Section 7: Combined Treatment -- Medication Plus Psychotherapy

The clinical question of whether to start medication, therapy, or both simultaneously is a common decisions in anxiety treatment. The evidence provides clear guidance, though the answer depends heavily on the patient's specific diagnosis, severity, developmental stage, and whether they are already taking a benzodiazepine.

Two patients are referred to your practice on the same day. The first is a 9-year-old girl with severe separation anxiety who screams and clings to her mother at school drop-off, has begun refusing to attend birthday parties, and wakes her parents nightly with fears that "something bad will happen" if they leave. Her SCARED separation subscale is in the clinical range. The second is a 14-year-old boy with social anxiety who avoids raising his hand in class, eats lunch alone to avoid conversation, and has declined invitations to join the debate team despite strong interest. Both families ask the same question: "Should we add medication to the therapy?"

7.1 Combined Treatment vs. Monotherapy: The Overall Evidence

Does adding psychotherapy to medication (or vice versa) produce better outcomes than either alone? The short answer is yes, with moderate-sized benefits. A landmark meta-analysis of 52 RCTs involving 3,623 adults with depression or anxiety demonstrated a moderately large advantage for combined pharmacotherapy and psychotherapy over pharmacotherapy alone (Hedges' g = 0.43, NNT = 4.20). (Cuijpers et al., 2014, World Psychiatry) The effects of medication and psychotherapy appear largely independent and additive, with each contributing roughly equally.

More recent data confirms this pattern while suggesting the advantage may be somewhat smaller than early estimates. A 2022 umbrella review of 102 meta-analyses across 3,782 RCTs and over 650,000 patients produced a more conservative but still significant aggregate effect (SMD = 0.31, 95% CI: 0.19-0.44). (Cuijpers et al., 2022, PMC)

The critical long-term finding: a 2024 systematic review of 19 RCTs (N=1,154) tracking outcomes at least 12 months post-treatment demonstrated that both psychotherapy alone and combined treatment significantly outperform pharmacotherapy alone in preventing relapse (combined vs. medication-alone RR = 0.60, 95% CI: 0.37-0.97; psychotherapy alone vs. medication-alone RR = 0.58, 95% CI: 0.38-0.89). (Frontiers in Psychiatry, 2024) The enduring protective effect is primarily mediated by the cognitive and behavioral skills acquired during psychotherapy, not by the ongoing presence of medication.

7.2 The CAMS Trial: The Anchor for Pediatric Treatment Decisions

The Child/Adolescent Anxiety Multimodal Study remains the strongest available evidence guiding pediatric anxiety treatment. This multisite RCT randomized 488 youth aged 7-17 with GAD, social phobia, or separation anxiety to 12 weeks of Coping Cat CBT, sertraline, combination, or pill placebo. (Walkup et al., 2008, New England Journal of Medicine)

Acute results. Combination treatment demonstrated clear superiority: response rates of 80.7% for combination, 59.7% for CBT alone, 54.9% for sertraline alone, and 23.7% for placebo. (Walkup et al., 2008, New England Journal of Medicine) Combination was statistically superior to both monotherapies. CBT and sertraline monotherapies did not differ from each other (p = 0.41), establishing them as co-equal first-line options.

Who benefits most from combination. Detailed moderator analyses found that the principal diagnosis was the only baseline variable significantly moderating outcomes. (Compton et al., 2014, PMC)

For separation anxiety disorder, combination generated large advantages over both monotherapies: effect sizes of -0.98 versus sertraline alone and -0.91 versus CBT alone (both "large"). For GAD, combination maintained a large advantage over sertraline (-0.69) and moderate advantage over CBT (-0.48). For social phobia, the advantage was markedly attenuated: only a small effect size difference of -0.16 over sertraline, suggesting that for social anxiety, the pharmacological addition provides minimal incremental benefit. (Compton et al., 2014, PMC)

This finding has direct clinical implications: youth with severe separation anxiety derive transformative benefit from starting with combination treatment. For social anxiety, where pathology is deeply rooted in behavioral avoidance and cognitive distortions, robust CBT is paramount and adding an SSRI provides only a modest boost.

Higher baseline severity, greater functional impairment, and higher caregiver strain predicted poorer outcomes regardless of treatment modality, suggesting these patients need combination treatment from the start rather than failing monotherapy first. (Compton et al., 2014, PMC)

Pause and think: Return to the two patients referred at the start of this section. The 9-year-old with severe separation anxiety and the 14-year-old with social anxiety both have families asking whether to add medication to therapy. Based on the CAMS moderator data, would your recommendation differ between them, and why?

See a model response

The CAMS moderator analyses found that the principal diagnosis was the only baseline variable significantly moderating treatment outcomes, and the differential is substantial. For separation anxiety, combination treatment generated large advantages over both monotherapies: effect sizes of -0.98 versus sertraline alone and -0.91 versus CBT alone. For this 9-year-old with severe separation anxiety, the evidence strongly supports starting with combination treatment rather than trying monotherapy first. Her high severity and functional impairment reinforce this: the CAMS data show that greater baseline severity and caregiver strain predict poorer outcomes regardless of modality, suggesting these patients need combination from the start.

For the 14-year-old with social anxiety, the picture is different. The combination advantage over sertraline alone was only -0.16, a small effect. Social anxiety pathology is rooted in behavioral avoidance and cognitive distortions that respond primarily to robust CBT, specifically behavioral experiments designed to test social predictions. Adding an SSRI provides only minimal incremental benefit. Starting with individual CBT and reserving medication for insufficient response is the better-supported approach.

The convergence phenomenon. At 24 and 36-week follow-up, the three active treatments converged on most categorical outcomes including remission rates. (Piacentini et al., 2014, PMC) Over 80% of acute responders maintained their response across all active arms. The convergence was driven by concomitant treatment-seeking: 36.4% of sertraline monotherapy patients and 29.9% of CBT monotherapy patients sought outside treatments during follow-up, effectively building their own combination therapy, compared to only 13.5% of the combination group. (Piacentini et al., 2014, PMC) While monotherapy is a reasonable starting point, persistent non-remission naturally drives patients toward combination over time.

7.3 Treatment Sequencing: Stepped Care and the SMART Trial

Stepped care models. Current guidelines uniformly advocate stepped-care approaches. The NICE guidelines structure treatment hierarchically: Steps 1-2 (identification, psychoeducation, low-intensity self-help), Step 3 (high-intensity CBT or pharmacotherapy for non-responders), Step 4 (specialist multi-agency treatment for complex, refractory cases). (NICE, 2013) For social anxiety, NICE deviates by recommending individual CBT as the primary first-line, positioning it ahead of medication. (NICE CG159)

The APA pharmacological sequencing hierarchy: first-line SSRIs/SNRIs with adequate trial of at least 2 months; second-line alternative SRI if the first fails; third-line older agents (TCAs, MAOIs, atypical antipsychotics) only after verifying the adequacy of CBT engagement. Escalating to severe polypharmacy without a robust trial of evidence-based psychotherapy is considered poor practice. (APA/Psychopharmacology consensus, 2025)

The SMART trial. The Sequential Multiple Assignment Randomized Trial (NCT04760275), led by Ginsburg and colleagues, directly addresses the sequencing question in pediatric anxiety. Youth aged 8-17 are randomized to fluoxetine or Coping Cat CBT for 12 weeks; non-remitters are then re-randomized to either optimizing their current treatment or adding the alternative modality. (Ginsburg et al., 2021, PMC) While final results are pending (2025-2026), the hypothesis is that non-remitters will benefit more from adding the alternative modality than from optimizing a failing monotherapy.

The Anxiety Action Plan (AxAP) provides a brief pediatrician-delivered intervention pathway. Open trials demonstrated significant reductions in child anxiety severity (Cohen's d = 1.06), impairment (d = 0.75), and caregiver burden (d = 0.60), demonstrating that primary care clinicians can effectively initiate early exposure-based care before specialist referral. (Ginsburg et al., 2015, PMC)

7.4 Pharmacological Augmentation of Extinction Learning

A distinct research frontier investigates pharmacological agents intended not to suppress anxiety continuously, but to selectively enhance neuroplasticity during exposure therapy sessions.

D-cycloserine (DCS). A partial agonist at the NMDA receptor, which is critical for neural plasticity and memory consolidation. Preclinical models showed that DCS powerfully facilitates fear extinction when paired with exposure. Early human trials for acrophobia and social anxiety generated impressive results. (Norberg et al., 2008, Biological Psychiatry)*

However, a comprehensive 2017 meta-analysis of 20 RCTs involving 957 participants found only a small, non-significant overall benefit (Hedges' g = 0.15 for treatment efficacy, RR = 1.08 for treatment response). (McGuire et al., 2017, Journal of Clinical Psychiatry) A moderate effect persisted only within the anxiety disorders subgroup (g = 0.33, p = 0.03), with zero benefit for OCD or PTSD.

The resolution to the DCS paradox lies in its mechanism: DCS is a memory enhancer, not an anxiolytic. It amplifies whatever learning was dominant at the end of the session. If a patient achieves low fear at session end, DCS robustly enhances the safety memory. If the patient remains highly fearful, DCS actually worsens outcomes by consolidating the fear memory (a process of fear reconsolidation). (Smits & Hofmann, 2013, Biological Psychiatry) Additional moderators: concurrent SRI use severely and negatively impacts DCS efficacy, as chronic SRI treatment may alter the glutamatergic neuroplasticity pathways DCS targets. Comorbid depression also significantly dampens extinction learning. (McGuire et al., 2017, Journal of Clinical Psychiatry)

DCS remains an experimental adjunct requiring expert clinical judgment about session quality. It should not be prescribed as a routine augmentation strategy.

Glucocorticoid augmentation. Hydrocortisone enhances consolidation of new memories while inhibiting retrieval of old emotional memories, a profile theoretically suited to exposure therapy. RCTs by Soravia and de Quervain demonstrated augmented exposure effects for spider phobia and acrophobia. (Soravia et al., 2006, PNAS; de Quervain et al., 2011, PNAS) However, clinical translation remains hindered by inconsistent dosing protocols and substantial individual differences in baseline neuroendocrine function.

MDMA-assisted therapy for social anxiety. The Social Anxiety MDMA-Assisted Therapy Investigation (SAMATI, NCT05138068) is evaluating MDMA-assisted psychotherapy for treatment-resistant social anxiety. A pilot study in autistic adults demonstrated rapid, durable reductions in social anxiety. (Danforth et al., 2018, Psychopharmacology) MDMA may facilitate a state of temporary neuroplasticity where deeply entrenched, shame-based core beliefs can be safely accessed, but this remains an experimental frontier without phase-III validation.

7.5 Benzodiazepines and Exposure Therapy: The Extinction Learning Problem

This is a clinically consequential topics in anxiety psychotherapy, and the evidence demands careful attention. The full benzodiazepine pharmacological framework is covered in Module 3 (Pharmacotherapy). This section addresses the specific question of how concurrent benzodiazepine use interacts with exposure-based psychotherapy.

The animal data. Foundational research by Bouton and colleagues (1990) demonstrated that fear extinction conducted under benzodiazepine influence fails to transfer to a drug-free state in a dose-dependent manner. (Bouton et al., 1990, Behavioral Neuroscience) Animals appeared fear-free during drugged extinction, but fear aggressively returned upon drug washout. The critical insight: the benzodiazepine acts as an "interoceptive context," meaning the brain associates safety not with the external environment but with the internal pharmacological state. When the drug is removed, the interoceptive context shifts and the original fear memory is "renewed." (Bouton et al., 1990, Behavioral Neuroscience)

Human experimental data. Grillon and colleagues (2006) used fear-potentiated startle paradigms to show that alprazolam selectively reduces sustained contextual anxiety (the diffuse dread of an unpredictable environment) but completely fails to suppress the phasic fear response to specific, predictable threat cues. (Grillon et al., 2006, Biological Psychiatry) This means that a patient taking a benzodiazepine before an exposure session may feel generally calmer but their acute fear response to the specific phobic stimulus remains intact, creating a contaminated learning environment where the patient is sedated but still reactively fearful.

How benzodiazepines may interfere with exposure. The proposed mechanisms operate at multiple levels. At the most basic level, benzodiazepines artificially suppress physiological arousal, which reduces the mismatch between the patient's catastrophic prediction and their actual experience. If the patient does not feel the feared physical sensations during the exposure, the expectancy violation that drives inhibitory learning is blunted. The patient attributes survival to the drug rather than to the situation's inherent safety. (Craske et al., 2014, Behaviour Research and Therapy)

A related problem is state-dependent learning: safety memories formed under benzodiazepine influence may fail to generalize to unmedicated states. The brain associates safety with the internal pharmacological state rather than with the external environment. When the drug is tapered, that internal context shifts and the safety memory becomes less accessible.

At a neurobiological level, benzodiazepines suppress the stress response system (the HPA axis) in a dose-dependent manner. When the brain encounters a fear-provoking situation, it normally releases cortisol, a hormone that plays a key role in consolidating new memories, including the safety memories formed during exposure. Benzodiazepines block this cortisol spike, which appears to be required for extinction memory consolidation. (Gram & Christensen, 1986, via Soravia et al., 2006, PNAS)

Finally, the pill itself can function as a safety behavior. Taking a benzodiazepine "just in case" before an exposure session converts the exposure into an exercise in avoidance, the patient endures the situation but credits the medication, not their own capacity to tolerate distress.

Clinical outcome data by disorder:

For panic disorder and agoraphobia, the evidence most clearly demonstrates harm. The Marks et al. (1993) trial randomized 154 patients to alprazolam plus exposure, alprazolam plus relaxation, placebo plus exposure, or placebo plus relaxation, using mean alprazolam doses of 5.8 mg/day (higher than standard contemporary clinical practice). During acute treatment, the combination appeared slightly superior. Upon drug discontinuation, these gains entirely reversed: the relative risk for clinical response dropped from 1.25 during treatment to 0.62 at follow-up, while the exposure-alone group maintained their gains. (Marks et al., 1993, British Journal of Psychiatry) Patients who attributed their improvement to medication rather than to their own efforts suffered significantly higher relapse rates and more severe withdrawal. (Basoglu et al., 1994, British Journal of Psychiatry) [Note: This trial was funded by Upjohn, the manufacturer of alprazolam, yet the independent analysis demonstrated that alprazolam ultimately impaired long-term recovery.]

For PTSD, the picture is more nuanced. A post hoc analysis of 283 female veterans found that benzodiazepine users did not have a weaker response to Prolonged Exposure therapy. (Rosen et al., 2013, Journal of Clinical Psychiatry) However, BZD use was strongly associated with poorer post-treatment maintenance for those receiving the less intense Present-Centered Therapy. The authors concluded that PE is "sufficiently robust" to overpower BZD interference, but less intensive protocols cannot. (Rosen et al., 2013, Journal of Clinical Psychiatry) The mean BZD dose in this trial (approximately 1.9 mg/day clonazepam equivalent) was lower than the Marks study's extreme doses. Many veterans consumed BZDs primarily at bedtime, potentially minimizing overlap with waking exposure sessions.

PRN (as-needed) use is a notable concern. A naturalistic study by Westra and colleagues (2002) found that PRN benzodiazepine use during CBT for panic disorder was a significant negative predictor of CBT outcome. (Westra et al., 2002, Journal of Anxiety Disorders) PRN users showed the fewest clinical gains, less reduction in anxiety sensitivity, and less reduction in anxious arousal compared to unmedicated peers. Patients on scheduled daily doses did not differ significantly from unmedicated subjects. The proposed mechanism: PRN use reinforces a learned loop between the onset of anxiety and pharmacological relief (each time anxiety rises and the pill brings relief, the brain strengthens the connection between "feel anxious → take pill → feel better"), functionally converting the medication into a safety behavior. This is a single naturalistic (non-randomized) study, and the finding has not been tested in a controlled trial designed to isolate the PRN variable. However, the clinical reasoning is consistent with the broader inhibitory learning framework, and the finding aligns with the animal data on state-dependent learning.

Important Note: Managing BZDs During Active Exposure Therapy
If a patient is initiating exposure therapy, PRN benzodiazepine use warrants careful discussion. The available evidence (a single naturalistic study, supported by the inhibitory learning framework) suggests PRN use is associated with poorer CBT outcomes because taking a pill "just in case" before an exposure session may blunt the expectancy violation that drives learning and convert the exercise into avoidance. This is a clinical judgment that should be weighed against the patient's individual risk profile.
For patients already on a stable, scheduled benzodiazepine: the evidence supports neither abrupt discontinuation (which risks dangerous withdrawal and guarantees dropout) nor indefinite maintenance (which undermines long-term gains). The evidence-based approach is concurrent CBT combined with a gradual taper.
The 2025 Joint Clinical Practice Guideline on Benzodiazepine Tapering endorses hyperbolic tapering: non-linear sequential dose reductions of 5-10% of the current dose every 2-4 weeks, never exceeding a 25% reduction in any two-week period. CBT, particularly interoceptive exposure targeting the overlap between withdrawal symptoms and panic symptoms, significantly increases the rate of successful long-term discontinuation. Cross-reference Module 3 for the full benzodiazepine framework.

In plain language for families:
*If you or your child is taking a benzodiazepine (like Xanax, Klonopin, or Ativan) and also doing exposure therapy, there is evidence that the two treatments may work against each other in some cases. The anxiety medication reduces the physical feelings of fear, but the therapy may need those feelings to teach the brain that the situation is safe.
Think of it like wearing earplugs during a fire alarm drill: you might feel calm, but you are not fully learning that the alarm is not dangerous. If your provider recommends tapering the medication, this is not because the medication is "bad," but because the therapy may work better when the brain can fully experience and process the fear.
The best approach is usually a slow, gradual reduction of the medication while therapy continues, so the brain can build its own confidence. This is a conversation to have with your treatment team, who can weigh the benefits and risks for your specific situation

PsychHQ

Combined Treatment: Medication Plus Psychotherapy