key: cord-0870280-tcaglk0u
authors: Nadkarni, Girish N.; Lala, Anuradha; Bagiella, Emilia; Chang, Helena L.; Moreno, Pedro; Pujadas, Elisabet; Arvind, Varun; Bose, Sonali; Charney, Alexander W.; Chen, Martin D.; Cordon-Cardo, Carlos; Dunn, Andrew S.; Farkouh, Michael E.; Glicksberg, Benjamin; Kia, Arash; Kohli-Seth, Roopa; Levin, Matthew A.; Timsina, Prem; Zhao, Shan; Fayad, Zahi A.; Fuster, Valentin
title: Anticoagulation, Mortality, Bleeding and Pathology Among Patients Hospitalized with COVID-19: A Single Health System Study
date: 2020-08-26
journal: J Am Coll Cardiol
DOI: 10.1016/j.jacc.2020.08.041
sha: 51b81f576a761e69ce59379db7b96bccad07c951
doc_id: 870280
cord_uid: tcaglk0u

Background Thromboembolic disease is common in coronavirus disease-19 (COVID-19). There is limited evidence on association of in-hospital anticoagulation (AC) with outcomes and postmortem findings. Objective To examine association of AC with in-hospital outcomes and describe thromboembolic findings on autopsies. Methods A retrospective analysis examining association of AC with mortality, intubation and major bleeding. We also conducted sub-analyses on association of therapeutic vs prophylactic AC initiated ≤48 hours from admission. We describe thromboembolic disease contextualized by pre-mortem AC among consecutive autopsies. Results Among 4,389 patients, median age was 65 years with 44% female. Compared to no AC (n=1530, 34.9%), therapeutic (n=900, 20.5%) and prophylactic AC (n=1959, 44.6%) were associated with lower in-hospital mortality (adjusted hazard ratio [aHR]=0.53; 95%CI: 0.45-0.62, and aHR=0.50; 95%CI: 0.45-0.57, respectively), and intubation (aHR 0.69; 95%CI: 0.51-0.94, and aHR 0.72; 95% CI: 0.58-0.89, respectively). When initiated ≤48 hours from admission, there was no statistically significant difference between therapeutic (n=766) vs. prophylactic AC (n=1860) (aHR 0.86, 95%CI: 0.73-1.02; p=0.08). Overall, 89 patients (2%) had major bleeding adjudicated by clinician review, with 27/900 (3.0%) on therapeutic, 33/1959 (1.7%) on prophylactic, and 29/1,530 (1.9%) on no AC. Of 26 autopsies, 11 (42%) had thromboembolic disease not clinically suspected and 3/11 (27%) were on therapeutic AC. Conclusions AC was associated with lower mortality and intubation among hospitalized COVID-19 patients. Compared to prophylactic AC, therapeutic AC was associated with lower mortality, though not statistically significant. Autopsies revealed frequent thromboembolic disease. These data may inform trials to determine optimal AC regimens.

Coronavirus disease-19 has led to >22 million affected, (1) and > 784,000 deaths worldwide. Among hospitalized patients, new thromboembolism has emerged as an important disease manifestation. (2) (3) (4) (5) Autopsy studies have corroborated these observations by demonstrating a high incidence of macro and microthrombi. (6) (7) (8) Accordingly, it has been hypothesized that inflammation associated with SARS-CoV2 infection leads to a "COVID-19 related coagulopathy", (9) resulting in increased thrombosis. (6) Observational analyses have suggested potential benefit for in-hospital use anticoagulation (AC) in COVID-19 treatment. (10, 11) Yet, practice patterns vary significantly due to lack of rigorous evidence for optimal regimens. Specifically, anticoagulant choice, dosing, and treatment duration are not well understood. In a preliminary analysis of 2700 patients admitted to the Mount Sinai Health System (MSHS) in New York, we found an association between in-hospital therapeutic AC and lower mortality compared to patients on no/prophylactic AC. (10) The present analysis expands upon those results in a larger cohort to explore the impact of therapeutic and prophylactic AC, as well as choice of agent, on survival, intubation, and major bleeding compared to no AC. We also review the first consecutive autopsies performed at our institution and describe their pre-mortem management as related to AC.

Data were retrieved from the electronic health record (EHR). Variables collected included demographics, laboratory measurements, vital signs, disease diagnoses, comorbidities, procedures, and outcomes (death, intubation, and hospital discharge). The Mount Sinai Institutional Review Board approved this study.

We included all patients >18 years old admitted with laboratory confirmed SARS-CoV-2 infection between March 1 st -April 30 th , 2020 to five New York City hospitals. Patients who left the hospital within 24 hours of admission as well as those patients treated with both therapeutic and prophylactic regimens of AC during their hospitalization were excluded. If treated for <48 hours total with a therapeutic or prophylactic dose, they were conservatively categorized as "not treated with AC" unless AC was stopped due to major bleeding. (Supplemental Figure 1) .

Details on how patients were categorized into therapeutic/ prophylactic AC are in the Supplemental Appendix.

The primary exposure of interest was therapeutic or prophylactic AC compared to no AC.

We also conducted a sub-analysis of patients initiated therapeutic or prophylactic anticoagulants within 48 hours of admission.

The primary endpoint was in-hospital mortality. Secondary endpoints were intubation and major bleeding. Consistency checks were performed to properly align these data tables and minimize missing data. If the amount of missing data was less than 1% patient was considered as not having the condition (e.g. for comorbidities). Missing values were mostly present for the vitals and the labs for which we used a "missing" category in the propensity score models to account for the missing data. (Supplemental Appendix) Major bleeding was defined using International Classification of Diseases-10 (ICD-10) codes (Supplemental Table 1 (perioperative or symptom improvement). We also ascertained the bleeding site.

Autopsies were performed at the Mount Sinai Hospital after obtaining appropriate consent and verifying SARS-CoV-2 infection status by nasopharyngeal swab unless already appropriately documented. Examinations were carried out in a negative pressure room with enhanced airborne precautions. Histological processing of tissue blocks was performed in standard fashion after extended formalin-fixation. Slides were reviewed by a team of pathology subspecialists.

General characteristics of the sample were summarized using appropriate descriptive statistics for continuous and categorical variables. Some continuous variables (e.g. body mass index [BMI], age, D-dimer, respiratory rate and oxygen saturation) were categorized using clinically meaningful cut-points to improve interpretability. Patients were divided into three groups according to whether they were treated with a therapeutic or prophylactic regimen, or no anticoagulant. Patients receiving both therapeutic and prophylactic anticoagulants were excluded.

J o u r n a l P r e -p r o o f Inverse probability treatment weighted (IPTW) models, were used to correct for the potential bias brought about by AC indication. A multinomial logistic model was fit with therapeutic, prophylactic or no use of AC during the hospitalization as the dependent variable, and age, sex, race and ethnicity, BMI, history of hypertension, atrial fibrillation, heart failure, chronic kidney disease or renal failure, use of anticoagulants or antiplatelet agents prior to hospitalization, month of admission, intubation during hospitalization, time of implementation of institutional guidelines for AC at Mount Sinai, respiratory rate, oxygen saturation and D-dimer at admission as the predictors. These predictors were chosen based on clinical judgment and model fit. We derived stabilized inverse IPTW by multiplying the inverse of the predicted probability of treatment from the propensity score model by the observed probability of treatment. The IPTW approach was used in all analyses. A robust variance was estimated in all models to account for the clustering effect resulting from IPTW. Standardized differences were calculated to determine the level of adjustment induced by the IPTW. To account for residual confounding, all models were adjusted for variables with absolute standardized differences greater than 0.2 (Supplemental Figure 1) . Regarding missing data, if the amount of missingness was less than 1%, a patient was considered as not having the condition (e.g., for comorbidities).

Missing values were mostly in vitals and labs (e.g., D-dimer) for which we used a "missing" category in the propensity score models to account for the missing data.

The primary analysis used IPTW Fine and Gray's sub-distribution hazard models to determine AC association with in-hospital mortality. (12) Survival in days was calculated as time from hospital admission to in-hospital death, discharge, or the date of dataset lock (May 7th, 2020). Patients who were still hospitalized at the time of the data lock were censored. Discharge alive was considered a competing risk. To minimize immortal time bias, therapeutic or J o u r n a l P r e -p r o o f prophylactic AC use were entered in the model as time dependent variables and similarly for intubation status. The multivariable model also accounted for admission respiratory rate and oxygen saturation.

For the time to intubation analysis, the time between hospital admission and intubation was considered in IPTW competing risk models using the method of Fine and Gray. Death and hospital discharge were considered competing events and patients who were in hospital but not intubated at the time of data lock were censored. AC use was entered as time dependent variables with the same covariate adjustment made previously. The hazard ratios (HR) and their respective 95% confidence intervals (CI) are reported for all time-to-event models. Frequency tables were used to describe the association between AC use and bleeding events. A similar approach was used for the subgroup of patients treated with therapeutic or prophylactic anticoagulants within 48 hours of admission.

Landmark analyses were considered at 3 different timepoints: days 2, 3 and 4 after hospital admission (Supplemental Appendix). All analyses were conducted using SAS 9.4 (SAS Institute, Inc, Cary, North Carolina).

A total of 4,389 patients met inclusion criteria for analysis (Supplemental Figure 2) .

The median age was 65 (IQR, 53 to 77 years), 44% were female, 26% self-identified as African American and 27% as Hispanic/Latino. Table 1 shows baseline characteristics and laboratory values stratified by therapeutic AC (n=900), prophylactic AC (n=1,959), and no AC (n=1,530).

Pre-hospital medications of angiotensin converting enzyme (ACE) inhibitors or angiotensin receptor blockers, prior AC and antiplatelet therapy by group are also shown in Table 1 .

Approximately one-tenth of the total cohort were on AC or antiplatelet medications prior to admission (1.8% and 8.5% respectively). On hospital presentation, patients in the therapeutic AC group had higher blood pressures, faster heart and respiratory rates, and lower oxygen saturation (Table 1) . D-dimer concentrations were highest in the patients who received therapeutic AC (2.3; 1.2-5.8 μg/ml). Elevated inflammatory markers including ferritin, lactate dehydrogenase, and c-reactive protein increased progressively from the no AC to prophylactic AC and then therapeutic AC patient groups.

Overall 1,073 (24.4%) patients died during the study period, 2892 (65.9%) were discharged alive and 424 (9.7%) were still hospitalized by dataset freeze date. Among the no AC group, 931 (60.8%) patients were discharged alive; 392 (25.6%) expired in the hospital; and 207 (13.5%) were still hospitalized. In the prophylactic AC group, 1472 (75.1%) patients were discharged alive; 424 (21.6%) expired in the hospital; and 63 (3.2%) were still hospitalized.

Finally, in the therapeutic AC group, 89 (54.3%) patients were discharged alive; 257 (28.6%) expired in the hospital; and 154 (17.1%) were still hospitalized. Therapeutic AC was associated with a 47% reduction in the hazard of in-hospital mortality (aHR=0.53; 95% CI: 0.45-0.62; p<0.001; Figure 1A ) compared to no AC. Similarly, prophylactic AC was associated with a lower hazard of mortality (aHR=0.50; 95% CI: 0.45-0.57; p<0.001) compared to no AC. Overall, 467 (10.6%) patients required intubation and mechanical ventilation during hospitalization.

Therapeutic AC was associated with a 31% reduction in the hazard of intubation (aHR 0.69; 95% CI: 0.51-0.94; p=0.02; Figure 1B ) compared to no AC. Prophylactic AC was also associated with similarly reduced incidence of intubation (adjusted HR 0.72; 95% CI: 0.58-0.89, p=0.003) J o u r n a l P r e -p r o o f compared to no AC. Landmark analyses showed similar associations (Supplemental Tables 2   and 3 ).

We conducted a sub analysis for patients initiated on therapeutic (n=766) or prophylactic doses (n=1,860) of AC <48 hours of admission. Baseline characteristics are presented in Supplemental Table 4 . Patients who received therapeutic AC were older, had more comorbid conditions and were more likely to be on an anticoagulant prior to admission compared to those receiving prophylactic AC. Patients on therapeutic AC also presented with more altered vital signs, and inflammatory markers, in particular D-dimer (2.4 vs. 1.4 μg/ml) compared to those receiving prophylactic AC. In adjusted analyses, therapeutic AC was associated with lower inhospital mortality (aHR 0.86; 95% CI: 0.73-1.02; p=0.08; Figure 2A) although not statistically significant. There was no difference in incidence of intubation (aHR 0.94; 95% CI: 0.74-1.21); p=0.63; Figure 2B ). Table 5 ) Among patients on a single therapeutic agent, bleeding rates were higher in those on low molecular weight heparin (LMWH) compared to novel anticoagulants (NOACs) (2.6% vs 1.3% respectively) and among those on a single prophylactic agent, bleeding rates were higher in those on unfractionated heparin (UFH) J o u r n a l P r e -p r o o f compared to LMWH (1.7% vs. 0.7% respectively). The site of bleeding was determined in 67/89; 75%, with gastrointestinal being most common (50.7%), followed by mucocutaneous (19.4%), bronchopulmonary (14.9%) and then intracranial (6%).

A sizable proportion of patients were on more than one AC agent over the course of their hospitalization preventing direct comparisons between anticoagulants. In a descriptive analysis, we present differences in cumulative incidence of mortality and intubation among individuals 

Autopsies were performed on COVID-19 positive patients at MSHS starting on 3/20/2020, with 72 completed by 5/7/2020. 8 Of these, the first 26 sequential cases were evaluated microscopically by a team of subspecialty pathologists across organ systems. These cases are presented with a focus on thromboembolism and contextualized by pre-mortem AC regimens ( Table 2) In total, 11/26 (42%) had evidence of thromboembolic disease, including 4 pulmonary emboli (15%, Figure 3A ,B), 2 cerebral infarctions (8%, Figure 3C ,D) and 5 patients with microthrombi in multiple organs including the heart (n=4, Figure 3E) , liver (n=1, Figure 3F ), kidneys (n=2, not shown) and lymph nodes (n=2, not shown). The lungs were examined and revealed an extensive burden of fibrin thrombi visible on hematoxylin and eosin stain (15/26), however, this was not counted towards the thrombotic burden as it is an expected and frequently encountered finding in diffuse alveolar damage. Two of the four patients with pulmonary emboli were on prophylactic AC throughout, one was not on AC and one was given AC using UFH to treat disseminated intravascular coagulation but at subtherapeutic levels. More generally, 8/11 (73%) patients with thromboemboli were not on therapeutic AC. There was no pre-mortem suspicion of thromboemboli in 25/26 patients. There was only one major bleeding complication, which was a retroperitoneal bleed on presentation in a patient taking warfarin for atrial fibrillation prior to admission.

Thromboembolic disease has emerged as an important complication among hospitalized patients with COVID-19. In the present report of nearly 4,400 patients, we demonstrate the following (Central Illustration): first, AC is associated with lower hazards of in-hospital mortality and intubation compared to no AC after controlling for relevant clinical factors.

Second, after restricting analysis to those in whom AC was initiated within 48 hours of admission, no statistically significant difference in in-hospital mortality or intubation for J o u r n a l P r e -p r o o f therapeutic vs. prophylactic AC was observed. Third, overall rates of major bleeding were low.

Finally, these observations were corroborated by autopsy findings, wherein 11/26 of patients had thromboembolic disease not otherwise suspected pre-mortem. The majority of these patients were not treated with therapeutic AC.

Mechanisms by which thrombotic disease may occur in the setting of COVID-19 infection include inflammation, hypoxia, and potentially pharmacotherapeutic interactions. (2, 4, 13, 14) As such, the potential benefit of AC in the treatment of COVID-19 is based on the prevention and treatment of micro and macrovascular thrombosis. In addition, AC agents may exert antiviral and anti-inflammatory properties affording further benefit. (15, 16) In our cohort of patients hospitalized with COVID-19, a strong association of AC with approximately 50% reduced hazard of in-hospital mortality was observed ( Figure 1A) . Both therapeutic and prophylactic doses of AC were associated with better in-hospital survival compared to no AC. As mortality rates for patients with COVID-19 who undergo intubation for respiratory failure range from 30-80%, (17) (18) (19) we analyzed the association between AC and intubation. Both therapeutic and prophylactic AC were associated with an approximately 30% reduced hazard of intubation compared to patients on no AC ( Figure 1B) . Landmark analyses were performed to minimize immortal time bias and revealed similar associations (Supplemental Tables 2, 3) .

Due to variation in timing of initiation and administration of AC across patients, a subanalysis to patients who received either therapeutic or prophylactic AC within 48 hours of admission, showed therapeutic AC was associated with a 14% reduction in hazard of mortality compared to J o u r n a l P r e -p r o o f prophylactic AC that did not reach statistical significance (p=0.08). There was no difference in intubation risk between the two doses (Figures 2A, B) .

In entirely descriptive analyses examining individual agents, potential benefit with prophylactic LMWH compared to UFH may exist for mortality but differences in intubation appear minimal. Therapeutic NOACs visually may be associated with lower mortality and intubation risk compared to LMWH (Supplemental Figure 4) . No conclusions can be drawn from these purely descriptive comparisons however, and randomized trials comparing specific agents are needed to inform whether comparative benefit exists.

Bleeding rates were low overall, but as expected, slightly higher in the therapeutic AC group compared to the prophylactic and no AC groups ( Table 2 ). In patients on a single therapeutic agent, the bleeding rates were higher in patients on LMWH vs. NOACs. Further studies and trials are required however to better understand this observation. As always, the benefit-risk tradeoff, here between AC and bleeding, needs to be evaluated on an individual basis and discussed as part of shared-decision making.

We show a high prevalence of thrombotic complications mostly occurring in patients receiving prophylactic/ no AC, consistent with a recent autopsy study demonstrating thrombotic burden in 58%. (6, 20) Though lung microthrombi were not counted towards overall burden but rather as a feature of diffuse alveolar damage, it is worth noting that this finding emphasizes the endothelial dysfunction at play. Finally, in all except for one case of stroke, there was no clinical suspicion of thromboembolic disease prior to autopsy, suggesting that clinical estimates of thromboembolic disease may be under-estimating the actual burden.

Our study has several limitations. As an observational study, there may have been confounders leading to differences in the outcomes for the treatment groups. Though we minimized their potential impact through IPTW modeling, unmeasured confounders and residual bias may have been present. Despite a two-physician manual review of different AC regimens for the purposes of categorizing patients, there may have been discrepancies between regimens of NOACs and LMWH wherein doses may not have accurately represented therapeutic and prophylactic AC. Patients who were on both therapeutic and prophylactic doses of AC were excluded due to an inability to definitively categorize them. Patients with hospital stay <24 hours were also excluded. Nonetheless, we adopted a conservative approach wherein individuals receiving <48 hours of AC were considered in "no AC" group. To minimize immortal time bias, we analyzed AC as a time dependent variable and conducted landmark sensitivity analyses.

However, we cannot rule out residual bias even after using IPTW. We included UFH infusion in the therapeutic group, but patients may not be in the therapeutic aPTT range. Since manual validation of each outcome was not feasible in the whole sample size, there exists the possibility of misclassification of outcomes. We did not conduct analysis on novel antiviral treatments (Remdesivir, IL-1 antagonists) since these were still under investigation and administered in the context of clinical trials at our institution. The generalizability of the autopsy data may be limited due to small sample size and fact that these were not consecutive deaths. Finally, we may have encountered higher proportions of patients on AC due to the fact that Mount Sinai initiated a system-wide protocol wherein at least prophylactic AC was strongly encouraged with guidance provided for consideration of therapeutic AC based on various factors (Supplemental Thromboembolic disease is a complication of COVID-19. Prophylactic and therapeutic anticoagulation are associated with better outcomes in hospitalized patients with COVID-19. randomized controlled trials evaluating different AC regimens in COVID-19 are needed. 

COVID-19 Map -Johns Hopkins Coronavirus Resource Center

COVID-19 and Thrombotic or Thromboembolic Disease: Implications for Prevention, Antithrombotic Therapy, and Follow-Up: JACC State-of-the-Art Review

Thromboembolic risk and anticoagulant therapy in COVID-19 patients: emerging evidence and call for action

Coagulation abnormalities and thrombosis in patients with COVID-19

COVID-19 and its implications for thrombosis and anticoagulation

Autopsy Findings and Venous Thromboembolism in Patients With COVID-19

American College of Physicians

Pulmonary and cardiac pathology in African American patients with COVID-19: an autopsy series from New Orleans

Pathophysiology of SARS-CoV-2: targeting of endothelial cells renders a complex disease with thrombotic microangiopathy and aberrant immune response. The Mount Sinai COVID-19 autopsy experience. medRxiv

COVID-19 and its implications for thrombosis and anticoagulation

Association of Treatment Dose Anticoagulation with In-Hospital Survival Among Hospitalized Patients with COVID-19

Anticoagulant treatment is associated with decreased mortality in severe coronavirus disease 2019 patients with coagulopathy

A Proportional Hazards Model for the Subdistribution of a Competing Risk

ISTH interim guidance on recognition and management of coagulopathy in COVID-19

Hypoxia downregulates protein S expression

Murine coronavirus with an extended host range uses heparan sulfate as an entry receptor

Activation of the SARS coronavirus spike protein via sequential proteolytic cleavage at two distinct sites

Covid-19 in Critically Ill Patients in the Seattle Region -Case Series

Characteristics of Hospitalized Adults With COVID-19 in an Integrated Health Care System in California

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the

Pulmonary Arterial Thrombosis in COVID-19 With Fatal Outcome: Results From a Prospective, Single-Center

cells/mm

Abbreviations: IQR, interquartile range; ACE, Angiotensin Converting Enzyme, ARB, Angiotensin Receptor Blocker Values at baseline are within 48 hours of admission *Chi-squared test used for categorical variables. Kruskal-Wallis test used for continuous variables