Cumulative illness severity and progression from moderate to severe retinopathy of prematurity

Transcription

1 (2007) 27, r 2007 Nature Publishing Group All rights reserved /07 $30 ORIGINAL ARTICLE Cumulative illness severity and progression from moderate to severe retinopathy of prematurity JI Hagadorn 1, DK Richardson 2,{, CH Schmid 3 and CH Cole 2 1 Division of Neonatology, Department of Pediatrics, Connecticut Children s Medical Center, University of Connecticut School of Medicine, Hartford, CT, USA; 2 Department of Neonatology, Beth Israel-Deaconess Medical Center, Harvard University School of Medicine, Boston, MA, USA and 3 Institute for Clinical Research and Health Policy Studies, Tufts-New England Medical Center, Tufts University School of Medicine, Boston, MA, USA Objective: To test cumulative neonatal illness severity (IS) and IS fluctuation as predictors of progression from moderate to severe retinopathy of prematurity (ROP). Study Design: Data from research databases and medical record review were collected for infants from four neonatal intensive care unit (NICUs) admitted between 1995 and 2001 and diagnosed with prethreshold ROP. Cumulative neonatal IS measured using daily Scores for Neonatal Acute Physiology (SNAP) for the first 28 days of life, and IS fluctuation as assessed by summing changes between daily SNAP scores, were tested as predictors of progression to threshold ROP using logistic regression. Result: Infants progressing to threshold (n ¼ 79), compared to those not progressing to threshold (n ¼ 130), had significantly (P<0.05) lower gestational ages (25.2±1.1 versus 25.8±1.4 weeks), higher cumulative neonatal SNAP (255±77 versus 224±63 weeks) and had more severe hospitalizations as indicated by diagnoses and medical management. In regression analysis, gestational age, chronological age and presence of plus disease at first diagnosis of prethreshold were associated with development of threshold. After adjusting for these factors, cumulative neonatal SNAP was significantly associated with progression to threshold. However, addition of cumulative SNAP to the model only increased receiver-operating characteristic curve area from 0.77 to 0.78 (NS). Other factors, including SNAP fluctuation, were not associated with progression to threshold after adjustment using this model. Conclusion: Cumulative neonatal IS, as measured by cumulative SNAP, is an independent risk factor for progression from moderate to severe ROP. However, cumulative IS does not enhance assessment of risk for ROP progression after adjusting for simpler clinical factors. (2007) 27, ; doi: /sj.jp ; published online 14 June 2007 Keywords: retinopathy of prematurity; illness severity; predictive model Correspondence: Dr James I Hagadorn, Division of Neonatology, Connecticut Children s Medical Center, 282 Washington St, Hartford, CT 06106, USA. jhagadorn@ccmckids.org { Deceased Received 25 January 2007; revised 23 April 2007; accepted 11 May 2007; published online 14 June 2007 Introduction Retinopathy of prematurity (ROP) is a common disorder of premature infants characterized by abnormal retinal vascular proliferation. Its incidence and severity are highest among infants born most prematurely, decreasing with advancing gestational age at birth. 1 While most cases regress with resumption of normal retinal development, ROP may progress to severe retinal vascular proliferation and subsequent visual compromise with or without retinal detachment. The mechanisms that determine whether ROP will resolve or progress are under active investigation. Refinement of risk models for identifying infants at greatest risk for progression to severe ROP may facilitate evaluation of experimental interventions and influence clinical management. Published ROP risk models have seldom adjusted for variations in severity of hospital course using a standardized scoring method. However, clinical observations and studies of ROP risk factors indicate that critically ill premature infants are at increased risk for developing severe ROP. 2,3 The strong associations reported between neonatal diagnoses or therapies and subsequent ROP suggest that severe illness, or an unstable fluctuating hospital course in general, are associated with risk for developing severe ROP Standardized illness severity (IS) measurements are widely used to account for variations in neonatal outcomes among hospitals. 15 The Score for Neonatal Acute Physiology (SNAP score) 16 on the first day of life appears associated with development of any ROP 17 and severe ROP. 18 Summation of serial illness severity scores, or cumulative illness severity, a measure of sustained exposure to physiological instability, was used in two studies of premature infants to account for differences in neurodevelopmental and cognitive outcomes. 19,20 No study has explored the relationship between cumulative illness severity and development of severe ROP. Moreover, fluctuation of illness severity scores over time has not been explored as a possible risk factor for ROP. Cumulative illness severity and illness severity fluctuation, as events occurring over a period of weeks, may be helpful in assessing risk for subsequent progression of retinal disease.

2 Cumulative illness severity and ROP 503 This study tested the hypothesis that cumulative neonatal illness severity and fluctuation of illness severity, as measured with SNAP scores, are significant risk factors for subsequent progression of moderate ROP to severe ROP. The specific aims testing this hypothesis were (1) to evaluate the association of cumulative illness severity in the first 28 days of life with subsequent progression of moderate to severe ROP; (2) to evaluate the association of illness severity score fluctuations in the first 28 days of life with subsequent progression of moderate to severe ROP; and (3) to evaluate the improvement in the performance of a risk model for progression of moderate to severe ROP produced by inclusion of cumulative illness severity and illness severity fluctuation as risk factors. Methods Ethics Institutional Review Board approval was obtained at all participating institutions. Study population This multicenter observational cohort study used existing prospectively collected databases supplemented with medical record review. Infants were included if they were admitted on the first day of life to any of four level III NICUs between 1 January 1995 and 31 December 2001, and subsequently were diagnosed with moderate or severe ROP. The four participating centers (The Floating Hospital for Children/Tufts-New England Medical Center, Beth Israel Deaconess Medical Center, Brigham and Women s Hospital, and The Children s Hospital) comprised the Boston Consortium of the Supplemental Therapeutic Oxygen for Prethreshold Retinopathy of Prematurity (STOP ROP) trial. 21 STOP ROP definitions of prethreshold and threshold ROP (Table 1) were used in the current study to identify moderate and severe ROP, as these definitions guided management of ROP at the participating hospitals during the study period. Subjects were identified through a prospectively managed research database, which contained results of all eye examinations performed on all premature infants in the neonatal intensive care units at the four study hospitals. Ascertainment was verified with written logs. All subjects had sufficient medical documentation to calculate daily SNAP scores for the first 28 days of life. For multiple gestations, all siblings who developed prethreshold ROP were included. This study tested cumulative neonatal illness severity and illness severity fluctuation as predictors of subsequent ROP progression at the time of first prethreshold ROP diagnosis. Infants were therefore excluded if they progressed rapidly to threshold ROP without a preceding diagnosis of prethreshold. One eye with prethreshold ROP per subject was selected for analysis. If prethreshold ROP initially developed unilaterally, the eye in which it was first diagnosed was included. When Table 1 ICROP criteria used to define prethreshold and threshold ROP, from the STOP ROP Trial 21 Prethreshold In zone 1 In zone 2 Threshold In zone 1 In zone 2 Any ROP not meeting criteria for threshold Any stage 3 less than threshold severity OR Any stage 2 with plus disease Any ROP with plus disease OR Any stage 3 Five contiguous or 8 cumulative clock hours of stage 3 with plus disease Abbreviations: ICROP, international classification of ROP; ROP, retinopathy of prematurity; STOP ROP, supplemental therapeutic oxygen for prethreshold retinopathy of prematurity. prethreshold ROP was present bilaterally, one eye was selected randomly for inclusion. The primary outcome was development of threshold ROP using STOP ROP criteria. 21 Infants who received retinal ablation for rapidly progressive prethreshold ROP just short of threshold criteria were included in the threshold group. Threshold ROP was considered not to have developed if threshold ROP was not observed on any eye examination, and if prethreshold ROP regressed with (1) mature retinal vascularization; (2) extension of retinal vessels into zone 3; or (3) regression of prethreshold ROP to less than prethreshold on two consecutive examinations 1 week apart. Infants with prethreshold ROP who were lost to follow-up before reaching definitive outcome were assumed not to have developed threshold, but analyses were conducted with and without inclusion of these infants to test this assumption. Data collection Eye examination results were obtained from the database described above. Examinations were performed by experienced, STOP ROP-certified ophthalmologists using the International Classification of ROP guidelines. 22 These data were verified during medical record review. Infants discharged or transferred to other hospitals before final retinal outcome was reached were tracked by written log and telephone follow-up as part of data collection for the database, with eye examinations performed at outside hospitals or outpatient clinics obtained by facsimile. STOP ROP enrollment status and treatment group assignment, if any, were collected during record review and verified through examination of STOP ROP trial logs. Medical record review and data entry were performed by research associates who used standardized data manuals and definitions and were trained by a single investigator. Research assistants were unaware of ophthalmic outcomes. For

3 504 Cumulative illness severity and ROP all subjects, data were collected regarding patient characteristics, perinatal and postnatal events, diagnoses, clinical management and illness severity. Data sufficient to calculate the SNAP score daily were collected for the first 28 days of each subject s hospitalization. The physiology-based SNAP score is calculated from 34 biological parameters including vital signs and laboratory values; each parameter is assigned a point value based on the severity of its departure from physiological normal. 16 Cumulative SNAP score for the first 28 days of life (SNAP28) was calculated by addition of daily SNAP scores for each subject, using a modification of the method described by Mattia and deregnier. 19 Thus, an infant with sequential daily scores of 20, 5, 17, 5 and 23 would have a 5-day cumulative score of 70. As described previously, 19 SNAP points for individual parameters were interpolated for up to 3 days where physiologically appropriate, if new values were not recorded. Thus, if the subject had a hematocrit meriting a daily SNAP score of 1 point on hospital day 20, was not transfused and no subsequent values were obtained, 1 SNAP point was assigned for hematocrit on hospital days 20, 21 and 22. This approach was used for hematocrit, serum sodium, potassium, calcium, ionized calcium, direct bilirubin and bicarbonate measurements, as well as for white blood cell and platelet counts and neutrophil immature/ total ratios. Because not all components of the SNAP score are ascertained daily during routine clinical care, parameters (other than when interpolated as described above) were assumed to be normal if not evaluated by the infant s clinical caretakers. This approach is well validated with this method of illness severity scoring. 16 For days when no arterial blood gas was obtained for infants on mechanical ventilator, points were added to the daily SNAP score for maximum and minimum fraction of inspired oxygen (FiO 2 ), using the scales for these parameters in the Clinical Risk Index for Babies score. 23 For infants with no arterial blood gas who were not on mechanical ventilation, but were still on supplemental oxygen, one point was assigned for maximum FiO 2 greater than 0.7 and for minimum FiO 2 greater than 0.6. Fluctuation of illness severity was quantified by summing the absolute value of changes in sequential daily SNAP scores. For example, for SNAP scores of 20 on day 1 and 5 on day 2, the fluctuation score would be 15 ¼ 15. For sequential daily SNAP scores of 20, 5, 17, 5 and 23 (cumulative SNAP score ¼ 70), the cumulative fluctuation score would be 15 þ 12 þ 12 þ 18 ¼ 57. Statistics Subjects and eyes were described with respect to patient, clinical and ophthalmic characteristics. Infants or eyes with prethreshold ROP not progressing to threshold were compared to those developing threshold using univariate analyses with w 2 with Yates correction or Fisher s exact tests as appropriate for categorical variables, and the independent samples t-test or Mann Whitney U test for continuous variables. Multivariable logistic regression modeling. Selected risk factors were designated a priori as high priority for regression modeling, to test study hypotheses while adjusting for factors found to be significant in previous large trials. 21,24,25 This First Priority Group was intentionally limited to eight factors to reduce risk of overfitting the model to the data set, 26 and included gestational age, birth weight, STOP ROP treatment group, vessels in zone 1 at first ROP diagnosis, chronological age at first prethreshold diagnosis, plus disease present at first prethreshold diagnosis and two factors testing study hypotheses (SNAP28, cumulative fluctuation from the first 28 days of life). These factors were tested using backward selection with a significance level of P ¼ Other potentially important risk factors were designated a priori to have lower likelihood of association with outcome. This larger Second Priority Group was not limited to a specified number of covariates. However, to avoid overfitting, its variables were tested at P ¼ All associations were described using odds ratios (OR) with confidence intervals (CI). Model discrimination was measured using ROC curve areas. 27 Alternative models with and without adjustment for cumulative illness severity and fluctuation of illness severity were compared directly using the drop-in-deviance w 2, and by comparison of ROC curve areas. 28 We repeated regression modeling without inclusion of infants with prethreshold ROP who were lost to follow-up before definitive ophthalmic outcome, to ascertain the impact of these infants upon results. Results Two hundred and fifty-one infants were admitted to the participating hospitals during the study period and developed prethreshold ROP in at least one eye. Of these, 24 did not meet all inclusion criteria and were excluded: 17 admitted after the first day of life, 1 discharged before day 28 and 6 missing large portions of their medical record. Threshold ROP developed without a preceding prethreshold diagnosis in 18, who were also excluded. For the remaining 209 infants, prethreshold ROP was diagnosed in at least one eye, with no threshold ROP in either eye at the time of first diagnosis of prethreshold disease. This group of 209 infants constituted the cohort used for this analysis. The 18 infants with threshold ROP who were excluded were not detectably different from the 209 infants in the study cohort as a whole with respect to birth weight, gestational age, gender, race, SNAP score on first day of life, cumulative neonatal SNAP scores, or occurrence of examined diagnoses and interventions. Excluded infants were slightly more mature at the time of first retina exam than were infants with prethreshold ROP progressing to threshold (median postmenstrual age (PMA) ¼ 31.1 versus 31.0 weeks,

4 Cumulative illness severity and ROP 505 P ¼ 0.067) and were clearly more mature at first diagnosis of ROP (PMA ¼ 34.5 weeks versus 32.9 weeks, P<0.001). The latter was due to different chronological ages at first diagnosis of ROP (64 days versus 56 days, P ¼ 0.001), rather than different gestational ages at birth. ROP was diagnosed in zone 1 more frequently in excluded infants, but this difference was not statistically significant. Among the 209 infants included in the study cohort, prethreshold ROP was first diagnosed in the right eye in 50 infants, in the left eye in 29 infants and bilaterally in 130 infants. For the group with bilateral prethreshold at first diagnosis, random selection designated 67 right eyes and 63 left eyes for analysis. Analyses were thus performed on 209 independent eyes with prethreshold ROP. Threshold ROP developed in 79. Univariate analysis Subjects who progressed to threshold ROP did not differ from those without progression with respect to gender, race, antenatal steroid exposure, route of delivery, incidence of multiple gestation or first day SNAP score. Infants progressing to threshold had significantly lower birth weights (median 710 g versus 740 g, P ¼ 0.02, Mann Whitney U) and shorter gestations (25 weeks versus 25.6 weeks, P ¼ 0.003). Ten (13%) were black race among infants progressing to threshold, 21 (16%) among those not developing threshold (NS). Regarding diagnoses and interventions in the first 28 days of life (Table 2), infants with threshold were more likely to require ductus arteriosus ligation. They received larger volumes of transfused red blood cells, and more days on mechanical ventilator. We did not detect a difference with respect to frequency or severity of intraventricular hemorrhage, days parenteral nutrition, or exposure to systemic steroids (dexamethasone or hydrocortisone) in the first 28 days. All received 1 to 4 doses of exogenous surfactant for respiratory distress syndrome. Illness severity data during the first 28 days of life are shown in Figure 1. For the whole cohort, median SNAP28 was 229 (range 83 to 458). Those progressing versus not progressing to threshold ROP had similar cumulative SNAP scores for the first 7 days of life, but those progressing to threshold had significantly higher scores thereafter. Median SNAP28 for the threshold group was 248 (interquartile range 192, 311) compared to 222 (180, 258) for those not progressing to threshold (P ¼ 0.004, Mann Whitney U). Fluctuation of daily SNAP scores in the first 28 days of life did not differ, with cumulative fluctuation 95 (interquartile range 81, 111) Figure 1 Cumulative illness severity scores during the first 28 days of life, by ophthalmic outcome. Table 2 Clinical characteristics of study cohort, first 28 days of life Prethreshold (n ¼ 130) Threshold (n ¼ 79) P* Any intraventricular hemorrhage 61 (47) 35 (44) 0.82 Intraventricular hemorrhage grade 3 or 4 or cerebral white matter damage (periventricular leukomalacia) 24 (18) 10 (13) 0.36 Patent ductus arteriosus, any clinically significant 83 (64) 60 (76) 0.09 Days indomethacin therapy, first 28 days 2 (0, 2) 2 (1, 3) 0.03 Patent ductus arteriosus, ligated 14 (11) 17 (22) 0.05 Parenteral nutrition, days, first 28 days 20.5 (15, 25) 22 (16, 26) 0.35 Enteral intake, ml, first 28 days 977 (259, 1480) 780 (151, 1307) 0.06 Phototherapy days, first 28 days 2 (0, 2) 2 (1, 3) 0.22 Oxygen, days, first 28 days 28 (27, 28) 28 (28, 28) 0.09 Continuous positive airway pressure, days, first 28 days 2 (0, 9) 0 (0, 4) 0.01 Any mechanical ventilation, days, first 28 days 25.5 (18, 28) 28 (27, 28) High-frequency ventilation, days, first 28 days 2 (0, 6) 6 (0, 13) 0.01 Systemic steroid therapy, days, first 28 days 3 (0, 14) 5 (0, 16.5) 0.17 Inhaled steroid therapy, days, first 28 days 3 (0, 14) 5 (0, 16.5) 0.29 Packed red blood cells transfused, cc, first 28 days 58 (43, 82) 63 (45, 100) 0.01 Continuous variables given as median (25th, 75th percentiles), categorial variables as n (%). *Mann Whitney U or w 2 test with Yates correction.

5 506 Cumulative illness severity and ROP for the threshold group compared to 91 (75, 106) for those not progressing to threshold (P ¼ 0.09). At first diagnosis of prethreshold ROP, infants progressing to threshold were more likely than those not progressing to be on supplemental oxygen (55 (70%) versus 62 (48%), P ¼ 0.003, w 2 ) and mechanical ventilation (11 (14%) versus 6 (5%), P ¼ 0.03). We did not detect differences between the two groups with regard to receipt of systemic steroids or general anesthesia before prethreshold, or occurrence of necrotizing enterocolitis, bacterial sepsis/meningitis or fungal sepsis before prethreshold.. Thirty-nine infants (19%) in the study cohort were enrolled in the STOP ROP trial. There was no detectable difference in enrollment status or treatment group assignment between those progressing versus not progressing to threshold. Ophthalmic characteristics at first prethreshold diagnosis are compared in Table 3 for eyes that did versus did not progress to threshold ROP. Prethreshold ROP was first diagnosed at between 37 and 176 days of age, and followed completion of the cumulative neonatal illness severity scoring period in all cases. Eyes progressing to threshold were diagnosed with ROP and with prethreshold ROP at earlier postmenstrual ages than eyes that did not progress to threshold. At first prethreshold diagnosis, eyes progressing to threshold ROP were more likely to have prethreshold ROP in the contralateral eye and were much more likely to have plus disease present at first prethreshold diagnosis. The two groups did not differ with respect to chronological age at the time of first examination, chronological age at first ROP diagnosis or proportion of zone 1 ROP or stage 3 ROP at first prethreshold diagnosis. Multivariable logistic regression modeling The risk assessment model based on a priori selected factors is shown in Table 4. Gestational age, the presence of plus disease at first prethreshold diagnosis and chronological age at prethreshold diagnosis were significantly associated with progression to threshold. The ROC curve area for a model based on these three factors was 0.77 (95% CI 0.71, 0.83). After adjusting for these factors, cumulative illness severity as measured with SNAP28 was significantly associated with progression of ROP from prethreshold to threshold. Each increase of one point in average Table 3 PMA, chronological age and ophthalmic characteristics of studied eyes at first diagnosis of prethreshold ROP Prethreshold (n ¼ 130) Threshold (n ¼ 79) P* PMA first retina exam (weeks) 31.6± ± PMA first diagnosis ROP (weeks) 33.6± ± PMA first diagnosis prethreshold (weeks) 36.3± ±1.9 <0.001 Chronological age first diagnosis ROP (weeks) 7.9± ± Chronological age first diagnosis prethreshold (weeks) 10.5± ± Interval from ROP to prethreshold (days) 15.5 (0, 28) 14 (0, 21) 0.08 Bilateral prethreshold 73 (56) 55 (70) 0.07 ROP in zone 1 at first diagnosis of prethreshold 25 (19) 16 (20) 1.0 Stage 3 ROP at first diagnosis of prethreshold 63 (48) 28 (35) 0.09 Plus disease present at first diagnosis prethreshold 39 (30) 50 (63) <0.001 Abbreviations: PMA, postmenstrual age; ROP, retinopathy of prematurity. Continuous variables given as mean±s.d. or median (25th, 75th percentiles), categorical variables given as n (%). PMA ¼ gestational age+chronological age. *Student s t-, Mann Whitney U or w 2 test (Yates correction for 2 2 comparisons). Table 4 Risk model for progression from prethreshold to threshold ROP OR 95% CI P Gestational age, per week Plus disease present at first diagnosis prethreshold <0.001 Chronological age at first diagnosis prethreshold, per week Cumulative daily SNAP scores, first 28 days, per point per day Abbreviations: CI, confidence intervals; OR, odds ratios; ROP, retinopathy of prematurity; SNAP, scores for neonatal acute physiology. N ¼ 209, with 79 threshold eyes. Second Priority Group factors tested: gender, black race, presence of grade 3 or 4 intraventricular hemorrhage (IVH) or cerebral white matter damage, presence of clinically significant patent ductus arteriosus (PDA), total volume of enteral feeds in first 28 days, days of mechanical ventilation in first 28 days, chronological age at first ROP diagnosis, interval in days between first ROP diagnosis and first prethreshold diagnosis, stage 3 ROP present at first prethreshold diagnosis, bilateral prethreshold ROP present at first prethreshold diagnosis, presence of necrotizing enterocolitis (NEC)Xstage 2 by Bell s criteria 42 before prethreshold diagnosis, any sepsis or meningitis before prethreshold diagnosis, any fungal sepsis before prethreshold diagnosis, on supplemental oxygen at diagnosis of prethreshold, or on mechanical ventilation at diagnosis of prethreshold.

6 Cumulative illness severity and ROP 507 daily SNAP score for the first 28 days increased odds of developing threshold by 15.9%. However, addition of SNAP28 to the regression model increased its ROC curve area insignificantly, from 0.77 to 0.78 (NS). After adjusting for the factors in this model, fluctuation of illness severity scores in the first 28 days of life was not significantly associated with progression to threshold. None of the Second Priority Group factors was significant at P ¼ 0.01 or P ¼ There were trends toward association between progression to threshold ROP and fungal sepsis before prethreshold (P ¼ 0.08) and requirement for mechanical ventilation at the time of prethreshold diagnosis (P ¼ 0.06). Regression modeling was repeated, excluding eyes lost to followup before developing a definitive retinal outcome. Nineteen infants developed prethreshold ROP and had neither threshold ROP nor significant improvement in retinal findings at the time of last documented examination. In addition, 28 subjects with prethreshold ROP showed regression to less than prethreshold severity on two consecutive examinations at least 1 week apart, but at the time of last examination had ROP in zone 2. Replication of regression modeling while excluding these infants resulted in no change in the significant factors in the model in Table 4. No Second Priority Group covariates were retained. Estimated relationships between significant risk factors and outcome differed minimally. When logistic regression analysis was repeated including the 18 infants diagnosed with threshold ROP without a preceding diagnosis of prethreshold ROP, significant variables were not changed. The association between presence of plus disease at first diagnosis prethreshold/threshold and progression to threshold was strengthened (OR increased from 4.8 to 5.7, P<0.001 in both cases). The adjusted estimate of relationship between cumulative SNAP score and progression to threshold was not changed dramatically (OR increased from with P ¼ 0.03 to 1.161, P ¼ 0.028). ROC curve area remained 0.78, not significantly changed. Discussion Enhancements to existing risk models for development or progression of ROP could improve patient selection and thereby the efficiency and safety of research into mechanisms and treatments of ROP. This study demonstrates that cumulative neonatal illness severity, measured with daily SNAP scores from the first 28 days of life, was significantly and independently associated with subsequent progression of moderate to severe ROP. However, the addition of this measure of cumulative illness severity did not significantly enhance the performance of a risk model in discriminating between eyes that did or did not subsequently progress to severe ROP. A fluctuating clinical course, as reflected by cumulative change in daily SNAP scores, was not significantly associated with subsequent progression to threshold. Mattia et al. 19,20 reported an association between cumulative illness severity and neurodevelopmental outcome in premature infants p30 weeks gestation at birth. Those in the highest quartile for cumulative SNAP score from their entire newborn hospitalization were more likely to have abnormal gross motor and psychomotor development at 2 to 3 years and 4 to 5 years of age than infants with a less severe hospital course. Cumulative SNAP score was a stronger risk factor for neurodevelopmental impairment than was gestational age or grade of intraventricular hemorrhage. These results suggest that objective measurement of sustained physiological instability with cumulative illness severity scoring may be useful in assessing neonatal risk for lasting deleterious effects on developing organ systems. This was not true when examining progression to severe ROP using a 28-day cumulative score, however, as the gain in model performance contributed by SNAP28 was modest. Simpler clinical variables appeared to function as well, either because they are adequate proxies for neonatal illness severity or because they are more specifically associated with progression of ROP. Fluctuation of SNAP scores was not a significant risk factor for progression to threshold in this study. Fluctuation in clinical status as assessed with daily illness severity scores captured change occurring from one 24-h period to the next. It is possible that fluctuations occurring over much shorter periods of time or in clinical factors not captured by SNAP scores may be more pertinent to progression of ROP. Short periods of hyperoxemia or hypoxemia, for example, may influence the risk for progressive ROP, 10,29 yet would not be reflected in changes of daily SNAP scores. Normal retinal vessel development is stimulated by production of vascular endothelial growth factor (VEGF) in the avascular retina, with insulin-like growth factor-1 (IGF-1) playing a permissive role. 30 Following premature birth, IGF-1 level may fall in response to many factors, halting vessel development. 3,31 Because IGF-1 is expressed by many organs and tissues, serum levels may be depressed by numerous factors adversely affecting tissue health, including acidosis and poor nutrition. 3,31 Thus, low IGF-1 levels in premature infants may be a marker for severe clinical illness. The cumulative illness severity data collected for this study correspond to the reported period during which IGF-1 levels may be low. Babies with higher cumulative neonatal illness severity may be at higher risk for depression of serum IGF-1, hyperoxia and VEGF overproduction, increasing the risk of abnormal retinal development. Further exploration of the relationship between illness severity and growth factors affecting retinal development is warranted. The Early Treatment for ROP (ETROP) study 32 used estimated risk of progression from prethreshold to poor retinal structural outcome to guide treatment, using the risk management of retinopathy of prematurity (RM-ROP2) model derived with data from the Multicenter Trial of Cryotherapy for Retinopathy of

7 508 Cumulative illness severity and ROP Prematurity Study. 33 The ETROP study generally corroborated RM-ROP2 risk estimates, demonstrating improved structural and functional outcomes in the early treatment group. However, the study authors estimated that early treatment based on the RM-ROP2 risk estimates alone would have resulted in the unnecessary treatment of 37% of high-risk prethreshold eyes that would have regressed without reaching threshold. 32 Continued refinement of ROP risk models is therefore warranted. Future studies may test improvements in the precision of predicting ROP progression associated with other potential risk factors such as cumulative oxygen exposure and oxygenation lability, biochemical markers such as VEGF and serum IGF-1 level, or proteomic or genomic factors. Several studies have found that black race is protective against progression to threshold. 21,34,35 Lower rates of severe ROP in black infants may be due to milder illness severity in black infants compared to nonblack infants of otherwise apparently equal risk for severe eye disease. 36,37 The results of the current study do not confirm this speculation. Black infants trended toward higher admission SNAP scores than nonblacks (P ¼ 0.056, data not shown), but the two groups were not different with respect to cumulative SNAP score in the first 28 days of life. Analysis of cumulative illness severity in a cohort where black race is found to be associated with risk of severe ROP might clarify this issue further. This study could not confirm a strong independent relationship between fungal sepsis and progression from prethreshold to threshold ROP after adjusting for the model in Table 4. It is possible that the small number of infants with fungemia in our cohort prevented detection of a real relationship between this condition and progression to severe ROP. However, it is also possible that fungal sepsis served as a non-specific marker for sustained severe illness in studies reporting an association between fungemia and ROP, with cumulative illness severity adjusting for this risk in this study. In summary, cumulative illness severity during the neonatal period, as measured with daily SNAP scores, is an independent risk factor for progression from moderate ROP to severe ROP. However, after adjusting for other clinical factors, measures of neonatal cumulative illness severity and fluctuation of illness severity do not significantly enhance the performance of a risk assessment model for ROP progression. Continued improvement of currently available models assessing risk for progressive ROP is warranted. Acknowledgments The authors gratefully acknowledge the efforts of Ralph Beltran, Jennifer Cho, Brian Costner, Julia Deanehan, Julie Hayner, Sarah Little, Kathleen O Connell, Christina Patoine, Sonja Payne, Alaka Ray, Erika Vargas, Tamra Carter and Lisa Valente. This study was funded by NEI K23 EY/HD00420, the Society for Pediatric Research Student Research Program, and the Tufts University School of Medicine Minority Student Research Program. References 1 Good W, Hardy R, Dobson V, Palmer E, Phelps D, Quintos M et al. The incidence and course of retinopathy of prematurity: findings from the early treatment for retinopathy of prematurity study. Pediatrics 2005; 116: Phelps DL. Retinopathy of prematurity. Pediatr Clin North Am 1993; 40: D Ercole A. ROP-forme fruste. J Perinatol 2002; 22: Manzoni P, Farina D, Leonessa M, Gomirato G, Arisio R. Risk factors for severe retinopathy of prematurity among very preterm infants: a unit-based or populationbased approach? Pediatrics 2005; 116: Karna P, Muttineni J, Angell L, Karmaus W. Retinopathy of prematurity and risk factors: a prospective cohort study. BMC Pediatr 2005; 5: Termote J, Donders A, Schalij-Delfos N, Lenselink C, Derkzen van Angeren C, Lissone S et al. Can screening for retinopathy of prematurity be reduced? Biol Neonate 2005; 88: Karlowicz MG, Giannone PJ, Pestian J, Morrow AL, Shults J. Does candidemia predict threshold retinopathy of prematurity in extremely low birth weight (</ ¼ 1000 g) neonates? Pediatrics 2000; 105: Brown DR, Biglan AW, Stretavsky MM. Retinopathy of prematurity: the relationship with intraventricular hemorrhage and bronchopulmonary dysplasia. J Pediatr Ophthalmol Strabismus 1990; 27: Higgins RD, Mendelsohn AL, DeFeo MJ, Ucsel R, Hendricks-Munoz KD. Antenatal dexamethasone and decreased severity of retinopathy of prematurity. Arch Ophthalmol 1998; 116: York J, Landers S, Kirby R, Arbogast P, Penn J. Arterial oxygen fluctuation and retinopathy of prematurity in very low birth weight infants. JPerinatol2004; 24: Hussain N, Clive J, Bhandari V. Current incidence of retinopathy of prematurity, Pediatrics 1999; 104: e Dani C, Reali M, Bertini G, Martelli E, Pezzati M, Rubaltelli F. The role of blood transfusions and iron intake on retinopathy of prematurity. Early Hum Dev 2001; 62: Englert J, Sounders R, Purohit D, Hulsey T, Ebeling M. The effect of anemia on retinopathy of prematurity in extremely low birth weight infants. J Perinatol 2001; 21: Cuculich PS, DeLozier KA, Mellen BG, Shenai JP. Postnatal dexamethasone treatment and retinopathy of prematurity in very-low-birth-weight neonates. Biol Neonate 2001; 79: Richardson D, Tarnow-Mordi WO, Lee SK. Risk adjustment for quality improvement. Pediatrics 1999; 103: Richardson D, Gray J, McCormick M, Workman K, Goldmann D. Score for neonatal acute physiology: a physiologic severity index for neonatal intensive care. Pediatrics 1993; 91: Anderson C, Richardson D, Stewart J, Leviton A. Predicting retinopathy of prematurity from NICU admission data. Pediatr Res 1994; 35: 263A. 18 Cole C, Hagadorn J, Rand W, Richardson D, Stewart J. Prediction of retinopathy of prematurity progression to threshold. Pediatr Res 2001; 49: 390A. 19 Mattia F, deregnier R. Chronic physiologic instability is associated with neurodevelopmental morbidity at one and two years in extremely premature infants. Pediatrics 1998; 102: E Merchant J, Mattia F, deregnier R. Gestational age and prolonged physiological instability are correlated with preschool cognitive function in premature infants p32 weeks gestation. Pediatr Res 1999; 45: 251A. 21 The STOP-ROP Multicenter Study Group. Supplemental Therapeutic Oxygen for Prethreshold Retinopathy of Prematurity (STOP-ROP), a randomized, controlled trial. I: primary outcomes. Pediatrics 2000; 105: Committee for the Classification of Retinopathy of Prematurity. An international classification of retinopathy of prematurity. Arch Ophthalmol 1984; 102: The International Neonatal Network. The CRIB (clinical risk index for babies) score: a tool for assessing initial neonatal risk and comparing performance of neonatal intensive care units. Lancet 1993; 342:

8 Cumulative illness severity and ROP Hardy R, Palmer E, Dobson V, Summers C, Phelps D, Quinn G et al. Risk analysis of prethreshold retinopathy of prematurity. Arch Ophthalmol 2003; 121: Hardy R, Palmer E, Schaffer D, Phelps D, Davis B, Cooper C. Outcome-based management of retinopathy of prematurity. J AAPOS 1997; 1: Harrell Jr FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996; 15: Hanley J. Receiver operating characteristic (ROC) methodology: the state of the art. Crit Rev Diagn Imaging 1989; 29: Hanley J, NcNeil B. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983; 148: McColm J, Cunningham S, Wade J, Sedowofia K, Gellen B, Sharma T et al. Hypoxic oxygen fluctuations produce less severe retinopathy than hyperoxic fluctuations in a rat model of retinopathy of prematurity. Pediatr Res 2004; 55: Smith L. IGF-1 and retinopathy of prematurity in the preterm infant. Biol Neonate 2005; 88: Verhaeghe J, Van Herck E, Billen J, Moerman P, Van Assche F, Giudice L. Regulation of insulin-like growth factor-i and insulin-like growth factor binding protein-1 concentrations in preterm fetuses. Am J Obstet Gynecol 2003; 188: Early Treatment For Retinopathy Of Prematurity Cooperative Group. Revised indications for the treatment of retinopathy of prematurity: results of the early treatment for retinopathy of prematurity randomized trial. Arch Ophthalmol 2003; 121: Palmer E, Flynn J, Hardy R, Phelps D, Phillips C, Schaffer D et al. Incidence and early course of retinopathy of prematurity. Ophthalmology 1991; 98: Tadesse M, Dhanireddy R, Mittal M, Higgins RD. Race, Candida sepsis, and retinopathy of prematurity. Biol Neonate 2002; 81: Schaffer D, Palmer E, Plotsky D, Metz H, Flynn J, Tung B. Prognostic factors in the natural course of retinopathy of prematurity. Ophthalmology 1993; 100: Yang M. Ethnic variation in the incidence and severity of retinopathy of prematurity: possible explanations, including racial differences in illness severity. Int Ophthalmol Clin 2003; 43: Berman S, Richardson D, Cohen A, Pursley D, Lieberman E. Relationship of race and severity of neonatal illness. Am J Obstet Gynecol 2001; 184: Manzoni P, Maestri A, Leonessa M, Mostert M, Farina D, Gomirato G. Fungal and bacterial sepsis and threshold ROP in preterm very low birth weight neonates. J Perinatol 2006; 26: Parupia M, Dhanireddy R. Association of postnatal dexamethasone use and fungal sepsis in the development of severe retinopathy of prematurity and progression to laser therapy in extremely low-birth-weight infants. J Perinatol 2001; 21: Noyola D, Bohra L, Paysse E, Fernandez M, Coats D. Association of candidemia and retinopathy of prematurity in very low birthweight infants. Ophthalmology 2002; 109: Mittal M, Dhanireddy R, Higgins R. Candida sepsis and association with retinopathy of prematurity. Pediatrics 1998; 101: Bell M, Ternberg J, Feigin R, Keating J, Marshall R, Barton L et al. Neonatal necrotizing enterocolitis. Therapeutic decisions based upon clinical staging. Ann Surg 1978; 187: 1 7.