Conflict/competing interest: No stated conflict of interest.
Funding sources: None declared.
Professor Glen Gole, Discipline of Paediatrics and Child Health, Royal Children's Hospital, Herston, Qld 4029, Australia. Email: firstname.lastname@example.org
Background: To report the incidence of retinopathy of prematurity (ROP) in a subgroup of extremely premature infants admitted to an Australian tertiary centre over an 18-year period.
Design: Retrospective study. Royal Brisbane and Women's Hospital Neonatal Intensive Care Unit.
Participants: Five hundred and fifty-four infants admitted between 23 and 25.6 weeks gestational age (GA).
Methods: The 18-year study was divided into three 6-year periods for analysis: period 1 (1992–1997), period 2 (1998–2003) and period 3 (2004–2009). Infants were compared based on their week of GA and by the study period in which they were born.
Main Outcome Measures: GA, birthweight, incidence and severity of ROP.
Results: Three hundred seventy-three (67.3%) infants survived until ROP screening, and 351 (63.4%) survived until discharge. ROP incidence increased from 78.2% in period 1 to 86.1% in period 3. Over the entire study, 90.5% of 23-week GA infants had ROP compared with 89.7% of 24-week GA infants and 76.1% of 25-week GA infants. Mean birthweight was significantly lower in infants with any ROP (725.1 g) and ROP of at least stage 3 (720.8 g) compared with infants without ROP (806.5 g) (P < 0.0001). Twenty-three–week GA infants had more severe ROP (28.6%) than 24 weeks (18.3%) and 25 weeks GA (11.9%).
Conclusions: There has been increased survival and incidence of ROP in extremely premature infants over the past 20 years. Lower birthweight and GA are both associated with higher incidence and more severe ROP.
Over the past two decades, there has been progressive improvement in neonatal practice, which has led to increased survival of extremely premature (EP) infants.1–3 This group of patients with gestational age (GA) less than 28 weeks are at particularly high risk for the development of retinopathy of prematurity (ROP).4,5 However, there is a shortage of data directly comparing ROP trends in current neonatal units with that of the early 1990s. The aim of the present study was to examine changes in the incidence and severity of ROP over an 18-year period in an Australian neonatal intensive care unit (NICU) in the context of declining infant mortality.
A retrospective review was conducted on EP infants admitted to the Royal Brisbane and Women's Hospital NICU, a tertiary referral centre in Queensland Australia between 1 January 1992 and 31 December 2009. All neonates between 23 and 25.6 weeks GA admitted to the NICU were identified from the neonatal unit database. A single paediatric ophthalmologist (GAG) performed all ROP screening examinations and any subsequent treatments.
Screening examinations were performed on all infants with GA < 32 weeks or birthweight < 1500 g, and continued biweekly or weekly if significant ROP was detected. Initial examinations were performed on infants at approximately 32 weeks postmenstrual age. Criteria for staging of disease were according to the International Classification of ROP ‘International Classification of Retinopathy of Prematurity’.6,7 Thirty minutes prior to the examination, pupils were dilated with two drops of topical cyclopentolate 0.2% and phenylephrine 1%. After application of topical anaesthetic (amethocaine 0.5%), an assistant would provide head fixation, while the examiner utilized a wire speculum and lens loop scleral depressor to view the fundus. Criteria for treatment were development of threshold disease8 or pre-threshold disease with high likelihood for progression that in later years was based on the early treatment of ROP study (ETROP) type 1 disease.9 Details of treatment and outcomes are discussed in a subsequent paper (in preparation).
The 18-year study was divided into three 6-year periods for comparison: period 1 (1992–1997), period 2 (1998–2003) and period 3 (2004–2009). Infants were compared based on their week of GA and by the study period in which they were born.
Statistical analysis was performed using the commercial software package IBM SPSS Statistics 19.0 (IBM Corporation, Armonk, NY, USA). Birthweights were analysed using one-way ANOVA with Bonferroni post-tests to correct for multiple statistical comparisons. Data were compared with contingency tables and the use of Fisher's exact test. Risks for development of ROP were modelled with univariate and multivariate logistic regression analyses.
From 1992 to 2009, 554 infants between 23 and 25.6 weeks GA were admitted to Royal Brisbane and Women's Hospital NICU. Details of the cohort by GA and study period are included in Table 1. Three hundred seventy-three (67.3%) infants survived to undergo screening for ROP, and 351 (63.4%) survived until NICU discharge. There were no infants in the cohort who survived without undergoing screening. Survival rates for each period are displayed in Figure 1. Overall, EP infant survival increased slightly over each period of the study, from 60.0% in period 1 to 64.9% in period 3 (P < 0.01). However, survival of 23-week GA infants improved substantially from period 1 (16.7%) to period 2 and period 3 (45.5%). Because of low sample size in each group (n = 55), statistical significance was not reached.
Table 1. Characteristics of EP infant cohort admitted to NICU and screened for ROP (%)
EP, extremely premature; GA, gestational age; NICU, neonatal intensive care unit; ROP, retinopathy of prematurity.
Mean birthweight (g)
Mean GA (weeks)
GA 23 weeks
GA 24 weeks
GA 25 weeks
The incidence of ROP in the cohort is presented in Table 2. As the study period progressed, there was increased incidence of ROP; between 1992 and 1997, 78.2% of screened infants developed ROP compared with 86.1% after 2004 (P < 0.05). The incidence of severe ROP was relatively stable over the study periods (P = 0.44).
Table 2. Incidence of ROP in screened EP cohort (%)
Mean GA (weeks)
Mean birthweight (g)
EP, extremely premature; GA, gestational age; ROP, retinopathy of prematurity.
Severe ROP (ROP ≥ 3)
Mean birthweight of the entire screened cohort was 740.1 g. This decreased from 745.5 g in period 1 to 724.2 g in period 3 (P < 0.0001). Mean birthweight was significantly lower (P < 0.0001) in infants with any ROP (725.1 g) compared with infants without ROP (806.5 g). More specifically, infants with stage 2 (722.5 g) and at least stage 3 (720.8 g) disease had significantly lower birthweights than disease-free individuals (P < 0.0001) (Fig. 2).
Table 3 displays the incidence of different stages of ROP in the cohort divided into groups by week of GA. Each additional week of GA was associated with a fall in the incidence of ROP. Results show that 90.5% of 23-week GA infants had ROP as compared with 89.7% of 24-week GA and 76.1% of 25-week GA (P < 0.01). Also, 23-week GA infants had more severe ROP (28.6%) than 24-week GA (18.3%) and 25-week GA (11.9%) (P < 0.05).
Table 3. Maximum stage of ROP detected in infants screened for ROP (%)
All screened (n = 373)
GA 23 weeks (n = 21)
GA 24 weeks (n = 126)
GA 25 weeks (n = 226)
GA, gestational age; ROP, retinopathy of prematurity.
Without ROP (%)
With ROP (%)
Severe ROP (≥3 (%))
ROP stage 1 (%)
ROP stage 2 (%)
ROP stage 3 (%)
ROP stage 4a (%)
ROP stage 4b (%)
ROP stage 5 (%)
There were 185 males and 188 females in the 373 screened infants. ROP was found in 83.2% (154/185) of males and 79.8% (150/188) of females. There were 86 twins or triplets in the cohort, 84.9% (73/86) of whom developed ROP compared with 80.5% (56/231) in non-twins. Fisher's exact test showed no significant difference in ROP incidence between genders (P = 0.425) or twin status (P = 0.430).
Logistic regression analyses were performed to compare individuals who developed ROP with those who did not.10 A univariate analysis was conducted with GA as a continuous variable with the range of 23 weeks to 25 weeks and 6 days, and found a log-linear relationship with ROP development. This model found that for each additional week of GA past 23 weeks, the odds ratio for development of any ROP, compared with 23 weeks, was 0.401. (95% confidence interval 0.25–0.66, P < 0.001). The model was well fitted to the dataset and described the relationship significantly better than the null hypothesis (Wald 13.2, Hosmer–Lemeshow χ2 (8) = 5.1, P = 0.75). When considering the dependent variable as development of severe ROP, each additional week of GA was associated with an odds ratio of 0.524 (95% confidence interval 0.34–0.81). This relationship was significant (P < 0.005, Wald 8.25, Hosmer–Lemeshow χ2 (8) = 8.11, P = 0.42). Repeating this logistic regression analysis considering each week of GA as a class variable showed odds ratios of 2.9 and 1.6 for 23 and 24 weeks compared with 25 weeks GA. Additionally, a multivariate analysis was performed, considering GA and birthweight as continuous variables, and gender as a simple categorical variable. Birthweight and gender were both found to be insignificant predictors of development of severe ROP (P = 0.90 and P = 0.45), with GA still being a significant predictor considering these influences (P = 0.014).
Of the 373 infants examined, 128 eyes from 66 individuals (17.6%) subsequently required laser diode treatment, while 2.1% of individuals (8/373) displayed Rush disease, later designated as aggressive posterior ROP (ICROP II).6,7 These results will be discussed in a further paper.
With progressive advances in neonatal medicine, the limit of GA that is compatible with life has decreased in recent years.1–3 Our findings support this trend, showing increased survival of EP infants from 1992 to 2009 with a particular rise in survival of 23-week GA infants (16.7% to 45.5%). There now exists a developing population of surviving infants with extremely low birthweight who are at high risk for the development of ROP with the potential consequence of lifelong poor vision.2 With continued improvement of neonatal services in middle-income and developing countries, ROP is increasing in incidence and is becoming a major cause of childhood blindness in these countries.11,12 Recent national-based population studies agree that as GA decreases under 26 weeks, the risk of developing ROP and more severe stages of ROP increases substantially.4,5,13–15 Our single-centre study adds further information to these large national studies by comparing the current trends in EP infant ROP with those of almost 20 years ago. Table 4 presents our data in comparison with these studies.
Table 4. Incidence of ROP in 25.6-week infants compared with recent population studies
The incidence of ROP in EP infants in our study was 81.2% and is compared with other studies in Table 4. The incidence was higher than the 47.5% seen in the Norwegian population study14 and the 69.0% seen in GA ≤ 27 weeks in the ETROP study.16 A more recent Swedish national study showed very similar overall ROP incidence at 81.7%.5 Perhaps of more clinical significance is the 15.0% incidence of severe ROP (stage ≥ 3) seen in our study, a population that is at higher risk of poor visual outcomes and more likely to receive treatment. Recent single-centre studies have presented incidence rates comparable with our own. For infants with birthweight <1000 g, severe ROP was seen in 17.0% in a Brazilian single-centre study17 and 13.7% (threshold ROP) in a Singapore study.18 The rate was also comparable with the 14.1% of the Norwegian14 and 21.9% of the Australia/New Zealand population studies13 but was substantially less than the Belgian, Austrian and Swedish studies.4,5,15
When comparing ROP incidence studies, there are a variety of potential confounding factors to consider. First, EP infants account for only a very small proportion of NICU patients, and consequently, statistical analyses are weakened by small sample sizes. This is particularly evident in GA 23- and 24-week infants. Population studies benefit from larger cohorts that strengthen significance and may minimize geographical selection bias. However, the majority of recent ROP population studies have been conducted in European populations with high proportions of Caucasians who are at increased risk of developing severe ROP.19 Also, despite the existence of the International Classification system, multicentre trials are more likely to be confounded by inconsistencies between examiners in terms of disease staging and treatment. The definition of severe ROP varies between studies from ROP stage ≥3, threshold ROP or ROP requiring treatment. Few of the studies included in Table 4 publish the specifics of ROP screening examination techniques that may potentially confound results.13–15 There may be variation in philosophies around initial resuscitation and end-of-life issues at the limit of viability between study centres that potentially select for later GA individuals with decreased incidence and stage of disease.20 For instance, recent recommendations in France are for GA < 24 weeks infants to be offered palliative services only, with 24–25 weeks deemed a ‘grey zone’ where treatment is discretionary and >26 weeks as full resuscitation.21 In the USA, recent guidelines suggest <23 weeks or <500 g as palliative only, 23–24.6 weeks as the grey zone and >25 weeks as full resuscitation.22 A recent consensus study in Australia defined 23–25.6 weeks as the grey zone for initial resuscitation.23 In practice, a complex variety of factors influence decisions around resuscitation, including condition at birth, response to initial treatment, infant gender, and the views of parents and clinicians.24 Finally, the recent date of the last cohort of our study must also be considered to account for the increased number of EP infants surviving in later periods who are at increased risk of developing ROP and severe ROP.
Although it is difficult to make direct comparisons between studies in terms of specific incidence of ROP, consistent relationships between variables have been found. Population5,13–15 and single-centre studies25–28 generally agree that with decreased GA and birthweight, there is increased frequency of ROP and increased incidence of severe ROP. Our findings are consistent with these trends, as seen in Figure 2 for birthweight and Table 4 for GA. Logistic regression analyses in this study revealed a strong inverse relationship between advancing GA and risk of development of ROP and severe ROP. For each additional week of GA, the odds ratio of developing severe ROP compared with that at 23 weeks is 0.524. This compares well with a recent Swedish national study that found an odds ratio of 0.51 for 22- to 26-week GA infants.5 A similar log-linear relationship was seen between GA and risk of development of severe ROP in the Australian and New Zealand study, although this study did not distinguish infants below GA 24 weeks.13 Our multivariate analysis found GA to be a stronger predictor of development of severe ROP than birthweight and gender, which were both found to be insignificant covariates in our modelling.
The strengths of this study include regular data collection over a significant period, a large sample size for this specific cohort, the consistency of an experienced single examiner (GAG) and a study conducted in a single institution. Because of the small number of live born infants with GA less than 26 weeks, it is difficult for retrospective single-centre case control studies to obtain sufficient EP numbers for statistical significance. Three hundred seventy-three individuals were included in our study, which is a sample size consistent with national population studies, albeit over a much longer period. All infants admitted to the NICU over the period were screened, and there were no exclusion criteria. A single paediatric ophthalmologist (GAG) performed all examinations in our study, utilizing ICROP and later revised ICROP II to minimize variation.6,7 Royal Brisbane and Women's Hospital NICU is a large 66-cot unit serving most of Queensland and some of Northern New South Wales. This is a geographically large region with a draining population of around 5 million people. The unit is typical of the few tertiary level NICUs in Australia, and figures presented in this paper are consistent with that of the Australian and New Zealand Neonatal Network study.13
In the Australian and New Zealand network population study, male gender was found to be a risk factor for development of ROP.13 Although there was increased incidence of ROP in males in our study, this was not found to be significant (P = 0.425). Similar non-significant differences have been seen in other studies.4,18,26,27
In the 23- to 25.6-week GA cohort, the incidence of ROP was seen to increase over the study period from 78.2% in period 1 to 86.1% in period 3. This could be accounted for by increased survival rates of infants with lower birthweights and younger GA. Measured comparison between the 1986–1987 Cryotherapy for Retinopathy of Prematurity trial and 2000–2002 ETROP trial reveals similar trends to those seen in our paper. For individuals with GA < 27 weeks, 83.4 % of individuals developed ROP in cryotherapy for ROP8 compared with 89.0% in ETROP. There was also an observed increase in zone I and more severe stages of ROP.16 A recent report from a single German institution found that there was no increase in ROP over a 30-year period but rather a decrease in incidence and rate of operation for infants less than 28 weeks.29 In that study, however, 26- and 27-week infants were included in whom ROP incidence is lower, and the sample size was relatively small with limited 23- to 25.6-week GA representation.
As our study progressed, it may have been predicted that severe ROP would increase in incidence, considering more premature infants survived. However, we found that the incidence of severe ROP remained relatively stable. This is likely explained by the influence of shifts in practice secondary to the ETROP study that promoted treatment of high-risk pre-threshold disease. Individuals that may have later developed more severe stages of ROP were treated early with the aim of preventing such outcomes.9
There have been significant changes in neonatal medicine in the past 20 years. With increased survival of EP infants at the edge of viability, there is a growing population of individuals at particularly high risk of development of sight-threatening ROP. Further research into the assessment and management of ROP in this group is of growing importance.