Introduction

Systemic vascular endothelial growth factor (VEGF) blockade has been the top adjunctive chemotherapy since 1990 (Hanna et al. 2019). Anti-angiogenic agents are widely used to treat various malignancies, such as non-small cell lung cancer, renal cell carcinoma, and colorectal cancer (Hanna et al. 2019; Phadke et al. 2021). VEGF inhibitors are powerful tools used to delay neovascularization and prevent retinal damage (Phadke et al. 2021). Intravitreal anti-VEGF therapy is widely used for the treatment of proliferative diabetic retinopathy/diabetic macular edema, age-related macular degeneration, and central retinal vein occlusion (Hanna et al. 2019; Hanna et al. 2022). Bevacizumab was the first anti-VEGF antibody used off-label via intravitreal injections to treat retinal neovascularization (Hanna et al. 2022). Aflibercept and ranibizumab have also been approved by the U.S. Food and Drug Administration for ophthalmological use (Hanna et al. 2019, 2022).

Compared to systemic administration, intravitreal use of anti-VEGF agents was initially considered to be a safer choice (Hanna et al. 2019). However, systemic exposure was reported to be more impactful in later studies (Avery et al. 2014, 2017). Bevacizumab, aflibercept, and ranibizumab are immediately circulated in the body (Avery et al. 2014, 2017). Specifically, ranibizumab is rapidly cleared from the bloodstream and exerts the weakest anti-VEGF effects (Avery et al. 2014, 2017). Aflibercept is the most effective in decreasing the plasma-free VEGF levels (Avery et al. 2014, 2017). Moreover, multiple intravitreal injections prolong the period of detectable systemic exposure of the drug (Avery et al. 2014).

Systemic toxicity of intravitreal VEGF blockade is a serious concern. Several studies have reported that intravitreal anti-VEGF therapy is associated with increased hypertension (Rasier et al. 2009), proteinuria (Chung et al. 2020), and thrombotic events (Schmid et al. 2015; Avery and Gordon 2016). Anti-VEGF therapy has also been associated with worsened renal function in some patients (Diabetic Retinopathy Clinical Research N et al. 2007; Jalalonmuhali et al. 2020). Serial retrospective studies have reported that intravitreal bevacizumab increases the risk of mortality in patients with age-related macular degeneration (Hanhart et al. 2017, 2018a, b). However, the association between patient outcomes and use of intravitreal VEGF inhibitors remains controversial, as some studies have reported no significant effects of these inhibitors on hypertension (Risimic et al. 2013; Glassman et al. 2018), proteinuria (Glassman et al. 2018; O’Neill et al. 2019; Bagheri et al. 2018), renal function (O’Neill et al. 2019; Bagheri et al. 2018; Kameda et al. 2018), and mortality (Dalvin et al. 2019) in patients.

Among currently available intravitreal VEGF inhibitors, ranibizumab is less potent and considered a safer option to treat patients with comorbidities (Hanna et al. 2022). Moreover, ranibizumab exhibited the least systemic exposure and decrease in plasma VEGF levels after intravitreal injection compared to bevacizumab and aflibercept (Avery et al. 2014, 2017). Ranibizumab also reduced nephrotoxicity, without any significant effect on glomeruli, in an animal study (Tschulakow et al. 2014). In contrast, several studies have reported that ranibizumab decreases renal function, worsens proteinuria, and causes thrombotic microangiopathy (Phadke et al. 2021; Pelle et al. 2011; Morales et al. 2017). As it is prioritized in the treatment of patients with comorbidities, it is necessary to determine the action mechanism and long-term renal effects of ranibizumab (Hanna et al. 2019; Estrada et al. 2019).

In this nationwide cohort study, we used propensity score matching to determine the risk of chronic kidney disease (CKD) in patients receiving intravitreal ranibizumab injections in Taiwan.

Methods

Data source

Since March, 1995, > 99% of the Taiwanese population has been insured via the National Health Insurance (NHI) program of Taiwan. National Health Insurance Research Database (NHIRD) is one of the largest global databases that are extensively used for epidemiological research. NHIRD contains the medical information of all insureds, such as inpatient, outpatient, emergency, and traditional Chinese medicine records. Patient diagnoses were recorded according to the International Classification of Diseases, 9th Revision, Clinical Modification, and International Classification of Diseases, 10th Revision (ICD-9-CM and ICD-10-CM). All analyses were conducted at the China Medical University branch center of the Ministry of Health and Welfare. This study was approved by the Institutional Review Board of China Medical University (CMUH110-REC3-133).

Subject inclusion and exclusion criteria

We selected subjects aged ≥ 20 years recently administered ranibizumab for the ranibizumab cohort, and the index date was defined as the first date of ranibizumab treatment between 2008 and 2018. Non-ranibizumab cohort consisted of subjects who did not receive ranibizumab, and the index date was a random date between 2008 and 2018. We excluded subjects with missing sex and age records and those in which the date of primary outcome was before the index date. The two cohorts were matched via 1:1 propensity score matching based on sex, age, index year, comorbidities, and medications. In the following section, definitions for comorbidities and medications will be expounded upon.

Primary outcomes, comorbidities, and medications

The primary endpoint was CKD in this study. We followed up the patients until CKD was diagnosed at the end of 2019 or until they withdrew from the NHI. CKD was defined as ICD-9-CM code 585 and ICD-10-CM code N18. Comorbidities with potential confounding factors were extracted from the database from January 1, 2008 to the index date of the subject. CKD-related comorbidities consisted of hypertension (ICD-9-CM code 401–405 and ICD-10-CM codes I10, I11, I12, I13, I15, and N26.2), diabetes mellitus (ICD-9-CM code 250 and ICD-10-CM code E08-E13), hyperlipidemia (ICD-9-CM code 272 and ICD-10-CM codes E71.30, E75.21, E75.22, E75.24, E75.3, E75.5, E75.6, E77, E78.0, E78.1, E78.2, E78.3, E78.4, E78.5, E78.6, E78.70, E78.79, E78.8, and E78.9), stroke (ICD-9-CM code 430–438 and ICD-10-CM code I60-I69), coronary artery disease (CAD; ICD-9-CM code 410–414 and ICD-10-CM code I20-I25), alcoholism (ICD-9-CM codes 291, 303, 305.0, 571.0–571.3, 790.3, V11.3, and V79.1 and ICD-10-CM codes F10, K70, R78.0, and Z65.8), chronic obstructive pulmonary disease (COPD; ICD-9-CM codes 490–496 and 504–506 and ICD-10-CM codes J40-J47 and J64-J68), and age-related macular degeneration (AMD; ICD-9-CM codes 362.52, ICD-10-CM: H35.32), retinal vein occlusion (RVO; ICD-9-CM: 362.36, ICD-10-CM: H34.83), and diabetic macular edema (DME; ICD-9-CM: 362.01, ICD-10-CM: E11.311). Medical confounders were angiotensin I-converting enzyme inhibitors, statins, corticosteroids, VEGF inhibitors including bevacizumab and aflibercept, lithium, amphotericin B (AmB), adefovir, NSAIDS, cisplatin, and calcineurin inhibitors (CNIs). Medication usage was defined as receiving a prescription after the index date.

Statistical analyses

Standardized mean difference (SMD) was used to estimate the differences in baseline characteristics between the case and control groups. Density of CKD events per 1000 person-years was calculated for both cohorts during the study period. Cox model was used to compare the risk of CKD between the case and control groups. Model 1 included a crude estimate of the hazard ratio and a 95% confidence interval (CI). Model 2 was adjusted for age, sex, comorbidities, and medications, and it estimated the adjusted HR (AHR). Moreover, Kaplan–Meier analysis and log-rank tests were used to estimate the difference in the cumulative incidence of CKD between the two groups.

Results

Among 48,272 participants aged ≥ 20 years, 24,136 (50%) received ranibizumab (13,565 men [56.20%] and 10,571 women [43.80%]). Moreover, 24,136 participants who did not receive ranibizumab were matched by age, sex, comorbidities, and medications (Table 1), and the baseline characteristics were found to be well-balanced. Mean (standard deviation) ages of subjects in the case and control groups were 66.23 (12.13) and 66.42 years (12.79), respectively. In the non-ranibizumab and ranibizumab cohorts, the top three comorbidities were hypertension (69.84 vs. 72.65%), diabetes mellitus (61.09 vs. 61.71%), and hyperlipidemia (59.40 vs. 64.36%). Compared with those in the control group, subjects in the case group had similar proportions of comorbidities and medications.

Table 1 Characteristics of individuals with and without ranibizumab treatment
Full size table

Table 2 presents the association between CKD incidence and patients with and without ranibizumab treatment. Subjects who received ranibizumab exhibited a significantly higher risk of CKD than those who did not receive ranibizumab (AHR = 1.88, 95% CI = 1.79–1.96). Compared with the female subjects, male subjects had higher risk of CKD (AHR = 1.29, 95% CI = 1.23–1.35). Compared with subjects in the 20–49 age group, subjects in the 50–64 age group and those > 65 years had a lower risk of CKD. The adjusted HRs were 0.76 (0.70–0.83) and 0.82 (0.75–0.89), respectively. In addition, patients with hypertension (adjusted HR = 1.86, 95% CI = 1.73–1.99), diabetes mellitus (adjusted HR = 2.37, 95% CI = 2.21–2.54), hyperlipidemia (adjusted HR = 1.19, 95% CI = 1.13–1.26), stroke (adjusted HR = 1.16, 95% CI = 1.10–1.23), CAD (adjusted HR = 1.24, 95% CI = 1.18–1.30), and COPD (adjusted HR = 1.10, 95% CI = 1.04–1.15) exhibited significantly higher risks of CKD than the corresponding groups (AHR > 1; p < 0.05). Moreover, patients who received statin, VEGF inhibitor, lithium, and NSAIDS had significantly lower risks of CKD than the corresponding groups (aHR < 1; p < 0.05).

Table 2 Risk factor analyses of all subjects for chronic kidney disease (CKD)
Full size table

Table 3 presents the subgroup analysis results: subjects with varying sex, age, comorbidities, and medications; ranibizumab users tended to have a lower risk of CKD than non-ranibizumab users (adjusted HR > 1, p-value < 0.05).

Table 3 Incidence and hazard ratio of CKD in individuals with and without ranibizumab treatment based on age, gender, comorbidities, and medications
Full size table

As shown in Fig. 1, the cumulative incidence of CKD in the ranibizumab cohort was significantly higher than that in the non-ranibizumab cohort (log-rank test p < 0.001).

Fig. 1
figure 1

Comparison of cumulative incidence of CKD between patients receiving ranibizumab and those not receiving the treatment

Full size image

Discussion

Our findings provide evidence that intravitreal ranibizumab, a VEGF inhibitor, is associated with a higher risk of CKD. Using propensity score matching, we created a control group with a comparable distribution to the ranibizumab group regarding gender, age, comorbidities, and medication. Multivariate analysis revealed increased risk of CKD in the ranibizumab group. Subgroup analysis also revealed that ranibizumab was associated with a higher risk of CKD.

To the best of our knowledge, this is the first population-based study to highlight the correlation between intravitreal anti-VEGF therapy and long-term renal patient outcomes. Although several case studies suggested the nephrotoxicity of intravitreal VEGF inhibitors, only a few focused on their chronic effects on renal function (Hanna et al. 2019, 2022). In a retrospective cohort of 85 diabetic macular edema cases, O’Neill et al. reported no association between glomerular filtration rate decline and intravitreal injections of VEGF inhibitors, mainly ranibizumab, in a mean duration of 31 months (O’Neill et al. 2019). Another randomized study of 660 participants with diabetic macular edema revealed that urine albumin–creatinine ratio does not change significantly after intravitreal aflibercept, bevacizumab, and ranibizumab treatment for 52 weeks (Glassman et al. 2018). The first study cohort may be too small to detect rare renal toxicities (O’Neill et al. 2019). In the second study cohort, chronic renal outcomes may not be affected due to the short following period (Glassman et al. 2018). Most studies only focused on diabetic patients with macular edema. Therefore, our study provides robust evidence for the association between intravitreal use of ranibizumab and increased risk of CKD over a long period using a nationwide database. Meanwhile, ranibizumab was selected as the representative treatment in our study because intravitreal bevacizumab is non-reimbursable, and its data is lacking in NHID. Ranibizumab was the first intravitreal anti-VEGF therapy reimbursed by NHI, and it was more clinically experienced than aflibercept before 2018. Although some aflibercept users might be included in the control group, the risk of ranibizumab was still significant. Since our study showed that even less potent and safer ranibizumab was associated with an increased risk of CKD, it is worthwhile to include aflibercept to compare their effects on kidney function and investigate whether drug class effects of adverse long-term renal outcomes exist by using an updated database in the future study.

In Taiwan, old age is associated with a high risk of CKD (Kuo et al. 2007), and CKD is more prevalent in group aged > 65 years (Wen et al. 2008). Age distribution of our cohort revealed that the patients treated with ranibizumab who were eligible for reimbursement, such as those with diabetic macular edema, age-related macular degeneration, polypoidal choroidal vasculopathy, and central or branch retinal vein occlusion with macular edema, were mainly elderly patients. As the elderly are more susceptible to renal injury, several measures have been proposed to minimize the nephrotoxicity of VEGF inhibitors (Estrada et al. 2019). Interestingly, our study revealed a higher risk of developing CKD in the group aged < 50 years after adjusting several covariates including diabetes. Here, we inferred that the young ranibizumab users mostly consisted of patients with diabetic macular edema based on previous epidemiological reports (Ho et al. 2008; Chang et al. 2018; Chang and Wu 2009). Since the development of diabetic retinopathy is correlated with the duration of diabetes (Zhang et al. 2010; Lee et al. 2015), the age of diabetic onset in young diabetic patients using ranibizumab may be considered earlier than old diabetic users. Several studies have shown that the young age of diabetes onset might be an important risk factor for renal complications (Magliano et al. 2020; Wu et al. 2021; Lee et al. 2023). One recent large prospective cohort study in China revealed that the young age of onset of diabetes synergistically enhanced the risk of CKD among the influence of diabetes duration (Wu et al. 2021). Another large population study in Korea also showed that patients with young-onset diabetes (age < 40 years) had 70% higher risk of developing CKD than the late-onset group (Lee et al. 2023). Our finding was consistent with the previous literatures. Current evidence proposed that more rapid β cell failure and obesity in young-onset patients than in late-onset patients leads to worse disease progression (Magliano et al. 2020). Besides, the correlation between intravitreal VEGF antagonist-related renal damage and diabetic nephropathy has also been reported (Morales et al. 2017; Hanna et al. 2020; Shye et al. 2020). Because our cohort was lacking in several parameters such as oral glucose tolerance test, albuminuria, or body mass index, we were unable to determine the possible mechanisms of higher risk of CKD in young ranibizumab users. Nonetheless, our result highlighted the need for clinical awareness, aggressive treatment, and closer collaboration between nephrologists and ophthalmologists to reduce renal progression in ranibizumab users aged < 50 years.

Mechanisms underlying the associations reported here remain unknown. Anti-VEGF nephrotoxicity is associated with various renal pathological manifestations, such as focal segmental glomerulosclerosis, minimal changes disease, membranous nephropathy, acute interstitial nephritis, and thrombotic microangiopathy (Hanna et al. 2019, 2022; Estrada et al. 2019). Downstream effects of VEGF inhibition on podocytes and glomerular endothelial cells have been proposed (Hanna et al. 2022; Estrada et al. 2019). For example, sequestration of VEGF leads to complement activation and increased nuclear factor-κB signaling in both podocytes and endothelial cells, subsequently causing thrombotic microangiopathy (Estrada et al. 2019). Because patients receiving intravitreal ranibizumab are not required specific monitoring, delayed recognition of subclinical renal injury may potentially occur. Acute kidney injury and CKD are intercorrelated (Okusa et al. 2009; Coca et al. 2012). Notably, patients recovering from acute kidney injury may be at risk of CKD due to nephron loss, incomplete repair, inflammation, fibrosis, and epigenetic changes (Wang and Zhang 2022; Tanemoto et al. 2022). As our study lacked serial clinical and laboratory parameters, such as blood pressure or urine protein levels, further prospective studies are needed to examine the causality and time course of the association between intravitreal VEGF inhibitors and risk of CKD.

This study has several limitations. Although we used the propensity score matching method and adjusted for extensively available covariates, we could not account for other residual confounding factors that contributed to the development of CKD. For example, data on cigarette smoking and obesity, two important risk factors for age-related macular degeneration and CKD (Wen et al. 2008; Chakravarthy et al. 2010), were lacking in our cohort. In addition, control group was selected according to the propensity score, and hence, did not reflect the actual health situation of the general population. This may have resulted in an underestimation of CKD risk. Furthermore, our findings may have surveillance bias as patients receiving ranibizumab were more likely to undergo laboratory examinations that detected CKD due to more frequent contact with the medical care system than those who did not receive ranibizumab. However, Taiwan NHI program enrolled > 99.99% of residents, removed some barriers, and provided free health care in the low urbanization areas (Cheng and Chiang 1997; Huang et al. 2006). Therefore, surveillance bias may be limited.

Using ICD codes to define comorbidities and incident CKD may have decreased the sensitivity of diagnosis and resulted in inaccurate administrative data. Therefore, a sampling bias may have occurred. In addition, we could not analyze whether dose accumulation is associated with CKD as reducing the dose and frequency of intravitreal VEGF inhibitors is necessary for treating high-risk patients (Hanna et al. 2019). Finally, we could not determine whether any distinct disease in ranibizumab users increased the risk of CKD.

The main strength of this study is the use of a large community-based cohort with a 10-year follow-up period. We examined the relationship between drug exposure and long-term patient outcomes in a real-world setting. The robustness of our findings was supported by consistent results after adjusting for various medications and comorbidities. Our study is the first to provide evidence for the adverse effects of intravitreal VEGF inhibitors, including low-potency ranibizumab, and their association with the incidence of CKD (Phadke et al. 2021). As it has the highest incidence and prevalence of end-stage renal disease worldwide (Wen et al. 2008), this association should be further validated to aid the public health in Taiwan.

In summary, our findings revealed that exposure to intravitreal ranibizumab is an independent risk factor for CKD. Therefore, physicians and ophthalmologists should make the patients aware of such a correlation to increase patient safety and decrease the CKD burden. However, the specific causal relationships and underlying mechanisms require further investigation.