This is a chapter from Design and Implementation of N-of-1 Trials: A User's Guide. The full report can be downloaded from the Overview page.
N-of-1 trials are prospectively planned multiple crossover trials conducted in a single individual.1 They can be used to evaluate a wide range of conditions including neurological, behavioral, rheumatologic, pulmonary, and gastrointestinal conditions. They are useful when the patient's symptoms are stable (or frequently occurring), and the treatment takes effect quickly, with few or no residual carryover effects. Further discussion concerning the features and indications for these trials can be found in Chapter 1 (Introduction) of this User's Guide.
Whether n-of-1 trials are a form of systematic learning with the aim of promoting evidence-based clinical care (and therefore a form of "quality improvement") or experiments that aim to produce generalizable knowledge (and therefore "research"), depends on the intent of the trial. In this chapter, we will consider issues that influence whether n-of-1 trials should be treated as clinical care or research. Settling this question is a critical prelude to addressing a number of major ethical concerns and providing guidance for discussions with institutional research ethics boards. We start by considering how n-of-1 trials compare to trials of therapy in routine clinical practice and to traditional randomized controlled trials (RCTs).
N-of-1 Trials Compared to "Trials of Therapy" in Usual Care
“Trials of therapy” or “therapeutic trials” have been utilized extensively in clinical practice to evaluate therapeutic effectiveness in individual patients for a wide variety of therapies, including medication (such as dose determination), device, behavioral and lifestyle therapies, etc. Such informal trials are part of usual care, are unblinded, have no control conditions, and involve no formal validated assessment of effectiveness. As a result, they are vulnerable to bias and uncertainty. Unlike trials of therapy, n-of-1 trials utilize multiple comparisons with a control condition (active or placebo) and a priori decisions about choice and timing of outcome assessment. As such, n-of-1 trials (compared to informal trials of therapy) reduce the risk of drawing invalid conclusions about the effectiveness of a therapy in an individual patient. More specifically, both informal trials of therapy and n-of-1 trials can be utilized in clinical care; however, n-of-1 trials can lead to better therapeutic decisions and outcomes.
N-of-1 Trials Compared to Traditional RCTs
RCTs, the gold standard of clinical research, protect against bias by utilizing blinding, randomization, control conditions, and a priori decisions about outcome measure assessment. While these elements protect internal validity, typical parallel group RCTs have been criticized for their limited external validity and generalizability.2 For example, restrictive inclusion and exclusion criteria may limit RCT enrollment to less than 10 percent of individuals with the disease in question.3 Unlike the usual parallel group RCTs, n-of-1 trials can be tailored to the condition and treatment in question, as well as the outcomes most relevant to the patient. As a result, it has been suggested that the n-of-1 trial design has the potential to provide the strongest evidence for individual treatment decisions and should therefore occupy the pinnacle of the evidence pyramid.4 Furthermore, a series of n-of-1 trials testing the same intervention and conducted in similar patients with identical outcome measures may be pooled for meta-analysis, potentially generating estimates of treatment effect that are relevant for a population.5 Thus, although n-of-1 trials can be utilized as individualized trials of therapy in clinical care settings, the same trial design can also be utilized as a research tool to extend the scope of the usual parallel group RCTs.
N-of-1 Trials: Clinical Care Versus Clinical Research
Differentiating clinical care from research employing experimental therapies can be difficult.6 Quite apart from research, clinical innovation may involve the use of novel therapies or existing therapies for new indications. The application of these therapies is determined by clinical judgment and overseen by all the usual channels for supervising clinical patient care.
Research with experimental therapy can also involve a single individual and is administered by the researcher, preferably in close collaboration with clinical expert(s), and is overseen by institutional research ethics boards. Careful consideration of the features that distinguish clinical innovation from research is needed, especially in the context of chronic disease management (see Table 2–1 for further consideration of the differences between clinical care, n-of-1 trials, and clinical research).
Discourse on the ethics of n-of-1 trials depends on the trial’s intention: research versus learning for clinical care. For example, in research, the goal is to produce a generalizable result; any benefit gained by individual participants is secondary. In clinical care, however, the primary goal is to determine treatment effectiveness for the individual patient. The two activities are fundamentally different in their intent and therefore require different ethical considerations.
More specifically, the U.S. Department of Health and Human Services' Office for Human Research Protections (HHS/OHRP) defines research that is subject to human subjects regulations as follows:
Research means a systematic investigation, including research development, testing and evaluation, designed to develop or contribute to generalizable knowledge. Activities which meet this definition constitute research for purposes of this policy, whether or not they are conducted or supported under a program which is considered research for other purposes. For example, some demonstration and service programs may include research activities.7
Using this definition, n-of-1 trials designed to evaluate therapeutic effectiveness in a single individual are not research (Figure 2–1).
The U.S. HHS/OHRP clarifies the distinction between human subjects research and quality improvement for clinical care as follows:
Protecting human subjects during research activities is critical and has been at the forefront of HHS activities for decades. In addition, HHS is committed to taking every appropriate opportunity to measure and improve the quality of care for patients. These two important goals typically do not intersect, since most quality improvement efforts are not research subject to the HHS protection of human subjects regulations. However, in some cases quality improvement activities are designed to accomplish a research purpose as well as the purpose of improving the quality of care, and in these cases the regulations for the protection of subjects in research (45 CFR part 46) may apply (HHS, 2009).
To determine whether these regulations apply to a particular quality improvement activity, the following questions should be addressed in order:
- Does the activity involve research (45 CFR 46.102(d))?
- Does the research activity involve human subjects (45 CFR 46.102(f))?
- Does the human subjects research qualify for an exemption (45 CFR 46.101(b))?
- Is the nonexempt human subjects research conducted or supported by HHS or otherwise covered by an applicable FWA approved by OHRP?
For those quality improvement activities that are subject to these regulations, the regulations provide great flexibility in how the regulated community can comply. Other laws or regulations may apply to quality improvement activities independent of whether the HHS regulations for the protection of human subjects in research apply (HHS, 2009). (HHS, 2009: https://answers.hhs.gov/ohrp/questions/7281)
Most importantly, the distinction here lies in the primary objective of the n-of-1 trial. If the primary interest is to produce local knowledge to inform treatment decisions for individual patients, n-of-1 trials so conducted should be interpreted as clinical care, and in our view are not subject to the HHS protection of human subjects regulations. Alternatively, if the primary interest is to produce generalizable knowledge to inform treatment decisions for future patients, such n-of-1 trials should be interpreted as human subjects research and required to comply with the standards of such research.
Figure 2–1 illustrates five ways in which an organization can conduct n-of-1 trials and use the resulting data. The specific intentions driving the use of a given model of n-of-1 trials inform whether the use of information gleaned from patients should be considered clinical care or human subjects research. We present three case examples to explore the models in greater depth.
Figure 2–1. Five models of n-of-1 trials
Abbreviations: IRB = Institutional Review Board, P = patients/participants
The parents of an 8-year old girl diagnosed with attention deficit/hyperactivity disorder (ADHD) come into her physician’s office concerned about their child’s sleep problems. The physician is aware that increased sleep onset latency is a major side effect of stimulant medications. Since there is no approved pharmacologic intervention for sleep problems in children, the physician believes a popular natural health product, melatonin, would benefit this patient. The physician decides to evaluate the effectiveness of melatonin in an n-of-1 trial in which the patient undergoes randomly alternating weeks of 3 mg/day melatonin and identical placebo. Neither the physician nor the parents nor the child will be aware of which treatment the child will be on each week. The parents are asked to monitor their child’s sleep and note in a daily sleep diary how long it takes the child to fall asleep over the 6-week period. If the child complains of any side effects throughout the trial, they should notify their physician’s office to determine if she should be seen. After 6 weeks, the physician unblinds the random assignments and graphs the results of the trial using the data recorded in the sleep diary. The physician explains the results of the trial to the parents and the child, and together they decide how to proceed. As is, this case exemplifies Model A. If the physician does the same with other patients and draws a general conclusion about whether to continue this approach, this case is an example of Model B. If a researcher later includes this case in a secondary aggregate analysis, it exemplifies Model C.
Case 2 (an example of Model D)
Inflammatory bowel diseases (IBD) such as Crohn's or ulcerative colitis manifest a set of symptoms that are not always correlated to measures of disease activity for which existing therapies have been developed. Patients often try complementary therapies to manage symptoms such as abdominal bloating, urinary urgency, or nighttime stooling patterns. Little is known beyond anecdotes about the efficacy of therapies such as probiotics, dietary manipulation, and herbal medications. A hospital is interested in evaluating whether the introduction of an n-of-1 trial service to IBD clinicians improves the quality-of-life measures for patients who are trying to manage poorly understood symptoms or trying complementary therapies. A researcher at the hospital designs a two-armed study to measure quality-of-life outcomes alongside primary disease activity measures, randomizing clinicians to a control arm (measurement only) and a treatment arm (n-of-1 trial service). The specific n-of-1 study design is determined by the individual clinician-patient dyads, as in Case 1, but the training, data collection procedures, and analysis are submitted to the institution's review board for approval. A secondary analysis can be done separately to assess whether there is evidence of efficacy of a given complementary therapy that may indicate a more structured, parallel group trial of that therapy for symptom management.
Case 3 (an example of Model E)
Chronic pain is a common condition that has considerable effects on an individual’s quality of life. A group of researchers are trying to determine which type of nonsteroidal anti-inflammatory drug (NSAID) will be most effective for chronic pain in adults with osteoarthritis. They decide to conduct a study in which each participant will be enrolled and offered his/her own n-of-1 trial. Each participant will undergo three pairs of 1-week periods of 3,000 mg/day acetaminophen or 1,200 mg/day ibuprofen, for a total duration of 6 weeks. The order of treatments will be randomized for each participant, according to a computer-generated randomization schedule. Patients, doctors, and research assistants will be blinded to treatment order. Participants will be required to mark their pain on a 10-point visual analog scale daily for 6 weeks. At the end of each 6-week trial, participants will receive their individual results. After all n-of-1 trials have been completed, these data will be aggregated to provide an overall estimate of treatment effect.
N-of-1 Trials as Clinical Care (Model A)
In case example 1, the n-of-1 trial is being used to advance clinical innovation, that is, the patient’s health and well-being are of primary interest. Rather than using a novel therapy, the clinician takes a novel approach to assess therapeutic effectiveness (the n-of-1 trial) rather than the usual trial of therapy undertaken by most clinicians. Although randomization, blinding, and use of placebos are unusual in clinical care, their presence alone does not mean the patient’s interests are not foremost, as these should be in any clinical encounter.
Learning in Clinical Care (Model B)
The results of n-of-1 trials of clinical care are typically stored in electronic medical records managed by the clinical care provider or by the trial service itself. The availability of electronically accessible data provides opportunities for learning from experience in clinical care, also referred to as evidence farming8,9,10 or using an evidence macrosystem.10 Duan8 characterized evidence farming as a “bottom-up” paradigm for clinical practices to incorporate practice data systematically as a source of evidence, or an articulated form of clinical experience. Hay et al.9 reported that most physicians participating in a pilot acceptability study saw evidence farming as a promising way to track experience, making scientific evidence more relevant to their own clinical practices. This learning paradigm is especially pertinent for n-of-1 trials, the design and implementation of which are managed primarily in clinical practices.
Learning in clinical care can occur in many ways, with distinct implications for human subjects procedures. In Model B, learning is focused on utilizing the experience from the operations of earlier trials to inform future trial operations, such as the selection of assessment instruments, the number of treatment periods to be tested, etc. This limited learning paradigm does not directly influence clinical care and therefore should be exempt from IRB review.
Enhanced Learning in Clinical Care (Model C)
In Model C, learning requires the major extra step of outcome analyses using de-identified data aggregated from previous n-of-1 trials to inform clinical care decisions in future trials. Here the individual cases being combined are prospectively planned and often randomized and blinded, making them more rigorous in terms of estimates of treatment effect than standard chart reviews of trials of therapy. Since the results of such analyses are used to make decisions about efficacy and not just operations, it is appropriate to seek institutional ethics approval to do secondary analysis for research purposes. These analyses are appropriate for expedited review, like any chart review.
Study Delivery System + Secondary Analysis (Model D)
The more challenging case to be made to an IRB is when individual n-of-1 trials are used as part of a larger intervention on care delivery, for example, studying the impact of an n-of-1 trial service on a hospital or care network. (This is illustrated in Case 2.) In such cases, the entire n-of-1 trial platform should be subject to full ethics review, but the individual trials would not be subject to ethics review, since they would be developed on a clinical basis. Data produced by these trials can also be used for improving clinical care without review and for secondary analysis or meta-analysis with expedited review. If the introduction of a trial service into care is the sole purpose of giving individual clinicians better tools to care for individual patients, and no larger research agenda is addressed, it may be reasonable to assume that no external IRB review is needed.
Use of N-of-1 Trials To Produce Generalizable Insights (Model E)
Finally, it is increasingly common to use a set of identically designed n-of-1 trials to answer questions typically posed in the context of conventional population-based trials, as in Case 3. Analytical techniques to aggregate the data have been developed to facilitate these kinds of trials (see Chapter 4 for details). In these cases, the entire framework of patient recruitment, trial design, data collection, and analysis should be reviewed by the IRB. The critical distinction here is that the design of individual patient trials is dictated by a larger research agenda. Under these circumstances, the autonomy of individual clinicians and patients is limited to ensure that the aggregation of trial outcomes meets the research design goals.
|Characteristic||Routine Clinical Carea||N-of-1 Clinical Servicea,b,c||N-of-1 Trials Conducted as Researchd,e|
|aCorresponds with Figure 2–1, model A
bCorresponds with Figure 2–1, model B
cCorresponds with Figure 2–1, model C
dCorresponds with Figure 2–1, model D
eCorresponds with Figure 2–1, model E
Abbreviation: IRB = Institutional Review Board.
Intent is to help patient
Intent is to help patient
|Altruism (greater good)
May or may not be helpful to patients
|Goal||Optimal patient care (individualized)||Optimal patient care (individualized)||Generalizable data (population estimates of treatment effect)|
|Population||Based on clinical expertise
|Based on clinical expertise
Recruit (i.e., advertise)
|Yes, n-of-1 approach
is a choice
NB: Secondary analysis will require separate IRB approval
|Yes, participation in research is a choice|
|Intervention (dose, duration, frequency, route)||Individualized||Individualized||Standardized|
|Outcomes||Informal||Formal outcomes (will be part of informed consent)||Formal outcomes (data collection)|
|Publish results||Yes (case reports, series)||Yes (suggest obtain consent a priori)||Yes|
|Cost of product (discussed further in Chapter 3: Financing)||Varies per jurisdiction||Varies; optimally no charge to patient||No charge to patient|
|Oversight||Physician licensing board or regulatory college Ensures standard of care||Physician licensing board or regulatory college oversees standard of care; IRB would be involved for secondary analysis||IRB|
Summary: Role of the IRB Review
Practically speaking, perhaps the most important issue for the implementation of n-of-1 trials is the role of the IRB review. Depending on the primary goal for the trials (research or clinical care), various scenarios are possible:
- No IRB involvement at all, as the n-of-1 trials are conducted for purely clinical purposes. In this instance, the intervention dose, choice of control, and period length would be individualized to meet the needs of the specific patient. In addition, the use of prior n-of-1 outcomes to improve the design and execution of trials would also be exempt from regulations for human subjects research.
- IRB approves platform and procedures of the n-of-1 trials and leaves subsequent treatment selection and design decisionmaking to informed patients and clinicians.
- IRB approves the n-of-1 trials protocol for a specific condition and specific set of treatments (A, B, etc.). This is appropriate for novel therapies and use of n-of-1 in traditional group trials.
- IRB reviews and approves each entry into an n-of-1 trial (case by case). This scenario is likely to be prohibitively costly and time consuming (further diminishing the "value proposition" discussed in Chapter 3 on finance), but undoubtedly some will advocate for this. We believe this approach is inconsistent with how research is defined by the U.S. HHS/OHRP. Furthermore, it would create a level of burden that would preclude the use of n-of-1 trials and act in practice to reduce patient choice, which ethically may be considered a kind of harm.
Informed consent is required from all patients participating in n-of-1 trials. However, the scope of that consent depends on the primary goal of the trial (human subjects research or quality improvement for clinical care).
Equipoise is reached when a rational, informed person has no preference between two (or more) available treatments.11 While equipoise is usually considered in the aggregate for parallel group RCTs, more specific equipoise on the individual level may be warranted for n-of-1 trials, especially for applications of n-of-1 trials to inform treatment decisions for individual patients. A prerequisite for conducting an n-of-1 trial for an individual patient is that there is substantial uncertainty, given the clinical knowledge available regarding the specific patient, regarding the pros and cons for the treatment options under consideration. More specifically, if there is good reason for the clinician to believe that treatment A is superior to treatment B for the specific patient, it might be unethical to conduct an n-of-1 trial for this specific patient to inform his/her treatment decision. At the same time, such a conundrum should not occur if the informed consent adequately presents the knowledge available, informing the patient of the clinical rationale for preferring treatment A over treatment B.
It is of course possible that the patient, even after receiving careful explanation of the clinical knowledge available, might still have a strong preference for treatment B over treatment A, and requests that an n-of-1 trial be conducted to determine whether the a priori clinical knowledge in favor of treatment A indeed applies to him/her specifically. The clinician could honor the patient’s preference in such a situation, as an informed choice by the patient.
Another reason for conducting an n-of-1 trial might be to satisfy a payer regarding treatment effectiveness. In this circumstance, the patient may have a preferred treatment but still be willing to participate in an n-of-1 evaluation so as to gather rigorous data that will allow a payer to be satisfied that the treatment expense is worthwhile.
Although many IRBs might interpret the intention to publish study findings as a criterion for research being subject to human subjects regulations, it is important to note that the U.S. HHS/OHRP does not necessarily hold this interpretation:
Planning to publish an account of a quality improvement project does not necessarily mean that the project fits the definition of research; people seek to publish descriptions of nonresearch activities for a variety of reasons, if they believe others may be interested in learning about those activities. (https://answers.hhs.gov/ohrp/questions/7286)
Therefore, it is conceivable that a publication may be derived from a series of n-of-1 trials conducted for the purpose of quality improvement for clinical care, without necessarily subjecting these trials to requirements, such as informed consent for human subjects research, beyond what is required within the realm of clinical care. Whether designed and conducted as research or clinical care, n-of-1 trials may be suitable for publication. Existing trial registries (e.g., clinicaltrials.gov) are compatible for registration of n-of-1 trials, so as to reduce potential for bias from selective publication. Published n-of-1 trial reports should be CONSORT compliant (see CONSORT Extension for N-of-1 Trials).12
In summary, whether n-of-1 trials represent clinical care, quality improvement, or research depends on their intent. If they are designed to improve the care of an individual patient, it is reasonable that they be considered clinical care. Quality improvement can be applied to n-of-1 trials, just as it is in usual care. However, n-of-1 trials may also be considered research if they were designed to answer a larger question for a population of patients. The following checklist was created to help clinicians, investigators, and institutional research ethics boards determine the most appropriate approach in determining which model of n-of-1 trial to use, and what kind of ethics review and approval is required.
|Clarify intent for n-of-1 trial||Is the primary reason for designing an n-of-1 trial to improve clinical care for a single patient? Or is it intended to help improve care for the population of patients with that condition? If the primary intent is generalizable data, then the n-of-1 trial should be considered research.|
|Select model of n-of-1 trial design||
|Does informed consent need to be obtained?||
|Equipoise||There should not be a clinical preference for or against one of the treatments based on health outcomes; however, there can be a difference in preference based on cost or convenience.|
- Guyatt G, Heyting A, Jaeschke R, et al. N-of-1 randomized trials for investigating new drugs. Control Clin Trials. 1990;11:88-100.
- Rothwell PM. External validity of randomized controlled trials: “to whom do the results of this trial apply?” Lancet. 2005;365(9453):82-93.
- Larson EB. N-of-1 clinical trials- a technique for improving medical therapeutics [Specialty Conference]. West J Med. 1990;152:52-56.
- Guyatt G, Jaeschke R, McGinn T. Therapy and Validity: N-of-1 Randomized Controlled Trials. Users' Guides to the Medical Literature: A Manual for Evidence-Based Clinical Practice. Chicago, IL: American Medical Association; 2002: 275-290
- Zucker DR, Ruthazer R, Schmid CH. Individual (n-of-1) trials can be combined to give population comparative treatment effect estimates: methodological considerations. J Clin Epidemiol. 2010; 63(12):1312-1323
- Kass NE, Faden RR, Goodman SN, et al. “The research-treatment distinction: A problematic approach for determining which activities should have ethical oversight,” Ethical Oversight of Learning Health Care systems, Hastings Center Report Special Report 43, no.1 (2013): S4-S15. DOI: 10.1002/hast.133.
- U.S. Department of Health and Human Services. (2009). Human Subjects Research (45 CRF 46). Washington, DC. https://www.hhs.gov/ohrp/policy/ohrpregulations.pdf.
- Duan N. A Quest for evidence beyond evidence-based medicine: Unleashing clinical experience through evidence framing. Presented at UC Davis School of Medicine, October 17, 2002, Sacramento, CA.
- Hay MC, Weisner TS, Subramanian S, et al. Harnessing experience: Exploring the gap between evidence-based medicine and clinical practice. J Eval Clin Pract. 2008;14(5):707-713.
- Sim L. Evidence framing: Implications for open architecture. Presented at Mobile Health, Stanford University, California, May 5, 2011. Slides available online at https://www.slideshare.net/openmhealth/evidence-farming-implications-for-open-architecture.
- Lilford RJ, Jackson J. Equipoise and the ethics of randomization. J R Soc Med. 1995;88(10):552-559.
- Shamseer L, Sampson M, Bukutu C, et al. P05.50 CONSORT extension for N-of-1 trials (CENT) guidelines. BMC Complement Altern Med. 2012;12(Suppl 1):410.
Punja S, Eslick I, Duan N, Vohra S, the DEcIDE Methods Center N-of-1 Guidance Panel. An Ethical Framework for N-of-1 Trials: Clinical Care, Quality Improvement, or Human Subjects Research? In: Kravitz RL, Duan N, eds, and the DEcIDE Methods Center N-of-1 Guidance Panel (Duan N, Eslick I, Gabler NB, Kaplan HC, Kravitz RL, Larson EB, Pace WD, Schmid CH, Sim I, Vohra S). Design and Implementation of N-of-1 Trials: A User’s Guide. AHRQ Publication No. 13(14)-EHC122-EF. Rockville, MD: Agency for Healthcare Research and Quality; January 2014: Chapter 2, pp. 13-22.