The study protocol was evaluated and approved by an external ethical committee; patients gave informed consent to participate in the study according with the guidelines from the International Conference on Harmonisation and WHO good clinical practice standards. The study design has been previously described in detail.30 Briefly, the investigators implemented a randomized 2x2 design to address the effectiveness of different education strategies involving the health provider and people with diabetes).
Sample size: for its determination in each of the four groups, the investigators considered the change in haemoglobin A1c (from baseline to the end of the study) as the primary outcome variable. This was done using a two-sided test at the 5% level of significance and 80% power using a paired t-test. The sample size was increased by 25% to account for non-independence and dropout, resulting in the number of providers and patients listed in table 1.
Educational strategies The investigators used the Diabetes Training Course for Physicians and Diabetes Structured Education Courses for People with T2DM, whose characteristics and usefulness have been previously reported.
Data collection Patients' clinical, psychological, biochemical, therapeutic and economic data before and after PRODIACOR were recorded using the Physician Data Form, the Annual and Bi-annual Clinical Record Form and the Feedback Report Form. The investigators developed software that compares the first two mentioned forms and generates a Biannual Feedback Report for physicians and patients; it compares the values recorded for haemoglobin A1c, serum lipids and blood pressure, and the goal proposed for each parameter according to international standards. It also includes recommendations about appropriate treatment to achieve therapeutic goals.31 Additionally, the investigators used the Patient Questionnaire, based on the one previously used in the Diabetes Advantage Program (DAP) which includes questions about the disease, and the WHO-5 well-being questionnaire. Data collection was completed with the Personalized Checkbook, a tool initially developed for the PROPAT study, whose aim and content have been previously described. It serves to order procedures, consultations, laboratory tests, prescription of drugs and strips for glucose self-monitoring, to record and communicate results, and as a payment voucher for all those items.
Study and Data management The Local Coordinator (city of Corrientes) together with the Central Cordinating Center (CENEXA), oversaw the education courses and the overall trial and maintained periodic contact with the participating physicians. Performance of the physicians and their patients was monitored at regular intervals (6 months) during the whole study period. Every physician was visited twice a year by a physician monitor who assessed the quality of the data he has recorded.30 Participating physicians collect their own patients´ data, sent it to the Local Coordinator who controlled its completeness and forwarded to the Central Coordinating Centre.
Resource utilization and costs The data recorded include all direct medical items used by each of the patients ascribed to the programme, and were obtained from the checkbook and the utilization and cost records from each of the participating health coverage entities.
Utilization was classified into five groups: i) hospitalizations, ii) drugs and supplies, iii) diagnostic tests, iv) special studies, and v) physician office visits. Drugs were identified by their generic and commercial name and their corresponding presentation. Their costs were assessed from retail prices published by Alfabeta.net and adjusted to July 2004 using the health chapter of local Consumer Price Index.
Estimated drug costs and haemoglobin A1c reduction experienced by each treatment group were used to construct marginal figures of cost-consequence ratio. For this purpose, the investigators calculated such ratio figures for each experimental group, i.e., the investigators expressed the total drug cost associated to 1% haemoglobin A1c decrease to 10 mm Hg of systolic blood pressure or 10 mg/dL of triglyceride levels. The investigators did not currently estimate the costs of the educational interventions themselves, namely, all the physical and human resources involved in the management and administrative activities of the programme.
Statistical analysis To estimate the effects of the different educational interventions tested, intention to treat analysis was used. Initial univariate differences among groups for quantitative data were analyzed by one-way ANOVA (Bonferroni post hoc test). Factorial ANOVA and two-way ANOVA models (Bonferroni post hoc test) were used to assess the differences among groups along the study. ANOVA was also used to explore initial vs. end-of-study differences among quantitative data. Differences among qualitative measures were explored by Chi-squared (Yates corrected). P values of less than 0.05 were considered as significant (two tailed). Missing data strategies applied to the primary endpoint were last observation carried forward and complete case analysis. Both strategies were compared to evaluate the impact of missing data on the efficacy analysis of the primary endpoint.
