Rationale for a factorial cluster crossover randomized trial
Individual-level RCTs are considered to be the gold standard in the evaluation of interventions [
25] and are needed for regulatory purposes to show the efficacy of a drug under highly controlled conditions in a highly selected group of patients. Therefore, findings from these RCTs do not necessarily translate into effectiveness in a real-world setting. In particular, individual-level RCTs are not well suited for studying system changes, and a cluster design may be more appropriate as this design aligns research with clinical care, allowing enrolment of a representative patient population in a real-world setting [
23,
26,
27]. Cardiac surgery is conducted in specialized centers using highly standardized procedures, an approach that lends itself to a cluster RCT design. In our study, the antibiotic prophylaxis regimen to be studied becomes part of the standardized operating procedures for a certain period for all patients undergoing cardiac surgery at the participating centers.
In addition to the study goal to evaluate effectiveness in a real-world environment rather than efficacy, a trial randomizing individual patients would likely not be feasible due to financial constraints, considering the large sample size needed to power the study properly, in the absence of any interest from industry to support such a trial. In contrast, if the randomization occurs at the level of the health-care center, and therefore, the study intervention becomes the standard operating procedure for that center, the resources required are significantly reduced.
Similarly, we chose to incorporate a factorial component to evaluate two different interventions (choice and duration of antibiotics, respectively) without requiring a substantial increase in sample size. This is an important advantage of a factorial design, especially when resources are limited [
28]. Furthermore, it has been suggested that using a factorial design is particularly beneficial when the purpose of the study involves choosing an optimal intervention [
29].
Finally, the crossover element of the study allows each center to act as its own control. This approach is proposed to protect from trends over time that are unrelated to antibiotic prophylaxis, such as possible shifts in the epidemiology of s-SSIs, while avoiding a negative impact on the power of the study.
Ethical considerations
Obtaining individual consent for this study would jeopardize one of the main advantages of the study design, that is, testing the comparative effectiveness of the interventions in a real-world setting rather than testing efficacy in a subgroup of highly selected patients. Similarly, obtaining individual patient consent would not allow assessment of the intervention in important groups of patients, such as emergent and urgent surgery patients or those from diverse ethnic backgrounds who may not be proficient in the official languages of Canada.
The Canadian Tri-Council Policy Statement allows research ethics boards to approve an alteration to the informed consent process, such as a waiver of consent, if the following criteria are met: (1) there is no more than minimal risk to participants, (2) the alteration to consent requirements is unlikely to affect the welfare of participants adversely, (3) it is impossible or impracticable to carry out the research properly given the research design if prior consent is needed, and (4) there is a plan to offer participants the possibility of having their data deleted from the study database [
30]. Similar criteria are used in the United States [
31].
Considering that both antibiotics as well as the durations used in this study are considered the standard of care in some centers, we strongly believe that our trial meets the first criterion required for a waiver of individual informed consent as the risk to participants is perceived to be minimal. The duration and antibiotics chosen in this clinical trial represent either current practice at some centers or are in keeping with current guidelines. The variety of antibiotic practices across Canada and elsewhere supports the clinical equipoise surrounding antibiotic choice and duration in cardiac surgery. Thus, the risk for participants will be minimal and the risk related to the surgical procedure is identical in all groups. Cefazolin and vancomycin are routinely used and have a favorable safety profile when used for a short of time (see
Discussion section on
Safety).
A maximum of 48 h of antibiotic prophylaxis was chosen for the long-term strategies as previous studies have shown the risk of developing an infection with resistant pathogens increases with antibiotic use longer than 48 h [
32]. Similarly, the extended duration of the use of antibiotics potentially increases the risk of other adverse events, such as diarrhea or infection with
Clostridium difficile. However, there is no clear evidence that the risk is significantly higher with a 48-h course versus a shorter course of antibiotic prophylaxis. This possible adverse effect highlights the need to conclusively compare short- and long-term antibiotic prophylaxis.
The second requirement to waive consent is that this alteration is unlikely to affect the welfare of the participants adversely. This criterion is met in the study as the antibiotic regimens being investigated are in use across Canada or follow current guidelines. The choice and duration of antibiotics are not decisions made by patients, but by surgeons or anesthesiologists, usually by following internal guidelines and policies. As such, participation in the trial does not negatively affect the patient’s autonomy. If patients have a strong opinion about the antibiotic regimen they would like to receive, the clinician is free to overrule the recommended study regimen.
Thirdly, obtaining consent from individual participants for the antibiotic prophylaxis strategy would not be feasible in a cluster randomized controlled trial, as the protocol would need to be applied to all patients undergoing cardiac surgery during the study period. The resources required to obtain individual consent would be quite significant, including increased study personnel for recruitment and resulting costs. Furthermore, obtaining consent can be difficult in emergent and urgent procedures. Such patients are often excluded from data collection and analyses, which could affect the extrapolation of study results to this population of participants. For these reasons, it would be impracticable to carry out the research properly given the study design if prior informed consent was required from every participant.
Lastly, to meet the final criteria to waive informed consent, patients will be informed of the study while still in hospital before their data has been collected by study personnel, and again during the telephone follow-up call at 3 months after surgery. The patients are provided with contact details for the study coordinator, whom they can contact any time with questions around the study as well as to opt out from data collection.
Therefore, a waiver of consent or an opt-out option for patients is considered appropriate in this study of the clinical effectiveness of the current standard of practice for antibiotic regimens [
33]. The research ethics boards at the pilot sites have agreed that these criteria are met and have waived the need for individual patient consent. However, the board at the first site insisted on an information letter with an option to opt out of data collection, which was also found acceptable by the other two research ethics boards. However, two ethics boards in Europe (Zurich, Switzerland, and Bavaria, Germany) concluded that the criteria for a waiver of consent are not met and would have insisted on essentially applying the same rules to this clinical effectiveness trial as for a phase III clinical trial.
In summary, the acceptable consent model for a clinical effectiveness trial using a cluster RCT design may differ from one jurisdiction to the next, and researchers must be aware that the approval given by one research ethics board will not necessarily translate into approval in other jurisdictions. Given the research question and study design, we aim to recruit only study sites that are able to obtain a waiver of consent for the reasons summarized above.
In accordance with good clinical practice and expectations by research ethics boards, all protocol modifications are communicated to investigators and ethics boards, and the registration on
clinicaltrials.gov is updated.
Safety
Vancomycin has well-known adverse effects of nephrotoxicity and ototoxicity; however, these adverse effects have virtually never been reported with trough levels lower than 15 mg/L and 50 mg/L, respectively. Trough levels after one single dose will remain below this toxic level; thus, a single dose is very unlikely to cause significant renal injury or ototoxicity. Even when three more doses are administered in the long-term strategy, these side effects are highly unlikely and a steady state would not yet have been reached. However, due to concerns by some surgeons, perioperative physicians, and intensivists, vancomycin levels have been occasionally been measured in patients that they deemed at risk of kidney injury. In the light of the availability of these data and a recent study that found an increase in acute kidney injury with combination prophylaxis [
15], we will conduct a review of postoperative vancomycin levels that had been measured by intensivists during our longer-term combination treatment strategy at one of our study sites.
Given that only antibiotics that are routinely used for perioperative prophylaxis are administered as part of this study, a data safety and monitoring committee was not deemed to be necessary for the pilot study.