Treatment of choroidal neovascularization using intravitreal bevacizumab


Wael Soliman MD
Department of Ophthalmology
Glostrup Hospital
Nordre Ringvej 57
DK-2600 Glostrup
Tel: + 45 43 23 48 17
Fax: + 45 43 23 46 69


Purpose:  This study aimed to assess the pharmacodynamic profile of intravitreal bevacizumab in relation to best corrected visual acuity (BCVA), foveal thickness, and other aspects of macular morphology after intravitreal injection of bevacizumab in eyes with subretinal choroidal neovascularization (CNV).

Methods:  A retrospective observational, uncontrolled case series including 26 eyes in 25 patients followed for up to 6 months after intravitreal injection of bevacizumab 1 mg repeated as deemed necessary after monthly assessments by biomicroscopy, optical coherence tomography, colour fundus photography, fluorescein angiography and BCVA determination. At follow-up, cases were classified by morphological treatment response (reduction or elimination of pathological neovascular leakage, retinal thickening or serous retinal detachment) or absence of response (deterioration or lack of improvement). Primary disease entities included age-related macular degeneration (22 eyes, four of which had evidence of retinal angiomatous proliferation), idiopathic peripapillary neovascularization (one eye), and angioid streaks (three eyes in two patients).

Results:  One month after the first injection, apparent morphological improvement was observed in 24/26 eyes and mean BCVA had improved by 3.1 ± 7.8 letters (p = 0.05). Of these 24 responders, which included all primary diagnoses, 11 (46%) demonstrated BCVA improvement of ≥ 5 letters. The two non-responders (7.7%) had lost > 3 lines of vision at 2 months follow-up. Overall, 18 eyes completed 6 months follow-up, with a mean BCVA improvement of 0.5 ± 12.7 letters, and 22 eyes completed 3 months follow-up, with a mean BCVA improvement of 2.0 ± 11.0 letters. Two months after the first injection, 11 (46%) of the 24 responders demonstrated signs of recurrent CNV activity, defined as decreased BCVA and/or increased retinal thickness and/or fluorescein angiographic CNV leakage. No serious drug-related adverse events were observed during the course of the study.

Conclusions:  Overall mean BCVA remained stable throughout the study. Morphological signs of reduced CNV activity were seen in the majority of eyes at 2–4 weeks after intravitreal bevacizumab injection. Half the responders showed signs of renewed CNV activity at 2 months after their first injection. All first-injection responders were also second-injection responders.


Vascular endothelial growth factor is an important mediator of subretinal choroidal neovascularization (CNV). Phase III studies of ranibizumab (Genentech, Inc., South San Francisco, CA, USA), an antibody fragment with broad vascular endothelial growth factor (VEGF)-subtype affinity, has shown remarkable results in neovascular age-related macular degeneration (AMD) (Gaudreault et al. 2005; Brown et al. 2006; Rosenfeld et al. 2006). Preliminary information about the efficacy and safety of ranibizumab released before regulatory approval prompted off-label use of a related broad-spectrum VEGF-inhibitor, the full-size antibody bevacizumab (Avastin; Genentech, Inc.) (Michels et al. 2005; Avery et al. 2006; Maturi et al. 2006; Rich et al. 2006; Spaide et al. 2006). Improved visual function and reduced CNV leakage following intravitreal bevacizumab administration has been reported in the majority of study patients in uncontrolled short-term case reports, but, as expected, most patients need retreatment. The safety of intravitreal bevacizumab has not been systematically assessed and the optimal retreatment interval has yet to be determined. Potential safety issues include the observation of retinal pigment epithelium tears following intravitreal bevacizumab treatment of CNV in AMD (Meyer et al. 2006). In addition to CNV in AMD, intravitreal bevacizumab has also been used in peripapillary and myopic CNV (Nguyen et al. 2005; Laud et al. 2006; Soliman et al. 2006; Yamamoto et al. 2006).

The present report describes our experience with intravitreal bevacizumab in a consecutive series of patients with subretinal neovascularization of varying aetiologies.

Materials and Methods

The study included 26 eyes in 25 consecutive patients who underwent intravitreal bevacizumab treatment at the Copenhagen County Department of Ophthalmology at Herlev Hospital or, from June 2006, at the Glostrup Hospital, Copenhagen County, Denmark. One patient, whose data are not included in this paper, has been previously reported (Soliman et al. 2006). The study comprised all patients who followed a predefined clinical protocol for assessments and injection procedures. The criteria for offering patients intravitreal bevacizumab treatment were failure to respond to prior therapy by involution of the CNV, the patient requesting such therapy, or lack of evidence from a controlled trial that other therapy could be of benefit for the patient's type of CNV-lesion.

The study comprised 26 eyes in 25 patients, 11 male and 14 female, aged 73 ± 14 years (mean ± standard deviation [SD]). The apparent primary aetiology was AMD (22 patients, four of whom demonstrated retinal angiomatous proliferation), angioid streaks in pseudoxanthoma elasticum (two patients, one of whom was treated in both eyes) or aggressive peripapillary neovascularization (one patient) that was unlikely to benefit from photocoagulation therapy because of massive haemorrhage and subfoveal extension.

Treatment consisted of intravitreal injection through the pars plana with 1 mg bevacizumab, a dose chosen on the basis of the favourable response observed in our first published case report (Soliman et al. 2006). The clinical protocol comprised follow-up at 2 weeks, 1 month and 2 months after each injection, with a repeated injection tentatively scheduled after 2 months, provided that signs of CNV activity were present at this time. The tentative retreatment interval was a conservative choice, compared with the 4-week interval applied in the pivotal ranibizumab trials (Brown et al. 2006; Rosenfeld et al. 2006). The choice was motivated by the heightened need to justify repeat intravitreal injection of a medication that has not been approved for this purpose and a shortage of injection procedure capacity at our clinic, as well as the longer half-life of bevacizumab compared with ranibizumab (Lucentis; Genentech, Inc., South San Franciso, CA, USA) (Gaudreault et al. 2005; Beer et al. 2006). The patients' needs for retreatment were evaluated every second month according to the following algorithm:

  • 1evidence of overall improvement compared with baseline plus no recurrence (no activity)→observe and follow up after 1 month;
  • 2evidence of overall improvement or stability compared with baseline plus signs of recurrence (activity)→retreatment with 1 mg bevacizumab, or
  • 3evidence of no improvement compared with baseline plus progression→stop treatment and shift to other treatment (Lucentis or intravitreal triamcinolone acetonide[IVTA] and/or photodynamic therapy[PDT]).

Patients were fully informed about the potential adverse effects of the injection, the limited efficacy and safety data on the intravitreal use of bevacizumab for CNV and the observation of thrombosis and arterial hypertension in cancer patients who received intravenous bevacizumab infusion.

Best corrected visual acuity (BCVA) was assessed according to the ETDRS (Early Treatment Diabetic Retinopathy Study) protocol at a standard distance of 4 m. If < 20 letters were read, the patient was moved to a distance of 1 m and given the opportunity to read the first six lines at this distance, which was then added to the total BCVA score. If > 20 letters were read at 4 m, 30 letters were added to the nominal score. At baseline, biomicroscopy, fluorescein angiography, red-free and colour fundus photography (FF 450 + Zeiss, Jena, Germany) and transfoveal optical coherence tomography (OCT) (Stratus III OCT; Carl Zeiss, Dublin, CA, USA) were performed in both eyes. Except for fluorescein angiography, which was not carried out at 2 weeks, these procedures were repeated at each follow-up visit. In some patients it was not possible to obtain an OCT scan of sufficient quality due to poor fixation.

Intravitreal injection was administered under aseptic conditions. The conjunctival sac was instilled with a local anaesthetic (0.4% oxybuprocaine) and a topical antibiotic (0.5% chloramphenicol). Before draping, the conjunctival sac and the surroundings of the eye were washed using 5% povidone iodine. After draping, a speculum was used to separate the eyelids. After subconjunctival injection of 1% lidocaine, bevacizumab 1 mg (0.04 ml of 25 mg/ml) (Avastin®; Roche A/S, Hvidovre, Denmark) was injected through the pars plana of the ciliary body. The patient was instructed to use chloramphenicol four times daily for 3 days after the injection. Seven days after the injection, the patient was contacted by telephone and interviewed to identify any symptoms of post-injection complications.

A repeat injection was tentatively scheduled 8 weeks after injection and performed if the treating physician found that a comprehensive evaluation of clinical data supported the existence of CNV activity. Protocol-defined signs of CNV activity included intraretinal and subretinal fluid accumulation, new intraretinal or subretinal haemorrhage, and CNV growth. Decreasing visual acuity (VA) was not considered sufficient evidence to diagnose CNV activity, but in the presence of ambiguous morphological signs of CNV activity, decreasing VA was considered to support the diagnosis of CNV activity and hence to warrant retreatment.

At every visit, patients were routinely interviewed about changes in visual performance, ocular and systemic symptoms, changes in medication, and hospital admissions.


Differences in VA scores and foveal retinal thickness between baseline and the various visits were analysed using a paired two-tailed t-test analysis. The p-value was considered significant at p ≤ 0.05.


All 25 patients completed 1 month of follow-up (26 eyes); 21 completed 3 months of follow-up (22 eyes), and 17 completed 6 months of follow-up (18 eyes). The median follow-up on bevacizumab-only treatment was 6 months and the median number of injections per eye at the latest visit was two. Seven eyes in six patients had received either PDT or pegaptanib prior to the initiation of bevacizumab treatment (Table 1). The treated eye was the eye with the better BCVA in 16/25 patients.

Table 1.   Characteristics of eyes included in the study, showing primary disease, subtype subretinal neovascular lesion and prior treatment.
Eye numberDiseaseMembrane classificationPrior treatment
  1. AMD = age-related macular degeneration; AS = angioid streaks; RAP = retinal angiomatous proliferation; PPM = peripapillary membrane; Pred C = predominantly classic; Min C = minimally classic.

1AMDPred CNo
3AMDPred CNo
6AMDPred CNo
10AMDPred CNo
17AMDPred CNo
21AMDPred CNo
22AMDMin CMacugen
25AMDMin CNo

Two patients with predominantly classic CNV and pigment epithelial detachment (PED) appeared to demonstrate no morphological or functional improvement after bevacizumab injection. Unambiguous disease progression was seen in both. One was subsequently treated with PDT and intravitreal triamcinolone injection, and the other with PDT only. The remaining 6 patients who left the study before 6 months were all shifted to Lucentis therapy.

Major adverse events included extensive postoperative subretinal and intravitreal haemorrhage in one eye of a patient with AMD and retinal angiomatous proliferation. The patient was being treated with warfarin and the haemorrhage occurred 9 weeks after his first injection (Fig. 1). Other ocular adverse events included mild ocular pain during or after the injection, conjunctival hyperaemia, subconjunctival haemorrhage, foreign body sensation and floaters. No adverse systemic effects considered to be of relevance were noted during the study.

Figure 1.

 (A) Colour fundus photography, (B) fluorescein angiography, and (C) horizontal optical coherence tomography cross-sections of the macula (width 6 mm) in the right eye of a 70-year-old man with age-related macular degeneration, retinal angiomatous proliferation with intraretinal oedema, serous retinal detachment, and subfoveal detachment of the retinal pigment epithelium. Best corrected visual acuity (BCVA) was 0.9 (80 ETDRS letters) at baseline. The patient received intravitreal bevacizumab 1 mg in the left eye. One month later, the retinal profile was normal and VA was unchanged. At 2 months after the initial injection, subfoveal CNV activity had recurred and BCVA was reduced to 0.4 (64 ETDRS letters).

Mean BCVA scores were 54.6 ± 14.8 letters (n = 26, Snellen equivalent 0.25) at baseline, 57.0 ± 12.7 letters (n = 22, Snellen equivalent 0.3) at 2 weeks, 57.7 ± 15.6 letters (n = 26, Snellen equivalent 0.3, p = 0.05) at 1 month, 57.5 ± 14.6 letters (n = 21, Snellen equivalent 0.3) at 2 months, and 55.8 ± 15.6 letters (n = 22, Snellen equivalent 0.25) at 3 months. The 18 eyes that completed 6 months of follow-up had an average BCVA score at this time-point of 53.7 ± 17.1 letters (Snellen equivalent 0.25) (Table 2, Fig. 2A, B). Only the changes in BCVA between baseline and 1 month proved to be significant (p = 0.05). Two patients lost during the study > 3 lines of vision (7.7%).

Table 2.   Mean visual acuity in relation to duration of intravitreal bevacizumab treatment in patients with choroidal neovascularization.
BCVABaseline2 weeks1 month2 months3 months6 months
  1. BCVA = best corrected visual acuity.

ETDRS54.6 ± 14.857.0 ± 12.757.7 ± 15.657.5 ± 14.655.8 ± 15.653.7 ± 17.1
(Snellen equivalent)(0.25)(0.3)(0.3)(0.3)(0.25)(0.25)
Number of eyes262226222218
Figure 2.

 Changes in ETDRS best corrected visual acuity (BCVA) in relation to time after first intravitreal injection of bevacizumab 1 mg in patients with choroidal neovascularization. (A)  Patients who completed 3-months follow-up (n = 22 eyes). (B)  Patients who completed 6-months follow-up (n = 18 eyes). Baseline visual acuity (time 0) was determined ≤ 1 week before the first injection.

Retinal thickening was found to be reduced or eliminated in the majority of patients 1 month after the most recent bevacizumab injection (Figs 1 and 3–10), as were angiographic CNV leakage, haemorrhage, serous detachment of the neurosensory retina, and pigment epithelium detachment (Table 3, Figs 1 and 7–10). When normalized to the value found immediately before the first injection, foveal thickness was 86 ± 21% at 1 month after injection (p = 0.002), 79 ± 32% at 2 months after injection (p = 0.009), 81 ± 23% at 3 months after injection (p = 0.0003) and 61 ± 51% at 6 months after injection (p > 0.05).

Figure 3.

 Thickness of the foveal subfield by optical coherence tomography (OCT) 1 month after intravitreal bevacizumab injection (n = 24).

Figure 4.

 Thickness of the foveal subfield by optical coherence tomography (OCT) 2 months after intravitreal bevacizumab injection (n = 23).

Figure 5.

 Thickness of the foveal subfield by optical coherence tomography (OCT) 3 months after intravitreal bevacizumab injection (n = 24).

Figure 6.

 Thickness of the foveal subfield by optical coherence tomography (OCT) 6 months after intravitreal bevacizumab injection (n = 16).

Figure 7.

 (A) Colour fundus photography, (B) fluorescein angiography and (C) horizontal optical coherence tomography (OCT) cross-sections of the macula (width 6 mm) in the right eye of a woman aged 75 years. The patient presented with subfoveal choroidal neovascularization secondary to age-related macular degeneration. Best corrected visual acuity (BVCA) was 0.5 Snellen in the left eye and hand movements in the right. The patient received two courses of verteporfin photodynamic treatment delivered 3 months apart. Seven months after the initial treatment, BCVA was 0.1 Snellen (34 ETDRS letters) (baseline). The patient then underwent intravitreal injection of bevacizumab 1 mg. One month later, the patient presented with reduced intraretinal and subretinal fluid, reduced angiographic leakage, and BCVA 0.125 Snellen (39 ETDRS letters). Two months after injection, biomicroscopic and OCT thickening revealed renewed CNV activity and BCVA was 0.125 (35 ETDRS letters). Repeated injection of bevacizumab was followed by flattening of the retina and improvement of BCVA to 0.16 Snellen (46 ETDRS letters) at 3 months.

Figure 8.

 (A) Colour fundus photography, (B) fluorescein angiography, and (C) horizontal optical coherence tomography (OCT) cross-sections of the macula (width 6 mm) in the left eye of a man aged 37 years who presented with subfoveal choroidal neovascularization (CNV) secondary to angioid streaks and best corrected visual acuity (BCVA) of 0.3 in both eyes. The patient received repeated courses of verteporfin photodynamic therapy in both eyes followed by only temporary reduction of CNV activity. Loss of BCVA to 0.2 in the right eye and 0.1 in the left eye prompted treatment with intravitreal pegaptanib 0.3 mg every 6 weeks in both eyes. Five injections were given in the right eye and four in the left. CNV activity and BCVA stabilized at about 0.2 in both eyes, but resumption of CNV activity nasal to the fovea then prompted a change to treatment with intravitreal bevacizumab 1 mg in both eyes. Compared with the examination at 6 weeks after the last pegaptanib injection (baseline), follow-up 1 month later demonstrated reduced CNV leakage and flattening of the macula. BCVA remained stable (46 ETDRS letters). Two months after the first bevacizumab injection, a mild recurrence of leakage and thickening was observed, together with the loss of seven ETDRS letters of BCVA in the left eye. Retreatment with bevacizumab was performed on the same day. One month later, reduced leakage and reduced thickening were observed together with BCVA recovery to 0.2 (50 ETDRS letters).

Figure 9.

 (A) Colour fundus photography, (B) fluorescein angiography (FA), and (C) horizontal optical coherence tomography (OCT) cross-sections of the macula (width 6 mm) in an 82-year-old caucasian woman referred with an idiopathic peripapillary subretinal neovascular membrane in the right eye. Snellen best corrected visual acuity (BCVA) was 0.9 at presentation in the right eye, but deteriorated to 0.2 within a few months without treatment because of the subretinal haemorrhage, argon laser photodynamic therapy were deferred. The patient was then treated with 1 mg bevacizumab intravitreally. Baseline BCVA was 66 in the right eye. After 1 month, leakage from the membrane was decreased, as seen on late-stage FA, and subretinal fluid had been resorbed. BCVA had fallen to 56. Two months after injection, BCVA had improved to 65 ETDRS letters and the morphological improvement seen at the 1-month visit persisted. Three months after the first injection, signs of progression in the disease were observed, with the appearance of subretinal fluid accumulation and increased leakage on late-stage FA. Despite this, vision remained stable and the patient was re-injected.

Figure 10.

 Transfoveal optical coherence tomography (width 6 mm) in the right eye of a woman aged 75 years with occult-only subfoveal choroidal neovascularization secondary to age-related macular degeneration. At baseline, a prominent retinal pigment epithelium detachment (PED) with overlying thickening and surrounding serous detachment of the neurosensory retina was seen. One month after intravitreal injection of bevacizumab 1 mg, the height of the PED was reduced and the thickness of the fovea had normalized. Partial recurrence was seen 2 months after injection, prompting renewed injection, which was followed by the best-yet status after treatment 1 month later. At baseline BCVA was 0.3 Snellen (59 ETDRS letters). Three months later, this improved to 0.4 Snellen (68 ETDRS letters).

Table 3.   Characteristics of eyes with subretinal choroidal neovascularization at baseline and after intravitreal bevacizumab injection. Assessment of change was based on open-label visual comparative grading of characteristics at baseline and at conclusion of study follow-up. Mean time to conclusion of study follow-up was 4.8 months.
CNV lesion elementBaseline
(no. of eyes)
Evolution by lesion element status at completion of study follow-up
  1. CNV = choroidal neovascularization; FA = fluorescein angiography; RPE = retinal pigment epithelium.

FA leakage2519(76%)3(12%)3(12%)
Hard exudate111(9%)4(36%)6(55%)
RPE detachment169(56%)3(19%)4(25%)
Subretinal fluid1710(59%)3(18%)4(24%)


The present case series documents the outcome of CNV after intravitreal bevacizumab treatment in AMD and other primary diseases of the retina or choroid. Our observations in this uncontrolled, retrospective study support previously published case series in that the outcome from 1 month onwards was better than has been reported with monotherapy using photodynamically activated verteporfin (Kaiser 2006; Pieramici et al. 2006).

We also found evidence that, in the majority of our patients, morphological abnormalities were occasionally normalized 1 month after injection, whereas, at 2 months after injection, definite recurrence could be seen in a large proportion of cases. Nevertheless, most recurrences were effectively managed by renewed injection and one case stabilized for the duration of the study after a single injection (Table 1, case 8). This suggests that a large proportion of patients need administration of bevacizumab more frequently than once every second month, although bi-monthly injections may suffice for some patients. Consequently, individually titrated retreatment may become an essential task in the bulk implementation of anti-VEGF therapy for subretinal CNV. An initial course of injections repeated at short intervals, with or without verteporfin PDT, followed by individually titrated retreatment is an attractive alternative, in theory, to the 24-month regimen of fixed 4-weekly retreatment intervals used in the pivotal ranibizumab trials (Brown et al. 2006; Rosenfeld et al. 2006).

Previous case reports, involving a total of 412 patients, did not identify serious adverse events related to intravitreal use of bevacizumab at dosages close to that used in this study (Avery et al. 2006; Bashshur et al. 2006; Rich et al. 2006; Spaide et al. 2006). A recent survey of adverse events in relation to ocular bevacizumab treatment, collected from 61 centres worldwide and including data on 5228 patients, suggest that the treatment is relatively safe (Fung et al. 2006). Obviously, the lack of a formalized prospective, randomized, placebo-controlled study means that adverse events may be under-reported. Preclinical safety data are scarce, but have identified no safety issues (Luke et al. 2006; Manzano et al. 2006; Shahar et al. 2006). Retinal tears have been observed after intravitreal bevacizumab and after intravitreal pegaptanib. This is a rare event that is known to occur spontaneously, and, in view of the large number of patients who have received one or both of the two medications, no conclusion can be drawn from the few case reports available (Dhalla et al. 2006; Meyer et al. 2006).

Although the majority of patients treated in this study demonstrated morphological improvement after the first injection, this was not always accompanied by functional improvement. It has to be considered, however, that several patients had mainly extrafoveal disease activity (their subfoveal CNV component had involuted to some extent) and that several had suffered considerable delay between their first contact with a physician and the initiation of treatment. Measures of morphological change such as haemorrhage, angiographic leakage and presence of PED all supported treatment benefit. Hard exudate was unchanged in most cases and increased in 40% of cases. The kinetics of hard exudate formation by lipoprotein precipitation and subsequent resolution by macrophage phagocytosis is complex. Effective closure of a source of retinal leakage may be followed shortly afterwards by increased amounts of hard exudate, presumably as a result of precipitation following early extraction of water and salt whereby slowly resorbed high-molecular weight components increase in concentration and precipitate in the extracellular space of the retina (Christoffersen et al. 1998; Taarnhoj et al. 2003).

The pharmacodynamics of intravitreal anti-VEGF treatment have not been fully elucidated (Rosenfeld 2006). This is also true for ranibizumab and pegaptanib. The optimal doses of bevacizumab and ranibizumab may be higher than those currently used (Bashshur et al. 2006). Compared with ranibizumab, a 48-kDa antibody binding-site fragment with enhanced VEGF-affinity, bevacizumab is a complete 149-kDa IgG antibody that may differ in yet unknown aspects from ranibizumab, for which safety data have been collected under regulatory supervision.


The study was supported by the Velux Foundation. The first two authors contributed equally to the study.