Diabetic retinopathy (DR) is the primary cause of visual impairment and blindness among working-age individuals in developed countries (1). Two important factors affecting the vision of DR patients are complications associated with proliferative diabetic retinopathy (PDR) and diabetic macular edema (DME), which are thought to occur as a result of vascular endothelial growth factor (VEGF) and other cytokines into the vitreous cavity (2, 3). PDR, the most advanced stage of DR, is characterized by neovascularization and proliferative membrane formation, which may cause vitreous hemorrhage and tractional retinal detachment (4–7). Those patients with PDR often require pars plana vitrectomy (PPV) treatment and emergency management to prevent further vision loss. Although DME can occur at any DR stage, the prevalence of DME was associated with diabetes mellitus (DM) duration and DR severity. PDR and DME frequently occur together and the prevalence of DME in PDR patients varied from 30% to 72.6% (8, 9). The risk factors for DME have been widely studied, including duration of diabetes (10), hypertension (11, 12), glycosylated hemoglobin (HbA1c) level (13–15), insulin (15–19), and other factors (11).

There were divergent findings of the association between different hypoglycemic regimens and DME. Several independent clinical studies (14–16, 18–21) and meta-analysis (17) confirmed that insulin use increased the risks of DME in patients with type 2 diabetes mellitus (T2DM) compared with oral hypoglycemic agents. What’s more, this finding was supported by another experimental research that insulin increased retinal vascular permeability in diabetic mice (22). While several studies pointed to a potential effect of insulin on reducing risks of DME (23, 24). The stable glycemic control induced by insulin might be one possible explanation. Insulin itself has a weak to moderate stimulatory effect on the proliferation of human retinal pigment epithelium (RPE) cells and could promote RPE wound healing (25), which might be another possible explanation. Apart from insulin, different oral hypoglycemic agents may have different effects on DME. Thiazolidinediones (TZDs) including pioglitazone and rosiglitazone, are insulin-sensitizing medications that can be used for glycemic control in T2DM. A growing body of studies has highlighted that TZDs might contribute to increase risks of DME (26–29). In addition to the aforementioned, the association between sulphonylureas, metformin, and the risk of DME has also been investigated (14, 30).

Treatments for DME have gradually evolved from the initial grid macular coagulations to intravitreal injection including triamcinolone, Ozurdex, and anti-VEGF agents. In addition to the aforementioned, PPV surgery is proven to be associated with structural benefits compared with the natural history of DME (9, 31, 32). PPV treatment in PDR aims at removing vitreoretinal traction and vitreous hemorrhage, clearing various cytokines from the vitreous cavity, reattaching detached neuroretina, maintaining media transparency, and improving ocular circulation (7, 33). Even if patients with PDR have undergone PPV treatment, some patients still develop postoperative DME. How to make those patients with postoperative DME preserve their vision and improve their quality of life is a problem that needs to be solved urgently. In addition to investigating the specific treatments of postoperative DME, more attention we should pay to the risk factors of postoperative DME. However, first-hand clinical data regarding the role of different hypoglycemic regimens in postoperative DME is scarce. In this study, we performed a secondary data analysis based on existing data that comes from the published paper (34) to investigate the association between different hypoglycemic regimens and postoperative DME.

This study demonstrated that compared with oral hypoglycemic medication, only insulin treatment had an association with the reduced risk of postoperative DME in all adjusted models. These stable and consistent results indicated that long-term insulin therapy seems to be beneficial in reducing the risk of postoperative DME. While the association between insulin therapy and reduced risk of postoperative DME seems to have nothing to do with HbA1c. In the stratified analysis, we can see that between different stratifications (age, gender, different HbA1c, gas tamponade, hypertension, posterior vitreous detachment, reoperation, MIVS, postoperative complications), oral medication plus insulin may exacerbate DME, although there is no statistical difference, the direction of the effect value is biased towards exacerbating DME. It may be necessary to increase the sample size to further verify whether the results are stable between different stratifications and whether the stratification will lead to different results.

Most previous studies (14–21) have confirmed that insulin might contribute to a higher risk of DME, which is different from the results of postoperative DME in this study. We think that there are the following reasons: Firstly, differences in the study population may lead to such inconsistent results. The population of previous studies included patients with DME from all stages of DR (14, 18, 19, 30). In this study, the population is all patients with PDR who have undergone PPV treatment, and their DR stages are comparable. We believe that it is more reasonable to study the association between different hypoglycemic regimens and DME under the unified classification of diabetes and the DR stage. Secondly, compared with DME without PPV surgery, postoperative DME has been relieved within a certain period by removing vitreoretinal traction and clearing various cytokines including VEGF which had a synergistic effect with insulin for higher risks of DR development in patients using insulin (32). Thirdly, insulin therapy is generally advocated when oral medications are insufficient in controlling glycemic. In the early stage of insulin treatment, there may be a series of edema reactions in the body due to drug switching or glycemic control (20, 35). While chronic insulin therapy, compared with oral hypoglycemic agents, does not modify the anatomic or functional effectiveness of DME treatment (23, 36, 37). In this study, the PDR patients treated with insulin had a longer course of diabetes. Long-term insulin therapy might reduce the risk of postoperative DME by lowering glycemia and HbA1c persistently which is termed metabolic memory (29). Metabolic memory was shown to persist through 10 years of follow-up and intensive glycemic control could reduce the risks of clinically significant DME in type 1 DM patients (29). In this study, due to metabolic memory, the advantage of strict glycemic control in the onset of postoperative DME might be reflected. At the same time, we found that although there is no statistical difference, HbA1c was lower in patients with oral medication than in patients with oral medication plus insulin and insulin. Apart from HbA1c, there may be other mechanisms to explain this phenomenon.

Except for insulin, different oral hypoglycemic medications have different effects on DME (27, 28). Most previous studies have shown that TZDs contributed to the increased risk of DME (26–28, 38). While another study (39) demonstrated that TZDs do not cause subclinical DME in a demographically diverse T2DM population, whether the TZDs are combined with other agents or not. Therefore, the relationship between TZDs and DME is still inconsistent. In addition to TZDs, sulphonylureas treatment was also related to a high risk of DME (14), while metformin did not have a relationship with the occurrence of clinically significant DME (30). Taken together, because different hypoglycemic medications may have different effects on DME and their mechanisms on DME are not fully understood, it is very important to clarify the proportion of different hypoglycemic mechanisms in hypoglycemic medications. Poor glycemic control applied by oral hypoglycemic medications might contribute to a higher risk of DME.

PPV itself has been widely used to treat tractional or refractory DME (40–43), which relieves DME through multiple mechanisms, including the elimination of traction factors (44), improving intravitreal oxygenation, removing pathological cytokines (such as VEGF) in the vitreous cavity, and accelerating the half-life of intravitreal cytokines (45). Whereas the low incidence of postoperative DME after PPV, there are few studies on the risk factors for postoperative DME in patients with PDR. A study (46) revealed that central macular thickness of the vitrectomized eyes was significantly correlated with atherogenic index of plasma, total cholesterol, low-density lipoprotein cholesterol, and uric acid. Kojima et al. (47) found that in DME patients, preoperative low HbA1c and postoperative pseudophakia were independently associated with the decrease in foveal thickness of the vitrectomized eyes. The two studies aforementioned confirmed that systemic factors might play an important role in the pathogenesis of postoperative DME. Based on clinical experience and observation, we believed that early postoperative DME may be related to surgical procedures (48, 49) and inflammatory reactions related to surgical procedures (49). In the late stage of postoperative DME may be related to systemic factors such as preoperative HbA1c (47), and the ocular accumulation of cytokines related to systemic factors may be another explanation. Therefore, we speculated that glucose control as one of the systemic factors may attribute to postoperative DME most. In the present secondary analysis, the association between the different regimens for glucose control and postoperative DME is circumstantial evidence for our hypotheses. We will further validate the hypotheses in our prospective PDR cohort (50).

Our study has some strengths. First, we treated different hypoglycemic regimens as a categorical variable and tested the P for trend, which is useful in evaluating the robustness of data analysis. Second, this was a historical cohort study and was thus susceptible to potential confounding. However, we used strict statistical adjustment to minimize the effect of residual confounders. Last, the results of present study provide a new idea of the relationship between DME secondary to vitrectomy with different hypoglycemic drugs, which is worthy of further exploration in future research.

However, the current study has several limitations. First, because of the retrospective cohort design, the proportion of different hypoglycemic mechanisms in oral DM medications and other drugs which might reduce the risks of DME such as aspirin and angiotensin-converting enzyme inhibitor is not clear. A prospective cohort study aiming at the different hypoglycemic regimens on postoperative DME is required. Second, because it is a retrospective cohort study, whether the presence of preoperative DME secondary to media opacification, and the definition and macular thickness of postoperative DME is unclear. The incidence of postoperative DME may be underestimated. It is noteworthy that the potential lack of information on outcomes resulting from such errors would bias toward the null and thus results in an underestimation of the association between different hypoglycemic regimens and postoperative DME. Third, the specific measurement time of postoperative DME is not known, and the pathogenesis of DME at different time points may be different. Besides, the period of oral medication or insulin that was prescribed for glycemic control should be considered. Finally, the generalization of the results may be limited, as the study population is all Japanese and the study had limited power to identify statistically significant differences in different stratifications, because of the small number of subjects included.

In conclusion, our results of the secondary analysis should be interpreted as a significant association between insulin treatment and reduced risks of DME in Japanese PDR patients with PPV surgery, compared with oral medications. Well glycemic control with longstanding insulin therapy may be beneficial to reduce the risks of postoperative DME in PDR patients with T2DM. Our investigation calls for large-scale and long-term prospective clinical studies for a full evaluation of the exact role of insulin in the progression of postoperative DME.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material .

This study was performed by Japanese researcher Nishi et al. at the Yamagata University Hospital, Yamagata, Japan. In the previously published article [34], Nishi et al. have clearly stated that: the study was performed following the Declaration of Helsinki and reviewed and approved by the Ethics Committee of the Yamagata University Faculty of Medicine (approval number: H26-21). All data were fully anonymized before we accessed them and the IRB waived the requirement for informed consent. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Conceptualization:CL, YZ, MZ; Formal analysis: CL; Validation: YZ; Supervision: MZ; Writing – original draft: CL, YZ, MZ; Writing – review & editing: CL, YZ, MZ. All authors contributed to the article and approved the submitted version.

This work was supported by 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (ZYJC21025), and the Sichuan Provincial Science and Technology Support Project (no.2019YJ0129), and China Postdoctoral Science Foundation (2021M692273).

Thanks for the comprehensive raw data provided by professor Koichi Nishitsuka’s team. They completed the entire study. They are (the rankings and institutions of these researchers were ranked according to the “reference (34)“) Katsuhiro Nishi, Koichi Nishitsuka (corresponding author) (Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine, Yamagata, Yamagata, Japan), Teiko Yamamoto, Hidetoshi Yamashita.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.