PPCV distribution showed high reproducibility between ophthalmologists with an ICC of 0.937. The analysis from one ophthalmologist was repeated twice to determine repeatability (κ = 0.986).

Patient characteristics

The study included 101 healthy subjects and 140 type 2 diabetic mellitus patients with DR. The demographics and baseline characteristics of the two groups are shown in Table 1.

Table 1 Demographics of healthy and diabetic subjectsDistribution of PPCV in normal subjects

The subjects were divided into four age groups (18–30, 31–50, 51–70, and > 70 years old; Fig. 3a1, a2). When the PPCV distribution was compared among different quadrants in the same age group (Fig. 3a1), the distributions in the temporal inferior and superior quadrants were significantly higher than those at the nasal superior and inferior quadrants in all four age groups (all P < 0.05, SNK test). In addition, when PPCV distribution was compared among different age groups in the same quadrant (Fig. 3a1), PPCV distribution in the over-70 age group was significantly lower than that in any of the other three age groups, and the 51–70 age group’s value was significantly lower than those of the 18–30 and 31–50 age groups (all P < 0.05, SNK test).

Fig. 3

Distribution of PPCV in normal subjects and severe nonproliferative DR cases. Comparison of PPCV distribution in normal subjects among different age groups (a1) and quadrants (a2), and in severe nonproliferative DR cases among different age groups (b1) and quadrants (b2). *P < 0.05 (SNK test). a1 There was no significant difference in PPCV distribution between the 18–30 and 31–50 years age groups in any of the four quadrants (all P > 0.05, SNK test). a2 There were no significant PPCV distribution differences between the temporal superior and inferior quadrants or between the nasal superior and inferior quadrants in any age group (all P > 0.05, SNK test). b2 There were no significant differences in PPCV distribution between any other quadrants in the 31–50 and 51–70 age groups (all P > 0.05, SNK test), and there were no significant differences in PPCV distribution between the four quadrants in subjects over 70 years (all P < 0.05, SNK test). c Scatterplot of PPCV distribution vs. age in normal subjects showing a best-fit simple regression line. The overall fit is significant (P < 0.001). PPCV, peripapillary capillary volume; DR, diabetic retinopathy; y, years

Multiple linear regression analysis of factors concerning PPCV distribution in normal subjects

The regression variables were age, gender, and quadrant location of PPCV (temporal and nasal hemisphere). Gender and quadrant location of PPCV were regarded as binary variables. Since, as mentioned above, there was no significant difference in PPCV distribution in any age groups between the superior and inferior quadrants in either the temporal or nasal region, PPCV distributions in the temporal superior-inferior quadrants were regarded as variables of quadrant locations of PPCV at the temporal hemisphere, while PPCV distributions in the nasal superior-inferior quadrants were regarded as variables of quadrant locations of PPCV at the nasal hemisphere. The variance inflation factors (VIF) for gender, age, and quadrant location of PPCV were all 1.000, and no evidence of multicollinearity was found. A scatter plot (Fig. 3c) showed a linear correlation between PPCV distribution and age. In multiple regression analysis, age and quadrant location of PPCV were shown to be significantly correlated with PPCV distribution (β =  − 20790.03, and β =  − 1.129 × 106, respectively, all P < 0.001). However, the effect of gender (P = 0.855) was negligible from a statistical point of view.

Distribution of PPCV in severe nonproliferative DR cases

In consideration of the small number of youth (younger than 30 years) with type 2 diabetes, the subjects were divided into three age groups (31–50, 51–70, > 70 years) (Figs. 2b2 and 3b1). When the PPCV distribution was compared among different quadrants in the same age group (Fig. 3b2), the PPCV distribution at the temporal inferior quadrant was significantly higher than those at the nasal superior and nasal inferior quadrants (all P < 0.05, SNK test). In addition, when PPCV distribution was compared among different age groups in the same quadrant (Fig. 3b1), it was significantly higher in the 31–50 age group than in the other two age groups. In the over-70 age group, it was significantly lower than in the 51–70 age group in all four quadrants (all P < 0.05, SNK test). PPCV distributions were compared between severe nonproliferative DR cases and normal subjects (Fig. 4). In all age groups, the PPCV distribution in each quadrant was significantly lower in severe nonproliferative DR patients than in normal subjects (all P < 0.05, t-test).

Fig. 4

Comparison of PPCV distribution between severe nonproliferative DR cases and normal subjects. Groups were compared using generalized estimating equation to correct for intereye associations. PPCV, peripapillary capillary volume; DR, diabetic retinopathy; TS, temporal superior; TI, temporal inferior; NS, nasal superior; NI, nasal inferior; y, years; *P < 0.05 (t-test)

Changes in PPCV distribution in DR patients before and after treatment

The change in PPCV distribution over time after DR treatment is illustrated in Fig. 5. In the anti-VEGF treatment group, PPCV distribution at weeks 1 and 2 was significantly reduced compared with baseline (all P < 0.001, LSD test). Pairwise comparison among the post-treatment groups showed that PPCV distribution at weeks 1 and 2 were significantly lower than that at months 1, 2, and 3 (all P < 0.05, SNK test). In the PRP treatment group, PPCV distribution at months 2 and 3 was significantly increased compared with baseline (all P < 0.001, LSD test). Pairwise comparison among the post-treatment groups showed that PPCV distribution at weeks 1 and 2 and month 1 were significantly lower than at months 2 and 3 (all P < 0.05, SNK test). In the surgery treatment group, the PPCV distribution at week 1 was significantly lower than baseline (P < 0.001, LSD test), and at month 3, it was significantly higher than baseline (P < 0.001, LSD test). Pairwise comparison post-treatment showed that the PPCV distribution at week 1 was significantly lower than that at week 2 and at months 1, 2, and 3 (all P < 0.05, SNK test), while at month 3, it was significantly higher than at week 2 and months 1 and 2 (all P < 0.05, SNK test).

Fig. 5

Post-treatment change in PPCV distribution in DR patients. Patients came for clinical follow-up at 1 and 2 weeks and 1, 2, and 3 months post-treatment. *P < 0.05 (LSD test and SNK test). PRP, pan-retinal photocoagulation; DR, diabetic retinopathy; PPCV, peripapillary capillary volume

In the anti-VEGF treatment group (Fig. 6a), no correlation was observed in the variation in PPCV distribution and the variation in LogMAR BCVA before and 3 months after treatment (P = 0.940). A scatter plot (Fig. 6b, c) showed that the variation in PPCV distribution was negatively and significantly correlated with the variation in LogMAR BCVA before and 3 months after PRP or surgery (β =  − 9.67 × 105, and β =  − 6.83 × 105, respectively, all P < 0.003).

Fig. 6

Scatterplots and linear regression analyses of variation in PPCV vs. variation in LogMAR BCVA before and 3 months after anti-VEGF (a), PRP (b), or surgery (c). BCVA, best-corrected visual acuity; PPCV, peripapillary capillary volume; PRP, pan-retinal photocoagulation; anti-VEGF, anti-vascular endothelial growth factor