Suboptimal diabetic eye disease screening is a major cause of preventable vision loss. Screening barriers include mydriasis and the need for dedicated screening appointments. The Clearsight trial assessed whether nonmydriatic ultra-widefield (NM UWF) screening on the day of a diabetes clinic visit improved detection of clinically important eye disease versus usual screening.
This single-center, randomized, parallel-group controlled trial was conducted at St. Joseph’s Health Care, London, Ontario, Canada. Adults with diabetes due for screening were randomized to same-day, on-site screening (NM UWF imaging) on the day of a scheduled diabetes clinic visit or usual screening (encouraged to arrange optometrist screening). The primary outcome was detection of actionable eye disease (AED), defined as the need for an ophthalmology referral or increased ocular surveillance. The primary analysis (modified intention-to-screen) compared the proportions of AED between groups within 1 year of enrollment.
Of 740 participants randomized between 7 March 2016 and 17 April 2019, 335 on-site screening and 323 usual screening participants met criteria for the primary analysis. More AED was detected in the on-site screening group than in the usual screening group (50 of 335 [14.9%] vs. 22 of 323 [6.8%]; adjusted odds ratio 2.51; 95% CI 1.49–4.36). The number needed to screen by on-site screening in order to detect 1 additional patient with AED was 13 (95% CI 8–29).
Same-day, on-site screening by NM UWF imaging increased the detection of clinically important diabetic eye disease versus usual screening. Integration of NM UWF imaging into routine diabetes clinic visits improved screening adherence and has the potential to prevent vision loss.
Introduction
Diabetic eye disease includes diabetic retinopathy (DR) and diabetic macular edema (DME) and affects >100 million people (1). Because diabetic eye disease may be asymptomatic, regular screening can improve outcomes by linking early detection to vision-protecting treatments, including laser photocoagulation, intraocular anti-vascular endothelial growth factor treatment, and strict glycemic and blood pressure control (2). Practice guidelines therefore recommend screening with mydriasis in common places where there is a need for many people to have dedicated eye examinations. For example, Diabetes Canada recommends dilated pupil direct ophthalmoscopy or indirect slit-lamp fundoscopy, or digital fundus photographs through dilated or undilated pupils, by an optometrist or ophthalmologist (2). The American Diabetes Association recommends ophthalmologist or optometrist dilated pupil examination and, as a second choice, retinal photographs read by trained eye care providers (3). The U.K. national screening programs use one- or two-field 45° digital fundus photography, with or without mydriasis (4–6).
Despite the benefit, diabetic eye screening adherence is poor (7–9). Barriers include mydriasis and the time required for dedicated eye appointments (10). Nonmydriatic ultra-widefield (NM UWF) imaging captures ∼80% of the retina in one image (11) and can be obtained by a trained technician in minutes and read remotely by a retinal specialist. These advantages are not offset by unacceptable false-negative and false-positive errors as those concern clinically important DR. Studies comparing NM UWF imaging with other screening tests, including the reference of standard 7-field stereoscopic fundus photography, have found that the capacity of NM UWF imaging to detect DR is strong based on sensitivities, specificities, positive and negative predictive values, and agreement levels by the κ statistic (12–15). On the other hand, while NM UWF imaging can detect clinically significant DME, it may not have enough sensitivity to forego other tests, including, as the reference standard, optical coherence tomography (OCT) (16).
The potential for NM UWF imaging to remove screening barriers without undue losses in diagnostic accuracy suggests it could better improve diabetes eye care versus other screening strategies. We hypothesized that NM UWF imaging at routinely scheduled diabetes clinic visits (“same-day” screening) in patients due for screening would improve detection of clinically important eye disease over usual screening. We report here the Clearsight trial, in which, as our primary aim, we tested this hypothesis by randomizing patients to NM UWF imaging or to usual screening. We designed the trial to answer the pragmatic question “In the real world, is screening by NM UWF imaging more effective than usual screening?” As a secondary aim, we assessed the capacity of NM UWF imaging to detect DME by also randomizing patients in the NM UWF imaging group to undergo OCT or not.
Research Design and Methods
Study Design and Participants
We conducted a randomized, parallel-group, superiority trial at St. Joseph’s Health Care London (SJHC) in London, Ontario, Canada. SJHC is a tertiary care facility (catchment population ∼2 million) to which adults with diabetes are referred and followed by endocrinologists. A study assistant recruited patients at routinely scheduled diabetes clinic visits. Patients due for screening by Diabetes Canada guidelines were eligible. We included patients age ≥18 years with type 1 diabetes for ≥5 years or type 2 diabetes of any duration, whose last eye examination was ≥12 months before entry according to patient report. We excluded patients under active ophthalmology follow-up, those judged unlikely to adhere to study requirements, including projected life expectancy of <12 months, and those unable to provide informed consent. For all randomized patients, we contacted their optometrist after randomization to confirm their last eye examination date. Patients whose last screen occurred <12 months pre-randomization remained in the study but were removed from the primary analysis. The trial protocol (see the Supplementary Appendix) was based on a pilot study (17). The Western University Health Sciences Research Ethics Board approved the trial.
We randomized patients 1:1 to on-site screening or usual screening stratified by glycated hemoglobin (HbA1c) (≤7.0% [53 mmol/mol] vs. >7.0% [53 mmol/mol]) and prior diabetic eye disease treatment (one or more of laser photocoagulation, vitrectomy, or intraocular anti-vascular endothelial growth factor vs. none). Patients allocated to on-site screening were also randomized 1:1 to OCT or no OCT. Both randomization schedules were computer generated by the study biostatistician (N.S.K.) and uploaded to Research Electronic Data Capture (REDCap), a secure web application. No one else had access to the schedules. The study assistant randomized patients using REDCap’s randomization module (18,19).
Patients and study personnel responsible for obtaining and reading NM UWF and OCT images could not be masked to group allocation. To limit bias during follow-up because of unequal use of treatments that affect diabetic eye disease or unequal eye surveillance, clinicians involved in patients’ care (endocrinologists, family physicians, optometrists, and ophthalmologists) were not informed of group assignments, and we encouraged patients to not disclose this information.
Procedures
For patients randomized to usual screening, we replicated what normally happens in our region. In Ontario, annual diabetes eye examinations are paid for by a publicly funded health care system, with most done by optometrists (7). The public system covers an examination that includes mydriasis and indirect ophthalmoscopy; the cost of any retinal imaging is off-loaded to patients. For usual screening patients, we explained how eye examinations help prevent vision loss and encouraged them to visit an optometrist. Patients without an optometrist were given a local optometrist contact list. We had no input into optometrists’ screening methods. At each diabetes clinic visit within 12 months post-randomization, patients’ screening status was reviewed, and if still overdue, patients were encouraged to arrange optometry examination.
On-site screening group participants underwent NM UWF 100° and 200° imaging on the day of randomization using the Optos 200Tx (Optos PLC) in the SJHC Ophthalmology Department. NM UWF images were taken by a trained assistant per protocol (Fig. 1) and read by the study ophthalmologist (J.R.G.). To help isolate any effect of NM UWF imaging over usual screening, we gave the same encouragement to arrange an eye examination to both usual screening and on-site screening patients.
On-site screening patients randomized to OCT underwent OCT within 3 months post-randomization in the SJHC Ophthalmology Department (limited OCT access precluded same-day imaging). OCT images were obtained using the ZEISS Cirrus (Zeiss) by a trained technician per protocol (Fig. 1) and read by the study ophthalmologist. The study ophthalmologist read NM UWF and OCT images in batches, where for most patients who underwent both procedures, the NM UWF images were read several days before the OCT images.
Study data were collected and managed using REDCap (18,19). The study assistant collected baseline data from patients and medical records. We followed patients for the primary outcome for up to 12 months post-randomization. Patients’ diabetes management including interventions that affect diabetic eye disease (antihyperglycemic, antihypertensive, and lipid-lowering medications, including fenofibrate; smoking cessation), was left to their diabetes care team. The study assistant collected data on medications, HbA1c, blood pressure, and smoking at follow-up clinic visits.
For patients in the on-site and usual screening groups who indicated they had received an eye examination within 12 months of randomization, we requested a facsimile copy of the examination report from their optometrist. For patients in both groups for whom we had no record of an eye examination 12 months post-randomization, we contacted their optometrist and confirmed that screening was not done. For those patients without a named optometrist at entry, we telephoned the patient and confirmed they had not seen an optometrist within 12 months post-randomization, or if they had, we obtained the report from their new optometrist.
We randomized patients between 7 March 2016 and 17 April 2019. On 17 March 2020, the Ontario government enacted a Declaration of Emergency in response to the coronavirus disease 2019 (COVID-19) pandemic. There were 36 on-site screening patients and 34 usual screening patients randomized between 17 March 2019 and 17 April 2019. Because the Declaration led to optometry office closures, we extended the eye examination follow-up to 31 December 2020 for the 34 usual screening patients.
Outcomes
The primary outcome was actionable eye disease (AED), defined as diabetes-related eye disease for which ophthalmology referral or increased optometrist surveillance (reexamination in <12 months) was indicated. The primary outcome could be ascertained at any point between randomization and 12 months later, but the specific AED definition differed by group.
For usual screening patients, we required a written report that included optometrist findings and disposition recommendation. In our pilot study, we found that most local optometrists reported findings with a template based on the widely used Diabetic Retinopathy Disease Severity Scale (DRDSS) (20) (Supplementary Appendix, p. 18). One or more of the following was required for AED: 1) moderate or severe non-proliferative DR; 2) any proliferative DR; 3) clinically significant macular edema; 4) ophthalmologist referral for diabetes-related indications; and 5) recommendation for reexamination in <12 months. For on-site screening patients, we based AED on our study ophthalmologist’s NM UWF image interpretation. One or more of the following was required for AED: 1) for retinopathy: microaneurysms/hemorrhages in 4 quadrants, intraretinal microvascular abnormalities ≥1 quadrant, venous beading ≥2 quadrants, neovascularization elsewhere or neovascularization of the disc, or vitreous hemorrhage; and 2) for maculopathy: microaneurysms, retinal hemorrhages, or exudates within one disc diameter of the fovea.
We prespecified secondary outcomes of interest related to screening adherence and DME. For screening adherence, we assessed 1) the proportions of patients with screening completed within 12 months of randomization by on-site and usual screening strategies; and 2) the proportion of on-site screening patients who also had an optometry examination within 12 months of randomization. For DME we assessed 1) the proportions of patients with DME detected within 12 months of randomization in the usual screening and on-site screening (NM UWF imaging without regard to OCT) groups; and 2) the proportions of on-site screening patients with DME detected by NM UWF imaging alone and by NM UWF imaging plus OCT. The criteria for DME by OCT were the presence of ≥1 of intraretinal cysts, intraretinal exudates, and subretinal fluid.
We foresaw no risks from the study procedures. For on-site screening patients in whom NM UWF imaging or OCT detected AED, we sent letters to their optometrist and endocrinologist to inform them.
Statistical Analysis
The analysis plan (Supplementary Appendix, pp. 22–42) was set before the data were locked. We planned to randomize 740 patients to identify a 5% absolute increase (5% two-tailed α error, 80% power) in AED detection rates between the on-site screening and usual screening groups. We judged this difference as clinically important based on a projected number needed to screen by on-site screening versus usual screening of 20 patients to detect 1 additional patient with AED. We also determined that with 370 on-site screening patients randomized to OCT or not, we could identify a 14% increase (5% two-tailed α error, 80% power) in DME detection by NM UWF imaging plus OCT versus NM UWF imaging alone.
To control for multiplicity, we followed a hierarchy of five sequential significance tests requiring P ≤ 0.05 to advance to the next test. The primary analysis was based on AED detection within 12 months post-randomization. All patients were analyzed in the group to which they were randomized, unless they withdrew consent, or their last eye examination was confirmed to be <12 months pre-randomization (modified intention-to-screen [mITS] population). We compared proportions of AED between on-site screening and usual screening groups using unadjusted and adjusted logistic regressions (adjustment variables: baseline HbA1c, prior diabetic eye disease treatment, and smoking) with profile likelihood based CIs. We estimated the number needed to screen in order to detect 1 patient with AED according to on-site screening versus usual screening. We examined potential subgroup differences by testing for interactions between the adjustment variables and screening method.
The second hierarchal test compared screening adherence rates between the on-site screening and usual screening groups using logistic regression with modified Poisson regression. The remaining tests used logistic regression to compare DME detection rates between 1) the on-site screening versus usual screening groups (third test); 2) the on-site screening patients who had OCT versus usual screening patients (fourth test); and 3) the on-site screening patients who had OCT versus the on-site screening patients who did not (fifth test). All other reported CIs used 95% coverage and were not adjusted for multiplicity.
We assessed the robustness of the primary analysis results in sensitivity analyses. In one analysis, we included patients with confirmed last eye examinations <12 months pre-randomization. In a second analysis, we assessed the impact of the COVID-19 pandemic on AED detection by including optometrist examination findings up until 31 December 2020 for usual screening patients who fulfilled the mITS criteria but whose follow-up ended after the COVID-19 Declaration of Emergency. In a third analysis, we assessed the impact of missing AED data. We used the missing indicator method to analyze missingness patterns (21). We compared AED rates under assumptions about AED in patients missing data that completely favored on-site screening (“best case”) or usual screening (“worst case”). We also assessed the impact of missing data on AED rates using multiple imputations by the chained equation approach (22).
We did not constitute a data monitoring committee or perform interim analyses. All analysis were done using R 4.1.2 software. The trial was registered with ClinicalTrials.gov, NCT02579837.
Results
From 7 March 2016 to 17 April 2019, we screened 16,939 medical records and randomized 740 individuals to on-site screening and usual screening (Fig. 2). Within the on-site screening group, 184 patients were randomized to NM UWF imaging alone, and 186 were randomized to NM UWF imaging plus OCT. A total of 82 patients had a confirmed last eye examination <12 months pre-randomization (n = 79) or withdrew consent post-randomization (n = 3), leaving 323 usual screening and 335 on-site screening patients in the mITS analysis.
Consolidated Standards of Reporting Trials (CONSORT) participant flow diagram for participant inclusion—intention-to-treat and mITT populations. mo, months.
Consolidated Standards of Reporting Trials (CONSORT) participant flow diagram for participant inclusion—intention-to-treat and mITT populations. mo, months.
The presence or absence of AED could be determined in 332 of 335 on-site screening patients (99.1%) based on adequate NM UWF images and in 274 of 323 usual screening patients (84.8%) based on optometrist examination within 12 months post-randomization. Among the 166 on-site screening patients in the mITS population randomized to OCT, 140 (84.3%) completed OCT within 3 months of their NM UWF imaging.
Patient baseline characteristics in the on-site screening and usual screening groups were similar (Table 1). Across both groups, the mean age was 50 years, 17% were smokers, 8% had previous diabetic eye disease treatment, 78% had HbA1c >7.0% (53 mmol/mol), 75% were taking insulin, and 47% had type 1 diabetes. Baseline use of antihyperglycemic, antihypertensive, and lipid-lowering medications, including fibrates, was also similar between groups.
Participant baseline characteristics
. | On-site screening . | Usual screening . | Total (mITS population) . | Excluded* . |
---|---|---|---|---|
. | n = 335 . | n = 323 . | N = 658 . | n = 79 . |
Age (years) | 49.5 (35.7, 60.6) | 50.0 (33.5, 60.2) | 49.7 (34.8, 60.4) | 56.2 (44.7, 68.0) |
Male sex | 192 (57.3) | 186 (57.6) | 378 (57.4) | 41 (51.9) |
Type 1 diabetes | 156 (46.6) | 155 (48.0) | 311 (47.3) | 38 (48.1) |
Duration of diabetes (years) | 15.9 (8.2, 23.3) | 13.5 (8.0, 22.9) | 15.0 (8.2, 23.2) | 18.4 (12.1, 29.7) |
Prior treatment for diabetes-related eye disease | 24 (7.2) | 26 (8.0) | 50 (7.6) | 6 (7.6) |
Interval–baseline EE to enrollment (months) | 22.2 (15.3, 33.3) | 22.0 (15.3, 32.6) | 22.1 (15.3, 33.2) | 10.8 (8.2, 11.4) |
Fundoscopy done at enrollment visit | 16 (4.8) | 21 (6.5) | 37 (5.6) | 3 (3.8) |
Baseline HbA1c (%) | 8.2 (7.1, 9.2) | 8.2 (7.2, 9.5) | 8.2 (7.2, 9.3) | 7.7 (6.9, 8.5) |
HbA1c ≤7% (53 mmol/mol) | 74 (22.1) | 70 (21.7) | 144 (21.9) | 23 (29.1) |
HbA1c >7% (53 mmol/mol) | 261 (77.9) | 253 (78.3) | 514 (78.1) | 56 (70.9) |
Interval–baseline HbA1c to enrollment (months) | 0.8 (0.3, 3.8) | 0.9 (0.3, 4.4) | 0.8 (0.3, 4.1) | 0.6 (0.3, 3.7) |
Systolic blood pressure (mmHg) | 129 (117, 144) | 129 (118, 142) | 129 (118, 143) | 124 (115, 134) |
Diastolic blood pressure (mmHg) | 76 (68, 84) | 77 (71, 85) | 77 (70, 85) | 73 (68, 77) |
Current smoking | 53 (15.8) | 58 (18.0) | 111 (16.9) | 9 (11.4) |
Amount (packs per day) | 0.5 (0.2, 1.0) | 0.5 (0.5, 1.0) | 0.5 (0.5, 1.0) | 0.5 (0.4, 0.6) |
Prior smoking | 89 (26.6) | 77 (23.8) | 166 (25.2) | 27 (34.2) |
History (pack-year) | 10 (5, 25) | 8 (4, 23) | 10 (4, 25) | 16 (7, 30) |
On insulin | 249 (74.3) | 246 (76.2) | 495 (75.2) | 63 (79.7) |
Other diabetes medication | ||||
Metformin (or Glumetza) | 115 (34.3) | 115 (35.6) | 230 (35.0) | 32 (40.5) |
Dipeptidyl peptidase 4 inhibitor | 58 (17.3) | 64 (19.8) | 122 (18.5) | 15 (19.0) |
Sulfonylurea | 45 (13.4) | 47 (14.6) | 92 (14.0) | 14 (17.7) |
Meglitinide | 1 (0.3) | 0 (0) | 1 (0.2) | 0 (0) |
Sodium–glucose cotransport 2 inhibitor | 44 (13.1) | 48 (14.9) | 92 (14.0) | 14 (17.7) |
Glucagon-like peptide 1 agonist | 16 (4.8) | 20 (6.2) | 36 (5.5) | 8 (10.1) |
α-Glucosidase inhibitor | 1 (0.3) | 2 (0.6) | 3 (0.5) | 1 (1.3) |
Thiazolidinedione | 1 (0.3) | 3 (0.9) | 4 (0.6) | 0 (0) |
Lipid-lowering medication | ||||
Statin | 167 (49.9) | 149 (46.1) | 316 (48.0) | 50 (63.3) |
Fibrate | 8 (2.4) | 7 (2.2) | 15 (2.3) | 3 (3.8) |
Ezetimibe | 27 (8.1) | 16 (5.0) | 43 (6.5) | 9 (11.4) |
Bile acid sequestrant | 4 (1.2) | 10 (3.1) | 14 (2.1) | 1 (1.3) |
Niacin | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
Antihypertensive medication | ||||
ACE-inhibitor or ARB | 147 (43.9) | 125 (38.7) | 272 (41.3) | 41 (51.9) |
Calcium-channel blocker | 47 (14.0) | 37 (11.5) | 84 (12.8) | 15 (19.0) |
Diuretic | 46 (13.7) | 35 (10.8) | 81 (12.3) | 12 (15.2) |
Other | 62 (18.5) | 56 (17.3) | 118 (17.9) | 14 (17.7) |
Lipid profile (mmol/L) | ||||
Total cholesterol | 4.0 (3.5, 4.9) | 4.3 (3.5, 4.9) | 4.2 (3.5, 4.9) | 3.7 (3.2, 4.7) |
LDL-cholesterol | 2.0 (1.5, 2.7) | 2.1 (1.5, 2.6) | 2.1 (1.5, 2.7) | 1.7 (1.3, 2.4) |
HDL-cholesterol | 1.2 (0.9, 1.6) | 1.3 (1.0, 1.6) | 1.2 (1.0, 1.6) | 1.2 (1.0, 1.5) |
Triglycerides | 1.4 (0.9, 2.4) | 1.4 (1.0, 2.2) | 1.4 (0.9, 2.3) | 1.2 (0.8, 1.9) |
Total cholesterol-to-HDL ratio | 3.2 (2.5, 4.4) | 3.2 (2.6, 4.1) | 3.2 (2.5, 4.2) | 2.9 (2.3, 3.6) |
. | On-site screening . | Usual screening . | Total (mITS population) . | Excluded* . |
---|---|---|---|---|
. | n = 335 . | n = 323 . | N = 658 . | n = 79 . |
Age (years) | 49.5 (35.7, 60.6) | 50.0 (33.5, 60.2) | 49.7 (34.8, 60.4) | 56.2 (44.7, 68.0) |
Male sex | 192 (57.3) | 186 (57.6) | 378 (57.4) | 41 (51.9) |
Type 1 diabetes | 156 (46.6) | 155 (48.0) | 311 (47.3) | 38 (48.1) |
Duration of diabetes (years) | 15.9 (8.2, 23.3) | 13.5 (8.0, 22.9) | 15.0 (8.2, 23.2) | 18.4 (12.1, 29.7) |
Prior treatment for diabetes-related eye disease | 24 (7.2) | 26 (8.0) | 50 (7.6) | 6 (7.6) |
Interval–baseline EE to enrollment (months) | 22.2 (15.3, 33.3) | 22.0 (15.3, 32.6) | 22.1 (15.3, 33.2) | 10.8 (8.2, 11.4) |
Fundoscopy done at enrollment visit | 16 (4.8) | 21 (6.5) | 37 (5.6) | 3 (3.8) |
Baseline HbA1c (%) | 8.2 (7.1, 9.2) | 8.2 (7.2, 9.5) | 8.2 (7.2, 9.3) | 7.7 (6.9, 8.5) |
HbA1c ≤7% (53 mmol/mol) | 74 (22.1) | 70 (21.7) | 144 (21.9) | 23 (29.1) |
HbA1c >7% (53 mmol/mol) | 261 (77.9) | 253 (78.3) | 514 (78.1) | 56 (70.9) |
Interval–baseline HbA1c to enrollment (months) | 0.8 (0.3, 3.8) | 0.9 (0.3, 4.4) | 0.8 (0.3, 4.1) | 0.6 (0.3, 3.7) |
Systolic blood pressure (mmHg) | 129 (117, 144) | 129 (118, 142) | 129 (118, 143) | 124 (115, 134) |
Diastolic blood pressure (mmHg) | 76 (68, 84) | 77 (71, 85) | 77 (70, 85) | 73 (68, 77) |
Current smoking | 53 (15.8) | 58 (18.0) | 111 (16.9) | 9 (11.4) |
Amount (packs per day) | 0.5 (0.2, 1.0) | 0.5 (0.5, 1.0) | 0.5 (0.5, 1.0) | 0.5 (0.4, 0.6) |
Prior smoking | 89 (26.6) | 77 (23.8) | 166 (25.2) | 27 (34.2) |
History (pack-year) | 10 (5, 25) | 8 (4, 23) | 10 (4, 25) | 16 (7, 30) |
On insulin | 249 (74.3) | 246 (76.2) | 495 (75.2) | 63 (79.7) |
Other diabetes medication | ||||
Metformin (or Glumetza) | 115 (34.3) | 115 (35.6) | 230 (35.0) | 32 (40.5) |
Dipeptidyl peptidase 4 inhibitor | 58 (17.3) | 64 (19.8) | 122 (18.5) | 15 (19.0) |
Sulfonylurea | 45 (13.4) | 47 (14.6) | 92 (14.0) | 14 (17.7) |
Meglitinide | 1 (0.3) | 0 (0) | 1 (0.2) | 0 (0) |
Sodium–glucose cotransport 2 inhibitor | 44 (13.1) | 48 (14.9) | 92 (14.0) | 14 (17.7) |
Glucagon-like peptide 1 agonist | 16 (4.8) | 20 (6.2) | 36 (5.5) | 8 (10.1) |
α-Glucosidase inhibitor | 1 (0.3) | 2 (0.6) | 3 (0.5) | 1 (1.3) |
Thiazolidinedione | 1 (0.3) | 3 (0.9) | 4 (0.6) | 0 (0) |
Lipid-lowering medication | ||||
Statin | 167 (49.9) | 149 (46.1) | 316 (48.0) | 50 (63.3) |
Fibrate | 8 (2.4) | 7 (2.2) | 15 (2.3) | 3 (3.8) |
Ezetimibe | 27 (8.1) | 16 (5.0) | 43 (6.5) | 9 (11.4) |
Bile acid sequestrant | 4 (1.2) | 10 (3.1) | 14 (2.1) | 1 (1.3) |
Niacin | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
Antihypertensive medication | ||||
ACE-inhibitor or ARB | 147 (43.9) | 125 (38.7) | 272 (41.3) | 41 (51.9) |
Calcium-channel blocker | 47 (14.0) | 37 (11.5) | 84 (12.8) | 15 (19.0) |
Diuretic | 46 (13.7) | 35 (10.8) | 81 (12.3) | 12 (15.2) |
Other | 62 (18.5) | 56 (17.3) | 118 (17.9) | 14 (17.7) |
Lipid profile (mmol/L) | ||||
Total cholesterol | 4.0 (3.5, 4.9) | 4.3 (3.5, 4.9) | 4.2 (3.5, 4.9) | 3.7 (3.2, 4.7) |
LDL-cholesterol | 2.0 (1.5, 2.7) | 2.1 (1.5, 2.6) | 2.1 (1.5, 2.7) | 1.7 (1.3, 2.4) |
HDL-cholesterol | 1.2 (0.9, 1.6) | 1.3 (1.0, 1.6) | 1.2 (1.0, 1.6) | 1.2 (1.0, 1.5) |
Triglycerides | 1.4 (0.9, 2.4) | 1.4 (1.0, 2.2) | 1.4 (0.9, 2.3) | 1.2 (0.8, 1.9) |
Total cholesterol-to-HDL ratio | 3.2 (2.5, 4.4) | 3.2 (2.6, 4.1) | 3.2 (2.5, 4.2) | 2.9 (2.3, 3.6) |
Data are presented as medians (quartile 1, quartile 3) or n (%). ARB, angiotensin receptor blocker; EE, eye examination.
*79 with baseline eye examination <12 months preenrollment excluded (does not include 3 withdrawals).
Table 2 shows the primary outcome results. AED detection was significantly higher among on-site screening versus usual screening patients (50 of 335 [14.9%] vs. 22 of 323 [6.8%], respectively; 95% CI on the absolute difference 3–12%; P = 0.001). This difference yielded a number needed to screen of 13 patients (95% CI 8–29) in order to detect 1 additional patient with AED by on-site screening over usual screening. Adjustment for baseline HbA1c, prior diabetic eye disease treatment, and smoking did not materially change the observed increase in AED detection by on-site screening nor were there apparent differences in AED detection between the on-site screening and usual screening groups within the prespecified subgroups (Supplementary Fig. 1). Similarly, in almost all of the preplanned sensitivity analyses, the point estimates for AED detection rates favored on-site screening and had 95% CIs that did not overlap a cut point of “no effect” (Supplementary Table 1).
Primary and secondary outcomes: on-site screening versus usual screening groups
. | On-site screening n = 335 . | Usual screening n = 323 . | Adjusted* OR (95% CI) . | Adjusted* RD (95% CI) . | P value for adjusted OR . | |
---|---|---|---|---|---|---|
Primary outcome | ||||||
AED detected within 12 months of enrollment | 50/335 (14.9) | 22/323 (6.8) | 2.51 (1.49–4.36) | 0.08 (0.03–0.12) | 0.001 | |
RR (95% CI) | P value for hierarchical testing | |||||
Secondary outcomes | ||||||
Adherence to the primary screening method within 12 months of enrollment | 332/335 (99.1) | 274/323 (84.8) | 1.17 (1.11–1.22) | <0.001 | ||
DME detected by primary screening method within 12 months of enrollment | 38/335 (11.3) | 7/323 (2.2) | 5.23 (2.37–11.55) | <0.001 | ||
UWF alone n = 169† | UWF+OCT n = 166‡§ | |||||
DME detected by UWF imaging or OCT† (on-site screening group) vs. follow-up eye examination within 12 months of enrollment (usual screening group) | NA | 21/166 (12.6) | 7/323 (2.2) | 5.84 (2.53–13.45) | <0.001 | |
DME detected by UWF imaging alone vs. UWF or OCT† | 17/169 (10.1) | 21/166 (12.6) | NA | 1.26 (0.69–2.30) | 0.46 |
. | On-site screening n = 335 . | Usual screening n = 323 . | Adjusted* OR (95% CI) . | Adjusted* RD (95% CI) . | P value for adjusted OR . | |
---|---|---|---|---|---|---|
Primary outcome | ||||||
AED detected within 12 months of enrollment | 50/335 (14.9) | 22/323 (6.8) | 2.51 (1.49–4.36) | 0.08 (0.03–0.12) | 0.001 | |
RR (95% CI) | P value for hierarchical testing | |||||
Secondary outcomes | ||||||
Adherence to the primary screening method within 12 months of enrollment | 332/335 (99.1) | 274/323 (84.8) | 1.17 (1.11–1.22) | <0.001 | ||
DME detected by primary screening method within 12 months of enrollment | 38/335 (11.3) | 7/323 (2.2) | 5.23 (2.37–11.55) | <0.001 | ||
UWF alone n = 169† | UWF+OCT n = 166‡§ | |||||
DME detected by UWF imaging or OCT† (on-site screening group) vs. follow-up eye examination within 12 months of enrollment (usual screening group) | NA | 21/166 (12.6) | 7/323 (2.2) | 5.84 (2.53–13.45) | <0.001 | |
DME detected by UWF imaging alone vs. UWF or OCT† | 17/169 (10.1) | 21/166 (12.6) | NA | 1.26 (0.69–2.30) | 0.46 |
Data are presented as n/N (%). OR, odds ratio; RD, risk difference; RR, risk ratio; NA, not applicable.
*Adjusted for baseline HbA1c, prior treatment for eye disease, baseline smoking status. †167 UWF images in 169 participants. ‡165 UWF images in 166 participants. §140 OCT images taken within 3 months of UWF images in 166 participants (excluded 3 OCT images taken >3 months from UWF images).
The secondary outcome results are also shown in Table 2. Screening adherence was significantly higher among on-site screening versus usual screening patients (332 of 335 [99.1%] vs. 274 of 323 [84.8%], respectively; P < 0.001). On-site screening patients were 1.17 times (95% CI 1.11–1.22) more likely to complete screening versus usual screening patients within 12 months of enrollment. Among on-site screening patients, 296 of 335 (88.3%) also had an optometry examination during follow-up; this did not differ significantly from the screening rate (84.8%) in the usual screening patients (P = 0.18).
Detection of DME was significantly higher among on-site screening versus usual screening patients (38 of 335 [11.3%] vs. 7 of 323 [2.2%], respectively; P < 0.001). DME was 5.23 times (95% CI 2.37–11.55) more likely to be detected by NM UWF imaging versus usual screening. DME detection was also significantly higher among on-site screening patients who had both NM UWF imaging and OCT versus usual screening patients (21 of 166 [12.6%] vs. 7 of 323 [2.2%], respectively; P < 0.001). DME was 5.84 times (95% CI 2.53–13.45) more likely to be detected by NM UWF imaging plus OCT versus usual screening. The difference in DME detection between patients who had NM UWF imaging plus OCT versus NM UWF imaging alone was not statistically significant.
Post hoc analyses are included in the Supplementary Appendix. For the primary outcome, adjusting for diabetes duration and insulin use did not improve model fit. In subgroup analysis by diabetes type, AED detection did not differ, although the relative risks of DME detection were higher in type 2 diabetes (Supplementary Table 2). On-site screening identified more patients with diabetic eye disease across all levels of DR severity versus usual screening (Supplementary Table 3).
Conclusions
In this randomized controlled trial of patients due for diabetic eye disease screening, we found that NM UWF imaging on the same day as a routinely scheduled diabetes clinic visit (on-site screening) significantly increased detection of clinically important diabetic eye disease compared with encouraging patients to book an eye care professional examination (usual screening). We defined clinically important diabetic eye disease based on DR and DME findings for which ophthalmologist referral or more frequent optometrist eye surveillance was indicated (AED). Compared with usual screening over 12 months of follow-up, we found that same-day on-site screening identified 1 additional patient with AED for every 13 screened patients. We also found in prespecified secondary hypothesis testing that on-site screening resulted in significantly higher screening adherence versus usual screening, and that on-site screening by NM UWF imaging alone and in combination with OCT significantly increased DME detection versus usual screening, but that there was no difference in DME detection between patients randomized to NM UWF plus OCT imaging and NM UWF imaging alone.
To our knowledge, this is the first randomized trial comparing the effectiveness of same-day NM UWF imaging versus usual screening to detect clinically important diabetic eye disease. At least two other trials have compared nonmydriatic retinal photography using smaller-field (not UWF) imaging to usual screening. In both, the primary outcome was not eye disease as we defined, but screening adherence (23,24). Conlin et al. (23) compared nonmydriatic 45° imaging at a primary care clinic plus usual screening to usual screening alone in 448 participants and showed that adherence to dilated eye examination was higher in the imaging group compared with usual screening (87% vs. 77%, P < 0.01). Mansberger et al. (24) compared nonmydriatic 45° imaging at routine primary care visits plus usual screening to usual screening alone in 567 participants. More imaging participants had a subsequent eye examination than usual screening participants within 6 months (94.6% vs. 43.9%, P < 0.001) and between 6 and 18 months (53% vs. 33.2%, P < 0.001) (24). The proportion of imaging participants requiring ophthalmology referral was 19.2% (<6 months) and 26.2% (6–18 months) (24), higher than our trial’s AED detection rate among on-site screening patients; however, ≥50% of these referrals were due to ungradable images, possibly because they used smaller-field imaging (24).
One reason on-site screening increased AED detection over usual screening in our trial was almost certainly because of higher screening among on-site screening (99%) versus usual screening patients (85%). Based on our pilot study patient surveys, we suspect the higher screening among on-site screening patients was because they did not need to arrange and attend a dedicated screening appointment (17). We doubt the between-group difference in screening adherence was because our approach to encourage usual screening patients to schedule eye examinations was below standard of care. We note that our usual screening group’s screening rate (85%) is comparable to that reported elsewhere (23,24) and is also much higher than Ontario screening rates (7). Because we randomized patients, it is also unlikely that a difference in tendency to adhere to screening recommendations between on-site screening and usual screening patients affected our results. The higher proportion of on-site screening patients (88%) who had an optometry examination during follow-up compared with usual screening patients (85%) may be explained by an increase in AED detection by NM UWF imaging that led to more patients being advised to see an optometrist in <12 months.
Our study strengths include the randomized design, which helped limit confounding that could lead in an observational study to a false conclusion that on-site screening was more effective than usual screening. As a second strength, we had information to assess the primary outcome in almost all randomized patients, including follow-up in >95% of usual screening patients to determine whether they had an optometrist examination. A third strength was the steps we took to limit potential biases because of unequal eye surveillance or use of treatments that affect AED between the on-site screening and usual screening groups, as a consequence of not masking patients to their assigned groups.
Our study has important limitations. It was not designed to test the effect of the screening strategies on vision. To do so would require a larger and longer trial. However, we think it a reasonable assumption that increasing AED detection will lead to better vision because of earlier vision-saving therapy use. Another limitation is the lack of a gold standard to which the accuracy of NM UWF imaging to diagnose AED could be compared. As a third limitation, we cannot be sure our results are generalizable to older adults (most participants were <60 years old, and thus more likely to have successful nonmydriatic images), to primary care settings (e.g., our proportion of type 1 diabetes [47%] was high), or to jurisdictions where access to and funding for diabetic eye screening differs from ours. Notwithstanding this limitation, it is possible NM UWF screening may be even more effective in detecting eye disease in regions with limited access to usual screening based on an eye care professional examination. Finally, we have yet to assess the cost-effectiveness of on-site screening compared with usual screening.
We saw no difference in DME detection among on-site screening patients randomized to NM UWF imaging plus OCT versus NM UWF imaging alone. This goes against evidence that indicates the sensitivity of NM UWF imaging to detect DME may not be adequate (15). However, because DME was a secondary outcome, we lacked the power to detect smaller but clinically important increases in DME detection when OCT was added to NM UWF imaging. Although OCT is currently considered an ancillary test for diabetes eye screening (25,26), a recent study showed that the incremental cost-effectiveness ratio of including OCT for DME screening in patients who had fundus photography is attractive (16).
In designing our trial, we focused more on whether on-site screening was better than usual screening under routine conditions (pragmatic trial) than ideal conditions (explanatory trial) (27). By enrolling patients at their scheduled diabetes clinic appointment, we replicated how NM UWF imaging could be offered during routine follow-up to patients due for an eye screening. We also used nonrestrictive entry criteria to enroll patients who would be eligible for NM UWF screening if it were to be included in routine care. Our approach to usual screening patients matched what currently happens in standard practice in many jurisdictions. Our primary outcome—while not a measure of visual acuity, which is of most importance to patients—was based on NM UWF imaging or standard diabetic eye screening examination findings that would be expected to lead to a change in patient care. These features may make our results more readily applicable to routine care, at least as that is provided in diabetes clinics similar to ours.
Further studies are needed, including a cost-effectiveness analysis of same-day on-site NM UWF imaging versus usual screening. NM UWF photography incurs major costs (e.g., equipment, specialized image-reading personnel) that may not be justified by the benefit.
In summary, in a referral diabetes clinic we found that NM UWF retinal imaging on the same day of a routinely scheduled appointment in patients due for an eye screening was superior to usual screening in detecting clinically important diabetic eye disease. This was associated with a higher screening rate among patients undergoing NM UWF imaging. Provision of NM UWF imaging in diabetes clinics to patients who need eye screening should be considered.
Clinical trial reg. no. NCT02579837, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.21514095.
Article Information
Acknowledgments. The authors thank Nour Abu-Romeh, Andrew Ghorashi, and Mohammed Dakroub for their assistance in recruiting participants. The authors also would like to thank their colleagues for allowing them to recruit patient participants from their diabetes clinics: Drs. Amanda Berberich, Kristin Clemens, Tisha Joy, Charlotte McDonald, Ruth McManus, Deric Morrison, Terri Paul, and Tamara Spaic. Most importantly the authors thank all of the patients who participated in the Clearsight trial.
Funding. The trial was funded by Physicians’ Services Incorporated (PSI) Foundation (grant no. 15-20). S.L.L., J.R.G., N.S.K., I.M.H., and J.L.M. report grant support from PSI Foundation for the Clearsight trial. I.M.H. reports grants from the Canadian Medical & Surgical Knowledge Translation Research Group.
The funding organization had no role in the study design, data collection, analysis, or interpretation, writing of the report, or in the decision to submit for publication.
Duality of Interest. S.L.L. reports grant support from Novartis for the Clearsight pilot study, payment or honoraria from Merck and the University of Ottawa, support for travel to meetings from Merck, and grants from Eli Lilly and Novo Nordisk (outside the submitted work). J.R.G. reports grant support from Novartis for the Clearsight pilot study. I.M.H. reports grants or contracts from Eli Lilly, Novo Nordisk, and Sanofi, consulting fees and payment or honoraria from CCRN, Insulet Corp., Boehringer Ingelheim/Eli Lilly (BI-LILLY joint venture), Eli Lilly & Co., Medtronic, Merck & Co. Inc., Novo Nordisk, and Sanofi (all outside the submitted work). J.L.M. reports participation on a Data Safety Monitoring Board for VERDICT (In ActiVE Ulcerative Colities – a RanDomIzed Controlled Trial for determination of the Optimal Treatment Target) (outside the submitted work). No other potential conflicts of interest relevant to this article were reported.
Author Contributions. S.L.L. wrote the first draft of the trial protocol. S.L.L., A.U., and J.L.M. accessed and verified the data. S.L.L. wrote the first draft of the manuscript with input from J.L.M. S.L.L., and J.L.M. conceived the study hypothesis. J.R.G., N.S.K., I.M.H., and J.L.M. provided critical revision of the protocol. E.O. recruited participants and was responsible for data collection. A.U. performed the statistical analysis. All authors had full access to all the data in the study. All authors reviewed and edited the manuscript. S.L.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented as a poster at the 2022 Diabetes Canada/Canadian Society of Endocrinology & Metabolism Professional Conference, Calgary, Alberta, Canada, 9–12 November 2022.