7 versus 14 days of antibiotic treatment for critically ill patients with bloodstream infection: a pilot randomized clinical trial

Both antibiotic use and the acquisition of antibiotic-resistant organisms are high in intensive care units (ICUs), where critically ill patients are vulnerable to bacterial infections and antibiotic complications. Shortened treatment durations offer an appealing opportunity to maximize the benefits while minimizing the harms and reducing the costs of antibiotic therapy [1]. Randomized clinical trials (RCTs) have established that shorter-duration treatments are sufficient for a wide range of bacterial infections [1‐5], including some infections among critically ill patients [6], but the optimal treatment duration for bloodstream infections remains understudied.

Our systematic review of the academic literature revealed no trials of shorter- versus longer-duration treatment among adult patients with bloodstream infection, but did uncover 24 studies (of 7595 patients) of shorter versus longer antibiotic treatment for infections commonly complicated by bacteremia. Cure rates were similar in patients receiving shorter (3–7 days) versus longer (7–21 days) treatment (risk ratio 1.00, 95% confidence interval (CI) 0.98–1.01) [7]. Although, subgroup outcome data were only available for small numbers of bacteremic patients across these studies (n = 115), shorter-duration treatment was associated with similar clinical cure (45/52 versus 47/49, risk ratio 0.88, 95% CI 0.77–1.01) and survival rates (15/17 versus 26/29, risk ratio 0.97, 95% CI 0.76–1.23) as longer-duration treatments. Given the absence of evidence to guide treatment durations for these bacteremic patients, we conducted a national practice survey of infectious disease physicians and critical care physicians [8]. Among these clinicians, 14 days was the most commonly reported treatment duration, but shorter treatments (usually 7 or 10 days) were recommended by nearly half of practitioners [8]. In single [9] and multicenter [10] observational studies of actual practice, we confirmed that critically ill patients with bacteremia receive a median of 14 days of treatment, but with wide variability (interquartile range (IQR) 9–17.5 days), further justifying the need for a trial comparing 7 versus 14 days of treatment. If such a trial could establish that 7 days is non-inferior to 14 days of treatment, this could lead to massive reductions in antimicrobial use, resistance and complications, and an estimated annual cost savings of $678 to 798 million in North America and $1.4 to 1.6 billion across Europe [11].

Prior to embarking on a large multicentre trial, we conducted a pilot RCT to test the feasibility of this trial design. We hypothesized that the co-primary feasibility outcomes (recruitment rate and adherence to treatment duration protocol) would confirm that a definitive trial could be conducted to compare 7 versus 14 days of antibiotic treatment for critically ill patients with bloodstream infections.

The BALANCE pilot RCT has confirmed the feasibility of conducting a large, multicenter RCT comparing 7 versus 14 days of antibiotic treatment for critically ill patients with bloodstream infection. It was feasible to recruit patients into this trial, and adherence to treatment duration protocols was adequate to ensure wide separation in antibiotic treatment between groups. No patients were withdrawn from the trial, and there were no losses to follow-up in ascertainment of any of the study outcomes in ICU, in hospital, or at 90 days.

Although, there was some heterogeneity in recruitment rates across sites, we achieved our overall target recruitment average of one patient per ICU per month. The principle BALANCE trial will seek to test whether 7 days of treatment is associated with a non-inferior survival rate as compared to 14 days of treatment, with a non-inferiority margin of 4%; after inflation for two interim analyses and potential losses to follow-up, the sample size requirement will be 3598. Therefore, at the rate of recruitment in our pilot RCT we will require approximately 60 ICUs, recruiting patients over a 4- to 5-year period. A study of this magnitude cannot be conducted within a single country. Through Canadian Institutes of Health Research (CIHR) funding, we have initiated enrollment in 20 Canadian ICUs (http://​balance.​ccctg.​ca). The BALANCE research program has also garnered enthusiastic collaboration from critical care trials groups in other countries including the United States, Australia, New Zealand, Saudi Arabia, France, Germany and the United Kingdom, where enrollment and potential parallel funding applications will help to foster accrual of the remaining 40 ICUs.

The adherence to treatment duration protocol was lower than our initial target of 90%, but it was similar to the landmark PneumoA study of 8 versus 15 days of treatment for ventilator-associated pneumonia, which detected 18% protocol non-adherence in the 8-day treatment arm [6], and was far lower than the non-adherence rates of as high as 50% seen in some studies of biomarker-guided treatment for infections in critically ill patients [24]. Still, non-adherence to treatment duration protocol, particularly in patients in the 7-day treatment arm, represents the main threat to the validity of the principle BALANCE trial. Non-adherence cannot be completely eliminated because some ICU patients will develop secondary infections or persistent sepsis for which antibiotics will be re-initiated or continued, and it is appropriate for those patients to receive treatment, as would occur outside of a trial. We aim for a pragmatic approach to intervention duration, and for patient care. However, several lessons learned from this pilot RCT will enable us to minimize non-adherence in the larger RCT, including: use of delayed randomization when imaging is pending to rule out abscesses or confirm success of initial source control procedures; daily research coordinator contact with the clinical team to monitor for premature or delayed antibiotic discontinuation; and real-time involvement of both infectious diseases and ICU site co-investigators in discussing cases of potential non-adherence. The fact that adherence rates were higher in the second versus the first half of the enrollments at each site suggests that adherence rates may improve significantly over time as the main RCT unfolds. Nevertheless, in the principle BALANCE RCT it will be important to conduct both a primary intention-to-treat analysis and a secondary per-protocol analysis; the strength of the inference of non-inferiority will depend on whether the results of the two analytic approaches are in harmony.

We have not measured treatment group-separated clinical and safety outcomes: the pilot RCT is inadequately powered to assess these outcomes. Spurious differences between groups could lead to unnecessary concerns about embarking on the main RCT, while similar outcomes between groups could lead to premature acceptance of non-inferiority and the adoption of shorter-duration treatment [23]. Instead, this is an internal pilot and the patients enrolled to date will be included and analyzed within the principle RCT. Nevertheless, from the aggregate clinical outcomes in the cohort of 115 randomized patients we have still been able to confirm low rates of antibiotic-related adverse events, and a wide bi-modal distribution of antibiotic-free days. Importantly, the overall rates of Clostridium difficile and secondary infections with antibiotic-resistant organisms are high enough that we may be able to detect potentially important benefits of shorter treatment in the principle RCT. For example, the main trial will have 85% power to detect a reduction in Clostridium difficile to 3% from 5% with 7 versus 14 days of antibiotics.

One important limitation of the BALANCE pilot and main RCT is that we will only be able to compare outcomes among patients with two fixed durations of treatment, 7 versus 14 days, rather than giving individual patients the minimum duration necessary for curing their particular infection. However, we believe that there are currently no adequate clinical indicators of cure in critically ill patients with bloodstream infection since host signs of sepsis can persist beyond cure of the triggering infection. Although biomarkers, such as PCT, offer a promising means to individualize and reduce antibiotic treatment durations [25, 26], there has been poor adherence to PCT guidance in key trials [24], and this technology has not yet been widely adopted. The results of the BALANCE RCT could be potentially synergistic with prior and subsequent biomarker-focused trials of treatment duration; if 7 days is non-inferior to 14 days of treatment this may lead to greater physician adherence to biomarker-guided stopping rules in the future. Moreover, the blinded PCT results from our study were above the 0.25-IU/mL treatment threshold in the majority of patients at both day 7 and at day 14, suggesting that PCT-guided treatment has the potential to lead to treatment prolongation, rather than reduction, in this clinical context. Finally, as in other studies that compare two arbitrary and fixed intervention thresholds (such as the TRICC trial of transfusion thresholds) [27], an adequately powered trial will provide a benchmark duration of therapy around which to make individualized treatment decisions.