Introduction

The number of people with diabetes mellitus (DM) is increasing worldwide, resulting in an increase in the prevalence of diabetic microvascular complications, which mainly include diabetic retinopathy (DR) and diabetic nephropathy (DN) [1]. DR is a leading cause of vision impairment and blindness in working-age people [2, 3], and it is estimated that without prompt treatment, the number of people with DR will increase rapidly worldwide from 126.6 million in 2010 to 191 million by 2030 [4]. Progressive microvascular damage is also commonly observed in glomeruli which may lead to microalbuminuria, a decrease in the glomerular filtration rate (GFR) and eventual development of chronic kidney disease (CKD) [5].

There is an increasing interest in the literature about the relationship between DR and DN due to the shared risk factors and similar pathophysiological features [5]. Numerous studies have consistently demonstrated a strong association between GFR and DR in type 1 diabetes mellitus (T1DM) patients [6, 7], however, this association has not been well established in type 2 diabetes mellitus (T2DM) patients, with some studies indicating a negative association between the GFR and DR [8, 9], while others showing no association [10, 11].

Microalbuminuria is a marker of pathophysiological vascular dysfunction in patients with kidney and cardiac disorders [12]. However, the role of microalbuminuria as a biomarker for microvascular damage in patients with DR is debatable because some studies have indicated that microalbuminuria may regress to normal over time [13]. Additionally, studies have found a dissociated relationship between GFRs and microalbuminuria levels in DM patients [14] and about 50% of T2DM patients with normal microalbuminuria levels exhibited decreased GFR levels [15], possibly due to nondiabetic causes [16], aging [17], kidney damage, or aggressive antihypertensive treatment [15]. Studies investigating the joint relationship of GFR and microalbuminuria with DR and which of these two markers of DN is more closely related to DR development in T2DM are not only limited, but also have yielded mixed results [10, 11, 18].

Therefore, we conducted a cross-sectional study to investigate the association of GFR with the presence of DR and diabetic macular oedema (DMO) in Chinese patients with T2DM. The effect of the presence of microalbuminuria on the association between GFR and DR was also assessed.

Methods

Study population

This cross-sectional study was conducted from October 2017 to April 2019 at the Zhongshan Ophthalmic Centre (ZOC), Guangzhou, and Southern China. Individuals with T2DM aged between 30 and 80 years and registered in the community health system near ZOC were consecutively recruited to participate in the study. Subjects were not recruited if any of the following conditions were present: (1) a previous history of cerebrovascular disease, cardiovascular disease, malignancy or other severe systemic disease that was not conducive to comprehensive examination; (2) pregnancy; (3) cognitive disorders or mental diseases; (4) ungradable bilateral fundus photographs; or (5) refusal of or intolerability to mydriasis due to a shallow anterior chamber (a van Herick value below 25°), an intraocular pressure ≥21 mmHg or angle closure glaucoma. The study adhered to the tenets of the Declaration of Helsinki, and approval was obtained from the Institutional Ethical Committee of ZOC (2017KYPJ094). Written informed consent was obtained from all participants.

Study procedure

Assessment of diabetic retinopathy and diabetic macular oedema

Standardised seven-field fundus photographs based on Early Treatment Diabetic Retinopathy Study (ETDRS) criteria were obtained using a digital camera (Canon CR-2, Tokyo, Japan) 30 minutes after mydriasis (0.5% topicamide plus 0.5% phenylephrine eye drops), and a swept-source optical coherence tomography (SS-OCT) device (DRI-OCT2 Triton, Topcon, Tokyo, Japan) was used to scan the fovea area using a 6 × 6 mm raster scan protocol.

Fundus photographs were independently graded by two trained graders according to the guidelines of the United Kingdom National Diabetic Eye Screening Programme (UK NDESP) [19]. Thus, R0 was defined as the condition with no diabetic retinal lesions. R1 was defined as the presence of at least one of the following retinal lesions: microaneurysm, dot haemorrhages, exudates, cotton wool spots, and venous loops. R2 was defined as the presence of any retinal lesion including venous beading, venous reduplication, multiple blot haemorrhage, or intraretinal microvascular abnormalities. R3 was defined as the presence of new vessels on the disc retina or elsewhere; preretinal fibrosis, preretinal or vitreous haemorrhage, or evidence of previous peripheral retinal laser treatment [19].

DMO was defined as the presence of exudates within a one disc diameter of the fovea centre, a collection of exudates within the macula, focal photocoagulation scars in the macular area, a central retinal thickness ≥300 μm or the presence of DMO features using SS-OCT [19, 20]. Participants presenting with any of R2, R3 or DMO were considered to have vision-threatening diabetic retinopathy (VTDR).

Assessment of blood and urine chemistry

A venous blood sample was collected to estimate serum uric acid (UA), the serum creatinine clearance rate (CCr), C-reactive protein (CPR), total cholesterol (TC), triglycerides (TGs), low-density lipoprotein cholesterol (LDL-c), and high-density lipoprotein cholesterol (HDL-c) using a chemical analyser (Cobs 8000; Roche Diagnostics; Mannheim, Germany). Haemoglobin A1C (HbA1c) was tested by a Sysmex G8 glycosylated haemoglobin analyser (Sysmex Corporation; Kobe, Japan) at the clinical laboratory of ZOC. A 50-ml midstream urine sample was collected to assess microalbuminuria with a Cobs 8000 biochemistry analyser (Roche Diagnostics; Mannheim, Germany). The estimated GFR was calculated according to the widely used Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [21]. Renal function was defined as normal (GFR > 90 mL/min/1.73 m2), mildly impaired (GFR 60-89 mL/min/1.73 m2), or indicative of CKD (GFR < 60 mL/min/1.73 m2) [22].

Assessment of other examinations

All participants underwent other ocular examinations, including visual acuity, IOP, auto-refraction, ocular biometry test and slit lamp examinations of anterior and posterior segments. Systolic (SBP) and diastolic (DBP) blood pressures, height and weight were measured by a trained nurse using a standard operating procedure. Mean arterial pressure (MAP) was defined as one-third of SBP plus two-thirds of DBP. The body mass index (BMI) was calculated as the weight (kg) divided by the square of the height (m). A standard questionnaire was administered by a trained interviewer to collect information about demography, lifestyle risk factors, systemic and eye medical histories.

Statistical analysis

DR was graded based on the worse eye in those with bilaterally qualified fundus images. All statistical analyses were performed using Stata (ver. 12.0; Stata Corp; College Station, TX). A Student’s t test or one-way ANOVA was used to compare continuous variables, while the chi-squared test was used to compare categorical data. Multiple logistic regression models were used to test the association between the presence of DR, VTDR, or DMO and renal function (microalbuminuria and GFR) after adjusting for confounders. Odds ratios (OR) and 95% confidence intervals (CI) are presented. A P value of <0.05 was considered statistically significant.

Results

A total of 1877 participants with T2DM were enrolled in the current study. Of the 1877 eligible participants, 380 (20.3%) exhibited any DR, and 78 (4.16 %) exhibited DMO. Participants with any DR were less likely to be male (P = 0.005), and more likely to have a longer DM duration (P < 0.001), use insulin (P < 0.001), have a history of hyperlipidaemia (P < 0.001), and have higher levels of HbA1c (P < 0.001), SBP (P < 0.001), GFR (P < 0.001), and microalbuminuria (P < 0.001) than those without any DR. Participants with DMO tended to be younger (P < 0.001), have a longer DM duration (P = 0.007), be more likely to use insulin (P < 0.001), be less likely to have a history of hyperlipidaemia (P < 0.001) and have higher level of HbA1c (P < 0.001), SBP (P < 0.001), LDL-C (P = 0.040) and lower levels of GFR (P = 0.009) than those without DMO (Table 1).

Table 1 Comparison of basic characteristic of participants with and without diabetic retinopathy.
Full size table

Participants with impaired renal function were more likely to be older (P < 0.001), male (P < 0.001), have a history of hypertension (P < 0.001), have a longer DM duration (P < 0.001), have higher levels of BMI (P = 0.016), UA (P < 0.001), CPR (P = 0.009) and microalbuminuria (P < 0.001), and higher risks of any DR (P = 0.012), VTDR (P = 0.006) and DMO (P = 0.022) than those with normal GFR (Table 2).

Table 2 Distribution of basic characteristic stratified by GFR among participants with diabetes.
Full size table

Participants in the higher quartile of microalbuminuria level tended to be more likely to be older (P < 0.001), male (P < 0.001), use insulin (P < 0.001), have histories of hypertension (P < 0.001) have a longer DM duration (P < 0.001), have higher levels of HbA1c (P < 0.001), BMI (P < 0.001), MAP (P = 0.015), CRP (P = 0.002), UA (P < 0.001) and TGs (P < 0.001 and lower levels of GFR (P < 0.001) and HDL-c (P < 0.001), and have higher risks of any DR (P < 0.001), VTDR (P < 0.001) and DMO (P < 0.001) than those in the first quartile (Table 3).

Table 3 Distribution of basic characteristic stratified by microalbuminuria level among participants with diabetes.
Full size table

In the multiple logistic regression model, a low GFR was significantly associated with the presence of DR (OR = 0.98 per 1 mL/min/1.73 m2 increase, P < 0.001) and VTDR (OR = 0.98 per 1 mL/min/1.73 m2 increase, P < 0.001).We also assessed the associations of DR with different GFR levels, and the results showed that CKD was an important risk factor for any DR (OR = 1.89, P = 0.017) and VTDR (OR = 2.76, P = 0.009) compared to the case with a normal GFR, and that mildly impaired renal function (OR = 1.39, P = 0.031) was significantly associated with any DR. A high microalbuminuria level was significantly associated with any DR (OR = 1.01 per 1 mg/dl increase, P < 0.001) and VTDR (OR = 1.39 per 1 mg/dl increase, P = 0.003) in the multiple logistic regression model, and participants in the fourth quartile had a 3.09-fold (P < 0.001) higher risk for any DR and a 6.41-fold (P < 0.001) higher risk for VTDR than participants in the first microalbuminuria quartile (Table 4).

Table 4 Logistic regression of GFR and microalbuminuria with presence of any DR and VTDR.
Full size table

To further assess the influence of microalbuminuria on the relationship between GFR and the presence of DR and VTDR, participants were divided into six groups according to their GFR level and the presence of microalbuminuria. Our results showed that a low GFR was only associated with any DR in the presence of microalbuminuria (OR = 2.40 for participants with GFR > 60 and microalbuminuria (+), P < 0.001; OR = 3.37 for participants with CKD and microalbuminuria (+), P = 0.001), whereas the GFR was an independent risk factor for VTDR regardless of whether microalbuminuria was present (all P < 0.05) (Table 4).

A low GFR (OR = 0.97 per 1 mL/min/1.73 m2 increase, P < 0.001) and CKD (OR = 4.23, P < 0.001) were associated with DMO in the age- and gender-adjusted logistic model, and this association remained statistically significant after adjusting for additional confounders (all P < 0.05). A significant association between DMO and microalbuminuria was found in the third and fourth microalbuminuria quartiles in the multivariate logistic model, and this association increased from an OR of 3.22 in the third quartile (P = 0.017) to an OR of 5.25 in the fourth (P = 0.001) (Table 5).

Table 5 Logistic regression of GFR and microalbuminuria with diabetic macular oedema.
Full size table

Discussion

In the current study, we found that both a low GFR level and a high level of microalbuminuria were significantly associated with the presence of any DR, VTDR and DMO in Chinese T2DM participants. We further assessed the influence of microalbuminuria on the association of GFR with the presence of DR, and the results showed that a low GFR was associated with any DR only in the presence of microalbuminuria, but that GFR was an independent risk factor for VTDR even when microalbuminuria was not present.

The exact relationship between GFR and DR is still controversial, with some studies demonstrating a strong negative association [8, 9, 23,24,25,26] and other studies indicating either no association or an indirect one [10, 11, 27]. Some studies even indicated a GFR-DR association only under certain scenarios [9, 25]. One study from Italy showed an association between a low GFR and advanced DR, regardless of whether microalbuminuria was present, although the association was strongest in the microalbuminuria group [24], whereas the Singapore Prospective Study Program showed that low GFR levels were associated with any DR and advanced DR in particular only in the presence of microalbuminuria [11]. In the current study, we also found that a low GFR was associated with any DR only in the presence of microalbuminuria, while GFR was independently associated with VTDR even without the presence of microalbuminuria. Our study further confirmed the hypothesis of shared pathogenesis mechanisms involving microvascular damage to retinal and renal vessels, leading to progression to DR and DN in diabetic patients [28]. Further longitudinal studies should be performed to verify the exact causal relationship between GFR and DR in future.

We found that microalbuminuria is strongly associated with any DR and VTDR, which is consistent with previous studies [10, 18, 27, 29,30,31]. Furthermore, the present study showed that microalbuminuria was associated with DR even in the presence of a normal GFR level. Microalbuminuria was also shown to be associated with other diabetic micro- and macrovascular complications, suggesting its role as an indicator of generalised vascular damage [32]. These findings indicated that microalbuminuria may be a useful biomarker for DR development, and that close monitoring of microalbuminuria may help clinicians to detect and treat DR early.

Few studies have investigated the association of DMO with renal function. A clinic-based cross-sectional study from Melbourne observed no associations between GFR and DMO in 61 Caucasian patients with DMO [9], whereas another population-based study indicated that CKD was significantly associated with incident DMO during 10-year follow-up examination [33]. The positive association of DMO and GFR was also observed among 413 Southern Chinese patients with T2DM in a single-centre retrospective observational study [31]. The microalbuminuria-DMO relationship was also inconclusive. One study from Korea showed that microalbuminuria was not associated with DMO in 971 Korean DM patients [26], however, another study have shown a positive relationship between microalbuminuria and DMO [34]. Our study found that CKD and high microalbuminuria level were both independent risk factors for DMO, and that the risk increased along with the microalbuminuria level. The reported disparity of DMO-renal function associations may be partly due to differences in study designs, subject races and study sample sizes. Thus, further longitudinal studies on DMO and renal function with different races and disease stage severities are needed to verify our finding.

The strengths of this study include its large sample size and detailed ocular examinations using a standardised protocol. Each participant underwent seven-field fundus imaging to ensure the possible maximal grade for the presence of DR; this was done because some features of DR occur in the peripheral retina, which may be missed when using two-field fundus imaging. The main limitation is that our study involved community-based recruitment from the Yuexiu Family Doctor Project, and that selection bias may exist. Thus, this study’s conclusions may not be generalisable to the general population. However, the relatively large study population obtained by consecutive recruitment makes our results reliable. Second, a causal relationship between DR and renal function could not be determined due to the cross-sectional design of our study, this will need to be further examined in longitudinal studies. Third, we used the UK NDESP grading system instead of the commonly used ETDRS grading system which comprises different definitions. However, these two grading systems perform similarly when used to assess the presence and severity of DR, meaning that this would have little effect on the results.

In conclusion, the current study indicated that a low GFR was associated with any DR only in the presence of microalbuminuria, while the GFR was independently associated with VTDR regardless of the microalbuminuria status among Chinese diabetic participants. CKD was an independent risk factor for DMO. A high level of microalbuminuria was significantly associated with the presence and severity of DR as well as the presence of DMO. Our findings may provide evidence for potential mechanisms underlying DR and DN and have clinical implications for the management of DM. Further longitudinal studies on the association between renal function and DR are needed to evaluate the causal relationship and potential mechanisms.

Summary

What was known before

  • 1, There was a strong association between GFR and diabetic retinopathy in type 1 diabetes mellitus patients 2, However, studies on the association of the GFR with diabetic retinopathy among type 2 diabetes mellitus patients were equivocal.

What this study adds

  • GFR was associated with any DR only under the premise of microalbuminuria, while GFR was associated with VTDR regardless of microalbuminuria status. CKD and high microalbuminuria were independent DMO risk factors.