published, indicating that a combination
of factors contributed to these changes during the study period. The gained experience
over time, i.e., improved general care, ventilation adjustment as well as fewer patients
intubated, contributed to an improved outcome in hospital in cases of severe COVID-19
in the first phase of the pandemic. Antiviral medication and cortisone could theoretically
have had some impact but were not systematically given to patients at Karolinska with
severe COVID-19 at that time.[46]
[47] In numbers, only 14 patients were given Remdesivir (10 in ICU, 4 in ward) and none
of the patients received dexamethasone. Undoubtedly, the new hospital-wide recommendations
of enhanced prophylaxis with LMWH coincided with the improved outcome pattern with
a decline in risk of both death and thrombosis. Previous studies have pointed in different
directions, both in line with our results and the opposite. This discrepancy depends
on several factors, most importantly heterogeneity between studies, and their results
are, therefore, unsuited to compare with ours.[12]
[21]
[22]
[23] We chose to compare time periods before and after hospital-wide guideline modification,
using our updated recommendations as a proxy for true dose escalation. We also show
how the use of higher maximum doses than ordinary prophylaxis increased over time
in [Fig. 2]. This gave us an opportunity to evaluate the effect of anticoagulation with an alternative
approach generating supporting evidence of the importance of anticoagulant in treatment
of thromboinflammation in COVID-19, strengthened by the impact on D-dimer levels and
platelet count.[48] LMWH does not only halter thrombin generation but also seem to have nonanticoagulant
effects, favorable in treatment of COVID-19.[48] Studies have suggested that LMWH may inhibit heparinase that contributes to vascular
leakage and inflammation in COVID-19.[49] Further, in vitro study of plasma from COVID-19 patients found an activation of
platelets and neutrophils and formations of neutrophil extracellular traps (NETs),
whereas NETs correlated with thrombosis and were blocked by LMWH but not by aspirin
or dipyridamole.[50] Aspirin has been found to have some favorable effect in COVID-19, but we are still
awaiting results from randomized control trials.[21]
Strengths and Limitations
This study has several strengths, with a relatively large single-center cohort, including
patients in both the ICU and wards, with complete information and follow-up until
discharge or in-hospital death. In total, the cohort generated 8,534 hospital days,
and the data were collected systematically including by reviewing the medical records
on a daily to weekly basis during follow-up. We have in our statistical analysis been
vigilant with the effect that the gained experienced over time had in the early phase
of the pandemic on both mortality and thrombotic events. A couple of different approaches
were utilized to address this effect in our analysis. First, we cautiously adjusted
to admission date for each patient in our regression analysis. Second, when evaluating
the influence on our hospital-wide recommendations for intensified anticoagulation,
we added interrupted time series regression. However, there are limitations in the
study that must be considered. We identified our patients as those having IL-6 results
in the KUL database, being a test included in the centralized COVID biobank and supposed
to be performed in all cases of severely ill in-hospital COVID-19 patients. Even so
we may have missed some patients. We also observed a noticeable amount of missing
D-dimer levels within the first 3 days at hospital. Instead of imputing values, we
decided to keep the patients with missing data in the analysis, categorizing the patients
with missing D-dimer levels to their own group. We did not find any significant association
with either death or thrombosis comparing the group with missing data with the reference
group ([Table 4]). This supports the hypothesis that the missing results occurred at random and the
group with missing levels had little effect on the outcome of the analysis, except
with losing statistical power. In contrast, we had neglectable number of missing results
for platelet count. We only report the highest-dose LMWH given to each patient, a
consequence of limited information of exact date of dose escalation in many patients.
Of course, this restricted us to go further into our analysis of anticoagulation,
but we have cautiously adjusted for timing of the first given dose in relation to
sampling regarding D-dimer level and platelet count. Finally, we cannot ignore residual
confounding influencing our results regarding the effect on the use of intensified
anticoagulation. Nevertheless, this study is observational, so only associations can
be established and not causalities. Implications for future research are, foremost,
randomized controlled trials regarding anticoagulant treatment doses and treatment
periods.
Conclusion
D-Dimer levels and platelet counts are key markers of thromboinflammation and severity
in COVID-19. From our study we suggest following the rate of change of levels during
the first week in hospital in connection with risk prediction of death and thrombosis.
When the daily mean platelet counts increased and levels of D-dimer levels decreased
during the study period, the thrombotic burden and deaths declined. This coincided
with the introduction of a higher dosage of anticoagulation to patients with severe
COVID-19, underscoring anticoagulation as a treatment cornerstone and the need for
dose adjustment in relation to disease severity.
What is known about this topic?
-
High level of D-dimers is associated with death and thrombosis in COVID-19.
-
Platelets seem to play a pivotal role in pathogenesis of severe COVID-19.
-
Thrombosis prophylaxis with low-molecular-weight heparin (LMWH) appears to improve
survival; randomized clinical trials are lacking providing guidance on optimal dosage.
What does this paper add?
-
The rate of change in platelet count in consort with D-dimer over the first week in
hospital was associated with poor outcome in COVID-19, supporting platelets' pivotal
role.
-
Recommendations on intensified anticoagulation with LMWH were significantly associated
with a drop in daily mean values of D-dimers while the platelet count rose. This also
coincided with a risk reduction in mortality and thrombosis.