Session Title: Vascular Diseases and Diabetic Retinopathy IV
Session Date/Time: Sunday 20/09/2015 | 09:00-10:30
Paper Time: 10:04
First Author: : M.Ribeiro PORTUGAL
Co Author(s): : C. Oliveira C. Neves J. Ramos H. Ferreira J. CunhaVaz
PURPOSE:To describe the procedures of a non-mydriatic diabetic retinopathy (DR) screening program in the Central Region of Portugal and the added value of the usage of the Retmarker automated “disease”/“no disease” analysis (Retmarker SA, Coimbra, Portugal), combined with a simplified grading scale focusing on the early identification of patients with signals of treatable pathology for which there is immediate, well defined, treatment available (maculopathy or proliferative DR).
DR screening program covering the Central Region of Portugal, performed by photographers equipped with portable non-mydriatic cameras that relocate between local health centres (12 months turnaround aim). Diabetic patients, type 1 and 2, are selected by their primary care unit/physician, excluding those with previous laser treatments or unable to collaborate.
The photographer performs visual acuity, fills in a diabetic condition form and acquires two 45-degree fundus images per eye (macular/F2 and nasal/F1). This information is sent to the Coimbra Ophthalmology Reading Center where it is analysed (including existing images from previous visits) using the Retmarker automated “disease”/“no disease” classification software. “Disease” cases and quality controls are then graded by trained and experienced non-ophthalmological graders, supervised by senior graders (ophthalmologists). Human grading uses an intentionally simplified grading scale whose aim is to focus on identifying patients for which there is immediate, well defined, treatment available (maculopathy or proliferative DR): R0 – no retinopathy; RL – non-proliferative DR (without maculopathy); M – maculopathy; RP – proliferative DR. Cases can also be marked as NC – not classifiable, and an optional repetition procedure can be solicited. Patients with M or RP are referred for an ophthalmologist appointment for potential treatment. Those who do not need immediate referral should return 1-year later for rescreening. To ensure the safety of the automated system, a configurable percentage of the patients not requiring human grading are nevertheless classified blindly. The human intra- and inter-graders agreement and the specificity and sensibility of the grading system are also evaluated.
This DR screening program has been continuously running since 2001 (208.498 screened patients). Since the introduction of the automated analysis in July 2011, 55.397 patients were screened with the following distribution: R0 – 72.8%; RL – 21.1%; M – 2.2%; RP – 0.2% and NC – 3.7%. The automated system alone demonstrated initially a sensitivity of 95.94% and a specificity of 35.75%, but with the introduction of new cameras more recent results show a sensitivity of 100% and specificity of 69.84% on the fully automated technology. In terms of the overall screening program where additional factors (eg. quality controls) are considered, the new cameras have translated into an increase of 60% on the patients requiring only automated analysis. The intra-grader analysis showed an agreement of 98.92% and a sensitivity and specificity of 100% and 99.51%, respectively. On the other hand, the inter-grader analysis showed an overall agreement of 96.83% with a sensitivity and specificity of 97.88% and 98.65%, respectively.
Our results show adequate levels of detection of maculopathy (2.2%) and proliferative retinopathy (0.2%). In the literature, the rates of referable retinopathy have been consistent, ranging from 1.9 to 5.1%. A 2015 study from Wales reports (type 2 diabetes) a sight-threatening DR prevalence of 2.9%. The results obtained are similar before and after the introduction of the automated system, indicating that Retmarker does not decrease sensitivity/specificity of the screening program. It has also the advantage of creating an auditable database of screening episodes and it helps to monitor key performance indicators on the photographers’ performance and potential camera issues (allowing timely repairs). Our experience and results show that screening for DR using automated “disease”/“no disease” analysis, combined with a simplified grading scale, identifies well DR sight-threatening complications (maculopathy and proliferative DR), while decreasing human burden significantly and consequently at lower costs. The results were obtained on a daily basis, on a real life setting with a methodology that could be applied elsewhere.