3 The company's submission
The Appraisal Committee considered evidence submitted by Santen Pharmaceutical and a review of this submission by the Evidence Review Group (ERG).
Clinical effectiveness
3.1
The company identified 2 multicentre (including the UK) double-masked, randomised controlled clinical trials relevant to the decision problem, SANSIKA and SICCANOVE. These trials compared ciclosporin with a vehicle in people with dry eye disease that had not improved despite treatment with artificial tears. The company presented results from both SANSIKA and SICCANOVE but considered SANSIKA to be most relevant to the decision problem, because it included only people with severe dry eye disease (whereas SICCANOVE included people with moderate to severe dry eye disease). Only details and results of SANSIKA are presented here.
3.2
SANSIKA (n=246) included patients with severe keratitis and severe dry eye disease defined as having a Corneal Fluorescein Score (CFS) of 4 on the modified Oxford scale, a Schirmer score (without anaesthesia) of 2 mm to 10 mm and an Ocular Surface Disease Index (OSDI) score of 23 or more. The trial compared ciclosporin in combination with artificial tears with the vehicle plus artificial tears. The vehicle contained the excipient cetalkonium chloride and patients were allowed to use preservative-free artificial tears as needed. SANSIKA was divided into 2 parts: part 1 studied the efficacy of ciclosporin over 6 months (n=245) and part 2, a 24‑week open-label extension, assessed the long-term safety of ciclosporin up to 12 months (n=207). Randomisation was stratified by centre. Treatment compliance was measured by the number of used and unused containers of ciclosporin in relation to the duration of the follow-up interval.
3.3
The primary end point was change from baseline in CFS‑OSDI, a composite variable combining the CFS and OSDI scores, at month 6. The definition of response using CFS‑OSDI was:
-
improvement of 2 points or more from baseline in CFS
-
improvement of 30% or more from baseline in OSDI.
Secondary end points were: change from baseline in CFS, ocular discomfort and CFS‑OSDI analysed at other time points, use of concomitant artificial tears, investigator global evaluation of efficacy, Schirmer test (without anaesthesia) in both eyes, human leukocyte antigen‑DR (HLA‑DR) expression on the conjunctival cell surface by impression cytology, tear break‑up time in both eyes, corneal and conjunctival staining assessed using the Van Bijsterveld grading system (Lissamine Green Staining), tear film osmolarity in both eyes, and quality of life measured with the EuroQoL‑5D Questionnaire (EQ‑5D) and the National Eye Institute Visual Function Questionnaire (NEI‑VFQ‑25). Adverse events were separated into ocular and systemic adverse events.
3.4
The efficacy end points were analysed based on the full analysis set (n=245 in SANSIKA part 1 and n=207 in SANSIKA part 2), which included all patients who had any amount of study drug and for whom post-baseline data were available. Statistical significance was set at a significance level of 5% (p≤0.05). The analyses for the safety end points were based on the safety analysis set (n=244 in SANSIKA part 1 and n=207 in SANSIKA part 2), which included all patients for whom there was evidence that they used the study medication. The company carried out several post hoc subgroup analyses including of the primary efficacy end point CFS‑ODSI response rate (setting CFS improvement at 3 grades instead of 2).
3.5
The company presented the results from SANSIKA for the primary end point and noted that none of the results presented were statistically significant. The company stated that there are many possible explanations for this, including the lack of correlation between signs and symptoms of dry eye disease and the possible beneficial effects of the vehicle itself.
3.6
The company presented an analysis of CFS score change from baseline over time in SANSIKA, which showed a statistically significant decrease in both treatment groups (p<0.001). It noted that there was a statistically significant benefit with ciclosporin compared with the vehicle over the 6‑month treatment period (p=0.017). At 6 months, the decrease in CFS score from baseline was statistically significantly greater with ciclosporin than with the vehicle (p=0.037).
3.7
From its post hoc analysis of CFS‑OSDI in SANSIKA (using an improvement of 3 grades or more in CFS as criteria for improvement), the company noted that there was a statistically significantly higher response with ciclosporin (imputed data: 18.8%; observed data: 21.4%) compared with the vehicle (imputed data: 7.7%; observed data: 8.5%; p=0.016 and p=0.012 based on imputed and observed data respectively).
3.8
Results of HLA‑DR in SANSIKA showed that at 6 months, ciclosporin was associated with a statistically significant decrease in HLA‑DR from baseline compared with the vehicle (p=0.021). This demonstrated that ciclosporin had an anti-inflammatory effect. The company noted that this is important because dry eye disease is characterised by inflammatory changes on the ocular surface.
3.9
The company presented the median use of artificial tears instead of the mean because the data distribution was skewed. It stated that there were no differences in the use of artificial tears between treatment groups during all visits in part 1 in SANSIKA but noted that the number of missing data was high. The company stated that considering all available data, there was a progressive decrease in the use of artificial tears over time in both treatment groups. The results in part 2 showed a steady decrease in the use of artificial tears during the first 6 months in both treatment groups (-3.8 drops per day per eye in people who had ciclosporin in both parts of SANSIKA, and -2.6 drops per day per eye in people who had the vehicle alone in part 1 and ciclosporin in part 2).
3.10
The company also analysed CFS‑OSDI response rates in part 2 of SANSIKA. It noted that responses were similar in both treatment groups at months 9 and 12. At month 12, for people who had ciclosporin in both parts of SANSIKA, the response rate was 39.1%; for those who had the vehicle alone in part 1 and switched to ciclosporin in part 2, the response rate was 38.0%.
3.11
The company presented the health-related quality-of-life results from SANSIKA using the NEI‑VFQ‑25 and EQ‑5D questionnaires. The results using NEI‑VFQ‑25 were similar between treatment groups at baseline and at 6 months but there was an increase in the mean NEI‑VFQ‑25 composite score over time in both treatment groups. There were no differences in the EQ‑5D summary index and the EQ‑5D VAS score between baseline and at 6 months in both treatment groups, or between treatment groups. The company noted that the tariff used to estimate the health utility values was based on UK data from 1993 (Rabin et al. 2011).
3.12
The company presented the results of meta-analyses of SICCANOVE and SANSIKA for the composite end point CFS‑OSDI response rate at 6 months for:
-
all patients: 21.6% for ciclosporin compared with 13.1% for the vehicle (p=0.015)
-
patients with severe dry eye disease: 29.5% for ciclosporin compared with 18.3% for the vehicle (p=0.038)
-
patients with Sjögren's syndrome: 19.2% for ciclosporin compared with 11.6% for the vehicle (p=0.113)
-
patients with Sjögren's syndrome and severe dry eye disease: 23.4% for ciclosporin compared with 9.4% for the vehicle (p=0.036).
3.13
The company presented pooled adverse effects results from SANSIKA and SICCANOVE. The company explained that treatment-emergent adverse effects represent any event occurring after the baseline visits, related or not to the study medication, whereas treatment-related adverse effects represent an event considered by the investigator to be related to the study medication. The most frequent treatment-emergent adverse effects with ciclosporin were instillation site pain, eye irritation, instillation site irritation and eye pain. The most frequent treatment-emergent adverse effects with the vehicle were eye pain, meibomianitis (an inflammation of the meibomian glands, a group of sebaceous glands in the eyelids) and reduced visual acuity. The company concluded that the observed adverse effects of ciclosporin were mild to moderate and temporary and that overall ciclosporin is safe and well tolerated.
Cost effectiveness
3.14
The company presented a de novo Markov economic model that assessed the cost effectiveness of ciclosporin compared with standard care (artificial tears) in patients aged over 18 years with dry eye disease and severe keratitis whose disease had not adequately responded to artificial tears. The company stated that the cost-effectiveness analysis was conducted from an NHS and Personal and Social Services perspective, costs and outcomes were discounted at 3.5% per year, the time horizon was 30 years and the cycle length was 3 months. The company noted that because patients in SANSIKA represent the licensed population, inputs in the model were derived from this trial where possible. Because the comparator in SANSIKA (vehicle, which contained the excipient cetalkonium chloride) is not commercially available and artificial tears represent established clinical practice in the NHS for this population, the company viewed the response or reduction in the use of artificial tears in the vehicle group as a regression to the mean. The baseline use of artificial tears in SANSIKA was assumed to be reflective of standard care in the NHS. The model included 7 different states: treatment induction, treatment responders, non‑responders, temporary punctal plugs, permanent punctal plugs, post plugs and death. Patients were assumed to be aged 61 years, they could die at any time, and the model included equal numbers of men and women.
3.15
Treatment response was represented using the observed data from the post hoc analysis of CFS‑OSDI response rate from part 1 of SANSIKA (defined as improvement of 3 points or more from baseline CFS and improvement of 30% or more from baseline OSDI). Response rates from the vehicle group were used to derive response rates for the artificial tears group in the model. People whose disease responded to the 6‑month induction period continued treatment until there was no response. These response rates were derived from part 2 of SANSIKA. Patients who had the vehicle in part 1 of SANSIKA and ciclosporin in part 2 were not included in the estimates for the model. The company assumed that transition probabilities were constant over time. The probability of stopping treatment with ciclosporin after 6 months (the end of SANSIKA) was taken from the rate of patients stopping treatment with ciclosporin between 6 and 12 months in part 2 of SANSIKA. For the artificial tears group, the rate of patients who stopped having the vehicle during part 1 of SANSIKA was used as a proxy for the estimates after the end of the trial. The annual rate of temporary punctal plugs was assumed to be 0.01 based on a study by Clegg (2006) and only 10% of people who had temporary punctal plugs were assumed to then have permanent punctal plugs. The response rate to permanent punctal plugs was assumed to be 100%. Patients with temporary or permanent punctal plugs were assumed to not use artificial tears. Mortality rates were derived from the general population aged 61 years, which was the mean age of patients in SANSIKA.
3.16
The composition of preservative-free artificial tears was polyvinyl alcohol, carbomers and paraffin. The company assumed that administration, monitoring and testing costs with ciclosporin or artificial tears were zero, because all treatments were self-administered and it was assumed that the rate of ophthalmologist visits, tests and monitoring were similar in both treatment groups irrespective of the response status of the disease. It was assumed that people with severe dry eye disease have treatment in both eyes. The company assumed that the average number of drops per eye per day at baseline was similar in both treatment groups as in SANSIKA. The company incorporated the change in artificial tear use at 6 months to the ciclosporin and artificial tears groups in SANSIKA in the model, noting that the vehicle could have had an effect on the reduction of artificial tears use in the comparator group. For patients whose disease did not respond to treatment, the number of artificial tears per eye per day was similar to this use at baseline. Because treatment-related adverse effects were of low severity and transient, these were not included in the model other than through a reduction in the treatment continuation rates. The source of the costs for punctal plugs was NHS Reference Costs 2013. Unit costs were taken from the BNF (month not stated).
3.17
The company used utility data from SANSIKA in the model (utility for response: 0.74; utility for no response: 0.66). It noted that patients whose disease responds need fewer artificial tears and have a higher utility, which was assumed to be constant during response. Patients with punctal plugs had the same utility as patients whose disease responds with ciclosporin or artificial tears.
3.18
The company's cost-effectiveness analysis produced an incremental cost-effectiveness ratio (ICER) for ciclosporin plus artificial tears compared with vehicle plus artificial tears of £19,156 per quality-adjusted life year (QALY) gained, with an associated incremental cost of £713 and 0.037 additional QALYs.
3.19
The company conducted deterministic and probabilistic sensitivity analyses, which showed that varying the utility value for responders had the largest effect on the ICER. When varying the utility value for responders between 0.67 and 0.81, the ICER for ciclosporin plus artificial tears compared with artificial tears alone ranged from £165,654 to £10,166 per QALY gained. Other variables that had a notable effect on the ICER were the acquisition cost of ciclosporin and the response probabilities to ciclosporin and the vehicle at 6 months. The probabilistic analysis results gave an ICER of £18,835 per QALY gained for ciclosporin plus artificial tears compared with vehicle plus artificial tears. The company noted that ciclosporin had a probability of 46.4% to be considered a cost-effective use of NHS resources at a maximum acceptable ICER of £20,000 per QALY gained. It also noted that a number of simulations were associated with incremental benefits close to zero, meaning that the probabilistic results should be interpreted with caution.
3.20
The company presented results from several scenario analyses including:
-
using the primary end point definition for CFS‑OSDI from SANSIKA (that is, improvement of 2 points or more from baseline CFS and improvement of 30% or more from baseline OSDI): ICER for ciclosporin plus artificial tears compared with artificial tears alone, £19,156 per QALY gained
-
using utility values from Schiffman et al. (0.72 for non-responders and 0.78 for responders): ICER for ciclosporin plus artificial tears compared with artificial tears alone, £33,291 per QALY gained
-
varying the time horizon (showing that the ICER increases above £20,000 per QALY gained when the time horizon is less than 10 years)
-
assuming that only 1 eye is treated: ICER for ciclosporin plus artificial tears compared with artificial tears alone, £23,290 per QALY gained.
3.21
The company did not present a subgroup analysis for patients with Sjögren's syndrome. It noted that SANSIKA was not powered to assess the benefit of ciclosporin in this subgroup, and any inference would have meant using published literature in different patient groups or clinical input which would have added uncertainty to the model.
Company's response to the Committee's request
3.36
The company, in response to consultation, provided a response to all the Committee's requests described in the appraisal consultation document. It did an updated systematic review with the aim of conducting an indirect treatment comparison of the clinical effectiveness of ciclosporin plus corticosteroids (if needed) and artificial tears, and that of corticosteroids (if needed) plus artificial tears. However, the company stated that a robust indirect comparison was not possible because of methodological problems and the evidence available.
3.37
The company also presented a revised economic analysis of the cost effectiveness of ciclosporin plus corticosteroids (if needed) and artificial tears, and corticosteroids (if needed) plus artificial tears addressing the Committee's request. This cost-effectiveness analysis included:
-
the original SANSIKA CFS‑OSDI definition of response (that is, improvement of 2 points or more from baseline CFS and improvement of 30% or more from baseline OSDI)
-
evidence-based treatment stopping rates with ciclosporin plus corticosteroids (if needed) and artificial tears
-
changes to resource use and costs reflecting:
-
that artificial tears may be used alongside punctal plugs
-
both a baseline average and a 6‑month average for the number of artificial tear drops used per day, for both treatment groups
-
the assumption that ciclosporin is dispensed and costs are incurred monthly
-
sensitivity analyses using different utility values for response by treatment group
-
a subgroup analysis for people with Sjögren's syndrome and severe dry eye disease.
3.38
The company did regression analysis to determine which CFS‑OSDI definition of response (the original or the post‑hoc definition of response) was a stronger predictor of change from baseline utility at 6 months and found that the effect on utility was greater with the post‑hoc definition of response. The company therefore concluded that it was more appropriate to use the post-hoc definition of response in the economic analysis.
3.39
The company also did a regression analysis to determine the impact of Sjögren's syndrome on utility and concluded that it did not affect health-related quality of life. The company also did a regression analysis to determine whether Sjögren's syndrome had an impact on response. The results showed that Sjögren's syndrome was a statistically significant predictor of response at baseline and at 6 months using the original CFS‑OSDI response definition and at 3 months using the post‑hoc CFS‑OSDI response definition.
3.40
The company noted that because of the lack of clinical evidence for the comparison of ciclosporin plus corticosteroids (if needed) and artificial tears, with corticosteroids (if needed) and artificial tears, corticosteroids were included in the revised model as a cost parameter only. Based on clinical opinion, the company assumed the composition of corticosteroids to be flouromethalone and prednisolone and duration of treatment with corticosteroids to be 8 weeks. The company also assumed that people whose disease responded to treatment were less likely to need corticosteroids (10% of patients whose disease responded to treatment and 30% of patients whose disease did not respond to treatment).
3.41
The company's revised cost-effectiveness analysis when using the post‑hoc CFS‑OSDI response definition and applying a stopping rule based on the assessment of CFS‑OSDI response at 6 months as per its original analysis produced an ICER for ciclosporin plus corticosteroids (if needed) and artificial tears compared with vehicle plus corticosteroids (if needed) and artificial tears of £14,517 per QALY gained, with an associated incremental cost of £709 and 0.05 additional QALYs. When using the trial CFS‑OSDI response definition the ICER was £45,554 per QALY gained, with an associated incremental cost of £1,161 and 0.03 additional QALYs.
3.42
The company also presented a revised cost-effectiveness analysis applying a stopping rule based on the assessment of CFS‑OSDI response at 3 months instead of at 6 months as per its original analysis. The ICER for ciclosporin plus corticosteroids (if needed) and artificial tears compared with vehicle plus corticosteroids (if needed) and artificial tears when this assumption was applied was £33,432 per QALY gained, with an associated incremental cost of £425 and 0.01 additional QALYs. When using the trial CFS‑OSDI response definition the ICER was £24,696 per QALY gained, with an associated incremental cost of £627 and 0.03 additional QALYs.
3.43
The company noted that a gain in utility was seen for people whose disease responded to treatment, regardless of the treatment regimen or response definition. Treatment did not show a significant effect on utility (p=0.935), and the company considered that the observed differences in utility between the ciclosporin and vehicle groups were circumstantial. Therefore, the company concluded that it was more appropriate to use pooled utility values in the model.
3.44
The company presented a subgroup analysis for people with Sjögren's syndrome which resulted in an ICER for ciclosporin plus corticosteroids (if needed) and artificial tears compared with vehicle plus corticosteroids (if needed) and artificial tears of £16,231 per QALY gained when using the post‑hoc CFS‑OSDI response definition, and of £44,874 per QALY gained when using the original trial definition of response.
3.45
The company also presented a cost-minimisation analysis comparing different formulations of ciclosporin, assuming that each had equivalent efficacy, adverse event profiles and secondary costs. It noted that the 2% ciclosporin drops formulation developed by Moorfields Pharmaceuticals is no longer available and so the company did not include it in the analysis. The results showed that ciclosporin (Ikervis) is less costly (£72.00 monthly) than Restasis (£454.20 monthly) and Optimmune 0.2% ointment (£227.10 monthly).