Personalized medicine: Shortcomings and an alternative approach

Personalized medicine has developed in response to the observation of significant variability in disease expression and treatment response across diseased individuals. The terms “personalized” or “precision” medicine are commonly used to refer to the application of pharmacogenomics to tailor treatment decisions based on an individual’s genetic profile. Leveraging our understanding of how specific genetic variants might affect the body’s response to a drug can, in theory, allow treatment choices to be personalized for the individual patient. Significant innovation in pharmacogenetics has been made in several clinical fields, including cardiovascular disease and oncology [Reference Ross, Krebs, Paré and Milani9,Reference Miteva-Marcheva, Ivanov, Dimitrov and Stoyanova10].

Despite major advances over the last two decades in human genome sequencing, mapping of interindividual genetic variations, and identification in gene-drug interactions, this field has yet to produce the widespread benefits and broad transformation of clinical medicine that have been predicted [Reference Joyner and Paneth11]. Beyond genetic factors, variation in treatment response is also influenced by unique physiological and biochemical characteristics as well as environmental exposures. This complex interplay leads to heterogenous treatment effects that generally will not be predictable based on any single clinical characteristic, genetic variation, or biomarker.

Under certain circumstances, this heterogeneity can be addressed in treating the individual patients using the therapy shown to be best for them in an n-of-1 clinical trial. When there is clinical equipoise on which treatment (or whether any treatment) is optimal for the individual patient, an n-of-1 clinical trial may be employed under certain conditions to determine the preferred treatment for that individual.

N-of-1 trials, also called “personalized trials,” employ a repeated crossover design in a single patient to allow direct comparison of treatment effects in the participant. Treatment order is preferably randomized and the repetition of treatment periods improves the ability to distinguish more precise differences in the estimates of treatment effects within the participant. Many design elements, including blinding, protocolized and objective outcome assessment, and washout periods between treatments, may be borrowed from traditional parallel group trials to minimize bias. Bayesian analyses may also be used to identify the probability of benefit, which is not directly assessed from conventional frequentist analyses [Reference Wijeysundera, Austin, Hux, Beattie and Laupacis12]. As interventions are alternated, data are systematically collected on the individual’s response to the interventions, allowing for an unbiased comparison of treatment effects that can directly inform treatment decisions for the individual. When a series of these trials are conducted in similar patients, the results can be aggregated across the participants to produce population-level estimates of treatment effects [Reference Punja, Xu and Schmid13].

The interested reader is referred to various examples of n-of-1 trials, including ours in treating children and adolescents with hypertension and other circumstances beyond assessing treatment effectiveness, modification of personal health behaviors, assessing behavioral interventions, and evaluating treatment harms [Reference Samuel, Bell and Samuels14–Reference Shaffer, Kronish, Falzon, Cheung and Davidson20]. The evidence base on n-of-1 trial methodology has grown rapidly in the last decade and addresses advanced issues in the design [Reference Kravitz and Duan21,Reference Moise, Wood and Cheung22], analysis [Reference Schmid and Yang23], and reporting quality [Reference Shamseer, Sampson and Bukutu24,Reference Li, Gao and Punja25].

Indications for n-of-1 trials

N-of-1 trials are appropriate when there is clinical equipoise, and the comparative effectiveness of the treatment options is in doubt. This occurs commonly in clinical practice, particularly when well-performed comparative effectiveness studies are not feasible (as for rare diseases), or their results have limited generalizability. N-of-1 trials may be particularly important when the treatment risks and benefits are most uncertain, as for populations that are often excluded from large randomized trials, such as vulnerable populations (ex. children and the elderly) or those with complicated comorbidities or rare diseases. For these patients, trial data may not be available, and decisions would otherwise be based on biological plausibility, observational studies, or anecdotal experience.

N-of-1 trials are most worthwhile for chronic conditions, which will require long-term maintenance therapy. Examples of previous use cases include hypertension, attention-deficit hyperactivity disorder, chronic pain, osteoarthritis, and atrial fibrillation [Reference Gabler, Duan, Vohra and Kravitz26,Reference Samuel, Wootton, Holder and Molony27]. A self-limited disease or one with infrequent clinical manifestations would not be appropriate for this study design. Participation in n-of-1 trials has the potential to promote self-efficacy and health literacy, especially when they are used to evaluate health behaviors including physical activity, stress reduction, and dietary changes. Additionally, by directly involving patients in the decision-making process, adherence to the selected treatment may be improved. Examples of various uses of n-of-1 trials are shown in the Table 1.

Table 1. Examples of n-of-1 trials for various use cases

The approach requires a patient who is eager to collaborate in an n-of-1 trial. As most patients will be unfamiliar with the concept, caregivers should be forthcoming about the notion of therapeutic uncertainty, and should review the risks and benefits of rotating between treatment options to repeatedly assess treatment response. Patient perspectives on participation may be influenced by cultural factors, and understanding these may help improve health equity in n-of-1 trials. Some barriers to participation identified in qualitative studies among racial and ethnic minority patients include mistrust in healthcare systems, language barriers, and a reluctance to undergo blinded treatment [Reference Marrast, Conigliaro and Chan28].

This trial design is most feasible when the treatment effects have a rapid onset and offset of action. As in traditional crossover trials, carryover effects must be considered, and are more likely to occur when the outcome is assessed before the effects of the previous treatment have sufficiently diminished. This can be addressed with the use of washout periods between treatments. The treatment periods should be long enough to capture a difference in the outcome, but short enough to be practical for the clinician and the patient. The number of times a given treatment is repeated will affect the precision (or power) in differentiating the relative effects of each treatment. Finally, this approach requires the use of measurable and clinically relevant treatment targets. The measurement chosen to assess treatment effects should not only be valid and reliable but also convenient enough that patients would be willing to undergo the assessment repeatedly.

Improving the usefulness of n-of-1 trials

N-of-1 trials are not immune to many of the same issues that plague some randomized trials and limit their informativeness to clinical practice. If the study designs are susceptible to bias or the results are poorly reported, their usefulness to all stakeholders, including patients, will be hindered. The Consolidated Standards of Reporting Trials extension for n-of-1 trials provides guidance on the preparation and appraisal of reports of n-of-1 trials in the medical literature [Reference Shamseer, Sampson and Bukutu24]. Trial design should be reported explicitly, including the planned number of periods, the duration of each period, whether a run-in or washout period is used, and whether the design was individualized to each participant. The outcome measurement tools should be comprehensively described, including measurement properties such as validity and reliability. If the order of treatments is randomized, the method used to generate the allocation sequence should be described. The statistical analysis should include a description of how carryover effects, period effects, and intra-subject correlation are handled. Changes from the original plan for treatment periods, including the number and sequence of periods completed, should be explained. Comprehensive reporting on the design and execution of n-of-1 trials will improve their applicability to clinicians who are interested in employing these trials to inform the care of their patients.

Increasing uptake

Several barriers to widespread implementation in clinical practice should be addressed, including the doubt among clinicians and patients about whether the costs and burdens of n-of-1 trials outweigh the benefits. Very few randomized trials have been conducted to assess whether n-of-1 trials, when compared to usual care, produce improved clinical outcomes [Reference Samuel, Wootton, Holder and Molony27]. While most studies were equivocal for the primary outcome, several studies identified significant differences in secondary outcomes, including satisfaction with care and increased shared decision-making. However, most of the studies did not reach their predefined sample size targets, and insufficient sample sizes may have reduced their power to identify the benefit of n-of-1 trials.

Digitizing n-of-1 trials

Another major constraint to the widespread application of this approach in routine clinical practice is the lack of tools and infrastructure that would allow a clinician to conduct n-of-1 trials in a busy clinical setting. For this reason, most n-of-1 trial activity has been limited to funded clinical research conducted in academic settings. The momentum to incorporate this approach into everyday clinical practice is usually stalled by loss of research funding or diminished interest among clinicians without a research background. Open-access digital platforms and mobile applications to support the design, conduct, and analysis of n-of-1 trials would reduce logistical barriers and might increase use. Integration of such digital platforms into the electronic health record would facilitate the implementation of n-of-1 trials into clinical practice. Additionally, online platforms would promote sharing of resources and knowledge, allowing protocols to be utilized across multiple institutions and populations. N-of-1 trial apps allow patients to self-track symptoms and may also integrate with other mobile health technologies to allow for passive outcome data collection. Further development of these technologies is critical to advancing the integration of n-of-1 trials into clinical practice [Reference Davidson, Cheung, McGinn and Wang29–Reference Chandereng31].

Ethical and regulatory oversight

Additional considerations in the implementation of n-of-1 trials include the issue of the ethical and regulatory oversight requirements. Currently, there are no national guidelines that address how institutional review boards should approach consent for n-of-1 trials, but some consensus is forming [Reference Cen, Hussain and Pak32,Reference Stunnenberg, Deinum and Nijenhuis33]. As we have previously outlined [Reference Samuel and Wootton34], if certain conditions are met, a less formal consent process could be sufficient. These criteria include 1) a patient’s well-being remains the primary motivation, 2) there is no incremental increase in risk relative to usual care, and 3) the patient understands the process and agrees to participate. Some n-of-1 trials may qualify as quality improvement whereas others may only require a streamlined consent approach (verbal discussion accompanied by an information sheet without the need for signature).