What actually works for antidepressant sexual dysfunction?

And why open-label trials of these treatments report success three times as often as placebo-controlled ones.

Are you finding it in the first place?

And asking is only half of it. The instruments the entire quantitative literature rests on — ASEX, CSFQ and PRSexDQ — were built around the desire, arousal and orgasm cycle. Three of the complaints patients actually describe are not in any of them.

The three the instruments miss are the three that most distinguish persistent dysfunction from ordinary treatment-emergent dysfunction. A patient describing genital numbness is describing something the literature has no denominator for — which is a reason to ask about it by name, not a reason to doubt it.

What has actually been tested?

The Cochrane review of strategies for antidepressant-induced sexual dysfunction enumerates watchful waiting, behaviour change, dose reduction, delayed dosing, switching, drug holidays, and adjuvant bupropion, buspirone or a PDE5 inhibitor — and notes that most of them have never been tested in a trial. Sorting them by what has been is the useful exercise.

Why does bupropion work in some trials and not others?

Four data points, and they do not line up as neatly as the usual summary suggests.

So once-daily failed — but one twice-daily trial failed its primary endpoint too, and it is the modest trend that drives the heterogeneity in the pooled estimate. A meta-analysis of 859 women found the twice-daily regimen improved desire, arousal and orgasm, at low GRADE quality.

So 300 mg a day is the reasonable target. But be clear with yourself about what that inference is: it comes from comparing separate trials, not from a dose-ranging study. No dose-ranging randomised trial and no dose meta-regression exists. The Cochrane pooled figure of 1.60 itself mixes both doses across three trials, with heterogeneity of 86% driven by two large effect sizes against one modest trend.

Should you switch the antidepressant?

That said, if you are going to switch, one target is better supported than the rest — not by switching trials, but by head-to-head comparison. Two identically designed eight-week double-blind trials randomised patients with normal baseline sexual function to bupropion XL, escitalopram or placebo.

A second randomised trial of 360 patients found significantly more orgasmic dysfunction on sertraline than on either bupropion or placebo. And a placebo-controlled imaging study in 18 healthy men found paroxetine both increased subjective dysfunction and blunted reward-region response, while bupropion did not.

Does sildenafil work in women?

It is also one trial. Broader reviews of female sexual dysfunction rate PDE5 inhibitor evidence in women as limited, and practice in this area is largely extrapolated from men. Offering it is reasonable. Presenting it as established is not.

Will it settle on its own?

The usual reassurance is that these effects settle with time. The study behind that claim followed 108 patients started on SSRIs, of whom 26 developed clinically relevant orgasm delay. Here is where those 26 were at six months.

High baseline severity significantly predicted failure to remit. Note the bottom row: the report characterises 16 of the 26, so the commonly quoted "about a third get better" rests on a study that does not account for the rest.

So the honest version is not that it settles. It is that it settles fully for about a third, partly for another sixth, and not at all for a sixth — and the patients whose symptoms are worst at the outset are the ones least likely to be in the first group. That last clause is the part the standard reassurance leaves out, and it is the part that changes what you tell the person in front of you.

Why does so much of this look better than it is?

A network meta-analysis pulled together 57 citations covering 27 randomised trials, 27 open-label trials and 3 crossover studies — 66 treatment arms, 33 interventions, 3,108 patients.

Note what that is and is not. It counts two different sets of trials across all 33 interventions; it does not track the same treatments from one design into the other, so it does not give you a conversion rate for any particular drug. What it does give you is a calibration: in this field, open-label results should be read as provisional until a placebo arm has had a go at them.

So what do you actually do?

1. Ask, with the symptoms named. In the fluoxetine registration trials, where patients had to volunteer it, decreased libido was reported by 4%. A structured interview using a validated questionnaire, in 1,022 outpatients with previously normal function, found 59%. Name the domains out loud — desire, erection or lubrication, time to orgasm, pleasure at orgasm, and genital sensation. The last two are not captured by any standard instrument and patients almost never volunteer them.
2. Work out whether it is the drug. Delayed orgasm separates clearly from placebo in randomised data. Reduced desire does not — in trials it is not cleanly distinguishable from the depression itself.
3. Exclude the rest. Testosterone and prolactin if indicated, thyroid, glycaemic status, alcohol, co-prescribed antipsychotics — and a cardiovascular assessment in a man with new erectile dysfunction, which is the item most likely to matter to his life expectancy.
4. Reduce the dose if the depression is fully remitted and the continuation phase is complete. Lowest-cost move, reversible, and it carries relapse risk, which is why the sequence matters.
5. Augment before switching if the antidepressant is working. Bupropion at 300 mg a day; sildenafil in men.
6. If you switch, switch to bupropion — and say that you are acting on head-to-head comparison rather than on a switching trial, because no switching trial for this indication exists.
7. Do not promise that stopping fixes it. Ordinary treatment-emergent dysfunction is expected to improve after stopping, but no controlled study gives a percentage resolved by any timepoint, and a small minority report symptoms that persist.