Synergising Macular Function and Structure in Diabetic Macular Oedema to Improve Patient Care: An International Delphi Consensus

Abstract

Introduction

The correlation between diabetic macular oedema (DMO) structural and functional assessments and their reflection of patient impact lacks consensus. This study aims to understand the ophthalmology community’s thoughts on optimising patient care by improving functional and structural assessments.

Methods

The process employed a modified Delphi methodology. Following a literature review, a steering committee of six experts developed 40 statements under three consensus topics. The statements were globally distributed via an online, four-point Likert scale survey using a snowball sampling technique.

Results

Survey responses (n = 93) were collated anonymously and independently analysed. Consensus was achieved in 35 statements (88%). There was strong agreement that routinely used assessments for DMO may not accurately reflect the level of vision impairment experienced by the patient and that there is a need for a composite functional assessment (of two or more tests) to demonstrate better correlation with structural changes. The correlation of anatomical features like central subfield thickness, epiretinal membrane, intraretinal fluid, and subretinal fluid volumes with visual acuity outcomes remain uncertain.

Conclusion

This consensus supports the need to develop a composite assessment to improve patient care. Utilising contrast sensitivity, distance and near low luminance, and novel patient-reported outcome measures may improve patient-centric assessments to better understand the patient’s quality of life and functional experience.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40123-026-01336-x.

Key Summary Points

Why carry out this study?

This study aimed to reach expert agreement on how current structural and functional assessments should be used in diagnosing and managing diabetic macular oedema (DMO), including whether best-corrected visual acuity (BCVA) alone adequately represents a patient's visual function and which outcome measures best indicate disease control.

A modified Delphi approach was used, as it is a recognised method for generating expert consensus in healthcare.

What was learned from the study?

The consensus highlights the need for a composite assessment to enhance patient care.

Incorporating contrast sensitivity, low-luminance distance and near measures, and new patient-reported outcomes may provide a more patient-centred understanding of quality of life and functional vision.

Introduction

For many decades, best-corrected visual acuity (BCVA) has served as the main functional outcome and primary endpoint in clinical trials evaluating therapies for diabetic macular oedema (DMO) [1]. Current international guidelines recommend a minimum examination consisting of BCVA testing and retinal examination sufficient to classify diabetic retinopathy (DR) to facilitate timely prevention and treatment of sight-threatening conditions [2].

When initiating treatment for centre-involving DMO, a reduction in BCVA and/or increase in central subfield thickness (CST) and the presence of subretinal or intraretinal fluid are commonly seen as signs of disease activity and used to monitor recurrence [3]. Most clinicians admit to treatment decisions and interval extensions being driven primarily by anatomical features on ocular coherence tomography (OCT). However, there is overwhelming evidence that BCVA correlates only moderately with CST at initiation and during treatment [4].

Other functional assessments, such as contrast sensitivity (CS) and patient-reported outcome measures (PROMs), may provide valuable insights into DMO progression and the functional impact. These assessments could be key to measuring the impact of DMO on vision, informing treatment initiation, and supporting the development of sight-saving therapies [5]. Cost and practicality limit the clinical use of these assessments, even in high-resource environments [2], and knowledge gaps remain regarding their application to DMO [6, 7].

This study sought to establish expert consensus on the role of structural and functional assessments in DMO diagnosis and management, specifically addressing whether BCVA alone reflects patient visual function and identifying optimal outcome measures for disease control. To achieve this, a modified Delphi technique was selected as an established method for determining expert agreement in healthcare.

Methods

The aims and scope of the project were designed by the steering committee (SC) (n = 2). These individuals (CD and AA) are experienced ophthalmologists, with a history of research, publications, and engagement in developing standards in ophthalmology. The process followed a modified Delphi methodology following a literature review (Fig. 1), guided by an independent facilitator with experience conducting Delphi studies (Triducive Partners Limited). The study was not registered, and all reporting follows the ACCORD (ACcurate COnsensus Reporting Document) guidelines [8].

Fig. 1.

Modified Delphi consensus process followed throughout the study. BCVA best-corrected visual acuity, CS contrast sensitivity, DMO diabetic macular oedema

SC Meeting 1 (Round 1)

Prior to the first Steering Committee (SC) meeting, a questionnaire, informed by the literature review, gathered input on the advantages, disadvantages, and potential clinical challenges of anatomical and functional tests for DMO. Facilitated by an independent practitioner, the SC met virtually in April 2024 and, after reviewing questionnaire responses, defined three focus topics: (Topic 1) Treatment burden of DMO as a chronic disease for patients, clinicians, and caregivers, (Topic 2) Treatment goal and disease control, and (Topic 3) Appropriate markers of structure and function and their correlation.

A consensus threshold of 75% was set, and a minimum target of 50 survey responses over a 4-week period, with 90% of statements (excluding grouped statements) meeting the consensus threshold of 75%, was agreed to as stopping criteria. The 75% threshold, a widely accepted benchmark, was used to define consensus, while statements with ≥ 90% agreement were categorized as achieving ‘very strong’ consensus.

Statement Agreement (Round 2)

Following a discussion, the SC generated 39 statements, which were independently rated as “accept”, “remove”, or “reword”. One statement was removed, two were modified, two new statements were added, and 36 statements were accepted for inclusion without modification. A final review resulted in 40 accepted statements which were incorporated into a four-point Likert scale survey. Grouped statements, using standardized language, were developed to test which assessments respondents considered correlated most strongly to BCVA (Statements 22–29) and visual impairment (Statements 32–38).

Two ranking questions were included to assess the perceived importance and utility of assessments. Respondents were asked to rank biomarkers by their strength of correlation with BCVA outcomes and visual function tests by their reflection of patient-experienced impairment (Table 1).

Table 1.

(A) Biomarkers included to rank according to strength of correlation with BCVA outcomes, and (B) visual function tests included to rank according to strength of reflection of patient-experienced impairment

(A) Structural biomarkers selected to correlate with best-corrected visual acuity (BCVA) in DMO	(B) Functional tests selected to reflect visual impairment experienced by patients
Central subfield thickness (CST) measured by optical coherence tomography (OCT)	BCVA
Epiretinal membrane (ERM)	Contrast sensitivity
Ellipsoid zone integrity (EZI)	Distance and near low luminance (DNLL)
Disorganisation of retinal inner layers (DRIL) assessed by OCT	Electrophysiology, e.g., multifocal electroretinography (mfERG) and focal electroretinography (FERG)
Disorganisation of retinal outer layers (DROL) assessed by OCT	Near visual acuity
Presence and extent of macular ischemia assessed by imaging modalities such as OCT angiography	Microperimetry
Quantification of intraretinal fluid (IRF) volumes using OCT	Reading speed
Quantification of subretinal fluid (SRF) volumes using OCT	Patient-reported outcome measures (PROMs) such as the National Eye Institute Visual Functioning Questionnaire (NEI VFQ-25)

Survey (Round 3)

The survey was distributed by the SC to a wider panel of peers using a snowball sampling method. Inclusion criteria for survey respondents were set as follows: (1) they must be employed in a relevant role (retina specialist, retina fellow, or general ophthalmologist) and (2) have experience in managing patients with DMO. The survey was disseminated by the SC across Asia, Africa, Europe, North and South America, and Oceania. All communications were undertaken in English, and responses were anonymised. A snowball sampling technique was used to maximise international reach within a specialist community. Sampling was stopped once predefined stopping criteria were met, including a minimum of 50 responses, representation across multiple continents, and ≥ 90% of statements (excluding grouped statements) achieving consensus.

Results Analysis

Demographic data was captured, but no personal information beyond this was gathered. Only aggregated results and demographic summaries were shared with the SC. A statement of consent was included at the start of the survey, and consent was implied by completion of the survey. Ethical approval was not required, as patient-specific data was not captured. The identity of respondents was not known to either the SC or facilitator. No incentives were provided to the panel for providing responses to the survey. See disclosures regarding funding details. Survey analysis produced an arithmetic agreement score for each statement. Ranked data were collated using Microsoft Forms (Microsoft Corporation), and survey results analysis was completed using Microsoft Excel (Microsoft Corporation) in September 2024.

Steering Group (SG) Meeting 2 (Round 4)

An extended expert group (n = 4) joined the SC (n = 2) and convened in October 2024 to agree that stopping criteria were met, no further survey rounds were required, and to discuss results. Draft recommendations were produced by the combined steering group (SG) (n = 6), independently refined, and amended by each member of the SG during the process of manuscript development.

Ethical Approval

Ethics approval was not sought because neither the assigned interventions nor the outcomes assessed were related to the health of participants, and the study employed a non-interventional opinion-based Delphi process involving healthcare practitioners, with no collection of sensitive or personally identifiable data. In accordance with the UK Governance Arrangements for Research Ethics Committees (GAfREC) section 2.3, research involving clinical staff recruited by virtue of their professional role does not require research ethics committee review unless it involves access to confidential information or raises issues of professional performance. In accordance with the Declaration of Helsinki, all respondents involved in the survey within study were informed of the research purpose and that their data would remain anonymous. Their consent was assumed through the completion and submission of their survey responses.

Results

The survey yielded 93 responses, primarily from retina specialists (94.6%), three (3.2%) from general ophthalmologists, and two (2.2%) from retina fellows. Role-based sub-analysis was not performed due to the disproportionate number of retina specialists. Regional analysis spanned six continents, including Europe (n = 38), Asia (n = 29), North America (n = 10), Africa (n = 9), and South America and Australia/Oceania; however, the last two were excluded due to small numbers (n = 2).

The survey revealed consensus (≥ 75%) for 35 statements, with 23 reaching very strong agreement (≥ 90%) and five failing to reach consensus (Table 2, Supplementary Figs. 1, 2).

Table 2.

Statement breakdown with mean agreement score

Green signifies > 90% agreement, colourless is between 75 and 89%, and red indicates < 75% agreement

Very strong agreement (> 90%) was reached for all statements in Topic 1, reflecting clear agreement on the burden of DMO for patients, clinicians, and caregivers. Notably, 100% of respondents agreed that tailoring treatment plans to individual patient-centred needs is essential for optimising treatment adherence and outcomes (S4). Meanwhile, respondents also agreed that the management of DMO places a significant burden on clinics and healthcare systems (S7, 95%).

Regarding treatment goals (Topic 2), most statements reached very strong agreement, with only two statements on anatomical outcomes (S11 and S13) achieving < 90% agreement. There was very strong agreement that treatment goal should be to improve the patient’s visual function (S10, 99%). Very strong agreement was recorded across statements emphasising the importance of considering ethnic variations in structure–function relationships: within diagnostic criteria and treatment protocols (S19, 94%), in clinical trial representation (S20, 97%), and in practice to inform research and understanding (S21, 100%).

Topic 3, concerning structure–function biomarker correlation, revealed very strong consensus that BCVA is insufficient for assessing patient-experienced visual impairment (S31, 96%), and supporting the potential for a composite functional assessment (of two or more tests) to demonstrate better correlation with structural changes (S30, 95%).

Sub-analyses explored consensus differences by region, experience, and patient volume are presented in Supplementary Figs. 3–5. It was agreed that the range of 0–50 patients seen per month was too broad to draw general assumptions. Most clinicians, location-dependent, would manage between 0 and 50 patients per month, but specialists who have an experienced managerial role (i.e., spending less time in clinic) would fall to the lower end of the range.

Ranking Results of Structural and Functional Outcomes

Ellipsoid zone integrity (EZI) was ranked as the structural biomarker believed to have the strongest correlation with BCVA, while epiretinal membrane (ERM) had the weakest (Fig. 2).

Fig. 2.

Order of structural biomarkers selected to correlate strongly with VA outcomes in DMO and a breakdown of the number of respondent choices for each marker. CST central subfield thickness, DRIL disorganisation of retinal inner layers, DROL distribution of retinal outer layers, ERM epiretinal membrane, EZI ellipsoid zone integrity, IRF intraretinal fluid, MI macular ischemia, OCT optical coherence tomography, OCT-A optical coherence tomography angiography, SRF subretinal fluid

All experience levels agreed that EZI was strongly correlated with BCVA. Regional variation sub-analysis also scored EZI highly in Europe, Asia, and North America but ranked it lower in Africa and South America (Supplementary Figs. 6–9). Patient-reported outcome measures (PROMs) and BCVA were ranked as the functional biomarkers having the strongest correlation with visual impairment experienced by patients, whilst electrophysiology scored the lowest (Fig. 3).

Fig. 3.

The order of functional biomarkers selected to correlate strongly with visual impairment experienced by patients and a breakdown of the number of respondent choices for each marker. BCVA best-corrected visual acuity, CS contrast sensitivity, FERG focal electroretinography, IVI impact of vision impairment, mfERG multifocal electroretinography, NEI VFQ-25 National Eye Institute Visual Functioning Questionnaire 25, PROMs patient-reported outcome measures

Those with > 11 years of experience in the specialty ranked distance and near low luminance as having the strongest correlation with patient-experienced visual impairment, whereas those with < 5 years and 5–10 years of experience ranked distance and near low luminance sixth and fourth, respectively (Supplementary Figs. 6–9).

Discussion

Our findings highlight a critical variance between current clinical practice for DMO and the provision of optimal patient-centred care. While BCVA and CST have been imperative in the approval of DMO therapies and use in clinical practice, this study confirms that the wider ophthalmology community agrees these measures are insufficient to deliver optimal care. This has far-reaching implications for every stage of DMO management, from diagnosis, treatment decisions, monitoring, and ultimately to patient outcomes.

Table 3 lists the key recommendations for optimal care identified from the results, and the subsequent discussion highlights the practical challenges faced in clinical settings.

Table 3.

Key recommendations

Topic	Recommendations
Treatment burden of chronic disease for patients, clinicians, and caregivers	1	Patients and caregivers urgently require individualised DMO treatment strategies that are not only efficacious but also easy to access and manage, while minimising the burden on healthcare systems
Treatment goals and disease control	2	Clinicians should incorporate ethnicity into patient records and provide culturally sensitive care to address the unique needs of diverse populations
	3	Researchers should investigate correlations between functional tests and structural changes in DMO through longitudinal studies to better understand the temporal relationship
Appropriate biomarkers of structure and function and their correlation	4	Considering emerging evidence of the value of CS, PROMS, DNLL, and OCT-A metrics, further research is required to validate the role of these functional tests and novel structural biomarkers in the assessment of DMO and potential inclusion in composite assessments
	5	Given the strong correlation between macular ischemia and visual acuity outcomes, OCT-A should be considered a valuable tool to validate quantitative metrics for predicting visual prognosis in patients with DMO
	6	A composite assessment (using two or more tests) may be required to provide a full picture that is more representative of a change in function

CS contrast sensitivity, DMO diabetic macular oedema, DNLL distance and near low luminance, OCT-A optical coherence tomography angiography, PROMS patient-reported outcome measures

Topic 1: Treatment Burden of Chronic Disease for Patients, Clinicians, and Caregivers

Recommendation 1: Patients and caregivers urgently require individualised DMO treatment strategies that are not only efficacious but also easy to access and manage, while minimising the burden on healthcare systems.

Clinicians agreed that frequent and repeated DMO treatments pose a significant burden on patients (Statement 1, 100%), caregivers (Statement 6, 100%), and the healthcare system (Statement 7, 95%). As statements were based on established background literature on treatment burdens and barriers [9], benefits of individualised care [10], and shared decision-making [3], agreement scores indicate an understanding and appreciation of the existing knowledge base. Delivering effective DMO treatment in the real world is often met with substantial challenges, leading to discrepancies between clinical trial and real-world outcomes [11, 12]. A London-based multi-ethnic study showed that anti-vascular endothelial growth factor (anti-VEGF) injection effectiveness was lower than in clinical trials, with patients receiving 2.5 fewer injections over a 12-month period [11]. Undertreatment, a known real-world limitation [13], underscores the need for patient-centred care that maximises adherence and minimises the strain on patients, caregivers, and healthcare systems. Shared decision-making (SDM), while effective, may be limited by two key barriers in real-world practice: low patient health literacy and appointment time constraints [14]. Initiatives and efforts targeted at improving disease and treatment knowledge empowers patients to participate more actively in SDM [14], leading to improvement in treatment adherence and better vision-related outcomes [9]. When combined with longer-acting therapies for DMO, SDM can enhance quality of life for patients and caregivers by reducing treatment burden and supporting better visual outcomes through greater patient ownership and engagement in care [9].

Beyond patient-level and clinician-level barriers, system-level factors substantially influence real-world DMO management. Variations in national reimbursement policies, access to anti-VEGF agents, availability of diagnostic technologies, and clinic capacity can constrain treatment choices independently of clinical efficacy [15]. In many settings, limitations on drug coverage, visit frequency, or imaging access necessitate deviations from evidence-based regimens, potentially contributing to undertreatment and variable outcomes [16]. These healthcare system constraints differ widely across regions and should be considered when interpreting the applicability of the present consensus.

Clinician burnout was identified as a key issue (Statement 8, 94%), especially amongst those with fewer years of experience. However, it is not known whether this burnout was due solely to the burden of DMO management or a broader reflection of caseload and capacity within healthcare systems, and this requires further work to understand. Regional differences in Africa were noted, where clinicians were less likely to perceive DMO management as a significant burden on healthcare systems (Statement 7, 67%). Limited access to advanced diagnostic tools such as OCT [17] and lower uptake of intravitreal therapies due to lack of infrastructure and cost coupled with smaller clinic populations may have contributed to this finding.

Topic 2: Treatment Goals and Disease Control

Recommendation 2: Clinicians should incorporate self-reported ethnicity and/or ancestry into patient records and provide culturally sensitive care to address the unique needs of diverse populations.

Globally, people of South Asian and African ethnicities and/or ancestry have a significantly higher risk of DR and sight-threatening diabetic retinopathy (STDR) than the white population [18, 19]. There is evidence of disparities in access to care, and DMO treatments may be less effective in these populations [20]. Data gaps on race/ethnicity/ancestry limits our understanding of the impact on pathophysiology and treatment efficacy in DMO. Enhanced data collection and inclusive clinical trials are needed to address these disparities and help determine whether screening and treatment protocols are fit for purpose for all demographics.

Recommendation 3: Researchers should investigate correlations between functional tests and structural changes in DMO through longitudinal studies to better understand the temporal relationship.

Sub-analysis by experience showed a divide between clinicians, especially when considering the importance of restoring eye anatomy. Clinicians with ≥ 20 years’ experience had higher agreement (94%) than clinicians with ≤ 5 years (50%) that restoring the normal anatomy of the retina should be a primary treatment goal. The consensus variance may relate to an experience-level shift in clinical perspectives. This is particularly evident in the evolving understanding of conditions like DMO, where the necessity of a completely dry retina is being challenged [21].

Many patients with DMO may have good vision at diagnosis, and the DRCR.net protocol V study [22] evaluated management of this group of patients, recommending that they can be observed until there is significant BCVA reduction. A survey found that 68% of retinal specialists agreed that OCT findings are the most important considerations when initiating treatment for DMO, compared with 23% of respondents choosing BCVA [23]. This is unsurprising given the high objective reliability and repeatability of central retinal thickness (CRT) measurements compared with variable BCVA [24, 25].

Newer and experienced clinicians closely agreed that anatomical outcomes predict long-term vision, possibly due to emerging evidence on retinal fluid measures [26]. Evidence suggests presence of retinal fluid by itself shows low to moderate correlation with BCVA, but persistent and/fluctuating fluid over time can cause damage to retinal neurons and correlates with poorer visual outcomes.

Topic 3: Appropriate Biomarkers of Structure and Function and Their Correlation

Recommendation 4: Considering emerging evidence of the value of CS, PROMS, distance and near low luminance (DNLL), and optical coherence tomography angiography (OCT-A) metrics, more research is required to validate the role of these functional tests and novel structural biomarkers in the assessment of DMO and potential inclusion in composite assessments.

Recommendation 5: Given the strong correlation between macular ischemia and BCVA outcomes, OCT-A should be considered a valuable tool to validate quantitative metrics for predicting visual prognosis in patients with DMO.

Function

Patient surveys suggest that vision is the most important consideration for those with DMO [27]. CS, particularly quantitative contrast sensitivity function (qCSF), and low luminance visual acuity (LLVA) demonstrate potential as early indicators of visual impairment and disease progression, offering advantages over traditional BCVA [1, 28]. They appear to be affected earlier in the disease process and may more closely correlate with worsening disease, potentially providing a more accurate guide for treatment decisions [1, 28]. LLVA, a modified version of BCVA performed in low light [29], was ranked highly by specialists with more experience. More experienced clinicians appear to have greater confidence in interpreting results and awareness of its use within the broader therapy area [1]. CS is an emerging test that has shown promising evidence of stronger correlation with CST and quality of life (QoL), offering valuable insight on patient experience [30].

qCSF offers a faster and more practical approach for clinical implementation than CS [1, 28]. Similarly, LLVA, a modified BCVA, is readily adaptable for routine clinical use, making it a feasible addition to standard assessments [1]. However, LLVA has not been robustly studied in the context of DMO, and it is appreciated that immediate implementation of qCSF and LLVA may be limited based on knowledge gaps around test–retest variability in DMO and longitudinal responsiveness as well as time cost, equipment availability, and personnel requirements. Therefore, these tests are currently limited to research settings but may hold promise for routine practice over time.

Most respondents felt that PROMs would best correlate with patient-experienced functional impairment. The National Eye Institute Visual Functioning Questionnaire (NEI VFQ-25), which is the most used measure of vision-related quality of life in ophthalmology, does correlate with changes in BCVA over time [31], although it is not specifically designed for diabetic retinal disease (DRD), lacks cut-off scores for use in early DRD staging, and has sizeable psychometric limitations [5]. It is, therefore, not routinely used in clinical practice. In comparison, DR and DMO computerised adaptive testing (CAT) PROMs provide high measurement precision of the impact of DRD, associated visual impact, and effectiveness of related treatments on QoL [5]. DMO-PROM development holds the potential to offer more sensitive and patient-centric assessments [32].

The effectiveness of electrophysiological tests in reflecting visual impairment (VI) in patients with DMO compared to standard VA tests was debated. The results show an experience-level divide in responses to the utility of novel testing in routine practice, with junior clinicians more in favour of exploring novel testing. The observed experience-level divide does not appear to reflect resistance to innovation among senior clinicians, but rather differences in practical responsibility and accountability for service delivery. Junior clinicians may be more open to exploratory technologies, whereas senior clinicians’ concerns regarding feasibility, workflow integration, cost, and patient acceptance reflect the realities of implementing new tools in routine practice. These perspectives are complementary and underscore the importance of considering real-world constraints when translating emerging functional assessments into clinical care.

Structure

Consistent with previous reports indicating that the correlation between BCVA and CST is moderate at best, Statement S22 did not reach consensus in this study (S22, 63%) [33]. This study also suggests that other OCT biomarkers, such as the ellipsoid zone (EZ) and disorganization of the retinal outer layers (DROL), are better anatomical parameters to guide visual impact than CST. Additionally, the correlation of anatomical features like epiretinal membrane (EM), intraretinal fluid (IRF), and subretinal fluid (SRF) volumes with VA outcomes remains uncertain and did not reach consensus, reflecting findings in the literature [33].

Determining the correlation between these higher-order OCT parameters and functional tests, such as CS and LLVA, may enable more visual function-informed decision-making [34]. Whilst these structural biomarkers can be assessed qualitatively in clinics, the development of automated tools within existing OCT equipment and artificial intelligence (AI)-based algorithms would facilitate real-world validation, correlation with other functional assessments, and adoption in the clinic [35].

Recommendation 6: A composite assessment (using two or more tests) may be required to provide a full picture that is more representative of a change in function.

This study demonstrated that clinicians across various geographies agree that BCVA may not always capture the level of visual impairment experienced by patients with DMO (S31, 96%). It is well understood that BCVA alone provides a limited view of overall functionality [30] and gives inadequate data on how DMO is affecting the function of the retina (especially during early disease [1]) or the patient [30].

In the clinical trial setting, composite or multi-component endpoints may be more useful to capture functional changes. Composite endpoints combine two or more distinct endpoints to provide an overall composite event rate. The individual components of the composite must be clinically meaningful and validated to correlate with structural change in DMO, with the goal of having a composite measure that more closely tracks structural changes. From the literature and the current study, it is evident that there is the scope for such a composite endpoint to be implemented in clinical practice, providing a better representation of changes in function across retinal diseases [2]. This assessment should aim to use existing methods of testing in the clinic, whilst adapting existing patient care pathways. Given their complementary nature, potential for rapid clinic administration, and shared chronological trajectory with DMO development, CS and LLVA may be well suited for validation as a prioritised composite functional measure in DMO, especially in the early stages. DMO-PROM may also be considered for incorporation, given the strength of the patient voice and the value placed on this measure by clinicians as shown in this study. PROMS can add complementary information to inform shared decision-making during treatment initiation or escalation for example. Composite endpoints can also comprise anatomical–functional endpoints—for example, a composite endpoint comprising CS and EZ integrity may provide a more robust measure of disease impact and response to treatment over time.

Finally, it is worth noting the advances in AI and machine learning (ML) for use in DMO management. ML models accurately predicted the status of diabetic eye disease using only age, sex, and visual function measures [36]. Employing interpretable ML methods allows the identification of distinct profiles of visual function associated with different disease stages [36]. With the widespread adoption of AI, it is feasible that structural–functional composite measures incorporating fluid fluctuations, for example, would be feasible and beneficial in DMO.

Strengths and Limitations

The strong levels of global agreement seen show that there is robust consensus with regard to treatment burden and goals, with some discordance around which structural biomarkers best correlate with BCVA. Importantly, there was strong consensus regarding the utility of a composite patient-centred assessment. However, potential confirmation and sample biases, limited regional responses, and survey length may have influenced results. Specifically, the 0–50-year patient age range was deemed too broad, and qualitative measures were suggested for nuanced understanding. Leading BCVA correlation statements were also noted as a potential limitation. The use of snowball sampling introduces a risk of sampling bias, and while this approach facilitated global participation, it may limit representation. Additionally, the consensus reflects expert opinion across diverse regions but does not explicitly stratify recommendations by healthcare system, reimbursement environment, or resource availability, which may limit generalisability across all practice settings.

A further limitation relates to panel composition. While the overwhelming majority of respondents were retinal specialists, a small proportion (5.4%) were general ophthalmologists or retina fellows. Although modest, their inclusion may have influenced agreement levels for statements close to the consensus threshold (≥ 75%). To mitigate this effect, borderline consensus statements should be interpreted with caution. Sensitivity analyses excluding non-retina specialists were considered; however, the small number precluded meaningful subgroup comparison. Importantly, statements achieving very strong consensus (≥ 90%) are unlikely to have been materially affected. Finally, the absence of patients in the study design could be considered a limitation of the study.

Conclusion

This consensus presents a call to action to recognise the current significant treatment burden associated with the management of DMO, the knowledge gap in structure–function correlation in the trajectory of DMO, and the need to explore the development of robust composite assessments that capture the visual impact of DMO to improve patient care. Functional assessments such as CS, LLVA, and novel PROMs in conjunction with higher-order structural metrics from OCT and OCT-A hold the potential to offer more sensitive and patient-centric assessments that capture the quality-of-life impact and functional experience with DMO. Improved functional assessments could enhance DMO care by enabling earlier detection and better-matched treatment, and reducing vision-related decline in QoL and overall burden.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

All authors developed the initial statements, contributed to the analysis and discussion of results equally, and read and approved the final manuscript. Christiana Dinah: Supervision, Conceptualization, Investigation, Methodology, Analysis, Writing–review & editing, Final approval. Andrew Chang: Conceptualization, Investigation, Methodology, Analysis, Writing–review & editing, Final approval. Pierre-Henry Gabrielle: Conceptualization, Investigation, Methodology, Writing–review & editing, Final approval. Rishi Singh: Conceptualization, Investigation, Methodology, Writing–review & editing, Final approval. Javier Zarranz-Ventura: Conceptualization, Methodology, Final approval. Insar Saffar: Conceptualization, Methodology, Final approval. Marloes Bagijn: Conceptualization, Methodology, Writing–review & editing, Final approval. Aude Ambresin: Conceptualization, Investigation, Methodology, Writing–review & editing, Final approval.

Funding

The study was initiated by practicing ophthalmologists who are members of the F. Hoffmann-LaRoche Limited (Roche) Ophthalmology Network. The project was funded by Roche, who commissioned Triducive Partners Limited to facilitate the project and analyse the responses to the consensus statements in line with the Delphi methodology. Roche oversaw the design and conduct of the study and participated in the review of the manuscript but was not involved in the collection, management, analysis, and interpretation of the data and preparation of the manuscript. The Ophthalmology Network is funded and facilitated by Roche. The opinions of the Ophthalmology Network are those of its members and do not necessarily reflect the opinions of Roche. The journal’s Rapid Service Fee was funded by Roche.

Data Availability

All relevant data are included within the manuscript and supplementary information. The raw dataset can be accessed upon reasonable request to the corresponding author.

Declarations

Conflicts of Interest

All authors received honoraria from Roche while undertaking this study. Roche commissioned Triducive Partners Limited to facilitate the project and analyse the responses to the consensus statements in line with the Delphi methodology. The authors state the following conflicts of interest: Christiana Dinah: Consultancy and Advisory boards: Roche, Boehringer Ingelheim, Bayer, AbbVie, Johnson & Johnson, Eyepoint Pharmaceuticals, Ocular Therapeutix, Alimera Sciences, Astellas, Apellis. Speaker fees: Roche, Boehringer Ingelheim, Bayer, Topcon Ltd, AbbVie, Apellis, Astellas. Travel and conference support: Roche, Boehringer Ingelheim, Apellis. Andrew Chang: Consultancy and Advisory boards: Roche, Bayer, Novartis, Alcon, Zeiss, Apellis, Astellas, Opthea. Pierre-Henry Gabrielle: Received travel expenses from AbbVie, Bayer, and Novartis, and is a medical consultant for Astellas, Novartis, Horus pharma, Roche, Allergan/Abbvie, and Bayer. Rishi Singh: Personal fees from Apellis, Iveric Bio, Eyepoint, Regenxbio, Genentech, Bausch & Lomb, Zeiss, Alcon, and Regeneron, and research grants from Janssen. Javier Zarranz-Ventura: Abbvie, Adverum, Alcon, Alimera Sciences, Bausch & Lomb, Bayer, Brill Pharma, DORC, Esteve, Novartis, Roche, Sandoz, Topcon, Zeiss. Insar Saffar: Personal fees from Apellis, Iveric Bio, Eyepoint, Regenxbio, Genentech, Bausch & Lomb, Zeiss, Alcon, and Regeneron, research grants from Janssen, employee of F. Hoffmann-La Roche Ltd. Marloes Bagijn: Employee of F. Hoffmann-La Roche Ltd. Aude Ambresin: Abbvie, Apellis, Astellas, Bayer, Novartis, Optovue, RetinAI, Roche.

Ethical Approval

Ethics approval was not sought because neither the assigned interventions nor the outcomes assessed were related to the health of participants, and it was a non-interventional opinion-based Delphi process involving healthcare practitioners, with no collection of sensitive or personally identifiable data. In accordance with the UK Governance Arrangements for Research Ethics Committees (GAfREC) section 2.3, research involving clinical staff recruited by virtue of their professional role does not require research ethics committee review unless it involves access to confidential information or raises issues of professional performance. In accordance with the Declaration of Helsinki, all respondents involved in the survey within study were informed of the research purpose and that their data would remain anonymous. Their consent was assumed through the completion and submission of their survey responses.