The association between premorbid beta blocker exposure and mortality in sepsis—a systematic review

Background The effect of premorbid β-blocker exposure on clinical outcomes in patients with sepsis is not well characterized. We aimed to examine the association between premorbid β-blocker exposure and mortality in sepsis. Methods EMBase, MEDLINE, and Cochrane databases were searched for all studies of premorbid β-blocker and sepsis. The search was last updated on 22 June 2019. Two reviewers independently assessed, selected, and abstracted data from studies reporting chronic β-blocker use prior to sepsis and mortality. Main data extracted were premorbid β-blocker exposure, mortality, study design, and patient data. Two reviewers independently assessed the risk of bias and quality of evidence. Results In total, nine studies comprising 56,414 patients with sepsis including 6576 patients with premorbid exposure to β-blockers were eligible. For the primary outcome of mortality, two retrospective studies reported adjusted odds ratios showing a reduction in mortality with premorbid β-blocker exposure. One study showed that premorbid β-blocker exposure decreases mortality in patients with septic shock. Another study showed that continued β-blockade during sepsis is associated with decreased mortality. Conclusion This systematic review suggests that β-blocker exposure prior to sepsis is associated with reduced mortality. There was insufficient data to conduct a bona fide meta-analysis. Whether the apparent reduction in mortality may be attributed to the mitigation of catecholamine excess is unclear. Trial registration PROSPERO, CRD42019130558 registered June 12, 2019. Electronic supplementary material The online version of this article (10.1186/s13054-019-2562-y) contains supplementary material, which is available to authorized users.


Introduction
The Sepsis-3 consensus defines sepsis as a life-threatening organ dysfunction caused by a dysregulated host response to infection [1]. While our understanding of sepsis pathophysiology is increasing, controversies in haemodynamic management persist [2,3]. The most recent surviving sepsis guidelines recommend noradrenaline as the first-choice vasopressor because of its vasopressor and positive inotropic properties [strong recommendation, moderate quality of evidence [4]]. In contrast, the concept of 'decatecholamisation' emerged in the last decade stemming from the recognized negative effects of catecholamines in sepsis [3,5,6]. Interestingly, the β-adrenergic blockade has emerged as a possible treatment option for blunting the adrenergic response in early sepsis with potential effects on the modulation of cardiogenic, metabolic, immunologic, and coagulopathic derangements in sepsis [7].
Early administration of the short-acting β-blocker esmolol in a recent trial showed a reduction in 28-day sepsis mortality [8,9]. Furthermore, some studies have suggested a benefit of premorbid β-blocker exposure on sepsis outcomes [10,11]. Multiple systematic reviews have since concluded that there is limited preliminary evidence for the use of β-blockers during sepsis [12][13][14], while others are skeptical [15]. However, to date, no published systematic review exists on the effects of premorbid β-blocker exposure on sepsis outcomes, including mortality. Therefore, we set out to systematically examine the evidence from all human studies on premorbid β-blocker exposure and sepsis.

Materials and methods
This study follows the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [16] and was registered with the international prospective register of systematic reviews (PROSPERO; CRD 42019130558). The MOOSE checklist is appended as Additional file 1: Table S1.

Data sources and searches
Three databases, EMBase, MEDLINE, and Cochrane were searched on 30 January 2019 for records dating from database conception to the date of search that was last updated on 22 June 2019. The search was only limited to human research. Duplicates were removed using the Ovid platform and checked for any incorrect removal. Hand searching from reference lists was also performed. The full search strategy is appended as Additional file 4: Figure S1.

Study selection
Inclusion criteria for this review were guided by the 'Patient, Population, or Problem, Intervention, Comparison, Outcome, Study Design or Setting' (PICOS) framework [17] (Table 1). Patients exposed to β-blockers prior to an episode of sepsis or septic shock and were cared for in the emergency department (ED) or intensive care unit (ICU) were included in this review. Observational studies were eligible. Excluded were case studies/small series (< 20 patients overall) and review articles. The abstracts were assessed by two investigators (KT, MH) independently, and disagreements were resolved with a third investigator (MN).

Data extraction and quality assessment
Data from eligible studies were independently extracted by two investigators (KT, MH). Where required, study authors were contacted directly to kindly provide missing research data. The Risk Of Bias In Non-randomized Studies -of Interventions (ROBINS-I) tool [18] was used to independently assess (KQ, MH) the quality of studies.

Data synthesis and analysis
Adjusted outcome data were combined using the inverse variance method [19]. Heterogeneity between studies was measured by Higgin's and Thomson's I 2 [20]. Statistical analyses were performed using Review Manager version 5.3 (Copenhagen: The Cochrane Collaboration, 2014)

Study selection
The initial search returned 2128 abstracts, all in English. Two thousand sixty-four abstracts were manually screened after removal of 64 duplicates. After screening, 16 studies were initially selected for data extraction. Where required, the corresponding authors were contacted to obtain necessary data for statistical analysis. Seven studies were excluded for not meeting all inclusion criteria. The list of studies excluded is appended (Additional file 2: Table S2). Overall, a total of nine studies were found to be eligible, comprising 56,414 patients with sepsis, including 6576 patients with premorbid exposure to β-blockers (Fig. 1).

Characteristics and type of studies
All studies were retrospective cohort studies, and the data were collected between 1999 and 2017.

Risk of bias assessment
All observational studies of premorbid medication use are at risk of bias because of confounding. Five studies included in this review [10,11,24,26,27] were judged to be of moderate risk of bias for the primary outcome of mortality as they reported adjustment of confounding variables via statistical analysis. Four studies [21][22][23]25]  were judged as having serious risk of bias due to confounding as the authors did not perform statistical analysis to correct for confounders. The risk of bias assessment using ROBINS-1 tool for each trial is appended (Additional file 5: Figure S2) with reasonings attached (Additional file 3: Table S3).  Table 3). Subgroup analysis investigating the effect of age, gender, organ dysfunction, and previous comorbidities did not alter the results. Adjustment for previous medication used including calcium channel blockers, amiodarone, angiotensin-convertingenzyme inhibitors, diuretics, or any nonsteroidal anti-inflammatory drugs also did not alter the results. The authors also conducted a propensity matching analysis, which led to similar results (OR = 0.72; 95% CI [0.57-0.91]; p = 0.04).
The study by Hsieh et al. showed that premorbid βblocker exposure was not associated with a significant decrease in hospital mortality in patients with sepsis and septic shock (aOR = 0.89; 95% CI [0.76, 1.04]; p = 0.1484; Table 3). However, subgroup analysis of patients with septic shock showed that premorbid βblocker exposure was significantly associated with decreased hospital mortality (aOR = 0.68; 05% CI [0.56, 0.82]; p = 0.0001). In patients without septic shock, premorbid β-blocker exposure was associated with     We compared the mortality data from the three studies that adjusted for potential confounders. Pooled analysis of the three studies showed an average odds ratio, aOR = 0.79; 95% CI (0.67, 0.92), p = 0.004; Fig. 2. However, there was substantial heterogeneity (i 2 = 74%) between the studies, indicating that a meta-analysis is premature and that further studies and subgroup analyses are needed to validate the results.
Our systematic search also included grey literature in the form of conference abstracts. Mortality data from three studies reported as conference abstracts showed a trend towards a decrease in mortality with premorbid βblocker exposure. However, the results were not statisti-   Table 3).

Clinical parameters
Only four studies by Contenti et al. [21], Charles et al. [23], Sharma et al. [25], and Fuchs et al. [26] provided clinical parameter data. However, reporting of parameters was inconsistent. There was no significant difference in the requirements for vasopressor infusion across all four studies. Contenti et al. and Charles et al. found that premorbid β-blocker exposure was associated with decreased heart rate; Sharma et al., did not report heart rate data. Continuation of β-blockade during sepsis was not associated with a decrease in heart rate in the first 24 h [26]. Premorbid β-blocker use was found to be associated with lower initial plasma lactate levels by Contenti et al., but not by Charles et al.. The continuation of β-blockade during sepsis was associated with lower plasma lactate levels in the first 24 h [26].
There were no significant differences in all other relevant parameters including mean arterial pressure, Sequential Organ Failure Assessment (SOFA) score, Acute Physiology and Chronic Health Evaluation (APACHE)-II or III score, and incidence of mechanical ventilation. The clinical parameter data are presented in Table 4.

Discussion
This is the first systematic review examining the role of premorbid β-blocker exposure on mortality outcomes in patients with sepsis. While there was not enough data to conduct a meta-analysis, pooled adjusted odds ratio from three studies indicated a potential decrease in mortality associated with premorbid β-blocker use, albeit with substantial heterogeneity. Our results provide preliminary evidence of a potential association between premorbid β-blocker use and mortality in sepsis and add to the emerging evidence suggesting harmful effects of adrenergic stress on mortality in sepsis. We discuss the effects of premorbid β-blocker exposure on the adrenergic response in early sepsis.
Cardiac dysfunction in sepsis is common and has both systolic and diastolic components [5]. However, only diastolic dysfunction seems to be associated with mortality [28,29]. While being on premorbid βblockers may reduce systolic function, the reduction of adrenergic response in sepsis (decreasing heart rate, prolongation of diastolic time, and improved coronary perfusion) can lead to mitigation of diastolic dysfunction [28,29]. Further, the risks of myocardial ischemia may be decreased due to reduced myocardial oxygen consumption [14].  Patients with septic shock are often treated with large doses of exogenous catecholamines for haemodynamic stabilization. The most recent Surviving Sepsis Campaign guidelines recommend using noradrenaline as the firstline agent for vasopressor therapy, with adrenaline or lowdose vasopressin as second-line agents [4]. Increased dosage and duration of noradrenaline administration has been associated with higher incidence of new onset atrial fibrillation [3]. Excessive catecholamine levels may also play an important role in sepsis-related cardiac dysfunction by causing cardiomyopathy and cardiomyocyte necrosis [5,7]. β-adrenergic blockade could reduce the amount of exogenous catecholamines used by restoring sepsis-induced downregulation of β-adrenergic receptors [12,30]. Four of the included studies in this systematic review, however, found that premorbid β-blocker exposure was not associated with a significant difference in vasopressor requirements during sepsis. Similarly, Fuchs et al. found that continuing chronic beta blockers during acute phase of sepsis is not associated with increased use of catecholamines.
Interestingly, Singer et al. reported that patients with premorbid exposure to non-selective β-blockers had lower mortality rates compared to patients with premorbid cardioselective β-blocker exposure [11]. This suggests that β-blocker modulation of non-cardiac adrenergic responses to sepsis may also have an important role. Furthermore, β-blockers may potentially positively modulate the disturbed autonomic (sympatheticparasympathetic) balance in sepsis [31].
Adrenergic response to sepsis induces a hypermetabolic state characterized by increased energy expenditure, hyperglycaemia, lipolysis and proteolysis, supressed ketogenesis, and negative nitrogen balance resulting in eventual loss of lean body mass [32]. β2-adrenergic blockade appears to have the potential to reverse hyperglycaemia and reduce proteolysis [7]. For example, the use of propranolol in children with severe burns appears to attenuate hypermetabolism and reverse muscle catabolism [33].
The immune system is also modulated by the adrenergic responses to sepsis [34]. The β-adrenergic system regulates apoptosis, mitochondrial function, and inflammatory cytokine production. β-blockers influence the pattern of cytokine synthesis with β1 blockers downregulating a proinflammatory response, whereas β2-antagonization seems to have an opposite effect, at least in chronic heart failure [35].
In sepsis, β2-adrenergic stimulation selectively inhibits CD4 + lymphocyte Th1 function and favours the Th2 responses that inhibit macrophage activation, T cell proliferation, and proinflammatory cytokine production [7]. CD8 + lymphocyte function may also be suppressed by β2-adrenergic stimulation [36]. The derangement in lymphocytic function induced by catecholamines is thus reminiscent of sepsis-induced immune suppression and could even be considered as one of the mechanisms. However, to date, the evidence for any beneficial use of β-adrenergic blockade on immune function in sepsis has been conflicting [7].
Sepsis results in a pro-thrombotic state with increases in plasma tissue factor and von Willebrand factor levels [37]. Platelets also express adrenergic receptors on their surface [38]. However, there are conflicting effects of β1 and β2 pathways on platelet function [7]. The use of βadrenergic blockade led to decreased endothelial cell damage in a murine model of shock coagulopathy [39]. This suggests that premorbid β-blocker therapy might mitigate shock-induced endotheliopathy (SHINE), attenuating sepsis-associated coagulopathy [40].
Nonetheless, multiple questions on the role of β-adrenergic blockade in sepsis remain unanswered. On top of safety and efficacy concerns, the duration and dosage at which β-blockade should be performed remain to be elucidated. Furthermore, the timing of therapeutic β-adrenergic blockade initiation is also controversial. The results of our systematic review suggest that we should not discount β-blockers during sepsis. Instead, we may consider continuing chronic β-blockers and perhaps introduce β-blocking drugs early in the sepsis management, especially the non-cardioselective ones.

Strengths and limitations
This study analysed data from nine observational studies, four of which were reported as conference abstracts. There was not enough data to conduct a meta-analysis. By nature of observational studies, systematic confounding and risk of bias cannot be ruled out. The risk of bias can be reduced by adjusted analysis. Analysis of pooled adjusted odds ratio revealed a significant decrease in sepsis mortality with premorbid β-blocker exposure, but adjusted data were available only from three studies. Despite the three studies providing data on the majority of patients included in this review, substantial heterogeneity is present and residual confounding is likely. Potential sources of confounding include the variable definitions of premorbid β-blocker exposure used by the included studies, the appropriate prescription of β-blockers to all included patients, and patient compliance to treatment.
The conclusions that can be drawn from this study are also hampered by the lack of clinical parameter data, limiting our ability to decipher the likely mechanism/s by which premorbid β-blocker exposure may lower sepsis mortality.

Conclusion
This systematic review suggests that β-blocker exposure prior to an episode of sepsis could have a role in reducing sepsis mortality. More evidence, however, is needed