Mind and Mood Medicines

TL;DR. Antidepressants work better than placebo, and the effect is modest on average and larger in severe depression. The "chemical imbalance" story was always a marketing simplification, and the drugs' failure to fit it does not mean they do not work. Benzodiazepines are effective, are among the most dependence-forming drugs in medicine, and should almost never be used for more than four weeks. Stopping antidepressants causes withdrawal in a substantial minority, which was under-recognised for years, and tapering slowly matters. Lithium remains the most effective drug in bipolar disorder and requires blood monitoring.

1. Depression and the antidepressants

The classes

How they work, honestly

SSRIs block the serotonin reuptake transporter, raising synaptic serotonin within hours. But the clinical effect takes 2 to 6 weeks, which means raising serotonin is not itself the therapeutic mechanism. The current understanding involves downstream adaptations: receptor changes, increased BDNF, and neuroplasticity in the hippocampus and prefrontal cortex.

The "chemical imbalance" framing was a simplification used in patient communication and pharmaceutical marketing from the 1990s. A widely publicised 2022 umbrella review by Joanna Moncrieff and colleagues found no consistent evidence that depression is caused by low serotonin. That finding was accurate and was widely misreported as showing antidepressants do not work.

Both things are true: depression is not simply a serotonin deficiency, and SSRIs outperform placebo in randomised trials. A drug does not have to correct a deficiency to work; paracetamol is not correcting a paracetamol deficiency.

What the evidence shows

The largest analysis, a 2018 Lancet network meta-analysis by Cipriani and colleagues covering 522 trials and 116,477 participants, found all 21 antidepressants studied were more effective than placebo, with odds ratios of roughly 1.4 to 2.1.

The size of the effect is genuinely contested. Average effect sizes are modest (standardised mean difference around 0.3), which is small in statistical terms. Two important qualifications:

• Benefit increases with baseline severity. In mild depression the drug-placebo difference is small; in severe depression it is substantial. Guidelines reflect this by recommending psychological therapy first for mild cases.
• Average effect sizes obscure individual response. Some people respond dramatically; some do not respond at all. The average is not the experience.

Roughly 50 to 60 percent respond to the first antidepressant they try, and about a third achieve remission. The STAR*D study showed that sequential switching raises cumulative remission rates, with diminishing returns.

Choice of drug is driven mainly by side effect profile, interactions, and previous response rather than by efficacy differences, which are small between agents.

Side effects

Early (first 1 to 2 weeks, usually settling): nausea, headache, agitation, insomnia or drowsiness, increased anxiety.

The increased anxiety and agitation in the first fortnight matters. In people under 25 there is a documented small increase in suicidal thoughts and behaviour in the early weeks, which is why regulators require warnings and why close monitoring is advised when starting or changing dose in young people. It does not mean the drugs should not be used; untreated depression carries a far higher risk. It means the first month needs support.

Persistent:

• Sexual dysfunction in 40 to 70 percent: reduced libido, delayed orgasm, erectile difficulty. This is the most common reason people stop and the least often discussed. Options include dose reduction, switching to bupropion or mirtazapine (much lower rates), or adding an agent. A small number of people report persistent symptoms after stopping (post-SSRI sexual dysfunction), which is now formally recognised by the EMA as a possible outcome and is poorly understood.
• Weight change: mirtazapine and paroxetine most; fluoxetine and bupropion least.
• Emotional blunting in perhaps 40 to 60 percent: reduced intensity of both negative and positive feelings. Frequently under-discussed and a common reason for stopping.
• Hyponatraemia (low sodium), particularly in older adults, especially with diuretics.
• Increased bleeding risk, because platelets use serotonin. Combined with NSAIDs this roughly triples GI bleeding risk (Chapter 66).
• QT prolongation with citalopram and escitalopram at higher doses.

Stopping: withdrawal is real

This was minimised for years and is now better recognised.

Discontinuation/withdrawal symptoms occur in a substantial minority, with estimates varying widely by methodology (a contested 2019 review suggested over half; a 2024 Lancet Psychiatry meta-analysis found around 15 percent experience symptoms attributable to withdrawal beyond placebo). Symptoms include dizziness, "brain zaps" (electric shock sensations), nausea, flu-like symptoms, insomnia, vivid dreams, irritability, and anxiety.

Risk is highest with short half-life drugs: paroxetine and venlafaxine are the worst; fluoxetine, with a half-life of days and an active metabolite, tapers itself and causes the least (Chapter 62).

How to stop: slowly, and more slowly at the end than the beginning. Hyperbolic tapering is now recommended in UK guidance: receptor occupancy is non-linear with dose, so the final small reductions produce the biggest change in receptor occupancy and need the smallest steps and the longest time. Liquid formulations and tapering strips exist for this. Months, not weeks, for people who have taken them for years.

Withdrawal is often mistaken for relapse, leading to indefinite continuation. The distinguishing features: withdrawal starts within days of a dose reduction, includes physical symptoms (dizziness, brain zaps) not typical of depression, and improves within hours of reinstating the dose. Relapse develops over weeks and looks like the original illness.

Physical dependence is not addiction (Chapter 62). Antidepressants are not craved or misused; they do produce adaptation that must be unwound carefully.

2. Anxiety

First-line is psychological therapy (CBT has the strongest evidence) and SSRIs or SNRIs, which work for generalised anxiety, panic disorder, social anxiety, OCD, and PTSD. Anxiety often worsens for the first two weeks, so doses are usually started lower than for depression.

Propranolol for the physical symptoms of situational anxiety (tremor, palpitations) works well and does nothing for the psychological component.

Pregabalin is licensed for generalised anxiety, is effective, and is now a controlled drug in the UK because of dependence and misuse, and because it substantially increases respiratory depression risk with opioids.

Buspirone is a non-dependence-forming option with modest efficacy.

3. Benzodiazepines and Z-drugs

Diazepam, lorazepam, temazepam, alprazolam, clonazepam; zopiclone, zolpidem, zaleplon

Mechanism: enhance GABA-A receptor function, increasing the brain's main inhibitory signal. Result: anxiolysis, sedation, muscle relaxation, anticonvulsant effect, and anterograde amnesia.

They work extremely well and extremely quickly, which is the whole problem.

The rules, which exist for good reason:

• Maximum 2 to 4 weeks, including tapering, for anxiety or insomnia.
• Tolerance develops within weeks, particularly to the hypnotic effect.
• Dependence can develop in as little as 4 weeks of regular use.
• Withdrawal is dangerous. Unlike opioid withdrawal, benzodiazepine withdrawal can cause seizures and delirium and can kill. Along with alcohol, it is the withdrawal you must never do abruptly.
• Withdrawal symptoms can persist for months in long-term users, and tapering over many months (often using the "Ashton manual" approach, converting to long-acting diazepam and reducing slowly) is standard.

Other harms: falls and hip fractures in older adults, cognitive impairment, road traffic accidents, and respiratory depression when combined with opioids or alcohol, which is a major contributor to overdose deaths (Chapter 68).

Legitimate uses: short-term severe anxiety or crisis, alcohol withdrawal (where they prevent seizures and save lives), status epilepticus, procedural sedation, and muscle spasm.

Z-drugs were marketed as a safer alternative for insomnia and are not meaningfully different in dependence potential. Zolpidem is associated with complex sleep behaviours (sleepwalking, sleep driving, sleep eating) carrying a boxed warning.

For insomnia, the first-line treatment is CBT-I (cognitive behavioural therapy for insomnia), which outperforms medication in the long term and is recommended ahead of it in every major guideline. It is available in digital form and remains under-provided.

4. Bipolar disorder

Lithium remains the most effective mood stabiliser and the only drug with reasonable evidence for reducing suicide risk specifically. It is an element, not a designed molecule, and its mechanism is still uncertain.

It requires monitoring, which is the main practical burden:

• Narrow therapeutic index: therapeutic range 0.4 to 1.0 mmol/L; toxicity above 1.5.
• Blood levels every 3 to 6 months, plus kidney and thyroid function, because it causes hypothyroidism in a substantial minority and can affect renal function long term.
• Toxicity presents with coarse tremor, vomiting, diarrhoea, confusion, ataxia, and seizures.
• Precipitated by dehydration, low sodium intake, NSAIDs, ACE inhibitors, ARBs, and diuretics, all of which reduce lithium clearance. This is a genuinely dangerous and common interaction set.
• Contraindicated in pregnancy in the first trimester where possible (Ebstein's anomaly risk), though the risk is lower than once believed and the decision is individual.

Other mood stabilisers: valproate (highly effective and absolutely contraindicated in anyone who could become pregnant, because of major congenital malformations in around 10 percent and neurodevelopmental disorders in around 30 to 40 percent of exposed children; regulatory restrictions are now strict and it is one of the most serious medicines scandals of recent decades), lamotrigine (good for bipolar depression; must be titrated very slowly because of Stevens-Johnson syndrome risk), and quetiapine and other antipsychotics.

5. Antipsychotics

Used in schizophrenia, bipolar disorder, and, at low doses, for agitation and as adjuncts in depression.

Mechanism: dopamine D2 receptor blockade, plus serotonin receptor effects in the newer agents.

Key side effects:

• Metabolic syndrome: olanzapine and clozapine are worst; weight gain can be substantial and is a major cause of the reduced life expectancy in serious mental illness. Requires active monitoring.
• Akathisia: an intensely unpleasant inner restlessness, frequently misread as agitation and treated by increasing the dose, which makes it worse. It is associated with suicidality and is under-recognised.
• Tardive dyskinesia: involuntary movements, often of face and tongue, after long-term use; may be irreversible.
• Neuroleptic malignant syndrome: rare, life-threatening. Fever, rigidity, altered consciousness, autonomic instability. A medical emergency.
• Hyperprolactinaemia (risperidone especially): galactorrhoea, menstrual disruption, sexual dysfunction, bone loss.
• QT prolongation.

Clozapine is uniquely effective in treatment-resistant schizophrenia and requires mandatory regular blood monitoring because of agranulocytosis (loss of white cells) in around 1 percent. It also causes severe constipation that can be fatal, and its levels rise sharply if a smoker quits, because tobacco smoke induces CYP1A2 (Chapter 62).

In dementia, antipsychotics for agitation carry a boxed warning: increased mortality and stroke. They are used only when non-drug approaches have failed and symptoms are severe, at the lowest dose for the shortest time.

6. ADHD medicines

Stimulants (methylphenidate, lisdexamfetamine, dexamfetamine) increase dopamine and noradrenaline in the prefrontal cortex. They have large effect sizes, among the largest in psychiatry, with response rates around 70 percent.

Side effects: reduced appetite (and, in children, a small effect on growth trajectory that largely catches up), insomnia, headache, raised heart rate and blood pressure, and irritability as the dose wears off.

On the addiction question: stimulants are controlled drugs and are diverted and misused. The evidence in treated ADHD points the other way: several large studies find that treatment is associated with lower, not higher, rates of subsequent substance use disorder, plus reduced accidents, injuries, and criminality. Untreated ADHD carries substantial risks of its own.

Non-stimulants: atomoxetine, guanfacine. Slower onset, smaller effect, useful where stimulants are unsuitable.

7. Interactions worth knowing

8. Practical guidance

• Give an antidepressant 4 to 6 weeks at an adequate dose before judging it. Stopping at two weeks because of early side effects is the commonest reason treatment "fails."
• Report side effects rather than stopping. Most are manageable by switching, and sexual dysfunction in particular has good alternatives.
• Never stop abruptly. Taper slowly, more slowly at the end, and expect months rather than weeks after long-term use.
• Benzodiazepines: two to four weeks, and no more. If you have been on them longer, do not stop suddenly, and ask for a structured taper.
• CBT-I before sleeping tablets, always.
• Combining medication and psychological therapy outperforms either alone for moderate to severe depression.
• Alcohol worsens depression and anxiety and interacts with every drug in this chapter.
• Tell any prescriber about St John's wort. It is sold as a harmless herbal and it is one of the most interaction-prone substances in the pharmacy.

9. The bottom line

• Antidepressants beat placebo in the largest analyses, with modest average effects that grow with severity. The "chemical imbalance" explanation was always a simplification, and its collapse does not mean the drugs do not work.
• Sexual dysfunction and emotional blunting are common, under-discussed, and often fixable by switching.
• Withdrawal on stopping is real, worst with paroxetine and venlafaxine, and is frequently mistaken for relapse. Taper slowly and hyperbolically.
• Benzodiazepines: maximum four weeks. Dependence develops fast, and withdrawal can cause seizures and kill, so never stop abruptly.
• Lithium is the most effective bipolar treatment and needs blood monitoring; NSAIDs, ACE inhibitors, and dehydration cause toxicity. Valproate must not be used in anyone who could become pregnant.
• Antipsychotics trade movement side effects for metabolic ones; akathisia is under-recognised and distressing.
• ADHD stimulants have large effects, and treatment is associated with lower subsequent substance use, not higher.

Sources and notes

Doses, cautions, and interactions follow the British National Formulary and the electronic Medicines Compendium; US figures follow FDA labelling. Antidepressant efficacy follows Cipriani et al., The Lancet, 2018, a network meta-analysis of 522 trials and 116,477 participants. The serotonin hypothesis umbrella review is Moncrieff et al., Molecular Psychiatry, 2022, with the subsequent correspondence about what it does and does not imply. Severity-dependent effect follows Fournier et al., JAMA, 2010, and Kirsch's analyses. STAR*D sequential remission follows Rush et al. Early suicidality in under-25s follows the FDA's 2004 and 2007 meta-analyses and the resulting boxed warning. Sexual dysfunction rates follow Montejo et al.; post-SSRI sexual dysfunction was recognised as a possible outcome by the EMA's PRAC in 2019. Emotional blunting follows Goodwin, Price, and colleagues' surveys. Withdrawal incidence is contested between Davies and Read, Addictive Behaviors, 2019, and Kalfas et al.'s 2024 Lancet Psychiatry meta-analysis; hyperbolic tapering follows Horowitz and Taylor, Lancet Psychiatry, 2019, and the 2022 NICE guidance. Benzodiazepine dependence, withdrawal seizures, and the Ashton approach follow the Ashton Manual and BNF guidance; CBT-I first-line follows NICE and AASM guidelines. Lithium monitoring, toxicity precipitants, and suicide reduction follow Cipriani et al., BMJ, 2013, and NICE CG185. Valproate teratogenicity and the Pregnancy Prevention Programme follow the MHRA's 2018 and 2023 regulatory actions. Antipsychotic metabolic and movement effects, akathisia under-recognition, and the dementia mortality warning follow the CATIE trial and MHRA and FDA advisories. Clozapine monitoring and the smoking-cessation interaction follow BNF. ADHD stimulant effect sizes and reduced substance use follow Cortese et al., Lancet Psychiatry, 2018, and Chang et al.'s Swedish register studies.

Open questions. The size of the antidepressant effect, and how much of it is clinically meaningful as opposed to statistically detectable, remains genuinely disputed between researchers reading the same trials. Withdrawal incidence estimates differ several-fold depending on methodology.

👉 Next: inhalers and steroids.