Introduction

The use of n-of-1 trials to improve therapeutic decisionmaking and clinical outcomes has been studied and reported upon for over 25 years.¹ Selected reports suggest successful resolution of therapeutic uncertainty in specific patients when the underlying condition and drugs are amenable to the n-of-1 approach: specifically, chronic conditions that do not change rapidly over time, with noncurative interventions, clear symptoms that can be tracked, and treatment effects that wash out relatively rapidly (see Chapter 1 of this User's Guide). In a time of increasing interest in personalized medicine, the n-of-1 trial presents a theoretically feasible and cost-effective method of determining the best therapeutic option for a particular person.^2-5 As patients and clinicians recognize that an individual's response to a medication may not be well represented by a population mean, the use of n-of-1 trials to distinguish true responses would seem logical. Nonetheless, after more than 25 years of sporadic reports on n-of-1 trials, largely from academic settings, to our knowledge the service is not generally available to patients and doctors anywhere in the world. This chapter will explore what is understood about costs, benefits, and possible financing of n-of-1 trials based on the literature and the authors' (WDP, EBL) experience.

Although health care providers have access to an array of tools that lend a high degree of confidence to diagnoses, few if any widely available tools help providers determine which medication (or behavioral health treatment) is best for a specific patient. Providers rely on several imperfect strategies for therapeutic decisionmaking. First, they interpret the evidence from randomized controlled trials, which present the average benefits and risks of a particular drug. Such evidence sometimes requires clinicians to find and interpret a large number of studies, then assess the extent to which their patient resembles or differs from the narrow population that qualified for inclusion in the study⁶ and the degree to which the benefits and risks of the drug matter to that patient.⁷ Second, clinicians may adopt a "trial of therapy" approach, in which they start a patient on a drug and wait to see how it works. The biases and potential problems of this approach have been well described.^7-9 At times, clinicians may simply give a patient two or more drugs in a similar class to take home and try at the patient's convenience (essentially an open-label n-of-1 trial without any control for washout periods, placebo effect, or numbers of crossovers required for clinical decisionmaking). The therapeutic decisions that result from these methods are imperfect at best, and at worst may lead to unnecessary costs and higher than necessary rates of adverse effects.

Beyond improving initial therapeutic decisionmaking, an n-of-1 trial has a number of other potential longer term benefits. In theory, the risk-benefit ratio of a drug would be improved because only medications with demonstrated effectiveness for a particular patient (as shown through an n-of-1 trial) would be prescribed. In addition, short-term side effects are typically clearly demonstrated in n-of-1 trials. Long-term adverse events, of course, are not immediately known and would need to be factored into a risk-benefit model using population-based data. Current population-based information from randomized controlled trials may make it difficult to extrapolate the full benefit of medications in a heterogeneous population. Subgroup analyses can help overcome some of these issues, but studies are often not large enough for these subanalyses, nor are data typically available at the patient level across studies to allow others to examine the heterogeneity issue. When they are feasible, n-of-1 trials eliminate concerns about population-based heterogeneity of responses.

N-of-1 Methods Not Yet Part of Routine Care

Despite their many potential benefits, n-of-1 trials have not become part of mainstream clinical medicine,¹⁰ and to our knowledge have never been a covered benefit in any insurance plan (private or government run) in the United States or Canada. A 2010 systematic review found 108 unique trial protocols from the years 1986 to 2010; the vast majority had authors from Canada (35%), Europe (26%), or the United States (22%).¹¹ N-of-1 trial services have been run almost exclusively by academic centers with little reach into community practice in the United States; somewhat broader reach has been achieved in Australia.¹² As academically run services, most have been supported by grants and local institutional funds. A systematic review by Gabler et al. found that most trials (69%) reported receiving Institutional Review Board (IRB) approval, and more than half (52%) received external funding.¹¹ No articles reported charging patients or insurance companies for the service. The involvement of IRBs in the majority of trial activities indicates the low acceptance of these activities as a component of routine clinical care. A more complete discussion of the role of IRBs in n-of-1 trials is presented in Chapter 2. Considering just the impact on financing, the involvement of IRB review for many services (even if the final decision is that n-of-1 trials are not “research”) highlights the “experimental” nature of the process and makes insurance coverage less likely, as insurance companies rarely pay for research activities.

Cost Data for N-of-1 Trials

While there is no literature on third-party or patient payments for n-of-1 trials, three studies have explored the costs of conducting trials (see Table 3–1). The reported costs vary widely, partly perhaps because of differences in costs between countries (one article from the United States, one from Canada, and one from Australia) as well as inflationary differences (1993 U.S .dollars versus 2008 Canadian dollars, for instance). Beyond these variables, different approaches have been advocated for conducting n-of-1 trials. Many trial centers develop new trial instruments for each patient, based on the patient’s stated preference or importance of one symptom or sign over another. Others report on multiple trials based around a single clinical question, using a standardized set of assessment instruments. Some trials provide feedback to the referring physician, who is then expected to develop a treatment plan with the patient based on the trial results. Other trials include the final treatment decision discussion in the trial itself. These and other differences in approaches would be expected to affect the overall cost of a single trial. In this report, we do not attempt to standardize costs to a particular reference point but simply express costs as reported in the papers we found.

Scuffham et al. evaluated the detailed costs of two multipatient n-of-1 trial series conducted by the University of Queensland.³ Using classic economic approaches, they initially divided costs into fixed startup costs and variable per-patient or per-trial costs. The costs were considered within the context of a “research” activity conducting two sets of n-of-1 trials using the same medications, the same outcome and side effect instruments, and the same patient problems within each set. The research approach clearly affected the costs incurred and may have also determined whether some costs were considered fixed or variable. The following items were considered fixed costs:

• Seeking funding
• Developing the research protocol
• Obtaining ethical review
• Developing instruments/forms for data collection
• Developing treatment sequencing
• Blinding medications
• Design/preparation of medication packs
• Database development

Variable costs were categorized as follows:

• Patient recruitment
• Managing the operation of each trial
• Data collection
• Data analysis
• Generation of results and feedback to clinicians and/or patients

The Queensland trial service found a total fixed cost of $23,280 Australian (2005) to set up two different n-of-1 trial protocols. Various components of these costs would not be applicable when operating n-of-1 trials primarily for clinical purposes. For example, the cost analysis included as “fixed costs” the applications for grant funding and ethical approval, which accounted for $7,730, or 33 percent of the total startup costs. While patient recruitment could also be considered a “research” expense, one would imagine that a commercially available n-of-1 trial system could incur major costs marketing the services to clinicians or patients, which would likely markedly exceed the relatively low “recruitment costs” assigned to this analysis. The cost of preparing medications is listed as a fixed cost, though if medication acquisition costs were included and a broad set of medications were included for potential n-of-1 trials, this would more logically be a variable cost. The investigators considered the developed protocols to be reasonably applied to 200 people, with resultant fixed costs per patient of $116. Variable costs were $610 for a trial of celecoxib versus long-acting acetaminophen and $577 for a trial of gabapentin versus placebo. The overall cost per trial based on this study is in line with many other diagnostic tests. However, this trial did not include costs related to the development of an electronic data collection system, which would be essential for any present-day commercial or clinically based system in the United States or Canada. Even though development of such a system could run into the hundreds of thousands of dollars (U.S.), if the system were used for enough trials, the overall cost per patient could still be kept in line with complex diagnostic tests such as advanced imaging modalities.

N-of-1 trials performed outside of a research study can provide further insight into the costs of the method. One of this chapter’s authors (EBL) worked with colleagues to explore the costs of operating an n-of-1 trial service in an academic institution. This service was operated for clinical purposes and therefore did not “recruit” patients as a research protocol would. After initial interactions with the local IRB, the service was declared to be a component of clinical care, therefore not requiring IRB review of each new trial protocol.

Larson’s group designed each single-patient trial in their series individually.⁵ Their cost assessment then focused on assessing the direct costs of operating a single trial. They estimated 16.75 hours of staff time per trial, which included a physician lead, nursing, data entry, analysis, and feedback time. Of note, none of the staff were solely devoted to work on trials but charged time to the n-of-1 trial service alongside other job tasks. In 1990 U.S. dollars this was estimated at approximately $500/study plus the cost of the medications. In 2013 dollars, just the staff time would likely rise to between $1,500 and $2,000.

Additional experience comes from a commercial application of the n-of-1 model. One of the authors (WDP) worked as an independent evaluator for a commercial venture that sought to bring n-of-1 trials to clinicians in a much more automated form. The group’s systems were tested initially with two treatment periods (medication 1 then medication 2, or vice versa) over three treatment cycles.^13,14 This approach was adjusted to five treatment cycles, generally running 5 to 7 days per treatment period, depending on the medication being studied. The group developed a Web-based data collection system and used a validated set of symptom and side-effect questionnaires for the drugs they offered for study. They offered all three primary types of n-of-1 trials: active versus placebo, active drug A versus active drug B, and dose A versus dose B of the same drug. The system allowed clinicians to simply write a prescription for the study of interest from a predetermined set of medications. The company contacted the patient and established a secured Web account. A contracted pharmacy prepared the medication unit dose packs with over-encapsulation to achieve patient blinding. The initial medications available for study were H2 blockers, proton pump inhibitors, and antihistamines. The underlying study design was set by default as requiring five treatment cycles (i.e., AB or BA, where both A and B represent either study medication 1 or study medication 2), with the ordering of treatment periods within each cycle established by random assignment. The number of days per treatment period was determined by the longest half-life of the medications under study, allowing for adequate time to assess symptoms and side effects after a washout period for each medication. If two active comparator drugs were used, patients were crossed over from one active drug to another, without a placebo washout period. To account for the lingering effects of the previous active medication, patient data gathered during a predetermined washout period were ignored in the analysis. A randomly selected crossover pattern was sent to the pharmacy, which prepared the medications for each participant. Analytics were built into the database as a report feature. Clinicians could receive reports as a hard copy or log in to the Web site for the information, including which medication improved symptoms the best, which had lower side-effect profiles, and whether there was a clinically meaningful effect versus placebo.

Unfortunately, the evaluation of the system was stopped early due to financial problems. Prior to that, a total of 64 patients were enrolled; 34 were enrolled in one of two n-of-1 drug trials comparing two active medications using the same data collection system, but only three patients completed full evaluations. This unwelcome experience in the clinical setting differed sharply from the initial, shorter testing, which had very high completion rates.¹³ Qualitative feedback indicated that patients did not see enough value in the added certainty provided by the trial results, given that they needed to complete daily logs on symptoms and adverse events for approximately 2 months. Patients indicated they could easily conduct their own open-label trials quickly and inexpensively to determine which medication worked best for them. This finding may have been influenced by the fact that all the medications being studied became available over the counter by the time the evaluation was underway. Interestingly, the side-effect rates (which study data showed were clearly caused by the medication in question, based on study completers or partial completers) were much higher than reported in the literature or package inserts for the medications, approaching 30 percent of proton pump inhibitor users, for instance. This experience is consistent with reports of new or more common significant side effects when drugs are approved by the Food and Drug Administration (FDA) and used in the more general population compared with the highly selected persons typically enrolled in studies meeting FDA efficacy standards.¹⁵

Cost Offset

To examine cost offsets, Scuffham et al. examined data from two separate multipatient n-of-1 studies conducted in Queensland, Australia: one study compared cox-2 inhibitors versus acetaminophen for osteoarthritis, and the other compared gabapentin versus placebo for neuropathic pain.³ They constructed a decision analysis model with two arms: "n-of-1 trial" and "no trial." In both studies, the n-of-1 arms ended up costing more per patient than the "no trial" groups, even taking into account savings from individuals who were able to stop taking ineffective medications. After estimated average per-patient cost offsets of Australian $569 for the gabapentin trial and Australian $221 for the celecoxib trial, the final estimated 5-year additional costs of n-of-1 trial versus no trial for these medications were Australian $869 for gabapentin and Australian $1,152 for celecoxib. This finding could be due to the small differences in outcomes between the n-of-1 and no-trial groups and the low responder rates for both active medications: 17 percent for celecoxib and 24 percent for gabapentin in the n-of-1 trial groups. Both groups demonstrated small improvements in quality of life for the n-of-1 trial participants, resulting in a cost per quality-adjusted life year (QALY) gained in the first 12 months of Australian $36,958 for the gabapentin group and Australian $126,661 for the celecoxib group. If therapy was maintained until the end of life, the cost per QALY gained dropped to Australian $1,725 and Australian $10,278, respectively. The variables most responsible for cost differentials were calculated. These included the underlying variable costs of conducting n-of-1 trials, the number of individuals among whom the fixed costs are shared, the probability that the n-of-1 trial will result in use of the study medication, the time horizon for which the results are valid, and the cost differential of the medications being studied. The longer the patients in this report were credited with taking medication of no value or causing undesirable side effects, the more value would issue from an n-of-1 trial, implying greater cost effectiveness. The paper examined time horizons of 5 years and lifetime, though other studies have used time horizons of less than 1 year following n-of-1 trials.^4,16 The model also indicates that the greater the effect differences between two medications or medication and placebo, the greater the cost efficacy of n-of-1 trials. This analysis used an imputed usual-care group and thus may not entirely capture the impact of an n-of-1 trial at the patient level if a higher percent of people remain on an ineffective drug than imputed.

In examining other reports of multipatient n-of-1 trials (i.e., series of n-of-1 trials entering multiple patients into the same n-of-1 protocol), it is evident why cost offsets can be hard to demonstrate. In the Queensland trials the pain difference for the n-of-1 trial participants versus no-trial group at the end of the celecoxib trial was 0.28 points on a 10-point scale, while the gabapentin trial demonstrated a 0.11-point drop in pain compared to the no-trial group.^17,18 Similarly, in a study of theophylline in patients with chronic obstructive pulmonary disease (COPD), Mahon et al. found that in 68 patients randomized to an n-of-1 trial versus usual care, 7 of the 34 n-of-1 trial patients benefited from theophylline, while 11 elected to continue theophylline at 3 months (35%).¹⁶ By the end of the trial at 12 months, 16 of 34 n-of-1 participants were using theophylline (47%). In the usual-care group, where theophylline effectiveness was determined through open-label on-off usage, 13 of 30 (43%) were using theophylline at 3 months and 15 (50%) were using theophylline at 12 months. Furthermore, there was no difference across study populations (responders and nonresponders in both groups included in the intent-to-treat analysis) in chronic respiratory disease questionnaire scores or 6-minute walk times.

In a study of the use of nonsteroidal anti-inflammatory medications (NSAIDs) in osteoarthritis, Pope et al. found no differences in use of NSAIDs between n-of-1 trial participants and usual-care participants (81% n-of-1 vs. 79% usual care).⁴ This relatively small trial (N = 51) found no significant differences in an overall health assessment scale, osteoarthritis pain and function scale, or SF-36 scores between the two groups. The total costs of care (osteoarthritis treatment), including the n-of-1 trial, at 6 months was $551.66 +/- $154.02 for the n-of-1 trial versus $395.62 +/- $226.87 for the usual-care group (2003 Canadian dollars). Since n-of-1 trials, even if taken to scale, will always cost more than open-label clinical trials, it will require demonstrations of greater effect from the trials themselves to demonstrate reasonable cost offsets.

Value Proposition

In general our review concludes that it is difficult to demonstrate a value proposition for n-of-1 trials based on the current literature. Trials reported to date have found limited differences in outcomes between n-of-1 participants and usual care, a tendency of both groups to end up with similar medication usage patterns over time, and small sample sizes. Kravitz et al.¹⁰ have postulated the potential for greater value where treatment costs are higher, such as with biological agents. Furthermore, where risk-benefit equations are very different between various treatments (e.g., low-dose methotrexate vs. biological agents for rheumatoid arthritis), demonstrating clear benefits to higher risk medications may improve the overall value proposition as the population of medication users is enlarged and serious side effects from high-risk medications appear. These issues are not considered in any of the current literature which directly examines costs of n-of-1 trials; given the small sample sizes and short followup timeframes, major side effects from medications were not encountered. A more general issue is that chronic disease effects and available treatments change over time. These changes may lower the enduring value of the results of an individual treatment, given changes in symptom patterns or a patient’s preference or physician recommendation based on availability of different treatments. For n-of-1 trials to be valuable in the face of seemingly inevitable changes, the methods would need to be relatively straightforward and efficient and meet patients’ timeliness expectations.

Karnon et al. have explored the use of n-of-1 trials to study the economic impact of various medication choices at the individual patient level.¹⁹ The authors consider adding questions related to total cost of care, cost of alternative medications used, and/or quality of life to better understand the cost/benefit of various medication choices. The paper considers the ethical issues of basing decisions on overall improvement versus the cost per unit of improvement. It concludes that clear patient preferences should drive clinical decisions and that economic considerations should come into play only when the clinical decision is ambiguous. The use of a series of n-of-1 trials with additional data collection could help researchers more precisely understand the economic and quality-of-life impact of various medication choices in responders. This rationale could also arguably be applied to diagnostic tests, which are typically adopted and paid for without a clear demonstration of a value proposition other than improved diagnostic accuracy.

Influence of Personalized Medicine

Personalized medicine is an area in which n-of-1 trials may help us study outcomes for commonly prescribed drugs. With growing concern about the overall safety and risk-benefit profile of many medications, n-of-1 trials could be used to personalize this information. N-of-1 trials seem particularly well suited to understanding side effects associated with a medication at the personal level. Could this drive interest in the method, if it were better understood? Similarly, n-of-1 trials are well suited to study herbal preparations, dietary supplements, and behavioral treatments (including lifestyle, behavioral, and complementary/alternative interventions, as discussed in Chapter 2 of this User’s Guide). There are many “natural” supplements available for a wide variety of conditions, most of which will never be submitted to rigorous population-level randomized controlled trials. Through crowd-sourcing, could a subgroup of individuals interested in trying supplements form a grassroots user group interested in the therapeutic precision of n-of-1 trials? The Patient-Centered Outcomes Research Institute²⁰ is developing Patient Powered Research Networks that could form a basis for a patient-centered n-of-1 trial network.

As we move toward personalized medicine based on genomic or proteomic data, combining n-of-1 trials for appropriate conditions and medications may be the one rational way to study outcomes associated with commonly prescribed drugs for both individuals (personalized medicine) and general populations. We can assume that the attractiveness of personalized medicine will grow, and as science-based personalized medicine disseminates, n-of-1 trials seem elegantly suited to become a regular part of personalized medicine.

Potential Financing Options

We have identified a number of potential ways in which the n-of-1 trial could be paid for. It is conceivable that large pharmacy chains could take on the conduct of n-of-1 trials. Most of these companies already have a strong Internet and mobile presence, the ability to prepare the medications for trials, and established financial relationships with payers.

If n-of-1 trials demonstrated positive financial offsets for selected medications, or greater levels of patient satisfaction and improved outcomes with low marginal costs, would Accountable Care Organizations (ACOs) consider contracting with commercial vendors or pharmacy chains for the service for selected medications? It is conceivable that with ACOs and cost bundling, n-of-1 trials would have a value proposition as a strategy to manage expenses while reducing side effects and adverse effects of drugs, especially for commonly used or expensive drugs. The Centers for Medicare & Medicaid Services (CMS) or the new Innovation Center within CMS could be a source for funding that would help elucidate the impact of n-of-1 trials taken to scale in usual clinical care. N-of-1 trials could be considered cognitive services, which do not involve a capital investment in a machine, device, provision, or procedure and thus have little potential in a fee-for-service world to cover implementation or facility costs.

No self-interested group has yet been inclined to develop a business case for n-of-1 trials. If anything, pharmaceutical companies have previously had a disincentive in a fee-for-service world to consider n-of-1 trials, since they typically reduce overtreatment and highlight side effects. In an ACO world, n-of-1 trials could be part of a risk-mitigation strategy to reduce overtreatment and medication side effects. Given the precision of information on short-term side effects developed through n-of-1 trials and the current FDA priority to find better ways of detecting adverse effects post marketing, the FDA might consider whether developing an infrastructure for an n-of-1 enterprise might be a worthwhile way to improve assessment of medications for symptomatic treatment of chronic diseases. If a trial registry were available that contained both standardized methods and outcome assessment toolkits as well as a repository for trial results, data derived from potentially thousands of individually conducted n-of-1 trials could be an added source of information to assess benefits and risks of drugs. Patients in a clinical trial registry would likely represent a broader population in regard to age and secondary morbidities than typically seen in phase 3 randomized controlled efficacy trials, allowing a better understanding of the impact of medications in everyday practice through secondary analysis of the pooled results. We believe this idea is worth exploring, especially for medications that are likely to be taken long term for chronic conditions.

Innovations That May Increase the Appeal of the N-of-1 Trials

Several innovations could increase the reach and appeal of n-of-1 trials. Interactive technology (discussed in Chapter 5 of this User’s Guide, which covers information technology) could incorporate patient preferences for the most important outcomes to them (a potentially variable cost) while still maintaining a “standard” data collection format. Validated instruments, for both outcomes and adverse effects, could be built into the data collection system, with patients indicating the most important personal outcome as well as the side effects they consider least tolerable or most troublesome. While the initial costs of development could be substantial, the per-trial cost could still be reasonable if amortized over thousands of patients. However, overall usage of n-of-1 trials would need to expand greatly for this model to be cost effective.

As mentioned, a national n-of-1 trial registry could improve shared decisionmaking based on an individual n-of-1 trial over time. Such a registry would store and analyze the combined results of n-of-1 trials using standardized processes and include well-developed assessment methods and outcome scales for appropriate medications, particularly those with narrow therapeutic windows, moderate population-level efficacy, or high cost-to-benefit ratios. This advance would increase the reach and use of n-of-1 trials greatly (but would likely require substantial ongoing support).

Conclusion

The long-term financing of n-of-1 trials will be determined by a value proposition that is more attractive to patients, clinicians, and other providers, including perhaps pharmaceutical companies, payers, and possibly regulators. Presently the limited use of n-of-1 trials may reflect that the value proposition for clinicians and patients lies with the rapid acquisition of data to guide diagnosis and treatment. N-of-1 trials, with their prolonged timeframe, are relatively unattractive compared to other clinical activities that produce rapid results, even though n-of 1 trials could fundamentally change the way that medicine is practiced. Can n-of-1 trials become more standardized, more efficient, and more patient and physician friendly? Most importantly, can they be moved from the rarefied world of the academic medical center and faculty with keen interests in clinical epidemiology and research to the everyday world of clinical practice and the rapidly changing world of consolidating delivery systems?

Larger scale, more efficient services aimed at enhancing patient-centered outcomes through more precise therapeutics could be a way to demonstrate value. The outcomes of greatest interest would be improved effectiveness of treatment, reduced side effects, and improved patient and physician satisfaction, along with reduced or improved management of costs through avoidance of adverse events and an ability to use less expensive drugs of proven effectiveness for individual patients. For n-of-1 trials to reach a broader audience, it will be important to develop methods that reduce patient reporting burdens. The use of small, Internet-connected personal devices should make this a possibility.

As with diagnostic interventions, an understanding of the characteristics of the intervention is important in determining when it will benefit patients and when it is contraindicated. For diagnostic interventions, these characteristics include specificity, sensitivity, prior probabilities, and positive and negative predictive values. For n-of-1 trials, a better understanding of the impact of different characteristics of the treatment differentials would help advance the concept. For instance, what are the impacts of different probabilities of a positive response to treatment on the utility of an n-of-1 trial? At what level of population response is an n-of-1 trial no longer indicated? What are the impacts of various levels of cost differentials of the final treatments on the potential benefits of an n-of-1 trial? Clinicians need information that will help them understand where n-of-1 trials would be of greatest value.

Overall, we conclude that the limited data currently available suggest that n-of-1 trials can be conducted for a reasonable per-patient cost (not considering the cost of the drug or drugs to be tested) and that these costs could be further lowered with modern technology such as interactive data collection systems. Furthermore, modern technology should be able to blend standardized data collection instruments with patient preference and modern testing theory to reduce data collection from nonuseful questions for a particular patient. The value proposition, from both the financial and patient outcome perspectives, is where the most uncertainty exists at present. Until this value proposition is better defined, it is unlikely that commercial payers will include coverage for n-of-1 trial activities.

Checklist

References

1. Mahon J, Laupacis A, Donner A, et al. Randomised study of n of 1 trials versus standard practice. BMJ. 1996 Apr 27. 1996;312(7038):1069-1074.
2. Kravitz RL, Duan N, White RH. N-of-1 trials of expensive biological therapies: a third way? Arch Intern Med. May 26 2008;168(10):1030-1033.
3. Scuffham PA, Yelland MJ, Nikles J, et al. Are N-of-1 trials an economically viable option to improve access to selected high cost medications? The Australian experience. Value Health. Jan-Feb 2008;11(1):97-109.
4. Pope JE, Prashker M, Anderson J. The efficacy and cost effectiveness of N of 1 studies with diclofenac compared to standard treatment with nonsteroidal antiinflammatory drugs in osteoarthritis. J Rheumatol. Jan 2004;31(1):140-149.
5. Larson EB, Ellsworth AJ, Oas J. Randomized clinical trials in single patients during a 2-year period. JAMA. 1993 Dec 8. 1993;270(22):2708-2712.
6. Rothwell PM. External validity of randomised controlled trials: "to whom do the results of this trial apply?" Lancet. Jan 1-7 2005;365(9453):82-93.
7. Ross JM. Commentary on applying the results of trials and systematic reviews to individual patients. ACP Journal Club. Nov-Dec 1998;129(3):A17.
8. Jaeschke R, Guyatt GH, Sackett DL. Users' guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? The Evidence-Based Medicine Working Group. JAMA. 1994 1994;271(9):703-707.
9. Jaeschke R, Guyatt G, Sackett DL. Users' guides to the medical literature. III. How to use an article about a diagnostic test. A. Are the results of the study valid? Evidence-Based Medicine Working Group. JAMA. 1994;271(5):389-391.
10. Kravitz RL, Duan N, Niedzinski EJ, et al. What ever happened to n-of-1 trials? Insiders' perspectives and a look to the future. Milbank Q. 2008;86(4):533-555.
11. Gabler NB, Duan N, Vohra S, et al. N-of-1 trials in the medical literature: a systematic review. Med Care. Aug 2011;49(8):761-768.
12. Nikles CJ, Mitchell GK, Del Mar CB, et al. An n-of-1 trial service in clinical practice: Testing the effectiveness of stimulants for attention-deficit/hyperactivity disorder. Pediatrics. 2006 Jun;117(6):2040-2046.
13. Reitberg DP, Del Rio E, Weiss SL, et al. Single-patient drug trial methodology for allergic rhinitis. Ann Pharmacother. 2002 Sep. 2002;36(9):1366-1374.
14. Wolfe B, Del Rio E, Weiss SL, et al. Validation of a single-patient drug trial methodology for personalized management of gastroesophageal reflux disease. J Manag Care Pharm. Nov-Dec 2002;8(6):459-468.
15. Greener M. Drug safety on trial. Last year's withdrawal of the anti-arthritis drug Vioxx triggered a debate about how to better monitor drug safety even after approval. EMBO Rep. Mar 2005;6(3):202-204.
16. Mahon JL, Laupacis A, Hodder RV, et al. Theophylline for irreversible chronic airflow limitation: a randomized study comparing n of 1 trials to standard practice. Chest. Jan 1999;115(1):38-48.
17. Yelland MJ, Nikles CJ, McNairn N, et al. Celecoxib compared with sustained-release paracetamol for osteoarthritis: a series of n-of-1 trials. Rheumatology. Jan 2007;46(1):135-140.
18. Yelland MJ, Nikles CJ, McNairn N. Single patient trials of gabapentin for chronic neuropathic pain. World Congress on Pain; 2005; Sydney, Australia.
19. Karnon J, Qizilbash N. Economic evaluation alongside n-of-1 trials: getting closer to the margin. Health Econ. Jan 2001;10(1):79-82.
20. Patient-Centered Outcomes Research Institute. https://www.pcori.org/. Accessed July 8, 2013.

Citation

Pace WD, Larson EB, Staton EW, the DEcIDE Methods Center N-of-1 Guidance Panel. Financing and Economics of Conducting N-of-1 Trials. 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 3, pp. 23-32.