Health systems have to decide what to fund. Online CBT is attractive to funders because it is cheap and scales easily, but cheap is only useful if the treatment does something. A Dutch trial published in 2010 measured both sides at once.
Gerhards and colleagues recruited 303 people from the general population in the south of the Netherlands through online screening, all with at least mild to moderate depressive symptoms lasting three months or more. They were randomly assigned to one of three arms: an unsupported online CBT program with no professional contact, usual care from their GP, or both together. Costs, depression severity and quality of life were tracked for twelve months, counting costs to society rather than to the health system alone.
Costs were lowest in the online CBT arm. On the clinical side there were no significant differences between the three groups in depression severity or quality of life, and quality of life across all three stayed close to where it started. Adherence was low everywhere, with online CBT users completing an average of five lessons.
Taken together, the economic analyses favoured online CBT, and the authors concluded it was the most efficient of the three strategies. The reason is not that it worked better. It is that nothing worked especially well and it cost the least.
That is a genuine finding about unsupported self-help, and it makes the case for keeping a person involved.
The question
Depression is common and expensive, and many people with it in primary care do not receive adequate treatment. Evidence for medication in mild to moderate depression is limited, and psychological treatment in primary care is scarce and costly. Computerised CBT, abbreviated CCBT, is a candidate solution, and NICE had already recommended two such programs for depression and anxiety. What was missing was evidence on whether it represents good value.
The authors had already compared the clinical effectiveness of unsupported online CCBT against usual care and against the two combined. This paper reports the economic evaluation of the same trial.
Design
In a randomised trial, 303 participants were recruited from the general population by large-scale internet-based screening in the south of the Netherlands. Inclusion criteria were age 18 to 65, home internet access with a broadband connection, a Beck Depression Inventory II score of at least 16, depressive complaints lasting three months or more, no current psychological treatment for depression, no continuous antidepressant treatment for at least three months before entry, fluency in Dutch, no alcohol or drug dependence, and no severe psychiatric comorbidity.
Three arms were compared. The CCBT program, Colour Your Life, is an online multimedia interactive program consisting of eight weekly sessions plus a ninth booster session, with homework assignments and a mood diary, and no professional assistance of any kind. Treatment as usual meant contacting the participant's own GP, whose practice guideline allows for four to five biweekly consultations with antidepressants if indicated. The third arm received both.
Participants completed monthly internet questionnaires over a 12-month follow-up.
How costs and outcomes were measured
The evaluation took a societal perspective, covering healthcare sector costs, costs to patient and family, and productivity costs. Because it is often unclear whether healthcare use or lost productivity in depression is attributable to the depression itself, all related and unrelated costs were included. Healthcare use was captured by a monthly questionnaire, and CCBT usage came from the program's own login records.
Effectiveness was measured by the BDI-II at 12 months. Utility was measured by the EQ-5D and SF-6D and converted into quality-adjusted life-years, or QALYs. Uncertainty was handled with 5000 bootstrap replications and with sensitivity analyses that varied the baseline correction method, the time horizon, the cost perspective, the productivity costing method, the tariff used to value the EQ-5D, the utility measure and the outcome definition. Because the real-world price of self-help CCBT is uncertain, its cost price was varied across a range of fixed prices per user and per lesson.
Results
Costs were lowest for the CCBT group.
There were no significant group differences in effectiveness or quality of life. In the base-case analysis, mean QALYs were 0.71 for CCBT, 0.71 for CCBT plus usual care and 0.72 for usual care. Those figures held at about 0.70 regardless of correction method or utility measure used, and were very close to the mean baseline utilities of about 0.70 in each group, which indicates little improvement in quality of life over the year in any arm.
The cost-utility analysis was expressed as the probability of each strategy being the most efficient at different willingness-to-pay thresholds. CCBT had about a 65% probability of being most efficient at a threshold of €0 per QALY, falling to about 40% at €80,000. Usual care ran the other way, from about 25% at €0 to about 40% at €80,000, so at the top of that range the choice between them becomes close to indifferent. The combined arm had the lowest probability throughout, between 10% and 20%.
Adherence was low in all groups, with CCBT users completing an average of five lessons of the nine.
What the conclusion actually says
The authors conclude that on balance CCBT constitutes the most efficient treatment strategy, while stating in the same sentence that all treatments showed low adherence rates and modest improvements in depression and quality of life.
Reading only the first half of that sentence would be a mistake. This is an efficiency result driven substantially by cost, in a trial where none of the three strategies produced a clear clinical gain. Unsupported CCBT looks good here relative to alternatives that also did not achieve much.
Why adherence keeps appearing
Five lessons out of nine, in a program with no professional contact, is consistent with what the wider literature reports about unsupported digital treatment. The programs are not usually the weak point. Sustained engagement is. The comparison that this trial does not make, and that matters most practically, is between unsupported CCBT and CCBT delivered with some form of human support.
Limits
Participants were recruited by internet screening from the general population rather than from clinical services, and were required to have home broadband, so they may differ from people presenting to a GP with depression. Usual care was whatever each GP provided, which varies. The follow-up was 12 months, and cost prices for self-help software in real-world use remain uncertain, which is why the authors varied them.
If a self-directed online program has not shifted things for you, that experience is reflected in this trial's data. Treatment with a psychologist involved is a different proposition, and a reasonable next step.