Does music help with anxiety before surgery?

Surgery is unusual among the places I read about, because the anxiety is expected and there's an established drug for it, which gives a trial something to compare against. The Cochrane Collaboration reviewed the music trials in 2013, and two larger meta-analyses have covered the ground since. I'll take them in order, because they don't say quite the same thing.

Who was in the 2013 review

Twenty-six randomised or quasi-randomised trials and 2,051 people, all adults scheduled for a surgical procedure, with samples from 9 to 327. In every trial the intervention was listening to prerecorded music, most often a single session of 20 to 30 minutes before the operation, and in all but one it was added on top of the usual care. The reviewers draw a line between that, which they call music medicine, and music therapy, where a trained therapist runs a tailored process with the patient. They looked for music therapy trials and found none they could include, so the review is entirely about headphones and a recording.

The numbers, and how the reviewers graded them

The main measure is the state half of the State-Trait Anxiety Inventory, a 20-item questionnaire scored from 20 to 80 that asks how you feel right now. Thirteen trials with 896 people between them used it, and in those the music groups came out 5.72 points lower than the standard-care groups after the session, with a confidence interval from 7.27 to 4.17. Control groups in those trials averaged somewhere between 37.6 and 44.4, so the difference is roughly a seventh of what the control groups were reporting. Seven more trials, 504 people, used other anxiety scales and pooled to 0.60 standard deviations in the same direction.

The reviewers graded both of those results as low certainty on the GRADE scale, which in their words means further research is very likely to have an important impact on their confidence in the estimate and is likely to change it. The reason is blinding. Twenty-one of the 26 trials didn't blind the people running them, and the outcome is a form the patient fills in about their own feelings while knowing which group they're in.

The body measures came out weaker still. Heart rate was 2.77 beats lower across 16 trials, but the trials disagreed with one another, and when the reviewers reran the analysis without the ten trials that had inadequate or unreported randomisation the difference stopped being significant. Diastolic blood pressure dropped about 2.4 points and systolic showed no clear effect. Breathing rate went the other way, one breath a minute faster with music, which the reviewers call clinically insignificant. Every one of those physiological results is graded very low.

The trial with the sedative in it

One trial in the set was built differently and the reviewers report it on its own. Bringman and colleagues in Sweden randomised 372 people awaiting elective surgery to either oral midazolam, a standard preoperative sedative given at 0.05 to 0.1 milligrams per kilogram, or a period of relaxing music with no sedative at all. The 327 who completed the protocol scored 30 on the state-anxiety scale after music and 34 after midazolam, and the drop from baseline was larger in the music group. Because midazolam was withheld from the music arm, the trial couldn't go into the pooled 5.72. The review's authors report it on its own and lean on it, alongside three earlier Cochrane reviews of music for anxiety in other medical settings, when they conclude that music may be a viable alternative to sedatives for this purpose.

I'd hold that more loosely than they do. It's one trial, the primary outcome is the same self-report form as everywhere else, and of the 372 people randomised, 45 are missing from the comparison. Most of those were called to theatre before the second questionnaire, but nine of them were midazolam patients who were too sedated to fill it in, which removes the people the drug worked hardest on from the anxiety score it's being compared against.

What the blank CD tells you

The obvious objection to all of this is that a person handed headphones by a nurse expects something to happen, and the form they fill in afterward records the expectation as much as the music. Three of the 26 trials gave the control group headphones too, with a blank CD or no sound, so that both groups had the same thing put on their heads. Only one of the three measured anxiety with the same questionnaire as the pooled result, a Chinese trial of 93 people waiting for laparoscopic surgery, and it is worth reading closely. The music group reported less anxiety than the silent-headphone group, so the questionnaire result held up against a blank CD in that one trial. The same trial also measured skin conductance, which tracks sweat and is used as a marker of arousal, and there the two groups came out the same. The reviewers had not planned a subgroup analysis by control design, so the review can't say how the pooled 5.72 would look across blank-CD trials in general. What the one trial that ran both measures shows is the same shape I found in the sleep trials, where the questionnaire moved and the instruments barely did.

Since 2013

Two larger reviews have come after. Hole and colleagues in the Lancet in 2015 pooled 73 trials of music before, during or after surgery, against routine care, silent headphones, white noise or bed rest, and found postoperative anxiety 0.68 standard deviations lower, pain 0.77 lower, patient satisfaction higher, and no difference in length of stay. Kühlmann and colleagues in the British Journal of Surgery in 2018 pooled 92 trials and 7,385 patients and found anxiety 0.69 standard deviations lower, which they translate to about 21 millimetres on a 100-millimetre scale, with the choice of music and the timing making no measurable difference. The 2018 authors rate the risk of bias in their trials as moderate to high.

So the direction has held up across three reviews and a few more years of trials, and the size has stayed in the same range. The confidence problem is the same in all three, because the person on the trolley always knows whether the headphones are playing.

If you're the one waiting

I read trials; I don't treat anyone, and the person to tell about preoperative anxiety is the anaesthetist, whose premedication decision is theirs to make. The 2013 review's background says the sedatives regularly used for this often have side effects and may prolong recovery, and the same review identified no adverse effects across its 26 music trials. Those are two separate observations and I'd keep them separate. None of the 26 trials measured wound healing, infection, time to discharge or satisfaction, so whether calmer patients before surgery do better after it is a question the 2013 review couldn't answer; the 2015 Lancet review found satisfaction higher and length of stay unchanged.

The result itself is small in scope and has been reproduced. Thirteen trials, 896 people, a short session of recorded music in the waiting bay, and a questionnaire that came back about six points calmer than the one filled in by the people who waited in silence, with the one trial that also wired its patients up finding no difference on the skin.