The best medicine for anxiety depends on the anxiety disorder, but for most adults with generalized anxiety disorder, panic disorder, or social anxiety disorder, a selective serotonin reuptake inhibitor (SSRI) is first-line treatment. Sertraline, escitalopram, and paroxetine carry FDA approval for at least one anxiety disorder and typically produce a 30 to 50 percent reduction in symptom scores over 8 to 12 weeks. Benzodiazepines like alprazolam work within 30 minutes but are recommended for short-term use only, generally under 4 weeks, because of dependence risk. For most people, the slow-acting SSRI is the better long-term medicine.
How Anxiety Medicines Work Inside the Brain
Anxiety disorders involve overactivity in the amygdala and insula, paired with underactivity in the prefrontal cortex that normally dampens threat signals. Medicines do not switch off the fear response. They remodel the circuits that generate it.
SSRIs block the serotonin transporter protein, raising serotonin in the synapse within hours. Yet symptom relief takes 2 to 6 weeks. That gap is the central puzzle of psychiatric pharmacology, and it tells you the drug is not simply topping up a chemical.
The leading explanation is downstream adaptation. Sustained serotonin elevation triggers changes in receptor density and gene expression, particularly in the hippocampus and prefrontal cortex. Brain imaging studies show amygdala reactivity falling by roughly 20 to 30 percent after 6 to 8 weeks of SSRI treatment, roughly matching the timeline of clinical response. In plain terms: the pill starts working on day one, but the brain takes weeks to rewire around it.
Benzodiazepines use a different route. They bind GABA-A receptors and increase the frequency of chloride channel opening, which hyperpolarizes neurons and calms excitatory signaling within 15 to 30 minutes. The effect is real and fast, but the receptor adapts. Tolerance to the sedative effect can develop within 2 to 4 weeks.
The Quantified Comparison: Which Drug Class Wins
Head-to-head trials and network meta-analyses give a clearer ranking than most consumer pages admit. The 2018 Cipriani network meta-analysis in The Lancet, covering 522 trials and over 116,000 participants, ranked escitalopram and sertraline among the best tolerated and most effective antidepressants overall.
| Drug class | Example | Typical onset | FDA anxiety approval | Withdrawal risk | Main downside |
|---|---|---|---|---|---|
| SSRI | Sertraline 25-200 mg/day | 2-6 weeks | Panic, PTSD, social anxiety (sertraline); GAD (escitalopram, paroxetine) | Moderate (5-20% discontinuation symptoms) | Nausea, sexual dysfunction (30-50%) |
| SNRI | Venlafaxine XR 75-225 mg/day | 2-6 weeks | GAD, social anxiety | High | Blood pressure rise above 150 mg |
| Benzodiazepine | Alprazolam 0.25-0.5 mg | 15-30 min | Panic, GAD | High | Dependence in 4-6 weeks |
| Buspirone | 15-60 mg/day | 2-4 weeks | GAD | Low | Modest effect size |
| Hydroxyzine | 25-50 mg as needed | 30-60 min | GAD (short-term) | Low | Sedation, anticholinergic effects |
| Pregabalin | 150-600 mg/day | 1 week | GAD (EU, not US) | Moderate | Scheduling, misuse potential |
The practical read: sertraline and escitalopram deliver the best balance of evidence and tolerability for most people. Venlafaxine XR matches them on efficacy but raises blood pressure in a dose-dependent way, with roughly 3 to 5 percent of patients showing sustained elevation at 225 mg.
One non-obvious correction: benzodiazepines are not "stronger" than SSRIs. In panic disorder trials, alprazolam reduces panic attack frequency faster, but after 8 to 12 weeks the two classes converge, and SSRIs win on relapse prevention once tapered. The 2020 American Psychiatric Association panic disorder guideline explicitly recommends SSRIs or SNRIs as initial treatment, with benzodiazepines as an adjunct.
Anxiety Disorder by Disorder: The Dose Changes
The right drug depends on which anxiety disorder you have. Doses differ by indication even for the same molecule, and insurers and prescribers often default to the lowest labeled dose.
Generalized anxiety disorder. Escitalopram (10-20 mg) and duloxetine (60-120 mg) carry FDA approval. Sertraline is widely used off-label with strong trial support.
Panic disorder. Sertraline (25-200 mg) and paroxetine (10-60 mg) are FDA-approved. Starting doses must be low, around 12.5 to 25 mg of sertraline, because panic patients are unusually sensitive to activation in the first week.
Social anxiety disorder. Paroxetine, sertraline, venlafaxine XR, and fluvoxamine are approved. Effect sizes cluster around 0.4 to 0.6, meaning moderate benefit.
Obsessive-compulsive disorder. This is anxiety-adjacent but needs higher doses. Fluoxetine up to 60 mg and sertraline up to 200 mg, for at least 8 to 12 weeks, per the 2013 American Psychiatric Association OCD practice guideline. A common error is stopping at 40 mg of fluoxetine, a depression dose, and declaring the drug a failure.
Post-traumatic stress disorder. Sertraline and paroxetine are the only FDA-approved medicines. The 2023 VA/DoD guideline places trauma-focused psychotherapy first, with SSRIs as a strong secondary option.
The Timeline: What Happens Week by Week
Expectation management is half of adherence. Patients who know the curve are far less likely to quit at week two, when side effects peak and benefit has not arrived.
| Time point | What changes | What to expect |
|---|---|---|
| Day 1-3 | Serotonin rises in synapse | Possible nausea, headache, jitteriness |
| Week 1-2 | Side effects peak, benefit minimal | Highest dropout risk; stay the course |
| Week 2-4 | Early anxiolytic signal in 50-60% | Sleep and appetite often improve first |
| Week 4-6 | Amygdala reactivity begins falling | Full responders emerge |
| Week 6-8 | Plateau of first-level response | This is the standard decision point |
| Week 8-12 | Maximum benefit with continued titration | Non-responders should switch or augment |
| Month 6-12 | Continued consolidation | Minimum 12 months of treatment after remission |
| Taper phase | 25-50% dose reduction every 2-4 weeks | Prevents discontinuation symptoms |
The week 6 to 8 mark is where "the medicine isn't working" should become a clinical conversation about dose escalation, not abandonment. Roughly 40 to 50 percent of patients do not achieve remission on the first SSRI.
Misconceptions That Cost People Months
The biggest one is the "chemical imbalance" framing. Serotonin levels in untreated anxiety patients overlap almost entirely with healthy controls. The drugs work, but the "low serotonin" story is a simplification that has been retired by most academic psychiatrists. Believing it leads patients to assume a pill alone is the whole answer, which it isn't.
The second is the dependence fear. SSRIs do not cause addiction in the pharmacological sense. They produce discontinuation symptoms in roughly 20 to 30 percent of users, but no craving-driven escalating use. Benzodiazepines do cause physiological dependence, and that distinction matters.
A third is the belief that natural options substitute for medication. Silexan, a lavender oil extract, has supportive trial data with effect sizes around 0.4 in subthreshold GAD. That is real but below SSRI performance. Kava carries hepatotoxicity warnings and should not be self-administered.
Red flags that demand a call to a clinician, not a forum:
- Suicidal thoughts, which can emerge early in treatment, especially in people under 25
- Sudden severe agitation, hallucinations, or fever (possible serotonin syndrome)
- Heart palpitations with chest pain on a new medicine
- Seizure activity after a benzodiazepine taper
- Rash, facial swelling, or difficulty breathing (allergic reaction)
- Worsening depression after week 4
What to Actually Do If You or Someone You Know Needs Help
Start with an assessment, not a prescription. A primary care doctor can diagnose and treat uncomplicated anxiety, but co-occurring depression, substance use, or trauma symptoms are common and change the choice of drug. Roughly 60 percent of people with an anxiety disorder have at least one comorbidity.
Ask about cognitive behavioral therapy alongside medication. The 2014 Cochrane review of 41 studies found combined treatment outperforms either alone for anxiety disorders, with pooled effect sizes around 0.5 to 0.7. Many insurers now cover 12 to 16 sessions.
Insist on the 8 to 12 week window before a switch. Ask for a target dose, not a starting dose. Sertraline at 50 mg is a starting dose; the anxiety trials used up to 200 mg.
Plan the taper before you start. If you begin at 50 mg, expect a taper of roughly 4 to 8 weeks at the end, not abrupt cessation.
Track symptoms with a validated scale, such as the GAD-7, at baseline, week 4, and week 8. The GAD-7 score of 10 or above is the usual treatment threshold and lets you see change objectively.
The Limits: What the Evidence Cannot Tell You
Treatment response is individually unpredictable. No blood test, genetic panel, or clinical feature reliably predicts which SSRI will work for a given person. Commercial pharmacogenomic tests claim to guide selection, but the 2023 FDA safety communication advised against using them to select psychiatric medications because the evidence base does not yet support it.
Relapse after stopping is common. About 25 to 30 percent of people with GAD relapse within 12 months of discontinuing effective medication, and that figure climbs in panic disorder. The 12-month minimum duration after remission is not arbitrary.
Anxiety medicines also carry a pregnancy context that changes the analysis. Sertraline has the largest safety dataset in pregnancy, but all SSRIs carry a small absolute increase in persistent pulmonary hypertension of the newborn, roughly 2 to 3 per 1,000 exposed pregnancies versus 1 to 2 per 1,000 unexposed. That small absolute risk is far below the risk of untreated maternal anxiety and depression.
In plain terms: the best anxiety medicine is the one matched to your disorder, dosed to target, continued long enough to judge, and paired with therapy. For most people that is an SSRI, usually sertraline or escitalopram, with a benzodiazepine reserved for a bridge, not a foundation.