Definition, objects, and research boundary
A human consumption trial of a sleep-related functional food asks eligible participants to consume a sufficiently characterised food, beverage, dietary supplement, or constituent according to a protocol and evaluates prespecified sleep-related outcomes, safety, and acceptability. Objects may include nutrients, botanical ingredients, amino acids, fatty acids, probiotics or prebiotics, and multi-ingredient formulations. Descriptions such as “natural,” a long history of consumption, or lawful sale as ordinary food do not replace ethics review, product quality, or risk assessment. A consumption trial differs from a sensory preference test, general consumer survey, observational dietary study, or therapeutic drug trial. Product classification, regulatory permission, and whether a study may investigate diagnosis, treatment, mitigation, or prevention of disease depend on its objective, formulation, dose, claims, and jurisdiction. Results support only conclusions matched to the tested product, batch, use, population, comparator, duration, and outcome. One trial does not establish effects for a broadly named ingredient, a different formulation, commercial batches, long-term use, or treatment of insomnia or sleep-disordered breathing.
Product characterisation, controls, and supply chain
Before enrolment, lock the product name, formulation, amount per serving, dosage form, consumption instructions, manufacturing site, batch, packaging, storage, and expiry. Confirm chemical identity and content for a discrete compound. For materials of botanical, animal, microbial, or fermentation origin, also record species or strain, part, provenance, processing or extraction, standardised constituents, and appropriate authentication evidence. A multi-ingredient product should not be reported only by brand or as a “proprietary blend”; disclose composition sufficient to reproduce and interpret the study and use suitable methods to verify identity, purity, strength, composition, uniformity, contaminants, and stability throughout the study. Match placebo or control food as far as practical for appearance, odour, taste, texture, energy, and consumption procedure, while confirming that it contains no component capable of affecting sleep or the target pathway. No-intervention, usual-diet, and active controls answer different questions. Maintain receipt, release, temperature, dispensing, return, accountability, and disposal records. Assess comparability after changes in ingredient source, processing, batch, or packaging instead of pooling different test articles by default.
Protocol, ethics, registration, and participant protection
Before recruitment, the protocol should define the question, design, primary and secondary outcomes, time points, sample-size basis, randomisation and masking, intervention and comparator, adherence, safety monitoring, statistical methods, and stopping rules, with prospective registration in an applicable public registry. Obtain review from an authorised ethics committee and valid informed consent that explains the product and research nature, consumption requirements, foreseeable risks and discomforts, uncertainty of benefit, alternatives, compensation, privacy, withdrawal, and arrangements for research-related injury. Recruitment and payment must not present a food as guaranteed treatment or exert undue influence on economically disadvantaged people, people with limited decision-making capacity, children, or other vulnerable groups. Eligibility should be relevant to the question and safety, addressing allergy, pregnancy or lactation, hepatic or renal status, sleep disorders, mental or neurological conditions, medicines and other supplements, alcohol and caffeine, shift work, and time-zone travel. Severe sleepiness, suspected sleep-disordered breathing, or other clinical risks need a predefined referral route; study participation cannot replace necessary care.
Design, bias control, and real consumption behaviour
Choose a randomised parallel, crossover, or other design suited to onset and washout, outcome variability, feasibility, and the research question. Document sequence generation, allocation concealment, and masking of delivery staff and outcome assessors separately. Colour, flavour, odour, texture, or gastrointestinal effects can reveal allocation and should be assessed rather than addressed only by the word “double-blind.” A crossover study needs a justified washout interval and evaluation of period and carryover effects. Multiple doses, ingredients, time points, and outcomes increase chance findings, so comparisons and multiplicity handling should be prespecified. At baseline and follow-up, capture meal timing, overall diet, weight change, alcohol, caffeine, nicotine, physical activity, naps, light, screens, time in bed, shift work, jet lag, and concomitant treatment because these can alter sleep. Assess adherence through dispensing and returns, diaries, timestamps, or suitable biomarkers, distinguishing no use, partial use, mistimed use, and discontinuation. Co-interventions, rescue treatment, and lifestyle advice should be equal between groups or transparently recorded; an intensively controlled protocol should not conceal conditions of ordinary use.
Sleep outcomes, measurement, and clinical meaning
Match the primary outcome to the claim and population, defining the variable, instrument, algorithm, time window, and meaningful difference before the study. Sleep diaries and validated questionnaires capture participant experience; actigraphy estimates multi-day sleep and rhythmic patterns; laboratory or home polysomnography provides specific physiological information. These methods measure different constructs and are not interchangeable. If a consumer device's sleep stages or composite score is used, identify the device, firmware, algorithm version, wear protocol, missing-data rules, and independent validation boundary. An algorithm update must not silently change an endpoint. Standardise or record workdays and free days, number of measurement nights, first-night effects, season, time zone, environment, and interval from consumption to bedtime, applying the same assessment to all groups. Report raw group values, changes, effect sizes, confidence intervals, and data completeness rather than statistical significance alone. A brief questionnaire improvement, change in one sleep stage, or higher device score does not necessarily establish a clinically meaningful improvement in daytime function, quality of life, or health.
Statistics, missing data, safety, and data governance
Base sample size on the primary outcome, target difference, variability, power, significance level, expected attrition, and design effects rather than analysing however many people happen to enrol. Before unmasking, the analysis plan should define analysis populations, models, covariates, repeated measures, crossover features, outliers, adherence analyses, multiplicity, subgroups, and sensitivity analyses, and explain how post-randomisation discontinuation, dose modification, rescue measures, and missing data will be handled. Excluding “non-compliant” participants or those without improvement can break randomisation; post hoc responder searches are exploratory. Safety surveillance should ask actively rather than depend only on spontaneous reports and address worsened sleepiness, next-day impairment, gastrointestinal effects, allergy, medicine or supplement interactions, laboratory abnormalities, and driving or work risk, recording severity, timing, action, and causality assessment. Escalation, pausing, emergency unmasking, and independent monitoring should be proportionate to risk. Identity, sleep, diet, and specimen data should be minimised, access-controlled, auditable, and governed for later use by consent. Withdrawing from the study, stopping consumption, and deleting data already collected are separate matters.
Reporting, interpretation, replication, and claim boundaries
Report participant flow, recruitment sites and dates, baseline data, intervention actually received, adherence, accompanying behaviours, every prespecified outcome, numbers with data at each time point, missingness, deviations, adverse events, effect estimates, and uncertainty in line with the registered protocol, clearly distinguishing prespecified from post hoc work. The protocol, statistical plan, registry record, product specifications, batch tests, funding, roles of investigators and suppliers, and conflicts of interest should be cross-checkable. Negative, inconclusive, and early-terminated studies also warrant timely disclosure. A replication should use a comparable test article, not merely the same generic ingredient name. Interpretation should address sensory unmasking, selected populations, short follow-up, small samples, multiple outcomes, adherence, and real-world dietary differences, and avoid moving directly from an associated biomarker to a sleep benefit. Human consumption evidence is not automatically authorisation for a health claim in any jurisdiction and cannot support “works for everyone,” “works tonight,” “no side effects,” “replaces treatment,” “cures insomnia,” or diagnosis based on a food study. Market claims must match the total evidence, actual formulation, target population, intake, and local rules.