Diabetic retinopathy screening in Denmark: Validation, attendance and complications in relation to systemic treatment

Background and aims:
The global occurrence of diabetes is increasing and has been described as hav-ing pandemic proportions. The prevalence is estimated to reach 643 million af-fected individuals in 2030(2). In Denmark a similar tendency is observed and the Danish Diabetes Association estimates that 467,000 Danish citizens will be affected by 2030(3). Diabetic retinopathy (DR) is the most frequent complication to diabetes and a significant concern as it can lead to severe vision loss or blindness(4). In Denmark it affects approximately 24% of individuals with diabe-tes(5). To ensure proper management and early detection of DR, Denmark has had a nationwide screening program for DR since 2013. Individuals with diabe-tes are offered to attend screening either at a practicing ophthalmologist or at a hospital-based screening unit. Results from screening are reported to the na-tional clinical quality database the Danish Registry of Diabetic Retinopathy (Dia-Base)(6). In this thesis, we have examined different aspects of the Danish screening program for DR including the quality of the contents of DiaBase (Substudy A, Paper I), the attendance in the screening program (Substudy A Paper II) as well as the potential relation between systemic interventions and DR development (Substudy B).

Methods:
We performed a clinical reliability study to examine the inter-rater agreeability in the Danish screening program for DR, to establish the validity of DR levels given at screening and represented in DiaBase (Substudy A, Paper I). The anal-ysis was done by re-grading a nationwide sample of retinal images, represent-ing randomly selected eyes of individuals who attended DR screening at both practicing ophthalmologists and hospitals. In addition to images, we collected information on primary graded DR level, screening facility and geographical screening region. All information except for the images themselves were blinded until re-graded. We utilized prevalence- and bias-adjusted kappa (PABAK) and Gwet’s agreement coefficient (AC) to determine agreement between primary and secondary grader. The remaining sub-studies were prospective register-based matched cohort studies, in which we utilized the Danish nationwide regis-ters, of which DiaBase defined our cohorts with data from 2013-2022 (2018 for Substudy A, Paper II). We characterized the individuals attending DR screening in Denmark, determined by their attendance patterns (Substudy A, Paper II). We categorized individuals as either attending, delayed or one-time attending and used a multinomial regression model to analyze risk factors for varying de-grees of non-attendance reported as odds ratios (OR) as well as a Cox regres-sion model to examine the risk of DR progression associated with delayed screening reported as hazard ratios (HR) both with 95% confidence intervals. Finally, we examined systemic interventions and their potential relation to wors-ening of DR in individuals attending the Danish screening program (Sub study B). We examined continuous subcutaneous insulin infusion (CSII) in individuals with type 1 diabetes and bariatric surgery in individuals with type 2 diabetes with logistic regression and Cox regression models resulting in OR and HR respec-tively, with 95% confidence intervals.

Results:
To establish the validity of DR levels available in DiaBase, we collected images representing 230 individuals (458 eyes). Overall agreement amongst primary and secondary grader were 93% (k=0.83). When stratified by screening facility agreement was 96% (k=0.89) and 90%(k=0.76) for practicing ophthalmologists and hospitals respectively. A sub-analysis allowing one-step-difference in grad-ing resulted in an agreement of 95.5% (k=0.93).
When examining attendance in the Danish screening program, we concluded that 53.0% followed the program as intended whereas 35.5% had one or more delayed screenings and 11.5% only attended screening once, despite the rec-ommendation to continue screenings. Individuals who had any delay in screen-ings were more than twice as likely to get a clinically significant progression of DR and individuals who had three or more delayed appointments had almost 13 times higher risk of clinically significant progression. When examining systemic treatments in relation to DR, we examined both treat-ments used for individuals affected by type 1 (CSII) and type 2 diabetes (bari-atric surgery). We identified 674 individuals with type 1 diabetes, who had re-ceived CSII treatment from 2013-2022 after their first screening in DiaBase. Our cohort were primarily young and female, and cases had comparable glycemic stability (HbA1c), other comorbidities and diabetes duration compared to con-trols. We found no difference in risk of DR worsening (HR 1.05 [95%CI 0.91;1.22], p=0.48, in individuals treated with CSII.
We identified 553 individuals with type 2 diabetes who underwent bariatric sur-gery after first screening in DiaBase. The cohort was primarily young and fe-male and cases had more comorbidities, shorter diabetes duration as well as more frequent use of antidiabetic and antihypertensive medication compared to individuals using MDI. We found no increased risk of DR worsening in individu-als who underwent bariatric surgery, neither short- nor long-term (OR 0.41 [CI 95% 0.13-1.33] p=0.14 and OR 0.71 [CI 95% 0.34-1.46] p=0.35 respectively) compared to individuals who did not.

Conclusions:
The Danish screening program for DR can be regarded as effective in aiding prevention of sight-threatening progression in DR as long as individuals attend screenings as recommended, as delays in screenings are associated with in-creased risk of progression. The data collected from screening are of high valid-ity and it can be trusted that screening ophthalmologists grade DR levels to with a high degree of correctness. This ensures that individuals are referred for timely treatment when needed, as well as attest to the high quality of data in DiaBase, which can be used confidently for register-based studies. Systemic treatments in both type 1 (insulin pumps) and type 2 diabetes (bariatric surgery) were found to be safe in regards to DR, with no increased risk of worsening. However, pre-surgical glycemic control was relatively good, in both groups which must be taken into account, when interpreting the results and for the sake of reproducibility in different geographical populations.