This study evaluated the cost-effectiveness of three treatments for chronic fatigue, lasting more than three months. The authors concluded that no clear economic advantage was found for any therapy; counselling was not cost-effective, and graded exercise needed a high value for a clinically significant improvement, for it to be cost-effective. The study was generally well reported, with clear assumptions. The conclusions are reasonable, but should be considered to be uncertain, due to a few limitations.

Cost-effectiveness analysis

This study evaluated the cost-effectiveness of three treatments for chronic fatigue, lasting more than three months.

The three therapies were graded exercise, counselling, and usual care with a self-help booklet. Graded exercise comprised eight personalised exercise sessions at a local general practice, followed by two follow-up phone calls. Counselling comprised eight 60-minute counselling sessions, designed to build patient understanding of their condition through discourse.

UK/primary care.

Analytical approach:

The evaluation was based on a single-centre, three-arm, randomised controlled trial. A total of 222 people were randomised. The authors stated that they took a health service perspective.

Effectiveness data:

All the effectiveness data were from the randomised controlled trial. The primary measure was the Chalder Fatigue Scale score, which had 13 items assessed using Likert scales (0, 1, 2, or 3) to produce a score ranging between 0 and 33. Patients were followed-up for six months. Based on the published literature, it was assumed that a six-month change of 4 on the Chalder Fatigue Scale was clinically significant. To calculate clinically significant changes, the Chalder scores at six months were divided by four. Analyses were carried out for patients with complete data, for clinical outcomes and costs, at six-month follow-up. Multiple regression was used to control for differences in baseline statistics, including cost differences.

Monetary benefit and utility valuations:

Not relevant.

Measure of benefit:

The measure of benefit was an improvement of 4 points on the Chalder Fatigue Scale.

Cost data:

The resource use was collected retrospectively from trial participants, using the Client Service Receipt Inventory. Measures were taken at six months and at baseline. The data included health and social care services, complementary health care, antidepressant medication, anti-anxiety medication, sleeping medication, and the duration of appointments, where available. The unit costs were from the British National Formulary, NHS reference costs, and Personal and Social Service Research Unit costs. For complementary medicine, service costs were from published studies. The price year was 2006 to 2007, and costs were in UK £.

Analysis of uncertainty:

Probabilistic sensitivity analysis was conducted, by bootstrapping from the trial data, to create cost-effectiveness acceptability curves.

Only 163 of the 222 participants randomised (51 graded exercise, 58 counselling, and 54 usual care) had full economic and outcome data at six months, representing a 27% loss to follow-up. Loss to follow-up was comparable across the three trial groups.

The total cost of graded exercise was £473.60, and the total improvement in Chalder score was 10.06. The total cost of counselling was £650.80, and the total Chalder score was 8.62. The total cost of usual care was £212.90, and the total improvement in Chalder score was 8.56. The differences in effectiveness, between the interventions, were not statistically significant.

Based on these non-significant differences, counselling was dominated by both other interventions, as it was less effective and more costly. Graded exercise, compared with usual care, had an incremental cost-effectiveness ratio of £987 per clinically significant change in Chalder score.

At a threshold of £1,000 per clinically significant improvement in Chalder score, graded exercise had the highest likelihood of cost-effectiveness at 55%. At willingness-to-pay thresholds up to £100,000 per clinically significant change in Chalder score, the likelihood of graded exercise being the most cost-effective treatment never exceeded 65%, while the likelihood of counselling being the most cost-effective treatment never exceeded 17%.

The authors concluded that no clear economic advantage was found for any of the therapies. Counselling was not cost-effective, and graded exercise needed a high value for a clinically significant improvement, for it to be cost-effective.

Interventions:

The interventions were sufficiently described and appear to have been appropriate. No comparisons from outside the trial were sought. It was not clear whether there were other relevant comparators.

Effectiveness/benefits:

The definition of a clinically significant change in Chalder score was based on an author's reply to a previous study. It was not clear whether there was a broad consensus on this, and the authors acknowledged its uncertainty. Only a few trial methods were reported. It was not clear how patients were randomised, and it seems that no blinding took place. The authors acknowledged the limitations of the per-protocol analysis, but imputation of the missing information could have created more issues. Due these limitations, there may be bias in the results.

Costs:

The perspective was clearly reported and all the relevant costs appear to have been included. The sources for the unit costs appear to have been appropriate, and the resource use categories were relevant. Both of these were well reported. Gathering the resource use retrospectively, for the six months before each patient survey, introduced potential for recall bias, and further validation of the resource use was necessary to reduce this uncertainty.

Analysis and results:

The outcome measure was clinically meaningful, but lacks generalisability for decision making. The incremental cost-effectiveness analysis and the probabilistic sensitivity analyses appear to have been appropriately conducted and were clearly reported.

Concluding remarks:

The study was generally well reported with clear definitions of the assumptions. The conclusions are reasonable, but should be considered to be uncertain, due to the limitations outlined.

Funded by the Wellcome Trust.