In this study, we observed that administration of early empiric antibiotic therapy to stable patients with suspected nosocomial infection was independently associated with an increase in the odds of the development of the composite primary outcome composed by sepsis, septic shock and/or death. When matched analysis for confirmed infection (by positive cultures) was performed, Early Empiric Antibiotic Therapy had no association with the primary outcome.
In critically-ill patients, common signs indicating the presence of an infection like fever and leukocytosis are non-specific and may occur in many non-infectious conditions, particularly in trauma patients [11, 12]. Moreover, the diagnostic criteria for sepsis are under scrutiny, with data showing the use of Systemic Inflammatory Response Syndrome criteria (Sepsis 2.0) has great sensitivity but low specificity while Sequential Organ Failure Assessment (Sepsis 3.0) is more accurate and specific [13]. Therefore, the decision to initiate antibiotic therapy to critically ill patients is strongly dependent on clinical suspicion and/or the presence of new organic failures attributed to an infection and must take into account the balance between the benefits and harms of the therapy.
Some observational studies have investigated whether early antibiotic administration may reduce mortality in patients with suspected infections. This relationship appears to hold true for septic and septic shock patients admitted to emergency services. Among 49,331 patients at 149 hospitals in New York, Seymour et al. observed that a longer period before the administration of antibiotics was associated with higher risk-adjusted in-hospital mortality [2]. In another large multicenter cohort of patients presenting with suspected infection in the emergency department, Liu et al. observed that hourly delays in antibiotic administration were associated with a marginally significant increase in the odds of hospital mortality even in the absence of organ dysfunction or shock [14].
On the other hand, evidence to the contrary was published in critically ill patients by Hranjec et al. showing that early empiric antibiotic treatment was associated with a lower chance of receiving initially appropriated treatment, a prolonged duration of antimicrobial treatment, and a significantly lower survival rate in ICU patients with suspected nosocomial infections without shock [3]. Furthermore, Seymour et al. observed that the time to the administration of antibiotics were not significantly associated with greater in-hospital mortality in a subgroup analysis that included 32610 septic patients without vasopressors [2]. Recently, Abe et al also did not find any association between earlier antibiotic administration and reduction in in-hospital mortality in 1124 patients with severe sepsis (Sepsis 2.0 criteria) admitted to ICU [15].
In our sample, early empirical antimicrobial therapy administered to critically ill surgical stable patients without an obvious source of infection was independently associated with an increase in the incidence of sepsis, septic shock or death within 14 days from the initial suspicion of infection with an OR of 1.83 (1.13–2.82). The other factors independently associated with the incidence of the primary outcome were non-trauma admission (OR 2.32 95% CI: 1.4–3.9) and mechanical ventilation (OR 2.09–95% CI: 1.31–3.35).
The groups were well balanced regarding the cause of admission, age, SAPS 3 and SOFA at inclusion, ICU length of stay before inclusion, and clinical markers of inflammation (heart rate and body temperature). However, laboratory markers of inflammation (White Blood Cell Count and C-Reactive Protein) were higher in Early Empiric Antibiotic Group, as well as confirmed infection (69% vs. 45% in the Conservative Group).
When we performed a sensitivity analysis, adjusting by confirmed infection, early empiric antibiotic therapy had no association with the primary outcome neither in patients with confirmed infection (OR 1.39–95% CI: 0.77–2.48), nor in patients without confirmed infection (OR 1.73 95% 0.63–4.86, p = 0.287). Although these sensitivity analyses were underpowered and showed no significant statistical association between early empiric antibiotic therapy and the development of the primary outcome, neither of them indicates any trend in favor of better outcomes if early empiric antibiotic therapies were administered. Our results are in line with those published by Hranjec et al evaluating a similar cohort of patients [3]. The reasons why early empiric antibiotics in non-septic critically ill patients with suspected but not yet confirmed infection might be detrimental, as shown in our study and others, are a question worth speculating.
Firstly, it is important to highlight that robust evidences for the benefit of an early empiric antibiotic therapy, often disseminated as “the sooner, the better” by guidelines might be true only for sepsis and septic shock [1, 2]. In other clinical scenarios, this potential benefit is not supported by hard evidence. In a multicentric, open label, randomized trial, Alam et al. did not observed any beneficial effect on survival of early antimicrobial therapy improved survival rate in pre-hospital patients with infection suspicion [16].
Secondly, when infection is not evident, overuse of antimicrobial agents may delay the determination of the correct etiology of the inflammatory process and lead to misdiagnosis. In our study, 31% of patients in the early antibiotic treatment group did not show any positive cultures supporting the diagnosis of nosocomial infection. According to Tamma et al., at least 20% of patients exposed to antibiotics are expected to experience significant systemic adverse side effects [17]. Consequently, we can speculate that these adverse effects were not counterbalanced by the overall benefits of these drugs in some patients. The odds ratio for colonization with carbapenem-resistant Enterobacteriaceae in patients exposed to antibiotics is five times higher than in non-exposed individuals[ 18]. Colonization by multidrug-resistant bacteria can last for months [19], and infection caused by these strains carries a mortality rate as high as 44%[ 20]. Data also suggest that the use of broad-spectrum antibiotics is associated with more severe microbiome disruption increasing the risk of subsequent sepsis within 90 days of discharge[ 21]. Hence, patients in the aggressive group are more prone to new sepsis, and these infections have higher chances to be caused by multidrug-resistant strains. Indeed, Guidry et al. followed 190 patients for four years from the initial cohort published in 2012 [3, 22]. Adjusted analysis showed higher long-term mortality for patients in the aggressive antimicrobial therapy at both one and four years, when compared with a conservative approach.
Thirdly, inflammatory markers are not specific for infection, particularly in the critically ill trauma patient. In this population the incidence of Systemic Inflammatory Response Syndrome may reach 91% in the first week post injury [23]. Furthermore, 94% of multiple-trauma patients may present an increase in the systemic inflammation marker C-reactive protein that may persist for more than 7 days [24]. In our study, 43% of patients in the conservative group did not receive any antibiotic therapy within 14 days after suspected infection. Hence, a more conservative led to a more thorough diagnostic investigation and detection of the condition causing hyperthermia, tachycardia and elevated laboratory inflammatory markers.
Even when infection is indeed present, early administration of antibiotics may not the main factor associated with better outcomes in the critically ill non-septic patient [25]. Delay on the diagnosis of the correct source of infection is associated with significant increase in mortality (> 10% increase in in-hospital mortality) [26]. Therefore, efforts should be made to establish a diagnosis as soon as possible, since treatment will probably fail if this premise is wrong [27]. In the meantime, early empiric antimicrobial therapy should be prescribed for septic or deteriorating patients.
Our study has weaknesses and limitations. It was a single center observational trial, and confounders such as selection and information biases might have influenced the results. Also, there was no standardization in the criteria for initiation of antibiotic therapy, with the attending physician being responsible for the decision. This might explain the higher number of culture-confirmed infection in the early empirical antibiotic group. Although there was no difference between groups when using Sequential Organ Failure Assessment, Simplified Acute Physiology Score 3 scores or others markers of disease severity (vasopressor or mechanical ventilation use), we cannot underestimate the clinical impression of the treating physician, which takes into account numerous known and intangible factors in the process of decision making and might have been the explanation for the early initiation of antibiotic treatment. Even though, early empiric antibiotic treatment was also not associated with better outcomes in matched analysis of patients with confirmed infection by positive cultures. This subgroup analysis, however, is probably underpowered. Lastly, our results are valid only for a strict number of critically ill patients in a specific situation.