It is well-known that diabetes mellitus is a group of metabolic diseases characterized by hyperglycemia. Meanwhile, it is a major risk factor for microvascular disease. Nowadays, diabetes mellitus has become one of the world’s fastest-growing chronic diseases. Since 2000, the International Diabetes Federation has reported the national, regional, and global occurrence of diabetes mellitus [
DR is a sight-threatening microvasculature impairment that seriously impacts the lives of diabetic patients. The global prevalence of DR, for the period 2015 to 2019, was 27.0% for any DR comprising of 25.2% nonproliferative DR and 1.4% proliferative DR [
The burden of an increasing population of diabetes mellitus will be more and more heavy based on the above reports. Since it has become an unstoppable trend, it is extremely necessary to take prevention seriously to decrease the incidence of associated complications. Therefore, our study tended to develop a risk nomogram for the prediction of DR.
This study was based on a cross-sectional study conducted in 6 different communities from Yangpu District and Pudong New District in Shanghai from September 2015 to December 2018. A randomized stratified multiple-stage sampling method was used to select a representative sample. At first, Yangpu District belonging to urban areas and Pudong New District belonging to suburban areas were randomly selected. Then, 6 communities were randomly chosen from the two districts, including Huamu Community, Jinyang Community, and Sanlin Community in Yangpu District and Yinhang Community, Siping Community, and Daqiao Community in Pudong New District. The cross-sectional study is aimed at investigating the situation about type 2 diabetes mellitus (T2DM) from community grassroots in Shanghai by cooperating with the community health centers who were responsible for the management of chronic diseases and had a health registration system for the residents with diabetes within its range of services. The study was approved by the Institutional Review Board of Shanghai Jiao Tong University School of Medicine and was performed abiding by the principles of the Declaration of Helsinki of 1975 which was revised in 2008. Therefore, based on the data from the cross-sectional study, this study was about to discover the risk factors associated with DR and develop a predictive model to present the influence of those risk factors visually and quantitatively.
4170 T2DM patients were included in this study in total. And the criteria for inclusion were as follows: (1) T2DM patients own a registered regional household or live in the community for more than 6 months and have been included in community health registration system; (2) T2DM patients must be at least 18 years old; (3) T2DM was diagnosed according to the international diagnostic criteria and classification declared by the WHO in 1990 which defined diabetes according to a fasting plasma glucose concentration of 7.0 mmol/L (126 mg/dL) or higher, or 2 h postglucose load venous plasma glucose of 11.1 mmol/L (200 mg/dL) or higher, or random plasma glucose concentration of 11.1 mmol/L (200 mg/dL) or higher; and (4) all participating patients were provided written informed consent and willing to participate in the study.
All subjects received questionnaire survey, biochemical indicator examination, physical examination, and fundus examination. The questionnaire survey was used to receive the basic information of the subjects, including age, gender, and course of disease. Biochemical indicator examination was conducted in the morning when all subjects were on fasting state and postprandial state, so fasting blood, 2 h postprandial blood, and morning urine were collected. It focused on the examination of fasting blood glucose (FBG), postprandial blood glucose (PBG), glycosylated haemoglobin A1c (HbA1c), total cholesterol (TC), total triglycerides (TG), low-density lipoprotein (LDL), high-density lipoprotein (HDL), serum creatinine (SCR), blood urea nitrogen (BUN), uric acid (UA), urine creatinine (UCR), and urinary microalbumin (UMA). Physical examination was focused on the measurement of blood pressure, height, weight, waist circumference, hip circumference, and the body mass index (BMI), and waist-to-hip ratio (WHR) was calculated. To assess the severity of DR, fundus examination was performed for all subjects by using a nonmydriatic fundus camera (CR-2 AF, Canon Inc., Tokyo, Japan) to get their digital, color, and nonstereoscopic retinal photographs. 45° digital retinal photographs were captured on the posterior pole of each eye (Figures
The normal fundus characteristics of T2DM patients (a) and the abnormal fundus characteristics of DR patients (b).
Statistical analysis was performed using the R software (version 3.6.1;
Further, several kinds of validation methods were used to estimate the accuracy of the risk prediction model by using the data of the training set and the validation set, respectively. The area under the receiver operating characteristic (ROC) curve was used to provide good discrimination for the quality of the risk nomogram to separate true positives from false positives [
A total of 4170 T2DM patients including 2412 females and 1758 males [
Differences in characteristics between the training set and the validation set.
Variables | Total ( | Training set ( | Validation set ( | |
---|---|---|---|---|
Age (years) | 0.106 | |||
T2DM duration (years) | 0.665 | |||
Gender | 0.102 | |||
Female | 2412 (57.8%) | 1833 (58.6%) | 579 (55.7%) | |
Male | 1758 (42.2%) | 1297 (41.4%) | 461 (44.3%) | |
FBG (mmol/L) | 0.552 | |||
PBG (mmol/L) | 0.481 | |||
HbA1c (%) | 0.652 | |||
TC (mmol/L) | 0.681 | |||
TG (mmol/L) | 0.202 | |||
LDL (mmol/L) | 0.729 | |||
HDL (mmol/L) | 0.770 | |||
SCR ( | 0.412 | |||
BUN (mmol/L) | 0.217 | |||
UA ( | 0.837 | |||
UCR (mmol/L) | 0.118 | |||
UMA (mg/L) | 0.038 | |||
SBP (mmHg) | 0.041 | |||
DBP (mmHg) | 0.661 | |||
BMI (kg/m2) | 0.729 | |||
WHR | 0.682 |
In all 19 associated characteristic variables, 7 potential predictors were selected on the basis of the data from the training set (Figures
Variable selection by LASSO binary logistic regression model. A coefficient profile plot was produced against the log(lambda) sequence (a). Seven variables with nonzero coefficients were selected by optimal lambda. By verifying the optimal parameter (lambda) in the LASSO model, the partial likelihood deviance (binomial deviance) curve was plotted versus log(lambda) and dotted vertical lines were drawn based on 1 standard error criteria (b).
The results of the logistic regression analysis among age, course of disease, PBG, HbA1c, UCR, UMA, and SBP are given in Table
Predictors for the risk of DR in T2DM patients.
Intercept and variables | Prediction model | |||||
---|---|---|---|---|---|---|
Odds ratio | CI (2.5%) | CI (97.5%) | ||||
Intercept | -3.037 | -5.097 | <0.001 | 0.048 | 0.015 | 0.154 |
Age | -0.031 | -4.442 | <0.001 | 0.969 | 0.956 | 0.983 |
Course | 0.050 | 7.416 | <0.001 | 1.051 | 1.038 | 1.066 |
PBG | 0.027 | 2.349 | 0.019 | 1.027 | 1.004 | 1.050 |
HbA1c | 0.258 | 7.058 | <0.001 | 1.294 | 1.205 | 1.390 |
UMA | 0.002 | 3.256 | 0.001 | 1.002 | 1.001 | 0.995 |
UCR | -0.027 | -2.415 | 0.016 | 0.973 | 0.951 | 1.003 |
SBP | 0.007 | 2.959 | 0.003 | 1.007 | 1.002 | 1.012 |
Development of the risk nomogram (a) and the dynamic nomogram for an example (b). The DR risk nomogram was developed with the predictors including age, course of disease, PBG, HbA1c, UCR, UMA, and SBP.
For the predictive model, the pooled area under the ROC curve of the nomogram was 0.700 in the training set and 0.715 in the validation set (Figures
ROC validation of the DR risk nomogram prediction. The
Calibration curves of the DR risk nomogram prediction. The
The decision curve showed that it would be more accurate to use this nomogram in the current study to predict the risk of DR when the risk threshold probability was between 21% and 57%, and in the validation set, it was between 21% and 51% (Figures
Decision curve analysis for the DR risk nomogram. The
Nomograms are considered reliable and pragmatic prediction tools, with the ability to generate an individual probability of a clinical event by integrating diverse prognostic and determinant variables [ [
In this study, about 21.3% of the T2DM patients have been complicated with DR. In the risk factor analysis, age, course of disease, PBG, HbA1c, UCR, UMA, and SBP are associated with the risk of DR in T2DM patients. Based on that, we built and validated a novel prediction tool for the risk of DR among T2DM patients using these 7 available variables. This predictive model suggested that younger age and longer course of disease, higher PBG and HbA1c, lower UCR and higher UMA, and higher SBP were the key individual factors that determined the risk of DR for T2DM patients. By introducing basic information and biochemical and physical examination indicators into the DR risk nomogram, it was beneficial and convenient for the T2DM individualized prediction of risk of DR. This study provided a relatively accurate prediction tool of risk of DR for T2DM patients. It demonstrated relatively good discrimination and calibration power, which identified that this nomogram could be widely and accurately used for its large sample size [
Age and course of disease are unmodifiable risk factors for T2DM patients. The result showed that younger age and longer course of disease were strongly connected with the incidence of DR. They should be as a pair of risk factors when we consider the association in it. Once T2DM is diagnosed, abnormally elevated blood sugar induces oxidative stress and causes microinflammation [
Hyperglycemia is a well-known factor closely associated with DR development. Our study also indicates that higher PBG and HbA1c may contribute to the risk of DR. One study reported that by categorizing PBG and HbA1c by deciles, with the prevalence of DR calculated in each decile, the prevalence of DR increased sharply in the 10th decile, when HbA1c exceeded 6.4%. The threshold of HbA1c for detecting DR is nearly consistent with the criteria for diagnosing diabetes from the World Health Organization [
As we all know, UMA and UCR are biochemical indicators reflecting renal function, and T2DM is a kind of chronic disease with systemic metabolic disorders. So the abnormal renal function metabolic indicators not only can indicate the renal disease but also suggest the risk of indirectly associated lesions like eye disease. Previous studies have proved that UMA is a strong predictor for DR and hard exudate formation in type 2 diabetics even after correcting for the duration of diabetes and other systemic risk factors [
The relationship between DR and hypertension has been suggested in some previous studies [
T2DM is a kind of chronic lifelong disease, so it is extremely important to control disease development and prevent the occurrence of associated complications. Controlling blood glucose is always the first measure to decrease the risk of associated complications. However, most of the T2DM patients cannot achieve well-control blood glucose due to the lack of preventive consciousness or the lack of knowledge about T2DM, which indicates the increasing risk of associated complications. Therefore, it is meaningful to develop risk prediction tools to help doctors and T2DM patients raise vigilance and take preventive measures. DR is one of the diabetic complications with high incidence and great damage to health. So we developed a valid prediction tool, which assisted clinicians with early identification of patients at high risk of DR through the developed nomogram. Moreover, early interventions such as changing therapeutic scheme will benefit to decrease the risk of DR.
There are some limitations to this study. Firstly, this study is based on the epidemiological data from a screening on community T2DM and its complication, so that the diagnosis of DR may lack strictness. So fundus photography with larger vision is needed and beneficial to ensure the accuracy of DR diagnosis. Secondly, that the predictive model showed medium prediction accuracy may suggest that more other indicators should be included such as inflammation-related indicators and oxidative stress-related indicators based on the mechanism of inflammation and oxidative stress [
Combining with age, course of disease, PBG, HbA1c, UCR, UMA, and SBP, this study built a novel nomogram with relatively good accuracy to help clinicians and T2DM patients estimate the risk of DR. According to the evaluation result, clinicians and patients can take more targeted measures on medical interventions in time. As for the limitation of the predictive model, there is still a large space to improve the nomogram and increase the clinical utility.
The data used to support the findings of this study are available from the corresponding author upon request.
The author and coauthors declare no conflict of interest associated with this study.
This research was financially supported by the fourth round of Shanghai Public Health Three-Year Action Plan Key Discipline Construction–Health Education and Health Promotion (Grant No. 15GWZK1002). Meanwhile, thanks are due to the directors and medical staff of the 6 community health centers in Shanghai for participating in this study, as well as all T2DM patients who participated in this study and to the Shanghai Municipal Health and Family Planning Commission for their funding support.