Personalized Trials have the potential to allow investigators to conduct paradigm-shifting comparative effectiveness research at the individual patient-level. The results of individual Personalized Trials can be pooled to determine population-level insights in an efficient manner. Below, you can learn about which conditions and treatments are most suitable for Personalized Trials, and how patient preferences can influence Personalized Trials design decisions.
Personalized Designers
For Clinicians
Benefits*†
- Identify treatments that maximize benefits for individual patients*
- Identify ineffective therapies, thus reducing polypharmacy, minimizing adverse effects, and conserving health care resources.*
- Increase patient engagement in care by helping patients attend to their own outcomes and think critically about treatments
- Awaken patients’ “inner scientist” and give them a greater stake in the process of their clinical care*
- Enhance systematic data collection on the comparative effectiveness of treatments in a pragmatic setting*
- Obtain a better understanding of treatments effects within a population by pooling results across personalized trials
* Kravitz RL, Duan N, editors, 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, M Agency for Healthcare Research and Quality; 2014. http://www.effectivehealthcare.ahrq.gov/N-1-Trials.cfm.
†Kronish IM, Alcantara C, Duer-Hefele J, St Onge T, Davidson KW, Carter EJ, et al. Patients and primary care providers identify opportunities for personalized (N-of-1) trials in the mobile health era. J Clin Epidemiol. 2017 Jun 23 [Epub ahead of print].
What does Personalized Trials mean for research?
Personalized Trials, commonly referred to as N-of-1 trials in the published literature, belong to the family of Single Case Designs that have been widely used in psychology, education, and social work. Personalized Trials is a specific form of randomized or balanced designs characterized by periodic switching from active treatment to placebo or between active treatments (also known as multiple crossover or “withdrawal-reversal” designs). The experiments are designed to determine the relative benefits and harms of the candidate treatments for that one patient. In typical Personalized Trials, an individual is assigned to receive a sequence of treatments, often in a random order. For example, a patient might receive treatment A for 2 weeks, followed by treatment B for 2 weeks, and then treatment A again for 2 weeks, based on a random sequence of treatments A and B. Often, neither the patient nor the clinician knows which treatment that patient is receiving. Throughout the trial, the patient closely monitors the effects of the treatments. At the end of the trial, the treatment “code” is broken, the patient and clinician jointly review the results to see if the patient did better with A or B, and they then decide which one(s) would be best for that patient. By prescribing multiple episodes of treatment, Personalized Trials increases the precision of measurement and control for treatment-by-time interactions, that is, the possibility that the relative effects of two treatments vary over time.
Why should I consider designing Personalized Trials?
The underlying rationale for Personalized Trials is that, in many cases, there is substantial heterogeneity in how treatments affect individual patients, hence, the need to compare treatments within individuals. If conducted in a series of patients, Personalized Trials can be used to gain a better understanding of the extent of heterogeneity of treatment effects. Personalized Trials can also be pooled together to gain an understanding of the group-level effect of treatments. According to some estimates, pooling data across Personalized Trials can be a more efficient methodologic design than a parallel-group randomized trial.
How are Personalized Trials different than conventional parallel group randomized trials?
Parallel group randomized trials compare the effects of treatments (or treatment versus control) across groups of patients. Most conventional clinical trials are conducted by enrolling a large number of patients. Patients are assigned to one of two or more a which compare different treatments or medication for the same condition. In the end, success is measured by comparing how the groups respond to the different treatments, and one obtains an estimate of the expected treatment effect in the average patient enrolled in the trial. Unfortunately, when there are substantial differences in treatment effects between different patients, it is challenging to apply the results of randomized clinical trials to individual patients.
How much of a role do patients play in the design and conduct of Personalized Trials?
Personalized Trials can be conducted with varying levels of patient involvement. For example, in some Personalized Trials, patients can select the types of treatments they wish to compare, the duration of the trial, and the outcomes they would like to assess during the trial. Other Personalized Trials designs are pre-specified in advance such that if a series of patients conduct the Personalized Trial, it will be easier to compare results across patients.
How do I know when it is appropriate to test treatments using Personalized Trials design?
The study design of Personalized Trials is not always appropriate or feasible. For example, if there is already strong evidence supporting the use of a treatment for a given patient, then it would not be ethical to compare this treatment with other treatments as part of Personalized Trials. The Table below highlights the circumstances that are required for Personalized Trials to be appropriate.
Guidelines for N-of-1
GUIDELINES FOR N-OF-1
- State of knowledge
- Nature of disorder
- Nature of treatment
- Outcome assessment
- Willingness of Stakeholders
- Availability of financial resources
N-of-1 appropriate
Clinical uncertainty or equipoise
N-of-1 NOT appropriate
Clear benefit of the treatment
Notes
Blinding ideal, but not always necessary
N-of-1 appropriate
Chronic stable or slowly progressing, Washout period known and safe
N-of-1 NOT appropriate
Rapid progression of symptoms, Possible harm discontinuing active treatment
Notes
Time trends for symptoms should be considered
N-of-1 appropriate
Rapid efficacy, Minimal carryover effects across time, Significant individual differences in treatment response expected
N-of-1 NOT appropriate
Slow efficacy onset, Substantial carryover, Too complex, Small individual differences expected
Notes
Blinding ideal, but not always necessary
N-of-1 appropriate
Valid outcome measures assessed over multiple items
N-of-1 NOT appropriate
Primary outcome assessed at a single time
Notes
Repeated assessments ideal
N-of-1 appropriate
Patients, clinicians, scientist, pharmacist and statistician willing to expend effort
N-of-1 NOT appropriate
Some stakeholders not available/willing
Notes
All design features must be decided a priori to allow for IRB approval and fully informed consent
N-of-1 appropriate
Measurement devices for outcome, cost of compounding drug and multiple visits with clinician can be procured
N-of-1 NOT appropriate
Resources not available
Notes
Time trends for symptoms should be considered
What conditions are suitable for conducting Personalized Trials?
1. Nature of the Problem
- Chronic stable, slowly progressive, or frequently recurring
- Symptomatic conditions (e.g., osteoarthritis) or asymptomatic conditions with biomarkers that can be tracked over time (e.g., hypertension)
2. Nature of the treatment
- Uncertainty about best treatment
- Substantial differences in individual responses to treatment
- Rapid onset
- Rapid and safe washout
3. Outcome Assessment
- Availability of valid, repeatable measures of treatment effects
4. Stakeholders
- Patients, healthcare providers, and health system willing to engage in Personalized Trials
Checklist for designing Personal Trials
Guidance
Key Considerations
- Indications include:
(a) substantial clinical uncertainty; (b) chronic or frequently recurring symptomatic condition; (c) treatment with rapid onset and minimal carryover. - Contraindications include:
(a) rapidly progressive condition; (b) treatment with slow onset or prolonged carryover; (c) patient or clinician insufficiently interested in reducing
Check
Guidance
Key Considerations
- Longer trial duration delivers greater precision, but completion can be difficult or tedious, with the potential for extended exposure to inferior treatment during trial
- Treatment period length should be adjusted to fit the therapeutic half-life (of drug treatments) or treatment onset and duration (of nondrug treatments)
- Simple randomization (e.g., AABABBBA) optimizes blinding (more difficult to guess treatment), while balanced sequencing (e.g., ABBABAAB) is a more reliable guarantor of validity
Check
Guidance
Key Considerations
- Washout is not necessary if treatment duration of action is short relative to treatment period
- Washout is contraindicated if patient could be harmed by cessation of active treatment.
Check
Guidance
Key Considerations
- Blinding is feasible for some drug treatments but infeasible for most nondrug treatments (behavioral, lifestyle)
- Adequate blinding allows investigators to distinguish between specific and nonspecific treatment effects
- In some circumstances, this distinction may not matter to patient and clinician; in others, participants may be primarily interested in the combined treatment effect (specific + nonspecific)
Check
Guidance
Key Considerations
- Patient preferences are preeminent, but clinicians’ goals and external factors should be accounted for and may occasionally supervene
- Valid and reliable measures are preferred when available, but patient-centeredness should not be sacrificed to psychometric imperatives.
Check
Guidance
Key Considerations
- There is a natural tension between identifying a single, primary outcome for decision making and coming to a full understanding of the data
- A reasonable approach is to select one or two primary outcome measures but present or use a variety of statistical and graphical methods to fully explicate the data.
Check
What design features should I consider?
There are a number of design features to consider after concluding that Personalized Trials is appropriate. One must consider the trial duration and the treatment sequence. One must compare the strengths and weaknesses of randomizing the treatment sequence versus choosing a balanced design. One must determine the appropriate allotment of time to allow for onset and washout of treatment effects. One must decide whether to blind or mask treatments and whether to include a placebo or sham option. One must select appropriate tools for measuring treatment effects over the course of the Personalized Trials. These decisions should be based on an understanding of the amount of data collection needed to have sufficient sample size to be able to identify clinically significant differences in treatment effects. Finally, one must determine the appropriate approach for visualizing data with patients and analyzing the data.
What is the best way to report the findings of Personalized Trials?
In 2015, a group of experts in Personalized Trials design developed guidelines for how to report and evaluate Personalized Trials in the peer-review literature. These guidelines are known as the CONSORT extension for N-of-1 trials, or CENT. Systematic reviews have revealed that the reporting of Personalized Trials in the literature is usually inadequate. CENT providers authors and reviewers with a checklist they can follow to assess and report the design of Personalized Trials. The checklist will also be helpful to researchers who are planning the design of their Personalized Trials.
What Personalized Trial design features are most preferred by patients?
As explained above, one needs motivated patients to have successful data collection during Personalized Trials. Thus, understanding the Personalized Trials designs of most interest to patients will be key to success. Accordingly, we conducted a separate national survey of another 500 patients with 2 or more chronic health conditions to learn which Personalized Trials design features were most preferred by patients. We used conjoint analysis to learn about patients’ preferences. The relative preferences for common design features is summarized in the Figure below. Overall, blinding was the least preferred design feature. Surprisingly, an out-of-pocket cost of $100 did not influence the attractiveness of Personalized Trials.
Clinician conducts trial
N-of-1 Service conducts trial
12 week trial
2 week trial
$100 cost
No cost
Clinician chooses Treatment**
Patient Chooses Treatment
Prescription option
Lifestyle option**
3 data points per day**
1 data point per day
30 minutes per day
5 minutes per day**
Blinding
No blinding**
Marginal Utility*
*Marginal utility is a term denoting patient satisfaction or preference** Significant preference (p<0.05) for design feature
Best Case
- No blinding
- No blinding
- Lifestyle option
- Spend 5 min/day monitoring
- Monitor 3 times/day
- Doctor chooses treatment
- + / – Two to Twelve week duration
- + / – Conducted by service/personal doctor
- + / – Cost
Worst Case
- Blinding
- Blinding
- CAM option
- Spend 30 min/day monitoring
- Monitor 1x/day
- Patient chooses treatments
- + / – Two to Twelve week duration
- + / – conducted by service/personal doctor
- + / – Cost
How can patients monitor effects of treatments and/or side effects?
- Traditional approaches include surveys, diaries, medical records, and administrative data
- Recent developments in information technology have opened the door to several new approaches, including ecological momentary assessment (EMA) and remote positional and physiologic monitoring. Mobile-device EMA cues the patient to input data at more frequent intervals (e.g., hourly, daily, or weekly) than is typical using traditional survey modalities.
For more information on designing Personalized Trials,
What health conditions do patients find amenable to Personalized Trials?
Willingness of patients to participate is an essential consideration before designing Personalized Trials. To be successful, Personalized Trials typically require patients to be engaged in data collection over the course of the study. Thus, identifying health conditions for which patients are interested in Personalized Trials is a key first step in deciding to pursue Personalized Trials design. Accordingly, we conducted an online survey of a national sample of 500 patients with 2 or more common chronic health conditions to learn which types of health conditions they were interested in trying Personalized Trials, as well as preferred goals and features of Personalized Trials.
The preferred conditions for Personalized Trial are shown below:
Pain/Back Pain
57.6%
Hypertension
(high blood pressure)
38.8%
Diabetes
28.8%
Trouble sleeping/Insomnia
27.4%
Depression
23%
Hyperlipidemia
(high cholesterol, high triglycerides)
19.4%
Asthma/Emphysema/Chronic Bronchitis
(breathing problems)
14.4%
Arthritis/Joint pain
14%
Obesity
8.2%
Osteoporosis
5%
Headaches
3.6%
Allergies
3.4%
Best Case
- No blinding
- Lifestyle option
- Spend 5 min/day monitoring
- Monitor 3 times/day
- Doctor chooses treatment
- + / – Two to Twelve week duration
- + / – Conducted by service/personal doctor
- + / – Cost
Worst Cast
- Blinding
- CAM option
- Spend 30 min/day monitoring
- Monitor 1x/day
- Patient chooses treatments
- + / – Two to Twelve week duration
- + / – conducted by service/personal doctor
- + / – Cost
What types of treatments are patients most interested in comparing during Personalized Trials?
Based on the results of our national survey, patients are most interested in comparing prescription medications in Personalized Trials.
80-90% moderately to very interested in monitoring
- Disease control (e.g., BP)
- Side effects
- Functional status
- Symptoms
What do patients hope to achieve from Personalized Trials?
Identify best treatment
Improve function/QoL
Improve health condition/sx
Reduce number of pills
Reduce side effects
FAQ
Aren’t Personalized Trials too time consuming for patients?
In the past, adoption of N-of-1 trials in clinical practice was limited by the need for pencil and paper assessments of treatment effects, and in-person visits. Some of the early exuberance for Personalized Trials was dampened when trialists realized that a minority of patients were interested in the effort needed to collect data for Personalized Trials as a means of being more precise about their treatment selection. With the advent of smartphones and wireless devices, the effort needed by patients to conduct personalized trials for use cases in which treatment effects can be monitored using these devices is much lower. Thus, there is currently much enthusiasm for re-launching Personalized Trials in the mobile health era.
How can clinicians and researchers afford the set up costs of initiating Personalized Trials?
In the past, each time someone wanted to do a personalized trial with a patient, it required building much of the infrastructure from the ground-up. While some examples of personalized trials services had some success in engaging a small number of patients, these services were not self-sustaining and dissolved after an initial period of start-up funding. Currently, however, with the advent of mobile apps that can make it more scalable to conduct personalized trials at low cost, once the initial app is built. On-boarding, guiding patients through the personalized trial, tracking of treatment effects, data visualizations and statistical analyses can all be conducted through well-designed apps. Thus, there is hope that in our current mobile health era, Personalized Trials can be more broadly delivered at lower cost.
Can the results of Personalized Trials be used to understand the effects of treatments in a population?
While the primary goal of Personalized Trials is to inform the selection of treatments in an individual patient, the results of Personalized Trials with the same protocol can be pooled together to gain an understanding of the effects of treatment in a population. There are two main approaches to pooling data: a Bayesian approach and a meta-analytic approach. For more information on pooling data from Personalized Trials, please see the following references:
- AHRQ Chapter
Do treatments need to be blinded during Personalized Trials?
In parallel-group RCTs of pharmacologic or device treatments, it is usually recommended to blind or mask the treatments so as to learn the true biological effect of the treatment in a population, distinct from expectancy effects related to other characteristics of the treatment. In Personalized Trials, the goal is often to understand the full treatment effect that the individual patient would experience after the Personalized Trials was over. For this reason, it may be preferable to have the patient compare treatments in an open-label manner as this may be the information of most importance to the patient. Removing the requirement to blind pharmacologic treatments can greatly lower the cost and increase the accessibility of Personalized Trials. In surveys of patients, blinding was one of the most undesirable features of Personalized Trials. Thus, removing the requirement for blinding is also likely to make Personalized Trials more desirable for patients. Of note, Personalized Trials can also be used to compare psychological or behavioral treatments; in such cases, it is not possible to blind the treatments being compared.
When should blinding be prioritized as a design feature of Personalized Trials?
If the goal of the Personalized Trials is to disentangle the biological effect from expectancy effects, then blinding may be necessary. One published example of such Personalized Trials was one in which patients perceived their statin medication to be causing myalgias. Such patients then compared statins to placebo, with both treatments blinded, to learn if the myalgias were truly attributable to the statin medication.
Do Personalized Trials need to be approved by an Institutional Review Board?
There has been a vigorous debate in the field about the need for institutional ethics review of Personalized Trials. The consensus is that if the Personalized Trials is solely conducted to guide treatment selection for an individual patient, then IRB approval should not be necessary, so long as standard approved tools and treatments are being compared and approved treatments are being tested in an ethical manner. If, however, there will be an interest in publishing the results of the Personalized Trials or pooling the findings with the results from other Personalized Trials, then informed consent and IRB approval will be necessary. Some Personalized Trials experts have been excited about creating registries comprised of de-identified data from Personalized Trials. Such registries may facilitate analyses of Personalized Trials with minimal IRB oversight.
Are rigorous statistical analyses that account for autocorrelations of treatment effects across time required to compare treatment effects?
In many published Personalized Trials, comparisons of treatment effects were made qualitatively by comparing data visualizations or using simple statistical tests (e.g., t-tests) to assess whether differences between treatments were statistically significant. Statistical tests that account for autocorrelations between treatment effects across time are recommended for a more rigorous understanding of the significance of differences between treatments. If differences between treatment effects are large and time effects are modest, then more sophisticated statistical techniques may not substantially contribute to decision making. Furthermore, conveying the meaning of such statistical tests to patients can be challenging. Thus, the choice of statistical testing may vary depending on the context of the Personalized Trials. Ideally, statistical experts will be consulted during the design of the Personalized Trials.