Retinal blood flow association with age and weight in infants at risk for retinopathy of prematurity

This prospective study evaluated the relationship between laser speckle contrast imaging (LSCI) ocular blood flow velocity (BFV) and five birth parameters: gestational age (GA), postmenstrual age (PMA), and chronological age (CA) at the time of measurement, birth weight (BW), and current weight (CW) in preterm neonates at risk for retinopathy of prematurity (ROP).38 Neonates with BW < 2 kg, GA < 32 weeks, and PMA between 27–47 weeks underwent 91 LSCI sessions. Correlation tests and regression analysis were performed to quantify relationships between birth parameters and ocular BFV. Mean ocular BFV index in this cohort was 8.8 +/− 4.0 IU. BFV positively correlated with PMA (r = 0.3, p = 0.01), CA (r = 0.3, p = 0.005), and CW (r = 0.3, p = 0.02). BFV did not correlate with GA nor BW (r=−0.2 and r=−0.05, p > 0.05). Regression analysis with mixed models demonstrated that BFV increased by 1.2 for every kilogram of CW, by 0.34 for every week of CA, and by 0.36 for every week of PMA (p = 0.03, 0.004, 0.007, respectively). Our findings indicate that increased age and weight are associated with increased ocular BFV measured using LSCI in premature infants. Future studies investigating the associations between ocular BFV and ROP clinical severity must control for age and/or weight of the infant.


INTRODUCTION
Retinopathy of prematurity (ROP) is the leading cause of preventable blindness in preterm infants in the US and worldwide.[1], [2] The two most well-established risk factors for ROP are early gestational age (GA) and low birth weight (BW).[3]- [7] The window of greatest risk for ROP progression dictates the timing of retinal screening eye examinations, which generally occur from 31 to 40 weeks or more of postmenstrual age (PMA).
The underlying pathogenesis of ROP is qualitatively described as an early ischemic phase followed by a late vasoproliferative phase.[8], [9] Imaging tools that can capture high resolution and dynamic retinal blood ow patterns in premature infants offer promise in further elucidating the quantitative blood ow changes that occur during the phases of ROP.Tools currently available include laser speckle contrast imaging, color doppler imaging (CDI), intravenous uorescein angiography, and optical coherence tomography (OCT) angiography.
Laser speckle contrast imaging (LSCI) permits the visualization and analysis of retinal blood ow dynamics in a noninvasive and non-contact manner.[10]- [14] Re ected laser illumination creates a "speckle" pattern during photography due to selective blurring from erythrocyte motion.[11]- [13] The selective blurring can be interpreted to simultaneously understand both the location and the speed of erythrocyte motion.Software creates a visual heatmap to visualize blood ow velocities within retinal vessels and numeric pro les of blood ow rates in a region of interest that can span one or more cardiac cycles.LSCI is particularly adaptable to infant imaging because it does not require the use of intravenous contrast agents, physical manipulation of the globe with a scleral depressor, a speculum, nor bright visible light, all of which are associated with stress, particularly in fragile preterm infants.[15] Several previous studies have investigated ocular blood ow velocity (BFV) in neonates at risk for ROP [16], [17].Studies using CDI demonstrated a higher BFV in the ophthalmic artery and retinal vessels of infants with ROP compared to those without ROP.[18]- [21] Studies using OCT angiography have identi ed quantitative differences in tortuosity and vessel density associated with ROP.[22], [23] Previous literature on LSCI has established an association between PMA and BFV [17] and identi ed a decline in BFV following ROP treatment, [24], [25] but no previous study had adequate power to formally test the hypothesis that birth parameters, age, and weight are associated with BFV.This study seeks to examine the association between LSCI-based BFV and ve birth parameters: GA, PMA, chronological age (CA), BW, and current weight (CW) in preterm neonates at risk for ROP.

METHODS
The study was conducted in compliance with protocols approved by the accredited Institutional Review Board of the University of Maryland, Baltimore in adherence with the guidelines established by the US Health Insurance Portability and Accountability Act (HIPAA).Informed consent was obtained from the parent of each subject after explanation of the nature and possible consequences of the study.

Participants
In this single-center prospective comparative study, 38 neonates admitted to the Neonatal Intensive Care Unit (NICU) of the University of Maryland Children's Hospital were included.Eligible subjects met the following inclusion criteria: 1) the attending neonatologist requested ophthalmology consultation for ROP based on GA ≤ 30 weeks, or 2) BW ≤ 1500g, 3) or BW 1500-2000 grams or GA > 30 weeks with risk for ROP due to medically unstable clinical course [26].Neonates were excluded from the study if they had media opacity, concurrent vascular or ocular disease other than ROP, a concurrent genetic syndrome, previous intraocular surgery, laser, or intravitreal pharmacologic treatment, or were otherwise evaluated as medically unstable per the attending neonatologist or bedside nurse.If eligible and amenable, the parent was invited to enroll their infant, and informed consent was obtained.

Prior to Imaging
Participants' pupils were dilated with cyclopentolate hydrochloride 0.2%, phenylephrine hydrochloride 1% one hour before imaging.LSCI was scheduled concurrently with the indicated ROP screening eye examination so that pharmacologic dilation was not solely for the purposes of the research study.One drop of 0.5% proparacaine was placed in each eye immediately before the examination.A pulse oximeter linked to the LSCI device was placed on the participant's foot.The infant was positioned supine, swaddled, and a 10-degree wedge was placed under the head and shoulders in all infants for comfort during exam.The wedge was selected because this allowed for optimal chin position during imaging and made infants with re ux more comfortable.The examining physician was at the infant's head and the ROP nurse was at the infant's feet.The eyelid of the eye was then gently opened by the gloved ngertips of the examining physician or ROP nurse.The nurse then tilted the head gently to the left or right to center the eye with the trajectory of the laser.
Imaging LSCI imaging was performed on a weekly basis, concurrent with standard ROP screening eye examination.Both eyes were imaged during each LSCI session, with the goal of one high quality image per eye per session.LSCI sessions were deferred if a subject was medically unstable.Retinal blood ow imaging with LSCI was conducted using the investigational XyCAM NEO prototype system (Vasoptic Medical, Inc., Columbia, MD, USA).The XyCAM prototype is an adaptation of the FDA-cleared XyCAM RI System (Vasoptic Medical, Inc., Columbia, MD, USA) that is mounted with a movable arm.The XyCAM NEO also has software customized for imaging retinal blood ow in supine infants in the NICU.XyCAM uses LSCI to obtain retinal blood ow data within a 25-degree eld of view.In this study, the eld of view was approximately centered on the optic nerve head by adjusting the relative positioning of the XyCAM with respect to the neonate's eye.High frame-rate image data was acquired for duration of six seconds from each eye under near-infrared laser illumination (peak wavelength of 785nm), with BFV index computation at every pixel.Focusing under laser illumination was assisted via visualization of blood ow velocities computed in real-time and depicted in pseudo-color.[13] The data was accepted if it contained a contiguous stretch of at least 2 seconds of 3 or greater artifact-free blood ow pulses (cardiac cycles) without signi cant motion of the eye.If spontaneous motion occurred, the image capture attempt was repeated up to 2 times, for a total XyCAM exam duration of no more than 10 minutes.Immediately following LSCI data collection protocol, the standard ROP eye screening exam was conducted with binocular indirect ophthalmoscopy followed by retinal photography using the RetCam 3 Shuttle (Natus Medical Inc., Middleton, WI, USA) if clinically indicated for advanced or progressive ROP.

Image analysis
The XyCAM image analysis software provides the user with analytical tools to choose speci c regions of interest, and select time-intervals over which the blood ow waveform will be analyzed.Detailed reports contain various retinal blood ow metrics associated with the user-chosen regions and time windows.To examine the retinal blood ow metrics at the optic nerve head, as a theoretical best indicator of total ocular blood ow, the outline of the optic nerve was drawn using a free-form region-selection tool while assisted by fundoscopic image overlay (Fig. 1).To compensate for any motion artifact, the analysis software registers all frames to a single image frame prior to computation of BFV indices.[27] The software also automatically detects peaks and dips in the blood ow waveform that can be manually adjusted if needed to de ne cardiac cycles.Time intervals chosen for analysis comprised of at least 2 continuous cardiac cycles, and the following 3 blood ow metrics were extracted and statistically analyzed: peak, mean, and dip BFV index.Final analysis was limited to the mean BFV index because results did not differ among peak, mean, and dip velocity indices.

Data and Statistical Analysis
Blood ow indices were provided by the XyCAM NEO software and tracked by subject identi cation number and eye across serial examinations ranging from 1 to 6 imaging sessions per eye.Birth parameters (BW and GA) were collected from the history and physical in the electronic medical record on the day of birth.PMA and CA were calculated based on date of birth and GA.Current weight (CW) was collected on the same day as the eye exam or one-day prior nursing report in the electronic medical record.Univariate analysis was used to determine the means and distributions of data.Kendall's correlation coe cients between birth parameters (GA, PMA, CA, BW, and CW) and XyCAM-derived waveform blood ow metrics were calculated based on a randomly selected single eye per subject (n = 38).Mixed-effects regression was used to model the association between blood ow and age and weight.
The use of mixed models allowed for the analysis of the entire dataset (n = 91 exams) while accounting for the inclusion of two eyes per subject and repeat measures over time per eye.R Studio version 4 The correlation between mean BFV and birth parameters are provided in Supplementary Table S1.The peak and dip blood ow velocities were the same as the mean BFV in correlation results, so the analysis only included mean BFV to avoid redundancy.Scatterplots demonstrating the association between mean BFV and birth parameters are provided in Fig. 2.
Table 2 demonstrates the linear regression relationship between individual birth parameters and mean BFV while accounting for repeat measures from the same eye and the association between paired eyes (clustered data) of the same subject.

DISCUSSION
This is the largest study to date to examine ocular blood ow associations with the age and weight of infants in an ROP screening cohort in the NICU setting.Our cohort included 38 infants from an American population of diverse backgrounds who underwent 91 separate eye imaging sessions.We con rmed the previously established association between PMA and ocular blood ow.[17] We further re ned the understanding of age association with blood ow by identifying and quantifying the analogous association between chronological age and ocular blood ow.Similarly, we quantify the signi cant association between weight and ocular blood ow.Given that weight increases with age, these variables are colinear and therefore were evaluated in our model individually.Future studies may justify inclusion of one or more age or weight variables in models associating blood ow and ROP severity, depending on the sample size and objectives of the study.
The correlation between blood ow volume and the size or age of an individual is intuitive, but nuances in a population of premature infants may be in uenced by the many unique features of their physiology including rapid growth rates (relative to adults), anemia of prematurity, exogenous uids, and oxygen supplementation.While it would be impossible to adequately control for all these features, our cohort re ects a pragmatic population among whom all of these factors are present and likely relatively consistent.
Our  [19], [33] Some CDI studies however did suggest increased blood ow with increasing age and weight, among infants in rst week of life.[20], [21] OCT angiography has similarly established quantitative vascular changes related to tortuosity and vessel area density in pre-term infants with more severe ROP.[22] No previous study using OCT angiography has evaluated the association between current age or weight and OCT angiographic features in infants at risk for ROP, Technical challenges limit clinical application of OCT and neonates because the tool is extremely sensitive to motion artifact.For example, previous literature using OCT angiography had sample size less than 10 infants.[22], [34] The strengths of our study include a diverse patient population, a relatively large sample size, and statistical analysis accounting for repeat measures.Our cohort was racially diverse, and we included high-risk neonates with much younger gestational ages (as young as 23 weeks) and extremely low birthweights (505 grams), compared to prior literature.Furthermore, our cohort was mostly incubatorbound and the majority of subjects were on respiratory support or supplemental oxygen.In this way, our study re ects a "real world" ROP cohort.This study also offers the largest sample size to date, compared to previous ROP studies using laser speckle contrast imaging.Our study included 38 subjects and 91 eye exams.We analyzed both eyes per subject and included repeat measures.We accounted for the inclusion of both eyes and repeat measures statistically.Our use of mixed models to account for repeat measures offers more power to detect relationships between variables that vary over time, highly appropriate to this use case looking at age and weight.
The limitations of this study included a challenging population of awake neonates, with slightly inferior reproducibility compared to adults.[17], [35] This study excluded neonates with signi cant systemic medical instability, but the cohort did include some infants with a history of various common comorbidities of prematurity (i.e.bronchopulmonary dysplasia), however we lacked adequate sample size to control for these comorbidities.Although our study examined the association of heart rate with BFV, we lacked reliable systemic blood pressure data.Previous related work did not nd any signi cant association between blood pressure and blood ow in similar cohorts [17], [18], as justi cation for the lack of inclusion of blood pressure in the data collection and analysis.Lastly, while LSCI provides indices of blood ow that are repeatable and reproducible, [11], [13], [14], [27], [35] it has never been validated to measure absolute blood in vivo in human eyes.Therefore, though comparisons can be made across subjects measured on the same device and relative changes can be compared across devices, direct comparisons are not possible between different LSCI devices.This differs from CDI, where absolute ow velocity in mm/s is available but not at a comparable resolution as LSCI.LSCI has key advantages over CDI including improved resolution, non-contact imaging, and ability to evaluate more anatomic regions of the retina.Also, LSCI offers concurrent retinal imaging and retinal blood ow evaluation; while CDI is limited to retrobulbar blood ow assessment.
Our ndings indicate that increased age and weight are associated with increased ocular BFV measured using LSCI in premature infants.This work highlights the importance of controlling for the age and/or weight of the infant in future studies investigating associations between ocular BFV and ROP clinical severity.

Declarations
and extracted data, analyzed images, statistically analyzed the data, interpreted the results, wrote the report, generated the initial hypothesis and design for the study, and acquired grant funding.All authors reviewed the manuscript.

Figures
Figure 1 Image of ocular blood ow using laser speckle contrast imaging, demonstrating blood ow heatmap, reconstructed fundus image, and dye-free angiography.Bottom panel demonstrates blood ow velocity waveform with concurrent pulse oximetry tracing.
.1.2(RFoundationforStatisticalComputing, Vienna) and SAS v9.4 (Statistical Analysis Software, North Caroline State University) were used to conduct the statistical analysis.RESULTSThis study included 38 preterm infants who underwent 91 LSCI examinations.Table1provides demographic and clinical features of the cohort.

Table 2
Models for the association between age and weight with mean blood ow.Analysis used mixed models to simultaneously account for the inclusion of both eyes per subject, and the inclusion of repeat measures of the same eye (n = 91 eye exams).Bold indicates p < 0.05.
[17]lts demonstrate that BFV changes in a predictable way during the early months of neonatal growth.Mean BFV index increases by 1.2 for every kilogram of weight gain, by 0.34 for every week of chronological age, and by 0.36 for every additional week of post menstrual age.Given that current age and post menstrual age are essentially the same variable, offset by gestational age, they cannot be meaningfully used in the same model.Similarly, age and weight are closely related and need not be included in the same model.In general, our results suggest that associations between blood ow and ROP severity will require controlling for at least one variable related to age or weight (postmenstrual age, current age, or current weight), and potentially all 3. Conversely, birthweight and gestational age did not show any meaningful association with mean BFV, and therefore should not be included in models evaluating ocular blood ow in ROP.Previous literature on blood ow in ROP includes data from LSCI, CDI, and OCT-angiography.[16],[28]-[32]Ourresultsareconsistent with prior work by Matsumoto et al., who reported signi cant association between post menstrual age and BFV as measured by LSCI.[17]Their group found a near signi cant association for current weight.They found no correlation of blood ow with chronological age but only with post menstrual age, possibly due to the choice of Pearson correlation rather than Kendall's correlation, which can be unduly in uenced by extreme values.Given the lack of normality and relatively small sample size, Kendall's correlation was appropriate for our dataset.CDI studies are inconclusive regarding the relationship of blood ow with age or weight in an ROP screening cohort.Several studies using CDI found relationships between blood ow and ROP severity, but most did not evaluate the association between blood ow and age or weight.[16], [18],