The Impact of Pregnancy on Diabetic Retinopathy: A Single-Site Study of Clinical Risk Factors

The doubling prevalence and incidence of diabetes over the past 2 decades is one of the most concerning global public health problems. Statistics in the USA are not better. In 2020, 26.9 million people of all ages in the USA had been diagnosed with diabetes, with 7.3 million individuals estimated to have undiagnosed diabetes; this number is expected to increase in the coming years [1]. When type 1 and 2 diabetes rates are pooled, African American populations have some of the highest rates, while non-Hispanic white populations have significantly lower rates [1]. Interestingly, when separated by type of diabetes, Caucasian individuals have the highest rates of type 1 diabetes while African American populations have the highest rates of type 2 [2]. Diabetes can cause complications in virtually every organ system; however, complications involving the eye, such as diabetic retinopathy (DR) and diabetic macular edema (DME), have significant and long-lasting impacts on patients’ lives. Nearly 12% of US adults diagnosed with diabetes report some form of vision disability [1] and a substantial burden on their functionality [3].

Certain populations, notably pregnant women, are more vulnerable to diabetes associated complications [4]. In a typical diabetic patient, DR is caused by hyperglycemia causing damage to retinal vessels [4]. In pregnancy, increased circulation forces may further potentiate damage, ischemia, and neovascularization in the delicate retinal vessels already at risk due to hyperglycemia [5, 6]. DR can be divided into two overarching categories: non-proliferative DR (NPDR) and proliferative DR (PDR) [4]. NPDR can be further split into three stages: mild, moderate, and severe. DME, a severe and debilitating complication of DR, can develop at any stage of DR and is characterized by macular edema – directly impacting central vision [4]. Regardless of available treatments, once the complications develop, vision frequently becomes permanently damaged. Consequently, the best way to mitigate the risk is to initiate tight glycemic and control other systemic risk factors.

The general screening guidelines for DR are well established. According to the American Academy of Ophthalmology, type I diabetics should undergo a retinal screening exam 5 years after the initial diagnosis, while type II diabetics should immediately undergo a retinal screening exam and follow yearly [7]. However, screening guidelines for vulnerable populations such as pregnant women are not as distinct and only suggest early and close examinations over the course of the pregnancy. New, more thorough counseling guidelines are needed to guide the management and counseling of modifiable risk factors (hypertension, treatment, smoking status, blood sugar control, etc.) that have been identified in this study and others [8‒10]. Achieving adequate control of modifiable risk factors can prevent potentiation of ocular complications as well as complications in the fetus [11]. Further studies are needed to determine the specific modifiable and non-modifiable risk factors affecting the development and progression of DR, DME, and worsening visual acuity (VA). This study aims to address this evidence-based gap by analyzing risk factors and clinical variables affecting the progression of DR, DME, and VA changes in pregnant patients.

This single-center study was done at Duke University Hospital. Institutional Review Board (IRB) approval was obtained, and a retrospective chart review was completed. Included patients were people with diabetes who entered prenatal care and had a delivery from January 2010 to December 2022. Maestro Care Slicer Dicer was used to identify 165 initial patients. Included patients who had a diagnosis of type 1 or 2 diabetes before pregnancy and patients with a diagnosis of gestational diabetes were excluded. Included patients had a recorded visit with a retinal exam before and after pregnancy (within 3 years) along with thorough documentation of risk factors and demographic information. Of the 165 patients initially identified, 22 were excluded due to diagnosis of diabetes after the completion of pregnancy, 20 patients were excluded due to lack of ophthalmologic documentation before pregnancy, and 20 patients were excluded due to lack of pregnancy documentation despite confirmed pregnancy. 53 patients were then identified as lost to follow-up (LTF) (did not have both a baseline and endpoint visit). Ultimately, 50 patients were used to analyze the progression and development of DR/DME and worsening of VA during pregnancy, while 53 patients were included in LTF analysis.

Variables of interest included age at pregnancy, race, type of diabetes, duration in years of diabetes at the time of pregnancy, gravida and parity numbers, number of comorbidities, HbA1c level pre- and postpregnancy, prepregnancy BMI, mean birth weight, gestational age, smoking status, preeclampsia development, hypertension treatment pre and during pregnancy, blood pressure measurements pre and during pregnancy, intravitreal injection treatment during pregnancy, pan-retinal photocoagulation (PRP) treatment received during pregnancy, dilation status during pregnancy, and number and class of fetal complications. Before, or prepregnancy, values were taken from documentation closest to the patient’s pregnancy within 3 years. After, or postpregnancy, values were taken from documentation closest to the end of the pregnancy, within 3 years. Every effort was made to draw data from the appointments closest to the pregnancy. Progression of DR was defined as change in NPDR stage (mild, moderate, severe) or a progression to PDR. Progression of DME was defined as development of DME or worsening of current DME-assessed by OCT and worsening vision. The number and type of fetal complications up to 12 months after birth were also recorded for all available patients.

Data were analyzed with Statistical Package for Social Sciences (SPSS). Patient demographics and baseline features were compared between those included in the study and those LTF. Patients were then analyzed by type of diabetes to assess for attributable differences in demographic and clinical variables between groups. Supplementally, patients in the LTF group that had at least two visits that did not qualify as baseline or endpoint during their pregnancy were split into “progression” and “no progression” groups and analyzed as well.

For the central analysis, patients were split into “progression” and “no progression” groups. Three primary outcomes were measured: development/progression of DR, development/progression of DME, and worsening of VA. Only 1 eye per patient was included in analysis; if a patient had DR or DME in both eyes, the more severe eye was included in analysis. Continuous variables were evaluated via independent samples t test while categorical variables were analyzed with χ² analysis. Statistically significant results for primary outcomes were further analyzed via binary logistic regression with simultaneous entry of all significant predictors.

A total of 50 patients were included in final analysis. Baseline characteristics by inclusion status can be found in Table 1. The mean age at pregnancy was 32.35 in included patients and 31.36 in patients LTF. Included patients had significantly higher rates of prepregnancy hypertension treatment (46.9%) when compared to LTF group (14.6%, p value <0.001). Included patients also had statistically significant higher rates of PRP treatment (16%, p value 0.002) and dilation during pregnancy (54%, p value 0.008).

Patients from the LTF group (n = 53) that had at least two visits over pregnancy (that did not qualify as a baseline or endpoint visit) were then split into DR “progression” and “no-progression” groups and analyzed (online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000533416). There were no statistically significant findings between the two groups.

Data were then compared for type 1 versus 2 diabetics (Table 2). The average age of type 1 diabetic patients (30.1 ± 5.8) was significantly lower than type 2 diabetics (33.7 ± 5.4, p value 0.004). The racial breakdown was also statistically significantly different between the groups. The type 1 diabetic group had 32/54 (59.3%) Caucasian, 19/54 (35.2%) African American, and 1/54 (1.9%) Pacific Islander. The type 2 diabetic group had 5/48 (10.4%) Caucasian, 28/48 (58.3%) African American, 10/48 (20.8%) Hispanic, and 4/48 (8.3%) Asian (p value <0.001). Duration of diabetes was significantly longer in type 1 diabetics (19 ± 8.0 years) than type 2 (13.9 ± 6.2 years, p value 0.002). The prepregnancy BMI was statistically significantly lower in type 1 (28.9 ± 6.2) when compared to type 2 (36.3 ± 8.1, p value <0.001). Hypertension treatment rates were significantly higher in type 2 diabetes both before pregnancy (42.6%) and during (44.7%) when compared to type 1. Rates of hypertension were significantly higher in type 2 (80%) when compared to type 1 (57.7; p value 0.019). Detailed tests and values are listed in Table 2 and demonstrate statistical differences between type 1 and 2 diabetes in age, race, duration of diabetes, prepregnancy BMI, and hypertension treatment and status after pregnancy.

Patients were split into “progression” and “non-progression” groups for each of the three primary outcomes: development/progression of DR, development/progression of macular edema (DME), and worsening of VA. Groups were then compared on background and clinical factors. The results are presented in Table 3. Findings show that HbA1c level before pregnancy and gestational age were different between groups for DR progression. Type of diabetes and prepregnancy hypertension treatment was statistically different between groups for DME progression. Intravitreal injection treatment during pregnancy was statistically different between groups for VA worsening. The average number of fetal complications was also statistically different between VA groups.

DR groups had similar ages and racial makeups between the “progression” and “non-progression” groups. The “progression” group had a slightly higher proportion of type 2 diabetics (55.6%) when compared to the “non-progression” group (47.8%), though this was not statistically significant. The “progression” group also had a statistically significantly higher HbA1c level at the beginning of pregnancy (9.9 ± 1.9) when compared to the “non-progression” group (8.5 ± 2.2; p value 0.028). The “progression” group also had a higher level at the end of pregnancy with a larger change over the pregnancy than the “progression” group, though this finding was not significant. The “progression” group also had higher prepregnancy BMI and higher mean birth weight along with increased rates of smokers and patients with preeclampsia, though these findings were not statistically significant. Gestational age was statistically significantly different in “progression” and “non-progression” groups (p value 0.028). Only 21.7% (5/23) of patients in the “non-progression” group had a full-term pregnancy while 51.9% (14/27) of patients in the “progression” group had a full-term pregnancy. The “progression” group had higher rates of intravitreal injections to treat DR during pregnancy with 14.8% (4/27) receiving either Eylea or Avastin while 1 patient in the “non-progression” received injections, though this was not significantly significant. Fetal complications were noted in both “progression” and “non-progression” groups; however, the average number or type of complications was not statistically different.

Macular edema “progression” and “non-progression” groups had similar ages. The DME “progression” group had a higher proportion of African American patients with a lower proportion of Caucasian patients when compared to “non-progression” groups, though this was not a significant finding. The “progression” group had higher prepregnancy BMI, higher rates of smoking, and preeclampsia, though these findings were not statistically significant. Of the patients in the “progression” group 100% (5/5) had type 2 diabetes, while only 30.9% (17/55) of patients in the “non-progression” group had type 2 diabetes (p value 0.029). The “progression” group’s HbA1c dropped by an average of 0.9 ± 1.0 over the pregnancy while the “non-progression” group’s HbA1c only dropped by an average of 0.7 ± 1.8 over the pregnancy, though this was not statistically significant. Hypertension treatment before the pregnancy was statistically significant between the groups with 100% (5/5) in the “progression” group and only 38.2% (21/55) in the “non-progression” group (p value 0.008).

VA in ‘worsening and “non-worsening” groups had similar ages and demographics. Rates of injections into the eye during pregnancy were significantly different between the groups with 22.2% (4/18) in the “progression” group and 3.6% (1/28) in the non-progression group (p value 0.047). The VA “progression” group had an average of 0.5 ± 0.76 fetal complications per patient while the “non-progression” group had 1.1 ± 1.7 complications (p value 0.04). Detailed tests and values are listed in Table 3.

All statistically significant predictors were then analyzed with logistic regression to determine associated risk for each variable. Results are shown in Table 4. When accounting for all other variables in the DR progression group, full-term pregnancies (OR: 3.43; CI: 0.908–12.94; p value: 0.069) and prepregnancy HbA1c level were not significant predictors of DR progression (OR: 1.3; CI: 0.97–1.83; p value: 0.075). When accounting for other variables, type of diabetes (OR: 5.40; CI: 0.515–55.78; p value: 0.160), change in HbA1c over pregnancy (OR: 0.74; CI: 0.447–1.234; p value 0.075), and prepregnancy hypertension treatment (OR: 3.4; CI: 0.447–25.88; p value: 0.237) were not found to be predictors of DME progression. Additionally, intravitreal injections during pregnancy were not a significant predictor of VA worsening (OR: 0.13; CI: 0.013–1.27, p value 0.080). Similarly, average number of fetal complications was not a significant predictor of VA progression (OR: 1.1; CI: 0.491–2.36; p value: 0.855).

To determine the effect of baseline retinopathy status on endpoint retinopathy status beyond “progression” or “non-progression,” rates of DR progression over the pregnancy were also measured via χ² analysis. Baseline and endpoint stage (none/NPDR/PDR) were collected for each patient. Progression rate based on baseline retinopathy status was found to be highly significant (p value <0.001). In patients that began their pregnancy with no DR, 23.8% had no progression (5/21), 66.7% (14/21) developed NPDR, and 9.5% (2/21) developed PDR. In patients that began the pregnancy with NPDR, 63.2% (12/19) had no progression, while 21.1% (4/19) had progression. 15.8% (3/19) patients that began pregnancy with NPDR had regression of stage during pregnancy. All patients that began their pregnancy with PDR remained in PDR at the end of the pregnancy (100%, 10/10).

Diabetic complications during pregnancy are often more severe than complications in nonpregnant diabetic patients. Pregnancy itself is a proven risk factor for complications for both mother and infant [12‒14] and has been detailed in previous studies [15] affecting almost every organ system [16‒18]. However, development and progression of severe ocular complications such as DR and macular edema have fewer studies, though recent literature has provided more detail on modifiable and non-modifiable risk factors [8]. These complications can lead to blindness [19] and severe changes in VA.

Continued analysis of risk factors is important and helps provide a clearer clinical picture to assist both clinicians and patients in lowering risk to both mother and fetus. Especially, important is seeing how well-known risk factors for development and progression of ocular complications apply to smaller populations and individual patients.

To ensure the validity and applicability of our results, we completed an analysis of baseline and demographic features of included patients versus patients LTF (Table 1). Though we sourced patients from a very large single center, it is surprising that only 50 patients had complete data. Additionally, many patients were LTF and did not have complete data. There may be many reasons patients are LTF; previous studies have found that demographic factors such as race and age can have an association with appointment no-shows in patients with chronic eye disease [20]. Demographics of North Carolina, the state of this medical center, show that Caucasians make up most of the state population at 70.1%, with the remainder made of 22.3% African American, 1.6% American Indian/Alaskan Native, and 3.4% Asian [21]. Acknowledging that socioeconomic factors can play a large role in patient compliance is important and can help physicians more accurately identify patients that are at an increased risk of becoming LTF. Interestingly, in our included patients and LTF, African American populations had the highest representation, differing from the state makeup. This further potentiates the importance of emphasizing appointment adherence to all patients at preconception visits. This study found that included patients had higher rates of hypertension treatment before pregnancy, PRP treatment during pregnancy, and dilation during pregnancy. Patients that were included in our analysis were likely already compliant before their pregnancy and because of increased visits had higher rates of medical intervention.

We also completed an analysis comparing data between type 1 and 2 diabetic patients (Table 2). In general, it is known that diabetes itself has a high disease burden and can contribute to development or potentiation of comorbidities, making it even more difficult to manage the root of DR and DME [22].

Most type 2 diabetics had progression of DR overall when all stages were pooled (15/27, 55.6%), while in type 1, a lower percentage of diabetics progressed (12/27. 44.4%); this was not a statistically significant finding. This same trend appeared to be true, and statistically significant for the progression of DME, 100% (5/5) type 2 diabetics had progression of DME, while no type 1 diabetics had DME progression. Ultimately, type of diabetes was not a significant predictor of DME progression after logistic regression analysis (Table 4).

The racial makeups of the type 1 diabetes and type 2 diabetes groups were significantly different – the type 1 patients were majority Caucasians while the type 2 group was mostly African American. The type 2 diabetic group also had a much higher prepregnancy BMI, falling into the obesity category. Obesity is well known as a risk factor for development of type 2 diabetes [23]. Blood pressure was significantly different between the two groups; many more of the type 2 diabetic patients had hypertension at the end of pregnancy when compared to type 1.

There seems to be little consensus as to the effect of race on development and progression of ocular complications. Previous studies have shown that certain ethnic groups can be at higher risk of DR and DME – African Americans, Hispanics, and south Asians have all been shown to have higher rates of DR than Caucasian populations [24, 25]. The Diabetes Control and Complications Trial (DCCT) was first to suggest a possible genetic component [26]; further studies have found the aldose reductase gene to be of specific significance with DR progression and development [27]. In our study, there were no significant differences in races between the “progression” and “no progression” groups.

Glycemic control is well known as a modifiable risk factor for DR and other ocular complications [28‒31]; rapid correction of blood glucose over a pregnancy can worsen DR [32, 33]. Glycosylated hemoglobin (HbA1c) level at the beginning of pregnancy was significantly different between the DR development/progression and no progression groups, giving further evidence to the importance of HbA1c control prepregnancy. This was not a statistically significant predictor when entered in multivariate analysis.

Higher BMI levels have been associated with higher risk of diabetic complications in both the general population and pregnant patients [28, 34]. In one animal study, epigenetic modifications causing increased damage to the retinal vasculature and increasing rate of DR progression were more common in diabetic animals fed a higher fat diet (leading to a higher BMI) than those with increased dietary restrictions [34, 35]. There is a direct relationship between maternal BMI and the fetus’s birth weight [36] and pregnant mothers that were overweight or obese were more likely to have macrocosmic neonates [36]. Our study found that the prepregnancy BMI of the mothers was higher in the DR, DME, and VA progression groups when compared to their no progression groups, though it was not statistically significant. Mean birth weight of the infants was higher in the DR, DME, and VA progression groups when compared to non-progressors, further suggesting that the patients with more severe diabetes were more likely to have higher birth weight infants. Generally, newborns from mothers with pregestational diabetes have increased rates of various congenital defects (such as limb deficiency, truncus arteriosus, and atrioventricular septal defect) [37, 38]. All groups analyzed in this study had a wide range of fetal complications; complications associated with prematurity were most common (retinopathy of prematurity, apnea of prematurity), followed by congenital heart defects (patent foramen ovale, atrial septal defect, ventricular septal defect), and hemorrhagic complications (subarachnoid and intraventricular hemorrhage).

The relationship between gestational age and development of ocular complications remains unclear. Our study found that a much higher percentage of patients with full-term pregnancies had development and progression of DR when compared to premature pregnancies. Though this did not reach statistical significance in multivariate analysis, a larger patient population likely would have yielded a significant result. This relationship could be in part due to increased duration of exposure to various growth factors and hormones known to be present in increased concentrations during pregnancy [39].

Smoking status, though it was not a significant finding in our study, could be linked to development and progression of ocular complications. Patients that were current or former smokers had a larger percentage of DR progression (6/27, 22.2%) compared to nonsmokers (3/23, 14.3%). Previous studies have associated smoking with increased levels of inflammatory interleukins and vascular endothelial growth factor (VEGF) production [40], lending a possible explanation for this mechanism of progression.

Hypertension and elevated blood pressure [8] are strongly associated risk factors for ocular complications during pregnancy. Our study found that a greater percentage of patients with preeclampsia had progression of DR (albeit not statistically significant). More patients that had received hypertension treatment before pregnancy had development/progression of DME when compared to those that did not have treatment. The percentage of patients that had hypertension at the end of pregnancy that developed/progressed in DME was significantly higher than the percentage of patients that were within normal blood pressure ranges; however, this was also not found to be a significant predictor of DME development/progression.

Treatment of diabetic ocular complications such as DR is multimodal and includes pharmacologic interventions such as anti-VEGF injections and laser therapy [41]. VEGF is widely regarded as the gold-standard target for treatment as it is mitigating angiogenesis seen in PDR [42]. Current treatments include various anti-VEGF molecules [42], and these treatments have proven highly efficacious in treating DR [43] and DME [44] while improving VA [45]. Regardless, they lack FDA approval for the treatment in pregnancy, due to possible adverse effects on the fetus [46]. Our study reports 5 patients overall receiving aflibercept (Eylea) or bevacizumab (Avastin) injection during pregnancy. These patients had no obvious deleterious events after administration; 4 patients had full-term births while one had a preterm birth at 28 weeks. All patients that received injections did so in the first or second trimester and 80% (4/5) had progression of DR during pregnancy and deterioration of their VA (Table 3). However, injection of anti-VEGF agents into the eye was not found to be a predictor of DR development and progression or VA deterioration. Eight patients underwent laser photocoagulation during pregnancy, and there were no significant differences in DR or DME development/progression or VA deterioration. We have included OCT and fundus photography for 1 patient that experienced progression (online suppl. Fig. 1). This patient began the pregnancy with DR in both eyes; both eyes experienced progression as noted in the figure. The patient had delivery at 31 weeks.

The effect of mydriatic agents during pregnancy (Pregnancy Category C from the FDA) is not sufficiently studied, and it is not recommended unless medically necessary [7]. In our study, mydriatics were used on 27 patients without any adverse events.

In conclusion, this study done on a diverse patient population draws attention to many risk factors of DR progression in pregnancy. Type of diabetes, level of HbA1c at beginning of pregnancy, gestational age, hypertension treatment prepregnancy, blood pressure postpregnancy, and intravitreal injections during pregnancy were all significantly different between the progression and no progression groups. This study provides a unique assessment by analyzing variables affecting DR, DME, and VA in pregnant diabetic patients. These findings might help establishing new standard of care for pregnant patients with diabetes, including recommending thorough counseling and screenings before initiating pregnancy and encouraging close follow-up by their entire care team through their pregnancy and after. There were shockingly few pregnant diabetic patients with complete data at our center; many patients were LTF. For a tertiary center that is often at the forefront of both complex and routine patient care, this is an unexpected finding.

Due to the single center, the data came from our study were limited by the small sample size. It is likely that this affected the significance of our multivariate analysis. The small sample size may have influenced the power of our study and findings-warranting additional research in the future (e.g., IRIS registry research). Future studies should be directed at the larger patient populations to elucidate the risk factors further and analyze long-term complications while using these findings as a guide.

No authors other than those listed here have contributed to authorship or development of this manuscript.

All planning, conduct, and reporting of human research were in line with COPE guidelines. All data were de-identified and are not identifiable. The need for informed consent was waived by the Duke University IRB. Consent is not required for this study in accordance with local or national guidelines. This study protocol was reviewed and approved by Duke University IRB on 6/10/21 (ID: 00108408). All aspects of human research followed the Helsinki Declaration and ICMJE recommendations.

No authors have conflict of interest to declare. There are no nonfinancial relationships that may have influenced the writing of this manuscript.

There was no funding of any research relevant to this study or any sponsors. There were no sponsors involved in the preparation of data or the manuscript.

M.H. and D.A. contributed to the design and idea of the project. J.R. retrieved patient information from chart review. S.S. and B.B. conducted statistical analysis of data. M.H., S.S., and J.R. all contributed to writing of the manuscript. All authors contributed to the manuscript revision, read, and approved the submitted version. All authors agree to be accountable for the content of the work.

All data generated or analyzed during this study are included in this article and its supplementary material files. Further inquiries can be directed to the corresponding author. This is primary data.

Jay Rathinavelu and Swara M. Sarvepalli contributed equally to this work.