Discussion

This real-world dataset collated from a large number of centres across the UK demonstrates the increased rate at which patients with diabetes develop DMO in the period after cataract surgery. The percentage of eyes developing first episode of ‘treatment-requiring DMO’ increased from 2.9% in the year immediately prior to surgery to 5.3% in the first year and 4.8% in the second year after surgery; the highest risk period was between 3 and 6 months after surgery with 1.7% of eyes developing first episode of ‘treatment-requiring DMO’ during that 3-month period, equating to an annualised risk of 6.8%. The percentage of ‘ranibizumab-requiring DMO’ which in the UK is based on an OCT definition of DMO with a minimum CMT of 400 microns, increased from 0.1% for the 2-year period prior to surgery to 1.5% of eyes for the 2-year period post-surgery.

Our findings provide additional data to suggest that there is a real increase in treatment-requiring DMO after cataract surgery. This has been a matter of considerable debate, especially since the ETDRS Report 25 failed to show a significant increase with 29% patients being noted to have CSME at their study visit occurring prior to cataract surgery and 31% having CSME at their study visit occurring post-surgery.9 It should be noted however that this study may now be less relevant to current practice due to the modernisation of cataract surgery (including the widespread adoption of phacoemulsification since that time) and advances in the care of diabetic eye disease. Additionally the population analysed in ETDRS Report 25 differed from most studies (including ours) in that it had a very high proportion of patients with more advanced retinopathy: PDR and/or vitreous haemorrhage was present in 40% of those who had a gradeable fundus photograph from the visit immediately preceding their cataract operation. A more recent prospective study looking at unilateral cataract surgery in 132 eyes (with the fellow eye acting as control) noted a trend towards higher rate of CSME in the operated eye versus the fellow eye (6.1% vs 4.5%) within the first 6 months.12 It should be noted that the difference between rates of CSME (both pre-operative and post-operative) may also reflect the inclusion of patients with milder levels of retinopathy in this later study. The findings of our study indicate that this previously observed trend towards higher levels of treatment for DMO in the post-operative phase, is indeed significant and provides, for the first time, evidence that there is a ‘high-risk’ period in the post-operative phase which peaks between 3 and 6 months. This dataset is taken using a single data capture platform across a large population in the UK, reflecting a wide ethnic-mix from urban and rural environments. The EMR mandates a structured assessment of DR resulting in an ETDRS grading of retinopathy and maculopathy based on the presence or absence of clinical signs at each clinic visit.

It is important to note that our study is not directly comparable to earlier studies in that we are not assessing CSME, but rather ‘treatment-requiring DMO’. ‘Treatment-requiring DMO’ is a clinically relevant group which, while subjective, provides an important metric of the burden of disease on the patient and the clinical service. The ‘ranibizumab-requiring DMO’ subgroup provides an objectively defined disease threshold for DMO due to the regulatory requirements in the UK requiring an OCT-defined threshold of DMO with a CMT of >400 microns before treatment with ranibizumab is funded. This avoids the subjectivity of clinically-defined CSME, which may show significant inter-observer variability. Both the ‘treatment-requiring DMO’ group and the ‘ranibizumab-requiring DMO’ subgroup show the same significant increase in treatment for DMO the post-operative phase, and both peak between 3 and 6 months post-operatively. It is recognised that in this analysis  we have considered neither the actual level of CMT nor which instrument it has been measured on (which may influence the threshold)16; these OCT data are not currently available for our dataset but may be available in future data extractions.

The relatively modest use of ranibizumab as first line treatment in this cohort probably reflects the fact that most UK centres only adopted its use routinely after the NICE guidance in 2013. The increased availability of this drug at later time points of the study must be borne in mind when interpreting any increase in its usage over time, such as the fact that the proportion of cases treated with ranibizumab increased from 2.5% for the 2 years prior to surgery to 20.8% for the 2 years after surgery. In contrast any decline in usage (such as after the observed peak in the 4 to 6 month period post-surgery) can be assumed to be real (since availability would have increased over this period). The treatment data therefore reliably inform us that the risk of developing DMO exceeding 400 microns peaks in the 4 to 6 month post-operative period, which is also in line with the incidence of cases of DMO overall. Ideally a future study would take a new cohort with study inception after 2013. This would show a much higher proportion of ranibizumab (and more recently aflibercept) usage both pre-surgery and post-surgery, and would provide a more complete estimate of the proportion of ‘treatment-requiring DMO’ which is associated with a macular thickness of greater than 400 microns.

A possible alternative explanation when assessing studies reporting progression of diabetic retinopathy and/or maculopathy after cataract surgery is whether this simply reflects improved visualisation of the fundus. An advantage of our study is that the ranibizumab endpoint reflects an OCT-defined measure in a subset of our cohort, which is usually possible pre-surgery and post-surgery. We would however recognise two limitations. First we do not know how rigorously clinicians obeyed the >400 micron limit before offering ranibizumab to their patients as we have not extracted the OCT data. Second, OCT was not compulsory at all visits so there may be cases of DMO with CMT >400 microns which were missed and not treated. It may be argued that this was more likely to happen pre-operatively due to worse visualisation. We do not think that this was a major issue however, as most patients with worsening vision and diabetic retinopathy being considered for cataract surgery are likely to have had an OCT, although we cannot confirm the exact rates of pre-operative OCT scanning in the current data extract.

It may be difficult to distinguish pseudophakic (postoperative, cystoid macular oedema), macular oedema from DMO post-operatively; the EMR mandated structured assessment only requires recording of the location of retinal thickening. Standard clinical practice in the UK would have been to use topical steroids and/or non-steroidal drops as first line treatment if the clinician believed the problem to be pseudophakic macular oedema followed by intravitreal or sub-Tenon’s triamcinolone acetonide as second line treatment. Analysis of our study suggests that this would only account for a small proportion of the cases seen. Intravitreal and sub-Tenon’s triamcinolone together accounted for around one-third of the ‘Other Treatment’ category (specifically 12% intravitreal, 20% sub-Tenon’s), and did not increase significantly after surgery with ‘Other Treatments’ comprising 13% vs 10.7% of total treatments (pre-surgery versus post-surgery respectively).

The lack of compulsory pre-operative and post-operative OCT and fundus fluorescein angiography (FFA) in all cases is one example of the differences between ‘real-world’ studies and prospective RCTs. An RCT can provide stringently quality-assured data, taken on a group of patients with the condition of interest, accompanied by minimal co-pathology and with carefully scheduled visits allowing treatment to be delivered and monitored under closely controlled conditions. Such studies have many advantages, and are valuable particularly for licensing studies as they may provide a high signal:noise ratio and, if well-designed, have a low risk of bias. They are however very expensive and generally deliberately exclude large sections of the population for whom the treatment is clinically indicated.

‘Real-world’ studies are of immense value in a number of situations. First, they provide some indication of how a treatment effect seen in the ideal but artificial environment of an RCT actually translate into benefits (and adverse events) in a much more heterogeneous group of patients in the less controlled environment of clinical practice. Second, they provide a mechanism of identifying and investigating rarer conditions, for which a suitably scaled RCT would be unaffordable.

Data from ‘real world’ studies can still be quality-assured, although not to the same level as an RCT. In our study, the design of the EMR system ensures that the data entry is structured, key fields are compulsory, and we have demonstrated previously low rates of missing non-compulsory fields.17 Fields have value cut-offs to stop major data entry errors; most minor errors may occur in either direction and so with a large sample size it is anticipated that these would become negligible. In summary, we would argue that any additional ‘noise’ in our study brought about by its real world, retrospective design, will be compensated for by the large sample size and multiple physicians and hospital.

One of the key questions arising from this study is whether this new evidence regarding the risk of post-operative DMO should alter practice, particularly regarding frequency (and location) of follow-up in the post-operative phase supported by the use of OCT to screen for DMO. The follow-up of patients with diabetic retinopathy after cataract surgery is very variable, being primarily dependent on the level of diabetic retinopathy. In the UK and many other countries, patients with mild NPDR may only receive the same level of hospital follow-up as patients without diabetes. Depending on local practice they receive one (or no) hospital reviews before being discharged to their community optometrist for post-operative refraction. Unless either the patient or the optometrist report specific concerns, patients may not receiveany further assessment of their fundus until their annual photographic review within the National Diabetic Eye Screening Service. It should be noted that this assessment is based on the grading of two-dimensional fundus photographs and the recorded level of visual acuity, and does not include stereo-photographs or OCT assessment which would assist in the detection of DMO. Our study suggests however that the highest risk for the development of DMO occurs in the period after the typical timing of the first post-operative visit and occurs at a significant rate even in those patients that have been assessed as having mild NPDR. It may however be possible to further stratify this risk. A DRCR.net study found that at sixteen weeks postoperatively, OCT-defined centre-involving ME developed in 0% of eyes with no preoperative DMO versus 10% if eyes had had pre-operative non–centre-involving DMO and 12% if eyes had been assessed as ‘possible’ centre-involving DMO.18  In this context it is also interesting to note that those patients with PDR appeared to be relatively protected from developing DMO in the post-operative phase when compared with NPDR. This is interesting but should be treated with caution as the numbers in the PDR group were significantly lower than in the other groups.

It may be argued that most patients will be aware of a decline in visual function, and so should be able to self-report visually significant DMO; indeed the NICE (UK) threshold is, in part, based on this equating to a visually significant difference. It is well known however that many patients may ignore the changes in one eye, particularly where reasonable vision is retained in the other eye.19 There may be therefore  an argument for providing a further OCT-based assessment, whether in the community or in the hospital eye clinic, at around 4 months post-operatively to screen for the development of DMO, and to provide the opportunity for early treatment.

In summary, this analysis of a real world dataset achieved at a large number of centres across the UK indicates an increased rate of developing visually significant DMO in the post-operative phase, with a particularly high risk from 3 to 9 months. The UK requirement for an OCT-defined threshold of DMO to access ranibizumab treatment, and the presence of a clear post-operative DMO spike in this cohort provides further evidence that this is a real increase in DMO and not just an issue of visualisation. We would recommend that assessment prior to cataract surgery for any patient with diabetes should include a record of DR severity status and a macular OCT of both eyes, while recognising that the lens opacities may limit assessment in some eyes. There may also be a role for increased monitoring of patients with even relatively mild diabetic retinopathy during the post-operative phase to help detect DMO and enable earlier treatment.