Purpose
To examine the risk of myocardial infarction and stroke with single and repeated doses of intravitreal bevacizumab in wet age-related macular degeneration (AMD).
Design
Nested case-control study and retrospective cohort study.
Methods
setting : Two patient cohorts from British Columbia, Canada. study population : Patients with wet AMD. intervention : For the cohort study, patients who received the first intravitreal bevacizumab; for the nested case-control study, repeated injections of intravitreal bevacizumab. main outcome measures : Myocardial infarction for the retrospective cohort study; myocardial infarction and stroke for the nested case-control study.
Results
In the cohort analysis, there were 2564 AMD subjects not on a vascular endothelial growth factor (VEGF) inhibitor and 5644 subjects receiving intravitreal bevacizumab. The rate of myocardial infarction (MI) among bevacizumab users was 11/1000 person-years, compared to 14.9/1000 person-years in nonusers. The adjusted rate ratio (RR) for MI was 0.70 (95% confidence interval [CI]: 0.50–1.00) and 0.74 (0.46–1.20) for the propensity score–adjusted analysis. In the nested case-control analysis there were 7452 new users of VEGF inhibitors, within which there were 313 cases of MI with 3130 matched controls. The adjusted RR for MI among those receiving 3 or more injections compared to those receiving fewer than 3 was 0.71 (95% CI: 0.41–1.22). Also in the nested case-control analysis, the adjusted RR for stroke was 0.81 (95% CI: 0.39–1.65) for those receiving ≥4 injections vs those receiving fewer than 4 injections.
Conclusion
Single or repeated doses of intravitreal bevacizumab were not shown to increase the risk of myocardial infarction or stroke in patients with wet AMD.
Vascular endothelial growth factor (VEGF) inhibitors have revolutionized the treatment of neovascular or wet age-related macular degeneration (AMD). VEGF has a pivotal role in the homeostasis of blood vessels outside of the retina including those in the heart and brain. Bevacizumab is a VEGF inhibitor approved for the treatment of cancer but also used in wet AMD, as it has shown to reduce neovascular leakage in this condition. Recently, there have been concerns regarding myocardial infarction (MI) and stroke with intravitreal VEGF inhibitors. Intravitreal bevacizumab is still used in patients with AMD despite the lack of regulatory approval because of substantially lower cost over other intravitreal VEGF inhibitors and equal efficacy to ranibizumab. Differences in cardiovascular (CV) safety profiles between intravitreal bevacizumab and other approved intravitreal VEGF inhibitors in wet AMD may result in serious ethical and policy implications if bevacizumab is used in this fashion and a higher risk of CV harm exists. Systematic reviews have attempted to answer the question of differential safety profiles between VEGF inhibitors. However, lack of statistical power, small numbers of events, and sporadic reporting of CV events in clinical trials included in these reviews make interpretation of cardiovascular risk challenging. Two large epidemiologic studies have compared cardiovascular event rates between intravitreal bevacizumab and ranibizumab. Neither study showed a statistically significant difference in the rate of MI or stroke between the 2 drugs. However, in both studies VEGF inhibitor use was assessed based on a single injection and the risk of adverse cardiovascular events with repeated injections was not examined. A recent within-patient crossover study that also examined single intravitreal injections of bevacizumab or ranibizumab found an increase in the risk of thromboembolic events with intravitreal bevacizumab and ranizibizumab.
VEGF is a subfamily of signaling proteins involved in both vasculogenesis and angiogenesis, the inhibition of which may lead to MI or stroke. Intravitreal injection of this drug could increase the risk of adverse CV events, but given that this is a localized administration, systemic absorption of repeated doses would be expected to increase the risk of drug-induced cardiovascular events more than a single dose. The recommended dose for intravitreal injection of bevacizumab is 150 times less than its systemic dose. Studies have shown that intravitreal bevacizumab has a higher systemic absorption than ranibizumab or aflibercept. If the systemic absorption of bevacizumab is of concern when used intravitreally, then it is probable that the antithrombotic effects of intravitreal bevacizumab might occur more frequently with multiple injections. To date, no study has examined the risk of CV events with repeated injections of intravitreal bevacizumab in wet AMD. Answering this question will be vital to clinicians who may opt to use bevacizumab over other VEGF inhibitors mainly owing to equivalent efficacy and significantly lower cost. This study examined the risk of myocardial infarction and stroke with both single and repeated doses of intravitreal bevacizumab in wet AMD.
Methods
Setting and Patient Population
This is a population-based observational study using the British Columbia (BC) Ministry of Health linked databases. All subjects in BC are entitled to receive VEGF inhibitors through The BC Provincial Retinal Disease Treatment Program. Intravitreal bevacizumab was the drug used more than 90% of the time during the study period. Clinical ophthalmic data for all patients are recorded in a comprehensive database that captures all intravitreal injections administered by retina specialists. These data are subject to frequent quality control and are linkable to other BC Ministry of Health Databases, which capture all hospitalizations through the Discharge Abstract Database, all physician visits obtained through the Medical Services Plan data file (MSP), and prescription drugs that are dispensed to patients, including drug strength and dose from the PharmaNet database. The positive predictive value (PPV) for MI and stroke in Canadian population health databases have been shown to be 95% and 92%. These databases have been used in many previous epidemiologic studies. We did not have information on death or smoking status. Ethics approval was granted by the University of British Columbia’s Clinical Research Ethics Board prior to the commencement of the study.
Study Design
Two study designs were used to answer 2 distinct questions. First, we used a retrospective cohort study to examine the risk of MI and stroke in patients with wet AMD who were administered intravitreal bevacizumab compared to those not taking VEGF inhibitors. Second, we conducted a nested case-control study within new users of intravitreal bevacizumab on the risk of MI or stroke. The nested case-control study was used because it is an optimal design for time-varying exposures such as prescription drugs and allows for the control of strong confounders such as age and calendar time through matching. The advantage of the retrospective cohort study is the ability to compare the risk of an MI among intravitreal bevacizumab users and nonusers. Since this design emulates a clinical trial, exposure in this design was defined as a patient having received a single intravitreal injection of bevacizumab. The advantage of the nested case-control design is its ability to compare the risk of MI and stroke among patients with different numbers of intravitreal injections, emulating clinical practice where repeated injections of VEGF inhibitors are the standard treatment strategy.
Retrospective Cohort Study
The exposed group was defined as all patients in British Columbia newly prescribed intravitreal bevacizumab for wet AMD between June 2009 and October 2013. Exposed subjects entered the cohort on the first date of injection of intravitreal bevacizumab and were followed to the first MI event, the last injection date, or termination of health services coverage, whichever came first. For the unexposed group, we used subjects from a separate dataset of all BC residents who had at least 1 ophthalmology visit between 2000 and 2007. Unexposed subjects were included if they had received at least 2 International Classification of Diseases, Ninth Revision (ICD-9) codes (362.52 or 362.5) for wet AMD in 1 year any time between 2000 and 2007 and had not received an intravitreal injection for a VEGF inhibitor during this period. Cohort entry for this group was deemed the first ICD-9 code for AMD. Both cohorts were followed to the first MI event (ICD-9 410) or the last date of follow-up. A Cox proportional hazards model was constructed to estimate rate ratios (RR) adjusting for the following covariates in the year prior to cohort entry: age, sex, hypertension, diabetes, deep vein thrombosis, stroke, peripheral vascular disease, and the use of statins, angiotensin system blockers (angiotensin converting enzyme inhibitors and angiotensin receptor blockers), anticoagulants, antiplatelet agents, calcium channel blockers, and diuretics. In order to better control for the imbalance between the groups, we also conducted a propensity score–adjusted analysis. Propensity scores were computed for both groups using logistic regression. Subsequently, the groups were matched by the propensity scores using the nearest available neighboring match. Standardized differences were used to assess the balance between the 2 groups with a difference of <10% considered adequate balance. Owing to the small number of stroke events in the cohort, only the outcome of MI was assessed in this study.
Nested Case-Control Study
For this study, cohort entry was defined as the first intravitreal injection between 2009 and 2013 and patients were followed to the first MI diagnosis (International Classification of Diseases, Tenth Revision [ICD-10] code I21-22), which was deemed the index date, or until the end of the study period, whichever came first. For each case, 10 controls that were at risk of developing the outcome were identified and matched to the cases by age, cohort entry date, and follow-up time using a density-based sampling approach. This approach allows for the close estimation of the RRs from the odds ratio (OR), which was derived from the conditional logistic regression model created. A similar methodological approach was used for the outcome of stroke (ICD-10 code I61-I64). Owing to the small number of patients receiving ranibizumab (<10% of patients), we only examined the risk of MI or stroke in patients receiving intravitreal bevacizumab. To compute the RRs for different injection risk periods, we first ascertained the median number of intravitreal injections from cohort entry to the index date. RRs were computed for injection counts above the median number compared to below the median number of injections. (To prevent any bias that may be introduced from patients who switched from bevacizumab to ranibizumab during the follow-up period, we excluded switchers from the analysis.) The effect of dose was also assessed as a continuous variable in the regression model. The adjusted model included the following covariates: sex, diabetes, deep vein thrombosis, peripheral vascular disease, hypertension, statins, angiotensin system blockers, anticoagulants, antiplatelet agents, beta blockers, calcium channel blockers, and diuretics.
Results
Retrospective Cohort Study Analysis
The retrospective cohort study included 5644 patients receiving intravitreal bevacizumab and 2564 wet AMD patients not on a VEGF inhibitor ( Table 1 ). The rate of MI among intravitreal bevacizumab users was 11/1000 person years compared to 14.9/1000 person years in nonusers. The adjusted RR for MI was 0.70 (95% confidence interval [CI]: 0.50–1.00). After propensity score matching, the standardized difference for the covariates between exposed and unexposed patients was <10%. Two exceptions were stroke and angiotensin receptor blockers, where the standardized difference was 11%, indicating a good balance between the 2 groups ( Table 2 ). The propensity–adjusted RR was 0.74 (95% CI: 0.46–1.20).
| Patient Characteristics | Bevacizumab | No Bevacizumab |
|---|---|---|
| Number | 5644 | 2564 |
| Demographics | ||
| Age (y), mean ± SD | 80.1 ± 10.1 | 77.9 ± 11.33 |
| Sex, male (%) | 38.4 | 37.4 |
| Event | ||
| Number of myocardial infarctions | 95 | 157 |
| Covariates (%) | ||
| Stroke | 23.4 | 2.7 |
| Hypertension | 44.5 | 34.2 |
| Diabetes | 29.3 | 9.9 |
| Deep vein thrombosis | 20.5 | 15.2 |
| Peripheral vascular disease | 26.9 | 1.6 |
| Statin use | 33.5 | 18.0 |
| Angiotensin system blockers | 28.4 | 33.5 |
| Anticoagulants or antiplatelets | 11.4 | 11.0 |
| Beta-blockers | 23.1 | 21.2 |
| Calcium channel blockers | 27.3 | 20.6 |
| Diuretics | 3.3 | 32.6 |
| Patient Characteristics | Bevacizumab | No Bevacizumab | Standardized Difference (%) a |
|---|---|---|---|
| Number | 2564 | 2564 | |
| Demographics | |||
| Follow-up (y) | 1.5 ± 1.1 | 4.1 ± 2.1 | |
| Age (y), mean ± SD | 77.9 ± 10.8 | 77.9 ± 11.3 | 0.00 |
| Sex, male (%) | 38.8 | 37.4 | 0.03 |
| Event | |||
| Number of myocardial infarctions | 46 | 157 | |
| Covariates (%) | |||
| Stroke | 1.2 | 2.7 | 11.0 |
| Hypertension | 25.9 | 28.2 | 7.0 |
| Diabetes | 11.2 | 10.5 | 4.0 |
| Deep vein thrombosis | 0.3 | 0.2 | 2.0 |
| Peripheral vascular disease | 1.9 | 1.6 | 2.0 |
| Statin use | 16.9 | 18.0 | 3.0 |
| Angiotensin system blockers | 28.5 | 33.5 | 11.0 |
| Anticoagulants or antiplatelets | 11.9 | 11.0 | 3.0 |
| Beta blockers | 24.7 | 21.2 | 8.0 |
| Calcium channel blockers | 21.3 | 20.6 | 2.0 |
| Diuretics | 33.2 | 32.6 | 1.0 |
a A standardized difference of 10% or less is indicative of good balance for a specific covariate between the 2 groups. Differences of 10%-20% are still considered an acceptable balance.
Nested Case-Control Analysis
In the nested case-control study, there were 313 cases of MI with 3130 matched controls. The median number of intravitreal injections in the year prior to MI was 3. More cases than controls had comorbidities ( Table 3 ). The adjusted RR for MI among those receiving 3 or more intravitreal bevacizumab injections compared to those receiving fewer than 3 was 0.71 (95% CI: 0.41–1.22). In the dose-response analysis, the RR of MI per additional intravitreal injection of bevacizumab was 0.97 (95% CI: 0.92–1.04). For the stroke analysis, there were 65 cases of strokes with 650 corresponding controls ( Table 4 ). The median number of injections in the year prior to the stroke was 4. The adjusted RR for stroke among those receiving 4 or more injections compared to those receiving fewer than 4 injections was 0.81 (95% CI: 0.39–1.65) ( Table 5 ). In the dose-response analysis, the RR for stroke per unit increase of bevacizumab injection was 1.00 (95% CI: 0.91–1.09).
