A decision tree model of the screening and treatment of gestational diabetes mellitus was used to evaluate the cost-effectiveness of the eight strategies. The model was an extension of one developed by the National Institute for Health and Clinical Excellence (NICE; National Collaborating Centre for Women’s and Children’s Health. 2008, see 'Other Publications of Related Interest' below for bibliographic details). The time horizon was the lifetime of the patient. The authors reported that the perspective was that of the NHS in England and Wales.
The clinical and effectiveness evidence was from a number of sources that included published studies, randomised controlled trials, and authors’ assumptions. The main effectiveness estimate was the effect of treatment on gestational diabetes. This was from a published meta-analysis of five studies (Horvath, et al. 2010, see 'Other Publications of Related Interest' below for bibliographic details). Only two studies were considered to have adequate randomisation, and the pooled results from these two studies were used.
Monetary benefit and utility valuations:
The utility estimates were from the two studies that supplied the clinical effectiveness data. Further studies were used for the utilities for other health states.
Measure of benefit:
The summary measure of benefit was quality-adjusted life-years (QALYs) gained. Future QALYs were discounted at an annual rate of 3.5%.
The authors reported that the direct costs were those of serious perinatal complications; pre-eclampsia, caesarean section, induction of labour; jaundice; and neonatal nursery care. More information on how these costs were estimated was available in electronic appendices. All costs were reported at 2009 prices in UK pounds sterling (£). They were all assumed to occur at, or close to, the time of screening or birth.
Analysis of uncertainty:
A probabilistic sensitivity analysis was undertaken to estimate the likelihood of each strategy being cost-effective at various willingness-to-pay thresholds. For each increment in individual risk, 10,000 simulations were produced, with the parameter values sampled from predefined probability distributions. A one-way sensitivity analysis was undertaken, using the effectiveness results of each of the two studies separately.