Survival analyses (Weibull regressions using STATA® software and formulas described by Briggs et al) 15 were used to transform the trial results into health state transition probabilities.
In the past, bioisosteric and scaffold hopping replacements were obtained experimentally using the trial-and-error approach, resulting in today's extensive literature available to the medicinal chemistry community [4].
For the BEV+IFN vs PAZ comparison the two extreme scenarios are based on the selected connector trials, whereby using the MRCRCC trial resulted in an ITC HR of 1.03 (95% CI: 0.61 1.74; P = 0.92) and using the 'proxy comparison' resulted in an ITC HR of 0.74 (95% CI: 0.40 1.37; P = 0.34).
Our model used the SUCCESS trial results as the basis for the key celecoxib and nsNSAID ulcer probabilities, whereas Spiegel used those for a coxib hybrid with predominant weightings from rofecoxib trials and CLASS, with its supratherapeutic doses of celecoxib.
Anonymised details labelled with the unique trial code will be transferred to a separate database which will be used for recording all of the trial results.
The effects of statin therapy have also been assumed to remain constant over the duration of the trials, but the trial results used to create the model had durations of less than the 5- to 20-year timescale of the simulated trials.
A probabilistic sensitivity analysis was conducted by combining 500 random samples from the trial results, obtained using bootstrapping, with 500 sample results from probabilistic sensitivity analysis each with a cohort size of 100 patients.
The trial results were pooled using meta-analysis, but a high level of heterogeneity was observed.
Preliminary estimates may, however, be of some use in planning till the time the trial results are available.
Thus, 3-year PFS, FFS and D-FFS could enable the early assessment of treatment effects on long-term survival using preliminary trial results, with 3-year PFS likely to provide the most accurate prediction of 5-year OS.
