The Appraisal Committee considered evidence submitted by the manufacturer of fluocinolone acetonide intravitreal implant and a review of this submission by the Evidence Review Group (ERG; section 4).
The manufacturer submitted evidence on the clinical and cost effectiveness of fluocinolone acetonide intravitreal implant compared with optimised standard of care and laser photocoagulation monotherapy. The manufacturer did not provide any specific analyses comparing fluocinolone acetonide intravitreal implant with triamcinolone alone or the anti‑vascular endothelial growth factor (anti‑VEGF) treatments bevacizumab and ranibizumab alone.
The main source of evidence in the manufacturer's submission was a preplanned analysis of data from the FAME A and B randomised controlled trials that evaluated the safety and efficacy of fluocinolone acetonide intravitreal implant for treating diabetic macular oedema. The preplanned analysis focused on duration of diabetic macular oedema, analysing patients who had had the condition for durations above and below the median separately. When the trial was unblinded the median duration was determined to be 3 years. The subgroup in the submission was patients with duration of diabetic macular oedema over 3 years (the manufacturer calculated the duration of diabetic macular oedema as the year of randomisation to treatment minus the year of diagnosis of the disease plus 1).
FAME A and B were 2 identical, randomised, double‑blinded, sham injection‑controlled multicentre trials conducted over 36 months. The results of the trials were combined and presented in the submission as a single analysis. Patients were randomised 1:2:2 to sham injection, 0.2 micrograms/day (low‑dose) or 0.5 micrograms/day (high‑dose) fluocinolone acetonide intravitreal implant. Participants in the trials were adults with diabetic macular oedema who were aged between 18 and 85, who had received at least 1 previous laser treatment, whose best corrected visual acuity (BCVA) was 19 to 68 letters on the Early Treatment of Diabetic Retinopathy Study (ETDRS) eye chart, and whose central retinal thickness was 250 microns or more at baseline. Exclusion criteria were intraocular pressure over 21 mmHg, and systolic blood pressure over 180 mmHg or diastolic blood pressure over 105 mmHg.
In both groups (sham injection and low‑ or high‑dose implant) additional treatment with laser photocoagulation was given as needed after week 6. Approximately 61% of the sham injection group and 41% of the fluocinolone acetonide intravitreal implant group received treatment with laser photocoagulation during the study. The mean number of laser treatments was 1.4 and 0.8 in the sham injection and fluocinolone acetonide intravitreal implant groups respectively. Re‑treatment with fluocinolone acetonide intravitreal implant was offered at any time after the month 12 assessments if patients experienced vision loss (5 letters or more) or retinal thickening of 50 microns or more compared with their best status in the previous 12 months.
Patients in both groups also received a range of therapies not allowed in the study protocol. These included intravitreal steroids (triamcinolone and dexamethasone), anti‑VEGF therapy, vitrectomies and posterior sub‑Tenon steroids. The number of off‑protocol treatments was higher in the sham injection group than in the fluocinolone acetonide intravitreal implant group (117 compared with 48); approximately 35% of patients in the sham injection group and 13% of patients in the fluocinolone acetonide intravitreal implant group received at least 1 off‑protocol treatment. Data from these patients were included in the analysis population.
The primary outcome reported in the FAME trials was the proportion of people with an improvement of 15 or more letters from baseline BCVA at month 24. Secondary outcomes included:
mean change in BCVA
mean change in excess retinal thickness
percentage with 3‑step (15 letters or more) worsening of ETDRS
percentage needing laser photocoagulation.
There were 956 patients enrolled in the FAME trials. Of these, 536 formed the subgroup of patients with chronic diabetic macular oedema for longer than 3 years. Of this subgroup of patients comprehensive data were presented by the manufacturer for the 0.2 micrograms/day implant group only, because only the low dose has been licensed. The resulting number of patients with chronic diabetic macular oedema in the subgroup that formed the basis of the manufacturer's submission was 321 (209 in the 0.2 micrograms/day implant group and 112 in the sham injection group).
The mean age of the patients was 62.9 years in the sham injection group (n=112) and 63.7 years in the 0.2 micrograms/day implant group (n=209). In the trials, the majority of patients had chronic diabetic macular oedema in both eyes at baseline, but only 1 eye was treated. In most cases, the worse‑seeing eye was treated.
At month 24, the proportions of patients with chronic diabetic macular oedema who had a 15.0 letters or more increase from baseline BCVA were 13.4% and 34.4% in the sham injection and 0.2 micrograms/day implant groups respectively (p<0.001). The proportions of patients who had a 15.0 letters or more improvement in BCVA at month 36 (13.4% and 34.0% respectively [p<0.001]) were comparable to those observed in month 24. There were numerical increases in mean change in BCVA from baseline in the 0.2 micrograms/day implant group compared with the sham injection group at all evaluations (12 through to 36 months); this was statistically significant at months 30 and 36. At month 36, there was a mean improvement of 7.6 letters in the 0.2 micrograms/day implant treatment group compared with 1.8 letters in the sham injection group (p<0.004).
The FAME trials included an assessment of health‑related quality of life using the Visual Function Questionnaire‑25 (VFQ‑25) at baseline and months 24 and 36. These data were provided in the clinical study reports of the FAME trials. The manufacturer stated that the VFQ‑25 was not used in the economic model because it measures overall visual function which is driven by vision in the better‑seeing eye, whereas in the FAME trials, the majority of patients had their worse‑seeing eye treated. The VFQ‑25 values are marked by the manufacturer as academic in confidence and therefore not presented here.
The manufacturer also included laser photocoagulation monotherapy as a relevant comparator for fluocinolone acetonide intravitreal implant. The manufacturer conducted a literature search and identified 1 relevant study: DRCR Protocol B (2008). This study was a phase 3, multicentre, randomised clinical trial conducted in the USA to compare intravitreal triamcinolone with focal/grid laser photocoagulation in patients with diabetic macular oedema. The manufacturer noted that the severity of diabetic macular oedema in the DRCR study was not as great as in the FAME trials. Of the DRCR population, approximately 40% of people had not had their disease treated with laser and there was no stipulation on duration of diabetic macular oedema at randomisation. A comparison of outcomes data as reported in the DRCR and FAME clinical trials was presented. The manufacturer did not use any statistical methods to compare the data indirectly. The manufacturer noted that the proportion of people with a 15 letters or more improvement in BCVA in the FAME trials at month 36 demonstrated a numerical difference in favour of fluocinolone acetonide intravitreal implant (34.0% compared with 18% for laser photocoagulation at 24 months in the DRCR Protocol B study).
The manufacturer's submission included data from the FAME trials for ocular adverse events in people with chronic diabetic macular oedema (duration of 3 years or longer). The data suggested that fluocinolone acetonide intravitreal implant is associated with the formation or progression of cataract and increased intraocular pressure. In the FAME trials, 34.4% (72/209) of patients in the fluocinolone acetonide intravitreal implant group experienced increased intraocular pressure or ocular hypertension compared with 14.3% (16/112) in the sham injection group. At baseline, 58.9% (66/112) of people in the sham injection group and 54.5% (114/209) in the fluocinolone intravitreal implant group were phakic (still had their natural lens). Of those who were phakic at baseline, 77.9% of the fluocinolone acetonide intravitreal implant group and 77.0% of the sham injection group had a pre‑existing cataract. Cataract surgery was needed by 85.1% (97/114) of the fluocinolone acetonide intravitreal implant group and 36.4% (24/66) of the sham injection group who were phakic at baseline.
The manufacturer provided a subgroup analysis of people with chronic diabetic macular oedema who had treatment in an eye with a pseudophakic lens (that is, they had already had an operation for cataract removal and had been fitted with an intraocular lens to replace the natural crystalline lens) at entry into the FAME trials. The manufacturer considered the overall treatment effect in this subgroup to be similar to that for the phakic population; however, increased benefits arose through the removal of a known adverse event (advancement of cataract development) and removal of the costs associated with cataract surgery. Of the people with chronic diabetic macular oedema, there were 46 (41.1%) who had a pseudophakic lens in the sham injection group and 95 (45.5%) in the 0.2 micrograms/day fluocinolone‑treated group. For this subgroup, at 36 months, 31.6% in the fluocinolone‑treated group and 17.4% in the sham injection group had a greater than 15‑letter increase from baseline BCVA, giving a difference of 14.2% (95% CI 28.6% to -0.2%).
The economic evidence provided by the manufacturer in its submission comprised a literature review (which identified no relevant published cost‑effectiveness studies) and a de‑novo cost–utility analysis. The manufacturer's economic evaluation compared fluocinolone acetonide intravitreal implant with the comparator ('optimised standard of care') in the FAME trials for a cohort of patients with chronic diabetic macular oedema. The model also included a comparison of fluocinolone acetonide intravitreal implant with laser photocoagulation using data for laser photocoagulation from the DRCR Protocol B study.
The manufacturer's model included 14 health states (13 BCVA health states and death) which were defined by bands of 5 ETDRS letters in the treated eye. Utility values associated with the 13 BCVA‑related health states captured the effect of varying degrees of visual gain or loss on patients' quality of life. The model structure made no distinction between treatment of the better‑seeing eye and the worse‑seeing eye. The model had a 15‑year time horizon and a quarterly cycle length, with costs and benefits both being discounted at 3.5%. For the first 3 years, the distribution of patients across health states was drawn directly from the FAME trials data. Beyond 3 years, a Markov model structure was adopted with transition probabilities being applied.
The base‑case analysis assumed that 1 fluocinolone acetonide intravitreal implant is needed every 3 years. Patients needed to have gained 5 or more ETDRS letters of visual acuity between baseline and month 36 to receive a further implant at month 36. In addition, using data from the FAME trials the manufacturer applied a drop‑out rate (for those who withdrew consent, were lost to follow‑up or died) to the patients receiving fluocinolone acetonide intravitreal implant who were re‑treated at the end of the first 36 months (these details are marked by the manufacturer as commercial in confidence and therefore not presented here). This adjusted re‑treatment rate was applied equally to each health state. Patients in both the fluocinolone group and the optimised standard of care group also received laser treatments based on rates from the FAME trials, and in the first 3 years could receive other therapies including triamcinolone, ranibizumab, bevacizumab and dexamethasone, again based on rates in the FAME trials.
In the base‑case analysis, it was assumed that 35% of patients would receive bilateral treatment in the optimised standard of care group. The manufacturer assumed that bilateral treatment would be contraindicated for patients treated in the first eye with fluocinolone acetonide intravitreal implant who had a subsequent rise in intraocular pressure greater than 30 mmHg. Therefore the bilateral treatment rate in the fluocinolone group was reduced based on the proportion of patients with raised intraocular pressure observed in the FAME trials.
The clinical efficacy data from the FAME trials were used directly to calculate the number of patients in each of the model health states in each quarter for the first 3 years. Changes after 3 years were extrapolated from the FAME trials data. The FAME data were divided into patients whose disease had responded to and not responded to treatment based on the ETDRS 5‑letter criteria. Data were then analysed to determine the numbers of patients whose vision improved or worsened by at least 5 letters each quarter. The average net changes in the last 4 quarters of the FAME trials were used in the model to extrapolate improvements in vision beyond 3 years. For patients receiving another implant at 36 months, the model assumed that 5% of patients in each health state would improve by 5 letters every quarter. For patients not receiving a further implant at 36 months, the model assumed that 3% of patients in each health state would experience a transition to a lower health state every quarter. In the optimised standard of care group and laser group, the model assumed that 3% of patients in each health state would have a worsening in vision of 5 letters and therefore move to a worse health state every quarter.
The manufacturer did not consider it appropriate to include the VFQ‑25 values in the economic model because VFQ‑25 is driven by vision in the better‑seeing eye, whereas in the FAME trials, the majority of patients had their worse‑seeing eye treated (these values are marked by the manufacturer as academic in confidence and are therefore not presented here). Furthermore, the manufacturer stated that a mapping exercise was not considered because there was not a universally accepted mapping process to convert VFQ‑25 data to utility scores.
The manufacturer conducted a systematic review to identify utility values reported in the literature for populations with visual impairment. The review of the articles included diabetic macular oedema and other disorders affecting visual acuity (such as age‑related macular degeneration). Based on the data available, the manufacturer chose to use time trade‑off data from Brown et al. (2000) as the source of utility values for its submission. Brown et al. (2000) was a US study that measured utility values in 5 groups according to visual acuity in the better‑seeing eye in a population of patients with age‑related macular degeneration. The values estimated by Brown et al. (2000) and the values used within the model ranged from 0.40 in the lowest health state (less than 20 ETDRS letters) to 0.89 in the highest health state (75 ETDRS letters or more). The Brown et al. (2000) study did not report utility weights in patients with BCVAs between 35 and 50, and therefore the unweighted averages of the utility weights above and below this range were assumed.
For patients who received treatment in both eyes, a 25% bilateral treatment quality‑adjusted life year (QALY) uplift was also applied to the aggregate QALYs.
The model did not consider utility decrements due to adverse events, or procedures and interventions for the adverse events. The manufacturer stated that because the utility was calculated for BCVA values, and the BCVA values were based on the trial data, the impact on patient vision of adverse events such as cataract formation was reflected in the BCVA of the treated eye.
The model included the costs of fluocinolone acetonide intravitreal implant as well as laser and other therapies at the rates observed in the FAME trials. Adverse event costs were also included. The manufacturer applied an annual cost of blindness of £6298 to the proportion of patients whose treated‑eye BCVA fell below 35 letters.
In its deterministic base case, based on an incremental cost of £11,330 and an incremental QALY value of 0.500, the manufacturer estimated an incremental cost‑effectiveness ratio (ICER) without the patient access scheme of £22,655 per QALY gained for fluocinolone acetonide intravitreal implant compared with optimised standard of care.
Following a request for clarification from the ERG the manufacturer provided a revised analysis. The manufacturer acknowledged that the health‑related quality of life values from Brown et al. (2000) may not apply to patients having their worse‑seeing eye treated. The manufacturer therefore used revised health‑related quality of life values to reflect a weighted average of values for people having their worse‑ and better‑seeing eyes treated, taken from a study by Heintz et al. (2012).
As well as changing health‑related quality of life values, the manufacturer's revised analysis also amended: male and female mortality rates to revise the pooled annual all‑cause mortality risk, the proportion of patients needing bilateral treatment, the percentage of patients needing bilateral treatment and for whom a second fluocinolone acetonide intravitreal implant treatment was not contraindicated, the quality of life uplift from bilateral treatment to 10%, and the unadjusted response rate in the fluocinolone group to a rate based upon a 10‑letter re‑treatment criterion.
The manufacturer's amendments reduced the estimate of cost effectiveness for fluocinolone acetonide intravitreal implant compared with optimised standard of care from £22,655 to £19,268 per QALY gained without the patient access scheme, based on an incremental cost of £11,927 and an incremental QALY value of 0.619.
The ERG commented that the 3‑year data used to inform the first 3 years of the economic model were robust, although the more usual modelling approach would have been to use transition probability matrices. The structure of the model means it cannot be manipulated during this 3‑year period to explore different scenarios.
The ERG commented that it would have been more appropriate for the manufacturer to use a model structure that modelled patients as having 2 eyes, rather than undertaking an ad hoc adjustment to the output of a model in which patients only had 1 eye. The ERG noted that the FAME trials had a reasonable proportion of patients who had their better‑seeing eye treated, and that the rate of chronic diabetic macular oedema in the other eye was high.
The ERG also noted that the distribution between health states for patients whose disease had responded to and not responded to fluocinolone acetonide intravitreal implant at 36 months was modelled as being a constant percentage of the overall patient distribution at 36 months. The ERG commented that this approach was not justified, and could lead to bias in the estimates of cost effectiveness for fluocinolone acetonide intravitreal implant.
The ERG noted that the manufacturer's base‑case model applied a re‑treatment criterion of a minimum 5‑letter improvement between baseline and 36 months. The ERG commented that a more realistic criterion might be a minimum 10‑letter improvement between baseline and 36 months, which the manufacturer applied in response to the clarification request from the ERG. However, the ERG commented that this only changed the number of patients not receiving another fluocinolone acetonide intravitreal implant at 36 months and did not affect their distribution across health states.
In the manufacturer's submission clinical effectiveness beyond 3 years was extrapolated from the FAME trials data. The ERG commented that rather than including the proportions of patients whose disease improved or worsened each quarter of a year, the proportions were netted. The ERG considered that the reasons for analysing the data in this way were unclear.
The ERG commented that because in the FAME trials patients had only 1 eye treated, assumptions were needed about rates of bilateral treatment. In the base‑case analysis, it was assumed that 35% of patients would need bilateral treatment; this percentage was increased in the revised analysis (these details are marked by the manufacturer as academic in confidence and therefore not presented here). The ERG considered that the manufacturer's revision was too high because a proportion of patients would not have visual impairment in both eyes because of diabetic macular oedema, and a proportion would not be able to have both eyes treated because of raised intraocular pressure or other reasons.
The ERG commented that there was considerable uncertainty about the appropriateness of the utility values used in the model. The original submission used utility values that related to sight in the better‑seeing eye. The ERG considered there were limitations in the data (based on Heintz et al. 2012) used in the revised analysis because these provided only 3 quality of life values over 13 health states, were based on small patient numbers, and were non‑monotonic.
Initially, the ERG made a series of revisions to address what it considered to be possible errors in the model. These were:
a change to the formulae for averaging mortality between male and female rates
a change to the formulae for applying the yearly natural discontinuation rate (the detail of which is commercial in confidence) in the cohort flow cells
a change to the formula for the percentage of patients remaining on fluocinolone acetonide intravitreal implant after year 9.
The cumulative impact of correcting the 3 errors was to increase the ICER without the patient access scheme from £22,655 to £26,526 per QALY gained for fluocinolone acetonide intravitreal implant compared with optimised standard of care.
The ERG also conducted additional exploratory analyses changing:
how the cost of fluorescein angiography needed before each laser administration was applied in the model
the number of laser administrations per patient
the unit costs for adverse event procedures
how the proportions of people whose disease improved and worsened each quarter year after 36 months was calculated
the rate of bilateral treatment
the cost of blindness
the cost and quality of life uplift applied for bilateral treatment for patients in the fluocinolone group whose condition needed bilateral treatment but for whom a second implant was contraindicated.
The cumulative impact of these changes resulted in ICERs of £37,740 for the 5‑letter response criterion (with an incremental cost of £14,569 and an incremental QALY of 0.386) and £35,940 for the 10‑letter response criterion (with an incremental cost of £12,736 and an incremental QALY of 0.354) without the patient access scheme.
The ERG also conducted further exploratory sensitivity analyses with a particular focus on the source of utility values. These additional analyses modelled the impact of changes in the BCVA of the better‑seeing eye using Brown et al. (1999). This US study of 325 patients measured utility values in 12 groups according to visual acuity in the better‑seeing eye in a population of patients with impaired vision in at least 1 eye. The values estimated by Brown et al. (1999) used within the model vary from 0.54 in the lowest health state (less than 20 ETDRS letters) to 0.89 in the highest health state (75 ETDRS letters or more), giving a range of 0.350 across the health states of the model. The ERG commented that if its cost‑effectiveness estimates without the patient access scheme for fluocinolone acetonide intravitreal implant compared with optimised standard of care were used as a starting point (£37,740 per QALY gained for the 5‑letter response criterion and £35,940 per QALY gained for the 10‑letter response criterion), applying the health‑related quality of life values taken from Brown et al. (1999) would suggest cost‑effectiveness estimates of £66,744 per QALY gained (5‑letter response criterion) and £64,249 per QALY gained (10‑letter response criterion).
The ERG explored the effect of using health‑related quality of life values from a regression analysis in the manufacturer's submission for NICE's technology appraisal guidance 237 on ranibizumab for the treatment of diabetic macular oedema (replaced by NICE's technology appraisal guidance 274 on ranibizumab for the treatment of diabetic macular oedema), and from Brown et al. (1999) and Brown et al. (2000). The ERG noted that there was uncertainty around the health‑related quality of life impact resulting from changes in the BCVA of the worse‑seeing eye, and presented 6 scenario analyses to take this into account in conjunction with the 3 different sources of utility values:
Scenario analysis 1: A flat health‑related quality of life function where changes in the BCVA of the worse‑seeing eye have no impact.
Scenario analysis 2: A health‑related quality of life function where changes in the BCVA of the worse‑seeing eye have 15% of the range of changes in the BCVA of the better‑seeing eye: that is, a range of 15% of 0.350 which equals 0.053.
Scenario analysis 3: A health‑related quality of life function where changes in the BCVA of the worse‑seeing eye have 30% of the range of changes in the BCVA of the better‑seeing eye: that is, a range of 30% of 0.350 which equals 0.105.
Scenario analysis 4: A health‑related quality of life function where changes in the BCVA of the worse‑seeing eye have 50% of the range of changes in the BCVA of the better‑seeing eye: that is, a range of 50% of 0.350 which equals 0.175.
Scenario analysis 5: A health‑related quality of life function where changes in the BCVA of the worse‑seeing eye have 70% of the range of changes in the BCVA of the better‑seeing eye: that is, a range of 70% of 0.350 which equals 0.245.
Scenario analysis 6: A health‑related quality of life function where changes in the BCVA of the worse‑seeing eye have 100% of the range of changes in the BCVA of the better‑seeing eye: that is, a range of 0.350.
The ERG also assumed that 20% and 40% of people received unilateral treatment in their better‑seeing eye and worse‑seeing eye respectively. For the fluocinolone group it was further assumed that 34% of people received bilateral treatment with fluocinolone acetonide intravitreal implant (that is, in both eyes), whereas 6% received treatment with fluocinolone acetonide intravitreal implant in the first eye but not in the second eye. In the optimised standard of care group, the ERG similarly assumed that 20% and 40% of people received unilateral treatment in their better‑seeing and worse‑seeing eye respectively; whereas 40% of people were assumed to receive treatment in both eyes.
The ERG completed sensitivity analyses without the patient access scheme. Using Brown et al. (1999), the ICERs (based on a 10‑letter response criterion and including a bilateral benefit) ranged from £48,533 per QALY gained (scenario analysis 6, where changes in the BCVA of the worse‑seeing eye have 100% of the range of changes in the BCVA of the better‑seeing eye, that is, a range of 0.350) to £110,730 per QALY gained (scenario analysis 1, where changes in the BCVA of the worse‑seeing eye are assumed to have no impact). Using Brown et al. (2000), the equivalent ICERs ranged between £30,910 and £61,942 per QALY gained for fluocinolone acetonide intravitreal implant compared with optimised standard of care. Using the utility values derived from the manufacturer's submission for NICE's technology appraisal guidance 237 on ranibizumab for the treatment of diabetic macular oedema, based upon the 10‑letter response criterion and including a bilateral benefit, the ICERs ranged from £69,802 (scenario analysis 6) to £251,686 per QALY gained (scenario analysis 1). Using a 10‑letter response criterion, including a bilateral benefit and assuming scenario analysis 3 the ICER values were £47,604 per QALY gained using Brown et al. (2000) utilities and £80,037 per QALY gained using Brown et al. (1999) utilities.
Full details of all the evidence are in the manufacturer's submission and the ERG report for NICE technology appraisal guidance 271 on fluocinolone acetonide intravitreal implant for the treatment of chronic diabetic macular oedema after an inadequate response to prior therapy.
In NICE's technology appraisal guidance 271, fluocinolone acetonide intravitreal implant was not recommended for treating chronic diabetic macular oedema. After publication, the manufacturer agreed a patient access scheme with the Department of Health and submitted revised analyses to be considered in a rapid review of the original guidance.
The manufacturer presented analyses for the full population of people with chronic diabetic macular oedema and for the subgroup of people with chronic diabetic macular oedema who had treatment in an eye with a pseudophakic lens. In the revised economic model the manufacturer included:
BCVA patient distributions in the extrapolation from month 36 onwards that were specific to a patient's response status
BCVA patient distributions in the extrapolation for the pseudophakic subgroup that were based on the final 30 months of observations from the FAME trials, and BCVA patient distributions for the full chronic diabetic macular oedema population that were based on the final 12 months of observations
analyses using utility values from Brown et al. (1999) and Brown et al. (2000), as well as an analysis using utilities from Czoski‑Murray et al. (2009) (used in NICE's technology appraisal guidance 237 on ranibizumab for the treatment of diabetic macular oedema, replaced by technology appraisal guidance 274 on ranibizumab for treating diabetic macular oedema)
an assumption that changes in vision for people treated in their worse‑seeing eye had 30% of the health‑related quality of life impact of the same change in vision from treating their best‑seeing eye
an assumption that 20% of patients are unilaterally treated in the best‑seeing eye, 40% of patients are unilaterally treated in the worst‑seeing eye, and the remaining 40% of patients receive bilateral treatment.
The manufacturer also clarified some characteristics of the people enrolled in the FAME trials. It stated that people in the trials were as severely affected as patients who would receive fluocinolone acetonide intravitreal implant in routine clinical practice. It stated that in clinical practice, a BCVA of 20/80 is generally considered poor and insufficiently responsive to treatment. It noted that in the FAME trials 59% of the patients with chronic diabetic macular oedema had vision of 20/80 or worse at baseline. Of these, 42.7% and 12.7% of patients had a 15‑letter gain in the fluocinolone acetonide intravitreal implant and optimised standard of care groups respectively (p<0.001). The manufacturer also stated that in the FAME trials all patients had at least 1 prior macular laser treatment before randomisation. It noted that in the optimised standard of care group, people with diabetic macular oedema for more than 3 years showed lower levels of response (13.4% of patients) than people with diabetic macular oedema for less than 3 years (27.8% of patients); whereas in the fluocinolone acetonide intravitreal implant group, people with diabetic macular oedema for more than 3 years showed higher levels of response (34% of patients) than people with diabetic macular oedema for less than 3 years (22.3% of patients). The manufacturer considered that fluocinolone acetonide intravitreal implant provided significant additive benefit for people with chronic diabetic macular oedema whose disease was responding insufficiently to other therapies.
The manufacturer presented the results for the comparison between fluocinolone acetonide intravitreal implant and optimised standard of care with the patient access scheme for the whole population with chronic diabetic macular oedema and for the subgroup of people with a pseudophakic lens. The ICERs for the whole population with the patient access scheme were £37,630 using the utilities from Brown et al. (2000) and £63,472 using the utilities from Brown et al. (1999) per QALY gained. The ICER using the utilities from Czoski‑Murray et al. (2009) was £42,663 per QALY gained. In the pseudophakic subgroup the ICERs were £17,639 using the utilities from Brown et al. (2000) and £30,296 using the utilities from Brown et al. (1999) per QALY gained. Using the utilities form Czosky‑Murray et al. (2009) the ICER was £19,884 per QALY gained.
The manufacturer carried out a sensitivity analysis assuming that patients would not be re‑treated with fluocinolone acetonide intravitreal implant automatically at month 36. Re‑treatment would only take place if there was a response to treatment and a BCVA less than 20/32 at month 36. These analyses resulted in small reductions in the ICERs. With the patient access scheme, the ICERs for the chronic diabetic macular oedema population were £34,668 using utilities from Brown et al. (2000) and £57,476 using utilities from Brown et al. (1999) per QALY gained. In the pseudophakic subgroup, the ICERs were £16,642 using the utilities from Brown et al. (2000) and £28,584 using the utilities from Brown et al. (1999) per QALY gained.
The ERG stated that there continued to be uncertainty about the best source of utilities to be used in the model. It noted that utilities from Brown et al. (1999) represented a more diverse group of patients with different eye conditions, all of whom were visually impaired with vision of at best 20/40 in at least 1 eye. The ERG considered that utilities from Brown et al. (1999) may be preferable to those from Brown et al. (2000) because Brown et al. (1999) included more patients, which allows a finer gradation of utility estimates for a given BCVA. The number of patients for a given BCVA band was also higher in Brown et al. (1999). The ERG also noted that one‑third of patients in Brown et al. (1999) had diabetes compared with none in Brown et al. (2000).
The ERG commented that it was unable to source the utility values from Czoski‑Murray et al. (2009) used by the manufacturer. The ERG applied the Czoski‑Murray utility function adjusted for an average age of 63 and found similar ranges to those used by the manufacturer. The ERG noted that patients included in Czoski‑Murray et al. (2009) used contact lenses to simulate different degrees of visual loss. It also noted that the duration of the simulated visual impairment was short, and so the utility values may not apply to patients with longer duration of visual loss.
The ERG commented that the differences between the ICERs for the whole chronic diabetic macular oedema population and the pseudophakic subgroup were in part driven by differences in the distribution of patients across health states. The ERG noted that there were uncertainties around these distributions because for the pseudophakic subgroup, the difference in the baseline distributions between the fluocinolone group and the optimised standard of care group may indicate a breakdown in randomisation. It also noted that there were differences in the 36‑month patient distributions within the optimised standard of care group, between the full chronic diabetic macular oedema population and the pseudophakic subgroup. The ERG commented on the importance of these patient distributions because they are the basis for the extrapolations in the model.
The ERG checked the manufacturer's ICERs and found minor differences in the values. The ERG commented that there were some small changes made to response rates and drop‑out rates that had not been previously applied in the model. The ERG also commented that there were some errors in the adverse effect cost calculations, but correcting for these had little impact on the ICERs. The deterministic ICERs from the ERG check with the patient access scheme for the chronic diabetic macular oedema population were £64,549 and £37,996 per QALY gained using Brown et al. (1999) and Brown et al. (2000) utilities respectively. For the pseudophakic subgroup, the ICERs with the patient access scheme were £30,025 (using Brown et al. [1999] utilities), £21,027 (using Czoski‑Murray utilities) and £17,487 (using Brown et al. [2000] utilities) per QALY gained.
The ERG noted that the manufacturer presented a sensitivity analysis that assumed that patients would be re‑treated at month 36 only if their disease responded to fluocinolone and they had a BCVA of less than 20/32. The ERG commented that this change only affected costs in the fluocinolone acetonide intravitreal implant group without any clinical impact of these patients stopping treatment. It also stated that the impact on costs may be underestimated as the re‑treatment adjustment was only applied to the second re‑treatment with fluocinolone acetonide intravitreal implant (at month 36) but not to subsequent re‑treatments.