Distribution of diabetic retinopathy in diabetes mellitus patients and its association rules with other eye diseases

The study aims to explore the distribution characteristics and influencing factors of diabetic retinopathy (DR) in diabetes mellitus (DM) patients and association rules of eye diseases in these patients. Data were obtained from 1284 DM patients at Henan Provincial People’s Hospital. Association rules were employed to calculate the probability of the common occurrence of eye-related diseases in DM patients. A web visualization network diagram was used to display the association rules of the eye-related diseases in DM patients. DR prevalence in people aged < 40 years (≥ 58.5%) was higher than that in those aged 50–60 years (≤ 43.7%). Patients with DM in rural areas were more likely to have DR than those in urban areas (56.2% vs. 35.6%, P < 0.001). DR prevalence in Pingdingshan City (68.4%) was significantly higher than in other cities. The prevalence of DR in patients who had DM for ≥ 5 years was higher than in other groups (P < 0.001). About 33.07% of DM patients had both diabetic maculopathy and DR, and 36.02% had both diabetic maculopathy and cataracts. The number of strong rules in patients ≥ 60 years old was more than those in people under 60 in age, and those in rural areas had more strong rules than those in urban areas. DM patients with one or more eye diseases are at higher risks of other eye diseases than general DM patients. These association rules are affected by factors such as age, region, disease duration, and DR severity.

Definitions and diagnostic criteria. DM was diagnosed according to the Standards of Medical Care in Diabetes established by the American Diabetes Association 9 . Fundus photographs of each participant were taken according to unified standards by an ophthalmologist using a Zeiss non-mydriatic fundus camera (VISUCAM 224, Germany), with the macula as the center and 45° Color mode shooting. Five fields were captured in each eye: macula center, temporal side, nasal side, upper and lower quadrants of the retina. Then, we selected the clearest picture from the three photographs as the final fundus photograph of the patient.
DR diagnosis and grading were conducted by two clinicians (with ≥ 5 years of experience) based on international clinical diagnosis and classification standards 10 . DR was divided into five stages: (0) absence of DR (NDR), (1) mild non-proliferative diabetic retinopathy (NPDR), (2) moderate NPDR, (3) severe NPDR and (4) proliferative diabetic retinopathy (PDR). When the grading results of the two clinicians were not in agreement, adjudication by a panel of four retina specialists served as the reference standard. Patients with DR in both eyes were regarded as one case and included in the analysis. Those with different degrees in each eye were graded according to the more severe eye. Cataracts 11 , glaucoma 12 , age-related macular degeneration 13 , diabetic maculopathy 14 and refractive errors 15 were diagnosed by ophthalmologists according to standard clinical diagnostic criteria. Microaneurysm 16 , optic atrophy 17 and tessellated retina 18 were determined based on the fundus photographs and references.
Vision was measured using the standard logarithmic visual acuity chart with the participants standing 5 m in front of the chart. The smallest line of vision that could be recognized by the examined eye was recorded as the value of the vision, which is the visual acuity of the examined eye. The vision of the eye with better vision was considered to be the participant's vision for the final analysis. The intraocular pressure of the eyes was measured using a non-contact tonometer (Canon TX-20P) according to the manufacturer's guidelines. The average of three measurements of each eye was taken as the value for each eye. Bergmeister papilla was observed using optical coherence tomography (OCT; Carl Zeiss Meditec, Inc.) focused on the optic cup and optic disc. The vertical and horizontal diameters of the optic cup and optic disc were measured using the measuring tool of the OCT machine. The average diameter of the optic cup or optic disc was calculated as the sum of the vertical and horizontal diameters divided by two. were used for the statistical analysis. For the qualitative variables, frequency was descripted. The χ 2 test was used for comparisons between groups. For the quantitative variables, mean ± standard deviation (SD) was determined, and t-tests were used for between-group comparisons. To observe the age distribution of DR, most of the patients were divided into seven age groups, spanning 10 years each, except for the first group, which included individuals aged 18.0-29.9 years. Those ≥ 80 years were merged into one group due to the small sample size. Regional distribution was created using free online software (http://c. dituh ui. com/). Multiple correspondence analysis was employed to show the relative relationship of eye-related diseases in DM patients. Apriori algorithm of association rules was used to determine the strength of the rule through the support (S) and the confidence (C) 19 , which were used to calculate the probability of co-occurrence of eye-related diseases in DM patients. Conditional support was given as S(X) = P(X), where P is the probability of the occurrence of X; and rule support S(X ⟹ Y) = P(X ∪ Y) is the probability of X and Y occurring at the same time. In this study, X and Y refer to eye-related diseases. Confidence, C(X ⟹ Y) = P(X⁄Y) = S(X ⟹ Y)/S(X), is the conditional probability of Y occurring under the condition that X occurs. Lift, P(Y|X)/P(X), reflects the correlation between X and Y in the association rules. Lift > 1 indicates a higher positive correlation; lift < 1 indicates a higher negative correlation; and lift = 1 indicates no correlation. In this study the minimum support, confidence, and lift were set to 15%, 80%, and 1.10, respectively. A web visualization network diagram was employed to show the association rules among the eye-related diseases of DM patients. All the P-values were two-tailed, and the level of significance was set at α = 0.05. Distribution characteristics of DR in DM patients. While there were more DR patients in the 50-60 age group, the prevalence of DM in the population under 40 years old was higher than that in the 50-60 age group (58.5-66.7%) (Fig. 1A). There was no statistically significant difference in the prevalence of DR between males and females with DM (χ 2 = 1.777, P = 0.680) (Fig. 1B). Among the 580 DR patients, 112 had mild NPDR, 144 had moderate NPDR, 170 had severe NPDR and 154 had PDR (Fig. 1C). DM patients in Pingdingshan (68.4%), Shangqiu (62.5%), Anyang (60.0%), and Zhoukou (58.2%) cities had a higher risk of DR (χ 2 = 43.582, P < 0.001) than those from other areas (Fig. 1D). There was a higher risk of DR in DM patients with a disease duration of > 5 years (42.9-60.0%) (χ 2 = 114.219, P < 0.001) than those with a disease duration of ≤ 5 years (Fig. 1E). There was a statistically significant difference in the prevalence of DR between rural and urban patients (χ 2 = 57.199, P < 0.001) (Fig. 1F).

Multiple correspondence analysis. Multiple correspondence analysis was performed to intuitively
understand the relationship between eye-related diseases and DR in DM patients, in which, age, gender, DM duration, hometown, cataracts, glaucoma, age-related maculopathy, diabetic maculopathy, hypertension, optic atrophy, and DR were employed as analysis variables. The results showed that glaucoma, age-related macular disease, refractive error, and optic atrophy had no obvious aggregation phenomenon from the joint plot of the category points, while the attributes of being a rural patient, blindness, hemangioma, and an age ≥ 40 years had obvious aggregations with DR ( Fig. 2).

Association rules of eye diseases in DM patients. A web visualization network was developed to
visually observe the association rules among eye-related diseases in DM patients (Fig. 3). Typical fundus photographs in patients are shown in Fig. 4 Results of association rules showed that the co-occurrence of DR, diabetic macular degeneration, and cataracts among the DM patients was very high, and the lift was greater than 1.1, indicating that the prevalence of eye diseases was positively correlated ( Table 2).
In patients under 40 years, the support and confidence of the rule "Tessellated retina, diabetic maculopathy → DR" were 25% and 100%, respectively, that is, 25% of those younger than 40 years had both tessellated retinas and diabetic maculopathy, and the prevalence of DR in this group was 100%, which is 1.54 times that found in the DM patients without tessellated retinas and diabetic maculopathy. The number of strong rules in patients ≥ 60 years old was more than that in those under 60 years old. The co-occurrence rate of optic atrophy and tessellated retina was 25%, and the prevalence of diabetic maculopathy in this population was 86.49%, which is 1.19 times that in the DM patients without optic atrophy and tessellated retina. In urban areas, 24.55% of patients had refractive errors, and 91.18% of this group had cataracts, which is 1.35 times the rate for the general www.nature.com/scientificreports/ population. Rural patients had more rules than those in urban areas. In DM patients with a disease duration of less than 10 years, there was no strong rule. In DM patients with a disease duration of more than 10 years, the probability of tessellated retina and DR occurring at the same time was 15.19%, and the prevalence of diabetic maculopathy was 100% in patients with both tessellated retinas and DR, which is 1.15 times that found in the population without tessellated retinas and DR. In addition, the results showed that the number of strong rules in male patients with DM was fewer than in female patients.

Discussion
DR can very easily cause blindness, which seriously affects a patient's quality of life 20 . In this study, we observed 1284 DM patients with different ocular diseases at the Henan Provincial People's Hospital, and explored the distribution and association rules of DR in these populations using multiple correspondence analysis and association rules. The results showed that: (1) among the included DM patients, the prevalence of DR varied with ages, regions, and DM durations, but there was no statistical difference in the prevalence of DR between males and females; (2) patients with one or more eye diseases are at a higher risk of other eye diseases than those in the general population, and these association rules are affected by age, disease duration, and grade of DR classification. The proportion of DR in DM patients aged 40-60 years was significantly higher than that in people > 60 years old. This may be due to the fact that young and middle-aged people face more stress factors from comprising the main workforce in society 21 . These factors include a fast work pace, a high level of mental stress, high levels of endocrine disorders, and increased catecholamines; these can easily trigger retinal vasoconstriction causing ischemia and hypoxia of the retina 22 . In addition, the prevalence of type 1 diabetes in young patients is higher.  The study showed that the prevalence of DR in type 1 DM was higher than in patients with type 2 DM, which may be associated with the higher prevalence of diabetic retinopathy in young patients 23 . Ineffective blood glucose control in patients with DM is often accompanied by lipid metabolism disorders 24 . The combined effect of all these factors causes a reduction in barrier function of vascular endothelial cells and the ability of red blood cells to carry oxygen, which will lead to the onset of other eye diseases in DR patients 25 . Therefore, patients with DM should pay strict attention to their lifestyle and eating habits, including learning how to relieve stress, using medication responsibly, regularly checking their blood glucose and the functional status of other tissues and organs, maintaining blood glucose stability, and avoiding the further deterioration of DM. In addition, the regional distribution of the included DM patients was observed. The results showed that DM patients in rural areas of Henan were more likely to develop DR than those in urban areas. However, a meta-analysis by Song et al. 26 showed the prevalence of DR was higher in urban than in rural settings. This inconsistency may be because the population structure of our study is different from those of other researches. Epidemiological surveys have shown that DM patients in rural areas of Henan have a significantly lower level of cognition and self-management ability for DM than those in urban areas 27 , which can lead to an increase in the possibility and speed of progression of DM to DR. Therefore, strengthening the public health services of rural grassroots medical staff and guiding patients in effective self-management may be key in comprehensively improving the effectiveness of DR prevention and control in rural patients. It is noteworthy that in the 17 cities in Henan Province, the prevalence of DR in DM patients in Pingdingshan City was significantly higher than in patients in other regions. This may be due to the diet in this region. Studies have shown that the daily dietary fat intake of residents in this area is significantly higher than those in other areas in Henan Province; in particular, www.nature.com/scientificreports/ the intake of edible oil is higher than the recommended amount, and the intake of cereals and beans is lower 28 . Therefore, strengthening the self-management of the diet of patients with DM and ensuring a balanced and controlled diet is essential to reducing the incidence of DM and, consequently, the prevalence of DR 29,30 . Consistent with the results in the literature 31,32 , this study found that the occurrence of DR was closely related to the duration of DM. In particular, there was a higher proportion of DR in patients with a diabetes duration of ≥ 5 years than those in other groups. This result is expected given the fact that the longer the duration of the disease, the longer patients are exposed to many risk factors and the higher the incidence of various DM-related chronic complications 33 , which may be caused by the interaction of multiple factors. Some studies have shown that in DM patients with good glycemic control, the occurrence and development of DR can still be observed www.nature.com/scientificreports/ when the disease duration is long, which may be related to the memory of blood glucose metabolism 34 . Therefore, it is necessary to regularly monitor the fundus photographs of DM patients. Regardless of whether blood glucose control is stable, the frequency of eye follow-ups should be increased with time 35 . DM-related fundus changes should be found early and treated as soon as possible to prevent the further development of the disease. The results of multiple correspondence analyses and studies showed that macular edema, hemangioma, hard exudates and retinal hemorrhages were the most common signs of DR. Traditionally, DR is regarded as a microvascular disease, and its main pathogenic links are closely related to pathological cascades, such as retinal inflammation triggered by hyperglycemia, blood-retinal barrier destruction, neovascularization, increased vascular permeability, and bleeding. Therefore, the strict control of blood glucose concentration is a key factor in controlling the occurrence of many diseases, including DR.
Finally, we analyzed the comorbidity of the eye diseases in DM patients. The results showed that the probability of comorbidity of DR, diabetic maculopathy, and cataracts in DM patients was high with a lift > 1.1, which suggests that the prevalence of these diseases is positively correlated. In addition, we found that for the same rule, the correlation strength was different in people with different characteristics. For example, "refractive error → cataract" was a strong correlation in DM patients who were > 60 years old, lived in urban areas, and had DR grades ≤ 3, while in DM patients younger than 60 years who lived in rural areas, and had DR grades > 3, there was no such association rule. There was no association rule in patients with a diabetes duration ≤ 10 years, while there were five strong association rules in patients who had DM for > 10 years. It is noteworthy that in DM patients with a duration of ≥ 10 years and combined with tessellated retina, the prevalence of diabetic maculopathy was 100%. Tessellated fundus was recognized as one of the markers in identifying high myopia. In high myopia, with the increase of axial length, sclera becomes thinner 36 , and the retina and choroid are mechanically stretched, which leads to the thinning of visual ganglion cells, resulting in ischemia and hypoxia of retinal and choroidal, which may be an important reason for maculopathy in high myopia 37,38 . In addition, DM patients with a duration of over 10 years, fundus lesions gradually aggravated and involved in macular area. Moreover, the patients included in this study often had obvious symptoms, and the prevalence of diabetic maculopathy was higher than that of the community population. Given these reasons, a very high-level prevalence was observed in this study.
In patients with a DR grade < 3, "DR → diabetic maculopathy" was a strong rule, but patients with grades > 3 did not have this strong rule. Studies have shown that for every year an individual has DM, the risk of DR increases 1.29 times 39 . Research conducted by Jeng showed that the risk of non-proliferative DR in cataract surgery patients was 1.48-4.11 times that of the population without cataracts 40 . Our results are similar to those of that study, indicating that the co-occurrence of ocular diseases and hyperglycemia has a certain correlation: one disease will increase the risk of another disease. This may be due to the fact that statins and angiotensinconverting enzyme inhibitors used by cataract patients change the blood-retinal barrier, causing increases in vascular endothelial growth factor, monocyte chemoattractant protein-1, interleukin-1β (IL-1β), and IL-6, thereby increasing the risk of DR 40,41 .
This study has several limitations. First, the participants included only DM patients in the Ophthalmology Department of Henan Provincial People's Hospital, and the sample size was small. Therefore, its representativeness may be poor, meaning there is a need to further increase the sample size and/or conduct multi-center research. Second, our data are lack of the diagnosis of dyslipidemia and cardiovascular diseases, the impact of these systemic diseases on DR could not be analyzed. Third, as this was a cross-sectional study, it was impossible to judge the sequence and evolution of the various eye diseases, that is, the causality could not be determined. Therefore, in order to reflect the nature of the diseases, prospective studies in a large population are still needed to verify these results.

Conclusion
The prevalence of DR in DM patients varies with ages, regions, duration of DM and the grade of DR. Patients with one or more eye diseases are at higher risks of developing other eye diseases than those in the general population, and this association rule can be affected by factors such as age, region, duration of the disease, and the severity of DR. This research may provide a valuable scientific basis for understanding the distribution and association rules for the screening, diagnosis, and treatment of DR.

Data availability
All relevant data are included in the papers. Contact corresponding author for additional information regarding data access.