Anti‐vascular endothelial growth factors in combination with vitrectomy for complications of proliferative diabetic retinopathy

Abstract

Background

Vitrectomy is an established treatment for the complications of proliferative diabetic retinopathy (PDR). However, a number of complications can occur during and after vitrectomy for PDR. These include bleeding and the creation of retinal holes during surgery, and bleeding, retinal detachment and scar tissue on the retina after surgery. These complications can limit vision, require further surgery and delay recovery. The use of anti‐vascular endothelial growth factor (anti‐VEGF) agents injected into the eye before surgery has been proposed to reduce the occurrence of these complications. Anti‐VEGF agents can reduce the amount and vascularity of abnormal new vessels associated with PDR, facilitating their dissection during surgery, reducing intra‐ and postoperative bleeding, and potentially improving outcomes.

Objectives

To assess the effects of perioperative anti‐VEGF use on the outcomes of vitrectomy for the treatment of complications for proliferative diabetic retinopathy (PDR).

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL; which contains the Cochrane Eyes and Vision Trials Register; 2022, Issue 6); Ovid MEDLINE; Ovid Embase; the ISRCTN registry; ClinicalTrials.gov and the WHO ICTRP. The date of the search was 22 June 2022.

Selection criteria

We included randomised controlled trials (RCTs) that looked at the use of anti‐VEGFs and the incidence of complications in people undergoing vitrectomy for PDR.  

Data collection and analysis

Two review authors independently assessed and extracted the data. We used the standard methodological procedures expected by Cochrane.

The critical outcomes of the review were the mean difference in best corrected visual acuity (BCVA) between study arms at six (± three) months after the primary vitrectomy, the incidence of early postoperative vitreous cavity haemorrhage (POVCH, within four weeks postoperatively), the incidence of late POVCH (occurring more than four weeks postoperatively), the incidence of revision surgery for POVCH within six months, the incidence of revision surgery for recurrent traction/macular pucker of any type and/or rhegmatogenous retinal detachment within six months and vision‐related quality of life (VRQOL) measures. Important outcomes included the proportion of people with a visual acuity of counting fingers (1.8 logMAR or worse), the number of operative retinal breaks reported and the frequency of silicone oil tamponade required at time of surgery.

Main results

The current review includes 28 RCTs that looked at the pre‐ or intraoperative use of intravitreal anti‐VEGFs to improve the outcomes of pars plana vitrectomy for complications of PDR. The studies were conducted in a variety of countries (11 from China, three from Iran, two from Italy, two from Mexico and the remaining studies from South Korea, the UK, Egypt, Brazil, Japan, Canada, the USA, Indonesia and Pakistan). The inclusion criteria for entry into the studies were the well‐recognised complications of proliferative retinopathy: non‐clearing vitreous haemorrhage, tractional retinal detachment involving the macula or combined tractional rhegmatogenous detachment. The included studies randomised a total of 1914 eyes. 

We identified methodological issues in all of the included studies. Risk of bias was highest for masking of participants and investigators, and a number of studies were unclear when describing randomisation methods and sequence allocation.

Participants receiving intravitreal anti‐VEGF in addition to pars plana vitrectomy achieved better BCVA at six months compared to people undergoing vitrectomy alone (mean difference (MD) ‐0.25 logMAR, 95% confidence interval (CI) ‐0.39 to ‐0.11; 13 studies, 699 eyes; low‐certainty evidence).

Pre‐ or intraoperative anti‐VEGF reduced the incidence of early POVCH (12% versus 31%, risk ratio (RR) 0.44, 95% CI 0.34 to 0.58; 14 studies, 1038 eyes; moderate‐certainty evidence).

Perioperative anti‐VEGF use was also associated with a reduction in the incidence of late POVCH (10% versus 23%, RR 0.47, 95% CI 0.30 to 0.74; 11 studies, 579 eyes; high‐certainty evidence).

The need for revision surgery for POVCH occurred less frequently in the anti‐VEGF group compared with control, but the confidence intervals were wide and compatible with no effect (4% versus 13%, RR 0.44, 95% CI 0.15 to 1.28; 4 studies 207 eyes; moderate‐certainty evidence). Similar imprecisely measured effects were seen for revision surgery for rhegmatogenous retinal detachment (5% versus 11%, RR 0.50, 95% CI 0.15 to 1.66; 4 studies, 145 eyes; low‐certainty evidence). 

Anti‐VEGFs reduce the incidence of intraoperative retinal breaks (12% versus 31%, RR 0.37, 95% CI 0.24 to 0.59; 12 studies, 915 eyes; high‐certainty evidence) and the need for silicone oil (19% versus 41%, RR 0.46, 95% CI 0.27 to 0.80; 10 studies, 591 eyes; very low‐certainty evidence).

No data were available on quality of life outcomes or the proportion of participants with visual acuity of counting fingers or worse.

Authors' conclusions

The perioperative use of anti‐VEGF reduces the risk of late POVCH, probably results in lower early POVCH risk and may improve visual outcomes. It also reduces the incidence of intraoperative retinal breaks. The evidence is very uncertain about its effect on the need for silicone oil tamponade. The reported complications from its use appear to be low. Agreement on variables included and outcome standardisation is required in trials studying vitrectomy for PDR.

Plain language summary

Anti‐VEGF for prevention of complications after vitrectomy for proliferative diabetic retinopathy

Why is this question important?

Proliferative diabetic retinopathy (PDR) is the most advanced stage of diabetic retinopathy, which is a complication of diabetes. It is caused by abnormal new blood vessels that grow on the retina, the light sensitive layer at the back of the eye. Pars plana vitrectomy is an established surgical treatment for the complications of PDR. During this operation, the gel‐like substance that exists inside the eye, called the vitreous, is removed. The most common reason for performing vitrectomy is non‐clearing vitreous haemorrhage, where there is bleeding in the inner part of the eye called the vitreous cavity. When this occurs, vitreous loses transparency and the light cannot pass through it, causing vision loss. Anti‐vascular endothelial growth factor agents (anti‐VEGFs) are drugs that can stop these abnormal blood vessels growing and control the leaking blood, when they are injected inside the eye. The injection of anti‐VEGF agents before vitrectomy for the complications of PDR may make surgery easier, reduce intra‐ and postoperative bleeding and improve outcomes. However, there is no clear evidence about the different types of anti‐VEGF agents, the best time to use them or their effect on other outcomes. We reviewed published research to determine whether anti‐VEGF injections around the time of surgery have an effect on the outcomes of vitrectomy for the treatment of complications of PDR.

What did we want to find out?

We wanted to find out if the additional use of anti‐VEGF agents either before or during diabetic vitrectomy surgery was better than diabetic vitrectomy alone in terms of:

‐ visual outcomes;
‐ incidence of early and late POVCH;
‐ incidence of revision surgery for POVCH in the first six months postoperatively;
‐ incidence of revision surgery for recurrent traction/macular pucker (structural alteration in the macula, the part of the eye which is responsible for sharp, central vision, thus causing decline of visual function) in the first six months postoperatively;
‐ vision‐related quality of life measures;
‐ proportion of people with poor visual acuity (counting fingers or worse);
‐ number of intraoperative retinal breaks (tears in the light sensitive layer of the eye);
‐ frequency of silicone oil tamponade (a tamponade agent used to push the retina towards the eye wall and keep it stable). 

What did we do?

We searched for studies that compared any anti‐VEGF combined with vitrectomy versus vitrectomy with no anti‐VEGF or with sham treatment or with any other anti‐VEGF in people undergoing vitrectomy for the complications of PDR. We compared and summarised the results of the studies and rated our confidence in the evidence, based on factors such as study methods and sizes.

What did we find?

We found 28 studies that involved 1914 eyes undergoing vitrectomy for complications of proliferative diabetic retinopathy. Sample sizes varied from 20 to 214 participants. The studies were from China (11), Iran (three), Italy (two) and Mexico (two), and the remaining studies were from South Korea, UK, Egypt, Brazil, Japan, Canada, USA, Indonesia and Pakistan. One study was multi‐centre, including 13 clinical sites from nine countries. Two studies declared receiving funding from the manufacturer of the anti‐VEGF drug studied. 

Main results 

The additional use of anti‐VEGFs in diabetic vitrectomy may improve visual outcomes at six months postoperatively. Their use also reduces the risk of late POVCH and probably reduces the risk of early POVCH. Anti‐VEGFs reduce intraoperative retinal breaks and may reduce the requirement for silicone oil tamponade.

Anti‐VEGF use probably results in a lower need for revision vitrectomy for POVCH, and may reduce the need for revision surgery for recurrent traction and/or retinal detachment.

We found no studies to help us answer our question about quality of life and the proportion of people with visual acuity of counting fingers or less.

What are the limitations of the evidence?

We are confident that anti‐VEGF injections reduce the incidence of late POVCH and intraoperative retinal breaks (high‐certainty evidence), fairly certain that they reduce the incidence of early POVCH and the need for revision surgery for POVCH (moderate‐certainty evidence) but less certain about the effects on visual outcomes, revision surgery for retinal detachment and silicone oil use (low‐certainty evidence).

How up‐to‐date is this evidence?

This review updates our previous review. The evidence is up‐to‐date to 22 June 2022.

Summary of findings

Summary of findings 1. Summary of findings.

Anti‐vascular endothelial growth factors in combination with vitrectomy for complications of proliferative diabetic retinopathy
Patient or population: people with proliferative diabetic retinopathy undergoing vitrectomy Settings: hospital Intervention: anti‐VEGF Comparison: no anti‐VEGF
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Anti‐VEGF
Mean BCVA (logMAR) 6 ± 3 months   BCVA logMAR score of 0 corresponds to a Snellen score of 6/6. Negative logMAR scores correspond to BCVA better than Snellen 6/6, whereas higher positive logMAR scores correspond to BCVA worse than Snellen 6/6.	The mean visual acuity ranged across control groups from 0.51 to 2.07 logMAR	The mean visual acuity ranged across anti‐VEGF groups from 0.29 to 1.10 logMAR	MD ‐0.25 (‐0.39 to ‐0.11)	699 (13)	⊕⊕⊝⊝ Low1	Anti‐VEGF administration may result in an improvement in visual acuity postoperatively compared to control
Early POVCH	312 per 1000	137 per 1000 (106 to 181)	RR 0.44 (0.34 to 0.58)	1038 (14)	⊕⊕⊕⊝ Moderate2	Anti‐VEGF administration probably results in fewer cases of early POVCH compared to control
Late POVCH	231 per 1000	108 per 1000 (69 to 171)	RR 0.47 (0.30 to 0.74)	579 (11)	⊕⊕⊕⊕ High3	Anti‐VEGF administration results in fewer cases of late POVCH compared to control
Revision surgery for POVCH	130 per 1000	57 per 1000  (20 to 166)	RR 0.44 (0.15 to 1.28)	207 (4)	⊕⊕⊕⊝ Moderate4	Anti‐VEGF administration probably results in fewer cases of revision surgery for POVCH compared to control
Revision surgery for recurrent traction and/or RRD	114 per 1000	57 per 1000 (17 to 189)	RR 0.50 (0.15 to 1.66)	145 (4)	⊕⊕⊝⊝ Low5	Anti‐VEGF administration may result in fewer cases of revision surgery for recurrent traction and/or RRD compared to control
Intraoperative retinal breaks	311 per 1000	115 per 1000  (75 to 183)	RR 0.37 (0.24 to 0.59)	915 (12)	⊕⊕⊕⊕ High6	Anti‐VEGF administration results in fewer cases of intraoperative retinal breaks compared with control
Silicone oil use	412 per 1000	190 per 1000  (111 to 330)	RR 0.46 (0.27 to 0.80)	591 (10)	⊕⊝⊝⊝ Very low7	Anti‐VEGF administration may result in less use of silicone oil compared with control
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BCVA: best corrected visual acuity; CI: confidence interval; MD: mean difference; RR: risk ratio; POVCH: postoperative vitreous cavity haemorrhage; RRD: rhegmatogenous retinal detachment; TRD: tractional retinal detachment; VEGF: vascular endothelial growth factor
GRADE Working Group grades of evidence High certainty: Further research is very unlikely to change our confidence in the estimate of effect. Moderate‐certainty: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low‐certainty: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low‐certainty: We are very uncertain about the estimate.

¹Downgraded one level for inconsistency (I² = 75%) and one level for study limitations because most of the studies were graded as unclear/high risk of bias.

²Downgraded one level for possible publication bias (asymmetry in funnel plot). Many studies in this analysis suffer from poor masking, however this probably is not an issue for this particular outcome and thus we decided not to downgrade another level.

³Similarly, many studies in this analysis suffer from poor masking, however this probably is not an issue for this particular outcome and thus we decided not to downgrade for risk of bias.

⁴Downgraded one level for imprecision (confidence intervals include 1 ‐ no effect).

⁵Downgraded one level for imprecision (confidence intervals include 1 ‐ no effect) and downgraded one level because the majority of the included studies in the analysis suffer from serious risk of bias (selection and detection bias).

⁶A few studies in this analysis suffer from questionable masking. However, we do not feel that this might have affected this particular outcome (i.e. surgical performance/difficulty of surgery).

⁷Downgraded one level for inconsistency (minimal overlap of confidence intervals between studies), one level for possible publication bias (asymmetry in funnel plot) and one level for study limitation (questionable masking).

Background

Pars plana vitrectomy is an established and successful treatment for the complications of proliferative diabetic retinopathy (PDR) (Ho 1992; Jackson 2015; McLeod 1991). Overall, up to 10% of people presenting with PDR require vitrectomy within one year, depending on the study and the patients investigated (Kaiser 2000; Vaideanu 2014). The most common indication for surgery is non‐clearing vitreous haemorrhage (Ho 1992; McLeod 1991). Other common indications are macula‐involving or macula‐threatening tractional retinal detachment (TRD) or combined rhegmatogenous/tractional retinal detachment (Jackson 2015).

Description of the condition

The aim of vitrectomy is to maximise visual function in the short and long term. However, diabetic vitrectomy can be a complex surgical procedure, particularly if delamination and/or segmentation of the fibrovascular membranes is required and there is retinal detachment present (Hutton 1980; Jackson 2015). Surgery can be complicated by a range of intra‐ and postoperative problems that can affect the visual outcome and limit recovery. These can be summarised as follows: 

1. Intra‐ and early postoperative bleeding causing early postoperative vitreous cavity haemorrhage (POVCH).

2. Retinal break formation during surgery and postoperative rhegmatogenous retinal detachment.

3. Persistent or recurrent traction causing postoperative macular traction or tractional retinal detachment requiring repeat vitrectomy.

4. Recurrent intraocular bleeding later following surgery causing late POVCH.

5. Retinal status at the end of surgery requiring the insertion of silicone oil as tamponade, which can be associated with a range of postoperative problems including visual loss.

Aetiology of complications

Postoperative vitreous cavity haemorrhage (POVCH)

POVCH can be a significant complication, occurring in approximately 20% to 30% of cases, although the reported range is large (5% to 80% of cases) (Benson 1988; Blankenship 1986; Liggett 1987; Nagbal 2017; Novak 1984; Tolentino 1989; Virata 2001; Yorston 2008).

Although there can be overlap, POVCH occurs in two main forms:

1. Early POVCH, being present from the first few postoperative days and delaying visual recovery by non‐clearance.

2. Late POVCH, occurring later during follow‐up, commonly at two to six months postoperatively, after a postoperative period during which the vitreous cavity was clear.

Persistent haemorrhage can result from operative and postoperative oozing of the remnants of new vessels or dissected tissue, or directly from the sclerostomies used to perform surgery. It can also occur from clot lysis in the first few postoperative days. Leaching of red blood cells can occur from retained old haemorrhages in residual anterior vitreous, causing apparent POVCH (McLeod 1991; Novak 1984; Tolentino 1989). Intraocular pressure at the end of vitrectomy has also been reported to have an effect on anatomical and visual recovery. Higher pressures were reported to result in better visual and anatomical outcomes (Zhou 2020). Consequently, ensuring waterproof sclerostomies at the end of surgery, even if that means suturing them when in doubt, can be beneficial in preventing postoperative hypotony.

Recurrent POVCH can occur from recurrent traction on residual new vessels, or indeed postoperative new vessel growth in the posterior retina. PDR is well known to be related to a range of angiogenic factors, such as vascular endothelial growth factor (VEGF), and these can remain raised postoperatively (Stefansson 2009), with postoperative new vessel growth in the posterior retina. Studies have also shown that a common cause of recurrent haemorrhage is new anterior vessel growth at the inner sclerostomy sites associated with fibrous traction (Bhende 2000; Hershberger 2004; Hotta 2000; Kreiger 1993; Sawa 2000; Steel 2008; West 2000). Termed 'entry site neovascularisation' (ESNV) (McLeod 2000), this is thought to be an aberrant wound‐healing response related to the presence of retinal and pars plana ischaemia (Kreiger 1993; Yeh 2005). The presence of ESNV is difficult to observe clinically because of the extreme anterior location but can be confirmed at the time of revision surgery with deep scleral indentation or endoscopic techniques (West 2000). It can also be localised, with anterior segment high‐resolution ultrasonography of the inner sclerostomy sites (Bhende 2000; Hershberger 2004; Hotta 2000; Steel 2008).

Retinal breaks and postoperative rhegmatogenous retinal detachment

Retinal breaks may exist at the time of vitrectomy in the case of combined rhegmatogenous tractional retinal detachment but can also be caused by surgery. They can occur in three main types:

1. Posterior breaks related to traction and membrane dissection.

2. Peripheral surgical entry site breaks partly related to instrument exchange at the pars plana entry ports. These have significantly reduced with the advent of cannulated transconjunctival vitrectomy systems (Issa 2011).

3. Retinal breaks created as a result of  vitreous separation, an important part of surgery. Occasionally, they are associated with predisposing retinal lesions including lattice degeneration.

If inadequately treated or associated with persistent or recurrent traction, retinal breaks can result in retinal detachment.   

Persistent or recurrent traction

Preretinal traction can be complex to completely remove and can be occult with the phenomenon of vitreoschisis, a splitting of the posterior vitreous cortex into multiple layers with adherence to the retina, as a consequence of diabetes (Kroll 2014; Sebag 1996). Epiretinal membrane is observed postoperatively after diabetic vitrectomy in 20% to 50% of cases; it can be associated with macular thickening and recurrent tractional detachment, can affect visual outcome and has been reported to require revision surgery in 7% to 22% of cases (Blankenship 1985; Gupta 2012; Mehta 2020; Oshima 2009; Schiff 2007; Yorston 2008). The exact cause of this is uncertain, although surgical and proteomic risk factors have been described (Hsu 2014; Yoshida 2017).

Consequences and management of complications after occurrence

POVCH

People with POVCH suffer a delay in their visual recovery which, in some cases, results in a worse level of visual acuity than preoperatively. Intraocular pressure can become raised from trabecular meshwork obstruction. If maculopathy is present, then additional laser treatment cannot be applied and intravitreal treatment may not be possible to continue, with the risk of worsening foveal function in the long term. Repeated clinic visits may be required to monitor POVCH, which may be a significant burden both for patients and health systems.

The initial treatment for POVCH is observation (Tolentino 1989). Spontaneous clearance occurs in some cases if no further bleeding occurs. This is because the red blood cells are no longer trapped in the gel structure of the vitreous and can circulate and clear more freely from the vitreous cavity. Clearance is related to the amount of haemorrhage, its recurrence and the degree of communication between anterior and posterior segments, allowing red blood cells to enter the anterior chamber and be cleared via the trabecular meshwork (McLeod 2000).

Non‐clearing POVCH necessitates revision surgery in approximately one‐third to one‐half of those who experience POVCH and approximately 10% to 20% of all patients undergoing surgery (Brown 1992; Han 1991; Martin 1992; Jackson 2015; Steel 2008; Yorston 2008). Non‐clearing POVCH is treated with surgery to remove haemorrhage and to treat any underlying cause that may have been identified. The timing of the surgery will depend on a variety of factors, including social situation, fellow‐eye status, maculopathy needing treatment, intraocular pressure, etc. Simple in‐office fluid‐air exchange has been proposed as an alternative to traditional revision surgery (Berrocal 2016).

In some patients, spontaneous clearing or revision surgery is followed by recurrent haemorrhage, which frustrates visual rehabilitation, especially if this occurs in the better eye.

Recently, intravitreal injection of anti‐VEGF agents has been used to promote resolution of POVCH (Antoszyk 2020; Chatziralli 2020).

Retinal breaks and rhegmatogenous retinal detachment

Retinal breaks are routinely successfully treated with laser or cryotherapy retinopexy; however, if undetected, obscured or in areas of unrelieved traction they can result in postoperative rhegmatogenous retinal detachment. Silicone oil is used in some cases to reduce the risk of postoperative rhegmatogenous retinal detachment. Silicone oil can be associated with peri‐retinal fibrosis, emulsification and a range of resultant complications, and can also be associated with unexplained visual loss (Chen 2021). 

Epiretinal membrane (ERM) and recurrent traction

Persistent or recurrent traction can be associated with macular oedema and tractional detachment. These can affect vision and have been reported to require revision surgery in 7% to 22% of cases (Blankenship 1985; Gupta 2012; Oshima 2009; Schiff 2007; Yorston 2008).

Description of existing interventions to reduce the incidence of postoperative complications and how the interventions may work

At the time of the initial vitrectomy surgery, several surgical strategies have been used to prevent complications and avoid the need for repeat surgery. These strategies include the following:

1. Established surgical interventions that are generally regarded as standard clinical practice:

   1. Ensuring adequate haemostasis at the time of vitrectomy with laser or intraocular bipolar coagulation or transient raised intraocular pressure, to maximise retinal view for adequate dissection and reduce postoperative oozing of dissected blood vessels.

   2. Removal of peripheral haemorrhagic vitreous to reduce leaching of sequestered red blood cells postoperatively into the vitreous cavity.

   3. Identification and removal of all posterior vitreoretinal traction. Vitreoschisis is known to occur in people with PDR, and identification of this and dissection in the true vitreoretinal plane are important to avoid recurrent traction and postoperative bleeding from neovascular tissue (Schwartz 1996).

   4. Applying supplementary posterior panretinal photocoagulation to reduce neovascular tissue.

2. Other strategies to prevent POVCH that have been reported but not routinely adopted:

   1. Surgical:

      1. Applying additional retinal photocoagulation up to the ora serrata to ablate retro‐oral ischaemic retina and reduce production of postoperative neovascular growth factors (Liggett 1987; Mason 1978; Yeh 2005).

      2. Direct treatment to the sclerostomy sites themselves with either cryotherapy or laser (Yeh 2005). This is thought to reduce the occurrence of entry site neovascularisation by inhibiting cellular migration through the sclerostomy wounds and causing focal atrophy of the ciliary epithelia.

      3. Removal of Wieger’s ligament and thorough anterior vitrectomy, especially around the inner sclerostomy wounds (McLeod 2000; McLeod 2003). This may be effective by increasing the egress of red blood cells and growth factors from the posterior segment to the anterior segment for rapid clearance of any POVCH, reducing any concentration of growth factors around the sclerostomy sites themselves, and removing the vitreous scaffold along which anterior new vessels could grow.

      4. Agents with physical actions thought to possibly reduce the rate of POVCH inserted into the eye during surgery, such as air (Joondeph 1989), gas (Koutsandrea 2001; Yang 2007), or viscoelastic substances (Packer 1989).

   2. Pharmacologic:

      1. Preoperative vascular endothelial growth factor (VEGF) inhibitor bevacizumab and alternative anti‐VEGF agents have been used in order to reduce vascular proliferation, permeability and vascularity of neovascular tissue (Romano 2009a; Romano 2009b; Yang 2008).

      2. Intraoperative triamcinolone has been injected in order to reduce inflammation and vascular proliferation (Faghihi 2008).

      3. Oral tranexamic acid, which inhibits fibrinolysis and hence clot dissolution, administered to the patient postoperatively (Laatikainen 1987; Ramezani 2005).

Description of the intervention

The use of anti‐VEGF agents injected into the vitreous cavity either before or at the time of surgery has been proposed to reduce the occurrence of a range of complications at the time of and after surgery.

How the intervention might work

Anti‐VEGF agents can reduce the amount and vascularity of abnormal retinal neovascularisation associated with PDR, facilitating surgical dissection, reducing intra‐ and postoperative bleeding and potentially improving outcomes. They may also reduce the occurrence of neovascularisation after surgery, reducing complications including bleeding. These effects may also reduce the need for silicone oil to be used, and reduce the need for re‐intervention for POVCH, recurrent traction or retinal detachment.

Why it is important to do this review

Diabetes mellitus is an increasing health problem. It is estimated that 7% (4.7 million) of the UK population are currently diabetic and that this will exceed 5.5 million by 2030. In the UK in 2010‐2011, 10% (GBP £10 billion) of the total NHS budget was spent treating the acute costs of diabetes and its complications, and this is predicted to increase to £17 billion by 2035 (Diabetes UK 2020). Diabetic retinopathy is one of the leading causes of blindness in the working age group in the UK, accounting for 14.4% of blind and partially sighted registrations (Liew 2014).

Within 20 years of diagnosis nearly all people with type 1 and almost two‐thirds of people with type 2 diabetes have some degree of retinopathy (Scanlon 2008); after 15 years, 30% and 10% of type 1 and type 2 diabetics, respectively, develop PDR (Klein 1984a; Klein 1984b). These patients are at risk of severe visual loss resulting from the complications of PDR. A study of patients at a large eye unit in the USA suggested that approximately 10% of patients presenting with PDR require vitrectomy surgery within one year (Kaiser 2000). Estimates based upon data from England suggest that approximately 4000 vitrectomies for the complications of PDR are currently performed annually in the UK (Vaideanu 2014). If 10% require revision surgery, this equates to 400 to 800 patients per annum in the UK.  Repeat interventions expose the patient to further operative risk and anxiety and add to the overall cost of care. 

The use of anti‐VEGF treatment is becoming increasingly common for many ophthalmic conditions and has significant revenue consequences.  As well as reducing the vascularity of fibrovascular membranes they can increase fibrosis, resulting in new areas of tractional retinal detachment (El‐Sabagh 2011). A previous Cochrane Review has shown the benefit of pre‐ and intraoperative anti‐VEGF use in reducing the occurrence of early POVCH after diabetic vitrectomy (Smith 2015). However, the effectiveness of different types of anti‐VEGF agents, their optimum timing of use and their effect on other outcomes is unclear. An up‐to‐date review of the literature will aid in optimum patient management and in the design of future studies.

Objectives

To assess the effects of perioperative anti‐VEGF use on the outcomes of vitrectomy for the treatment of complications of proliferative diabetic retinopathy.

Methods

Criteria for considering studies for this review

Types of studies

This review included randomised controlled trials (RCTs). We imposed no language, publication status or date restrictions.

Types of participants

Participants in the trials were people with PDR undergoing vitrectomy for the complications of diabetic retinopathy for the first time. To provide an overall estimate of the effect of anti‐VEGF agents in a broad population, we applied no restrictions based on participants' residence, race, ethnicity, occupation, education or socio‐economic status in general. 

Types of interventions

Intervention

Any anti‐VEGF combined with vitrectomy (within two weeks before or at time of vitrectomy). We considered any of the available anti‐VEGF treatments when given pre‐ or intraoperatively, including but not limited to:

• bevacizumab 1.25 mg/0.05 mL;

• ranibizumab 0.5 mg/0.05 mL;

• pegaptanib 0.3 mg/0.05 mL;

• aflibercept 2.0 mg/0.05 mL;

• conbercept 0.5 mg/0.05 mL.

Comparator

Vitrectomy with no anti‐VEGF treatment or with sham treatment, or with any other anti‐VEGF.

Types of outcome measures

As the main aim of this review was to look at outcomes and complications following vitrectomy, we excluded studies that did not report any postoperative outcomes.

Critical

1. Mean difference in best corrected visual acuity (BCVA) between study arms at six (± three) months following the primary vitrectomy. BCVA values were collected regardless of the method that was used for assessment (e.g. Snellen, Early Treatment Diabetic Retinopathy Study (ETDRS) letters logMAR). We pooled data only if reported measurements could be converted to logMAR. According to the US Food and Drug Administration, a 15‐letter change (equating to 0.3 logMAR) in BCVA would be clinically significant (Csaky 2008).

2. The incidence of early POVCH after surgery. We defined early POVCH as haemorrhage present from the first postoperative day or within the first four weeks postoperatively, being at least grade 2 in severity as measured by the Diabetic Retinopathy Vitrectomy Study criteria (Anonymous 1985).

3. The incidence of late POVCH after surgery. We defined late POVCH as haemorrhage occurring more than four weeks postoperatively after a period during which the vitreous cavity was clear. The minimum length of follow‐up for this outcome was six months.

4. The incidence of revision surgery for POVCH in the first six months postoperatively.

5. The incidence of revision surgery for recurrent traction/macular pucker of any type and/or rhegmatogenous retinal detachment in the first six months postoperatively. Preoperative administration of intravitreal bevacizumab has been stated to reduce the revision vitrectomy rate in patients primarily operated for vitreous haemorrhage (VH) (17.4% versus 7.7%). However, in patients primarily operated for tractional retinal detachment (TRD), preoperative administration of intravitreal bevacizumab reportedly led to an increase in re‐vitrectomy rate (54.5% versus 78.6%) (Hu 2019b).

6. Vision‐related quality of life (VRQOL) measures using any validated quality of life questionnaire performed at least six months following vitrectomy. Many tools have been developed in order to assess VRQOL (de Boer 2004). The 25‐item National Eye Institute Visual Function Questionnaire (NEI VFQ‐25) has been validated in patients with diabetic retinopathy (Mangione 2001). A clinically significant gain of at least 15 letters in BCVA corresponds to a clinically significant gain of at least seven points in the overall composite score of the (NEI VFQ‐25 (Suñer 2017). However, we did not expect all studies to have used the same tool in order to assess VRQOL.

Important

1. Proportion of people with a visual acuity of counting fingers (1.8 logMAR or worse).

2. Number of operative retinal breaks reported.

3. Frequency of silicone oil tamponade required at time of surgery.

Follow‐up

Follow‐up for the included studies' outcome measures varied from one month to 12 months post vitrectomy. We looked at studies with any follow‐up period, using six (± three) month follow‐up data as a minimum for late POVCH, revision surgery and visual acuity. Follow‐up periods for the remaining outcomes varied and are described in the Characteristics of included studies section.

Search methods for identification of studies

Electronic searches

The Cochrane Eyes and Vision Information Specialist searched the following databases for randomised controlled trials and controlled clinical trials. There were no restrictions to language or year of publication. The date of the search was 22 June 2022.

• Cochrane Central Register of Controlled Trials (CENTRAL 2022, Issue 6) (which contains the Cochrane Eyes and Vision Trials Register) in the Cochrane Library (searched 22 June 2022) (Appendix 1).

• MEDLINE Ovid (1946 to 22 June 2022) (Appendix 2).

• Embase Ovid (1980 to 22 June 2022) (Appendix 3).

• LILACS (Latin American and Caribbean Health Science Information database (1982 to 22 June 2022) (Appendix 4).

• ISRCTN registry (www.isrctn.com/editAdvancedSearch; searched 22 June 2022) (Appendix 5).

• US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov; searched 22 June 2022) (Appendix 6).

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp; searched 22 June 2022) (Appendix 7).

We ran the initial searches both with and without RCT search filters, as we wished to identify as much literature as possible. For subsequent updates, however, we will only run searches with RCT filters, as the review is focusing on evidence from RCTs only.

After editorial input, we decided to revise and broaden the search strategies for the review. The searches were constructed of terms for the following four components: diabetic retinopathy, vitrectomy, haemorrhage and anti‐VEGF drugs. The searches have now been amended whereby the terms for vitrectomy and for haemorrhage have been combined into one set of search terms. This has made the search broader and will help us to identify additional relevant studies.

Searching other resources

We manually searched the reference lists of the studies included in the review for additional trials. We did not specifically handsearch journals or conference proceedings for this review.

Data collection and analysis

Selection of studies

Two of the review authors independently assessed abstracts to ascertain which studies met the inclusion criteria for the review. We labelled the abstracts as included, unclear or excluded. We obtained full‐text copies of all included and unclear studies, and the two authors independently determined which studies met the inclusion criteria. We again labelled each study as included, unclear or excluded. We discussed any unclear articles and contacted the authors of the relevant article for further details when necessary. We also collected information on study design and setting, participant characteristics (including disease severity and age), study eligibility criteria, details of the intervention(s) given, the outcomes assessed, the source of study funding and any conflicts of interest stated by the investigators.

Data extraction and management

Two of the review authors independently extracted data using the standard methodological procedures recommended by Cochrane. We compared each data set and resolved any discrepancies by discussion. We independently entered data into RevMan Web and then rechecked the data (RevMan Web 2023). 

We collected and recorded the following variables from the included studies:

• Primary author

• Title

• Country where the study was conducted

• Healthcare setting

• Potential conflicts of interest

• Eligibility criteria

• RCT design

• Randomisation timing

• Randomisation method

• Number of participants randomised

• Number of eyes studied

• Number of eyes studied per arm

• Number of eyes analysed per arm

• Number of participants lost with reasons

• Baseline data provided

• Baseline data differences

• Type of intervention in each study arm

• Follow‐up examination schedule

• Outcomes with definitions

• Outcome comparisons by group

• Possible confounders

• Quality of evidence for each outcome using the GRADE classification (GRADEpro GDT)

• Potential sources of bias

Assessment of risk of bias in included studies

Two review authors structured the criteria for assessment of risk of bias using Cochrane's risk of bias tool (RoB 1) and assessed the risk of bias in each included study as per Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). We assessed each domain area, giving a judgement of 'low' risk of bias, 'high' risk of bias or 'unclear' risk of bias. A judgement of 'unclear' indicated that we were uncertain of the risk of bias. We looked specifically at the following domain areas:

• Methods used to generate the study groups.

• Methods used to conceal the allocation of the study groups.

• Methods used to mask (blind) study participants and personnel.

• Review of the study outcome data, looking specifically at the completeness of each study group outcome and selective outcome reporting.

Moreover, we performed assessment of reporting bias by testing for funnel plot asymmetry if there were at least 10 studies included in an analysis, as per Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). Asymmetry in the funnel plot may indicate small study effects and potential publication bias.

Measures of treatment effect

We presented dichotomous outcomes as a risk ratio with 95% confidence interval (CI) and continuous outcomes as mean difference with 95% CI.

The review's critical outcomes included dichotomous data on the incidence of persistent or early POVCH, the incidence of late POVCH following the primary surgery, the incidence of revision vitrectomy for POVCH, and also separately for recurrent traction/macular pucker/rhegmatogenous retinal detachment (RRD) up to six months postoperatively. We intended to analyse data regarding VRQOL as a critical outcome, whereas we analysed the mean difference in visual acuity at six (± three) months following the primary vitrectomy as continuous data. We also intended to analyse data on the proportion of people with visual acuity of counting fingers (CF) or worse as dichotomous data. We presented dichotomous data regarding the use of silicone oil tamponade at the time of primary surgery and continuous data regarding the number of intraoperative retinal breaks.

Unit of analysis issues

In the 2023 update, the eye receiving pars plana vitrectomy (PPV) for the complications of diabetic retinopathy was the unit of analysis. We included a few studies that included a small number of participants who had both eyes randomised (Ahn 2011; di Lauro 2009; Hu 2019; Manabe 2015; Ou 2021; Sohn 2012; Zhou 2017). Due to the number being small, we considered that the effect of interpersonal variance on the outcome would be negligible. Since the intervention given relates to the results of the specific eye being operated, risk of a confounding cross‐over effect is minimal. 

Dealing with missing data

In clinical trials it is expected that reports will contain missing data (Dziura 2013). In our review, we tried to investigate the reasons for missing data because this might introduce bias into the study. Where data were missing in the included studies, we contacted the authors of the studies to request the missing data. In studies for which the authors could not provide missing data, we assessed whether the missing data would introduce 'low' or 'high' risk of bias, as per the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). We performed an intention‐to‐treat (ITT) analysis, using imputed data completed by the trial investigator. However, we did not impute missing data ourselves. When ITT analysis was not possible, we performed a complete case analysis, based on the assumption that data are missing randomly. We assessed this assumption for each included study separately based on factors like exclusion criteria and the proportion of participants that were lost to follow‐up.

Assessment of heterogeneity

We assessed clinical heterogeneity by review of the study manuscripts and presented a descriptive summary of the results. We assessed heterogeneity between studies included in the meta‐analysis using the I² statistic. We took an I² result of more than 50% as an indication of high study heterogeneity.

Assessment of reporting biases

To avoid the presence of reporting biases within the review, we searched trial registry databases, which allowed us to identify all registered trials, published and unpublished. We contacted the investigators of registered trials that were listed as having completed participant recruitment to ask for any unpublished data that may be relevant to the review outcomes.

Data synthesis

We performed analysis of data from the included RCTs for the critical and important outcomes where feasible. We used a random‐effects model in every comparison including at least three studies, otherwise we used a fixed‐effect model.

If we came across multi‐arm studies, we only included those outlined in our pre‐specified eligibility criteria. Using eligible studies, we planned to combine intervention groups to make a comparison in patients treated with anti‐VEGFs (within two weeks before or at the time of vitrectomy) versus vitrectomy with no anti‐VEGF treatment or with sham treatment, as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021).

Subgroup analysis and investigation of heterogeneity

Regarding the effect of different agents when administered at different time points peri‐operatively, we planned to perform pre‐specified subgroup analyses where possible, according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021). 

It is possible that the interval of preoperative anti‐VEGF injection may influence postoperative outcomes. An interval of at least five days, preoperatively has been reported to be more effective in terms of better visual outcomes and fewer postoperative complications (Castillo 2017).

Moreover, the indication for vitrectomy may affect the effectiveness of perioperative anti‐VEGF therapy based on surgical complexity and we therefore planned to assess for heterogeneity of response according to two broad indications, namely vitreous haemorrhage (VH) and traction retinal detachment (TRD), if possible. 

We limited the number of subgroup analyses to avoid multiplicity issues. More specifically, we planned prespecified subgroup analyses for the following:

• Mean difference in BCVA in all cases for anti‐VEGF administration < 5 days versus ≥ 5 days before surgery.

• Incidence of early POVCH in all cases for anti‐VEGF administration < 5 days and ≥ 5 days before surgery.

• Mean difference in BCVA in cases with main indication traction retinal detachment versus vitreous haemorrhage.

• Mean difference in BCVA for bevacizumab (the most commonly used agent) versus all other agents.

Sensitivity analysis

Sensitivity analysis helps to assess the robustness of an analysis after the inclusion of studies graded as high risk of bias. In the original analysis we did not exclude any study on the basis of high risk of bias. However, to assess the strength of our findings, we performed additional sensitivity analyses to examine the effect of performance, detection and attrition bias by excluding studies at high risk in any one of these domains. Moreover, we performed a random‐effects meta‐analysis both with and without outlying studies as part of a sensitivity analysis when outliers were identified, according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021).

Summary of findings and assessment of the certainty of the evidence

We prepared a summary of findings table presenting relative and absolute risks. We graded the overall quality (certainty) of the evidence for each outcome using the GRADE classification (GRADEpro GDT). We included the following outcomes in the summary of findings table.

1. Mean difference in visual acuity, logMAR (six ± three months)

2. Early POVCH (occurring within four weeks postoperatively)

3. Late POVCH (occurring more than four weeks postoperatively after a period during which the vitreous cavity was clear)

4. Incidence of revision vitrectomy for POVCH

5. Incidence of revision vitrectomy for recurrent traction/macular pucker (six months)

6. Number of intraoperative retinal breaks

7. Frequency of silicone oil tamponade required at time of surgery

8. Quality of life

9. Proportion of people with a visual acuity of counting fingers (1.8 logMAR or worse)

These outcomes were not defined a priori because the original protocol and first edition of this review were published before the summary of findings methodology was adopted by Cochrane Eyes and Vision.

Results

Description of studies

Results of the search

The electronic searches in the original review yielded a total of 472 titles and abstracts and six reports of ongoing studies (Smith 2011). After de‐duplication, the Information Specialist scanned 397 records and discarded 311 records that were not relevant to the scope of the review. We screened the titles and abstracts of the remaining 86 references. We rejected 70 abstracts as not eligible for inclusion in the review. We obtained and screened full‐text copies of 16 references, of which we included four studies and excluded 12 studies from the review.

The amended searches run in May 2015 retrieved a total of 1414 records (Figure 1). After removing duplicate records, we screened 1138 references and excluded 1110 records that were not relevant to the scope of the review. We assessed 28 full‐text reports of studies for potential inclusion in the review. In addition to the four previously included studies (Ahmadieh 2009; di Lauro 2009; Modarres 2009; Rizzo 2008), we have now included eight new studies (Ahn 2011; El Batarny 2008; Farahvash 2011; Hernández‐Da Mota 2010; Manabe 2015; Sohn 2012; Zaman 2013). We excluded 12 studies (Berk Ergun 2014; Bhavsar 2014; Demir 2013; Goncu 2014; Gupta 2008; Gupta 2012a; Gupta 2012b; Jirawison 2012; Li 2010; Pakzad‐Vaezi 2014; Sato 2013). We previously excluded da R Lucena 2009 and Yeh 2009 at the initial stage of screening abstracts. However, these two studies have been included in a meta‐analysis by Zhao 2011, so in the interests of transparency we have now added these studies to our list of excluded studies with reasons for exclusion.

1.

The updated electronic searches run in June 2022 yielded a total of 662 records (Figure 1). After removal of 208 duplicates, the Cochrane Information Specialist screened the remaining 454 records and removed 317 records that were not relevant to the scope of the review. We screened the remaining 137 records and after scanning for double and irrelevant records, we excluded outright 72 records. We obtained the full‐text reports of 65 records for further assessment and have included the following 16 new studies (Aleman 2019; Arevalo 2019; Comyn 2017; Cui 2018; Dong 2019; Ferraz 2013; Hu 2019; Jiang 2020; Li 2022; Ou 2021; Ren 2019; Su 2016; Wildan 2019; Yang 2016; Yu 2015; Zhou 2017). We excluded 49 new studies: see Characteristics of excluded studies for details. 

In the previous version of this review there were 12 included studies (Smith 2015). In this 2023 update, Elwan 2013, which was previously an included study, has now been excluded as the study has been retracted. Pakzad Vaezi 2014, which was excluded in the previous version of the review, has now been included in the review because this study reports outcomes of interest.

Ongoing studies

On 11 September 2022, we checked the status of the eight ongoing trials previously listed in this review. Two trial reports have been moved to included and excluded studies within this review. Six reports remain as ongoing trials and we will assess them for potential inclusion when data become available (ISRCTN79120387; NCT00516464; NCT00931125; NCT01091896; NCT01151722; NCT01306981).

Included studies

We identified 28 completed studies from the electronic database searches that met our inclusion criteria; see the Characteristics of included studies table for details. All the data presented in the review were obtained from published literature.

Setting and participants

The included studies were conducted in a number of countries. Eleven were from China, three were from Iran, two were from Italy, two were from Mexico, and the remaining studies were from South Korea, the UK, Egypt, Brazil, Japan, Canada, the USA, Indonesia and Pakistan. One study was multi‐centre, including 13 clinical sites from nine countries. The included studies randomised a total of 1914 eyes. Sample size varied from 20 (Sohn 2012) to 214 (Arevalo 2019). 

Intervention

The most frequently used anti‐VEGF regime was intravitreal bevacizumab (IVB) (1.25 mg/0.05 mL) within one week prior to pars plana vitrectomy. The studies with a different protocol were the following:

• Ahn 2011 had an intervention group in which 1.25 mg IVB was given at the end of surgery (37 participants).

• Aleman 2019 had an intervention group that received 1.25 mg Ziv‐aflibercept (IVZ) 1 to 10 days before vitrectomy (99 participants).

• Comyn 2017 had an intervention group that received 0.5 mg ranibizumab (IVR) within one week of pars plana vitrectomy (15 participants).

• Cui 2018 had an intervention group that received 0.5 mg conbercept (IVC) within one week of surgery (20 participants) and one group that received 0.5 mg IVR within one week of surgery (20 participants).

• di Lauro 2009 had an intervention group that received 1.25 mg IVB three weeks before vitrectomy.

• Hernández‐Da Mota 2010 gave 1.25 mg IVB 48 hours before surgery.

• Hu 2019 had an intervention group that received 0.5 mg IVR one week before surgery (51 participants).

• Jiang 2020 had an intervention group that received 0.5 mg IVC at the end of surgery (15 participants).

• Li 2022 had an intervention group that received 0.5 mg IVR one day before surgery (16 participants) and another group that received 0.5 mg IVR three days before surgery (16 participants).

• Manabe 2015 gave 0.16 mg (0.05 mL) IVB one day before surgery.

• Modarres 2009 gave 2.5 mg IVB three to five days before surgery.

• Ou 2021 had an intervention group that received 0.5 mg IVC five days before surgery (32 participants).

• Pakzad Vaezi 2014 had an intervention group that received 0.5 mg IVR one week before surgery (15 participants).

• Ren 2019 had an intervention group that received 0.5 mg IVC immediately after surgery (25 participants).

• Su 2016 had an intervention group that received 0.5 mg IVC one week before surgery (18 participants).

• Wildan 2019 had an intervention group that received 0.5 mg aflibercept (IVA) four to seven days before surgery (12 participants).

• Yang 2016 had an intervention group that received 0.5 mg IVC three days before surgery (54 participants).

• Yu 2015 had an intervention group that received 0.5 mg IVC two to six days before surgery (61 participants).

• Zhou 2017 had an intervention group that received 0.5 mg IVC three to seven days before surgery (18 participants).

Outcomes

The main outcomes in the studies included best‐corrected visual acuity (BCVA), feasibility of surgery (operating time, intraoperative bleed and type of surgical steps required) and postoperative complications. The main postoperative complications were rate of POVCH, iris rubeosis and retinal detachment. Follow‐up varied significantly across the studies, ranging from one month (Ahmadieh 2009; Manabe 2015) to one year (Arevalo 2019; Jiang 2020). Li 2022 did not report clearly the exact follow‐up period. Most of the studies had a final follow‐up visit at six months.

A number of studies reporting the incidence of POVCH included participants who received silicone oil endotamponade, which has made interpretation of the study results difficult. The presence of silicone oil within the vitreous cavity precludes an accurate diagnosis of POVCH, as any blood would be localised to the far periphery of the posterior chamber. Grading of POVCH is not possible with silicone endotamponade present. In light of this, we have carried out sensitivity analyses on the incidence of POVCH excluding participants with silicone oil endotamponade. Two studies excluded participants who received silicone oil and gave an accurate grading system for POVCH (Ahmadieh 2009; Ahn 2011).

In addition, we identified nine ongoing RCTs and have provided details in the Characteristics of ongoing studies table. We cannot comment on whether these studies' characteristics meet the inclusion criteria until they are published and we can undertake a review of their methodology.

Excluded studies

We excluded 76 studies identified by the search that were either non‐randomised prospective studies or retrospective studies; see the Characteristics of excluded studies table. We excluded one study that did not collect data on POVCH (da R Lucena 2009).

Risk of bias in included studies

See also Figure 2.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Eleven studies clearly stated how the random sequence generation was performed and we graded the methods as 'low' risk for randomisation and allocation (Ahmadieh 2009; Ahn 2011; Aleman 2019; Arevalo 2019; Comyn 2017; Hu 2019; Jiang 2020; Ou 2021; Pakzad Vaezi 2014; Ren 2019; Rizzo 2008). 

Ten studies stated that participants were randomly assigned into the groups, but provided no description on how the randomisation or sequence allocation was performed (Cui 2018; di Lauro 2009; Dong 2019; Ferraz 2013; Su 2016; Wildan 2019; Yang 2016; Yu 2015; Zaman 2013; Zhou 2017). Six studies did not provide any details about the randomisation methods (El Batarny 2008; Farahvash 2011; Hernández‐Da Mota 2010; Li 2022; Modarres 2009; Sohn 2012). We graded the methods used for randomisation in these studies as 'unclear'. 

We assessed five studies at 'low' risk of selection bias, as a clear description of how this process was performed was provided (Ahmadieh 2009; Ahn 2011; Arevalo 2019; Comyn 2017; Manabe 2015).

We graded the remaining studies as unclear, because they did not comment on the allocation procedure.

Blinding

We assessed seven studies as at 'low' risk of bias for masking (blinding), as a clear description of how this process was performed was provided (Ahmadieh 2009; Arevalo 2019; Comyn 2017; di Lauro 2009; Manabe 2015, Pakzad Vaezi 2014; Sohn 2012). We graded Aleman 2019 as low risk for performance bias since both participants and personnel were masked, but as high risk for detection bias since unmasked personnel provided postoperative examinations. We graded Cui 2018 as high risk for performance bias since participants were not masked and as low risk for detection bias since surgeries were performed by masked surgeons. We graded seven studies that had an open‐label design as 'high' risk (Ahn 2011; El Batarny 2008; Farahvash 2011; Hernández‐Da Mota 2010; Hu 2019; Su 2016; Zaman 2013). We graded Jiang 2020 and Li 2022 as high risk for performance bias since participants were not masked and low risk for detection bias since all the assessors were masked to the group. We graded nine studies that did not comment on the masking of participants or investigators as 'unclear' (Dong 2019; Ferraz 2013; Modarres 2009; Ou 2021; Rizzo 2008; Wildan 2019; Yang 2016; Yu 2015; Zhou 2017). Ren 2019 reported that all the surgeries were performed by the same masked surgeon, however it is not clarified who performed the postoperative examinations. As a result we graded this study as having an unclear risk of bias.

Incomplete outcome data

Ahmadieh 2009, Ahn 2011, Farahvash 2011 and Sohn 2012 clearly accounted for each participant who failed to complete the study protocol. Moreover, attrition rates and loss to follow‐up rates were similar between arms in these studies. It should be noted that in Ahmadieh 2009, less than half of the participants (16 of 35) within the treatment group failed to complete the protocol, and 18 of 35 participants within the control group completed the protocol. Manabe 2015 reported participant flow and all participants were followed up to one month. Li 2022 did not clearly report the exact follow‐up period, thus rendering the study at high risk of bias. Although the remaining included studies did not comment directly on the attrition rate, it is clear from the results that all participants completed follow‐up.

Selective reporting

Ferraz 2013 did not report the incidence of late POVCH. Li 2022 did not provide a list of prespecified outcome measures. For this reason we graded these studies as high risk. Ahmadieh 2009, Ahn 2011, Aleman 2019, Arevalo 2019, Comyn 2017, di Lauro 2009, Jiang 2020, Manabe 2015, Pakzad Vaezi 2014, Ren 2019 and Sohn 2012 reported all the outcomes stated in the original trial register and we graded them as low risk. The remaining studies did not provide information about a study protocol or trial register in order to confirm whether all prespecified outcome measures were reported (Cui 2018; Dong 2019; El Batarny 2008; Farahvash 2011; Hernández‐Da Mota 2010; Hu 2019; Modarres 2009; Ou 2021; Rizzo 2008; Su 2016; Wildan 2019; Yang 2016; Yu 2015; Zaman 2013; Zhou 2017). As a result, we have graded these studies as at unclear risk of bias.

Other potential sources of bias

Ahmadieh 2009 recorded a high number of participants who failed to complete the trial in both the control and treatment groups: 15 of 35 in the control group and 19 of 35 in the treatment group. The main reasons for failing to complete follow‐up were vitreous haemorrhage reabsorption, silicone oil endotamponade, gas tamponade and loss to follow‐up.

Ahn 2011 included 107 eyes of 91 participants. The 16 bilateral participants were included in the data analysis; no comment was made as to whether adjustment for within‐person correlation was performed during statistical analysis.

In Comyn 2017 and Pakzad Vaezi 2014, the authors declared receiving funding from the manufacturer of the anti‐VEGF studied.

In the intravitreal bevacizumab (IVB) group of Manabe 2015, participants with bilateral proliferative diabetic retinopathy (PDR) received IVB in one eye only, and the other eye was excluded from the study. The authors considered that IVB injected in one eye may have an effect on the contralateral eye. In the control group, both eyes of bilateral cases were included in the study.

Effects of interventions

See: Table 1

We have described the effect of each intervention in order of outcome type.

Critical outcomes

1. Visual acuity at six (± three) months following the primary vitrectomy

Analysis 1.1 includes all studies that compared any anti‐vascular endothelial growth factor (VEGF) given preoperatively to control. Thirteen studies including 699 participants (392 in the anti‐VEGF arm and 307 in the control arm) were included in the analysis. The use of anti‐VEGF preoperatively was shown to improve visual acuity (mean difference (MD) ‐0.25 logMAR, 95% confidence interval (CI) ‐0.39 to ‐0.11). The studies Ahn 2011 and Hernández‐Da Mota 2010 appear to be outliers. After excluding these two outlying studies, anti‐VEGF administration was associated with a significant improvement in postoperative best corrected visual acuity (BCVA) (MD ‐0.23 logMAR, 95% CI ‐0.32 to ‐0.14). These results were robust to all further sensitivity analyses. There were a few studies that reported visual outcomes, however they did not contribute data that could be included in the analysis. Ahmadieh 2009 and Yang 2016 reported a significant improvement in BCVA in the anti‐VEGF arm versus the control, however they only measured visual acuity at one month postoperatively. Similarly, Arevalo 2019 reported better outcomes at 12 months with preoperative anti‐VEGF compared to pars plana vitrectomy (PPV) alone. Ferraz 2013 only graphically presented the visual outcomes, however they did not establish a statistically significant benefit of anti‐VEGF. Manabe 2015 and Su 2016 only reported the change in BCVA during the first month postoperatively, without a statistically significant difference between the groups. Rizzo 2008 did not report standard deviations. However, the improvement in BCVA was slightly significant in the anti‐VEGF group, while in the control group it was not. Su 2016 reported better visual outcomes for the anti‐VEGF group at six weeks compared to the control group. When investigating possible publication bias, the funnel plot seemed to be symmetrical (Figure 3).

1.1. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 1: Mean BCVA

3.

We carried out subgroup analyses to compare the effect on visual acuity of any anti‐VEGF given less than five days preoperatively to at least five days postoperatively (Analysis 2.1). Analysis 2.1.1, consisting of seven studies with 396 participants (209 in the anti‐VEGF arm and 187 in the control arm), showed that administration of anti‐VEGF less than five days preoperatively leads to an improvement in visual acuity (MD ‐0.30 logMAR, 95% CI ‐0.53 to ‐0.07). Analysis 2.1.2 consists of four studies with 207 participants (116 in the anti‐VEGF group and 91 in the control group) and shows that administration of anti‐VEGF more than five days preoperatively (range 5 to 21 days) also leads to an improvement in visual acuity (MD ‐0.13 logMAR, 95% CI ‐0.20 to ‐0.06). These results were robust to all further sensitivity analyses. The test for subgroup differences showed no statistically significant results (P = 0.20).

2.1. Analysis.

Comparison 2: Anti‐VEGF versus control: subgroup analysis ‐ time of administration of anti‐VEGF, Outcome 1: Mean BCVA

Only one study included participants that were listed for surgery solely due to tractional retinal detachment. Hernández‐Da Mota 2010 compared 20 participants that received bevacizumab 48 hours prior to vitrectomy to controls. The authors concluded that preoperative bevacizumab leads to an increase in visual acuity (MD ‐1.31 logMAR, 95% CI ‐1.76 to 0.86), however this was not statistically significant. Analysis 3.1 includes studies with participants listed for surgery solely due to vitreous haemorrhage. Three studies with 137 participants (77 in the anti‐VEGF group and 60 in the control group) were included. The analysis shows that preoperative anti‐VEGF leads to a lower but statistically significant increase in visual acuity compared to tractional retinal detachment cases (MD ‐0.25 logMAR, 95% CI ‐0.37 to ‐0.13). The test for subgroup differences showed statistically significant results (P < 0.01). These results were robust to all further sensitivity analyses.

3.1. Analysis.

Comparison 3: Anti‐VEGF versus control: subgroup analysis ‐ main indication for vitrectomy, Outcome 1: Mean BCVA

We carried out subgroup analyses to compare bevacizumab, the most commonly used agent, to control and all other anti‐VEGFs to control, respectively (Analysis 4.1). Analysis 4.1.1 includes six studies with 307 participants and shows that preoperative bevacizumab leads to an increase in visual acuity that was not statistically significant (MD ‐0.35 logMAR, 95% CI ‐0.79 to 0.08). However, two out of six studies presented considerable heterogeneity, with a mean difference in BCVA ranging from 1.31 logMAR better vision in the anti‐VEGF group (Hernández‐Da Mota 2010) to 0.14 logMAR better vision in the control group (Ahn 2011). As a result, a pooled estimate of effect for this comparison may not be informative. On the other hand, Analysis 4.1.2 includes six studies with 272 participants and it shows that all other anti‐VEGF agents combined, when given preoperatively, lead to a statistically significant increase in visual acuity (MD ‐0.19 logMAR, 95% CI ‐0.28 to ‐0.10). These results were robust to all further sensitivity analyses. No statistically significant difference was shown when testing for subgroup differences (P = 0.48).

4.1. Analysis.

Comparison 4: Anti‐VEGF versus control: subgroup analysis ‐ bevacizumab versus other anti‐VEGF, Outcome 1: Mean BCVA

2. The incidence of early postoperative vitreous cavity haemorrhage (POVCH) after surgery

Fourteen studies with 544 participants receiving preoperative anti‐VEGF and 494 control group participants were included in Analysis 1.2. This shows that preoperative administration of anti‐VEGF significantly reduces the risk of early (less than four weeks postoperatively) POVCH (risk ratio (RR) 0.44, 95% CI 0.34 to 0.58). These results were robust to all further sensitivity analyses. A funnel plot showed asymmetry, thus implying possible publication bias and heterogeneity due to differences in the size of each study (Figure 4).

1.2. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 2: Early POVCH

4.

We additionally carried out subgroup analyses that divided the studies according to when the anti‐VEGF agent was administered (Analysis 2.2). Analysis 2.2.1 includes studies that administered any anti‐VEGF less than five days preoperatively. It includes eight studies with 618 participants and shows that anti‐VEGF administration leads to a statistically significant decrease in the risk of early POVCH (RR 0.44, 95% 0.29 to 0.66). Similarly, Analysis 2.2.2 includes five studies that examined the use of any anti‐VEGF at least five days prior to surgery (range 5 to 21 days). This shows that anti‐VEGF leads to a comparable decrease in the risk of early POVCH (RR 0.26, 95% 0.14 to 0.51). These results were robust to all further sensitivity analyses.

2.2. Analysis.

Comparison 2: Anti‐VEGF versus control: subgroup analysis ‐ time of administration of anti‐VEGF, Outcome 2: Early POVCH

3. The incidence of late POVCH after surgery

Eleven studies with 579 participants were included in Analysis 1.3. This shows that administration of anti‐VEGF leads to a reduction in the risk of late (at least four weeks postoperatively) POVCH (RR 0.47, 95% CI 0.30 to 0.74). These results were robust to all further sensitivity analyses and a funnel plot did not reveal any asymmetry (Figure 5).

1.3. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 3: Late POVCH

5.

4. The incidence of revision surgery for POVCH at six months

Four studies with 207 participants (115 treated preoperatively with anti‐VEGF and 92 in the control group) were included in Analysis 1.4. Anti‐VEGF administration was associated with a non‐statistically significant decrease in the incidence of revision surgery after six months (RR 0.44, 95% CI 0.15 to 1.28). When we performed a sensitivity analysis by excluding studies with a high risk of performance bias from the analysis, anti‐VEGF administration was associated with a statistically significant reduction in the incidence of revision surgery after six months (RR 0.18, 95% CI 0.04 to 0.85). 

1.4. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 4: Revision surgery for POVCH

5. The incidence of revision surgery for recurrent traction/macular pucker of any type and/or rhegmatogenous retinal detachment at six months

Four studies with 145 participants (75 treated with anti‐VEGF and 70 in the control group) were included in Analysis 1.5. Anti‐VEGF administration reduced the rates of revision surgery for these indications, however the confidence intervals were wide and compatible with no effect (RR 0.50, 95% CI 0.15 to 1.66).

1.5. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 5: Revision surgery for recurrent traction and/or RRD

6. Quality of life measures performed at least six months following vitrectomy

There are currently no studies that report on this outcome.

Important outcomes

1. Proportion of people with visual acuity of counting fingers (1.8 logMAR or worse)

No studies reported visual outcomes in a way that could be used for extracting the proportion of people with this visual acuity.

2. Number of intraoperative retinal breaks

We synthesised 12 studies with 915 participants (468 receiving anti‐VEGF and 447 receiving control) for this outcome. Analysis 1.6 shows that anti‐VEGF administration leads to a statistically significant reduction in the risk of intraoperative retinal breaks (RR 0.37, 95% CI 0.24 to 0.59). These results were robust to all further sensitivity analyses. A funnel plot was symmetrical (Figure 6).

1.6. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 6: Intraoperative retinal breaks

6.

3. Frequency of silicone oil tamponade required at the time of surgery

Ten studies with 591 participants were included in the analysis, which shows that preoperative use of anti‐VEGF leads to a decrease in the need for silicone oil tamponade (RR 0.46, 95% CI 0.27 to 0.80) (Analysis 1.7). Of note, when performing sensitivity analysis by excluding studies with high risk of performance bias, this decrease is not statistically significant (RR 0.58, 95% CI 0.28 to 1.19). When examining for possible publication bias and differences in the size of effect, a funnel plot showed asymmetry (Figure 7).

1.7. Analysis.

Comparison 1: Anti‐VEGF versus control, Outcome 7: Silicone oil use

7.

Discussion

Summary of main results

The aim of this review was to establish whether the use of anti‐vascular endothelial growth factors (VEGFs) as an adjunct to pars plana vitrectomy, either pre‐ or intraoperatively, is of benefit. Currently the review includes 28 randomised controlled trials (RCTs), which look at the use of bevacizumab, ranibizumab, aflibercept, Ziv‐aflibercept and conbercept. The critical outcomes were the effect of anti‐VEGF on best corrected visual acuity (BCVA), early and late postoperative vitreous cavity haemorrhage (POVCH), revision surgery for POVCH, revision surgery for recurrent traction/macular pucker of any type and/or rhegmatogenous retinal detachment and vision‐related quality of life (VRQOL). We summarised the effect of anti‐VEGFs by outcome type.

Participants receiving intravitreal anti‐VEGF in addition to pars plana vitrectomy achieved better BCVA at six months compared to people undergoing vitrectomy alone (mean difference (MD) ‐0.25 logMAR, 95% confidence interval (CI) ‐0.39 to ‐0.11; 13 studies, 699 eyes; low‐certainty evidence). Subgroup analyses showed that perioperative anti‐VEGF might be more beneficial in terms of visual outcomes when administered less than five days preoperatively (MD ‐0.30 logMAR, 95% CI ‐0.53 to ‐0.07; 7 studies, 396 eyes) compared to ≥ 5 days preoperatively (MD ‐0.13 logMAR, 95% CI ‐0.20 to ‐0.06; 4 studies, 207 eyes), although the test for subgroup effect was not statistically significant. Clinically significant regression of neovascularisation has been reported to happen during the first week following treatment with anti‐VEGF in proliferative diabetic retinopathy (PDR) patients. Moreover, the half‐life of bevacizumab in a non‐vitrectomised eye has been reported to be 6.7 days. The occurrence of increased traction following anti‐VEGF injections prior to vitrectomy has also been widely reported and related to the time period between injection and surgery, with longer periods increasing the risk of visually significant macular traction (Russo 2019). As a result, administration of anti‐VEGF closer to the surgery date might still facilitate the surgery, minimise the risk of macular traction and provide better visual outcomes (Jorge 2006; Zhang 2013; Zhu 2008). 

Pre‐ or intraoperative anti‐VEGF reduced the incidence of early POVCH (12% versus 31%, risk ratio (RR) 0.44, 95% CI 0.34 to 0.58; 14 studies, 1038 eyes; moderate‐certainty evidence). Both administration < 5 days preoperatively (15% versus 35%, RR 0.44, 95% CI 0.29 to 0.66; 8 studies, 618 eyes) and ≥ 5 days preoperatively (6% versus 30%, RR 0.26, 95% CI 0.14 to 0.51; 5 studies, 259 eyes) reduced the incidence of early POVCH.

Perioperative anti‐VEGF use was also associated with a reduction in the incidence of late POVCH (10% versus 23%, RR 0.47, 95% CI 0.30 to 0.74; 11 studies, 579 eyes; high‐certainty evidence).

The need for revision surgery for POVCH occurred less frequently in the anti‐VEGF group compared with the control group but the confidence intervals were wide and compatible with no effect (4% versus 13%, RR 0.44, 95% CI 0.15 to 1.28; 4 studies 207 eyes; moderate‐certainty evidence). Similar imprecisely measured effects were seen for revision surgery for rhegmatogenous retinal detachment (5% versus 11%, RR 0.50, 95% CI 0.15 to 1.66; 4 studies, 145 eyes; low‐certainty evidence). Anti‐VEGF facilitated the surgery by reducing the incidence of intraoperative retinal breaks (12% versus 31%, RR 0.37, 95% CI 0.24 to 0.59; 12 studies, 915 eyes; high‐certainty evidence). The evidence is very uncertain about the effect of perioperative anti‐VEGF on the need for silicone oil (19% versus 41%, RR 0.46, 95% CI 0.27 to 0.80; 10 studies, 591 eyes; very low‐certainty evidence).

No data were available on quality of life outcomes or the proportion of participants with visual acuity of counting fingers (CF) (1.8 logMAR or worse).

Overall completeness and applicability of evidence

This review includes 28 studies conducted in Europe, North America, Africa and Asia. The results of these studies are likely to be applicable to standard clinical practice. The studies consistently showed that intravitreal anti‐VEGF applied around the time of vitrectomy to reduce the occurrence of complications of proliferative diabetic retinopathy decreases the risk of both early and late POVCH. Anti‐VEGF use was also associated with statistically significant better postoperative BCVA. However, the evidence was less complete and could not be analysed with respect to other outcomes: the effect of anti‐VEGF on VRQOL and the proportion of patients with a low postoperative BCVA (less than counting fingers).

The included studies used a variety of regimens, including bevacizumab, ranibizumab, aflibercept, Ziv‐aflibercept and conbercept. Most studies used a similar dose of bevacizumab (1.25 mg), but a variety of regimens were used in terms of the time of administration. No data were available on other anti‐VEGF agents.

Quality of the evidence

The design of RCTs to look at the effect of adjunctive use of anti‐VEGF in diabetic vitrectomy is difficult. There are a number of indications for vitrectomy in proliferative retinopathy. These vary from non‐clearing vitreous haemorrhage in older patients with previously treated inactive proliferative diabetic retinopathy, through to younger patients with active untreated proliferative diabetic retinopathy and severe tractional changes. Many of the included RCTs had a heterogenous group of participants, as above, and the effect of anti‐VEGF on POVCH and BCVA may differ depending on the patient type.

Ten of the included studies appear to be at high risk of performance bias due to the personnel and participants not being masked. This is a significant proportion and may limit confidence in the estimate of the treatment effect of anti‐VEGF.

Potential biases in the review process

We followed our pre‐specified review protocol and documented the rationale for any changes from protocol. 

Agreements and disagreements with other studies or reviews

Zhao 2011 is a systematic review looking at the potential benefits of preoperative intravitreal bevacizumab on both intraoperative outcomes (including intraoperative bleeding, endodiathermy, iatrogenic retinal tears and mean surgical time) and postoperative outcomes (including BCVA, recurrent vitreous haemorrhage, reabsorption time of blood and other complications).

Data analysis of included studies within the Zhao 2011 review, which looked at the incidence of early POVCH, showed an almost statistically significant result in favour of the treatment group (odds ratio (OR) 5.48, 95% CI 0.97 to 31.02; P = 0.05). This result is in agreement with the data analysis in this review. However, it should be noted that many of the included studies within the Zhao 2011 data analysis contained participants who received silicone oil endotamponade, which may be a significant source of bias. Also, the authors of Zhao 2011 commented that due to varying definitions of POVCH, study heterogeneity within the data analysis was high (I² = 93.6%), so the results should be interpreted with caution.

Zhang 2013 reported similar outcomes in a meta‐analysis of the reported data on preoperative intravitreal bevacizumab use in vitrectomy for proliferative diabetic retinopathy. The study reported a beneficial effect of intravitreal bevacizumab on the rate of early POVCH (OR 0.35, 95% CI 0.21 to 0.58) and no effect on the rate of late POVCH (OR 0.55, 95% Cl 0.25 to 1.21).

Veliz 2014 summarised the results of a previously published version of this review. Zhang 2013, Zhao 2011 and Veliz 2014 concluded that bevacizumab probably decreases intraoperative bleeding, but its effect on visual acuity is uncertain. They concluded that it might increase the risk of cardiovascular events, but this was based on a RR of 2.2 with a 95% CI from 0.11 to 44.3, so there is considerable uncertainty in this estimate as well.

In a RCT, Castillo 2017 reported better outcomes when administering bevacizumab 5 to 10 days before vitrectomy as compared with a shorter interval. In our review, although both comparisons showed a statistically significant reduction in the incidence of early POVCH, this was greater when anti‐VEGF was administered < 5 days preoperatively rather than ≥ 5 days preoperatively.

In a recent meta‐analysis, Dervenis 2021 reported similar outcomes in terms of the effect of intravitreal bevacizumab on iatrogenic retinal breaks, POVCH and visual outcomes.

Authors' conclusions

Implications for practice.

This review provides support for the use of perioperative anti‐vascular endothelial growth factor (VEGF) to improve visual outcomes, reduce the risk of POVCH after vitrectomy for proliferative diabetic retinopathy and facilitate surgery. The current analysis suggests that pre‐ or intraoperative anti‐VEGF use may improve best corrected visual acuity (BCVA), likely reduces the risk of early postoperative vitreous cavity haemorrhage (POVCH) and it reduces the risk of late POVCH, and minimises the incidence of intraoperative retinal breaks. The evidence is very uncertain about the effect of anti‐VEGF on the need for silicone oil tamponade.

Implications for research.

Further randomised studies looking at the effect of anti‐VEGF agents on vision‐related quality of life (VRQOL) are needed. Furthermore, the spontaneous clearance of vitreous haemorrhage or clearance following the injection of an anti‐VEGF should be reported clearly in future studies. A minimum follow‐up period of at least six months is important to ensure that the long‐term effect of anti‐VEGF use is documented. Visual acuity should be analysed at defined time points and measured before and after further surgery to remove POVCH. We recommend grading of POVCH using the criteria within the Diabetic Retinopathy Vitrectomy Study, making the outcome of randomised controlled trials (RCTs) more comparable and easier to subject to further analysis. There is a need for agreement on variable and outcome reporting standardisation in trials with patients undergoing vitrectomy for proliferative diabetic retinopathy (PDR). Finally, more studies looking at surgical time and choice of final tamponade are needed. With a trend towards shorter tamponade agents, we need more studies looking at the effect that these agents may have on the outcomes of diabetic vitrectomy (Dervenis 2022; Govers 2022; Rush 2022).

What's new

Date	Event	Description
31 May 2023	New citation required but conclusions have not changed	The title has been changed from 'Anti‐vascular endothelial growth factor for prevention of postoperative vitreous cavity haemorrhage after vitrectomy for proliferative diabetic retinopathy'. The PICO was also amended.
31 May 2023	New search has been performed	New trials have been included that met the inclusion criteria.

History

Protocol first published: Issue 1, 2010
Review first published: Issue 5, 2011

Date	Event	Description
22 October 2014	New citation required and conclusions have changed	Issue 8, 2015: Eight new RCTs were identified that met the inclusion criteria (Ahn 2011; El Batarny 2008; Elwan 2013; Farahvash 2011; Hernández‐Da Mota 2010; Manabe 2015; Sohn 2012; Zaman 2013).
22 September 2014	New search has been performed	Issue 8, 2015: Electronic searches were updated.

Appendices

Appendix 1. CENTRAL search strategy

#1 MeSH descriptor: [Diabetic Retinopathy] explode all trees
#2 (diabet* or proliferative or non‐proliferative) near/4 retinopath*
#3 DR near/3 (eye* or vision or visual* or sight)
#4 #1 or #2 or #3
#5 MeSH descriptor: [Vitrectomy] explode all trees
#6 vitrectom*
#7 PPV*
#8 #5 or #6 or #7
#9 MeSH descriptor: [Angiogenesis Inhibitors] explode all trees
#10 MeSH descriptor: [Angiogenesis Inducing Agents] this term only
#11 MeSH descriptor: [Endothelial Growth Factors] this term only
#12 MeSH descriptor: [Vascular Endothelial Growth Factors] explode all trees
#13 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 faricimab or brolucizumab or leizumabor or abicipar pegol
#14 anti near/2 VEGF*
#15 anti near/1 angiogen*
#16 endothelial near/2 growth near/2 factor*
#17 VEGF TRAP*
#18 #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17
#19 #4 and #8 and #18

Appendix 2. MEDLINE Ovid search strategy

1. randomized controlled trial.pt.
2. (randomized or randomised).ab,ti.
3. placebo.ab,ti.
4. dt.fs.
5. randomly.ab,ti.
6. trial.ab,ti.
7. groups.ab,ti.
8. or/1‐7
9. exp animals/
10. exp humans/
11. 9 not (9 and 10)
12. 8 not 11
13. diabetic retinopathy/
14. ((diabet$ or proliferative or non‐proliferative) adj4 retinopath$).tw.
15. diabetic retinopathy.kw.
16. (DR adj3 (eye$ or vision or visual$ or sight$)).tw.
17. or/13‐16
18. exp vitrectomy/
19. vitrectom$.tw.
20. PPV$.tw.
21. or/18‐20
22. exp angiogenesis inhibitors/
23. angiogenesis inducing agents/
24. endothelial growth factors/
25. exp vascular endothelial growth factors/
26. (macugen$ or pegaptanib$ or lucentis$ or rhufab$ or ranibizumab$ or bevacizumab$ or avastin$ or aflibercept$ or conbercept$ or OPT 302 or Opthea$ or RTH258 or faricimab or brolucizumab or leizumabor or abicipar pegol).tw.
27. (anti adj2 VEGF$).tw.
28. (anti adj1 angiogen$).tw.
29. (endothelial adj2 growth adj2 factor$).tw.
30. VEGF TRAP$.tw.
31. or/22‐30
32. 12 and 17 and 21 and 31

The search filter for trials at the beginning of the MEDLINE strategy is from the published paper by Glanville 2006.

Appendix 3. EMBASE Ovid search strategy

1. exp randomized controlled trial/
2. exp randomization/
3. exp double blind procedure/
4. exp single blind procedure/
5. random$.tw.
6. or/1‐5
7. (animal or animal experiment).sh.
8. human.sh.
9. 7 and 8
10. 7 not 9
11. 6 not 10
12. exp clinical trial/
13. (clin$ adj3 trial$).tw.
14. ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
15. exp placebo/
16. placebo$.tw.
17. random$.tw.
18. exp experimental design/
19. exp crossover procedure/
20. exp control group/
21. exp latin square design/
22. or/12‐21
23. 22 not 10
24. 23 not 11
25. exp comparative study/
26. exp evaluation/
27. exp prospective study/
28. (control$ or prospectiv$ or volunteer$).tw.
29. or/25‐28
30. 29 not 10
31. 30 not (11 or 23)
32. 11 or 24 or 31
33. exp diabetic retinopathy/
34. ((diabet$ or proliferative or non‐proliferative) adj4 retinopath$).tw.
35. (proliferat$ adj3 retinopath$).tw.
36. (DR adj3 (eye$ or vision or visual$ or sight$)).tw.
37. or/33‐36
38. exp vitrectomy/
39. vitrectom$.tw.
40. PPV$.tw.
41. or/38‐40
42. angiogenesis/
43. angiogenesis inhibitors/
44. angiogenesis factor/
45. monoclonal antibody/
46. exp endothelial cell growth factor/
47. vasculotropin/
48. (macugen$ or pegaptanib$ or lucentis$ or rhufab$ or ranibizumab$ or bevacizumab$ or avastin$ or aflibercept$ or conbercept$ or OPT 302 or Opthea$ or RTH258 or faricimab or brolucizumab or leizumabor or abicipar pegol).tw.
49. (anti adj2 VEGF$).tw.
50. (endothelial adj2 growth adj2 factor$).tw.
51. (anti adj1 angiogen$).tw.
52. VEGF TRAP$.tw.
53. or/42‐52
54. 37 and 41 and 53
55. 32 and 54

Appendix 4. LILACS search strategy

diabetic retinopathy [Words] and vitrectom$ or PPV$ [Words] and ranibizumab or bevacizumab or avastin or aflibercept or conbercept or OPT 302 or Opthea or RTH258 or faricimab or brolucizumab or leizumabor or abicipar pegol [Words]

Appendix 5. ISRCTN search strategy

(Diabetic Retinopathy) AND (vitretomy OR PPV) AND (Ranibizumab OR Bevacizumab OR Avastin OR Aflibercept OR Conbercept OR faricimab OR brolucizumab OR leizumabor)

Appendix 6. ClinicalTrials.gov search strategy

(Ranibizumab OR Bevacizumab OR Avastin OR Aflibercept OR Conbercept OR OPT 302 OR Opthea OR RTH258 OR faricimab OR brolucizumab OR leizumabor OR abicipar pegol ) AND (Diabetic Retinopathy) AND (vitretomy OR PPV) 

Appendix 7. WHO ICTRP search strategy

Diabetic Retinopathy = condition  AND Ranibizumab OR Bevacizumab OR Avastin OR Aflibercept OR Conbercept OR OPT 302 OR Opthea OR RTH258 OR faricimab OR brolucizumab OR leizumabor OR abicipar pegol = intervention

Data and analyses

Comparison 1. Anti‐VEGF versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Mean BCVA	13	699	Mean Difference (IV, Random, 95% CI)	‐0.25 [‐0.39, ‐0.11]
1.2 Early POVCH	14	1038	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.34, 0.58]
1.3 Late POVCH	11	579	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.30, 0.74]
1.4 Revision surgery for POVCH	4	207	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.15, 1.28]
1.5 Revision surgery for recurrent traction and/or RRD	4	145	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.15, 1.66]
1.6 Intraoperative retinal breaks	12	915	Risk Ratio (M‐H, Random, 95% CI)	0.37 [0.24, 0.59]
1.7 Silicone oil use	10	591	Risk Ratio (M‐H, Random, 95% CI)	0.46 [0.27, 0.80]

Comparison 2. Anti‐VEGF versus control: subgroup analysis ‐ time of administration of anti‐VEGF.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Mean BCVA	11	603	Mean Difference (IV, Random, 95% CI)	‐0.26 [‐0.41, ‐0.10]
2.1.1 Anti‐VEGF administered < 5 days preoperatively	7	396	Mean Difference (IV, Random, 95% CI)	‐0.30 [‐0.53, ‐0.07]
2.1.2 Anti‐VEGF administered ≥ 5 days preoperatively	4	207	Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.20, ‐0.06]
2.2 Early POVCH	13	877	Risk Ratio (M‐H, Random, 95% CI)	0.39 [0.28, 0.55]
2.2.1 Anti‐VEGF administered < 5 days preoperatively	8	618	Risk Ratio (M‐H, Random, 95% CI)	0.44 [0.29, 0.66]
2.2.2 Anti‐VEGF administered ≥ 5 days preoperatively	5	259	Risk Ratio (M‐H, Random, 95% CI)	0.26 [0.14, 0.51]

Comparison 3. Anti‐VEGF versus control: subgroup analysis ‐ main indication for vitrectomy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Mean BCVA	4	177	Mean Difference (IV, Random, 95% CI)	‐0.42 [‐0.72, ‐0.13]
3.1.1 Main indication for surgery: vitreous haemorrhage	3	137	Mean Difference (IV, Random, 95% CI)	‐0.25 [‐0.37, ‐0.13]
3.1.2 Main indication for surgery: tractional retinal detachment	1	40	Mean Difference (IV, Random, 95% CI)	‐1.31 [‐1.76, ‐0.86]

Comparison 4. Anti‐VEGF versus control: subgroup analysis ‐ bevacizumab versus other anti‐VEGF.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Mean BCVA	12	579	Mean Difference (IV, Random, 95% CI)	‐0.22 [‐0.36, ‐0.08]
4.1.1 Bevacizumab versus control	6	307	Mean Difference (IV, Random, 95% CI)	‐0.35 [‐0.79, 0.08]
4.1.2 All other anti‐VEGFs versus control	6	272	Mean Difference (IV, Random, 95% CI)	‐0.19 [‐0.28, ‐0.10]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Ahmadieh 2009.

Study characteristics
Methods	Prospective, randomised, double‐masked clinical trial Country: Iran
Participants	Number: 68 eyes of 68 participants undergoing PPV Age: mean (± SD) age was 55.2 ± 11.1 years (range 21 to 76 years) Sex: 34 (50%) males; 34 (50%) females Inclusion criteria: Non‐clearing vitreous haemorrhage Tractional retinal detachment involving or threatening the macula Active progressive PDR Exclusion criteria: BCVA of 20/40 or better Pregnancy History of IVB injection Intraoperative use of long‐acting gas or silicone oil Simultaneous intraocular surgery such as cataract extraction Monocular participants
Interventions	Intervention: IVB injection (1.25 mg) 1 week before surgery (n = 35) Comparator: Sham injection (n = 33)
Outcomes	Primary outcome measure: Incidence of early (≤ 4 weeks) postoperative vitreous haemorrhage Secondary outcome measures: Mean change in BCVA IVB‐related adverse events Follow‐up: 1 day, 1 week and 1 month after surgery
Notes	Trial registration number: NCT00524875 Date study conducted: January 2007 to December 2007 Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed by random block permutation according to a computer‐generated randomisation list.
Allocation concealment (selection bias)	Low risk	Details of the allocation were unknown to the investigators.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants and personnel were masked to the treatment method.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Visual acuity was measured by an optometrist who was masked to the groups. All pre‐ and postoperative examinations were performed by one of the authors (NS), who was also masked to the study group identification.
Incomplete outcome data (attrition bias) All outcomes	Low risk	A flow chart included within the results section accounted for all participants who failed to complete the study protocol.
Selective reporting (reporting bias)	Low risk	All primary and secondary outcomes were reported, based on both the intention‐to‐treat analysis and per‐protocol analysis.

Ahn 2011.

Study characteristics
Methods	Prospective, randomised, open‐label trial Country: South Korea
Participants	Number: 107 Age: mean age (± SD). Preoperative IVB group 51.0 (± 9.5), intraoperative IVB group 55.6 (± 10.6), and control 55.0 (± 11.4) Sex: (male/female): preoperative IVB group 23/13, intraoperative IVB group 21/16 and control 16/18 Inclusion criteria: Non‐clearing vitreous haemorrhage Macula‐involving or macula‐threatening TRD or fibrovascular proliferation with vitreoretinal adhesions Exclusion criteria: Intraoperative use of long‐acting gas or silicone oil Repeat vitrectomy after first vitrectomy for retinal diseases other than vitreous haemorrhage Previous history of vitrectomy Uncontrolled hypertension History of blood coagulopathy
Interventions	Intervention: IVB injection (1.25 mg/0.05 mL) 1 to 14 days before PPV IVB injection (1.25 mg/0.05 mL) at the end of PPV Comparator: No IVB injection
Outcomes	Primary outcome measures: Incidence of early (< 4 weeks) POVCH Incidence of late (> 4 weeks) POVCH Secondary outcome measures: Initial time of vitreous clearing BCVA at 6 months after surgery Follow‐up: 1 day, 1 week, 1, 3 and 6 months after surgery if there were no postoperative events
Notes	A total of 107 eyes of 91 participants, of which there were 16 bilateral participants, were included for analysis Trial registration number: NTC00745498 Date study conducted: June 2008 to October 2010 (from trials registry entry) Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was carried out using a permuted block analysis.
Allocation concealment (selection bias)	Low risk	Allocation was performed by a biostatistician.
Blinding of participants and personnel (performance bias) All outcomes	High risk	The study was open‐label.
Blinding of outcome assessment (detection bias) All outcomes	High risk	The study was open‐label.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The distribution and attrition rate of study participants was clearly described in a flow chart.
Selective reporting (reporting bias)	Low risk	All outcomes were reported.

Aleman 2019.

Study characteristics
Methods	Prospective, randomised, double‐masked clinical trial Country: Mexico
Participants	Number: 206 Age: mean age was 58.0 years (range 56.0 to 60.1 years) in IVZ group and 55.8 (range 53.9 to 57.8 years) in IVB group Sex: (male/female): preoperative IVZ group 44/38, preoperative IVB group 47/44 Inclusion criteria: Type I or type II diabetic patient Age between 18 and 85 years Active PDR Decreased VA principally from a non‐clearing vitreous haemorrhage, TRD, fibrous proliferation or a combination of them Snellen BCVA between light perception and 20/40 in the study eye at baseline Exclusion criteria: History of anterior or posterior vitrectomy in the study eye Significant lens or corneal opacity leading to 2 or more lines of reduced Snellen BCVA Significant retinal/optic nerve disease unrelated to diabetes mellitus leading to 2 or more lines of reduced Snellen BCVA Macula‐involving RD older than 6 months in the study eye IOP > 30 mmHg secondary to neovascular glaucoma Macular ischaemia observed as enlargement of FAZ on FFA leading to 2 or more lines of reduced Snellen BCVA Uncontrolled systemic hypertension (systolic > 200 mmHg or diastolic > 120 mmHg) History of intravitreal anti‐VEGF to the study eye within 3 months of enrolment
Interventions	Intervention: IVZ injection (1.25 mg/0.05 mL) 1 to 10 days before surgery (n = 99) Comparator: IVB injection (1.25 mg/0.05 mL) 1 to 10 days before surgery (n = 107)
Outcomes	Primary outcome measure: BCVA at 6 months Secondary outcome measures: Perioperative TRD development rates Surgical times Intraoperative complications Postoperative complications Rates of unplanned PPV during the study period Follow‐up: 15 ± 5 days, 40 ± 10 days, 185 ± 15 days (6 months)
Notes	206 participants randomised and 173 participants completed follow‐up Trial registration number: NCT02590094 Date study conducted: October 2016 to August 2018 Funding: none Conflict of Interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was carried out using a simulated coin toss programme.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Both participants and personnel were masked to the treatment method.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Unmasked vitreoretinal department faculty and fellows rendered the preoperative and postoperative care.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The distribution and attrition rate of study participants was clearly described in a flow chart.
Selective reporting (reporting bias)	Low risk	All primary and secondary outcomes were analysed and reported.

Arevalo 2019.

Study characteristics
Methods	Prospective, randomised, double‐masked, multi‐centre clinical trial Countries: multi‐centre (13 clinical sites from 9 countries) – PACORES (Pan‐American collaborative retina study group)
Participants	Number: 214 Age: mean age (± SD) was 59.5 ± 11.0 years in IVB group and 61.3 ± 10 in sham group Sex: (Male %): preoperative IVB group 62%, preoperative sham group 54% Inclusion criteria: Age: 18 years or older Type 1 or 2 diabetic patients PDR and TRD with or without rhegmatogenous component and with or without VH Exclusion criteria: Ocular: TRD due to cause other than diabetes Other ocular condition predisposing to poor visual outcomes postoperatively Treatment for DMO or DR at any time in the past 4 months with anti‐VEGF Major ocular surgery within the prior 4 months of randomisation History of Nd:YAG capsulotomy within 2 months before randomisation IOP ≥ 25 mmHg External ocular infection Systemic: Renal disease A condition precluding participation in the study Participation in an investigational trial within 30 days of randomisation that involved treatment with a non‐approved drug Known allergy to any component of the study drug Blood pressure >180/110 mmHg Major surgery within 28 days before randomisation or planned during the following 6 months to randomisation Myocardial infarction or other cardiac events requiring hospitalisation, stroke, TIA within 4 months before randomisation Pregnant or lactation or intending to become pregnant within 12 months from inclusion in the study Moving out of the area of the clinical centre during the first 12 months of the study Blood disease associated with abnormal coagulation and patients on anticoagulant therapy
Interventions	Intervention: IVB injection (1.25 mg/0.05 mL) 3 to 5 days before surgery (n = 102) Comparator: Sham injection 3 to 5 days before surgery (n = 112)
Outcomes	Primary outcome measure: Intraoperative bleeding Total surgical time Early (< 1 month) postoperative VH Mean change in BCVA at 12 months Secondary outcome measures: Number of endodiathermy applications Intraoperative retinal breaks Change in central macular thickness Proportion of eyes gaining at least 2 lines (10 letters) of ETDRS BCVA Follow‐up: 12 months
Notes	Trial registration number: NCT01976923 Date study conducted: November 2013 to July 2015 Funding: none reported Conflict of Interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed using an Internet‐based randomisation generator using the method of randomly permuted blocks.
Allocation concealment (selection bias)	Low risk	Allocation was performed by a retina specialist not involved in patient treatment or follow‐up.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	All follow‐up examinations were performed by personnel masked to the type of treatment. Patients were masked too.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Personnel performing follow‐up examinations were masked to the type of treatment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No comment was made. However, it is clear from the results that all participants completed follow‐up.
Selective reporting (reporting bias)	Low risk	All outcomes were reported.

Comyn 2017.

Study characteristics
Methods	Pilot, prospective, randomised, double‐masked, single‐centre clinical trial Country: United Kingdom
Participants	Number: 30 Age: mean age (± SD) was 48.7 ± 18 years in IVR group and 57.1 ± 14 in preoperative saline group  Sex: (male/female): preoperative IVR group 6/9 and preoperative saline group 3/12 Inclusion criteria: Patients having PPV and delamination for advanced PDR with fibrovascular complexes and/or TRD Exclusion criteria: Combined PPV and cataract surgery Patients with only a single focal point of vitreoretinal attachment Patients with advanced cataract or uncontrolled glaucoma Patients with BCVA in the fellow eye of 3/60 or poorer Patients with hypersensitivity in one active substance or excipients Cerebrovascular, cardiovascular or peripheral vascular disease
Interventions	Interventions: IVR injection (0.5 mg/0.05 mL) 7 ± 1 days before surgery (n = 15) Subconjunctival saline control 7 ± 1 days before surgery (n = 15)
Outcomes	Primary outcome measure: BCVA at 12 weeks Secondary outcome measures: Extent of TRD and macular perfusion at the time of surgery Ease of vitrectomy (duration, instrument usage and intraoperative haemorrhage) Presence of postoperative vitreous cavity haemorrhage Follow‐up: 12 weeks
Notes	Trial registration number: NCT01306981 Date study conducted: March 2011 to March 2013 (from ClinicalTrials.gov) Funding: by an unrestricted research grant from Novartis Pharmaceuticals UK Ltd; the National Institute for Health Research Biomedical Research Centre at Moorfields Eye Hospital National Health Service Foundation Trust and University College London Institute of Ophthalmology; and a multi‐user equipment grant from The Wellcome Trust Conflict of Interest: OC has received travel support from Novartis. None declared from the remaining authors.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed using random permuted blocks of varying sizes.
Allocation concealment (selection bias)	Low risk	The allocation was held by the trial statistician.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Both participants and personnel performing assessments were masked to the interventions.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Masked assessing investigators performed all postoperative assessments.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All patients completed follow‐up.
Selective reporting (reporting bias)	Low risk	All outcomes were reported.

Cui 2018.

Study characteristics
Methods	Prospective, randomised clinical trial Country: China
Participants	Number: 60 Age: mean age (± SD) was 60.74 ± 2.63 years in IVC group, 55.28 ± 5.16 in the IVR group and 57.49 ± 4.22 in IVTA group  Sex: (male/female): preoperative IVC group 9/8, preoperative IVR group 13/6 and preoperative IVTA group 10/5 Inclusion criteria: Patients having PPV for VH, PDR or TRD and other serious PDR Exclusion criteria: Previous intravitreal injection Previous vitreous or retinal surgeries Glaucoma Abnormal blood coagulation indexes Diseases of surgical contraindication
Interventions	Interventions: IVC injection (0.5 mg/0.05 mL) 3 to 7 days before surgery (n = 20) IVR injection (0.5 mg/0.05 mL) 3 to 7 days before surgery (n = 20) IVTA (4 mg/0.5 mL) during surgery (n = 20)
Outcomes	Primary outcome measure: BCVA monthly Operation time Incidence of iatrogenic retinal breaks Endodiathermy rate Silicone oil tamponade Secondary outcome measures: Average vitreous clearing time Frequency of intraoperative and postoperative bleeding PRP completion rate Reoperation probability Intraocular pressure Follow‐up: 6 months
Notes	60 patients randomised and 58 patients completed the follow‐up Date study conducted: January 2015 to December 2015 Funding: from the First Hospital of Qiqihar City, Heilongjiang, China, the National Natural Science Foundation of China, the US Department of Veterans Affairs and the Veterans Administration Medical Center at Memphis, TN Conflict of Interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“Patients were randomly divided…”
Allocation concealment (selection bias)	Unclear risk	Not stated in the report.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants were not masked to their study group.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The surgeries were performed by two masked surgeons.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The distribution and attrition rates of the participants are clearly presented in a flow chart.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

di Lauro 2009.

Study characteristics
Methods	Prospective, interventional randomised controlled trial Country: Italy
Participants	Number: 72 eyes of 68 participants Age: not given Sex: not given Inclusion criteria: Vitreous haemorrhage TRD Exclusion criteria: Neovascular glaucoma or cataract, or both Combined traction and rhegmatogenous retinal detachment (diagnosed either before or during the surgery)
Interventions	Intervention: IVB injection (1.25 mg) 1 week before the vitrectomy (n = 24) IVB injection (1.25 mg) 3 weeks before the vitrectomy (n = 24) Comparator: Sham injection: subconjunctival injection of 0.05 mL balanced salt solution 3 weeks before the vitrectomy (n = 24)
Outcomes	Primary outcome measures: Clearing of vitreous haemorrhage Incidence of adverse effects and the need for other procedures during the surgery Secondary outcome measures: Change in BCVA Duration of surgery Follow‐up: 1, 6, 12 and 24 weeks after surgery, and ultrasonography was performed before and 24 weeks after the injection
Notes	Trial registration number: NCT01025934 Date study conducted: October 2005 to May 2007 Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"We randomly assigned eligible patients... " and "The patients were randomly divided into three treatment groups (A, B, C)".
Allocation concealment (selection bias)	Unclear risk	Not stated in the report.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	A sham injection was used in the control group.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	A sham injection was used in the control group.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not stated in the report, however clear from the results that all participants completed follow‐up.
Selective reporting (reporting bias)	Low risk	Data for all main outcomes were analysed and presented.

Dong 2019.

Study characteristics
Methods	Prospective, randomised clinical trial Country: China
Participants	Number: 120 eyes of 120 participants undergoing PPV Age: mean (± SD) age was 56.61 ± 5.32 years (range 20 to 82 years) in the vitrectomy alone group and 57.31 ± 5.62 years (range 21 to 84) in the IVR group  Sex: male/female: 33/27 in the vitrectomy alone group and 35/25 in the IVR group Inclusion criteria: Meet the diagnostic criteria of PDR as set in the “Guidelines for Clinical Diagnosis and Treatment of Diabetic Retinopathy in China” Exclusion criteria: People with disabilities or mental illness that are unable to communicate normally Uncontrolled blood pressure and blood sugar Severe heart, liver, kidney and other systemic disease
Interventions	Intervention: IVR injection (0.5 mg) 3 to 5 days before surgery (n = 60) Comparator: Vitrectomy alone (n = 60)
Outcomes	Duration of operation Intraoperative complications Postoperative complications BCVA IOP Central macular thickness Follow‐up: 6 months
Notes	Date study conducted: October 2014 to November 2017 Funding: not reported Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“120 eyes of 120 patients were randomly divided into two groups…”
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not stated in the report, however clear from the results that all participants completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All primary and secondary outcomes were reported.

El Batarny 2008.

Study characteristics
Methods	Prospective, randomised, open‐label, controlled trial Country: Egypt
Participants	Number: 30 Age: mean (± SD). Control group 46 (± 12) and intervention group 44 (± 11) Sex: not given Inclusion criteria: TRD threatening or involving the macula TRD with vitreous haemorrhage Non‐clearing vitreous haemorrhage (present for more than 1 month) Massive preretinal/subhyaloid blood covering the posterior pole Exclusion criteria not stated
Interventions	Intervention: IVB (1.25 mg) 5 to 7 days before surgery (n = 15) Comparator: No treatment (n = 15)
Outcomes	Primary outcomes: Feasibility of surgery Postoperative complications Follow‐up: 1 day, 7 days, 2 weeks, 1 month and then monthly until end of study (all participants completed a minimum of 6 months)
Notes	Trial registration number: not reported Date study conducted: not reported Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Participants were randomised by diagnosis. Not clear statement on randomisation sequence.
Allocation concealment (selection bias)	Unclear risk	Not stated.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants were not masked to their study group.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Not mentioned, and treatment groups were different.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Farahvash 2011.

Study characteristics
Methods	Randomised, single‐masked (surgeon), controlled trial Country: Iran
Participants	Number: 35 Age: mean (range) 58.5 (37 to 73) Sex: (male/female): 18/17 Inclusion criteria: Vitreous haemorrhage > 1 month in a participant with no history of panretinal photocoagulation Non‐clearing vitreous haemorrhage in a participant with complete panretinal photocoagulation Vitreous haemorrhage with neovascularisation of the iris Vitreous haemorrhage with glaucoma Vitreous haemorrhage with retinal detachment Exclusion criteria: Vitrectomy or any intraocular injection in the study eye History of IVB in either eye Previous myocardial infarction or thromboembolic event Uncontrolled hypertension Use of anticoagulation (except aspirin, which was discontinued 1 week before injection) or coagulation abnormalities
Interventions	Intervention: IVB (1.25 mg) 7 days before surgery (n = 18) Comparator: No IVB (n = 17)
Outcomes	Primary outcomes: Severity of intraoperative bleeding and break formation Secondary outcomes Visual acuity at month 3 postsurgery Complications of surgery: retinal detachment, neovascular glaucoma Early and late postoperative vitreous haemorrhage Follow‐up: 1 day, 1 week, every 3 months (minimum follow‐up 3 months)
Notes	We contacted the authors to further discuss the severity of POVCH in each arm of the study. Although no formal grading scale was used to assess the POVCH, the authors described the clinical findings as follows: "We had two cases of early vitreous cavity haemorrhage (within 4 weeks of surgery), both in non injection group. The haemorrhage was dense and fundus obscuring, but resolved spontaneously during follow up. Both of them have been undergone vitrectomy for vitreous haemorrhage without TRD, with no air, SF6 or silicone as a tamponade. We had five cases of late vitreous cavity haemorrhage before writing of paper (occurring after more than 4 weeks), two in non injection group and three in injection group. Only in one participant in injection group, vitreous cavity haemorrhage was dense and fundus obscuring. In other participants, it presented as recurrent bouts of non fundus obscuring vitreous cavity haemorrhage". Trial registration number: not reported Date study conducted: January 2008 to January 2009 Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Participants were randomly assigned..." Not enough information on random sequence generation.
Allocation concealment (selection bias)	Unclear risk	Not stated.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study participants were not masked to which group they had been randomised to.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	All surgeons were masked regarding treatment group, but it was unclear if the outcome assessors were masked.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Participants who failed to complete the study follow‐up are clearly accounted for in the results.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Ferraz 2013.

Study characteristics
Methods	Prospective, interventional consecutive, randomised, clinical trial Country: Brazil
Participants	Number: 27 eyes of 27 participants undergoing PPV Age: mean (± SD) age was 52.2 ± 13.9 years  Sex: 17 males; 10 females Inclusion criteria: Severe PDR Persistent VH for 6 months Exclusion criteria: TRD TRD with RRD component Previous vitrectomy in the study eye
Interventions	Interventions: IVB injection (1.25 mg) 5 to 7 days before surgery (n = 9) IVTA (4.0 mg) 5 to 7 days before surgery (n = 9) PPV alone (n = 9)
Outcomes	Primary outcome measures: Incidence of early (< 4 weeks) recurrent VH Incidence of late (> 4 weeks) recurrent VH Follow‐up: 1 week, 1 month and 6 months after surgery
Notes	Trial registration number: not reported Date study conducted: not reported Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“The patients were randomized in order to assign…”
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not stated in the report, however clear from the results that all participants completed follow‐up.
Selective reporting (reporting bias)	High risk	The incidence of late (> 4 weeks) postoperative VH is not reported, although it is mentioned in the methods section of the report as one of the primary outcomes.

Hernández‐Da Mota 2010.

Study characteristics
Methods	Prospective, randomised, open‐label study Country: Mexico
Participants	Number: 40 Age: mean 55.7 (both groups) Sex: not given Inclusion criteria: Presence of advanced PDR Presence of TRD threatening the macula HbA1c < 7 Exclusion criteria: Macular involvement of retinal detachment Iris rubeosis Cataract ‐ when precluding proper evaluation of the fundus No perception of light vision Macular ischaemia
Interventions	Intervention: IVB (1.25 mg/0.05 mL), 48 hours before vitrectomy (n = 20) Comparator: No IVB (n = 20)
Outcomes	Primary outcomes: BCVA Feasibility of surgery (intraoperative bleeding, break formation) Postoperative complications (POVCH, iris rubeosis, and retinal detachment) Follow‐up: 1 week, 3 and 6 months
Notes	No response from authors to correspondence asking for clarification of POVCH definitions Trial registration number: not reported Date study conducted: not reported Funding: reported none Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Patients were randomly assigned into two groups",
Allocation concealment (selection bias)	Unclear risk	Not stated,
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants were not masked to treatment group,
Blinding of outcome assessment (detection bias) All outcomes	High risk	"For all patients, visual acuity and intraocular pressure were measured and recorded before PPV in a standardized, masked manner" However, not clearly stated whether outcome assessment was always masked and whether other outcomes, such as intraocular bleeding, were masked,
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although not directly stated, it is clear from the results that all participants completed the stated follow‐up,
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported,

Hu 2019.

Study characteristics
Methods	Prospective, randomised clinical trial Country: China
Participants	Number: 106 eyes of 84 participants undergoing PPV Age: mean (± SD) age was 62.5 ± 5.4 years for the IVR group and 63.0 ± 4.6 years for the control group Sex: male/female: 22/20 in the IVR group and 25/17 in the control group Inclusion criteria: Diabetic patients as per diagnostic criteria formulated by the Endocrinology Society of China Patients with PDR as per American Academy of Ophthalmology clinical guidelines No previous treatment for PDR Exclusion criteria: Glaucoma Eye infection Cataract Hypertensive patients Coagulation systemic diseases Other contraindications
Interventions	Intervention: IVR injection (0.5 mg) 1 week before surgery (n = 51) Comparator: Only PPV group (n = 55)
Outcomes	Outcome measures: BCVA Foveal thickness VEGF levels GAS6 levels SDF‐1 levels Surgical complications Follow‐up: 3 months
Notes	Trial registration number: not reported Date study conducted: January 2016 to January 2018 Funding: Yancheng Medical Science and Technology Development Program (No.YK2014016) Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed by random number table method.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants were not masked to their study group.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Surgeons were not masked to the intervention.
Incomplete outcome data (attrition bias) All outcomes	Low risk	It is not reported, however from the results it is clear that all participants completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Jiang 2020.

Study characteristics
Methods	Prospective, randomised pilot study Country: China
Participants	Number: 60 Age: mean age (± SD) was 55.54 ± 9.94 years in intraoperative IVC group and 53.5 ± 9.59 in control group Sex: male/female: intraoperative IVC group 6/9, control group 10/5 Inclusion criteria: PPV for PDR‐relate complications (non‐clearing VH, fibrovascular proliferation and macula‐involving or macula threatening TRD) Exclusion criteria: History of vitreoretinal surgery History of intravitreal injection of long‐acting gas or silicone oil at the end of surgery Ocular disease other than diabetic VH History of intravitreal anti‐VEGF within the 3 previous months Ocular surgery within the previous 6 months Uncontrolled hypertension History of coagulopathy Follow‐up less than 12 months
Interventions	Intervention: IVC injection (0.5 mg/0.05 mL) at the end of surgery (n = 15) Comparator: Only PPV (n = 15)
Outcomes	Primary outcome measure: Incidence of postoperative VH Secondary outcome measures: Initial time of vitreous clearing BCVA CRT Follow‐up: 1 day, 1 week, 1 month, 3 months, 6 months and 12 months
Notes	60 patients were randomised and 30 completed the 12‐months follow‐up Trial registration number: ChiCTR1800015751 Date study conducted: not reported Funding: National Natural Science Foundation of China (Grant No. 81870670 and Grant No.81500734), the Shanghai health bureau project (20184Y0215) and Shanghai science and technology commission project (Grant No. 18411965100). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Conflict of interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation according to randomised table.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants were not masked to the group.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessors were masked to the group.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The distribution and attrition rates of the participants are clearly presented in a flow chart.
Selective reporting (reporting bias)	Low risk	All outcomes were reported.

Li 2022.

Study characteristics
Methods	Pilot, randomised clinical trial Country: China
Participants	Number: 48 eyes of 48 participants undergoing PPV Age: mean (± SD) age was 49.9 ± 9.8 years Sex: 27 (56.3%) males; 21 (43.7%) females Inclusion criteria: Vitreous haemorrhage resulting from active PDR Exclusion criteria: NVG RRD Previous laser treatment or vitrectomy in the study eye History of thromboembolic events Systemic inflammatory, autoimmune or immunosuppressive disease Pre‐existing ocular disease (RVO, RAO, ARMD) Uncontrolled hypertension Previous ocular surgery Coagulation abnormality or use of an anticoagulative medication other than aspirin
Interventions	Intervention: IVR injection (0.5 mg) 1 day before surgery (n = 16) IVR injection (0.5 mg) 3 days before surgery (n = 16) Comparator: Sham subconjunctival injection 3 days before vitrectomy (n = 16)
Outcomes	Intraoperative outcomes (bleeding, iatrogenic tears, relaxing retinotomy, endodiathermy use, use of silicone oil)  BCVA VEGF, CTGF concentrations Follow‐up: not clearly reported
Notes	Trial registration number: ChiCTR‐ONC‐16009520 Date study conducted: November 2016 to July 2018 Funding: National Natural Science Foundation of Hunan Province and the Innovation and Development Project of Hunan Development and Reform Commission Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants were not masked to the intervention.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	All assessments were performed by one of the authors (JZ), who was masked to the study group identification.
Incomplete outcome data (attrition bias) All outcomes	High risk	The exact follow‐up period is not clearly stated.
Selective reporting (reporting bias)	High risk	There is no clear list of prespecified outcome measures for this study.

Manabe 2015.

Study characteristics
Methods	Prospective, double‐masked randomised controlled trial Country: Japan
Participants	Number: 62 people (66 eyes) Age: average 60 years (range 20 to 82) Sex: 82% male Inclusion criteria: Diagnosed as having an indication for primary vitrectomy because of persistent vitreous haemorrhage for over 3 months and TRD caused by PDR; or fundus findings of proliferative changes or vitreous haemorrhage in people with iris neovascularisation or neovascular glaucoma Exclusion criteria: Intraocular surgery or retinal photocoagulation within 3 months before the study Intravitreal injection of drugs or sub‐Tenon injection of steroids within 3 months before the study Cerebrovascular infarction or myocardial infarction within 3 months before the study
Interventions	Intervention: IVB 0.16 mg/0.05 mL 1 day before vitrectomy (n = 32) Comparator: Sham injection 1 day before vitrectomy (n = 34)
Outcomes	Primary outcome: Frequency of reoperation due to postoperative recurrent vitreous haemorrhage within 4 weeks after vitrectomy Secondary outcome measures: Numbers of intraoperative laser and endodiathermy spots Frequency of postoperative recurrent vitreous haemorrhage within 4 weeks after vitrectomy Concentration of VEGF in vitreous at the beginning of surgery
Notes	Trial registration number: NCT01854593 Date study conducted: June 2012 to August 2013 Funding: not reported Conflict of interest: reported "none"
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported.
Allocation concealment (selection bias)	Low risk	"...were randomised using the envelop method" "Both the investigators and patients were masked to the treatment assignment".
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"Both the investigators and patients were masked to the treatment assignment".
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"All postoperative examinations were performed by three vitreoretinal surgeons (H.S., T.H., H.N.) who also were masked to the study group allocation".
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised participants were followed up for 1 month.
Selective reporting (reporting bias)	Low risk	All outcomes on trial registry entry reported.

Modarres 2009.

Study characteristics
Methods	Prospective, single‐masked (surgeon), randomised clinical trial Country: Iran
Participants	Number: 40 Age: mean (± SD). Intervention group 55.8 (± 11.3) and control group 53.2 (± 11.7) Sex: not given Inclusion criteria: Diabetics who were candidates for vitrectomy Exclusion criteria: Presence of significant cataract to cause impairment of vision Previous IVB injection Other vitreoretinal pathology: past history of uveitis, retinal artery or vein occlusion Follow‐up: mean follow‐up 7 ± 3.6 months
Interventions	Intervention: IVB injection (2.5 mg) 3 to 5 days before operation (n = 22) Comparator: No treatment (n = 18)
Outcomes	Primary outcome measure: Facilitation of surgery and decrease in complications Secondary outcome measure: Anatomic and visual outcomes at 6 months Follow‐up: mean follow‐up 7 ± 3.6 months
Notes	Trial registration number: not reported Date study conducted: not reported Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"40 eyes of 40 diabetic participants who were candidates for vitrectomy were randomly assigned to receive ..."
Allocation concealment (selection bias)	Unclear risk	Not stated in report.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not stated in report.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	"The surgeons were masked as to the injection", but not clear if the outcome assessors were masked.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not stated in report, however clear from the results that all participants completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All data discussed within the primary outcome for facilitation of surgery were analysed.

Ou 2021.

Study characteristics
Methods	Prospective, randomised control trial Country: China
Participants	Number: 75 eyes of 64 patients Age: mean (± SD) age: preoperative IVC group 56.71 (± 12.65), only vitrectomy group 57.35 (± 10.98) Sex: male/female: preoperative IVC group 15/17, vitrectomy only group 14/18 Inclusion criteria: Diabetics who were candidates for vitrectomy Exclusion criteria: not reported
Interventions	Intervention: IVC injection (0.5 mg/0.05 mL) 5 days before PPV (n = 32) Comparator: No IVC injection, PPV only (n = 32)
Outcomes	Outcome measures: BCVA at 1 week, 1 month and 3 months CMT at 3 months VEGF, PIGF, bFGF concentrations Operation time Intraoperative blood loss Postoperative retinal oedema Complications within 3 months Follow‐up: 3 months
Notes	Trial registration number: not reported Date study conducted: January 2016 to October 2019 Funding: not reported Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was carried out using a random number table method.
Allocation concealment (selection bias)	Unclear risk	Not stated in the report.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not stated in the report.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not clear if the outcome assessors were blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	It is not clearly stated in the text, however from the results it is obvious that all patients completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Pakzad Vaezi 2014.

Study characteristics
Methods	Prospective, randomised, double‐masked pilot study Country: Canada
Participants	Number: 29  Age: mean (± SD) age: preoperative IVR group 51.6 (± 12.3), preoperative IVB group 52.8 (± 10.5) Sex: male/female: preoperative IVR group 6/15 and preoperative IVB group 11/14 Inclusion criteria: PDR with nonclearing VH and/or TRD requiring PPV Exclusion criteria: VH secondary to ocular disease other than diabetes
Interventions	Intervention: IVR injection (0.5 mg/0.05 mL) 1 week before surgery (n = 15) Comparator: IVB injection (1.25 mg/0.05 mL) 1 week before surgery (n = 14)
Outcomes	Outcome measures: Total surgical time Presence or absence of TRD Intraoperative bleeding Iatrogenic tears Endodiathermy use Endolaser use Silicone oil use
Notes	Trial registration number: NCT01760746 Date study conducted: July 2012 to April 2015 (from trial registry) Funding: not reported Conflict of interest: Drs. Albiani, Kirker, Merkur, Kertes, Eng, and Forooghian receive honoraria from Novartis. Drs. Pakzad‐Vaezi, Albiani, Kirker, Merkur, Kertes, Eng, and Forooghian receive research funding from Novartis. Dr. Fallah has no financial or proprietary interest in the materials presented herein.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A stratified randomisation algorithm was used.
Allocation concealment (selection bias)	Unclear risk	It is not stated who performed the allocation and whether they were masked or not.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Both investigators and participants were masked to the identity of the study drug.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Both investigators and participants were masked to the identity of the study drug.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although it is not clearly stated, it is obvious from the results that all patients completed follow‐up.
Selective reporting (reporting bias)	Low risk	All outcomes on trial registry entry reported.

Ren 2019.

Study characteristics
Methods	Prospective, randomised, controlled trial Country: China
Participants	Number: 50  Age: mean (range). Preoperative IVC group 57 (28 to 69) and control group 62 (46 to 80) Sex: male/female: preoperative IVC group 16/9 and control group 14/11 Inclusion criteria: Patients with non‐clearing VH and fibrovascular proliferation without TRD undergoing PPV Exclusion criteria: Previous PPV Stroke, thromboembolic event within 6 months prior to surgery Myocardial infarction in the previous 6 months contraindicating the withdrawal of platelet and anti‐coagulant medications Silicone oil use Long or complex surgery
Interventions	Intervention: IVC injection (0.5 mg/0.05 mL) immediately after surgery (n = 25) Comparator: No IVC injection (n = 25)
Outcomes	Outcome measures: BCVA CRT VH recurrence
Notes	Trial registration number: NCT03426540 Date study conducted: January 2017 to March 2018  Funding: this article is supported by National Natural Science Foundation of China (81600723) Conflict of interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients were randomly allocated according to randomised table.
Allocation concealment (selection bias)	Unclear risk	It is not stated who performed the allocation and whether they were masked or not.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	The surgeon performing the operations was masked. It is not stated whether the participants were masked.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The surgeon performing the operations was masked. However, it is not clearly stated who performed the postoperative examinations.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although it is not clearly stated, it is obvious from the results that all patients completed follow‐up.
Selective reporting (reporting bias)	Low risk	All outcomes on trial registry entry reported.

Rizzo 2008.

Study characteristics
Methods	Interventional, consecutive, randomised, prospective study Country: Italy
Participants	Number: 22 Age: mean (range). 52 years (24 to 63) Sex: not given Inclusion criteria: TRD Tractional rhegmatogenous retinal detachment TRD complicated by vitreous haemorrhage Exclusion criteria: History of vitrectomy History of thromboembolic events Major surgery in the 3 previous months or planned within 28 days Uncontrolled hypertension Known coagulation abnormalities or current use of anticoagulative medication other than aspirin
Interventions	Intervention: IVB (1.25 mg) 5 to 7 days before PPV (n = 11) Comparator: No IVB (n = 22)
Outcomes	Primary outcome measure: Feasibility of the surgery Secondary outcome measure: Visual and anatomic outcome at 6 months Follow‐up: up to 6 months
Notes	Trial registration number: not reported Date study conducted: not reported Funding: not reported Conflict of interest: reported that there were none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"We used a table of random numbers in order to assign each study participant to group 1 or 2".
Allocation concealment (selection bias)	Unclear risk	Not stated in report.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not stated in report.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not stated in report.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Not stated in report, however it is clear from the results that all participants completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	Results for the feasibility of surgery were stated, but no analysis was performed.

Sohn 2012.

Study characteristics
Methods	Prospective, double‐masked, interventional study Country: USA
Participants	Number: 20 eyes of 19 participants Age: median 52 years Sex: male/female: 12/7 Inclusion criteria: TRD involving the macula Combined tractional/rhegmatogenous detachments Non‐clearing or recurrent vitreous haemorrhage precluding panretinal photocoagulation, with tractional detachment not involving the macula Exclusion criteria: History of vitrectomy Dense vitreous haemorrhage preventing preoperative grading of fibrovascular membranes Inability to return for PPV within 3 to 7 days after randomisation History of stroke, thromboembolic event or myocardial infarction within 6 months Younger than 18 years Pregnancy
Interventions	Intervention: IVB (1.25 mg/0.05 mL) 3 to 7 days before vitrectomy (n = 10) Comparator: No treatment (n = 10)
Outcomes	Clinical outcomes: Median visual acuity at 3 months postsurgery Surgical complications: POVCH, cataract, neovascular glaucoma and retinal detachment Follow‐up: up to 3 months
Notes	—
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not stated in report.
Allocation concealment (selection bias)	Unclear risk	Not stated in report.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants in the control group underwent a sham injection.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Participants in the control group underwent a sham injection.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The number of participants lost to follow‐up was clearly stated.
Selective reporting (reporting bias)	Low risk	All outcomes were reported.

Su 2016.

Study characteristics
Methods	Randomised controlled trial Country: China
Participants	Number: 36 eyes of 36 participants Age: not reported Sex: not reported Inclusion criteria: Patients with nonclearing VH, fibrocellular proliferation with or without TRD undergoing PPV  Exclusion criteria: Previous PPV Menstruation Stroke Thromboembolic event and heart attack within 6 months thus contraindicating the withdrawal of antiplatelet and anticoagulative medications
Interventions	Intervention: IVC (0.5 mg) 7 days before vitrectomy (n = 18) Comparator: No treatment (n = 18)
Outcomes	Primary outcomes: VH clearing Incidence of adverse effects Need for other procedures during surgery Secondary outcomes: Changes in BCVA Operating time
Notes	Trial registration number: not reported Date study conducted: not reported Funding: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“Patients were randomly assigned…”
Allocation concealment (selection bias)	Unclear risk	Not stated in the report.
Blinding of participants and personnel (performance bias) All outcomes	High risk	The surgeon was masked to the intervention. However, participants were not blinded.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although not clearly stated, it is obvious from the results that all patients completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Wildan 2019.

Study characteristics
Methods	Randomised controlled trial Country: Indonesia
Participants	Number: 24 Age: mean (± SD). IVA group 51 (± 12) and IVB group 53.50 (± 9). Sex: not reported Inclusion criteria: Age 40 to 60 years Type II diabetic patients No photocoagulation, intravitreal anti‐VEGF, vitrectomy and other surgery in the past Exclusion criteria: Cloudy refractive media Stroke Heart disease Inflammation or intraocular infection
Interventions	Intervention: IVA (0.5 mg) 4 to 7 days before vitrectomy (n = 12) Comparator: IVB (1.25 mg) 4 to 7 days before vitrectomy (n = 12)
Outcomes	Clinical outcomes: BCVA at 1 week and 1 month Surgical complications
Notes	Trial registration number: not reported Date study conducted: January 2015 to July 2016 Funding: not reported Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“Subjects who met the inclusion criteria were then randomized.”
Allocation concealment (selection bias)	Unclear risk	Not stated in the report.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	The surgeon was masked to the intervention. However, there is no report about the participants.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although not clearly stated, it is obvious from the results that all patients completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Yang 2016.

Study characteristics
Methods	Prospective, single‐centre, single‐dose, randomised, controlled clinical trial Country: China
Participants	Number: 107 Age: mean (± SD). IVC group 48.63 (± 8.24) and control group 49.64 (± 8.71) Sex: male/female: IVC group 27/27 and control group 24/29 Inclusion criteria: Age ≥ 18 years Type I or II diabetic patients HbA1c ≤ 10% Visual loss only due to nonabsorbent VH, fibrovascular proliferation with vitreoretinal adhesions or TRD Exclusion criteria: Coagulopathy Hypertension Stroke or TIA Intraocular treatment with steroids or anti‐VEGF or intraocular surgery at any time IOP ≥ 21 mmHg Glaucoma Iris neovascularisation in either eye Severe retinopathy requiring long‐acting gas or silicone oil use Follow‐up less than 3 months
Interventions	Intervention: IVC (0.5 mg/0.05 mL) 3 days before vitrectomy (n = 54) Comparator: Only vitrectomy (n = 53)
Outcomes	Clinical outcomes: Severity of intraoperative bleeding Incidence of recurrent VH Vitreous clearance time BCVA levels BCVA improvement IOP Endophthalmitis Rubeosis TRD Systemic adverse events
Notes	Trial registration number: not reported Date study conducted: March 2014 to January 2015 Funding: supported by Yunnan Applied Basic Research Projects Items (no. 2014FB093), Science and Technology for People‐Benefit Projects of Yunnan Province (no. 2014RA070), and Special‐Purpose Fund of Wang Longde Academician Workstation Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“Patients were randomised into two categories.”
Allocation concealment (selection bias)	Unclear risk	Not stated in the report.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	The surgeons were masked to the intervention. However, there is no report about the participants.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessors were masked to treatment group.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Participant flow is clearly presented in a flow chart.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Yu 2015.

Study characteristics
Methods	Randomised controlled trial Country: China
Participants	Number: 125 Age: mean (range) 47.27 (27 to 77) Sex: 59 males and 66 females Inclusion criteria: T1DM or T2DM > 3a Significantly decreased VA  Exclusion criteria: not reported
Interventions	Intervention: IVC (0.5 mg) 2 to 6 days before vitrectomy (n = 61) Comparator: No treatment (n = 64)
Outcomes	Primary outcomes: Operation time Iatrogenic retinal holes Postoperative VA  Postoperative complications
Notes	Trial registration number: not reported Date study conducted: January 2011 to December 2014 Funding: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	It is stated in the manuscript that patients were randomly assigned to treatment groups.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although not clearly stated, it is obvious from the results that all patients completed follow‐up.
Selective reporting (reporting bias)	Unclear risk	All outcomes were reported.

Zaman 2013.

Study characteristics
Methods	Randomised, open‐label, controlled trial Country: Pakistan
Participants	Number: 54 Age: mean (± SD) 52.07 (± 5.54); range 39 to 67 Sex: male/female: 32/22 Inclusion criteria: Non‐clearing vitreous haemorrhage of at least 1 month TRD involving or threatening the macula Pre‐retinal subhyaloid bleeding covering the macula Exclusion criteria: not stated
Interventions	Intervention: IVB (1.25 mg/0.05 mL) 1 week prior to surgery (n = 24) Comparator: No treatment (n = 30)
Outcomes	Primary outcome measures: BCVA after surgery Postoperative complications, including rate of early and late POVCH, frequency of hyphaema and rubeosis Follow‐up: days 1, 7, 14 and then monthly for up to 6 months
Notes	We attempted to contact the corresponding author to clarify details in the methodology and results, but have received no response Trial registration number: not reported Date study conducted: September 2010 to August 2011 Funding: not reported Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Patients were randomised into two categories". No further details provided.
Allocation concealment (selection bias)	Unclear risk	Not commented on within the article.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Although no specific comment was made, the study appears to be open‐label.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Although no specific comment was made, the study appears to be open‐label.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants completed the study protocol.
Selective reporting (reporting bias)	Unclear risk	All stated outcomes were presented in the results.

Zhou 2017.

Study characteristics
Methods	Randomised, controlled trial Country: China
Participants	Number: 42 eyes from 35 patients Age: mean. IVC group 72.46 and no IVC group 52.07  Sex: male/female: IVC group 8/10 and no IVC group 8/9 Inclusion criteria: PDR Persistent non‐absorbing VH > 3 weeks No intravitreal injection 12 months prior to PPV No history of laser and eye surgery Good blood sugar control Good liver and kidney function  Exclusion criteria: not reported
Interventions	Intervention: IVC (0.5 mg) 3 to 7 days prior to surgery (n = 18) Comparator: No treatment (n = 17)
Outcomes	Outcome measures: Operating time Iatrogenic retinal tears Intraoperative bleeding BCVA Central foveal thickness Follow‐up: 6 months
Notes	Trial registration number: not reported Date study conducted: October 2014 to August 2016 Funding: not reported Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	It is stated in the manuscript that patients were randomly assigned to treatment groups.
Allocation concealment (selection bias)	Unclear risk	Not commented on within the article.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not commented on within the article.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not commented on within the article.
Incomplete outcome data (attrition bias) All outcomes	Low risk	From the results it is obvious that all participants completed the study protocol.
Selective reporting (reporting bias)	Unclear risk	All stated outcomes were presented in the results.

ARMD: age‐related macular degeneration
BCVA: best‐corrected visual acuity
bFGF: basic fibroblast growth factor
CMT: central macular thickness
CRT: central retinal thickness
CTGF: connective tissue growth factor
DMO: diabetic macular oedema
DR: diabetic retinopathy
FAZ: foveal avascular zone
GAS6: growth arrest‐specific protein 6
IOP: intraocular pressure
IVB: intravitreal bevacizumab
IVC: intravitreal conbercept
IVR: intravitreal ranibizumab
IVTA: intravitreal triamcinolone acetonide
IVZ: intravitreal Ziv‐aflibercept
NVG: neovascular glaucoma
PDR: proliferative diabetic retinopathy
PIGF: placental growth factor
POVCH: postoperative vitreous cavity haemorrhage
PPV: pars plana vitrectomy
PRP: panretinal photocoagulation
RAO: retinal artery occlusion
RD: retinal detachment
RRD: rhegmatogenous retinal detachment
RVO: retinal vein occlusion
SD: standard deviation
SDF‐1: stromal cell‐derived factor 1
T1DM/T2DM: type I/type II diabetes mellitus
TIA: transient ischaemic attack
TRD: tractional retinal detachment
VA: visual acuity
VEGF: vascular endothelial growth factor
VH: vitreous haemorrhage

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Abd 2020	Study group did not receive vitrectomy, it received panretinal photocoagulation
Abdelhakim 2011	This prospective case series included 20 eyes and found the following results: Mean pre‐injection BCVA (logMAR) was 1.460 ± 0.439 Mean BCVA on day 1, week 1, months 2 and 3 after surgery were 1.645 ± 0.422, 1.300 ± 0.413, 1.065 ± 0.538, and 1.065 ± 0.538 logMAR, respectively (P = 0.078, 0.123, 0.002, and 0.002, respectively) Intraoperative bleeding was minimal in most cases (85%, n = 17) Postoperatively, 16 participants had no bleeding (80%), 4 had minimal bleeding (20%) and 1 had recurrent fibrovascular proliferation (5%) The study was excluded as it was non‐randomised
Albuquerque 2014	Anti‐VEGF agent was administered again 6 weeks postoperatively
Antoszyk 2020	Study arm did not receive vitrectomy
Antoszyk 2021	Study arm did not receive vitrectomy
Berk Ergun 2014	This study was excluded as it is retrospective
Bhavsar 2014	Vitrectomy was an outcome rather than an intervention
Castillo 2017	No control arm; all arms received the same anti‐VEGF
Castillo 2018	No control arm; all arms received the same anti‐VEGF
Chakraborty 2021	Retrospective design
Chatziralli 2020	Study arm did not receive vitrectomy after ranibizumab
Cheema 2010	This study was excluded as it is retrospective
Chelala 2018	No vitrectomy performed
Chen 2020	Non‐randomised study
Comyn 2019	No outcomes of interest reported
da R Lucena 2009	Did not assess POVCH
Demir 2013	This study was excluded as it is retrospective
Dirani 2014	Vitrectomy was an outcome rather than an intervention; initially none of the groups received vitrectomy
Doganay 2010	This prospective case‐controlled study was excluded due to lack of randomisation. Participants were divided into 2 groups: Group 1 (n = 32) received 1.25 mg of IVB 1 week before surgery; Group 2 (n = 50) did not. Results: Recurrent vitreous haemorrhage was found in 4 participants (12.5%) from Group 1 and in 14 participants (28%) from Group 2 An increase in the mean BCVA value was determined in 24 participants (75%) from Group 1 and in 32 participants (64%) from Group 2
Elwan 2013	Study has been retracted
Feng 2018	Control arm included eyes with non‐diabetic ocular disease
Forooghian 2016	No outcomes of interest reported
Gao 2020	All arms received anti‐VEGF injection
Goncu 2014	Excluded as not randomised
Gupta 2008	Excluded as not randomised
Gupta 2012a	This study was excluded due to its retrospective design
Gupta 2012b	Excluded as this is a retrospective observational study
Haseeb 2019	Not randomised
Hu 2021	Retrospective design
Hu 2022	No outcomes of interest reported
Ishikawa 2009	This study was excluded because of small numbers and no control group; also no comment made on POVCH
Jiao 2014	No outcomes of interest reported
Jin‐An 2014	No outcomes of interest reported
Jirawison 2012	Excluded as this is a retrospective observational study
Jorge 2021	Both groups received the same anti‐VEGF; one group did not receive vitrectomy
Karimov 2021	Retrospective design
Kasetty 2021	Both study arms received aflibercept; no control group
Kawano 2013	Not randomised
Leonard 2014	All patients received the same anti‐VEGF preoperatively
Li 2010	This study was excluded due to its retrospective design
Li 2015	Study group included macular hole patients as well as patients with proliferative diabetic retinopathy
Li 2016	Retrospective design
Li 2020	Control group included non diabetic patients
Liu 2019	All study arms received anti‐VEGF injection; outcomes of interest not reported
Lo 2009	This study was excluded due to its retrospective design
Ma 2016	Non randomised study
Marcus 2017	All study arms received the same anti‐VEGF; no control group
Meng 2016	Vitrectomy was an outcome rather than an intervention; initially no treatment group was supposed to receive vitrectomy
NCT01589718	Withdrawn prior to enrolment
Oshima 2009	This study was excluded because of the retrospective enrolment of participants, the absence of randomisation and the different study periods of the two groups
Ozer 2022	No control group ‐ everyone received the same anti‐VEGF; no outcomes of interest reported
Papavasileiou 2017	Retrospective design
Park 2010	This study was excluded due to its retrospective design
Park 2021	Vitrectomy was an outcome rather an intervention; initially no patient was supposed to have vitrectomy
Pokroy 2011	This study was excluded due to its retrospective design
Romano 2009a	Found the following results: No reduction in the rate of early grade 2 and 3 POVCH (2/32 (6%) at 7 days and 1 month) following preoperative bevacizumab Rates of late POVCH (grade 2 and 3): 2/32 (6%) at 3 months and 3/32 (9%) at 6 months A significant improvement in 29/32 eyes (mean BCVA 1.6 (SD 1.2) to 0.4 (SD 0.8) logMAR, P = 0.02) with surgery. There was no improvement secondary to POVCH (grade 2 to 3) in the remaining 3 eyes. No clinically significant intraocular inflammation on the first postoperative day. Cataract formation occurred in 12/22 (54%). The incidence of grade 2 and 3 POVCH at 1 day, 1 week, 1, 3 and 6 months was 9%, 6%, 6%, 6% and 9%, respectively (n = 32) This prospective study was excluded due to the lack of a control group. Participants also received an additional dose of bevacizumab at the end of surgery, which makes interpretation of the effect of the preoperative bevacizumab on POVCH difficult.
Romano 2009b	Found the following results: The incidence of grade 2 and 3 POVCH in the initial postoperative period was 4/30 (14%) at 7 days and 11/30 (36%) at 1 month Rates of late POVCH (grade 2 and 3) were 9/30 (30%) at 3 and 6 months Visual improvement in 21/30 treated eyes (mean BCVA 1.00 (SD 0.9) to 0.4 logMAR). There was no improvement secondary to POVCH in the remaining eyes (9/30, P = 0.2). No clinically significant intraocular inflammation on the first postoperative day. Cataract formation occurred in 9/22 (40%) of phakic participants during the 6‐month follow‐up. The incidence of grade 2 and 3 POVCH was 20%, 14%, 36%, 30% and 30% (n = 30) This prospective study was excluded due to the lack of a control group
Sato 2013	This study was excluded due to its retrospective design
Shi 2020	No control group; all arms received anti‐VEGF injection
Starnes 2019	All study arms received the same anti‐VEGF agent; no control group
Sun 2017	No outcomes of interest reported
Sun 2022	Did not report any outcomes of interest (it only measures levels of cytokines)
Tsubota 2019	Non randomised study; no outcomes of interest reported
Turkseven Kumral 2021	Retrospective study
Victor 2014	No outcomes of interest reported
Vidinova 2020	Non randomised study
Wang 2019	No outcomes of interest were reported
Wang 2020	No outcomes of interest reported
Yang 2008	Found the following results: A significant reduction in the vitreous clear‐up rate postoperatively in participants who received preoperative bevacizumab and C3F8 endotamponade, when compared with their historical control group (7.2 (SD 5.6) vs 15.2 (SD 11.4) days (P = 0.04)). No significant reduction in the rate of early POVCH (defined as recurrent haemorrhage that obscured the retinal vessels ‐ grade 2 or above in the Diabetic Retinopathy Vitrectomy Study, for more than 1 week after vitreous clear‐up) was observed between the treatment and control groups (0/16 vs 1/24 (P = 0.41)). The rate of late POVCH in participants who received preoperative bevacizumab and C3F8 endotamponade was 1/16 (6.25%) and in the historical control group 1/24 (4%). This study was excluded due to the small numbers and lack of randomisation (participants in the control group were selected retrospectively). Also C3F8 was used in addition to preoperative anti‐VEGF in all participants in the treatment group.
Yeh 2009	Excluded as participants allocated by alternation
Yeoh 2008	Found the following results: No significant reduction in the overall rate of POVCH (7/18 (38.8%)), in 18 cases treated with 1.25 mg of bevacizumab within 2 weeks of surgery. When the results are divided into early and late POVCH, 2/18 cases had early POVCH (11.1%) and 5/18 had late POVCH (27.7%). A late rebleed rate of 27.7% (5/18). The authors suggested that this rate may be secondary to later reperfusion of fibrovascular tufts, once the effect of bevacizumab has worn off, which would not have been apparent at the time of surgery. Visual acuity had improved in 14/18 treated eyes, was unchanged in 1/18 and was worse in 3/18. The mean improvement in postoperative visual acuity was not quantified because of the significant number of participants with hand movement vision preoperatively. Of the 3 participants who had reduced visual acuity, 2 had extensive macular ischaemia on fluorescein angiography, and the remaining eye still contained silicone oil. 7/18 (38.8%) treated eyes had POVCH within the 6‐month follow‐up; 6 of the 7 eyes required surgical washout (6/18 (33.3%)). No adverse events in any participant treated with bevacizumab, local or systemic. This study was excluded due to the lack of a control group.
Yeung 2010	This study was excluded as it is retrospective
Yin 2017	Not randomised
Zeng 2017	Non‐randomised study
Zhou 2018a	Control group included eyes without diabetic pathology
Zhou 2018b	Non‐randomised

BCVA: best‐corrected visual acuity
IVB: intravitreal bevacizumab
logMAR: logarithm of minimal angle of resolution
POVCH: postoperative vitreous cavity haemorrhage
SD: standard deviation
VEGF: vascular endothelial growth factor

Characteristics of ongoing studies [ordered by study ID]

ISRCTN79120387.

Study name	A randomised, single‐masked, phase IV pilot study of the efficacy and safety of adjunctive intravitreal Avastin® (bevacizumab) in the prevention of early postoperative vitreous haemorrhage following diabetic vitrectomy
Methods	Randomised, single‐masked, controlled trial
Participants	30
Interventions	Avastin® will be administered intravitreally in a single‐ or dual‐dose regimen of 1.25 mg (in 0.05 mL) 2 weeks prior to vitrectomy and at the end of vitrectomy if internal tamponade (oil, air or gas) was not used. No sham intravitreal injections will be given before or after vitrectomy if participant is randomised to the usual treatment group.
Outcomes	Primary: Postoperative vitreous haemorrhage based on 2 masked clinical assessments at 6 weeks and 6 months Secondary: Recruitment and dropout rates Rates of re‐operation for recurrent vitreous haemorrhage Rate of postoperative rubeosis and rubeotic glaucoma Rate and severity of intraoperative bleed Mean change in ETDRS acuity and MNRead acuity Serum Avastin® and growth factor levels at 2 and 4 to 6 weeks after injection. Vitreous levels at 2 weeks after injection
Starting date	2007
Contact information	Mr Lyndon da Cruz Moorfields Eye Hospital, 162 City Road, London
Notes	No reply from lead investigator

NCT00516464.

Study name	Evaluation of ranibizumab on the ease of procedure and complication rate in proliferative diabetic retinopathy (PDR) requiring vitrectomy
Methods	Randomised, single‐masked, controlled trial
Participants	40
Interventions	40 participants will be enrolled and randomised 3:1 to ranibizumab or vitrectomy alone. 30 consented, enrolled participants will receive 3 intravitreal injections of 0.5 mg ranibizumab administered 1 to 3 weeks pre‐procedure, intraoperatively, and 1 month post‐vitrectomy
Outcomes	To evaluate the effect of ranibizumab on vitrectomy complications such as recurrent vitreous haemorrhage, postoperative retinal detachment rates, and development of intraoperative retinal breaks (time frame: 6 months)
Starting date	August 2007
Contact information	Retina Associates, PA. Shawnee Mission, Kansas, United States, 66204 Contact: Lexie Manning     913‐831‐7400 lmanning@kcretina.com [mailto:lmanning%40kcretina.com?subject=NCT00516464, FVF4295, Evaluation of Ranibizumab in Proliferative Diabetic Retinopathy (PDR) Requiring Vitrectomy]
Notes	Principal Investigator: Gregory M Fox, MD. Contacted: Lead author clarified that they are unlikely to have any data relevant to this review and will probably not go on to publish data from this trial

NCT00931125.

Study name	Randomized, double blinded, controlled, two‐center study assessing the safety and efficacy of intravitreal ranibizumab as a preoperative adjunct treatment before vitrectomy surgery in proliferative diabetic retinopathy (PDR) compared to vitrectomy alone
Methods	Randomised, double‐masked, controlled trial
Participants	Active, not recruiting Estimate 70
Interventions	This is a randomised, double‐masked, controlled, 2‐centre study assessing the feasibility, efficacy and safety of intravitreal ranibizumab injection applied as a preoperative adjunct treatment before vitrectomy surgery in severe PDR. Comparator arm consists of participants receiving standard vitrectomy alone with sham intravitreal injection preoperatively.
Outcomes	Primary outcome measures: Efficacy of preoperative intravitreal ranibizumab (time frame: OP day) (designated as safety issue: no) Efficacy, measured by surgical time, number of intraoperative bleedings, intraoperative retinal breaks, required endodiathermy Secondary outcome measures: Change in BCVA (time frame: 6 months) (designated as safety issue: no) Effect in anatomical changes (time frame: 3 ± 1 days after injection) (designated as safety issue: no) Safety (time frame: over 6 months) (designated as safety issue: yes) Retinal circulation integrity (time frame: month 1, 3, 6) (designated as safety issue: yes). Evaluating the circulation of original retinal vessels, evaluating the size of proliferative vessels (size of leaking areas and number of leaking points measured by fluorescein angiography)
Starting date	March 2009 Estimated study completion date: December 2013
Contact information	University of Debrecen, Medical and Health Science Center, Faculty of Medicine, Department of Ophthalmology
Notes	Debrecen, Hungary, H‐4012

NCT01091896.

Study name	Adjuvant intravitreal bevacizumab in pars plana vitrectomy for vitreous hemorrhage secondary to diabetic retinopathy
Methods	Randomised, single‐masked, controlled trial
Participants	Not stated
Interventions	No intervention: 1 ‐ no bevacizumab. Participants will neither receive bevacizumab before nor during vitrectomy Experimental: 2 ‐ bevacizumab before vitrectomy. Participants will receive intravitreal injection of 1.25 mg of bevacizumab (0.05 mL) 7 days before vitrectomy Experimental: 3 ‐ bevacizumab after vitrectomy. Participants will receive intravitreal injection of 1.25 mg of bevacizumab (0.05 mL) at the end of vitrectomy
Outcomes	Primary outcome measure: recurrent vitreous haemorrhage incidence after vitrectomy Secondary outcome measure: visual outcome
Starting date	2010 Estimated study completion date: March 2011
Contact information	felipeppalmeida@yahoo.com.br [mailto:felipeppalmeida%40yahoo.com.br?subject=NCT01091896, IBEViH, Adjuvant Intravitreal Bevacizumab in Pars Plana Vitrectomy for Vitreous Hemorrhage]
Notes	Principal Investigator: Felipe Almeida

NCT01151722.

Study name	Adjuvant intravitreal bevacizumab in pars plana vitrectomy for diabetic vitreous hemorrhage (ABeVi)
Methods	Randomised, single‐masked, controlled trial
Participants	Not stated
Interventions	No intervention: no injection, no bevacizumab Experimental 2: bevacizumab before vitrectomy Experimental 3: bevacizumab after vitrectomy
Outcomes	Primary outcome measure: vitreous haemorrhage recurrence (time frame: 3 months)
Starting date	2009 Estimated study completion date: December 2009
Contact information	felipeppalmeida@yahoo.com.br [mailto:felipeppalmeida%40yahoo.com.br?subject=NCT01151722, HCRP22, Adjuvant Intravitreal Bevacizumab in Pars Plana Vitrectomy for Diabetic Vitreous Hemorrhage]
Notes	Principal Investigator: Felipe Almeida, MD

NCT01306981.

Study name	A prospective, randomised controlled trial of ranibizumab pre‐treatment in diabetic vitrectomy ‐ a pilot study
Methods	Randomised, double‐masked, controlled study
Participants	Not stated
Interventions	Participants will be randomised 1:1 to receive either ranibizumab intravitreal injection (0.5 mg in 0.05 mL) or saline subconjunctival injection at the time of surgery
Outcomes	Primary outcome measures: BCVA (time frame: 12 weeks postoperative) Secondary outcome measures: Ease of performing vitrectomy surgery Incidence of postoperative vitreous haemorrhage (time frame: 6 weeks postoperative) Extent of retinal neovascularisation (time frame: 6 weeks postoperative) Extent of tractional retinal detachment (time frame: 1 week postinjection) Extent of macular perfusion (time frame: 12 weeks postoperative) Vitreous and serum levels of ranibizumab and related cytokines (time frame: 1 week postinjection) Incidence of postoperative vitreous haemorrhage (time frame: 12 weeks postoperative) Extent of retinal neovascularisation (time frame: 12 weeks postoperative)
Starting date	2011 Estimated study completion date: March 2013
Contact information	Moorfields Eye Hospital NHS Foundation Trust
Notes	Principal Investigator: Mr James W Bainbridge MA, PhD, FRCOphth. Contacted: Currently the study is in the write‐up stages. Unpublished data on the rate of POVCH are as follows: Vitreous cavity haemorrhage occurred in 4/15 in the ranibizumab (Lucentis®) arm and 6/15 in the control arm At 12‐weeks postoperative persistent vitreous cavity haemorrhage was observed in 0/15 in the ranibizumab arm and 2/15 in the control arm

BCVA: best‐corrected visual acuity
ETDRS: Early Treatment Diabetic Retinopathy Study
PDR: proliferative diabetic retinopathy
TRD: tractional retinal detachment
VEGF: vascular endothelial growth factor

Differences between protocol and review

Due to the importance of postoperative retinal detachment, we included an additional analysis to enable patients and surgeons to make a more informed choice when deciding to use intravitreal bevacizumab as an adjunct to vitrectomy.

Our protocol specified that we would exclude participants who received silicone oil endotamponade in the review analysis. Peer review comments on our review questioned whether this would give rise to bias, so we presented analyses of all participants and sensitivity analyses excluding silicone oil cases. These additional analyses were not originally planned within the protocol.

Our protocol specified that we would present odds ratios or risk ratios. We have chosen to present risk ratios as they are easier to interpret.

For the 2023 update, in response to a request from the Cochrane Eyes and Vision (CEV) editorial base, we changed the title of the review from "Anti‐vascular endothelial growth factor for prevention of postoperative vitreous cavity haemorrhage after vitrectomy for proliferative diabetic retinopathy" to "Anti‐vascular endothelial growth factor in combination with vitrectomy for the treatment of complications of proliferative diabetic retinopathy". The CEV editors had noted overlap between the current review and another CEV review "Anti‐vascular endothelial growth factor for proliferative diabetic retinopathy". After discussion with the review author teams of both reviews, it was agreed to modify the PICO framework for each updated review to ensure that relevant indications and outcomes were covered, but with no overlap between the two reviews. We also updated the outcomes following updated guidance from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021).

Contributions of authors

For the 2023 update:

JS, ND, PD,  DS

• Co‐ordinating the review

• Screening search results

• Data management

• Screening retrieved papers against inclusion criteria

• Appraising quality of papers

• Abstracting data from papers

• Analysis of data

• Interpretation of data

• Entering data into RevMan

• Writing the review

• Updating the review

Sources of support

Internal sources

• No sources of support provided

External sources

• Public Health Agency, UK

  This review update 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.

Declarations of interest

For the 2021 update (taken from CoI forms completed on 24 March 2021):

JS: No conflicts to declare
ND: No conflicts to declare
PD: No conflicts to declare
DS: No conflicts to declare

New search for studies and content updated (no change to conclusions)