Cost effectiveness of an intensive blood glucose control policy in patients with type 2 diabetes: economic analysis alongside randomised controlled trial (UKPDS 41)

Radcliffe Infirmary, Oxford; Royal Infirmary, Aberdeen; Birmingham General Hospital; St George's Hospital, London; Hammersmith Hospital, London; Belfast City Hospital; North Staffordshire Royal Infirmary, Stoke on Trent; Royal Victoria Hospital, Belfast; St Helier Hospital, Carshalton; Whittington Hospital, London; Norfolk and Norwich Hospital, Norwich; Lister Hospital, Stevenage; Ipswich Hospital; Ninewells Hospital, Dundee; Northampton Hospital; Torbay Hospital; Peterborough General Hospital; Scarborough Hospital; Derbyshire Royal Infirmary, Derby; Manchester Royal Infirmary; Hope Hospital, Salford; Leicester General Hospital; Royal Devon and Exeter Hospital, Exeter.

Testing the model validity

The validity of the model was examined in two ways. Firstly, a separate model was built for each of the end points. Each of these models was validated individually and then combined into a single competing risks model. Estimates of time to first event from this model were found to agree to the second decimal place with those from the twofold model. Secondly, a form of temporal cross validation was used. The twofold model was refitted to the trial data using only the first 12 years of follow up. Figure w1 compares predictions made by this submodel for years 13-19 with survival probabilities observed within the trial. The success of this submodel, based on 12 years of data, at extrapolation over a further six years, suggests that the full model can accurately project beyond the trial follow up period.

Fig w1

  
 Comparison of observed (life table) survival function with estimates from a cross validation submodel

Secondary analysis

Secondary comparisons of the mean cost per patient in the conventional policy group compared with the insulin treated and sulphonylurea treated groups separately showed no significant differences. Patients in the intensive policy group who were randomised to insulin had mean trial costs of £7371 compared with £7170 in the conventional group, a non-significant difference of £202 (- £413 to £816) in favour of the conventional group. In a non-trial setting this difference becomes £603 (- £5 to £1210) in favour of the conventional group. Patients in the intensive policy arm who were randomised to sulphonylurea had mean trial costs of £6654 compared with £7170 in the conventional group, a difference of £515 (- £59 to 1089) in favour of the intensive group. In a non-trial setting this difference becomes £143 (- £426 to £712) in favour of the intensive group.

The difference in mean total trial costs per patient between the insulin treated and sulphonylurea treated groups was £717 (£273 to £1161) in favour of the sulphonylurea group (P=0.002). This difference was entirely attributable to the higher costs of insulin treatment, including drugs and blood glucose tests, with no significant differences in the costs of complications. In a non-trial setting, there was a difference of £459 (£22 to £896) in favour of sulphonylurea (P=0.039).

We also compared the costs for patients randomly allocated to chlorpropamide with those randomly allocated to glibenclamide. Treatment costs were significantly lower in the chlorpropamide group (mean difference £197, £7 to £387), but there were no significant differences in complication costs or in total costs.