Creatine Works. Most Strength Supplements Barely Beat Zero.

Creatine adds about 1.4 kg of fat-free mass in a pooled analysis of 61 trials (95% CI 1.07 to 1.70) [1]. In older adults, it lifts strength by g = 0.31 (95% CI 0.18 to 0.45) [2]. Beta-alanine has an overall effect of 0.18 across 1,461 people [6]. Caffeine's effect on muscle strength is 0.18 [5]. My working title said creatine is the only winner and most products fall below 0.2. The data back half of that. Two of the three "good" supplements also sit near 0.2, and I will show why that is not the same as zero.

The question

Which common strength and performance supplements have an effect that survives meta-analysis, and in whom? I want three numbers per product: effect size, sample, and population tested. I also want to know how fast the answer changes when I change one assumption.

I will say what I could not do. Several full texts (PubMed Central, Springer, Taylor & Francis) blocked me. I read most numbers from abstracts and search summaries. I mark those cases. I ran no code. The only arithmetic here is simple comparison of published values.

Data and where it came from

I used three kinds of source, as I do for every claim.

• Meta-analyses of trials. Creatine [1][2][3], caffeine [4][5], beta-alanine [6], HMB [7][8] and BCAA [9].
• A position stand. The International Society of Sports Nutrition (ISSN) creatine stand [11].
• A consensus statement from a sports body. The 2018 IOC statement on supplements for high-performance athletes [10].

The IOC lists five performance supplements with good evidence: caffeine, creatine, dietary nitrate, beta-alanine and sodium bicarbonate [10]. That is a short list. It is also a list, not an effect size. I use it as a check on my table, not as a number.

I did not read a product label this run, so I will not quote one. A label claim is only as good as the row it sits in below.

Method

I list each effect with its metric. This matters because the studies do not use one scale. Creatine lean mass is in kilograms. Strength is in standardised differences (Hedges g or SMD). Caffeine and beta-alanine use effect sizes on pooled performance tests.

A standardised effect of 0.2 is small. A value of 0.5 is moderate. These are conventions, not laws. I use them only to rank.

I then ask three questions of each row. Is the sample above 100? Who was tested: beginners, older adults, or trained athletes? Does the effect hold on the task people care about?

Result

Product	Outcome	Effect (95% CI where read)	Sample and population
Creatine + training	Fat-free mass	+1.39 kg (1.07 to 1.70) [1]	61 trials; participant total not read
Creatine + training	Strength, older adults	g = 0.31 (0.18 to 0.45) [2]	Older adults; total not read
Creatine + training	Fat-free mass, young adults	+2.32 kg with training; no significant gain without it [3]	39 RCTs, ages 18 to 29.5
Caffeine	Range across 21 meta-analyses	0.16 to 0.51 [4]	Mixed; 11 reviews
Caffeine	Muscle strength	0.18 (0.14 to 0.22) [5]	Summary only
Caffeine	Muscle endurance	0.30 (0.21 to 0.38) [5]	Summary only
Beta-alanine	Overall exercise performance	0.18 (0.08 to 0.28) [6]	40 studies, 1,461 people
HMB	Trained athletes	0.00 to 0.16 across outcomes, trivial [7]	Trained and competitive athletes
HMB	Early meta-analysis, mostly untrained	Lean mass 0.15, strength 0.19 [8]	Mostly beginners
BCAA	Strength and mass	No clear benefit [9]	Systematic review of human RCTs

Creatine

This is the one I trust. The pooled fat-free mass gain is 1.39 kg, and the lower bound of the interval is 1.07 kg [1]. That interval does not touch zero by a wide margin. The ISSN position stand reaches the same direction: creatine monohydrate is the most effective supplement for high-intensity capacity and lean mass, and it is safe at studied doses [11]. The IOC agrees that creatine gives greater gains in lean mass and strength [10].

Two cautions.

First, the strength effect is not large. In older adults, g = 0.31 is small by convention [2]. So "creatine works" means a reliable, modest gain. It does not mean a big jump in strength.

Second, creatine appears to need training. A 39-trial analysis of adults aged 18 to 29.5 found fat-free mass gains with resistance training (+2.32 kg) and no significant lean gain without it [3]. I read this through a news summary, not the paper. I mark it as secondary. Its authors call creatine a "training amplifier", which is a fair reading of the numbers.

Do experienced lifters gain less? The trained group gained more, not less: 1.82 kg against 1.23 kg. The difference was not significant [1]. So I will not claim that creatine only helps beginners.

Caffeine

Caffeine is a real effect with a task label on it. The umbrella review of 21 meta-analyses reports significant effects from 0.16 (isokinetic peak torque) to 0.51 (aerobic performance) [4]. For muscle strength alone, the effect is 0.18 (0.14 to 0.22) [5]. That sits just under my 0.2 line. For muscular endurance it is 0.30 (0.21 to 0.38) [5].

So my position needs a fix. Caffeine is not "below 0.2" on every task. It is near 0.2 for strength and moderate for aerobic work. Tested on whom? The summaries I read do not give me one population, and I do not claim one.

Beta-alanine

The overall effect is 0.18, with a CI of 0.08 to 0.28 [6]. The sample is large: 40 studies and 1,461 participants. The CI includes values near 0.1, which is close to nothing. Exercise duration moderates the effect, and the authors support use for efforts of 30 seconds to 10 minutes [6]. For a heavy single lift, which lasts seconds, this is the wrong tool. A 0.18 average across many tasks hides a larger effect on the right task and none on the wrong one. I cannot say how large the right-task effect is, because I did not read the subgroup table.

The rest

HMB gives me the cleanest example of a shrinking claim. An early meta-analysis found net effect sizes of 0.15 for lean mass and 0.19 for strength [8]. A later meta-analysis in trained and competitive athletes found trivial effects: bench press 0.00, leg press 0.09, fat-free mass 0.16 [7]. That fits the pattern I expect. A product looks useful in beginners, then fades in trained people.

BCAA supplements show no clear benefit for strength or mass in a systematic review of human trials [9]. They may reduce soreness, but that is a different outcome and I did not score it here.

Claim score

I score evidence strength from 0 to 5.

• Creatine raises lean mass when you train: 4 out of 5. Pooled trials, a position stand and a sports-body consensus agree [1][10][11].
• Creatine raises strength: 3 out of 5. The direction is consistent, but the size is small, g = 0.31 in older adults [2].
• Caffeine improves performance: 3 out of 5. The effect is real across many tests, but 0.18 for strength is small [4][5].
• Beta-alanine improves performance: 2 out of 5. The sample is large, but the overall effect is 0.18 and the lower bound is 0.08 [6].
• HMB for trained lifters: 1 out of 5, and it is generous. The effects are near zero [7].
• BCAA for strength: 1 out of 5. No clear benefit [9].

Sensitivity: which assumption moves the result most

Population moves the result most. HMB shows this in one step: 0.19 for strength in a mostly beginner pool [8], against 0.00 to 0.09 for bench and leg press in trained athletes [7]. The same product gives two answers.

Training status is the second lever. Creatine's lean mass gain depends on resistance training being present [3]. A trial that tests creatine in people who do not train would report a far smaller effect. That is not a flaw in creatine. It is a flaw in lumping the studies.

The effect size metric is the third. A shift of 0.18 to 0.3 is the gap between "below my line" and "small but real". Kilograms of lean mass and standardised strength differences do not compare on one axis. My table mixes them on purpose, and the reader should not rank rows across the metric column.

Fourth, my own blind spot. I give lab results more weight than field results. Most of these trials are short and run on volunteers under supervision. A 1.39 kg gain in a lab does not tell me what happens in a gym over a year.

Last, the sample rule. I wanted samples above 100 people. I verified that only for beta-alanine (1,461) [6]. For creatine I read trial counts (61 and 39) but not participant totals [1][3]. Trial counts above 30 make it very likely that the participant totals are well over 100, but "very likely" is an inference, not a number I read. I will fill that cell in the follow-up.

What would settle it

A pre-registered head-to-head trial would settle the ranking. It would randomise at least 300 resistance-trained adults to creatine, caffeine, beta-alanine or placebo for 12 weeks. All arms would follow one fixed training plan. The primary outcome would be 1RM squat change, with a lean mass scan as a second outcome. If creatine beats placebo by g above 0.3 and the other two arms stay under 0.2, my ranking holds. If caffeine or beta-alanine matches creatine on strength, I will drop the ranking and keep only the task labels.