The FIRST trial

Motivated by a widely publicised patient death,8 evidence that sleep deprivation leads to poor cognitive function and fatigue-related errors and accidents and ‘growing public opinion that long duty hours compromise patient safety and resident well-being’,9 the Accreditation Council for Graduate Medical Education (ACGME) in USA introduced national regulations to limit resident working hours in 2003. The regulations limited duty-hours to 80 hours per week, capped overnight shift lengths and mandated minimum time off between shifts.1 In 2011, persistent concerns over patient safety and resident well-being led to further restrictions,10 including shorter maximum shift lengths for interns and increased time-off after overnight shifts for residents.11 The ACGME policy changes are based on the idea that alleviating resident fatigue should reduce errors and improve patient safety. Yet, evidence for these claims is weak. Indeed, some have suggested that shorter shifts and longer time off actually may make patients less safe.2 Duty-hours restrictions increase the frequency with which residents ‘hand-off’ patients under their care, ‘which could jeopardise patient safety by forcing residents to leave at critical times’.1 Increased transfers may also undermine the goals of resident training if they are ‘unable to follow patients through critical aspects of their care’.1

The FIRST trial investigators sought to address this knowledge gap and rigorously determine the impact of ACGME duty-hour policies on patient safety and resident well-being. The FIRST trial employed a pragmatic, non-inferiority, cluster randomised design. One hundred and eighteen general surgery resident programmes in USA were randomised, as clusters, to two policy interventions. Programmes assigned to the restricted duty-hours arm were to continue to adhere to the ACGME duty-hour policies. Programmes assigned to the flexible duty-hours arm were granted a waiver by the ACGME for four of the duty-hour reforms that concerned maximum shift lengths and minimum time off between shifts, specifically: (1) allowing interns to exceed duty periods of 16 hours; (2) allowing residents to exceed 28 hours duty periods; (3) not requiring residents to have 8–10 hours off between shifts; (4) not requiring residents to have 14 hours off after 24 hours of continuous duty.5 Nevertheless, residents in the flexible duty-hour arm were required to adhere to the ACGME duty-hour requirements of limiting work to 80 hours per week, mandated day-off every 7 days and on-call duty no more frequently than every third night.

The primary patient outcome was a composite of 30-day postoperative mortality or serious complication.1 Patient data were obtained from information routinely collected by the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP). The primary resident outcomes were satisfaction with quality of education and well-being. Resident data were collected from a survey administered on completion of the American Board of Surgery In-training Examination (ABSITE), mandatory for all surgical residents. The study protocol was reviewed by the Northwestern University’s REC, ‘which determined the trial to be non-human-subjects research’, thus exempting the trial from research regulations.1 The REC based its decision on the supposition that the trial involved only ‘the collection or study of existing data, documents, records’.12

The results of the FIRST trial showed that flexible, ‘less-restrictive duty-hour policies for surgical residents were non-inferior to standard ACGME duty-hour policies’, on the outcomes of patient safety and resident well-being.1 The researchers nonetheless concluded that flexible duty-hours are associated with benefits to ‘nearly all aspects of patient safety… [and] continuity of care’;1 their conclusion was based on the presumption that ‘patient care and resident education can be compromised by interruptions in continuity of care’, suggesting that flexible duty-hour policies are better for patient safety, although at the cost of resident time away from work.1

Debate about the FIRST trial

Following the publication of the FIRST trial, Public Citizen and the American Medical Student Association filed complaints accusing the FIRST trial investigators of ‘egregious ethical and regulatory violations’.3 Broadly speaking, critics and defenders identify three ethical issues, on which they are sharply divided.

First, critics of the FIRST trial reject the claim that the trial posed an acceptable balance of benefits and harms to participants. Residents in the flexible duty-hour intervention, they argue, were knowingly subjected to increased fatigue, which rendered them susceptible to a ‘greater risk of motor vehicle accidents, percutaneous injuries and exposure to bloodborne pathogens, depression and, possibly, poorer obstetric outcomes’.3 13–20 Because the risk of harms associated with flexible duty-hour policies were purportedly known to be greater than those of restricted duty-hour policies at the outset of the trial, clinical equipoise—an ethical precondition that requires there to be uncertainty within the relevant expert community regarding the comparative benefits and harms of the study interventions21—did not obtain between the policy interventions.4

Defenders of the FIRST trial disagree. They argue that the trial did in fact pose an acceptable balance of benefits and harms, as they contend it was not known at the start of the trial whether the flexible duty-hour intervention posed greater risk to residents or patients.4 Indeed, defenders argue that it was precisely the uncertainty surrounding the evidence for the policies that prompted the need for a large-scale randomised controlled trial. Accordingly, they argue clinical equipoise did obtain at the outset of the FIRST trial.

Second, critics claim that conditions for a waiver of resident consent were not satisfied. A waiver of consent requires that study participation pose no more than minimal risk to research participants and the study would be infeasible if informed consent was required. However, critics argue study participation posed serious risk, including ‘increased risks of medical errors, death and serious morbidity’, and therefore does not meet the first criterion for a waiver of consent.3

Defenders argue, conversely, that the waiver of consent was justified.4 They claim that risks to participants were minimal because the study interventions were ‘unequivocally’ in equipoise.4 Further, requiring informed consent would have led to serious scientific bias potentially rendering the trial invalid. Given the likelihood of a strong preference for flexible duty-hours among surgical residents, the two study arms would have had substantially different recruitment rates, a source of selection bias in CRTs.

Third, critics claim that even if informed consent was sought and obtained from surgical residents, it would have been invalid.3 They argue that any attempt to obtain consent from residents would amount to duress and coercion, owing to the power differential between staff surgeons and their trainees:

‘[O]btaining the voluntary informed consent of all medical resident subjects who would be enrolled in a trial such as the FIRST trial would never be feasible because the prospective resident subjects would be exposed to significant undue influence and coercion.’3

Hence, critics argue that residents ought to be considered a vulnerable population, given the manifest pressure they are under to conform in residency programmes.

Defenders simply reject the claim that socially important trials that inform evidence-based policy are ethically impossible to conduct. Indeed, defenders seem to suggest that such an ethical conclusion and any argument in support of it, serves as its own reductio ad absurdum.

Applying the Ottawa statement to the FIRST trial

In the face of such stark disagreement about the ethics of the FIRST trial, how should researchers and RECs proceed? Plainly, a systematic approach to the ethics of CRTs is required. In CRTs, the units of randomisation are groups, or ‘clusters’, of individuals, such as schools, neighbourhoods, medical practices or hospitals. CRTs thus differ from individually randomised trials, where the individual study participants themselves are randomised to differing study arms. CRTs can be very complex, as they often ‘evaluate multifaceted interventions targeted at clusters, professionals or individual cluster members’.5 Moreover, the unit of randomisation may differ from the unit of intervention, which may in turn differ from the unit of data collection. For example, ‘in a CRT of teaching a new hand-washing technique to help avoid transmitting infection on hospital wards, the unit of allocation may be the hospital, the intervention may be delivered to health professionals and data may be collected about or from patients’.5 This complexity can make identifying and addressing the ethical issues paradigmatic of CRTs challenging.

The Ottawa Statement on the Ethical Design and Conduct of Cluster Randomized Trials seeks to diminish these challenges and promote consistency in the ethical analysis of CRTs by providing 15 recommendations for their ethical design, conduct and review, including: (1) guidance on REC review; (2) identification of research participants; (3) informed consent procedures; (4) the role and authority of gatekeepers; (5) benefit-harm analysis and (6) the protection of vulnerable participants.5 In what follows, we assess the utility of the Ottawa Statement by applying its recommendations to the FIRST trial.

Ethical issues requiring further analysis

As demonstrated, the Ottawa Statement systematically identifies and provides recommendations to address the ethical issues common to a broad class of CRTs. Moreover, the Ottawa Statement grounds its recommendations in internationally accepted ethical principles, as opposed to national research regulation, thereby ensuring its guidance is of global relevance. As such, the Ottawa Statement provides much-needed clarity to the debate about the FIRST trial. The confusion inherent to the debate about the FIRST trial is perhaps predictable, as US federal regulations were not written with CRTs in mind and neither critics nor defenders cite the recommendations outlined in the Ottawa Statement.

Nonetheless, the application the Ottawa Statement to the FIRST trial makes it clear that there remain important issues to be addressed, specifically pertaining to policy trials involving healthcare providers and trainees. Here, we briefly enumerate three ethical issues requiring further work.

First, a clearer understanding of the conditions that satisfy equipoise for trials involving policies is needed. Application of equipoise to policy CRTs seems to lead to solutions that are at once too restrictive and too permissive. Critics of the FIRST trial effectively argue that clinical equipoise did not obtain with respect to a policy once it has been implemented in practice. On this view, once a policy has been enacted, it becomes the standard of care and depriving individuals of the protection of the policy violates equipoise. However, this position would undermine our ability to study existing policies that lack sound support and would hinder the development of evidence-based policy. Conversely, if a paucity of evidence—characteristic of many policies—is sufficient to establish equipoise, then it appears that the conduct of almost any CRT involving policy interventions is ethically permissible, even in cases in which policies clearly protect people from substantial risks. Neither view seems entirely palatable. What, then, is the appropriate application of equipoise to CRTs involving policy interventions?

Second, a clearer account of the duties of health providers in the research setting is needed. The main REC determined that the FIRST trial was not research involving human participants and failed to recognise the fact that the surgical residents were research participants. Indeed, it is not uncommon for RECs to miss that healthcare providers, including trainees, are research participants. This phenomenon may stem, in part, from the belief that healthcare providers have an obligation to participate in research.25 This raises the following important questions, which are also applicable to a broader class of CRTs: What additional guidance might help researchers and RECs recognise when healthcare providers are research participants? Do healthcare providers have an obligation to participate in research? If so, are healthcare providers obligated to participate in research categorically or is their participation subject to their own discretion? What are the implications for informed consent procedures?

Third, and finally, when ought health providers be considered vulnerable participants? As we have seen, some argue that CRTs are simply ethically impossible to conduct when they involve vulnerable healthcare providers or trainees. However, it plainly does not follow from the contention that trainees are vulnerable, that research involving them is ethically impermissible. Rather, research involving vulnerable participants may be conducted with the appropriate protections in place.24 Nevertheless, these concerns do raise the following important questions: Are trainees and other individuals in subordinate positions ‘vulnerable participants’, as the Ottawa Statement contends? If so, what protections should be implemented to alleviate concerns about vulnerability, such as duress and coercion?

With the increased prevalence of CRTs involving policy interventions that target healthcare providers, a definitive answer to these important questions is urgently required. The answers to these questions are central to the debate about the FIRST trial and to CRTs of policy interventions involving healthcare providers generally. Future guidance on the design and conduct of CRTs must therefore adequately address these questions if important research—critical to informing evidence-based policy for healthcare providers and trainees—is to proceed both ethically and expeditiously.