How to read a clinical study on a food supplement

Eclo Science

Randomization · placebo · endpoints · bias

RCTread before concluding

In brief. A study being “published” does not automatically make it strong evidence. To understand what a trial actually allows us to conclude, we need to look at who was studied, with which ingredient, at what dose, for how long, against what comparator, using which endpoints and analysis. For Eclo, the most important filter is the identity of the evidence: a study on an ingredient never becomes, by simple association, a study on the finished product containing it.

What is the first question to ask?

Before even looking at the result: “What exactly was the study conducted on?” A raw material may have a standardized composition, a specific process and a studied dose. A result obtained with that ingredient should not automatically be attributed to another ingredient in the same category.

A trial on ExceptionHYAL® Star at 200 mg/day is therefore relevant to that ingredient at that dose. It does not become a trial of Ressource™, which combines several actives.

The 14 questions to ask before interpreting a result

Population: who participated? Age, sex, skin condition, inclusion/exclusion criteria.
Sample size: how many people were enrolled, randomized, analyzed and completed the study?
Randomization: was allocation to groups genuinely random?
Comparator: placebo, another intervention, or simply before/after?
Blinding: did participants, investigators and/or assessors know which group was received?
Exact intervention: which raw material, form, standardization or formulation?
Dose: what daily amount was actually studied?
Duration: four weeks, eight weeks, three months? An effect at eight weeks does not prove an effect after three days.
Primary endpoint: which result was defined as the priority before analysis?
Measurement: objective instrument, clinical grading, standardized photography, questionnaire or subjective perception?
Relevant comparison: did the active group change differently from the comparator, or only from its own baseline?
Magnitude and uncertainty: what is the effect size and confidence interval, beyond the p-value alone?
Missing data: how many dropouts, for what reasons, and how were they analyzed?
Funding and conflicts of interest: who funded the work, who analyzed it, and what relationships were declared?

Why does randomization matter?

Randomization aims, on average, to distribute known and unknown factors between groups. It reduces the risk that people receiving the active ingredient are systematically different from those receiving the comparator. But the word “randomized” does not guarantee a high-quality study: sequence generation, allocation concealment, protocol deviations, missing data and analysis also matter.

Does double-blind mean perfect evidence?

No. Blinding reduces certain behavioral and assessment biases, particularly for subjective endpoints. But a trial can be double-blind and still be too small, too short, poorly analyzed or based on an unhelpful endpoint. Some protocols also make blinding difficult; that risk should be discussed rather than ignored.

Why is placebo important in beauty studies?

Skin changes with season, humidity, skincare habits, sun exposure and many other factors. Taking part in a study can also change behavior. A placebo group followed under the same conditions helps distinguish changes associated with the intervention from these broader variations.

Before/after is not the same as active versus placebo

A significant improvement from day 0 to day 56 in the active group may be interesting, but the stronger causal question is often: did the active group change differently from the placebo group? An analysis based only on within-group change can exaggerate the strength of a result.

What is a primary endpoint?

A rigorous trial defines its primary endpoints before seeing the results. Secondary and exploratory endpoints remain useful, but should be identified as such. Measuring many parameters and then communicating only those with favorable p-values increases the risk of selective conclusions.

What does “statistically significant” mean?

A p-value below a conventional threshold measures neither the size of a benefit nor its practical importance. A tiny difference can be statistically significant in a large sample; an interesting estimate can be too imprecise to cross the threshold in a small study.

We therefore need to consider effect size, confidence intervals, endpoint relevance and consistency with other studies. “Not significant” does not prove “no effect”; it means that the protocol did not provide sufficient statistical evidence for that endpoint under the planned analysis.

How do the main study designs compare?

Design	What it adds	Typical limitation
Open before/after study	Preliminary signal and change over time	No placebo control and many possible biases
Non-randomized controlled study	Comparison between groups	Possible baseline differences between groups
Randomized controlled trial	Stronger protection against confounding	Quality still depends on execution and analysis
Randomized, placebo-controlled, blinded trial	Robust design for many efficacy questions	Does not compensate for low power, poor endpoints or inappropriate extrapolation
Systematic review / meta-analysis	Structured synthesis of several studies	Depends on the quality and comparability of the included studies

CONSORT, SPIRIT and RoB 2: three different tools

CONSORT 2025 is a reporting guideline for randomized-trial results. It helps assess whether the paper clearly describes what was done and found; it is not an efficacy score or a guarantee that bias is absent.

SPIRIT 2025 concerns reporting of a randomized-trial protocol: what is planned before or during the conduct of the study. A public protocol makes it easier to compare planned analyses with the results eventually reported.

Cochrane RoB 2 is used to assess the risk of bias in a randomized-trial result across several domains, including randomization, deviations from intended interventions, missing data, outcome measurement and selection of the reported result. These tools are complementary, not interchangeable.

Should a trial be registered?

Prospective registration improves transparency. For trials within its scope, the ICMJE requires registration in a public registry at or before the time of first patient consent for enrolment.

But a registration number is not enough on its own: the endpoints, population, duration and published analyses still need to be compared with what was planned. Trial registration, a public protocol and a statistical analysis plan are related but distinct levels of transparency.

Does peer-reviewed publication guarantee quality?

No. Peer review generally makes methods and results more inspectable, but it does not eliminate errors, bias, low power or conflicts of interest. Conversely, an unpublished study can contain useful data, but it is less verifiable when the full protocol and results are not accessible.

Who funded the study?

In food supplements, ingredient manufacturers often fund research on their own raw materials. That does not make the results false. But funding, the sponsor’s role and authors’ relationships should be visible. Independent replication generally increases confidence.

Can a supplier study be useful?

Yes, provided it is labelled correctly. A proprietary study may document a formulation, dose and measurements that have not yet been published. Its level of verifiability is nevertheless different from an article whose complete methodology is accessible and peer reviewed.

Eclo therefore distinguishes a published study on the ingredient from a proprietary study on the ingredient, rather than presenting them as equivalent.

Was the finished product studied?

This is the question that prevents the most exaggeration. A formula containing five ingredients may draw on five different bodies of literature, but that does not create a sixth study on their combination. Interactions, doses and use of the complete product have only been tested if a protocol actually used the complete product.

The evidence framework used by Eclo

Authorized regulatory claim: a nutritional function authorized within the legal framework and under specified conditions.
Published study on the ingredient: an identifiable scientific article concerning the relevant ingredient.
Proprietary study on the ingredient: supplier data explicitly presented as such.
Mechanism / preclinical data: biological information supporting a hypothesis without, on its own, proving a clinical benefit.
Study on the finished Eclo product: only if the final product itself was tested.

Frequently asked questions

Does a published study mean the result is proven?

No. Publication makes the study inspectable; the strength of the evidence then depends on design, execution, analysis, uncertainty and replication.

Is a supplier-funded study invalid?

No. Funding is context and a potential source of bias to disclose, not an automatic verdict.

Does a study on an ingredient prove a product containing it works?

No. It documents that ingredient under the studied conditions. The finished product is only “studied” if the complete formula itself was tested.

Does “statistically significant” mean “important for skin”?

Not necessarily. We need the magnitude of the difference, its confidence interval and the practical relevance of the endpoint.

What can we conclude?

Reading a study well means avoiding both “it is published, therefore it is true” and “it is industry-funded, therefore it is false”. It means examining the design, intervention identity, measurement quality, analysis, uncertainty, transparency and replication — then limiting the conclusion precisely to what the protocol tested.

Cart 0

No more products available for purchase