A randomised controlled trial of adjunctive triamcinolone acetonide in eyes undergoing vitreoretinal surgery for open globe trauma the ASCOT study

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 12/35/64. The contractual start date was in April 2014. The draft report began editorial review in November 2021 and was accepted for publication in October 2022. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

Copyright © 2023 Charteris et al. This work was produced by Charteris et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaption in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2023 Charteris et al.

Background

Trauma is an important cause of visual impairment and blindness worldwide and a leading cause of blindness in young adult males. 1 Globally, it has been estimated that 1.6 million people are blind as a result of ocular trauma, with 2.3 million suffering bilateral low vision. 2 Ocular trauma is the most common cause of unilateral blindness in the world today, with up to 19 million with unilateral blindness or low vision. 2 It is estimated that almost one million people in the United States live with trauma-related visual impairment. 3 Ocular trauma has extensive socioeconomic costs; patients with open globe injuries lose a mean of 70 days of work. 4 In the United States, work-related eye injuries cost over $300 million per year (www.preventblindness.org), which equates to an annual cost to the UK economy (for which no comparable data exist) of £37.5 million.

In the UK, it is estimated that 5000 patients per year sustain eye injuries serious enough to require hospital admission and, of these, 250 will be permanently blinded in the injured eye. 5 Recent European studies document incidences of 2.4 and 3.2 per 100,000 per year6,7 for open globe injuries, which suggests an annual incidence for the UK of between 1500 and 2000.

Ocular injuries that result in visual loss invariably affect the posterior segment of the eye, and prevention of visual loss involves posterior segment (vitreoretinal) surgery. It is clear from recent published data that although vitreoretinal surgical techniques have improved, outcomes remain unsatisfactory, and that development of the intraocular scarring response proliferative vitreoretinopathy (PVR) is the leading cause. 8–11

Proliferative vitreoretinopathy

Eyes sustaining penetrating or open globe trauma (OGT) are a group at high risk of severe visual impairment. Retinal detachment (RD) is common in these eyes and multiple surgical interventions are often necessary. PVR is the most common cause of recurrent RD and visual loss in eyes with OGT. It is documented to occur in 10–45% of all OGT,8–11 its incidence varying with the nature of the penetrating injury. 8

Proliferative vitreoretinopathy is a process of fibrocellular scar tissue formation, which complicates 5–12% of cases of primary RD, 16–41% of giant retinal tears and 10–45% of cases of posterior segment trauma. 12 PVR represents a difficult vitreoretinal surgical challenge and although final retinal attachment may now be achieved in many cases, multiple surgeries are often needed and visual results are frequently very poor. 12,13 Binocular vision outcomes are notably unsatisfactory in PVR. 14 PVR management is costly in patient time and healthcare resources. 13

Preclinical data

Clinical observation and laboratory investigations undertaken on eyes with PVR and surgical specimens have identified potential targets for pharmacological adjuncts to its surgical management. 14 The cellular components of PVR peri-retinal membranes (retinal pigment epithelium, glial, inflammatory and fibroblastic cells) proliferate and may also be contractile and are thus targets for antiproliferative agents. There is a notable inflammatory component to the PVR process, with marked blood–retinal barrier breakdown and a greater tendency to intraocular fibrin formation. 15 Macrophages and T lymphocytes have been identified in PVR membranes14 and, although relatively small in number, they may play an important role in membrane development and contraction through growth factor production. Thus, both cellular proliferation and the intraocular inflammatory response are realistic targets for adjunctive treatments in PVR.

Steroid treatment has the potential to influence both the inflammatory and proliferative components of the pathological process of PVR. Previous experimental work has suggested that triamcinolone acetonide (TA) can reduce the severity of PVR. 16 It has also been demonstrated that periocular corticosteroids can reduce the severity of experimental PVR. 17 Laboratory work has indicated that TA appears to have no significant retinal toxicity,18 although in vitro it downregulates the proliferation of retinal cells.

Clinical data

Intravitreal TA has been used extensively clinically to treat macular oedema, intraocular inflammation and subretinal neovascularisation without demonstrable retinal toxicity but with a raised incidence of elevated intraocular pressure (IOP) and cataract. Previous small-scale clinical studies of PVR have suggested that systemic prednisolone,19 infused dexamethasone20 and intravitreal TA21–23 can reduce the severity of PVR, although none of these studies was of sufficient power to provide a definitive answer.

Rationale and risks/benefits

The Adjunctive Steroid Combination in Ocular Trauma (ASCOT) project was a phase 3, multicentre, randomised clinical trial to test the hypothesis that adjunctive steroid (TA) given locally at the time of surgery can improve the outcome of vitreoretinal surgery for OGT (both visually and anatomically). Open globe ocular trauma complicated by intraocular scarring (PVR) is a relatively rare, blinding, but potentially treatable, condition for which surgery is often unsatisfactory and visual results frequently poor. To date, no pharmacological adjuncts to surgery have been proven to be effective. Analysis of the costs and economic effectiveness of the trial intervention were also undertaken.

Assessment and management of risk

For the purposes of this study, TA was used outside the terms of its licence.

Risk–benefit analysis

We classified this trial as type A (no higher than the risk of standard medical care in the ADAMON project classification) based on the following analysis.

Triamcinolone acetonide has been used off-label in clinical ophthalmic practice for many years. Ophthalmologists have experience of its periocular administration for over 50 years, with administration via the intraocular route being adopted for over 30 years. It has been used to treat a variety of posterior segment ocular inflammatory pathology. 24–27 Its use as an intraocular surgical adjunctive tool for visualisation of the posterior hyaloid during pars plana vitrectomy has been well established. 28 Additionally, intraocular TA has been found to reduce postoperative inflammation following vitrectomy surgery. 29 It has been investigated specifically to determine its effect on vitreoretinal scarring (PVR), with varying success. 21–23

Triamcinolone acetonide has an extremely well documented safety profile with the most common significant adverse effect recorded as elevated IOP. 30 Data from the pilot study31 performed at the principal site found a similar incidence of elevated IOP 35% (n = 7) in patients who received intravenous TA compared with 25% (n = 5) in those patients who received standard care.

An audit of study sites revealed that over half (54%) of sites have used intraocular TA in vitrectomy surgery following OGT, with 25% of sites using it routinely in this patient population. The investigators concluded that there is extensive clinical experience with the product and had no reason to suspect a different safety profile in the trial population.

Trial design

The ASCOT study was a multicentre, prospective, individually randomised, patient and outcome assessor masked controlled trial that tested the superiority of the intervention at six months. The trial design was formulated in consultation with the accredited clinical trials unit at King’s College London, two trial statisticians, a methodologist at the Research Design Service and in partnership with patients who have suffered severe ocular trauma. 32 A total of 28 vitreoretinal surgery centres throughout the UK agreed to take part. Some 300 adult patients with OGT were scheduled to be randomised one to one to receive adjunctive intraocular and periocular steroid (TA) versus standard care (surgery without adjunctive treatment). Operating surgeons were masked until the end of surgery (when the adjunct is given). Patients and primary outcome observers were masked throughout.

The primary outcome was the proportion of participants with a clinically meaningful improvement in corrected VA in the study eye, defined as having a change in 10 letters or more in Early Treatment Diabetic Retinopathy Study (ETDRS) score (measured using validated ETDRS vision charts at a starting distance of 4 m) between baseline and at six months. The sample size was based on detecting a 19% increase (55–74%) in the proportion of participants who have a meaningful minimum improvement in VA the primary outcome. Secondary outcomes were the change in ETDRS score at six months after surgery, the development of scarring (PVR), RD, IOP abnormalities and other complications in the study eye between initial surgery and six months after initial surgery. Quality of life assessments were undertaken using the EQ-5D and VFQ-25 tools. Using these data, cost effectiveness and cost-consequence analyses to investigate the impact of injury and recovery were carried out. Data collection was undertaken at baseline (prior to initial study surgery), three months after initial surgery and six months after initial surgery.

Timetable

The project was projected to run for 48 months, with time allocated as follows: months 0–5 project set-up, 6–40 patient recruitment (6–11 stage 1 internal pilot, 12–18 stage 2 internal pilot, 19–40 phase III), 41–45 final follow-up, 46–48 write up and results dissemination.

Owing to slow patient recruitment, a 33-month extension to the recruitment phase was granted.

Selection of patients

Patients with an open globe injury undergoing vitrectomy either as a primary or secondary procedure were investigated. OGT was classified as one of the following: a full thickness eyewall injury in the form of a rupture caused by a blunt object, laceration caused by a sharp object or an intraocular foreign body (IOFB). Patients were recruited at vitreoretinal outpatient or emergency clinics at 28 participating vitreoretinal surgical centres throughout the UK. The worse injured eye in patients with bilateral eye injuries was selected as the study eye for randomisation and the better eye received standard treatment. As binocular trauma is a rare occurrence, the study was not stratified by binocularity.

Inclusion criteria

1. Adult subjects (aged 18 years or over at the time of enrolment).

2. Full thickness, open globe ocular trauma undergoing vitrectomy.

3. Ability to give written informed consent.

4. Willingness to accept randomization and attend follow-up for six months.

Exclusion criteria

1. Children (<18 years of age at time of enrolment).

2. Pre-existing uncontrolled uveitis.

3. Definitive diagnosis of previous steroid induced glaucoma – these patients were considered to be at risk of steroid related pressure rise and will be excluded (this did not include patients in whom a query of previous steroid-induced raised IOP has been postulated).

4. Pregnant or breastfeeding women. Women of childbearing potential were advised to use an effective method of contraception (hormonal or barrier method of birth control; true abstinence) from the time consent was signed until six weeks after their completion of the trial. Women of childbearing potential had to have a negative urinary pregnancy test within seven days prior to being registered for trial treatment (patients were considered not of childbearing potential if they were permanently sterile; i.e. they had undergone a hysterectomy, bilateral tubal occlusion or bilateral salpingectomy) or they were postmenopausal.

5. Allergy or previous known adverse reaction to TA.

6. Inability to attend regular follow-up.

7. Inability to give written informed consent.

8. Current or planned systemic corticosteroid use of a dose above physiological levels (e.g. >10 mg prednisolone).

Recruitment

The study was a multicentre trial involving 28 UK sites. Ethically approved trial specific adverts were distributed to all sites for display. Recruitment was monitored closely at regular intervals.

Study procedures and schedule of assessments

Intraoperative randomisation

1. The operating surgeon confirmed retina satisfactory (i.e. final confirmation of eligibility).

2. The surgeon completed the surgical procedure and confirmed that they were ready to randomise.

3. A theatre staff member telephoned eSMS global and communicated the patient information.

4. Treatment allocation was revealed and communicated to the operating surgeon such that patient was unaware and remained masked (i.e. if under local anaesthesia).

5. The surgeon administered the investigational medicinal product according to protocol, depending on treatment allocation.

If a participant was eligible for the study out of hours (overnight or at weekends) when their condition required urgent treatment and a member of the research team is unavailable, the site principal investigator or operating surgeon followed the above procedure for randomisation.

Masking

Assessments

Investigational treatment

The following medications were administered to participants allocated to the treatment arm of the study:

1. 4 mg/0.1 ml TA administered into the vitreous cavity by the operating surgeon at end of procedure.

2. 40 mg/1 ml of TA administered into the sub-Tenon’s space at the end of the procedure.

Usual care/control arm

Patients underwent standard vitreoretinal surgery and postoperative care appropriate to their ocular trauma without the addition of adjunctive medication.

Adverse events

Adverse events were recorded with clinical symptoms and accompanied with a simple, brief description of the event, including dates as appropriate. Adverse events were reported on the eCRF. Serious adverse events were reported in an expedited manner to the sponsor for each participant for their duration in the trial.

Data management and quality assurance

Data handling and analysis

InferMed MACRO eCRF version 4 was used to record the study data. Staff were allocated data entry or monitor roles to access the system by KCTU. At the end of the trial, after queries were resolved and all data fields completed, the database was locked for analysis. All study documents are to be retained for a period of five years following conclusion of the study. Data query reports were generated monthly by the data management team at KCTU and passed to the trial manager to be actioned. The trial manager raised data queries with sites via the eCRF during or between site monitoring visits.

Outcomes

Sample size calculation

Published and pilot data indicate that VA, defined as ETDRS letter score at six months is skewed in this population and that the shape of the distribution of compared with VA differs between treatment arms. In a pilot randomised controlled trial (RCT), we observed small difference in mean VA between treatment arms but a sizable difference in proportion of patients (80% vs. 55%) with a meaningful improvement in VA (a change of ETDRS letter score of at least 10, which is widely accepted to be clinically meaningful in research studies of eye disease). 13,15,19–22 We defined the primary outcome as the proportion of patients with a meaningful improvement in VA. With 140 participants per group and a statistical significance of 5%, there is at least 90% power to detect a 19% increase (55–74% corresponding to an OR of 2.33) in participants who have a meaningful minimum improvement in VA of at least 10 letters. We therefore allowed for a 7% dropout rate and the target sample size for ASCOT was 300 participants (150 per arm).

Following slower than anticipated recruitment, the recruitment period was extended to 75 months. Over the full recruitment period, 280 eligible patients were recruited and are included within this analysis. Based on the original sample size parameters outlined above, it was established that the trial would still be adequately powered with a sample size of 280. A sample size of 280, assuming loss to follow-up of 7%; that is, 260 completers at 6 months provided 89.7% power to detect a 19% increase (55–74%) in meaningful improvement in VA (≥10 letters).

Statistical methods

Analysis of secondary outcomes

Linear (Gaussian) mixed regression models were used for the analysis of the principle secondary outcome (change in ETDRS) and other continuous secondary outcomes (VFQ-25). Binary secondary outcomes were analysed using mixed logistic regression models. For count outcomes, a mixed-effect negative binomial model was fitted, which allowed for overdispersion. Similar to the primary analysis model, the models for secondary outcomes included centre as a random intercept and a fixed effect for treatment group. For continuous secondary outcomes, models additionally included a fixed effect for the baseline value of the outcome.

The continuous change in the six-month ETDRS (principle secondary outcome) was also analysed using a Bayesian linear mixed regression model analysis, fitted using Markov chain Monte Carlo (MCMC) methods. Similar to the main frequentist analysis for this principle secondary outcome, the model included fixed effects for baseline ETDRS, treatment group and a random intercept for centre. Uninformative priors were used for each model parameter, specifically for each fixed regression parameter Normal (0, 100) and igamma (0.01, 0.01) for the variance parameters. A burn in of 2500 and 10,000 MCMC iterations were used. Convergence of model parameters was assessed visually using diagnostic plots. From the model, we present the estimated average treatment effect for the six-month change in ETDRS with an accompanying 95% credible interval, together with posterior probabilities for the change in ETDRS being greater than 0–50, by treatment group and for the treatment group difference.

Adverse events were collected by means of spontaneous reports from participants, clinical observation and clinical examinations and blood tests. Data on safety outcomes are summarised for all randomised participants who underwent surgery. Adverse events were summarised by type: adverse events, adverse reactions (a subset of the adverse events), unexpected adverse reactions (a subset of the adverse reactions), serious adverse events, serious adverse reactions (a subset of the serious adverse events) and unexpected serious adverse reactions (a subset of the serious adverse reactions) and by treatment arm. Adverse events were tabulated by treatment group for both the number of events and the number of participants with the type of event. Adverse events were also summarised by event term and intensity (subjectively assessed by local clinical investigators as mild/moderate/severe). To identify the events with the strongest evidence for between-arm differences, adverse events were summarised visually in a dot plot, which display the proportions of individuals experiencing each type of event by treatment arm and the relative risk difference with 95% CI.

Subgroup analysis

Preplanned subgroup analysis investigated whether the treatment effect on the primary outcome differed by:

1. RD: attached

2. RD: tractional

3. RD: rhegmatogenous

4. fovea involvement: yes

5. fovea involvement: no

6. fovea involvement: splitting

7. presence of PVR: yes

8. presence of PVR: no

9. presence of retinal incarceration: yes

10. presence of retinal incarceration: no

11. lens status at baseline: clear (phakic)

12. lens status at baseline: cataract (phakic)

13. lens status at baseline: anterior chamber intraocular lens implant (ACIOL) and posterior chamber intraocular lens implant (pseudophakic)

14. lens status at baseline: aphakic.

Each subgroup analysis was performed by adding the relevant treatment-by-subgroup interaction term to the same analysis model as for the primary outcome; p-values for each interaction term were presented. No adjustment for multiple tests was made and the results were hypothesis generating only. The consistency of estimates was depicted visually by means of a forest plot.

Health economics analyses

The ASCOT study took an NHS perspective in terms of the identification, measurement and valuation of costs and outcomes.

Health economics analyses

We analysed the incremental cost-effectiveness of the trial intervention (intraocular and periocular steroid) in eyes undergoing vitreoretinal surgery for ocular trauma compared with surgery alone in terms of changes in VA. To enable this analysis, from an NHS perspective, we:

1. fully costed the vitreoretinal surgery and follow up

2. recorded study participant primary and secondary care health service use and social care use over the six-month follow-up period (using a research nurse interviewer administered CSRI, costed using national unit costs) and making use of routine hospital data on surgical and postoperative care as part of the CRF

3. conducted a primary cost effectiveness analysis (using the trial primary outcome measure of VA, ≥10-letter improvement in ETDRS score, as our measure of effectiveness)

4. conducted a secondary cost utility analysis to explore the cost per QALY of adjunctive steroid intraocular and periocular steroid (TA) as compared with usual treatment relative to the £20,000–30,000 NICE payer threshold. NICE currently uses this threshold to determine whether the health benefits provided from a new drug or healthcare intervention is greater than the health likely to be lost as services are displaced to accommodate for the new intervention

5. through bootstrapping, generated CEACs to communicate to policy makers the probability that the intervention is cost-effective.

Patient and public contribution to study development

The trial was initially conceived with a primary outcome of anatomical success (retinal attachment without PVR). This was the approach used in previous clinical trial undertaken by the same investigators. 31,42–44 The investigator team undertook a series of meetings with patients who had suffered RD, ocular trauma and PVR, as well as patient groups affected by ocular trauma (Blind Veterans UK). There was a clear preference among patients and patient groups for a visual measure as the primary outcome. The primary outcome and the calculation of sample size for the study was therefore based on a measure of visual outcome (VA).

Data description

Recruitment and participant flow

Between December 2014 and March 2020, a total of 792 patients were screened; 317 were assessed for eligibility and 280 eligible participants across 27 sites in England and Scotland were randomly allocated to standard surgery (137 participants) or surgery plus adjunctive TA (143 participants, Table 1). One site failed to recruit. Figure 1 is the CONSORT flowchart for the trial, which summarises the flow of participants through the trial. One participant in each arm received the alternative treatment. The three-month follow-up was obtained for 269 participants (standard surgery n = 135 and surgery plus adjunctive TA n = 134) and for 259 at the six-month follow-up (standard surgery n = 129 and surgery plus adjunctive TA n = 130).

TABLE 1 - Randomisation by centre and treatment arm

Study centre	Standard surgery, N (%)	Surgery + TA, N (%)	Total, N (%)
Total randomised	137 (49)	143 (51)	280
Birmingham	5 (4)	4 (3)	9 (3)
Bristol	8 (6)	7 (5)	15 (5)
Cambridge	1 (1)	1 (1)	2 (1)
Canterbury William Harvey Hospital	2 (1)	1 (1)	3 (1)
Derby	2 (1)	3 (2)	5 (2)
Edinburgh	1 (1)	1 (1)	2 (1)
Frimley Park	2 (1)	0 (0)	2 (1)
Glasgow	2 (1)	3 (2)	5 (2)
Hull	2 (1)	2 (1)	4 (1)
King’s College London	2 (1)	3 (2)	5 (2)
Liverpool	1 (1)	0 (0)	1 (0)
Maidstone	7 (5)	8 (6)	15 (5)
Manchester	1 (1)	2 (1)	3 (1)
Moorfields	56 (41)	57 (40)	113 (40)
Newcastle	6 (4)	6 (4)	12 (4)
Oxford	0 (0)	1 (1)	1 (0)
Plymouth	1 (1)	1 (1)	2 (1)
Portsmouth	4 (3)	5 (3)	9 (3)
Sheffield	0 (0)	1 (1)	1 (0)
South Tees	7 (5)	7 (5)	14 (5)
Southend	3 (2)	2 (1)	5 (2)
St Thomas’ London	2 (1)	3 (2)	5 (2)
Stoke Mandeville Stoke Mandeville Hospital	1 (1)	2 (1)	3 (1)
Sunderland	5 (4)	6 (4)	11 (4)
Western Eye London	12 (9)	11 (8)	23 (8)
Whipps Cross London	0 (0)	2 (1)	2 (1)
Wolverhampton	4 (3)	4 (3)	8 (3)

FIGURE 1.

Consort flow diagram. (a) Includes four participants who were randomised in error but were ineligible and immediately withdrawn on date of randomisation. Numbers withdrawn are cumulative. See Tables 3 and 4 for more details on withdrawals.

Primary outcome: meaningful improvement (≥10) in the Early Treatment Diabetic Retinopathy Study at six months

Primary outcome analysis

A total of 259 participants (standard surgery, n = 129; surgery plus adjunctive TA, n = 130) who had a six-month follow-up were included in the inferential analysis of the primary outcome. A total of 56 (43.4%) participants in the standard surgery arm experienced a clinically meaningful improvement in VA (6 month change in ETDRS ≥ 10) compared with 61 (46.9%) in the surgery plus adjunctive TA arm (unadjusted difference in proportion 3.5%, 95% CI –8.6% to 15.6%; Figure 2). The adjusted OR for a clinically meaningful change in VA for surgery plus adjunctive TA relative to standard surgery was 1.03 (95% CI 0.61 to 1.75, p = 0.908), indicating no difference between the treatment arms. The population averaged marginal probability of clinically meaningful improvement in VA was 41.8% in the standard surgery arm compared with 44.2% in the surgery plus adjunctive TA arm.

FIGURE 2.

Clinically meaningful improvement (≥10) in ETDRS at six months.

Missing data sensitivity analysis

Sensitivity analysis was conducted to explore the impact of the missing data on the primary ETRDS outcome. Table 5 summarises the missing data by treatment arm. Sensitivity analysis initially explored the robustness of the primary analysis results to two extreme MNAR assumptions (Table 8):

1. Participants in the standard surgery arm have meaningful change; participants in surgery plus adjunctive TA arm do not.

2. Participants in the standard surgery group do not have meaningful change; participants in surgery plus adjunctive TA arm do have meaningful change.

In comparison with the primary treatment effect (OR 1.03, 95% CI 0.61 to 1.75), in scenario 1, the point estimate was more in favour of standard surgery (0.74, 95% CI 0.45 to 1.23) and in scenario 2, the point estimate was more in favour of surgery plus adjunctive TA (1.46, 95% CI 0.89 to 2.40). However, in both sensitivity analyses, inferences remained consistent with the primary analysis and did not identify a significant between treatment group difference.

Further MNAR scenarios were explored using a range of plausible assumptions of the odds of clinically meaningful improvement among those with missing data being 0 to 1 times the odds of clinically meaningful improvement amongst the observed (Figure 3). In all additional MNAR analyses, the treatment effect (OR) remained close to 1 and the 95% CIs continued to contain OR = 1, indicating that in all evaluated scenarios there was no evidence of a significant treatment effect, consistent with the primary inference.

FIGURE 3.

Sensitivity analysis exploring the impact of data MNAR. (OR >1 indicates higher event rate in the adjunct arm).

Out of visit window sensitivity analysis

Sensitivity analysis excluded data collected outside the visit window (6 months ± 4 weeks). An additional sensitivity analysis where data collected outside the four-week window was included, but where patients with data outside the four-week window were weighted by one-half was also be performed; Participants with data within the allowed visit window had a weight of one. In both sensitivity analyses, results were consistent with the primary analysis (Table 9).

Subgroup analysis

Subgroup analysis was performed for the primary outcome to explore the uniformity of the treatment effect found overall.

The p-values provide a test for whether there is any evidence to suggest that there is a difference in the overall treatment effect by the subgroups of interest. Subgroup effects should be interpreted with caution and viewed as exploratory, as the trial was not powered to detect subgroup effects; This is reflected by the large widths of the CIs for each subgroup effect depicted in Figure 4. For this reason, emphasis should be placed on the consistency of the treatment effect across the subgroups and not the individual effect within each subgroup. While there is some variation in the point estimates for individual subgroups, for all explored subgroup effects, the 95% CI overlap and the tests for interaction are not significant (p = 0.106 or larger).

FIGURE 4.

Forest plot showing the effect on meaningful change in ETDRS at six months of adding TA within subgroups. (a) Subgroup effects are not estimable for fovea involvement – splitting as only one participant observed in surgery plus TA arm to have splitting (no meaningful change). OR represents the baseline ETDRS adjusted odds of meaningful change for surgery plus TA relative to standard surgery for the associated subgroup. OR >1 means a higher event rate in the adjunct arm.

Principal secondary outcome

A total of 259 participants (standard surgery n = 129; surgery plus TA n = 130) were included in the analysis of the change in ETDRS at six months as a continuous outcome. The baseline adjusted mean difference in the month 6 change in ETDRS for surgery plus TA compared with standard surgery was –2.65 (95% CI –9.22 to 3.92, p = 0.430), with the point estimate in favour of standard surgery.

Bayesian analysis

In secondary Bayesian analysis using non-informative priors, the baseline adjusted mean difference in the month 6 change in ETDRS for surgery plus adjunctive TA compared with standard surgery was –1.01 (95% CI –7.13 to 5.75), with the point estimate in favour of standard surgery. The estimated probabilities of the six-month change in ETDRS and treatment group differences being greater than or equal to 0 to 50 letters are shown in Table 10. The probability of the six-month change in ETDRS being greater for surgery plus adjunctive TA relative to standard surgery was 0.372; this is a fairly low probability. The probability of the treatment group difference in six-month change in ETDRS being greater or equal than 10 points was very small, at 0.0001.

There was high autocorrelation in the Bayesian MCMC sample, resulting in low precision for some model estimates. It was identified that the random centre effect was contributing to the high autocorrelation. Therefore, post hoc, the analysis model was simplified and a single-level linear regression model including fixed effects for baseline ETDRS and treatment group was also fitted with uninformative priors. In the simplified model, the adjusted mean difference in the month 6 change in ETDRS for surgery plus adjunctive TA compared with standard surgery was –0.19 (95% CI –6.34 to 5.94), with the point estimate in favour of standard surgery. Comparable to the results of the more complex Bayesian model, there was a low probability (0.495) of the six-month change in ETDRS being greater for surgery plus adjunctive TA compared with standard surgery and the probability of the treatment group difference in six-month change in ETDRS being greater or equal to 10 points higher for surgery plus adjunctive TA compared with standard surgery was very small, at 0.0001 (Table 11).

Secondary outcomes

Post hoc sensitivity analysis

In accordance with the ASCOT statistical analysis plan, the above analysis of hypotony included data recorded on the secondary outcome forms at three and six months. Additional adverse events of hypotony (<6 mmHg) were recorded for eight participants (two standard surgery and six surgery plus adjunctive TA) on the adverse event form that were not captured on the three- or six-month secondary outcome form. This group of eight participants included one participant who later withdrew from the surgery plus adjunctive TA arm.

A post hoc analysis combined the data on hypotony across the two data sources. A total of 28/124 participants in the standard surgery arm had hypotony within six months of the study vitrectomy compared with 31/125 in the surgery plus adjunctive TA arm (22.6% vs. 24.8%) as recorded on the secondary outcome forms at three or six months or on the adverse event form (Table 18). The OR of hypotony for surgery plus adjunctive TA relative to standard surgery was 1.13 (95% CI 0.63 to 2.03, p = 0.680), which was comparable with the main analysis of hypotony with the point estimate in favour of surgery plus TA.

Post hoc sensitivity analysis

In accordance with the ASCOT statistical analysis plan, the analysis of raised IOP included data recorded on the secondary outcome forms at three and six months. Additional adverse events of raised IOP (>25 mmHg) were recorded for seven participants (five standard surgery and two surgery plus adjunctive TA) on the adverse event form that were not captured on the three- or six-month secondary outcome form. This group of seven participants included two participants who withdrew from the trial sometime after the recording of IOP (one from each arm).

A post hoc analysis combined the data on IOP across the two data sources. A total of 40/127 participants in the standard surgery arm experienced raised IOP within six months of the study vitrectomy compared with 58/125 in the surgery plus adjunctive TA arm (31.5% vs. 46.4%) as recorded on the secondary outcome forms at three or six months or on the adverse event form. The OR of raised IOP for surgery plus adjunctive TA relative to standard surgery was 1.88 (95% CI 1.13 to 3.15, p = 0.016), which was comparable with the main analysis of raised IOP in favour of standard surgery (Table 20).

Visual Function Questionnaire-25

A total of 213 participants (standard surgery n = 105; surgery plus adjunctive TA n = 108) had baseline and six-month VFQ-25 available and were included in the analysis of the VFQ-25. The adjusted mean difference in the month 6 VFQ-25 for surgery plus adjunctive TA relative to standard surgery was 0.78 (95% CI –3.53 to 5.10, p = 0.723), with the point estimate in favour of surgery plus TA (Table 22).

Safety event reporting

Data on safety outcomes are summarised for all randomised participants who underwent surgery (n = 280) by type of adverse event and ophthalmic category in Tables 23 and 24 and Figures 5 and 6. The majority of adverse events were not serious. There were more non-serious adverse events in the surgery plus TA arm (171 events in 80 participants) in comparison with the surgery only arm (142 events in 66 participants). Tables 25 and 26 summarise the non-serious and serious adverse events by treatment arm, relatedness and intensity. There were a total of five serious adverse events, which all occurred in the surgery plus TA arm (see Table 26 for details).

FIGURE 5.

Dot plot of adverse events. This figure plots displays the proportions of individuals experiencing each type of event by treatment group in the left-hand panel, the risk difference with 95% CI in the middle panel and numbers of participants experiencing each event and event totals in right-hand panel. Additional cases of raised IOP (>25 mmHg) were recorded on the secondary outcome forms at 3 and 6 months (n = 23 participants; 5 in group A and 18 in group B) that were not also captured on the adverse event form due to poor reporting (see Tables 19 and 20); 7 participants who had raised IOP recorded on an adverse event form did not have raised IOP recorded on the secondary outcome form (5 group A and 2 group B). Additional cases of hypotony (<6 mmHg) were also recorded on the secondary outcome forms at 3 and 6 months (n = 9; 3 in group A and 6 in group B) that were not also captured on the adverse event form due to poor reporting (see Tables 17 and 18); 8 participants who had hypotony recorded on an adverse event form did not have hypotony recorded on the secondary outcome form (2 group A and 6 group B). The participants experiencing hypotony or raised IOP on secondary outcomes but not included in the adverse events are included in Figure 6.

FIGURE 6.

Dot plot of adverse events including raised IOP and hypotony events from secondary outcomes. This figure plots displays the proportions of individuals experiencing each type of event by treatment arm in the left-hand panel, the risk difference with 95% CI in the middle panel and numbers of participants experiencing each event and event totals in the right-hand panel.

Health economics results

Analysis of baseline data

For the health economic evaluation, a baseline test was conducted using the Mann–Whitney U test. Table 27 details the outcomes (ETDRS, EQ-5D, VFQ and cost), which showed that there was no significant difference in the data at baseline.

TABLE 27 - Baseline equivalence of test of the ETDRS, EQ-5D, VFQ and cost using the Mann–Whitney U test

Standard surgery			Intervention
	Mean	SD	Mean	SD	p-value
ETDRS	16.92	30.87	10.09	22.24	0.35
EQ-5D	0.68	0.27	0.68	0.30	0.63
VFQ	67.52	21.51	65.30	23.02	0.52
Cost (£)	2534.97	4332.49	2840.37	3767.76	0.25

The ICER value of the primary outcome was £8,362.32 per 10 or more letter improvement, as shown in Table 28. We have no payer threshold against which to compare this value. Figure 7 shows the ICER plot of the ETDRS ≥10-letter improvement. The mean effect of the ASCOT cohort for the EQ-5D was slightly lower, with a value of 0.003 QALY, but for the VFQ value, the standard care had a lower mean effect of 0.03. The ICER values for the EQ-5D show that the ASCOT intervention was dominated by the control while the VFQ was £10,897.20 per vision improvement quality of life.

FIGURE 7.

A deterministic sensitivity analysis of the ICER: tornado diagram for a multiple one-way sensitivity analysis.

TABLE 28 - Cost-effectiveness results for the ASCOT intervention compared with standard care

Outcome	Standard surgery		Intervention		Incremental change		ICER
	Mean cost (£)	Mean effect	Mean cost (£)	Mean effect	Cost	Effect
ETDRS:					£324.12		8362.32
≥10	4767.10	0.43	5774.92	0.47
<10	4501.67	0.57	4192.96	0.53		0.04
EQ-5D:							Dominated
≥10	4767.10	0.37	5774.92	0.36
<10	4501.67	0.36	4192.96	0.37		–0.00
VFQ:							10,897.20
≥10	4767.10	34.83	5774.92	34.03
<10	4501.67	34.16	4192.96	34.90		0.03

It is worth noting that the outcome ETDRS and VFQ-25 are not preference-based measures, thus cannot be compared with respect to the NICE threshold of £20,000–30,000 per QALY gained.

The EQ-5D is a preference-based measure; see Figure 7 for the ICER scatter plot, which shows a slightly denser plot in the northwest quadrant. This result is further emphasised in the CEAC curve (Figure 8).

FIGURE 8.

A comparison of the incremental cost-effectiveness with respect to changes in the intervention cost, a one-way sensitivity analysis. Each variable is analysed for a ±50% variability.

The CEAC plot shows the intervention has a very low probability of cost-effectiveness; approximately 30% at a willingness to pay of £0–30,000 per QALY gained compared with the control intervention.

A deterministic sensitivity analysis was conducted, and the results are shown in the tornado plot in Figure 7. The tornado plot shows changes in the value of the ICER when variables influencing the outcome of the health economic evaluation are varied. In the tornado plot, we have varied each variable (cost of intervention, probability of EDTRS ≥10, mean cost in the intervention arm of participant with ETDRS ≥10 and mean cost in the intervention arm of participant with ETDRS <10), by ±50% of its original value to investigate the change in ICER value. We see that the greatest influencers of the ICER value are the mean costs incurred during the follow-up phase. A reduction of the cost of intervention by 50% gives a lower ICER of £6653 per EDTRS ≥10 while an increase of 50% gives an ICER value of £10,071 per ETDRS ≥10.

Further investigation into the intervention cost showed that there is a positive linear correlation between the intervention cost and the ICER values (Figure 8). An extrapolation at no intervention cost gives an ICER of £4944 per EDTRS ≥10, while an intervention cost of £500 gives an ICER value of £17,844 per EDTRS ≥10.

Other secondary effects, EQ-5D and VFQ, were also investigated. The correlation between the VFQ and the EQ-5D for the two arms of the intervention and the combined data is very high, for the ASCOT intervention is 0.81, for the control is 0.77 and for the combined dataset it is 0.79.

Table 28 demonstrates that using the secondary results of the EQ-5D from the ASCOT intervention is dominated by the minimal negative change in QALY value of 0.003 and a positive change in the cost in comparison to the control/standard arm. A probability sensitivity analysis of the QALY result is shown in the scatter plot (Figure 9).

FIGURE 9.

Scatter plot for probabilistic sensitivity analysis showing the cost-effectiveness of intervention (TA included) compared with the control (TA not included).

The scatter plot is distributed in the four quadrants but the north-west is the more densely populated region, hence we proceed to the CEAC curve, as shown in Figure 10. The CEAC curve shows lower probability, ranging around 0.3 for the ASCOT intervention in comparison with the control/standard care (see Figure 10). This confirms the results previously obtained that there is little or no advantage of implementing the ASCOT intervention over and above standard care.

FIGURE 10.

Cost-effectiveness acceptability curves for the intervention (top) vs. standard/control group (bottom).

Tables 29 and 30 detail the costs of standard care and the ASCOT intervention.

TABLE 29 - Summary table of hospital-based service use mean (SD) costs by patients (N = 259) in the ASCOT trial over three time points (baseline, three months and six months).

Department	Standard care (£)(n = 129)			ASCOT intervention (£)(n = 130)
	Baseline	3 months	6 months	Baseline	3 months	6 months
Accident and emergency	164.16 (133.51)	31.27 (110.66)	20.85 (72.77)	166.73 (134.72)	36.21 (100.28)	16.81 (58.61)
Ophthalmology inpatient	389.06 (1024.06)	28.84 (165.00)	31.71 (154.31)	511.84 (846.10)	97.22 (332.29)	25.77 (105.41)
Other inpatient ward	37.07 (334.16)	0.00 (0.00)	8.64 (98.17)	32.06 (327.09)	0.01 (0.09)	0.02 (0.18)
Ophthalmology outpatient	151.37 (361.28)	180.35 (564.90)	65.12 (111.95)	164.93 (344.77)	173.39 (280.86)	116.67 (341.87)
Other outpatient	43.58 (149.46)	44.49 (429.39)	27.98 (227.85)	31.81 (74.62)	15.18 (97.82)	19.44 (145.52)
Other hospital-based services	106.17 (21.96)	3.56 (31.87)	0.00 (0.00)	80.96 (16.82)	4.06 (34.98)	32.05 (327.26)

TABLE 30 - Summary table of community-based service use mean (SD) costs by patients (n = 259) in the ASCOT trial over three time points (baseline, three months and six -months)

Community-based service	Standard care (n = 129)			ASCOT intervention (n = 130)
	Baseline	3 months	6 months	Baseline	3 months	6 months
GP visit	42.83 (77.29)	49.38 (102.06)	38.30 (76.05)	51.46 (95.31)	49.38 (113.26)	53.39 (146.80)
Practice nurse	22.64 (91.17)	15.12 (53.35)	20.35 (87.32)	17.31 (54.08)	15.00 (61.52)	15.00 (145.09)
District nurse	0.66 (7.48)	5.19 (46.96)	1.24 (14.89)	37.69 (400.40)	17.23 (196.46)	26.73 (233.29)
Social worker	0.00 (0.00)	5.11 (50.74)	0.00 (0.00)	20.39 (216.80)	2.08 (23.68)	1.39 (11.12)
Counsellor	10.58 (96.31)	17.91 (126.66)	27.52 (145.82)	4.00 (37.80)	15.31 (107.07)	12.85 (62.48)
Dietician	0.89 (10.13)	6.24 (54.38)	0.00 (0.00)	0.89 (10.09)	5.31 (60.52)	0.00 (0.00)
Optician	13.53 (41.94)	15.50 (44.22)	31.98 (97.93)	20.15 (59.82)	15.38 (53.95)	80.73 (724.01)
Dentist	24.03 (59.47)	17.79 (63.99)	29.30 (80.86)	34.19 (67.78)	10.38 (46.39)	16.62 (47.59)
Physiotherapist	9.88 (83.02)	8.99 (78.49)	7.87 (55.30)	16.34 (126.75)	4.46 (30.95)	15.62 (85.16)
Occupational health therapist	6.01 (35.74)	6.01 (30.04)	3.61 (23.45)	12.85 (101.21)	14.31 (138.43)	5.96 (44.87)
Alternative therapist	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)
Other community-based services	15.12 (148.40)	5.58 (52.28)	10.43 (72.59)	5.98 (31.69)	7.65 (45.53)	17.11 (120.13)

Contributions of authors

David G Charteris (https://orcid.org/0000-0001-6267-4180) (Professor of Ophthalmology, Consultant Ophthalmologist) Devised the investigation, coordinated study development, wrote research application and protocol and acted as chief investigator for the study.

Philp Banerjee (https://orcid.org/0000-0002-6113-1173) (Consultant Ophthalmologist) Supported study development and protocol writing.

Edward Casswell (https://orcid.org/0000-0002-8837-9910) (Research Fellow, Ophthalmology) Coordinated patient recruitment and publication writing.

Rhiannon Tudor Edwards (https://orcid.org/0000-00003-4748-5730) (Professor of Health Economics) Initiated and oversaw health economics analysis.

Victory Ezeofor (https://orcid.org/0000-0002-4211-8942) Health economics analysis.

Bethany Anthony (https://orcid.org/0000-0002-2593-1069) Health economics analysis.

Suzie Cro (https://orcid.org/0000-0002-0935-3713) (Trial Statistician) Carried out statistical analysis throughout the trial and final statistical analysis.

Victoria R Cornelius (https://orcid.org/0000-0002-2880-2086) (Trial Statistician) Provided statistical analysis plan and oversaw statistical analysis at all stages of the trial.

Catey Bunce (https://orcid.org/0000-0002-4389-5284) (Trial Statistician) Provided statistical analysis plan and oversaw statistical analysis at all stages of the trial.

Elizabeth Robertson (https://orcid.org/0000-0001-7547-8998) (Trial Manager) Managed all aspects of the study day-to day.

Joanna Kelly (https://orcid.org/0000-0001-6687-426X) Provide trail coordination support.

Caroline Murphy (https://orcid.org/0000-0002-0080-1065) Oversaw all aspects of trial conduct and safety.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure to protect everyone’s privacy and it is important that there are safeguards to make sure that they are stored and used responsibly. Everyone should be able to find out about how patient data is used. #datasaveslives. You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Confidentiality/anonymity

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Publications

Banerjee PJ, Cornelius VR, Phillips R, Lo JW, Bunce C, Kelly J, et al. Adjunctive intraocular and peri-ocular steroid (triamcinolone acetonide) versus standard treatment in eyes undergoing vitreoretinal surgery for open globe trauma (ASCOT): study protocol for a phase III, multi-centre, double-masked randomised controlled trial. Trials 2016;17(1):339.

Lo JW, Bunce C, Charteris D, Banerjee P, Phillips R, Cornelius VR. A phase III, multi-centre, double-masked randomised controlled trial of adjunctive intraocular and peri-ocular steroid (triamcinolone acetonide) versus standard treatment in eyes undergoing vitreoretinal surgery for open globe trauma (ASCOT): statistical analysis plan. Trials 2016;17:383.

Disclaimers

This report presents independent research funded by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care.