Six approaches have been studied in depth for health anxiety. We evaluate each by how strong the evidence is, what tempers it, and what trials it rests on.
The short answer: cognitive behavioural therapy (CBT), including exposure and response prevention (ERP), has the strongest evidence and is what the NHS offers. Nothing has been shown to beat it in a fair comparison. If you cannot reach a therapist, internet-delivered CBT has been tested against face-to-face therapy and was not found to be worse over the course of treatment. There is no dedicated NICE guideline for health anxiety, so NHS provision follows guidance written for related conditions rather than this one; NHS advice is to try self-help materials and see a GP if things do not improve or start affecting daily life.
Sample sizes and outcome rates read straight out of the abstracts. Each figure is a median across every paper that reported one, so it describes the literature rather than any single study.
213,203
People across all trials
1,398 trials reporting a sample size
72
Median trial size
smallest 10, largest 12,600
56%
Median response rate
reported by 185 papers
0.84
Median effect size
across 356 trials and meta-analyses
"Strength of evidence" here means three things: how many trials exist, whether they compared the treatment against something active rather than a waiting list, and whether the benefit was still there at follow-up. It does not mean effect size alone, and this is why.
A treatment measured against a waiting list will always look stronger than one measured against a credible alternative, because waiting produces no improvement while any plausible treatment produces some. In this exact population, one meta-analysis found CBT at g = −1.32 against waitlist and g = −0.58 against placebo. Same treatment, same disorder, half the effect. Ranking treatments on headline numbers without checking what they were compared against produces a league table that mostly measures study design.
So where a comparison is against a waiting list, we say so. Where a finding comes from routine care rather than a trial, we say that too — those numbers run lower, and they are the more useful guide to what to expect.
The most tested treatment for health anxiety by a wide margin, and the one the NHS offers. Multiple meta-analyses find large effects at post-treatment, with benefits still present at follow-up. The CHAMP trial followed patients for eight years and found the advantage held, though it shrank from 2.98 to 1.83 points on the Health Anxiety Inventory.
What tempers it: Two things temper this. Effects against a waiting list run roughly twice those against an active comparison — in the same population, one meta-analysis found g = −1.32 against waitlist and g = −0.58 against placebo. And publication bias is documented here: adjusting for two likely-missing trials pulls the pooled effect from −0.70 to −0.60.
“Based on 19 RCTs, the pooled between-group effect on health anxiety was moderate to large (g = 0.79; 95% CI: 0.57-1.01; adjusted for publication bias: g = 0.62), with small to moderate effects on secondary symptoms and effects largely sustained 12-18 months after treatment.”
“The pooled CBT response rate was 66%, and the remission rate 48%.”
“RESULTS: Meta-analysis was conducted on 21 comparisons and a large effect size for CBT compared with a control condition was found at post therapy d = 1.01 (95% confidence interval 0.77-1.25), as well as at 6- and 12-month follow-up.”
“Results indicated that CBT outperformed control conditions on primary outcome measures at post-treatment (Hedges's g = 0.95) and at follow-up (Hedges's g = 0.34).”
“CBT also outperformed control conditions on measures of depression at post-treatment (Hedges's g = 0.64) and at follow-up (Hedges's g = 0.35).”
“In the original investigation, 17% reported at least one adverse event, 0-10% met criteria for overall symptom deterioration, and 10-19% dropped out.”
Usually delivered as part of CBT, though the trials tested it as its own treatment: facing the trigger while not performing the checking, googling or reassurance-seeking that normally follows. Tested head-to-head against cognitive therapy in a randomised trial, where both worked and exposure held a small edge on safety behaviours specifically — the compulsions themselves.
What tempers it: The head-to-head trial compared both against a waiting list, so the effect sizes are the inflated kind. In routine outpatient care rather than trial conditions, response rates were 51–63% and remission 29–43%, which the authors note is lower than trials report.
“We analysed putative mediators of change based on 13 weekly assessments in a randomised controlled trial (N = 132) of exposure-based minimal-contact CBT (guided Internet-delivered CBT, unguided Internet-delivered CBT and bibliotherapy) vs. a waitlist control for severe health anxiety.”
“In the present study, latent process growth modeling showed that treatment condition had a significant effect on the FFMQ-NR growth trajectory (α-path), estimate=0.18, 95% CI [0.04, 0.32], p=.015, indicating a larger increase in mindful non-reactivity among participants receiving exposure-based CBT compared to the BSM group.”
“The FFMQ-NR growth trajectory was significantly correlated with the SHAI trajectory (β-path estimate=-1.82, 95% CI [-2.15, -1.48], p<.001.”
“RESULTS: Linear mixed effects models analysis showed a significant interaction effect of group and time indicating superior improvements in SRH in exposure-based CBT compared to the control condition (Z=2.69, p=0.007).”
“The controlled effect size was moderately large (d=0.64) and improvements were stable at 1-year follow-up.”
“RESULTS: Compared with the WL, both CT and ET reduced the frequency of somatic attributions regarding severe diseases (CT: Hedges's g = 1.12; ET: Hedges's g = 1.03) and increased the frequency of normalizing attributions (CT: Hedges's g = 1.17; ET: Hedges's g = 1.24).”
The most practically important finding for anyone who cannot access a therapist. A noninferiority trial found internet-delivered CBT was not worse than face-to-face CBT for health anxiety over the treatment period, with a between-group difference of 0.00 points. That trial was not powered to test the follow-up period, and at 12 months face-to-face held a small lead. Internet-delivered ACT has also been tested against an active control rather than a waiting list, which is the harder test.
What tempers it: These programmes generally involve a therapist — through messaging, review, or a supervising clinician. That matters for how far the result transfers to unsupported self-help, which is the next section.
“Post-treatment, iCBT showed no significant effect difference vs. other psychotherapies (Hedges'g [95% CI]: general symptoms = 0.09 [-0.11; 0.28]; health anxiety: = 0.18 [-0.17; 0.52]), but outperformed waitlisting across most symptoms (Hedges'g [95% CI]: general symptoms = 0.30 [0.12; 0.48]; gastrointestinal = 0.62 [0.35; 0.88]; fatigue = 0.47 [0.25; 0.68]; health anxiety = 0.71 [0.26; 1.16]).”
“The effect of iCBT was not significantly different or slightly higher than treatment as usual after treatment (Hedges'g [95% CI]: gastrointestinal = 0.41 [0.23; 0.58]; fatigue = 0.65 [-0.51; 1.82]; fibromyalgia = 0.27 [0.00; 0.55]).”
“The results showed that internet-based psychological interventions had a moderate to large between-group effect on health anxiety (g = 0.70) that was significant both at post-intervention (g = 0.74) and follow-up (g = 0.64).”
“Furthermore, these interventions were significantly more effective than passive control conditions including waitlist, usual care, and placebo at post-treatment (g = 1.07), but had effects comparable to active control groups at both post-intervention and follow-up.”
“The random-effects meta-analysis revealed a significant positive correlation between online health information seeking and health anxiety (r = 0.28, 95% confidence interval [0.16, 0.41], p < .0001), despite considerable heterogeneity.”
“Random effects meta-analyses showed that there was a positive correlation between health anxiety and online health information seeking [r = 0.34, 95% CI (0.20, 0.48), p < .0001], and between health anxiety and cyberchondria [r = 0.62, 95% CI (0.52, 0.71), p < .0001].”
Third-wave CBT: rather than arguing with the fear, making room for it and acting anyway. Two randomised trials in severe health anxiety specifically, one group-delivered against a waiting list and one internet-delivered against an active control, both with large effects and a number needed to treat under three.
What tempers it: Fewer trials than CBT, and the group trial used a waiting-list comparator. The internet trial's active control makes it the more informative of the two.
“Also, the intervention group compared to the control group had better QOL after the intervention in general health perceptions (3.03±0.96 vs. 2.43±0.95, P=0.01), mental health (7.12±2.25 vs. 6.34±1.85 and P=0.01) and mental component summary) 16.78±3.75 vs. 15.43±3.05, P=0.01), physical component summary (16.06±2.66 vs. 15.19±2.25, P=0.01), and total score of SF-12 (32.84±5.39 vs. 30.62±4.34, P=0.004).”
“RESULTS: The results showed that the intervention group compared to the control group had a significant decrease in all subscales of HAI after the intervention including worry about consequences (5.78±2.66 vs. 7.37±1.34, P=0.004) and awareness of bodily sensation or changes (8.90±2.77 vs. 11.75±2.30, P=0.001), worry about health (10.94±2.38 vs. 13.09±1.92, P=0.001), and total score of HAI (25.62±4.93 vs. 32.25±3.93, P=0.001).”
“The number needed to treat was 2.8 (95% CI 1.8-6.1, p < 0.001), and twice as many patients in iACT were no longer clinical cases (35% v. 16%; risk ratio 2.17, 95% CI 1.00-4.70, p = 0.050).”
“A mean difference in change over time of 19.0 points [95% confidence interval (CI) 10.8-27.2, p < 0.001] was shown on the WI-7, and a large standardized effect size of d = 0.80 (95% CI 0.38-1.23) at 6-MFU.”
“RESULTS: Statistically significant indirect effects (IE) of ACT-G on the outcome of illness worry 6 months after treatment were found for PF (IE = -5.5, BCa 99% CI -12.3;-1.2) and one mindfulness subscale, namely 'non-react' (IE = -6.5 BCa 99% CI -15.3: 1.0).”
“RESULTS: Intention-to-treat analysis showed a statistically significant mean difference of 20.5 points [95% confidence interval (CI) 11.7-29.4, p < 0.001] on the WI between the groups at 10 months, and the between-group effect sizes were large (Cohen's d = 0.89, 95% CI 0.50-1.29).”
Tested in a randomised trial against unrestricted usual services, with lower health anxiety in the MBCT group at both post-treatment and one-year follow-up. Fewer people still met criteria for hypochondriasis afterwards.
What tempers it: A single trial in health anxiety specifically, with a treatment-as-usual comparator. The effect size (d = 0.48) is moderate rather than large.
“MBCT + Usual Services: n=36; Usual Services alone: n=38; Post-Treatment ITT d=0.48; 1-Year Follow-Up ITT d=0.48; Post-Treatment: MBCT 50.0% vs Usual Services 78.9%; 1-Year Follow-Up: MBCT 36.1% vs Usual Services 76.3%”
“Significantly fewer participants allocated to MBCT than to US met criteria for the diagnosis of hypochondriasis, both immediately following the intervention period (ITT 50.0% vs. 78.9%; PP 47.1% vs. 78.4%) and at 1-year follow-up (ITT 36.1% vs. 76.3%; PP 28.1% vs. 75.0%).”
“RESULTS: In the intention-to-treat (ITT) analysis (N = 74), MBCT participants had significantly lower health anxiety than US participants, both immediately following the intervention (Cohen's d = 0.48) and at 1-year follow-up (d = 0.48).”
A small number of trials, showing a modest effect against pill placebo. One meta-analysis found no statistically significant difference between CBT and SSRIs where they were compared directly.
What tempers it: Only three SSRI arms across the meta-analysis, totalling 193 participants, against more than 1,200 in the CBT arms. The direct CBT-versus-SSRI comparison rested on two trials, and its small numerical trend favoured medication rather than CBT — a difference too small and too thinly evidenced to act on. Medication decisions belong with a doctor.
“Four of the studies showed the pooled estimates of mean difference in the change of HAMA scores between acupuncture plus paroxetine or duloxetine group and medication alone group were statistically significant at week 4 (-1.94, 95%CI: -3.71 to -0.17; p = 0.03) with borderline significance at week 6/8 (-3.17, 95%CI: -6.38 to 0.04; p = 0.05) from baseline.”
“RESULTS: The predicted pattern of response was statistically significant, as shown by the following responder rates: joint treatment group, 47.2%; single active treatment group, 41.8%; and placebo group, 29.6%.”
“Because approximately 50% of patients did not respond to the study treatments, new or more intensive approaches are needed.”
“The primary analysis assessed outcome at week 24 among the intent-to-treat sample, with responders defined as having a 25% or greater improvement over baseline on both the Whiteley Index and a modified version of the Yale-Brown Obsessive Compulsive Scale for hypochondriasis (H-YBOCS-M).”
“More patients responded with improvement in hypochondriasis when given fluoxetine compared with placebo, starting at week 8 (50.0% vs 19.0%, P = 0.03) and continuing to week 12 (62.5% vs 33.3%, P = 0.05).”
“The acute treatment response was maintained to week 24 with more responders in the fluoxetine compared with the placebo group (54.2% vs 23.8%, P = 0.04).”
We make one, so here is the position plainly. No randomised controlled trial of any smartphone app for health anxiety exists, ours or anyone else's. That is a fact about the whole category, and we are not going to imply otherwise.
There is no NICE guideline for health anxiety. NHS advice is to try self-help materials and see a GP if things do not improve or start affecting daily life. Condri was built to fill the self-help gap.
The other gap is the one therapy leaves. CBT happens for an hour a week; health anxiety happens at 2am when you have found a lump and the search bar is open. Condri is built for that moment — the same CBT, ERP and ACT techniques a therapist would teach, in a form you can actually reach when the urge to check is strongest. Used that way it sits alongside treatment rather than in place of it.
Treatment does not work for everyone, and what happens when it does not is worth reading before you start.
If you are already working with a therapist or a GP, the sensible thing is to show them. Ask whether the exercises fit what you are doing together, and whether daily practice between sessions would help or get in the way. They know your situation; we do not.
We want to study this properly, and are working towards a trial with academic partners so this section can eventually rest on evidence rather than reasoning. Until it does, the honest summary is that the methods are well evidenced and the delivery is not yet.
If you can access CBT — through NHS Talking Therapies, which you can refer yourself to in England without going through a GP — take it.
Every finding above is drawn from a corpus of 10,657 papers and checked against the source abstract. The full evidence base, including prevalence, cost and mechanism, is at the evidence base, and the data is available as JSON.