Article Text


Submacular haemorrhages after intravitreal bevacizumab for large occult choroidal neovascularisation in age-related macular degeneration
  1. S V Goverdhan,
  2. J Lochhead
  1. Ophthalmology Department, St Mary’s Hospital, Newport, Isle of Wight, UK
  1. Mr J Lochhead, Consultant Ophthalmologist, Department of Ophthalmology, St Mary’s Hospital, Newport PO30 5TG, Isle of Wight, UK; jonathan.lochhead{at}


Objective: To report the occurrence of submacular haemorrhages following intravitreal bevacizumab for occult choroidal neovascularisation (CNV) in age-related macular degeneration (AMD).

Methods: Retrospective chart review of 53 patients with occult CNV who had received intravitreal bevacizumab 1.25 mg. Analysis was done in three groups based on mean CNV lesion size: <10 mm2 (n = 17), ⩾10 to <15 mm2 (n = 17) and ⩾15 mm2 (n = 18). ETDRS derived acuity, incidence of fresh macular haemorrhages and haemorrhage size (pre-existing or fresh) were documented and analysed.

Results: The median injection number was 1.0 (range: 1 to 3) with a minimum follow-up of 6 months (range: 4 to 12 months). The mean presenting size of occult lesions was 13.4 mm2 (range: 3.0 to 30.3 mm2). Submacular fresh haemorrhages were seen in the absence of pre-existing haemorrhage in four out of 10 patients in the ⩾15 mm2 CNV size group (40%) but none in the remaining groups with CNV sizes <15 mm2 (OR = 20.1, p = 0.01, 95% CI = 0.99 to 409.3). These haemorrhages developed at a median of 14 days.

Conclusions: Submacular haemorrhages seem to be a significant adverse event following intravitreal bevacizumab in large occult choroidal neovascularisation and may affect visual outcomes. Prospective studies are required to establish the optimal dose of bevacizumab for larger lesion sizes or to identify the most appropriate anti-VEGF agent in large occult CNV with fibrovascular and serous PED lesions.

Statistics from

Recently there was a rapid uptake in the off-label use of intravitreal bevacizumab, a 149 kDa full-length humanised antibody against vascular endothelial growth factor- A (VEGF-A) isoforms for treating choroidal neovascularisation (CNV) in age-related macular degeneration (AMD). This was despite the availability of FDA approved anti-VEGF agents which are currently approved and being used in the treatment of wet AMD. Small retrospective studies have shown bevacizumab to be clinically effective both for vision and for reducing macular thickening.13 However, there are still concerns about its safety, toxicity and adverse effects in the eye. Significant reported complications following intravitreal bevacizumab in AMD include the development of acute retinal pigment epithelial tears4 5 especially in CNV associated with pigment epithelial detachment (PED). In this study, we report the occurrence of acute subretinal haemorrhages in larger occult CNV lesions following intravitreal bevacizumab and correlate visual outcomes in these patients.


All retrospectively selected cases had thorough clinical examination, colour fundus photographs and fluorescein angiography (FFA). Any use of anticoagulants including aspirin was noted, and blood pressure measurements taken. The off-label nature of the treatment and the potential risks and benefits were discussed in detail. A signed informed consent was obtained from all patients who then received 1.25 mg of intravitreal bevacizumab injected through pars plana under sterile conditions. Topical antibiotics were prescribed postoperatively for 5 days. Best-corrected Snellen visual acuity was measured at each visit, and an equivalent Early Treatment Diabetic Retinopathy Study (ETDRS) letter score was derived. Follow-up data were collected from day 1 and 2-week visits initially and 4-weekly thereafter. Repeat injections were offered if CNV activity persisted, fresh haemorrhage or subretinal fluid was observed. A change in acuity also influenced decisions for re-treatment.

Fifty-three eyes of 53 consecutive patients with occult CNV lesions treated with intravitreal bevacizumab over a 6-month period were retrospectively selected for the study. All patients included in the study had unilateral active CNV at presentation. The occurrence of sudden and significant haemorrhage within 4 weeks of intravitreal Bevacizumab was considered a positive result, and these cases were included for analysis. Occult CNV lesion sizes were measured, excluding the area of haemorrhage, if any. In cases where pre-injection haemorrhage was documented with the occult CNV lesion, the area of haemorrhage was measured, and the occurrence of any additional or fresh postinjection haemorrhage was also documented. Measurements were taken using Zeiss digital angiography software (VISUPAC version 4.1).

For statistical analysis, the total study sample was analysed in three subgroups based on the mean size of CNV lesions: <10 mm2 (n = 17), ⩾10 to <15 mm2 (n = 17) and ⩾15 mm2 (n = 18). The prevalence of submacular haemorrhages in these three groups was then calculated. The significance of submacular haemorrhage occurrence in the groups, the size of haemorrhages and change in ETDRS acuity in the three groups was tested by two-tailed Mann–Whitney tests. The level of statistical significance in the study was set at p<0.05. All statistical analyses were performed using the SPSS statistical software package version 12.0


Data were available from 53 patients (53 eyes) with occult CNV (table 1). The mean patient age was 79.5 years with a mean follow-up of 5.9 months (range = 2 to 12 months). All patients had received at least one injection of Bevacizumab (mean = 1.36, range = 1 to 3). The average CNV lesion size in the overall group was 13.4 mm2 (range = 3.0 to 30.3). The mean visual acuity at presentation was 35.1 ETDRS letters in the overall group; this had improved to a mean of 39.1 letters at final follow-up (table 1). In those patients with pre-existing haemorrhage, the mean haemorrhage size increased postinjection in all the three groups, but this was not found to be statistically significant. No difference was found in the demographics, number of injections and follow-up period between the three subgroups (data not shown).

Table 1 Demographic and clinical characteristics of the study population

In group 3 (lesion sizes ⩾15 mm2, n = 18), four of the 10 patients (40%) without any pre-existing haemorrhage had developed fresh submacular haemorrhages all within 3 weeks following intravitreal bevacizumab (see tables 1, 2 and fig 1). Three of these patients displayed additional clinical and angiographic characteristics of fibrovascular PED, and one patient had a serous occult PED. Fresh submacular haemorrhages were not seen in any patients in the other two groups. This occurrence of fresh haemorrhage in group 3 (CNV sizes ⩾15 mm2) was statistically significant with an odds ratio of 20.1 (p = 0.01, 95% CI = 0.99 to 409.3) when compared with the remaining groups with CNV sizes <15 mm2. The timing of onset of these haemorrhages following the injection was acute and ranged from 10 to 21 days (median = 14 days). None of these patients were on medication with aspirin or warfarin. Eight patients with pre-existing haemorrhages in group 3 (n = 18) developed an increase in the size of these haemorrhages, but these were excluded from the above analysis because of the perceived difficulty in differentiating postinjection haemorrhage from disease natural history in this group.

Figure 1 (A, B, C) Colour fundus photograph and fluorescein angiograph of patient 3 with a large occult fibrovascular PED which developed fresh subretinal haemorrhage in the macula extending infero-temporally from the edge of the lesion at day 7.
Table 2 Clinical characteristics of patients who developed fresh submacular haemorrhages in group 3 (⩾15 mm2 CNV size)

In the overall treated group, visual acuity improved or remained stable in 47/53 patients (89%), eight 8/53 patients (15%) had an improvement of three or more lines with one patient improving over six lines (mean acuity improved by 3.2 to 5.3 ETDRS letters in the four groups). Of the four patients who developed fresh submacular haemorrhages in group 3, one patient lost six lines or 30 ETDRS letters, acuity improved by 15 letters in one patient and two patients maintained their initial acuities of 35 and five ETDRS letters (table 2). No significant correlation was found between haemorrhage size and ETDRS acuity. Also, no systemic adverse events were recorded in the patients treated with intravitreal bevacizumab. Calculations revealed that the sample size had 80% power to detect a 40% difference in proportions between two groups at an alpha significance level of 0.05 with 95% confidence.


This series is the first documented report of acute submacular haemorrhages after bevacizumab therapy for large occult CNV lesions in AMD. Previous bevacizumab-related ocular adverse events have all been mainly related to the intravitreal injection procedure (including endophthalmitis, traumatic lens injury and retinal detachment).6 7 The actual incidence of submacular haemorrhages cannot be determined from this small case series due to a large confidence interval. However, in larger lesions measuring ⩾15 mm2, the prevalence of fresh submacular haemorrhages in this study was 40%. In a recently published internet survey of 4576 patients after intravitreal bevacizumab, possible adverse events reported included uveitis (0.14%), central retinal artery occlusion (0.01%), tears of the retinal pigment epithelium (0.06%) and new or progressive subretinal haemorrhages (0.06%).8 It is worth noting, however, that this survey included data from all CNV types and not just occult lesions as in this study. The risk of subretinal haemorrhages in this study was 20-fold higher in the group with larger occult CNV lesions (⩾15 mm2) than in those with smaller lesions.

Subretinal haemorrhages have also been reported after procedures like photodynamic therapy (PDT) and transpupillary thermotherapy (TTT) for neovascular AMD associated with PED. However, the mechanism here may be different from that due to anti-VEGF effect on CNV. Haemorrhages have been reported in five out of 34 (14%) PDT-treated eyes in one study9 and in four of 821 (0.4%) patients treated with PDT for exudative AMD. In the latter series, these haemorrhages occurred within 1 week of treatment.10 The timing of haemorrhages in our series at 2 weeks seems to be similar to that seen after PDT for occult CNV. In a study of TTT for occult CNV, three out of 48 (6.2%) treated eyes were found to develop large submacular haemorrhages all within 2 months of treatment.11

During the natural history of occult CNV in AMD, the development of haemorrhagic CNV has been reported in 17/100 cases with PED (17%) over a period greater than 12 months in one study.12 However, this report had not evaluated PED lesion size in relation to development of haemorrhage. The 40% rate of fresh acute submacular haemorrhages seen in group 3 of this study over a span of 3 weeks is significant and cannot be explained by natural history. Patients with pre-existing haemorrhage who had a significant increase following the injection were excluded from the final subgroup because of the obvious difficulties in differentiating the features of postinjection haemorrhages in this subgroup from that due to disease natural history. The presence of pre-existing haemorrhage itself did not, however, seem to be a risk factor for increase in postinjection haemorrhage.

The actual mechanism of subretinal haemorrhages in all these cases is unknown. However, previous suggestions for the development of acute RPE tears following bevacizumab injections for occult PED lesions could in theory also be important factors for subretinal haemorrhage.4 5 These include possible spontaneous ruptures or increased contraction of the CNV due to the antiangiogenic activity of bevacizumab. The timing of this effect on CNV may be further corroborated from recent studies of bevacizumab for managing persistent new vessels in diabetic retinopathy. Although this has a very different pathophysiology, the new vessels and their area of active leakage have been shown to decrease significantly by the first month.13 Another possible mechanism might involve short-term changes in systolic blood pressure following bevacizumab. Increased blood pressure has been linked to increased incidence of submacular haemorrhages.14 There have been concerns raised relating to possible increased blood pressure following bevacizumab intravitreal injections. This is a recognised complication of systemic bevacizumab,15 but so far this has not been a significant observation associated with intraocular use of the drug.8

In summary, a significant number of eyes with large occult CNV ⩾15 mm2 developed submacular haemorrhages following intravitreal bevacizumab injections. Occult lesion size ⩾15 mm2 is a risk factor for acute submacular haemorrhage following intravitreal bevacizumab. Occult CNV lesions ⩾15 mm2 possibly require less aggressive anti-VEGF suppression with either lower doses of the drug or alternative anti-VEGF therapy. In the future, more comprehensive use of optical coherence tomography in the initial evaluation of CNV will enable additional calculation of volume measurements of these lesions. This may in fact be more critical in quantifying the risk of haemorrhage in these eyes. Long-term studies are needed to establish the risks, benefits and optimal safe dose of anti-VEGF agents in treating large CNV in AMD.


View Abstract


  • Competing interests: None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • At a glance
    Harminder S Dua Arun D Singh, Editors-in-Chief