N=1 Experimentation and Personal Science
Individual-level experimentation — patients systematically testing interventions on themselves, tracking outcomes, and sharing results with the community — represents a form of knowledge production that operates outside the formal research system but has a defined methodology and community infrastructure.
The N-of-1 paradigm as a formal research method uses a single patient as their own control: the patient cycles through intervention and control (or washout) periods, with repeated measurements of relevant outcomes (Wicks 2018). The methodology has particular advantages for conditions like ME/CFS, where heterogeneity makes group-level RCTs difficult to interpret (a treatment that helps one subtype may harm another, producing a null group effect) and where patients have intimate, longitudinal knowledge of their own symptom patterns that can inform measurement design.
McDonald and colleagues developed a patient-centred N-of-1 protocol specifically for ME/CFS symptom fluctuations — designing a framework in which patients track symptom triggers, responses, and PEM patterns over time in a structured observational design (McDonald et al. 2022). The protocol is significant because it was designed for ME/CFS, not adapted from a generic N-of-1 template — it accounts for PEM as a delayed outcome variable, fluctuating baselines, and the patient’s own hypothesis about their triggers.
The citizen science dimension is that these methods have been adopted by patient communities outside formal research: patients tracking their response to supplements, dietary modifications, medications, and activity protocols, and sharing aggregated results in online forums. This creates a distributed, informal knowledge base — preliminary, uncontrolled, but real. (Certainty: 0.68 for Wicks 2018 paradigm (conceptual framework, PatientsLikeMe employee author, general-population evidence), 0.55 for McDonald 2022 (protocol only, no results).)
Consequence: The N-of-1 paradigm represents a methodological bridge between the individual patient’s experiential knowledge — “this worked for me” — and the formal research system’s demand for controlled evidence. It does not replace RCTs, but it provides a structured way to generate preliminary evidence, identify promising candidate interventions for group-level study, and capture individual heterogeneity that group-level designs miss. For a disease with high heterogeneity and limited research funding, N-of-1 methods leverage the community’s most abundant resource — longitudinal self-observation — as a research tool. Severity applicability: N-of-1 methods are most feasible for mild/moderate patients who can self-administer interventions and track outcomes; severe/very severe patients may require caregiver-assisted data collection, which introduces additional measurement challenges.
The N-of-1 paradigm has well-documented limitations. Self-experimentation generates biased data — patients testing a treatment they hope will work are not blinded, and placebo effects, natural fluctuations, and confirmation bias all operate without the controls that group-level RCTs provide (Wicks 2018). Low retention and uncertain data quality are documented challenges in mHealth-based citizen science projects (Schmitz et al. 2018).
The translation gap from N-of-1 findings to generalizable knowledge is substantial. A patient who reports improvement on supplement X during N-of-1 testing may have experienced a placebo response, a naturally occurring remission, a co-intervention effect, or a genuine treatment effect — and the N-of-1 design cannot reliably distinguish these. Patients using N-of-1 methods to guide their own treatment are making decisions under uncertainty — a condition that also applies to clinicians making treatment decisions in the absence of RCT evidence. The N-of-1 method structures the uncertainty; it does not eliminate it.
Most critically, the community-level aggregation of N-of-1 results — patients sharing their findings in forums — produces a knowledge base that is richer than individual anecdotes but methodologically weaker than controlled studies. A forum archive documenting that 60% of patients who tried intervention X reported improvement is not equivalent to an RCT finding of 60% efficacy — the sample is self-selected, the outcome measurement is unstandardized, and negative results are less likely to be reported. (Certainty: 0.65 — these limitations are documented in the Wicks 2018 and Schmitz 2018 sources; the assessment of their severity for ME/CFS specifically is inferential, not directly studied.)
Consequence: N-of-1 methods are epistemically intermediate — more structured than anecdote, less controlled than trial. Their appropriate role is hypothesis generation, not conclusion. But in a disease with minimal RCT infrastructure, the community-level accumulation of N-of-1 results is already functioning as a de facto evidence base that patients use to make treatment decisions. The challenge is not to suppress this knowledge base — it exists and is used — but to structure it so that its limitations are transparent and its findings are testable. Severity applicability: same as above.