Retinopathy prevalence, incidence and trajectories in type 2 diabetes: The Fremantle diabetes study phase II

Abstract

Aims

To determine diabetic retinopathy (DR) prevalence, incidence, and whether distinct trajectories are associated with DR‐complicating Type 2 diabetes.

Methods

Retinal photographs from Fremantle Diabetes Study Phase II (FDS2) participants with Type 2 diabetes recruited in 2008–2011 and who attended biennial assessments for up to 6 years were graded as no DR, mild non‐proliferative DR (NPDR), moderate NPDR or severe NPDR/proliferative DR. Baseline DR prevalence, and the cumulative incidence of moderate NPDR or worse in those without DR at baseline, were calculated. Group‐based DR trajectory modelling was performed. Logistic regression determined independent associates of incident moderate NPDR or worse and trajectory group membership.

Results

Of 1521 participants (mean age 65.6 years, 52.1% males, median diabetes duration 9.0 years; 98% of all FDS2 participants with Type 2 diabetes) with gradable baseline photographs, 563 (37.0%) had DR. During a median 6.1 years of follow‐up, 23 (3.2%) without baseline DR developed at least moderate NPDR (crude incidence 6.1/1000 person‐years) with HbA_1c the sole independent predictor (odds ratio [95% CI]: 1.62 [1.30–2.02] per 1% [11 mmol/mol] increase). Trajectory analysis showed two distinct groups, those with baseline/persistent DR (20%) and those remaining DR free (80%). Longer diabetes duration, insulin use, higher mean HbA_1c, higher mean systolic blood pressure and higher mean urinary albumin: creatinine ratio all increased the odds (p ≤ 0.014) of being in the persistent DR trajectory group.

Conclusions

The low incidence of at least moderate NPDR reflects the trajectory analysis. The currently recommended biennial retinal screening frequency for individuals without DR could potentially be extended.

Novelty statement.

What's new

• Whether the incidence of diabetic retinopathy (DR) complicating type 2 diabetes is decreasing is uncertain.

• In the community‐based longitudinal Fremantle Diabetes Study Phase II, there was a low incidence of DR (7.5% over 6 years), with few participants (0.4%) progressing to more severe DR.

• Participants without DR at baseline were likely to remain DR free over 6 years in trajectory analysis, supporting biennial rather than annual screening for retinopathy‐free people with type 2 diabetes.

1. INTRODUCTION

The incidence of most chronic vascular complications of diabetes has been decreasing in recent years in developed countries, ¹ likely reflecting earlier diabetes diagnosis and improved clinical management. ² Diabetic retinopathy (DR) is a common microvascular complication and a leading cause of visual impairment and blindness globally. ³ Current evidence as to whether DR incidence is also decreasing is inconclusive. ¹ The Wisconsin Epidemiological Study of Diabetic Retinopathy (WESDR) in the 1980 s showed that 97% of people with young‐onset insulin‐treated diabetes had developed retinopathy 20 years after diagnosis and that there was a high incidence of DR in those with older‐onset diabetes. ⁴ These observations suggested a substantial lifetime DR risk but this may no longer apply given temporal trends in other complications. A recent review showed that the incidence of DR in people with Type 2 diabetes was lower in studies conducted after versus before 2000, ⁵ and a UK study showed a significant decrease in the incidence of DR between 2013 and 2016. ⁶ Furthermore, the global prevalence of DR appears to have decreased from 34.6% in a pooled analysis of studies conducted between 1980 and 2008 ⁷ to 22.2% in a more recent analysis in which studies up to 2020 were included. ⁸ Although suggestive, these trends may be influenced by temporal changes in DR ascertainment and the accuracy and coverage of administrative datasets.

Since DR is typically asymptomatic until late in the disease process, screening programmes are vital as they prevent vision loss and are cost‐effective. ⁹ , ¹⁰ In line with the likely decreasing prevalence of DR, recent publications have questioned whether the interval between DR screening visits could be lengthened to reduce costs without compromising outcomes, ¹¹ especially as most DR treatment is for proliferative disease or macular oedema. This suggestion is supported by some studies showing that the progression of DR is often slow, especially with shorter diabetes duration. ⁵ , ¹¹ Indeed, personalized DR screening has been suggested, and an algorithm developed and validated by Icelandic researchers with those at low risk requiring screening only every 5 years. ¹² , ¹³ There are, however, few contemporary community‐based studies assessing the incidence and progression of DR which would allow screening programmes to be developed on an appropriate evidence base. ¹⁴

The aim of the present study was, therefore, to ascertain DR prevalence and the incidence of moderate non‐proliferative DR (NPDR) or worse in well‐characterised community‐based people with Type 2 diabetes, and to determine whether there are sub‐groups with distinct DR trajectories whose characteristics could be used in the development of novel screening programmes.

2. METHODS

2.1. Patients and approvals

Data from participants in the longitudinal observational Fremantle Diabetes Study Phase II (FDS2) were used. ¹⁵ In brief, 1551 people with clinically diagnosed Type 2 diabetes were recruited from a postcode‐defined urban Australian population of approximately 150,000 between 2008 and 2011. All participants underwent comprehensive baseline and then biennial face‐to‐face assessments including an interview, questionnaires, a physical assessment and fasting biochemical tests. Each biennial assessment included single‐field, 45° retinal photographs of each eye taken using a Canon CR‐DGi Non‐Mydriatic Retinal Camera. If required, pupils were dilated using 0.5% tropicamide drops. The South Metropolitan Area Health Service Human Research Ethics Committee approved FDS2 and written informed consent was obtained from each participant. The study protocol conformed to the ethical guidelines of the Declaration of Helsinki.

2.2. Retinopathy assessment

Diabetic retinopathy was assessed according to the modified Airlie House Classification system for the Early Treatment Diabetic Retinopathy Study. ¹⁶ , ¹⁷ Baseline fundus photographs were assessed by a single external grader accredited by the Centre for Eye Research Australia, University of Melbourne, who was blinded to other participant data. Severity was classified as none, mild non‐proliferative diabetic retinopathy (NPDR), moderate NPDR, severe NPDR or worse. ¹⁸ Level 10 was classified as no DR, levels 11–31 as mild NPDR, levels 41 and 84 as moderate NPDR, and levels 51–80 and 86 as severe NPDR or worse. For participants who did not have fundus photography or whose photographs were ungradable, further information regarding DR status was sought from hospital records and/or attending optometrists/ophthalmologists. Participants without information on DR status at baseline were excluded from the analysis.

As the original grader was unavailable, follow‐up fundus photographs were first assessed independently by an experienced ophthalmologist under the same system used for the baseline images. A second ophthalmologist (AT) graded all photographs assessed as showing DR by the first, and he also graded a randomly‐selected sample of the remaining images. In discrepant cases, a consensus grading was obtained in consultation with a diabetologist with experience in DR (TMED). The worse eye from each assessment was used in the analysis.

2.3. Statistical analysis

Statistical analyses were conducted using the computer package IBM SPSS for Windows (Version 25.0. IBM Corp) and Stata (Version 15.1, Stat Corp.). Data are presented as proportions, mean ± SD, geometric mean (SD range) or median [interquartile range]. For independent samples, two‐way comparisons for categorical variables were by Fisher's exact test, for normally or log‐normally distributed continuous variables by independent sample t‐test, and for variables not conforming to normal or log‐normal distribution by Mann–Whitney U‐test. A two‐tailed significance level of p < 0.05 was used throughout.

Baseline DR prevalence and the cumulative incidence of at least moderate NPDR in those without DR at baseline were calculated. Group‐based trajectory modelling, which employs finite mixture modelling to approximate unknown distributions of trajectories across a study population, identified whether distinct trajectories are associated with DR status (no DR or any DR). Logistic models were used to estimate trajectories of DR over 6 years in those participants with DR status available for ≥2 study visits. This was implemented using the procedure ‘traj’ in Stata (Version 15.1, Stat Corp). ¹⁹ To assist model selection, the Bayesian Information Criterion (BIC) was utilised to determine the optimum number of groups and their functional form (linear, quadratic or cubic). ²⁰ Other selection criteria included: (i) adequate numbers of subjects in each group, (ii) distinct trajectories (non‐overlapping confidence intervals), (iii) acceptably narrow confidence intervals, (iv) average posterior probabilities of group membership >0.70, (v) odds of correct classification based on posterior probabilities of group membership >5 and (vi) close correspondence between group estimated probability and the proportion of participants classified to that group according to the maximum posterior probability assignment rule.

When comparing differences between trajectory groups, the updated mean (the average of values at each assessment) of pertinent clinical values including HbA_1c, supine systolic blood pressure and urinary albumin: creatinine ratio (uACR) were used. Backward conditional logistic regression models determined the independent associates of the different trajectory groups and incident moderate NPDR or worse in those with no DR at baseline and at least one follow‐up visit. Variables with p < 0.20 in the bivariable analyses were considered for model entry.

3. RESULTS

3.1. Participant disposition and baseline characteristics

Baseline DR status was available for 1521 (98%) of FDS2 participants with Type 2 diabetes. Their mean ± SD age was 65.6 ± 11.5 years, 52% were males, and their median (inter‐quartile range [IQR]) diabetes duration was 9.0 [2.9–15.8] years. Their median [IQR] HbA_1c was 6.8[6.2–7.7]% (51[44–61] mmol/mol) and mean supine systolic blood pressure was 146 ± 22 mmHg. A participant flow diagram is shown in Figure 1. The baseline DR prevalence was 37%, most (79%) of which was mild. Of 958 (63%) without DR at baseline, the 726 (76%) who had DR status ascertained as part of at least one follow‐up visit were included in the analysis of DR incidence.

FIGURE 1.

Consort diagram showing participants included in analysis.

3.2. Incident diabetic retinopathy

Of the eligible 726 participants without DR at baseline, 54 (7.5%) developed DR during a median of 6.0 years of follow‐up (crude incidence 14.5/1000 person‐years). Twenty‐three (3.2%) developed at least moderate NPDR during a median of 6.1 years of follow‐up (crude incidence 6.1/1000 person‐years). Of these, only 3 (0.4%) developed severe NPDR or PDR during follow‐up. Bivariable analysis comparing baseline characteristics of those who developed at least moderate NPDR during follow‐up to those who did not are shown in Table 1. Higher fasting serum glucose, HbA_1c, and uACR were associated with the development of at least moderate NPDR. In logistic regression modelling, HbA_1c at baseline was the sole independent predictor (odds ratio (95% CI): 1.62 (1.30–2.02) per 1% (11 mmol/mol) increase) of new moderate NPDR or worse during follow‐up.

TABLE 1.

Bivariable statistics comparing those who developed at least moderate non‐proliferative diabetic retinopathy (NPDR) during follow‐up to those who did not, in participants with type 2 diabetes who had no diabetic retinopathy at baseline and had at least one follow‐up visit

Variables at baseline	No moderate NPDR or worse during follow‐up	At least moderate NPDR during follow‐up	p‐value
Number (%)	703 (97)	23 (3.2)
Age at FDS entry (years)	64.5 ± 10.9	66.1 ± 11.1	0.507
Sex (% male)	350 (50)	10 (44)	0.673
Ethnic background (%)			0.985
Anglo‐Celt	409 (58)	15 (65)
Southern European	64 (9.1)	≤5 (22)*
Other European	47 (6.7)	≤5 (22)*
Asian	29 (4.1)	≤5 (22)*
Indigenous Australian	27 (3.8)	≤5 (22)*
Mixed/other	127 (18)	≤5 (22)*
Not fluent in English (%)	49 (7.0)	≤5 (22)*	0.673
Education beyond primary level (%)a	633 (91)	23 (100)	0.250
Currently married/de facto (%)	460 (65)	13 (57)	0.381
Private health insurance (%)	456 (65)	14 (61)	0.665
Duration of diabetes (years at study entry)b	5.0 [1.4–12.0]	8.0 [5.0–15.1]	0.102
On insulin (with or without other OGA as %)b	83 (12)	6 (26)	0.052
Fasting glucose (mmol/L)b	7.0 [6.1–8.3]	8.2 [6.7–9.9]	0.005
HbA1c (%)	6.6 [6.1–7.4]	7.7 [7.0–9.2]	<0.001
HbA1c (mmol/mol)	49 [43–57]	61 [53–77]	<0.001
Self‐monitors blood glucose (%)c	601 (86)	18 (78)	0.354
BMI (kg/m2)	31.2 ± 6.1	32.0 ± 5.5	0.520
Obesity (% by waist circumference)	488 (69)	19 (83)	0.248
Heart rate (bpm)	68 ± 12	69 ± 10	0.892
Supine systolic blood pressure (mmHg)	143 ± 20	147 ± 22	0.294
Supine diastolic blood pressure (mmHg)	80 ± 12	80 ± 16	0.912
Total cholesterol (mmol/L)b	4.4 ± 1.1	4.8 ± 1.3	0.066
HDL‐cholesterol (mmol/L)b	1.2 ± 0.3	1.3 ± 0.3	0.722
Serum triglycerides (mmol/L)b	1.5 (0.9–2.4)	1.7 (1.1–2.9)	0.134
Urinary albumin: creatinine (mg/mmol)d	2.5 (0.8–7.9)	5.2 (1.3–21.0)	0.004
eGFR (CKD‐EPI) categories (%)e			0.976
≥90 ml/min/1.73 m2	296 (42)	9 (39)
60–89 ml/min/1.73 m2	330 (47)	12 (52)
45–59 ml/min/1.73 m2	42 (6.0)	≤5 (22)*
<45 ml/min/1.73 m2	31 (4.4)	≤5 (22)*
Any ischaemic heart disease (%)	169 (24)	9 (39)	0.135
Cerebrovascular disease (%)	37 (5.3)	≤5 (22)*	0.128
Peripheral arterial disease (%)	130 (19)	≤5 (22)*	0.999
Peripheral sensory neuropathy (%)	386 (55)	13 (57)	0.999
Any depression (%)f	66 (9.8)	≤5 (22)*	0.472
Alcohol (standard drinks/day)g	0.1 [0–1.2]	0.1 [0–0.3]	0.232
Current smoker (%)b	53 (7.5)	≤5 (22)*	0.691

Note: Data are presented as number (percentage), mean ± standard deviation, geometric mean (standard deviation range) or median [inter‐quartile range]. Missing data: ^a n = 8 (1.1%); ^b n = 1 (0.1%); ^c n = 5 (0.7%); ^d n = 2 (0.3%); ^e n = 4 (0.6%); ^f n = 31 (4.3%); ^g n = 28 (3.9%).

^*Exact numbers withheld to ensure participant privacy and confidentiality.

3.3. Trajectory analysis

Trajectory analysis included 1115 participants (72% of the Type 2 diabetes cohort) who had DR assessed at ≥2 study visits. At entry, these participants had a mean ± SD age of 64.6 ± 10.9 years, 54% were males and their median [IQR] diabetes duration was 8.0 [2.0–15.0] years. There were 313 with mild NPDR at baseline and at least one follow‐up visit, of whom 60 (14%) progressed to at least moderate NPDR. The best model (with the lowest BIC) identified two groups, those with baseline and persistent DR (Persistent DR group), and those without DR at baseline who remained DR free (No DR group; Table 1). The No DR group was the largest (n = 893 [80%]). Both groups had a generally flat cubic trajectory. Figure 2 illustrates the two trajectories and their narrow 95% confidence intervals. Although the BIC was similar for a three‐group model, there were inadequate numbers (<10%) in two of the three groups.

FIGURE 2.

The results from the trajectory analysis show two distinct retinopathy groups, the Persistent DR group (above) and no DR group (below) with 95% confidence intervals (dashed grey lines) for both.

There was close agreement between observed and predicted DR status (Table S2). The average posterior probabilities for the trajectory groups were 0.97 (No DR) and 0.96 (Persistent DR), both exceeding the recommended 0.7 cut‐off. The odds of correct classification based on the posterior probabilities of group membership were >5 for both groups (Table S3). Each group's estimated probability corresponded closely to the proportion of participants allocated to that group according to the maximum posterior probability assignment rule (Table S4). Group‐based trajectory modelling describes the progression of a phenomenon (DR) over time and identifies variations in its progression between groups of individuals. ¹⁹ However, this is a generalisation; some participants in the No DR group developed DR and some in the Persistent DR group improved (Figure 2 and Supplemental Table S2).

3.4. Trajectory group characteristics

The bivariable analysis of the characteristics of the two trajectory groups is summarised in Table 2. Persistent DR group members were less likely to be of Anglo‐Celt ethnicity and to speak English fluently. They were more likely to be of Asian, Indigenous Australian or Southern European ethnicity, to have been insulin‐treated at baseline, and they had a higher baseline and updated mean HbA_1c, uACR and systolic blood pressure, and a higher baseline heart rate. They were more likely to be obese (by waist circumference) and have cerebrovascular and peripheral arterial disease at baseline. Persistent DR group members were also more likely to have a lower baseline eGFR. In multiple logistic regression analysis, longer diabetes duration, insulin use, higher updated mean HbA_1c, higher updated mean systolic blood pressure and higher updated mean uACR independently increased the odds of being in the Persistent DR trajectory group (p ≤ 0.014; Table 3). Those in the no DR group were more likely to have had DR assessed at all four study visits when compared to the Persistent DR group (57% and 43%, respectively, p < 0.001).

TABLE 2.

Bivariable analysis showing differences between the two diabetic retinopathy trajectory groups

Variables at baseline	Group 1: no diabetic retinopathy group	Group 2: persistent diabetic retinopathy group	p‐value
Number (%)	893 (80)	222 (20)
Age at FDS entry (years)	64.6 ± 10.8	64.5 ± 11.5	0.911
Sex (% male)	474 (53)	128 (58)	0.229
Ethnic background (%)			0.001
Anglo‐Celt	525 (59)	106 (48)
Southern European	86 (9.6)	38 (17)
Other European	64 (7.2)	19 (8.6)
Asian	35 (3.9)	13 (5.9)
Indigenous Australian	35 (3.9)	17 (7.7)
Mixed/other	148 (17)	29 (13)
Not fluent in English (%)	61 (6.8)	31 (14)	0.001
Education beyond primary level (%)a	808 (92)	192 (88)	0.067
Currently married/de facto (%)	589 (66)	145 (65)	0.874
Private health insurance (%)	575 (64)	139 (63)	0.640
Duration of diabetes (years at study entry)b	5.0 [1.7–12.1]	15.8 [11.0–20.3]	<0.001
On insulin (with or without other OGA as %)b	116 (13)	110 (50)	<0.001
Fasting glucose (mmol/L)c	7.0 [6.1–8.4]	8.1 [6.6–10.3]	<0.001
HbA1c (%)	6.7 [6.1–7.4]	7.4 [6.8–8.8]	<0.001
HbA1c (mmol/mol)	50 [43–57]	57 [51–73]	<0.001
Self‐monitors blood glucose (%)d	755 (85)	191 (87)	0.668
BMI (kg/m2)b	31.3 ± 6.1	31.9 ± 5.5	0.182
Obesity (% by waist circumference)b	768 (86)	204 (92)	0.013
Heart rate (bpm) c	68 ± 12	72 ± 13	<0.001
Supine systolic blood pressure (mmHg)b	144 ± 20	151 ± 23	<0.001
Supine diastolic blood pressure (mmHg)b	80 ± 12	82 ± 12	0.089
Total cholesterol (mmol/L)e	4.4 ± 1.1	4.3 ± 1.1	0.398
HDL‐cholesterol (mmol/L)e	1.2 ± 0.3	1.2 ± 0.3	0.507
Serum triglycerides (mmol/L)e	1.5 (0.9–2.5)	1.6 (0.9–2.6)	0.274
Urinary albumin: creatinine (mg/mmol)e	2.6 (0.8–8.4)	5.3 (1.1–25.2)	<0.001
eGFR (CKD‐EPI) categories (%)f			0.008
≥90 mL/min/1.73 m2	370 (42)	85 (39)
60–89 mL/min/1.73 m2	418 (47)	93 (42)
45–59 mL/min/1.73 m2	63 (7.1)	20 (9.1)
<45 mL/min/1.73 m2	38 (4.3)	22 (10)
Any ischaemic heart disease (%)	224 (25)	64 (29)	0.266
Any cerebrovascular disease (%)	48 (5.4)	24 (11)	0.006
Peripheral arterial disease (%)b	158 (18)	56 (25)	0.013
Peripheral sensory neuropathy (%)b	494 (55)	138 (62)	0.058
Any depression (%)g	83 (9.7)	30 (14)	0.060
Alcohol (standard drinks/day)h	0.1 [0–1.2]	0.1 [0–1.3]	0.126
Smoking status (%)b			0.672
Never	415 (47)	98 (44)
Ex‐smoker	402 (45)	102 (46)
Current smoker	75 (8.4)	22 (9.9)
Other data
Updated mean HbA1c (%)	6.9 [6.3–7.6]	7.7 [6.9–8.7]	<0.001
Updated mean urinary albumin: creatinine (mg/mmol)b	3.1 (0.9–10.7)	7.6 (1.6–36.6)	<0.001
Updated mean supine systolic blood pressure (mmHg)	141 ± 16	148 ± 17	<0.001

Note: Data are presented as number (percentage), mean ± standard deviation, geometric mean (standard deviation range) or median [inter‐quartile range]. Missing data: ^a n = 15 (1.3%); ^b n = 1 (0.1%); ^c n = 2 (0.2%); ^d n = 10 (0.9%); ^e n = 3 (0.3%); ^f n = 6 (0.5%); ^g n = 52 (4.7%); ^h n = 44 (3.9%).

TABLE 3.

Multiple logistic regression analysis showing the variables independently associated with being in the persistent DR trajectory group

Variable	Odds ratio (95% confidence interval)	p‐value
Diabetes duration at baseline (per 1‐year increase)	1.12 (1.09–1.15)	<0.001
On insulin at baseline	2.52 (1.72–3.70)	<0.001
Updated mean HbA1c (per 1% or 11 mmol/mol increase)	1.50 (1.30–1.72)	<0.001
Ln (updated mean urinary albumin: creatinine ratio) a	1.19 (1.04–1.36)	0.009
Updated mean supine systolic blood pressure (per 10 mmHg increase)	1.14 (1.03–1.28)	0.014

^a A 2.72‐fold increase in the updated mean urinary albumin: creatinine ratio corresponds to an increase of 1 in ln(updated mean urinary albumin: creatinine ratio).

4. DISCUSSION

In our community‐based Australians with Type 2 diabetes, just over one‐third had DR at baseline and the incidence of moderate NPDR or worse in those without DR at baseline was low. Trajectory analysis showed little change in DR status during 6 years of follow‐up. This suggests that most people with Type 2 diabetes without DR are very likely to remain DR free. Hyperglycaemia, assessed from fasting serum glucose and/or HbA_1c, was independently associated with both the incidence of moderate NPDR and Persistent DR trajectory group membership. Additionally, longer diabetes duration, hypertension and higher uACR were independently associated with being in the Persistent DR group. These findings have implications for clinical management.

The prevalence of DR varies greatly between studies, from 5% to 40% in recent reports. ¹¹ , ¹⁴ , ²¹ , ²² , ²³ , ²⁴ The global prevalence of DR has been estimated at 22% ⁸ which is lower than the 37% prevalence in our study. The United Kingdom Prospective Diabetes Study (UKPDS) also had a 37% prevalence of any DR at baseline in newly diagnosed Type 2 diabetes ²⁵ and the baseline prevalence in the WESDR was similar at 39% in the older onset group not on insulin even if there was a higher 70% prevalence in the older onset insulin‐treated group. ⁴ A more recent Saudi Arabian study using data from a chronic illness clinic found a 38% prevalence for any DR in people with Type 2 diabetes, ²³ similar to our results. In a 2018 Thai study, ²⁴ a 5% prevalence of DR in people with Type 2 diabetes was lower than in many other studies, likely reflecting the lack of data from university hospitals including tertiary medical centres. ²⁴

Differences in study sampling methods and ascertainment of DR status could contribute to the large variations in reported DR prevalence. Lowering of the fasting serum glucose required for diabetes diagnosis in 1999 and greater diabetes awareness has led to earlier diabetes diagnosis and therefore earlier DR screening, which would very likely affect DR prevalence and incidence. Furthermore, emphasis on the importance of intensive cardiovascular, including glycaemic, management has resulted in better blood pressure and improved glucose levels which would lower DR prevalence. ⁶ Notwithstanding these considerations, our DR prevalence is at the upper end of the range of published rates and shows that more than one in three Australians with Type 2 diabetes has DR.

The incidence of DR also varies greatly between studies. A UK retrospective cohort analysis found that 46% of those with newly diagnosed Type 2 diabetes who were retinopathy‐free developed DR after 7 years. ²² This study used the Royal College of General Practitioners (RCGP) Research and Surveillance Centre (RSC) database and ascertained DR status using diagnosis codes, ²² although the validity of the database is unclear, particularly regarding those without DR. Another UK population‐based study reported an 8.5% annual incidence of new DR in 2016 in those with Type 2 diabetes using electronic medical records. ⁶ In a review of population‐based studies of Type 1 and/or Type 2 diabetes conducted after 2000, the annual incidence of DR ranged from 2.2% to 12.7%. ⁵ Our study found that 7.5% of participants developed DR during a median of 6 years suggesting a rate below this range. In the WESDR there was a much higher incidence in the older onset cohort, with 47% of those on insulin and 34% of those not on insulin developing any DR over a four‐year period, ²⁶ whilst 22% of UKPDS participants developed DR over a six‐year period. ²⁵ Despite the FDS2 prevalence being similar to these older studies, our incidence rates were much lower which likely reflects more intensive contemporary risk factor management.

Our novel trajectory analysis showed two distinct groups. The largest comprised those without DR who were likely to remain DR free over 6 years. This observation lends support to current Australian guidelines recommending biennial DR screening for those without risk factors ²⁷ but suggests that even longer intervals could be considered for low‐risk people with Type 2 diabetes. The Icelandic personalised DR risk equation has shown that screening intervals can be up to 60 months without compromising outcomes, ¹² , ¹³ with biennial screening reduced by 23% and annual screening by 61% in the Netherlands validation cohort. ¹³ Danish DR screening guidelines, updated in 2018, now recommend up to 48‐month intervals for those at lower risk and up to 24 months for those with moderate NPDR, depending on risk factors. ²⁸ Our trajectory analysis data and these European studies support lengthening Australian DR screening intervals to 48–60 months in those without DR and at low risk, specifically individuals with short‐duration, non‐insulin‐treated Type 2 diabetes without nephropathy and with adequate glycaemic and blood pressure control.

As just 14% of our participants progressed from mild to moderate NPDR or worse during follow‐up, screening intervals in those with DR could potentially be extended beyond annual reviews. Those with DR were less likely to attend all follow‐up visits, but participants may have attended DR screening elsewhere as part of usual care. A recent publication found those at increased risk of DR progression undergoing 18‐month screening had greater adherence to screening than those with a 36‐month interval, but adherence rates were > 90% in both groups. ²⁹ Nevertheless, a history of non‐adherence should not influence screening intervals. ²⁸

More frequent DR screening can be tailored to individual risk factors. In the UK, a randomised controlled trial determined that individualised risk‐based screening at 6‐, 12‐ or 24‐month intervals were non‐inferior to the recommended annual screening. ³⁰ In an Australian context, Aboriginal or Torres Strait Islanders with diabetes who have longer duration disease, suboptimal glycaemia, hypertension, renal disease or dyslipidaemia are recommended to have annual screening. ²⁷ Those with DR are recommended for even more frequent reviews. ²⁷ Our study findings support these guidelines since the variables associated with the DR group in our trajectory analysis were longer diabetes duration, higher HbA_1c, insulin use (which may reflect sustained hyperglycaemia), hypertension and uACR, all well‐known DR risk factors. ⁴ , ⁷ , ¹⁴ , ²² , ²⁵

The strengths of this study include using novel trajectory analysis to examine temporal changes in DR, the representative community‐based sample and comprehensive face‐to‐face assessments. Limitations include the use of single‐field retinal images which may have underestimated the DR presence and/or severity, potentially biasing the results towards the null. There may be inter‐rater bias as different graders performed the baseline and follow‐up grading, despite using the same classification system. Additionally, we did not include macular oedema grading in our analyses.

In conclusion, our novel trajectory analysis showed that FDS2 participants with Type 2 diabetes without DR were likely to remain DR‐free during 6 years of follow‐up. This was supported by a low incidence of at least moderate NPDR. Our results are in accord with current Australian DR screening guidelines but suggest that the screening interval in low‐risk people with Type 2 diabetes could be extended beyond the currently recommended two‐year period.

FUNDING INFORMATION

The FDS2 was funded by the National Health and Medical Research Council of Australia (project grants 513781 and 1042231). TMED is supported by a Medical Research Future Fund Practitioner Fellowship. These funding bodies had no involvement in the study design, data collection, analysis and interpretation of results or writing of this manuscript.

CONFLICT OF INTEREST

None to declare.

Supporting information

Table S1–S4

Drinkwater JJ, Davis TME, Turner AW, Davis WA. Retinopathy prevalence, incidence and trajectories in type 2 diabetes: The Fremantle diabetes study phase II . Diabet Med. 2023;40:e15032. doi: 10.1111/dme.15032

DATA AVAILABILITY STATEMENT

The datasets generated and analysed during the current study are not publicly available. However, anonymised data are available from the corresponding author upon reasonable request, excluding prevalent coronary heart disease and cerebrovascular disease. The latter were identified through health data linkage with permission from the Western Australian Department of Health and cannot be shared without approval.