To perform 11- and 2-year health care sector (ophthalmic) and societal cost perspective reference case, cost-utility analyses comparing bevacizumab, ranibizumab, and aflibercept monotherapies for neovascular age-related macular degeneration (NVAMD).
The authors performed 11-year and 2-year ophthalmic and societal cost perspective, cost-utility analyses comparing bevacizumab, ranibizumab, and aflibercept monotherapies for neovascular age-related macular degeneration (NVAMD). We employed patient utilities, bilateral outcomes, 2018 U.S. dollars, vision-related mortality, a Medicare fee schedule, and CATT (Comparison of Age-Related Macular Degeneration Treatments) study and VIEW (VEGF Trap-Eye: Investigation of Efficacy and Safety in Wet AMD) trial. Cochrane data were also used. S etting : Center for Value-Based Medicine. P atient / study population : patients with NVAMD. I ntervention : Cost-utility analyses using published data. Data-modeled 10-year vision outcomes were modeled forward to year 11. M ain outcome measurement : These included cost-utility ratios (CURs), costs, and quality-adjusted life-years (QALYs) gained. $100,00/QALY was considered the US cost-effectiveness upper limit.
Bevacizumab and ranibizumab each conferred an 11-year, 1.339 QALY gain versus observation. Aflibercept conferred a 1.380 QALY gain. Aflibercept conferred greater QALY gain for less cost than ranibizumab but was not cost-effective compared to bevacizumab ($1,151,451/QALY incremental CUR). The average ophthalmic cost perspective CUR for bevacizumab was $11,033/QALY, $79,600/QALY for ranibizumab, and $44,801/QALY for aflibercept. Eleven-year therapies saved a 1.0 year-of-life loss without treatment from the 11.0-year life expectancy. Early treatment was 138%-149% more cost-effective than late treatment. Two-year therapy prevented a 1-month-of-life loss, and revealed bevacizumab, ranibizumab, and aflibercept conferred 0.141, 0.141, and 0.164 QALY gains, respectively, with corresponding average CURs of $40,371/QALY, $335,726/QALY, and $168,006/QALY, respectively.
From an ophthalmic (medical) cost perspective, bevacizumab, ranibizumab, and aflibercept NVAMD monotherapies were all cost-effective over 11 years, with bevacizumab 6.21× more cost-effective than ranibizumab and 3.06× more cost-effective than aflibercept. Two-year modeling revealed bevacizumab was cost-effective, whereas ranibizumab and aflibercept were not. Early treatment was critical for obtaining optimal vision and cost-effectiveness, as is long-term follow-up and adherence to treatment.
Bevacizumab, ranibizumab, and aflibercept have similar patient preference-based vision and safety outcomes for treating wet age-related macular degeneration (AMD).
The 3 study VEGF inhibitor therapies are all cost effective over 11 years and prevented loss of life from untreated wet AMD.
Bevacizumab therapy cost effectiveness was found to be 621% greater than ranibizumab therapy and 306% greater than aflibercept therapy.
U.S. Food and Drug Administration (FDA)-approved vascular endothelial growth factor-inhibitor (VEGF-I) monotherapies in the United States for treating neovascular age-related macular degeneration (NVAMD) include intravitreal ranibizumab (Lucentis, Genentech-Roche, South San Francisco, California; approved 2006), , intravitreal aflibercept (Eylea, Regeneron, Eastview, New York; approved 2011), , and intravitreal brolucizumab (Beovu, Novartis, Basel, Switzerland; approved 2019). A fourth drug, bevacizumab (Avastin, Genentech-Roche, South San Francisco, California), has not been approved by the FDA but has been shown in a multicenter National Eye Institute-sponsored clinical trial to be therapeutically equivalent to ranibizumab and has assumed widespread acceptance within the vitreoretinal community. , Brolucizumab was not included in the present study because follow-up data were only available through 48 weeks.
Ranibizumab for treating subfoveal NVAMD (n = 712), treating predominantly classic NVAMD (n = 423), a continuation of these 2 trials, a large study comparing aflibercept and ranibizumab (n = 2,457), , the multicenter CATT (Comparison of Age-Related Macular Degeneration Treatments Trials) (n = 1,185) study comparing bevacizumab and ranibizumab, , meta-analyses comparing bevacizumab and ranibizumab, and ranibizumab and aflibercept, , Cochrane Database Systematic Reviews comparing aflibercept, bevacizumab and ranibizumab, , and a large U.S. database comparing aflibercept and ranibizumab showed that monotherapy with each of the 3 VEGF-I drugs yielded similar vision results. Nonetheless, the longest randomized clinical trial portion of any study was 24 months. , , , A review of the bevacizumab, ranibizumab, and aflibercept therapy regimens using real-world data from 13,859 patients in the American Academy of Ophthalmology IRIS (Intelligent Research in Sight) registry showed that monotherapy using each of the 3 VEGF-I drugs yielded equivalent vision results at 1 year. Although some studies have found minor discrepancies, adverse event profiles for the 3 drugs have also been shown to be similar. , ,
An American Academy of Ophthalmology Ophthalmic Technology Assessment Committee recently evaluated the 3 VEGF inhibitors for NVAMD. Although the vision outcomes were similar among the drugs at 24 months, the Committee believed that longer term follow-up was needed.
The HORIZON (Open-Label Extension Trial of Ranibizumab for Choroidal Neovascularization Secondary to Age-Related Macular Degeneration) study enrolled predominantly 2-year MARINA (Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular Age-Related Macular Degeneration) and ANCHOR (Anti-VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in Age-Related Macular Degeneration) patients to 2 additional years of as-needed ranibizumab therapy at the discretion of the investigator from months 25 to 48. The eyes receiving ranibizumab for 4 years averaged a mean letter gain of +2.0 from months 0 to 48 and a −0.1 letter for some patients followed 0-60 months. The SEVEN-UP (Seven-Year Observational Update of Macular Degeneration Patients Post-MARINA/ANCHOR and HORIZON Trials) study, noted that eyes treated with >11 injections of ranibizumab from month 49 to a mean of 7.3 years (88 months) gained a mean 3.9 letters according to Early Treatment Diabetic Retinopathy criteria over that time, whereas eyes receiving 6-10 post-HORIZON injections lost a mean of 6.9 letters. Combining the 2 sub-cohorts resulted in a loss of −0.6 letters from months 49 to 88 and a mean vision outcome of 20/63-2. The SEVEN-UP study was able to recall and examine 65 of 357 patients (18%) who completed the 2-year MARINA or ANCHOR trial, then the additional 2-year HORIZON follow-up study. More recent 10-year data for patients treated with 10 injections per year for 10 years in the study by Suner and associates revealed a mean vision outcome of 20/63-2 from years 6-10, virtually identical to late SEVEN-UP study outcomes.
Cost-effectiveness analyses of each of the 3 medications in the treatment of NVAMD have been performed, but the authors were unaware of a U.S. cost-effectiveness analysis comparing the 3 together. A recent study compared the long-term, historical drug costs of the more expensive ranibizumab and aflibercept versus bevacizumab. The authors noted that, from 2008 to 2015 in the U.S. Medicare Fee-for-Service population, $13.8 billion would have been saved by Medicare and $3.5 billion would have been saved by patients if bevacizumab had been substituted for ranibizumab and aflibercept for NVAMD therapy.
These authors were unaware of any cost-utility analyses comparing the use of bevacizumab, ranibizumab, and aflibercept monotherapy for NVAMD using 1) ophthalmic patient utilities, 2) bilateral treatment outcomes and costs, 3) an average, national Medicare Fee Schedule cost basis, 4) vision loss mortality data, and 5) ophthalmic (direct medical) and societal cost perspectives. Therefore, the current investigation was undertaken.
Subjects and Methods
Reference case, ophthalmic cost perspective and societal cost perspective, and average and incremental cost-utility analyses were performed for intravitreal bevacizumab therapy, ranibizumab therapy, and aflibercept therapy for the treatment of NVAMD. Wills Eye Hospital Institutional Review Board (IRB) approval was waived because no new patients were enrolled or identified patient data were used. The research adhered to the Declaration of Helsinki, and no state or federal regulations were violated.
A list of cost-utility analysis parameters is shown in Supplemental Table 1 . Clinical study parameters are shown in Supplemental Table 2 . , ,
The methodology used for this study has been described previously , , and agrees with recommendations of the Second Panel on Cost-Effectiveness in Health and Medicine. CATT data , were used for reference case analysis. CATT was a National Eye Institute-supported multicenter, randomized clinical trial comparing intravitreal bevacizumab and ranibizumab therapy for NVAMD therapy. VIEW (VEGF Trap-Eye: Investigation of Efficacy and Safety in Wet AMD) 1 and 2 trial data, , modeled after ranibizumab trials, , (as was CATT , ) were used to compare ranibizumab and aflibercept therapies.
Vision: 2-Year Analysis
According to clinical trials, , , , meta-analyses, , Cochrane Database Systematic Reviews, , a U.S. database, and a large IRIS registry study, the vision results of the 3 drugs under study were the same through the 24 months of randomized clinical trial data.
Vision: 11-Year Analysis
There were statistical differences at 2 years after baseline between the CATT drug cohorts that received monthly injections versus as-needed injections. Combining the ranibizumab and bevacizumab cohorts, there was a 24-month, 2.4 greater letter gain with monthly injections than with the as-needed injections ( P = .046). The present authors therefore modeled the more visually favorable monthly injection regimen for bevacizumab and ranibizumab during the first 24 months after baseline treatment. Because it was noted in the VIEW trials that bimonthly aflibercept injections yielded vision outcomes equivalent to monthly aflibercept and ranibizumab injection results after 3 initial monthly injections, our study modeled half the administration frequency for aflibercept versus ranibizumab and bevacizumab after the first 24 months ( Supplemental Table 2 ). , , ,
For months 25-60 for ranibizumab and bevacizumab, vision results were modeled after the 5-year CATT study results, which were very similar to those of the HORIZON trial, an open-label extension of ranibizumab for the treatment of NVAMD, and the 5-year follow-up, Australian retrospective review. Averaging these results resulted in a 5-year outcome of 20/63-1.
The authors assumed that, during this period, aflibercept was administered with half the frequency of ranibizumab and bevacizumab. Table 1 shows the vision treatment results, their evidence-based vision basis, and the untreated control cohort vision results. ,
|Time||Bevacizumab, Ranibizumab, or Aflibercept Monotherapy||No Treatment (Control Cohort)|
|One month||20/50 −1||20/63 −1|
|Three months||20/50 +2||20/80 +2|
|Six months||40/40 −2||20/80 −1|
|12 months||20/40 −1.5||20/125 −2|
|2 years||20/40 −1.5||20/200|
|3 years||20/50 +1||20/250 +1|
|4 years||20/50 −2||20/250 −2|
|5 years||20/63 −1||20/320|
|6 years||20/63 −2||20/400|
|7 years||20/63 −2||20/500|
|8 years||20/63 −2||20/500 −2|
|9 years||20/63 −2||20/630 +1|
|10 years||20/63 −2||20/630|
|11 years||20/63 −2||20/630|
|Treatment References for VISION||Control references for Vision|
|Years 0 to 2: CATT study ,||Years 0 to 2: CATT study ,|
|Years 3 to 5.0: CATT study |
Australian 5-year study
|Years 3 to 11: Shah and Del Priore meta-analysis of 6 Macular Photocoagulation Study trial control arms|
|Years 5.1 to 7.4: SEVEN-UP Study|
|Years 7.5 to 11: SEVEN-UP Study, last observation carried forward|
|Years 5-11: Suner et al. 10-year, real-world, fixed-interval treatment—vision outcome, the same as in the SEVEN-UP study|
From months 61-88 (years 5-7.4), vision results were modeled after the SEVEN-UP (post-MARINA/ANCHOR and HORIZON trials) study for all 3 drugs. It should be noted that the authors biased against more favorable SEVEN-UP vision results by integrating data from the top 2 sub-cohort vision outcomes rather than just the top cohort of 11 or more injections; the latter was used to model VEGF-inhibitor-treated subjects in our cost-utility analysis. This amalgamation resulted in a letter loss of 0.6 from the end of year 5 to year 11, with a resultant vision of 20/63-2 during that time. From months 89-132, we used a last-observation-carried-forward model from the SEVEN-UP vision results at 88 months (7.4 years). Months 61-132 were modeled with the 10-year treatment described in the study by Suner and associates, with vision outcomes identical to those in SEVEN-UP through 88 months.
After 5 years, the authors arbitrarily assumed that bevacizumab and ranibizumab were administered 3 times a year from years 6 to 8 and twice yearly from years 9 to 11 ( Supplemental Table 2 ). Thus, 11 bevacizumab and ranibizumab injections were given after 4 years, the minimum number needed in SEVEN-UP to obtain the best visual result. Aflibercept injections from years 3 to 11 were given half as often as ranibizumab and bevacizumab.
For controls, the authors used 2-year, MARINA, and ANCHOR sham-treatment, ranibizumab trial data for the 2-year analysis, , followed by a Shah and Del Priore Lineweaver-Burke plot meta-analysis of the natural history of untreated NVAMD for years 3 to 11 in the 11-year analysis. Shah and Del Priore integrated the untreated control cohort vision results from 6 previous, randomized clinical trials. Those authors found that mean vision in a cohort correlates highly with the time since the development of NVAMD. The visual acuities at specific times are shown in Table 1 .
The reference case herein was based primarily upon the ophthalmic cost perspective cost-utility ratio (CUR) which was calculated by dividing the ophthalmic direct medical costs expended for an intervention by the quality-adjusted life-years (QALYs) gained. The cost-utility methodology was similar to that used in other value-based medicine ophthalmic cost-utility analyses using patient utilities, an average national Medicare Fee Schedule, and a 3% annual discount rate for QALY outcomes and costs (the last recommended by the Second Panel on Cost-Effectiveness in Health and Medicine ). The time tradeoff utilities were obtained from a visual utility database of 1,400 ophthalmic patient interviews. The utilities have been validated, are reproducible, correlated most highly with vision in the better-seeing eye, and are typically unaffected by comorbidities, sex, age, ethnicity, and Western country of origin. They are generally, but not always, unaffected by the underlying cause of vision loss. ,
The authors also integrated newer data for vision-related mortality into the current analysis (see Mortality below).
A combined-eye model cost basis was used. , , , This model included a weighted average of first-eye and second-eye models. , , The second-eye model assumes that the first eye already has vision loss from NVAMD or another cause. Thus, patient value gain begins immediately at the time of initiation of therapy in the second eye.
With the first-eye model, it is assumed that the fellow eye has good vision. We have been unable to demonstrate a significant improvement in quality-of-life (QOL) when vision improves in 1 eye (e.g., 20/200 to 20/63 or 20/63 to 20/40) and the fellow eye has normal vision. , In that instance, QALYs are gained when the second eye with good vision eventually converts to NVAMD and is treated. The authors have modeled using CATT data showing that 60% of cases at baseline underwent second-eye treatment (second-eye model) and that 40% had the first eye treated (first-eye model), with good vision at baseline in the fellow eye. With the first-eye model, fellow eyes were treated if they converted from atrophic AMD to NVAMD over the 2- or 11-year model ( Table 2 ). With the first-eye model, QALY gains were assumed to occur only when second eyes were treated for conversion for NVAMD and experienced decreased vision.
|Time (beginning of year)||First-Eye Model Total: (Fellow Eye With Good Vision and no NVAMD)||Percent of Remaining Fellow Eyes in First-Eye Model Converting to NVAMD Over Time||Second-Eye Model (First Eye Already With NVAMD)|
The authors did not use a methodology that has been noted by some that decreases the interventional utility gain in individuals who are elderly or have disabilities, thus biasing against these groups. That methodology has little chance of becoming public policy in the United States.
Using the ophthalmic cost perspective, the direct ophthalmic costs expended typically result in a positive CUR because the ophthalmic direct medical costs expended generally exceed ophthalmic direct medical costs saved (e.g., low-vision costs). Use of a societal cost perspective more commonly has a negative CUR because the direct ophthalmic medical costs expended are much more likely to be exceeded by the costs saved (caregiver costs, medical costs such as trauma and depression, wage loss prevented, and others). Thus, the overall costs associated with the intervention are negative. A negative CUR can also be encountered when one intervention dominates another, meaning that it delivers greater QALY gain for less cost than the comparator intervention.
Improvement in QALY and/or Length of Life Gains: Mean vision measurements in the VEGF-I cohorts and control cohorts each year were converted to utility format, ranging from a utility of 0.80 for 20/40 vision in the better-seeing eye to 0.538 associated with 20/630 vision, to calculate the yearly QALY accrual associated with therapy versus none ( Table 3 ). , The QALY loss from adverse events was subtracted from the total QALY accrual in each treated eye. Adverse events included cases of endophthalmitis per 1,756 intravitreal injections, , 1 day of post-injection ocular discomfort (utility = 0.89) and 2 days of post-injection ocular erythema (utility = 0.96). The 11-year QALY loss from adverse events was 0.070 QALY per eye in the bevacizumab and ranibizumab cohorts and 0.035 QALY per eye in the aflibercept cohort receiving fewer intravitreal injections. For the 2-year model, the respective adverse event QALY losses were 0.065 per eye and 0.032 per eye. Systemic adverse events, such as death and cardiovascular events, were not considered due to the similar incidence rates among the 3 drugs and the uncertainty as to whether they exceeded those found in an age-matched population without NVAMD. , ,
|Scenario||Bevacizumab QALY Gain (Value Gain)||Ranibizumab QALY Gain (Value Gain)||Aflibercept QALY Gain (Value Gain)|
|First-eye, 11-year model, QALY gain with mortality factored in||0.425 (8.3%)||0.425 (8.3%)||0.465 (9.0%)|
|Second-eye, 11-year model QALY gain with mortality factored in||1.948 (37.9%)||1.948 (37.9%)||1.990 (38.7%)|
|Combined-eye, 11-year model QALY gain with mortality treatment decrease factored in (60% 2nd-eye model + 40% 1st-eye model)||1.339 (26.1%)||1.339 (26.1%)||1.380 (26.9%)|
|Control cohort, 10.0-year model, QALY accrual with increased mortality factored in||5.099||5.099||5.099|
|Combined-eye model QALY gain with 10.0-year life expectancy||1.080 (21.1%)||1.080 (21.1%)||1.114 (21.7%)|
|QALY gain from increasing life expectancy from 10.0 years to 11.0 years, combined eye model||0.259 (5.0%)||0.259 (5.0%)||0.266 (5.2%)|
|Combined-eye, 2-year model with mortality factored in||0.141 (10.4%)||0.141 (10.4%)||0.157 (12.1%)|
Christ and associates calculated the risk of premature death associated with vision loss in the better-seeing eye ( Supplemental Table 3 ). Their comprehensive methodology and statistical analysis are well described in a study of 2,520 ophthalmic patients with baseline ages of 65-84 years and 20-year follow-up in the Salisbury Eye Evaluation Study. Referent to 20/20 vision with a hazard ratio of 1.0 for the chance of dying within 8 years, the hazard ratio for 20/40 vision was 1.03, 1.08 for 20/80, and 1.18 for 20/200. The calculation by Christ and associates in our present study demonstrated that, for 20/630 vision, the hazard ratio was 1.33. Thus, there is a 33% higher chance that someone with 20/630 vision will die within 8 years than a person with 20/20 vision. When the annual mean vision changes between the present cohorts treated with VEGF-I and those of the control cohort were analyzed ( Table 1 ) and correlated with the hazard ratios, the mean treated patient in each of the bevacizumab, ranibizumab, and aflibercept cohorts had a life expectancy of 11.0 years, whereas the mean person in the untreated control cohort was calculated to have a mean life expectancy of 10.0 years. Thus, VEGF-inhibitor monotherapy resulted in the prevention of approximately 1.0 year of life lost for the average untreated NVAMD patient, a 0.259 QALY gain for bevacizumab and ranibizumab and a 0.266 QALY gain for aflibercept, each consisting of 19.2% of the total QALY gains, respectively, of 1.339 and 1.380, respectively. In the 2-year model, VEGF-inhibitor therapy prevented 1 month of life loss over 2 years while gaining 8.1% of the total bevacizumab and ranibizumab QALY gains and 6.6% of the QALY gain associated with aflibercept therapy.
The ophthalmic direct medical costs (paid by insurers and patient out-of-pocket dollars) were taken directly from the 2018 national average Medicare fee schedule ( Table 4 ). Although only 1 eye was enrolled in the clinical trials studied, the authors also included the costs associated with treating fellow eyes with atrophic AMD (age-related macular degeneration) that converted to NVAMD with the same drug during the 11-year ( Table 5 ) and 2-year ( Table 6 ) models to simulate clinical practice. The direct nonophthalmic medical costs were taken from a study by Javitt and associates, and the direct nonmedical costs (caregiver costs) were taken from a study by Brown and associates, and the indirect medical costs were taken from Brown and associates, The U.S. Bureau of Labor Statistics, and the Household Economic Studies from the US Census Bureau. Direct nonmedical, or caregiver, costs included 27% for paid caregivers and 73% for unpaid caregivers, similar to what has been noted by others.
|National Average Medicare Fee Schedule Costs|
|Entity (2018 U.S. Nominal $ Cost per 1 Service)||Medicare HCPSC #||Bevacizumab (1.25 mg/Dose)||Ranibizumab (0.5 mg/dose)||Aflibercept (2.0 mg/Dose)|
|Initial eye examination ($154)||92,004||$154||$154||$154|
|Office examination ($129)||92,014||$2,066||$2,066||$6,226|
|Optical coherence tomography ($42)||92,134||$2,593||$2,593||$2,593|
|Fundus photographs ($58)||92,250||$114||$114||$114|
|Fluorescein angiography ($88)||92,235||$173||$173||$173|
|Intravitreal injection ($104)||67,028||$4,879||$4,879||$2,423|
|Drug ASP cost a||See below||$3,717||$87,727||$45,005|
|Total cost per treated baseline eye||NA||$13,565||$97,573||$56,546|
|Direct ophthalmic medical cost per fellow eye converting to NVAMD||NA||$1,207||$9,009||$5,266|
|Total direct ophthalmic medical cost per average patient||NA||$14,772||$106,582||$61,811|
|Bilateral drug cost/direct ophthalmic medical cost per average patient||NA||27.4%||89.7%||79.3%|
|Total cost per treated second-eye model eye||NA||$5,526||$45,961||$25,584|
|Direct ophthalmic medical cost per fellow eye converting to NVAMD||NA||$410||$3,395||$1,972|
|Total direct ophthalmic medical cost per treated first-eye model eye patient||NA||$5,937||$49,356||$27,557|
|Total direct ophthalmic medical cost per weighted average of first-eye model eye (40%) and second-eye model eyes (60%)||NA||$5,690||$47,319||$26,377|