Key messages

• Preoperative intravitreal injection of bevacizumab reduces the incidence of vitreous hemorrhage in the early postoperative period and probably do so in the late postoperative period.
• We are uncertain whether the use of bevacizumab prior to vitrectomy leads to better visual acuity as the certainty of the evidence has been assessed as very low.
• Preoperative use of bevacizumab probably reduces total surgical time.

What is the evidence.
See evidence matrix  in Epistemonikos later

We found five systematic reviews [2], [3], [9], [10], [11] including 16 primary studies [5], [6], [7], [8], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], of which 14 are randomized trials [4], [5], [6], [7], [12], [13], [15], [16], [17], [19], [20], [21], [22], [23].

This table and the summary in general are based on the latter,  since the observational studies did not increase the certainty of the existing evidence, nor did they provide relevant additional information.

What types of patients were included*

Most trials included patients without distinction by sex or age. Only one trial excluded patients under 18 years of age [4]. The age reported by the different trials ranged from 44 [12] to 62 years [5]. 

All trials included patients with indication of pars plana vitrectomy for complications of proliferative diabetic retinopathy, mainly vitreous hemorrhage [5], [7], [8], [12], [13], [15], [19], [21], [22], [23] and tractional retinal detachment [5], [7], [8], [12], [13], [15], [16], [17], [19], [20], [21], [22], [23]. Only one trial excluded patients with vitreous hemorrhage of any grade [16], while another trial excluded cases of dense vitreous hemorrhage preventing preoperative grading of fibrovascular membranes [5]. One trial did not consider cases of tractional-rhegmatogenous retinal detachment [19] and one trial excluded retinal detachment with macular involvement [20]. 

There was great variability in the exclusion criteria reported. It should be noted that most trials excluded patients with previous intraocular surgery, especially vitreoretinal surgery [5], [6], [8], [13], [16], [17], [21], [22]; five trials excluded patients with history of myocardial infarction, stroke or thrombotic events [4], [7], [13], [16], [22]; four trials excluded patients with blood coagulopathy, alterations in coagulation tests or use of antithrombotics other than aspirin [13], [16], [21], [22]. Only two trials excluded pregnant women [5], [15] and two excluded previous intravitreal injection of bevacizumab [15], [17]. 

The trials included 674 eyes in 649 patients, all of them adults. Four trials included more than one eye per patient [5], [8], [19], [21].

What types of interventions were included*

All trials compared intravitreal injection of bevacizumab before vitrectomy against vitrectomy alone. Only three trials used simulated injection in the control group [5], [15], [19]. 

Intravitreal bevacizumab was used in the intervention group in different administration protocols. The dose was mostly 1.25 mg in 0.05 ml [5], [6], [7], [12], [13], [15], [19], [20], [21], [22], [23]. Two trials used higher doses of bevacizumab in the same concentration: one of them used 2.5 mg [17] and another 1.5 mg [16]. The concentration of bevacizumab was the same in all trials except in the most recent one that used a significantly lower concentration with a dose of 0.16 mg in 0.05 ml [8]. 

The timing of the intervention ranged from 1 to 20 days before surgery. Most trials administered bevacizumab at least 7 days prior to vitrectomy [5], [6], [7], [8], [12], [13], [15], [17], [20], [22]. One trial administered bevacizumab 14 days before the procedure [16]. 

In two trials patients were randomized to one of  three groups: control and two intervention groups, which varied in the preoperative time intervals [19], [21]: 7 or 20 days before vitrectomy respectively [19] and 1 to 14 days prior vitrectomy, with an average of 4.9 days [21]. Data obtained from the 37 eyes of this study [21] receiving intraoperative bevacizumab were excluded when meta-analyses were performed.

What types of outcomes
were measured

There is great variability in the outcomes measured by the trials. These can be classified into intraoperative variables, postoperative course and adverse effects:

• Intraoperative variables: total surgical time, intraoperative bleeding, use of endodiathermy (cauterization of retinal blood vessels), iatrogenic retinal tears and use of tampons such as silicone, oil, gas, perfluorocarbon.
• Postoperative outcomes: Best corrected visual acuity, early (< 4 weeks) or late (> 4 weeks) postoperative vitreous hemorrhage, resorption time of vitreous hemorrhage, attachment of the retina.
• Injection-related and surgery-related adverse events: systemic adverse events and systemic thrombosis, mainly myocardial infarction or stroke, local adverse effects such as endophthalmitis, retinal detachment, increased intraocular pressure, new cases of rubeosis, neovascular glaucoma, cataract progression, corneal erosions, anterior chamber reaction, uveitis, among others.

To whom this evidence does and does not apply

• This evidence applies to adults with indication of vitrectomy for complications of proliferative diabetic retinopathy, particularly vitreous hemorrhage and tractional retinal detachment.
• Patients with previous intraocular surgeries other than cataract surgery, especially vitreoretinal surgeries, were excluded.
• This evidence does not apply to patients with history of major cardiovascular events such as myocardial infarction, stroke or systemic thrombosis, and patients with coagulopathy, abnormal coagulation tests and/or use of anticoagulants.
• Pregnant women were also excluded.

• The outcomes included in this summary are consistent with those considered critical for decision-making by most trials, systematic reviews and the main clinical guidelines.
• The evidence on how patients value different outcomes in diabetic retinopathy shows that visual acuity is the most important [24], [25]. Unfortunately, the effect of using bevacizumab prior to vitrectomy on this outcome is still unclear.

• It is not possible to estimate risk/benefit balance properly since there is uncertainty with respect to visual acuity. However, there is evidence that shows benefits in postoperative vitreous hemorrhage and the most important surgical variables.
• None of the trials reported cases of systemic adverse events for either of the two treatment arms. However, the follow-up carried out by the trials was limited; 6 months in most trials, but in some cases as short as one month [8], [15].
• The systemic application of bevacizumab in patients with colorectal cancer has been associated with higher treatment-related mortality, serious cardiovascular adverse effects, hypertension and arterial thrombotic events [26]. Considering plasmatic presence of the drug has been described after intravitreal administration [27], it is reasonable to take caution when considering the evidence presented.
• However, this risk has not been demonstrated in the intravitreal application of bevacizumab. Systematic reviews that assessed the safety of this intervention in different ophthalmological conditions have not shown an increased risk of serious adverse effects, such as major cardiovascular events, but more evidence is needed to carry out an adequate evaluation [28], [29] [30], [31]. A recent publication by the Canadian Agency for Drugs and Technology in Health, concludes that it is still not clear whether the risk of thromboembolic events is significantly higher in a population using intravitreal anti-VEGF compared to a similar population (that has a high basal risk) that does not use this treatment, and that the risk would not be greater in one type of anti-VEGF compared to another [32].
• Therefore, with the intravitreal application of bevacizumab, the cardiovascular and systemic risks evidenced with its systemic application have not been reported so far. However, the current evidence is insufficient to obtain definitive conclusions.
• Regarding to local adverse events, these occurred with low frequency in the included trials (mainly cases of early transient elevation of intraocular pressure) and there was no difference between the intervention and control groups. No cases of endophthalmitis were reported in the six trials that reported it. However, this is probably due to the small sample size. It is known that the risk of endophthalmitis is very low, but it exists [32]. A recent meta-analysis, which analyzed 503,890 intravitreal injections of various anti-VEGF, reports an average incidence of 0.039% for bevacizumab, with no significant differences when compared with ranibizumab or aflibercept [33]. This is a point to bear in mind, but it should also be noted that the included patients with severe ophthalmological conditions requiring vitrectomy that have a high-risk of complications such as endophthalmitis, which occurs on average in 0.05% of cases [34].

• Because the use of bevacizumab prior to a vitrectomy has been recently proposed and its clinical effects are still under investigation, no studies assessing the cost-effectiveness of this intervention have been conducted.
• The positive effect of bevacizumab on intraoperative variables such as total surgical time could contribute to lower costs and make the intervention more efficient. The favorable effect on clinical outcomes could contribute to improve the quality of life. However, its effect on visual acuity is still unclear.
• Since bevacizumab is currently a high-cost drug, its use prior to a vitrectomy should be assessed in detail in contexts where there are major resource constraints.

• Due to the great importance that patients attribute to visual acuity, in scenarios without major resource constraints, some patients may be inclined to use a therapy with a potential benefit in this outcome, although the certainty of the evidence is very low.
• Physicians may be inclined to use this option for its favorable effects on intraoperative variables such as total surgical time in addition to the decrease in the incidence of postoperative vitreous hemorrhage.

Differences between this summary and other sources

• In general terms, our conclusions are concordant with the included systematic reviews identified regarding the outcomes of early postoperative vitreous hemorrhage, visual acuity, total surgical time, use of endodiathermy, iatrogenic retinal tears and adverse effects.
• With regard to late postoperative vitreous hemorrhage, there are differences in the results obtained by our review and three of the systematic reviews [2], [3], [10]. These are mainly explained by differences in the primary studies included. Our meta-analysis is the only one that considers all randomized trials with information available on this outcome. However, the certainty of the evidence is moderate so it could change with future trials.
• It is expected that the effect of preoperative bevacizumab on the incidence of vitreous hemorrhage in the late postoperative period (> 4 weeks) is less than in the early postoperative period (< 4 weeks) or even null, as the half-life of the drug is short. Pharmacokinetic studies, both in animal models and humans, have shown the half-life of bevacizumab in the vitreous humour would be approximately 4 to 5 days [35], [36], [37], and would be minimally detectable up to 30 days. However, it could be detected in the retina for up to 14 days [18], and there could be effects that persist independent of the presence of the drug. New formulations of bevacizumab are being studied, such as sustained-release models, that could eventually increase the effect of the drug on the late postoperative period [37]. The half-life would be modified with the dose administered [38], so the most effective dose would have to be assessed. Furthermore, the half-life of the drug in vitrectomized eyes would be considerably lower [36] and the clinical effect might also be less in previously vitrectomized eyes, cases that were excluded by most of the randomized trials. All these pharmacokinetic aspects should be investigated in advance with adequate clinical studies that provide sufficient evidence to formally assess if they might have real clinical implications or not.
• With regard to the occurrence of intraoperative bleeding, there are subtle differences in relation to the results and interpretation of the evidence, which are due mainly to the primary studies included and the statistical analysis model used for comparison (in our case we use relative risk). Similar to the other three systematic reviews [2], [9], [10], our results show that there is no certainty whether preoperative bevacizumab  decreases the occurrence of intraoperative bleeding because the certainty of the evidence is very low. In three [5], [12], [20] of the six trials reporting data for this outcome, intraoperative bleeding occurred in almost all patients in both treatment groups. A probable interpretation is that the effect of bevacizumab decreases the severity of intraoperative bleeding rather than the occurrence of it, which would be difficult to avoid given the context of neovascularization presented by these patients. In this sense, the frequency of use of endodiathermy was considered by several trials as an indicator of bleeding severity, and the evidence shows with more certainty that the use of preoperative bevacizumab would decrease the number of intraoperative applications of endodiathermy.
• The key messages of our summary are concordant with the main guidelines identified, which suggest bevacizumab should be considered preoperatively in eyes with proliferative diabetic retinopathy undergoing vitrectomy [39], [40]. Likewise, it is not opposed to other guidelines that indicate that this therapeutic application is still in investigation [41], since the certainty of the  available evidence is not the optimal regarding several outcomes.

• It is probable that the conclusions associated with the effect of the preoperative use of bevacizumab in eyes with proliferative diabetic retinopathy undergoing vitrectomy will change in the future, particularly in relation to their effect on the visual acuity.
• Three randomized trials responding this question were identified in the International Clinical Trials Registry Platform of the World Health Organization, but they had planned to complete the recruitment more than 5 years ago and still do not report results [42], [43], [44].
• We did not identify ongoing systematic reviews in PROSPERO (International prospective register of systematic reviews) assessing this question.