Real-world therapy and persistence of patients with neovascular age-related macular degeneration and diabetic retinopathy or diabetic macular edema: a German claims data analysis

Abstract

Purpose

Vascular endothelial growth factor (VEGF) inhibition is the current and high-volume standard-of-care for patients with neovascular age-related macular degeneration (nAMD) and diabetic retinopathy (DR) with diabetic macular edema (DME). This study assessed the impact of non-persistence in anti-VEGF treatment using claims data from two German states.

Methods

This study identified adults with nAMD or DR/DME and incident anti-VEGF treatment (= index) in January 2015-June 2019 using the German AOK PLUS claims database (January 2014-June 2021, ~ 3.5 million insured). Baseline characteristics were observed within 12 months before index. Patient follow-up lasted ≥ 24 months or until death. Non-persistence (gap of ≥ 180 days) was calculated using Kaplan–Meier estimation. Cox regression identified variables linked to non- persistence. The study analysed reimbursed anti-VEGF treatments, thus excluding off-label use of bevacizumab.

Results

5,498 patients diagnosed with nAMD (mean age, 80.09 years; male, 37.50%; mean Charlson Comorbidity Index [CCI] score, 3.07) and 484 patients with DR/DME (mean age, 67.14; male, 58.88%; mean CCI score, 4.54) were identified. Non-persistence to anti-VEGF treatment within 12 months after index occurred in 51.38% of nAMD patients and 62.60% of DR/DME patients, with mean times to first gap of 11.28 and 8.98 months, respectively. Cox regression revealed factors associated with non-persistence, including higher age, female gender, higher care needs, longer AMD history, and the use of ranibizumab.

Conclusion

Epidemiologic and ophthalmologic factors associated with anti-VEGF non-persistence were successfully identified in the first year of therapy. The analyzed dataset can potentially be enriched with additional health insurance database sets under the used criteria to gain more understanding of anti-VEGF non-persistence.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00417-024-06690-9.

Key messages

What is known

• Observational data on anti-vascular endothelial growth factor (VEGF) therapy has shown that there is a high rate of non-persistence among patients with neovascular age-related macular degeneration (nAMD) or diabetic retinopathy (DR) with diabetic macular edema (DME); Nonetheless, no study in Germany had examined the non-persistence of nAMD or DR/DME patients using administrative data.

What is new

• This study is the first to provide real-world insights into the non-persistence of anti-VEGF injections in patients with nAMD or DR/DME using German billing data while examining possible underlying factors, patient characteristics, and treatment patterns.

• Non-persistence to anti-VEGF treatment within 12 months after index occurred in more than half of nAMD patients and more than 60% of DR/DME patients. Factors associated with non-persistence included higher age, female gender, higher care needs, longer AMD history, and the use of ranibizumab rather than aflibercept.

• Based on our data, non-persistence to anti-VEGF therapy in clinical practice is high in nAMD and DR/DME patients, calling for closer evaluation of patient parameters and needs, particularly in patients who do not attend routine assessments in clinical practice.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00417-024-06690-9.

Background

Neovascular, exudative, or "wet" age-related macular degeneration (nAMD) is a progressive and degenerative disease in elderly people [1], primarily characterized by macular neovascularization, i.e., the abnormal growth of blood vessels under the macula [2, 3], which is mainly caused by the upregulation of vascular endothelial growth factor (VEGF)-A [4]. While it occurs in about 10% of the approximately 7.0 million AMD patients in Germany [5], it accounts for approximately 90% of all vision loss caused by AMD [6, 7].

Diabetic macular edema (DME) is a chronic disease and complication of diabetic retinopathy (DR) involving microvascular abnormalities. Patients with DR and DME suffer chronic inflammation, and/or ischemia in the retina. This typically results in the breakdown of the blood-retinal barrier, accumulation of extracellular fluid, and the subsequent swelling of the macula [8]. DR and DME represent a major cause of impaired visual acuity (VA) and vision loss in middle-aged and elderly patients with diabetes mellitus [9, 10] and are thought to affect approximately 2.5 million patients in Germany [11].

Over the past 16 years, new therapies targeting the VEGF have been approved throughout Europe to improve VA outcomes and stop neovascularization in patients with nAMD, DR, and DME. Intravitreal therapy (IVT) using anti-VEGF or, more recently, anti-VEGF/angiopoietin 2 (ANG2) agents currently represent the standard of care [12, 13], and the widespread use of VEGF-inhibitors in the German healthcare system has established a high therapeutic standard [14]. Within the context of anti-VEGF therapy, patients oftentimes receive monthly IVT injections for the first three (nAMD) to five (DR and DME) loading doses and then on a bi- or tri-monthly basis thereafter [15, 16]. At present, four anti-VEGF agents are authorized for IVT therapy of nAMD and DR and DME in Europe, including ranibizumab, receiving approval from the European Medicines Agency (EMA) in 2007, aflibercept approved in 2012, brolucizumab approved in 2020 and faricimab approved in 2022 (the latter being an anti-VEGF/anti-Ang2 agent, falls outside the observation period of this study) [17–20]. Bevacizumab, a humanized anti-VEGF monoclonal antibody approved for the treatment of several forms of malignant tumors, is frequently applied as an inexpensive, off-label therapy in patients with nAMD throughout Europe [19, 21]. Clinical trials consistently demonstrate the positive impact of regular anti-VEGF injections, improving vision and preventing vision loss in patients with nAMD or DR and DME [22–26]. However, observational data on anti-VEGF therapy has shown that there is a high rate of non-persistence, meaning that patients discontinue their therapy prematurely [27]. Factors that have been reported to influence the rate of non-persistence include transportation issues, poor general health, or lack of organizational efficiency (i.e., long waiting times) [27–30]. The percentage of patients who discontinue anti-VEGF therapy within 12 months of initiation can be as high as 57% [31], which might have implications on the therapeutic outcomes demonstrating an unmet medical need.

Moreover, previous studies have sought to measure non-adherence to anti-VEGF therapy, defined as deviation from planned therapy intervals [32]. Depending on the definition, reported non-adherence rates varied widely in nAMD, ranging between 32% and 95% [33–35]. However, though some experts have made suggestions for the exact frequency of maintenance anti-VEGF dosing [36, 37], no written consensus across therapies and indications exists, and physicians may use different methods (e.g., standard schedule, pro re nata, treat-and-extend) to prevent disease relapse without overwhelming the patient [38, 39]. Because our understanding of reasons for variations in real-world therapy within claims data is limited and considering that patient-related non-adherence has a lesser impact on the patient's clinical outcome and, consequently, a reduced socio-economic impact than non-persistence [40], it is not analyzed in this study.

As outlined above, prior real-world studies have evaluated the degree and causative factors behind non-persistence in anti-VEGF therapy among nAMD or DR and DME patients [23, 27, 29, 31, 34]. Similarly, in prior German studies for non-ophthalmologic indications, non-persistence was determined by identifying gaps in recorded billing codes, mirroring the approach used in this study [41–43]. Nevertheless, no study in Germany examined the non-persistence of nAMD or DR and DME patients using claims data. Thus, this study aims to delineate patient characteristics and therapy patterns, assess non-persistence prevalence, and elucidate the factors influencing their persistence to anti-VEGF IVT injections in two neighboring German states.

Methods

Study design and data source

This study is a retrospective, non-interventional cohort analysis utilizing German claims data from 01/01/2014–30/06/2021. Patients included in this study were followed longitudinally for at least 24 months from the date of the first IVT anti-VEGF injection until the end of the study period (30/06/2021) unless a patient was censored due to death.

Data was provided by AOK PLUS, the sixth-largest German sickness fund covering approximately 3.5 million insured persons in the regions of Saxony and Thuringia in central-eastern Germany. This corresponds to about 5% of the overall population of statutory insured patients in Germany. The AOK PLUS dataset contains data on the sociodemographic profile of patients, their therapy with prescription aids and medications, as well as all billable data on outpatient and inpatient care.

Study population

Two cohorts of patients who started therapy with anti-VEGF agents were identified. One cohort contained patients with nAMD, identified through the International Statistical Classification of Diseases And Related Health Problems, 10th revision, German Modification (ICD-10-GM) code H35.3 for AMD, and the receipt of IVT injections specifically indicated for nAMD. This approach was necessitated by the absence of a specific ICD-10-GM code for nAMD at the time of patient identification. The other cohort consisted of patients with DR/DME. The rationale for including patients with DR or DME rather than DME alone results from the absence of a distinct ICD-10-GM code for DME. As a result, a proxy was utilized, and ICD-10-GM codes for DR (ICD-10 H36.0) or diabetes mellitus associated with ophthalmologic complications (ICD-10-GM E10.3, E11.3, E12.3, E13.3, E14.3) were identified, as outlined below. However, this way, patients with DME and patients with DR cannot be distinguished from each other.

Patients in the nAMD cohort were defined as follows:

1. Continuously insured from 01/01/2014–30/06/2021

2. At least one inpatient or two confirmed outpatient diagnoses of AMD from an ophthalmologist (physician key 5) from 01/01/2014–30/06/2019

3. At least one documented EBM code related to IVT injections from 01/01/2015–30/06/2019, in association with an Anatomical Therapeutic Chemical Classification System (ATC) code or operations and procedures key ("Operationen- und Prozedurenschlüssel", OPS) for an anti-VEGF agent (ranibizumab, aflibercept, brolucizumab, or bevacizumab) prescribed within 30 days before the first observed IVT-related EBM code (first EBM code served as the index date; codes utilized can be found in Supplementary Table 1)

4. At least one additional IVT administration in combination with an anti-VEGF agent

5. No anti-VEGF prescription from 01/01/2014 until 31 days before index and no IVT injection from 01/01/2014 until the day before index

6. Age of at least 50 years at index

7. No documentation of a DME implant (dexamethasone or fluocinolone) or DME focal laser therapy during the study period (Supplementary Table 1)

8. No inpatient or outpatient diagnosis of ophthalmic complications during the entire study period, including DME, as well as retinal vein occlusion (ICD-10-GM H34)

Patients with DR/DME were described as follows:

1. Continuously insured from 01/01/2014–30/06/2021

2. At least one inpatient or two confirmed outpatient diagnoses of DR or at least one inpatient or two confirmed outpatient diagnoses of diabetes mellitus with ophthalmic complications from an ophthalmologist from 01/01/2014–30/06/2019

3. At least one documented IVT injection code (EBM) from 01/01/2015–30/06/2019, in association with an ATC- code or OPS-code for an anti-VEGF prescribed within 30 days before the first observed IVT-related EBM code (first EBM code served as the index date)

4. At least one additional IVT administration in combination with an anti-VEGF agent

5. No anti-VEGF prescription from 01/01/2014 until 31 days before index and no IVT injection from 01/01/2014 until one day before index

6. Age of at least 18 years at index

7. No inpatient or outpatient diagnosis of a confounding ocular disease during the entire study period, including AMD, choroidal neovascularization (ICD-10-GM H31.8), as well as retinal vein occlusion

Data analysis

Baseline characteristics and outcomes were analyzed separately for the nAMD and the DR/DME cohorts. Patient profiles were descriptively assessed during the patient-individual 12-month baseline period. Age, sex, and comorbidity status based on the Charlson Comorbidity Index (CCI) and care level were reported. The German care level determines the disability grade of a patient. The higher the care level, the higher the monetary and nursing care benefits [44].

Treatment rates and patterns, as well as persistence outcomes, were examined during the follow-up period. Loading dose conformity was examined and defined as at least three loading doses for nAMD and five loading doses for DR/DME within 21 to 36 days on average. Switching to other anti-VEGF therapies was specified as a change in the agent-specific ATC/OPS code (Supplementary Table 1) within 180 days of the anti-VEGF code previously received. However, the off-label use of drugs like bevacizumab — frequently used off-label for nAMD and DR/DME — was not included in this analysis.

Non-persistence was defined as a 180-day gap between IVT injections documented by EBM codes in the outpatient setting [45, 46]. A gap was observed from the last previously documented EBM code date. It was not a reason for censoring, meaning patients were followed up after a gap to monitor outcomes, including resumption of treatment. In addition, time to therapy discontinuation was assessed during the patient-individual follow-up period through Kaplan–Meier (KM) analysis, and drivers of non-persistence were determined by performing a Cox regression analysis considering age, gender, CCI, average time between preceding anti-VEGF injections, index year, index agent, number of ophthalmologist visits, number of all-cause hospitalizations, and level of care as explanatory variables.

To observe whether disease progression and response to therapy were regularly examined, an effort was made to measure the number and percentage of patients with optical coherence topographies (OCTs) using outpatient EBM and inpatient OPS codes (Supplementary Table 1). However, as OCTs were not introduced to the EBM catalog until October 2019 in Germany, and most OCTs are performed in the outpatient setting, these numbers could only be reported in patients who received their index IVT injection in 2019.

Additionally, efforts were made to assess VA or blindness. In ophthalmology and social law, a person is considered blind if they have a complete loss of sight or if their VA is reduced to less than two percent of normal vision. However, for most patients in the nAMD and the DR/DME cohorts, no information about the VA score could be determined, as no corresponding ICD-10-GM code was identified in the database. Therefore, blindness was reported only in patients with at least one coded VA during the baseline or follow-up period, respectively.

Frequency analysis was performed for all categorical variables and reported by number and percentage for each category. Summary statistics were specified for continuous outcomes, including mean and standard deviation (SD). All reported p-values were based on 2-sided tests, and variables with a corresponding p-value < 0.05 were considered significant.

Regulatory aspects and general considerations

Because of the retrospective nature of this study and the completely anonymized data set used, neither the consultation of an ethical board nor informed patient consent was required. The study protocol was thoroughly reviewed by a scientific steering committee and the data owner. All aspects of data analysis adhered to the regulations set forth by data protection guidelines described in the Social Code (SGB X §75).

All relevant data queries and analyses were performed using Microsoft SQL Server 2019, Microsoft Excel 2305, and Stata version 17.0.

Results

Study cohorts

5,498 patients diagnosed with nAMD and initiating anti-VEGF IVT injections were identified (Supplementary Fig. 1). These patients were, on average (SD), 80.09 (8.03) years old (Table 1) and had a mean (SD) CCI of 3.07 (2.66). The 484 identified patients (Supplementary Fig. 2) with DR/DME were comparatively younger (mean [SD] age: 67.14 [11.07]) but more comorbid, with a mean (SD) CCI amounted to 4.54 (2.46). Patients with nAMD were more frequently female (62.50%) compared to those with DR/DME (female: 41.12%). The most common cardiovascular comorbidities were coronary heart disease (nAMD: 33.18%; DR/DME: 30.99%) and heart failure (nAMD: 21.21%; DR/DME: 24.38%). Further, 83.63% of patients with nAMD and 88.64% of patients with DR/DME were diagnosed with hypertension; 48.31% of patients with nAMD showed dislipidemia compared to 53.10% in the DR/DME cohort. Of all patients with at least one documented VA score (nAMD: n = 556, 10.11%; DR/DME: n = 28, 5.79%), 26.65% of patients with nAMD and 17.65% of patients with DR/DME had records of blindness prior to index.

Table 1.

Baseline characteristics

		Patients with nAMD	Patients with DR/DME
		N = 5,498	N = 484
Age in years (at index date)	Mean (SD)	80.09 (8.03)	67.14 (11.07)
Gender
Male	n (%)	2,062 (37.50)	285 (58.88)
Female	n (%)	3,436 (62.50)	199 (41.12)
Index year (year of first IVT injectio n code)
2015	n (%)	1,501 (27.30)	124 (25.62)
2016	n (%)	1,275 (23.19)	116 (23.97)
2017	n (%)	1,147 (20.86)	95 (16.63)
2018	n (%)	1,135 (20.64)	95 (16.63)
2019 (until June 30)	n (%)	440 (8.00)	54 (11.16)
Follow-up period in months	Mean (SD)	3.97 (1.47)	3.80 (1.56)
Censorship reason
Death	n (%)	1,239 (22.54)	122 (25.21)
End of follow-up	n (%)	4,259 (77.46)	362 (74.79)
Charlson Comorbidity Index (12 months BL)	Mean (SD)	3.07 (2.66)	4.54 (2.46)
Care level (12 months BL)
Low care level (“Pflegestufe” 0&1, “Pflegegrad” 1&2)	n (%)	659 (11.99)	53 (10.95)
High care level (“Pflegestufe” 2&3, “Pflegegrad” 3, 4&5)	n (%)	187 (3.40)	21 (4.34)
Top-5 comorbidities per cohort
I10: Essential (primary) hypertension	n (%)	4,598 (83.63)	429 (88.64)
H35: Other retinal disorders	n (%)	3,768 (68.53)	-*
H52: Disorders of refraction and accommodation	n (%)	2,748 (49.98)	-*
E78: Disorders of lipoprotein metabolism and other lipidaemias	n (%)	2,656 (48.31)	257 (53.10)
M54: Dorsalgia	n (%)	1,807 (32.87)	-*
E11: Type 2 diabetes mellitus	n (%)	-*	465 (96.07)
H36: Retinal disorders	n (%)	-*	382 (78.93)
E14: Unspecified diabetes mellitus	n (%)	-*	285 (58.88)
Diabetes mellitus
E10 Diabetes mellitus type 1	n (%)	-†	117 (24.17)
E11 Diabetes mellitus type 2	n (%)	-†	465 (96.07)
E12-E14 Other/unspecified diabetes	n (%)	-†	292 (60.33)
Cardivascular comorbidities
I20-I25 Coronary heart disease	n (%)	1,824 (33.18)	150 (30.99)
I60-I64 Stroke	n (%)	239 (4.35)	30 (6.20)
I50 Heart failure	n (%)	1,166 (21.21)	118 (24.38)
I46 Cardiac arrest	n (%)	7 (0.13)	1 (0.21)

nAMD neovascular Age-related Macular Degeneration, DR diabetic retinopathy, DME diabetic macular edema, IVT intravitreal, N/n number, SD standard deviation, BL baseline

^* Not among the top-5 comorbidities for this indication; ^† Not assessed for nAMD

Among patients with DR/DME, in the 12 months prior to their index IVT injection, 96.07% had at least one diagnosis of type II diabetes, 24.17% had type I diabetes, and 60.33% had unspecified or other forms of diabetes (patients may have received multiple diagnoses).

Treatments and treatment outcomes

The proportions of included patients with nAMD receiving their first IVT injection in conjunction with anti-VEGF treatment decreased from 2015 (27.30%) to 2018 (20.64%), and patients received either ranibizumab (64.79%) or aflibercept (35.21%) as index agents.

In patients with DR/DME, 25.62% were included in 2015 and 16.63% in 2018. In addition, 55.79% of patients were prescribed ranibizumab, and 44.21% received aflibercept.

Almost all patients (98.31%) assigned to the nAMD cohort received all three IVT loading doses defined for the cohort, and the mean (SD) interval between loading dose injections was 31.14 (9.47) days. 4,258 patients (77.45%) had label conform loading doses with an average interval between injections of 21–36 days. Of patients with nAMD and non-label conform loading doses (n = 1,240), 7.50% did not receive all three doses, while another 60.56% had an average injection interval above 36 days, and 31.94% maintained an average schedule of less than 21 days. At least one post-loading dose, i.e., > 3 injections, was received by 74.83% of patients overall at an average (SD) time interval of 54.69 (19.23) days, and patients had an average (SD) of 6.73 (8.68) post-loading doses.

Moreover, 78.93% of patients with DR/DME received all five IVT loading doses defined for the cohort, with the average (SD) interval between doses being 30.82 (13.76) days. 134 patients (29.55%) had label conform loading doses with an average interval between injections of 21–36 days. Of those with non-confirm loading doses (n = 350), 29.14% did not receive all five doses, while 40.00% of patients had average injection intervals exceeding 36 days and 30.86% maintained an average injection interval of less than 21 days. Post-loading doses, i.e., > 5 injections, were received by 60.74% of patients on an average (SD) schedule of 46.40 (25.04) days, with an average (SD) of 5.08 (8.66) post-loading doses administered.

Most patients in both cohorts maintained their index treatment throughout the follow-up period. In the nAMD cohort, most patients did not switch treatments (78.52%), while 14.48% switched treatment once, and 7.00% underwent two treatment switches. In contrast, a higher percentage (88.64%) of patients with DR/DME did not switch treatments during follow-up, while 8.26% switched treatment once and 3.10% underwent two treatment switches (Fig. 1).

Fig. 1.

Treatment status at 0, 6 and 12 months after index (Sankey diagram). *Please note that two nAMD patients received bevacizumab at 6 and 12 months respectively. In addition, four patients received brolucizumab after 12 months. However, due to the extremely low proportions, the treatment pathways of these could not be shown in the Sankey diagram. nAMD, neovascular Age-related Macular Degeneration; DR, Diabetic Retinopathy; DME, Diabetic Macular Edema; N, number; M, months

In addition to aflibercept and ranibizumab, brolucizumab was administered in 4.00% of patients with nAMD, while 0.15% received a bevacizumab prescription code during the follow-up period. In the DR/DME cohort, only 0.21% of patients received bevacizumab. Other treatments observed during follow-up included verteporfin for the nAMD cohort (0.04%). For patients with DR/DME, dexamethasone and fluocinolone IVT implant refills were received by 15.29% and 1.03% of patients, respectively. Lastly, 40.08% of these patients had at least one OPS code for focal laser treatment during the follow-up period.

Persistence outcomes

During the entire follow-up period, 77.45% of nAMD patients had a gap in any IVT anti-VEGF injections of ≥ 180 days, with 51.38% being non-persistent after 12 months. KM estimates showed a median time-to-treatment gap of 11.28 months (Inter Quartile Range [IQR]: 3.68–32.45). Of those who discontinued, 38.02% resumed treatment during follow-up. For DR/DME, 86.78% had at least one gap of ≥ 180 days until the end of follow-up (62.60% within the first 12 months). The median time to gap was 8.98 (IQR: 2.79–18.54) months, with 29.52% reinitiating treatment (Fig. 2).

Fig. 2.

Time to treatment gap (Kaplan–Meier). *As 27 patients experienced a gap on their index prescription date, only 5,471 patients have been included in the KM analysis. **As four patients experienced a gap on their index prescription date, only 480 patients have been included in the KM analysis. nAMD, neovascular Age-related Macular Degeneration; DR, Diabetic Retinopathy; DME, Diabetic Macular Edema; N, number; CI, Confidence Interval

When conducting the Cox proportional hazards model for nAMD patients, the hazard of non-persistence increased with age (HR: 1.01, p < 0.001), was higher in females (HR: 1.10, p = 0.029), and patients with a registered care level compared to those without (HR low care level: 1.21, p = 0.007; HR high care level: 2.07, p < 0.001). In contrast, the risk decreased with increasing index year, with the lowest hazard in 2019 (HR 2015 vs. 2019: 0.57, p < 0.001). For DR/DME, a high care level (HR: 1.83, p = 0.017) and receiving ranibizumab as an index agent compared to aflibercept (HR: 1.39, p = 0.001) increased the risk of discontinuation (Table 2).

Table 2.

Drivers of non-persistence (Cox regression)

	Patients with nAMD N = 5,498				Patients with DR/DME N = 484
	HR	SE	p-value	95% CI	HR	SE	p-value	95% CI
Age (at index, continuous)	1.01	0.00	< 0.001	1.01–1.02	1.01	0.01	0.221	1.00–1.02
Female gender (reference: male, binary)	1.10	0.05	0.029	1.01–1.20	0.97	0.10	0.797	0.79–1.19
CCI (during 12-month BL, continuous)	1.00	0.01	0.724	0.98–1.01	1.02	0.02	0.309	0.98–1.07
Care level (reference: No care level, categorical)
Low care level	1.21	0.09	0.007	1.05–1.40	1.07	0.19	0.675	0.77–1.51
High care level	2.07	0.25	< 0.001	1.63–2.63	1.83	0.46	0.017	1.12–3.01
Index year (reference: 2015, categorical)
2016	0.87	0.05	0.012	0.77–0.97	0.95	0.13	0.723	0.72–1.25
2017	0.77	0.05	< 0.001	0.69–0.87	0.96	0.14	0.772	0.72–1.28
2018	0.63	0.04	< 0.001	0.55–0.71	1.00	0.15	0.991	0.75–1.34
2019	0.57	0.05	< 0.001	0.47–0.68	1.05	0.19	0.793	0.73–1.50
Index agent Ranibizumab (reference: Aflibercept, binary)	0.98	0.04	0.656	0.90–1.07	1.39	0.14	0.001	1.14–1.70

nAMD neovascular Age-related Macular Degeneration, DR diabetic retinopathy, DME diabetic macular edema, CCI charlson comorbidity index, BL baseline, HR hazard ratio, SE standard error, CI confidence interval, N number

OCTs and long-term outcomes

Among nAMD patients starting anti-VEGF therapy in 2019, 65.45% had received at least one OCT within 12 months after starting treatment, and 80.45% had OCTs throughout follow-up. Among patients deemed non-persistent within the first 24 months of initiating therapy, 69.65% had at least one OCT documented during this timeframe (91.80% among persistent patients). Among DR/DME patients starting anti-VEGF therapy in 2019, 48.15% had received at least one OCT within 12 months, and 66.67% had at least one OCT throughout follow-up. Among patients with DME deemed non-persistent within the first 24 months of initiating therapy, 59.09% had at least one OCT documented during this timeframe (100.0% among persistent patients).

Furthermore, 15.42% of nAMD patients had at least one recorded VA score after their first IVT injection (DR/DME: 7.02%), with 33.25% of those with a recorded measurement experiencing blindness within 12 months (DR/DME: 17.65%) and 67.45% (DR/DME: 44.12%) during follow-up. Additionally, 52.95% of nAMD patients and 45.45% of DR/DME patients had at least one fracture or injury (Supplementary Table 2) during the observational period.

Both cohorts' most common adverse events were cataracts (nAMD: 78.92%; DR/DME: 73.14%) and elevated intraocular pressure (nAMD: 9.29%; DR/DME: 7.85%). Codes for OCT and other long-term outcomes can be found in Supplementary Table 1.

Discussion

This study is the first to provide real-world insights into the non-persistence of IVT injections in patients with nAMD or DR with DME using German billing data while examining possible underlying factors, patient characteristics, and treatment patterns. AOK PLUS claims data were particularly suitable for this study because, unlike other health insurance companies, it has not entered into any selective IVT contracts with outpatient practices. As a result, the invoiced uniform valuation scheme (EBM) codes used to identify IVT treatments reflect all injections performed.

Our study showed that, on average, patients with nAMD were older (mean age: 80.09) and more frequently female (62.50%) compared to those with DR/DME (mean age: 67.14; female: 41.12%). Despite the younger age of DR/DME patients, they had a slightly higher CCI, partly due to diabetic comorbidity (nAMD: 3.07 vs. DR/DME: 4.54). Furthermore, a somewhat higher percentage of DR/DME patients were diagnosed with hypertension (nAMD: 83.63% vs. DR/DME: 88.64%) and dyslipidemia (nAMD: 48.31% vs. DR/DME: 53.10%). These findings are consistent with previous studies, suggesting an onset of DR/DME at a younger age and a tendency towards more male patients in a population with prevalent cardiovascular risk factors, which, in turn, adversely affect disease outcomes [47–49]. Notably, patients included in this real-world study were generally more comorbid and older than those in clinical trials, which must be considered when interpreting the results.

The percentage of patients receiving all loading doses, according to EMA approved posology described in the Summary of Product Characteristics, was high in the nAMD cohort with 98.31% of patients receiving all 3 loading doses. This was about 20% less in patients with DR/DME (78.31%). Notably, 15.05% (51.24%) of nAMD (DME) patients received loading doses with an average injection interval of > 36 or < 21 days, non-conforming to the recommended posology. One reason for the difference in results between the indications could be the requirement of five IVT injections for the DR/DME cohort to adhere to the label recommended posology, compared to three injections required for nAMD patients [15–17]. Future studies should investigate these differences further to better understand what influences treatment adherence and outcomes.

The percentage of nAMD patients receiving their first IVT injection with anti-VEGF treatment declined over time, from 27.30% in 2015 to 20.64% in 2018, reflecting a downward trend in treatment numbers.

In general, non-persistence was high in both patient cohorts, with a non-persistence rate observed within 12 months of IVT initiation of 51.38% (62.60%) for nAMD (DME). According to previous European studies, 12-month non-persistence rates in nAMD patients range from about 4% to 51% [31, 33]. While several studies have explored non-persistence in nAMD, limited research has focused on patients with DR or DME. Weiss et al. stand out as one of the few studies in this context, revealing a 46% treatment interruption rate in patients with DME [48]. Notably, lower non-persistence rates were observed when defined as a lack of ophthalmologist visits rather than a lack of anti-VEGF treatment only [35, 45], which is an important consideration as in this study, non-persistence was defined as a lack of anti-VEGF treatment. The high rate of OCT measurements after index (where available) indicates that while some patients may stop receiving anti-VEGF injections, their care for the condition continues. As treatment goals and reasons for therapy discontinuation are not recorded in claims data, patients who discontinued due to medical reasons such as adverse events or inefficacy were counted as non-persistent in this study. This includes patients who experienced blindness (VA < 2%) after initiating therapy, who, in clinical practice, may switch to supportive care rather than continuing anti-VEGF treatment as treatment effectiveness will be limited in these patients. In our study, 5.13% (1.24%) of nAMD (DME) patients experienced blindness within 12 months of treatment initiation, which may be underestimated, as VA scores were only available for 15% (7%). All this indicates that at least a proportion of patients discontinued therapy for medical reasons rather than patient-related reasons. Accordingly, the non-persistence rates derived in this study represent a conservative upper limit of non-persistence in nAMD/DME.

Several factors contribute to non-persistence – The multifactorial nature of non-persistence to anti-VEGF IVT injections is well-described [27, 29, 31]. Our study reveals that advanced age and female gender significantly contribute to non-persistence in nAMD patients. In both cohorts, higher care levels were associated with treatment gaps, aligning with prior research linking poor health status (connected to higher age) to non-persistence [29, 30]. This aligns with our findings of poorer persistence outcomes in DME patients: The prevalence of systemic comorbidities in this patient group may have contributed to poorer non-persistence outcomes. Prior research highlighted poor health status as a significant factor in treatment discontinuation [29, 31]. Another reason for a higher proportion of non-persistence in DME might be the younger average age in combination with necessitating the balancing of work commitments with physician visits. Previous studies have identified the burdensome nature of frequent and lengthy journeys to physicians as a contributing factor to treatment discontinuation [27, 31, 45]. This could be one of the reasons why about 17% fewer patients with DR/DME received OCTs within 12 months of follow-up compared to patients with nAMD. Additionally, both cohorts exhibited over 70% eye-related comorbidities, alongside a notable high occurrence of fractures and injuries within the 12-month follow-up, suggesting that elderly, highly comorbid patients may struggle to comply with treatment requirements without support. Previous studies showed that lacking a caretaker to help with transportation further compounds the problem [27, 29, 31]. Other factors, such as fear, anxiety about treatment and adverse events, unmet expectations, and dissatisfaction with VA improvement, were mentioned in previous research [27, 31]. Prior clinical studies consistently demonstrated the positive effects of regular anti-VEGF injections on VA in nAMD or DR/DME patients [22–26]. While our study showed a 7% increase in the number of nAMD patients diagnosed with blindness after starting treatment, it can only be speculated whether lowered VA led to discontinuation of therapy or whether blindness was the result of poor therapy adherence/persistence. Addressing the above-mentioned challenges could involve improving information transfer, organizing and coordinating follow-up visits, and exploring therapies with more flexible dosing. Considering each patient's unique needs, IVT therapy should strive to optimize dosing intervals and patient management. Therapies that allow flexible treatment regimens (e.g., treat and extend) can lessen the treatment burden by considering the patients' capacity and willingness to attend regular anti-VEGF appointments and could, therefore, reduce non-persistence [28, 50].

Moreover, our study recorded low bevacizumab prescriptions (< 1% in both cohorts), likely attributed to the fact that this treatment is not specifically approved for nAMD or DME [21]. As we focus on reimbursed prescriptions, off-label use of bevacizumab has not been observed and was considered beyond the scope of this investigation. (Off-label) real-world use of bevacizumab may be higher than reported in this study. However, this should not have affected the non-persistence outcomes since IVT codes and not prescription codes determined non-persistence.

Overall, the multifaceted nature of non-persistence suggests that it is an overlooked aspect of patient care, requiring more attention and investigation. Evaluating patient parameters and needs, particularly in patients who do not attend routine assessments, poses a unique challenge in understanding and addressing non-persistence.

Strengths and limitations

This study's strength lies in the extensive dataset encompassing around 3.5 million insured individuals, providing a representative sample of the German population over 6.5 years. In Germany, around 90% of the population is covered by statutory health insurance [51], which offers comprehensive data collection while avoiding selection bias.

However, it's important to acknowledge certain limitations in our analysis. There's a potential misclassification of diseases, as nAMD is not yet specifically coded in ICD-10-GM during the observational period. However, accurate identification of AMD patients was ensured by combining ICD-10-GM coding and clinical context [52]. A proxy was also needed for DR/DME identification, using DR or diabetic mellitus with ophthalmic complications ICD-10-GM codes in conjunction with anti-VEGF treatment, to ensure accurate cohort identification. Accordingly, DME patients without anti-VEGF treatment were disregarded in our study due to the inability to unambiguously identify these patients.

Since the study focused solely on licensed treatments, the off-label use of bevacizumab, which is not specifically approved for nAMD or DME, was not specifically observed. The dataset of the AOK PLUS is likely to underreport the real-world usage of bevacicumab for IVT in Germany. Moreover, non-persistence outcomes vary based on gap definitions, as seen in previous studies. In this study, the reasons for a gap or discontinuation and the patient's treatment plan are also not observable, posing a potential for overestimating non-persistence. However, given the chronic and progressive nature of nAMD, DR, and DME, physicians are unlikely to opt for treatment termination.

Claims databases, while valuable for routine practice insights, may have missing data and coding errors, especially in outpatient diagnoses. Consequently, long-term outcomes like VA, blindness, visual aid prescriptions, and OCT results were limited. Further, the low number of observed bevacizumab prescriptions suggests gaps in capturing off-label treatment administrations. However, all available results have been reported comprehensively.

Lastly, the specific codes used to identify nAMD, DR/DME patients may reflect German-specific clinical practices and may not be generalizable to other countries.

Conclusion

In conclusion, this study yields valuable insights into the demographic characteristics, treatment patterns, and persistence outcomes of individuals with nAMD or DR with DME. Non-persistence is notably pronounced, especially among older females with nAMD and individuals requiring higher levels of care. To meet individual needs, IVT therapy should adopt flexible regimens, considering patients' capacity and willingness for regular anti-VEGF appointments, potentially reducing non-persistence. These findings significantly enhance our understanding of current practices and offer a foundation for future initiatives to improve patient care. Nevertheless, further research is warranted to explore the underlying factors contributing to non-persistence and to compare and categorize various approaches for assessing treatment discontinuation.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

Anti-VEGF

Anti-Vascular Endothelial Growth Factor

ATC

Anatomical Therapeutic Chemical code

CCI

Charlson Comorbidity Index

DME

Diabetic Macular Edema

DR

Diabetic Retinopathy

EBM

Uniform valuation scheme (“Einheitlicher Bewertungsmaßstab”)

EMA

European Medicines Agency

HR

Hazard Ratio

ICD-10-GM

International Classification of Diseases, 10th revision, German Modification

IVT

Intravitreal Therapy

IQR

Inter Quartile Range

KM

Kaplan-Meier

nAMD

Neovascular Age-related Macular Degeneration

OCT

Optical Coherence Tomography

OPS

Operations and Procedures key (“Operationen und Prozedurenschlüssel “)

SD

Standard Deviation

SGB

Social Code (Sozialgesetzbuch)

VA

Visual Acuity

Author contribution

Julia Krieger was an employee of Cytel Inc. at the time of study conduct. Cytel Inc. was contracted by F. Hoffmann-La Roche Ltd. to conduct the research.

Oliver Cox is an employee of F. Hoffmann-La Roche Ltd. and has nothing to declare.

Jan-Paul Flacke and Lena Beilschmidt are employees of Roche Pharma AG and have nothing to declare.

Sabrina Mueller is an employee of IPAM e.V. and has nothing to declare.

Ulf Maywald received honoraria from Roche for advisory boards.

Michael Janusz Koss has the following to declare for the past 12 months: Roche: paid lectures; Heidelberg Engineering: paid lectures; Glaukos: paid lectures; Alcon: paid lectures.

Funding

Data collection, statistical analysis and medical writing support were provided by Cytel, Inc. (Berlin, Germany), which was funded by F. Hoffmann-La Roche AG (Roche).

Declarations

Ethical approval

Due to the non-interventional, retrospective nature of the analyzed data and because our analysis involved an anonymized dataset, neither ethical review nor informed consent of the patients was required according to German national legislation (§ 75 SGB X). As the responsible authority, the involved sickness fund (AOK PLUS) approved the use of the data for the purpose of this study.

Conflict of interest

Oliver Cox, Jan-Paule Flacke, and Lena Beilschmidt are employees of Roche. Julia Krieger was an employee of Cytel. Sabrina Mueller is an employee of IPAM. Cytel and IPAM work was financially supported by Roche. Ulf Maywald has no potential conflict of interest, except those potentially related to his employer, AOK PLUS. Michael Janusz Koss is an employee of the Augenklinik Herzog Carl Theodor and has declared no conflicts of interest.