Budget Impact of Faricimab in Neovascular Age-Related Macular Degeneration in the Netherlands: A Systematic Review and Meta-Analysis of Injection Count

Abstract

Introduction

Frequent anti-vascular endothelial growth factor (anti-VEGF) injections for the treatment of neovascular age-related macular degeneration (nAMD) burden patients and healthcare systems. Faricimab may reduce this burden, but robust data are lacking. This study aimed to systematically quantify the injection frequency reduction with faricimab compared to anti-VEGF agents and estimate Dutch budget impact.

Methods

A systematic review of studies on patients with nAMD switching to faricimab was conducted in PubMed. A hybrid approach using artificial intelligence (NotebookLM) and manual verification was employed for data extraction and risk of bias assessment. A random-effects meta-analysis determined the pooled mean difference in annual injections. A budget impact analysis estimated direct medical costs (drug and administration costs) over a 1-year time horizon using Dutch data.

Results

A meta-analysis of 19 real-world studies (2231 patients) was conducted. Patients switched to faricimab for persistent fluid or to extend treatment intervals, resulting in a significant mean reduction of 2.65 injections in the first year (from 9.70 to 7.05; 95% confidence interval − 3.36 to − 1.93). The base-case analysis projected annual savings of approximately €79 million, corresponding to 96,235 fewer injections nationwide. Scenario analyses showed that substantial savings (€16 to €75 million) can be achieved when using faricimab in second- and third-line settings, although replacing first-line bevacizumab would increase costs.

Conclusions

Switching patients to faricimab reduced the injection frequency by two to three injections in the first year. Although evidence certainty was limited by statistical heterogeneity, the reduction was consistent across studies. Although replacing first-line bevacizumab increases costs, substantial savings are achievable in later lines. Strategic positioning of faricimab in the second-line yields significantly higher savings compared to third-line use, and could significantly lower the clinical, patient, and economic burden of nAMD care in the Netherlands. These findings provide quantified, real-world evidence to inform Dutch clinical practice and healthcare policy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40123-025-01301-0.

Key Summary Points

Why carry out this study?

Robust real-world data on faricimab injection frequency and its economic implications in the Netherlands were lacking, despite its potential to reduce the burden of neovascular age-related macular degeneration (nAMD) care.

This study aimed to systematically quantify the reduction in injection frequency with faricimab compared to anti-vascular endothelial growth factor agents for nAMD and to translate these differences into economic implications for the Dutch healthcare system.

What was learned from the study?

Switching patients with nAMD to faricimab reduced the injection frequency by two to three injections in the first year. Although replacing off-label first-line bevacizumab increases costs, substantial savings (approx. €16 to €75 million) and reduced treatment burden are achievable when using faricimab in second- and third-line settings.

Strategic positioning of faricimab maximizes economic benefits. Utilizing faricimab in the second-line setting yields significantly higher annual savings (up to €75 million) compared to restricting its use to the third-line setting (approx. €16 million).

Introduction

Age-related macular degeneration (AMD) is a primary cause of irreversible vision impairment globally, particularly affecting older populations in developed nations and considerably impacting quality of life. This progressive disease damages the macula, the central portion of the retina, leading to a loss of the sharp, detailed vision required for reading and recognizing faces. Although most cases are of a slowly progressing “dry” form, most vision loss is caused by neovascular AMD (nAMD), a subtype characterized by the growth of abnormal blood vessels under the macula [1]. Globally, AMD affected an estimated 196 to 200 million people in 2020, which is projected to rise to nearly 288 to 300 million by 2040 due to aging populations [2]. Although age-standardized incidence rates may be stabilizing or slightly decreasing, the absolute number of affected individuals and associated disability continues to grow. The burden increases considerably with age and is higher in women and certain ethnicities, and is significantly exacerbated by modifiable risk factors like smoking [3].

In the Netherlands, AMD is also a leading cause of vision loss among the elderly [4]. Over 146,000 prevalent cases were estimated in 2021, with approximately 13,700 new cases annually [5]. Dutch data from the Rotterdam Study, included in a large-scale European meta-analysis, confirm the strong age dependency of the disease. That analysis specifically observed a decreasing prevalence of late-stage AMD (i.e., the most advanced, vision-threatening stage of the disease) after 2006, a trend the authors attribute to healthier lifestyles, for example [6]. However, the aging population means the absolute number of Dutch patients will increase in the upcoming years, similar to projections for the whole of Europe where 3.9 to 4.8 million people with late-stage AMD are expected by 2040 [6].

Managing AMD places a considerable strain on the Dutch healthcare system, contributing to long ophthalmology waiting lists (average 12.2 weeks) and significant eye care costs, estimated at approx. €75 million annually for drug costs alone [7, 8]. The required frequent anti-vascular endothelial growth factor (VEGF) injections (therapies that inhibit abnormal blood vessel growth in the eye) and monitoring impose a substantial burden on patients, caregivers, as well as clinical resources [5]. Healthcare costs in the Netherlands are heavily concentrated among older adults and those with chronic conditions needing ongoing management, making (cost-)effective treatment pathways for conditions such as AMD crucial for system-level sustainability.

Intravitreal anti-VEGF treatment is the standard of care for nAMD [9–11]. Available agents towards management of nAMD in the Netherlands include off-label bevacizumab, aflibercept (2 mg or 8 mg dose), faricimab, ranibizumab, and brolucizumab [5]. Dutch national guidelines recommend initiating nAMD treatment with off-label bevacizumab, primarily for cost reasons. For non-responders to first-line treatment with bevacizumab, subsequent options include aflibercept 2 mg or ranibizumab as second-line treatment, followed by aflibercept 2 mg or 8 mg, ranibizumab, or faricimab as options (no preference) as third-line, and brolucizumab as an option in fourth-line treatment [5, 10, 11].

Treatment typically follows proactive treat-and-extend (TAE) regimens [5, 10, 11]. However, maintaining vision often still requires frequent monitoring and injections, leading to high treatment burden and potential adherence issues [5, 10–12]. Results from Dutch Real-World Data (RWD) indicate a higher injection count when using bevacizumab as a first-line treatment, compared to other high-income countries where other anti-VEGF agents such as aflibercept or ranibizumab are used as first-line treatments (e.g., 8.5 injections per year vs. 5.8 per year) [13]. This suggests that the current treatment pathway, while aiming to reduce costs, incurs a considerable burden to the healthcare system (e.g., long waiting lists), patients, and caregivers [13–15].

Faricimab is a novel bispecific antibody simultaneously inhibiting VEGF-A and Angiopoietin-2 (Ang-2), aiming for greater vascular stability and durability [16, 17]. Phase III trials (TENAYA/LUCERNE) showed that faricimab achieved non-inferior best-corrected visual acuity (BCVA) gains compared to aflibercept (2 mg) every 8 weeks (Q8W), with the potential for extended dosing up to every 16 weeks (Q16W) [16]. At 2 years, approximately 60% to 67% of faricimab patients achieved Q16W dosing, and approximately 74% to 81% achieved a treatment interval of longer than 12 weeks (≥ Q12W) [16]. Systematic reviews and meta-analyses confirm non-inferior visual outcomes with significantly fewer injections for faricimab versus comparators [18].

Faricimab’s potential for extended dosing offers the prospect to reduce the high treatment burden in treatment of Dutch patients with nAMD. Emerging real-world evidence supports these findings, showing durability and effectiveness in both naïve and previously treated patients, including those refractory to prior anti-VEGFs [19]. Although initial economic evaluations performed for the Dutch context have provided valuable insights into the direct medical costs of faricimab alongside other anti-VEGFs, these analyses often utilize injection counts for newer agents such as faricimab that are primarily derived from their pivotal clinical trial data [5, 16, 20]. This is a notable limitation, as the highly controlled nature of clinical trials, with strict protocols and homogenous patient populations, may not accurately reflect the more variable treatment patterns and outcomes observed in routine clinical practice. Evaluating the use of treatments in the real-world setting can provide valuable additional information which supplements results from pivotal trials. Although some studies have begun systematically quantifying injection count based on clinical trial data [18], a comprehensive understanding of the actual number of injections needed in real-world settings based on a broader range of evidence remains lacking. Evidently, clinical trials are essential for establishing initial efficacy and safety and remain the gold standard but may not fully reflect the treatment patterns, patient heterogeneity, or long-term outcomes observed in daily clinical practice [21]. Other studies explored real-world data [19, 22], but did not focus on injection count specifically. Thus, specific information on how many injections are needed with faricimab compared to other anti-VEGF agents for nAMD, based on real-world data, is needed. Additionally, the specific cost savings for the Dutch healthcare system, as a result of a decrease in the number of injections, have not yet been determined.

Therefore, the aim of this study was first to systematically quantify the potential reduction in injection count of faricimab compared to other anti-VEGF agents for nAMD by systematically reviewing findings from published, peer-reviewed observational and other real-world studies. Second, this research aimed to translate the differences in injection count into the budget impact for the Dutch healthcare context. For this, Dutch costs for medication and procedures were applied to the synthesized data.

Methods

This systematic review, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, examined studies that assessed faricimab treatment intervals and injection frequency in patients with nAMD [23]. A literature search was performed in PubMed (MEDLINE fully indexed, in process and ahead of print). No limits were applied to make sure all studies evaluating faricimab were included in the search. This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Search Strategy

The PubMed database was searched on May 22, 2025 with terms related to the disease area (“macular degeneration”) and the drug (e.g., “faricimab” and “Vabysmo”) in title, abstract, and MeSH headings or keywords. The full PubMed search is available in the Supplementary Material. The electronic search was supplemented by searching reference lists of relevant review articles and of the retrieved full texts. During the search, a search log was kept consisting of keywords used, searched databases, and search results. Titles and abstracts of the retrieved studies were stored in an electronic database using ReadCube®.

Study Screening and Selection

Three reviewers (ME, CJvdL, JPFdG) independently and manually assessed titles and abstracts to determine study eligibility. Studies were included if they involved adult patients with nAMD treated with intravitreal faricimab. Eligible studies compared faricimab to any other anti-VEGF agent (standard of care) or involved patients who switched to faricimab from prior anti-VEGF treatment. Studies also needed to report the mean or median number of injections or treatment intervals at specific time points. Both randomized controlled trials and various types of observational or real-world studies were eligible. Only full papers published in English were included. Papers not clearly reporting mean or median injections or treatment intervals were excluded. In addition, papers that did not include a comparator group or previous treatment with anti-VEGF were excluded. Although reviews and cost-effectiveness or costing studies were excluded, they were flagged for reference checking and for comparison with the cost calculation presented in this study. Table 1 describes the population, intervention, comparator, and outcome (PICO) criteria for study inclusion.

Table 1.

Population, intervention, comparator, and outcome (PICO) criteria for study inclusion

Criteria	Inclusion criteria
Population (P)	Adult patients diagnosed with neovascular age-related macular degeneration (nAMD)
Intervention (I)	Intravitreal injection of faricimab (6.0 mg)
Comparator (C)	All other anti-vascular endothelial growth factors used in clinical practice for treatment of nAMD
Outcomes (O)	Mean or median treatment interval achieved at specific time points Mean or median number of injections administered over a specific time period
Study design (S)	Randomized controlled trials (RCTs) and observational studies

Full texts were retrieved when studies fulfilled the inclusion criteria or if uncertainty remained about the inclusion of a specific study. All full texts were read and checked for eligibility by three independent reviewers (ME, CJvdL, JPFdG). To resolve disagreement between the reviewers, a consensus procedure was used. A fourth reviewer was consulted when disagreements persisted (JGFV).

Data Extraction

For data extraction from selected studies, a hybrid approach involving an AI-powered research assistant (NotebookLM) and manual human verification was employed to improve efficiency and ensure accuracy.

NotebookLM is an AI-powered research assistant developed by Google [24]. It is a personal workspace that allows one to upload source documents such as research articles, interview transcripts, or reports. The tool utilizes a large language model to analyze the provided content, allowing users to summarize documents, interact with the data, and synthesize information across multiple sources. In contrast with general-purpose chatbots, such as ChatGPT or Gemini, NotebookLM exclusively works with the user-provided source library, providing in-line citations that link directly back to the original text within the documents. Compared to other general-purpose chatbots, this reduces the possibility of hallucinations [25, 26]. Although AI-powered tools offer great improvements over traditional ways of performing reviews, human oversight and verification of AI-generated output remain necessary [27].

The retrieved full-texts of the selected studies were uploaded as source documents to NotebookLM. In addition, a comprehensive structured prompt template was uploaded as a source document in which AI’s role was defined. NotebookLM was instructed to act as a systematic review expert and research assistant for this specific study. The prompt detailed a series of tasks to be conducted on each selected study. The primary task was to perform the data extraction. Other tasks mainly included internal quality and audit checks, where NotebookLM was instructed to review its own work to mitigate error.

The following data were extracted: authors, publication year, journal, geographical area, setting, study design, population, sample size, treatment status, treatment regimen, intervention, loading phase, prior anti-VEGF treatments, indication, age, gender, follow-up duration, treatment interval and injection count for faricimab, and treatment interval and injection count for anti-VEGF.

The data extraction table, as generated by NotebookLM, including the quality and audit check, was considered a preliminary draft table. Subsequently, one reviewer (ME) independently verified each data point by comparing the AI-generated extraction table directly against the source document. Any discrepancies were flagged and discussed among all reviewers (ME, CJvdL, JPFdG, JGFV) until consensus was reached, ensuring the final data extraction table was fully correct and human-verified.

Risk of Bias Assessment of Included Studies

For the risk of bias assessment of the selected studies, a similar hybrid approach involving NotebookLM and manual human verification was employed. For assessing the risk of bias of included studies the Joanna Briggs Institute (JBI) Critical Appraisal Tool for Quasi-Experimental Studies was used [28]. This is a checklist consisting of nine items including temporal precedence, selection and allocation, confounding factors, administration of intervention, assessment, detection, and measurement of the outcome, participant retention and statistical analysis.

Studies were uploaded, one by one, to NotebookLM. This was done to avoid contamination between studies. The primary task was to score each item and to extract the exact piece of text (verbatim) that justified this score. If a study appropriately adhered to an item or sufficiently described how it handled this item it was scored a “yes”. Other tasks included internal quality and audit checks, where NotebookLM was instructed to review its own work to mitigate error, similar to the approach used for data extraction.

Subsequently, one reviewer (ME) independently verified each score on each item using the exact verbatim piece of text that was extracted by NotebookLM as justification for the score. These scores were then summarized within a risk of bias assessment table and an overall score was calculated. Studies adhering to more than half of the items (5 out of 9) were included in the meta-analysis. Any discrepancies were flagged and discussed among all reviewers (ME, CJvdL, JPFdG, JGFV) until consensus was reached, ensuring the final risk of bias assessment table was fully correct and human-verified.

Data Synthesis and Statistical Analysis

R version 4.3.2 and RStudio were used to perform statistical analysis using the packages “mice” and “metafor” [29, 30]. Based on the data extraction table, reported injection numbers were standardized to a uniform metric of mean number of injections in the first year of initiating faricimab. The mean was selected as the primary metric to enable the calculation of total aggregate resource use required for the budget impact analysis and to maximize data inclusion using validated transformation methods [31]. For both faricimab and anti-VEGF treatments, the mean number of injections in the first year of initiating faricimab was calculated for each study using a prioritized approach. The methods for calculation of annualized mean injections were applied in the following order of priority:

1. From mean treatment interval: This method was prioritized as most studies reported treatment intervals which aids in comparability between studies with varying follow-up durations. When the mean interval between injections was provided, the annualized mean was calculated as (e.g., if the treatment interval was reported in weeks, the numerator was 52 weeks):

   $$Meaninjectionsperyear=\frac{Timeunitsinayear}{Meantreatmentinterval}$$ 1

2. From explicitly reported annual mean injections: If a study directly reported the mean number of injections over a 1-year period, this value was used without modification. This method had a lower priority as many studies did not report a mean number of injections over a 1-year period.

3. From mean total injection count: If treatment interval data was not reported, but a mean total number of injections over a specifically stated follow-up period was reported, the annualized mean was calculated as:

   $$Meaninjectionsperyear=\frac{Meantotalinjections}{Follow\text{-}upinyears}$$ 2

This method had a lower priority as many studies did not explicitly state a specific period over which injections were administered, making this calculation less frequently possible.

In cases where only median treatment intervals or injections were reported, these medians were converted to an estimated mean using one of the following methods (depending on the available data):

1. From median (m), minimum (a), and maximum (b), based on the method of Hozo et al. [32]:

   $$Estimatedmeaninjectionsperyear\approx \frac{(a+2m+b)}{4}$$ 3
2. From median and interquartile range, based on the method of Wan et al. [33]:

   $$Estimatedmeaninjectionsperyear\approx \frac{(Q_1+m+Q_2)}{3}$$ 4
3. From median only [32–34]:

   $$Estimatedmeaninjectionsperyear\approx Medianinjectionsperyear$$ 5

The specific calculation method was reported, including which priority, formula, and values were used. Furthermore, to be able to perform a subsequent meta-analysis, standard deviations (SDs) also needed to be scaled to a 1-year period [31]. The following methods were used:

1. From directly reported SD: when a direct SD was available, error propagation formulas were applied to standardize the variance to a 1-year period [31]. The annualized SD was calculated based on scaling time or the Delta method (first-order Taylor series approximation for error propagation) [35].

For data scaled by time:

$$NewSD=ReportedSD\ast \sqrt{Timeratio}$$ 6

For data derived from division:

$$ApproximatedSD\approx \frac{Timeunits}{{\mathit{Mean}}^2}\ast ReportedSD$$ 7

• 2.
  From reported interquartile range (IQR): if a direct SD was not reported, but an IQR was available, the SD was estimated assuming a normal distribution in line with recommendations described in the Cochrane Handbook for Systematic Reviews of Interventions [31]. The annualized SD was calculated as:

  $$EstimatedSD\approx \frac{\mathit{IQR}}{1.35}$$ 8

This estimated SD was then used in the appropriate error propagation formula (formulas 6 and 7).

If insufficient data was available to be able to estimate SDs, the value was recorded as “not calculable” and considered to be missing. A table was developed reporting the annualized means and SDs for faricimab and anti-VEGF, including which methods were prioritized and used in the estimation of the means and SDs as well the calculations themselves.

Before performing the meta-analysis, multiple imputation using predictive mean matching was used to impute the missing SDs [29, 36]. Missing SDs were assumed to be missing at random [37]. The number of imputed datasets was increased until loss of efficiency due to missing information was less than 5%, resulting in five imputed datasets [38–40]. The imputation model incorporated data that was available in the table with the calculated means and SDs, including sample size, mean injections, and treatment group—to produce plausible estimates for the missing SDs. Results were pooled using Rubin’s rules [37]. Rather than conducting a complete-case analysis, and discarding studies from the meta-analysis, this approach allows more evidence to be incorporated into the meta-analysis. In addition, multiple imputation results in grand means that are less biased than those obtained from complete-case analysis [41].

A random-effect meta-analysis was conducted to estimate the pooled mean difference (MD) in the number of injections in the first year of using faricimab vs. other anti-VEGF treatments [23, 30, 31]. Additionally, two separate one-group meta-analyses were performed to estimate the pooled mean number of injections for each treatment individually (faricimab vs. other anti-VEGF). A random-effect model was used to account for heterogeneity across studies, including variations in populations, treatment protocols, and follow-up durations. For all models, the restricted maximum-likelihood (REML) estimator was used [30]. Results from the five imputed datasets were pooled into a single overall estimate and 95% confidence intervals using Rubin’s rules to account for imputation uncertainty [37]. Statistical heterogeneity was quantified using the chi-squared Q test and the I² statistic [42]. To illustrate the impact of heterogeneity and quantify the range of expected effects in a future study, a 95% prediction interval was also calculated from the pooled results. Results were visualized using forest plots, and potential publication bias was assessed by visual inspection of funnel plots and formally with Egger’s regression test. A P value of < 0.05 was considered statistically significant.

Ultimately, the overall certainty of the evidence for the primary outcome (mean injections) was evaluated using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) framework [43]. Using this framework, we systematically assessed the body of evidence considering five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias.

To assess the robustness of the primary meta-analysis, three sensitivity analyses were conducted. First, to evaluate the impact of the multiple imputation procedure, a complete-case meta-analysis was performed which means that studies with the imputed SDs were excluded [41]. Second, two subgroup analyses were conducted to investigate the impact of prior off-label bevacizumab use, as it is known to require a higher injection frequency than other anti-VEGFs [13]. One analysis excluded studies where patients had received prior off-label bevacizumab, and the other analysis exclusively included studies that reported prior off-label bevacizumab use.

Budget Impact Analysis

To translate the findings from the meta-analysis into potential economic implications for the Dutch healthcare system, a cost analysis, in the form of a budget impact analysis, was performed. This was performed from a direct medical cost perspective, including medication and administration costs, over a 1-year time horizon. Costs were expressed for the year 2025.

Base Case

The base case provides a direct translation of the meta-analysis results. Within this base case, the costs per patient for a faricimab injection were based on the public list price (€730 per injection) [44]. The costs per patient for an injection with any other anti-VEGF were calculated using a weighted average. This average was derived from the treatment mix based on the included studies that reported the number or proportions of prior anti-VEGF agents used: aflibercept, ranibizumab, bevacizumab, and brolucizumab. Of note, the number of studies reporting this data varied for each specific anti-VEGF agent. The average proportional use of each anti-VEGF agent was subsequently normalized to sum to a 100%. This final treatment mix was used to estimate the weighted average costs per injection by multiplying the proportions by their respective public list prices (Z-index, July 2025) [44]. See Table 2.

Table 2.

Proportional use of anti-VEGF and weighted average cost per anti-VEGF injection

Anti-VEGF	Average proportional use of anti-VEGF	Normalized percentage	Public list price (Z-index) [44]	Costs
Aflibercept	80.99%a	70.93%	€723.72	€513
Ranibizumab	16.38%b	14.35%	€479.48	€69
Bevacizumab	12.31%c	10.78%	€35.00	€4
Brolucizumab	4.49%d	3.93%	€686.77	€27
Total	114.17%	100%	€613

VEGF vascular endothelial growth factor

^aAverage based on 12 included studies

^bAverage based on 10 included studies

^cAverage based on 4 included studies

^dAverage based on 7 included studies

For both faricimab and anti-VEGF injections, total yearly costs were calculated by multiplying the pooled mean number of yearly injections, as reported in the results section, by the respective costs per injection per patient, plus the pooled mean number of injections multiplied by the standard administration tariff, as determined by the Dutch Health Authority (€520 per administration) [45]. This is expressed as:

$$C_{annual\_patient}=N_{\mathit{inj}}\ast (C_{\mathit{agent}}+C_{\mathit{admin}})$$ 9

where:

• $C_{annual\_patient}$ = Total yearly costs per patient on a specific regimen.

• $N_{\mathit{inj}}$ = The pooled mean number of injections per year for faricimab or anti-VEGF (see results)

• $C_{\mathit{agent}}$ = The costs per injection, either a single price or the calculated weighted average for a treatment mix.

• $C_{\mathit{admin}}$ = The standard administration tariff per injection. ​

The potential population-level implications were estimated by applying the yearly per patient costs for faricimab or anti-VEGF agents to an estimated Dutch nAMD population of 36,315 patients in 2025. This number was based on the number of patients (30,000) that were treated for nAMD in the Netherlands, as reported by the Dutch National Health Care Institute in 2019 [46]. A population growth factor (21%) for individuals aged 75 years and older from 2019 to 2024, based on data from Statistics Netherlands (CBS), was used to extrapolate the number of patients that were treated for nAMD [47]. The population costs were expressed as:

$$C_{pop\_total}=C_{annual\_patient}\ast P_{\mathit{size}}$$ 10

where:

• $C_{pop\_total}$ = Total annual cost for the population on a specific regimen.

• $C_{annual\_patient}$ = The calculated yearly costs per patient on a specific regimen (formula 9).

• $P_{\mathit{size}}$ = The size of the relevant patient population.

To estimate the difference between the two regimens at a population level, the budget impact in the first year after switching to faricimab was calculated. This was calculated by subtracting the total yearly cost per patient of the new treatment regimen from the prior treatment regimen and multiplying this result by the relevant patient population size. This is expressed as:

$$\mathrm{\Delta }{C}_{\mathit{pop}}{=(C}_{annual\_patient,prior}-C_{annual\_patient,new})\ast P_{\mathit{size}}$$ 11

where:

• Δ $C_{\mathit{pop}}$= The annual population budget impact (cost difference).

• $C_{annual\_patient,prior}$​ = The calculated total yearly cost per patient for the prior treatment regimen.

• $C_{annual\_patient,new}$​ = The calculated total yearly cost per patient for the new treatment regimen.

• $P_{\mathit{size}}$ = The size of the relevant patient population.

To increase transparency, the total budget impact was divided into its two main parts: yearly drug costs and yearly administration costs, which were reported separately next to the total budget impact.

Scenario Analyses

Although the base case directly reflects the synthesized evidence, the used treatment mix from the included (international) studies does not fully align with the stepwise Dutch treatment guidelines that recommend specific agents for first-, second-, third-, and fourth-line therapy [11]. Therefore, to better inform Dutch decision-makers, several scenario analyses were performed to estimate the population budget impact within these specific treatment lines. For each of these treatment pathway specific scenarios, the costs per patient for an anti-VEGF injection were recalculated to reflect the different treatment mix of anti-VEGF agents used in that particular treatment line, while the population size was also adjusted accordingly. These scenarios, however, operate under the key assumption that the pooled mean difference in yearly injections is generalizable across the different anti-VEGF agents used in each specific treatment line within the Dutch treatment pathway. The scenarios are described in Table 3.

Table 3.

The explored scenarios for estimating the budget impact of using faricimab within the Dutch context

Scenario	Prior treatment regimen	Costs per injection*	New treatment regimen	Costs per injection*	Number of patients	Description
First-line therapy	100% bevacizumab	€35	100% faricimab	€730	36,315	This scenario compares the current first-line therapy, off-label bevacizumab, against a future state where faricimab is assumed to fully replace bevacizumab for all first-line patients Patient population: the analysis is applied to the total estimated patient population with nAMD of 36,315
Second-line therapy	85% aflibercept and 15% ranibizumab	€687	100% faricimab	€730	21,426	This scenario compares the current second-line treatment mix, consisting of aflibercept (85%) and ranibizumab (15%), against a future state where faricimab is assumed to fully replace both agents Patient population: the analysis is applied to an estimated second-line patient population of 21,426. This figure is derived by assuming that 59% of the total patient population with nAMD progresses to second-line treatment [48]
Second-line therapy	85% aflibercept and 15% ranibizumab	€687	33.3% aflibercept, 33.3% ranibizumab, and 33.3% faricimab	€644	21,426	This scenario compares the current second-line treatment mix of aflibercept (85%) and ranibizumab (15%) against a future competitive market where aflibercept, ranibizumab, and faricimab each hold an equal (33.3%) share Patient population: the analysis is applied to an estimated second-line patient population of 21,426. This figure is derived by assuming that 59% of the total patient population with nAMD progresses to second-line treatment [48]
Third-line therapy	33.3% aflibercept, 33.3% ranibizumab, and 33.3% faricimab	€644	100% faricimab	€730	6321	This scenario compares the current third-line treatment mix, modeled as an equal (33.3%) market share between aflibercept, ranibizumab, and faricimab, against a future state where faricimab is assumed to be the exclusive therapy Patient population: the analysis is applied to an estimated third-line patient population of 6321. This figure is derived by applying a progression rate of 29.5%—assumed to be half of the 59% progression rate from first- to second-line—to the second-line patient cohort of 21,426

*When a treatment mix was assumed, a similar approach as presented in Table 2 for estimating the weighted average costs per injection was used

nAMD neovascular age-related macular degeneration

Sensitivity Analysis

To address the limitation that the base-case analysis relies on public list prices, a sensitivity analysis was performed to account for the uncertainty of confidential rebates. Specifically, a two-way threshold analysis was conducted to identify the conditions of cost indifference between faricimab and prior anti-VEGF therapies.

A break-even plot was created, with the possible percentage rebate for faricimab on the x-axis and the possible rebate for anti-VEGF (treatment mix) on the y-axis. These rebates represent potential confidential price reductions from the public list price, which are often negotiated between manufacturers and healthcare payers such as hospitals, health insurers, or the Ministry of Health. A linear equation was derived to calculate the exact discount on anti-VEGF needed to equal the total costs of faricimab at any rebate level. This threshold analysis was performed for the base case and then repeated for each of the treatment lines scenarios (Table 3) to understand how the results vary depending on the specific comparator.

The analysis was designed as a practical tool for stakeholders. Since actual drug costs are subject to confidential rebates, the break-even plots allow decision-makers to input their own negotiated prices and assess the financial implications under real-world pricing conditions. The plots provide an easy-to-interpret guide: rebate combinations falling below the break-even line represent a cost advantage for faricimab, while combinations above the line represent a cost advantage for the prior anti-VEGF therapies.

Results

The electronic search identified 226 potentially eligible studies. All of these 226 studies were screened on title and abstract. A total of 175 studies were excluded based on title and abstract, of which two studies could not be retrieved. Eventually, full texts of 51 studies were screened. Of these full texts, 32 studies were excluded because they did not report sufficient data to annualize injection count (n = 13), there was no comparison with prior anti-VEGF (n = 10), it considered only the loading phase of faricimab (n = 5), reported only prior anti-VEGF and no switch to faricimab (n = 3), and having an insufficient sample size (n = 1). Ultimately, 19 studies did meet the inclusion and exclusion criteria and were included in the review [49–67]. See Fig. 1 for the PRISMA flowchart.

Fig. 1.

PRISMA flowchart. Anti-VEGF anti-vascular endothelial growth factor

Study Characteristics

The 19 included studies were published between 2023 and 2025 (Table 4). All together, these studies represented 2231 patients with nAMD who were switched to faricimab after being on a prior anti-VEGF therapy. The primary reasons for switching to faricimab reported in these studies included a suboptimal response to previous treatments, often indicated by persistent fluid, and the goal of reducing treatment burden by extending the dosing interval. The majority of the studies were retrospective observational studies (n = 18), while one study was prospective in design (n = 1) [53].

Table 4.

Data extraction table

Authors	Year	Journal	Geographical area	Setting	Study design	Sample size	Treatment status	Treatment protocol	Reason for switch	Number of eyes
Aljundi et al.	2024	Pharmaceutics	Germany	Single-center	Observational (retrospective)	33 eyes of 33 patients	Previously treated with other intravitreal anti-VEGF injections	Four IVF injections (6 mg/0.05 mL) as upload phase; thereafter, treatment interval extended to 8 or 12 weeks if disease activity not recorded	Patients were switched to faricimab because their nAMD was refractory, meaning they had persistent fluid (IRF, SRF, and/or PED) despite receiving treatment with at least two previous anti-VEGF agents	33
Ambati et al.	2025	Retina	USA	Single-center	Observational (retrospective cohort)	263 eyes from 217 patients	Previously treated with other intravitreal anti-VEGF injections	TAE approach; IVF initiated at the same interval (± 1 week) as the previous longest tolerated injection interval on the current agent	The primary clinical reasons for switching to faricimab (IVF) were: persistent or worsening fluid (41.1%), insufficient interval (representing frequent treatment burden) (39.2%), and worsening visual acuity (3.4%). Note that 16.7% had no documented reason	263
Bantounou et al.	2025	BMC Ophthalmology	UK	Single-center	Observational (retrospective case series)	297 eyes from 237 participants (Loaded cohort: 144 eyes, Interval-Matched cohort: 153 eyes)	Previously treated with other intravitreal anti-VEGF injections	Loaded cohort: Minimum of 4 doses at 4-week intervals before extending; Interval-matched cohort: Same interval as previous regimen or longer than standard 4-week loading; All treated with TAE approach	Patients were switched to faricimab because their nAMD showed an inadequate response to previous anti-VEGF therapy. This inadequacy was clinically defined by either persistent disease activity (the presence of any amount of IRF and/or SRF on OCT) or the requirement for frequent injections (four weekly injections) to maintain a dry macula	297
Borchert et al.	2024	Eye	UK	Single-center	Observational (retrospective)	151 eyes from 116 patients (107 nAMD and 44 DMO)	Previously treated with other intravitreal anti-VEGF injections	nAMD: Same interval as previous anti-VEGF without loading	Patients were switched to faricimab because their nAMD or DMO was treatment resistant and had shown a partial response. This meant there was persistence of IRF or SRF (in both nAMD and DMO), or PED or subretinal hyperreflective material (specifically in nAMD), despite receiving prior anti-VEGF treatment on a four to six weekly injection schedule	151, 107 nAMD and 44 DMO
Cancian et al.	2024	Ophthalmololgy and Therapy	Switzerland	Single-center	Observational (prospective cohort)	33 eyes of 33 patients	Previously treated with other intravitreal anti-VEGF injections	Loading dose of four 4-weekly injections, followed by a TAE regimen	Patients were switched to faricimab because of an unsatisfactory treatment response to previous anti-VEGF therapies (aflibercept or ranibizumab), which meant they had a maximum fluid-free interval of 8 weeks or less or exhibited persistent IRF and/or SRF despite fixed 4-weekly injection intervals	33
Goodchild et al.	2024	Eye	UK	Single-center	Observational (retrospective)	98 eyes of 79 patients	Previously treated with other intravitreal anti-VEGF injections	Loading dose of four 4-weekly injections, followed by a TAE regimen	Patients were switched to faricimab because their nAMD was sub-optimally responsive to prior anti-VEGF treatment (aflibercept 2 mg in all eyes), meaning they had disease activity, defined as persistent IRF and/or SRF at a treatment interval of 8 weeks or less, or they maintained disease control only at short intervals of 6 weeks or less (high re-treatment burden)	98
Grimaldi et al.	2025	Ophthalmology Retina	Switzerland	Multicenter	Observational (retrospective)	353 eyes of 325 patients	Previously treated with other intravitreal anti-VEGF injections	TAE protocol for previous anti-VEGF (faricimab protocol not explicitly stated as general rule for whole cohort, but variability including loading dose mentioned)	Patients were switched to faricimab because their nAMD was active (i.e., insufficiently controlled), meaning they had the presence or recurrence of retinal fluid (RF, exudates, or hemorrhage) limiting treatment interval extension before switch, despite receiving continuous prior anti-VEGF therapy	353
Hafner et al.	2025	International Journal of Retina and Vitreous	Germany	Single-center	Observational (retrospective)	46 eyes from 41 patients	Previously treated with other intravitreal anti-VEGF injections	Loading dose of four 4-weekly injections, followed by a TAE regimen	Patients were switched to faricimab because their nAMD was treatment-resistant, meaning they had an insufficient response to ranibizumab or aflibercept characterized by the persistence of IRF or SRF despite 4-weekly anti-VEGF injections, or the inability to extend treatment intervals beyond 6 weeks	46
Hang et al.	2024	Clinical Ophthalmology	USA	Single-center	Observational (retrospective)	88 eyes from 73 patients	Previously treated with other intravitreal anti-VEGF injections	TAE protocol; Interval extended by 1 to 2 weeks at discretion of treating retina physician	Patients were switched to faricimab because of recalcitrant IRF and/or SRF on OCT and/or a desire to extend the treatment interval	88
Hikichi	2023	Japanese Journal of Ophthalmology	Japan	Single-center	Observational (retrospective)	48 eyes of 48 patients	Previously treated with other intravitreal anti-VEGF injections	Same interval as last previous medication after switch; adjusted for 2 weeks based on disease activity; TAE-protocol	Patients were switched to faricimab because they had persistent exudative changes despite regular injections based on a TAE regimen, or their treatment interval failed to extend to 12 weeks despite continued TAE treatment for more than 1 year	48
Janmohamed et al.	2025	Eye	UK	Single-center	Observational (retrospective cohort)	215 eyes of 184 patients	Previously treated with other intravitreal anti-VEGF injections	Loading dose of four 4-weekly injections, followed by a TAE regimen	Patients were switched to faricimab because their nAMD was treatment-experienced with a suboptimal response to existing anti-VEGF therapies, meaning they were unable to extend treatment intervals beyond 8 weeks, with 42.3% of eyes having demonstrated suboptimal responses to more than one anti-VEGF agent	215
Kataoka et al.	2024	Graefe’s Archive for Clinical and Experimental Ophthalmology	Japan	Multicenter	Observational (retrospective)	130 eyes of 124 patients	Previously treated with other intravitreal anti-VEGF injections	Loading dose of four 4-weekly injections, followed by a TAE regimen	Patients were switched to faricimab because their nAMD was refractory, meaning they had persistent fluid (SRF, IRF, and/or sub-RPE fluid) despite the necessity for monthly aflibercept treatment	130
Khodor et al.	2024	Journal of Vitreoretinal Diseases	USA	Two-center	Observational (retrospective chart review)	135 eyes of 119 patients	Previously treated with other intravitreal anti-VEGF injections	Not specified, at the discretion of healthcare provider	Patients were switched to faricimab because their nAMD was refractory or treatment-resistant, meaning they had the persistent presence of intraretinal (IRF or SRF despite receiving treatment with previous anti-VEGF agents	135
Leung et al.	2023	Clinical Ophthalmology	USA	Single-center	Observational (retrospective)	190 eyes in 186 patients	Previously treated with other intravitreal anti-VEGF injections	Often on a TAE protocol	Patients were switched to faricimab because their nAMD was treatment-resistant, meaning the decision was most commonly due to persistent fluid (64%) and/or an inability to extend the treatment interval beyond 4–6 weeks	190
Ng et al.	2024	Life	UK	Single-center	Observational (retrospective)	63 eyes of 54 patients	Previously treated with other intravitreal anti-VEGF injections	Interval extension at discretion of clinicians, usually initially 4 weeks when fluid-free on SD-OCT after loading doses, then 2 or 4 weeks thereafter	Patients were switched to faricimab because their nAMD was treatment-refractory, meaning they had persistent fluid (SRF and/or IRF) on SD-OCT assessment despite receiving multiple previous anti-VEGF treatments	63
Savant et al.	2024	Journal of Ophthalmology	USA	Single-center	Observational (retrospective case series)	71 eyes from 62 patients (65 nAMD, 6 DMO)	Previously treated with other intravitreal anti-VEGF injections	Intravitreal injections administered in a TAE fashion; patients with > 4 weeks interval prior were not reduced to 4 weeks for loading	Patients were switched to faricimab because their nAMD or diabetic macular edema was treatment resistant, meaning they had persistent or worsening IRF or SRF on OCT, or the inability to extend treatment intervals despite receiving monthly intravitreal injections. (The clinical reasons for switching included refractory IRF or SRF (59.2% of eyes), inability to extend the treatment interval (5.6%), or a combination of both (35.2%)	71
Sim et al.	2024	Ophthalmology Retina	UK	Multicenter	Observational (retrospective cohort study)	117 eyes and 117 patients	Previously treated with other intravitreal anti-VEGF injections	Loading phase of 4 monthly intravitreal faricimab followed by TAE approach	Patients were switched to faricimab because their nAMD required intensive treatment, meaning they had a high treatment burden defined as being on a 4-weekly treatment interval with ranibizumab or aflibercept 2 mg in the last 3 visits	117
Sutter et al.	2024	Cureus	USA	Single-center	Observational (retrospective chart analysis)	19 eyes from 17 patients	Previously treated with other intravitreal anti-VEGF injections	Not reported	Patients were switched to faricimab because they were suboptimal responders to anti-VEGF treatment, meaning they had an anti-VEGF injection history of more than 3 months and the presence of fluid (SRF or IRF) after 3 or more anti-VEGF injections	19
Szigiato et al.	2024	Ophthalmology Retina	USA	Single-center	Observational (retrospective cohort study)	126 eyes of 106 patients	Previously treated with other intravitreal anti-VEGF injections	Switching and administration schedule at discretion of clinician	Patients were switched to faricimab because their nAMD exhibited a suboptimal response to prior anti-VEGF therapy, which was defined by the presence of persistent macular fluid (including IRF, SRF, and/or PED), despite receiving at least two anti-VEGF injections in the previous 6 months	126

Authors	Intervention	Loading phase	Prior anti-VEGF agents	Indication	Age	Gender	Follow-up duration	Faricimab treatment interval	Anti-VEGF treatment interval	Faricimab injection count	Anti-VEGF injection count
Aljundi et al.	IVF (6 mg/0.05 mL)	Four IVF injections (loading phase)	Previous total intravitreal injections: 44 ± 21; Previous intravitreal bevacizumab: 12 ± 10; Previous intravitreal ranibizumab: 11 ± 9; Previous intravitreal aflibercept: 21 ± 16	nAMD	82 ± 5 years	Male: 55%, Female: 45%	1 year (52 weeks)	Not reported	Not reported	7.1 ± 2.4 injections (mean ± SD)	8.6 ± 1.2 injections (mean ± SD), in the last year prior to switch
Ambati et al.	IVF (6 mg/0.05 mL)	No loading phase	Most frequent IVI was aflibercept with a median of 20 injections; bevacizumab, ranibizumab, aflibercept, brolucizumab	nAMD	81.8 ± 7.4 years	Male: 31.3%, Female: 68.7%	1 year (52 weeks)	7.6 ± 2.4 weeks (mean ± SD)	5.9 ± 1.8 weeks (mean ± SD)	6.4 ± 2.3 injections (mean ± SD)	39.7 ± 30.5 injections (mean ± SD), over a period of 5.8 ± 3.5 years prior to switch
Bantounou et al.	IVF (6 mg/0.05 mL)	Loaded cohort: Minimum of 4 doses at 4-week intervals; Interval-matched cohort: Not a standard loading phase	Aflibercept was the most frequently prescribed anti-VEGF treatment (n = 275, 92.6%), followed by ranibizumab (n = 21, 7.1%) and brolucizumab (n = 1, 0.3%)	nAMD	80.7 ± 7 years	Male: 43.9%, Female: 56.1%	9.6 ± 4.1 months (mean ± SD)	7.2 ± 2.8 weeks (mean ± SD)	6.0 ± 2.3 weeks (mean ± SD)	7.5 ± 1.9 injections (mean ± SD)	32.5 ± 24.4 injections (mean ± SD), prior to switch (period not reported)
Borchert et al.	IVF (6 mg/0.05 mL)	nAMD: Without a loading course (but 27 eyes (25%) received 4 injections at 4-weekly intervals)	nAMD: Aflibercept (95%), brolucizumab (5%)	nAMD and DMO	nAMD: 79 ± 7 years	nAMD: male 41%, female 59%	6 months	nAMD: 6.9 ± 2.3 weeks (mean ± SD)	nAMD: 5.2 ± 1.7 weeks (mean ± SD)	nAMD: 5 ± 1 injections (mean ± SD)	nAMD: Mean of 26 ± 18 injections (mean ± SD), prior to switch (period not reported)
Cancian et al.	IVF (6 mg/0.05 mL)	Upload phase with four 4-weekly injections	Aflibercept or ranibizumab (not by percentage)	nAMD	Median 82 years (IQR 77, 85)	Male: 42%, Female: 58%	Median 72 weeks (IQR 61, 76) (approx. 1.4 years)	8 weeks (median)	5 weeks (median)	12 injections (median, IQR: 10, 13)	8 injections (median, IQR, 8, 10), prior to switch (period not reported)
Goodchild et al.	IVF (6 mg/0.05 mL)	Upload phase with four 4-weekly injections	Aflibercept 2 mg (100%)	nAMD	81 ± 7.26 years	Male: 47%, Female: 53%	Minimum 6-month follow-up	8 ± 2.42 weeks (mean ± SD)	5 ± 1.51 weeks (mean ± SD)	5.96 injections (calculated by 584 total faricimab injections divided by in total 98 eyes)	8 ± 1.83 injections (mean ± SD), in the last year prior to switch
Grimaldi et al.	IVF (6 mg/0.05 mL)	Not explicitly reported; Variability in treatment protocols adopted across different centers, including loading dose	Ranibizumab, aflibercept 2 mg, bevacizumab, or brolucizumab (percentages not given)	nAMD	Mean 79.9 ± 7.9 years	Male: 43.1%, Female: 56.9%	12 months	Mean 8.3 ± 4.2 weeks (mean ± SD)	5.9 ± 2.2 weeks (mean ± SD)	8.7 ± 2.0 injections (mean ± SD)	8.7 ± 2.8 injections (mean ± SD), in the last year prior to switch
Hafner et al.	IVF (6 mg/0.05 mL)	Upload phase with four 4-weekly injections	Ranibizumab or aflibercept; Mean prior anti-VEGF: Total 30.22 ± 25.51, RBZ 12.26 ± 12.80, AFL 19.30 ± 19.71	nAMD	Mean 74.70 ± 7.04 years	Male: 36.6%, Female: 63.4%	9 months	56 days (median, IQR: 20 days)	35 days (median, IQR: 15 days)	7.43 ± 0.78 injections (mean ± SD)	8.50 ± 3.06 injections (mean ± SD), in the last year prior to switch
Hang et al.	IVF (6 mg/0.05 mL)	No upload phase	Bevacizumab, ranibizumab, aflibercept, brolucizumab; 1 type: 27.3%, 2 types: 53.4%, 3 types: 13.6%, 4 types: 5.7%; Immediate prior: Aflibercept (71.6%), ranibizumab (9.1%), bevacizumab (6.8%), brolucizumab (4.5%), alternating aflibercept/brolucizumab (8%)	nAMD	82 ± 9 years	Male: 34.1%, Female: 65.9%	30.1 ± 13.5 weeks (mean ± SD)	7.4 ± 2.6 weeks (mean ± SD)	6.06 ± 2.0 weeks (mean ± SD)	5.1 ± 2.4 injections (mean ± SD)	27.5 ± 26.6 injections (mean ± SD), over a period of 41.9 ± 39.4 months (mean ± SD) prior to switch
Hikichi	IVF (6 mg/0.05 mL)	No upload phase	Ranibizumab (20.8%), aflibercept (79.2%)	nAMD	81.1 ± 1.1 years	Male: 58%, Female: 42%	6 months	10.45 ± 0.44 weeks (mean ± SD)	6.72 ± 0.34 weeks (mean ± SD)	Not reported	38.7 ± 3.25 injections (mean ± SD), over a period of 64.9 ± 6.0 months (mean ± SD) prior to switch
Janmohamed et al.	IVF (6 mg/0.05 mL)	Upload phase with four 4-weekly injections	Aflibercept (71.2%), ranibizumab biosimilar (19.5%), ranibizumab (7%), brolucizumab (2.3%) as immediate prior; 2 types: 30.2%, 3 types: 12.1%	nAMD	81.36 ± 7.85 years	Male: 44%, Female: 56%	12.19 ± 2.70 months (mean ± SD)	7.49 ± 2.64 weeks (mean ± SD)	4.71 ± 1.38 weeks (mean ± SD)	8.63 ± 2.2 injections (mean ± SD)	18 injections (median, IQR 10–28.5), over a period of 5.02 ± 11.82 years (mean ± SD) prior to switch
Kataoka et al.	IVF (6 mg/0.05 mL)	Upload phase with four 4-weekly injections	Aflibercept (mean 39.1 ± 30.3 injections), ranibizumab (mean 6.5 ± 10.6 injections), brolucizumab (mean 0.8 ± 2.7 injections); All eyes receiving monthly aflibercept (prior to switch)	nAMD	77.8 ± 8.4 years	Male: 66.2%, Female: 33.8%	6 months	8.7 ± 1.7 weeks (mean ± SD)	4.4 ± 0.5 weeks (mean ± SD)	Not reported	43.5 ± 35.1 injections (mean ± SD), prior to switch (period not reported)
Khodor et al.	IVF (6 mg/0.05 mL)	Not reported	Aflibercept 2 mg, investigational higher dose aflibercept 3 mg or 4 mg, bevacizumab, ranibizumab, alternating bevacizumab and aflibercept 2 mg (percentages not provided)	nAMD	79.8 ± 7.3 years	Male: 46.2%, Female: 53.8%	11.6 ± 2 months (mean ± SD)	5.5 ± 2 weeks (mean ± SD)	4.8 ± 1.3 weeks (mean ± SD)	3.3 ± 2.0 injections (mean ± SD)	10.7 ± 2.6 injections (mean ± SD), in the last year prior to switch
Leung et al.	IVF (6 mg/0.05 mL)	Not reported	Ranibizumab or aflibercept (not by percentage)	nAMD	80.1 ± 8.1 years	Male: 59%, Female: 41%	34.88 ± 8.2 weeks (mean ± SD)	7.64 ± 6.2 weeks (mean ± SD)	5.33 weeks (mean, SD not reported)	6.99 ± 2.3 injections (mean ± SD)	34.2 ± 23 injections (mean ± SD), over a period of 182.41 ± 128 weeks (mean ± SD) prior to switch
Ng et al.	IVF (6 mg/0.05 mL)	Not reported	Number of types of previous anti-VEGF injections in each eye: 1.87 ± 0.75 (range 1–3)	nAMD	79.2 ± 7.8 years	Male: 38.8%, Female: 61.1%	6.98 ± 1.75 months (mean ± SD)	5.25 ± 1.99 weeks (mean ± SD)	5.24 ± 1.92 weeks (mean ± SD)	4.81 ± 1.16 injections (mean ± SD)	41.5 ± 22.4 injections (mean ± SD), over a period of 65.4 ± 39.0 months (mean ± SD) prior to switch
Savant et al.	IVF (6 mg/0.05 mL)	No upload phase	Bevacizumab or aflibercept (100%); bevacizumab AND aflibercept (67.6%); brolucizumab (6.2%); steroids (33.3%, DMO only)	nAMD and DMO	82 years (mean, SD not reported)	Male: 36.6%, Female: 63.4%	218.7 ± 81.2 days (mean ± SD)	45.2 ± 16.6 days (mean ± SD)	37.6 ± 10.8 days (mean ± SD)	Not reported	26.1 injections (mean, SD not reported), prior to switch (period not reported)
Sim et al.	IVF (6 mg/0.05 mL)	Upload phase with four 4-weekly injections	Aflibercept 2 mg (87.2%), ranibizumab (12.8%)	nAMD	74.2 ± 7.9 years	Male: 49.6%, Female: 50.4%	1 year (52 weeks ± 4 weeks)	6.9 ± 2.3 weeks (mean ± SD)	4.2 ± 0.3 weeks (mean ± SD)	9.1 ± 1.5 injections (mean ± SD)	10.1 ± 1.6 injections (mean ± SD), in the last year prior to switch
Sutter et al.	IVF (6 mg/0.05 mL)	Not reported	Aflibercept (16 eyes), ranibizumab (1 eye), brolucizumab (1 eye), bevacizumab (1 eye)	nAMD	78 ± 10 years	Male: 58.0%, Female: 42.1%	After 3 and 4 injections	9.3 ± 3.9 weeks (mean ± SD)	7.6 ± 2.8 weeks (mean ± SD)	Not reported	Not reported
Szigiato et al.	IVF (6 mg/0.05 mL)	Not reported	Aflibercept (87%), bevacizumab, ranibizumab, brolucizumab (percentages not provided for others)	nAMD	80.7 ± 8.2 years	Male: 42%, Female 58%	24.3 ± 5.2 weeks	6.8 ± 2.1 weeks (mean ± SD) in table, text report 6.3 weeks (mean, SD not reported)	5.6 ± 1.6 weeks (mean ± SD)	Not reported	Aflibercept 20.0 ± 18.4 injections (mean ± SD), bevacizumab 7.0 ± 8.9 injections (mean ± SD), ranibizumab 1.9 ± 8.5 injections (mean ± SD), brolucizumab 0.3 ± 1.6 injections (mean ± SD), prior to switch (peroiod not reported)

AFL aflibercept, anti-VEGF anti-vascular endothelial growth factor, DMO diabetic macular oedema, IQR interquartile range, IRF intraretinal fluid, IVF intravitreal faricimab, nAMD neovascular age-related macular degeneration, OCT optical coherence tomography, PED pigment epithelial detachment, RBZ ranibizumab, SD standard deviation, SRF subretinal fluid, TAE treat and extend, UK United Kingdom, USA United States of America

Treatment regimens varied across the included studies, especially regarding the use of a faricimab loading phase. Seven studies reported a mandatory loading phase, usually consisting of four monthly injections [49, 53, 54, 56, 59, 60, 65], while four studies explicitly initiated treatment with faricimab without a loading phase [50, 57, 58, 64]. In the other studies the loading phase was mixed or not clearly reported. The follow-up duration was also heterogeneous, ranging from 3 months to more than a year, with 6 and 12 months being the most commonly assessed durations.

The studies were primarily conducted in the USA (n = 7) [50, 57, 61, 62, 64, 66, 67] and the UK (n = 6) [51, 52, 54, 59, 63, 65]. Additionally, there were two studies each from Germany, Japan, and Switzerland. Across the entire patient cohort, the average age was 80.9 years and approximately 59.9% of patients were female. In all included studies, patients had been previously treated with anti-VEGF therapy, where aflibercept was the most commonly reported prior therapy before switching to faricimab (70.93%). Other anti-VEGF therapies included ranibizumab (14.35%), bevacizumab (10.78%), and brolucizumab (3.93%). These percentages are reported in Table 2.

Risk of Bias Assessment

The risk of bias for the 19 included studies was assessed using the 9-item JBI Critical Appraisal Tool (Table 5). The overall quality scores ranged from 6 to 8 on a scale of 0 to 9. All included studies met the pre-specified quality threshold to be included in the meta-analysis.

Table 5.

Risk of bias assessment according to Joanna Briggs Institute (JBI)

Authors	Temporal precedence	Selection and allocation	Confounding factors	Administration of intervention/exposure	Assessment, detection, and measurement of the outcome			Participant retention	Statistical analysis validity	Overall score (0–9)
	Item 1	Item 2	Item 3	Item 4	Item 5	Item 6	Item 7	Item 8	Item 9
Aljundi et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7
Ambati et al. 2025	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Bantounou et al. 2025	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Borchert et al. 2024	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	No	7
Cancian et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7
Goodchild et al. 2024	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	8
Grimaldi et al. 2025	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7
Hafner et al. 2025	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Hang et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Hikichi 2023	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Janmohamed et al. 2025	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Kataoka et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7
Khodor et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Leung et al. 2023	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7
Ng et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Savant et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7
Sim et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Sutter et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	No	Yes	6
Szigiato et al. 2024	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	7

There were several areas that were well reported and can be considered a methodological strength. For example, all studies clearly established temporal precedence (item 1) and confounding factors (the participants included in the comparison between anti-VEGF and faricimab were similar since they are the same patients that switched from anti-VEGF to faricimab; item 3). In addition, studies used reliable methods for assessing the outcome (item 5) and had complete follow-up data (item 6).

However, a number of weaknesses were identified for all 19 included studies, which were primarily related to the non-randomized nature of the studies. None of the studies included a control arm selected in a similar fashion (item 2), nor was the administration of the treatment standardized across participants (item 4). All patients with nAMD received a variety of anti-VEGF therapies before switching to faricimab. Although these weaknesses indicate the potential for selection and performance bias, the overall quality of the included studies was deemed sufficient to be quantitatively synthesized.

Estimation of Mean Injections Faricimab vs. Anti-VEGF

The majority of mean injections were calculated using priority 1—using the reported treatment intervals (Table 6). Three standard deviations were not calculable based on the reported information and were imputed using multiple imputation. On average, patients who switched to faricimab had fewer injections administered compared to when they were treated with anti-VEGF agents.

Table 6.

Calculated mean yearly number of injections for faricimab and anti-VEGF

Study	Faricimab				Anti-VEGF			Other anti-VEGF
	Mean injections per year	Method for calculating mean	Calculated SD	Method for calculating SD	Mean injections per year	Method for calculating mean	Calculated SD	Method for calculating SD
Aljundi et al. 2024	7.10	Priority 3; reported annual mean	2.40	Method 1 (from direct SD)	8.60	Priority 3; reported annual mean	1.20	Method 1 (from direct SD)
Ambati et al. 2025	6.84	Priority 1; from interval: (52/7.6)	2.16	Method 1 (for division): (52/7.62) × 2.4	8.81	Priority 1; from interval: (52/5.9)	2.69	Method 1 (for division): (52/5.92) × 1.8
Bantounou et al. 2025	7.22	Priority 1; from interval: (52/7.2)	2.81	Method 1 (for division): (52/7.22) × 2.8	8.67	Priority 1; from interval:(52/6.0)	3.32	Method 1 (for division): (52/6.02) × 2.3
Borchert et al. 2024	7.54	Priority 1; from interval: (52/6.9)	2.51	Method 1 (for division): (52/6.92) × 2.3	10.00	Priority 1; from interval: (52/5.2)	3.27	Method 1 (for division): (52/5.22) × 1.7
Cancian et al. 2024	6.50	Priority 1; from interval: (52/8) using mean = median	3.17	Not calculable SD imputed using multiple imputation	10.40	Priority 1; from interval: (52/5) using mean = median	3.30	Not calculable SD imputed using multiple imputation
Goodchild et al. 2024	6.50	Priority 1; from interval: (52/8)	1.97	Method 1 (for division): (52/82) × 2.42	10.40	Priority 1; from interval: (52/5)	3.14	Method 1 (for division): (52/52) × 1.51
Grimaldi et al. 2025	6.27	Priority 1; from interval: (52/8.3)	3.17	Method 1 (for division): (52/8.32) × 4.2	8.81	Priority 1; from interval: (52/5.9)	3.29	Method 1 (for division): (52/5.92) × 2.2
Hafner et al. 2025	6.50	Priority 1; from interval: (52/56 days converted to weeks) using mean = median	1.72	Method 2: Est. SD = IQR/1.35 = 20/1.35 = 14.81 days or 2.12 weeks (assuming normality). Propagate error (division): (52/82) × 2.12	10.40	Priority 1; from interval: (52/35 days converted to weeks) using mean = median	3.30	Method 2: Est. SD = IQR/1.35 = 15/1.35 = 11.11 days or 1.59 weeks (assuming normality). Propagate error (division): (52/52) × 1.59
Hang et al. 2024	7.03	Priority 1; from interval: (52/7.4)	2.47	Method 1 (for division): (52/7.42) × 2.6	8.58	Priority 1; from interval:(52/6.06)	2.83	Method 1 (for division): (52/6.062) × 2.0
Hikichi 2023	4.98	Priority 1; from interval: (52/10.45)	0.21	Method 1 (for division): (52/10.452) × 0.44	7.74	Priority 1; from interval: (52/6.72)	0.39	Method 1 (for division): (52/6.722) × 0.34
Janmohamed et al. 2025	6.94	Priority 1; from interval: (52/7.49)	2.45	Method 1 (for division): (52/7.492) × 2.64	11.04	Priority 1; from interval: (52/4.71)	3.24	Method 1 (for division): (52/4.712) × 1.38
Kataoka et al. 2024	5.98	Priority 1; from interval: (52/8.7)	1.17	Method 1 (for division): (52/8.72) × 1.7	11.82	Priority 1; from interval: (52/4.4)	1.34	Method 1 (for division): (52/4.42) × 0.5
Khodor et al. 2024	9.45	Priority 1; from interval:(52/5.5)	3.44	Method 1 (for division): (52/5.52) × 2.0	10.83	Priority 1; from interval: (52/4.8)	2.93	Method 1 (for division): (52/4.82) × 1.3
Leung et al. 2023	6.81	Priority 1; from interval: (52/7.64)	5.52	Method 1 (for division): (52/7.642) × 6.2	9.76	Priority 1; from interval:(52/5.33)	3.29	Not calculable SD imputed using multiple imputation
Ng et al. 2024	9.90	Priority 1; from interval: (52/5.25)	3.75	Method 1 (for division): (52/5.252) × 1.99	9.92	Priority 1; from interval: (52/5.24)	3.64	Method 1 (for division): (52/5.242) × 1.92
Savant et al. 2024	8.05	Priority 1; from interval: (52/45.2 days converted to weeks)	2.96	Method 1 (for division): (52/(45.2/7)2) × (16.6/7)	9.68	Priority 1; from interval: (52/37.6 days converted to weeks)	2.78	Method 1 (for division): (52/(37.6/7)2) × (10.8/7)
Sim et al. 2024	7.54	Priority 1; from interval: (52/6.9)	2.51	Method 1 (for division): (52/6.92) × 2.3	12.38	Priority 1; from interval: (52/4.2)	0.88	Method 1 (for division): (52/4.22) × 0.3
Sutter et al. 2024	5.59	Priority 1; from interval: (52/9.3)	2.35	Method 1 (for division): (52/9.32) × 3.9	6.84	Priority 1; from interval: (52/7.6)	2.52	Method 1 (for division): (52/7.62) × 2.8
Szigiato et al. 2024	7.65	Priority 1; from interval: (52/6.8)	2.36	Method 1 (for division): (52/6.82) × 2.1	9.29	Priority 1; from interval: (52/5.6)	2.65	Method 1 (for division): (52/5.62) × 1.6

Anti-VEGF anti-vascular endothelial growth factor, IQR interquartile range, SD standard deviation

Meta-Analysis

All 19 studies were included in a random-effects meta-analysis to determine the pooled mean difference in the number of injections during the first year of patients with nAMD that switched from any anti-VEGF therapy to faricimab. In Fig. 2, the forest plot shows that switching to faricimab resulted in a statistically significant reduction in the mean number of injections of − 2.65 injections (95% CI − 3.36 to − 1.93), favoring faricimab. This indicates that on average patients required 2–3 fewer injections in the first year after switching to faricimab compared to their prior treatment with anti-VEGF. The overall effect was statistically significant (z = − 7.85, p < 0.0001).

Fig. 2.

Forest plot of the difference in mean injections in the first year after switch from anti-VEGF to faricimab. Anti-VEGF anti-vascular endothelial growth factor, CI confidence interval, SD standard deviation

There was a large degree of heterogeneity observed among the studies, as indicated by the Cochran Q test (Q(18) = 543.65, p < 0.0001) and a high I² statistic of 96.6%. These values suggest that most of the variability in the observed effects is due to true differences between the studies rather than sampling error. The 95% prediction interval ranged from 5.48 and 0.18, suggesting that a future study could potentially show a small or no difference.

Two separate one-group meta-analyses showed a pooled mean of 7.05 injections (95% CI 6.50–7.61) per year for patients treated with faricimab and 9.70 injections (95% CI 9.03–10.36) per year for patient treated with any other anti-VEGF. Visual inspection of funnel plots did not suggest the presence of significant publication bias, and Egger’s regression test was not statistically significant (p = 0.13). See the Supplementary Material.

Subgroup and Sensitivity Analyses

To assess robustness of the primary finding, a subgroup analysis was performed. The subgroup analysis of the eight studies that included patients with prior bevacizumab use showed a statistically significant reduction of − 1.81 injections (95% CI − 2.26 to − 1.37) (Fig. 3). This subgroup demonstrated moderate heterogeneity (I² = 41.6%), indicating more consistency among these studies. In contrast, the subgroup of 11 studies excluding studies with prior bevacizumab use showed a much larger reduction of − 3.33 injections (95% CI − 4.41 to − 2.25), but with high heterogeneity (I² = 97.7%) (Fig. 4).

Fig. 3.

Forest plot of the difference in mean injections in the first year after switch from anti-VEGF to faricimab including studies with prior bevacizumab use. Anti-VEGF anti-vascular endothelial growth factor, CI confidence interval, SD standard deviation

Fig. 4.

Forest plot of the difference in mean injections in the first year after switch from anti-VEGF to faricimab including studies without prior bevacizumab use. Anti-VEGF anti-vascular endothelial growth factor, CI confidence interval, SD standard deviation

A complete-case sensitivity analysis excluding the (two) studies with imputed data was performed to assess the impact of the imputation (Fig. 5). This analysis yielded a pooled mean difference of − 2.56 injections (95% CI − 3.35 to − 1.77), confirming the primary finding was not substantially driven by the imputation procedure. Heterogeneity in this analysis remained high (I² = 97.1%).

Fig. 5.

Forest plot of the difference in mean injections in the first year after switch from anti-VEGF to faricimab using complete-case analysis. Anti-VEGF anti-vascular endothelial growth factor, CI confidence interval, SD standard deviation

Overall Certainty of Evidence

The certainty of evidence for the primary outcome of this review, the mean difference in the number of injections in the first year after switching from a prior anti-VEGF therapy to faricimab, was evaluated using the GRADE framework [43]. All included studies in the meta-analysis, the body of evidence, consisted entirely of observational studies. As a result, the grading started with an initial certainty rating of “low”. On the basis of the following assessment the certainty rating was modified:

• Risk of bias: A detailed risk of bias assessment (see Sect. “Risk of Bias Assessment”) identified a number of methodological weaknesses that observational studies are characterized by. First, none of the studies included a control group. Second, the administration of treatment was not standardized across participants. In addition, there was a large degree of heterogeneity in the reporting of the primary outcome: some studies reported injection counts that included a loading phase, whereas other studies only focused on treatment intervals and did not report injection count. Considering these issues, there is a serious risk of bias, which led to a one-level downgrade.

• Inconsistency: Although all included studies in the meta-analysis showed an effect in the same direction, there was a substantial amount of heterogeneity, with an I² statistic of 96.6% (p < 0.0001). This large degree of heterogeneity indicates that the variability in results is due to a “real” difference between studies, not chance. This is likely because studies differed in how they measured and reported injection count as well as treatment intervals. In addition to these differences, other differences between studies such as varying follow-up periods, varying loading phases, and varying prior treatments with anti-VEGF agents, amongst others, also could have contributed to this large degree of heterogeneity. Given these significant differences and unexplained heterogeneity, the inconsistency was serious, which led to one-level downgrade.

• Indirectness: Based on the PICO in this study, the evidence was judged to be direct. The population (patients with nAMD), intervention (switch to faricimab), comparator (prior anti-VEGF), and outcome (treatment/injection frequency) in the included studies directly align with the primary question of the meta-analysis.

• Imprecision: This domain was carefully judged. The 95% prediction interval [− 5.48 to 0.18] crossed the line of no effect. However, a decision was made not to downgrade this domain. This decision was based on the fact that the interval was highly asymmetrical and skewed towards fewer injections; the upper bound of 0.18 is clinically irrelevant (i.e., approx. 1/5th of an injection) while the 95% confidence interval for the pooled mean effect is precise and entirely on the side of less injections [− 3.36 to − 1.93].

• Publication bias: As previously mentioned, visual inspection of the funnel plots and the formal Egger’s regression test (p = 0.078) did not indicate the presence of significant publication bias.

Starting from a “Low” level of certainty, the evidence was downgraded twice: once for serious risk of bias and once for serious inconsistency. No factors were identified that would suggest upgrading the evidence. Therefore, the final overall quality of the evidence is rated as “Very low” (Table 7). This indicates that there is very little confidence in the effect estimate, i.e., the true effect is likely to be substantially different from the estimated effect.

Table 7.

Overall certainty of evidence

Outcome	Number of studies	Mean difference (95% CI)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty of evidence
Mean injections per year	19	− 2.65 (− 3.36; − 1.93)	Serious	Serious	Not serious	Not serious	Not serious	⊕ ◯◯◯ Very low

Budget Impact Analysis

Base-Case and Scenario Analyses

The budget impact analysis suggests that switching from a prior anti-VEGF to faricimab could lead to substantial cost savings for the Dutch healthcare context (Table 8). In the base case, which reflects the treatment mix and injection count derived directly from the meta-analysis, switching to faricimab is associated with a cost saving of approximately €79 million. This is mainly driven by a reduction in the total yearly costs per patient from €10,989 to €8813, primarily due to the lower number of required injections and administrations.

Table 8.

Budget impact

	Drug costs pre-switch	Drug costs post-switch	Administration cost pre-switch	Administration costs post-switch	Total costs pre-switch	Total costs post-switch	Total population costs pre-switch	Total population costs post-switch	Drug costs budget impact	Administration costs budget impact	Total budget impact
Base case	€5945	€5147	€5044	€3666	€10,989	€8813	€399,071,326	€320,025,938	− €29,003,318	− €50,042,070	− €79,045,388
First-line scenario	€340	€5147	€5044	€3666	€5384	€8813	€195,501,803	€320,025,938	€174,566,205	− €50,042,070	€124,524,135
Second-line scenario	€6665	€5147	€5044	€3666	€11,709	€8813	€250,869,167	€188,815,303	− €32,529,043	− €29,524,821	− €62,053,864
Second-line scenario with equal share between drugs (33%)	€6665	€4538	€5044	€3666	€11,709	€8204	€250,869,167	€175,787,893	− €45,556,453	− €29,524,821	− €75,081,274
Third-line scenario	€6244	€5147	€5044	€3666	€11,288	€8813	€71,349,937	€55,700,514	− €6,939,600	− €8,709,822	− €15,649,423

*All costs are reported on a yearly basis

In the scenario analysis, where the specific lines of therapy according to the Dutch nAMD treatment guidelines were modelled, results varied. If faricimab were to replace bevacizumab as a first-line therapy, it would increase costs by approximately €124.5 million per year. Thus, the higher drug price of faricimab is not outweighed by the savings from fewer injections compared to bevacizumab. For the other scenarios, when using faricimab in later treatment lines, considerable savings were found. Shifting to faricimab from the current treatment mix in second-line therapy results in a saving of €62.1 million. When faricimab enters a competitive second-line market, achieving an equal one-third market share, the savings are greater at around €75.1 million. Exclusively using faricimab in its current third-line position would result in a saving of nearly €16 million.

Subgroup and Sensitivity Analyses

A sensitivity analysis was conducted to assess the impact of using subgroup-specific pooled injection reductions rather than the overall pooled mean difference. Since the lower reduction rate (− 1.81 injections) was observed specifically in patients with prior bevacizumab use, it was applied exclusively to the first-line scenario. This adjustment increased the projected cost for the first line by approximately €38.1 million, resulting in a total budget impact of €162.6 million. Conversely, for the second- and third-line scenarios, where patients switch from aflibercept or ranibizumab, the relevant comparator is the non-bevacizumab subgroup. Applying the higher reduction rate observed in this group (− 3.33 injections) resulted in additional savings. For the second-line scenario, savings increased by approximately €18.2 million to a total of approximately €80.3 million, and for the third-line scenario, savings increased by approximately €5.4 million to a total of €21.0 million.

Figure 6 presents the two-way threshold analysis for the base case, illustrating the break-even point where faricimab and prior anti-VEGF treatments have equal annual costs. At the Dutch public list price (a 0% rebate for faricimab), cost neutrality is achieved only if the price of the anti-VEGF comparator is reduced by 36.6%. The line’s positive slope demonstrates that as the rebate on faricimab increases (moving right along the x-axis), the required rebate on the anti-VEGF comparator must also increase to maintain cost neutrality. For example, if faricimab were offered at a 50% rebate, the price of anti-VEGF would require a decrease of approximately 80% to remain cost-equivalent. Any price combination falling below this break-even line indicates a cost advantage for faricimab, while any combination above the line favors the anti-VEGF comparator.

Fig. 6.

Break-even threshold analysis for the base-case scenario. The line represents the point of cost indifference between faricimab and the weighted anti-VEGF mix treatment strategies in the Dutch context and setting. Costs consist of both medication and administration costs. Anti-VEGF anti-vascular endothelial growth factor

The pricing conditions for cost neutrality changed significantly across the different treatment-line scenarios. In the first-line scenario, where bevacizumab is the primary comparator, its low cost means that cost neutrality with faricimab cannot be achieved under any realistic price scenario. However, for second- and third-line scenarios, break-even thresholds similar to the base case were identified, indicating that faricimab can be a cost-neutral or cost-saving option. Detailed break-even plots for each scenario are provided in the Supplementary Material.

Discussion

This systematic review and meta-analysis of 19 real-world studies is the first to systematically quantify the reduction in injection frequency and its associated budget impact when switching patients with nAMD from prior anti-VEGF therapies to faricimab within the Dutch context. The primary analysis in this study, which included over 2200 patients, showed that there was a reduction of 2.65 injections in the first year after switching to faricimab (95% CI − 3.36 to − 1.93). This implies that switching to faricimab leads to a considerable reduction in treatment burden for both the healthcare system and patients.

This reduction in injection frequency is expected to lead to considerable financial implications. In the base case, which reflects the treatment mix as identified in the included international studies, switching the nAMD population to faricimab could potentially generate a cost saving of approximately €79 million in the first year after switching to faricimab. This includes savings on both drug and administration costs.

Although replacing the Dutch-specific off-label first-line treatment with bevacizumab with faricimab would increase costs, the majority of scenario analyses showed that employing faricimab in second- and third-line settings could lead to significant cost savings. These cost savings for these lines were approximately €75 million in the second-line setting and approximately €16 million in the third-line setting in the first year after switching to faricimab. However, as the scenario analyses have shown, the economic impact of faricimab is highly dependent on its placement in the Dutch treatment pathway. Strategic positioning in later lines of therapy is important to optimize both clinical and economic outcomes in the Netherlands.

Comparison with Literature

Previous reviews and meta-analyses of both clinical trials and real-world studies have primarily focused on clinical outcomes and durability outcomes such as treatment intervals of anti-VEGF therapies, often including both peer-reviewed and grey literature, without a specific focus on quantifying injection frequency or the associated economic impact [19, 22, 68]. Although several cost-effectiveness and cost-minimization analyses of anti-VEGF therapies exist for the Dutch context, these analyses largely focused on injection frequency derived from pivotal clinical trials (e.g., TENAYA/LUCERNE) and include various assumptions, rather than systematic and broad synthesis of real-world observational evidence [5, 20]. This study tackles this knowledge gap by basing its economic analysis on a systematically derived and real-world injection count.

Although our study differentiates from the previous studies by focusing on injection frequency and its associated economic impact for the Dutch context, our results are in line with the broader body of evidence. For example, our results of lower injections are in line with the outcomes of the pivotal TENAYA/LUCERNE trials, which reported that a majority of the patients could achieve extended treatment intervals of 12 weeks with faricimab [16, 17]. In addition, the pooled mean of 7.05 yearly injections of faricimab is close to the mean yearly number of injections reported in a recent network meta-analysis of Wojciechowski et al. [69]. Overall, our findings are consistent with results from other reviews, even though they focus on different metrics, that also concluded that faricimab significantly reduces treatment burden compared to anti-VEGF agents [18, 19, 22]. These reductions in injection frequency are further corroborated by recent observational data showing rapid anatomical response and fluid resolution in the loading phase [70], which supports the clinical feasibility of the extended intervals captured in our economic analysis. Furthermore, our budget impact analyses support the conclusions of previous Dutch economic evaluations, that the cost-effectiveness and hence the economic impact of faricimab are highly dependent on the comparator and their positions within the Dutch treatment pathway [5, 20]. Our study corroborates these earlier findings with a robust, real-world evidence base, and strengthens the conclusion that faricimab offers durable and economically viable treatment options, especially in later lines of therapy, with the greatest savings observed when using faricimab in the second-line setting.

Strengths and Limitations

The study included several notable strengths, of which two were novel methodological approaches. First, to the best of our knowledge, this is the first study that directly links the findings of a systematic review and meta-analysis of real-world evidence to a national budget impact analysis. This approach allows the economic analysis to be grounded on a broad and systematically synthesized evidence base from 19 real-world studies, moving beyond the commonly adopted approach of using single studies to inform parameters in budget impact analysis and economic evaluations. Second, a hybrid approach was pioneered using a generative AI tool (NotebookLM) to enhance efficiency and accuracy of the systematic review process. Unlike fully automated methods which require statistical validation (e.g., inter-rater reliability) to quantify error, our protocol employed a human-in-the-loop verification model where every data point was manually cross-referenced against the source text. This ensured full human oversight and verification of all extracted data [27]. Furthermore, other strengths include the large aggregated patient cohort of more than 2200 patients and the use of advanced statistical methods such as multiple imputation to handle missing data, which prevented study exclusion and reduced potential bias compared to complete-case analysis [41].

However, the study also has several limitations that must be acknowledged. The most important limitation is the generalizability of the international evidence to the unique Dutch context, especially regarding the first-line off-label use of bevacizumab. Generalizability of the base-case budget impact is limited, since the treatment mix of the included studies predominantly used aflibercept as the prior therapy. This does not align with Dutch guideline recommendations [11]. However, the findings are highly relevant for the subsequent lines of therapy, where aflibercept and ranibizumab are the standard of care. Since Dutch patients in these later lines are typically switching from these specific agents, our meta-analysis, which largely consists of patients switching from aflibercept and ranibizumab, accurately reflects this specific clinical population, unlike data derived from treatment-naïve trials. A related limitation is the key assumption, as stated in the methods, that the pooled mean difference of − 2.65 is generalizable to all treatment-line scenarios. This assumes that the reduction in injection frequency is constant, regardless of the prior anti-VEGF therapy. To illustrate this complexity, the sensitivity analysis revealed a counterintuitive finding related to prior bevacizumab use. The subgroup of studies that included these patients showed a smaller reduction in injection frequency (− 1.81) than the subgroup that excluded them (− 3.33). A plausible explanation is that patients on bevacizumab in these international studies represent a more challenging group to treat, possibly with more persistent or treatment-resistant nAMD. Although these patients still benefited from switching to faricimab, their improvement was less pronounced than that observed in the broader patient population. Consequently, our sensitivity analysis demonstrated that utilizing these subgroup-specific rates widens the economic divergence. It increases the projected costs for the first-line setting but substantially increases the projected savings for the second- and third-line settings. This confirms that our base-case savings estimates for the recommended Dutch treatment pathway are conservative.

However, it is plausible that our pooled estimate is a conservative estimate for the Dutch setting. Given that Dutch patients initiate treatment with off-label bevacizumab, a therapy that requires higher injection frequency compared to other anti-VEGF therapies, the true mean difference in the treatment-line scenarios could be larger than the pooled estimate. This is especially likely for patients moving towards second-line therapy [13, 48].

Furthermore, large statistical heterogeneity was observed. Although a limitation, this was not unexpected considering the inclusion of real-world observational studies. Sensitivity analyses demonstrated that prior anti-VEGF therapy was a significant source of this variance. Excluding studies that enrolled patients with prior bevacizumab use substantially reduced the overall heterogeneity, reinforcing that these patients represent a clinically distinct subgroup. Included studies also varied considerably, for example with regard to differences in follow-up durations, patient populations and the use (or non-use) of a mandatory loading phase for faricimab. Seven of the 19 included studies clearly reported using a loading phase, while four explicitly reported initiating treatment without one. The inclusion of studies with a loading phase could have potentially inflated the pooled mean number of injections in the first year of treatment with faricimab. Variability in loading phase protocols (mandatory vs. treat-and-extend initiation) likely contributed to the observed statistical heterogeneity. Consequently, a formal subgroup analysis was limited by inconsistent reporting across studies. However, it is expected that in subsequent years, the number of injections for faricimab would be lower than reported. This potential inflation in the first year does not alter the direction of our findings. The projected lower injection count in subsequent years supports the long-term sustainability of these results and would rather strengthen the study’s overall conclusion regarding treatment burden and cost savings.

Another limitation is that the budget impact analysis relies on public list prices for faricimab and anti-VEGF agents. Usually, actual costs of drugs are subject to confidential discounts. To address this limitation, a two-way threshold analysis was developed as a practical tool to allow stakeholders to input their own negotiated prices and assess the financial implication, which ensures the relevance of the budget impact analysis as the market evolves. For example, aflibercept 2 mg is on its way to patent expiration, which would allow lower-cost generic drugs to enter the market. The break-even plots included in this study remain relevant as they allow for assessing the financial implications of a low-cost generic alternative for aflibercept 2 mg.

Ultimately, the overall certainty of the evidence for the primary outcome was rated as “Very low” according to the GRADE framework. This was driven by the serious risk of bias in observational studies and the serious inconsistency (large heterogeneity) observed. Although this rating suggests that the true effect could be substantially different from the point estimate, it is important to note that the direction of effect was consistent across all 19 studies, and the pooled estimate is well in line with the broader body of evidence, including high-quality clinical trials.

Implications and Suggestions for Further Research

The findings from this study have direct implications for healthcare policy and clinical practice in the Netherlands. The analysis suggests that while using faricimab as a first-line therapy is not a cost-saving strategy compared to bevacizumab, adopting and increasing the use of faricimab in the second- and third-line settings offers potential benefits in terms of reducing patient treatment burden and lower direct healthcare costs. Furthermore, the reported reduction in injections would alleviate significant strain on Dutch ophthalmology clinics. With a projected reduction of over 96,000 injections annually, this efficiency gain represents a substantial release of capacity that could be redirected to shorten waiting lists, which currently average more than 12 weeks [7].

Other benefits include benefits beyond the direct healthcare cost perspective of this analysis. A broadened societal perspective is much needed to quantify the indirect benefits of a reduced treatment burden, including lower costs for both patients and their caregivers (e.g., travel costs, informal care costs, lost productivity costs). Although this study, including international evidence, provided the most robust estimate to date, the ultimate answer to the question which nAMD treatment strategy truly lowers costs and alleviates strain on the Dutch healthcare system lies in the prospective collection and analysis of local, real-world data. Establishing or leveraging existing national registries to capture data on actual injection frequencies, treatment intervals, and the direct impact on healthcare resources, such as on waiting lists, is crucial.

Conclusion

This systematic review and meta-analysis of real-world observational studies demonstrated that switching patients with nAMD from prior anti-VEGF therapy to faricimab can reduce the yearly treatment burden by approximately two to three injections per patient. Although this estimate reflects the variability inherent in real-world observational data (I² = 96.6%), within the Dutch healthcare context, this reduction in injection frequency can lead to substantial cost savings when faricimab is used in second- and third-line therapy. However, as a result of the low cost of off-label bevacizumab, these economic benefits do not apply to the first-line setting. Ultimately, these findings provide valuable and quantified evidence for both clinicians and policymakers that need to navigate treatment decisions for patients with nAMD.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Mohamed El Alili was responsible for the conceptualization, methodology, formal analysis, and data curation, as well as writing the original draft, reviewing and editing, and project administration. Celine J. van de Laar contributed to the methodology, data curation, and writing—review & editing. Jeroen P.F. de Greeff & Johannes G.F. Vromans contributed to review & editing. Freekje van Asten and Judith E. Bosmans were involved in writing—review & editing the manuscript. All authors approved the final manuscript.

Funding

All authors declare that they have not received any financial support to perform this study. The journal’s Rapid Service Fee was funded by Roche.

Data Availability

Data sharing is not applicable to this article as no datasets were generated or analyzed. Data from previously conducted and published studies is available in the article.

Declarations

Ethical Approval

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Conflict of Interest

Mohamed El Alili, Celine J. van de Laar, Jeroen P.F. de Greeff, and Johannes G.F. Vromans are employees and shareholders of Roche. Freekje van Asten and Judith E. Bosmans report no conflict of interests.