Relationship Between Incremental Glucose Peak and Diabetic Retinopathy in Tertiary Hospital: Liaoning Diabetic Microvascular Complications Study (LD-MCS)

Background: Glucose variability may contribute to macrovascular complications of diabetes but evidence for its association with diabetic retinopathy (DR) is limited. This study is aimed to investigate the correlation between incremental glucose peak (IGP) and diabetic retinopathy (DR) in Chinese population with type 2 diabetes. Methods: A hospital-registry, case-control study collecting data from Liaoning Diabetic Microvascular Complications Study (LD-MCS) during 2012-2018. 795 adults with DR were included and matched with two diabetic controls per case (n = 1590) by gender and age. IGP was dened as the peak minus baseline plasma glucose value during 5-point oral glucose tolerance test (OGTT). DR was diagnosed by two-eld fundus photographs under Early Treatment Diabetic Retinopathy Study criteria and stratied into mild non-proliferative DR (NPDR), moderate NPDR, and vision-threatening DR (VTDR). Logistic regression analysis was applied to determine the relationships. Results: Among 2385 participants (age, 58.40±11.52 years; 57.11% men), higher IGP was independently associated with the presence of any DR (odds ratio [OR], 1.05; 95% condence interval [CI], 1.02-1.07; P < 0.001), moderate NPDR (OR, 1.06; 95% CI, 1.01-1.12; P = 0.020) and VTDR (OR, 1.13; 95% CI, 1.07-1.19; P < 0.001), respectively. Moreover, estimated ratio for any DR and VTDR from low-quartile IPG increased by 1.33 (95% CI, 1.04-1.70; P for trend = 0.021) and 3.64 (95% CI, 1.97-6.72; P for trend < 0.001), respectively. Conclusions: OGTT-derived IGP was independently associated with DR severity scales in Chinese patients with type 2 diabetes. More attention by health care personnel needs to be paid to glucose variability of this population. similar

Glycemic variability (GV), the dynamic evaluation of plasma glucose changes, deepened the concept of glycemic control beyond HbA1c. However, evidence for the effect of GV on retinopathy of type 2 diabetes was limited. Sonoda et al. suggested the mean amplitude of glycemic excursions (MAGE) had no signi cant correlations with DR. [10] However, some evidences certi ed direct or indirect damage to retina caused by GV based on 72-hour continuous glucose monitoring system (CGM) or 1,5-anhydroglucitol (an indicator related to postprandial hyperglycemia and GV), especially in acute uctuation and hyperglycemia. [11][12][13] Clinically, CGM use is still considered experimental and limited, we attempted to focus on incremental glucose peak (IGP) derived from oral glucose tolerance test (OGTT), which had be considered as an appropriate alternative for approximation of GV. [14,15] IGP refers to the glycemic difference between baseline and peak during OGTT, which has been found with signi cant associations with some macrovascular events. [16,17] However, its association with microvascular complications, such as DR, has not been explored yet. Accordingly, this study aimed to investigate relationships between IGP and DR as well as its severity strata in Chinese population with type 2 diabetes.

Research design and study population
We used data from Liaoning Diabetic Microvascular Complications Study (LD-MCS). The LD-MCS is a series of hospital-based observational studies and aimed to identify risk factors and biomarkers of diabetic microvascular complications for Chinese adults with type 2 diabetes. Theory basis and methodology of the LD-MCS were reported in the Appendix (see Additional le 1).
Cases and controls de ned as participants with DR or without DR, respectively. The control group sample was calculated by power of 80%, ratio of cases to controls: 1:2 and percentage of controls exposed to higher GV: 10% and Odds ratio 1.20 (both provided based on pilot investigation). Added 10% for losses, sample calculation resulted in 620 cases and 1240 controls.
After assessment in LD-MCS, 1469 cases with DR were collected from the database and subsequently some participants with incomplete data, either because an OGTT contraindication (n = 568; i.e. plasma glucose levels > 11.0 mmol/L before initiation of the OGTT), or ambiguous fundus grades (n = 106) were removed. Each remaining case was matched randomly with two controls without any history of DR by same gender and age in the database. Finally, 795 con rmed cases (54.1%) with DR and 1590 corresponding controls amounting to 2385 participants were recruited in this study.

Assessment Of Diabetic Retinopathy
Two ophthalmologists of the LD-MCS were invited to independently evaluate fundus situation of participants and con rm the diagnosis and severity scales of DR under guidelines of the Early Treatment for Diabetic Retinopathy Study (ETDRS). [18] Each image was a two-eld colorful one taken by nonmydriatic digital retinal photography (CR6-45NM; Canon, Inc., Tokyo, Japan).  and insulin was expressed in µIU/mL, to evaluate insulin resistance and pancreatic β-cell function. [20] Estimated glomerular ltration rate (eGFR) was calculated by the equation of Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) to evaluate renal function. [21] Statistical analysis All statistical analyses were performed by the SPSS version 24.0. (IBM for Windows). Continuous and categorical data were expressed as mean and SD (mean ± SD) and percentage (No. [%]), respectively.
Data was veri ed for the normal distribution by Kolmogorov-Smirnov test. The Chi-square test and Fisher exact test was applied for discontinuous data comparison and the independent Student's t-test was applied for continuous variables comparison across different study population groups. One-way ANOVA test and the Chi-square test were used to compare variables between various subgroups. Three logistic regression models adjusted for confounding factors were carried out to estimate the odds ratio (OR) with a 95% con dence interval (95% CI). IGP and HOMA-IR were transformed into categorical data by quartiles which were segregated by the 25th, 50th and 75th percentage as Quartile 1 (Q1), Q2, Q3, and Q4. Multiple R-squared liner regression analysis was performed to reveal the correlations between variables and IGP quartiles. In all analyses, P value less than 0.05 was de ned as statistical signi cance.

Study population characteristics
The characteristics of 2385 participants strati ed by the presence and severity of DR were presented in , subjects with DR were like to be with longer duration of diabetes, and had higher level of weight, HR, SBP, PP, HbA1c, 120-min PG, 180-min PG, post-load plasma glucose peak, and IGP (all P < 0.05). Blood glucose variability before and after OGTT among participants with different DR severity scales were described on Figure S1 in Additional le 2. Along with the advanced retinopathy condition, HbA1c, 120-min PG, 180-min PG, post-load plasma glucose peak, and IGP showed an upward tendency (HbA1c: P = 0.014; 180-min PG: P = 0.002; all others: P < 0.001). In view of confounders, Fig. 1 presented bivariate and multivariate logistic regressive analysis performed to further explore the relationships between IGP and DR as well as its severity strata. Adjusted for gender and age (model 1), IGP levels had a statistically signi cant correlation with the presence of DR (OR, 1.05; 95% CI, 1.02-1.07; P < 0.001), moderate NPDR (OR, 1.06; 95% CI, 1.01-1.12; P = 0.020) and VTDR (OR, 1.13; 95% CI, 1.07-1.19; P < 0.001), respectively. Model 2 with adjustment for variables related to severity of DR and model 3 with additional potential adjustment for FPG, HOMA-IR, serum creatinine and eGFR showed similar relationships.

Additional Analyses
Multiple R-squared liner regression including variables about DR severity, OGTT-derived parameters, HOMA-IR, and an index of renal function were used for further assessment of the association between IGP and demographic as well as biomedical variables (Table 4). After stepwise selection of variables in model 1 and 2, only pre-and post-load index entered the models (both adjusted R 2 = 0.908) with remarkable signi cance (all P < 0.001). Model 2: includes gender, age, BMI, SBP, PP, antidiabetic agents, duration of diabetes, HbA1c, FPG, post-meal glucose (30 min, 60 min, 120 min, and 180 min), HOMA-IR, serum creatinine, and eGFR.
Abbreviation: IGP, incremental glucose peak; BMI, body mass index; SBP, systolic blood pressure; PP, pulse pressure; HbA1c, glycated hemoglobin A1c; FPG, fasting plasma glucose; HOMA-IR, the homeostasis model assessment of insulin resistance index; eGFR, estimated glomerular ltration rate. * Using stepwise method to select variables. Stepwise method ordinarily included the most signi cant variable (P < 0.05) and then excluded the most indistinctive variable (P > 0.10), which was aimed to obtain an optimal combine of variables based on Forward method and Backward method.

Discussion
In current study, we investigated case-control associations of IGP with DR as well as its severity in Chinese participants with type 2 diabetes. Firstly, our ndings revealed the relationships between OGTTderived indicators and the risk of retinopathy. HbA1c and 120-min PG value has previously been found to be independently associated with DR which conformed with our ndings, [7,22] while the evidence on the association between 180-min PG value and DR were limited which might partly depend on 120-min PG.
Post-load plasma glucose peak, which was described as a risk factor of cardiovascular diseases with postprandial peak, [23] also had impact of hyperglycaemia on endothelial function and retinal vascular reactivity. Previous studies had proposed that FPG could screen high risk diabetes with microvascular events and 60-min PG could prompt the onset of DR due to theirs close connection with insulin secretion. [22,24] By contrast, in our study no independent association between FPG and 60-min OGTT glucose concentration and presence of DR was found. However, their use of cross-sectional design and 1-or 2-h plasma glucose concentrations by Americans or Indians entails a major different nding.
Additionally, adjusting for gender, age, and potential clinical-based confounders for DR, this present study rst reported the association of IGP index with the presence of any-DR, moderate NPDR, and VTDR. This nding was also consistent with the results when IGP level was considered into categorical variable. Although there is no direct evidence on IGP levels with DR, high IGP was found to be associated with several macrovascular impairments, such as aortic stiffness, carotid intima-media thickness (CIMT), and maladaptive carotid remodeling. [16,17] Moreover, evidences showed that when CGM was unavailable, IGP could be chosen to assess GV in replacement of coe cient of variation (CV; SD / mean * 100%), [14] which was recommended as the leading measurement of variability by the international consensus of the Advanced Technologies & Treatments for Diabetes (ATTD) Congress. [25] The association between variability assessed by IGP and retinopathy was consistent with the results of the GV metrics assessed by CGM, [11,13] while the results of some researches remained in debate and further longitudinal, population-based studies are needed to con rm our ndings . [10] Mechanistic Explanations The biological pathway underlying the relationship between IGP and DR remains to be elucidated. In order to ll this gap, HOMA-IR which was associated with DR, was graded by quartiles, and higher level of insulin resistance increased FPG and 3-h post-load plasma glucose peak (both P value < 0.001), but decreased the level of IGP (see Figure S2 in Additional le 3). In contrast, Murata et al. got a positive correlation between glucose uctuation and HOMA-IR due to de nitional discrepancy between glucose uctuation and IPG. [26] In our study, IGP could re ect an integrated situation of pre-load and 3-h postload plasma glucose and seem to be related to insulin resistance, then form a relationship with DR.
In addition, DR has been traditionally recognized as a micro -vascular disease. Nevertheless, retinal neurodegeneration (neuronal apoptosis and glial activation) was also proved to be involved in the pathogenesis of DR. [27] The activation of Müller cells, the representative glial component of retina, was observed in response to GV. [28] Besides, Picconi et al. and Stem et al. demonstrated that increased glycemic excursion was connected with early structural impairments of neural retina in patients with type 1 diabetes, resulting a stable glycemic status and a decrease in GV might prevent the damage to structure and function of retina. [29,30] Although molecular mechanisms induced by hyperglycemia in the incidence and progression of DR have been described in previous studies, such as the hexosamine pathway, increased polyol pathway ux, protein kinase C (PKC) activation, free radicals formation and advanced glycation end (AGE) products accumulation, the metabolic mechanism of GV was unclear yet. [9] Some evidences indicated that oscillating glucose played a more destructive role on oxidative stress and endothelial function than sustained hyperglycemia. [31,32] And a further mechanism revealed oscillating glucose left a metabolic memory, which suggested the persistence of intracellular pro-oxidant environment after plasma glucose normalization. [33] More observational and experimental evidence for the effect of GV on retinopathy was expected.

Clinical Relevance
Diabetes mellitus causes some macrovascular complications such as cardiovascular diseases (CVD) as well as microvascular complications such as retinopathy, nephropathy, and neuropathy, of which, DR is a major cause of vision loss in middle-aged and elderly people. Previous study reported that IPG was associated with aortic stiffness, which is an independent determinant of CVD. Currently, we found that after adjustment for relevant confounders risk of DR was increased with IGP increasing. Our results may imply that, even in case of well-controlled obesity, blood pressure, HbA1c, and renal function, the harmful effects of IPG on DR are still present. Future studies should investigate whether our ndings translate to other microvascular complications. If they are replicated in nephropathy, and neuropathy, it would further justify therapeutic interventions that speci cally target IGP.

Strengths And Limitations
In current study, a relative large sample size, standardized information collection, and a comprehensive mixed model analysis strengthened the reliability of our ndings. However, some limitations should be noted. Firstly, the random measurement error might be induced by IGP, which was calculated by the indices derived from the 3-h OGTT with moderate reproducibility. Secondly, GV was an integrated de nition by amplitude, frequency, and duration of the excursions, [25] but IGP merely emphasized the amplitude as a substitution of professional CGM. Thirdly, individuals this study only included Chinese hospitalized adults, which restricted the generalizability of the outcomes. Fourthly, a relatively large number of subjects were excluded due to uncompleted OGTT data. Fifth, the cross-sectional design renders us unable to rule out reverse causality.

Conclusions
In brief, the present study showed that 3-h OGTT-derived IGP was signi cantly correlated with the onset and progression of DR in in Chinese patients with type 2 diabetes. This nding suggested that glycemic control of DR patients should be additionally targeted at IPG to approach the normally glycemic condition. Future research is needed to elucidate the effect of IGP on DR and to what extent DR could be prevented by reducing IPG.