Retinal fundus examinations have been part of routine diabetes care for decades. When the concept of annual reviews became established, direct dilated ophthalmoscopy became an integral part of the annual review. An annual review was an important way to organize services in a reliable and reproducible way, and to ensure that all appropriate care was provided efficiently and effectively. Thus, annual retinal examinations were established as a pragmatic option, but not an evidence-based one.
Over the last 20 years, digital retinal photography and national retinal screening programmes have been implemented, both supported by information technology (IT) and call–recall systems. This permits the opportunity for more detailed and sophisticated approaches to organizing routine retinal screening. However, what is the evidence that would support the possibility of varying screening intervals for some patients?
The Wisconsin trial from the 1970s showed that the incidence of proliferative retinopathy was 1.2% after 10 years for individuals assumed to have Type 1 diabetes , and 2% at 5 years  for those with presumed Type 2 diabetes. In Iceland, 9 years after diagnosis, virtually no patients with Type 1 diabetes developed proliferative retinopathy, and even at 15–19 years the rate was only 2% . For patients with Type 2 diabetes, the rate was approxiately 8% at 10–14 years.
These data resulted in a biennial slit-lamp screening programme being introduced for patients with no baseline retinopathy in Iceland. Ten years after introducing this system, a review of 296 patients identified 23 cases of pre-proliferative disease, with four cases of proliferative retinopathy and four cases of clinically significant macular oedema. All cases had already been referred to the ophthalmology clinic by the time laser photocoagulation was required . This indicated that no cases of sight-threatening eye disease had been missed and the slit-lamp screening programme was continued.
The UK Prospective Diabetes Study (UKPDS) showed that, in 2316 newly diagnosed patients with Type 2 diabetes, only 0.2% progressed to need laser photocoagulation at 3 years, whilst the number was 1.1% at 6 years . This again demonstrated a very slow progression rate of diabetes-related sight-threatening retinopathy, admittedly in a group of patients with newly diagnosed diabetes.
In Lund, Sweden, the incidence of sight-threatening diabetic eye disease was very low for patients with no baseline retinopathy. In 1322 patients with Type 2 diabetes and no baseline retinopathy, only one patient required laser photocoagulation for macular oedema at 3 years of follow-up . The mean HbA1c for these patients was 46 mmol/mol (6.4%), indicating these patients had excellent control. However, on the basis of these figures, a 2-yearly screening programme using retinal photography was reaffirmed in Lund, with consideration of a 3-year screening programme for such patients.
In the Australian Blue Mountain Study of 139 patients with newly diagnosed diabetes of 50 years of age or older, only 4.1% developed sight-threatening retinopathy at 5 years, but all such patients had some baseline retinopathy .
These studies were relatively small. The Liverpool eye study was the first study to look at this systematically in large numbers of patients in a standard routine clinical population. In 9890 patients with Type 2 diabetes who had no baseline retinopathy, 0.5% developed proliferative retinopathy or treatable macular disease at 1 year, 1.4% at 2 years and 2.8% at 3 years . The numbers were remarkably similar for 501 patients with Type 1 diabetes, with rates of progression being 0.6, 1.6 and 2.6% at 1, 2 and 3 years, respectively . This was a surprise given data from other studies which showed faster progression in patients with Type 1 diabetes .
In early 2012, two large studies reported, one from the Welsh and one from the English retinal screening programmes [11,12]. The Welsh retinal screening programme has followed up 57 199 patients with Type 2 diabetes and no baseline retinopathy for 4 years. The number who subsequently developed referable retinopathy was 0.2% at 1 year, with a total of 1.16% after 4 years. For a subgroup of patients with Type 2 diabetes on insulin and duration of 10 years or more, the rates were much greater .
Similar data were obtained from England in 16 444 patients with Type 2 diabetes who had no baseline retinopathy, 4% developed pre-proliferative retinopathy, 0.6% sight-threatening maculopathy and 0.7% proliferative retinopathy at 5 years. Rates of progression were much greater for patients who had background retinopathy at baseline . Further preliminary data from Scotland in 112 000 patients show similar results (Unpublished data).
Thus, in summary, the proportion of patients with diabetes who progress towards referable eye disease is very low at about 0.2% after 1 year and the number with treatable disease is even less. The risk of missing important treatable retinopathy would be low if a 2-year screening interval was introduced for patients with no baseline retinopathy. The risk may be reduced further by continuing annual screening for patients with diabetes duration more than 10 years, and possibly for patients with Type 1 diabetes where the evidence is less clear. Glycaemic control may be another factor that could be considered, but microvascular risk is probably more closely related to the historic HbA1c of 5–10 years previously, rather than the current HbA1c .
Many of the referrals to the eye clinic are as a result of changes near the fovea, but without any macular oedema. Surrogate markers are currently required in retinal photography to identify patients with clinically significant macular oedema (CSMO). Patients without macular oedema do not require treatment and thus ‘referable’ disease should not be confused with ‘treatable’ disease, where the number of patients will be even less. Use of optical coherence tomography (OCT) within a screening programme may help improve the efficiency of screening for diabetic macular oedema, with fewer false negative referrals into the eye clinic, and could improve the sensitivity and specificity of referable disease .
Although it appears the risks of biennial screening may be small, the risks need to be clearly defined for the individual patient. Patients will have a natural anxiety about changes in retinal screening intervals. A dialogue with patient groups would be important if such a strategy was considered, exploring the risks, but also possible benefits. However, recent reports suggest patients would accept increased screening intervals if it was evidence based, especially if other aspects of service provision improved [15,16]. One way forward may be to introduce 2-yearly screening, possibly restricted to those with Type 2 diabetes and diabetes duration of less than 10 years. Selected ‘high-risk’ patients could be screened more frequently than annually, more attention could be given to recurrent defaulters from areas of social deprivation and, if optical coherence tomography is proven to be helpful in screening, then 2-yearly screening could be introduced at the same time as introducing optical coherence tomography into screening programmes. There is accumulating evidence to support a more sophisticated and varied approach to retinal screening intervals and a UK-based four nations research group has now been established to look at the evidence critically.