Prophylactic heparin and risk of orotracheal intubation or death in patients with mild or moderate COVID-19 pneumonia

Prophylactic low molecular weight heparin (pLMWH) is currently recommended in COVID-19 to reduce the risk of coagulopathy. The aim of this study was to evaluate whether the antinflammatory effects of pLMWH could translate in lower rate of clinical progression in patients with COVID-19 pneumonia. Patients admitted to a COVID-hospital in Rome with SARS-CoV-2 infection and mild/moderate pneumonia were retrospectively evaluated. The primary endpoint was the time from hospital admission to orotracheal intubation/death (OTI/death). A total of 449 patients were included: 39% female, median age 63 (IQR, 50–77) years. The estimated probability of OTI/death for patients receiving pLMWH was: 9.5% (95% CI 3.2–26.4) by day 20 in those not receiving pLMWH vs. 10.4% (6.7–15.9) in those exposed to pLMWH; p-value = 0.144. This risk associated with the use of pLMWH appeared to vary by PaO2/FiO2 ratio: aHR 1.40 (95% CI 0.51–3.79) for patients with an admission PaO2/FiO2 ≤ 300 mmHg and 0.27 (0.03–2.18) for those with PaO2/FiO2 > 300 mmHg; p-value at interaction test 0.16. pLMWH does not seem to reduce the risk of OTI/death mild/moderate COVID-19 pneumonia, especially when respiratory function had already significantly deteriorated. Data from clinical trials comparing the effect of prophylactic vs. therapeutic dosage of LMWH at various stages of COVID-19 disease are needed.

www.nature.com/scientificreports/ a pandemic 2 . To date, over 50.7 million COVID-19 cases and 1.2 million deaths have been reported to WHO. Currently, there are more than 3.6 million new cases and over 54 000 new deaths reported 3 . COVID-19 might be commonly complicated with some hemostatic changes including mild thrombocytopenia 4 and increased D-dimer levels 5,6 , indicating some forms of coagulopathy 7-10 that may predispose to thrombotic events, associated with a higher risk of requiring mechanical ventilation, intensive care unit (ICU) admission, or death 6,9,10 . These hemostatic changes are a specific effect of SARS-CoV-2 and a consequence of a cytokine storm that alters the onset of the systemic inflammatory response syndrome as observed in other viral disease 8 . Generally, a correlation between inflammation and coagulation exists: several inflammatory cytokines lead to an impairment of the coagulation pattern, with a consequent imbalance between the procoagulant and anticoagulant states 11 . In the severe acute respiratory syndrome induced by coronavirus, vascular endothelial damage in small and medium sized pulmonary vessels, disseminated intravascular coagulation (DIC), deep venous thrombosis, and pulmonary thromboembolism have been described 12,13 . Hospitalized patients with acute medical illness, including infections such as pneumonia, are at increased risk of thrombotic events 10 and it is well known that prophylactic anticoagulation reduces that risk 14,15 . Interestingly, heparin and its related derivatives have shown antiviral and anti-inflammatory activities and seem to be beneficial for patients with other diseases 16,17 . As inflammation, atherogenesis, thrombogenesis, and cell proliferation are joint with each other, the pleiotropic effects of heparin and derivatives may have a therapeutic effect and might be relevant in this setting 6,16,18 . Nowadays, physicians treating patients with COVID-19 are facing challenges and one of these is related with the therapeutic utility of heparin 17 . The use of prophylactic-doses of low molecular weight heparin (pLMWH) is now recommended by the WHO 19 and others guidelines [20][21][22][23][24] for all hospitalized COVID-19 patients, unless of clinical contraindications. However, there are conflicting opinions regarding the optimal dose of prophylactic anticoagulation to prevent thrombotic events in COVID-19 patients and to induce a potential anti-inflammatory activity because of the lack of solid evidences.
The aim of this analysis was to assess the effectiveness of prophylactic dose of LMWH vs. no heparin in reducing the risk of orotracheal intubation and death in a real-life setting of patients hospitalized for COVID-19.

Results
Patients' characteristics. A total of 449 patients with COVID-19 mild/moderate pneumonia was included in this analysis. Over 48 h from the date of admission, 210 (46.8%) patients started pLMWH and 239 (53.2%) did not. Overall, 39% were female, with a median (Inter-Quartile Range, IQR) age of 63 (50-77) years and a median of 8 days from onset of symptoms to hospital admission (IQR 4-12).
The main characteristics of the study population at admission, overall and according to pLMWH treatment at admission, are shown in Table 1.
53% of patients did not receive pLMWH and they were more frequently hospitalized in the first pandemic period (196/239 in March, 33/239 in April and 10 between June and July 2020; p < 0.001).
Overall, only 16 (3.5%) pulmonary thrombosis occurred, of whom 12 were in participants who started pLMWH close to admission and 4 in those who did not. We observed 5 major bleeding events which occurred 4 in people who were treated with heparin and 1 in untreated (p at Fisher exact test 0.076), more in detail: two intramuscular hematomas, 1 cerebral haemorrhage, 1 cerebellar bleeding and a vascular bleeding from the ascending aorta.
As to other treatments, patients on pLMWH received immunomodulant therapy, steroids and remdesivir over follow-up more frequently than patients not treated with pLMWH.
At multivariable analysis, a first model was adjusted only for time-fixed confounders (model 1) and a second one which included also time-varying confounders concerning concomitant treatment (model 2). www.nature.com/scientificreports/ Overall, crude and adjusted marginal hazard ratio for OTI/death showed a reduced risk for patient who received pLMWH but data were highly compatible with the null hypothesis of no difference (model 1: aHR = 0.89, 95% CI 0.34-2.29, p = 0.806; model 2: aHR = 0.66, 95% CI 0.28-1.57, p = 0.352).
The ITT analysis showed similar risk for treated and not treated in the group with PaO 2 /FiO 2 > 300 and higher risk for treated if the baseline PaO 2 /FiO 2 was ≤ 300 mmHg (supplementary table 1).
Similar results were obtained after the exclusion of 16 patients with pulmonary thromboembolic events from the study population patients treated with pLMWH showed higher risk of OTI/death versus those not treated if their PaO 2 /FiO 2 at admission was ≤ 300 mmHg (HR 1.26; 95% CI 0.45-3.53), and they showed a lower risk if they were admitted at hospital with PaO 2 /FiO 2 > 300 mmHg (HR 0.29; 95% CI 0.04-2.17) (Supplementary table 2). Table 1. General characteristics of study population. pLMWH prophylactic dose of low molecular weight heparin, IQR inter quartile range, COPD chronic obstructive pulmonary diseases, LPV/r lopinavir/ritonavir, HCQ hydroxychloroquine. a Defined by the presence of at least two of the following criteria: (a) blood lymphocytes < 1000/mmc; (b) ferritin > 500 ng/mL; (c) LDH > 300 U/L; (d) D-dimers > 1000 ng/mL; (e) C-reactive protein > 3 mg/dL. b Available for 130 patients.

Discussion
This cohort of patients hospitalized for COVID-19 pneumonia at the National Institute for Infectious Diseases L. Spallanzani in Rome, Italy, was mainly enrolled during the first pandemic time-window of the hospitalizations for COVID-19 in Rome. The fact that the evidence was insufficient to determine the risks and benefits of   www.nature.com/scientificreports/ prophylactic anticoagulants for people hospitalized with COVID-19 because of the lack of randomized comparisons on pLMWH versus no treatment and of the availability of few observational studies with no converging results 25 were the main triggers to perform this analysis. Patients receiving LMWH prophylaxis (39%) appeared to be older than those who did not receive LMWH prophylaxis, with at least 1 comorbidity; specifically, cardiovascular diseases, hypertension, kidney diseases, COPD/Asthma were the more prevalent.
In our study population, there was a non-negligible proportion of patients, mainly those hospitalized in March/April 2020 (53%) in whom prophylactic LMWH was not prescribed. This finding reflects the fact that in the early stages of the epidemic the risk of thromboembolic events in people with COVID-19 disease had not been clearly recognized. As soon as recommendations were made on prophylactic anticoagulation in COVID-19, since May 2020 [19][20][21][22][23][24][25][26] , all hospitalized patients with pneumonia at our COVID-hospital were administered prophylactic dose of LMWH in order to prevent SARS-CoV-2-related thrombotic events. Therefore, this type of analysis will be no longer possible for people enrolled during the second wave of the pandemic.
This analysis reveals that a significant higher proportion of patients receiving LMWH prophylaxis had an impaired respiratory function and a hyperinflammation pattern, which have a known potential prognostic value 5 . These findings highlight that clinicians might have been more prone to use anticoagulant prophylaxis in Table 2. Hazard Ratio of oro-tracheal intubation/death (OTI/death) in all population and according to PaO 2 / FiO 2 at admission. pLMWH prophylactic dose of low molecular weight heparin. *Adjusted for time-fixed factors: age, gender, time from symptoms onset, comorbidities (cardiovascular diseases, hypertension, COPD/ Asthma, diabetes), PaO 2 /FiO 2 at admission. **Adjusted for time-fixed and time varying factors: age, gender, time from symptoms onset, comorbidities (cardiovascular diseases, hypertension, COPD/Asthma, diabetes), PaO 2 /FiO 2 at admission, time-varying use of immune-therapy, antiviral and steroids and censoring using IPW. a Initiation of invasive mechanical ventilation or death.  www.nature.com/scientificreports/ patients admitted in severe clinical conditions and that respiratory function was the main driver in prescribing LMWH prophylaxis. Our findings are only partially consistent with those of a meta-analysis showing that adjunctive LMWH use appeared to reduce 7-day and 28-day mortality [RR 0.52 (0.31-0.87 and 0.63 (0.41-0.96), respectively)] as well as improved the PaO 2 /FiO 2 ratio [by weighted mean difference 74.8 mmHg (52. 18-96.78)] in individuals with acute lung injury/acute respiratory distress syndrome (ALI/ARDS) not caused by SARS-Cov-2 27 . The results of this meta-analysis were similar after excluding two studies including more severe patients. Furthermore, our results are also in conflict with those of another observational study in which a better in-hospital survival was shown even in a population with saturation of oxygen < 90% and fever 28 .
More recently, other studies have emerged regarding the risk of mortality in patients treated with heparin such as the experience of the Multicenter Italian CORIST observational study which showed a 40% lower risk of death in patients receiving LMWH or unfractionated heparin [UFH] vs. no heparin (hazard ratio = 0.60; 95% confidence interval: 0.49-0.74; E-value = 2.04) association particularly evident in patients with a higher severity of disease or strong coagulation activation 29 .
The results of a randomized trials, also only recently published 30,31 , add evidence against dose-escalated thromboprophylaxis in critically ill patients with COVID-19. A large observational cohort study of 2,809 critically ill patients with COVID-19 from 67 centers in the US found no benefit of therapeutic dose anticoagulation initiated within 2 days of intensive care unit (ICU) admission compared with standard-dose thromboprophylaxis 32 . Similarly, another Italian study found that the use of a prophylactic dosage of enoxaparin appears to be associated with similar in-hospital overall mortality compared to higher doses in patients hospitalized for COVID-19 33 .
Unfortunately our data do not provide elements to contribute to this debate as we only compared prophylactic dosage with no heparin at all.
Interestingly, in our study the risk of a clinical worsening in patients receiving prophylactic LMWH seemed to vary by the PaO 2 /FiO 2 ratio at admission. In particular, there appeared to be a qualitative interaction with some evidence that treatment with pLMWH was beneficial in reducing the risk of OTI/death in participants who started the drug with a PaO 2 /FiO 2 > 300 mmHg but even potentially harmful in those who started with PaO 2 /FiO 2 ≤ 300 mmHg.
The results for the outcome death alone, were similar to those of the main analysis although with a reduced power to detect the potential interaction with levels of PaO 2 /FiO 2 and again only partially consistent with those of other observational studies [33][34][35] .
Regarding the risk of bleeding events, although it was expected, the frequency was very low and there was no difference between pLMWH and no pLMWH.
Our study has some limitations. Firstly, the analysis is retrospective and conducted in the observational setting, therefore residual confounding bias is likely to be an issue. Secondly, this observation includes mainly patients hospitalized in the early stages of the epidemic only in one COVID-hospital in central Italy and may have disproportionately included more patients with better outcomes. Thirdly, only the prophylactic dose of LMWH was evaluated so our data do not contribute to the current debate regarding the identification of the optimal dosage. Last, although an interesting signal was detected regarding a possible role of PaO 2 /FiO 2 as an effect modifier, the analysis was not powered to detect this interaction.
In conclusion, our results carry little evidence that prophylactic doses of pLMWH can lead to a reduction in risk of OTI/death in patients with mild/moderate COVID-19 pneumonia. Therefore, overall it seems that prophylactic doses are not sufficient to contrast the hypercoagulable state established in many severe COVID-19 patient, as an obvious consequence of the hyperinflammation and the cytokine storm syndrome and that higher dosage might be needed in people showing generally hyper-inflamed status, impaired respiratory function or suspected high risk of a thrombotic event.
Nevertheless, we have also shown a signal for some clinical benefit of using pLMWH in participants who initiated the drug with a PaO 2 /FiO 2 > 300 and these data are important to guide future research and the design of randomized studies evaluating the impact of prophylactic heparin vs. higher doses in COVID-19 disease. Our data are compatible with the null hypothesis of no interaction although the effect sizes in the strata are so different that lack of power is a likely explanation for the large p-value. Indeed, the role of prophylactic vs. therapeutic doses LMWH for reducing the risk of thrombosis in hospitalized patients with COVID-19 is currently under evaluation in randomized studies. www.nature.com/scientificreports/ CT scans were performed on a multi-detector CT scanner (Bright Speed, GE Medical Systems, Milwaukee, WI). The non-contrast scans were reconstructed with sub-millimetric thicknesses and spacing, high-contrastresolution algorithm and evaluated to assess the residual pulmonary volume with automatic segmentation of lung areas on dedicated workstation (expressed in Liters).
Patients were included if they were followed-up for at least 2 days after admission. Patients who started a standard prophylactic dose of heparin within 48 h from admission, non-randomly, according to local protocol (intermediate dosage of 100 UI/Kg/day) 25 were included in the intervention group and compared to the remaining patients who did not receive the drugs.
Patients who started a prophylactic dose of heparin more than 48 h after admission or started a therapeutical dose were excluded from the analysis dataset.
Definitions. Hyperinflammation condition was defined by the presence of at least two of the following criteria at any time from admission: (a) blood lymphocytes < 1000/mmc; (b) ferritin > 500 ng/mL; (c) LDH > 300 U/L; (d) D-dimers > 1000 ng/mL; (e) C-reactive protein > 3 mg/dL 5 . The Padua score is a tool used to stratify patients and to guide management of the risk of pulmonary embolism 36 . Endpoints. The primary endpoint of this analysis was the time to the first event between orotracheal intubation and death (OTI/death). Time to death was analyzed as secondary endpoint.
Statistical analysis. Patients' characteristics were described at baseline, non-parametric Mann-Whitney test was used to compare continuous variables and Chi-Square test to compare categorical variables between treatment groups (pLMWH vs. not). Shapiro-Wilk test was used to check for the normality of distribution and the Spearman correlation coefficient was calculated and tested for the correlation analysis. Baseline for the survival analysis was the admission for patients not treated and heparin initiation for treated group. Standard survival analysis by means of weighted Kaplan-Meier (KM) curves were performed to estimate the cumulative proportion of people experiencing the primary endpoint from baseline.
The main analysis was performed using a Cox marginal structural model. The causal HR and corresponding 95% CI of the primary outcome for heparin treated vs not treated participants were estimated by Cox regression model weighted by (i) inverse probability of treatment weights and (ii) censoring weights.
Participants' follow-up accrued from baseline until the occurrence of the outcome or last in-hospital observation. The follow-up was censored if participants changed the heparin dose from prophylactic to therapeutic. Confounders included for the construction of the weights were: gender, age, duration of symptoms, type of comorbidities, PaO 2 /FiO 2 measured at admission as time-fixed factors, and the initiation of any antiviral therapy, any immunomodulating agents, any steroids as time-varying factors.
To test the hypothesis of a beneficial effect pLMWH solely via reduction of inflammation, a sensitivity analysis was performed after exclusion of participants who experienced pulmonary thromboembolic events.
The analysis was stratified according to the severity of disease at admission defined as a) PaO 2 /FiO 2 ratio ≤ or > 300 mmHg. The interaction between PaO 2 /FiO 2 ratio level and heparin use was formally tested.
The analysis was conducted following both OT and ITT principle, the latter ignored any change in dosage of heparin during observation.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.