Why doing your own research does not always work, and why evidence needs interpretation
Key takeaways
- Access to skincare research does not automatically mean understanding what the evidence actually shows.
- Many skincare and supplement claims are based on real studies, but the results are often exaggerated in marketing.
- Common issues include small studies, short follow-up periods, surrogate measurements (e.g. hydration instead of visible anti-ageing), and selective reporting of positive results.
- Collagen supplements are a good example, with evidence suggesting modest hydration benefits but far weaker support for broad anti-ageing claims.
- AI tools can help summarise research but often miss important nuances, limitations and patient-specific factors.
- The most important question is not whether a study exists, but whether the claim accurately reflects what the study actually proved.
- Good clinical advice combines evidence with individual factors such as skin type, medical history, goals and suitability.
Patients arrive at clinic better informed than at any point in the history of dermatology. They come having read the studies, compared the ingredient lists, watched the explainer videos, saved the before-and-after images and, increasingly, having asked an artificial intelligence tool to summarise the literature before they sit down. Most of them have done a genuine amount of work, and almost all of them are trying to make a sensible decision about their own skin. None of that is the problem. The problem is quieter and more interesting than that, and it is the thing this piece is about.
There is a difference between having access to scientific information and being able to translate it, and that difference is where a great deal of skincare and supplement marketing now operates. The issue is rarely that the underlying study is fake or fraudulent. Far more often the study is real, carefully conducted and perfectly honest about its own limitations, and the difficulty arises only later, when a narrow, short-term, instrument-measured finding is translated into a broad consumer promise that the data was never designed to support. Hydration becomes rejuvenation, a biomarker becomes proof of anti-ageing, elasticity becomes collagen rebuilding and a p-value becomes certainty. The science does not change in that journey, but the claim does, and patients are left holding the claim rather than the science.
At Self London we remain consultation-led for precisely this reason. Reading the abstract is not the same as understanding what the result means for the person in the chair. What follows is an attempt to show, with real published examples, where the translation tends to break down, so that the next time you encounter a confident claim you have a way of asking whether the claim and the evidence are actually describing the same thing.
The honest objection, answered first
There is an obvious retort to everything that follows, and it is worth meeting head on rather than hoping nobody raises it. An argument that says evidence needs expert interpretation, written by an expert who sells consultations, can sound conveniently self-serving, as though the conclusion were chosen to flatter the author. That suspicion is healthy, and the right response is not to ask for trust but to hand the reader the tools, so that everything claimed here can be checked rather than taken on authority. What follows names the specific mechanisms by which a study can appear to show an effect that is not really there, and applies them without exception, including to research that happens to support the case being made. If the argument is sound, it survives being turned on its own sources, and where it does not survive, that is said plainly. The aim is not to make you defer to a clinician. It is to make the next confident claim you meet easier to interrogate on your own.
How a study can show an effect that is not really there
Most misleading skincare and supplement claims do not rest on fabricated data. They rest on real studies whose findings have been quietly inflated or stretched, and the inflation usually happens through one of a small number of recurring mechanisms. Learning to recognise them is most of what separates reading a study from understanding it, and once you can name them you start to see them everywhere.
The first is multiplicity, which is the simple arithmetic that if you measure enough things, something will reach statistical significance by chance. A study that records a dozen outcomes across several sites and several time points is almost guaranteed to find a scattering of positive results even if the product does nothing, and unless the analysis was planned in advance with a single primary endpoint and a correction for all those comparisons, a lone significant finding among many tells you very little. The honest question is never whether a study found something, but whether it found the thing it set out to test.
The second is the surrogate endpoint, where a measurement that is easy to move stands in for the outcome you actually care about. A hydration reading, an elasticity instrument or a biomarker in a biopsy can shift convincingly while the visible appearance of the skin barely changes, because biological activity and visible benefit are not the same thing, and a result that is real at the level of an instrument can be invisible at the level of a face.
The third is the missing control, or the reliance on a within-group comparison rather than a between-group one. Skin improves for all sorts of reasons that have nothing to do with the active ingredient, including the moisturising effect of any cream base, the daily ritual of application, regression to the mean and the patient’s own wish to see a result. The only way to isolate what the active actually contributed is to compare it against a placebo or vehicle, and a study that reports how much better people got than their own starting point, without telling you how the control group did, has not yet shown that the product did anything at all.
The fourth is the small or selective sample, which exaggerates and destabilises. A trial of thirty or fifty carefully chosen participants can detect a signal, but it produces unstable effect sizes, is vulnerable to chance imbalances, and frequently fails to replicate, and when its participants are healthy volunteers in a narrow age band with none of the complicating skin that real patients have, even a clean result may not transfer to anyone outside the study.
The fifth is the short follow-up that cannot match a long-term claim. A twelve-week study can speak to hydration or tolerability, and it cannot speak to collagen remodelling, durable pigment control or anything describable as longevity, because that biology unfolds over years rather than weeks. The timescale of the evidence has to match the timescale of the promise, and very often it does not.
The sixth is the funding and quality effect, where the apparent answer depends on who paid and how rigorously the work was done. The largest and most flattering effects, looked at across a whole literature, tend to cluster in the lowest-quality and most conflicted studies, and a result that vanishes when you restrict the analysis to independent, well-conducted trials was never as solid as the pooled figure suggested.
The seventh is the relative percentage, which makes a trivial change sound dramatic. A thirty per cent improvement from a tiny baseline may be completely invisible, and a figure that omits the absolute change, the starting point and the threshold at which a human being would actually notice is engineered to impress rather than to inform.
The eighth is the absence of a defined threshold for meaningful change, the minimum difference that a patient or an independent assessor would actually register. A study that never specifies, in advance, how much change would matter can report a statistically significant difference that is real and simultaneously too small to see, which is how a trial proves a difference without ever proving a benefit.
None of these mechanisms requires anyone to lie. Each is a way an honest study can be honestly conducted and then dishonestly translated, and the sections that follow work through the most important of them using real published examples, including the awkward cases where the study supporting a point turns out to have weaknesses of its own.
“Clinically proven” does not always mean what people think it means
The phrase “clinically proven” is one of the most powerful in beauty, and one of the least informative. It carries the reassuring weight of medical rigour while telling you almost nothing of what you would need to know to act on it. It does not tell you what was proven, how it was measured, in how many people, against what comparator, over what period, by whom it was funded, or whether the result would ever be visible to the person using the product.
A serum may be clinically proven to improve a corneometer hydration reading after four weeks, which is a real and measurable thing, and entirely different from being proven to reverse the visible signs of ageing. A supplement may produce a statistically significant change in an elasticity instrument over twelve weeks without that change translating into a face that looks lifted, or laxity that looks improved, or collagen that has meaningfully been rebuilt. A study answers the specific question it was designed to answer. Marketing tends to answer a more flattering question instead, and the gap between the two is where patients are most often misled.
The UK advertising regulator has made this concrete on more than one occasion. In a 2023 ruling against a collagen supplement brand, the Advertising Standards Authority accepted that a hydration claim could stand but found that claims about maintaining skin elasticity and firmness and reducing the appearance of fine lines and wrinkles had not been substantiated, and were therefore misleading. The same ruling noted that a summary of an in vitro collagen-synthesis study was not relevant evidence for those claims, which is a useful detail, because it shows that the existence of “a study” is not the same as the existence of evidence for the particular thing being sold. A laboratory observation about collagen synthesis in a dish is a long way from a visible improvement in a real human face, and the regulator was unwilling to let one stand in for the other.
When you measure enough things, something becomes significant
Ageing studies tend to measure a great deal at once. Hydration, transepidermal water loss, elasticity, firmness, wrinkle depth, pigmentation, radiance, roughness, pore count, dermal density, investigator grading, patient satisfaction and sometimes biomarkers, often across several sites and several time points. This breadth is not in itself a flaw, and there are good reasons to capture more than one outcome, but it creates a statistical vulnerability that the consumer-facing claim rarely acknowledges. The more endpoints a study tests, the more likely it becomes that one or two will reach statistical significance by chance alone, unless the analysis is designed in advance with a defined primary endpoint and a proper correction for multiple comparisons.
This matters because a single positive finding among many can quietly become the whole story. A statistically significant improvement in one hydration parameter, at one site, at one time point, can be marketed as “clinically proven to improve the signs of ageing,” which sounds vastly broader than what actually happened. The useful question, when you see a claim, is not whether the study found something, but what it found, whether that finding was the one the study set out to test, and whether it survived honest statistical scrutiny rather than emerging from the noise of everything that was measured. The more outcomes a study reports, the more this question matters, not less.
A biomarker is not a face: the surrogate endpoint problem
A surrogate endpoint is a measurement that stands in for the thing you actually care about. In skincare these are everywhere, and they are seductive precisely because they sound so scientific. Hydration readings, transepidermal water loss, cutometer elasticity, ultrasound dermal density, histological collagen and fibrillin markers all belong to this category. They can be genuinely informative about mechanism, and they can show that an ingredient is doing something biologically plausible, but they are not the same as the outcome a patient experiences, which is whether the skin actually looks better, and whether that improvement lasts.
The cleanest published illustration of this gap is a 2009 study in the British Journal of Dermatology by Watson and colleagues, which subjected an over-the-counter cosmetic anti-ageing product to a proper double-blind, vehicle-controlled trial. The biology was convincing. The product produced significant deposition of fibrillin-1, a structural protein of the dermal matrix, both in a short patch-test assay and after six months of use, and fibrillin-1 is a well-regarded biomarker for repair of photoaged skin. The clinical picture was more nuanced, and this is the part that tends to be forgotten. At six months the treated skin improved significantly in facial wrinkles compared with its own starting point, while the vehicle did not, but the head-to-head comparison between the active product and the vehicle at that stage was only a non-significant trend. The between-group difference, which is the comparison that actually controls for the placebo effect, had not yet reached significance, and it was only after a further open period, with everyone using the active product, that clinical improvement became clearer.
This is not a criticism of the study, which was unusually rigorous for a cosmetic, and which its authors reported with admirable honesty. It is an almost perfect teaching case, because it lays out the full distance between a solid biological signal, a within-group change that looks impressive in isolation, and the placebo-controlled clinical difference that a patient would actually notice in the mirror. A biomarker can support a mechanism. It should not be allowed to become the entire claim, and the moment a fibrillin result or a collagen marker is presented to you as though it were a visible rejuvenation, the translation has already gone wrong.
The collagen supplement story, and why funding changes the answer
Oral collagen is the clearest example of how a confident consensus can rest on shakier ground than it appears to, and it is worth setting out carefully because the structure of the problem repeats across the whole category. For several years the pooled evidence looked persuasive. A large 2022 systematic review in Frontiers in Nutrition, covering sixty-six randomised controlled trials, found that oral collagen and ceramides produced statistically significant improvements in skin hydration and reductions in transepidermal water loss compared with placebo, with hyaluronan and procyanidin also showing positive effects on hydration. Read quickly, that sounds like settled science, and it is the sort of result a brand can cite with a straight face.
A closer look at that same body of work introduces the first caution, and the same scrutiny this article asks you to apply elsewhere has to be applied here too, including to a source that happens to support the argument. That 2022 review was co-authored by someone affiliated with a pharmaceutical company, and its own quality assessment was unflattering, since fewer than half the included trials adequately described how they randomised participants and fewer than a third reported proper allocation concealment, with most conducted in a single region. None of that makes its central finding wrong, and the finding is in any case a cautious one, but it does mean the review should be read as suggestive rather than authoritative, and it would be inconsistent to demand that standard of the marketing and not of a paper merely because the paper agrees with me. The outcomes that reached significance were in any event overwhelmingly hydration and water-loss measures, which are real but modest, and a long way from the wrinkle-reversing, collagen-rebuilding, ageing-defying language that tends to accompany the products. Hydration is not rejuvenation, and a meta-analysis of hydration outcomes, however large or however many trials it pools, does not become evidence of visible structural anti-ageing simply by being big.
The decisive analysis arrived in 2025, in The American Journal of Medicine, where Myung and Park pooled twenty-three randomised controlled trials involving 1,474 participants and did something most reviews do not, which was to break the results down by study quality and by who paid for them. Pooled across all twenty-three trials, collagen again appeared to improve hydration, elasticity and wrinkles, which is the headline a marketer would stop at. The subgroup analysis is where it falls apart. The effect on hydration, elasticity and wrinkles disappeared in the trials that were not funded by industry, while the significant effects were confined to the industry-funded studies, and a sensitivity analysis that removed a handful of outlier trials with extreme results further weakened the apparent benefit. The authors concluded that there is currently no clinical evidence to support the use of collagen supplements to prevent or treat skin ageing, which is a remarkable conclusion to reach about a product the public has been confidently sold for a decade.
Intellectual honesty requires noting that this paper has its critics, and that the criticism is instructive rather than merely defensive. Industry bodies and some independent commentators have argued that the conclusion overreaches its own data, pointing out that the abstract concedes a significant benefit across all studies before the subgroup analysis dismantles it, and that some trials the authors classed as independent were in fact internal research by collagen manufacturers, which would blur the clean funding line the headline depends on. These are interested parties and their objections are not neutral, but the objection is fair on its own terms, and it would be hypocritical for an article about scrutinising evidence to wave it away. The honest position is that the funding split is suggestive rather than decisive, and that it does not need to be decisive to make the point that matters, because every reading of this literature, generous or severe, leaves you with the same conclusion. The benefits that survive are modest and concentrated in hydration, the strongest effects cluster in the lowest-quality and most conflicted studies, and the confident anti-ageing language on the packaging is supported by markedly less than it implies. You do not have to accept the most damning interpretation to see that the marketing has outrun the science.
The broader lesson is not that every collagen supplement is worthless, nor that industry funding automatically invalidates a study, because much good research is industry-funded for the simple reason that companies have the strongest incentive to test their own products. The lesson is that the consumer-facing certainty had run far ahead of the independent evidence, and that the apparent answer shifted depending on who was asking the question and how carefully. When a claim rests on a meta-analysis, the meta-analysis is not the end of the conversation. The better questions begin there. Who funded the trials inside it, did the effect survive in the higher-quality studies, and was the outcome that reached significance the one being advertised, or merely the one that was easiest to move.
Meta-analysis organises evidence, it does not rescue it
A systematic review and meta-analysis sits near the top of the evidence hierarchy, and the phrase carries an authority that can short-circuit further thought, which is exactly why it deserves more scrutiny rather than less. A meta-analysis is only ever as good as the trials it pools, and if those trials are small, short, heterogeneous, industry-funded, conducted at different doses in different populations measuring different outcomes at different time points, then combining them does not manufacture certainty. It can simply organise a collection of limitations into a single, more impressive-looking number.
The collagen literature demonstrates this precisely, because earlier meta-analyses produced favourable pooled effects that the more careful, funding-sensitive 2025 analysis then unpicked. Nothing about the underlying biology had changed in the interval. What changed was the willingness to look inside the pooled figure and ask whether it survived contact with study quality and independent funding, and it largely did not. A meta-analysis can tell you what a body of evidence collectively suggests. It cannot turn weak primary evidence into a strong clinical truth, and the words “systematic review” at the top of a marketing page should prompt you to ask what was inside it, not to stop asking.
Short studies cannot carry long-term claims
A great deal of skincare and supplement research runs for eight, twelve or sixteen weeks, which is an entirely reasonable timeframe for assessing tolerability, barrier function, hydration or early changes in surface texture. It is a far less reasonable basis for claims about collagen remodelling, structural change, durable pigment control or the increasingly fashionable language of “skin longevity.” Skin ageing is slow biology, unfolding over years through photodamage, hormonal change, collagen decline and pigmentary instability, and a twelve-week study simply cannot speak to whether a product changes that trajectory over time, however good its short-term numbers look.
The honest principle is that the timescale of the evidence should match the timescale of the claim. A short trial that improves a hydration reading has earned the right to say something about hydration over a short period, and nothing more, and it should not be stretched into a promise about how your skin will age across the next decade.
This is also why sunscreen occupies a genuinely different evidential category from most anti-ageing products, and serves as a useful illustration of what better evidence actually looks like. The Nambour trial, published by Hughes and colleagues in the Annals of Internal Medicine in 2013, randomised over nine hundred adults under the age of fifty-five to daily versus discretionary sunscreen use and followed them for four and a half years, with skin ageing graded blind from silicone microtopography impressions of the skin. The daily-use group showed no detectable increase in skin ageing over that period, and were found to have around twenty-four per cent less ageing than the discretionary group. Notice what makes this persuasive. The intervention matched the claim, the follow-up was measured in years rather than weeks, the outcome was skin ageing itself rather than a hydration proxy, the assessors were blinded, and the sample was far larger than the typical beauty supplement study. When a claim is about ageing over time, the evidence ought to look like ageing evidence, and most of it does not.
Even this trial repays careful reading rather than reverent citation, which is the whole point of the exercise. The ageing it measured was assessed on the back of the hand rather than the face, so the strict finding concerns photoageing at that site rather than facial wrinkles directly, and the sunscreen used was an SPF 15, lower than would now be recommended, which means the result supports the principle of daily broad-spectrum use rather than endorsing that particular strength. Neither caveat undermines the conclusion, because the trial remains far stronger evidence than the typical anti-ageing study, and noticing them is exactly the habit this article is asking you to develop. A good study earns trust precisely because it survives this kind of scrutiny, and a claim that cannot survive it was never resting on much.
The average patient is not the patient in front of you
Most studies report the mean change across a group, which is statistically sensible and clinically incomplete, because the average can conceal an enormous spread of individual responses. A supplement that improves hydration by a small mean amount might have done so by nudging almost everyone slightly, or by producing a large effect in a responsive subgroup while leaving most participants unchanged, and the headline figure looks identical in both cases. If the responders were the drier, older, more photodamaged or post-menopausal participants, then the result tells you something about them and very little about a younger patient with oily, acne-prone skin who buys the product on the strength of the average.
In clinic this is the whole game, because we do not treat an average data point. We treat a particular person with a history, a skin type, a diagnosis, a tolerance profile, a hormonal context and a set of expectations, and a trial result is useful background to that conversation rather than a substitute for it. The same caution applies to aesthetic procedures, where an impressive mean improvement across a cohort guarantees nothing about the individual outcome. Some patients respond beautifully, some modestly, some need combination treatment and some are not suitable at all, and the work of a good consultation is deciding which of those a given person is likely to be, rather than quoting them the cohort average and hoping.
Small trials detect signals, they do not define universal truths
A large proportion of cosmetic and supplement research is small, often enrolling thirty, fifty or a hundred carefully selected participants over a few weeks, and frequently excluding anyone with active skin disease, pregnancy, medication use or recent treatment. Small studies are not worthless, and in early research they are exactly how a promising signal is first detected and a case for larger work is built, but they are more vulnerable to unstable effect sizes, baseline imbalance and positive findings that fail to replicate, and they become genuinely misleading the moment a forty-person trial is used to justify a sweeping claim about everyone.
The general principle is unglamorous but reliable, which is that the more ambitious the claim, the stronger the evidence behind it ought to be. A small short-term study can reasonably suggest that a product deserves further investigation, or that it was well tolerated over the period studied, and it cannot establish a durable, visible, universal anti-ageing benefit across the real-world population that will eventually buy it. Beauty marketing is the place where the confidence of the language tends to rise considerably faster than the quality of the evidence, and that mismatch is worth watching for.
Relative numbers sell, absolute numbers inform
One of the most common consumer-facing distortions is the relative percentage, because it reliably makes a small change sound dramatic. “Wrinkles reduced by thirty per cent” and “hydration improved by forty per cent” are persuasive precisely because they omit the information you would need to judge them, namely the absolute change, the baseline it started from and the threshold at which a human being would actually notice. A thirty per cent improvement from a very low baseline measurement may be completely invisible, and a statistically significant movement in an instrument score may sit well below the level of perceptible change, while the percentage still reads impressively on the packaging.
This is closely tied to a deeper and more technical gap, which is that many skincare studies never define in advance what size of change would actually matter to a patient. Statistical significance tells you only that a result is unlikely to be due to chance under the assumptions of the analysis, and says nothing about whether the change is large enough to see, to value or to justify the cost and the risk of irritation. The concept missing from much of this research is the minimum clinically important difference, the pre-specified threshold separating a change that registers on an instrument from a change a person would care about. A trial with no such threshold defined ahead of time can demonstrate a difference without ever demonstrating value, and a great deal of “clinically proven” sits in exactly that space.
A study can be statistically valid and clinically narrow
Trial populations rarely resemble clinic populations. Skincare and supplement studies tend to recruit healthy adults, often predominantly women, frequently within narrow age bands, and they commonly exclude people with active acne, melasma, rosacea, eczema, pregnancy, recent retinoid use, recent procedures, significant medical history or polypharmacy, while darker skin types are persistently underrepresented. A trial conducted this way can be entirely valid on its own terms and still tell you very little about a real patient, because internal validity and real-world applicability are not the same property.
The patients we actually see rarely arrive with a single, tidy problem. They present with acne and pigmentation together, rosacea alongside sensitivity, melasma in the context of perimenopause, active breakouts on top of old scarring, a history of isotretinoin, a plan to conceive, a course of prescription skincare already underway, or skin of colour with a meaningfully higher risk of post-inflammatory pigmentation. A study can be statistically immaculate and still fail to represent any of them, which is why the relevant question about a piece of evidence is not only whether it exists, but whether it applies to this skin, with this history, in this context. That filtering is most of what clinical judgement consists of, and it is precisely the step that reading alone cannot perform.
Patient perception matters, and it is not immune to expectation
None of this is an argument for dismissing how a patient feels about their own skin, because in aesthetics perception genuinely matters, and a product that makes someone more comfortable, more confident or more consistent with a routine has delivered something real. The difficulty is that self-reported improvement is unusually vulnerable to influences that have nothing to do with biology, including the texture and fragrance of a product, its packaging, its price, the authority of whoever recommended it and the simple human wish to see a return on money and hope already invested. The more expensive and emotionally loaded the purchase, the stronger those effects tend to be.
This is why patient-reported outcomes are most useful when read alongside more objective measures rather than in place of them, and why blinded assessment, standardised photography and validated scales exist. It is also why before-and-after imagery, however compelling, should be treated with care, since lighting, angle, expression, camera settings, skin preparation, makeup and the natural daily fluctuation of skin can all manufacture an apparent transformation that owes more to the photographer than to the product. At Self London we use VISIA imaging not because a machine replaces clinical judgement, but because consistent, objective documentation strengthens it, and gives both clinician and patient a more honest record than memory and a flattering photograph can provide.
Where artificial intelligence fits, and where it does not
Artificial intelligence tools have become part of how patients research their skin, and they are genuinely useful for summarising papers, explaining ingredients and making dense literature more approachable. They also flatten nuance in ways that are easy to miss, because a confidently written summary can present weak, heterogeneous or inapplicable evidence with exactly the same assurance as strong evidence, list studies without weighing their quality, conflate a hydration finding with an anti-ageing one, treat a mechanistic claim as a clinical outcome, or recommend something for “skin” without any sense that a pigmentation approach suited to one skin type may be actively risky in another.
The result is a newer version of an old problem, in which a patient arrives having not only done their own research but had it organised into a fluent, confident narrative by a tool that has never examined their skin, cannot take responsibility for the outcome and will not be there to manage a complication. This is not a reason to avoid these tools, which are not going anywhere, but it is a reason to use them for the right task. A good question to put to an AI is what the limitations of a body of evidence are, what the primary endpoint of a study was, whether it was controlled, who funded it, and whether it applies to skin with melasma or acne or a darker Fitzpatrick type. A poor question, and the one most people actually ask, is simply what they should use on their face, because that question quietly bypasses the diagnosis on which a safe answer depends.
What this means for how we work
The role of a good clinic is not to provide access to treatments, since access has never been easier, but to interpret evidence, diagnose accurately, choose appropriate interventions, decline inappropriate ones and explain uncertainty honestly. This is why we do not practise menu medicine at Self London, and why we resist the model in which a patient selects from a list of treatments before anyone has understood the underlying problem. Pigmentation is not one diagnosis. Acne scarring is not one entity. Redness may be vascular, inflammatory, barrier-related or rosacea-driven, and skin ageing is never simply wrinkles, but an interacting set of changes across pigment, texture, laxity, vascularity and structure that has to be read before it can be treated. The treatment follows the diagnosis, and not the other way around.
The same logic governs how we think about skincare and supplements, where the right question is never merely whether an ingredient has studies behind it, but what those studies actually show and whether they apply to the person asking. For one patient a simple moisturiser and a reliable sunscreen will do more than any expensive active. For another, a prescription retinoid is appropriate, while for a third the barrier needs stabilising before any active can be tolerated at all. For someone with melasma the wrong brightening regime makes things worse, for someone with rosacea a product marketed as barrier-strengthening may still sting, and for someone considering supplements the honest conversation includes medical history, interactions, cost and a realistic account of what the evidence can and cannot promise. None of this is as satisfying as a confident claim, and all of it is safer, more intelligent and more genuinely useful.
The real problem with doing your own research
The phrase “doing your own research” has become loaded, and at its best it describes something admirable, namely curiosity, engagement and a wish to make informed decisions, which is exactly the kind of patient we most enjoy working with. The trouble is not that patients read, because patients should read, and an engaged patient is almost always easier to help than a passive one. The trouble is that reading alone does not supply context, and context is most of what turns information into judgement.
A patient may know perfectly well that hyaluronic acid improves hydration without appreciating the difference between a topical humectant, an oral supplement and an injectable skin booster, which are three quite different propositions wearing the same name. They may know that collagen declines with age without understanding why swallowing collagen peptides is not the same as rebuilding facial dermal collagen, or that retinoids stimulate collagen without grasping the irritation, the pigmentary risk, the pregnancy restrictions and the need to introduce them carefully, or that lasers remodel collagen without any sense of why wavelength, fluence, density, downtime and skin type make one device entirely unlike another. Information is not judgement, and in an era when wellness culture, beauty marketing and AI summaries all conspire to make complex biology sound simple, the gap between the two has rarely mattered more, because skin very seldom behaves as simply as the claim suggests.
A more intelligent way to be an informed patient
The most useful patients are not the most compliant ones, and certainly not the most passive, but the most genuinely curious, the ones who ask questions, who are open to nuance, who accept that not every treatment will suit them and that a responsible clinician will sometimes say no. That posture is very different from arriving with a conclusion already formed and asking the clinic to ratify it, and the distinction tends to predict who does well.
So the next time you meet a skincare or supplement claim, it is worth pausing before you accept the headline, and asking what the study actually showed. Was it tested in humans, was there a control group, how many people took part, how long did it run, what was the primary endpoint, and was that endpoint something visible or merely something measurable. Were multiple outcomes tested, and was the analysis corrected for that. Who funded it. Did the participants resemble you, and were darker skin types or people with acne, rosacea or pigmentation included or excluded. Was the effect large enough to see, and did it last, and did the final claim stay within the limits of what the evidence could support. These questions do not make you cynical. They make you literate, and good science does not collapse under careful questioning. It becomes clearer.
The deeper point is that skincare science is not useless, and the answer is emphatically not to dismiss all of it, because some products help, some ingredients are genuinely well supported, some supplements have modest value for selected people, and some cosmetic studies are careful, thoughtful and biologically interesting. The answer is to translate properly, which means letting a hydration signal be called hydration, a biomarker be called a biomarker and a twelve-week study be treated as a twelve-week study, and refusing to let a small trial become a universal truth, a subjective impression masquerade as an objective endpoint, a control-free comparison pass as proof, or a meta-analysis conceal the weakness of the studies inside it.
A patient does not live inside a confidence interval. The questions that actually decide whether a product was worth it are whether the result was visible, whether it lasted, whether it was tolerable and safe, whether it justified the cost, and whether it matched what was promised, and none of those questions is answered by the word “significant.” The skill this article has been describing is simply the ability to hold a claim up against the evidence behind it and see whether the two are describing the same thing, and it is a skill anyone can learn, because it is built from the handful of recurring tricks set out near the start rather than from any specialist knowledge of skin. Learn to spot the missing control, the surrogate standing in for the outcome, the dozen endpoints quietly mined for one success, the short study carrying a long claim and the effect that survives only in the studies someone paid for, and most of the noise resolves itself. What remains, the genuinely useful science, is quieter and more modest than the marketing, and considerably more trustworthy for it.
This article is intended as general education and not as individual medical advice. If you would like an assessment tailored to your own skin, history and goals, our consultant-led team at Self London is glad to help.
Selected references
Hughes MCB, Williams GM, Baker P, Green AC. Sunscreen and prevention of skin aging: a randomized trial. Annals of Internal Medicine 2013;158(11):781–790.
Watson REB, Ogden S, Cotterell LF, Bowden JJ, Bastrilles JY, Long SP, Griffiths CEM. A cosmetic ‘anti-ageing’ product improves photoaged skin: a double-blind, randomized controlled trial. British Journal of Dermatology 2009;161(2):419–426.
Myung S-K, Park Y. Effects of collagen supplements on skin aging: a systematic review and meta-analysis of randomized controlled trials. The American Journal of Medicine 2025;138:1264–1277.
Shu Q, et al. Effectiveness of dietary supplement for skin moisturizing in healthy adults: a systematic review and meta-analysis of randomized controlled trials. Frontiers in Nutrition 2022;9:895192.
Advertising Standards Authority. Ruling on Kollo Health Ltd, 22 November 2023.
Advertising Standards Authority. Ruling on HealthArena Ltd (Dermacoll), 19 June 2019.
