655

Impact of Postnatal Systemic Corticosteroids on Mortality and Cerebral Palsy in Preterm Infants: Effect Modification by Risk for Chronic Lung Disease Lex W. Doyle, Henry L. Halliday, Richard A. Ehrenkranz, Peter G. Davis and John C. Sinclair Pediatrics 2005;115;655-661 DOI: 10.1542/peds.2004-1238

The online version of this article, along with updated information and services, is located on the World Wide Web at: http://www.pediatrics.org/cgi/content/full/115/3/655

PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since 1948. PEDIATRICS is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, 60007. Copyright © 2005 by the American Academy of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.

Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

Impact of Postnatal Systemic Corticosteroids on Mortality and Cerebral Palsy in Preterm Infants: Effect Modification by Risk for Chronic Lung Disease Lex W. Doyle, MD*‡; Henry L. Halliday, MD§; Richard A. Ehrenkranz, MD㛳; Peter G. Davis, MD*‡; and John C. Sinclair, MD#¶ ABSTRACT. Objective. In preterm infants, chronic lung disease (CLD) is associated with an increased risk for cerebral palsy (CP). However, systemic postnatal corticosteroid therapy to prevent or treat CLD, although effective in improving lung function, may cause CP. The objective of this study was to determine the effect of systemic postnatal corticosteroid treatment on death and CP and to assess any modification of effect arising from risk for CLD. Methods. Randomized, controlled trials of postnatal corticosteroid therapy for prevention or treatment of CLD in preterm infants that reported rates of both mortality and CP were reviewed and their data were synthesized. Twenty studies with data on 1721 randomized infants met eligibility criteria. The relationship between the corticosteroid effect on the combined outcome, death or CP, and the risk for CLD in control groups was analyzed by weighted meta-regression. Results. Among all infants who were randomized, a significantly higher rate of CP after corticosteroid treatment (typical risk difference [RD]: 0.05; 95% confidence interval [CI]: 0.02, 0.08) was partly offset by a nonsignificant reduction in mortality (typical RD: ⴚ0.02; 95% CI: ⴚ0.06 to 0.02). Consequently, there was no significant effect of corticosteroid treatment on the combined rate of mortality or CP (typical RD: 0.03; 95% CI: ⴚ0.01 to 0.08). However, on meta-regression, there was a significant negative relationship between the treatment effect on death or CP and the risk for CLD in control groups. With risks for CLD below 35%, corticosteroid treatment significantly increased the chance of death or CP, whereas with risks for CLD exceeding 65%, it reduced this chance. Conclusions. The effect of postnatal corticosteroids on the combined outcome of death or CP varies with the level of risk for CLD. Pediatrics 2005;115:655–661; infant, preterm, survival, cerebral palsy, corticosteroids.

From the *Division of Paediatrics, Royal Women’s Hospital, Melbourne, Australia; ‡Departments of Obstetrics and Gynaecology and of Paediatrics, University of Melbourne, Melbourne, Australia; §Department of Child Health, Queen’s University, Belfast, Northern Ireland; 㛳Department of Pediatrics, Yale University School of Medicine, New Haven, Connecticut; ¶Department of Pediatrics, McMaster University, Hamilton, Ontario, Canada; and #Department of Pediatrics, Center for Clinical Research and Evidence-Based Medicine, University of Texas Medical School, Houston, Texas. Accepted for publication Aug 5, 2004. doi:10.1542/peds.2004-1238 No conflict of interest declared. Reprint requests to (L.W.D.) Department of Obstetrics and Gynaecology, Royal Women’s Hospital, 132 Grattan St, Carlton 3053, Australia. E-mail: [email protected] PEDIATRICS (ISSN 0031 4005). Copyright © 2005 by the American Academy of Pediatrics.

ABBREVIATIONS. CLD, chronic lung disease; CP, cerebral palsy; RCT, randomized, controlled trial; RR, relative risk; RD, risk difference; CI, confidence interval; IQR, interquartile range.

I

mproved survival of preterm infants has been accompanied by an increased rate of chronic lung disease (CLD),1 most often defined as a dependence on supplemental oxygen beyond 36 weeks’ postmenstrual age. From the mid-1980s, postnatal corticosteroids were increasingly prescribed for prevention or treatment of CLD, supported by evidence of benefit on some short-term outcomes, including earlier weaning from mechanical ventilation and a reduction in CLD.2–4 In the era before postnatal corticosteroid treatment, the long-term neurodevelopmental outcome for survivors with CLD was worse than that in similar infants without CLD.5,6 Therefore, a reduction in CLD might be expected to reduce the rate of adverse long-term sequelae. However, corticosteroids can have direct toxic effects on the developing brain, including neuronal necrosis, interference with healing, and inhibition of brain growth.7,8 There is currently concern about possible adverse long-term effects of postnatal corticosteroids, cerebral palsy (CP) having been reported to occur more frequently in some randomized, controlled trials (RCTs)9–11 and meta-analyses of RCTs.2,12 Recently, the European Association of Perinatal Medicine issued a warning about postnatal corticosteroids,13 and the American Academy of Pediatrics and the Canadian Pediatric Society stated that routine postnatal use of systemic dexamethasone (the most commonly prescribed corticosteroid) for prevention or treatment of CLD is not recommended.14 Clinicians currently have a dilemma when faced with an infant who is increasingly ventilator dependent: should they start corticosteroids in an attempt to reduce ventilator and oxygen dependence and possibly to improve long-term outcome, or should they avoid corticosteroids because they may be independently harmful? Interpreting the RCTs of postnatal corticosteroids is difficult as there are differences in the baseline risks for CLD, the timing and the dose of corticosteroids according to the protocol, and the rate of open-label corticosteroids given to infants in the placebo group (the “contamination” rate). The follow-up data also differ regarding the age of children at follow-up, the follow-up rate, and

PEDIATRICS Vol. 115 No. 3 March 2005 Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

655

the expertise of the personnel involved in the follow-up assessments. Early death and CP among survivors are competing outcomes, and both must be considered when evaluating the effect of corticosteroid treatment. The aims of this report were to synthesize the data concerning mortality and CP from all RCTs of postnatal systemic corticosteroids for treatment or prevention of CLD in preterm infants and to test the hypothesis that the effect of corticosteroids is modified by the level of risk for CLD. METHODS The literature was searched for all RCTs of postnatal corticosteroids, and additional unpublished data were sought from authors of all known RCTs.2–4 The Medline search, updated through May 2004, used the terms “adrenal cortex hormones” or “dexamethasone” or “betamethasone” or “hydrocortisone” or “steroids” or “corticosteroids”; limits “randomized controlled trials,” “human,” and “all infant: birth to 23 months.” Studies of only inhaled steroids were excluded. All RCTs of postnatal corticosteroid treatment versus no treatment for prevention or treatment of CLD in preterm infants that reported mortality and the long-term outcome of CP were included. When there was ⬎1 report of long-term outcomes from 1 RCT, we used data from the report with outcomes described at the latest age of the study subjects. We sought data on the rate of CLD in the control group (defined as requiring oxygen therapy at 36 weeks’ postmenstrual age), the contamination rate (percentage of control group ultimately given corticosteroids), the dose of corticosteroids according to the protocol (in mg/kg dexamethasone equivalent), the age of the infants when corticosteroids were started (first week ⫽ early; after first week ⫽ later), the age of children at follow-up (in months), whether they were assessed by experts who were blinded to treatment group allocation, and the follow-up rate of survivors (%). Descriptions in the RCTs of the types of participants, interventions, short-term outcome measures, study design, and validity assessment of the studies have already been reported.2–4 Data were analyzed in RevMan 4.2.2, with the endpoints being mortality before follow-up; CP; and, to allow for the competing risks of death and long-term morbidity, the combination of CP or death. The measures of treatment effect for individual trials included relative risk (RR); risk difference (RD); and, for the metaanalyses the typical RR and typical RD, all with their 95% confidence intervals (CIs). A fixed-effects model was assumed for meta-analysis, with significant statistical heterogeneity defined as P ⬍ .05. In addition to examining the effect of corticosteroids on death and CP in all of the trials taken together, we conducted subgroup analyses that were based on the postnatal age of starting corticosteroids (early and later) and on the contamination rate (no contamination, ⬎0% to ⬍35% contamination, and ⱖ35% contamination). We examined the relationship between the corticosteroid effect on the combined outcome, death or CP, and the risk for CLD (indicated by the rate of CLD in the control group) by weighted meta-regression analysis in Stata (version 7.0)15 to test the hypothesis that the treatment effect on survival free of CP would be less favorable in trials that were conducted in populations at low risk for CLD and more favorable in trials that were conducted in populations at high risk for CLD. Meta-regression fits models with 2 additive components of variance, one representing the variance within studies, the other the variance between studies. Variance within studies is related to sample size, and larger studies contribute more to the overall results. We then adjusted for other important study design variables, including timing (early versus later) and total dose (ⱖ3.0 mg/kg, ⬍3.0 mg/kg) of corticosteroids, the contamination rate (⬍35%, ⱖ35%), the follow-up rate (ⱖ90%, ⬍90%), assessment by blinded experts (yes, no), and age of follow-up (ⱖ36 months, ⬍36 months). These cutpoints were chosen either to be around the median (dose, contamination rate, and age of follow-up) or because they represented desirable features of follow-up studies (follow-up rate and outcome assessment). We then repeated the meta-regression analyses to examine the relationship between the rate of CLD in the control group and the corticosteroid effect on death and CP separately, without adjust-

656

ing for the other study variables. For all analyses, P ⬍ 0.05 was considered statistically significant.

RESULTS Description of Studies

Thirty-six RCTs of systemic postnatal corticosteroid therapy enrolling 4269 infants for whom mortality data were reported were reviewed. Of these, 20 studies enrolling 2064 infants assessed effect on CP either in which the data were published9–11,16–39 or in which unpublished follow-up data were obtained by personal communication with authors.40–45 Among 6 published follow-up studies, additional details were clarified by the authors.10,11,18,20,24,28,36 In 3 multicenter studies, follow-up data were available for infants who were randomized from only 1 of the study sites23,44,45; hence, there were potentially 1721 randomized infants with follow-up data. In the 1721 infants randomized, there were 433 deaths before follow-up, 1151 (89%) of 1288 of survivors were examined, and there were 220 diagnoses of CP. Of the 20 studies with follow-up data, in 14, data on the rate of CLD in the control group were obtained.11,18,24,25,28,30,32,34,37,38,42–45 In the 6 studies without data on CLD,16,19,21,36,40,41 the corticosteroids were started beyond the first week of life in all but 1 study.16 Clinical characteristics of the studies varied widely (Table 1). In 9 studies, therapy started in the first week of life, and in 11, therapy started after the first week of life. The median cumulative dose of steroids (in mg/kg dexamethasone equivalent) according to the protocol was 3.0 (interquartile range [IQR]: 1.1– 5.0), the median contamination rate (available for 17 studies only) was 33% (IQR: 16%– 49%), and the median control group rate of CLD at 36 weeks’ postmenstrual age in the 14 studies that reported this outcome was 50% (IQR: 28%– 69%). Follow-up method also varied widely (Table 1); the median age at assessment was 36 months (IQR: 19 –58 months); the follow-up rate was at least 90% in 70% (n ⫽ 14) of studies; and in 70% (n ⫽ 14) of studies, children were assessed by experts who were blinded to treatment allocation. The numbers of infants who were randomized, infants who died, survivors assessed, and survivors with CP are shown in Table 2. Most studies had small sample sizes (median: 43), and only 5 studies with available long-term follow-up data randomized ⬎100 infants. Effect on Death and CP: All Trials, Early Treatment and Later Treatment Subgroups

There was no significant effect of corticosteroid treatment on death before latest follow-up in any individual study or in the meta-analyses, whether those were based on all trials or on either the early treatment or later treatment subgroups of trials (Table 3). There was no evidence of benefit of corticosteroid treatment on the incidence of CP (Table 3). Instead, a significantly higher rate of CP among children who had been randomized to corticosteroid treatment was found in 2 individual trials,10,11 in the meta-analysis of all trials (in which the typical abso-

POSTNATAL CORTICOSTEROIDS AND CEREBRAL PALSY Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

TABLE 1.

Description of Studies That Assessed Effect of Postnatal Corticosteroids on CP

Study (First Author)

Postnatal Age at Start of Treatment, d

Steroid Dose,* mg/kg

Contamination Rate,† %

CLD (Control) at 36 Weeks, %

Age at Follow-up, mo

Follow-up Rate, %

Assessed by Blinded Experts

Baden16,17 Ariagno40 Cummings18 Harkavy41 CDTG19,20 Ohlsson21,22 Kari23,24 Sanders42 Brozanski25,26 Shinwell11,27 Subhedar28,29 Yeh30,31 Vincer36 Kovacs43 Romagnoli32,33 Watterberg44 Romagnoli34,35 Kothadia37,10 Sinkin45 Stark38,39

1 21 13–15 30 21 21–35 10 ⬍1 7 ⬍1 4 1 28 7 3 ⬍2 10 15–25 ⬍1 1

1.0 4.0 8.0 5.0 4.2 5.6 3.5 1.0 1.5 1.5 4.5 6.2 2.4 1.5 2.4 0.4 4.9 7.8 1.0 0.9

NS 17.6 0 25.0 43.0 46.2 33.3 14.3 NS 25.9 41.7§ 11.4 NS 60.0 52.0 75.0§ 33.3 0 36.9 51.4

NS NS 72.7 NS NS NS 50.0 23.8 59.0 9.5 61.9 28.5 NS 46.7 68.0 50.0¶ 73.3 73.8 24.6¶ 45.0

12‡ 12–36‡ 15 or 48‡ 5–24‡ 36 18 60 Mean 62 12‡ Mean 53 30‡ 98 24‡ Up to 90 34–42‡ 11‡ 36–42‡ 12 and 60‡ 12‡ 18–22‡

93 96 100 32 94 96 100 100 68 83 95 92 100 70 100 86# 100 98 100# 87

Yes NS Yes Yes No NS NS Yes NS Yes Yes Yes Yes NS Yes Yes Yes Yes Yes Yes

First reference reports original RCT; second reference reports follow-up (if different). CDTG indicates Collaborative Dexamethasone Trial Group; NS, not specified. * Cumulative dose in dexamethasone equivalent. † Contamination rate refers to percentage of control group treated with open-label steroids. ‡ Age corrected for prematurity. § Not stated in original study in controls randomized, data from follow-up report. ¶ Rate of CLD in whole multicenter study: follow-up data from one center only. # Follow-up rate in 1 center of multicenter study. TABLE 2.

Frequency of Death, Follow-up Assessment, and CP in Individual Studies Study

Randomized, n

Treatment started early (first week of life) Baden16,17 Sanders42 Shinwell27,11 Subhedar28,29 Yeh30,31 Romagnoli32,33 Watterberg44 Sinkin45 Stark38,39 Treatment started later (after first week of life) Ariagno40 Cummings18 Harkavy41 CDTG19,20 Kari23,24 Ohlsson21,22 Brozanski25,26 Vincer36 Kovacs43 Romagnoli34,35 Kothadia37,10

Death Before Follow-up, n

Survivors Assessed, n

CP, n

CS

Controls

CS

Controls

CS

Controls

CS

Controls

22 19 132 21 132 25 13† (20) 32† (189) 111

22 21 116 21 130 25 13† (20) 27† (195) 109

8 2 31 11 53 2 3 11 26

9 7 27 9 50 3 2 7 30

13 17 79 10 72 23 10 21 76

12 14 80 11 74 22 8 20 67

2 3 38* 0 17 2 1 4 11

1 1 12* 2 9 3 2 1 12

17 25 9 143 11† (17) 12 39 11 30 15 57

17 11 12 142 12† (24) 13 39 9 30 15 61

5 7 1 33 1 1 4 2 8 0 7

5 6 2 29 2 0 9 1 5 0 16

11 18 3 100 10 11 25 9 15 15 48

12 5 3 109 10 13 19 8 18 15 45

1 5 1 20 3 1 5 4 1 2 12*

3 2 2 18 2 3 4 2 1 3 4*

First reference reports original RCT; second reference reports follow-up (if different). CS indicates corticosteroids. * Individual studies with statistically significant differences in rates between steroid and control groups in this variable (see text). † Number randomized in the 1 center of multicenter study reporting follow-up data, with number in parentheses indicating total randomized in complete study.

lute risk increase was 5% [95% CI: 2% to 8%]), and in the early treatment subgroup of trials. For the combined outcome, death before follow-up or CP, the meta-analyses found no significant benefit of postnatal corticosteroid treatment. In 1 trial,11 there was a

significant increase in the combined outcome death or CP in the corticosteroid group. In all studies combined and in the early treatment subgroup, there was a nonsignificant increase in the combined outcome death or CP associated with corticosteroid treatment.

Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

ARTICLES

657

TABLE 3.

Effects on Death and CP: Results of Meta-analyses No. of Studies

Steroids, % (n/N)

Controls, % (n/N)

RR* (95% CI)

RD† (95% CI)

20 9 11

24.7 (216/876) 29.0 (147/507) 18.7 (69/369)

25.9 (219/845) 29.8 (144/484) 20.8 (75/361)

0.94 (0.80–1.10) 0.98 (0.81–1.18) 0.87 (0.65–1.16)

⫺0.02 (⫺0.06 to 0.02) ⫺0.01 (⫺0.06 to 0.05) ⫺0.03 (⫺0.09 to 0.03)

20 9 11

15.2 (133/876) 15.4 (78/507) 14.9 (55/369)

10.3 (87/845) 8.9 (43/484) 12.2 (44/361)

1.45 (1.13–1.87) 1.70 (1.20–2.42) 1.20 (0.83–1.74)

0.05 (0.02 to 0.08) 0.06 (0.02 to 0.10)‡ 0.02 (⫺0.02 to 0.07)

20 9 11

39.8 (349/876) 44.4 (225/507) 33.6 (124/369)

36.2 (306/845) 38.6 (187/484) 33.0 (119/361)

1.09 (0.96–1.22) 1.15 (0.99–1.33) 0.99 (0.81–1.21)

0.03 (⫺0.01 to 0.08) 0.06 (0.00 to 0.12) 0.00 (⫺0.07 to 0.06)

20 9 11

22.7 (133/586) 24.3 (78/321) 20.8 (55/265)

15.4 (87/565) 14.0 (43/308) 17.1 (44/257)

1.46 (1.14–1.86) 1.75 (1.25–2.44) 1.16 (0.81–1.66)

0.07 (0.03 to 0.12)§ 0.11 (0.05 to 0.17)储 0.03 (⫺0.04 to 0.09)

Death before follow-up among all randomized in trials assessing CP All trials Early treatment Later treatment CP among all randomized All trials Early treatment Later treatment Death or CP among all randomized All trials Early treatment Later treatment CP among survivors assessed All trials Early treatment Later treatment

* Typical RR: value ⬍1 favors steroid group; value ⬎1 favors controls. † Typical RD: negative value favors steroid group; positive value favors controls. Where 95% CI excludes 1 for RR or excludes 0 for RD, P ⬍ .05. Statistically significant heterogeneity of treatment effect with fixed effects model: ‡ P ⫽ .03; § P ⫽ .01; 储P ⫽ .001.

For CP among survivors examined, 1 trial11 found a significant increase in CP in the corticosteroid group. Corticosteroid treatment was associated with a significant increase in CP among survivors examined in the meta-analysis of all trials and in the early treatment subgroup. Subgroups by Contamination Rate

In 2 studies,10,18 which randomized 154 infants, there was no contamination. Combining the results from these 2 studies, there was a significant reduction in mortality (typical RD: ⫺0.17; 95% CI: ⫺0.30 to ⫺0.03; P ⫽ .02), a significant increase in CP among all randomized (typical RD: 0.12; 95% CI: 0.01 to 0.23; P ⫽ .04), and no significant effect on the combined outcome of death or CP (typical RD: ⫺0.05; 95% CI: ⫺0.20 to 0.10; P ⫽ .55). In the subgroup analysis that was restricted to the studies that had a contamination rate from ⬎0% to ⬍35% (7 studies, 658 infants), there was no significant effect on mortality (typical RD: ⫺0.01; 95% CI: ⫺0.08 to 0.06; P ⫽ .75), a significant increase in CP among all randomized (typical RD: 0.09; 95% CI: 0.04 to 0.14; P ⬍ .001), and a significant increase in the combined outcome of death or CP (typical RD: 0.08; 95% CI: 0.01 to 0.16; P ⫽ .03). In the subgroup analysis that was restricted to studies that had a contamination rate ⱖ35% (8 studies, 767 infants), there was no significant effect on either mortality (typical RD: 0.02; 95% CI: ⫺0.04 to 0.08; P ⫽ .50) or CP among all randomized (typical RD: ⫺0.01; 95% CI: ⫺0.05 to 0.04; P ⫽ .76) or the combined outcome of death or CP (typical RD: 0.01; 95% CI: ⫺0.05 to 0.08; P ⫽ .69). Meta-Regression Analysis: Relation of Treatment Effect to Control Risk for CLD

Fourteen studies reported the rate of CLD at 36 weeks’ postmenstrual age in the control group; this ranged from 9.5% to 73.8% (Table 1). Meta-analysis 658

of these 14 studies found no significant effect on the combined rate of death or CP among children who were randomized (typical RD: 0.04; 95% CI: ⫺0.02 to 0.09). However, there was a significant negative relationship between the RD for death or CP and the rate of CLD in the control group in these studies (Fig 1): for every 10% that the rate of CLD increased in the control group, the RD for death or CP fell by 3.8% (95% CI: 1.4% to 6.2%; P ⫽ .002). The regression line crossed 0 (no effect on death or CP) at a rate of CLD of ⬃50%. At control rates of CLD below ⬃35%, the lower bound of the 95% CI for the regression line was ⬎0, indicating a significantly increased rate of death or CP with corticosteroid treatment. At control rates of CLD above ⬃65%, the upper bound of the 95% CI for the regression line was ⬍0, indicating a significant treatment benefit for this outcome. None of the potential confounders (time of starting treatment, dose, contamination, follow-up rate, age at follow-up, and assessment by experts), either individually or in combination, significantly affected the relationship between the corticosteroid effect on death or CP and the control risk for CLD. In an adjusted weighted meta-regression based on 13 studies for which the required data were available, after simultaneously adjusting for each of the candidate confounding variables, a statistically significant relationship persisted between the corticosteroid treatment effect on the combined rate of death or CP and the control rate of CLD. Indeed, in the adjusted metaregression, the absolute size of the regression coefficient was increased, although the CIs were wider: for every 10% that the rate of CLD increased in the control group, the RD for death or CP fell by 9.1% (95% CI: 1.0% to 17.2%; P ⫽ .028). Meta-regression of RD for death alone against the rate of CLD in the control group showed a nonsignificant negative relationship (Y ⫽ 3.7 ⫺ 0.17X; P ⫽ .11). For every 10% increase in the rate of CLD in the

POSTNATAL CORTICOSTEROIDS AND CEREBRAL PALSY Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

Fig 1. RD (%) for death or CP among all participants versus rate of CLD (%) in the control group. Each study is shown by a circle whose area is proportional to that study’s weight. The control rates for CLD in the individual studies are given in Table 1, column 5. Regression line and its 95% CI are shown. Regression equation: Y ⫽ 18.7 ⫺ 0.38X; P ⫽ .002. CS, corticosteroids.

control group, the RD for death fell by 1.7% (95% CI: ⫺0.4% to 3.9%). Meta-regression of RD for CP alone against the rate of CLD in the control group showed a significant negative relationship (Y ⫽ 14.1 ⫺0.23X; P ⫽ .023). For every 10% that the rate of CLD increased in the control group, the RD for CP fell by 2.3% (95% CI: 0.3% to 4.3%) DISCUSSION

Corticosteroid therapy was initially introduced in chronically ventilator-dependent infants with the hope of saving lives and possibly improving longterm neurodevelopmental outcome if the severity and duration of CLD could be reduced. As early reports of short-term benefits appeared, there was an increasing tendency to start corticosteroid treatment on infants who were less critically ill and earlier in their nursery stay. This tendency has been tempered considerably with reports of an increase in CP among survivors. However, it is possible that corticosteroids are now being withheld in some infants who might benefit. There may be a point, defined by the chance of developing CLD, below which the risks of corticosteroid treatment on long-term outcome outweigh the benefits and above which the benefits outweigh the risks. The 20 studies included in this review were all randomized trials. Although all studies reported effects on some long-term outcomes, none was designed primarily to assess long-term effects of treatment. We attempted to obtain as complete a data set as possible by contacting the investigators and requesting any unpublished data concerning longterm outcomes. In the included studies, children were assessed mostly early in childhood, before the diagnosis of CP can be certain,46 not always by experts who were blinded to treatment group, not always using explicit and reproducible criteria for the diagnosis of CP, and sometimes with potentially important (⬎10%) loss to follow-up. Corticosteroids that are given after birth to pre-

term infants with lung disease have proved shortterm benefits, including reductions in ventilator and oxygen dependence and in incidence of CLD.2–4 These short-term benefits might be expected to lead to a reduction in adverse neurodevelopmental outcomes associated with CLD in preterm infants. However, among all trials, we found no evidence of benefit of postnatal corticosteroid treatment on the rate of CP, regardless of whether the competing risk for death before follow-up was taken into account. In fact, the direction of effects was generally that of harm, reaching statistical significance for CP among all randomized and among survivors examined. The subgroup analyses that were based on contamination rates of control groups in the various trials support the interpretation of a causal relationship, overall, between postnatal corticosteroid treatment and CP, in that this relationship was stronger in the trials with no or less contamination and weaker or nonexistent in the trials with high rates of contamination. In trials with low or no contamination, there was a tendency for corticosteroid treatment to reduce mortality. This pattern suggests that, overall, postnatal corticosteroid treatment, although maybe reducing mortality, causes CP in survivors. Corticosteroid treatment was associated with a significant increase in CP in the early treatment but not the later treatment subgroup; however, even in the later treatment subgroup, there was no trend toward benefit. Given that preterm infants are more unstable in the first days after birth, when events such as cerebrovascular hemorrhage are prone to occur, it is possible that any adverse effects of corticosteroids might be exacerbated at that time and that the marked adverse effects noted in the early treatment subgroup could be attributed to postnatal age per se. However, the result of the meta-regression shown in Fig 1 suggests an alternative explanation for the adverse effects observed in the early treatment trials. Postnatal corticosteroids in the doses used might not only have had direct toxic effects on the developing brain7,8 but also an indirect benefit to the brain by improving lung function. The early treatment trials enrolled infants who were generally at lower risk for CLD. Thus, infants who were treated early might have experienced net harm because they had less to gain in terms of the indirect benefit, but all were exposed to the direct harmful effects of corticosteroid treatment. The significant negative association between treatment effect on death or CP and risk for CLD supports this interpretation. None of the potential confounding variables was significantly related to the corticosteroid effect on the combined outcome of death or CP. Moreover, adjustment for these variables, including the time of starting treatment, did not eliminate the statistically significant negative relationship between the risk for CLD in controls and the effect on the rate of death or CP. We recognize that there are limitations of our meta-regression analysis to explore effect modification arising from risk for CLD. First, the analysis was limited to a subset of studies (14 of 20) in which the control rate of CLD was reported. Second, in taking the control rate of CLD as a measure of control risk,

Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

ARTICLES

659

we used data that were unavailable at trial entry. Future application of these results in identifying suitable candidates for corticosteroid treatment would require accurate prediction of CLD in at-risk populations. Third, we did not have access to individual patient data, which would allow more sensitive analysis of the relations between risk factors known at baseline, subsequent development of CLD, and corticosteroid effect on death or CP. Fourth, the list of potential confounding variables that we examined was limited by the data available and does not include other possible confounders, such as differences between studies in mechanical ventilation strategies used. The main focus of our study was on CP because this is the outcome that has been most controversial. We have also taken into account the competing outcome of death. Data on the effects of steroids on other neurologic outcomes are also important but are much more limited and are discussed in the existing Cochrane reviews.2–4 There were too few studies that reported both the effects of corticosteroids on neurologic outcomes other than CP and the rate of CLD in the control group to allow us to examine those associations in meta-regression analyses. Future research should be directed toward trying to prevent CLD and hence the need for corticosteroids, to more accurate early prediction of CLD47 to allow targeting of any postnatal corticosteroids to those most likely to experience long-term benefit, and to determination of the lowest dose of corticosteroid that is effective in reducing ventilator and oxygen dependence and the incidence and severity of CLD, without causing long-term harm. The analyses reported here support future randomized trials of postnatal corticosteroids specifically in preterm infants who are accurately predicted to be at high risk for CLD, with effects on neurodevelopment and mortality as primary outcomes. ACKNOWLEDGMENTS This study was supported in part by Project Grant No. 108700 from the National Health and Medical Research Council (NHMRC) of Australia and Contract N01-HD-6-3252 from the National Institute of Child Health and Human Development. Peter Davis is supported in part by a Practitioner Fellowship from the NHMRC.

REFERENCES 1. Young TE, Kruyer LS, Marshall DD, Bose CL. Population-based study of chronic lung disease in very low birth weight infants in North Carolina in 1994 with comparisons with 1984. The North Carolina Neonatologists Association. Pediatrics. 1999;104(2). Available at: www.pediatrics.org/cgi/content/full/104/2/e17 2. Halliday HL, Ehrenkranz RA, Doyle LW. Early postnatal (⬍96 hours) corticosteroids for preventing chronic lung disease in preterm infants Cochrane Database Syst Rev. 2003;(1):CD001146 3. Halliday HL, Ehrenkranz RA, Doyle LW. Moderately early (7–14 days) postnatal corticosteroids for preventing chronic lung disease in preterm infants. Cochrane Database Syst Rev. 2003;(1):CD001144 4. Halliday HL, Ehrenkranz RA, Doyle LW. Delayed (⬎3 weeks) postnatal corticosteroids for chronic lung disease in preterm infants. Cochrane Database Syst Rev. 2003;(1):CD001145 5. Teberg AJ, Pena I, Finello K, Aguilar T, Hodgman JE. Prediction of neurodevelopmental outcome in infants with and without bronchopulmonary dysplasia. Am J Med Sci. 1991;301:369 –374 6. Skidmore MD, Rivers A, Hack M. Increased risk of cerebral palsy among very low-birthweight infants with chronic lung disease. Dev Med Child Neurol. 1990;32:325–332

660

7. Taeusch HW Jr. Glucocorticoid prophylaxis for respiratory distress syndrome: a review of potential toxicity. J Pediatr. 1975;87:617– 623 8. Weichsel ME Jr. The therapeutic use of glucocorticoid hormones in the perinatal period: potential neurological hazards. Ann Neurol. 1977;2: 364 –366 9. Yeh TF, Lin YJ, Huang CC, et al. Early dexamethasone therapy in preterm infants: a follow up study. Pediatrics. 1998;101(5). Available at: www.pediatrics.org/cgi/content/full/101/5/e7 10. O’Shea TM, Kothadia JM, Klinepeter KL, et al. Randomized placebocontrolled trial of a 42-day tapering course of dexamethasone to reduce the duration of ventilator dependency in very low birth weight infants: outcome of study participants at 1-year adjusted age. Pediatrics. 1999;104:15–21 11. Shinwell ES, Karplus M, Reich D, et al. Early postnatal dexamethasone treatment and increased incidence of cerebral palsy. Arch Dis Child Fetal Neonatal Ed. 2000;83:F177–F181 12. Barrington KJ. The adverse neuro-developmental effects of postnatal steroids in the preterm infant: a systematic review of RCTs. BMC Pediatr. 2001;1:1 13. Halliday HL. Guidelines on neonatal steroids. Prenat Neonat Med. 2001; 6:371–373 14. American Academy of Pediatrics, Committee on Fetus and Newborn, and Canadian Paediatric Society, Fetus and Newborn Committee. Postnatal corticosteroids to treat or prevent chronic lung disease in preterm infants. Pediatrics. 2002;109:330 –338 15. Intercooled Stata 7.0 for Windows. College Station, TX: Stata Corp; 2000 16. Baden M, Bauer CR, Colle E, Klein G, Taeusch HW Jr, Stern L. A controlled trial of hydrocortisone therapy in infants with respiratory distress syndrome. Pediatrics. 1972;50:526 –534 17. Fitzhardinge PM, Eisen A, Lejtenyi C, Metrakos K, Ramsay M. Sequelae of early steroid administration to the newborn infant. Pediatrics. 1974; 53:877– 883 18. Cummings JJ, D’Eugenio DB, Gross SJ. A controlled trial of dexamethasone in preterm infants at high risk for bronchopulmonary dysplasia. N Engl J Med. 1989;320:1505–1510 19. Collaborative Dexamethasone Trial Group. Dexamethasone therapy in neonatal chronic lung disease: an international placebo-controlled trial. Pediatrics. 1991;88:421– 427 20. Jones R, Wincott E, Elbourne D, Grant A. Controlled trial of dexamethasone in neonatal chronic lung disease: a 3-year follow-up. Pediatrics. 1995;96:897–906 21. Ohlsson A, Calvert SA, Hosking M, Shennan AT. Randomized controlled trial of dexamethasone treatment in very-low-birth-weight infants with ventilator-dependent chronic lung disease. Acta Paediatr. 1992;81:751–756 22. Ohlsson A. Randomized Controlled Trial of Dexamethasone Treatment in Very-Low-Birth-Weight Infants With Ventilator-Dependent Chronic Lung Disease. Master’s thesis. Hamilton, Ontario, Canada: McMaster University; 1990 23. Kari MA, Heinonen K, Ikonen RS, Koivisto M, Raivio KO. Dexamethasone treatment in preterm infants at risk for bronchopulmonary dysplasia. Arch Dis Child. 1993;68:566 –569 24. Mieskonen S, Eronen M, Malmberg LP, Turpeinen M, Kari MA, Hallman M. Controlled trial of dexamethasone in neonatal chronic lung disease: an 8-year follow-up of cardiopulmonary function and growth. Acta Paediatr. 2003;92:896 –904 25. Brozanski BS, Jones JG, Gilmour CH, et al. Effect of pulse dexamethasone therapy on the incidence and severity of chronic lung disease in the very low birth weight infant. J Pediatr. 1995;126:769 –776 26. Hofkosh D, Brozanski BS, Edwards MD, Williams LA, Jones JG, Cheng KP. One year outcome of infants treated with pulse dexamethasone for prevention of BPD. Pediatr Res. 1995;37:259A 27. Shinwell ES, Karplus M, Zmora E, et al. Failure of early postnatal dexamethasone to prevent chronic lung disease in infants with respiratory distress syndrome. Arch Dis Child Fetal Neonatal Ed. 1996;74: F33–F37 28. Subhedar NV, Ryan SW, Shaw NJ. Open randomised controlled trial of inhaled nitric oxide and early dexamethasone in high risk preterm infants. Arch Dis Child Fetal Neonatal Ed. 1997;77:F185–F190 29. Subhedar NV, Bennett AJ, Wardle SP, Shaw NJ. More trials on early treatment with corticosteroids are needed. Br Med J. 2000;320:941 30. Yeh TF, Lin YJ, Hsieh WS, et al. Early postnatal dexamethasone therapy for the prevention of chronic lung disease in preterm infants with respiratory distress syndrome: a multicenter clinical trial. Pediatrics. 1997;100(4). Available at: www.pediatrics.org/cgi/content/full/100/4/e3 31. Yeh TF, Lin YJ, Lin HC, et al. Outcomes at school age after postnatal dexamethasone therapy for lung disease of prematurity. N Engl J Med. 2004;350:1304 –1313

POSTNATAL CORTICOSTEROIDS AND CEREBRAL PALSY Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

32. Romagnoli C, Zecca E, Vento G, De Carolis MP, Papacci P, Tortorolo G. Early postnatal dexamethasone for the prevention of chronic lung disease in high-risk preterm infants. Intensive Care Med. 1999;25:717–721 33. Romagnoli C, Zecca E, Luciano R, Torrioli G, Tortorolo G. Controlled trial of early dexamethasone treatment for the prevention of chronic lung disease in preterm infants: a 3-year follow-up. Pediatrics. 2002; 109(6). Available at: www.pediatrics.org/cgi/content/full/109/6/e85 34. Romagnoli C, Zecca E, Vento G, Maggio L, Papacci P, Tortorolo G. Effect on growth of two different dexamethasone courses for preterm infants at risk of chronic lung disease. A randomized trial. Pharmacology. 1999;59:266 –274 35. Romagnoli C, Zecca E, Luciano R, Torrioli G, Tortorolo G. A three year follow up of preterm infants after moderately early treatment with dexamethasone. Arch Dis Child Fetal Neonatal Ed. 2002;87:F55–F58 36. Vincer M, Allen AC. Double blind randomized controlled trial of 6-day pulse of dexamethasone for very low birth weight infants (VLBW ⬍ 1500 grams) who ventilator dependent at 4 weeks of age (sic). Pediatr Res. 1998;43:201A 37. Kothadia JM, O’Shea TM, Roberts D, Auringer ST, Weaver RG 3rd, Dillard RG. Randomized placebo-controlled trial of a 42-day tapering course of dexamethasone to reduce the duration of ventilator dependency in very low birth weight infants. Pediatrics. 1999;104:22–27 38. Stark AR, Carlo WA, Tyson JE, et al. Adverse effects of early dexamethasone in extremely-low-birth-weight infants. National Institute of Child Health and Human Development Neonatal Research Network. N Engl J Med. 2001;344:95–101

39. Stark AR, Carlo WA, Tyson JE, et al. Neurodevelopmental outcome and growth at 18 –22 months in infants treated with early dexamethasone. Pediatr Res. 2001;49:388A 40. Ariagno RL, Sweeney TJ, Baldwin RB, Inguillo D, Martin D. Dexamethasone effects on lung function and risks in 3 week old ventilatory dependent preterm infants. Am Rev Respir Dis. 1987;135:A125 41. Harkavy KL, Scanlon JW, Chowdhry PK, Grylack LJ. Dexamethasone therapy for chronic lung disease in ventilator- and oxygen-dependent infants: a controlled trial. J Pediatr. 1989;115:979 –983 42. Sanders RJ, Cox C, Phelps DL, Sinkin RA. Two doses of early intravenous dexamethasone for the prevention of bronchopulmonary dysplasia in babies with respiratory distress syndrome. Pediatr Res. 1994;36: 122–128 43. Kovacs L, Davis GM, Faucher D, Papageorgiou A. Efficacy of sequential early systemic and inhaled corticosteroid therapy in the prevention of chronic lung disease of prematurity. Acta Paediatr. 1998;87:792–798 44. Watterberg KL, Gerdes JS, Gifford KL, Lin HM. Prophylaxis against early adrenal insufficiency to prevent chronic lung disease in premature infants. Pediatrics. 1999;104:1258 –1263 45. Sinkin RA, Dweck HS, Horgan MJ, et al. Early dexamethasone— attempting to prevent chronic lung disease. Pediatrics. 2000;105:542–548 46. Stanley FJ. Using cerebral palsy data in the evaluation of neonatal intensive care: a warning. Dev Med Child Neurol. 1982;24:93–94 47. Yoder BA, Anwar MU, Clark RH. Early prediction of neonatal chronic lung disease: a comparison of three scoring methods. Pediatr Pulmonol. 1999;27:388 –394

MYOPIC GLOBALIZERS

”We may now all be benefiting from the communications revolution, and some may see progress in the fact that the Egyptian Sphinx now stares directly at the neon sign of a Kentucky Fried Chicken franchise . . . [but] too many globalizers and policymakers have taken an Olympian view of the globalization process . . . they do not have the ability to comprehend, however dimly, how other people live.“

DeSoto H. The Mystery of Capital. New York, NY: Basic Books; 2000

Submitted by Student

Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009

ARTICLES

661

Impact of Postnatal Systemic Corticosteroids on Mortality and Cerebral Palsy in Preterm Infants: Effect Modification by Risk for Chronic Lung Disease Lex W. Doyle, Henry L. Halliday, Richard A. Ehrenkranz, Peter G. Davis and John C. Sinclair Pediatrics 2005;115;655-661 DOI: 10.1542/peds.2004-1238 Updated Information & Services

including high-resolution figures, can be found at: http://www.pediatrics.org/cgi/content/full/115/3/655

References

This article cites 35 articles, 15 of which you can access for free at: http://www.pediatrics.org/cgi/content/full/115/3/655#BIBL

Citations

This article has been cited by 27 HighWire-hosted articles: http://www.pediatrics.org/cgi/content/full/115/3/655#otherarticle s

Subspecialty Collections

This article, along with others on similar topics, appears in the following collection(s): Premature & Newborn http://www.pediatrics.org/cgi/collection/premature_and_newbor n

Permissions & Licensing

Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: http://www.pediatrics.org/misc/Permissions.shtml

Reprints

Information about ordering reprints can be found online: http://www.pediatrics.org/misc/reprints.shtml

Downloaded from www.pediatrics.org at Indonesia:AAP Sponsored on May 16, 2009