Article Text

Download PDFPDF

Clinical science
Predictive factors of visual and anatomical outcome after intravitreal bevacizumab treatment of neovascular age-related macular degeneration: an optical coherence tomography study
  1. R P Singh1,
  2. E X Fu1,
  3. S D Smith1,
  4. D R Williams1,2,
  5. P K Kaiser1,2
  1. 1
    Cole Eye Institute, Cleveland, Ohio, USA
  2. 2
    Digital OCT Reading Center, Cleveland, Ohio, USA
  1. Correspondence to Dr P K Kaiser, Digital OCT Reading Center, Cole Eye Institute, Cleveland Clinic, 9500 Euclid Ave, Desk i3, Cleveland, OH 44195, USA; pkkaiser{at}


Aim: To evaluate the baseline visual and optical coherence tomography (OCT) factors on outcomes after intravitreal bevacizumab treatment of subfoveal neovascular age-related macular degeneration (AMD).

Methods: A retrospective analysis of 73 eyes treated with intravitreal bevacizumab for subfoveal neovascular AMD was performed. Change in best corrected Snellen visual acuity (BCVA) and central retinal thickness (CRT) on OCT were the primary outcomes. Automated and manual measurements were made for all OCT characteristics.

Results: Seventy-three (100%) and 58 (79.5%) eyes were followed for 3 and 6 months, respectively. The mean BCVA improved from 20/177 to 20/160 (p = 0.03) at 3 months and to 20/143 (p = 0.04) at 6 months. The mean CRT decreased 93 µm (p<0.0001) and 105 µm (p<0.0001) at 3 and 6 months, respectively. Baseline BCVA worse than 20/100 was associated with greater visual improvement (p⩽0.04). Eyes with baseline CRT greater than 400 µm experienced a greater mean CRT reduction (p<0.05). Treatment-naïve patients had a greater mean CRT reduction than those previously treated with any modality (p<0.05)

Conclusions: Baseline BCVA and CRT positively influence mean visual and CRT improvement, respectively, after intravitreal bevacizumab in wet AMD. Any prior treatment predicted less CRT reduction.

Statistics from

Chorodial neovascularisation (CNV) is the primary cause of vision loss in exudative age-related macular degeneration (AMD). Recently, short-term retrospective case series noted visual improvement in patients treated with off-label intravitreal bevacizumab (Avastin, Genentech, South San Francisco, California).12 Bevacizumab is a full-length antibody that targets all isoforms of VEGF, and in the largest series to date, Spaide et al reported that approximately 30% of patients experienced visual improvement.2 These reports have led to increased use of intravitreal bevacizumab to treat subfoveal neovascular AMD. However, it remains unclear which patients benefit the most from this therapy. Thus, the purpose of this study was to evaluate the predictive factors for visual and anatomical outcomes after intravitreal bevacizumab treatment of AMD.

Patients and methods

After Institutional Review Board approval, a retrospective review of all patients who received intravitreal bevacizumab for subfoveal neovascular AMD was performed. Inclusion criteria included patients with subfoveal CNV due to exudative AMD who received at least one intravitreal injection of 1.25 mg bevacizumab and had a baseline fluorescein angiogram (FA), optical coherence tomography (OCT) and follow-up of at least 3 months. All lesion types and sizes were included. Only one eye of each patient was assigned as the study eye, and no patients were undergoing bilateral treatments. Exclusion criteria included patients treated with ranibizumab (Lucentis, Genentech, South San Francisco, California) after one bevacizumab injection or who had visual outcomes confounded by clinically significant ocular disease including clinically significant diabetic macular oedema, sterile endophthalmitis or patients who underwent cataract extraction.

Baseline demographic characteristics including age, gender, ethnicity, refractive error and presence of hypertension or diabetes were recorded. Prior type and number of AMD treatments were documented. In all patients, the baseline FA was independently assessed by two of the investigators (EXF, RPS) to confirm subfoveal CNV leakage, evaluate lesion composition and document subfoveal haemorrhage. Lesions obscured by severe subretinal haemorrhage were categorised as undetermined. Patients underwent a comprehensive baseline ophthalmic exam and OCT assessment at 6-week intervals. The OCT scans were obtained using an established protocol on the Stratus OCT scanner (Carl Zeiss Meditec, Dublin, California, software V.4.0), consisting of a fast macular thickness map, 3 and 6 mm horizontal and vertical linear cross hair scans and a customised posterior pole scan depending on the eye.

Each OCT scan was evaluated by at least two masked graders. OCT characteristics were measured using the caliper-assisted manual measurement approach previously described by Costa et al (table 1).3 Central retinal thickness was defined as the distance between the inner limiting membrane (ILM) of the retina and the inner border of the RPE/choriocapillaris band, inclusive of subretinal fluid, measured in the central 1 mm area of the 7 mm posterior pole scan. It excludes the CNV lesion and any sub-RPE fluid. Central retinal lesion thickness is the distance between the ILM and the inner border of the choriocapillaris, inclusive of subretinal fluid collections and any observable CNV or scar tissue. For measurement of PED thickness, an estimate was made from the OCT scans by evaluating the distance from the RPE/Bruch membrane complex to the optically empty bowl shaped depression produced when the software eliminates the true PED during post-processing. In addition, automated measurement of the centre subfield was recorded from analysis of the fast macular thickness map, and its correlation with manual CRT measurement was assessed.

Table 1

Definitions of optical coherence tomography (OCT) anatomical characteristics

All patients received intravitreal injections of bevacizumab (1.25 mg/0.05 ml) at baseline via the pars plana using a 30-gauge needle under aseptic conditions. Additional injections were given at 6-week intervals if any of the following were observed by the evaluating physician clinically or on OCT: (1) vision loss with OCT evidence of fluid in the macula; (2) persistence of or increase in central retinal thickness (CRT), subretinal fluid (SRF), pigment epithelial detachment (PED), and/or cystoid macular oedema; or (3) new macular haemorrhage. All criteria were based on comparisons with the previously scheduled visit.

Best-corrected visual acuity recorded in Snellen units was converted to logMAR for statistical analysis. Stable vision was defined as ±half of the visual angle, while improvement and decline were defined as changes of more than half of the visual angle. Subgroups were categorised based on demographic characteristics, prior treatments, angiographic lesion types, and baseline vision and OCT anatomical patterns. Univariate comparision of BCVA, CRT and secondary OCT anatomical outcomes between subgroups was performed using the t test. Comparison of BCVA, CRT and secondary OCT anatomical outcomes before and after intravitreal bevacizumab treatment was performed using a paired t test. Multivariate linear regression models were used to identify independent predictors of change in BCVA and OCT anatomical parameters after treatment. A backward elimination method was used to identify parsimonious models by excluding predictors not significantly associated with the outcome (p>0.15).


A total of 89 patients were screened, and 73 eyes of 73 patients were included in this study. Among those screened, 16 patients were excluded for the following reasons: less than 3 months of follow-up (three), switched to ranibizumab after one bevacizumab injection (seven), cataract extraction (four), clinically significant macular oedema (one) and sterile endophthalmitis (one). Seventy-three (100%) and 58 (79.5%) eyes were followed for 3 and 6 months, respectively. Of the 15 patients not included in the 6 month analysis, 12 had begun ranibizumab treatment, two were lost to follow-up, and one was enrolled in a clinical trial. The mean number of intravitreal bevacizumab injections per eye was 3.1.

Baseline demographics are shown in table 2. Sixty-six (90%) of the 73 eyes had undergone prior treatment for neovascular AMD. The mean number of prior treatments was 2.4 with PDT monotherapy, 2.0 with PDT combined with intravitreal triamcinolone acetonide, 3.5 with intravitreal pegaptanib or 1.2 with laser photocoagulation for extrafoveal CNV. The mean baseline BCVA and CRT were 20/177 (log MAR 0.95) and 433 (162) µm, respectively. Thirty-seven eyes (50.7%) had subretinal fluid (SRF), 58 (61.1%) had measurable CNV, and 23 (37%) had a pigment epithelial detachment (PED). Sixty-two eyes (85%) demonstrated retinal cysts with a mean CMO grade of 2.

Table 2

Baseline demographics and study eye characteristics

At 3 months, the mean visual acuity improved from 20/177 (log MAR 0.95) to 20/160 (log MAR 0.90) (p = 0.03) and was accompanied by a 93 µm decrease in mean CRT (p<0.0001; table 3). There was further improvement at 6 months with mean vision of 20/143 (log MAR 0.85) (p = 0.04) and a mean reduction in CRT of 105 µm (p<0.0001). Visual acuity was stable, defined as ±half of the visual angle, in 47 (64.4%) and 39 (67.2%) eyes at 3 and 6 months, respectively. An improvement of ⩾3 lines was observed in 18 (24.7%) eyes at 3 months and 12 (20.7%) eyes at 6 months. Eight (11%) and 7 (12.1%) eyes experienced worsening of more than half of the visual angle at 3 and 6 months, respectively (fig 1A). No significant association was seen between the mean change in BCVA and CRT for matching or staggered time points. The mean thickness decreased for all OCT characteristics at 3 and 6 months (p<0.0001) (fig 1B). Most of the improvement occurred within the first 6 weeks. No associations between visual acuity and changes in these OCT characteristics were found.

Figure 1

(A) Visual acuity stabilised (±half of the visual angle) in 47 (64.4%), improved > half of the visual angle in 18 (24.7%), and worsened more than half of the visual angle in eight (11%) eyes at 3 months (n = 73), p = 0.04. At 6 months (n = 58), 39 (67.2%) stabilised, 12 (20.7%) improved, and 7 (12.1%) worsened, p = 0.006. Error bars represent 95% confidence intervals. (B) Mean change in optical coherence tomography (OCT) characteristics at 3 and 6 months compared with baseline (p<0.0001). CNVT, choroidal neovascularisation thickness; CRLT, central retina lesion thickness; GLT, greatest lesion thickness; PED: pigment epithelial detachment thickness; SRF, subretinal fluid thickness. (A) Visual acuity at 3 and 6 months. (B) Mean change in OCT characteristics.

Table 3

Mean change in visual acuity and central retinal thickness

When evaluating baseline characteristics, patients with baseline BCVA⩽20/100 experienced greater visual improvement than those with baseline BCVA>20/100 at 3 (p = 0.04) and 6 (p = 0.006) months. At 6 months, 39.4% of patients with baseline BCVA⩽20/100 improved more than half of the visual angle versus 2.0% of those with BCVA>20/100 (p = 0.001) (fig 2). Moreover, fewer patients (9.1%) with BCVA⩽20/100 lost vision more than half of visual angle when compared with those with BCVA>20/100 (36.0%) (p = 0.001). A similar association between baseline vision and visual outcome was found at 3 months.

Figure 2

Follow-up visual acuity distribution based on baseline visual acuity. At 6 months, visual acuity stabilised (±half of the visual angle) in 17 (51.5%), improved more than half of the visual angle in 13 (39.4%), and worsened more than half of the visual angle in three (9.1%) eyes in patients with baseline vision worse than 20/100 (best-corrected visual acuity (BCVA) ⩽20/100), n = 33, p = 0.001. In patients with baseline vision better than 20/100 (BCVA>20/100), 14 (56.0%) stabilised, two (8.0%) improved, and nine (36.0%) worsened, n = 25, p = 0.001. Error bars represent 95% confidence intervals.

Differences in mean BCVA change based on CNV lesion types were significant at 6 weeks (p = 0.01) and 3 months (p = 0.03) (fig 3A). During the first 3 months, the mean BCVA improved in patients with predominately classic, minimally classic and undetermined lesions, while it worsened in occult with no classic CNV lesions. However, there were no differences in mean BCVA change between CNV lesion types at 6 months (p = 0.28).

Figure 3

(A) Mean visual acuity change based on baseline choroidal neovascularisation (CNV) lesion type (p = 0.01 at 1.5 months, p = 0.03 at 3 months, p = 0.19 at 4.5 months, p = 0.28 at 6 months). (B) Mean central retinal thickness (CRT) change based on baseline CRT (p = 0.02 at 1.5 months, p = 0.003 at 3 months, p = 0.02 at 4.5 months, p = 0.04 at 6 months). (C) Mean CRT change based on previous treatment history (p = 0.01 at 1.5 months, p = 0.01 at 3 months, p = 0.04 at 4.5 months, p = 0.005 at 6 months).

Baseline CRT (p<0.05) and any prior treatment (p⩽0.04) were significant predictors of mean CRT change in all regression models. Patients with baseline CRT greater than 400 µm experienced a greater mean CRT reduction than those with baseline CRT less than 400 µm at both 3 and 6 months (p<0.05) (fig 3B). There was no difference in mean CRT change between patients with baseline CRT⩽300 µm and CRT 301–400 µm. Treatment-naïve patients had a greater mean CRT reduction than those previously treated with any modality (p<0.05) (fig 3C). However, there was no significant difference when the patients were stratified by previous laser photocoagulation, PDT monotherapy, PDT and intravitreal triamcinolone acetonide, intravitreal pegaptanib or multiple modalities.

Patients’ age, gender, race, smoking history, refractive error and systemic medical conditions such as hypertension and diabetes had no impact on BCVA or CRT changes. Further, baseline SRF, CNV, PED, GLT (greatest linear thickness) and CMO measurements showed no predictive value. Analysis of mean change in any of the above OCT characteristics and BCVA or CRT change also revealed no significant association. While treatment-naïve patients showed a greater CRT reduction, prior treatments do not affect BCVA change. In contrast to BCVA, angiographic lesion type is not associated with CRT change.


Recently there has been considerable interest in the use of intravitreal bevacizumab to manage subfoveal neovascular AMD due to its low cost and apparent efficacy. A number of short-term retrospective case series have shown a significant visual improvement and mean CRT reduction in patients treated with intravitreal bevacizumab.12456789 Similar to these reports, we found that visual acuity stabilised in 65% and improved in 25% of patients, while it declined in only 10% of eyes at 3 months despite 90% of the patients in our study having received previous AMD treatment. This visual improvement was accompanied by reductions in mean CRT, SRF thickness, CNV thickness, PED thickness, GLT and CMO scores. These effects were sustained at 6 months. These results suggest that a significant improvement in vision and OCT characteristics can be expected even in previously treated patients.

Baseline BCVA and CNV lesion types were identified as predictors of mean BCVA change. Patients with baseline BCVA worse than 20/100 experienced a greater visual improvement than those with better baseline vision. Similarly, baseline vision was the most influential predictor in the MARINA subgroup analysis in which better baseline vision reflected a smaller mean improvement in patients receiving monthly ranibizumab injections.1011 One explanation for this finding is that patients with worse baseline BCVA have more potential for improvement. An important point from this finding is that intravitreal bevacizumab should be considered for patients with poor baseline vision because these patients may be more likely to benefit from treatment.

The mean change in BCVA was less in eyes having occult with no classic CNV lesions in comparison with predominately classic, minimally classic and undetermined lesions at 6 weeks and 3 months. This difference was not significant at 4.5 and 6 months (fig 3A). Our findings suggest a worse BCVA outcome in occult only lesions and a better outcome in predominately classic CNV in the short-term. A longer follow-up is needed to determine if these relationships persist. Conclusions regarding mean BCVA change in minimally classic and undetermined lesions are difficult to draw due to the limited number of patients in each subgroup. The MARINA subgroup analysis is the only other study that evaluated the effect of CNV lesion type on mean BCVA change with anti-VEGF agents. It found no difference in visual improvement between the minimally classic and occult with no classic CNV lesions at 24 months, although patients with predominantly classic lesions in the ANCHOR study did have better visual results.11

A worse baseline CRT predicted a greater mean CRT reduction. It may simply be that eyes with a thicker baseline CRT had more potential to improve than thinner lesions. Nonetheless, this finding suggests that patients with severely thickened baseline CRT are likely to experience a greater mean CRT reduction and, thus, may benefit from intravitreal bevacizumab. Although there was no difference based on previous treatment modalities, any prior treatment predicted less CRT reduction with bevacizumab treatment. Another recently published study also found that prior treatments of PDT or pegaptanib resulted in less improvement in central macular thickness.12 Compared with treatment-naïve patients, the duration of neovascular AMD is likely longer in previously treated patients. It is possible hypothetically that the more mature lesions had lost their sensitivity to anti-VEGF therapy due to the formation of pericytes on the endothelium.

Central retinal thickness and all other OCT anatomical characteristics were manually evaluated in our series because several studies have reported that automatically measured retinal thickness using the built-in software of the Stratus OCT is often erroneous, particularly in diseases with subretinal processes like AMD.131415 Ray et al identified several OCT image artefacts in AMD that led to errors in more than 60% of automated retinal thickness measurements.15 Most of these image artefacts can be avoided by careful scanning procedures by experienced photographers. In addition, operator-independent errors caused by software misidentification of the inner or outer retinal boundaries must be reviewed with caution. To avoid errors from misidentification of retinal boundaries by the automated analysis, we urge future studies of subretinal disease processes to adopt a manual caliper-assisted measurement approach to double-check automated measurements as we did in our series.

We recognise that the retrospective design, limited number of patients, non-ETDRS vision measurements and short follow-up are weaknesses of this study. Because of the approval of Ranibizumab during the course of the study, some patients were converted to this drug because of its FDA approval status with documented safety and efficacy data. This impacted patient follow-up. The strengths are manual-caliper-assisted OCT measurements and consistent pro re nata (PRN) treatment decisions by the same retina specialist. To date, this is the most comprehensive study to evaluate the relationships between visual and OCT anatomical outcomes and demographic characteristics, prior treatments, angiographic CNV types, baseline vision and baseline OCT characteristics in patients treated with intravitreal bevacizumab for subfoveal neovascular AMD.

Our study demonstrated a significant improvement in mean vision and OCT anatomical characteristics after intravitreal bevacizumab for exudative AMD. More importantly, we found that a lower baseline vision and thicker baseline CRT positively impacted mean visual and CRT improvement, respectively. In contrast, occult with no classic CNV lesions were associated with less improvement in vision at 3 months, and any prior AMD treatment predicted less CRT reduction. Larger, prospective studies with a longer follow-up are needed to substantiate these findings.



  • Competing interests None.

  • Provenance and Peer review Not commissioned; externally peer reviewed.

  • Ethics approval Ethics approval was provided by Cleveland Clinic Institutional Review Board.

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.