FormalPara Key Summary Points

The cardiovascular safety profile of intravitreal anti-vascular endothelial growth factor (anti-VEGF) therapy is unclear in patients with diabetes, who are susceptible to higher risks of cardiovascular complications.

The present study assesses the real-world links between incident short-to-medium-term cardiovascular disease (CVD) and intravitreal anti-VEGF therapy in patients with diabetes.

Our study shows that intravitreal anti-VEGF is associated with a numerical but non-statistically significant increased risk of cardiovascular events in patients with diabetes over 12 months, compared to other therapies.

Our study supports the safety of anti-VEGF therapy in patients with diabetes free of CVD at baseline.

Introduction

Intravitreal injections of anti-vascular endothelial growth factors (anti-VEGF) agents have revolutionised the treatment of many ophthalmic pathologies, including diabetic macular oedema and wet age-related macular degeneration (AMD), and they have been shown to be more effective and beneficial than laser or steroid treatments alone in treating these conditions [1,2,3,4]. Currently, four widely used anti-VEGF agents in the ophthalmology field are ranibizumab, bevacizumab, aflibercept, and faricimab, which work primarily by binding to VEGF protein directly and impeding its action [5].

VEGF-A, in particular, contributes to the formation of new blood vessels, or angiogenesis, which is essential for tissue development and function and implicated in pathological conditions [6]. Increased VEGF expression is a crucial factor in the pathogenesis of many ocular diseases, including neovascular AMD, diabetic retinopathy (DR)/macular oedema (DMO), retinal vein occlusion (RVO), glaucoma and retinopathy of prematurity [7]. The increased VEGF production and transcription are thought to be induced by hypoxia, which in turn promotes angiogenesis to restore tissue oxygen supply. However, these new vessels bleed and fibrose easily, leading to complications such as retinal detachment and haemorrhage. VEGF-A also causes inflammation and vascular barrier breakdown, leading to atrophy of the choriocapillaris and degeneration of photoreceptors [8]. In the past decade, anti-VEGF therapy, initially used to suppress tumour growth in cancer treatment, has been re-labelled to treat neovascular eye disease. In addition to inhibiting neovascularisation, anti-VEGF agents have been demonstrated to induce regression of pathological microvessels, stabilise normal vessels and prevent leakage and the concomitant inflammatory response [5].

Systemically delivered anti-VEGF therapy, bevacizumab in particular, has been widely reported to be linked to a variety of cardiovascular adverse events, including acute myocardial infarction (MI), stroke, and heart failure (HF) [5, 9,10,11]. Nevertheless, not all studies demonstrate this excess risk [12]. Inhibition of VEGF-A decreases the production of nitric oxide and prostaglandin-I 2, leading to vasoconstriction and hypertension. Vascular endothelial death caused by anti-VEGF also causes phospholipids to accumulate on the luminal plasma membrane, facilitating thromboembolic events [13].

However, the risk of cardiovascular adverse events with intravitreally delivered anti-VEGF is not clear. A large meta-analysis found that intravitreal anti-VEGF therapy was not associated with an increase in major cardiovascular events; however, there was a possible signal for mortality risk in patients with diabetic retinopathy (OR, 1.80; 95% CI, 1.03–3.16;) [14]. Given the suspected relatively low occurrence of cardiovascular adverse events with intravitreal anti-VEGF, clinical trials lack sufficient power to assess such risks as secondary endpoints. Understanding the risk of cardiovascular events of anti-VEGF therapy is relevant, particularly in patients with diabetes who are at higher risk of cardiovascular disease (CVD), to support clinicians and patients in making informed healthcare choices.

This study aimed to determine incident short-to-medium-term CVD in relation to intravitreal anti-VEGF therapy in patients with diabetes, compared to other therapies, using a large real-world global federated dataset.

Methods

Data

Data were utilised from TriNetX, an electronic global federated health research network of 108 large healthcare organisations (HCOs) across 16 countries (predominately US healthcare data). As part of the data ingestion process, when HCOs join the network, data are mapped to a common data model to reflect individual institution, country and regional standards with regard to electronic health record data. All data collection, processing and transmission are performed in compliance with all Data Protection laws applicable to the contributing HCOs, including the EU Data Protection Law Regulation 2016/679, the General Data Protection Regulation on the protection of natural persons regarding the processing of personal data and the Health Insurance Portability and Accountability Act, the US federal law, which protects the privacy and security of healthcare data. Analytics are performed at the HCOs with only aggregate results being surfaced and returned to the TriNetX platform. Data usage and publication agreements are in place with all HCOs. The TriNetX platform provides access to real-time, real-world electronic medical records, including diagnoses, procedures, medications, laboratory values and genomic information. The TriNetX platform is described in detail elsewhere [15, 16].

Cohort Building

The current study searched for all patients aged 18 years old and over as of the 9th of November 2023 (n = 110,048,154). Of these, patients with diabetes mellitus (ICD-10-CM E08-E13) were included, which consists of type 1, type 2, drug-induced and secondary causes. Patients with a history of CVD, i.e. acute myocardial infarction (MI) (ICD-10-CM I21), cerebral infarction (ICD-10-CM I63) or heart failure (HF) (ICD-10-CM I50) were excluded. Patients with at least one year of data after the index event were included. Patients were then categorised into two cohorts: (1) those who received intravitreal anti-VEGF therapy or (2) alternative therapies. Intravitreal anti-VEGF treatments included all the available intravitreal anti-VEGF agents in the database, which were bevacizumab (n = 73,343), aflibercept (n = 21,022), ranibizumab (n = 6111), brolucizumab (n = 101), and faricimab (n = 57). Control (comparator) treatments included photocoagulation of the retina, intravitreal steroid injection, or implants of fluocinolone and dexamethasone. This cohort was chosen so that patients with diabetes free of CVD would be compared. Triamcinolone was excluded because there was no coding specifically for its intravitreal injections, and its administration route could not be determined.

Index Event/Propensity-Score Matching

The index event is the time point at which data collection began. The index events for this study were defined as first receiving anti-VEGF treatment in cohort 1 or control treatment in cohort 2. The two cohorts for analyses were 1:1 propensity score-matched (PSM) for age, sex, ethnicity, body mass index (< or ≥ 25 kg/m2), systolic blood pressure (< or ≥ 140 mmHg), HbA1c (< or ≥ 7%), and cardiovascular medications (digoxin, beta-blockers, alpha-blockers, calcium channel blockers, antianginals, antiarrhythmics, anti-lipid agents, antihypertensives, peripheral vasodilators, diuretics, angiotensin-converting enzyme inhibitors, angiotensin II inhibitor and direct renin inhibitor). TriNetX performed a logistic regression utilising the scikit-learn package in Python version 3.7. Nearest-neighbour matching with a tolerance level of 0.01 and the difference between propensity scores ≤ 0.1 was used.

Outcome Comparison and Statistical Analysis

Outcomes analysis results were calculated in the TriNetX platform using R's Survival package v3.2–3 and validating those results with output from SAS version 9.4. Four outcomes were measured: (1) all-cause mortality and diagnoses of (2) acute MI, (3) cerebral infarction, and (4) HF. Outcome comparisons were made over 1 month, 6 months and 12 months from the index event. The relative risks (RR) were used to estimate the probability of the outcome at the respective time interval, and 95% confidence intervals (CIs) are presented. The t test and X2 statistical testing were conducted for differences in outcomes between cohorts. A p value < 0.05 was considered statistically significant.

Results

A total of 5,821,858 patients with diabetes were identified without a history of CVD. Of these, 24,060 (0.41%) received anti-VEGF treatment, and 2680 (0.05%) received alternative treatments (Fig. 1). In the anti-VEGF treatment group, there were 17,737 (73.7%) patients with type 2 diabetes and 2573 (10.7%) with type 1 diabetes. Within the control group, there were 1581 (59%) with type 2 diabetes and 375 (14%) with type 1 diabetes. After 1:1 propensity matching, the number of patients in both groups reduced to 2205. No patients were censored from either 1-month, 6-month or 12-month analyses. A summary of propensity score-matching (PSM) characteristics is summarised in Table 1.

Fig. 1
figure 1

Patient selection. Individuals with diabetes without pre-existing cardiovascular disease (CVD) received anti-vascular endothelial growth factors (anti-VEGF) therapy or alternative therapy. MI myocardial infarction, HF heart failure

Full size image
Table 1 Patient characteristics before and after propensity score matching (PSM)
Full size table

All-Cause Mortality

There were numerical but non-statistically significant differences in the risk of all-cause mortality over 1 month (RR 1; 95% CI 0.42,2.40), 6 months (RR 1.57; 95% CI 0.91,2.71) and 12 months (RR 1.28, 95% CI 0.90,1.83) between the anti-VEGF and control groups.

Acute Myocardial Infarction

No events of acute MI were found in either group over 1 month (RR n/a). There were ten events in each group over 6 months, with no differences in the risk of acute MI (RR 1; CI 95% 0.42, 2.40). The incidence of acute MI did not change over 12 months (RR 1; CI 95% 0.42, 2.40) between the anti-VEGF and control groups.

Cerebral Infarction

There were no events of cerebral infarction in the anti-VEGF group but ten events in the control group over 1 month (RR n/a). The incidences remained the same over 6 and 12 months (RR n/a).

Heart Failure

There were no events of HF in either group over 1 month (RR n/a). There were ten events in each group over 6 months, with no difference in the risk of heart failure (RR 1; 95% CI 0.42, 2.40). The incidence of HF did not change over 12 months (RR 1; 95% CI 4.42, 2.40) between the anti-VEGF and control groups (Fig. 2).

Fig. 2
figure 2

Summary of cardiovascular outcomes with anti-vascular endothelial growth factor (anti-VEGF) therapy in patients with diabetes over 1, 6 and 12 months. RR relative risk, CI confidence interval, MI myocardial infarction. Data expressed the number of outcomes and relative risk (95% confidence interval)

Full size image

Discussion

In this study, we have conducted a real-world retrospective propensity-matched cohort analysis to evaluate the putative association between cardiovascular events and intravitreal anti-VEGF therapy in patients with diabetes. Our study demonstrates that intravitreal anti-VEGF therapy is not associated with a significantly increased risk of cardiovascular events in patients with diabetes over 12 months.

Intravitreal anti-VEGF therapy has become a pillar of treatment for multiple eye conditions in recent years. Until now, their real-world safety profile in patients with diabetes has been unclear. Previously, there were also concerns that there may be greater susceptibility to cardiovascular events/complications [14].

A number of studies have evaluated the systemic adverse effects of intravitreal anti-VEGF therapy. However, the overall results are conflicting: a 2018 Cochrane review of 24 trials with 6007 patients with diabetes and DMO found intravitreal anti-VEGF therapy was not associated with increased risk of death with aflibercept (RR 1.01, 95% CI 0.34,3.03), bevacizumab (RR 1.61, 95% CI 0.45, 5.69) and pegaptanib (RR 0.81, 95% CI 0.16,4.03), and ranibizumab (RR 0.90, 95% CI 0.40, 2.01); there was also no increased risk of thromboembolic events at 24 months with any of the anti-VEGF drugs compared to the control group (alternative treatment, sham or no treatment). However, the analysis was incoherent, and the evidence was considered to be of low or very low certainty [17].

In contrast, a meta-analysis of four trials (n = 1328) has shown that those receiving maximum monthly doses of intravitreal anti-VEGF in DMO had an increased risk of death (OR 2.57, 95% CI 1.31,5.05) and cerebrovascular events (OR 2.33, 95% CI 1.04,5.22) but no increased risks of MI or arteriothrombotic events, compared with those receiving sham injections and laser treatments [18]. The regularity and frequency of the anti-VEGF administration in this study design are more relevant to clinical practice. Increased risk of stroke was also demonstrated in a review of five trials of ranibizumab compared to control (HR 2.2, 95% CI 0.80, 7.10). Further sensitivity analysis in a high-risk subgroup of patients, which included patients aged over 85 with a history of stroke or transient ischaemic attack, demonstrated increased risk (HR 2.12, 95% CI 0.75, 5.90) [18].

Inconsistent findings in the literature may be due to the differences in study design. Some trials do not exclude patients with a history of cardiovascular disease, which could lead to a higher occurrence of adverse events. Further, the anti-VEGF drugs and their dosage were also varied between studies. Overall, the incidence of adverse events was low among studies, and the CIs were wide. Additionally, most studies are likely underpowered to accurately detect differences in risks of intravitreal anti-VEGF, even if they exist.

It is well established that anti-VEGF therapy may enter the circulation when administrated intravitreally [5]. Plasma VEGF levels were shown to be reduced by bevacizumab and aflibercept after 7 days and 1 month in patients with AMD; however, no significant reductions of plasma VEGF levels were observed in patients receiving ranibizumab during follow-up compared to the other two anti-VEGF drugs [19]. Ranibizumab has a smaller molecular size and lacks Fc domain, leading to a different pharmacokinetic profile than other anti-VEGF agents [20]. These findings suggest that the distinctive anti-VEGF agents may influence the free VEGF levels differently.

In addition to angiogenesis, VEGF-A is important for cardiac morphogenesis, contractility and remodelling of the myocardium. Depletion of VEGF in animal models has been shown to lead to impaired myocardial angiogenesis with subsequent heart failure [21]. In patients who experienced MI, low plasma VEGF levels were found to be associated with a significantly increased risk for further major adverse cardiovascular and cerebrovascular events [22]. This paradoxical finding makes predicting anti-VEGF’s adverse effects challenging. The pharmacodynamic-pharmacokinetic profiles of anti-VEGF therapy appear to be complex and not well understood. Since most patients would require repeated injections for chronic VEGF inhibition, further longer-term studies with robust treatment regimens are needed to reflect the risk of cardiovascular in patients with diabetes receiving serial or long-term anti-VEGF therapy.

Strengths

Our study includes a large number of patients from a real-world database, potentially enhancing statistical power and generalisability. We also used propensity matching against appropriated patient characteristics and cardiovascular phenotype to address significant potential confounding factors.

Limitations

Our study is limited by the coding system of the platform. We could only select inclusion and exclusion criteria based on the ICD-10 codes. The respective indications for intravitreal injections were not reported. Given the limited event rate in the overall analysis, a sub-analysis of different types of diabetes and specific anti-VEGF agents was not undertaken. The details of the eye disease, such as type, duration, and severity, could not be entirely determined. The tobacco use was also not propensity-score matched as data availability for this variable was more limited. The index event was set as the initiation of either anti-VEGF agents or control treatment, however we could not determine the dose or frequency of the injections. Another source of uncertainty is that patients could have received cross-over therapies after the index event and additionally we are unable to account for residual effect from unknown confounders.

Conclusions

There was no statistically significant risk of cardiovascular-related events in the short or medium term in patients with diabetes who received intravitreal anti-VEGF therapy, despite a very small increase in the number of CVD events. Our study supports the real-world safety of intravitreal anti-VEGF therapy in patients with diabetes free of baseline CVD.