Input From the Clinical Experts Consulted by CADTH

The clinical expert consulted by CADTH indicated that the treatment goals of DME are to delay and, in some cases, reverse the progression of DME or DR, as well as to improve vision-related and general quality of life. Because most patients are currently required to attend treatment visits once every 1 to 3 months, the clinical expert noted that there is an unmet need for effective treatments that can be administered at longer treatment intervals, reducing the burden on patients and caregivers associated with frequent treatment visits and increasing adherence with treatment regimens.

The clinical expert noted that faricimab is expected to have a place as a first-line or later-line treatment in patients with DME, similar to other anti-VEGF drugs. The clinical expert indicated that if faricimab is reimbursed, a shift in the treatment paradigm is likely, given that faricimab is the first anti-VEGF approved for an extended interval of up to 16 weeks, which could help address the burden of frequent treatment visits. The clinical expert noted that the dual mechanism of faricimab, which targets both the VEGF-A and Ang-2 pathways, is particularly relevant to DR.

The clinical expert noted that patients with DR associated with vision loss secondary to centre-involved DME are suitable candidates for faricimab. The clinical expert indicated that faricimab can be used in patients who are treatment-naive or who require a change in therapy due to inadequate responses to other anti-VEGF drugs. Patients who may not be suitable for treatment include those who present with major structural damage to the macular retina (e.g., macular atrophy or fibrosis), according to the expert.

The clinical expert noted that clinical evaluation and optical coherence tomography (OCT) should be performed at dosing visits to determine prognosis and follow-up. Key assessment outcomes include change in visual acuity, retinal thickness, and the presence of retinal fluid. According to the expert, an optimal response to anti-VEGFs is generally achieved 6 to 12 months after the initiation of therapy.

The clinical expert indicated that faricimab should be discontinued in patients with treatment futility and proof of irreversible anatomic or functional damage, such as those with macular atrophy (schema) and fibrosis.

Regarding prescribing conditions, the clinical expert recommended retina subspecialty care as the most appropriate setting for the prescription and administration of faricimab in urban areas; in rural settings, trained comprehensive ophthalmologists with experience and expertise in the management of DME would be sufficient.

Clinician Group Input

CADTH received input from 1 clinician group: the Canadian Retina Society.

The clinician group input was consistent with the clinical expert CADTH consulted with respect to the unmet need for a durable treatment with fewer injections that could reduce treatment burden while maintaining maximal vision gain. The clinical group also noted the importance of minimizing side effects, such as injection-related complications, including inflammation, infection, bleeding, retinal detachment, cataract, and glaucoma.

Clinician group input was consistent with the clinical expert input regarding the potential place in therapy for faricimab and the suitable patient population. The clinician group also noted that patients without centre-involved DME should not be treated with faricimab, and those without vision loss secondary to DME (pre-symptomatic patients) should be monitored as long as very close follow-up can be maintained.

Clinician group input generally aligned with the clinical expert input on the assessment of response to treatment and discontinuation of treatment. The clinician group noted that clinically meaningful outcomes include improvement in vision, reduction or resolution of macular edema, regression in Diabetic Retinopathy Severity Scale (DRSS) score, and reduction in the frequency of treatment (intervals of 4 months or longer between treatments).

The clinician group broadly identified the setting for treatment administration as ophthalmology offices in the community setting and/or hospital setting.

Description of Studies

The YOSEMITE and RHINE studies met the inclusion criteria for the systematic review. They were identically designed phase III, multi-centre, randomized, double-blind, active-controlled, noninferiority trials that compared faricimab with aflibercept in patients with DME (YOSEMITE, n = 940; RHINE, n = 951) over 100 weeks. Patients were randomized in a 1:1:1 ratio to 1 of 3 arms: fixed-dose faricimab every 8 weeks; faricimab dosing on a personalized treatment interval (PTI); and fixed-dose aflibercept every 8 weeks. Patients in the 8-week faricimab arm received faricimab 6 mg intravitreally every 4 weeks for 6 loading doses, followed by maintenance doses every 8 weeks. Patients in the PTI faricimab arm received faricimab 6 mg intravitreally every 4 weeks for 4 loading doses, after which maintenance doses could be administered every 4, 8, 12, or 16 weeks, depending on patient outcome, determined by a predefined algorithm. Patients in the aflibercept arm received aflibercept 2 mg intravitreally every 4 weeks for 5 loading doses, followed by a fixed maintenance interval of every 8 weeks.

Both studies aimed to establish the noninferiority of faricimab to aflibercept for the primary outcome, which was change from baseline in best corrected visual acuity (BCVA) (measured using the Early Treatment Diabetic Retinopathy Study [ETDRS] chart) averaged over weeks 48, 52, and 56 in the intention-to-treat (ITT) population. The noninferiority margin was specified as 4 letters on the ETDRS chart. The proportion of patients with improvement from baseline of 2 or more steps on the ETDRS DRSS score at week 52 was a key secondary end point. The noninferiority margin for this outcome was specified as a difference of 10% between treatment arms. Other secondary outcomes included the frequency of administration for faricimab at a PTI, retinal thickness, presence of retinal fluids, and measures of HRQoL and vision-related function, all of which were analyzed without control for multiplicity. The primary analysis was conducted at week 56, and secondary analysis data were available up to week 100.

The baseline demographic and ocular characteristics of patients were, overall, balanced in the treatment arms in each study. The baseline characteristics were generally similar in the 2 studies, except median months since DME diagnosis was shorter for patients in the YOSEMITE trial than in the RHINE trial (3.1 months versus 6.6 months) and mean baseline central subfield thickness (CST) was slightly higher for patients in YOSEMITE than in RHINE (487.5 µm versus 471.6 µm). In both studies, the median age of patients at baseline was 62 to 64 years, and the majority were male (> 57%) and White (> 76%). At the start of the studies, most patients (around 70% to 75%) had DRSS scores of 35 to 47 (mild to moderately severe nonproliferative DR with an anti-VEGF drug.

Efficacy Results

A summary of the key efficacy results is provided in Table 2.

Change in Visual Acuity

The primary outcome of both studies was the change from baseline in BCVA (ETDRS letters) averaged over weeks 48, 52, and 56 in the ITT population. In the YOSEMITE trial, the mean difference in change between the 8-week faricimab group and the aflibercept group was –0.2 letters (97.5% confidence interval [CI], –2.0 to 1.6 letters), and between the PTI faricimab group and the aflibercept group was 0.7 letters (97.5% CI, –1.1 to 2.5 letters). In the RHINE trial, the mean difference in change between the 8-week faricimab group and the aflibercept group was 1.5 letters (97.5% CI, –0.1 to 3.2 letters), and between the PTI faricimab group and the aflibercept group was 0.5 letters (97.5% CI, –1.1 to 2.1 letters). The CIs for all these estimates did not cross the pre-established noninferiority margin of 4 letters. All the CIs in these comparisons crossed the line of no effect and, therefore, neither faricimab arm was superior to aflibercept for the change in BCVA. Results of the sensitivity analyses, the supplemental analyses, and the per-protocol population were congruent with the primary analysis.

The proportion of patients gaining 15 or more ETDRS letters in BCVA from baseline averaged over weeks 48, 52, and 56 (a secondary outcome) was comparable across treatment arms and studies: 29.2%, 35.5%, and 31.9% in the 8-week faricimab, PTI faricimab, and aflibercept groups, respectively, in YOSEMITE; and 33.6%, 28.3%, and 30.5%, respectively, in RHINE. Most patients (> 95% across treatment arms) avoided a loss of 15 or more ETDRS letters in BCVA from baseline during the studies. Comparable results were seen across all 3 treatment arms for patients gaining 10 or more, 5 or more, or 0 or more letters, and for patients avoiding a loss of 10 or more or 5 or more letters in BCVA from baseline in the 2 studies; these end points were other secondary outcomes in the trials.

Results were mostly congruent at year 2 and year 1 for these BCVA secondary outcomes, except a numerically lower adjusted proportion of patients in the PTI faricimab arm than in the aflibercept arm in RHINE gained 15 or more ETDRS letters in BCVA from baseline averaged over weeks 92, 96, and 100.

Table 2

Summary of Key Results From Pivotal and Protocol Selected Studies.

Harms Results

A summary of the key harms results is provided in Table 2.

Over 100 weeks in the safety-evaluable population in the YOSEMITE trial, the proportion of patients reporting at least 1 ocular adverse event (AE) in the study eye was comparable across treatment arms (47.0% in the 8-week faricimab arm, 46.6% in the PTI faricimab arm, and 46.3% in the aflibercept arm). In the RHINE study, a higher proportion of patients in the 8-week faricimab and PTI faricimab arms reported an ocular AE than in the aflibercept arm (52.4%, 51.7%, and 44.6%, respectively). In the RHINE study, the ocular AEs that occurred at a higher incidence in the 2 faricimab arms than in the aflibercept arm include cataract, dry eye, and blepharitis, and the ocular AEs that were numerically more common in the 8-week faricimab arm than in the aflibercept arm include conjunctival hemorrhage, increased intraocular pressure, vitreous floaters, cataract subcapsular, posterior capsule opacification, eye pruritis, and conjunctivitis allergic. The most common ocular AEs in both studies were cataract (9.9% to 17.6% in each treatment arm) and conjunctival hemorrhage (5.6% to 9.8% in each arm).

Ocular serious adverse events (SAEs) were reported with low frequency in both trials; however, there was a slightly higher frequency of ocular SAEs in the PTI faricimab arm than in the aflibercept arm in both YOSEMITE and RHINE, and the 8-week and PTI faricimab arms were somewhat higher than the aflibercept arm in YOSEMITE (3.8%, 4.5%, and 2.3%, respectively) and in RHINE (4.4%, 6.3%, and 4.1%, respectively). The most common ocular SAE reported in both studies was cataract (0.6% to 2.2% across treatment arms). The frequency of nonocular SAEs in any arm of the studies ranged from 20.1% to 31.6%, with COVID-19 (1.3% to 3.2%) and pneumonia (1.3% to 2.6%) being the most frequently reported.

In both studies, a small proportion of patients in all arms discontinued treatment due to AEs (1.6% to 2.9% per arm). The most common AE (≥ 1% in any arm) related to treatment discontinuation was uveitis (3 patients in the PTI faricimab arm of YOSEMITE). The proportion of patients in all arms that discontinued the study due to AEs ranged from 4.4% to 8.6% across treatment arms. The most common reason for study discontinuation was death (9 patients in the faricimab arms and 1 patient in aflibercept arm) and COVID-19 (8 patients in the faricimab arms and 1 patient in aflibercept arm).

In the pooled YOSEMITE and RHINE population, 81 patients died (4.4%, 4.7%, 3.7% in the 8-week faricimab arm, PTI faricimab arm, and aflibercept arm, respectively). The most common primary causes of death included gunshot wounds, falls, natural causes, advanced hepatocellular carcinoma with metastases to the bone, head injury, and unexplained death (3 patients, 6 patients, and 1 patient in the 8-week faricimab arm, PTI faricimab arm, and aflibercept arm, respectively). Study investigators did not consider any of the deaths to be related to the study treatment.

Cataract was the most common notable harm, occurring in 9.9% to 17.6% of patients across all treatment arms in the 2 studies. Over the course of both studies, 6 patients in the faricimab arms and 1 patient in the aflibercept arm reported endophthalmitis. Uveitis and iritis were the most commonly reported intraocular inflammation events. Uveitis occurred in 7 patients in the faricimab arms and no patients in the aflibercept arm. The occurrence of iritis was comparable across treatment arms. Nonocular arterial thromboembolic events were reported in 6.9% to 10.9% of patients across both studies, with comparable frequencies in the treatment arms. Vitreous floaters were reported in 1.9% to 5.4% of patients, and these events were numerically higher in the 8-week faricimab arm than in the aflibercept arm of both studies. Retinal detachment, retinal tear, glaucoma, retinal vascular occlusive disease events, eye irritation, ocular discomfort, and blurred vision occurred infrequently (< 2% for each harm across all treatment arms in both studies). A small number of patients in the faricimab arms reported retinal detachments (6 in the 2 studies) and retinal tears (3 in the 2 studies); in the aflibercept arm, there were 2 retinal detachments and no retinal tears in either study. There were no reports of retinal hemorrhage in either study.

Critical Appraisal

Description of Studies

One ITC was submitted by the sponsor and is included in this review. No additional ITCs were identified in the literature. The sponsor performed a Bayesian network meta-analysis (NMA) to estimate the efficacy of faricimab and of other anti-VEGFs in patients with DME.

Efficacy Results

For the outcome of BCVA at 1 year, 23 trials were analyzed using a random-effects model. In the ITC, no treatment ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

For the outcome of number of injections at 1 year, 11 trials were analyzed under a random-effects model. The ITC showed that that |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| these data are impacted by the administration of therapies with fixed intervals in clinical trials, according to protocols within the 1-year time frame of the randomized controlled trials (RCTs).

For the outcome of retinal thickness at 1 year, 23 RCTs were analyzed using a random-effects model. The results of |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||. However, 95% credible intervals (CrIs) are wide and heterogeneity in the methods to assess retinal thickness across studies adds considerable uncertainty to the results for this analysis and limit conclusions about the relative effect of faricimab on central retinal thickness (CRT).

The outcome of the proportion of patients gaining or losing 10 or 15 ETDRS letters at 1 year was analyzed, but poor model fit precludes conclusions about the effect of faricimab, compared with comparators, for this outcome.

Harms Results

There were limited data available for the NMAs conducted for ocular adverse effects and for discontinuation; therefore, fixed-effects models were used for these end points and there was a high degree of uncertainty in these models. Limitations of the NMAs preclude conclusions about ocular adverse effects and overall treatment discontinuation.

Critical Appraisal

There may be important sources of bias related to different study or patient characteristics that could impact conclusions about this ITC. The limitations described may pose a considerable challenge when trying to come to conclusive decisions about the validity of the results that can inform clinical practice. There were many trials with missing information about study and baseline characteristics and there was considerable heterogeneity among these characteristics. Most notably, there was heterogeneity in the methods used to assess retinal thickness and in the availability of information related to the presence of significant diabetic macular ischemia or systemic comorbidities. Additionally, there was a weak connection between faricimab and the rest of the network through aflibercept.

Although PTI faricimab may be more favourable than ranibizumab pro re nata (PRN) (i.e., as needed), treat and extend, every 4 weeks, and bevacizumab PRN for the outcome of BCVA, CrIs were very close to the null value for this analysis. The results of the analysis related to the number of injections will have been affected by the administration of therapies with fixed intervals in clinical trials, according to study protocols. Limitations to the NMA preclude conclusions about the proportion of patients gaining or losing 10 or 15 ETDRS letters and about retinal thickness.

There were limited data available for the NMAs conducted for ocular adverse effects and for treatment discontinuation; therefore, fixed-effects models were used for these end points and there was a high degree of statistical uncertainty in these models. Therefore, there are limited data from which to draw any conclusions about the effect of faricimab, compared with comparators, on ocular adverse effects and treatment discontinuation.

Unmet Needs

The clinical expert consulted by CADTH indicated that the treatment goals of current therapies are to delay DME and, in some cases, to reverse the progression of DME and/or retinopathy, as well as to improve vision-related and general quality of life. Although current anti-VEGF treatments for DME have been useful for the treatment of DME over the past 10 to 15 years, they need to be given intravitreally by trained clinicians once every 1 to 3 months on an ongoing basis, often for years. This frequent administration poses a significant burden to patients and caregivers, especially in Canada where travel distances can be long and challenging in the winter. The clinical expert noted that longer-acting treatments would fill a significant unmet medical need by improving convenience of the treatment regimen and reducing the burden on patients and caregivers. As well, they could improve outcomes, in part, by increasing adherence with treatment regimens. Additionally, the expert noted that not all patients respond to available treatments and, in some cases, patients may become refractory to current treatment options.

Place in Therapy

As faricimab is the first anti-VEGF to target the Ang-2 pathway in addition to the VEGF pathway, the expert noted that its mechanism of action is a rational approach, given the underlying disease process. The expert noted that the Ang-2 pathway of angiogenesis is relevant in DR. According to the clinical expert, faricimab is expected to have a place in therapy, along with other anti-VEGFs, as a first-line or later-line treatment in patients with DME. The clinical expert indicated that if faricimab is reimbursed, a shift in the treatment paradigm is likely, given that faricimab is the first anti-VEGF approved for an extended interval of up to 16 weeks, which could potentially address the unmet need related to frequent treatment visits.

In the clinical expert’s opinion, there are no clinical reasons to make patients try other treatments before faricimab is initiated. Faricimab is expected to be prescribed as a first-line (or later-line) treatment for DME and, as with any of the existing treatments, early initiation is important for the best clinical outcomes.

Patient Population

Patients with DR associated with vision loss secondary to centre-involved macular edema are the best candidates for faricimab, according to the clinical expert. The clinical expert indicated that faricimab can be used in patients who are treatment-naive or who require a change in therapy due to inadequate responses to other anti-VEGF drugs. Patients with better baseline visual acuity, centre-involved edema of recent onset, and better control of diabetes and comorbid conditions may be more likely to benefit from treatment. Patients who present with major structural damage to the macular retina (e.g., macular atrophy or fibrosis) may not be suitable for treatment. Suitability for treatment would be assessed using a clinical exam, IV fluorescein angiography and OCT, potentially with the addition of OCT angiography. In current clinical practice, OCT is unlikely to lead to misdiagnosis. OCT is used not only for diagnosis, but also for prognosis and follow-up.

Assessing Response to Treatment

The clinical expert noted that clinical evaluation and OCT should be performed at almost every dosing visit to assess treatment response, with a treat-and-extend approach to achieve the longest sustainable interval without recurrence, and that monitoring between dosing visits would not be required. Key assessment outcomes include change in visual acuity, retinal thickness, injection frequency, and the presence of retinal fluid.

The clinical expert reported that the optimal response to anti-VEGFs is generally achieved at least 6 to 12 months after initiation of therapy. In the experience of the expert, the majority of patients can achieve stabilized vision and improved quality of life, and about 50% to 65% of patients can achieve visual acuity improvement.

The clinical expert noted that when assessing the magnitude of change in visual acuity, it is crucial to keep in mind that patients with better vision at baseline generally have less room for improvement than those with poor vision at baseline. As such, the clinical expert reported that there is not an agreed-upon threshold that is indicative of a clinically meaningful change in visual acuity in patients with DME.

The clinical expert indicated that the presence of SRF or IRF is an indicator of active disease, which should prompt modification of the treatment plan (often a reduction in injection interval).

Discontinuing Treatment

The clinical expert indicated that faricimab should be discontinued in patients with treatment futility with proof of irreversible anatomic or functional damage, such as macular atrophy (ischemia) or fibrosis.

Prescribing Conditions

The clinical expert recommended retina subspecialty care as the most appropriate treatment setting for the prescription and administration of faricimab, especially in urban areas. In rural settings, trained comprehensive ophthalmologists with experience and expertise in the management of DME may be able to provide care.

Unmet Needs

Even though the current standard of therapy (i.e., anti-VEGF) is more effective than the previous standard of care (i.e., laser) for centre-involved DME, the clinician group stated that durability and a robust safety profile that improves long-term outcomes are unmet needs. Durability reduces treatment burden (with less frequent dosing) and maintains maximal vision gain (with improved compliance and monitoring). Durability will be translated into improved quality of life and independence, and will reduce the risk of falls, depression, and surgical intervention (vitrectomy). The minimization of side effects, such as injection-related complications like inflammation, infection, bleeding, retinal detachment, cataract, and glaucoma, is also important because side effects can compromise visual outcomes and result in blindness. The clinician group stated that all patients with DME have these unmet needs.

Place in Therapy

The clinician group said that faricimab can be considered a first-line treatment and/or a rescue treatment for patients who do not respond to the current therapies available for DME. The clinician group said that there is currently no standard of care in cases of treatment failure. Moreover, the clinician group noted that the vision of treatment-naive patients and of patients previously treated with other anti-VEGF drugs can benefit from a switch to faricimab. According to the clinician group, it is not necessary for patients to try other therapies before faricimab is initiated.

Patient Population

According to the clinician group, all patients with centre-involved DME will be suitable for treatment with faricimab. The clinician group stated that eligible patients can be identified with clinical exams and diagnostic tests (OCT, OCT angiography, IV fluorescein angiography) in the routine clinical practice setting. The clinician group said that patients without centre-involved DME should not be treated with faricimab and that those without vision loss secondary to DME (pre-symptomatic patients) should be monitored, as long as very close follow-up can be maintained. The clinician group also noted that patients with good baseline vision are likely to maintain good vision in the long term, although patients with all levels of vision would benefit from faricimab.

Assessing Response to Treatment

The clinician group said that response to treatment can be measured with visual acuity testing (subjective outcome), fluid can be measured with OCT testing (objective outcome), and macular thickening can be assessed with a clinical exam. According to the clinician group, improvement in vision, a reduction or resolution of macular edema, a regression in DRSS score, and a reduction in the frequency of treatment (4 months or longer between treatments) are considered clinically meaningful responses. Last, the clinician group said that response to treatment should be assessed at every clinical visit, which is determined by treatment need.

Discontinuing Treatment

The clinician group said that in the case of end-stage disease with significant atrophy, fibrosis, and/or no improvement despite regular treatments, discontinuation of faricimab should be considered.

Prescribing Conditions

The clinician group stated that ophthalmology offices in the community and/or hospital setting are appropriate for the administration of faricimab. The group added that an ophthalmologist is required to accurately diagnose, treat, and monitor patients being treated with faricimab.

Inclusion and Exclusion Criteria

The key inclusion criteria for both studies included macular thickening secondary to DME involving the centre of the fovea (CST ≥ 325 µm on Spectralis SD-OCT or ≥ 315 µm on Cirrus SD-OCT or Topcon SD-OCT) in patients 18 years and older; current use of oral or injectable antidiabetic medication for the treatment of type 1 or type 2 diabetes, with a hemoglobin A1C level of 10% or less; and BCVA scores of 73 to 25 letters using the ETDRS protocol (20/40 to 20/320 Snellen equivalent). Patients could have been previously treated with an anti-VEGF in the study eye or could be treatment-naive. Enrolment of participants with anti-VEGF experience was capped at a maximum 25%. Key exclusion criteria were untreated diabetes, uncontrolled blood pressure, stroke or myocardial infarction in the previous 6 months, high-risk PDR in the study eye, use of medicated intraocular implants in the previous 6 months, any previous use of Iluvien implants, PRP, or macular laser treatment in the previous 3 months, and concurrent ocular conditions that could affect vision in the study eye. The key inclusion and exclusion criteria for the 2 trials are shown Table 6. Only 1 eye was assigned as the study eye in the studies. If both eyes were eligible, the eye with the worst BCVA at baseline was selected (unless the other eye was deemed by investigators to be more suitable for treatment).

Baseline Characteristics

A summary of baseline characteristics of the ITT population in both studies is shown in Table 7. The baseline demographic and ocular characteristics of patients were, overall, balanced in the treatment arms within each study. The baseline characteristics were generally similar across the studies, except months since DME diagnosis was shorter in YOSEMITE than in RHINE (median [minimum to maximum] months = 3.1 [0 to 304] months and 6.6 [0 to 380] months, respectively) and baseline CST was slightly higher in YOSEMITE than in RHINE (mean [standard deviation] = 487.5 [132.5] µm and 471.6 [125.3] µm, respectively). In both studies, patients had a median age of 62 to 64 years and the majority were male (> 57%) and White (> 76%). Macular ischemia (nonperfusion) was present at baseline in more than 1-third of patients (37% to 43%), and most patients had macular leakage (> 93%). At the start of the studies, most patients (approximately 70% to 75%) had a diabetic retinopathy severity (DRS) level of 35 to 47 (mild to moderately severe nonproliferative DR), with mean baseline BCVA scores of around 62 letters. Slightly more than 1 in 5 patients had been previously treated with an anti-VEGF (20.4% in RHINE; 22.9% in YOSEMITE).

Table 7

Summary of Baseline Characteristics in the YOSEMITE and RHINE Trials (ITT Population).

Efficacy Outcomes

HRQoL and Vision-Related Function

Presence of IRF and/or SRF

The proportion of patients with an absence of IRF and/or with an absence of SRF at week 52 and over time were measured as secondary end points. SRF and IRF specifically in the central subfield (within the 1 mm diameter centre of the macula) were of interest.

Harms Outcomes

The safety analysis included the incidence and severity of ocular and nonocular AEs that occurring during the study period. The occurrence of AEs was assessed at all assessment time points.

Noninferiority Margin

In the YOSEMITE and RHINE studies, a noninferiority margin of 4 ETDRS letters was used in the primary outcome analysis, where noninferiority would be demonstrated if the lower limit of the 97.5% CI of the difference in change in adjusted mean BCVA from baseline between the faricimab arms (8-week and PTI) and the active comparator (aflibercept) was greater than –4 ETDRS letters. The noninferiority margin of 4 ETDRS letters was selected based on data from the VISTA and VIVID trials, which compared aflibercept with laser control in patients with DME. A margin of 4 ETDRS letters represents approximately 50% of the least estimated benefit of aflibercept over laser control at week 52 in the VISTA study (10.7 letters for aflibercept versus 0.2 for control) and the VIVID study (10.7 letters for aflibercept versus 1.2 for control). The investigators also considered that a loss of 5 letters (1 ETDRS line) between treatments is generally clinically relevant, and therefore inferred that the noninferiority margin of 4 ETDRS letters, being smaller than a loss of 5 letters, would be small enough to allow the conclusion that the new treatment is not inferior to active control to an unacceptable extent. If the noninferiority of faricimab (8-week or PTI) to aflibercept for the primary end point was met in the ITT population, tests for superiority would be conducted, first in the treatment-naive population and then in the overall ITT population. Further details on the order of testing are provided in Figure 4.

In the key secondary efficacy analysis, a noninferiority margin of 10% was used, in which noninferiority would be demonstrated if the lower limit of the 97.5% CI for the difference in weighted proportions of patients with an improvement in DRS of at least 2 steps from baseline on the ETDRS DRSS between the treatment group (8-week or PTI faricimab) and the active comparator group (aflibercept) at week 52 was greater than –10%. No justification was provided for this noninferiority margin. In the case that noninferiority of faricimab (8-week or PTI) to aflibercept for the key secondary end point was met in the ITT population, tests for superiority would be conducted, first in the treatment-naive population, then in the overall ITT population.

Type I Error Control

The noninferiority and superiority hypotheses for the primary end point were tested at an overall significance level of alpha of 0.0496, using a graph-based testing procedure to control for the overall type I error rate. Pairwise comparisons between each dose of faricimab and aflibercept were conducted according to the testing procedure order illustrated in Figure 4. If the tests for 1 treatment sequence were all positive at the alpha/2 ( = 0.0248) level, then alpha/2 was propagated to the beginning of the other treatment sequence, which was tested at a significance level of alpha of 0.0496.

Sample Size Calculation

A sample size calculation determined that approximately 300 patients per treatment arm were required to demonstrate noninferiority between faricimab and aflibercept in the ITT population (using pairwise comparisons between the active comparator and each of the faricimab arms) with respect to the change in BCVA from baseline averaged over weeks 48, 52, and 56 at a 1-sided type I error rate of 1.25% with a power of 90% using a 2-sample t-test, assuming a noninferiority margin of 4 ETDRS letters, a standard deviation of ETDRS 11 letters, and a 10% dropout rate.

Figure 4

Graph-Based Testing Procedure for the Primary End Point.

Statistical Analysis for Efficacy Outcomes

The primary outcome was change from baseline in BCVA averaged over weeks 48, 52, and 56. The primary analysis was based on the ITT population (and the treatment-naive population for the initial superiority test) and was performed using a mixed model for repeated measures (MMRM), which included the change from baseline at weeks 4 to 56 as the response variable and was adjusted for treatment group, visit, visit-by-treatment-group interaction, baseline BCVA (continuous), as well as randomization stratification factors as fixed effects (day 1 BCVA ETDRS letter score [64 letters or better versus 63 letters or worse], prior IVT anti-VEGF therapy [yes or no], and region [US and Canada, Asia, and the rest of the world]), assuming an unstructured covariance structure. In the primary estimand, ICEs not due to COVID-19 (study treatment discontinuation due to AEs or lack of efficacy, use of prohibited systemic treatment or therapy in the study eye) were handled using a treatment policy strategy where all observed values were used regardless of the occurrence of the ICE, whereas ICEs due to COVID-19 (study drug discontinuation, use of prohibited therapy, missed doses with a potentially major impact on efficacy, or death due to COVID-19) were handled with a hypothetical strategy in which all values were censored after the ICE. Missing data for continuous outcomes were implicitly imputed with the MMRM model, based on the assumption that data were missing at random (MAR). Nonstandard BCVA data (e.g., assessed by ETDRS BCVA testing with prior visit refraction, test performed by unmasked certified ETDRS BCVA assessor or by uncertified experienced ETDRS BCVA assessor) were excluded from the analyses.

Pre-specified subgroup analyses were conducted with respect to the primary end point and key secondary end point. Baseline BCVA subgroup (≥ 64 letters and ≤ 63 letters), prior IVT anti-VEGF therapy (yes and no), baseline DRS (< 47, 47 to 53 and > 53 ETDRS DRSS), and baseline hemoglobin A1C (≤ 8% and > 8%) were relevant to this review. Other subgroups identified as relevant to the systematic review in the protocol were not assessed in the studies, including history of ischemic (cerebrovascular or cardiovascular) disease, hypertension, dyslipidemia, and proliferative and nonproliferative DR.

A pre-specified sensitivity analysis was performed using the same estimand and analysis method as the primary analysis, except that a last observation carried forward imputation approach was used to account for missing BCVA data, as well as for BCVA assessments that were censored after COVID-19-related ICEs. Six supplementary analyses using the per-protocol population, a multiple imputation method, and different analysis methods (analysis of covariance, trimmed mean) and strategies for handling ICEs (treatment policy strategy only, hypothetical strategy only) were performed to further evaluate the robustness of the evidence from the primary analysis.

A summary of the statistical analyses of efficacy end points in both studies is shown in Table 9. The analysis for secondary outcomes assessed data in the ITT population (and the treatment-native population for the superiority test of the key secondary end point). Continuous secondary end points that were of interest in this review were analyzed using the same approach as the primary analysis, except that no sensitivity analysis was performed. Binary secondary end points that assessed the proportion of patients in each treatment group and the difference in proportion between treatment groups were calculated by applying Cochran-Mantel-Haenszel (CMH) weights and stratified by the following randomization stratification factors: day 1 BCVA ETDRS letter score (64 letters or better versus 63 letters or worse), prior IVT anti-VEGF therapy (yes versus no), and region (US and Canada, Asia, and the rest of the world). CIs for the proportion of patients in each treatment arm and the overall difference in proportions between treatment arms will be calculated using the normal approximation to the weighted proportions. Missing data were not imputed for secondary outcomes. Exploratory outcomes were analyzed using descriptive statistics (mean, standard deviation, median, and range for continuous end points, and counts and percentages for categorical end points).

Statistical Analysis for Harms Outcomes

The safety analysis was based on AEs recorded through week 100 and were summarized using descriptive statistics.

Table 9

Statistical Analysis of Efficacy End Points in the YOSEMITE and RHINE Trials.

Analysis Populations

Results are reported for the ITT, treatment-naive, per-protocol, and safety-evaluable populations in YOSEMITE⁹ and RHINE.¹¹

ITT population: All patients who were randomized in the study were included. Patients were assessed according to the treatment assigned at randomization. This analysis population served as the primary analysis set for all efficacy analyses.

Treatment-naive population: All patients who were randomized in the study who had not received any IVT anti-VEGF drugs in the study eye before day 1 were included. For analyses based on this population, patients were grouped according to the treatment assigned at randomization. This population was used in the initial test of superiority of faricimab over aflibercept.

Per-protocol population: All patients randomized in the study who received at least 1 dose of study treatment and who did not have a major protocol violation that affected the efficacy evaluation or treatment interval determination were included. Patients were assessed according to the actual treatment received. If a patient received a combination of different active treatments (faricimab and aflibercept) in the study eye, the patient’s treatment group was assessed as randomized. This population was used for supplementary analysis of the primary efficacy end point.

Safety-evaluable population: All patients who received at least 1 injection of either faricimab or aflibercept in the study eye were included. Patients were assessed according to the actual treatment received. If a patient received a combination of different active treatments (faricimab and aflibercept) in the study eye, the patient’s treatment group was as randomized. This population was used for safety analyses.

For this review, noninferiority will be assessed using the results from both the ITT and per-protocol analyses.

Change From Baseline in Visual Acuity

Change in CRT

Frequency of Injection

In YOSEMITE, the proportion of patients in the faricimab PTI arm on a 4-, 8-, 12-, and 16-week treatment interval at week 52, a secondary outcome, was 10.8%, 15.4%, 21.0%, and 52.8%, respectively; in RHINE, the proportions were 13.3%, 15.6%, 20.1%, and 51.0%, respectively (Table 14). An analysis of the number of patients in the PTI faricimab arm on 4-, 8-, 12-, and 16-week dosing at each monthly visit through week 52, prepared by FDA statistical reviewers, is presented in Figure 13 in Appendix 3.

Table 13

Change From Baseline in CST (ILM-BM) at 1 and 2 Years (ITT Population).

Table 14

Frequency of Injection Outcomes (ITT Population).

At week 96, the proportion of patients in the PTI faricimab arm on a 4-, 8-, 12-, and 16-week treatment interval was 7.0%, 14.8%, 18.1%, and 60.0%, respectively, in YOSEMITE; in RHINE, the proportions were 10.1%, 11.8%, 13.6%, and 64.5%, respectively (Table 14).

Results are descriptive and between-group analyses were not reported in either study.

HRQoL and Vision-Related Function

Proportion of Patients With a BCVA Snellen Equivalent of 20/200 or Worse (Legal Blindness)

The number of patients progressing to legal blindness (BCVA Snellen equivalent of 20/200 or worse [BCVA ETDRS ≤ 38 letters]) was small in all treatment arms in both studies (between 0 and 6 patients per arm), and the difference in the proportion of patients who progressed to legal blindness or worse was small (< 1%) between each faricimab arm and the aflibercept arm in both studies at 1 and 2 years (Table 15).

Table 15

NEI VFQ-25 and Other Vision Function Outcomes (ITT Population).

Intraretinal Fluid and/or Subretinal Fluid

Absence of SRF

In both studies, the proportion of patients with an absence of SRF in the ETDRS central subfield from baseline through week 100 was comparable in the 8-week faricimab arm, the PTI faricimab arm, and the aflibercept arm. At week 52, the difference in the adjusted proportion in patients without SRF between the 8-week faricimab arm and the aflibercept arm was –2.2% (95% CI, –5.2% to 0.8%) and between the PTI faricimab arm and the aflibercept arm was –2.5% (95% CI, –5.6% to 0.5%) in YOSEMITE; in RHINE, the differences were –3.1% (95% CI, –6.3% to 0.1%) and –2.0% (95% CI, –4.9% to 0.9%), respectively (Table 16).

Given the inconsistency between the results for absence of IRF and the results for absence of SRF individually, the proportion of patients with an absence of both IRF and SRF is not presented here.

Table 16

Proportion of Patients With an Absence of IRF or SRF Over Time (ITT Population, CMH Method).

Change From Baseline in DRS

Proportion of Patients Who Develop New PDR Over Time

Few patients who did not have PDR at baseline developed new PDR in the study eye over time (up to week 96) in the 2 studies (< 3% in any treatment arm). A comparable proportion of patients in the 8-week faricimab arm, the PTI faricimab arm, and the aflibercept arm had developed new PDR in YOSEMITE and RHINE at week 52 and week 96 (Table 17).

Table 17

Change in DRS From Baseline Over Time (CMH Method) .

Adverse Events

The proportion of patients reporting at least 1 ocular AE in the study eye during the study period was comparable across treatment arms in the YOSEMITE trial (47.0%, 46.6%, and 46.3% in the 8-week faricimab, PTI faricimab, and aflibercept arms, respectively). However, in the RHINE study, a higher proportion of patients in the 8-week faricimab arm and the PTI faricimab arm reported an ocular AE than in the aflibercept arm (52.4%, 51.7%, and 44.6%, respectively). The most common ocular AEs in both studies were cataract, conjunctival hemorrhage, vitreous detachment, vitreous floaters, elevated intraocular pressure, diabetic retinal edema, dry eye, eye pain, posterior capsule opacification, and punctate keratitis. In the RHINE study, conjunctivitis, blepharitis, cataract subcapsular, medication error, and DR were common ocular AEs (≥ 2% in any treatment arm). In the RHINE study, AEs likely contributing to the higher occurrence of ocular AEs in both faricimab arms than in the aflibercept arm include cataract, dry eye, and blepharitis. The frequency of cataract in the 8-week faricimab, PTI faricimab, and aflibercept arms, respectively, was 14.5%, 15.7%, and 9.9%; the frequency of dry eye was 5.7%, 6.0%, and 3.5%, respectively; and the frequency of blepharitis was 4.4%, 2.2%, 0.6%, respectively. The frequency (8-week faricimab, PTI faricimab, and aflibercept arms, respectively) of conjunctival hemorrhage (9.8%, 5.6%, and 6.7%), elevated intraocular pressure (5.7%, 3.8%, and 3.2%), vitreous floaters (5.0%, 2.2%, and 3.8%), subcapsular cataract (3.2%, 1.9%, and 1.3%), posterior capsule opacification (2.8%, 0.6%, and 0.6%), eye pruritis (1.6%, 0.6%, and 0.6%), and allergic conjunctivitis (1.6%, 0.3%, and 0.6%) was higher in the 8-week faricimab arm than in the aflibercept arm, which also contributed to the higher rate of ocular AEs in the 8-week faricimab arm in the RHINE study (Table 18).

In both the YOSEMITE and RHINE studies, A comparable proportion of patients reported at least 1 nonocular AE in the 8-week faricimab, PTI faricimab, and aflibercept arms in YOSEMITE (76.7%, 80.2%, and 77.8%, respectively) and in RHINE (69.4%, 68.3%, and 73.6%, respectively). The proportion of nonocular AEs suspected by the investigators to be related to treatment was low and comparable in the 8-week faricimab, PTI faricimab, and aflibercept arms in YOSEMITE (1.0%, 0.6%, and 1.3%, respectively) and in RHINE (0.3%, 0.9%, and 1.0%, respectively).

Serious Adverse Events

Ocular SAEs were reported at a low frequency in both trials; however, in both YOSEMITE and RHINE, there was a slightly higher frequency of ocular SAEs in the PTI faricimab arm than in the aflibercept arm, and in YOSEMITE, there was a slightly higher frequency of ocular SAEs in the 8-week faricimab and PTI faricimab arms than in the aflibercept arm (YOSEMITE: 3.8%, 4.5%, and 2.3%, respectively; RHINE: 4.4%, 6.3%, and 4.1%, respectively). The most common (≥ 1% in any treatment arm) ocular SAE reported during the studies was cataract. In YOSEMITE, endophthalmitis and uveitis were also common (≥ 1% in any treatment arm) ocular SAEs (3 patients each [1.0%] in the faricimab PTI arm) (Table 18).

In YOSEMITE, there were numerically more nonocular SAEs in the 8-week faricimab and PTI faricimab arms than in the aflibercept arm (31.6%, 31.0%, and 27.0%, respectively), whereas in RHINE, there were numerically fewer (24.0%, 20.1%, and 28.3%, respectively). The most common (≥ 2% in any treatment arm) nonocular SAEs in the trials across treatment arms were COVID-19 (1.3% to 3.2%), pneumonia (1.3% to 2.6%), cellulitis (0.3% to 2.5%), sepsis (0% to 2.3%), and osteomyelitis (0.3% to 2.3%).

Withdrawals Due to Adverse Events

In both studies, a small proportion of patients in all arms discontinued treatment due to AEs. In YOSEMITE, 2.6% of patients in the 8-week faricimab arm and 2.9% in the PTI faricimab arm discontinued treatment due to AEs, whereas in RHINE, 2.2% and 2.8%, respectively, did. In both YOSEMITE and RHINE, 1.6% of patients in the aflibercept arm discontinued treatment due to AEs. The most common reason for treatment discontinuation was uveitis (3 patients in the PTI faricimab arm in YOSEMITE) (Table 18).

A high proportion of patients in the 8-week faricimab, PTI faricimab, and aflibercept arms discontinued the study due to AEs in YOSEMITE (7.0%, 8.6%, and 5.8%, respectively) and in RHINE (5.0%, 4.4%, and 5.1%, respectively). The most common reasons for study discontinuation due to AE was death (9 patients in the faricimab arms, 1 patient in the aflibercept arm) and COVID-19 (8 patients in the faricimab arms, 1 patient in the aflibercept arm) (Table 18).

Mortality

Of the 50 patients who died during the YOSEMITE study, 16 were in the 8-week faricimab arm, 21 were in the PTI faricimab PTI arm, and 13 were in the aflibercept arm. Of the 31 patients who died during the RHINE study, 12 were in the 8-week faricimab arm, 9 were in the PTI faricimab arm, and 10 were in the aflibercept arm (Table 18). When data for the 81 deaths that occurred during the the 2 studies were pooled, they accounted for 4.4%, 4.7%, 3.7% of patients in the 8-week faricimab, PTI faricimab, and aflibercept arms, respectively. The most common primary causes of death were the reported term of death (a category that included gunshot wounds, falls, natural causes, advanced hepatocellular carcinoma with metastases to the bone, head injury, and unexplained death) (reported in 3 patients, 6 patients, and 1 patient in the 8-week faricimab, PTI faricimab, and aflibercept arms, respectively); COVID-19 (reported in 3 patients, 4 patients, and 1 patient, respectively); and myocardial infarction (reported in 2 patients, 2 patients, and 4 patients, respectively). According to the sponsor, none of the deaths were suspected by the investigator to be related to study treatment.⁶⁰

Table 18

Summary of Harms During the Study (to Week 100 in the Safety-Evaluable Population) .

Notable Harms

Cataract was the most commonly occurring notable harm, occurring in 9.9% to 17.6% of patients across treatment arms during both studies, followed by nonocular arterial thromboembolic events (6.9% to 10.9% per arm) and conjunctival hemorrhage (6.4% to 9.8% per arm). Over the course of both studies, 7 patients reported endophthalmitis: 3 in the PTI faricimab arm in YOSEMITE; and 2 in the 8-week faricimab arm, 1 in the PTI faricimab arm, and 1 in the aflibercept arm in RHINE. Intraocular inflammation was reported in 0.6% to 2.2% of patients across treatment arms in both studies, with uveitis being the most commonly reported intraocular inflammation event, occurring in 7 patients in the faricimab arms (6 in YOSEMITE and 1 in RHINE) and no patients in the aflibercept arm. There were 2 events of intraocular inflammation associated with a vision loss of at least 15 letters and 2 events associated with a loss of at least 30 letters during the YOSEMITE study in the PTI faricimab arm. In RHINE, 1 patient in the 8-week faricimab arm experienced mild and nonserious vitritis that was related to the study drug.

Retinal detachment, retinal tear, glaucoma, retinal vascular occlusive disease events (defined as arterial occlusive disease, retinal artery embolism, retinal vein occlusion, retinal artery occlusion, or venous occlusion), eye irritation, ocular discomfort, and blurred vision occurred infrequently (< 2% across all treatment arms in both studies). There were no reports of retinal hemorrhage as an AE in either study. A small number of patients in the faricimab arms experienced retinal detachment (6 patients in the 2 studies) and retinal tears (3 patients in the 2 studies); in the aflibercept arm, there were 2 retinal detachments and no retinal tears. Vitreous floaters, reported in 1.9% to 5.4% of patients, were numerically higher in the 8-week faricimab arm than in the aflibercept arm of the studies. Nonocular arterial thromboembolic events (including nonfatal stroke, nonfatal myocardial infarction, and vascular death) were reported in 6.9% to 10.9% of patients across treatment arms in the 2 studies, with comparable frequencies between treatment arms.

There was generally no discernable imbalance in other notable harms across treatment arms or studies (Table 18).

Internal Validity

YOSEMITE and RHINE were identically designed randomized, double-blind, active-controlled, noninferiority phase III trials that compared 2 dosing regimens of faricimab (8-week and PTI) with aflibercept (8-week). The overall trial design was appropriate for the objectives of the YOSEMITE and RHINE studies. There were no major concerns about the method of randomization, which involved stratification by baseline BCVA, prior anti-VEGF treatment, and geographic region, as well as the use of an interactive web-based response system for randomized assignment. The baseline characteristics of the study population were generally well balanced across treatment arms and studies, except that time since DME diagnosis was, on average, shorter in RHINE than in YOSEMITE, and CST was slightly lower CST at baseline; however, the clinical expert thought that these differences were unlikely to have an impact on the results of the studies. The methods of allocation concealment and blinding were appropriate. The use of sham injections to preserve blinding in patients was likely successful, according to the clinical expert, considering that the procedure was done on anesthetized eyes and patients were unlikely to feel the difference between the sham injection with the blunt end of a syringe and a real injection. Treatment assignment was inadvertently unmasked for a small number of patients (8 patients in YOSEMITE and 2 patients in RHINE) and, in some cases, a masked physician performed an unmasked physician task (16 in YOSEMITE) or an examiner of visual acuity was unmasked to a patient’s study eye (1 instance in YOSEMITE and 3 instances in RHINE); however, these numbers were infrequent, and there is no information indicative of wider issues with blinding in the studies.

Approximately half the patients in the studies had at least 1 major protocol deviation (46.3% to 50.5% across treatment arms at 56 weeks). The most common major protocol deviation was missed visits at weeks 44, 48, 52, or 56 (21.6% to 25.3% of patients across treatment arms in both studies), followed by major issues with images (e.g., missed images) (5.8% to 10.3% across treatment arms in both studies). These deviations were likely higher than expected given the COVID-19 pandemic, as 32% of deviations in YOSEMITE and 39% in RHINE were deemed to be related to COVID-19. Although the number of major protocol deviations is a limitation, these events were generally balanced between treatment arms within each study, and results of the sensitivity and supplementary analyses (including the per-protocol analysis) were consistent with the primary estimand.

Per the study design, different dosing schedules were used in the treatment arms in both the loading and maintenance phases. In the loading phase, patients in the aflibercept arm received 5 monthly doses, patients in the 8-week faricimab arm received 6 monthly doses, and patients in the PTI faricimab arm received 4 monthly doses. In the maintenance phase, the treatment interval could be modified after randomization for patients in the PTI faricimab arm, using pre-specified criteria, to 4-, 8-, 12-, or 16-weeks. Intervals in the aflibercept arm could not be adjusted in this way after randomization, and patients received doses at fixed 8-week intervals. Protocol-based differences in the dosing schedule should be taken into account when considering the number and frequency of injections patients received. The ability to reduce the dosing interval to every 4 weeks could potentially create a bias in favour of PTI faricimab for efficacy outcomes; however, treatment intervals longer than 8 weeks in the PTI faricimab arm could have the opposite effect. Around 11% to 13% patients in the PTI faricimab arm were on a 4-week dosing schedule at 1 year. However, according to the clinical expert, a certain proportion of patients with DME (approximately 10%) would be expected to have an inadequate response to anti-VEGF treatment in general.

The studies established the noninferiority of faricimab to aflibercept based on primary outcome analyses in the ITT population. A supplementary per-protocol analysis confirmed the noninferiority in the primary ITT population. As well, several sensitivity analyses conducted by the sponsor and by the FDA confirmed the findings of each study.⁴⁵

The noninferiority margin of 4 ETDRS letters for the primary end point, which was determined by the sponsor based on prior clinical trial data and clinical reasoning, aligned with recommended approaches. The clinical rationale was considered reasonable by the clinical expert consulted by CADTH. No rationale was provided for the noninferiority margin for the key secondary end point analysis, in which a margin of 10% was used to demonstrate noninferiority in the difference weighted proportions of patients with an improvement in DRS of 2 steps or more from baseline on the ETDRS DRSS; however, the clinical expert consulted by CADTH indicated that the 10% margin is a reasonable choice.

The enrolled sample sizes were adequate for the assessment of the primary outcome. Subgroup analyses were pre-specified, although, because of the lack of sample size considerations, control for multiplicity, and statistical testing for treatment-by-subgroup interaction, no conclusions related to subgroup effects can be drawn. Similarly, the secondary and exploratory end points should be interpreted in light of the lack of both sample size considerations and control for type I error.

The proportion of patients who discontinued the study treatment before week 56 was approximately 9% and 6% in YOSEMITE and RHINE, respectively, and the proportions were generally balanced in the treatment arms within each study, except the PTI faricimab arm in RHINE had a lower proportion of patients who discontinued treatment than the other arms (3.4%). ICEs occurred in approximately 9% and 10% of patients YOSEMITE and RHINE, respectively, through week 56, the majority of which were missed doses related to COVID-19, which could have had a potentially major impact on efficacy at weeks 44, 48, and 52. The proportion of missed doses was higher in the 8-week faricimab arm than in the PTI faricimab and aflibercept arms in YOSEMITE (9.8%, 5.4%, 4.8%, respectively) and in RHINE (10.4%, 6.6%, 7.0%, respectively), which could potentially create a bias against 8-week faricimab treatment. The few ICEs not related to COVID-19 were comparable in the treatment arms. Treatment policy strategy and hypothetical strategy were used to address non-COVID-19-related and COVID-19-related ICEs, respectively, in the primary estimand of the primary efficacy end point. The strategies were consistent with the approaches recommended by the FDA for ICEs.⁶¹

Although the hypothetical strategy (i.e., ICEs due to COVID-19 were censored and imputed using MMRM modelling and missing data were assumed to be MAR) is 1 of the approaches identified in the FDA guidance, the treatment policy strategy (i.e., including all data, regardless of ICEs) is the preferred approach to ensure that all data were used and because it is not clear that the MAR assumption would be met. The FDA statistical review of the faricimab studies, likewise, considered the treatment policy strategy to be the better of the 2.⁴⁵ However, results of the supplementary analyses confirmed those of the primary estimand, suggesting the approach used to handle ICEs unlikely introduced bias.

The studies used implicit imputation by MMRM that assumed a MAR mechanism to account for missing data for continuous outcomes, while observed data with no imputation were used on missing binary outcomes, including for change in DRSS, which was a key secondary end point in the studies. No sensitivity analyses were conducted to assess the impact of missing data on the secondary outcomes, thereby making unsubstantiated assumptions about the secondary analyses. The FDA statistical review also noted this as a limitation and conducted additional analyses.⁴⁵ The results of these additional analyses confirmed the original secondary results.

External Validity

Ten of the 174 RHINE study sites and none of the 179 YOSEMITE study sites were in Canada. The studies included patients who had been previously treated with an anti-VEGF and those who were treatment-naive. The inclusion and exclusion criteria were generally reflective of the eligibility criteria for anti-VEGF treatment in clinical practice; however, the expert consulted by CADTH noted that, in clinical practice, patients with less well controlled diabetes and/or a wider range of comorbidities would still be considered for treatment with faricimab. Treatment would not normally be withheld based only on a high level of hemoglobin A1C. Patients with high-risk PDR (excluded in the studies) would also receive treatment in practice, as would patients receiving tamoxifen (a prohibited therapy in the studies). However, the expert did not think that the inclusion or exclusion criteria would limit the generalizability of the studies’ results to the patient population seen in real-world settings. Patients with uncontrolled diabetes might have different response to treatment, but there is no direct evidence currently available for faricimab in this population.

A large percentage of patients failed to meet the eligibility criteria during the screening phase (38.6% in YOSEMITE and 44.5% in RHINE). The expert noted this was to be expected, given the patient population, and confirmed that the baseline characteristics of the study populations were similar to those of patients with DME in Canada.

The clinical expert indicated that aflibercept is an appropriate comparator, as it is the most commonly prescribed on-label anti-VEGF in Canada. Although direct comparative evidence against ranibizumab and bevacizumab would have been useful, the choice of aflibercept was seen as reasonable. The dosing regimen of aflibercept in the studies up to week 56 aligns with the product monograph dosing, but after year 1, the Canadian product monograph allows for a treat-and-extend approach, with treatment intervals extended by up to 2-week increments, based upon disease activity. However, aflibercept was given at a fixed interval of 8 weeks for the entire maintenance phase, which does not align with the treat-and-extend approach approved for the treatment of DME in Canada in year 2. This might bias the year 2 results, and the direction of the bias for efficacy would most likely favour aflibercept, given that an 8-week dosing interval would likely result in fewer injections than treat-and-extend dosing.

In terms of the clinical relevance of the outcomes assessed in the studies, BCVA change, retinal thickness measured by OCT, and the presence of retinal fluid are routinely assessed to evaluate treatment response in clinical practice, according to the clinical expert. The clinical group input also noted regression in DRSS as a clinically important outcome. Further, frequency of injection was identified as an outcome of key interest in both patient and clinical input. Although NEI VFQ-25 scores are infrequently measured outside of clinical research, HRQoL and vision function, which are partly captured in the questionnaire, are important outcomes to patients.

In the studies, patients were monitored monthly, but monthly monitoring is not mandated in the product monograph for faricimab, and the expert agreed that monitoring would only be required at dosing visits. The expert noted that the algorithm used to determine whether to reduce, maintain, or extend intervals of faricimab treatment was fairly rigid and would likely be applied in a more simplified manner in clinical practice, with more responsiveness to individual treatment outcomes based on CST and OCT measurements. It is unclear how different approaches to decision-making about treatment intervals would affect results.

The length of assessment in the primary analysis (56 weeks) was adequate for the assessment of efficacy and safety of faricimab in the context of a noninferiority trial, and data for up to 100 weeks were available. Longer-term studies may be needed to gain confidence on the durability of faricimab, and studies with a larger sample size would be needed to identify potential rare AEs.

YOSEMITE and RHINE are the only phase III studies to date that have provided direct evidence comparing faricimab with other anti-VEGF drugs in patients with DME. There is no direct evidence comparing faricimab with anti-VEGF drugs, other than aflibercept, currently used in Canadian practice (i.e., ranibizumab and bevacizumab), which represents an evidence gap.

The FDA review of faricimab noted that the impact of faricimab on corneal endothelial health has not been evaluated, and stated a need for a phase IV trial evaluating the corneal endothelial health of eyes treated with faricimab.⁴⁴

Objectives

The objective of the ITC was to assess the efficacy and safety of PTI faricimab compared with relevant interventions (listed in Table 20) given as monotherapy.

Study Selection Methods

A strategy was developed and a search was conducted of MEDLINE, EMBASE, the Cochrane Library, abstracts of relevant conferences, and relevant health technology assessment agencies, clinical trial registries, and key government or international bodies. The population of interest was patients with DME older than 18 years. Studies of both treatment-naive and treatment-experienced patients were included. The main intervention was defined as PTI faricimab dosing (6 mg IVT every 4 weeks to every 16 weeks).

Table 19

Study Selection Criteria and Methods for the Sponsor-Submitted ITC.

Table 20

Treatment Doses Considered in the ITC.

The criteria included studies published before the cut-off date of October 21, 2020; this search was updated in September 2021 (but no new trial data were added). Two reviewers independently screened the retrieved reports at 2 stages (titles and abstracts, and then full papers), and any disagreements were adjudicated by a third party. Final citations were verified by the project lead. Reasons for exclusion were documented. Data extraction was conducted by a single reviewer and quality was checked by a second reviewer. Details of the methods used to extract data from the included studies were described.

Quality assessment of the selected studies was carried out by 2 reviewers, and any disagreements were resolved by discussion or additional referees. A quality (risk-of-bias) assessment of studies was conducted using the 7-criteria checklist provided in section 2.5 of the National Institute for Health and Care Excellence single technology appraisal user guide.

Treatment doses considered in the ITC are listed in Table 20. Different dosing regimens and schedules were treated as different treatment arms in the NMA. Any arms that had the same regimen were pooled. PRN regimens were treated as similar treatment arms without regard for the number of loading doses.

Outcomes were considered at 12 months and are listed in Table 19. Visual acuity outcomes were defined as the mean change from baseline in BCVA, according to ETDRS letters. In addition, the proportion of patients gaining or losing 10 or more or 15 or more letters on the ETDRS scale from baseline was captured. This was defined as the proportion of patients in mutually exclusive letter categories (≥ –15; > –15 to ≥ –10; > –10 to ≥ –5; > –5 to < 5; ≥ 5 to < 10; ≥ 10 to < 15; ≥ 15). Anatomic outcomes included the mean change in retinal thickness, measured by CST. If the CST was missing from a study but 1 or more anatomic outcomes were reported, the other value was used in the following order: CST, central point thickness, CRT. Injection frequency was the mean number of injections given; this end point was considered at 12 months. Safety outcomes included overall ocular AEs, ocular SAEs, systemic AEs, and overall treatment discontinuation. No specific definition for overall treatment discontinuation, injection frequency, or AEs (overall ocular AEs and SAEs, or systemic adverse effects) was specified beyond the standard reporting in each trial.

The sponsor’s ITC reported outcomes at 12 months. Any result reported from week 48 to week 56, 12 months, or 1 year was classified as a 12-month outcome.

ITC Analysis Methods

An overview of the submitted ITC analysis methods is presented in Table 21.

The ITC compared faricimab with comparators for the available end points of visual acuity (BCVA and the proportion of patients gaining or losing ≥ 10 or ≥ 15 ETDRS letters) and anatomic outcomes (retinal thickness), number of injections, adverse effects, and overall treatment discontinuation at 12 months.

The ITC was an NMA performed using a Bayesian approach. The model was a Bayesian comparison using a generalized linear model framework; BCVA, retinal thickness, and number of injections used the identity link, the proportion of patients gaining or losing 10 or more or 15 or more ETDRS letters was modelled using the probit link, and all other end points were modelled using the logit link. Noninformative (vague) priors were planned for all parameters, and alternative priors were considered if the planned priors did not give sensible results or were too informative. Models generally used 10,000 iterations as a burn-in, with 40,000 iterations with a thinning parameter of 10; however, this was increased by 10 times for ocular adverse effects, serious ocular adverse effects, meta-regressions on patient characteristics for BCVA, and the number of injections and serious ocular AEs to improve convergence. There were at least 2 parallel chains run in all model fits. Convergence of the model was assessed using Brooks-Gelman-Rubin diagnostics.

Model fit was assessed using the diagnostic information criterion (DIC) (a 5-point difference was considered meaningful) and residual deviance. Laser PRN or ranibizumab 0.3 mg/0.5 mg IVT PRN was used as the reference treatment for computational efficiency, based on the best-connected nodes, depending on the network.

Table 21

Analysis Methods for ITC.

Meta-regression was conducted to investigate whether the treatment effect varied by the level of covariate. Patient characteristics of BCVA and CST (or, if not reported, CRT, central foveal thickness, central macular thickness) at baseline were investigated using standard network meta-regression methods to determine whether treatment effect varied by covariate. These covariates were investigated in separate models first, as they are correlated, and a model with all patient characteristics was considered only if both were important. In the meta-regression of BCVA and CST, aflibercept was used as the reference treatment for the interaction effects.

Different dosing schedules were treated as different treatment arms in the NMA.

Summary of Included Studies

Of the 135 publications from 83 unique studies included in the feasibility assessment, 33 were excluded for treatment-related reasons, 4 studies did not connect to faricimab through the network, 13 studies were excluded because they investigated unlicensed combination regimens, and 4 studies were excluded for other reasons. Of the remaining 29 studies considered for inclusion in the NMA, another 3 were excluded for reasons such as not reporting relevant data for outcomes at the time point of interest. Therefore, 26 trials were included in the NMA.

Table 22

Assessment of Homogeneity for the Sponsor-Submitted ITC.

An overview of the assessment of homogeneity for the ITC is presented in Table 22. A total of 12 of the 26 trials were head-to-head trials with active anti-VEGF comparators, 4 trials compared anti-VEGF drugs to dexamethasone, 1 trial compared anti-VEGF drugs to sham IVT, 1 trial compared immediate to deferred argon, and 10 trials compared anti-VEGF to laser therapy. A total of 11 of the 26 trials were double-masked, 3 were open-label, 6 were single-masked, and 6 did not report masking. Twelve of the 26 studies were phase III, 3 were phase II, 1 was phase IV, 3 were phase I/II, and 7 studies did not report phase. Eight of the 26 trials were multi-centre trials, 9 were muti-centre international trials, 7 were single-centre trials, and this information was not reported for 2 trials. Study size ranged from 20 to 2,244 patients randomized; 8 trials had fewer than 100 patients and 5 trials had more than 1,000 patients. Years of study ranged from 1985 to 2020.

Baseline patient characteristics were presented by treatment arm. Mean age at baseline ranged from 55 years to 69 years. The proportion of males in the study treatment arms ranged from 33.0% to 64.4%, although this information was not recorded for 5 trials. The proportion of study participants who were White ranged from 0.0% to 98.4%, although this information was not recorded for 5 trials. Mean BCVA (letters, ETDRS letters, ETDRS chart, ETDRS letter score, or ETDRS-like visual acuity) at baseline ranged from 29.2 to 70.4 letters. There was heterogeneity in the way retinal thickness was measured and defined in the trials, which included central foveal thickness, central macular thickness, CRT, and CST. There was also heterogeneity in the type of measurement used for retinal thickness. Retinal thickness at baseline ranged from 394 μm (CST) to 540 μm (CRT). The mean number of years with diabetes ranged from 11.1 years to 19.7 years (although there was 1 outlier at 1.31 years), but this was not reported for 13 trials. Mean intraocular pressure ranged from 14.9 mm Hg to 19.2 mm Hg; however, it was only reported for 6 trials. Mean hemoglobin A1C ranged from 7.3% to 8.4%, but this was not reported for 11 trials. The proportion of patients with type 2 diabetes ranged from 79.5% to 100%, but this was not reported in 8 trials. Mean time since diagnosis of DME was only reported for 5 trials, and ranged from 1.1 years to 20.7 years. Prior therapy for DME was reported as mixed for 15 trials, as previous treatment for 4 trials (of these, 2 were prior anti-VEGF), as treatment-naive for 5 trials, and was not reported in 2 trial.

Results

Risk of Bias

A quality-assessment diagram was reported for the 83 studies considered for inclusion in the feasibility assessment. Overall, included studies were rated to be of moderate to high quality. However, a quality assessment specifically for the 26 studies that were included in the NMA was not reported.

Figure 5

Network Diagram for the Outcome of Mean Change in BCVA at 12 Months.

BCVA

For the outcome of BCVA at 12 months, 22 trials were included in the analysis, which was conducted under a random-effects model, as this model had a lower DIC than the fixed-effects model (DIC not reported). A graphic representation of the evidence network is presented in Figure 5. Most trials compared active treatments, but the network did include 1 closed loop involving faricimab as an intervention.

||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

The comparative results are outlined in Table 23. There was no evidence of inconsistency from the comparison of the consistency and inconsistency model fits (DIC, mean residual deviance were compared).

Table 23

Mean Change in BCVA — ITC Results.

Retinal Thickness

For the mean change in retinal thickness at 12 months, 23 RCTs were included in the analysis, which was conducted with a random-effects model (the DIC for the random-effects model was larger than for the fixed-effects model but was not considered to be meaningful, as this difference was less than 5 points). A graphic representation of the evidence network is presented in Figure 6. The results ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||. However, 95% CrIs are wide. Results are outlined in Table 24. There was variability in the way retinal thickness was measured and reported in trials. If the CST was missing from a study but 1 or more anatomic outcomes were reported, the other value was used, in the following order: CST, central point thickness, CRT. It is not clear from the sponsor’s report how the values for this outcome compared across the trials. This variability in the way retinal thickness was defined, measured, and reported may contribute considerable heterogeneity to the ITC.

Table 24

Retinal Thickness — ITC Results.

Number of Injections at 12 Months

For the number of injections at 12 months, 11 RCTs were included in the analysis, which was conducted with a random-effects model, as the DIC for this model was lower than for the fixed-effects model. A graphic representation of the evidence network is presented in Figure 7. For the number-of-injections networks, ranibizumab 0.3 mg/0.5 mg IVT PRN was used as the computational reference, rather than laser PRN, as the network was formed without laser PRN. The ITC showed that |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| these data are affected by the administration of therapies with fixed intervals in clinical trials, according to protocols within the 1-year time frame of the RCTs. Results are outlined in Table 25.

Figure 6

Network Diagram for the Outcome of Retinal Thickness at 12 Months.

Patients Gaining or Losing at Least 10 or 15 ETDRS Letters

For the outcome of patients gaining or losing 10 or more or 15 or more ETDRS letters at 12 months, 22 trials were included in this analysis, which was conducted with a random-effects model, as the DIC for this model was lower than for the fixed-effects model. A graphic representation of the evidence network is presented in Figure 8. The results show that no treatment was favoured (95% CrI included the null value). Results are outlined in Table 26.

The technical report noted that various assessments showed the model for this outcome had a poor fit, likely because of limited data and heterogeneity. Various methods were used to adjust for the limitations, but these did not improve the model fit. This precludes the ability to draw conclusions from these data on the effect of faricimab, compared with comparators, on patients gaining or losing at least 10 or 15 ETDRS letters.

Table 25

Number of Injections 12 Months — ITC Results.

Figure 7

Network Diagram for the Outcome of Mean Number of Injections at 12 Months.

Figure 8

Network Diagram for the Outcome of Proportion of Patients Gaining or Losing at Least 10 or 15 ETDRS Letters at 12 Months.

Table 26

Patients Gaining or Losing 10 or More or 15 or More Letters ETDRS — ITC Results.

Ocular AEs

For the outcome of all ocular AEs at 12 months, 10 trials were included in the analysis, which was conducted with a fixed-effects model, as this model had a lower DIC than the random-effects model, despite not being considered meaningful in the technical report (DIC not reported). A graphic representation of the evidence network is presented in Figure 9. The results show that no treatment was favoured (95% CrIs included 1 for the odds ratio of ocular adverse effects) for ocular AEs for most comparators. In addition, PTI faricimab may be favourable (95% CrIs included 1 for the odds ratio of ocular adverse effects) for ocular adverse effects, compared with bevacizumab 1.25 mg IVT PRN and ranibizumab 0.3 mg/0.5 mg IVT PRN. However, the CrIs were close to the null value for these results. The comparative results are outlined in Table 27.

A NMA for serious ocular adverse effects or systemic adverse effects was not reported because there were limited data available for these events.

Figure 9

Network Diagram for the Outcome of Ocular Adverse Effects at 12 Months.

Table 27

Ocular Adverse Effects — ITC Results.

Overall Treatment Discontinuation

For the outcome of overall treatment discontinuation at 12 months, the sponsor included 14 trials in the analysis, which was conducted with a fixed-effects model, as this model had a lower DIC than the random-effects model, despite not being considered meaningful in the technical report (DIC not reported). A graphic representation of the evidence network is presented in Figure 10. The results show that no treatment was favoured (95% CrIs include 1 for odds of overall treatment discontinuation), although CrIs were wide for some comparisons. The comparative results are outlined in Table 28. The authors highlight the fact that a significant share of overall treatment discontinuation events in the YOSEMITE and RHINE trials were due to patient deaths, which were not considered to be treatment-related.

Table 28

Overall Treatment Discontinuation — ITC Results.

Figure 10

Network Diagram for the Outcome of Overall Treatment Discontinuation at 12 Months.

ClinicalTrials.gov

Produced by the U.S. National Library of Medicine. Targeted search used to capture registered clinical trials.

[Search -- Studies with results | faricimab]

WHO ICTRP

International Clinical Trials Registry Platform, produced by the World Health Organization. Targeted search used to capture registered clinical trials.

[Search terms -- faricimab]

Health Canada’s Clinical Trials Database

Produced by Health Canada. Targeted search used to capture registered clinical trials.

[Search terms -- faricimab]

EU Clinical Trials Register

European Union Clinical Trials Register, produced by the European Union. Targeted search used to capture registered clinical trials.

[Search terms -- faricimab]

Minimal Clinically Important Difference

To our knowledge, there has been no derivation of a minimal clinically important difference (MCID) for the ETDRS in DME. The FDA recommends a mean change of 15 letters or more on an ETDRS chart, or a statistically significant difference in the proportion of patients with ≥ 15 letter change in visual acuity, as clinically relevant outcomes in studies of DME.^50,51 The 15-letter reference point is still a topic of discussion for the FDA.

The test–retest variability (TRV) of the measure can help guide what would be considered a clinically meaningful change. Literature-based estimates of TRV range from ± 0.07 to ± 0.19 logMAR.⁷⁹ This suggests that any change in score between baseline and follow-up of approximately 4 to 10 letters results in insufficient certainty that the difference in letters is not just due to chance alone. When TRV is high, the ability to detect a real change in score is low. For example, for a TRV of ± 0.19, the sensitivity of a 0.1 logMAR change (5 letters) was 4% (0% to 14%). If the TRV is lowered to ± 0.11, the sensitivity of the test increases to 38% (25% to 53%). If the TRV remains at ± 0.11, and the threshold for change increases to a 0.2 logMAR change (10 letters), the sensitivity of the scale increases to 100% (93% to 100%). The baseline visual acuity of a sample population will affect the TRV of ETDRS letter scores⁴⁹ and as a result will also affect what would reasonably be considered an MCID. A TRV of ± 0.11 has been found in healthy participants,⁷⁹ while higher variability (± 0.15 to ± 0.20) has been cited for individuals with pathological changes in vision.⁸¹ For eyes with acuity better than 20/100, a change in visual acuity of ≥ 5 letters has > 90% probability of being a real change, while for eyes worse than 20/100, a change of ≥ 10 letters is required for the same reliability.⁴⁹ A threshold for clinically meaningful change in patients with advanced eye disease should be higher than in healthy individuals, and has been suggested to range between 10 and 15 letters.²⁵

Reliability

On an individual eye basis, complete inter-rater agreement on DRSS level in the ETDRS was observed in 53% of eyes and agreement within one level occurred in 88% of eyes.⁸² The unweighted kappa statistic was 0.42, which increased to 0.65 with a weighting of 1 for exact agreement, 0.75 for one-level disagreement, and 0 for all other disagreements.⁸² On a patient basis, there was complete inter-rater agreement in 38% of patients, agreement within one level in 71% of patients, and agreement within 2 levels in 87% of patients.⁸² The unweighted kappa statistics was 0.31 and the weighted kappa was 0.71 with a weight of 1 for exact agreement and weights of 0.9375 and 0.75 for disagreements of 1 and 2 levels, respectively.⁸²

Responsiveness to Change

Step progression refers to an increase in photographic level that can be used to describe change in DR over time.^65,82 In the ETDRS, the proportion of eyes with worsening of 2 or more levels was similar among all severity categories at the 1 year follow-up time point, but not for longer follow-up periods.⁸²

The Wisconsin Epidemiology Study of Diabetic Retinopathy evaluated whether fewer than 3 steps of ETDRS DRSS worsening using the patient-based scale were clinically meaningful by conducting a population-based study of patients with diabetes in 10 years of follow-up.⁶⁵ The results indicated that patients with 1 or more and 2 or more steps of ETDRS DRSS worsening over the first 4 years were significantly more likely to develop proliferative diabetic retinopathy in the last 6 years than those without ETDRS DRSS step progression.⁶⁵

Minimal (Clinically) Important Difference

An improvement of 3 or more steps is associated with a clinically meaningful improvement of 15 ETDRS letters in visual acuity⁵⁹ and has previously been accepted by the FDA as an efficacy end point for assessing improvement in diabetic retinopathy.

Reliability

Intra-device repeatability and inter-device reproducibility of measurements depend on a number of factors including retinal pathology, retinal region, region size, OCT model, equipment settings, manual or automated analysis, and operator experience.²⁹ In eyes with diabetic macular edema (DME), a comparison of measurements with 4 different OCT devices found good intra-device repeatability, but statistically significant differences in retinal thickness values across different devices.⁶⁸ Another study that compared the reproducibility of retinal thickness measurements from OCT images of eyes with DME obtained by TD-OCT and SD-OCT instruments found that SD-OCT devices demonstrated less TRV.⁶⁹

Validity

In patients with DME, the association between OCT-measured retinal thickness and BCVA has been evaluated. A moderate correlation between visual acuity and OCT centre point thickness has been observed (r = 0.52).⁵⁷ For every 100 µm decrease in centre point thickness, visual acuity increased by 4.4 letters (95% confidence interval [CI], 3.5 to 5.3).⁵⁷ Other studies have shown similarly modest correlations between visual acuity and central retinal thickness determined by OCT.^66,67

Responsiveness to Change

In eyes with DME treated by laser photocoagulation, changes in centre point thickness were associated with changes in visual acuity, with correlation coefficients of 0.44, 0.30, and 0.43 at 3, 5, 8, and 12 months, respectively.⁷⁰

Minimal (Clinically) Important Difference

MCID for OCT has not been estimated in patient population with DME.

Reliability

Both versions of the NEI VFQ were reported to be valid and reliable measures of health-related quality of life among patients with a wide range of eye conditions, including DME.^53,75,85,86 Internal consistency reliability was acceptable (Cronbach’s alpha α ≥ 0.7⁷⁴) for 6 of the 8 multi-item subscales. The internal consistency for peripheral vision and colour vision subscales was not available.^85,86 The near vision and distance vision subscales are 3-item domains on the NEI VFQ-25; their internal reliability as represented by Cronbach’s alpha was reported as 0.73 and 0.58, respectively. Limitations of internal consistency due to the presence of single-item domains were noted in a validation study specific for DME population.⁵³

Validity

Known groups validity was demonstrated by the higher mean VFQ-25 composite score in the quartile of patients with the best visual acuity (measured by ETDRS letters) compared with the quartile of patients with the worse visual acuity (mean ± SD: 72.1 ± 17.9 versus 56.1 ± 18.0, P < 0.001). Statistically significant differences were also demonstrated within all the subscales, except for ocular pain and colour vision.⁵³

Concurrent validity was assessed though correlations of the VFQ-25 subscales with the EQ-5D visual analogue scale (ranging from worse imaginable health to best imaginable health) and Pearson correlation coefficients ranged from 0.16 to 0.43. The correlation coefficient for the composite score and the EQ-5D visual analogue scale was 0.38.⁵³

Assessment of convergent validity yielded similar results, with correlations of the subscales with ETDRS letter score ranging from 0.10 to 0.41 in the study eye and from 0.01 to 0.51 in the fellow eye. In convergent validity analysis, the NEI VFQ-25 domains collectively showed low to moderate correlations with ETDRS visual acuity score for both the study and untreated eyes. The Pearson correlation with ETDRS total letters in the study eye was reported as 0.35 for the near vision subscale and 0.34 for the distance vision subscale. A slightly stronger correlation was observed between the NEI VFQ-25 and the EQ-5D Visual Analogue Scale (VAS), and the EQ-5D VAS along with ETDRS was a significant predictor of near and distance vision subscale scores, suggesting that general health-related quality of life was captured by the NEI VFQ-25 more so than strictly vision-related information. However, in support of known group validity, patients who saw more ETDRS letters also scored higher on the NEI VFQ-25 near and distance subscales as well as on the NEI VFQ-25 composite. Overall, the authors concluded that despite its documented limitations and the need for an improved instrument, the NEI VFQ-25 demonstrated a degree of validity to measure health-related quality of life in patients with DME.⁵³

Assessments of the psychometric validity of the NEI VFQ-25 using Rasch scoring and principal component analysis have identified issues with multidimensionality (measurement of more than 1 construct) and poor performance of the subscales.^71,72,73 The NEI VFQ-25 subscales were found to have too few items and were unable to discriminate among the population under measurement, and thus were not valid.^72,73 Re-engineering the NEI VFQ into 2 constructs (visual functioning and socio-emotional factors) and removing misfit items (e.g., pain around eyes, general health, and driving in difficult conditions) improved the psychometric validity of the scale in individuals with low vision.^72,73 Considering this recent evidence of multidimensionality, the validity of the single composite score of the NEI VFQ may be questioned.

Responsiveness to Change

All but 2 subscale scores (general health and ocular pain) have been shown to be responsive to changes in visual acuity in the better-seeing eye.^71,75

Minimal (Clinically) Important Difference

Determination of what constitutes a clinically meaningful change in the NEI VFQ appears to be linked to its correlation with visual acuity. A psychometric validation study of the NEI VFQ-25 specifically in patients with DME has more recently been conducted, and 2 distribution-based methods were employed to determine a minimal clinically important difference (MCID) from baseline to week 54.⁵³ Using a half-standard deviation–based approach, the MCID for each VFQ-25 domain ranged from 8.80 (general vision) to 14.40 (role difficulties), producing a composite score MCID of 6.13 points. A standard error of measurement (SEM) approach yielded similar MCID estimates from 8.79 (driving) to 14.04 (role difficulties), with a composite score MCID estimate of 3.33 points.⁵³ The MCID for the near vision and distance vision subscales were 10.24 and 11.07, respectively. A standard error of measurement (SEM) approach yielded similar MCID estimates from 8.79 (driving) to 14.04 (role difficulties), with a composite score MCID estimate of 3.33 points. This technique lowered the MCID estimates for the near and distance vision domains, which were reported as 9.17 and 10.19, respectively.