Keywords
preterm infant - retinopathy of prematurity - risk factors - gestational age - laser
therapy
Introduction
Retinopathy of prematurity (ROP) is a major cause of childhood blindness worldwide.[1] Despite advanced care in neonatology, approximately 16% of extremely low gestational
age neonates suffer from advanced ROP (≥stage 3) and 12% undergo ROP treatment.[2] The pathogenesis of ROP is multifactorial, and several risk factors have so far
been identified. Previous studies in larger cohorts with higher ROP prevalence have
shown that low gestational age (GA), low birth weight (BW), and oxygen use are the
most important risk factors for ROP; however, other associated factors have been recently
described (i.e., days of ventilation, sepsis, hyperglycemia, blood transfusions, severe
hypotension, and impaired postnatal weight gain).[3]
[4]
[5]
[6]
[7]
[8]
[9]
Even if the incidence of blindness among preterm infants suffering from ROP has decreased
over time owing to more effective therapies, the occurrence of severe (type 1) ROP
requiring treatment is still associated with significant long-term visual impairment,
such as myopia and strabismus.[10]
The objectives of this work were (1) to determine risk factors for type 1 ROP (stage ≥ 3
or plus disease) requiring laser therapy among preterm infants admitted at our low-incidence,
regional level III NICU, and (2) to find out if infants treated for severe ROP had
adverse visual and neurodevelopmental outcomes.
Patients and Methods
This study was a retrospective evaluation of prospectively collected data from all
infants born < 30 weeks of GA who were admitted to our regional NICU from January
2004 to December 2013. Infants were examined by the same pediatric ophthalmology service
following ROP international classification throughout the study period.[10] All infants were screened by indirect ophthalmoscopy starting on day 30 (GA at birth < 27
weeks) or day 15 (GA at birth between 27 and 32 weeks) and strictly followed up. Severe
(type 1) ROP was defined as stage 3, 4, 5, or plus disease.[11]
We reviewed patients' data collected in our database and ophthalmology surgical notes
from 2004 to 2013. Neurodevelopmental outcome was assessed at 2 years of corrected
age by means of Bayley scales (third edition) by a skilled physiotherapist and a neonatologist.[12] Composite motor and cognitive scores were calculated for each infant, and the occurrence
of visual impairment was assessed as reported by ophthalmology follow-up visits. The
local Institutional Review Board approved the study.
Sepsis was defined as a positive blood culture or suggestive clinical and laboratory
findings leading to treatment with antibiotics for at least 72 hours. The diagnosis
of patent ductus arteriosus (PDA) was assessed by heart ultrasound performed by skilled
neonatologists. Bronchopulmonary dysplasia (BPD) was defined according to the physiologic
definition.[13]
Demographics and clinical characteristics of patients who underwent laser therapy
for type 1 ROP were compared with controls (no ROP or ROP stage 1–2 without indication
for treatment) by means of univariate analysis: t-test for normally distributed numerical data, Mann–Whitney test for skewed numerical
data, and chi-squared analysis for categorical variables. A p-value < 0.05 was considered significant. A logistic regression analysis was then performed
including the aforementioned variables to find predictors of laser therapy for severe
ROP. Another logistic regression analysis was performed to find out predictors of
neurodevelopmental impairment (cognitive score < 85, motor score < 70). SPSS (version
19.0; IBM, Armonk, New York, United States) was used to perform statistics. Data were
presented as percentages for categorical data, mean, and standard deviation for normally
distributed numerical data and as median and interquartile range (IQR) for skewed
data.
Results
Five hundred fifty-eight infants were born during the study period; 73 died before
the scheduled ROP examination, 20 had major malformations, and 18 had incomplete data.
The study population consisted of 447 infants; 356 out of 447 (79.6%) had follow-up
data (Bayley test was assessed in 237 patients).
Type 1 ROP occurred in 10 patients out of 447 (2.2%) who underwent laser treatment.
These patients were compared with those who did not suffer from type 1 ROP (stage
1 ROP, n = 107; stage 2 ROP, n = 126; no ROP, n = 204). The results are shown in [Table 1].
Table 1
Comparison of clinical outcomes between infants with severe ROP and infants with mild/no
ROP using univariate analysis
|
Type 1 ROP
n = 10
|
Mild/no ROP
n = 437
|
p-Value
|
|
Birth weight (g), median (IQR)
|
818 (741–878)
|
982 (800–1,202)
|
0.090
|
|
Gestational age (days), median (IQR)
|
180 (177–186)
|
196 (187–203)
|
0.000
|
|
Male gender, n (%)
|
4 (40)
|
232 (53)
|
0.526
|
|
SGA, n (%)
|
1 (10)
|
26 (6)
|
0.454
|
|
Preeclampsia, n (%)
|
1 (10)
|
109 (25)
|
0.463
|
|
Antenatal steroids, n (%)
|
8 (80)
|
389 (89)
|
0.301
|
|
Intubation in the delivery room, n (%)
|
6 (60)
|
192 (44)
|
0.351
|
|
Surfactant administration, n (%)
|
9 (90)
|
284 (65)
|
0.176
|
|
FiO2 on day 7 of life, median (IQR)
|
0.21 (0.21–0.28)
|
0.21 (0.21–0.23)
|
0.690
|
|
FiO2 on day 14 of life, median (IQR)
|
0.28 (0.23–0.32)
|
0.21 (0.21–0.25)
|
0.061
|
|
Patent ductus arteriosus, n (%)
|
9 (90)
|
288 (66)
|
0.176
|
|
Early-onset sepsis, n (%)
|
2 (20)
|
26 (6)
|
0.133
|
|
Late-onset sepsis (no. of episodes), median (IQR)
|
1 (0–3)
|
0 (0–1)
|
0.003
|
|
Necrotizing enterocolitis, n (%)
|
2 (20)
|
13 (3)
|
0.046
|
|
Bronchopulmonary dysplasia, n (%)
|
4 (40)
|
114 (26)
|
0.297
|
|
Prevalent human milk feeding (>66% of the total amount during hospital stay), n (%)
|
3 (30)
|
214 (49)
|
0.341
|
|
Dopamine therapy, n (%)
|
5 (50)
|
118 (27)
|
0.143
|
|
Weight gain at week 6 of life (g/kg/d), mean (SD)
|
23.1 (14.8)
|
17.3 (6.0)
|
0.182
|
|
Blood transfusions per patient, median (IQR)
|
7 (2–10)
|
1 (0–3)
|
0.111
|
|
Insulin requirement, n (%)
|
3 (30)
|
31 (7)
|
0.033
|
Abbreviations: Fi, fraction of inspired oxygen; IQR, interquartile range; ROP, retinopathy
of prematurity; SD, standard deviation; SGA, small for gestational age.
Infants with type 1 ROP had worse visual (16.7 vs. 0.6%) and neurological development
compared with the controls ([Table 2]).
Table 2
Comparison of clinical outcomes between infants with type 1 ROP and infants with mild/no
ROP using univariate analysis
|
Type 1 ROP
n = 10
|
Mild/no ROP
n = 346
|
p-Value
|
|
Cognitive score, median (IQR)
|
85 (80–98)
|
95 (85–100)
|
0.615
|
|
Motor score, median (IQR)
|
97 (96–105)
|
100 (92–107)
|
0.963
|
|
Visual impairment[a], n (%)
|
1 (16.7)
|
1 (0.6)
|
0.000
|
|
Cerebral palsy, n (%)
|
2 (33.3)
|
22 (6.3)
|
0.055
|
Abbreviations: IQR, interquartile range; ROP, retinopathy of prematurity.
Note: Some patients had partial follow-up data.
a Strabismus, myopia.
A logistic regression analysis was performed including the variables that were significant
at univariate analysis to find predictors of laser therapy for type 1 ROP. The model
was statistically significant (Hosmer–Lemeshow = 0.816), and the results are presented
in [Table 3].
Table 3
Predictors of laser therapy for type 1 ROP using logistic regression analysis
|
Odds ratio
|
95% CI
|
p-Value
|
|
Increasing gestational age (d)
|
0.904
|
0.828–0.986
|
0.023
|
|
Sepsis (number of episodes)
|
1.813
|
0.906–3.629
|
0.093
|
|
Necrotizing enterocolitis
|
2.564
|
0.309–21.292
|
0.383
|
|
Insulin administration
|
2.336
|
0.517–10.557
|
0.270
|
Abbreviations: CI, confidence interval; ROP, retinopathy of prematurity.
Logistic regression analyses conducted to find out the predictors of neurodevelopmental
impairment revealed the following associations: BPD for impaired cognitive development
(odds ratio: 2.631, 95% confidence interval [CI]: 1.200–5.780, p = 0.016), type 1 ROP (odds ratio: 10.417, 95% CI: 1.149–90.909, p = 0.037), and surfactant administration (odds ratio: 3.952, 95% CI: 1.102–14.084,
p = 0.035) for impaired motor development.
Discussion
This study confirmed that the strongest predictor for type 1 ROP, requiring laser
therapy among preterm infants born at a low-incidence setting for ROP, was low GA.
The study is in agreement with other reports conducted in high-incidence settings.[3]
[9]
Other factors, such as insulin requirement, number of late-onset infections, and necrotizing
enterocolitis, which were recently reported as potential risk factors, were associated
with severe ROP after univariate analysis but were not confirmed after logistic regression
analysis.[4]
[5]
[6]
[7]
Immature GA might reflect the degree of retinal immaturity at birth and therefore
the retinal vulnerability to injury. Furthermore, low GA could increase the duration
of an infant's exposure to adverse extrauterine postnatal insults (i.e., oxygen and
sepsis) contributing to the risk of ROP.
Severe hyperglycemia requiring insulin therapy has been associated with increased
risk of death, sepsis, intraventricular hemorrhage, and ROP.[6] Hyperglycemia may influence ROP through a reduction of retinal blood flow and an
increase in vascular endothelial growth factor (VEGF) production.
Late- but not early-onset sepsis has been associated with severe ROP. The increased
risk may be due to systemic and local inflammation with consequent excessive production
of VEGF and angiogenesis.[7]
We found an association between severe ROP and necrotizing enterocolitis after univariate
analysis. A possible explanation could be that the generalized inflammatory state
observed during necrotizing enterocolitis could damage the developing retina due to
impaired VEGF production, alterations in local blood flow, and oxidative damage. An
alternative explanation of this association could be that necrotizing enterocolitis
and ROP represent comorbidities affecting the sickest infants.
Moreover, recent studies reported on the importance of predisposing genetic factors
(i.e., polymorphisms of VEGF factor A [VEGFA], endothelial NOS [eNOS]) in the occurrence
of ROP and other complications of prematurity.[14]
In a previous study, we showed that the adoption of a controlled oxygen administration
policy, with different saturation targets and depending on gestational age, drastically
reduced the incidence of severe ROP at our institution.[8] Thus, the findings of this study come from a low-incidence setting and may not apply
to other populations.
The main limitations of this study are the retrospective nature of the analysis and
the small number of patients (connected with the low-incidence of ROP at our institution),
which may limit the statistical power.
Infants treated for type 1 ROP compared with infants with mild/no ROP had increased
visual impairment at 2 years of corrected age; moreover, they tended to show worse
neurological development at the same age.
Possible explanations of these findings could be that ROP and brain damage share risk
factors; alternatively, some factors associated with severe ROP (i.e., excessive oxygen
administration, general anesthesia) may be neurotoxic to the developing brain of very
preterm infants.
Predictors of impaired neurodevelopment after logistic regression analysis were type
1 ROP, surfactant administration (for motor development), and bronchopulmonary dysplasia
(for cognitive development). The association between BPD and adverse neurodevelopment
has been documented[15] and might depend on the long-term mechanical ventilation of these infants. Surfactant
administration was (at the time of the study) performed by tracheal intubation and
sedation; we speculate that either the procedure itself could be harmful for the developing
brain or the need for surfactant may be a proxy for the severity of infants' clinical
conditions.
In conclusion, this study showed that low GA was the main factor associated with severe
ROP in a low-prevalence setting, and that infants with severe ROP had higher prevalence
of visual impairment and worse neurological outcome at 2 years of corrected age. Surfactant
administration and BPD were also associated with impaired neurodevelopment. Further
studies are warranted to better elucidate the risk factors for severe ROP in other
settings.
