Progression of Diabetic Retinopathy and Predictors of Its Development and Progression During Pregnancy in Patients With Type 1 Diabetes: A Report of 499 Pregnancies

Diabetic retinopathy (DR) is a major microvascular complication of diabetes mellitus and is often the first complication to appear. Pathophysiology involves retinal vascular lesions. DR is responsible for 2.6% of blindness cases worldwide (1). Type 1 diabetes is usually diagnosed in youth, and DR incidence has been reported to be as high as 30% after 5 years among all types of diabetes (2).

Pregnancy is known to interfere with many diseases, including DR, and has been associated with increased DR development and progression (3,4). Reported rates of DR progression range from 5 to 42% (5,6).

The main predictors of the progression of DR during pregnancy include diabetes duration (7,8), baseline severity of DR (5,9), poor glycemic control at the time of conception (10,11), high baseline HbA_1c level and marked HbA_1c reduction (3), and hypertension and preeclampsia (12,13).

Recently, the French Diabetes Society and the French Ophthalmology Society proposed new guidelines for DR management during pregnancy (2). They recommend that pregnancy preparation include DR screening. Severe nonproliferative DR (NPDR) and proliferative retinopathy usually require laser treatment before pregnancy. Guidelines also recommend ophthalmologic follow-up at least every trimester or every month in case of prepregnancy DR, arterial hypertension, long duration of diabetes, unplanned pregnancy, or rapid HbA_1c decrease during the first trimester, independent of the level of prepregnancy HbA_1c.

The aim of our study was to assess the progression rate (including development and worsening) of DR during pregnancy in a large cohort of women with type 1 diabetes followed in a single center by a multidisciplinary team and to compare our practice with the new French guidelines for women with type 1 diabetes during pregnancy.

This retrospective study included all consecutive pregnancies beyond 24 weeks in women with type 1 diabetes who delivered in Lille Hospital, Lille, France, between 1997 and 2015. The exclusion criteria were abortion (spontaneous or induced) and all other types of diabetes mellitus (including type 2 diabetes mellitus, maturity-onset diabetes of the young, and syndromic and secondary diabetes).

Maternal demographics, obstetric and ophthalmologic examinations, time of diabetes diagnosis, and presence of complications were collected in the patients’ records.

Diabetes and obstetric follow-up were performed monthly, and each patient was called twice a week by a specialized nurse to assess glucose control and adjust the insulin dose if needed. Most patients were treated with short-acting insulin analogs (aspart or lispro) before meals and long-acting insulin analogs (detemir) in the morning and/or at bedtime or with continuous subcutaneous insulin infusion (CSII). All patients performed self-monitoring of blood glucose, with blood glucose targets of <100 mg/dL before meals and <140 mg/dL after meals. HbA_1c was measured monthly using automated high-pressure liquid chromatography (normal range 4.0–6.0%, 20–42 mmol/mol).

Age, prepregnancy BMI, type of treatment (multiple daily injections [MDIs] or CSII), previous pregnancies, history of arterial hypertension (defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or treatment with antihypertensive drug), nephropathy (defined as albuminuria ≥30 mg/24 h or renal failure), and neuropathy were recorded, as were history and status of DR at last fundus examination before pregnancy and last prepregnancy HbA_1c measurement.

HbA_1c values for the first, second, and third trimesters were expressed as the mean of all HbA_1c tests during the given period. To assess the variation of HbA_1c values during pregnancy, we used the mean difference between HbA_1c value at each trimester and prepregnancy HbA_1c value.

Pregnancy-induced hypertension was defined as the development of hypertension after 20 weeks of amenorrhea. Preeclampsia was defined by the combination of systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg and proteinuria ≥300 mg/24 h after 20 weeks of amenorrhea.

Each of the participants in this study had one or more ophthalmologic examinations before the incident pregnancy. DR was assessed every 3 months by fundus examination or photography or every month when severe retinopathy was present or when DR worsening occurred. Examination and follow-up were performed by a single ophthalmologist (H.C.) who followed all patients during their pregnancies for the study period.

DR severity was divided into seven stages according to the simplified Early Treatment Diabetic Retinopathy Study (ETDRS) classification (14): absence of retinopathy, nonproliferative stage 1 (mild), nonproliferative stage 2 (moderate), nonproliferative stage 3 (severe), proliferative, laser treated stable, and laser treated with progression. Because there is no further classification of laser-treated DR in the simplified ETDRS classification, we separated stable laser-treated DR and laser-treated DR with progression.

Development of DR was defined as a DR diagnosis during pregnancy in a patient who had no DR before conception. Worsening of DR was defined as deterioration of at least one DR stage in a patient who already had DR before conception. Progression of DR was defined as either development or worsening of DR during pregnancy.

Two categories of risk predictors were studied: first, those preceding the pregnancy (age, BMI, nulliparity, duration of diabetes, CSII treatment, arterial hypertension, diabetic nephropathy, diabetic neuropathy, diabetic retinopathy, laser photocoagulation, prepregnancy HbA_1c), and second, decrease in HbA_1c during pregnancy.

These data were collected as part of usual management. All patients were informed that the results would be anonymized and used for statistical purposes.

Data are presented as number (percentage) for categorical variables and mean ± SD for continuous variables. Normality of distribution was checked graphically and by using the Shapiro–Wilk test. Predictors of DR progression during pregnancy were firstly determined in bivariate analyses using mixed logistic regression models (distribution binomial, log link function), with patients as a random effect, to include patients with multiple pregnancies during the study period. Continuous prognostic factors were analyzed as such (as linear terms) and after categorization by quartiles to explore the shape of the relationship. Because we found no evidence of non–log-linear relationships whatever the continuous variables (data not shown), we reported only estimates for continuous variables with no categorization. All possible first-order interactions with prepregnancy DR were further investigated by including the corresponding interaction term in the mixed logistic regression models. Factors and interactions with values of P < 0.10 in previous analyses were entered into backward stepwise mixed logistic regression analysis, with a value of P < 0.10 used as the cutoff for retention in the model.

Changes in HbA_1c over pregnancy from the prepregnancy value were compared according to DR progression using a linear mixed model for repeated measures (an unstructured covariance pattern model to account for the correlation between trimester measures during the same pregnancy); in this model, prepregnancy HbA_1c, trimester of measure, DR progression, and trimester of DR progression interaction were introduced as fixed effects and patients as a random effect (to account for the patients with multiple pregnancies); post hoc comparisons were performed using linear contrast, and the least squares (adjusted) means were derived from linear mixed models. Normality and homoscedasticity of the residuals were checked graphically. Data were analyzed using SAS software package, release 9.4 (SAS Institute, Cary, NC).

We identified 508 pregnancies in women with type 1 diabetes occurring during the study period. We excluded nine pregnancies because data on prepregnancy DR status were not present. Ultimately, 499 pregnancies among 375 patients were included in our study. Mean age at pregnancy was 29.7 ± 4.8 years, and 50.1% of patients were nulliparous (Table 1). Mean duration of type 1 diabetes was 13.6 ± 8.1 years. Before conception, prevalence of DR was 30.3% (n = 151 of pregnancies), and diabetic nephropathy was present in 7.6%, diabetic neuropathy in 2.2%, and arterial hypertension in 3.8%. Mean prepregnancy HbA_1c was 7.6 ± 1.6%. During pregnancy, 53.1% of the women were treated with CSII; the remaining patients received MDIs. Rates of pregnancy-induced hypertension and preeclampsia were 11.3% and 6.3%, respectively.

Of the cases of prepregnancy DR occurring among 151 pregnancies, most were stage 1 DR (n = 103; 68.2%) or photocoagulated retinopathy (n = 31). There were only a few pregnancies with stage 2 NPDR (n = 14). Stage 3 DR and proliferative DR (PDR) were rare, at 2 and 1 case, respectively (Table 2).

Of the 499 pregnancies in women with type 1 diabetes, 236 (47.2%) were complicated by DR. Retinopathy was identified before pregnancy in 151 (64.0%) of these 236 pregnancies, and retinopathy was first recognized during pregnancy in 85 (36.0%). Progression in DR stage during pregnancy occurred in 24 (15.9%) of the 151 cases of prepregnancy DR. In all 85 pregnancies in which DR was first found during pregnancy, progression by at least one DR stage occurred during pregnancy. Most of the progressions occurred during the first and second trimesters (11.0% and 9.2% of pregnancies, respectively), whereas few were observed during the third trimester (3.8%) (Table 2). In addition, third-trimester DR progression was observed only in pregnancies in which progression had already started during the first or second trimester.

Of the 85 cases of DR first recognized during pregnancy, 76 ended with mild DR (89.4% of the developments; 69.7% of the progressions), only 1 became severe, and 1 progressed to PDR during pregnancy.

Of pregnancies in which DR worsened, 17 (i.e., 15.6% of progressions) progressed only one stage, 12 of them (11% of progressions) from stage 1 to stage 2 (Table 2). Progression was seen in only two pregnancies with prepregnancy pan-retinal photocoagulated DR (8.3% of the cases with worsening DR [i.e., 1.8% of the cases with progression]). Among the 109 DR progressions, 93 (85.3%) progressed by only one stage. Among the 16 pregnancies with a change of two or more stages during pregnancy, 9 occurred in the 85 pregnancies in which DR was first recognized during pregnancy and 7 in the 151 pregnancies in which DR antedated pregnancy. Only four (3.7%) had to be treated with pan-retinal photocoagulation during pregnancy.

We had ophthalmologic data during the first postpartum year for 270 pregnancies, showing DR progression in only 11 (4.1%) of the pregnancies and regression in 25 (9.3%) (Table 2).

In bivariate analyses, nulliparity, duration of diabetes ≥10 years, and elevated prepregnancy HbA_1c were significantly associated with an increased risk of DR progression, whereas laser photocoagulation before pregnancy and existence of prepregnancy DR were significantly associated with a lower risk (Table 3).

All possible first-order interactions with prepregnancy DR were further investigated by including the corresponding interaction term in the mixed logistic regression models. Except for CSII treatment (P for interaction = 0.0004), no significant first-order interactions of prognostic factors with prepregnancy DR were found (data not shown). Among pregnancies with prepregnancy DR, CSII treatment was associated with a decreased risk of DR progression (odds ratio [OR] 0.24; 95% CI 0.09–0.62; P = 0.003), whereas it was associated with an increased risk in DR development in pregnancies without prepregnancy DR (OR 1.72; 95% CI 1.02–2.89; P = 0.041). Furthermore, the prognostic impact of prepregnancy DR was greater in pregnancies with CSII treatment (OR 0.23; 95% CI 0.10–0.50; P < 0.001) compared with MDI treatment (OR 0.39; 95% CI 0.17–0.87; P = 0.022).

In backward stepwise multivariate analysis (Table 4), nulliparity, diabetes duration, elevated prepregnancy HbA_1c, and interaction between prepregnancy DR and CSII treatment were selected as predictors of DR progression. This selected model had good discrimination (c-statistic 0.87) and calibration, as indicated by the Hosmer–Lemeshow goodness-of-fit test (P = 0.30). In this multivariate analysis, the effect of CSII treatment according to prepregnancy DR status was attenuated and remained nonsignificant within prepregnancy DR strata. The strong prognostic impact of prepregnancy DR toward a lower risk of DR progression remained significant in pregnancies with CSII treatment only (OR 0.12; 95% CI 0.05–0.30; P < 0.001).

During pregnancy, mean HbA_1c level decreased from the prepregnancy value of 7.6 ± 1.6% (60 ± 17 mmol/mol) to 7.0 ± 1.2% (53 ± 13 mmol/mol) during the first trimester, to 6.5 ± 1.05% (48 ± 11 mmol/mol) during the second trimester, and to 6.6 ± 0.9% (49 ± 10 mmol/mol) during the last trimester. As shown in Fig. 1, change in HbA_1c during the last trimester of pregnancy differed significantly according to DR progression status. After adjustment of prepregnancy HbA_1c values, the adjusted mean HbA_1c reduction during the last trimester was significantly greater in pregnancies with DR progression (−1.25; 95% CI −1.42 to −1.08) than in those without (−0.99; 95% CI −1.09 to 0.90; P = 0.008). After additional adjustments for nulliparity, diabetes duration, and interaction between prepregnancy DR and CSII, this difference remained significant (mean change −1.31 vs. −1.01; P = 0.002).

In our large cohort of 499 pregnant women with type 1 diabetes, we show that the rate of DR progression during pregnancy was 21.8%, including DR development in 24.4% of the 348 pregnancies without prepregnancy DR and DR worsening in 15.9% of the 151 pregnancies with prepregnancy DR. Although a retrospective study, it includes the largest population (499 pregnancies) published for the study of DR during pregnancy in type 1 diabetes. There were few exclusion criteria, and the results are likely to be representative for women with type 1 diabetes. The same ophthalmologist followed all women during their pregnancies over the entire period from 1997 to 2015. Before the 2000s, DR was reported to progress in 5–42% of pregnancies in women with diabetes (15,16). More recent cohort studies have shown a decrease in the DR progression rate to ∼20% (8,10–12) or even lower (6), probably because of better glycemic control during the conception period and the increase in planned pregnancies.

Our rate of progression (21.8%) was close to those reported in other recent studies (8,10,11), with more DR progression among patients without prepregnancy DR (24.4%) than among those whose DR was present before pregnancy (15.9%). In the Diabetes in Early Pregnancy study, progression of two or more DR levels was present in 10.3% of pregnancies with no DR at baseline and 54.8% of pregnancies with moderate DR or worse at baseline (9). Morrison et al. (17) extracted data from 14 studies where DR progression based on baseline level was reported. They reported that incident DR occurred in 13.1% of women with type 1 diabetes without retinopathy at baseline and that 33.8% of women with NPDR at baseline experienced progression during pregnancy (17). In our study, PDR rarely developed (n = 3) in the absence of DR or when only mild microaneurysms were present at the beginning of pregnancy, and sight-threatening deterioration of retinopathy was rare. These results are partly due to our population including only a few pregnancies with moderate to severe untreated DR, with cases of DR being laser treated before pregnancy. In our study, we separated NPDR into mild, moderate, and severe DR; however, only a few published studies have confirmed a trend of moderate or severe baseline DR progressing more than the mild form.

In our study, DR progression occurred mostly during the first and second trimesters. In addition, third-trimester DR progression was observed in pregnancies in which DR had already started to progress during the first or second trimester. Therefore, follow-up of pregnant women with type 1 diabetes during the third trimester could depend on the severity of the previously diagnosed DR. Although some guidelines recommend follow-up every trimester, this point should be considered according to DR status in the first and second trimesters. Therefore, unless there are significant confounding risk factors, it seems reasonable to discuss the next examination according to DR status in the second trimester. The observation that progression mainly occurred in early or midpregnancy is important for future recommendations regarding timing and frequency of screening for retinopathy.

As Arun and Taylor (10) showed in their study including 59 women, DR progression remains rare during the postpartum period (4.1%), whereas regression was twice as frequent during the year after delivery (9.3%). Although guidelines recommend close monitoring for DR detection during the 6–12 months after pregnancy, our results show that DR progression remains rare after delivery, even with DR regression of 9.3%. The observation of a low risk of deterioration of retinopathy the 1st year after delivery is novel, and this point is important for future guidelines for retinopathy screening.

Regarding the risk predictors for DR progression, our study confirms the high impact of diabetes duration (P < 0.001) and suggests a trend for nulliparity (P = 0.069). However, unlike findings in other studies (5,11–13), patients with complications such as arterial hypertension or diabetic nephropathy were not at risk for progression of DR. However, we did not consider systolic or diastolic blood pressure but only presence or absence of arterial hypertension. Our results are similar to those of Lauszus et al. (18), who showed that hypertension was not associated with DR progression in a cohort of patients with type 1 diabetes who had 24-h blood pressure monitoring.

As expected, glucose control remained a major factor in the prevention of DR development or worsening, and significant results were seen in our large population. Poor glycemic control has been shown to be a risk factor for DR progression during pregnancy in women with type 1 diabetes (3,9,11,13), with a lower risk of progression in those with tight glucose control before conception (3). In our study, prepregnancy HbA_1c was a significant predictor for DR progression, and we show, after adjustment for prepregnancy HbA_1c values, that the adjusted mean decrease in HbA_1c during the last trimester was significantly greater in pregnancies with DR progression than in those without. The importance of prepregnancy glucose control optimization should be highlighted, because it is associated with less DR progression. However, this retrospective study included a study period of 18 years. Treatment practice is likely to have changed over this time period. In particular, short-acting analogs were introduced in the early 2000s and long-acting analogs in 2012 in France. Use of diabetes technologies, including insulin pumps, has become more prevalent in recent years. One weakness of our study is that we had no information on differences over time regarding HbA_1c levels, types of insulin, or percentages of women using an insulin pump and thus could not determine how any differences may have affected study results.

Another interesting observation was the association between CSII treatment and DR progression depending on whether DR antedated pregnancy. Among women with prepregnancy DR, there was a trend for a protective effect of CSII treatment, whereas there was no effect in women without prepregnancy DR. We can speculate that there is a greater tendency to propose CSII treatment in women with prepregnancy DR compared with women without prepregnancy DR.

Women with prepregnancy DR had a lower risk of DR progression when they were treated with CSII compared with women treated with MDIs. This is the first report of data on DR progression according to type of treatment (CSII vs. MDIs). In women with prepregnancy retinopathy, insulin pump treatment was associated with a decreased risk of retinopathy progression. One may speculate whether this could be ascribed to more refined treatment during the later part of the study period (introduction of insulin analogs, more use of diabetes technology) or, in women participating more than once, to a change in treatment (change to insulin analogs, change from MDIs to insulin pump treatment) from the first to second pregnancy. Moreover, in some smaller and older studies, women with type 1 diabetes receiving insulin pump treatment had more diabetes complications (19), but this was not seen in the recent CONCEPTT trial (20). However, unlike in the CONCEPTT study (21), we had no data about metabolic profile (time in range, time in hypoglycemia, time in hyperglycemia).

In summary, we should consider ophthalmologic follow-up crucial for women with type 1 diabetes before and during pregnancy, but it could be less important during the third trimester in patients with no previously diagnosed DR. For women planning pregnancy, an HbA_1c target of ≤6.5% seems prudent. Not only would this lead to tight glucose control, it would also prevent a rapid decrease in HbA_1c, especially between prepregnancy and the third trimester. In women with a longer duration of diabetes, we should take care to treat severe DR before women consider pregnancy. Unplanned pregnancies and nulliparous patients would also benefit from intensive ophthalmologic follow-up. CSII could also be proposed for patients with prepregnancy DR to prevent its worsening. This new element could help us prevent DR from worsening during pregnancy in these patients.

Acknowledgments. The authors thank Sylvie Picard, endocrinologist, Dijon, France, for her help in the preparation of the English manuscript.

Funding. No separate funding was obtained for this project.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.B. collected data and wrote the manuscript. H.C. interpreted retinal pictures and reviewed the manuscript. N.R., E.D., and A.D. directed and conducted the statistical analyses. D.S. and P.D. reviewed the manuscript. A.V. initiated and directed the study and reviewed the manuscript. A.V. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.