With the increased understanding of the integral roles of both coagulation and inflammation (and the interaction between the two) in the pathophysiology of sepsis, attention has focused on the potential for anticoagulants as potential treatments for patients with sepsis, especially anticoagulants that also contain anti-inflammatory properties. Recombinant human activated protein C represented one such agent, but after initial promising results, subsequent data appeared less promising and it was removed from the market. Unfractionated heparin also has both anticoagulant and anti-inflammatory properties. Over the last decade, numerous articles have been written, and a few small trials have been conducted, around the potential for heparin to improve outcomes in sepsis. Unfortunately, none of the data have been overly convincing, and heparin (or other anticoagulants) has not been widely utilized specifically for the treatment of sepsis.
Until about five years ago, I never even thought of using therapeutic heparin (not venous thromboembolism prophylactic doses, but actual therapeutic doses) in the patients I was treating for sepsis. In fact, I didn’t even use heparin when my patients with sepsis also had atrial fibrillation, rationalizing that the potential harm from heparin and the risk of bleeding likely outweighed the benefit of systemic anticoagulation for a few days to a week. However, about five years ago, Walkey and colleagues published a fascinating article on the incidence of stroke and mortality in patients with sepsis and new-onset atrial fibrillation (Walkey AJ, et al. JAMA. 2011;306(20):2248). In this retrospective cohort study of almost 50,000 patients with sepsis induced organ dysfunction, they found that patients with sepsis and new-onset atrial fibrillation, defined as developing during the hospitalization in patients not previously known to have a history of atrial fibrillation, was associated with a 2.7 fold increased risk of having a stroke during that hospitalization and significantly higher in-hospital mortality. Interestingly, patients who developed sepsis with a history of chronic atrial fibrillation had similar risks of in-hospital stroke and death as those with sepsis but without atrial fibrillation. Given that this was a retrospective cohort study, the only relationship that can be drawn is an association – not evidence for new-onset atrial fibrillation causing these worse outcomes. Despite that limitation, and the possibility that these results were confounded by some of the “new onset atrial fibrillation cases being misclassified paroxysmal atrial fibrillation (which is known to have worse outcomes), the association still piqued my interest and made me wonder whether treating my septic patients with new-onset atrial fibrillation with anticoagulation (and specifically unfractionated heparin) would improve their outcomes.
This was followed last year by another cohort study that demonstrated a 37% incidence of venous thromboembolic (VTE) disease in critically ill patients with sepsis, despite those patients receiving appropriate VTE chemoprophylaxis (Kaplan D, et al. Chest. 2015;148(5):1224). And similar to the Walkey study, development of VTE was associated with worse outcomes, specifically longer lengths of stay and numerically higher mortality.
In addition, about the same time, the Canadians published a meta-analysis in Critical Care Medicine evaluating the safety and efficacy of heparin in sepsis (Zarychanski R, et al. Crit Care Med. 2015;43(3):511). While it has a number of limitations, including the fact that the dosages of heparin vary widely amongst the included studies, and that many of the included studies investigate heparin in addition to other anticoagulants, overall they found a small benefit in hospital mortality when heparin is used in patients with sepsis (RR = 0.88) with no statistically significantly increased risk of major hemorrhagic events.
Taking these three articles together, I really started wondering if I should be providing systemic anticoagulation to my critically ill patients with sepsis. Knowing that these data by no means demonstrated the cause and effect of improved outcomes, I still somewhat changed my practice – instead of not even considering it, I adopted a low threshold for starting systemic anticoagulation in my critically ill patients with sepsis, especially if they had another indication (no matter how soft the indication) like new-onset atrial fibrillation.
Well, as often happens in medicine, and especially critical care medicine, new data are available and the pendulum appears to be swinging back. This month, in JAMA Cardiology, Walkey and colleagues have published another retrospective cohort study, this one an analysis of outcomes of patients with atrial fibrillation and sepsis receiving anticoagulation (Walkey AJ, et al. JAMA Cardiol. Epub ahead of print August 3, 2016). This represents a logical follow-up to their previous study, which begged the question that if new-onset atrial fibrillation was associated with more in-hospital strokes and mortality in patients with sepsis, could we improve those outcomes with systemic anticoagulation in these patients. In this study, Walkey and colleagues investigate the use of heparin in about 40,000 patients with sepsis and atrial fibrillation. It turns out that a little more than a third of these patients received systemic anticoagulation. Now, one of the problems with retrospective cohort analyses of treatments is a specific type of bias termed indication bias. What this means is that since the choice to treat with anticoagulation is at the discretion of the treating clinician, patients who the clinician decides to anticoagulate are undoubtedly different than those in whom the decision is made to withhold anticoagulation. For example, patients with atrial fibrillation and sepsis and past history of transient ischemic attacks (TIAs) or a concurrent non-ST elevation myocardial infarction (NSTEMI) may be more likely to receive anticoagulation in their treatment course, as opposed to the patient who also has atrial fibrillation and sepsis but is actively bleeding. That patient is almost assuredly not going to receive systemic anticoagulation. This results in comparing two different populations of patients (i.e. those likely versus those unlikely to receive anticoagulation), who may have different outcomes based on differences in characteristics and not related to differences in treatment. One way of trying to deal with this indication bias (or “apples to oranges” comparison), is to adjust your analyses using a propensity score, which is what was done in this study. In order to adjust for differences in the populations, you calculate a propensity score for receiving the treatment of interest (anticoagulation in this case) and then you use this propensity score to adjust or match for the likelihood of receiving anticoagulation based on population factors.
So what did Walkey and colleagues find? As in their previous study, patients with sepsis and new-onset atrial fibrillation again had a higher rate of in-hospital ischemic stroke compared to patients with sepsis and pre-existing atrial fibrillation (nice consistency of results across studies). However, in both unadjusted and among propensity-score matched patients, the rates of in-hospital stroke did not differ between those who did and those who did not receive systemic anticoagulation (1.3% vs. 1.4%; RR 0.94). This was true for both patients with pre-existing (RR 1.12) and those with new-onset atrial fibrillation (RR 0.85). So, these data do not support anticoagulating patients with sepsis and atrial fibrillation in an effort to reduce their incidence of ischemic stroke. Furthermore, Walkey and colleagues found that clinically significant bleeding occurred more frequently in patients with sepsis and atrial fibrillation receiving parenteral anticoagulation (8.6% vs. 7.2%; RR 1.21). So they didn’t find a signal that providing anticoagulation improved outcomes but did find a signal of increased harm.
So, how will I incorporate these new (or cumulative) data into my clinical practice? Well, I think these data throw real caution to the wind around the practice of anticoagulating patients with sepsis and new-onset atrial fibrillation, especially if the intent is to prevent ischemic stroke. It seems that anticoagulation may not improve that outcome, and is likely associated with potential harm in the form of an increased risk of clinically relevant bleeding. While I still find intrigue in the potential of treating sepsis with an agent that has both anticoagulant and anti-inflammatory properties, it remains unclear to me if heparin, or other available anticoagulants, improve the outcome in patients with sepsis in general. Well designed and conducted, large, multicenter randomized trials need to be done evaluating the effect of systemic anticoagulation (starting with heparin) in patients with sepsis (with or without atrial fibrillation). Given all the data now available, I am back to not providing systemic anticoagulation to my patients with sepsis unless they have a pretty hard indication for anticoagulation (i.e. known clot, documented myocardial infarction, lupus anticoagulant, mechanical heart valve, etc).