Real‐world prognosis of eyes with diabetic macular oedema receiving treatment with vascular endothelial growth factor (VEGF) inhibitors

Objectives

This is a protocol for a Cochrane Review (prognosis). The objectives are as follows:

To assess one‐ to five‐year visual prognosis of eyes receiving intravitreal VEGF inhibitor treatments for DMO in routine clinical practice.

Investigation of sources of heterogeneity between studies

We will assess the source of heterogeneity in the outcome measures between studies.

Background

Description of health condition and context

The prevalence of adult diabetes is projected to increase from 415 million in 2015 to 642 million by 2040, comprising 10% of the global adult population aged between 20 to 79 years (Ogurtsova 2017). Approximately 193 million people with diabetes remain undiagnosed, predisposing them to the development of several long‐term complications of chronic hyperglycaemia. The rising prevalence of diabetes poses significant challenges to healthcare systems worldwide.

Diabetes is associated with a significant increase in morbidity and mortality through the direct and indirect effects of hyperglycaemia on large and small blood vessels (Garcia 1974). The chronic sequelae of diabetes are separated into macrovascular and microvascular disease (Crofford 1995). The microvascular complications of diabetes affect the retina (diabetic retinopathy and diabetic macular oedema (DMO)).

There are approximately 93 million people in the world with diabetic retinopathy, 17 million with proliferative diabetic retinopathy, 21 million with DMO, and 28 million with vision‐threatening diabetic retinopathy (Yau 2012). The overall global prevalence of DMO is 6.8% of people with diabetes (6.74 to 6.89). Diabetic retinopathy has been estimated to be responsible for vision impairment in 3.7 million people globally and blindness in 0.8 million, contributing to 2.6% of overall blindness and 1.9% of vision impairment (Leasher 2016).

Primary interventions such as improving blood glucose, blood pressure, and serum lipid profiles can reduce the risk of development of diabetic retinopathy and DMO (Yau 2012). All the secondary interventions to prevent loss of vision in eyes with established retinopathy that have been tested so far (aspirin, protein kinase C inhibitors (ruboxistaurin), aldose reductase inhibitors (sorbinil), inhibitors of growth hormone and insulin‐like growth factors (octreotide)) have failed (Aiello 2007; Chew 1995; Kirkegaard 1990; Sorbinil Retinopathy Trial Research Group 1990). Laser treatment with panretinal photocoagulation and focal/grid laser was the first intervention identified to reduce the risk of loss of vision from diabetic retinopathy and DMO (DRS2 1978; DRS8 1981; ETDRS19 1995). However, a significant number of patients treated with laser continued to lose vision (Schatz 1992).

Pivotal phase 3 clinical trials of the vascular endothelial growth factor (VEGF) inhibitors ranibizumab, bevacizumab, and aflibercept for DMO have shown excellent results in the short and medium term (Heier 2016; Michaelides 2010; Rajendram 2012; Schatz 1992). Clinical trials study new treatments in highly controlled conditions for a highly selected group of patients that may not be representative of the general population with the disease. The applicability of the results of pivotal clinical trials to the general population under real‐life conditions thus needs to be confirmed by population‐based post‐marketing observational studies (Sherman 2016).

Health outcomes

This review will assess the visual prognosis of eyes receiving VEGF inhibitors for DMO for at least 12 months in real‐world clinical practice. We will also assess whether age, visual acuity, and central macular thickness at baseline and the number of injections had any effects on vision changes with VEGF inhibitor treatment.

Why is it important to do this review?

Clinical trials with VEGF inhibitors for another common macular disease, neovascular age‐related macular degeneration (nAMD), showed remarkable results, with mean visual acuity improving for the first time for this condition (Brown 2009; Heier 2012; Rosenfeld 2006). However, there is good evidence showing that these results could not in fact be replicated for patients in most parts of the world. One large observational study that evaluated outcomes in eyes with nAMD receiving VEGF inhibitors in routine clinical practice reported poor outcomes in many European countries around 2011, where the mean improvements on a visual acuity chart were only 1 or 2 letters compared with 7 to 12 letters in the clinical trials (Finger 2013). However, other observational studies from roughly the same time demonstrated that the results of the clinical trials were being replicated in real‐world practice in Australia, highlighting that the problems in Europe may include restricted access to treatment and reactive rather than proactive treatment regimens which led to insufficient injection frequencies (Arnold 2015; Gillies 2013). The steps being taken to address this issue in Europe as a result of these studies will likely improve outcomes in hundreds of thousands of people.

This review will evaluate the overall prognosis of eyes with DMO receiving VEGF treatments in routine clinical practice. A clearer understanding of this will identify situations where outcomes are inferior and how they may be improved.  Data from this review will also improve patient compliance by providing an idea of what will be required when the patient commences their 'treatment journey'.

Objectives

To assess one‐ to five‐year visual prognosis of eyes receiving intravitreal VEGF inhibitor treatments for DMO in routine clinical practice.

Investigation of sources of heterogeneity between studies

We will assess the source of heterogeneity in the outcome measures between studies.

Methods

Criteria for considering studies for this review

Types of studies

We will include any retrospective or prospective study in routine clinical practice evaluating overall outcomes in eyes receiving treatment with VEGF inhibitors for DMO. We will exclude case‐control, cross‐sectional, and randomised studies.

Follow‐up period of at least 12 months from starting treatment.

Targeted population

People aged 18 years and over who received VEGF inhibitors (aflibercept, bevacizumab, or ranibizumab) for the treatment of DMO in one or both eyes in routine clinical practice.

Types of outcomes to be predicted

Primary outcome

• Change in visual acuity in logMAR letters (at 1 year, 2 years, and 3 to 5 years).

Table 1. The PICOTS format for the primary outcome of the review (Debray 2017; Moons 2014)
Population	Adults (18 years and over) with DMO in one or both eyes
Observed intervention	Eyes that received VEGF inhibitors for DMO
Comparator	None
Outcomes	Mean change in visual acuity
Timing	At least 12 months from the start of VEGF inhibitor treatment
Setting	To predict the outcomes of eyes receiving VEGF inhibitors for DMO in routine clinical practice
DMO: diabetic macular oedema; PICOTS: Population, Intervention, Comparator, Outcome(s), Timing, and Setting; VEGF: vascular endothelial growth factor

Secondary outcomes

• Proportion of eyes with 6/12 (driving) vision (at 1 year, 2 years, and 3 to 5 years).

• Proportion of eyes with visual gain of ≥ 5 letters (1 line), ≥ 10 letters (2 lines), and > 15 letters (3 lines) and those that lost ≥ 5 letters, ≥ 10 letters, and ≥ 15 letters on their last follow‐up visit compared to the baseline visit (at 1 year, 2 years, and 3 to 5 years).

• Change in central macular thickness (at 1 year, 2 years, and 3 to 5 years).

• Number of injections received (at 1 year, 2 years, and 3 to 5 years).

• Adverse events associated with the treatment.

Investigation of sources of heterogeneity between studies 

We will assess sources of heterogeneity in the outcome measures between the studies. We will assess whether there were associations of the mean change in visual acuity with age, visual acuity, and central macular thickness at baseline or the number of injections.

Search methods for identification of studies

Electronic searches

The Cochrane Eyes and Vision Information Specialist will search the following electronic databases for this prognostic review. We will not use a study design filter to limit the results. There will be no restrictions to language, but the searches will begin from 1997, the year that VEGF was first used.

• MEDLINE Ovid (1997 to present) (Appendix 1).

• Embase Ovid (1997 to present) (Appendix 2).

Searching other resources

We will also handsearch references of studies identified by the database obtained from searches. We will use the Science Citation Index to identify studies that have cited the individual studies included in the review.

Data collection

We will use the guidelines from the checklist for critical appraisal and data extraction for systematic reviews of prediction modelling studies (CHARMS) (Moons 2014).

Selection of studies

Two review authors will independently screen the titles and abstracts of records identified as a result of the electronic searches for potentially relevant studies, and then screen the full‐text reports of potentially relevant studies to identify studies for inclusion in the review. In the case of disagreement, a third review author will be consulted.

Data extraction and management

Two review authors will independently extract the data. Where necessary, we will contact the authors of the individual studies for additional information.

We will extract the following data from each study.

• Country and setting in which study was conducted

• Eligibility criteria of the included studies

• Study design

• Study dates

• Number of eyes

• Number of participants

• Mean age of participants

• Gender ratio

• Follow‐up duration (mean or median or interquartile range)

• Mean visual acuity at baseline, 1 year, 2 years, and ≥ 3 years

• Mean central macular thickness at baseline, 1 year, 2 years, and ≥ 3 years

• VEGF inhibitor type: bevacizumab, ranibizumab, or aflibercept

• Proportion of eyes gaining ≥ 5 and ≥ 10 letters

• Proportion of eyes losing ≥ 5 and ≥ 10 letters

• Proportion of eyes with 6/12 vision

• Mean number of intravitreal injections administered at 1 year, 2 years, and ≥ 3 years

• Mean central macular thickness at 1 year, 2 years, and ≥ 3 years

• Mean number of visits at 1 year, 2 years, and ≥ 3 years

• Type of adverse events

• Total number of adverse events

• Missing data

We will use a standard  MS Excel (Microsoft Excel) based data collection form (Appendix 3) for study characteristics and outcome data which we will pilot on three studies eligible for inclusion in the review. We will compare the extracted data and discuss potential issues to optimise the data extraction form.

Assessment of risk of bias in included studies

Two review authors will conduct the risk of bias assessment. The opinion of a third review author will be sought to resolve any disagreements. The review authors will not be masked to study authors, institution, or journal of publication due to feasibility. We will contact the study authors if any information needed to make a decision on risk of bias is lacking. We will document that we attempted to contact the study authors and mark the risk of bias as unclear when the study authors are unwilling or unable to provide us with additional information.

We will use a modification of the Quality in Prognostic Studies (QUIPS) tool for the assessment of risk of bias in this prognostic review (Hayden 2013). Details regarding the coding of risk of bias are provided in Appendix 4. We will judge risk of bias in each report by examining study participation, study attrition, outcome measurement, and statistical analysis and reporting. A few prompting items will be used to rate the adequacy of reporting by a study as yes, partial, no, or unsure. We will combine these ratings to provide an overall risk of bias for each domain and assign a risk of bias rating for the study. We will use risk of bias assessments to assess the quality of included studies and for sensitivity analyses.

We will pilot the risk of bias form on three eligible studies to compare the extracted data and discuss potential issues to optimise the risk of bias form (Appendix 4).

Data synthesis

We will pool primary and secondary outcomes by appropriate random‐effects (Hartung‐Knapp‐Sidik‐Jonkman) meta‐analysis for continuous and binary data. For the meta‐analysis of proportions, the pooled estimates are obtained as a weighted average by fitting the logistic regression model without covariate but an intercept using the binomial distribution to model the within‐study variability and 95% confidence intervals calculated by Wilson score procedure.

Participants with both eyes will be included in the analysis. We may perform a sensitivity analysis using only studies that reported outcomes for unilateral cases.

Assessment of heterogeneity

We will assess statistical heterogeneity by formal Chi² test and the proportion of variability in effect estimates across studies that is due to heterogeneity (rather than sampling error) quantified by the I² statistic (Debray 2017; Hayden 2013).

We will consider a random‐effects meta‐regression analysis, weighted using the inverse‐variance method, to explore the relationship between the number of injections, age, baseline visual acuity, and baseline central macular thickness on the visual acuity outcomes.

Dealing with missing data

In the case of missing standard deviation values, we will impute the values from other included studies or algebraically estimate them from other summary statistics (Weir 2018).

In the case of missing means, we will attempt to estimate them by medians, lower/upper quartiles, or other summaries and impute the calculated values.

We will use the formulas as suggested by Wan and colleagues to calculate the missing sample means (and standard deviation) using the available median, lower quartile and upper quartile performed best in preserving the precision of the meta‐analysis findings (Wan 2014).

We will also present a scenario analysis where we will perform meta‐analysis using complete‐case analyses.

Subgroup and investigation of heterogeneity 

In the case of sufficient studies, we will conduct subgroup analysis to examine whether treatments with any of the VEGF inhibitors have a benefit over the other. We will consider univariable random‐effects meta‐regression analyses to explore the relationship between the number of injections, age, baseline visual acuity, and baseline central macular thickness on the visual acuity outcomes.

Sensitivity analyses

We may perform a sensitivity analysis using only studies that reported outcomes for unilateral cases.

We also plan to run a sensitivity analysis including only studies at low risk of bias as defined in Appendix 4.

Conclusions and summary of findings

We will use the GRADE framework for judging the certainty of evidence and present our results in a summary of findings table (Appendix 5) (Guyatt 2011). We will evaluate the certainty of the evidence using the five GRADE considerations: study limitations (risk of bias), inconsistency of effect, imprecision, indirectness, and publication bias (Appendix 6) (Guyatt 2011; Hayden 2014; Iorio 2015). We will assess the certainty of the evidence as high, moderate, low, or very low (Appendix 7). We will justify all decisions to downgrade or upgrade by using footnotes and making comments to aid the reader’s understanding of the review where necessary.

Appendices

Appendix 1. MEDLINE Ovid search strategy

1. exp macular edema/     
2. (macula$ adj3 oedema).tw.     
3. (macula$ adj3 edema).tw.     
4. maculopath$.tw.     
5. (CME or CSME or CMO or CSMO).tw.     
6. (DMO or DME).tw.     
7. or/1‐6     
8. exp diabetes mellitus/     
9. diabetic retinopathy/     
10. diabetes complications/     
11. diabet$.tw.     
12. retinopath$.tw.     
13. or/8‐12     
14. exp angiogenesis inhibitors/     
15. angiogenesis inducing agents/     
16. endothelial growth factors/     
17. exp vascular endothelial growth factors/     
18. (macugen$ or pegaptanib$ or lucentis$ or rhufab$ or ranibizumab$ or bevacizumab$ or avastin$ or aflibercept$ or conbercept$ or OPT 302 or Opthea$ or RTH258 or Brolucizumab$ or abicipar pegol).tw.     
19. (anti adj2 VEGF$).tw.     
20. (anti adj1 angiogen$).tw.     
21. (endothelial adj2 growth adj2 factor$).tw.     
22. VEGF TRAP$.tw.     
23. or/14‐22     
24. 7 and 13 and 23

Appendix 2. Embase Ovid search strategy

1. diabetic macular edema/     
2. (diabet$ adj2 macula$ adj2 oedema).tw.     
3. (diabet$ adj2 macula$ adj2 edema).tw.     
4. (DMO or DME).tw.     
5. or/1‐4     
6. macula edema/     
7. (macula$ adj3 oedema).tw.     
8. (macula$ adj3 edema).tw.     
9. maculopath$.tw.     
10. (CME or CSME or CMO or CSMO).tw.     
11. or/6‐10     
12. exp diabetes mellitus/     
13. diabetic retinopathy/     
14. diabet$.tw.     
15. retinopath$.tw.     
16. or/12‐15     
17. angiogenesis/     
18. exp angiogenesis inhibitors/     
19. angiogenic factor/     
20. endothelial cell growth factor/     
21. bevacizumab/ or brolucizumab/ or ranibizumab/     
22. vasculotropin/     
23. (macugen$ or pegaptanib$ or lucentis$ or rhufab$ or ranibizumab$ or bevacizumab$ or avastin$ or aflibercept$ or conbercept$ or OPT 302 or Opthea$ or RTH258 or brolucizumab$ or abicipar pegol).tw.     
24. (anti adj2 VEGF$).tw.     
25. (anti adj1 angiogen$).tw.     
26. (endothelial adj2 growth adj2 factor$).tw.     
27. VEGF TRAP$.tw.     
28. or/17‐27     
29. 5 and 28     
30. 11 and 16 and 28     
31. 29 or 30

Appendix 3. Data collection spreadsheet

Item	Working definition
Study number
General information
Journal citation	Journal name, year, volume, pages
Language
Author	First author and/or corresponding author
Sources of data
Study design
Prospective
Retrospective
Start and end recruitment reported (Yes/No)
Start recruitment	[date]
End recruitment	[date]
Participants
Study population/group location
Size of population/group
Male: female
Age, mean (SD) years
Diabetes duration, mean (SD) years
VA baseline, mean (SD) logMAR letters
Proportion of eyes with VA ≥ 69 letters	(numerator/denominator or proportion)
CST baseline, mean (SD) mm
VEGF inhibitors
Monotherapy	[Yes/No]
Aflibercept/bevacizumab/ranibizumab
Outcomes
Follow‐up duration
VA final, mean (SD) logMAR letters
CST baseline, mean (SD) mm
Proportion of eyes with VA ≥ 69 letters	(numerator/denominator or proportion)
Proportion of eyes gaining ≥ 5 and ≥ 10 letters	(numerator/denominator or proportion)
Proportion of eyes losing ≥ 5 and ≥ 10 letters	(numerator/denominator or proportion)
Mean (SD) number of intravitreal injections
Mean (SD) number of visits
Type of adverse events
Total adverse events
Missing data
Number of participants lost to follow‐up
Handling of missing data
Complete‐case analysis
Multiple analysis
Other methods
CST: central subfield thickness, logMAR: logarithm of minimum angle of resolution, VA: visual acuity, SD: standard deviation, VEGF: vascular endothelial growth factor

Appendix 4. Risk of bias table based on QUIPS

Risk of bias table based on QUIPS
Variable	Bias domains
	1. Study participation	2. Study attrition	3. Outcome measurement	4. Statistical analysis and reporting
Optimal study or characteristics of unbiased study	The study sample adequately represents the population of interest.	The study data available (i.e. participants not lost to follow‐up) adequately represent the study sample.	The outcome of interest is measured in a similar way for all participants.	The statistical analysis is appropriate, and all primary outcomes are reported.
Prompting items and considerationsa	a. Adequate participation in the study by eligible individuals	a. Adequate response rate for study participants	a. A clear definition of the outcomes is provided	a. Sufficient presentation of data to assess the adequacy of the analytic strategy
	b. Description of the source population or population of interest	b. Description of attempts to collect information on participants who dropped out	b. Method of outcome measurement used is adequately valid and reliable	b. The selected statistical analysis is adequate for the design of the study
	c. Description of the baseline study sample	c. Reasons for loss to follow‐up are provided	c. The method and setting of outcome measurement is the same for all study participants	c. There is no selective reporting of results
	d. Adequate description of the sampling frame and recruitment	d. Adequate description of participants loss to follow‐up
	e. Adequate description of the period and place of recruitment	e. There are no important differences between participants who completed the study and those who did not
	f. Adequate description of inclusion and exclusion criteria
Ratingsb
High risk of bias	The relationship between the prognosis and the outcome is very likely to be different for participants and eligible non‐participants.	The relationship between the prognosis and the outcomes is very likely to be different for completing and non‐completing participants.	The measurement of the outcome is very likely to be different related to the baseline level of the participants.	The reported results are very likely to be spurious or biased related to analysis or reporting.
Moderate risk of bias	The relationship between the prognosis and the outcome may be different for participants and eligible non‐participants.	The relationship between the prognosis and the outcome may be different for completing and non‐completing participants.	The measurement of the outcome may be different related to the baseline level of the participants.	The reported results may be spurious or biased related to analysis or reporting.
Low risk of bias	The relationship between the prognosis and the outcome is unlikely to be different for participants and eligible non‐participants.	The relationship between the prognosis and the outcome is unlikely to be different for completing and non‐completing participants.	The measurement of the outcome is unlikely to be different related to the baseline level of the participants.	The reported results are unlikely to be spurious or biased related to analysis or reporting.
aPrompting items are to guide the user’s judgement regarding the risk of bias for each domain and are taken together to inform the overall judgement of potential bias and to facilitate consensus between review authors for each domain. Some items may not be relevant to the specific study or the review research question. bEach domain is rated as having high, moderate, or low risk of bias based on the prompting items.

Appendix 5. Summary of findings table

[Enter text]

Summary of findings table
Year	Outcomes	Aflibercept	Bevacizumab	Ranibizumab	Certainty of evidence and reason for downgrading
1	Change in visual acuity in logMAR letters, mean (95% CI)   Eyes with ≥ 10 letters gain, %   VA ≥ 6/12 at baseline/final, %   Eyes with ≥ 10 letters loss, %   Change in CST, mean µm (95% CI)   Number of injections, mean (95% CI)
2	Change in visual acuity in logMAR letters, mean (95% CI)   Eyes with ≥ 10 letters gain, %   VA ≥ 6/12 at baseline/final, %   Eyes with ≥ 10 letters loss, %   Change in CST, mean µm (95% CI)    Number of injections, median (Q1, Q3)
3 to 5	Change in visual acuity in logMAR letters, mean (95% CI)   Eyes with ≥ 10 letters gain, %   VA ≥ 6/12 at baseline/final, %   Eyes with ≥ 10 letters loss, %   Change in CST, mean µm (95% CI)   Number of injections, mean (95% CI)
CI: confidence interval, CST: central subfield thickness, logMAR: logarithm of minimum angle of resolution, VA: visual acuity

 

Appendix 6. GRADE assessment for judging the overall certainty of the evidence on prognosis

GRADE assessment for judging the overall certainty of the evidence for treatment outcomes
	Domain	Description
Grade down if:	Risk of bias	Most evidence is from studies with moderate or high risk of bias for most risk of bias domains.
	Inconsistency	Unexplained heterogeneity or variability in results (point estimates) across studies with differences in results that are not clinically meaningful
		Meta‐analysis: large I2 value (significant heterogeneity)
		Narrative analysis: variations in estimates across studies with points of effect on either side of the lines of no effect and confidence intervals showing little overlap
	Indirectness	The study sample or the outcomes in the study, or both, do not accurately reflect the population of interest, or the measured outcome does not capture what is believed to be important.
	Imprecision	This is based on the width of the 95% confidence interval around the pooled estimate/effect size (meta‐analysis) and the position of the confidence interval relative to clinical decision threshold, that is there is no precise estimate of the effect size in the meta‐analysis, and confidence intervals are excessively wide or overlap with the value of no effect.
		In narrative analyses, there is no precision in the estimation of the effect size within each study.
		Also graded down if there is across‐study imprecision: few studies and insufficient sample size (< 100 cases reaching follow‐up), or no justification provided for small sample size.
	Publication bias	Confidence downgraded unless specific outcomes have been repeatedly investigated.
Grade up if:	Large effect	Moderate or large effect reported by most studies or in pooled findings in the meta‐analysis.
	Dose‐response gradient	Gradient exists between studies for factors measured at different doses or an increase or decrease in events over time, which follows a well‐defined pattern (e.g. linear).
Footnotes

Appendix 7. Levels of certainty

Levels of certainty
Quality level	Definition
High	Very confident that the true prognosis lies close to that of the estimate
Moderate	Moderately confident that the true prognosis lies close to that of the estimate, but there is a possibility that it is substantially different
Low	Confidence in the estimate is limited. True prognosis may be substantially different from the estimate.
Very low	Very little confidence in the estimate. True prognosis is likely to be substantially different from the estimate.

Contributions of authors

Drafting the protocol: SB, VN, SFB, GLDT, MCG
Final approval of protocol: all authors
SB is the guarantor for the review.

Sources of support

Internal sources

• No sources of support provided

External sources

• National Institute for Health Research (NIHR), UK

  • Richard Wormald, the former Co‐ordinating Editor for Cochrane Eyes and Vision (CEV), received financial support for his CEV research sessions from the Department of Health through the award made by the NIHR to Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology for a Specialist Biomedical Research Centre for Ophthalmology.

  • Up to March 2021, this review was supported by the NIHR, via Cochrane Infrastructure funding to the CEV UK editorial base.

  The views expressed in this publication are those of the authors and not necessarily those of the NIHR, NHS, or the Department of Health.

• Public Health Agency, UK

  As of April 2021, the completion of this review was supported by the HSC Research and Development (R&D) Division of the Public Health Agency which funds the Cochrane Eyes and Vision editorial base at Queen's University Belfast.

• Queen's University Belfast, UK

  Gianni Virgili, Co‐ordinating Editor for Cochrane Eyes and Vision’s work, is funded by the Centre for Public Health, Queen’s University Belfast, Northern Ireland.

Declarations of interest

SB: none known
VN: none known
SFB: none known
GLDT: none known
MCG: none known

New