A review of therapies for diabetic macular oedema and rationale for combination therapy

Abstract

Diabetic macular oedema (DMO) is responsible for significant visual impairment in diabetic patients. The primary cause of DMO is fluid leakage resulting from increased vascular permeability through contributory anatomical and biochemical changes. These include endothelial cell (EC) death or dysfunction, pericyte loss or dysfunction, thickened basement membrane, loss or dysfunction of glial cells, and loss/change of EC Glycocalyx. The molecular changes include increased reactive oxygen species, pro-inflammatory changes: advanced glycation end products, intracellular adhesion molecule-1, Complement 5–9 deposition and cytokines, which result in increased paracellular permeability, tight junction disruption, and increased transcellular permeability. Laser photocoagulation has been the mainstay of treatment until recently when pharmacological treatments were introduced. The current treatments for DMO target reducing vascular leak in the macula once it has occurred, they do not attempt to treat the underlying pathology. These pharmacological treatments are aimed at antagonising vascular endothelial growth factor (VEGF) or non-VEGF inflammatory pathways, and include intravitreal injections of anti-VEGFs (ranibizumab, aflibercept or bevacizumab) or steroids (fluocinolone, dexamethasone or triamcinolone) as single therapies. The available evidence suggests that each individual treatment modality in DMO does not result in a completely dry macula in most cases. The ideal treatment for DMO should improve vision and improve morphological changes in the macular (eg, reduce macular oedema) for a significant duration, reduced adverse events, reduced treatment burden and costs, and be well tolerated by patients. This review evaluates the individual treatments available as monotherapies, and discusses the rationale and potential for combination therapy in DMO. A comprehensive review of clinical trials related to DMO and their outcomes was completed. Where phase III randomised control trials were available, these were referenced, if not available, phase II trials have been included.

Introduction

In 2002, it was reported that diabetes affected 220 million people worldwide,¹ and anticipated that the prevalence of diabetes will double within the next 10 years.² More recent estimates indicate that the prevalence of diabetes in adults (aged 20–79 years) worldwide was 382 million people in 2012, and that this would likely increase to 592 million in 2035.³

Diabetic retinopathy (DR) has been extensively studied over the years, and its incidence correlates with poor glycaemic control and hyperlipidaemia.^{4, 5} Diabetic choroidopathy is a less well-studied entity, and is thought to occur in the advanced stages of diabetic eye disease.^{6, 7, 8, 9} As such, the retinal and choroidal vascular beds seem to be affected differently by diabetes. Diabetes and hyperglycaemia have obvious effects on intraocular vascular endothelial cell (EC) permeability, adhesion to leukocytes, as well as angiogenesis.^{10, 11, 12} These alterations result in increased vascular leakage (increased permeability), vascular occlusions, ischaemia, and angiogenesis.^{13, 14} However, the exact mechanisms underlying these changes are not fully understood, and require further elucidation.

Diabetic macular oedema (DMO) is responsible for significant visual impairment in diabetic patients.^{1, 2, 15, 16} In the retina, leakage is due to increased permeability that occurs at the retinal ‘neurovascular' unit, which consists of single layer of tightly adherent ECs, basal lamina, surrounding pericytes, astrocytes, and microglia leading to increased EC trans- or paracellular permeability, as summarised in the recent review by Klaassen et al.¹⁷ There are contributory anatomical and biochemical changes that are interlinked (Figure 1). The anatomic changes include EC damage—death or dysfunction, pericyte loss or dysfunction, thickened basement membrane, loss or dysfunction of glial cells, and loss/change of EC Glycocalyx. The molecular changes include increased reactive oxygen species, pro-inflammatory changes: advanced glycation end products, intracellular adhesion molecule-1 (ICAM-1), Complement 5–9 (C5–9) deposition, cytokines, which result in increased paracellular permeability—small molecules and water: tight junction (TJ) disruption—and increased transcellular permeability—large molecules and water: caveolar transport, aquaporins, plasmalemmal vesicle-associated protein. The increased intraretinal fluid leads to progressive retinal dysfunction (see Klaassen et al¹⁷ review). The contributions of the choroidal vasculature to the clinical disease of DR are less well understood, but again will be largely contributed to by the choroidal EC (CEC) alterations.^{6, 7, 8, 9} It is known that retinal vascular leakage in DMO is contributed to by vascular endothelial growth factor (VEGF) upregulation as well as non-VEGF-dependent inflammatory pathways, so that chronic subclinical inflammation is important in the pathogenesis of DR.^{18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28} An early event in the pathogenesis of diabetic vasculopathy is leukocyte adherence to retinal vascular endothelium, resulting in EC death, vascular leakage, and capillary closure¹⁸ (Figure 2).

Figure 1.

Pathogenesis of diabetic vasculopathy. DMO, diabetic macular oedema; Ang-2, Angiopoietin-2; EC, endothelial cell; ICAM, intercellular adhesion molecule; IP, interferon-induced protein; MMP, metallic metalloproteinase; VEGF, vascular endothelial growth factor; VE, vascular endothelial; PDGF, platelet-derived growth factor; TJ, tight junction, bFGF, basic fibroblast growth factor.

Figure 2.

Pathophysiology and treatment mechanisms in diabetic macular oedema (DMO). AGE, advanced glycation products; Ang-2, Angiopoietin-2; EC, endothelial cell; C, complement; ICAM, intercellular adhesion molecule; IL, interleukin; MCP, monocyte chemotactic protein; IP, interferon-induced protein; MMP, metallic metalloproteinase; NO, nitric oxide; NOS, nitric oxide synthase; RAGE, receptor for advanced glycation products; VEGF, vascular endothelial growth factor; VE, vascular endothelial; ROS, reactive oxygen species; PDGF, platelet-derived growth factor; PKC, protein kinase C; BRB, blood–retinal barrier.

The current treatments for DMO at reducing vascular leak in the macula once it has occurred, and aim for a dry macular subsequently. No attempt is made at addressing the underlying pathology, although it is possible that some of these treatments may alter retinal function in other ways (such as neurodegeneration) in addition to reducing vascular leakage.

This review evaluates the individual treatments available, and discusses the rationale and potential for combination therapy in DMO.

Laser photocoagulation

Laser photocoagulation was the recommended treatment for DMO for several years.^{29, 30} The exact mechanisms of action of laser photocoagulation in reducing oedema in DMO are unknown. Plausible mechanisms include destruction of high-oxygen consuming photoreceptors, increased oxygenation of the retina through diffusion from the choroid, restoration of new retinal pigment epithelium (RPE) barrier, production of cytokines including transforming growth factor-β and pigment epithelium-derived factor from the stimulated RPE as discussed in the review by Bhagat et al.¹⁶ In the early treatment diabetic retinopathy study (ETDRS), DMO eyes treated with laser had lower rates of visual loss compared with control group (12% vs 24% at 3 years). This benefit was only noticeable in eyes with clinically significant DMO.^{29, 30} In eyes with diffuse DMO, response to grid laser photocoagulation was of limited benefit, with only 15% showing a visual improvement, 24% developing visual deterioration, and 61% unchanged.³¹ The average best corrected visual acuity (BCVA) change in laser-treated eyes in the diabetic retinopathy clinical research network (DRCRnet) and RESTORE (ranibizumab monotherapy or combined with laser vs laser monotherapy for DMO) studies were +2.7 to +3.2 letters at 12 months, and the fovea remained thickened in a large proportion of the laser-treated eyes. Although effective in some cases of DMO, ETDRS protocol photocoagulation may require placement of burns close to the centre of the macula. Over time, laser burns may develop into areas of progressive RPE and neuroretinal atrophy that become larger than the original laser spot size and encroach upon fixation, or subretinal membranes may occur.^{32, 33} Photocoagulation for DMO may be associated with loss of central vision, central scotomas, and decreased colour vision. In an attempt to reduce these adverse effects, many retinal specialists now treat with burns that are lighter and less intense than originally specified in the ETDRS (modified-ETDRS technique).³⁴ In the alternative approach of mild macular grid laser, mild, widely spaced burns are applied throughout the macula, avoiding the foveal region. By design, some burns could be placed in clinically normal retina if the entire retina was not abnormally thickened, including areas within the macula that are relatively distant from the area of thickening.³⁴ As such, laser photocoagulation is not advised in eyes where the leakage is close to the fovea and when the oedema is centre involving.

Subthreshold laser photocoagulation has recently been suggested as a better alternative in the treatment of DMO as the collateral damage to the retina-choroid complex is limited.^{16, 35} That is because subthreshold laser does not destroy the RPE on account of the much shorter duration. The role of subthreshold laser therapy in DMO has yet to be widely taken up, and requires further evaluation.

Pharmacologic treatments

Several pharmacologic agents are now available for the treatment of DMO including anti-VEGF agents^{36, 37, 38, 39, 40, 41} and corticosteroids.^{36, 42, 43, 44, 45, 46, 47, 48} These treatments are summarised in Table 1. These treatments are particularly useful in eyes with centre-involving DMO. Targeting VEGF has resulted in the use of anti-VEGFs, including pegaptanib, ranibizumab, aflibercept, and bevacizumab, in the treatment of DMO.^{36, 37, 38, 39, 40, 41, 43} Several clinical studies have investigated the efficacy of steroids, such as triamcinolone, fluocinolone, and dexamethasone, in the treatment of DMO.^{13, 42, 46, 47, 48, 49, 50, 51} Corticosteroids may reduce DMO by targeting the non-VEGF-dependent inflammatory pathways including blockage of the arachidonic acid pathway (reduction of prostaglandin synthesis), and inhibition of the release of pro-inflammatory mediators, including VEGF. This may be through modulation of EC TJ molecules.¹⁹

Table 1. Summary of outcomes of treatment with DMO with different agents.

Pharmacological treatment/study (study type)	Study groups	Size	VA outcomes	OCT outcomes	Adverse events
Aflibercept VIVID and VISTA (RCT, multicentre, masked, phase III, randomised)	1:1:1, aflibercept 2  mg every 4 weeks (2q4) vs 8 weeks after 5 initial monthly doses vs macular laser photocoagulation (control group; baseline laser plus sham injections every visit)	872 Eyes of type 1 and DM	Mean BCVA gains of 12.5 l and 10.7 l vs 0.2 l in the 2q4 and 2q8 groups vs the laser group at 12 m. Gain of 15 l of 41.6 and 31.1% vs 7.8% in VISTA, and 32.4 and 33.3% vs 9.1% in VIVID at 12 m	Mean CRT reduction of −185.9 and −183.1 u vs −73.3 u in VISTA, and −195.0 and −192.4 vs −66.2 u in VIVID at 12 m	Event rates (ocular and non-ocular, and SAEs) were similar in the treatment the groups
Ranibizumab (Ran) RISE and RIDE (RCT, masked, multicentre, phase III)	Ran 0.3 and 0.5 mg vs sham injection control with laser rescue	377 Eyes	Gain of >15 l in 33.6% of 0.3 mg Ran, 45.7% of 0.5 mg Ran treated vs 12.3% in laser control group at 24 m	CRT reduction of 250.6 u in 0.3 mg Ran, 253.1 u in the 0.5 mg Ran, and −133.4 u in the laser control	Slight increase in deaths and APTC events among the 0.5 mg monthly treated group cf controls and the 0.3 mg group
Ran RESOLVE (RCT, double masked, multicentre, phase II)	2 Doses Ran (0.3 and 0.5 mg) vs sham control	151 Patients	Gain of 10.3 l (11.8 l in 0.3 mg group and 8.8 l in 0.5 mg) Ran compared with −1.4 l in control at 12 m	CRT reduction by 200.7 u in the 0.3 mg Ran, 187.6 u in the 0.5 mg vs 48.4 u in the control group
Ran RESTORE (RCT, double masked, multicentre, phase III)	Laser controlled	354 Patients	BCVA score ≥15 l improvement in 22.6% and BCVA letter score level ≥73 (6/12) in 53% in the Ran-treated eyes, 22.9 and 44.9% Ran plus laser vs 8.2% and 23.6%, respectively, in the laser-only group	CRT reduction by by 118.7 u with Ran and 128.3 u with Ran plus laser compared with −63.3 u in the laser-treated group
Ran, Tri DRCRNet43 (RCT, multicentre, masked, controlled phase III)	4 Arms: 0.5 mg Ran plus prompt laser, 0.5 mg Ran plus deferred laser (≥24 weeks), 4 mg Tri plus prompt laser, or sham injection of Ran plus prompt laser Follow-up: 1–2 years	854 Eyes of 691 patients	Gain of 10 l in 50%, and 15 l in 30% of Ran and prompt laser eyes at 12 m compared with 47% and 28% in Ran and deferred laser, 33% and 21% in tri and laser, and 28% and 15% in the laser groups, respectively. Mean BCVA gain of 9.0 l in the Ran+prompt laser, or Ran+deferred laser at 12 m, cf 4 l in Tri plus laser, and +3 l in the sham injection +prompt laser control group. At 2 years, outcomes are +7, +9, +2, and +3 l, and a gain of >10 l occurred in 44%, 49%, 41%, and 36%, respectively, at 2 years. In Tri group, pseudophakia gave VA outcomes similar to Ran group. Results maintained at 24 and 36 m with deferred laser doing better	CRT improvements mirrored VA changes. CRT reduction of −102 u with prompt laser, −131 u in Ran with prompt laser, −137 u in Ran with delayed laser and −127 u with triamcinolone with prompt laser at 6 m. CRT reduction at 2 years was −138 u, −141 u, −150 u, and −107 u, respectively	Progression of cataracts and IOP elevation significant in the Tri-treated group
Bevacizumab (Bev) BOLT Study (RCT)	2 Arms, Bev vs macular laser; 12 m	80 Eyes	VA gain of +8.6 l in Bev group compared with −0.5 l in the laser control group. >10 l gain in 49% in Bev vs 7% in control, and >15 l loss in 32% control group cf to 4% Bev	CRT reduction of −146 u in the Bev group cf −118 u in the laser group
Bev Hong Kong Study56 (RCT)	2 Doses of Bev (1.25 and 2.5 mg Bev), 6 m	52 Eyes	VA gain of +5.5 l in the 1.25 mg group, and 6.5 l in the 2.5 mg group at	CMT reduction of 96 u in the 1.25 mg group cf 74 u in the 2.5 mg group
Bev Iran Study137 (RCT, masked)	2 Arms of Bev 1.25 mg with or without laser	80 Eyes of 40 patients	Mean change of BCVA of 0.138 LogMAR in the Bev group compared with 0.179 in the combination group. No difference in outcomes between the 2 groups	CMT reduction of −39 u in both groups
Pegaptanib (Peg)53 (RTC, phase 2)	4 Arms: 0.3, 1, and 3.0 mg Peg 6 weekly vs sham control	172 Eyes of 172 patients	VA gain of +4.7 l in the 0.3 and 1 mg, and +1.1 l in 3.0 mg Peg groups vs −0.4 l gain in controls at 36 weeks	CRT reduction of −68.0 , −22.7, and −5.3 u in the 0.3, 1.0, and 3.0 mg Peg, respectively, vs +3.7 u in controls
Peg54 (Multicentre, Phase II/III, RCT)	2 Arms: 0.3 mg Peg 6 weekly vs placebo controlled	260 Eyes of 260 patients	VA gain of +5.2 l in 0.3 mg Peg group cf 1.2 l in controls at 12 m, and +6.1 l and +1.3 l, respectively, at 24 m	CMT decrease of ≥25% in 31.7% vs 23.7%, and ≥50% in 14.6% and 11.9%, respectively, in the Peg and control groups at 12 m and ≥25%: 40.4% vs 44.6%, and ≥50%: 19.2% vs 26.1%, respectively, at 24 m
Dexamethasone MEAD Study (2 RCTs with identical protocols)	3 Arms: 0.7 mg, 0.35 mg Dexamethasone implants and sham treated (1:1:1), followed up for 3 years	1048 Eyes of 1048 patients	VA gain of ≥15 l was 22.2%, 18.4% and 12.0% in 0.7 mg, 0.35 Dex, and control groups, respectively	Mean reduction of CRT of −111.6 u, −107.9 u, and −41.9 u, respectively, in 0.7 mg, 0.35 mg Dex, and sham controls, respectively	Cataract 67.9%, 64.1%, and 20.4%, respectively. IOP rise requiring treatment occurred in 41.5%, 37.6% and 9.1% of the 0.7 mg, 0.35 mg Dex, and sham-treated controls. IOP increases were controlled with eye drops; 2 in the 0.7 mg and 1 in the 0.35 mg Dex groups required trabeculectomy
Fluocinolone FAME Study (Multicentre, RCT)	3 Arms placebo-controlled: Fluocinolone implant 0.2 μg/day, 0.5 μg/day, and control groups (2:2:1)	956 Eyes of 956 patients	VA gain of +4.4, +5.4, and +1.7 in the 0.2 μg, 0.5 μg, and control groups, respectively, at 24 m. ≥15 l gain in 29.0%, 29.0%, and 16.0%, respectively, in the 0.2 μg, 0.5 μg and control, groups at 24 m	CRT reduction of −167.8, −177.1, and −111.3 u in the 0.2 μg, 0.5 μg and control groups, respectively, at 24 m	Cataracts in 81.7% in 0.2 μg group cf 50.4% controls. Cataract surgery performed in 80.0% and 27.3%, respectively. IOP rise >30 mmHg in 18% of 0.2 μg Fluo group cf 4.3% in controls. IOP lowering medication required in 38.4 and 14.1%, trabeculectomy required in 4.8% of 0.2 μg and 0.5% of controls
Triamcinolone (Tri) DRCR Network Protocol (Multicentre, RCT)	3 Arms: 1 mg Tri repeated 3.5X, 4 mg, Tri repeated 3.1X, and macular laser (2.9X) over 2 years	840 Eyes of 693 patients	BCVA gain of −2 l, −3 l, and +1 l for the 1 mg, 4 mg Tri and laser groups, respectively, at 2 years. 10 l gain occurred in 25, 28, and 31%. 10 l loss occurred in 26, 28 and 19%	Reduction in CMT was −86, −77, and −139 u in the 1 mg Tri, 4 mg Tri, and laser groups, respectively, at 2 years	Progression of cataracts and elevated IOP esp in the 4 mg occurred in the Tri groups

Abbreviations: Ran, ranibizumab; Bev, bevacizumab; Peg, pegaptanib; Tri, Triamcinolone; VA, visual acuity; BCVA, best corrected visual acuity; CRT, central retinal thickness; CMT, Central Macular Thickness; IOP, intraocular pressure; APTC, Antiplatelet Trialists' Collaboration.

Ranibizumab

In the RISE and RIDE Study of two multicentre double-blind controlled trials, of intravitreal ranibizumab (0.3 and 0.5 mg; Lucentis, Genentech, South San Francisco, CA, USA/Novartis, Basel, Switzerland) was compared with laser photocoagulation, 45.7% in the ranibizumab group gained >15-letter improvement in visual acuity (VA) at 24 months compared with 12.3% in the laser-treated group.³⁹ There was a signal for increased deaths and acute thrombolytic events (as per the Antiplatelet Trialists' Collaboration criteria) in patients treated with the monthly 0.5 mg ranibizumab compared with the laser-treated group. However, these differences were not statistically significant, and have not been repeated in related studies. In RESOLVE, a 12-month, multicentre, sham-controlled, double-masked study of eyes with centre involving DMO (age 18 years, type 1 or 2 diabetes, central retinal thickness (CRT) 300 μm), patients received 3 monthly injections of ranibizumab followed by rescue laser or prn ranibizumab.³⁷ BCVA improved from baseline by a mean of 10.3±9.1 letters with ranibizumab and declined by 1.4±14.2 letters with sham (P<0.0001). A mean reduction of CRT of 194.2±135.1 μm with ranibizumab and 48.4±153.4 μm with sham (P<0.0001) was achieved. Similarly, the RESTORE Study³⁸ demonstrated superiority of the 0.5 mg ranibizumab monotherapy group or in combination with laser photocoagulation compared laser alone, with a mean of seven injections over 12 months. There was a reduction in CRT on optical coherence tomography (OCT) by −118.7 u with ranibizumab compared with −63.3 u in the laser-treated group.³⁸ In the DRCR.net Study (Protocol I), 50% of eyes treated with intravitreal ranibizumab showed >10-letter improvement at 24 weeks compared with baseline; 30% gained >15 letters.⁴³ Results were similar whether laser photocoagulation was prompt or deferred for >24 weeks. The results were maintained at 24 and 36 months, although eyes with deferred laser doing better in the longer term (2.9 letters better at 36 months).^{36, 43, 52} These VA changes were associated with parallel reductions in CRT as measured with OCT.³⁶

Aflibercept

The VIVID and VISTA Studies are two parallel studies that compared the efficacy and safety of aflibercept (Eylea, Bayer) given every 4 or 8 weeks after an initial monthly dosing for 5 months with focal laser photocoagulation at 52 weeks.⁴¹ The results showed that there was significant superiority of aflibercept in anatomical and functional end points when compared with laser. The efficacy of the 8 weekly dosing was similar that of 4 weekly dosing. The mean BCVA improvement at week 52 was 12.5 and 10.7 letters in the 4 weekly and 8 weekly treated groups, respectively, compared with 0.2 letters in the laser photocoagulation group of VISTA. In the VIVID Group, the corresponding improvements were 10.5, 10.7, and 1.2 letters, respectively.⁴¹ The mean reductions in CRT were 185.9 and 183.1 compared with 73.3 μm (P<0.0001) in VISTA and 195.0 and 192.4 compared with 66.2 μm in VIVID. The proportion of eyes gaining 15 letters or more were 41.6% and 31.1% compared with 7.8% in VISTA, and 32.4% and 33.3% compared with 9.1% in VIVID (P<0.0001).⁴¹

Pegaptanib

Two different randomised controlled studies evaluated the efficacy of pegaptanib (Macugen, Eyetech Pharma, New York, NY, USA/Pfizer, New York, NY, USA) in DMO. In the Cunningham study (a phase II study), the 0.3 and 1 mg injection groups had a better visual outcome with a gain of +4.7 letters at 6 months.⁵³ In another study (Phase II/III), Sultan et al⁵⁴ compared outcomes with 0.3 mg pegaptanib with sham. The BCVA at 2 years was significantly better in the pegaptanib-treated group (gain of +6.1 letters) compared with sham (+1.3 letters). However, there was no difference in proportions of eyes gaining 10 letters or more (38.3% pegaptanib vs 30.0% sham).⁵⁴ Pegaptanib has not received marketing authorisation for the treatment of DMO in the European Union.

Bevacizumab

Bevacizumab (Avastin, Roche/Genentech), which is unlicensed for any intraocular indication, has also been used in the treatment of DMO. The BOLT study reported a 10-letter improvement in BCVA at 24 months for eyes treated with bevacizumab compared with 7% in the multiple focal laser photocoagulation group.⁵⁵ Similarly, there was an 8.6-letter improvement in the bevacizumab-treated group compared with −0.5 letters in the laser group. More significantly, there was visual loss of more than 15 letters or more in the laser compared with the bevacizumab group.⁵⁵ Lam et al⁵⁶ compared two different doses of bevacizumab (1.25 and 2.5 mg) in DMO and found no difference in outcomes at 6 months.

Triamcinolone

Triamcinolone reduces macular oedema (in DMO), although the visual outcome is limited, and there is significant association with elevated intraocular pressure (IOP) and cataract formation.^{57, 58, 59} In addition, injections needed to be repeated at variable intervals. There was no evidence of long-term benefits of intravitreal injection of triamcinolone as treatment for DMO.^{58, 59} The commonly available Kenalog (Squibb) preparation of triamcinolone is unlicensed for intraocular administration. The preservative-free preparations of triamcinolone are not freely available. Results from the Gillies et al⁶⁰ study in Australia were not very different from those in the DRCNnet studies.

Fluocinolone

Fluocinolone 0.59 μg implant (Retisert, Bausch and Lomb, Rochester, NY, USA) resulted in an overall VA improvement of ≥3 lines (compared with sham injection/standard of care (SOC)) in 16.8% of implanted eyes at 6 months (P=0.0012), in 16.4% at 1 year (P=0.1191), in 31.8% at 2 years (P=0.0016; SOC, 9.3%); and in 31.1% at 3 years (P=0.1566; SOC, 20.0%). However, there was a significant increase in cataracts and IOP.⁴⁶

With a different design of implant, fluocinolone acetonide implant (Iluvien, Alimera Sciences, Alpharetta, GA, USA) was compared with sham injection (with laser rescue) in patients with DMO in order to evaluate the safety and efficacy of an intravitreal insert of fluocinolone acetonide for the treatment of DMO. Two doses of 0.2 and 0.5 μg of fluocinolone implant were compared with the sham implant group for 36 months. The primary outcome was the difference in the percentage of patients with BCVA improvement by ≥15 letters from baseline at month 24 between the treatment was 28.7 and 28.6 in the low- and high-dose insert groups, respectively, compared with 16.2 in the sham group (P=0.002). The mean improvement in BCVA letter score between baseline and month 24 was 4.4 and 5.4 in the 0.2 and 0.5 μg groups.⁴⁷ There was a significant reduction in macular oedema. Further analysis suggested that fluocinolone was more effective in eyes with long-standing DMO, where chronic inflammation was more important than the VEGF-mediated oedema. The fluocinolone implant lasts up to 2 years. However, there was a significant increase in the incidence of IOP elevation and cataracts.⁴⁷

Dexamethasone

In a prospective multicentre open label, 26-week study of 55 eyes to evaluate the safety and efficacy of dexamethasone intravitreal implant (Ozurdex, Allergan, Irvine, TX, USA) 0.7 mg in the treatment of DMO in vitrectomized eyes, there was a significant improvement in BCVA by 6.0 and 3.0 letters at 8 and 26 weeks, respectively, from baseline, and vascular leakage as shown by a reduction from baseline CRT (403 μm) of a mean −156 μm at week 8 (P<0.001) and −39 μm (−65, −13 μm) at week 26 (P=0.004). At week 8, 30.4% of patients had gained ≥10 letters in BVCA.⁴⁴ In their most recent report, Boyer et al⁴⁵ have evaluated the 3-year outcomes of treatment of DMO, and suggest that the dexamethasone implant was efficacious in the treatment of DMO over the medium term. The mean number of treatments given over 3 years was 4.1, 4.4, and 3.3 with dexamethasone implant 0.7 mg, dexamethasone implant 0.35 mg, and sham, respectively. The percentage of patients with ≥15-letter improvement in BCVA from baseline at study end was greater with dexamethasone implant 0.7 mg (22.2%) and dexamethasone implant 0.35 mg (18.4%) than sham (12.0% P<0.018). However, treatments needed to be given more frequently than 6 monthly, probably at 3-4 months intervals, to maintain a dry macula and retain good visual function.⁴⁵ The cataract-related adverse event rates in phakic eyes were 67.9%, 64.1%, and 20.4% in the dexamethasone implants of 0.7 mg, 0.35 mg, and sham groups, respectively. IOP rises associated with the dexamethasone implant were usually controlled with medication.

Gillies et al⁶⁰ recently reported that dexamethasone implant (Ozurdex, Allergan) was as efficacious as bevacizumab in reducing DMO, although progressive lens opacities may result in visual loss in the Ozurdex-treated group compared with those treated with bevacizumab.

Potential treatments

Angiopoietin (Ang)-2 blockage

Ang-2 is another cytokine with inflammatory and angiogenic properties that is expressed by EC. Ang-2 levels are regulated at the level of transcription by different cytokines including VEGF, as well as hypoxia and high-glucose levels,^{61, 62, 63, 64} and stored in Weibel–Palade bodies before rapid release.⁶⁵ It is elevated in eyes with clinically significant DMO^{66, 67, 68} as well as proliferative DR (PDR),^{69, 70} and is thought to induce loss of vascular endothelial (VE)-cadherin through phosphorylation.⁷¹ Ang-2 may act by exerting a permissive role for supplementing the action of other pro-inflammatory cytokines⁷² through sensitisation of ECs to tumour necrosis factor (TNF)-α and subsequent expression of ICAM-1. Similarly, in the presence of VEGF, Ang-2 increases vascular permeability.⁶¹ Recently, You et al⁷³ have reported that pan-retinal laser photocoagulation reduced the serum levels of Ang-2 and VEGF in diabetic patients with PDR compared with the significantly elevated pre-treatment levels. Another study⁷⁴ showed that the changes in serum VEGF and Ang-2 following laser photocoagulation were not significant. It has been suggested that benfotiamine, a vitamin B1 derivative may block three major pathways of hyperglycaemia damage and prevent experimental DR.⁷⁵ Yao et al⁶⁴ have suggested that high glucose increased Ang-2 transcription through methoxyglyoxal modifications. Benfotiamine has also been reported to have anti-inflammatory effects.^{76, 77} Ang-2 is not expressed by pericytes; pericytes express Ang-1 (and not Ang-2), which promote EC survival signalling, maintenance of endothelial barrier and vascular maturation and quiescence^{78, 79, 80, 81} and reduces VEGF-stimulated leukocyte adhesion to ECs.⁸² Taurine has been reported to inhibit high glucose-induced apoptosis in retinal pericytes through increased Tie-2 expression and downregulation of Ang-2.⁸³

Sirolimus

Sirolimus (rapamycin), a macrolide antibiotic, has immunosuppressive and anti-proliferative activities through its inhibition of the mammalian target of rapamycin (mTOR).⁸⁴ Sirolimus, through its inhibition of mTOR, inhibits VEGF expression and signalling, and downregulates HIF-1A.^{85, 86, 87, 88, 89, 90} In addition, sirolimus has been shown to inhibit VEGF-induced microvascular permeability, and reduces protein kinase C (PKC) activity,^{85, 86, 87, 88, 89, 90, 91} as well as reduces the expression of inflammatory cytokines.^{92, 93} Recently, a phase 1 study of subconjunctival and intravitreal sirolimus in DMO was reported by Dugel et al.⁹⁴ No non-ocular adverse events were noted, and the systemic exposure to sirolimus was low. There were no serious ocular adverse events reported. A median BCVA improvement of 4.0 letters was noted at day 14 and maintained for 90 days, associated with a reduction in the CRT of −33 u at day 14 and −68.3 u at day 90. Similarly, Krishnadev et al⁹⁵ in their phase I/II study reported sirolimus to be safe when administered subconjunctivally, although the treatment effect was variable. Progress in the development of sirolimus formulations for intraocular delivery (DE-109, Santen Pharma, Osaka, Japan) is encouraging.

Combination therapy

Scientific basis

The available evidence suggests that each individual treatment modality in DMO does not result in a completely dry macula in most cases. Furthermore, it takes a while to achieve a dry macula in most cases, requiring frequent treatments. The ideal treatment for DMO should improve vision and improve morphological changes in the macular (eg, reduce macular oedema) for a significant duration, reduced adverse events, reduce treatment burden and costs, and be well tolerated by patients.

It is known that retinal vascular leakage in DMO is contributed to by VEGF upregulation and non-VEGF-dependent inflammatory pathways, which may be chronic and subclinical.^{18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28} Accumulating evidence also suggests that chronic subclinical inflammation has an important role in the pathogenesis of DR. An early event in the pathogenesis of diabetic vasculopathy is leukocyte adherence to retinal vascular endothelium, resulting in EC death, vascular leakage, and capillary closure.¹⁸

Clinical evidence indicates that DMO results from breakdown of the inner blood–retinal barrier,⁹⁶ which is maintained by EC TJ complexes.⁹⁷ TJ proteins include occludins, claudins, and junctional adhesion molecules (JAMs), transmembrane molecules, which interact with intracellular molecules including zonula occludens (ZO) and cingulin. Adherens junctions are formed by VE-Cadherin, which promotes calcium-dependent homophilic cell to cell contacts, and interact intracellularly with actin through catenins. Dejana et al⁹⁸ have reported that the principal TJ proteins found in retinal ECs (RECs) are occludin and claudin-5. Phosphorylation of occludin and ZO1 contribute to the dysregulation of EC paracellular permeability.⁹⁹ It has been suggested that the hyper-permeability induced by diabetes and hyperglycaemia is VEGF mediated.^{100, 101} VEGF-A is thought to be the main factor responsible for increased retinal vascular permeability as upregulation of it results in phosphorylation of TJ proteins (occludin and ZO) and their disassembly.^{102, 103} Exposure of RECs to hyperglycaemia has also been reported to result in increased MMP-9 levels, which lead to active proteolytic degradation of EC TJs especially occludin.¹⁰⁴ Mukarami et al¹⁰⁵ and Felinski et al¹⁰¹ have attempted to elucidate the mechanisms of VEGF enhancement of vascular permeability. Other factors including TNFα and interleukin (IL)-1B,¹⁰⁶ nuclear factor-κB (NF-κB) receptor activator,^{21, 22} hepatocyte growth factor (HGF),²³ and the accumulation of advanced glycation end products²⁰ are known to induce vascular leakage independent of VEGF-A. Adhesion molecules such as CD44^{24, 25} may also contribute significantly to microvascular occlusions and angiogenesis.

JAMs also signal to cytoskeleton-associated proteins, as well as recruiting cellular polarity proteins. The preferential expression of different TJ and adherens junction molecules in different vascular beds may explain their specific physiologic functions, for example, the high expression of JAM-A in the blood–brain barrier (blood–retinal barrier),¹⁰⁷ and the relative high expression of JAM-A in the liver compared with its minimal expression of JAM-B and C, substitution or redundancy.¹⁰⁸ JAM-C is highly expressed in inter-EC junctions in lymph nodes.¹⁰⁹ Different JAMs including JAM-A, -B, -C are altered in diabetes and hyperglycaemia.¹¹⁰ Furthermore, different capillary beds may respond differently to different stimulatory or inhibitory factors.¹¹¹ Similarly, they may differ in their basal and TNFα-regulated expression of adhesion molecules and cytokines.¹¹² This is supported by the observation that although DR is a microvascular disease microangiopathy does not seem to have a major pathogenic role in the cerebral complications of diabetes and the blood–-brain barrier is not as susceptible to hyperglycaemia as the retinal microvasculature.¹¹³

It has been reported that hyperglycaemia can stimulate both NF-κB- and PKC-dependent pathways thus upregulating endothelial adhesion molecules, such as ICAM-1 (CD54) or E-Selectin (CD62E),^{114, 115, 116} vascular cell adhesion molecule-1 (CD106),¹¹⁷ and platelet endothelial cell adhesion molecule (CD31), in ECs. ICAM-1 upregulation is thought to be VEGF dependent, whereas P- and E-Selectin are not.¹¹⁸ Vitreous levels of several inflammatory cytokines (including IL-6, IL-8, and monocyte chemotactic protein (MCP)-1) are increased in eyes with DMO and PDR.¹¹⁹ In other studies, Funatsu et al^{120, 121, 122} have reported significant elevations of VEGF, ICAM-1, IL-6, MCP-1, and pigment epithelium-derived factor, which correlated with retinal vascular permeability changes. In addition to this pathway, stimulation of 6-N-acetyl glucosaminyl-transferase in leukocyte can result in more modification of glycans on its surface thus inducing leukocyte rolling interactions to endothelial selectins.^{123, 124, 125}

Nehme and Edelman¹²⁶ have reported that hyperglycaemia, TNFα, and IL-1 induce an inflammatory phenotype in human retinal perictyes, characterized by hypersecretion of inflammatory and angiogenic mediators. No such change was noted in the retinal microvascular ECs in this study. Dexamethasone at various potencies blocked hypersecretion of several proteins in pericytes.¹²⁶ It was further suggested that corticosteroids may inhibit PLA2 and downstream products, for example, prostaglandins and leukotrienes, as well as inhibit adhesion signalling pathway components, including ICAM-1, IL-6, VEGF-A, and stromal cell-derived factor-1, which are involved in vascular permeability and leucocytosis (inflammation).¹²⁷ Steroids may also inhibit TJ disassembly. Further elucidation of the cellular mechanisms underlying increased permeability in diabetes and hyperglycaemia, and the effects of different pharmacologic agents and their interactions is required.

Corticosteroids have significant effects on RECs.^{101, 128} It has been suggested that steroids may act by both suppressing inflammation and by directly affecting the ECs by regulating phosphorylation, organisation, and content of TJ proteins.¹⁰¹ Felinski et al¹²⁸ have reported that glucocorticoid treatment of primary RECs increases content of the TJ proteins occludin and claudin-5, in parallel with an increase in barrier properties of endothelial monolayers, therefore restoring endothelial paracellular permeability. However, the exact mechanisms of action are poorly understood.

Microarray analysis studies in our laboratory have determined that occludin and claudin-5 gene expression was higher in RECs than CECs under normal glucose culture conditions.¹¹⁰ We have further determined that there are differences in HGF expression between RECs and CECs, as well as c-Met (the HGF receptor). Specifically, CEC secrete HGF at higher levels than REC.¹²⁹ These results would support the concept that CEC and REC behave differently under hyperglycaemia, and may explain, at least in part, some of the clinical differences observed in the frequencies/prevalences of DR and choroidopathy.^{6, 7, 8, 9} Western blotting indicated that claudin-5 protein expression was also higher in RECs, whereas JAM-A and C and VE-Cadherin levels were similar.¹¹⁰ High-glucose conditions (equivalent to hyperglycaemia) significantly increased the permeability in both REC and CEC monolayers, although the increase was higher for RECs.¹¹⁰ In RECs exposed to high-glucose claudin-5, occludin, and JAM-A were found to be reduced, and JAM-C increased, whereas the expression of VE-Cadherin was relatively unchanged when evaluated with western blotting, immunofluorescence, and qPCR.

Recent studies have reported that there are significant changes in the cytokine profile in DMO. The elevated cytokines include VEGF, MCP-1, IL-6, IL-8, interferon γ-induced protein-10.^{119, 130, 131, 132} Funatsu et al^{120, 121, 122} have also previously reported increased levels of ICAM-1, IL-6, Angiotensin II, and VEGF in DR and DMO. Klein et al¹³³ reported that sVCAM-1 and TNFα are associated with higher odds of developing more severe DR in patients with renal disease and PDR. Other evidence suggests that anti-VEGF treatments have no effect on the cytokines other than VEGF levels in intraocular fluids.^{130, 131, 132}

The available evidence therefore implies that targeting VEGF alone or VEGF-independent pathways in isolation may not adequately block the increased permeability associated with diabetes and hyperglycaemia. A combination approach may be necessary in some cases. Steroids and anti-VEGF may have complementary and additive effects on retinal vascular permeability. This follows from the evidence of Nehme and Edelman¹²⁶ on the effect of dexamethasone on REC and pericyte permeability, and the reports from Felinski,^{101, 128} in concert with the anti-VEGF trials. The concept of combination pharmacotherapy in DMO has been given further impetus by preliminary results from our laboratory, which show that a combination of dexamethasone and ranibizumab reduced REC permeability more than either agent alone in vitro.¹¹⁰

Clinical evidence

A few trials have evaluated the combination of bevacizumab with triamcinolone.^{134, 135, 136, 137} Ahmadieh et al¹³⁴ evaluated bevacizumab alone in comparison to bevacizumab and triamcinolone, and determined that there was no further improvement with the combination compared with bevacizumab alone. Sohelian et al^{135, 136} compared combination of 1.25 mg bevacizumab and 2 mg triamcinolone with bevacizumab alone and laser photocoagulation alone. Bevacizumab alone improved BCVA at 36 months, compared with either the combination or the laser alone group.^{135, 136} Faghihi et al¹³⁷ compared intravitreal bevacizumab (1.25 mg) alone with combination of bevacizumab and focal laser, and determined at 6 months that both regimes had comparable vision improvement and that neither dosage was superior to the other. The outcomes reported by Faghihi et al¹³⁷ that combinations of bevacizumab and focal laser photocoagulation induced earlier recovery of VA compared with bevacizumab monotherapy, although this was not confirmed by Soheilian et al.¹³⁸ Sheth et al¹³⁹ did not find any difference in DMO reduction or post injection VA in eyes treated with intravitreal triamcinolone vs dexamethasone and bevacizumab. However, that is most likely because of the short duration of the dexamethasone used, as well as the small sample size.¹³⁹ This concept has been given further impetus by preliminary results from our laboratory, which show that a combination of dexamethasone and ranibizumab reduced REC permeability more than either agent alone in vitro.¹¹⁰ In the DRCRNet Study,^{43, 52} intravitreal 0.5-mg ranibizumab or 4-mg triamcinolone combined with focal/grid laser was compared with focal/grid laser alone for treatment of DMO. At 24 months, the mean change in the VA letter score from baseline was 3.7 letters greater (95% CI: −0.4 to +7.7) in the ranibizumab plus prompt laser group, 5.8 letters greater in the ranibizumab plus deferred laser group (95% CI: +1.9 to +9.8) and 1.5 letters worse in the triamcinolone plus prompt laser group (95% CI: −5.5 to +2.4).³⁶ These VA changes were associated with parallel reductions in CRT as measured with OCT.³⁶ Combination of dexamethasone implant with laser photocoagulation has been reported to result in more reduction in macular oedema and greater BCVA improvement up to 9 months compared with either agent on its own.¹⁴⁰ No significant difference was noted at 12 months in the number of eyes that gained 10 letters or more, however.

The way forward

It is possible that an initial combination of dexamethasone and anti-VEGF therapy (ranibizumab or aflibercept) will allow a faster VA rise to the plateau than achieved with anti-VEGF monotherapy, may induce a longer remission of DMO, and more permanently modulate the recurrence of DMO than after either agent on its own. This is important as complete dryness of the macular is only achieved in about 50–75% of eyes treated with anti-VEGF monotherapy. Furthermore, such a combination will reduce potential systemic adverse events as may be associated with frequent anti-VEGF therapies. As the effect of the dexamethasone implant only lasts 3–6 months, any associated IOP rise is short-lived. The longer-acting steroid, fluocinolone may also have a role in combination with anti-VEGF therapies especially in eyes that are pseudophakic. This follows from the fact that fluocinolone is more effective in eyes with longer duration of DMO than those with more recent onset. This is despite the side-effect profile of steroids including cataracts, and increased IOP. Potentially, if such a combination works, only one or two dexamethasone implant injections will be required, but at the same time reducing the frequency of intravitreal anti-VEGF injections.

Macular laser photocoagulation (focal) applied after treatment with anti-VEGF and/or dexamethasone implant (deferred laser) allows lower energy laser applications thus reducing collateral damage. This, as seen in the DRCRnet studies, gives better outcomes and results in reduction of the number of intravitreal injections required, with increased BCVA gain (2.9 letters better at month 36).⁵²

In the future, mTOR inhibition by subconjuctival or intravitreal sirolimus, or inhibition of Ang-2 alongside anti-VEG therapies may also more result in significant benefit in DR. For the anti-Ang-2 drug, the mode of delivery will, however, depend on the particular properties.

Similarly, other drugs that may interfere with any of the mechanisms in the development of DR, for example, benfotiamine are worth exploring as adjuncts to other therapies in DMO.

Fovista is an anti-platelet-derived growth factor agent currently under development as an intravitreal agent for the treatment of neovascular age-related macular degeneration in combination of anti-VEGF drugs.¹⁴¹ At the present time, it is not envisaged that it will have a role in the treatment of DMO, until it is shown to be unlikely to induce more pericyte loss in the diabetic retina. Furthermore, systemic platelet-derived growth factor-B was not found to be significantly upregulated in patients with DMO when compared with normal individuals and the significant upregulation in neovascular age-related macular degeneration.¹⁴²

Further elucidation of the cellular mechanisms underlying increased permeability in diabetes and hyperglycaemia, and the effects of different pharmacologic agents and their interactions are required. Similarly, cellular and physiological changes in the choroidal vasculature in diabetes require further investigation.

Systemic control of hyperglycaemia (Hb A1C), blood pressure and serum lipids, as well as the addition fenofibrates and/or statins in diabetics cannot be over-emphasised as these have a significant effect on the fluid leakage in the macula.^{143, 144, 145, 146}

Conclusion

This review has evaluated the individual treatments available for DMO, and discussed the rationale and potential for combination therapy in DMO. Systemic factors including blood sugar, blood pressure, and lipids cannot be over-emphasised.

The ideal treatment for DMO should improve vision and morphological changes in the macular (eg, reduce macular oedema) for a significant duration, reduce adverse events, reduce treatment burden and costs, and be well tolerated by patients. Laser photocoagulation is of limited use in centre-involving DMO, and requires augmentation in eyes where the foveal centre is threatened. No current individual treatment modality in DMO results in a completely dry macula in most cases. This is understandable as there are multiple mechanisms including VEGF pathway, and the non-VEGF inflammation, which becomes more important in the chronic stages of DMO. Eyes with chronic DMO are more likely to befit from intravitreal steroid therapy. A combination of anti-VEGF therapies with steroids will provide more optimal pharmacotherapy, and may be supplemented with macular laser photocoagulation as necessary in individual cases. Newer agents in the future will further enhance efficacy of our treatments for DMO.

WMK Amoaku has provided consultancy services to Alcon, Allergan, Bayer, Novartis and Thrombogenics. He has received travel grants from Allergan, Bayer and Novartis, and honoraria for lectures from Allergan, Novartis. He has participated in clinical trials for which his institution has received funding from Allergan, Novartis and Pfizer. His institution has further received research grants from Allergan and Novartis for non-clinical studies, and CentreVue (Italy) for clinical studies. The remaining authors declare no conflict of interest.