Informed consent

Informed consent to participate in the study was sought from potentially eligible participants by the ophthalmologist or a designee at each participating site. The treating ophthalmologist also obtained informed consent before carrying out the allocated laser procedure.

Participant withdrawal

Participants could withdraw from the study at any time without prejudice. In the event of consent withdrawal, patients were to be asked for their permission to use the data already collected. If this permission was declined, then any collected data on that patient would not be used in the trial analysis. Withdrawal of consent was recorded on the patient’s CRF.

Micropulse laser

Subthreshold micropulse laser was delivered using a 577 nm optically pumped diode laser (IQ 577™ laser system; IRIDEX Corporation, Mountain View, CA, USA). The IRIDEX Corporation laser system was used as IRIDEX was, at the time the trial was designed and initiated, the only manufacturer that produced a 577 nm wavelength laser that could have potential benefits when compared with other available wavelengths, including lack of absorption by the macular pigment, very good absorption by melanin and good penetration through lens opacities. SML was applied confluently to the macular area, using three 7 × 7 spot grids above and below the fovea (500 µm from its centre) and one 7x7 spot grid at each side (temporal and nasal) of the fovea (500 µm from its centre). In addition, treatment was also applied to areas of thickening located outside this central area. Firstly, a threshold was set by titrating the power of the laser upwards, starting from 50 mW, in 10 mW increments, in an area where oedema was present, around > 2 disc diameters from the foveal centre (if possible), and until a barely visible tissue reaction was seen. If a reaction was evident with 50 mW, the power was not increased. Then, the laser was switched to the micropulse mode with the power of the laser set at x4 the threshold identified (e.g. if a barely visible reaction was seen at 50 mW, then micropulse laser was applied with a 200mW power). SML was then undertaken using a spot size of 200 µm, duty cycle of 5%, and ‘on’ duration of 200 ms (composed of multiple 0.1 ms of ‘on’ pulses, with 1.9 ms of ‘off’ time in between ‘on’ pulses). A standard operating procedure (SOP) for SML was prepared for the trial (see Report Supplementary Material 1).

Standard laser

For patients allocated to SL, the laser was applied to areas of thickened retina, macular non-perfusion (away from and non-contiguous with the perifoveal capillaries) and leaking microaneurysms, in accordance with the ETDRS and the Royal College of Ophthalmologists guidelines.^5,27 FFA and SD-OCT were used to identify areas of non-perfusion and leakage (FFA) and thickening (SD-OCT) before treatment, at the discretion of the treating ophthalmologist. Treatment was applied to obtain a mild grey–white burn evident beneath leaking microaneurysms and in other areas of leakage or non-perfusion not affecting the perifoveal capillaries based on FFA, if FFA had been obtained, or to cover areas of thickening if treatment was given based on OCT findings, or both. The treatment was intended to spare the central 500 µm and the area within 500 µm from the optic nerve head. A standard operating procedure (SOP) for SL was prepared and used in the trial (see Report Supplementary Material 2).

The SL treatment in the DIAMONDS trial was performed using a modified ETDRS technique. In the ETDRS, argon laser was used, whereas in the DIAMONDS trial other types of lasers were allowed, given that argon laser is no longer widely available. The technique and parameters used for SL treatment in the DIAMONDS trial is representative of the technique used in other macular laser trials^6,7 and that used in standard clinical practice.

Retreatment

Where necessary, retreatments were carried out with the same laser allocated by randomisation. When retreating, treatment of areas within 300–500 µm from the centre of the fovea was allowed. Details of retreatments were recorded in the patient’s CRF.

Rescue treatment

Where necessary, rescue treatment, with anti-VEGF therapy or steroids, as appropriate based on judgement by the treating ophthalmologist, was allowed in both treatment groups if the CRT increased to ≥ 400 µm at any point during the patient’s follow-up or if a loss of ≥ 10 ETDRS letters occurred in relation to DMO. The type and date of any rescue treatment was recorded in the patient’s CRF.

Participating sites and experience

Sixteen NHS HES sites participated in the recruitment, management and follow-up of DIAMONDS trial participants. All treating ophthalmologists had extensive experience in diabetic retinopathy and DMO as well as in delivering laser treatment for DMO. The participating sites were: Royal Victoria Hospital, Belfast Health & Social Care Trust; Bristol Eye Hospital, University Hospitals Bristol NHS Foundation Trust; Frimley Park Hospital NHS Foundation Trust; Hinchingbrooke Hospital, North West Anglia NHS Trust; King’s College Hospital NHS Foundation Trust; Manchester Royal Eye Hospital, Central Manchester University Hospitals NHS Foundation Trust; National Institute for Health and Care Research (NIHR) Biomedical Research Centre at Moorfields Eye Hospital NHS Foundation Trust; Newcastle Eye Centre, Royal Victoria Infirmary, Newcastle upon Tyne Hospitals NHS Foundation Trust; John Radcliffe Hospital, Oxford University Hospitals NHS Foundation Trust; Sheffield Teaching Hospitals NHS Foundation Trust; Sunderland Eye Infirmary, South Tyneside and Sunderland NHS Foundation Trust; Bradford Royal Infirmary, Bradford Teaching Hospitals NHS Trust; James Cook University Hospital, South Tees Hospitals NHS Foundation Trust; Hull and East Yorkshire Hospital, Hull and East Yorkshire NHS Trust; Stoke Mandeville Hospital, Buckinghamshire NHS Trust; and Hillingdon Hospitals NHS Foundation Trust.

Randomisation

Following informed consent, eligible participants were randomised 1:1 to receive either SML or SL using a minimisation algorithm within the automated randomisation system Sealed Envelope (Sealed Envelope Ltd, London, UK; URL: www.sealedenvelope.com), with the allocation concealed to the ophthalmologist randomising the patient until the patient had joined the trial. The local ophthalmologist used this automated system to ensure post-randomisation masking of allocation to the outcome assessors. Although most patients received their allocated therapy at the baseline visit, it was acceptable for it to be performed at a later visit within two weeks of the baseline visit. If there was a longer interval between the baseline visit and the laser treatment, eligibility was re-confirmed before treatment.

The randomisation system used minimisation to balance allocation of patients across intervention groups for the following prognostic factors: centre, distance BCVA at presentation [≥ 69 ETDRS letters (Snellen equivalent of ≥ 20/40; log-MAR ≥ 0.3); 24–68 ETDRS letters (Snellen equivalent ≤ 20/50; log-MAR 0.4–1.2)] and previous administration of anti-VEGF therapy or macular laser in the study eye. A random element was used in the minimisation to provide a probability of 0.85 for assigning to the treatment group that minimised imbalance.

Masking

The DIAMONDS trial was a pragmatic RCT so that its results would be applicable immediately in an NHS setting once the trial was completed. For this reason, ophthalmologists undertaking laser treatments for DMO at each of the participating centres delivered the treatment for the trial and, thus, were not masked to the laser used. However, participants and outcome assessors (e.g. optometrists measuring visual function; photographers, technicians and nurses obtaining OCT images; and ophthalmic technicians obtaining visual fields) were all masked to treatment allocation. Patients were not informed before, during or after the laser treatment about which laser was used and remained masked until the trial ended. Similarly, investigators obtaining outcome measures had access to the CRF booklet only (and not to the notes of the patients) and this booklet did not contain information about the type of laser the patient had been allocated to or received (this information was held in a locked cabinet/locked room by the treating ophthalmologists).

Data quality

The chief investigator and the NICTU provided training to unit staff on trial processes and procedures including CRF completion and data collection. Monitoring during the trial included adherence to the protocol, trial-specific procedures and good clinical practice. Within the NICTU, the clinical data management process was governed by SOPs, which ensured standardisation and adherence to International Conference of Harmonisation Good Clinical Practice guidelines and regulatory requirements.

Data quality control checks were carried out by a data manager following the NICTU SOP. Data validation was implemented and discrepancy reports were generated following data entry to identify discrepancies such as out-of-range values, inconsistencies or protocol deviations based on data validation checks. Changes to data were recorded and fully auditable. Data errors were documented and corrective actions implemented.

A Data Monitoring and Ethics Committee (DMEC) and a Trial Steering Committee (TSC) were convened for the DIAMONDS trial, the former to carry out reviews of the accumulating data at regular intervals during the study, to ensure the safety of participants and the latter to monitor the progress of the trial, among other tasks.

Anticipated adverse events due to laser treatment

The following were listed in the protocol²² as potential AEs in the DIAMONDS trial:

• foveal burn
• central/paracentral scotomas
• epiretinal membrane formation
• choroidal neovascularisation related to the laser
• self-reported reduced colour vision
• self-reported metamorphopsia.

Pre-trial power calculation

The DIAMONDS trial was powered to demonstrate not only non-inferiority but also equivalence of SML compared with SL with respect to the primary outcome (i.e. mean change in BCVA in the study eye between baseline and month 24). This was because, based on the existing knowledge when the trial was designed, it was possible that no differences could be found in the primary outcome but that differences could exist in other important secondary outcomes, such as PROMs. The DIAMONDS trial was also sufficiently powered to determine superiority of one laser over the other, if this were to exist.

Based on a mean of 0.08 log-MAR [standard deviation (SD) 0.23 log-MAR] for BCVA change from baseline for the standard care laser⁹ and a permitted maximum difference of 0.1 log-MAR (± 5 ETDRS letters) between groups, we estimated that the DIAMONDS trial would require 113 randomised participants per group, at 90% power and 0.05 level of significance. Allowing for up to a 15% dropout rate during the 24 months of follow-up, as observed in other randomised trials on DMO with outcomes determined at 24 months,^28,29 the recruitment target was set at 266 patients.

A permitted maximal difference of 5 ETDRS letters between groups was chosen as the non-inferiority margin (± 5 ETDRS letters for equivalence) because a difference of this size or less is not considered clinically relevant or meaningful to patients.^{2,3,8,24–26}

In addition, 24-month data for 113 participants per group would also be sufficient to detect a mean difference between lasers of 37.7 µm in CRT (based on a SD of 86.8¹⁰ µm) and of 6.55 in NEI-VFQ-25 scores (based on a SD of 15.1 score as per Tranos et al.³⁰). These are important secondary outcomes and such differences in CRT and NEI-VFQ-25 scores have been shown to be clinically relevant.^31,32

Analysis principles

The DIAMONDS trial was designed with sufficient power to detect not only non-inferiority but also equivalence of SML when compared with SL and the primary statistical analysis was per protocol (PP), but an intention-to treat (ITT) analysis was also undertaken. ITT is recommended for superiority trials but, for non-inferiority or equivalence trials, a PP analysis is preferred because ITT increases the risk of a type I error for such trials. The main analyses were as pre-specified in the protocol, but some additional post hoc analyses were also undertaken (these are detailed in Post hoc analyses).

The difference between lasers for change in BCVA [with 95% confidence intervals (CIs)] from baseline to month 24 (primary end point) was compared with the permitted maximum difference of 5 ETDRS letters (0.1 log-MAR). SML would be deemed non-inferior to SL if the lower limit of the 95% CI of the treatment difference was above this non-inferiority margin. If the 95% CI of the treatment difference was wholly within both the upper and lower margins of the permitted maximum difference (± 5 ETDRS letters), then SML would be deemed to be equivalent to SL.

Change in BCVA from baseline to month 24 was compared between the two intervention groups using an independent two-sample t-test with a secondary analysis using an analysis of covariance (ANCOVA) model adjusted for baseline BCVA score, baseline CRT and minimisation factors/covariates comprising centre; distance BCVA at baseline of ≥ 69 ETDRS letters (Snellen equivalent of ≥ 20/40; log-MAR ≥ 0.3) or 24–68 ETDRS letters (Snellen equivalent ≤ 20/50; log-MAR 0.4–1.2); previous use of anti-VEGF therapies in the study eye; and previous use of macular laser in the study eye.

The primary analysis was based on data from the study eye only. When performing a secondary analysis on the subset of participants with both eyes eligible and treated, study eye was included as a random effect within the mixed model. The principal analysis was based on available case data with no imputation of missing values. Intention-to-treat analyses were used for all secondary outcomes because the aim was to assess superiority for these outcomes.

Side effects of laser treatment and use of additional treatments (e.g. steroids or anti-VEGF therapy) were analysed using logistic regression models with adjustment for the minimisation covariates. Analyses of secondary measures of visual function and anatomical outcomes (MD of the 10–2 visual field test, CRT and macular volume) and number of treatments required were undertaken using linear regression models adjusted for baseline BCVA score and minimisation variables. Analysis of ‘driving ability’ (i.e. meeting standards for driving) was undertaken using a logistic regression model adjusted for baseline BCVA and the minimisation variables. The number of AEs, adverse reactions (ARs), SAEs, serious adverse reactions (SARs) and suspected unexpected serious adverse reactions (SUSARs), and the number and percentage of participants experiencing these events are reported. The chi-square test (or Fisher’s exact test if appropriate) and proportion test were used to check if incidences of AEs differed between intervention groups. Relative risks with 95% CIs are reported. Baseline characteristics, follow-up measurements and safety data are presented using appropriate descriptive summary measures depending on the scale of measurement and distribution.

Statistical diagnostic methods were used to check for violations of the model assumptions.

Statistical significance was based on two-sided tests, with p < 0.05 taken as the criterion for statistical significance and with no adjustment for multiple testing.

Sensitivity analyses

Sensitivity analyses were undertaken to assess the impact of missing data by imputing extreme values (lowest and highest) and last observation carried forward; the impact of including patients who were not treatment naive (i.e. excluding those who have had previous laser treatment for DMO in the study eye or previous anti-VEGF therapy for DMO or PDR in the study eye); and the impact of using month-24 data that were collected outside of ± 14 days of the due date.

Subgroup analyses

We conducted pre-specified subgroup analyses of the primary outcome based on clinical rationale. These subgroups were centre; distance BCVA at baseline of ≥ 69 ETDRS letters (Snellen equivalent of ≥ 20/40; log-MAR ≥ 0.3) or 24–68 ETDRS letters (Snellen equivalent ≤ 20/50; log-MAR 0.4–1.2); previous administration of anti-VEGF therapy; and previous use of macular laser in the study eye. These analyses were carried out by including the corresponding interaction term in the regression model and 99% CI.

We also conducted analyses to identify whether or not any group of participants was at high risk of poor outcomes. High-risk participants were defined as participants with a baseline glycated haemoglobin type A1c (HbA_1c) value of ≥ 53 mmol/mol (≥ 7%). These were analysed in exploratory subgroup analyses.

Primary analysis

The primary analysis was performed on the mean change in BCVA in the study eye from baseline to month 24 and comprised the following: mean change in BCVA by intervention group (with SDs); difference in means with 95% CI; p-value from independent two-sample t-test; and secondary analysis using ANCOVA adjusted for baseline BCVA and minimisation variables. The non inferiority (and equivalence) margin was compared against the 95% CI for both PP and ITT analyses.

Secondary analyses

The secondary analyses, and how they were reported, were as follows:

• Mean change in binocular BCVA from baseline to month 24 – mean (SD) by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• mean change in CRT as determined by SD-OCT, from baseline to month 24 – mean (SD) by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• mean change in the MD of the Humphrey 10–2 visual field from baseline to month 24 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• change in the percentage of participants meeting driving standards from baseline to month 24 – number and percentage of participants meeting driving standards by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for baseline BCVA and minimisation variables
• number of participants experiencing side effects from baseline to month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables
• number of laser treatments used in the study eye from baseline to month 24 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• number of participants receiving at least one additional treatment (other than laser) from baseline to month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables.

Additional analyses specified in the statistical analysis but not in the published protocol

Additional analyses specified in the statistical analysis but not in the published protocol, and how they were reported, were as follows:

• mean change in BCVA in the study eye from baseline to month 12 – mean with SD by intervention group, difference in means with 95% CI and p-value from ANCOVA adjusted for baseline BCVA and minimisation variables
• mean change in binocular BCVA from baseline to month 12 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• mean change in CRT, as determined by SD-OCT, from baseline to month 12 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• mean change in the MD of the Humphrey 10–2 visual field from baseline to month 12 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• change in the percentage of people meeting driving standards from baseline to month 12 – number and percentage of participants meeting driving standards by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for baseline BCVA and minimisation variables
• number of participants experiencing side effects from baseline to month 12 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables
• number of laser treatments used in study eye from baseline to month 12 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• number of participants with at least one additional treatment (other than laser) from baseline to month 12 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables
• number of steroid injections as additional treatments from baseline to month 12 and from baseline to month 24 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for minimisation variables
• number of participants with at least one steroid injection from baseline to month 12 and from baseline to month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables
• number of anti-VEGF treatments as additional treatments from baseline to month 12 and from baseline to month 24 – mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for minimisation variables; and as number and percentage of participants by category (≤ 4, 5–10 and > 10)
• number of participants with at least one anti-VEGF treatment as additional treatment from baseline to month 12 and from baseline to month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables
• number of participants receiving rescue treatments from baseline to month 12 and from baseline to month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for minimisation variables
• number of participants satisfying rescue criteria from baseline to month 12 and from baseline to month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for adjusted for baseline BCVA and minimisation variables
• number of participants experiencing a loss of 10 or more ETDRS letters between baseline and month 24 – number and percentage of participants by intervention group, odds ratio with 95% CI and p-value from logistic regression adjusted for baseline BCVA and minimisation variables.

Post hoc analyses

Post hoc analyses, and how they were reported, were as follows:

• mean change in macular volume from baseline to month 24: mean with SD by intervention group, difference in means with 95% CI and p-value from linear regression adjusted for baseline BCVA and minimisation variables
• number of participants experiencing a loss of 5 or more ETDRS letters from baseline to month 24
• number of participants experiencing an increase in BCVA or a BCVA loss not greater than 5 ETDRS letters from the baseline value to month 24
• number of participants satisfying rescue criteria at least once at any time from baseline to month 24
• number of participants satisfying rescue criteria at least once and receiving rescue treatment at any time from baseline to month 24
• number of participants satisfying rescue criteria at least once and not receiving rescue treatment at any time from baseline to month 24
• number of participants who did not satisfy rescue criteria at all from baseline to month 24
• number of participants who did not satisfy rescue criteria at all but received rescue treatment from baseline to month 24.

Overview

The main objective of the health economics evaluation was to conduct a short-term (baseline to 2 years’ follow-up) within-trial analysis comparing the cost-effectiveness of SML with SL in patients with DMO fulfilling the DIAMONDS trial eligibility criteria. To achieve this, a systematic comparison of the costs of resource inputs used by participants in the two treatment groups and the consequences associated with the interventions was conducted. The primary analysis adopted an NHS and Personal Social Services (PSS) perspective. The economic evaluation took the form of a cost–utility analysis, expressed in terms of incremental cost per QALY gained. Costs and outcomes beyond the first year of follow-up were discounted at 3.5% in line with the NICE reference case.³³

For the health economics analysis, we adopted an ITT approach as reported in the health economics analysis plan. ITT requires that study participants are analysed according to their treatment assignment regardless of actual treatment received. This is the approach preferred by NICE for cost-effectiveness analyses as stated in their methods guide.³³ We report the results for the per-protocol analysis in a sensitivity analysis.

Measuring resource use and costs

Data were collected on resource use and costs associated with delivery of laser treatment (direct intervention costs) and broader health service resource use during the 24 months of follow-up. All costs were expressed in Great British pounds and valued in 2019–20 prices. Where appropriate, costs were inflated to 2019–20 prices using the NHS Cost Inflation Index.³⁴

Direct intervention costs

Direct intervention costs were costs associated with the delivery of laser treatment. These included staff costs and equipment costs (i.e. costs of laser machines) (see Appendix 1, Table 22 and Table 23). Unit costs for staff were obtained from the Personal Social Services Research Unit (PSSRU) Unit Costs of Health and Social Care 2020 compendium³⁴ and were multiplied by the time it took to perform specified procedures for each participant. The time it took to undertake various procedures was recorded on the DIAMONDS trial CRFs and comprised: (1) the time it took imaging technicians to obtain FFA and SD-OCT scans to guide laser treatment and (2) the time taken by consultants and associate specialists to perform the laser procedure, as well as the time invested in counselling the patient. Costs of laser machines were obtained directly from manufacturers. An annual equivalent cost of equipment was obtained by annuitising the capital costs of the item over its useful life span, applying a discount rate of 3.5% per annum. We derived a per-patient cost of equipment (including annual maintenance costs) by assuming that the machine will be used to perform laser procedures on 3000 patients per year (Professor Noemi Lois, Queen’s University Belfast, 2021, personal communication).

Measuring and valuing resource use

Resource use data were captured on trial CRFs at scheduled clinic visits over the 24-month follow-up period, at 4, 8, 12, 16, 20 and 24 months. On-site monitoring visits during the trial ensured the accuracy of entries in the CRF. The CRFs captured details related to the eye condition of inpatient and day case admissions, outpatient attendances, other tests or investigations, medication use including anti-VEGF therapy/steroids or other rescue treatments, and laser retreatments.

Resource inputs were valued by attaching unit costs derived from national compendia in accordance with NICE’s Guide to the Methods of Technology Appraisal.³³ The key databases for deriving unit cost data included the Department of Health and Social Care’s Reference Costs 2018–19 schedules,³⁵ the PSSRU’s Unit Costs of Health and Social Care 2019 compendium³⁴ and volume 80 of the British National Formulary.³⁶ Appendix 1, Table 22 gives a summary of the unit cost values and data sources for broader resource use categories identified within the follow-up questionnaires.

The following assumptions were made in costing outpatient attendances. Where an outpatient attendance was reported but no procedure undertaken, the average unit cost of an outpatient ophthalmology visit was used; this varied between £80 and £101 per consultation depending on whether the consultation was ‘non-consultant’ or ‘consultant-led’. We used data captured on trial CRFs documenting the grade of professional that attended to the patient to select the most appropriate unit costs. For example, if a consultant attended to the patient, then the consultant-led unit cost was applied. Where a procedure(s) was undertaken as part of the visit, the relevant Healthcare Resource Group (HRG) code was derived using the HRG4 + Reference Costs Grouper Software (NHS Digital, Leeds, UK). High-cost drugs (specifically anti-VEGF therapy) were separately costed as these are considered an unbundled HRG. We also costed steroid rescue treatment. Costing of laser retreatments followed the same approach as costing for the index (first-session) laser procedure.

Summary statistics were generated for resource use variables by treatment allocation and assessment point. Mean resource use and cost values were compared between groups using two-sample t-tests. Differences between groups, along with 95% CIs, were estimated using non-parametric bootstrap estimates (10,000 replications).

Health outcomes

The primary outcome of the within-trial economic evaluation is the QALY, as recommended in the NICE reference case.³³ The QALY is a measure that combines quantity and quality of life lived into a single metric, with one QALY equating to 1 year of full health. QALY estimates are generated from combining length (survival) and health-related quality of life (HRQoL) data from participants for the period covering the trial time horizon through an area under the curve (AUC) approach using a linear extrapolation.³⁷ Since AUC estimates are predicted to correlate with baseline scores (and thus potential baseline imbalances), AUC estimates were adjusted for baseline scores within regression analyses. HRQoL was converted into health-state utilities indexed at 0 and 1, where 0 represents death and 1 represents full health. Patients who died during the study were subsequently scored 0 at later scheduled follow-up visits for both cost and HRQoL scores and were included as observed data.

To calculate QALYs, it is imperative to obtain health state values for participants within the trial. The HRQoL of trial participants was assessed at baseline and at 12 and 24 months’ post randomisation using the EQ-5D-5L instrument. The EQ-5D-5L consists of the descriptive system and the visual analogue scale. The descriptive system includes five questions addressing mobility, self-care, usual activities, pain/discomfort and anxiety/depression, with each dimension assessed at five levels from no problems to extreme problems. Responses to the EQ-5D-5L instrument were converted into health utility scores using the EQ-5D-5L Crosswalk Index Value Calculator currently recommended by NICE, which maps the EQ-5D-5L descriptive system data onto the EuroQol-5 Dimensions, three-level version (EQ-5D-3L) valuation set.³⁸ A detailed description of the mapping methodology is provided in the study by van Hout et al.³⁸

Health-related quality of life was also assessed by determining the vision-related quality of life using two vision specific measures: the NEI-VFQ-25 and the VisQoL.^39–41 The NEI-VFQ-25 is a vision-specific patient-reported quality of life tool. This validated questionnaire has been used widely to evaluate visual outcomes in patients with eye diseases, including diabetic retinopathy. In addition to eliciting information about general health and vision, it specifically addresses difficulty with near vision, distance vision, driving and the effect of light conditions on vision, providing a comprehensive evaluation of vision-related quality of life. The NEI-VFQ-25 scoring is done in two-stages: (1) each item is scored on a scale of 0 (lowest) to 100 (highest), where a higher score represents better functioning; and (2) items within each subscale are averaged together (11 subscales in total for the NEI-VFQ-25). To obtain the combined score for the questionnaire, the average of the subscales (excluding the general health rating question) is undertaken. Averaging across the subscales scores, rather than individual items, gives equal weight to each subscale.

The VisQol questionnaire is shorter than the NEI-VFQ-25 with only six attributes: physical well-being, independence, social well-being, self-actualisation, planning and organisation. However, it has not been widely validated. The utilities for VisQoL were developed using a time trade-off exercise in people who were visually impaired including patients with age-related macular degeneration, diabetic retinopathy and glaucoma.⁴¹

The health utility values and QALYs accrued over the 24-month follow-up period were summarised by treatment group and assessment point and presented as means and associated standard errors (SEs). Between-group differences were compared using the two-sample t-test, in a similar way to the descriptive analyses of resource inputs and costs.

Handling of missing data

Multiple imputation by chained equations was used to predict missing health status (utility) scores and costs based on the assumption that data were missing at random (MAR). The MAR assumption was tested through a series of logistic regression analyses comparing participants’ characteristics for those with and without missing end-point data. Imputation was achieved using predictive mean matching, which has the advantage of preserving non–linear relationships and correlations between variables within the data.⁴² Twenty imputed data sets were generated and used to inform the base-case analyses. Parameter estimates were pooled across the 20 imputed data sets using Rubin’s rules to account for between- and within-imputation components of variance terms associated with parameter estimates.⁴³ Imputed and observed values were compared to establish that imputation did not introduce bias into subsequent estimation.

Cost-effectiveness analysis

The base-case cost-effectiveness analysis was performed using an ITT approach. Mean incremental costs and QALYs were estimated using seemingly unrelated regression methods that account for the correlation between costs and outcomes. The seemingly unrelated regression adjusted for the following covariates: baseline utilities, baseline body mass index (BMI), baseline BCVA, patient-reported previous use of anti-VEGF therapy at baseline and previous use of macular laser. The joint distributions of costs and outcomes were generated using non-parametric bootstrap methods and used to populate the cost-effectiveness plane. The incremental cost-effectiveness ratio (ICER) for SML compared with SL was calculated by dividing the between-group difference in adjusted mean total costs by the between-group difference in adjusted mean QALYs. Mean ICER values were compared against cost-effectiveness threshold values ranging between £15,000 and £30,000 per QALY gained, in line with NICE guidance.³³ The cost-effectiveness thresholds provide an indication of society’s willingness to pay for an additional QALY; lower ICER values than the threshold could be considered cost-effective for use in the UK NHS. The incremental net monetary benefit (NMB) of switching from SL to SML was calculated for cost-effectiveness thresholds ranging from £15,000 to £30,000 per QALY gained. NMB is calculated as the net benefit of an intervention (expressed in monetary terms) considering all costs associated with the intervention. A positive incremental NMB, suggests that, on average, SML is cost-effective compared with SL at the given cost-effectiveness threshold. In that case, the cost to derive the benefit is less than the maximum amount that the decision-maker would be willing to pay for this benefit.⁴⁴

Trial Management Group

The Trial Management Group (TMG) was chaired by the chief investigator and included representation from the NICTU and other investigators or collaborators involved in the study. The TMG had responsibility for the day-to-day operational management of the DIAMONDS trial and met monthly throughout the trial to discuss and monitor its progress.

Trial Steering Committee

The conduct of DIAMONDS was overseen by a TSC, on behalf of the sponsor and funder. The TSC comprised an independent chair, additional independent members (including a patient representative) and members of the trial team (including the chief investigator). TSC meetings were arranged so that they coincided with meetings of the DMEC, to allow members in both committees to discuss issues and recommendations raised by the DMEC.

Data Monitoring and Ethics Committee

The DMEC was responsible for safeguarding the interests of participants in the DIAMONDS trial. The DMEC monitored the main outcome measures including safety and efficacy. The DMEC comprised two clinicians and a statistician who were independent of the trial.

DIAMONDS patient and public involvement group

At the very early stages of the DIAMONDS trial conception, a DIAMONDS patient and public involvement (PPI) group was established. The DIAMONDS PPI group contributed to the trial design, including the selection of outcomes, preparation of patient-related materials for the trial, recruitment strategies, interpretation of trial results and preparation of the Plain English summary.