Systemic Drugs Associated With Maculopathy

Key Points

Question

Are there systemic drugs associated with potentially underrecognized maculopathies identified using pharmacovigilance signal detection and nationwide health claims analyses?

Findings

In this study, US Food and Drug Administration Adverse Event Reporting System (FAERS) disproportionality analysis of 15 748 maculopathy reports identified fingolimod, apixaban, paclitaxel, ibrutinib, and sildenafil as candidate signals. With the use of South Korean Health Insurance Review and Assessment Service databases, apixaban, paclitaxel, ibrutinib, and sildenafil were shown to have elevated incidence rate ratios and cumulative maculopathy incidence rates ranging from 6.0% to 15.7%, with dose-dependent risk increases for paclitaxel and ibrutinib.

Meaning

By integrating FAERS pharmacovigilance data with nationwide health claims, the cohort study revealed underrecognized maculopathy risks, supporting monitoring for maculopathies associated with drug use.

Abstract

Importance

Systemic medications may have unrecognized macular toxic effects; early identification might be important for vision preservation.

Objectives

To identify systemic drugs potentially associated with maculopathy via pharmacovigilance reporting and to evaluate their macular adverse effects in a nationwide encounter database.

Design, Setting, and Participants

This 2-part study included (1) disproportionality analysis for candidate identification, using 15 748 maculopathy-related individual case safety reports from the US Food and Drug Administration Adverse Event Reporting System (FAERS) between July 2014 and December 2023, and (2) a population-based cohort study for association evaluation, using the South Korean Health Insurance Review and Assessment Service (HIRA) database among users of the candidate drugs, covering approximately 50 million individuals. Data were analyzed from January 1, 2015, to December 31, 2023.

Exposure

Use of systemic drugs identified as candidate signals in FAERS.

Main Outcomes and Measures

Reporting odds ratios for signal detection in FAERS and incidence rate ratios, hazard ratios, and cumulative incidence of maculopathy in HIRA.

Results

Five systemic drugs with underrecognized macular toxic effects—fingolimod, apixaban, paclitaxel, ibrutinib, and sildenafil—were identified among the top 30 maculopathy signals in FAERS. In HIRA, the incidence rate ratios for maculopathy following exposure (vs preexposure) were 1.92 (95% CI, 0.62-5.96) for fingolimod, 3.08 (95% CI, 2.68-3.54) for apixaban, 2.85 (95% CI, 1.62-5.02) for paclitaxel, 3.71 (95% CI, 2.58-5.34) for ibrutinib, and 2.75 (95% CI, 2.17-3.48) for sildenafil. The cumulative incidence rates ranged from 4.4% (fingolimod) to 15.7% (apixaban). Dose-response relationships were observed for paclitaxel (hazard ratio, 2.01 [95% CI, 1.77-2.29] for third vs first quartile) and ibrutinib (hazard ratio, 4.82 [95% CI, 1.39-16.81] for fourth vs first quartile).

Conclusions and Relevance

These findings suggest that the integration of pharmacovigilance signal detection and nationwide health claims analysis identified associations of maculopathy with apixaban, paclitaxel, ibrutinib, and sildenafil. This combined approach offers a potentially cost-effective, robust method for identifying systemic drugs with possibly underrecognized macular adverse effects.

This study examines data to identify systemic drugs potentially associated with maculopathy via pharmacovigilance reporting and evaluates their macular adverse effects in a nationwide encounter database.

Introduction

Drug-induced maculopathy can be a relevant but potentially preventable condition that affects the macula.^1,2 It can lead to visual impairment ranging from subtle changes to irreversible vision loss, depending on the severity of damage. There are well-documented associations of drug-induced retinal toxic effects, such as hydroxychloroquine retinopathy or pentosan polysulphate maculopathy^3,4,5,6; however, the potential harm of other systemic drugs associated with macular toxic effects may be underexplored.

Integrating pharmacovigilance signal detection with evidence from the clinical practice setting provides a promising approach to identifying potentially novel or underrecognized associations, enhancing our understanding of drug safety and possibly contributing to the safe use of these drugs. Considering this, we hypothesized that combining pharmacovigilance signal detection with health claims database analysis can effectively identify and evaluate drug-related maculopathy, allowing for the validation of pharmacovigilance signals and the quantification of associated maculopathy risks in population-based or other large datasets.

Current adverse event reporting systems, widely used for pharmacovigilance signal detection,^7,8,9 have inherent limitations that include underreporting, lack of exposure denominator data, and potential confounding factors leading to spurious associations.^7,10 To overcome these limitations, this study used disproportionality analysis of the US Food and Drug Administration Adverse Event Reporting System (FAERS) database to identify systemic drugs potentially associated with maculopathy and conducted a nationwide population-based cohort study using the South Korean Health Insurance Review and Assessment Service (HIRA) database to evaluate these associations. By analyzing the risk of maculopathy before and after candidate drug use, as well as dose-response relationships, this study sought to establish a comprehensive framework for identifying and evaluating systemic drugs with macular adverse effects using 2 large public databases, potentially contributing to ocular safety for prescription drug users.

Methods

This study adopted a 2-phase approach, integrating pharmacovigilance signal detection from FAERS with population-based analysis using the HIRA database to identify and evaluate systemic drugs associated with maculopathy. This study adhered to the principles of the Declaration of Helsinki and was approved by the institutional review board of Hanyang University Hospital. Because it was a retrospective analysis of a fully deidentified dataset, the institutional review board waived the requirement for informed consent. The use of anonymized data from the FAERS and HIRA ensured compliance with the READUS-PV (Reporting of a Disproportionality Analysis for Drug Safety Signal Detection Using Individual Case Safety Reports in PharmacoVigilance) statement for disproportionality analysis in pharmacovigilance and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies.^10,11 See the eMethods, eTables 1 to 4, and eFigure 1 in Supplement 1 for the methods used for the identification of candidate drugs using FAERS.^{12,13,14,15,16,17}

Cohort Study for Evaluating Maculopathy Risk Using HIRA

We used the HIRA database, which encompasses comprehensive health care claims from nearly the entire South Korean population (97%; approximately 50 million) under the universal mandatory health insurance system, to identify patients prescribed 1 of the candidate drugs identified in the FAERS analyses from January 1, 2015, to December 31, 2023; its near-universal coverage inherently minimizes volunteer bias compared with voluntary registries. All demographic and key clinical variables used in our analyses, including age, sex, residence, prescriber specialty, indication, and ophthalmologic diagnoses (for individuals with ophthalmology-related claims), were captured as complete fields in the HIRA dataset. All eligible initiators were drawn from the full nationwide claims set without further sampling to minimize selection bias. The study schematic for the HIRA analysis is graphically depicted in Figure 1.

Figure 1. Schematic Diagram of the Study Design Using the Health Insurance Review and Assessment Service (HIRA) Database to Evaluate Maculopathy Risk for Candidate Drugs.

FAERS indicates US Food and Drug Administration Adverse Event Reporting System.

Statistical Analysis

Maculopathy incidence rates before and after drug exposure were calculated as events per person-year, stratified by drug type. Incidence rate ratios (IRRs) were computed to quantify the relative increase in risk after drug exposure. Cumulative incidence rates of maculopathy at 6 months, 1 year, and the end of the study period were calculated for each drug cohort. Kaplan-Meier survival analysis was performed to investigate the cumulative risk over time. Hazard ratios (HRs) with 95% CIs were calculated using the Cox proportional hazards model adjusted for age, with the lowest cumulative dose group serving as the reference for each exposure category to assess the dose-response relationship.

Continuous variables were presented as mean (SD) values. All analyses were performed using SAS Enterprise Guide, version 7.1 (SAS Institute) or R, version 4.0.3 (R Foundation for Statistical Computing), and all P values were 2-sided without adjustment for multiple analyses. P < .05 was considered statistically significant. Data were analyzed from January 1, 2015, to December 31, 2023.

Results

FAERS Analysis

Table 1 presents the 30 candidate drugs with the highest number of reported maculopathy-related adverse events in the FAERS database as of December 2023. It highlights a diverse range of medications with varying levels of reporting odds ratio (ROR). Pentosan polysulfate sodium ranked first with 2627 reports and an exceptionally high ROR (1081.8 [95% CI, 1030.8-1135.3]), indicating a strong association with maculopathy. Hydroxychloroquine had 230 reports and a notably high ROR (9.5 [95% CI, 8.3-10.8]). By applying the exclusion criteria (eTable 2 in Supplement 1), we identified 5 candidate drugs for further analysis: fingolimod (788 reports; ROR, 8.1 [95% CI, 7.6-8.8]), apixaban (262 reports; ROR, 2.5 [95% CI, 2.2-2.8]), paclitaxel (248 reports; ROR, 7.6 [95% CI, 6.7-8.6]), ibrutinib (175 reports; ROR, 2.4 [95% CI, 2.0-2.7]), and sildenafil (92 reports; ROR, 3.4 [95% CI, 2.8-4.2).

Table 1. Top 30 Candidate Drugs With the Highest Number of Reported Maculopathy Adverse Events (December 2023).

Rank	Medication	No. of reports	Reporting odds ratio (95% CI)
1	Pentosan polysulfate	2627	1081.8 (1030.8-1135.3)
2	Ranibizumab	1021	83.2 (78.1-88.8)
3	Insulin	879	4.5 (4.2-4.8)
4	Adalimumab	799	1.3 (1.3-1.4)
5	Fingolimod	788	8.1 (7.6-8.8)
6	Aflibercept	524	29.0 (26.5-31.6)
7	Bevacizumab	285	4.9 (4.3-5.5)
8	Apixaban	262	2.5 (2.2-2.8)
9	Paclitaxel	248	7.6 (6.7-8.6)
10	Hydroxychloroquine	230	9.5 (8.3-10.8)
11	Brolucizumab	208	45.1 (39.3-51.8)
12	Cefuroxime	194	30.5 (26.5-35.2)
13	Tiotropium	181	5.2 (4.5-6.0)
14	Ibrutitnib	175	2.4 (2.0-2.7)
15	Etanercept	173	0.5 (0.4-0.6)
16	Methotrexate	158	0.8 (0.7-0.9)
17	Lenalidomide	149	0.6 (0.5-0.7)
18	Teriperatide	114	1.2 (1.0-1.4)
19	Tofacitinib	111	0.7 (0.6-0.8)
20	Pregabalin	108	1.0 (0.9-1.2)
21	Verteporfin	108	201.4 (165.3-245.3)
22	Alprazolam	103	3.6 (2.9-4.3)
23	Dupilumab	101	0.5 (0.4-0.5)
24	Prednisolone	101	1.3 (1.1-1.6)
25	Denosumab	94	0.8 (0.7-1.0)
26	Sertraline	92	1.8 (1.5-2.2)
27	Sildenafil	92	3.4 (2.8-4.2)
28	Latanoprost	91	8.9 (7.2-10.9)
29	Dexamethasone	89	2.0 (1.6-2.5)
30	Rivaroxaban	87	1.0 (0.8-1.2)

eTable 5 in Supplement 1 summarizes the clinical characteristics of patients treated with the 5 candidate drugs, as reported in the FAERS database. Most reports for fingolimod (33.8% [266 of 788]) and sildenafil (31.5% [29 of 92]) use involved patients aged 40 to 59 years, whereas other drugs were more commonly reported in older populations. The sex distribution showed a predominance of female patients for most drugs, particularly fingolimod (70.6% [556 of 788]), apixaban (69.5% [182 of 262]), and paclitaxel (60.1% [149 of 248]), whereas sildenafil was primarily reported to be used by male patients (67.4% [62 of 92]). Geographically, most reports originated in North America, with the United States contributing the majority of patients. Health professionals were the primary reporters for all drugs except apixaban and ibrutinib, which had a higher proportion of consumer reports (79.0% [207 of 262] and 77.1% [135 of 175], respectively). Serious adverse events, including disability, were reported at low frequencies for all drugs, with sildenafil having the highest rate (3.3% [3 of 92]). The number of adverse events reported varied by year, with recent years (2020-2023) showing a relatively higher proportion of reports, particularly for apixaban and ibrutinib.

The specific types of maculopathy reported in the FAERS database highlight the varying patterns of maculopathy associated with fingolimod, apixaban, paclitaxel, ibrutinib, and sildenafil, as shown in eFigure 2 in Supplement 1. Macular edema was the most common condition for fingolimod (677 cases [85.9%]) and paclitaxel (224 cases [90.3%]), whereas apixaban and ibrutinib use was predominantly associated with macular degeneration (248 cases [94.7%] and 132 cases [75.4%], respectively). Sildenafil adverse reactions were more evenly distributed among macular degeneration (39 cases [42.4%]), macular edema (29 cases [31.5%]), and other maculopathies (24 cases [26.1%]).

HIRA Analysis

eTable 6 in Supplement 1 presents the clinical characteristics and prescription details of the patients in a population-based cohort using the HIRA database for the 5 candidate drugs. The mean (SD) age of the patients varied substantially across medications, with fingolimod users being the youngest (35.3 [10.4] years) and apixaban users being the oldest (73.0 [11.5] years). The distribution of age groups also differed, with a notable proportion of patients aged 60 years or older for apixaban (31.2% [104 665 of 335 968]) and ibrutinib (15.0% [139 of 924]), whereas fingolimod had a higher percentage of patients younger than 30 years (33.7% [62 of 184]). Sex distribution also varied, with a higher proportion of female patients in the fingolimod (63.0% [116 of 184]) and paclitaxel (59.0% [57 973 of 98 289]) cohorts and a predominance of male patients in the ibrutinib cohort (72.5% [670 of 924]). The median duration of use varied markedly across different drugs.

Table 2 presents the event rates and corresponding IRRs for maculopathy observed before and after the initiation of the 5 drugs. Fingolimod showed an IRR of 1.92 (95% CI, 0.62-5.96). The IRRs were 3.08 (95% CI, 2.68-3.54) for apixaban, 2.85 (95% CI, 1.62-5.02) for paclitaxel, 3.71 (95% CI, 2.58-5.34) for ibrutinib, and 2.75 (95% CI, 2.17-3.48) for sildenafil, indicating a pronounced increase in the rates of events from before to after use (all P < .05). eTable 7 in Supplement 1 further presents age-adjusted HRs for these drugs, with adjustments based on the differences between preexposure and postexposure periods, which are consistent with the IRR results. In addition, sensitivity analyses using t₀ at January 1, 2016, and January 1, 2017, yielded similar IRR estimates for all 5 candidate drugs (eTable 8 in Supplement 1), confirming stability of the effect estimates.

Table 2. Incidence Rate Ratios for Maculopathy Before and After Drug Use.

Drug	Before drug initiation				After drug initiation				Incidence rate ratio (95% CI)a
	Mean (SD) period, y	Median (IQR) period, y	Events, No.	Rate, No. of events/person-year	Mean (SD) period, y	Median (IQR) period, y	No. of events	Rate, No. of events/person-year
Fingolimod (n = 184)	6.7 (4.8)	6.9 (2.5-9.4)	12	0.15	3.5 (2.6)	3.9 (1.4-5.6)	4	0.29	1.92 (0.62-5.96)
Apixaban (n = 335 968)	6.6 (3.8)	7.1 (3.4-9.7)	61 710	0.15	2.2 (1.8)	1.8 (0.8-3.2)	19 652	0.45	3.08 (2.68-3.54)
Paclitaxel (n = 98 289)	7.3 (3.8)	7.8 (4.4-10.3)	8422	0.14	2.4 (2.0)	1.8 (0.7-3.5)	3020	0.42	2.85 (1.62-5.02)
Ibrutinib (n = 924)	7.0 (3.8)	7.1 (4.4-9.8)	150	0.14	1.9 (1.7)	1.4 (0.7-2.3)	36	0.53	3.71 (2.58-5.34)
Sildenafil (n = 3226)	6.9 (3.8)	7.1 (3.8-9.7)	201	0.15	2.5 (2.0)	2.1 (0.8-3.9)	105	0.40	2.75 (2.17-3.48)

^a The preexposure period served as the denominator.

Table 3 shows the cumulative incidence rate of maculopathy observed at 3 time points: 6 months and 1 year after drug initiation and up to the end of the study (up to 9 years). Among the 5 drugs analyzed, apixaban demonstrated the highest cumulative incidence at all time points, reaching 1.2% at 6 months, 2.4% at 1 year, and 15.7% by the end of the study (up to 9 years). In contrast, fingolimod exhibited the lowest cumulative incidence rate in each interval (0.6%, 0.6%, and 4.4%, respectively). The remaining drugs displayed intermediate trends, with ibrutinib showing a particularly steep increase over time, reaching 9.9% by the study’s conclusion. eFigure 3 in Supplement 1 illustrates the cumulative incidence rate of maculopathy over the entire study period for each drug, providing a visual representation of the temporal patterns and highlighting variations in the temporal risk patterns among the drugs. Apixaban, paclitaxel, and sildenafil showed a gradual trend with a slight decrease in the slope over time.

Table 3. Cumulative Incidence Rates of Maculopathy per Person-Year Up to 6 Months and 1 Year After Drug Initiation and the Study End Date^a.

Drug	Duration of medication use, median (range), mo	Cumulative incidence rate, % (95% CI)
		At 6 mo	At 1 y	Up to study end
Fingolimod	31.3 (0.3-83.0)	0.6 (0.1-4.2)	0.6 (0.1-4.2)	4.4 (1.1-16.4)
Apixaban	6.6 (0.2-108.0)	1.2 (1.1-1.2)	2.4 (2.3-2.4)	15.7 (14.7-16.8)
Paclitaxel	2.8 (0.2-107.5)	0.6 (0.5-0.6)	1.2 (1.1-1.2)	6.6 (6.1-7.0)
Ibrutinib	9.3 (0.2-108.0)	0.8 (0.4-1.8)	2.2 (1.3-3.5)	9.9 (5.5-17.3)
Sildenafil	3.7 (0.2-108.0)	0.6 (0.4-0.9)	1.1 (0.8-1.6)	6.0 (4.8-7.5)

^a First year of use: fingolimod (2017), apixaban (2013), paclitaxel (2009), ibrutinib (2016), and sildenafil (2013).

Dose-Response Relationship

Figure 2 illustrates the dose-response relationship between age-adjusted maculopathy risk and drug exposure, stratified by cumulative dose risk group (tertiles or quartiles). The results showed distinct trends across drugs, with varying HRs for maculopathy as the cumulative doses increased. Fingolimod and sildenafil showed no marked increase in HR in the higher exposure groups, with age-adjusted HRs of 0.57 (95% CI, 0.05-6.79) and 3.24 (95% CI, 0.44-24.06) in the highest exposure group, respectively. Apixaban showed only a mild increase in risk in the highest exposure group, with an HR of 1.16 (95% CI, 1.11-1.21). Paclitaxel demonstrated a marked increase in maculopathy risk with higher cumulative doses, peaking in the third quartile (HR 2.01 [95% CI, 1.77-2.29]). Ibrutinib also showed a dose-dependent increase in maculopathy risk, with the fourth quartile having the highest HRs, 4.82 (95% CI, 1.39-16.81).

Figure 2. Dose-Response Relationship Between Maculopathy and Drug Exposure, Stratified by Risk Group Based on Cumulative Dose.

Tertiles are reported for groups with less than 100, whereas the remainder are reported by quartile. HR indicates hazard ratio.

Discussion

This study identified systemic drugs with maculopathy associations that potentially have been underrecognized or underreported. Using a 2-step study design, we first identified drugs of interest in a pharmacovigilance reporting system and then assessed these drugs in a nationwide population-based cohort study using an encounter database. This study highlights a methodology that integrates signal detection from a pharmacovigilance reporting system with detailed evaluation in a large encounter database to both identify and evaluate drug-maculopathy associations. By combining these complementary data sources, our approach provides a potentially novel assessment of potential macular adverse reactions (eTable 9 and eDiscussion in Supplement 1) associated with systemic drugs.

The FAERS, one of the largest pharmacovigilance databases globally, provides a platform for signal detection through disproportionality analysis, leveraging spontaneous reports of adverse events to identify drugs that require further investigation.^8,12 Its strengths include global coverage, a large volume of reports, and utility as an early warning system for rare adverse events (eg, maculopathy). However, pharmacovigilance reporting systems, such as FAERS, have inherent limitations, including reporting biases, lack of denominator data, lack of consideration of confounding factors, and the inability to establish causality.^18,19 Moreover, only relative signals could be identified through disproportionality analysis, which should be further validated by more rigorous epidemiological studies, such as case-control or cohort studies. To address these limitations, we conducted further analyses using the HIRA database, which includes comprehensive and publicly available data on approximately 50 million individuals in South Korea. The HIRA database encompasses comprehensive data on drug exposure, health outcomes, and potential confounding factors, making it a robust, nationwide, population-based resource that has been validated in numerous studies.^5,6,20,21 This integration allows for a detailed evaluation of drug-adverse event associations identified in FAERS, providing a potentially stronger basis for considering causality and quantifying risks.

Our analysis of the FAERS database highlighted a diverse range of medications associated with maculopathy, underscoring the importance of further investigation of certain high-ROR candidates. Pentosan polysulfate sodium and hydroxychloroquine, with 2627 and 230 reports and notably high RORs of 1081.8 and 9.5, respectively, showed an association with maculopathy, reaffirming their established association with the condition. Moreover, in the HIRA analysis, application of the same self-controlled case-series model to hydroxychloroquine and pentosan polysulfate users in our initial cohort yielded elevated IRRs (eTable 10 in Supplement 1), further supporting the validity of our methodology and offering reference levels for postexposure maculopathy risk. In addition, we identified 5 candidate drugs from the top 30 list—fingolimod, apixaban, paclitaxel, ibrutinib, and sildenafil—that warrant further analysis due to their high ROR values, which suggest previously underrecognized associations with maculopathy. This evaluation lays the groundwork for further, more comprehensive studies on these potential drug-related risks.

The distinct patterns of maculopathy associated with each drug, as shown in Figure 1, offer insights into the potential drug-specific mechanisms and risks for various types of maculopathy. Among the 5 drugs identified in our candidate list, paclitaxel and fingolimod have been previously reported for macular edema,^22,23 although these associations have not been conclusively confirmed by large-scale studies. Fingolimod and paclitaxel were also predominantly associated with macular edema in our study, which is consistent with their potential effects on vascular permeability.^24,25 In contrast, apixaban, ibrutinib, and sildenafil were more frequently associated with a nonspecific macular degeneration, which may reflect differences in the underlying mechanisms. These patterns underscore the heterogeneity of maculopathy manifestations across drugs, highlighting the importance of considering drug-specific risks and clinical presentations when monitoring patients for adverse ocular events. These findings illustrate the value of pharmacovigilance systems in characterizing detailed clinical presentations, enabling the identification of drug-specific maculopathy patterns, and informing tailored monitoring strategies for users of each drug.

The observed IRR for maculopathy events before and after the use of the 5 medications (Table 2) indicated an increased risk following drug initiation for 4 of the 5 drugs. Although fingolimod showed the fewest events—likely associated with a smaller number of users—the other drugs demonstrated robust associations with increased IRR, indicating substantial increases in maculopathy incidence following drug initiation. Our results support the hypothesis that medications frequently reported and potentially associated with maculopathy, as identified through FAERS disproportionality analysis, may also be associated with maculopathy in clinical encounter database analyses. Moreover, the risk can be quantitatively assessed in a large population. These findings emphasize the hypothesis-generating role of a pharmacovigilance reporting system in identifying drug-adverse events’ association and hypothesis-strengthening role in studies using a large health claims database. This complementary approach highlights the importance of integrating these methods for comprehensive pharmacovigilance and population-based drug safety research, including identification of novel drugs associated with specific ocular adverse reactions.

Table 3 may be clinically helpful for understanding the cumulative incidence of maculopathy associated with each drug over time. These findings highlight the fact that certain drugs, particularly apixaban and ibrutinib, are associated with a progressively higher risk of maculopathy with greater exposure. The cumulative incidence rates, approaching or exceeding 10%, suggest that clinicians should remain vigilant in monitoring macular changes in patients prescribed these medications, especially if subsequent studies confirm these findings. These findings suggest that regular, long-term ocular monitoring should be integrated into clinical practice for candidate drug users to enable timely detection and intervention to manage associated maculopathy.

The dose-response relationship depicted in Figure 2 underscores the increasing risk of maculopathy with higher cumulative exposure to several drugs. For drugs such as paclitaxel and ibrutinib, the risk of maculopathy increased progressively with increasing cumulative doses. This reinforces the association between drug exposure and the development of maculopathy and highlights the potential need for dose adjustments or enhanced monitoring for those with higher exposure levels. Overall, these findings suggest that clinicians should be cautious when prescribing these drugs to patients with higher cumulative doses and implement regular ophthalmic screenings for the detection of maculopathy, especially if subsequent studies confirm these findings.

Limitations

This study has limitations. First, in drug identification, the study relied on adverse event reporting, which is susceptible to reporting bias because not all adverse events are reported or they may have been underreported, particularly for less severe or rare cases. Additionally, by focusing on the top 30 drugs, the analysis may have excluded medications with rare macular adverse reactions, potentially overlooking important associations. The designation of the primary suspect drug in FAERS, which can be made by nonexperts, is also a limitation; however, FAERS’s large volume and broad inclusivity facilitate early signal detection and screening. Accordingly, we first used FAERS to identify candidate maculopathy-associated drugs and then validated these signals using claims data. Another limitation is the low number of fingolimod users, which may have contributed to the lack of a clear dose-response pattern for maculopathy risk in this cohort. The selection of the preexposure start date is somewhat arbitrary and may introduce uncertainty in IRR estimates; however, our sensitivity analyses using alternative start dates demonstrated consistent results, supporting the robustness of the observed associations. Furthermore, the diagnoses of maculopathy were based on health claims data, which depended on assessments made by ophthalmologists but were not confirmed through direct medical records. Moreover, the study population in the encounter database, which consisted entirely of Korean patients, limits the generalizability of the findings to other populations, although reproducing FAERS signals, which includes populations around the world, in this independent population may help support the validity of our findings. These limitations underscore the need for further research with larger and more diverse cohorts and access to verified clinical data to strengthen the findings and ensure the robustness of the conclusions drawn. Additionally, although we mitigated confounding using a self-controlled case-series design and further adjusted for key demographic characteristics (age and sex) and concomitant retinotoxic medications, residual confounding by health care use patterns and other unmeasured factors may have influenced our findings. Finally, disproportionality analysis alone cannot distinguish between true, rare adverse drug reactions and nonspecific events. Furthermore, some rare conditions could not be validated in the claims data owing to potential mismatches between the Medical Dictionary for Regulatory Activities (MedDRA) Preferred Terms and the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10)–based diagnostic codes, as MedDRA is a terminology system focused on adverse events and clinical symptoms, whereas ICD-10 is a coding system designed for the diagnosis and classification of diseases. For more definitive causal inference, signals involving rare events may require target trial emulation or well-controlled cohort studies.

Conclusions

In summary, this cohort study identified several candidate drugs with previously underrecognized risks of maculopathy through both pharmacovigilance data and subsequent large-scale encounter database analyses. Our approach demonstrated the feasibility of detecting and evaluating drug-related maculopathy associations in a clinical setting. Specifically, apixaban, paclitaxel, ibrutinib, and sildenafil were associated with an increased risk of maculopathy, with different clinical presentations, particularly at higher cumulative doses. These findings highlight the potential importance of ongoing pharmacovigilance efforts and the need for regular ophthalmic monitoring of patients receiving these medications to enable early detection and timely intervention if subsequent studies confirm these findings.

Supplement 1.

eMethods. Identification of candidate drugs using FAERS

eTable 1. FAERS terms used for identifying maculopathy adverse events and their number of reports in the database

eTable 2. Criteria for excluding candidate systemic drugs associated with maculopathy from FAERS database

eTable 3. Definitions and Korean Classification of Diseases (KCD) codes (adapted from ICD-10) used to identify maculopathy

eTable 4. Summary of exclusion criteria with the initial, excluded, and final user counts for candidate systemic drugs in the HIRA nationwide health claims database

eTable 5. Clinical characteristics of patients treated with one of five candidate drugs in the FAERS database

eTable 6. Clinical characteristics and prescription details of the included patients in population-based cohort from the HIRA database

eTable 7. Age-adjusted hazard ratios (HRs) for maculopathy before and after the drug use

eTable 8. Sensitivity analyses of different pre-exposure start dates and incidence rate ratios (IRR) for maculopathy

eTable 9. Literature search for novelty of our two-step approach

eDiscussion.

eTable 10. Incidence rate ratios (IRRs) before and after drug exposure for pentosan polysulfate (PPS) and hydroxychloroquine (HCQ) users in the initial cohort

eFigure 1. Flowchart for identification of candidate drugs associated with maculopathy using FDA Adverse Event Reporting System (FAERS) data

eFigure 2. Distribution of maculopathy types by prescription drug among patients reporting maculopathy in the FAERS database

eFigure 3. Cumulative incidences of maculopathy over the study period

Supplement 2.

Data Sharing Statement