Spironolactone: the investigators' grading was strong; the patients' own ratings would not pool

The SAFA trial in Health Technology Assessment (2024) and a meta-analysis in J Cosmet Dermatol (2025).

SAFA first.

A pragmatic, parallel, double-blind, randomised superiority trial, recruiting from primary and secondary healthcare and from community settings.

The participants were women aged 18 and over with facial acne persisting at least 6 months, judged to potentially warrant oral antibiotic treatment. Of 1,267 assessed, 410 were randomised 1:1 (201 to spironolactone, 209 to control); 342 were in the primary analysis.

• dose — 50 mg a day, or matched placebo, until week 6, increasing to 100 mg a day until week 24
• participants continued their usual topical treatment. At baseline 82.9% were using topical treatments
• mean age 29.2; 7.9% were from non-white backgrounds
• baseline severity — mild 46%, moderate 40%, severe 13%
• over 95% in both groups tolerated the treatment and increased their dose

What the investigators recorded.

• Investigator's Global Assessment judged successful at week 12 — 31 of 201 (18.5%) on spironolactone, 9 of 209 (5.6%) on placebo
• adjusted odds ratio 5.18 (95% CI 2.18–12.28)

What the participants reported, alongside.

• reporting overall acne improvement — 72.2% vs 67.9% at 12 weeks (adjusted OR 1.16, 0.70–1.91); 81.9% vs 63.3% at 24 weeks (2.72, 1.50–4.93)
• satisfaction with treatment improved — 70.6% vs 43.1% (3.12, 1.80–5.41)

Now the meta-analysis.

It pooled five randomised trials, 563 patients, comparing oral spironolactone with placebo in women with acne (251 on spironolactone, 42.9%).

• objective assessment of acne improvement — OR 6.59 (95% CI 3.50–12.43), p < 0.00001, I² = 0%
• trial sequential analysis stated the required sample size had been reached

How to read it.

First, the two overlap. The meta-analysis appeared in 2025 and pooled five placebo-controlled trials. SAFA is very likely among those five. This site does not count them as two independent sources. This badge rests on the five trials the meta-analysis pooled, and SAFA is listed separately because it is the largest, and the only publicly funded one.

Second, the endpoint is graded by investigators. IGA and objective assessment. This is not a trial with self-report only. This site does not award strong evidence on self-report alone.

Third, I² = 0%. There was almost no disagreement between the five trials. That is a rare figure compared with other meta-analyses on this site. Read it next to the 85% in the entry just below.

Fourth, the funding. SAFA was paid for by the NIHR Health Technology Assessment programme. This site's condition — no strong badge if everything is industry-funded — does not even come into play here. The drug is old, and the people who profit from selling it did not run this trial.

Fifth, this is an add-on. 82.9% of SAFA participants were already using topical treatment, and the trial told them to continue. So the number means 'when added on top of topical treatment', not on its own. The question this site keeps asking — what was it added on top of — has an answer here.

Sixth, look at who was measured. 7.9% from non-white backgrounds. And all women. This site writes that beside the numbers.

Seventh, one sentence of the meta-analysis's conclusion we do not carry over. The authors wrote that the drug should be elevated from 'off-label' use to an officially recommended standard of care. That is the authors' recommendation, not a regulatory fact. This site does not mix the two.

So it sits at 'strong evidence'. With the entry below restating that the badge hangs on an endpoint the investigators graded.

A secondary endpoint of the same 2025 meta-analysis. And the reverse of what this site usually writes.

The result.

• subjective assessment — OR 5.22, 95% CI 0.62–44.24, p = 0.13, I² = 85%
• the confidence interval contains 1. That is, 'no difference' lies inside it
• disagreement between trials was 85%

SAFA points the same way.

• participants reporting overall improvement — 72.2% vs 67.9% at 12 weeks, adjusted OR 1.16 (0.70–1.91) — not significant
• the same measure at 24 weeks was 81.9% vs 63.3%, 2.72 (1.50–4.93) — significant

How to read it, and why this entry has to be on the page.

First, this is the opposite of this site's usual sentence. Usually we write 'only the self-report moved and the instrument did not'. Here the investigators' grading is strong and consistent, and the patients' own ratings would not pool.

Second, what 0.62 to 44.24 means. That width means not 'it does not work' but 'the trials said different things'. An interval open as far as 44-fold is a mark of knowing little. I² = 85% says the same thing another way.

Third, why they might have split. Within what we checked, three explanations are possible.

• the trials asked differently — how subjective improvement was asked is not standardised. I² = 85% points that way
• what a doctor sees as improvement and what you feel are not the same — inflammatory lesions can fall while the face in the mirror feels less changed, because of leftover scars and pigmentation, which are different problems
• time — in SAFA, self-report was not significant at 12 weeks and was at 24. A pool containing short trials blurs

Fourth, this entry does not delete the badge above. The investigator-graded result stands. But the sentence 'it clearly improves in your own eyes as well' cannot be written from this material. This site puts the difference side by side on one page.

So on this endpoint it sits at 'not established'.

The primary outcome of the SAFA trial. This entry is about the target the trial set for itself.

The primary outcome was the adjusted mean difference in the Acne-Specific Quality of Life symptom subscale at 12 weeks. A score the participant writes.

• baseline mean — 13.0 (SD 4.7) across both groups
• week 12 — spironolactone 19.2 (6.1), placebo 17.8 (5.6). Adjusted difference 1.27 (95% CI 0.07–2.46)
• week 24 — spironolactone 21.2 (5.9), placebo 17.4 (5.8). Adjusted difference 3.77 (95% CI 2.50–5.03)
• secondary outcomes also favoured spironolactone at 12 weeks, with greater differences at 24 weeks
• the paper's conclusion — better than placebo, with greater differences at week 24 than week 12

How to read it.

First, the lower bound is 0.07. The value the trial itself chose as its primary outcome is almost touching zero. 0.07 is not 'nothing', but it is close to 'we very nearly measured nothing'. This site always records a primary outcome that only just cleared. It is the same kind of note as the ITT and per-protocol split on the HIFU body contouring page.

Second, the 24-week figure is clear. 3.77 (2.50–5.03). Same people, same trial, twelve weeks more.

Third, so this drug is slow. That matters practically. Twelve weeks is exactly when many people decide it is not working and stop. And in this trial the figures at 12 and 24 weeks differ.

Fourth, why this site does not report only the better number. Showing the 24-week figure alone erases where the primary outcome was. The primary outcome is the target fixed before the trial began, and the result at that target is the most trustworthy value. A better value at a later time point is a secondary result. The two cannot carry equal weight, and that is why this entry stops at 'moderate'.

Fifth, this endpoint is patient-reported. It is a quality-of-life score, the same kind of value as in the entry above, and this site does not raise a grade on that endpoint alone.

So it sits at 'moderate evidence'.