Anxiety disorders

Anxiety helps us prepare for possible danger. It becomes a disorder when fear or worry is difficult to manage, causes substantial distress, or stops us doing ordinary things. Understanding it involves three different questions: why someone becomes vulnerable, what produces their symptoms, and what keeps those symptoms going. The answers overlap, but they are not interchangeable. A process that maintains anxiety today may be different from whatever first started it.

The strongest scientific account brings together inherited differences, early development, learning, experiences and present circumstances. Brain and body systems help us anticipate threats, interpret physical sensations and organise a response. Researchers understand many parts of this picture, but usually cannot identify one biological change that explains a particular person's anxiety. Uncertainty about its original cause does not prevent effective treatment. WHO overview of anxiety disorders

Anxiety takes several forms

Generalised anxiety disorder involves persistent, hard-to-control worry across several areas of life. Panic disorder involves recurrent unexpected panic attacks and concern about further attacks or their consequences. Social anxiety centres on scrutiny, embarrassment or rejection. A specific phobia concerns a particular object or situation. Agoraphobia concerns situations in which escape or help might feel difficult if distressing symptoms occur.

These diagnoses describe recognisable patterns; they do not identify five separate faults in the brain. They can overlap, and people with the same diagnosis can arrive there by different routes. Post-traumatic stress disorder and obsessive–compulsive disorder are classified separately. Research that groups them with anxiety disorders can be informative, but its results should not automatically be applied to every anxiety diagnosis. This essay principally concerns adults, with childhood findings identified separately. Review of anxiety disorders

Inheritance and experience shape vulnerability

Family and twin studies commonly estimate that inherited differences account for around 30–50% of the variation in susceptibility to anxiety within the populations studied. This does not mean that 40% of one person's anxiety is genetic. It also does not tell us how readily someone can recover. Such estimates depend on the population, environment, age and way anxiety is measured. Family resemblance can reflect shared experiences and learning as well as inheritance. Review of anxiety genetics

Many genetic differences each make a small contribution. A large 2026 study compared 122,341 people with anxiety diagnoses with 729,881 comparison participants. It identified 58 associated regions of the genome and replicated 51 in a separate sample. Some genetic influences overlapped with depression and a tendency to experience negative emotions. The results also pointed towards GABA, a chemical messenger involved in regulating nerve-cell activity. These are biological leads, not a genetic test that explains an individual's anxiety. The participants were of European ancestry, limiting how confidently the findings transfer to other populations. Research on changes in how genes are switched on and off remains less developed. Large genetic study of anxiety disorders

Early temperament can matter. Children who remain very wary or withdrawn in unfamiliar situations are more likely to develop anxiety, particularly social anxiety. A review of 20 distinct groups of children found that the odds of later anxiety were about 2.8 times higher among those with this temperament, and the odds of social anxiety about 5.8 times higher. Odds are not the same as a child's probability of developing a disorder. Many wary children do not develop anxiety disorders, and many people with anxiety were not markedly inhibited as children. Review of childhood wariness and later anxiety

Abuse, serious losses, bullying and other adverse experiences can increase vulnerability. Stronger studies compare siblings, use adoption research or exploit changes in circumstances to reduce the influence of alternative explanations. A review of 34 such studies found that the connection between childhood maltreatment and later mental-health problems became smaller after these adjustments, but remained. This is consistent with a causal contribution. The findings covered several mental-health conditions, however, rather than establishing a precise effect specifically on anxiety. Trauma is neither necessary nor sufficient: anxiety need not start with a dramatic event, and adversity does not make illness inevitable. Review of maltreatment using stronger causal designs

Present circumstances deserve the same attention as the past. Someone living with violence, financial insecurity or continuing instability may have substantial reasons to remain alert. An assessment needs to distinguish responses to real danger from fear that has spread beyond the original threat. Women are more frequently affected by anxiety disorders, but that observation alone does not tell us how much reflects biology, exposure, social conditions or differences in recognition. WHO overview

How learning and avoidance can keep fear going

Imagine someone who fears speaking in meetings. Staying away brings immediate relief. That relief makes avoidance more attractive next time, but it also removes opportunities to discover whether the feared outcome would occur or whether an awkward moment could be managed. Anxiety can therefore continue even after its original trigger has changed. Review of anxiety and learning

People can also learn to fear situations resembling an earlier threat. Laboratory studies find that anxious groups, on average, spread learned fear somewhat more widely than comparison groups. The average difference is small—about a quarter of a standard deviation, a statistical measure of how spread out scores are. This does not mean the process explains a quarter of cases. The studies include people with PTSD and OCD, and usually cannot show whether this difference preceded illness. They support a possible contributing process, not a claim that everyone with anxiety is unable to recognise safety. Review of the spread of learned threat

Repeated checking, concealment or reassurance-seeking can sometimes maintain fear too. Researchers call these actions “safety behaviours” when their purpose is to prevent a feared outcome. Experiments have deliberately increased or reduced them. A 2026 review found that reducing them improved symptoms, while increasing them worsened symptoms. Its combined symptom benefit was around 0.43 standard deviations; an exploratory adjustment for missing negative studies reduced it to about 0.31. The direction remained favourable, but the size was less secure. The studies involved varied problems and populations. Their strongest message concerns what can keep symptoms going, rather than what originally caused a disorder. Review of safety-behaviour experiments

The purpose of a behaviour matters. Avoiding a genuine hazard can be sensible. Treatment concerns patterns that unnecessarily restrict life or prevent useful learning, not abandoning ordinary precautions.

Uncertainty is especially relevant to generalised anxiety. A review of 115 studies and dissertations, involving 28,693 participants, found stronger connections between distress about uncertainty and generalised anxiety than between that distress and panic. Some questionnaires asked overlapping questions, which can strengthen apparent connections. These findings do not prove that difficulty with uncertainty always comes first, or that worry is the only route linking uncertainty to anxiety. Review of uncertainty, worry and anxiety

What happens in the brain

Anxiety involves communication between many brain systems. The amygdala and nearby structures help learn what matters and organise defensive responses. The insula and parts of the cingulate cortex help interpret bodily states, uncertainty and demands for action. The hippocampus helps remember context: a cue that matters in one setting may be harmless in another. Frontal networks contribute to attention, interpretation, flexible action and regulation. Reward and social-processing systems also matter, because approaching a situation involves weighing possible benefits as well as danger. Review of brain circuits in anxiety

Calling the amygdala an “anxiety centre” loses much of this complexity. It contains different groups of cells, sometimes with opposing functions, and participates in reward and social behaviour as well as threat. Findings change with the task, timing and part of the structure examined. Larger studies have challenged some striking early claims. Review of the amygdala and mental illness

Animal experiments can show that changing a particular circuit changes defensive behaviour. Human scans more often show associations between symptoms and activity. An animal's defensive response does not by itself explain a person's worry, shame or fear of rejection. A difference between two groups on a scan also does not show that it caused anxiety, occurs in everyone with the diagnosis, or can select the best treatment for one person.

Bodily sensations and chemical messengers

Panic can involve a powerful loop between sensation and interpretation. A pounding heart may be understood as evidence of imminent catastrophe. Fear then increases bodily arousal, which seems to confirm the danger. Research supports this account, although much of the evidence is of moderate quality. It does not mean every panic attack begins with an obvious conscious thought, or that breathing and other physical processes are irrelevant. Review of the cognitive account of panic

Being preoccupied with the body is also different from detecting its signals more accurately. A review of 55 studies found no overall relationship between anxiety and accuracy on heartbeat-perception tasks. Anxiety may change how sensations are noticed, judged or trusted without improving objective detection. This finding concerns those particular tasks, rather than every form of bodily awareness. Review of anxiety and heartbeat perception

The automatic systems controlling heart rate, breathing and other bodily responses help produce physical symptoms. But measures such as variation between heartbeats are influenced by many conditions and activities. A 2025 review judged the anxiety evidence for this proposed marker weak. The body's cortisol system also interacts with threat processing, yet studies find both raised and reduced responses. Neither a heartbeat measurement nor a simple “high cortisol” account supplies a general explanation. Review of heart-rate variability, review of cortisol-related systems

Serotonin, noradrenaline, GABA and glutamate all influence communication in relevant circuits. Their involvement does not establish one shortage or excess shared by anxiety disorders. A medicine can help by changing a system without reversing the original cause. The effectiveness of a serotonin-related medicine therefore does not prove that anxiety began with too little serotonin. Clinical review of anxiety

Direct chemical measurements remain unsettled. A review of 25 brain-spectroscopy datasets, involving 370 patients and 342 comparison participants, found lower concentrations of choline-containing compounds in parts of the outer brain, but no statistically clear pooled differences in GABA or glutamate. Small studies finding no difference cannot establish that the systems are identical. Tissue concentrations are also different from receptor activity or the timing of chemical signals. The choline finding does not show a dietary deficiency or justify taking a supplement. Review of brain chemical measurements

Promising explanations still need testing

Inflammation and the communities of microbes living in the gut may matter for some people. Human studies of diagnosed anxiety and gut microbes are generally small and measure people at one point in time. Diet, medicines, physical health and anxiety-related behaviour can all influence the results. Animal findings or lower anxiety questionnaire scores in otherwise healthy volunteers do not establish a common cause of clinical anxiety. Review of the gut microbiome in anxiety disorders

Computer-based models examine how people estimate danger, handle uncertainty and decide how much to trust bodily signals. These models can generate useful predictions, but no single faulty calculation has been established across anxiety disorders. One study found altered adjustment to new bodily information without strong evidence for unusually fixed prior beliefs; its anxiety-only group was small and its wider sample included several diagnoses. Evolutionary explanations likewise help explain why protective systems exist, but additional evidence is needed to explain why a persistent disorder develops. Study modelling interpretation of bodily signals

Some biological differences may be vulnerabilities; others may be consequences of prolonged anxiety, attempts to compensate, treatment effects or responses to continuing adversity. Routine scans, genetic profiles, cortisol tests and proposed biomarkers cannot currently provide a validated general explanation of an individual's anxiety. Clinical assessment also considers physical illness, medication and substances: thyroid, heart and breathing conditions can produce overlapping symptoms or coexist with anxiety. New or severe physical symptoms need medical assessment rather than an assumption that they are anxiety. Review of psychiatric biomarkers, clinical assessment review

Talking treatment should match the problem

Cognitive behavioural therapy, usually called CBT, helps people examine and change patterns of thought and behaviour that maintain difficulties. Its content matters more than the label. Exposure-based work involves approaching feared but sufficiently safe situations or sensations in a planned way, so that new learning becomes possible. It is not a demand to endure real danger.

For generalised anxiety, individual CBT and a structured approach called applied relaxation are guideline options. For panic, treatment addresses fear of attacks and associated avoidance. For social anxiety, the preferred initial psychological treatment is individual CBT specifically developed for fear of social judgement. It can include testing predictions, shifting attention away from constant self-monitoring, reducing unhelpful safety behaviours, and changing worry before and after social events. For uncomplicated specific phobias, exposure has the clearest role; continuing medication is not standard treatment. NICE guidance on generalised anxiety and panic, NICE guidance on social anxiety

A review of 65 trials with 5,048 adults with generalised anxiety found short-term benefits from conventional CBT and newer approaches incorporating acceptance or mindfulness. At three to twelve months after treatment, only conventional CBT retained a statistically clear advantage over usual care. This supports its position without proving that every alternative has worse lasting effects. The review did not find clear effectiveness differences between the psychological treatment categories themselves. A separate review of 101 trials involving 13,164 people with social anxiety supported individual CBT as the preferred initial intervention. Review of generalised-anxiety psychotherapies, review of social-anxiety treatments

Where medication helps, and its trade-offs

Antidepressants can treat anxiety even when the person is not depressed. The most established classes are SSRIs and SNRIs, which affect serotonin or serotonin and noradrenaline signalling. For generalised anxiety and panic, both appropriate medication and psychological treatment have substantial evidence. There is no established overall winner between high-intensity psychological treatment and medication for generalised anxiety. Preferences, previous response, side effects and access help determine a reasonable starting point.

A 2025 Cochrane review of 37 trials with 12,226 adults with generalised anxiety provides a concrete comparison. About 56 in 100 people receiving an antidepressant achieved at least a halving of anxiety symptoms, compared with about 40 in 100 receiving placebo. That is roughly 16 additional responders per 100 treated. The review's conventionally rounded estimate was one additional responder for every seven treated, with an uncertainty range of five to nine. Response did not necessarily mean symptoms had disappeared. About 11 in 100 taking antidepressants stopped because of adverse effects, compared with five in 100 taking placebo. Overall dropout was similar, partly because fewer people on medication left for lack of benefit. Cochrane review of antidepressants for generalised anxiety

Benefits usually develop over several weeks. Nausea, disturbed sleep, initial agitation and sexual difficulties are among possible burdens; some settle and others persist. Interactions, earlier treatment experiences and withdrawal effects matter when choosing a medicine. No individual antidepressant is universally best. Research on panic generally favours SSRIs when balancing benefit and harm, while precise drug rankings remain uncertain. NHS information on antidepressants, review of medication for panic disorder

NICE's suggestion to consider sertraline first for generalised anxiety reflects cost-effectiveness as well as clinical evidence. Pregabalin is another option when SSRIs or SNRIs cannot be tolerated, with specific precautions about dependence and misuse. For adults with social anxiety who prefer medication, sertraline or escitalopram are guideline options. These are choices for a prescribing discussion, not instructions to change treatment independently. NICE medication recommendations, NICE social-anxiety recommendations

Benzodiazepines can provide rapid relief, but dependence, withdrawal and limited evidence for sustained benefit restrict their continuing role. NICE permits short-term use in a crisis for generalised anxiety; routine use is generally discouraged. Beta-blockers have much weaker evidence for treating anxiety disorders themselves. A review found only five small studies, involving 179 participants, suitable for combined analysis and no clear advantage over placebo or benzodiazepines for social phobia or panic. That does not settle every use for an isolated physical symptom during a stressful event. Review of panic medication, review of beta-blockers

Combining treatment and helping children

An incomplete response can be a reason to add or switch treatment after reviewing what has actually been delivered. It is not evidence that anxiety is untreatable. For social anxiety, NICE recommends adding individual CBT after only a partial response to sertraline or escitalopram over ten to twelve weeks. Across adult anxiety disorders, however, evidence does not establish that everyone should begin therapy and medication together. Findings vary by diagnosis and by which single treatment provides the comparison. NICE social-anxiety guidance, international anxiety-treatment guideline

The childhood evidence provides a clearer example of combination benefit. In a trial of 488 participants aged seven to seventeen, twelve-week improvement rates were about 81% for CBT plus sertraline, 60% for CBT, 55% for sertraline and 24% for placebo. The combination outperformed either component. These results concerned selected childhood anxiety disorders and measured improvement, not permanent cure. They cannot simply be transferred to adults or used as a prescribing rule for every anxious child. Childhood anxiety treatment trial

Keeping improvement going

CBT benefits can last after sessions end. A review of 69 trials involving 4,118 participants found advantages over control conditions through follow-up lasting up to a year. Longer follow-up was sparse and results differed by diagnosis. Quality limitations prevent a confident promise of protection for many years. Review of longer-term CBT outcomes

Medication continuation also reduces relapse risk among people who have responded. Across 28 trials with 5,233 previous responders, relapse occurred in about 16% continuing antidepressants and 36% after discontinuation, during follow-up of up to a year. Some participants had OCD or PTSD, and the findings cannot identify an ideal stopping date or establish a need for lifelong medication. For effective generalised-anxiety medication, NICE recommends continuing for at least a year. Review of relapse after antidepressant discontinuation, NICE information for the public

Withdrawal can resemble returning anxiety. A later review criticised studies that used short reductions in dose or failed to distinguish withdrawal from relapse. That review mainly concerned recurrent depression, so it qualifies rather than replaces the anxiety findings. Stopping should usually be gradual and agreed with the prescriber, taking account of benefits, harms, earlier episodes and preferences. Cochrane review of stopping long-term antidepressants

Remote therapy, mindfulness and exercise

Structured CBT delivered remotely with a therapist can be effective. A review of 54 trials found little average difference from in-person treatment; seventeen trials concerned anxiety or related conditions. This supports a delivery option, not a claim that every app or unguided programme works equally well. Review of therapist-guided remote CBT

A structured mindfulness programme is a credible alternative for some adults. In one trial of 276 people, an eight-week programme performed within a prespecified acceptable margin of escitalopram on the main anxiety outcome. This does not prove exact equivalence. The programme involved weekly two-and-a-half-hour classes, a retreat and 45 minutes of daily practice, and the main completion-based analysis included 208 people. Its demands and the relatively limited comparative evidence matter when considering suitability. Mindfulness and escitalopram trial

Exercise can also reduce anxiety symptoms. A 2026 review of 30 trials found favourable results, particularly for resistance exercise and activities combining movement with attention to the body. Only four trials were judged at low risk of bias. This supports exercise within care, while leaving uncertainty about comparisons with established treatment and lasting benefit. Regular sleep, physical activity and reducing harmful alcohol use can support recovery; they should fit the person's situation rather than become another source of blame. Review of exercise for anxiety disorders, WHO self-care information

What the evidence can—and cannot—decide

Improvement, remission and everyday recovery are different outcomes. A person can halve a symptom score while still finding work or relationships difficult. Rapid relief, tolerability, completing a course and staying well after it ends are also different goals. A useful review therefore considers daily functioning and harms alongside symptoms.

Treatment rankings depend partly on the comparison. CBT often looks more effective against a waiting list than against a credible alternative therapy. Medicines have a smaller advantage over placebo than the total improvement observed after starting them. Comparing these different effect sizes cannot identify a winner. Psychotherapy cannot usually be hidden from participants or therapists; expectations, researcher preferences, dropout and selective reporting can influence findings. Longer follow-up brings missing participants and additional treatment. Trial participants may also have fewer other illnesses than people seen in everyday practice.

The same care is needed with causes. A symptom questionnaire, self-reported diagnosis, clinical interview and health record do not measure exactly the same thing. Large samples do not solve poor measurement or limited representation. Better research needs to follow diverse people before and through illness, test whether changing a proposed mechanism actually changes symptoms through that mechanism, and show that any biological or prediction tool improves real decisions. A tool predicting who improves in general is not enough to select treatment: it needs to identify who benefits more from one option than another. For treatment, the priorities include direct comparisons lasting at least two years, better withdrawal studies, and outcomes covering relationships, work, quality of life and treatment burden.

For someone seeking help, the practical questions are more immediate: which problem is being treated, whether the therapy contains the relevant techniques, what practice and support it requires, how benefits and side effects will be reviewed, and what happens if improvement is incomplete. Several established treatments help. Choosing among them is a continuing collaboration informed by evidence and the person's life, rather than a search for one universal explanation or cure.