Prescribing data were extracted from the Bennett Institute for Applied Data Science Open Prescribing Database and imported into Microsoft Excel in September 2026 by the lead researcher. The relative market share expressed as a percentage, for the Reference product and biosimilar were calculated for every month between September 2023 and March 2026 inclusive. A segmented regression analysis using a linear mixed-effects model was performed. Given that longitudinal prescribing data are likely to exhibit temporal autocorrelation, a first-order autoregressive [AR(1)] covariance structure was specified a priori to account for correlation between adjacent observations within each sublocation. To maintain a unified framework, this approach was applied across all device types. To accommodate for possible non-linear trends, a quadratic post-intervention term was added to the model. The time at which prescribing uptake approached saturation was estimated from the fitted quadratic post-intervention function by calculating the post-intervention time point at which the first derivative of the predicted uptake trajectory equalled zero. Cost savings were estimated by comparing observed prescribing with a counterfactual scenario based on continuation of the pre-intervention prescribing trend.