Research poses no greater net risk than the standard of care

The research setting can be very different from the real world context which guidelines address. For instance, guidelines may need to simplify complicated information for practitioners and policymakers, or suggest a standard of care that can be feasible on a large scale. These motivations sometimes make guidelines poorly-calibrated for the care and attention that can be provided in a research context. Moreover, because they have controlled conditions, close monitoring or interventions that could not be provided outside of the study, research can sometimes offer the same risk to benefit ratio as the standard of care without offering the standard of care itself.

Assuming that the deviation from standard of care is necessary to answer an important scientific question, when a study protects participants such that they will not be exposed to greater net risk than if the guideline had been followed, the study offers the functional equivalent of the standard of care and may not require a high level of justification. This view is consistent with influential ethics guidelines. The Declaration of Helsinki and the Council for International Organisations of the Medical Sciences both permit deviating from the best proven standard of care where there is scientific reason to do so and participants are not at risk of serious or irreversible harm.13 14

One example is a study testing whether HIV-positive pregnant women should receive preventative therapy for TB. In December 2010, WHO issued a strong recommendation that HIV-positive people in TB-endemic settings, regardless of their level of immunosuppression, should receive isoniazid preventative therapy to reduce their risk of developing TB.15 Recognising the limitations of TB programmes in developing countries, WHO did not require a positive tuberculin skin test result before initiating therapy, and indicated that the recommendation included pregnant.

At the time WHO recommendation was issued, a study on the safety and toxicity of isoniazid in pregnant women was in development. Isoniazid is associated with increased hepatoxicity. Given that pregnancy itself is a risk factor for hepatoxicity, the study was motivated by the concern that pregnant women might face higher risk from isoniazid than the general population. Furthermore, WHO recommendation might not be favourable for women who had not been exposed to TB. To address these concerns, the study was designed to randomise pregnant women to either immediate or delayed isoniazid (started at 12 weeks post partum). The risks were minimised to protect women in the control arm. The study required that all women with CD4 counts below 350 cells/mm³ be on ART, which would decrease the risk of being infected with TB, and that each visit would include TB screening. If TB was detected, the women and/or their infants would be referred to therapy. Thus, women in the study would not be exposed to greater net risk than if they had received guideline-driven isoniazid. As long as the study protections were the functional equivalent of providing pregnant women with isoniazid, the research could justifiably deviate from the guideline.

Continued need for additional evidence

A second justification for deviating from guidelines is when there is a continued, compelling need for additional evidence. This can be the case if a guideline has insufficient evidence to apply to a particular subpopulation, such as infants or children, and there is reason to suspect that the intervention poses different benefits and risks for that subpopulation. The previous example of isoniazid therapy in pregnant women may fall under this category.

More generally, guidelines may be promulgated without full evidentiary support for a variety of reasons. A recent article found that the Infectious Diseases Society of America practice guidelines are frequently issued based on expert opinion alone, and only 14% are based on evidence from randomised controlled trials.16 In cases like these, research can serve a critical function by filling gaps that remain even after much of the needed information is available, whether or not the results end up supporting what the guideline originally recommended.

For instance, complying with guidelines for treatment of hospital-acquired pneumonia recommending more frequent combination broad-spectrum treatment has been found to significantly increase mortality.17 This was likely because the guidelines should not have applied to all hospitalised patients. As one commentator explained, ‘Because of the fatal flaw in making of an accurate diagnosis of intensive-care unit pneumonia and the inherent inability to separate uninfected colonised patients from infected patients, it is probable that a notable number of uninfected patients received unnecessary broad-spectrum combination therapy’.18 Therefore, a study deviating from these guidelines to test a different strategy would likely have been justified.

Guidelines can be justified by the evidence but are controversial

Finally, if a guideline can be justified based on the evidence, but practitioners disagree about the interpretation of the evidence and therefore do not regularly comply with the guidelines, then it may be important for research to be conducted in certain cases. Two additional criteria for this category are that the research question is relevant to the source of disagreement and answering that question requires deviating from the guidelines. Additionally, this reason for deviation would not apply to cases where the failure to adopt the guideline can be addressed in other ways. If practitioners fail to adopt an intervention out of inertia, irrationality, insufficient reimbursement, lack of knowledge, lack of resources or lack of time, these issues are unlikely to be addressed by more research and cannot justify deviating from guidelines; rather, they may require increased educational efforts or policy changes.

It is very difficult to justify deviations from guidelines of this type. For instance, consider a study that tested whether treating people who are very immunosuppressed as if they had TB (without confirmation that they do have TB) could decrease their risk of death. This study planned to enrol people with CD4 counts below 50 cells/mm³, screen them for active TB and exclude those who screen positive, and randomise the remaining people to receive either standard of care monitoring or full TB treatment. As mentioned previously, WHO recommended isoniazid preventative therapy for HIV-positive people regardless of tuberculin skin test status in resource limited settings, which was controversial given that some people not at risk of TB would be receiving a drug with side effects. For this reason and because of cost constraints, WHO recommendations were not being implemented in many places around the world. This study attempted to develop a higher standard of care than what the guidelines required, but doing so required deviating from the guidelines: those in the experimental arm would receive more than WHO required, but those in the standard of care arm would receive less.

Failing to offer the standard of care was difficult to justify because participants with such advanced HIV disease were at significant risk of morbidity and mortality. Additionally, WHO recommendations were clear and beginning to be implemented: facts that made a placebo controlled trial much less relevant. Research comparing an experimental intervention against placebo would not help policymakers who were no longer seriously considering the option of not offering anything. The relevant question was whether to provide what WHO recommended or pick a more aggressive alternative. Therefore, this study would not satisfy the ethical criteria for deviating from the guidelines and was in fact redesigned to offer participants in the control arm isoniazid therapy.