Table 1.

Rationale for the methodology

Conditions to use a multiple N-of-1 trial (adapted from Nikles and Mitchell [16])

• The patient's condition is stable during the phases with minimal fluctuations over time  • The treatment does not change the condition but only treats the symptoms  • The intervention has a short wash-out, i.e. its effect disappears when it is removed  • The therapeutic effect is quick when the treatment is applied and quickly stops when it is removed, i.e. on/off effect  • There is no cumulative effect of the treatment

Advantages to use a multiple N-of-1 trial methodology

• Patients lost to follow-up or out of the study can still be analyzed on the phases already completed  • Possibility to work with heterogeneous samples for individual interpretation  • The 2 arms of the treatments are perfectly matched  • All patients test both treatments (no ethical problem)  • At group level: statistical significance with smaller sample size  • Patient-by-patient analysis takes into account inter-individual variability by analyzing each N-of-1 trial to determine is a patient is a responder, non-responder, or is aggravated  • Highlights possible inter-individual variability in the global results  • Allows to extract an effect even in presence of large intra-individual variability (e.g.: patient performing variably because of pain, fatigue, unusual activity…) by measuring repeatedly the patient over time.

Disadvantages to use a multiple N-of-1 trial methodology

• Only treatments with true on/off effects can be explored and the condition must be stable. Therefore, any prosthetic/orthotic inducing cerebral plasticity (e.g.: thumb opponent splints in children) or requiring a long time for the patient to adapt/learn cannot be explored. This limits the range of prosthetic interventions that can be explored by multiple N-of-1 trials.  • The population is smaller and may not be representative of the general populationa.  • Inability to explore statistically predictive factorsb.  • Inability to detect low-prevalence side effects.

^aHowever, this problem is not specific to N-of-1 trials. Randomized controlled trials (RCT) claim in theory to provide a generalizable population outcome but they mask variability in response to treatment. Even in the most successful RCT study there are individuals whose behavior is not affected or is worsened by treatment, but these results are drowned out by a group mean.

^bThis is also true for randomized controlled trials (RCT). Power is calculated to demonstrate the effect of a treatment and not to determine predictive factors and responder characteristics.