Prenatal Risk Factors for Severe Retinopathy of Prematurity Among Very Preterm Infants of the Australian and New Zealand Neonatal Network
Objective. To identify prenatal and perinatal risk factors for clinically severe (stage 3 or 4) retinopathy of prematurity (ROP).
Methods. Data were collected prospectively as part of the ongoing Australian and New Zealand Neonatal Network audit of high-risk infants (birth weight of <1500 g or gestational age [GA] of <32 weeks) admitted to a level III neonatal unit in Australia or New Zealand. Prenatal and perinatal factors to 1 minute of age were examined for the subset of infants with GA of <29 weeks who survived to 36 weeks’ postmenstrual age and were examined for ROP (n = 2105). The factors significantly associated with stage 3 or 4 ROP were entered into a multivariate logistic regression model.
Results. Two-hundred three infants (9.6%) had stage 3 or more ROP. Prematurity was the dominant risk factor, with infants with GA of <25 weeks having 20 times greater odds of severe ROP than infants with GA of 28 weeks. Birth weight for GA also had a “dose-response” effect; the more growth-restricted infants had greater risk, with infants below the 3rd percentile of weight for GA having 4 times greater odds of severe ROP than those between the 25th and 75th percentiles. Male gender was also a significant risk factor (odds ratio: 1.73; 95% confidence interval: 1.25–2.40).
Conclusions. These data, for a large, essentially population-based cohort, suggest that factors related to the degree of immaturity, intrauterine growth restriction, and male gender contribute to severe ROP.
Retinopathy of prematurity (ROP) remains a significant cause of morbidity among very preterm infants. Although there is some evidence that the incidence of the more severe stages of acute ROP (stage 3 or more) may be decreasing in the more developed world,1–6 other reports, particularly those concerning the most immature infants, show no decrease or even an increase in incidence.7,8 During the same period, there has been a documented increase in less developed countries, reaching almost epidemic proportions.9 Although many factors have been suggested to have some causal association with ROP, increasing prematurity, low birth weight, and prolonged exposure to supplementary oxygen are the most consistently identified.10,11 Most studies have been unit based or from hospital groupings, rather than being population based, and have involved cohorts of infants identified by birth weight; both factors may introduce bias into the results.
The Australian and New Zealand Neonatal Network (ANZNN) consists of all 29 level III NICUs in Australia and New Zealand. Since 1995, a dataset of 60 perinatal variables has been collected prospectively for all infants with gestational age (GA) of <32 weeks or birth weight of <1500 g who were admitted to a NICU. Both countries have a regionalized system of neonatal intensive care, meaning that the ANZNN dataset for surviving infants with GA of <29 weeks represents essentially population-based data. The aim of this study was to analyze this dataset for the 2-year period of 1998–1999, to identify significant risk factors for severe (stage 3 or more) ROP. The identified risk factors were then verified with a similar dataset for infants born in 2000–2001.
Data were collected prospectively as part of the ongoing ANZNN audit of high-risk infants admitted to neonatal nurseries.12 The database uses agreed-upon definitions authorized by the ANZNN Advisory Committee. GA is completed weeks assessed from early obstetric ultrasound scans, the first day of the last menstrual period, or clinical assessment of the infant. Weight-for-GA categories were determined from the Australian national birth weight percentile charts reported by Roberts and Lancaster.13 Birth weight-for-GA percentiles were divided into categories and coded as follows, to test for trend with the Mantel-Haenszel test: 0, >75th; 1, 25th to 75th (reference group); 2, 10th to 24th; 3, 3rd to 9th; 4, <3rd. Symmetry around the reference category was explored with divisions at the 75th, 90th, and 97th percentiles, but these were collapsed into 1 large-for-GA (>75th percentile) grouping because of similar parameter estimates. In testing for a linear trend in the effect of GA, codes 1, 2, 3, and 4 represented weeks 28, 27, 26, and 25, respectively, and ≤24 weeks was coded as 5.
Current recommended criteria for screening for acute ROP in Australia are GA of <32 weeks or birth weight of <1500 g14 and in New Zealand are GA of <31 weeks or birth weight of <1250 g.15 Screening is conducted by experienced ophthalmologists with indirect ophthalmoscopy, findings are reported according to international criteria,16 and eyes are examined until fully vascularized. Because some units in the ANZNN use the criterion of birth weight of <1250 g for screening for ROP and to avoid a bias toward small-for-GA (SGA) infants, the model was developed for the subset of infants born at <29 weeks’ GA between January 1, 1998, and December 31, 1999 (n = 2830), who survived to 36 weeks’ postmenstrual age (nonsurvivors: n = 544; 19.2%). Infants for whom no eye examination results were recorded were also excluded from the main analysis (n = 181, 8.0%), leaving 2105 infants.
Analysis of 21 perinatal variables up to 1 minute after birth was undertaken to formulate a predictive model for severe ROP. The variables identified as significant at P < .05 in univariate analyses were entered into a multivariate logistic regression model, and the least significant variables were removed sequentially.
Variables were retained in the adjusted model if they were significant at P < .01. To facilitate the identification of potential colinearity or confounding among the predictors, a process of refitting, with individual covariates deleted and verified at each stage, was performed. The stability of the estimated coefficients was also checked between the univariate and multivariate results.
The fit of the models was checked with the Hosmer-Lemeshow goodness-of-fit statistic,17 with additional verification that the models were not overfitted (indicated by very high P values). The discriminatory ability of the models was assessed with the area under the receiver operating characteristic (ROC) curve. The models were developed for the 1998–1999 cohort, and a temporal validation was performed by fitting a model with the same predictor variables to the equivalent 2000–2001 cohort (that is, infants of <29 weeks’ GA, registered with ANZNN in 2000–2001, who survived to 36 weeks’ postmenstrual age and were examined for ROP). Sensitivity analyses were performed to check the validity of excluding patients who were not examined for ROP, with best-case and worst-case scenarios. For the worst-case scenario, patients who were not examined for ROP were treated as having severe ROP (stage 3 or 4); for the best-case scenario, they were treated as being without severe ROP.
All analyses were performed with SAS statistical software, version 8.2 (SAS Institute, Cary, NC), and the process adhered to strict confidentiality guidelines. Approval for the project was given by the Ethics Review Committee of the Royal Prince Alfred Hospital and the Human Ethics Committee of the administering institution, the University of Sydney.
There were 2105 infants born in 1998–1999 who were eligible for inclusion in the study. The median birth weight of the cohort used in the analysis was 926 g (interquartile range [IQR]: 760–1096 g), and the median GA was 27 weeks (IQR: 26–28 weeks). The median birth weight for the validation cohort (born in 2000–2001) was 930 g (IQR: 756–1100 g), with a median GA of 27 weeks (IQR: 25–28 weeks). The male/female ratios were also similar in the 2 periods, ie, 1.2:1 in 1998–1999 and 1.1:1 in 2000–2001. Stage 3 or more ROP (highest stage in either eye) occurred for 203 (9.6%) infants in the 1998–1999 cohort.
The following variables were not related significantly to ROP in univariate analyses (Table 1): transferred into the tertiary unit or outborn, previous premature birth, previous perinatal death, hypertension during pregnancy (new episode of or existing hypertension), prepartum hemorrhage, preterm labor, fetal distress requiring obstetric intervention, maternal age group, birth order, plurality, any prenatal corticosteroids, maternal ethnicity, and presentation at birth. Within the ethnicity category, infants born to Asian mothers had an increased risk (odds ratio [OR]: 1.83; 95% confidence interval [CI]: 1.06–3.16), but this did not persist in the adjusted analysis. The unadjusted protective association for infants with breech presentation was statistically significant; however, because breech presentation was strongly associated with cesarean section without labor (37% of those with breech presentation had cesarean section without labor, compared with 17% of those with cephalic presentation) and presentation was not recorded for a large proportion of infants, the method of birth was retained for multivariate analyses and presentation was excluded.
Significant unadjusted risk for severe ROP was associated with younger GA, lower birth weight for GA, male gender, Apgar score at 1 minute of <4, and prolonged rupture of membranes of >48 hours (Table 2). Prelabor preterm rupture of membranes and cesarean section with no labor showed protective associations. In multivariate analyses, method of birth and prolonged and prelabor preterm rupture of membranes did not retain significance. Low Apgar score at 1 minute remained significant in multivariate analyses (P < .0001) but was excluded because of its strong relationship with GA (the proportion with 1-minute Apgar scores of <4 increased steadily from 17% at 28 weeks to 44% at 24 weeks; trend χ12 = 81.3, P < .0001) and the subjectivity of its measurement.
Multivariate analyses retained the variables GA, weight for GA, and gender, which were significant at P < .01 after simultaneous adjustment. The final model in Table 3 includes the 2105 infants with complete data for these 3 variables. GA was the dominant risk factor for severe ROP. The risk of severe ROP increased with decreasing GA (trend χ12 = 156.6, P < .0001), as shown in Fig 1. A “dose-response” effect was also apparent for weight for GA (trend χ12 = 28.8, P < .0001), indicating that, with lower weight at a given GA, there was greater risk of severe ROP. Male gender was also a significant risk factor (OR: 1.73; 95% CI: 1.25–2.40; χ12 = 10.7, P = .001) (Fig 1).
The model with 3 predictors validated well under the best-case scenario, assuming that patients not examined did not have severe ROP, with all 3 predictors being significant at P < .01 and with minimal impact on the estimated coefficients. Under the worst-case scenario, the variables weight for GA (P = .02) and gender (P = .01) did not retain significance in the multivariate model. Examination of the characteristics of infants not examined for ROP (n = 181) revealed that the majority of infants were larger and older, with 82% weighing at or beyond the IQR and 75% born at or beyond 27 weeks’ GA. These infants are less likely to have severe ROP, which provides a justification for the best-case scenario.
The predictive model from Table 3 yielded an area under the ROC curve of 0.80, indicating excellent discrimination, and the Hosmer-Lemeshow statistic indicated a good model fit (P = .37). The temporal stability also validated well for 2000–2001, with excellent discrimination (area under the ROC curve: 0.82) and goodness of fit (P = .23). Table 4 shows the details of the logistic regression model fitted to the full 4-year dataset.
We have reported data for a large cohort of infants of <29 weeks’ GA, which represent essentially population-based data for 2 countries, Australia and New Zealand. Birth registrations in both countries in 1995 identified 1.2% of infants born alive at <29 weeks who were not registered with ANZNN. In addition, since 1998, all New Zealand level II units have contributed data to ANZNN,5 and 1.3% of infants of <29 weeks’ GA who survived to 36 weeks’ postmenstrual age were registered to these units, rather than the level III NICUs included in this study. Although a few surviving infants who were admitted to a level III NICU were not examined for ROP, they represented a small proportion of all births. Because the majority of those infants were larger and more mature, assuming that those infants did not have severe ROP seems justifiable; this was verified by the minimal impact on the estimated coefficients under the best-case scenario. Both the size of the study and the fact that it was essentially population-based mean that biases inherent in some other etiologic studies are minimized.
We confirmed that prematurity is the most important risk factor for ROP, and we identified low birth weight for GA and male gender as significant risk factors. Virtually all studies of risk factors for ROP, whether unit based or population based, identified a measure of immaturity or the infant's size as having the greatest association with risk of ROP.10,11
In the first decade after the first description of ROP (then called retrolental fibroplasia) in 1942, the disease was noted more often among male infants than female infants.18,19 Since that time, however, few epidemiologic studies have identified gender as an independent risk factor for ROP, although 2 small studies20,21 did report a doubling of risk for male infants. A small case-control study that compared 8 infants with stage 2 ROP with 8 infants with stage 3 or 4 ROP also found that male gender was associated significantly with an increased risk of progression beyond stage 2.22 Being female more than doubled the chances of survival at extreme prematurity in a population-based cohort study from Australia.23 Similarly, Synnes et al,24 reporting on a large, institution-based cohort of infants born at 23 to 28 weeks’ GA, noted female infants to have a survival advantage equivalent to 1 additional week of gestation. Draper et al25 reported outcomes for all very preterm births in a United Kingdom health region over 4 years and confirmed a survival advantage for female infants. The explanation for the better outcomes for female infants is unclear, but the difference has been suggested to result from a different hormonal milieu associated with increased organ maturation, compared with male infants of the same GA.26 Our findings of a significantly increased risk of ROP for male infants, or a decreased risk for female infants, add weight to such theories.
It has long been evident that mortality rates at each GA are lower for infants with greater birth weights.24,25,27–29 An exception to this is that, in some series, infants who were significantly large for GA, and might have had conditions such as hydrops, had higher mortality rates.24 Our data included 3 infants with hydrops, 2 in the 1998–1999 cohort. These infants were very preterm and did not have a birth weight beyond the 75th percentile. The relationships between birth weight and GA and morbidity are less clear, in part because many reports on morbidity included cohorts selected on the basis of birth weight alone. In comparisons of infants on the basis of birth weight, it is perhaps not surprising that more mature but SGA infants have better survival rates.30 Now that cohorts are more commonly based on GA at birth, it is clear that growth-restricted infants have increased mortality24,31,32 and morbidity33–36 rates. However, there have been few reports on the impact of growth restriction on the development of ROP.
Bardin et al34 compared infants of GA of <27 weeks who were either SGA (birth weight of ≤3rd percentile for GA) or not, in an historical, institution-based cohort from Canada. SGA infants had greater risks of any ROP (90% vs 58%, P < .01) and stage 3 or 4 ROP (65% vs 12%, P < .01). Regev et al,37 reporting from the Israeli national very low birth weight (birth weight of ≤1500 g) database, confined analyses to infants of <32 weeks’ GA at birth. After adjustment for perinatal risk factors, infants with birth weights >2 SD below the mean for GA had twice the risk of stage 3 or 4 ROP, compared with non—growth-restricted infants. Allegaert et al38 reported on a case-control study, from a single European institution, of infants with threshold ROP, compared with those of the same GA without threshold disease. In a multivariate regression model, birth weight of <10th percentile for GA (SGA) and birth weight of <25th percentile for GA (growth restriction) were both associated with increased risks of threshold ROP (relative risk: 3.7 and 4.5, respectively). Those authors noted that others39 reported an association between poor postnatal weight gain and an increased risk of severe ROP and that, in the rat model of ROP, postnatal growth restriction increased the risk of retinal neovascularization. In their study, however, growth-restricted infants retained a greater risk of threshold ROP even if their postnatal weight gain was similar to that of control infants, which suggested that prenatal growth restriction was the more important factor. Zaw et al40 reported on a hospital-based cohort of infants with GA of <34 weeks who were born between 1993 and 2001. SGA was defined as weight of <10th percentile, according to Canadian neonatal birth weight standards or US fetal growth standards. Compared with appropriate-for-GA infants, SGA status was associated with an increased risk of ROP, whether neonatal (OR: 3.88; 95% CI: 2.33–6.48) or fetal (OR: 5.38; 95% CI: 2.87–10.90) standards were used.
In our large, population-based cohort, we observed a “dose-response” relationship between lower birth weight for GA and increased risk of ROP (trend P < .0001), rather than only SGA infants being at increased risk. Others34 speculated that several factors, including chronic intrauterine hypoxia, antioxidant deficiency, and abnormal growth factor levels might contribute to an increased risk of ROP among SGA infants. Our results suggest that the hormonal milieu related to growth in utero might have a crucial role. Recent studies in animal models demonstrated the role of oxygen-regulated vascular endothelial growth factor in retinal vasculogenesis (reviewed by Smith41 and Madan42). The non—oxygen-regulated hormones growth hormone and particularly insulin-like growth factor-1 (IGF-1) also are critical to normal vascular development,43 with IGF-1 controlling vascular endothelial growth factor activation. Hikino et al44 found that serum IGF-1 concentrations remained low after birth among infants born at <28 weeks’ GA but increased among more mature infants. Hellström et al45 measured prospectively serum IGF-1 concentrations weekly among 84 preterm infants. The mean serum IGF-1 concentration at 30 to 33 weeks’ postmenstrual age was lowest among infants with severe ROP, compared with moderate or no ROP, and the duration of low IGF-1 concentrations was strongly correlated with the severity of ROP. Additional investigations into the role of IGF-1 and other factors in normal and abnormal vasculogenesis among very preterm, and particularly growth-retarded, infants are warranted.
We investigated population-based data for a large cohort of infants of <29 weeks’ GA, to identify prenatal and perinatal risk factors, up to 1 minute of age, for severe (stage 3 or 4) ROP. In multivariate analyses, there was a significantly increased risk of severe ROP with decreasing GA, a “dose-response” effect of birth weight for GA, such that more growth-restricted infants exhibited greater risk of ROP, and a significantly increased risk associated with male gender. These data suggest that factors related to the degree of immaturity, intrauterine growth restriction, and male gender contribute to severe ROP.
This work was supported by the National Health and Medical Research Council of Australia (grant 211088 to D.J.H.-S., D.A.D., J.M.S., and N.J.E.) and the Centre for Perinatal Health Services Research, University of Sydney, Sydney, Australia.
The ANZNN Advisory Committee and Executive members were as follows: Australia: Centre for Perinatal Health Services Research (New South Wales): David Henderson-Smart and Deborah Donoghue; Flinders Medical Centre (South Australia): Peter Marshall; John Hunter Hospital (New South Wales): Chris Wake; King Edward Memorial and Princess Margaret Hospitals (Western Australia): Noel French, Ron Hagan, and Karen Simmer; Launceston General Hospital (Tasmania): Chris Bailey; Liverpool Health Service (New South Wales): Robert Guaran; Mater Mother's Hospital (Queensland): David Tudehope; Mercy Hospital for Women (Victoria): Andrew Watkins; Monash Medical Centre (Victoria): Kaye Bawden, Andrew Ramsden, and Victor Yu; National Perinatal Statistics Unit (New South Wales): Paul Lancaster; Nepean Hospital (New South Wales): Lyn Downe; Newborn Emergency Transport Service (Victoria): Michael Stewart; New South Wales Newborn and Pediatric Emergency Transport Service: Andrew Berry; Perinatal Research Centre (Queensland): Paul Colditz; Royal Children's Hospital (Victoria): Linda Johnstone and Peter McDougall; Royal Darwin Hospital (Northern Territory): Charles Kilburn; Royal Hobart Hospital (Tasmania): Peter Dargaville; Royal Hospital for Women (New South Wales): Kei Lui; Royal North Shore Hospital (New South Wales): Jennifer Bowen; Royal Prince Alfred Hospital (New South Wales): Nick Evans; Royal Women's Hospital (Queensland): David Cartwright; Royal Women's Hospital (Victoria): Lex Doyle, Colin Morley, and Neil Roy; Sydney Children's Hospital (New South Wales): Barry Duffy; Canberra Hospital (Australian Capital Territory): Graham Reynolds; Children's Hospital at Westmead (New South Wales): Robert Halliday; Townsville Hospital (Queensland): John Whitehall; Western Australia Neonatal Transport Service: Jenni Sokol; Westmead Hospital (New South Wales): William Tarnow-Mordi; Women's and Children's Hospital (South Australia): Ross Haslam; New Zealand: Christchurch Women's Hospital: Nicola Austin; Christchurch School of Medicine: Brian Darlow; Dunedin Hospital: Roland Broadbent; Gisborne Hospital: Graeme Lear; Hastings Hospital: Jenny Corban; Hutt Hospital: Robyn Shaw; Middlemore Hospital: Lindsay Mildenhall; National Women's Hospital: Carl Kushell; Nelson Hospital: Peter McIlroy; Palmerston North Hospital: Jeff Brown; Rotorua Hospital: Stephen Bradley; Southland Hospital: Paul Tomlinson; Taranaki Hospital: John Doran; Tauranga Hospital: Hugh Lees; Timaru Hospital: Philip Morrison; University of Auckland: Jane Harding; Waikato Hospital: David Bourchier; Wairau Hospital: Ken Dawson; Wanganui Hospital: John Goldsmith; Wellington Women's Hospital: Vaughan Richardson; Whakatane Hospital: Chris Moyes; Whangarei Hospital: Peter Jankowitz.
We thank Dr David Todd and Professor Lex Doyle, for helpful suggestions on the manuscript, and the directors of the NICUs of the ANZNN.
- Accepted August 23, 2004.
- Address correspondence to Brian A. Darlow, MD, FRACP, Department of Paediatrics, Christchurch School of Medicine and Health Sciences, PO Box 4345, Christchurch, New Zealand. E-mail:
No conflict of interest declared.
- ↵Hussain N, Clive J, Bhandari V. Current incidence of retinopathy of prematurity, 1989–1997. Pediatrics.1999;104(3) . Available at: www.pediatrics.org/cgi/content/full/104/3/e26
- Bullard SR, Donahue SP, Feman SS, Sinatra RB, Walsh WF. The decreasing incidence and severity of retinopathy of prematurity. JAAPOS.1999;3 :46– 52
- Rowlands E, Ionides AC, Chinn S, Mackinnon H, Davey CC. Reduced incidence of retinopathy of prematurity. Br J Ophthalmol.2001;85 :933– 935
- Blair BM, O'Halloran HS, Pauly TH, Stevens JL. Decreased incidence of retinopathy of prematurity, 1995–1997. JAAPOS.2001;5 :118– 122
- ↵Chow LC, Wright KW, Sola A. Can changes in clinical practice decrease the incidence of severe retinopathy of prematurity in very low birth weight infants? Pediatrics.2003;111 :339– 345
- ↵Allegaert K, de Coen K, Devlieger H, EpiBel Study Group. Threshold retinopathy at threshold of viability: the EpiBel study. Br J Ophthalmol.2004;88 :239– 242
- ↵Donoghue DA, Australian and New Zealand Neonatal Network. Report of the Australian and New Zealand Neonatal Network 2001. Sydney, Australia: Australian and New Zealand Neonatal Network; 2003
- ↵National Health and Medical Research Council. Clinical Practice Guidelines: Care Around Preterm Birth. Canberra, Australia: Australian Government Publishing Service; 1997
- ↵Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd ed. New York, NY: John Wiley and Sons; 2000
- ↵Wheatley CM, Dickinson JL, Mackey DA, Craig JE, Sale MM. Retinopathy of prematurity: recent advances in our understanding. Br J Ophthalmol.2002;86 :696– 700
- ↵Doyle L. Outcome at 5 years of age of children 23 to 27 weeks' gestation: refining the prognosis. Pediatrics.2001;108 :134– 141
- ↵Draper ES, Manktelow B, Field DJ, James D. Prediction of survival for preterm births by weight and gestational age: retrospective population based study. Br Med J.1999;319 :1093– 1097
- ↵Ingemarsson I. Gender aspects of preterm birth. Br J Obstet Gynaecol.2003;110(suppl 20) :34– 38
- ↵Horbar JD, Badger GJ, Lewit EM, Rogowski J, Shiono PH. Hospital and patient characteristics associated with variation in 28-day mortality rates for very low birth weight infants. Pediatrics.1997;99 :149– 156
- ↵Beeby PJ. Risk of death for small for gestational age very preterm infants. Pediatrics.1998;102 :1497– 1498
- ↵Tyson JE, Kennedy K, Broyles S, Rosenfeld CR. The small for gestational age infant: accelerated or delayed pulmonary maturation? Increased or decreased survival? Pediatrics.1995;95 :534– 538
- ↵Bardin C, Zelkowitz P, Papageorgiou A. Outcome of small-for-gestational age and appropriate-for-gestational age infants born before 27 weeks of gestation. Pediatrics.1997;100(2) . Available at: www.pediatrics.org/cgi/content/full/100/2/e4
- ↵Allegaert K, Vanhole C, Casteels I, et al. Perinatal growth characteristics and associated risk of developing threshold retinopathy of prematurity. JAAPOS.2003;7 :4– 7
- ↵Wallace DK, Kylstra JA, Philips SJ, Hall JG. Poor postnatal weight gain: a risk factor for severe retinopathy of prematurity. JAAPOS.2000;4 :343– 347
- ↵Zaw W, Gagnon R, da Silva O. The risks of adverse neonatal outcome among preterm small for gestational age infants according to neonatal growth standards. Pediatrics.2003;111 :1273– 1277
- ↵Smith LEH. Pathogenesis of retinopathy of prematurity. Acta Paediatr.2002;437(suppl) :26– 28
- ↵Hellström A, Engstrom E, Hard AL, et al. Postnatal serum insulin-like growth factor 1 deficiency is associated with retinopathy of prematurity and other complications of premature birth. Pediatrics.2003;112 :1016– 1020
- Copyright © 2005 by the American Academy of Pediatrics