Does finasteride regrow hair, or does it hold the line?

Most answers to this question are a sentence and a photograph. The trials behind the licence measured it four different ways over five years, and the four measures disagree in an instructive direction: the drug looks best when you compare it to what happened to the men who did not take it, and least impressive when you ask it to produce new hair on demand.

The clearest measure: photographs, rated blind

An independent panel rated standardised photographs of the head without knowing who had taken what. At five years, 48% of the men on finasteride were rated as having an increase in hair growth, 42% as having no change — described in the label as no further visible progression of hair loss from baseline — and 10% as having lost hair compared with the start. On placebo the same panel rated 6% as increased, 19% unchanged and 75% as having lost hair.

Now read the two columns as one picture. Roughly half the treated men gained visible hair; roughly nine in ten avoided visible further loss. Among untreated men, three quarters visibly lost. The drug's dominant effect over five years, on this measure, was preventing the slide rather than reversing it.

The counted measure: hairs in a circle

Hair counts were taken from photographic enlargements of a fixed patch of scalp — a circle one inch across, about 5.1 square centimetres. At twelve months the difference between the treated and placebo groups was 107 hairs in that circle. At two years it was 138 hairs. At five years, 277 hairs.

That widening gap is the single most misread statistic in this subject, because it looks like accelerating regrowth and is not. The label spells out what is happening: in treated men the maximum improvement over baseline was reached during the first two years, after which a slow decline set in, while hair count stayed above baseline through year five. The gap kept widening because the placebo group was falling faster than the treated group was drifting down.

The avoided-loss measure

The trials also counted how many men had any decrease in hair count from baseline. At twelve months that was 58% of the placebo group against 14% of the treated group. At two years, 72% against 17%. At five years, 100% of the placebo group had lost hair against 35% of the treated group.

That last pair is the plainest sentence in the whole dossier. Every untreated man in the study lost hair over five years. About a third of the treated ones did.

The judgement measures, which flatter the drug

Investigators rating scalp hair on a seven-point scale called 65% of treated men improved at twelve months against 37% on placebo, 80% against 47% at two years, and 77% against 15% at five years. Men rating themselves reported more hair, less loss and better appearance, with an overall improvement over placebo visible as early as three months.

Treat unblinded judgement measures with care in any trial. The blinded photograph panel and the hair counts are the harder evidence, and they are the more modest ones.

One study that counted growing hairs

A 48-week study in 212 men counted total and actively growing — anagen — hairs in a one-square-centimetre target area. Treated men gained 7 total and 18 anagen hairs; placebo men lost 10 and 9. The proportion of hairs in the growing phase moved from 62% at the start to 68% in the treated group.

Shifting a few percent of follicles back into a growing phase is a fair description of what this drug does mechanically. The pathway page explains why that follows from blocking one enzyme.

What the trials cannot tell you

• Whether it works for you. The participants were men aged 18 to 41 with mild to moderate loss, 88% of them white, and the label reports smaller mean hair-count changes in the smaller Black, Asian and Hispanic subgroups. Your starting point is probably not theirs.
• Anything about the temples. The label says efficacy in bitemporal recession has not been established, and the anterior mid-scalp study deliberately excluded the hairline and the receding corners from its counts.
• Anything about women. The one trial in the label, in 137 postmenopausal women, could not demonstrate effectiveness on any measure.
• What happens at year ten. The extensions ran five years. Beyond that the honest answer is that this evidence base stops.

What "working" means, and why people disagree about it

Much of the argument about whether this drug works is really an argument about the word. Four definitions are in circulation and the trials answer them differently.

• Visible new hair where there was none. The weakest claim the data supports: roughly half the treated men were rated as increased by the blinded panel at five years, and the gains peaked in the first two.
• A measurable increase in count. Supported, in a fixed one-inch circle, with the treated group staying above their own baseline through year five.
• Not getting worse. The strongest result in the dossier, and the one the photographs make plainest: 35% of treated men had lost hair by five years against 100% of untreated men.
• Looking better. Not measured by any instrument, measured instead by men rating themselves and by investigators rating scalps, both unblinded and both more generous than the photographs.

Two people can read the same table and come away with "it works" and "it barely does" because they are using different definitions, and neither is lying.

The part of the trial design that is easy to miss

Every man in these studies, on drug or on placebo, was told to use a specified medicated tar-based shampoo for the first two years — a design choice made to stop seborrheic dermatitis confounding the assessment of hair growth. It is a small detail with a useful implication: the comparison is between drug and no drug on top of a common baseline routine, not between a treatment and doing nothing at all.

So: regrow or hold?

Both, unevenly, and mostly hold. About half of treated men gained visible hair on the blinded panel; the larger group either gained nothing visible or avoided visible loss; a minority lost anyway. If the expectation is a restored hairline, the trial data does not support it. If the expectation is that the line stops moving for a few years, it does — for most, at a cost set out in full on the side-effect record, and with the whole medicine described on the finasteride page.

One more thing the same trials measured, which belongs to this question: stop, and the effect reverses. What happens when you stop has the numbers.