Manuscript Citation: Shaffer ML, Baud O, Lacaze-Masmonteil T, Peltoniemi OM, Bonsante F, Watterberg KL. Effect of prophylaxis for early adrenal insufficiency using low-dose hydrocortisone in very preterm infants: an individual patient data meta-analysis. Journal of Pediatrics 2018.
Neha Kumbhat, MD, MSEpi; Alexis S. Davis, MD, MSEpi; William E. Benitz, MD.
Institution: Stanford University, Palo Alto, CA
Neha Kumbhat, MD, MS Epi.
Neonatal-Perinatal Medicine Clinical Fellow.
Department of Pediatrics/Division of Neonatology.
Stanford University School of Medicine
750 Welch Road, Suite 315
Palo Alto, CA 94304
Phone: (650) 723-5711
Fax: (650) 725-8351
Keywords: Extremely preterm infants, Bronchopulmonary Dysplasia, Hydrocortisone, Adrenal Insufficiency
Type of Investigation
Individual patient data meta-analysis of four prospective randomized controlled trials.
Inextremely premature infants, does exposure to low dose hydrocortisone versus exposure to placebo increase survival without bronchopulmonary dysplasia (BPD) at 36 weeks postmenstrual age (PMA)?
Design: Retrospectively designed, multicenter, individual patient data meta-analysis of four prospective randomized controlled trials.
Allocation: In each trial that was studied, infants were randomized to one of the two groups: low-dose hydrocortisone or placebo
Characteristics of the studies included in the IPD meta-analysis: Characteristics of the studies included in the IPD meta-analysis are shown in Table 1.
Primary Outcome: Survival without bronchopulmonary dysplasia (BPD) at 36 weeks of post-menstrual age (PMA) as a binary variable.
Secondary Outcomes: Days of ventilation, continuous positive airway pressure, or oxygen; supplemental oxygen at discharge; medical or surgical treatment for patent ductus arteriosus (PDA).
Adverse Outcomes: Pneumothorax, spontaneous gastrointestinal perforation, necrotizing enterocolitis, severe IVH, cystic periventricular leukomalacia, severe retinopathy of prematurity, late onset sepsis, effects on weight or head circumference at 36 weeks PMA, 2-year neurodevelopmental outcomes.
Data sources, acquisition, processing and assessment: The authors provided a PRISMA-IPD, which is a checklist of items required for reporting meta-analysis of IPD. (1) From a MEDLINE search for the terms "hydrocortisone", "randomized", and "human" and two meta-analyses conducted in 2010 (2) and 2017 (3), eleven RCTs of early hydrocortisone therapy were identified. Six trials were excluded because they included larger, more mature subjects or used hydrocortisone for other specific indications (such as refractory hypotension), in combination with other drugs (triiodothyronine or dopamine), or for less than 7 days. Five trials were specifically designed to determine the efficacy of early treatment with hydrocortisone to improve survival without BPD.(4)(5)(6)(7)(8) IPD was not available from the first pilot trial conducted in 1999.(4) The IPD dataset was compiled by the corresponding authors onto a standard data capture template using a common data dictionary for available data elements; in the absence of data elements, alternative measures were included.
Analysis: The one-step approach to IPD meta-analysis, in which subjects from all trials are individually included in each regression, was used. Continuous outcomes were specified based on mixed linear models while binary outcomes were specified using logistic regression models. The Kenward-Roger variance estimator was used to derive the 95% CI for the summary effects. Assuming that the treatment effect was consistent (homogeneous) across studies, a fixed-effect model was used. To account for clustering of patients within the included studies, random effect models were compared. Since the approaches produced comparable results, results from the fixed-effect model were reported for simplicity. Statistical variation (heterogeneity) across studies was summarized as I2. For long-term developmental outcomes, a traditional aggregate random-effect meta-analysis was performed using the published data from three studies.
The unadjusted odds of survival without BPD were greater in the hydrocortisone-treated group (OR, 1.37; 95% CI 1.07-1.76), and remained significant after adjusting for sex, gestational age, and antenatal steroid exposure (aOR, 1.45; 95% CI 1.11-1.90; I2 = 0%). The hydrocortisone-treated group also had lower odds of BPD (aOR, 0.73; 95% CI 0.54-0.98; I2 = 0%). While a benefit was not seen for odds of death prior to 36 weeks PMA (aOR, 0.76; 95% CI 0.54-1.07; I2 = 0%), the hydrocortisone-treated group had lower odds of dying before hospital discharge (aOR, 0.70; 95% CI 0.51-0.97; I2 = 0%). There was no significant difference between treatment groups for respiratory support components such as days of ventilation, continuous positive airway pressure, and oxygen. Among infants who developed BPD, requirements for those respiratory support measures did not differ between treatment groups. Infants in the hydrocortisone-treated group had decreased odds of requiring medical treatment for PDA, including indomethacin or ibuprofen (aOR; 0.72; 95% CI 0.56-0.93, p=0.012), but odds of ligation similar to control subjects. Odds for spontaneous gastrointestinal perforation (SIP) were higher in infants exposed to hydrocortisone (aOR 2.50, 95% CI 1.33-4.69; I2 = 31.9%), but only in association with cotreatment with indomethacin. Infants in the hydrocortisone-treated group had increased odds of late onset sepsis (aOR 1.34; 95% CI 1.02-1.75; I2 = 0%), which was not associated with increased overall mortality, other in-hospital adverse outcomes, or poor 2-year neurodevelopmental outcomes. Among subjects with intrapartum exposure to chorioamnionitis, hydrocortisone treatment was also associated with an increased incidence of late-onset sepsis (aOR 1.91; 95% CI 1.18-2.08; p=0.009), but also with greater odds of survival without BPD (aOR 2.01 95% CI 1.19-3.39, p=0.009) and decreased odds of death prior to discharge (aOR 0.43 95% CI 0.23-0.82; p=0.01).
This individual patient data meta-analysis shows that exposure to early low-dose hydrocortisone therapy: 1) increased survival without BPD, 2) decreased the rate of medical treatment for PDA, and 3) improved survival to discharge. Two adverse outcomes, late-onset sepsis and spontaneous intestinal perforation were significantly increased but did not negate beneficial effects of hydrocortisone.
The authors have successfully conducted an international collaborative meta-analysis utilizing individual patient data (IPD), which has brought clarity to one of the most controversial questions in Neonatology: Does early, prophylactic hydrocortisone have a beneficial effect in survival without BPD without causing adverse effects? Utilizing a large pooled sample, the authors were able to address important confounders at the individual patient level, conduct a sub-group analysis and assess long term outcomes.
An aggregate meta-analysis utilizing published data from the four trials included in the IPD analysis, failed to demonstrate a significant effect of low dose hydrocortisone on outcome of survival without BPD. [aOR 1.2; 95% CI 0.98-1.48, p = 0.16]. Although, a time consuming and costly undertaking, IPD analysis has several advantages over the aggregate meta-analysis. Firstly, raw data from each trial is assessed for quality then pooled and analyzed as a larger dataset. The authors had access to individual patient data for 982 patients from four studies, providing a large pooled sample that improved the statistical power. Conducting a randomized controlled trial with 982 premature infants would be expensive and time-consuming. This IPD provides us with a large sample size of pooled data for four RCT's with results immediately available for interpretation and use in clinical practice. Secondly, it allows adjustments of covariates at an individual patient level and then a group level. For example, in this IPD analysis, the effect of predictors on outcomes were adjusted for the following covariates: birthweight, gestational age, sex and exposure to antenatal steroids. These adjustments avoid an inherent weakness of aggregate meta-analysis - ecological fallacy. In other words, conclusions about an individual's outcome is deduced from the individual characteristics adjusted for various confounders rather than inferences based on the group the subject belongs to. Finally, it allows a sub-group analysis of the primary outcome. This study also assessed the safety profile of using low dose hydrocortisone in a planned sub-group analysis. While the hydrocortisone-exposed infants had an increase in the rate of late-onset sepsis and SIP, overall mortality and 2-year neurodevelopmental outcomes were similar to infants in the non-exposed group. Infants who received indomethacin concurrently had a significantly higher odds of developing spontaneous intestinal perforation. The greatest benefit for survival without BPD was seen among premature infants exposed to chorioamnionitis, suggesting that some effects may be mediated by treatment of adrenal insufficiency and others by anti-inflammatory actions of hydrocortisone. Accordingly, it may also be that it should be avoided in certain subsets of population while providing benefit in others.
The studies were similar in hypothesis tested, study design, and primary outcome studied, with minor variations in eligibility criteria, dose, duration of hydrocortisone therapy, and completeness of respiratory support data from some trials. Statistical variation was not noted across studies for the primary outcomes of survival without BPD at 36 weeks PMA (i.e., I2 = 0%). This low statistical heterogeneity across the studies makes the results highly reliable.
Using individual patient data meta-analysis, the authors of this study, who are the authors of all the original trials involved, hoped to better delineate the role of hydrocortisone for improving survival without BPD without significant adverse effects. Maximizing nutrition and minimizing oxidative and ventilatory injury are known to decrease the incidence and severity of BPD, addition of early prophylactic low dose postnatal steroids might add to the beneficial effects of these lung protective strategies.
EBM Lesson: Statistical heterogeneity
Meta-analysis brings together studies that have similarities but may differ slightly and thus bring variability into the systematic review. This variability is termed as heterogeneity. There are three forms of heterogeneity: 1) clinical heterogeneity, also called clinical diversity, is variability introduced into the review by differences in patient populations, interventions and/or outcomes studies; 2) methodological heterogeneity introduced by differences in study design and risk of bias; and finally, 3) statistical heterogeneity which is a manifestation of clinical or methodological heterogeneity. Clinical and methodological heterogeneity are subjective and cannot be quantified. Statistical heterogeneity on the other hand, is a concept that can be quantified utilizing Cochran's Q-statistic and the I2-statistic. They determine whether variation in findings are due to chance alone or due to statistical differences. Higgins et al(9) developed a way to quantify of statistical heterogeneity within systematic reviews, called I2. It is the percentage of variation across studies that is due to heterogeneity rather than chance. This measure of heterogeneity does not depend on the number of studies considered in the meta-analysis and thus is not affected when the analysis includes a small number of studies. Data from published meta-analysis such as Q Cochrane statistic (the sum of squared deviations of the study's estimate from the overall pooled estimate where each study's contribution carries the same weight regardless of their sample size) and degrees of freedom may be used to calculate I2. I2 = 100% X (Q – degrees of freedom/Q). I2 is categorized as very low (<25%), low (25%- < 50%), moderate (50% - < 75%) and large (> 75%). Negative I2 is treated as a zero, so I2 values always lie between 0% - 100%. A large I2 suggests too large a heterogeneity for a meaningful meta-analysis. In the IPD meta-analysis by Shaffer et al, the statistical heterogeneity was very low [I2 = 0% for the primary outcome of survival without BPD at 36 weeks PMA) for the primary outcome rendering the results of the IPD analysis more reliable and generalizable.
The Journal Club is a collaboration between the American Academy of Pediatrics - Section of Neonatal Perinatal Medicine and the International Society for Evidence-Based Neonatology (EBNEO.org).
Conflict of interest: The authors declare that they have no conflict of interest.
1. Stewart LA, Clarke M, Rovers M, Riley RD, Simmonds M, Stewart G, et al. Preferred Reporting Items for a Systematic Review and Meta-analysis of Individual Participant Data: The PRISMA-IPD Statement. JAMA. 2015 Apr 28;313(16):1657.
2. Doyle LW, Ehrenkranz RA, Halliday HL. Postnatal Hydrocortisone for Preventing or Treating Bronchopulmonary Dysplasia in Preterm Infants: A Systematic Review. Neonatology. 2010;98(2):111–7.
3. Doyle LW, Cheong JL, Ehrenkranz RA, Halliday HL. Early (< 8 days) systemic postnatal corticosteroids for prevention of bronchopulmonary dysplasia in preterm infants. Cochrane Neonatal Group, editor. Cochrane Database Syst Rev [Internet]. 2017 Oct 24 [cited 2019 Feb 18]; Available from: http://doi.wiley.com/10.1002/14651858.CD001146.pub5
4. Watterberg KL, Gerdes JS, Gifford KL, Lin H-M. Prophylaxis Against Early Adrenal Insufficiency to Prevent Chronic Lung Disease in Premature Infants. :8.
5. Watterberg KL, Gerdes JS, Cole CH, Aucott SW, Thilo EH, Mammel MC, et al. Prophylaxis of Early Adrenal Insufficiency to Prevent Bronchopulmonary Dysplasia: A Multicenter Trial. PEDIATRICS. 2004 Dec 1;114(6):1649–57.
6. Peltoniemi O, Kari MA, Heinonen K, Saarela T, Nikolajev K, Andersson S, et al. Pretreatment cortisol values may predict responses to hydrocortisone administration for the prevention of bronchopulmonary dysplasia in high-risk infants. J Pediatr. 2005 May;146(5):632–7.
7. Bonsante F, Latorre G, Iacobelli S, Forziati V, Laforgia N, Esposito L, et al. Early Low-Dose Hydrocortisone in Very Preterm Infants: A Randomized, Placebo-Controlled Trial. Neonatology. 2007;91(4):217–21.
8. Baud O, Maury L, Lebail F, Ramful D, El Moussawi F, Nicaise C, et al. Effect of early low-dose hydrocortisone on survival without bronchopulmonary dysplasia in extremely preterm infants (PREMILOC): a double-blind, placebo-controlled, multicentre, randomised trial. The Lancet. 2016 Apr;387(10030):1827–36.
9. Higgins JPT. Measuring inconsistency in meta-analyses. BMJ. 2003 Sep 6;327(7414):557–60.
Tables and Figures:
Table 1: Characteristics of studies included in IPD meta-analysis
|Characteristics||Watterberg et al||Petonmiemi et al||Bonsante et al ||Baud et al|
|Design||Double blind RCT||Double blind RCT||Double blind RCT||Double blind RCT|
|Gestational age (weeks)||N/A||230/7-295/7||240/7-295/7||240/7-275/7|
|Birth weight (grams) ||500 - 999 gm||501 - 1250 gm||500 - 1249 gm||n/a|
|Enrollment||12-48 hours of age ||< 36 hours of age||< 48 hours of age||< 24 hours of age|
|Dose & duration||1 mg/kg/d for 12d then 0.5 mg/kg/d for 3d||Tapered from 2 - 0.75 mg/kg/d over 10d||1 mg/kg/d divided into 2 doses/d for 9d then 0.5 mg/kg/d for 3d||1 mg/kg/d divided into 2 doses/d for 7d then 0.5 mg/kg/d for 3d|