Hemoglobin mediates the link between the ‘weekend warrior’ physical activity pattern and diabetic retinopathy

Table of Contents

Characteristics of the participants

After strict screening, a total of 5102 participants with DM but not DR and 1043 participants with DR were included in this study. The median age was 62.00 years in the DR-free group and 63.00 years in the DR group. The median Hb concentration was 13.90 g/dL in the DR-free group and 13.40 g/dL in the DR group. The percentage of inactive participants was clearly greater in the DR group than in the DR-free group. Compared with the participants in the DR-free group, those in the DR group presented lower education levels, lower household income, higher alcohol consumption, and higher prevalence rates of hypertension and hyperlipidaemia. All baseline profile data are presented in Table 1.

Table 1 Basic participant characteristics.

Association between PA pattern and DR risk

Logistic regression analysis of PA pattern and DR risk was performed using inactive participants as a reference. After correcting for all the covariates, no significant association was demonstrated between the insufficiently active PA pattern and DR (Model 3: OR = 0.849, 95% CI = 0.690–1.045, P = 0.122), and no significant association was found between the WW (Model 3: OR = 0.631, 95% CI = 0.487–0.818, P < 0.001) and RA (Model 3: OR = 0.738, 95% CI = 0.554–0.949, P = 0.018) PA patterns, which are significant protective factors against DR (Table 2).

Table 2 Association between PA and DR.

To more clearly compare the effects of the RA and WW PA patterns on DR risk, further logistic regression analysis of PA pattern and DR risk was performed using RA participants as a reference. After correcting for all the covariates, no significant association was demonstrated between the insufficiently active and WW PA patterns and DR (insufficiently active Model 3: OR = 1.150, 95% CI = 0.852–1.553, P = 0.362; WW Model 3: OR = 0.855, 95% CI = 0.611–1.197, P = 0.362); the inactive PA pattern (Model 3: OR = 1.355, 95% CI = 1.054–1.742, P = 0.018) was found to be a significant risk factor for DR (Table 2).

Multicollinearity testing and subgroup analysis

Table S1 shows the results of a multicollinearity test between variables. The variance inflation factor (VIF) values of all the variables ranged from 1.09 to 1.80, with a mean VIF = 1.25, and it can be concluded that there was no multicollinearity between variables. Table S2 shows the results of the multicollinearity test between variables when using as subgroup variables age, race, sex, education level, marital status, drinking status, smoking status, hypertension status, hyperlipidaemia status, family PIR and BMI for further stratified analysis of the association between PA pattern and DR risk. After correcting for variables other than subgroup variables, we observed that the association between the RA PA pattern and DR risk was not affected by covariate interactions (P for interaction > 0.05).

Association between hemoglobin level and DR risk

All models found that a lower total Hb level was a significant risk factor for DR (Model 1: OR = 0.854, OR = 0.820–0.890, P < 0.001; Model 2: OR = 0.829, 95% CI = 0.792–0.869, P < 0.001; Model 3: OR = 0.835, 95% CI = 0.797–0.875, P < 0.001). Hb levels were divided into quartiles to investigate the associations between different Hb levels and DR risk. In Model 3, which was corrected for all covariates, there was a negative association between the Q2, Q3, and Q4 quartiles of Hb level and the prevalence of DR compared to the Q1 quartile (Q2 Model 3: OR = 0.767, 95% CI = 0.639–0.922, P = 0.005; Q3 Model 3: OR = 0.599, 95% CI = 0.492–0.729, P < 0.001; Q4 Model 3: OR = 0.488, 95% CI = 0.390–0.611, P < 0.001) (Table 3).

Table 3 Association between hemoglobin level and DR.

Association between PA pattern and hemoglobin level

Linear regression analysis of PA pattern and Hb level was performed using inactive participants as a reference. After correcting for all the covariates, no significant association was demonstrated between the insufficiently active and RA PA patterns and Hb level (insufficiently active Model 3: OR = 0.130, 95% CI=−0.025-0.236, P = 0.055; RA Model 3: OR = 0.095, 95% CI=−0.026-0.216, P = 0.123). A significant positive correlation was detected between the WW PA pattern and Hb level (OR = 0.178, 95% CI = 0.059–0.298, P = 0.004) (Table 4).

Table 4 Association between PA and hemoglobin level.

Mediation analysis

In the mediation model, PA pattern was the predictor, hemoglobin level was the mediator, and DR was the outcome factor. Bootstrap tests of the mediation effects revealed that the total (c=−0.017, SE=−3.542, 95% CI: −0.027 to −0.008, P < 0.001), the direct (c′=−0.016, SE=−3.346, 95% CI: −0.026 to −0.007, P = 0.001), and the indirect (a×b= −0.001, SE=−0.001, 95% CI: −0.033 to −0.000, P = 0.042) effects were significant. Suggesting that hemoglobin levels play a partial mediating role between PA and DR (Table 5; Fig. 2). The results of the Sobel–Goodman mediation test also supported the existence of a mediating effect (coefficient=−0.001, SE = 0.001, P = 0.044), with a mediation ratio of 5.95%.