To determine longitudinal rates of ocular complications after anti–vascular endothelial growth factor (VEGF) treatment for neovascular age-related macular degeneration (AMD) in a nationally representative longitudinal sample.
Retrospective, longitudinal case-control study.
Using the Medicare 5% claims database, diagnoses of neovascular AMD and anti-VEGF injections of ranibizumab, bevacizumab, or pegaptanib were identified from International Classification of Diseases and Current Procedural Terminology procedure codes. Six thousand one hundred fifty-four individuals undergoing anti-VEGF treatment for neovascular AMD (total of 40 903 injections) were compared with 6154 matched controls with neovascular AMD who did not undergo anti-VEGF treatment. Propensity score matching was used to match individuals receiving anti-VEGF injections with controls. Rates of postinjection adverse outcomes (endophthalmitis, rhegmatogenous retinal detachment, retinal tear, uveitis, and vitreous hemorrhage) were analyzed by cumulative incidence and Cox proportional hazards model to control for demographic factors and ocular comorbidities.
At the 2-year follow-up, the rates of endophthalmitis per injection (0.09%; P < .01), uveitis (0.11%; P < .01), and vitreous hemorrhage per injection (0.23%; P < .01) were significantly higher in the anti-VEGF treatment group. With Cox proportional hazards modeling, the anti-VEGF treatment group had a 102% higher risk of severe ocular complications overall and a 4% increased risk per injection, both of which were statistically significant ( P < .01).
Rates of endophthalmitis, uveitis, and vitreous hemorrhage were higher in the group treated with anti-VEGF injection than in the control group, although these nevertheless were rare in both groups. The overall risk of severe ocular complications was significantly higher in the anti-VEGF treatment group.
Age-Related Macular Degeneration (AMD) is the leading cause of blindness in developed countries and the third leading cause of blindness worldwide. Since 2005, the advent and widespread use of anti–vascular endothelial growth factor (VEGF) drugs for the treatment of neovascular AMD has changed dramatically the management of this disease. Although the visual prognosis for these patients has improved, each intravitreal injection poses a risk of infection, postinjection inflammation, retinal tear or detachment, and vitreous hemorrhage.
The 3 medications delivered via intravitreal injection for treatment of neovascular AMD are ranibizumab (Lucentis; Genentech, South San Francisco, California, USA), bevacizumab (Avastin; Genentech), and pegaptanib (Macugen; OSI-Eyetech, New York, New York, USA). Ranibizumab and pegaptanib have been approved by the Food and Drug Administration for treatment of neovascular AMD, and bevacizumab has been used off label for this indication with increasing frequency. A review of safety data from 4 randomized trials of ranibizumab (Minimally classic/occult trial of the Anti-VEGF antibody Ranibizumab In the treatment of Neovascular AMD [MARINA], ANti-VEGF Antibody for the Treatment of Predominantly Classic CHORoidal Neovascularization in AMD [ANCHOR], Safety Assessment of Intravitreal Lucentis fOR AMD [SAILOR], A Phase IIIb, Multi-center, Randomized, Double-Masked, Sham Injection-Controlled Study of the Efficacy and Safety of Ranibizumab in Subjects with Subfoveal Choroidal Neovascularization with or without Classic CNV Secondary to Age-Related Macular Degeneration [PIER]), in which 3252 patients received more than 28 500 injections, found a 0.05% rate of endophthalmitis per injection (Boyer DS, et al. A safety overview of ranibizumab in patients with wet AMD: ANCHOR, MARINA, PIER, and SAILOR studies. Paper presented at the AAO/SOE Joint Annual Meeting. November 8–11, 2008; Atlanta, Georgia). Retrospective reviews looking at bevacizumab, pegaptanib, and ranibizumab have found rates of endophthalmitis per injection of 0%, 0.02%, 0.077%, and 0.16%. Rates of serious intraocular inflammation have ranged widely from 0.03% per injection in the 4 randomized trials of ranibizumab, to 0.09% in a 12-month study of bevacizumab, to 1.5% in a retrospective review of ranibizumab and bevacizumab injections. Rates of mild or moderate intraocular inflammation have been much higher, occurring in up to 17.1% of patients in the ANCHOR study. Retinal tear and retinal detachment have been rare in published studies to date; the rate of retinal detachment was 0% in the MARINA trial, 0.03% in the ANCHOR study, and 0.16% in a retrospective review of bevacizumab injections.
To date, the ocular complications of intravitreal injections have not been studied in the Medicare population. This study examines the ocular complications of anti-VEGF injections given for neovascular AMD in a nationally representative longitudinal cohort of elderly persons. This represents the largest reported sample studying ocular complications of patients undergoing anti-VEGF treatment and offers the advantage of reducing surgeon- and center-specific factors.
For this retrospective, longitudinal cohort analysis, Medicare 5% inpatient, outpatient, Part B, and durable medical equipment claims files were used to identify a nationally representative sample of Medicare beneficiaries aged 68 years and older diagnosed with neovascular AMD. The data contained information on beneficiaries’ demographic characteristics, diagnoses (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM]), procedure codes (Current Procedural Terminology; Healthcare Common Procedure Coding System [HCPCS]), and United States Centers for Medicare and Medicaid Services provider physician specialties, submitted with claims that were used to identify whether an individual received anti-VEGF treatment and related adverse outcomes and to ensure that individuals had seen an eye specialist. Data were linked by a unique identifier, permitting construction of longitudinal, person-specific data from January 1, 2002 through December 31, 2008.
Individuals undergoing anti-VEGF treatment between 2005 and 2008 were identified using these codes: pegaptanib (HCPCS codes J3490, C9128, and J2503), bevacizumab (HCPCS codes J3490, J3590, and J9035), and ranibizumab (HCPCS codes J3490, J3590, C9399, and C9233). For the unclassified codes (HCPCS codes J3490, J3590, and C9399), price ranges and eye injection codes on the day of treatment were used jointly to ascertain those who were receiving anti-VEGF treatments and those who were not. The specific price ranges were: pegaptanib, $750 to $1000; bevacizumab, $1 to $200; and ranibizumab, $1500 to $2500.
We used a 3-year review period to identify comorbid eye conditions and to ensure that the individuals receiving anti-VEGF had at least 2 previous diagnoses of neovascular AMD (ICD-9-CM codes 362.52, 362.42, and 362.43) before anti-VEGF treatment. To ensure that we had a full 3-year review, individuals younger than 68 years were excluded from our sample, as were individuals who were enrolled in a health maintenance organization (HMO) or who lived outside the United States for more than 12 months during the 3 -year review period. Baseline date for individuals in the anti-VEGF group was the first date they received an anti-VEGF injection.
The control group was composed of individuals who had received at least 2 diagnoses of neovascular AMD within a 3-year period from January 1, 2006, and had not received any anti-VEGF treatments up to the end of our study period, December 31, 2008. The same age, HMO, and living outside the United States restrictions were applied to the control group. The baseline date was January 1, 2006, for the control sample. We further required individuals in the control group to have at least 1 visit to an ophthalmologist (Centers for Medicare and Medicaid Services code 18) or optometrist (Centers for Medicare and Medicaid Services code 41) during the follow-up period.
We followed up individuals for 730 days or until an eye complication developed or until they were censored. Censoring occurred when a beneficiary underwent cataract or glaucoma surgery, joined a HMO, moved outside of the United States, or died during the follow-up period.
Adverse events were endophthalmitis (ICD-9-CM code 360.0), rhegmatogenous retinal detachment (ICD-9-CM code 361.0), retinal tear (ICD-9-CM codes 361.30 and 361.31), uveitis (ICD-9-CM codes 364.00 to 364.05, 364.10, and 364.11), and vitreous hemorrhage (ICD-9-CM code 379.23). These diagnosis codes were grouped together to form a severe ocular complications category. We required that control and treatment persons never received a prior diagnosis of endophthalmitis, rhegmatogenous retinal detachment, retinal tear, uveitis, or vitreous hemorrhage before their date of first anti-VEGF treatment if they were in the treatment or before January 1, 2006, if they were in the control group. We also did not include any adverse events if they occurred after cataract or glaucoma surgery during the follow-up period.
Propensity Score Matching
In the first step of the matching process, we performed logit analysis to predict the probability of an individual undergoing anti-VEGF treatments. Covariates for the logit analysis were binary variables for gender, dry AMD, black race, other race, and continuous variables for age and Charlson index, a widely used measure of comorbidity.
In the second step of the matching process, we paired an individual undergoing anti-VEGF treatment to his or her nearest match in the control group using the predicted probability of an anti-VEGF treatment from the logit analysis. Propensity score matching reduces selection bias in the receipt of anti-VEGF treatments among individuals with neovascular AMD. The program, SAS Greedy 5 to 1 digit match macro (by Lori S. Parsons; available at: www2.sas.com/proceedings/sugi26/p214-26.pdf ; accessed April 20, 2009), made the best match first by pairing individuals in the treatment and control group on exact 5-digit matches of their predicted probability of receipt of anti-VEGF treatments. Considering all persons not previously matched, the macro then attempted to match individuals based on 4 digits of their propensity score, then 3, then 2, and then 1. Individuals unable to be matched on 1 digit were excluded. Standardized differences were calculated for the matched sample and revealed no differences of more than 10%, resulting in a well-matched sample.
Time-to-event analysis was performed on the resulting matched sample using a Cox proportional hazards model. Unadjusted and adjusted time to a severe ocular complication was studied. We performed 4 specifications: (1) having a complication using a binary indicator for anti-VEGF treatment unadjusted for other covariates; (2) having a complication using a binary indicator for anti-VEGF treatment adjusted for other covariates (the adjusted model consisted of controls for age, background diabetic retinopathy, proliferative diabetic retinopathy, Alzheimer disease or other dementia, cataract surgery, cataract, glaucoma, glaucoma surgery, male gender, white race, Charlson index, dry AMD, blindness or low vision, and duration of neovascular AMD); (3) having a complication based on the frequency of anti-VEGF treatment unadjusted for other covariates; and (4) having a complication based on the frequency of anti-VEGF treatment adjusted for other covariates. The Duke University Institutional Review Board approved this study.
During the study, there were 2 163 207 beneficiaries in the Medicare 5% inpatient, outpatient, Part B, and durable medical equipment claims files. Of these beneficiaries, 20 671 individuals (1.0% of the total) with a diagnosis of neovascular AMD between 2005 and 2008 were identified who met the inclusion criteria outlined in the Methods. Of the diagnoses, 96.4% came from ophthalmologist examinations and 3% of the data came from optometrist examinations. Of the 20 671 individuals with neovascular AMD, 6154 (29.8%) underwent at least 1 anti-VEGF injection during the study, receiving a total of 40 903 injections. Individuals were followed up for 2 years, until an eye complication developed, or until they were censored; mean follow-up time was 533 days for the control group and 435 days for the anti-VEGF treatment group. Beneficiaries were censored if they underwent cataract or glaucoma surgery during the follow-up period, joined an HMO, moved outside the United States, or died; 35 beneficiaries were censored in the control group and 72 patients were censored in the anti-VEGF treatment group.
To reduce selection bias, propensity score matching then was performed with the covariates of gender, race, age, history of dry AMD, and the Charlson comorbidity index to create a control group of 6154 individuals who had neovascular AMD but did not undergo an anti-VEGF injection during this period. In terms of the matched characteristics of age, gender, race, and mean value of the Charlson Index, the 2 groups were very similar ( Table 1 ). For ocular comorbidities, the control group was significantly more likely than the anti-VEGF treatment group to have cataract (47% vs 45%; P = .03) and significantly less likely to have undergone previous cataract surgery (44% vs 47%; P < .01) or glaucoma surgery (4% vs 6%; P < .01).
|Baseline Characteristics||Control Group (n = 6154)||Anti-VEGF Treatment Group (n = 6154)||P Value|
|Age (yrs)||82.23||82.15||.02 a|
|Charlson Index (0 to 18)||1.92||1.91||0.61|
|Background diabetic retinopathy||357 (6%)||318 (5%)||0.11|
|Proliferative diabetic retinopathy||75 (1.2%)||90 (1.5%)||0.24|
|Cataract||2868 (47%)||2746 (45%)||0.03 a|
|Cataract surgery||2701 (44%)||2866 (47%)||< .01 b|
|Glaucoma||1833 (30%)||1762 (29%)||0.16|
|Glaucoma surgery||251 (4%)||380 (6%)||< .01 b|
Between 2005 and 2008, the number of all types of anti-VEGF injections in the sample increased from 3393 in 2005 to 15 103 in 2008, and the number of bevacizumab and ranibizumab injections increased dramatically ( Table 2 ). By 2008, bevacizumab injections accounted for 54.5% of anti-VEGF injections overall. The number of pegaptanib injections declined precipitously, from 96.4% of anti-VEGF injections in 2005 to only 1.1% of injections by 2008.
|Year||Bevacizumab (%)||Ranibizumab (%)||Pegaptanib (%)||Total No.|
|2005||116 (3.42)||5 (0.15)||3272 (96.43)||3393|
|2006||3326 (40.91)||2807 (34.52)||1998 (24.57)||8131|
|2007||6814 (47.73)||7187 (50.34)||275 (1.93)||14 276|
|2008||8235 (54.53)||6704 (44.39)||164 (1.09)||15 103|
In terms of serious ocular complications, there were 38 cases of endophthalmitis in the anti-VEGF treatment group between 2005 and 2008, compared with only 6 cases in the control group ( Table 3 ). Uveitis was significantly more common in the anti-VEGF treatment group (0.73% vs 0.37%; P < .01). Vitreous hemorrhage also was significantly more common in the anti-VEGF treatment group (1.80% vs 0.94%; P < .01). There were no significant differences between the 2 groups in rates of rhegmatogenous retinal detachment or retinal tear. The rate of endophthalmitis per anti-VEGF injection was 0.09%; the rates of rhegmatogenous retinal detachment, retinal tear, uveitis, and vitreous hemorrhage per injection also were very low, ranging from 0.06% for retinal tear to 0.23% for vitreous hemorrhage. The rate of complication per injection in excess of the observed rate in the control group (the rate presumably attributable to anti-VEGF treatment alone) was 0.08% for endophthalmitis, 0.05% for uveitis, and 0.12% for vitreous hemorrhage.
|Ocular Complication||Control Group (n = 6154 patients), No. (%)||Anti–Vascular Endothelial Growth Factor Treatment Group (n = 6154 patients), No. (%)||P Value||Rate of Complication per Injection in Treatment Group (n = 40 903 Injections), No. (%)||Rate of Complication per Injection in Treatment group in Excess of Observed Rate in Control Group (n = 40 903 Injections) No.(%)|
|Endophthalmitis||6 (0.10)||38 (0.62)||< .01 a||37 (0.09)||33 (0.08)|
|Rhegmatogenous retinal detachment||40 (0.65)||41 (0.67)||0.91||41 (0.10)||1 (0.002)|
|Retinal tear||17 (0.28)||24 (0.39)||0.27||25 (0.06)||8 (0.02)|
|Uveitis||23 (0.37)||45 (0.73)||< .01 a||45 (0.11)||20 (0.05)|
|Vitreous hemorrhage||46 (0.75)||95 (1.54)||< .01 a||94 (0.23)||49 (0.12)|