This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.

Three technologies for weight loss were assessed:

a do-it-yourself workbook approach;

an expert computer system together with the workbook (intermediate intervention); and

staff consultation in addition to the computer system and workbook (intensive intervention).

All of the interventions used a cognitive-behavioural approach. A computer session lasted 20 to 30 minutes and was encouraged every week for the first 3 months and once a month thereafter. Staff consultation comprised 6 closed-group sessions and up to 18 telephone or face-to-face consultations with a registered dietician and/or cognitive-behavioural therapist.

Treatment of obesity and secondary prevention of obesity-related diseases.

Cost-effectiveness analysis.

The study population comprised individuals with a body mass index (BMI) of more than 25, or a BMI of 24 or more with one cardiovascular risk factor.

The setting was the community or primary care. The economic study was carried out in New York, USA.

The dates when the effectiveness and resource use data were collected were not reported. A price year was not reported.

The effectiveness evidence was derived from a single study.

The costing was carried out prospectively on the same sample of patients as that used in the effectiveness study.

The authors used assumptions from related research to determine the power needed to rule out the influence of chance on their results. They aimed to detect a weight loss of 8 pounds (lb) between the workbook and intensive interventions with 0.99+ power, a loss of 4.5 lb between the workbook and intermediate interventions with 0.82 power, and a loss of 3.5 lb between the intermediate and intensive intervention with 0.80 power. This gave sample sizes of at least 80 individuals in the workbook group, and 171 in each of the intermediate and intensive interventions. The participants were recruited from the HMO's patient population and surrounding community using a general advertising campaign. Interested individuals attended a recruitment session that provided study details and eligibility criteria. The eligibility criteria included the correct BMI level, and a willingness to follow study protocols such as paying a $100 refundable deposit. The exclusion criteria included the intention to move beyond a commutable distance within the next 12 months, and medical conditions that would interfere with study participation. Participants were also excluded if they did not complete brief surveys and computerised baseline questionnaires.

The initial sample was appropriate for the study question since it included patients who were obese and who might benefit from the technologies under assessment. A total of 588 individuals were randomised to the study arms. Of these, 116 were randomised to workbook only, 236 to workbook plus computer, and 236 to workbook, computer and staff consultation.

The basis of the analysis was a randomised controlled trial. The unit and method of randomisation were not reported. The study was conducted in an HMO located in Albert Einstein College of Medicine, Long Island campus, adjacent to the Long Island Jewish Medical Centre in New Hyde Park, NY. The participants were followed for a duration of one year, with 81% completing the 12-month study. The dropout rate was 16% (n=19) for the workbook intervention, 22% (n=53) for the intermediate intervention and 17% (n=42) for the most intensive intervention. The authors reported that those who completed the study were not significantly different from the dropouts in terms of their baseline characteristics. Blinding may have been possible for clinicians assessing the outcomes, but it was not reported to have been used.

It was not stated whether the analysis was conducted on an intention to treat basis or on treatment completers only. The primary health outcomes used were body weight, diabetes and cardiovascular risk, medication usage, dietary and exercise habits, and quality of life measures. The intervention groups were compared on a range of parameters and were reported to be "similar with respect to baseline demographic characteristics".

Extensive results were presented.

Changes in dietary intake and exercise were found not to be statistically significant between the groups.

The change in weight was -2.2 (+/- 1.26) lb for the workbook intervention, -4.7 (+/- 1.02) lb for the workbook and computer (intermediate) intervention, and -7.4 (+/- 1.15) lb for the workbook, computer and staff consultation (intensive) intervention, (p=0.003).

The BMI was -0.4 (+/- 0.21) for the workbook intervention, -0.8 (+/- 0.17) for the intermediate intervention and -1.2 (+/- 0.19) for the intensive intervention, (p=0.003).

Differences in waist measurements were not statistically different.

The proportion of body fat was -0.001% (+/- 0.441) for the workbook intervention, -0.161% (+/- 0.287) for the intermediate intervention and -1.233% (+/- 0.294) for the intensive intervention, (p=0.001).

The percentage weight loss was 0.9% (+/- 0.54) for the workbook intervention, 2.2% (+/- 0.48) for the intermediate intervention and 3.5% (+/- 0.49) for the intensive intervention, (p=0.002).

The proportion of people with any weight loss was 54% for the workbook intervention, 61% for the intermediate intervention and 70% for the intensive intervention, (p=0.037).

The percentage of people with greater than 5% weight loss was 15 for workbook, 23 for workbook and computer, and 31 for workbook, computer and staff consultation, p= 0.012.

Metabolic parameters and blood pressure changes were not found to be statistically significant.

The authors concluded "computer tailoring of weight-control programs may facilitate implementation of weight-loss programs in HMOs".

The summary measure of benefit was the number of pounds (lb) lost.

The costs were estimated from the perspective of an HMO. The analysis centred on differences in cost between the interventions. The authors considered the costs of the workbook, the computer system (including depreciation) and labour (clerical staff, dietician and social worker staff). Discounting was not carried out, which was appropriate due to the short time horizon of the study. Estimations were made using actual unit costs taken from the Bureau of Labour statistics and costs at the study setting. Resources appear to have been measured during the clinical study period, although a price year was not reported.

The authors did not report any statistical analysis of the costs despite extensive statistical analyses of the effectiveness estimates.

The authors did not estimate the indirect costs to society. These costs may have been relevant if productive work time was lost to attend consultation sessions, or if weight loss enabled some participants to become economically productive where they had not been previously.

US dollars ($).

The authors did not report any sensitivity analysis.

See the 'Effectiveness Results' section.

The mean cost per participant in the study was $12.33 for the workbook alone approach, $41.99 for the intermediate intervention and $133.74 for the intensive intervention.

In calculating the cost per lb lost, only participants who had complete data were included. The mean cost per lb lost was $6.23 for the workbook alone approach, $8.57 for the intermediate intervention and $18.78 for the intensive intervention.

The computerised tailoring of weight-control programmes may facilitate the implementation of weight-loss programmes in Health Maintenance Organisations (HMOs). The authors argued that the cost per participant was quite modest in each of the three interventions studied.

Three approaches were compared: workbook; expert computer system plus workbook; and staff consultation in addition to the computer system and workbook. The alternatives were justified with a thorough discussion of the available alternative weight-loss programmes. The authors highlighted existing uncertainty in the costs of providing programmes of varying intensity. This seems to have been the justification for the alternatives chosen.

The authors conducted a well-designed trial, taking many steps to improve the internal validity of the study. The basis of the analysis was a randomised controlled trial, which reduces the potential for systematic differences between participants in the groups. Power calculations were carried out to ensure the samples were of sufficient size to rule out the influence of chance. The participants were compared extensively at baseline in terms of both demographic and study-specific variables, and were shown to be comparable. Statistical analyses were carried out to assess whether the differences between the groups were statistically significant.

The summary measure of benefit used was the number of pounds lost in weight. This was an appropriate measure for comparing the three groups in this analysis and it will allow comparisons with similar weight-related studies. However, the specific measure will not allow a broader comparison with other health-related studies for the purposes of policy decision-making.

An HMO perspective was adopted in the analysis and, as such, the costs included in the study represented this perspective. The authors highlighted differences between the alternatives as a way of focusing the readers' attention on the cost of the workbook, computer system and staff consultation time. A statistical analysis of the costs was not carried out. Therefore, it is not possible to tell whether small omissions in the costs may alter the principle results and conclusions, especially given the relatively small absolute differences in cost per pound lost. The unit costs for the principle cost drivers were reported separately.

The authors made extensive comparisons of their findings with those from other studies, citing studies that found computerised systems to be valuable. The issue of generalisability to other studies was not addressed. However, the authors' very thorough reporting of the effectiveness study enables the reader to assess the likely generalisability of the study to their own setting. The results were not presented selectively. Indeed, the authors were again very thorough in their reporting. The conclusions drawn were limited and there was little discussion as to whether the same results might hold in an empirical setting. For instance, the dropout rates might be higher if real patients did not go through a tough screening process to meet the study protocols. No limitations to the study were highlighted.

The authors did not make any recommendations for policy or practice following their study, although they set out options facing HMOs. More specifically, "managed care organisations could potentially incorporate all three options as preventative health services", and "patients could opt for an approach based on availability and personal preference for learning method and time". The authors did not highlight any requirement for further research.

Supported by the National Heart, Lung and Blood Institute (grant number R01 HL50372) and the Diabetes Research and Training Center (grant number R01 DK20451).

Kristal AR, Curry AJ, Shattuck AL, Feng Z, Li S. A randomised trial of a tailored, self-help dietary intervention: the Puget Sound Eating Patterns Study. Preventive Medicine 2000;31:380-9.