“No more effective than placebo”

The gold standard in scientific research is the randomized control trial, whereby subjects are randomly allocated to receive either an intervention, such as a medical treatment or a fitness programme, or a placebo. If the subjects in the intervention group change to a greater extent that those in the placebo group, then the intervention can be said to work. If the two groups are similar, then the intervention can be said to be “no more effective than placebo”. This is often taken to mean that the intervention doesn’t work, but of course, it doesn’t mean that.


Why? Because placebos can be effective. It’s well established in the medical research, particularly around pain, that placebos can exert some biological influence; indeed, around one-third of patients receiving a placebo report improvements in pain, which is essentially the same number of subjects who report that morphine does not reduce their pain. The mechanisms underpinning this relationship are both complex and, often, quite poorly understood, but as a lay summary placebos can work through a number of ways, including expectancy (you think something will help, so it does), and regression to the mean (the person would get better anyway, but providing a placebo makes them feel like they are doing something positive about it). Given the complex biopsychosocial nature of pain, in which stress and anxiety can exacerbate feelings of pain and disability, reducing this stress and anxiety in and of itself can reduce the pain; and this can be achieved readily through placebo. I can think of two anecdotes (which I’m aware are the lowest form of evidence) from my life that illustrate this. The first is that my back pain was always greatest on days when I had a competition. It seems unlikely that my underlying back issues were, for some reason, much worse of days when I was competing; instead, it seems more likely that I was more stressed and anxious on my competition days, and as a result interpreted the signals from my back as more threatening, which in turn increased my feelings of pain.


The second example is something that a researcher would struggle to get past an ethics board, such is the measure of deception and subterfuge used by the principle investigator – in this case, my mum. Early in my teenage years, I developed an entirely irrational fear of flying, and, in particular, turbulence. This made the flight from the UK to Florida for a family holiday somewhat unpleasant, both for me, and, I imagine, for my parents. Fortunately, whilst we were in Florida, my mum managed to track down some anti-anxiety medicine (don’t ask me how) that I could use on the plane; whenever I felt a bit scared, I could put a lozenge in my mouth an allow it to slowly dissolve, easing my anxiety. The lozenges themselves were very nice, which an aniseed flavor that, whilst not to everyone’s liking, is certainly to mine. As such, the long return flight how was an altogether calmer affair, which my anxiety been much lower, and I was very grateful that my mum had managed to locate these tablets which eased my anxiety. It turns out that these were just aniseed sweets she had bought at a sweet shop.


My point is two-fold – firstly, that many of our physical feelings have a psychological components, and, secondly, that by targeting this psychological component we can influence our physical sensations. And there is where placebos, and their close cousin expectancy, come in.


Let’s look at caffeine in sport. Caffeine is definitely performance enhancing for most people, in most sports, most of the time, to the point where around 75% of athletes consume it specifically as an ergogenic aid. Get a group of cyclists, tell them that they are being given caffeine, and their performance improves – even if you actually gave them a placebo. Conversely, if you consume caffeine but think that you haven’t, then your performance improves to a lesser extent than if you correctly determined you had consumed caffeine.


It’s clear, therefore, that both placebo and expectancy can have real-world impacts on a number of measures. Part of this is the theatre surrounding the treatment; in many of the studies utilizing placebos in medicine, the very act of seeing a well-qualified human being who (hopefully) takes our problems seriously is likely to have a positive impact in and of itself. Similarly, taking a substance you know will improve your performance will cause you to work harder, even if you haven’t actually taken the substance.


All of this brings me to my key point; when you see “no better than placebo” written in a study, don’t be tempted to assume that this means that the substance doesn’t work. Instead, understand that it might well work, but that its effect is likely related to other aspects outside of the particular interventions physiological impact. Secondly, whilst it’s unethical to deliberately deceive athletes, understand that their beliefs will impact their performance. They may well be utilizing something for which there is no evidence of effectiveness, but, if they believe in it, it may well be more effective that doing nothing. Obviously, you should prevent them from doing something harmful, but they key is not to try and move them away from behaviors for which is no/limited evidence, if they think it works. Because, if they think it works, it just might.


Caffeine, CYP1A2 genotype, and exercise performance: A role for timing?


A recent paper has made some quite significant waves, both in the science and lay communities, by exploring the effect of a polymorphism in a gene called CYP1A2 on the ergogenic effects of caffeine on performance. The paper (Caffeine, CYP1A2 genotype, and endurance performance in athletes) was published in Medicine and Science in Sports and Exercise, where the authors put 101 competitive male athletes through a 10km cycle ergometer time trial under three experimental conditions; no caffeine (i.e. placebo), 2mg/kg caffeine, and 4mg/kg caffeine. The main finding was that caffeine was ergogenic for subjects with the AA version of the gene, likely neutral for people with the AC version, and potentially ergolytic (i.e. harmful to performance) for those with the CC version of this gene in the 4mg/kg caffeine trial (and neutral in the 2mg/kg trial).


This is what I wrote about CYP1A2 in my 2017 paper on inter-individual variation in caffeine ergogenicity:


The gene CYP1A2 encodes cytochrome P450 1A2, an enzyme responsible for up to 95% of all caffeine metabolism [35]. A SNP within this gene, rs762551, affects the speed of caffeine metabolisation. AA homozygotes (“fast” metabolisers) tend to produce more of this enzyme, and therefore metabolise caffeine more quickly. Conversely, C allele carriers (“slow metabolisers”) tend to have slower caffeine clearance. The variable effects of this SNP are most well-established in regards to health, with myocardial infarction and hypertension risk increased in slow metabolisers consuming moderate (3-4 cups) amounts of coffee, whilst fast metabolisers exhibit a protective effect of moderate coffee consumption.


These earlier medical studies prompted research into how the CYP1A2 polymorphism might modify the ergogenic effects of caffeine. Womack et al. put thirty-five trained male cyclists through two 40-km cycle time trials, following consumption of either 6 mg/kg of caffeine or placebo 60-minutes beforehand. There was a significant effect of CYP1A2 genotype on the ergogenic effects of caffeine, with AA genotypes (fast metabolisers; 4.9% improvement) seeing a significantly greater performance improvement than C allele carriers (slow metabolisers; 1.8% improvement). Within AA genotypes, caffeine improved performance by at least one minute for 15 out of 16 subjects, whilst in C allele carriers only 10 of 19 subjects saw an improvement greater than one minute. These findings allowed the authors to conclude that caffeine has a greater ergogenic effect for CYP1A2 AA genotypes than C allele carriers.


Since this initial paper, a small number of subsequent studies have been published. The same group published a paper hampered by a lack of CC genotypes, putting 38 recreational cyclists through four 3km time trials under different experimental conditions; placebo mouth rinse + placebo ingestion, placebo mouth rinse + caffeine ingestion, caffeine mouth rinse + placebo ingestion, and caffeine mouth rinse + caffeine ingestion [40]. Both AC (4.1%) and AA (3.4%) genotypes saw performance improvements in the combined caffeine mouth rinse and ingestion trial, but only AC (6%) genotypes saw a performance improvement in the caffeine ingestion trial. The conclusion was that AC genotypes saw greater performance enhancement with caffeine ingestion, in contrast to Womack et al. One potential cofounder was the short exercise trial duration (c.5 minutes), as caffeine shows greater ergogenic effects in events of longer duration. A second potential cofounder is that Womack et al. utilised trained subjects, whilst Pataky et al. did not. Exercise appears to increase CYP1A2 expression, such that trained and untrained subjects may metabolise caffeine differently. Algrain et al. reported no modifying effect of the CYP1A2 polymorphism on the ergogenic effects of caffeine; however, they noted the small subject number (n=20), the untrained status of these subjects, and the lower caffeine dose (approximately 255mg). Klein et al. and Salinero et al. found no effect of the CYP1A2 polymorphism on the effects of caffeine on tennis and Wingate test performance respectively, although with modest sample sizes (n=16 and 21).”


So, given the highly equivocal results of the previous research, Guest and colleagues study is both timely and exciting. Of all the studies exploring CYP1A2, caffeine, and performance, it has by far the largest sample size, which is important in studies exploring genetics, as the effect sizes of each individual gene can be quite low – meaning that large numbers of people are required to determine a “true” effect. It’s also the first study (to my knowledge) to specifically show that caffeine is harmful for individuals with a certain property, in this case the CC genotype of CYP1A2. Prior to this, we knew some people didn’t find caffeine ergogenic, but we weren’t sure what they had in common.


Alongside Guest’s study was another one (The effect of CYP1A2 genotype on the ergogenic properties of caffeine during resistance exercise: a randomized, double-blind, placebo-controlled, crossover study) that focused on resistance training. Here, 30 resistance-trained males undertook 3 sets to failure at 85% of 1RM in four different exercises (bench press, leg press, seated cable row, and shoulder press). Those with the AA genotype could perform more reps compared to placebo when they had 6mg/kg of caffeine pre-test; whilst those with the C allele saw no performance improvement – but, crucially, also no performance decrement – when consuming caffeine. The results of this study got considerably less fanfare than Guest’s results, and I’m not sure why, but suffice to say we now have a decent idea that caffeine appears to not be ergogenic for CYP1A2 C allele carriers when it comes to both resistance and endurance training. Furthermore, it might even be harmful for those with the CC genotype, but, because very few people actually have the CC genotype (approximately 10%), most studies just group the CC and AC genotypes together into one group.


So, do we really believe that caffeine has no performance benefits for the ~50% of people who have at least one C allele? Personally, I don’t, in part because I have the AC genotype and I use caffeine to enhance my performance, and cognitive dissonance means I hope I am not wasting my money. To be clear, I don’t think Guest or Rahimi, or their co-authors, have committed scientific fraud by making up their results (and I think their studies are great), but I do think that there may be a way for caffeine to be ergogenic for these people. This is what I wrote in a letter to the Irish Journal of Medical Science on the topic:


The results, however, promote significant cognitive dissonance. As an AC genotype who used caffeine extensively as an ergogenic aid during my professional athletic career, do I really believe that caffeine had no performance benefits for me? It’s hard for me to reach such a conclusion, even though the results of both Rahimi and Guest and colleagues’ research suggest as much. I’m sure many other practitioners feel the same, and indeed perhaps worry that the standard caffeine recommendations are harmful, or at best neutral, for a significant proportion of the athletes they work with.


The mechanism proposed by Guest et al. is that, because C allele carriers metabolise caffeine at a slower rate than AA genotypes, they experience prolonged vasoconstriction, which is likely to be performance limiting in endurance events where the transfer of oxygen and nutrients to the working muscle is crucial. Additionally, Womack and colleagues speculated that the metabolites of caffeine – paraxanthine, theobromine, and theophylline – have additional ergogenic effects; in this case, the presence of these ergogenic substances would be lower in C allele carriers at a given time point due to the slower metabolization of caffeine, reducing caffeine’s performance benefits. This proposed mechanism was echoed by Rahimi.


If the above mechanisms are indeed correct, then there remains the possibility that caffeine can still be ergogenic for C allele carriers, but that such individuals need to consume it a greater amount of time prior to exercise. In the majority of studies – including Guest et al. and Rahimi – subjects consume caffeine ~60 minutes pre-exercise. However, for C allele carriers, could the ergogenic effects of caffeine be restored by utilizing a caffeine dose 90 or 120 minutes pre-exercise? Such an hypothesis is, of course, speculative, and requires testing – but it does represent a potential way by which caffeine can indeed be ergogenic for all. The resolution of whether caffeine is truly ergolytic or neutral for CYP1A2 C allele carriers, or if it merely necessitates a different caffeine strategy, represents an important step on the journey towards more personalized sports nutrition guidelines.”


The next step is obviously to test this under experimental conditions; watch this space!