Discussion

Statement of principal findings

The DIAMONDS trial was a pragmatic, allocation-concealed, multicentre, randomised, non-inferiority clinical trial (but powered to detect equivalence and superiority, if it were to exist) in which outcome assessors and participants were masked to treatment allocation. The DIAMONDS trial recruited in full, for a total of 266 participants with 133 participants allocated to each laser group; one participant in the SL group withdrew consent for their data to be used and thus 265 participants were included in the analysis. Despite the COVID-19 pandemic, which occurred during the second year of follow-up for most participants, primary outcome data (measured at 24 months from trial initiation) was available in the majority of participants: 116 (87%) and 115 (86%) in the SML group and SL group, respectively. Participants were predominantly white males with uncontrolled (mean HbA_1c of 8.5%) type 2 diabetes and were mostly overweight, obese, or morbidly obese. Participants had a mean age of 62 years, a mean duration of DMO diagnosis of 2.5 years, a mean BCVA of 80 letters (6/7.5 Snellen Equivalent) and a mean CRT of 329.2 µm on SD-OCT.

Best-corrected visual acuity was selected as the primary outcome in the DIAMONDS trial. Central vision is very important to people with diabetes as it is used in many tasks including recognising faces and reading (many people with diabetes need to read medication labels and instructions specifying, for example, the amount of insulin needed for their treatment). The non-inferiority margin of 5 ETDRS letters (equivalence margin of± 5 ETDRS letters) was chosen as changes within this range are not considered to be clinically relevant and could be the result of test/re-test variability. The DIAMONDS trial found SML to be equivalent to SL, with an adjusted mean change in BCVA in the study eye from baseline to month 24 of –2.36 ETDRS letters (SE 0.67 ETDRS letters) in the SML group compared with –0.53 ETDRS letters (SE 0.67 ETDRS letters) in the SL group (mean difference –1.84 ETDRS letters, 95% CI –3.72 to 0.05 ETDRS letters; p = 0.056).

There were no statistically significant differences in the pre-defined secondary outcomes with the exception of the number of laser treatments performed, which was slightly higher (less than one further session) in the SML group. This finding appeared to be driven, however, by a small number of participants (n = 13) that required a larger number of laser treatments in the SML group. Most participants (approximately 80% and 90% of participants in SML group and SL group, respectively) required 1–3 laser sessions throughout the 2-year period (see Appendix 1, Table 30).

A similar number of participants in each laser treatment group (approximately one-third) met eligibility criteria to receive rescue treatment at any time during the 2-year period of the study. Fewer participants, however, actually received rescue treatment [24 (18%) in the SML group and 28 (21%) in the SL group; odds ratio 0.78, 95% CI 0.42 to 1.45; p = 0.44 and percentage point difference –2.8, 95% CI –13.1 to 7.5; p = 0.59]; all received anti-VEGF therapy and one, in the SL group, received intravitreal steroids in addition. Most participants maintained good vision throughout the trial, with only a small proportion [25 (9%) participants: 17 (13%) in the SML group and 8 (6%) in the SL group] experienced a drop of 10 ETDRS letters (which would be considered a clinically relevant change) from baseline to month 24 (see Appendix 1, Tables 31–33). Similarly, most participants (over 95% for each type of laser) met driving standards at the 24-month trial visit. Meeting driving standards was identified at the time of trial conception by the DIAMONDS PPI group and in consultation with patients with diabetic retinopathy and DMO to be very important to people with diabetes; as a result this was one of the secondary outcomes investigated. The DIAMONDS PPI group contributed not only to the design of the trial, elaboration of participant-related materials and planning of recruitment strategies, but also to the interpretation and dissemination of findings.

Most participants maintained good HRQoL and vision-related quality of life throughout the trial period. The total cost of the treatment, including the laser (first session and subsequent laser sessions), any additional treatments required and follow-up for the 2 years was £897.83 and £1125.66 for the SML group and SL group, respectively.

The DIAMONDS trial was designed as a pragmatic trial.49 On its conception, we followed the PRECIS-2 (PRagmatic Explanatory Continuum Indicator Summary 2) tool50 to ensure as much as possible that trial results would be generalisable and reproducible in clinical practice if implemented in the NHS.

Published randomised trials and systematic reviews comparing subthreshold micropulse laser with standard laser

We carried out a literature review at the time the DIAMONDS trial was designed. In addition, we ran auto-alerts during the duration of the trial, up to January 2021, to identify studies that could be pertinent to the DIAMONDS trial. We then extended the searches up to December 2021. We found six systematic reviews comparing SML with SL published since 2016: Chen et al.,16 Qiao et al.,17 Wu et al.,18 Jorge et al.,19 Scholz et al.51 and Blindbaek et al. 2018.52 These were quality assessed using the National Institutes of Health (NIH) checklist (see Appendix 1, Table 34).53 Based on this, four of these reviews (Chen et al.,16 Jorge et al.,19 Qiao et al.17 and Wu et al.18) were judged to be of good or high quality.

Chen et al.16 found statistically significantly better visual acuity (0.1 log-MAR) with SML laser treatment at 12 months follow-up only,16 but this was due to one trial by Lavinsky et al.9 and the high-density arm of that trial. The review concluded that SML laser is an effective therapy for treating DMO, but that compared with SL, the changes were of limited clinical significance.

Qiao et al.17 included seven RCTs; six of these seven trials were included in the review by Chen et al.16 mentioned above. Meta-analysis found no statistically significant differences in BCVA or central macular thickness at any time point. The results of meta-analyses for contrast sensitivity were mixed, with some trials favouring SL and others favouring SML.

The results in the Qiao et al.17 review differ somewhat from those of Chen et al.16 Qiao et al.17 reported mean BCVA at 12 months with three trials, Figueira et al.,11 Lavinsky et al.9 and Vujosevic et al.,10 whereas Chen et al. reported change from baseline in only two trials, Lavinsky et al.9 and Vujosevic et al.10 The Chen et al. review16 used only the high-density arm of Lavinsky et al.,9 whereas Qiao et al.17 used only the normal density arm of this trial. Their different conclusions between these two reviews appear to be due to the results of Lavinsky et al.,9 deemed to give a significant result in Chen et al.16 (based on change in BCVA from baseline) but not in Qiao et al.17 (based on mean BCVA at 12 months).

The review by Wu et al.18 included a Bayesian network meta-analysis of RCTs and quasi-randomised trials comparing any two treatments of interest, including SML or SL photocoagulation as monotherapy or SL combined with anti-VEGF therapy. The review included 18 studies: 15 reported BCVA and 16 reported CRT and were included in the network meta-analysis, and 11 studies compared SL alone with laser plus ranibizumab. Of seven trials comparing SL with SML, four (Casson et al.,54 Laursen et al.,14 Venkatesh et al.12 and Xie et al.15) had only six months’ follow-up. The remaining three were those by Figueira et al.,11 Lavinsky et al.9 and Vujosevic et al.10 Wu et al.18 found no statistically significant difference in visual acuity improvement between monotherapy laser photocoagulation with SL and SML.18

Scholz et al.51 reviewed the effects of SML for the treatment of macular disorders including central serous chorioretinopathy, DMO and retinal vein occlusion. Although a comprehensive search was undertaken, the review did not specify eligibility criteria or any other review methods and, thus, was subsequently rated as being of low quality. Eleven prospective studies in DMO were included, although the review stated that a larger number of studies were identified. The review calculated change scores for CRT and BCVA but the methods of these calculations were not reported. The conclusions made by the authors, who stated that included studies demonstrated efficacy of SML, should, therefore, be taken with caution.

A review by Blindbaek et al.52 of the potential role of focal/grid laser photocoagulation in DMO treatment was not rated as high quality. They noted that few high-quality studies have compared SML with SL therapy and that results were mixed. Quality of life was not discussed.

One of the problems with all available meta-analyses was that only three trials were analysed, each with a follow-up period of 12 months; none had a longer follow-up period. The mean CRTs in the trials by Lavinsky et al.9 and Vujosevic et al.10 were approximately 370 µm, so many patients would have had a CRT of > 400 µm, unlike those participants included in the DIAMONDS trial. None of the reviews reported effects on quality of life or costs.

Fazel et al.20 conducted a single centre trial that randomised 68 eyes of 68 patients to SML or SL. Inclusion criteria included a CRT of 300–449 µm. With SL there was no statistically significant change in BCVA whereas a significant change was observed with SML, with an improvement of 0.07 Log-MAR by 4 months. CRT was reduced by 13 µm following SML and by 5 µm following SL. The small number of patients included in each treatment group and the short follow-up period (maximum 4 months) limits greatly the usefulness of the results presented. There was no power calculation for the trial shown. Owing to its year of publication, this trial was not included in the reviews by Chen et al.,16 Qiao et al.17 and Wu et al.18

A review by Cooper et al.55 aimed to synthesise the evidence on the psychological, social and everyday visual impact of diabetic retinopathy and DMO from the patient perspective. Eighty-five studies with a range of study designs were included; 23 of these studies assessed quality of life. However, the studies appeared to focus on diabetic retinopathy; quality of life in DMO was not discussed. Visual functioning and psychological wellbeing measures were reported in other studies, including some in DMO; however, limited numerical data were reported and no data on DMO were summarised.

Assessment for meta-analysis

We considered adding the DIAMONDS 12-month results into a meta-analysis with the three previous RCTs with a 12-month follow-up. However, there were important differences in baseline characteristics such as CRT, previous treatments and BCVA, as well as in laser regimens used (Tables 20 and 21). The trial by Lavinsky et al.9 used two SML regimens, standard and high dose, and only the high-dose regimen showed any advantage.9 In the DIAMONDS trial the response to laser did not seem to be affected by previous treatments, so that may not prevent combining the studies in a meta-analysis. However, the high baseline mean CRTs in the trials by Lavinsky et al.9 and Vujosevic et al.10 suggest that these trials included many participants with CRTs well over 400 µm, which would be an impediment to amalgamating them in a meta-analysis with the DIAMONDS data. The baseline BCVA in Lavinsky et al.9 was much lower than that of other studies, so there would be potentially more scope for participants in this trial to gain vision. None of the previous RCTs comparing SL with SML used the 577 nm SML that was used in the DIAMONDS trial. It may be that all SML may have similar effects independently of the wavelength used, but we do not yet know if this is the case. For example, a study undertaken by Vujosevic et al.56 randomised 53 eyes from 53 patients to 810 nm infrared SML or 577 nm yellow SML and found no difference in effectiveness or safety.

We therefore concluded that owing to the heterogeneity among the trials, a meta-analysis was not appropriate.

Other recent studies

One retrospective study by Mansouri et al.57 compared the effects of SML in patients with DMO and a CRT of ≤ 400 µm with those with a CRT of > 400 µm. The study included 63 eyes from 58 patients. Patients could have repeated laser treatment if DMO had not improved at 3 months, and they could also have rescue anti-VEGF therapy with bevacizumab if, by 6 months, DMO remained. The group with a CRT of ≤ 400 µm achieved an average reduction of 55 µm in CRT and a 0.2 log-MAR gain in visual acuity at 12 months and none required rescue anti-VEGF therapy with bevacizumab. The group with a CRT of > 400 µm did not improve, despite 19 of the 30 patients receiving repeated laser treatment, and all required rescue anti-VEGF treatment with bevacizumab between 6 and 12 months.

Another retrospective study by Kim et al.,58 using ‘selective retina therapy’ with the Q-switched neodymium-doped yttrium lithium fluoride laser system, also found that eyes with a CRT of < 400 µm (n = 35) experienced a statistically significant improvement in BCVA and reduction in CRT at 6 months following treatment, whereas no statistically significant change in either of these outcomes was observed in eyes with a CRT of ≥ 400 µm (n = 37). These studies provide further support for the decision made by NICE1,2,4 to advise treatment with macular laser for patients with DMO where the CRT is < 400 µm.

A prospective study by Citirik59 included 80 eyes from 40 patients who had previously undergone ranibizumab treatment for DMO and, subsequently, had recurrent DMO and received SML treatment. Patients were grouped as having a CRT between 251 µm and 300 µm (Group 1; n = 20 eyes), between 301 µm and 400 µm (Group 2; n = 18 eyes) and > 400 µm (Group 3). There was also a group of 20 eyes with a CRT of 250–300 µm in patients who did not accept additional treatment (i.e. laser) and were used as controls (Group 4). At 2 months’ follow-up, eyes in Group 1 experienced a 37.43 µm reduction in CRT (p = 0.03) whereas eyes in groups 2, 3, and 4 experienced no statistically significant change (p = 0.47, p = 0.58 and p = 0.32, respectively). Similarly, the mean BCVA in Group 1 increased from Log-MAR 0.52 to Log-MAR 0.38 (p = 0.01), whereas there was no improvement in BCVA in groups 2, 3, and 4 (p = 0.74, p = 0.88 and p = 0.46, respectively). The results of this study, however, are limited by the very small number of participants in each group and the very short follow-up period and, thus, should be interpreted cautiously.

Adverse events

If SML has a similar effectiveness to SL, as most trials including the DIAMONDS trial suggest, the next question is whether or not it has significantly fewer adverse effects, and whether or not these are clinically meaningful.

In the DIAMONDS trial both lasers proved to be very safe, with only a very small number of participants (the highest proportion of participants for any of the following events being ≈ 2%) experiencing AEs potentially related to the laser treatment, including central/paracentral scotomas, epiretinal membrane, choroidal neovascularisation, and self-reported reduced colour vision and metamorphopsia. No statistically significant differences in the occurrence of AEs were found between laser groups. These potential AEs owing to laser treatment were identified a priori before the trial commenced; patients were questioned at each of the follow-up visits about their occurrence and ophthalmologists also evaluated participants to determine whether or not any of these had happened. None of these reported AEs, however, was regarded by the investigators as being definitively caused by the laser treatment.

In the RCT by Vujosevic et al.,10 it was found that contrast sensitivity measured by microperimetry increased after SML, but decreased after SL. Qiao et al.17 included in their meta-analysis three trials (Kumar et al.13 and Venkatesh et al.12 at 6 months, and Figueira et al.11 at 3 months) that had included contrast sensitivity and found no difference overall between SML and SL. There was considerable heterogeneity among trials.

Chhablani et al.60 randomised 30 eyes from 20 patients with non-centre-involving DMO to two SML strategies (5% and 15% duty cycle) or to SL. Patients had microperimetry, CRT and BCVA measured at baseline and at 6 and 12 weeks post treatment. Retinal sensitivity increased in the SML group but decreased in the SL group.

The majority of previously published studies9,10,13 reported a lack of visible retinal changes after SML, whereas laser scars are most often present after SL. Figueira et al.11 reported scars in 59% of eyes after SL versus 14% after SML after a follow-up of 12 months.

None of the above studies assessed patient-reported outcomes, with the exception of the DIAMONDS trial. The DIAMONDS trial showed no differences in health-related or vision-related quality of life, suggesting that differences in retinal sensitivity and the presence of retinal scars may not be perceived by and be relevant to patients or, alternatively, that the quality of life tools used are unable to detect them, if they were to exist.

Clinical implications

For many years, macular laser was the mainstay therapy to treat DMO, with the ETDRS demonstrating in 1985 that it reduced the risk of visual loss (loss of ≥ 3 ETDRS lines = loss of 15 ETDRS letters) by 50% at 3 years in people with CSMO.5 As only a small (≈ 3%) proportion of participants in the ETDRS experienced an improvement in vision of 15 ETDRS letters, it has been usually stated that macular laser does not improve vision. The great majority of eyes in the ETDRS (1903/2243; 85%) had excellent vision, of 20/40 or better at baseline,5 which may explain the limited improvement in vision observed in this large trial. Indeed, a Diabetic Retinopathy Clinical Research Network trial showed that 32% and 44% of patients with centre-involving DMO experienced a 10-letter visual acuity improvement at 2 and 3 years, respectively, following macular laser.6,7 Thus, macular laser can improve vision in people with centre-involving DMO. However, with the advent of anti-VEGF therapies the benefits of macular laser have been ignored by many, as shown in the published European Society of Retina Specialists (EURETINA) guidelines.61 The DIAMONDS trial shows that for patients with DMO fulfilling the DIAMONDS inclusion criteria, and as advised by NICE based on best available evidence, macular laser has still an important place in the management of people with centre-involving DMO.2,3 In addition, as highlighted by Bakri et al. in the American Society of Retinal Specialists guidelines,62 macular laser also remains the treatment of choice in patients with non-centre-involving DMO. Furthermore, it has been shown in a publicly funded RCT that 41–64% of people receiving anti-VEGF therapies (bevacizumab, ranibizumab or aflibercept) required macular laser to control DMO during the 2-year period following initiation of therapy.8 Thus, macular laser is still needed even in people treated with anti-VEGF therapy.

A retrospective study by Lai et al.63 which included 164 eyes from 164 patients with centre-involving DMO and a CRT of > 300 µm treated with either SML (86 eyes) or intravitreal aflibercept monotherapy (78 eyes) found that although at the 6-month follow-up a higher percentage of eyes in the aflibercept group experienced at least a 5-letter visual acuity improvement when compared with those in the SML group (56% vs. 38%; p = 0.044), this was no longer the case at 12 months (45% vs. 49%; p = 0.584) or at 24 months (49% vs. 57%; p = 0.227). Similarly, although at the 6-month visit the aflibercept group had a higher percentage of eyes with an improvement of at least 10% in CRT than the SML group (73% vs. 49%; p = 0.005), this was no longer the case at the 12-month (73% vs. 70%; p = 0.995) or 24-month visits (85% vs. 84%; p = 0.872).

A publicly funded RCT by the Diabetic Retinopathy Clinical Research Network compared observation, standard macular laser and anti-VEGF therapy (aflibercept) in patients with centre-involving DMO with good vision, with the primary outcome being a loss of ≥ 5 ETDRS letters from baseline to 2 years.64 A total of 625 of the 702 participants (89%) completed the 2-year visit. Participants had normal vision [mean ETDRS letter score 85 ETDRS letters (6/6)], and had better glycaemic control (median HbA_1c 7.6) and less severe DMO [mean CRT ≈ 300 µm (306  µm, 314  µm and 314 µm in aflibercept, laser and observation groups, respectively)] than those included in DIAMONDS. The percentage of eyes with a loss of ≥ 5-ETDRS letters at 2 years was 16% in the aflibercept arm, 17% in the macular laser arm and 19% in the observation arm, with no statistically significant differences among groups. Similarly, there were no statistically significant differences in change in CRT from baseline to 2 years among treatment groups. The median number of aflibercept injections over 2 years in the aflibercept group was eight (interquartile range 6–11) and macular laser was additionally required in 6% of eyes in this group. A small proportion of participants (9%, 7% and 7% in the aflibercept, macular laser and observation arms, respectively) experienced a loss of 10 or more ETDRS letters from baseline to month 24. The authors concluded that observation without treatment, unless visual acuity worsens, was a reasonable strategy for eyes with centre-involving DMO. Given that DIAMONDS trial participants had poor diabetes control (mean HbA_1c 8.5%), reduced vision [mean ETDRS score of 80 letters (6/7.5 Snellen equivalent)] and more severe DMO (mean CRT 329.2 µm), observation is unlikely to have been the right approach for the management of their DMO.

The DIAMONDS trial showed that SML had comparable efficacy and cost to SL, suggesting that either treatment could be equally offered to patients with centre-involving DMO with a CRT of < 400 µm that was suitable for macular laser. Given that SML has been shown to preserve photoreceptor cells (visual cells),21,56,65 RPE and neurosensory retina,10,21 and produce no burn or objective damage, it should be easier and safer to teach to, for example, junior ophthalmologists and non-medical staff, as a foveal burn would be avoided. The possibility of having trained non-medical staff contributing to the management of people living with complications of diabetic retinopathy could help the NHS to cope with the high and ever-increasing demand for care in HES. Specialist nurses and trained hospital optometrists are already providing intravitreal injections for people with DMO in the NHS.66 A new pathway involving allied health care professionals has been recently shown to have good sensitivity and acceptable specificity, as well as to be cost saving, for the surveillance of people with previously treated complications of diabetic retinopathy, including DMO.47,67–69 Thus, the possibility of expanding the role of allied non-medical staff to deliver macular laser treatment using SML seems to be reasonable and worth pursuing.

Strengths and limitations

The DIAMONDS trial was a robustly designed and appropriately powered multicentre RCT. It was powered to detect not only non-inferiority of SML when compared with SL, but also equivalence and superiority if these were to exist. It was powered to detect differences not only in the primary outcome but also in important secondary outcomes (CRT and vision-related quality of life). It was estimated at the trial conception that 113 participants in each laser arm were required to have completed primary outcome data at month 24 for the trial to answer the research questions formulated; at trial completion (24-month follow-up) a higher number of participants (116 and 115 in the SML and SL groups, respectively) were available. The trial was designed as a pragmatic trial so that its results could be reproduced in routine clinical practice. Unlike many RCTs evaluating treatments for DMO, in which the primary outcome was measured at 1 year, the DIAMONDS trial set the primary outcome at 2 years, as it is possible that benefits of treatments may wane overtime or, as shown in some laser trials, improve over time. Similarly, unlike many RCTs evaluating treatments for DMO, the DIAMONDS trial included patient-reported outcomes and, importantly, incorporated a clinical outcome that was suggested by people with diabetes and DMO, namely meeting driving standards. This highlights the importance of PPI in the design of research studies, including RCTs. Furthermore, and unlike most trials on treatments for DMO, it incorporated a prospective within-trial health economic evaluation to compare the cost-effectiveness of the alternative treatments investigated. Limitations include the fact that the great majority of participants enrolled had poorly controlled diabetes and, thus, it is possible that better outcomes in both treatment groups could be achieved in people with adequately controlled diabetes.

Implications for health care

The results of the DIAMONDS trial show that ≈ 80% of patients with centre-involving DMO and a CRT of < 400 µm can be managed successfully with macular laser alone and that either SML or SL can be used given that both laser treatments are equivalent. Only ≈ 20% of people in the DIAMONDS trial required additional anti-VEGF treatment. Despite this, one European group61 has recommended anti-VEGF therapy over laser treatment for people with DMO. However, because of the cost of these drugs and the frequency of patient visits required when anti-VEGF therapies are used, implementation of this anti-VEGF strategy for all patients with DMO would be much more expensive for, and would increase demand for care within, health-care systems. Moreover, it is likely that macular laser may be more convenient for patients, given that it requires fewer visits to the clinic for monitoring/repeating treatment when needed and is less painful.

Recommendations for research

The DIAMONDS trial showed that laser treatment remains an effective treatment strategy for patients with centre-involving DMO with a CRT of < 400 µm as determined by SD-OCT. A study determining which patient’s baseline characteristics, besides CRT, are associated with a higher likelihood of an adequate functional (visual acuity) and anatomical (clearance of DMO) treatment response to macular laser would be recommended, so that a more individualised treatment strategy could be offered to patients (i.e. those less likely to respond to macular laser could be offered alternative therapies).

A RCT comparing anti-VEGF monotherapy with anti-VEGF plus laser therapy once the CRT has decreased to less than 400 µm following anti-VEGF treatment would be of interest, as it may reduce the number of anti-VEGF injections required and may improve functional and anatomical outcomes more effectively than either treatment alone. Such a study has not been undertaken. In RCTs in which macular laser was combined with anti-VEGF therapy, it was not stipulated in the protocol that macular laser should be performed only once CRT had decreased to < 400 µm following anti-VEGF injections.

A study evaluating the feasibility of allied non-medical staff delivering SML to patients with DMO would be of benefit as this strategy, if safe and acceptable to patients, could increase capacity in the NHS.

Patient and public involvement

The DIAMONDS PPI group had a substantial input in the trial. As stated in Chapter 2, DIAMONDS patient and public involvement group, the PPI group contributed to the trial design, including the selection of outcomes, preparation of patient related materials for the trial and recruitment strategies. One outcome (meeting driving standards) came directly from this group (i.e. it had not been anticipated by the trial investigators) as it was identified as being a very important outcome for people living with diabetes and DMO. The PPI group participated in the interpretation of trial results and voiced the importance of the overall finding of the trial (i.e. the fact that SML, which does not cause any detectable damage in the retina, is as effective as SL in the treatment of people with centre-involving DMO suitable for macular laser and with a CRT of < 400 µm) and the subsequent importance of future implementation of trial data into clinical practice. The PPI group is currently taking part in the dissemination of the trial results.

Conclusions

Subthreshold micropulse laser was found to be equivalent to SL in clinical benefits, safety and associated costs and, thus, either laser can be used for the treatment of people with centre-involving DMO suitable for macular laser and with a CRT of < 400 µm.