Changes in Optical Coherence Tomography Angiography Precede Clinical Onset of Placental Insufficiency

Abstract

Purpose: This prospective cohort study examined the use of optical coherence tomography angiography (OCTA) to detect in vivo retinal and choroidal variations associated with placental insufficiency, manifesting as fetal growth restriction (FGR) and hypertensive disorders of pregnancy (HDP).

Methods: Pregnant women were imaged with OCTA from gestational week 16 and every 4 weeks until week 36. Pregnancy outcomes were categorized into three groups: uncomplicated, complicated by late FGR, or complicated by HDP after 32 weeks. OCTA metrics, including retinal perfusion density (PD), vessel length density (VLD), and choriocapillaris flow deficits (CCFDs), were compared between groups.

Results: In uncomplicated pregnancies, OCTA metrics remained stable throughout gestation. In contrast, the FGR group exhibited significant overall increases in superficial VLD (P = 0.016), decreases in deep VLD (P = 0.029), and increases in CCFDs (P = 0.002) throughout pregnancy when compared to the uncomplicated group. The HDP group showed significant overall decreases in both PD (P < 0.001) and VLD (P = 0.019) in the deep layer when compared to the uncomplicated group. Furthermore, CCFDs demonstrated strong potential for early prediction of FGR as early as week 16 (area under the curve = 0.73; P = 0.012).

Conclusions: Our pilot study highlights the potential of OCTA to identify retinal and choroidal variations as biomarkers for pregnancy complications, particularly increased CCFDs in FGR.

Placental insufficiency affects 10% to 15% of all pregnancies and occurs when nutrients are not properly transferred from the mother to fetus.^1,2 This leads to complications such as hypertensive disorders of pregnancy (HDP), including gestational hypertension, preeclampsia, and eclampsia; fetal growth restriction (FGR); preterm birth; and stillbirth.³ During pregnancy, hemodynamic and hormonal changes impact perfusion and vascular structure in the placental, renal, and retinal vasculature. Currently, the diagnosis of placental insufficiency primarily relies on obstetric ultrasound techniques, such as Doppler flow, to monitor placental vascular resistance. However, Doppler ultrasound is generally only utilized if there is suspected FGR and may underdiagnose placental insufficiency in fetuses who fall in the normal growth percentiles by conventional growth charts but are small for their genetic predisposition.^3,4 This limitation is significant given the crucial role of placental health in ensuring fetal well-being and preventing complications. Although magnetic resonance imaging can provide detailed information and has potential in research settings, it is not commonly used clinically due to its high cost, limited accessibility, and risks associated with the use of contrast dye during pregnancy. These diagnostic constraints highlight a critical gap in our ability to monitor and understand vascular changes that occur during pregnancy. 

Optical coherence tomography (OCT) and optical coherence tomography angiography (OCTA) enable non-invasive assessment and measurement of the retinal microvasculature during pregnancy without the risk or cost associated with contrast imaging techniques. Furthermore, OCT and OCTA do not require dilation, offer greater contrast for visualizing the retinal capillaries, and are depth resolved, allowing the different vascular layers of the retina and inner choroid to be separately visualized. A study by Lupton et al.⁵ found that women who developed preeclampsia exhibited a significant reduction in the diameter of retinal arterioles and venules before the onset of clinical symptoms. These reductions in retinal vascular calibers were observed at 13, 19, and 38 weeks gestation, along with increased peripheral vascular resistance compared to the control group. However, although a few studies have explored the utility of OCT/OCTA during pregnancy,^6–12 only one has investigated longitudinal imaging, conducting one session per trimester in uncomplicated pregnancies.⁶ 

Our study aims to bridge this knowledge gap by extending prior studies utilizing OCT/OCTA and focusing on the dynamics of retinal vascular changes during pregnancy complications such as FGR and hypertension disorder of pregnancy (gestational hypertension and preeclampsia) and imaging longitudinally starting from 16 weeks of gestational age. 

Materials and Methods

Patient Population, Recruitment, and Data Collection

In this prospective, longitudinal study, pregnant participants were recruited from the UCLA Health Westwood Obstetrics and Eisner Health clinics when they were at or before 16 weeks gestational age (GA). This study was approved by the Institutional Review Board of UCLA and complied with the tenets of the Declaration of Helsinki. Informed consent was obtained from all patients participating in this study. Recruitment was conducted through a screening approach that included automated EPIC MyChart messaging, strategic flyer placement, and in-person engagement. 

Participants’ pregnancy status was confirmed by first-trimester ultrasound. Inclusion criteria included healthy singleton pregnancies at 16 weeks gestational age (GA), without maternal pre-existing conditions (e.g., diabetes, chronic hypertension, autoimmune, eye diseases except for refractive errors), known fetal genetic disorders, or congenital anomalies. Participants were followed for the entire duration of pregnancy every 4 weeks and for an additional 3 to 6 months through the postpartum period. At the end of the observational period, participants were divided into three groups: (1) uncomplicated group, which demonstrated no acute events during or after pregnancy; (2) FGR group, which was defined as an estimated fetal weight or fetal abdominal circumference below the 10th percentile for gestational age, occurring in the third trimester (in this study, FGR cases were classified as late onset and attributed to placental insufficiency, with no indications of genetic or infectious etiologies); and (3) HDP group, which included cases of gestational hypertension and preeclampsia. Gestational hypertension was defined as systolic blood pressure (SBP) ≥ 140 mm Hg and/or diastolic blood pressure (DBP) ≥ 90 mm Hg occurring after 20 weeks of gestation in women who were normotensive at baseline. Preeclampsia was defined as an SBP ≥ 140 mm Hg and/or DBP ≥ 90 mm Hg after 20 weeks of gestation in previously normotensive women, accompanied by one or more signs of target organ involvement, including proteinuria, thrombocytopenia, elevated transaminase levels, renal insufficiency, pulmonary edema, or new-onset headache.¹³ 

Image Acquisition and Scanning Protocol

Enrolled participants were scheduled for OCTA sessions at regular intervals (every 4 weeks from 16 to 36 weeks of pregnancy). To mitigate the potential impact of diurnal variation, imaging sessions were scheduled at similar times occurring midday between 10 AM and 2 PM. Each subject underwent OCTA imaging using the standard desktop SPECTRALIS OCT2 with a chinrest and OCTA acquisition software (version 6.9.2.700) utilizing a probabilistic, full-spectrum OCTA algorithm (Heidelberg Engineering, Heidelberg, Germany). Two consecutive OCTA scans were performed. A 10° × 10° OCTA scan, consisting of 512 × 512 A-scans, centered on the fovea, was performed using ART 10. The scan with the best quality was selected for further measurements. Images with a quality of Q ≥ 30 with no major signs of motion artifact were utilize for image analysis. For each participant, the first successfully acquired image was used as the reference for subsequent imaging sessions. 

Image Processing

The following macular retinal and choroidal vasculature metrics were analyzed: perfusion density (PD), and vessel length density (VLD) for the retinal circulation and choriocapillaris (CC) flow deficits for the inner choroid. The instrument default automated segmentation was applied to delineate the superficial vascular complex (SVC), deep vascular complex (DVC), and CC. All B-scans were manually reviewed to assess for errors in the segmentation boundaries and manual adjustments were made wherever necessary. Projection artifact removal was applied to the DVC and CC slabs. 

The methods for analyzing PD and VLD were adapted from prior studies.^14,15 Briefly, SVC and DVC images were processed through two binarization methods to refine vessel features and reduce background noise. A Hessian filter with global thresholding enhanced vessel detection, and median local thresholding minimized noise. The final binarized image included only pixels common to both processing methods (Figs. 1A–F). PD was determined by quantifying the ratio of the vessel-occupied area to the total assessable area in binarized images. For VLD, skeletonized images were used to calculate the ratio of vessel length per unit area for the total assessable area. 

Figure 1.

 

Image processing for quantitative analysis of SVC, DVC, and CC metrics. (A) Default SVC OCTA en face slab exported from Heidelberg Eye Explorer 2 (HEYEX 2). (B) Binarized SVC OCTA en face slab used to compute PD. (C) Skeletonized SVC OCTA en face slab used to compute VLD. (D) Default DVC OCTA en face slab exported from HEYEX 2. (E) Binarized DVC OCTA en face slab used to compute PD. (F) Skeletonized DVC OCTA en face slab used to compute VLD. (G) Default CC OCTA en face slab exported from HEYEX 2. (H) Binarized CC OCTA en face slab used to calculate the percentage of CC flow deficits.

The approach for CCFDs analysis was modified from a previous study.⁸ In this method, the default CC slab underwent binarization using Phansalkar's local thresholding method (Figs. 1G–H). Flow deficits were quantified as the percentage of the analyzed area; any deficits having an equivalent diameter smaller than 24 µm were excluded from analysis. This threshold was selected based on prior findings indicating that distances below 24 µm reflect normal intercapillary spacing rather than true perfusion deficits.¹⁶ All quantitative analyses were performed using FIJI software, an extended version of ImageJ 1.51a (National Institutes of Health, Bethesda, MD, USA). 

Statistical Analysis

Statistical analyses were performed using SPSS Statistics 29 (IBM, Chicago, IL, USA). Fisher's exact test was used to assess the difference in the frequency of categorical variables. To detect differences in PD, VLD, and CCFDs among the groups over time, linear mixed-model analysis was used to adjust for correlations between the two eyes of the same subject and repeated measurements. Post hoc analysis with Bonferroni's correction was performed for multiple comparisons if any significant differences were found. Variables with P < 0.05 from the linear mixed model were selected for further evaluation. These variables underwent receiver operating characteristic (ROC) curve analysis to determine their sensitivity and specificity as predictors of the outcome. P < 0.05 was considered statistically significant. 

Results

Subject Demographics

A total of 64 eyes from 32 pregnant females (18 with uncomplicated pregnancies, five with FGR, and nine with HDP) were included in this pilot study. The demographic information for the three groups is shown in Table 1. There are no statistically significant differences in age at delivery, race, or the outcomes for their babies, except that the birth weight is lower in the FGR group compared to the other two groups (P = 0.002) (Table 1). No retinal disorders, including serous retinal detachment, were observed in any cohort throughout the follow-up period during pregnancy. 

Table 1.

 

Demographics of Study Population

 

Retinal Microvascular Changes Throughout Pregnancy

Dynamic Superficial Layer Retinal Vascular Changes

The superficial retinal vasculature in the uncomplicated pregnancy group demonstrated minimal fluctuation in PD (Fig. 2A, purple line). Similar changes in superficial VLD were observed in the uncomplicated group (Fig. 2B, purple line), with no significant differences between visits within the group. The FGR group exhibited an initial increase in PD and VLD from 16 to 28 weeks GA, followed by a decrease at 32 weeks GA and then a return to levels similar to those observed at 28 weeks GA by 36 weeks GA (Figs. 2A, 2B, orange lines). The trajectory of both PD and VLD in the FGR group showed dynamic behavior at individual time points throughout the pregnancy period but no significant differences between visits within the group. The HDP group experienced an initial increase in PD and VLD from 16 to 20 weeks GA, which plateaued at 28 weeks GA (Figs. 2A, 2B, blue lines). This was followed by a continued increase at 32 weeks GA and a slight decrease at 36 weeks. Although there were no significant differences between visits within the group for PD, there was a significant positive effect of visit on VLD (P = 0.034), indicating an overall increase in VLD with each visit throughout the pregnancy period. When compared to the uncomplicated group, both the FGR and the HDP groups demonstrated no overall statistical difference in PD. However, the FGR group demonstrated a statistically significant overall increase in VLD (P = 0.016) (Table 2). Despite not being statistically significant, the HDP group also demonstrated an overall increasing trend in VLD. 

Figure 2.

 

Dynamic changes in retinal vascular structure in pregnant women. (A, B) Retinal vascular changes in the layer of the SVC. (C, D) Retinal vascular changes in the layer of the DVC. In the uncomplicated pregnancy group, both the superficial and deep retinal changes (indicated by the purple line) demonstrated minimal fluctuation in PD and VLD. However, both the FGR group (orange line) and the HDP group (blue line) exhibited distinct dynamic changes throughout the pregnancy. Error bars represent the standard error of the mean.

Table 2.

 

Comparison of Retinal and Choroidal Metrics Averaged Over Study Period (16–36 Weeks GA) in Different Groups

 

Dynamic Deep Layer Retinal Vascular Changes

Similar to changes in the superficial retinal vasculature, changes in the deep retinal vascular layer (both PD and VLD) in uncomplicated pregnancies remained relatively stable throughout pregnancy, except for a notable peak at 28 weeks GA in both PD and VLD (Figs. 2C, 2D, purple lines). There were no significant differences between visits within the group. A decreasing PD in the deep layer was observed in the FGR group over time (P = 0.026), with a sharp decrease from 28 weeks GA to 36 weeks GA (Fig. 2C, orange line). This decreasing trajectory was also noticed in VLD (Fig. 2D, orange line), but it was not statistically significant over time within the group. A steady increasing trend in PD and VLD was observed in the deep retinal vascular layer in the HDP group (Figs. 2C, 2D, blue lines), but no statistically significant difference over time was found within the group. When compared to the uncomplicated group, the FGR group exhibited a statistically significant decrease in VLD across the study period (P = 0.029) (Table 2). Similarly, the HDP group showed a significant reduction in both PD and VLD compared to the uncomplicated group over the study period (P < 0.001 and P = 0.019, respectively) (Table 2). 

Dynamic Choriocapillaris Changes Throughout Pregnancy

CCFDs showed an overall pattern of increasing during the second trimester (<28 weeks GA), followed by stabilization in the later stages of pregnancy, with no significant differences observed over time within each group (Fig. 3). However, the FGR group, but not the HDP group, demonstrated a statistically significant higher average CCFDs compared to the uncomplicated group over the study period (P = 0.002) (Table 2). 

Figure 3.

 

Dynamic changes in choriocapillaris flow deficits across different gestational ages of pregnancy. CCFDs remained relatively stable, with a slight increase from early to late pregnancy, followed by stabilization across the three groups. No significant differences were seen at baseline. However, CCFDs increased significantly during pregnancy in the FGR group compared to the uncomplicated pregnancy and HDP groups (P = 0.002 and P < 0.001, respectively). Error bars show the standard error of the mean.

Potential Discriminatory Ability of CCFDs in Early Detection of FGR

ROC curve analysis was performed to assess the sensitivity and specificity of each variable identified as significant previously (Table 2) in discriminating between the FGR and uncomplicated groups. CCFDs was the only OCTA metric that demonstrated statistically significant discriminatory ability (area under the curve [AUC] = 0.74; P = 0.008) (Fig. 4). Further analysis revealed that CCFDs demonstrated the highest predictive power for FGR during the 16 to 24 weeks GA period, with an AUC of 0.75 (P = 0.004) (Fig. 4). Notably, this discriminatory power could be detected as early as 16 weeks GA, when the AUC was 0.73 (P = 0.012) (Fig. 4). Unfortunately, no significant predictive power was found for distinguishing HDP from uncomplicated groups. 

Figure 4.

 

Potential choriocapillaris flow deficit discriminatory capacity for early detection of FGR. The ROC curves illustrate the predictive performance of CCFDs for FGR across various GA intervals. The model shows that CCFDs have predictive power as early as 16 weeks GA, with the highest accuracy observed during the interval of 16 to 24 weeks GA, corresponding to the second trimester of pregnancy.

Discussion

There is a limited amount of information available regarding retinal and choroidal vascular changes throughout pregnancy, especially longitudinally, whether complicated or uncomplicated. In our longitudinal prospective study, we aimed to measure the retinal and choroidal variations associated with pregnancy complications through non-invasive OCTA imaging. Within the uncomplicated pregnancy group, we noted minimal fluctuations in PD, VLD, and CCFDs, indicating an adaptive mechanism preserving retinal vascular physiology throughout gestation. In contrast, the FGR group exhibited dynamic retinal vascular changes, with a upward trend in superficial retinal VLD, contrasted with a decreasing trend in deep retinal PD and VLD (Fig. 1, orange line). Conversely, the HDP group demonstrated a sustained upward trend in both superficial and deep retinal PD and VLD (Fig. 1, blue line). Notably, the most prominent change was the increased CCFDs observed in the FGR group, potentially serving as a unique image biomarker predictor for pregnancy complications. To our knowledge, this is the first study to investigate CCFDs longitudinally in the setting of pregnancy complications associated with fetal growth restriction. 

For women with uncomplicated pregnancies, the stability observed in PD, VLD, and CCFDs suggests an adaptive mechanism that maintains retinal vascular physiology throughout pregnancy. In the FGR group, there were significant increases in superficial retinal VLD compared to physiologic pregnancies, suggesting a compensatory response to systemic hypoxia caused by placental insufficiency. The increase in superficial VLD likely reflects greater vessel tortuosity, enhancing the vascular surface area to support retinal oxygenation. In contrast, the decrease in deep retinal VLD over time in subjects with FGR may be because the deep retinal layers are more susceptible to hypoxic damage. This opposite trend in superficial and deep retinal layers might indicate an adaptive redistribution of vascular resources in response to altered fetal demands or systemic changes. To date, no studies have specifically shown retinal vasculature alterations in pregnancies with FGR; however, a Doppler study of the ophthalmic artery and biomarkers of impaired placentation at 36 weeks’ gestation observed increased ophthalmic artery peak systolic velocity and altered biomarker levels,¹⁷ supporting our observations of systemic vascular adaptations in FGR pregnancies. The most significant change in the FGR group was an increase in CCFDs when compared to uncomplicated pregnancies. This suggests a lack of blood flow to the choroid, similar to the lack of blood flow to the placenta in placental insufficiency. Additionally, systemic hypoxia, endothelial dysfunction, and increased oxidative stress in FGR further impair choroidal perfusion, making the choroidal vasculature particularly vulnerable to these changes.¹⁸ Baseline CCFDs were also slightly elevated in FGR, suggesting potential preclinical vascular abnormalities that may emerge prior to 16 weeks. Moreover, the overall CCFDs exhibited potential predictive power for FGR, with the highest predictive accuracy during the second trimester (16–24 weeks GA), highlighting the importance of early detection. This early window is critical, as it enables timely interventions that could improve fetal outcomes. 

In the HDP group, significant decreases in both PD and VLD were observed in the deep retinal layer compared to uncomplicated pregnancies. This may be attributed to compensatory vasodilation and increased vascular complexity, aiming to maintain adequate blood flow and oxygen delivery despite elevated blood pressure. These findings align with previous research by Lupton et al.,⁵ which indicated decreased retinal perfusion in the third trimester of pregnancy in patients with preeclampsia. Lupton et al.⁵ demonstrated that women with preeclampsia exhibited reduced central retinal arteriolar equivalent (CRAE) and central retinal venular equivalent (CRVE) compared to normotensive women. Such changes in CRAE and CRVE indirectly suggest alterations in retinal perfusion density, consistent with our results showing lower perfusion density in the deep retinal layer in women with preeclampsia. 

Limitations of this study include a relatively small sample size and a relatively small effect size. Furthermore, we had a relatively homogeneous racial distribution. Environmental factors associated with our specific geographical location add an additional layer of complexity, potentially influencing normative data and limiting the application of our findings to the general population. Additionally, slight variations in refractive error and axial length within the included range may influence vessel perfusion density measurements. However, it is important to recognize that our study is the first to provide insight into retinal and choroidal changes in pregnancy complications associated with fetal growth restriction, which is crucial for comparative analyses in future studies. Despite these limitations, our study identified CCFDs as a potential image biomarker that can serve as a predictor for pregnancy complications associated with FGR. 

In summary, our pilot longitudinal prospective study utilizing non-invasive OCTA imaging demonstrates retinal and choroidal variations associated with pregnancy complications. The observed trends in retinal and choroidal biomarkers, such as PD, VLD, and CCFDs, particularly in pregnancies complicated by FGR, emphasize the potential utility of retinal imaging biomarkers in predicting pregnancy complications. The distinct retinal vascular adaptations observed in the FGR group, coupled with the elevated CCFDs, suggest the feasibility of using these biomarkers as predictors for adverse pregnancy outcomes, although further replicative studies will be required. Despite limitations, our study provides valuable insights into ocular adaptations during pregnancy with or without complications and highlights the promising utility of imaging biomarkers in identifying and managing pregnancy complications. 

Acknowledgments

Supported by grants from the National Eye Institute, National Institutes of Health (R21 EY030295, R01EY024378, R01EY034218), an unrestricted grant from Research to Prevent Blindness to Stein Eye Institute/Dr. H. James and Carole Free Catalyst Award for Innovative Research Approaches for Age-Related Macular Degeneration. 

Disclosure: Y. He, None; P. Heumann, None; M. Weilin Song, None; S. Kadomoto, None; S.R. Sadda, 4DMT (C), Abbvie (C), Alexion (C), Allergan (C), Alnylam Pharmaceuticals (C), Amgen (C), Apellis Pharmaceuticals (C), Astellas (C), Bayer Healthcare Pharmaceuticals (C), Biogen MA (C), Boehringer Ingelheim (C), Carl Zeiss Meditec (C, F, R), Catalyst Pharmaceuticals (C), Centervue (C, F), Genentech (C), Gyroscope Therapeutics (C), Heidelberg Engineering (C, F, R), Hoffman La Roche (C), Iveric Bio (C), Janssen Pharmaceuticals (C), Nanoscope (C), Nidek (F, R), Notal Vision (C), Novartis Pharma AG (C, R), Optos (C, F), Oxurion/Thrombogenics (C), Oyster Point Pharma (C), Regeneron Pharmaceuticals (C), Samsung Bioepis (C), Topcon Medical Systems (C, F, R); I.D. Pluym, None; I. Tsui, None