Table 1 Key features and assumptions of the model

Treatment phases

In order to fit the BCVA evolution profile observed in clinical trials [35, 36] treatment response was modelled through three phases over time. The first (efficacy) phase, lasting 3 months, was where the active treatments had a positive impact on BCVA levels. This was followed by the maintenance phase lasting until the end of year 6, characterised by slower improvement in BCVA over time. The final end-of-life phase, starting six years from model entry, assumed a constant decline of vision over time. Of note, although the response to treatment varied with these phases, treatment doses did not.

Treatment in one or both eyes

At baseline, patients were treated for DMO in the study eye only, or in both eyes. The latter group could receive different treatments in both eyes, and if their fellow eye did not receive the study treatment, it was assumed to be treated with usual care (UC).

Treatments included in usual care

Based on the ICE-UK study, UC was defined as a mixture of laser photocoagulation, ranibizumab 0.5 mg/injection, bevacizumab 1.25 mg/injection and aflibercept 2 mg/injection. For treatment mix see Table 2.

Implant discontinuation

Patients who discontinued treatment with implants in either eye were switched to UC in that eye.

Usual care administration

We did not consider any fixed schedule of laser administration and anti-VEGF injection, but the model took into account the average number of administrations / injections per year in clinical practice (see Table 2)

Anti-VEGF efficacy

The efficacy of anti-VEGFs in the study eye was assumed to be insufficient based on the population definition. Therefore, the costs of UC included anti-VEGF costs, but the efficacy of UC was conservatively assumed to be identical to that of sham plus usual care in the FAME study. We also performed a scenario analysis where UC costs included only the costs of laser photocoagulation, and BCVA was assumed to be constant.

Fellow eyes treated with UC were potentially responsive to anti-VEGFs.

Other treatment combinations

In clinical practice, patients may receive anti-VEGF injections alongside steroid implants. The costs and adverse effects of those injections were not explicitly considered in FAc and dexamethasone arms, but the numbers of anti-VEGF injections in the UC arm was reduced to avoid any bias in favour of steroids.