n-of-1 trials: increasing precision in therapeutics
ACP J Club. 1993 July-Aug;119:A16. doi:10.7326/ACPJC-1993-119-1-A16
Lewis Thomas, in his 1983 autobiography, The Youngest Science: Notes of a Medicine Watcher, describes how medical practice has changed in the 20th century. As a physician's son, he learned that his doctor-father's job was to diagnose, to stand by on call, and, most important, “to explain what had happened and was likely to happen” (1). As a student, Thomas learned the art of diagnosis, and, as an intern, he experienced that “medicine made little or no difference” (1).
Internal medicine has traditionally revered diagnosis and great diagnosticians like William Osler. In this tradition, we have emphasized diagnostic precision—often going to great expense and risk to rule out a diagnosis or to precisely define the extent and character of disease. In some fields, our ability and efforts to obtain precision in diagnosis may be out of proportion to therapies available to our patients. Medicine has been subject to “recurrent fads of therapy” (1) from the time of Lewis Thomas' father and through most of the 20th century. However, with 20th-century advances in our knowledge of human biology and the evaluation of therapeutics, particularly through the widespread use of the randomized clinical trial, our ability to define beneficial therapy has improved mightily. It is now ironic that we spend so much effort on precision in diagnosis and yet are relatively imprecise in therapeutics.
Therapeutics, like diagnosis, should be viewed as an exercise of discovery. In diagnosis, we try to discover the nature and cause of a patient's symptoms and signs; in therapeutics, we learn over time the effects of treatment on these symptoms and signs.
The recent revival of interest in n-of-1, or single-patient, trials offers the opportunity to enhance therapeutic precision for the individual patient (2-5). Arguments in favor of efforts to improve therapeutic precision include the prevalence and severity of adverse drug effects (6), the increasingly high cost of therapeutics (7), the better availability of beneficial drugs, and the increasing consensus that a physician's role is to provide care that improves patient-valued outcomes.
The n-of-1 trial increases precision by applying the principles of rigorous clinical trial experimentation—randomization, blinding, and assessment of predetermined outcomes—to the everyday experiments we conduct when we administer therapy to single patients (4, 8, 9). Randomization of treatment order is necessary to minimize biases that occur related to the order of treatment and to allow blinding. Double blinding of treatment type is needed to minimize observer bias and the placebo effect. The use of predetermined outcomes enables clinicians and patients to define the desired outcomes of treatment—ideally,on the basis of patient values. The n-of-1 trial properly forces doctor and patient to agree to measure 3 to 5 explicit outcomes, a critical step that, in our judgment, is easily bypassed in everyday treatment.
If therapeutic precision is desirable, why not routinely use the n-of-1 method? The reason to do an n-of-1 trial is to improve the ability of the doctor and the patient to determine therapeutic efficacy. The n-of-1 trial is useful only when there is important doubt about whether a new or existent treatment is helpful in a particular patient. The treatment in question is usually one that does not cure but is designed to ameliorate the effects of a chronic disease. The response to this treatment is somewhat unpredictable, so there is uncertainty about what the individual patient's response will be. Thus, for example, an n-of-1 trial to study H-2 blocker treatment of duodenal ulcer disease would likely be of little value, except perhaps to document a possible side effect of treatment. By contrast, many subjective symptoms, such as headache, non-ulcer dyspepsia, joint and muscle pain, and the signs and symptoms of chronic neurologic, behavioral, and cognitive diseases, display striking variations in the way an individual patient responds to a given treatment. If a treatment also has potentially toxic effects and will be used long term, the value of a method to increase therapeutic precision increases. If the physician and patient disagree, an n-of-1 trial may be an attractive way to resolve the disagreement.
The cost of an n-of-1 trial is considerable, but, typically, the direct costs are less than the costs of commonly used diagnostic tests, such as magnetic resonance imaging scans, angiograms, and the like. Our experience indicates that 15 to 20 hours of a practitioner's (pharmacist and physician) time are spent on a typical trial, containing 4 to 6 treatment periods, with 50% of the time used for trial preparation and 30% involved with patient visits (10). The difference between a diagnostic test and an n-of-1 trial is that the cost of the latter is usually played out over weeks and months, during which time uncertainty about treatment remains as costs accrue. By contrast, a diagnostic test usually provides a more immediate answer. For these reasons, it is unlikely that n-of-1 services will ever achieve the current activity level enjoyed by diagnostic services, although their product—improved knowledge about a single patient—can be viewed as comparable.
If clinicians and patients wish to enjoy the improved precision offered by results of n-of-1 trials, an organized service may be needed. At our institution, a typical trial requires the following steps: patient interview to determine interest, willingness, and desired outcomes; a review of the therapeutic literature, if needed; trial design; study medication preparation and procurement; informed consent; data collection during the trial; analysis of results; and discussion of results with physician and patient. At present, whether a service is developed and sustained is a function of the enthusiasm of a single investigator or a small cadre of interested clinicians, along with availability of funding.
Whether n-of-1 trials will become generally available to help clinicians improve therapeutic precision is uncertain. If advances in biology lead to therapeutics that are more predictably effective, need for n-of-1 trials will lessen. More objective methods to assess response to therapy and minimize observer bias will also reduce the need for n-of-1 trials. On the other hand, it is likely that both generalists and specialists will be caring for increasing numbers of patients with incurable, chronic diseases. These patients will usually be elderly, and one of the tenets of geriatrics is that variability in treatment response increases with age. To the extent that n-of-1 methods can meet the desire for patients and physicians to gain more precise knowledge of therapeutic efficacy, such trials logically should play a role in clinical practice. Our own work suggests that physicians are more likely than patients to gain confidence in therapeutic decisions following an n-of-1 trial (10). Thus, physicians will probably determine whether the current wave of interest remains localized and episodic or becomes more widespread in clinical practice.
Eric B. Larson, MD, MPH
Allan J. Ellsworth, PharmD