Most research on online therapy studies people who signed up for a trial, were screened, and knew someone would be checking on them. Farvolden and colleagues looked at what happens when a program is simply put on the internet for anyone to use.
The Panic Center was a free website offering a 12-session CBT program for panic disorder and agoraphobia. Between September 2002 and February 2004 it drew 99,695 users and nearly half a million visits. Of those, 1,161 registered for the program itself. Twelve people finished all twelve sessions, which is 1.03% of those who registered.
That number is the headline, and it is worth sitting with. Free, unsupported, self-directed treatment loses almost everyone who starts it.
The second finding is more encouraging. Among the people who stayed even briefly, panic attacks became less frequent and less severe, and the reductions were statistically significant. Of the 152 users who completed only three sessions, the average number of attacks per day fell from 1.03 in week two to 0.63 in week three.
Both findings are real and they belong together. A self-help program can genuinely reduce panic attacks for the people who use it. Very few people use it without support. If you have tried an online program and stopped, you were in the large majority, and it says nothing about whether treatment can work for you.
Why this study is unusual
By 2005 several controlled trials had shown that web-based self-help for anxiety disorders could work. Controlled trials, however, recruit and screen participants, follow them up, and by their nature create accountability. Very little was known about how freely available online interventions actually behave outside that setting.
Farvolden and colleagues at the Centre for Addiction and Mental Health in Toronto studied a free, publicly available, 12-session web-based CBT program for panic disorder and agoraphobia, hosted at a site called the Panic Center. They analysed cumulative anonymous usage data alongside a longitudinal survey of self-reported symptoms among users who registered for the program.
Method
Usage statistics were examined for the site as a whole. For registered program users, the primary outcome measures were self-reported panic attack frequency and severity, collected at the beginning of each session from session 2 through session 12. Because the measures were taken at the start of each session, the data for any given week exist only for those who returned that week.
Usage
Between 1 September 2002 and 1 February 2004 the Panic Center received 484,695 visits and 1,148,097 page views from 99,695 users. During the same period 1,161 users registered for the CBT program.
Twelve of those 1,161 registered users completed the 12-week program, which is 1.03%.
Symptom change
Despite the attrition, the symptom data among those who remained showed statistically significant reductions in self-reported panic attack frequency and severity, at a threshold of P<.002, when two weeks of baseline data were compared against data after 3, 6 or 8 weeks.
The authors give a concrete example. Among the 152 users who completed only three sessions, the average number of panic attacks per day fell from 1.03 at week 2 to 0.63 at week 3, a difference significant at P<.001.
How to interpret both halves
The two findings pull in different directions and both are true.
The attrition figure describes what happens when a treatment is made available with no cost, no screening, no clinician and no accountability. Registering for a program is not the same as committing to it, and the enormous gap between 99,695 site users and 1,161 registrations, then between 1,161 registrations and 12 completions, shows how many decision points a person passes through before finishing.
The symptom figure is subject to an obvious selection effect, which the authors acknowledge by describing those who stayed as a highly self-selected group. People who continue may be those who are improving, or those with more capacity to persist. Without a control group, ordinary fluctuation in panic frequency cannot be ruled out either. These are uncontrolled observational data.
What survives both caveats is that the material itself was capable of being associated with meaningful symptom reduction quite quickly, within a few sessions, for people who engaged with it.
What it means in practice
The study's conclusion is stated plainly by the authors: freely available web-based self-help will likely be associated with high attrition, but for the group who stay, significant improvements were observed.
For someone considering a digital program, the useful implication is that the content is not usually the problem. Continuing is. Whatever raises the chance of continuing, such as a scheduled review with a clinician, a set time each week, or someone who knows you are doing it, is likely to matter more than which program you pick.
Limits
There was no control group, no diagnostic verification of users, and no independent measurement. Outcomes were self-reported and collected only from those who returned, so the data are progressively drawn from a narrower and more persistent group. The site and its program date from the early 2000s, and the internet population then was not the internet population now.
Panic disorder responds well to CBT delivered with a clinician. If a self-directed program has not held your attention, that is an argument for adding support rather than for concluding that treatment will not work.