Results

Ten studies were selected for inclusion in the review, eight of which were full publications30-37; two were abstracts at the time of the review.38 39 One of the abstracts was in the process of being submitted for publication. Additional information was sought from the authors of these abstracts and also of the published studies as required. The characteristics of the trials, as reported, are summarised in tables 1 and 2.

Most of the trials allowed for open treatment with dexamethasone at the discretion of the clinician after the treatment period. Only ventilator dependent patients were randomised in all trials.

The study design of each trial is shown in table 3. Of the 10 trials reviewed in this systematic review, in nine the intervention was blinded and in eight the outcomes were blinded; all the trials having more than 90% follow up.

MORTALITY

There was no significant difference (table 4) in mortality in the neonates randomised to dexamethasone in the overall analyses, the subgroups where dexamethasone was started within 36 hours of age (table 5), and the subgroup where surfactant was administered30 37-39 [ERR 0.98 (0.70, 1.36)]. However, in the subgroup analyses where dexamethasone was started within 7 to 14 days of postnatal age (table 5) there was a significant reduction in mortality [ERR 0.35 (0.16, 0.74), NNT 8 (4, 30)]. This suggests that on average eight infants would need to be treated with dexamethasone between 7 to 14 days of postnatal age to prevent one death. There was no significant heterogeneity between the results of the trials.

CHRONIC LUNG DISEASE AT 28 DAYS

In the overall31 33-36 38 39 (table 6) and all three subgroup analyses there was a significant reduction in CLD in the neonates given dexamethasone. The results suggest that eight infants would need to be treated with dexamethasone to prevent one infant developing CLD at 28 days in the overall analyses. In the subgroup where dexamethasone was started within 36 hours,30 31 33 38 39 the ERR and NNT were 0.64 (0.49, 0.83) and 8 (5–16), respectively. The ERR and NNT of the subgroup where dexamethasone was started between 7 and 14 days34-36 were 0.83 (0.72, 0.95) and 6 (4, 20), respectively. The findings in the surfactant subgroup30 31 34 35 39 did not differ from the overall analysis [ERR 0.73 (0.60- 0.90)]. There was significant heterogeneity in the results of the trials.

CLD AT 36 WEEKS OF POSTMENSTRUAL AGE

In the neonates given dexamethasone, the overall analysis30-32 34 35 39 showed a significant reduction in CLD (table 7). Similar results were obtained in the subgroup where surfactant30-32 34 35 38 39 was used [ERR 0.57 (0.43, 0.75)]. In the subgroup analyses34 35 where dexamethasone was started between 7 and 14 days of age, the ERR was 0.63 (0.47, 0.85) and only three infants would need to be treated with dexamethasone to prevent one infant developing CLD. In the subgroup30-33 38 39 where dexamethasone was started within 36 hours of age the ERR was 0.53 (0.33, 0.83) and 13 infants would need to be treated to prevent one infant developing CLD. There was no heterogeneity in the results of the trials.

DEATH OR CLD AT 28 DAYS

There was a significant reduction in risk of death or CLD in the overall and in all three subgroup analyses in the dexamethasone group. In the overall analyses the ERR was 0.83 (0.74, 0.93) and the NNT was 8 (5, 17). The ERRs and NNTs of the subgroups where dexamethasone was begun within 36 hours and between 7 and 14 days of postnatal age were 0.83 (0.71, 0.97), 9 (5, 36), and 0.83 (0.73, 0.93), 9 (5,60), respectively. The ERR and NNT of the surfactant subgroup were 0.89 (0.73, 0.97) and 11 (5, 31), respectively.

DEATH OR CLD AT 36 WEEKS OF POSTMENSTRUAL AGE

The risk of death or CLD at 36 weeks postmenstrual age was significantly reduced in the overall analyses, ERR 0.78 (0.66, 0.93), NNT 9 (5, 25) and in the subgroup where dexamethasone was started between 7 and 14 days of postnatal age, ERR 0.60 (0.44, 0.81), NNT 3 (2, 7). Thus to reduce the incidence of death or CLD by one on an average, only three infants would need to be treated with dexamethasone, starting between 7 and 14 days of postnatal age. Results of the surfactant subgroup were similar to those of the overall analysis.

DAYS ON MECHANICAL VENTILATION

Table 8 shows that there is a decrease in days on mechanical ventilation in the overall as well as the subgroup analyses. Overall weighted mean difference with 95% CI,−2.05 (−2.33, −1.78). Only five studies30-32 35 39 in the overall analyses, four studies30-32 39 in the subgroup where dexamethasone was started within 36 hours of age, and one study35 in the subgroup where dexamethasone was started between 7 and 14 days of age assessed this outcome. Four studies assessed this outcome in the surfactant subgroup.30 32 35 39

DAYS ON OXYGEN

There was no significant decrease in the number of days on supplemental oxygen, (table 8) with the overall weighted mean difference being 0.66 (−0.08, 1.40). However, in the subgroup35 where dexamethasone was started between 7 and 14 days there was a significant decrease in the days in supplemental oxygen, with the weighted mean difference being −59 (−72, −46), but only one trial assessed this outcome in the subgroup.

DAYS IN HOSPITAL

There was a significant reduction in hospital stay in the group that received dexamethasone (table 8), which was evident in the overall as well as the subgroup analyses. Overall weighted mean difference with 95% CI was −2.06 (−3.51, −1.68). Four studies31-33 35 in the overall analyses, three31-33 in the subgroup where dexamethasone was started within 36 hours of age, one study ³⁵ in the subgroup where dexamethasone was started between 7 and 14 days and two studies in the surfactant subgroup32 35 assessed this outcome.

DAYS TO REGAIN BIRTHWEIGHT

Only two studies30 33 have recorded this outcome. There was an increase in the number of days to regain birth weight in the group that received dexamethasone. The weighted mean difference was 3.9 days with a 95% CI of 1.8, 6.0 days.

HYPERTENSION

There was a significant increased risk of hypertension in the group that received dexamethasone in the overall (table 9) and in the subgroup analyses where dexamethasone was started between 7 and 14 days of postnatal age: ERRs 3.04 (1.34, 6.86), 3.78 (1.15, 12.38) and NNHs 42 (19, 212) 65 (16, 344), respectively. In the surfactant subgroup there were no significant differences in risk of hypertension [ERR 1.91 (0.84, 4.35)]. In all the trials that showed a significant increase, the hypertension was transient, with most patients recording normal blood pressures by day 10 of starting dexamethasone. None of the patients required continued anti-hypertensive treatment after treatment or at discharge.

HYPERGLYCAEMIA

The risk of hyperglycaemia due to dexamethasone was increased, but the risk was low and only just reached significance in the overall analyses,30 31 34-37 39 the ERR being 1.32 (1.04, 1,69) and NNH 14 (8, 100). This was not significant in the subgroup analyses: the ERRs were 1.34 (0.94, 1.91) and 1.30 (0.93, 1.83), respectively.

INTRAVENTRICULAR HAEMORRHAGE (ANY GRADE)

The pooled results showed no significant difference in effect of dexamethasone on the incidence of IVH all grades and grades 3 or 4 in the overall30 32 35 39 and subgroup analyses, the ERRs being 0.91(0.65, 1.29) and 1.06 (0.69, 1.61), respectively.

RETINOPATHY OF PREMATURITY (ROP) ANY STAGE

There was no significant difference in the incidence of ROP in the overall30 32-34 36 37 39 or any subgroup analyses; the ERRs were 0.92 (0.67, 1.26), 0.86 (0.57, 1.30), 0.94 (0.41, 2.13) and 0.85 (0.55, 1.30), respectively.

NECROTISING ENTEROCOLITIS, AIR LEAK SYNDROMES, AND SEPSIS

Pooled estimates of NEC,30-35 39 air leak syndromes,30-32 34 35 37 39 and sepsis30-37 39 showed no significant difference; the ERRs being 0.64 (0.37, 1.12), 0.80 (0.57, 1.12) and 0.89 (0.63, 1.25). The results were similar in all the subgroup analyses.

INSULIN TREATMENT

Insulin treatment for hyperglycaemia as an outcome was assessed in two studies30 32: ERR 1.68 (0.75, 3.78)

LONG TERM NEURODEVELOPMENTAL AND PULMONARY OUTCOMES

Cummings et al 37 assessed neurodevelopmental outcomes at 6 and 15 months. At 15 months corrected age, none (0/9) of the survivors from the 42 day course of dexamethasone had abnormalities on neurological examination, whereas more than 56% (5/9) of the survivors from the 18 day group and 40% (2/5) of the survivors from the control group had marked truncal hypotonia or cerebral palsy. Thus good neurodevelopmental outcome (normal exam and Bayley indexes > 84) were more often present in infants in the 42 day group (78%) than in 18 day group (22%) or the control group (40%). Follow up at 4 years40 showed better results in the 42 day group on all the McCarthy subscales. The problem with long term pulmonary and neurodevelopmental outcome is that crossover was allowed in most of the trials following the study period. Recent trials have used shorter courses and smaller doses, so long term follow up from these trials would be valuable.

The sensitivity analyses were done to look at the effect of different doses and duration of dexamethasone treatment on mortality and CLD. Included in the analyses were all the trials which used dexamethasone in doses starting at 0.5 mg/kg/day for 7–14 days.30 33 34 36 39 This was done to see if there were differences in outcomes when trials which used higher doses and longer courses were not included in the analyses. The results were similar to those of the overall analyses; the ERRs and NNTs for CLD at 28 days30 33 34 36 39 and 36 weeks of postmenstrual age30 34 39 were 0.57 (0.45, 0.73), 4 (3, 8), and 0.29 (0.15, 0.54), 5 (4, 10), respectively.

Discussion

This systematic review and meta-analyses of early dexamethasone to prevent CLD was based on the methods recommended by the Cochrane Collaboration.27 This involved an extensive search of published data and rigorous methodology. Reviewer bias was minimised by independent assessment and extraction of data by two authors. Where subgroup analyses were planned the criteria were established before the search or data analyses, in keeping with the initial objectives of the review.

There were some potential sources of bias in this systematic review. No attempt was made to trace unpublished trials. The small number of patients in the few trials reporting on outcomes such as days to regain birth weight, days in supplemental oxygen, and days in hospital, may affect the pooled results if no benefit or only a small effect due to lack of statistical power, rather than the absence of true effect is shown. On the other hand, some significant results such as increased risk of hypertension, which was transient, may be of doubtful clinical importance. Another limitation was the heterogeneity of the doses and duration of dexamethasone administration among the trials. However, this was addressed by sensitivity analyses.

In the overall pooled analysis of death30-39 there was no significant difference in risk when dexamethasone was administered. In the pooled subgroup34-37 where dexamethasone was started between 7 and 14 days, despite the small number of patients (91 patients in the intervention group and 94 in the control group), there was a significant reduction in risk of death, NNT 8 (4, 30). This significant finding has not been reported before. This difference in mortality between the groups does not seem to be due to differences in populations, but it cannot be precluded. The differences were derived from randomised controlled trials so the potential for bias is minimal.

There was a significant reduction in the risk of CLD both at 28 days and 36 weeks of postmenstrual age in both the overall analyses and all the subgroup analyses. The risk reduction at 36 weeks was greater when dexamethasone was started between 7 to 14 days of postnatal age [NNT 3 (2, 9)]: only three infants need to be treated with dexamethasone to prevent one infant developing CLD. In this subgroup only two of the four trials assessed this outcome and in one trial steroids were administered in pulses of three days out of every 10 until 36 weeks of postmenstrual age or until supplemental oxygen was no longer required. Three out of the four trials assessed CLD at 28 days in this subgroup. The findings of risk reduction, however, were consistent in all the trials except that of Shinwell et al 29 where the ERR was 1.17 (0.69, 1.99), which suggested a trend towards an increase in CLD if dexamethasone was administered. This finding was not significant, but this was the largest trial with 132 patients in the intervention group and 116 in the control group. The pooled results indicate that despite the increased risk in that trial there was still a significant risk reduction in the incidence of CLD both at 28 days and 36 weeks postmenstrual age in both the overall and the subgroup analyses. The reasons for the increased risk of CLD in the trial by Shinwellet al 29 are unclear. They may be related to the dose, duration, or the timing of administration. The heterogeneity in the results of CLD at 28 days is due to the results of this trial. There was no significant difference in the continuous outcome such as days in supplemental oxygen, despite a significant reduction in CLD. This may be due to skewed data with wide standard deviations. It would be more appropriate to report these continuous variables with medians and ranges rather than means and standard deviations to derive more meaningful information from these continuous variables.

The pooled estimates showed an increased risk of hypertension in the overall and in the subgroup where dexamethasone was started within 7 to 14 days of age. The increased risk was significant as the NNH was 42 (19, 212) in the overall analysis. The hypertension tended to be transient and was more apparent in the trials that used dexamethasone for longer than 14 days as well as higher doses.

There was a significant increase in risk of hyperglycaemia with NNH 14 (8, 100). This shows the wide confidence intervals, reflecting the small number of patients assessed for this outcome and the lack of strong evidence.

The sensitivity analyses showed that when dexamethasone was administered for 7 to 14 days, starting at 0.5 mg/kg/day the results were similar to those of the overall analyses except that NNT were lower for CLD at 28 days and 36 weeks of postmenstrual age [4 (3, 8), 5 (3, 9)]. Thus it seems that lower doses and shorter duration of dexamethasone treatment may be as effective as longer courses. However this needs to be further tested in randomised controlled studies. It may even be possible to use smaller doses than those used in the current studies.

There are concerns that dexamethasone may reduce cerebral DNA accretion, causing functional impairment.20 One year follow up of 12 infants given only two doses of hydrocortisone at birth showed increased frequencies of neurological and electroencephalographic abnormalities.41 The shortcomings of most other trials are the subsequent treatment with dexamethasone in the control group after the study period. Thus it is difficult to get a true picture of the neurodevelopmental outcomes from these studies. Mammel et al 42 followed up eight ventilator dependent infants treated with a three week course of dexamethasone; at 1 year they showed a small but non-significant reduction in the Bayley mental developmental index. Ohlsson43 found no difference in Bayley scores at 2 years of age in seven treated and nine control infants who were not crossed over. There was only one trial37 from this review that has published long term outcomes at 6, 15 months, and 4 years. However, there were only 9/13, 9/12, and 5/11 survivors in the 42 day, 18 day, and control groups, respectively. With animal data suggesting that dexamethasone may interfere with cerebral DNA accretion, and conflicting evidence from small number of follow up patients it is still not clear whether dexamethasone adversely affects long term neurodevelopmental outcomes and whether it is related to dose or duration of treatment.

The results of this review and the meta-analyses of early postnatal dexamethasone for prevention of CLD suggest a significant reduction in risk of CLD at 28 days and 36 weeks of postmenstrual age if dexamethasone is started within 14 days of postnatal age. However, if begun between 7 to 14 days of postnatal age there is also a dramatic reduction in risk of death as well as an increase in the magnitude of risk reduction of CLD. There is the potential of substantial differences in the populations being treated early and those being treated late and thus caution is warranted in interpreting the differences between the groups. The magnitude of effect is also greater in reducing CLD at 28 days and 36 weeks of postmenstrual age when administered for 7 to 14 days at doses starting from 0.5 mg/kg/day. The only side effects found that were clinically significant were transient hypertension, hyperglycaemia, and an increase in days to regain birthweight. The clinical importance of the increased risk was low (NNH) compared with the benefits (comparing the NNT). However caution is warranted in routine administration of dexamethasone at this stage despite the above results for the following reasons:

(1)

There are no significant long term studies of neurodevelopmental outcomes, in infants who received dexamethasone. There are ongoing concerns about myelination and somatic growth with dexamethasone. There are no significant long term studies of neurodevelopmental outcomes in infants who received dexamethasone. Due to the lack of the above it is difficult to assess the risk benefit ratio of the dramatic reduction in mortalityversus the long term neurodevelopmental outcome;

(2)

The above studies used different doses and duration of dexamethasone administration, therefore it is not clear which would be the minimum effective dose;

(3)

There are few studies stratified according to gestational age so it is difficult to identify the group who would benefit the most with acceptable side effects. However, it can be recommended that if it is deemed necessary to give dexamethasone, it can be started between 7–14 days postnatal age in doses starting at 0.5 mg/kg/day and weaned over 7 to 14 days.

 Future studies need to target infants most at risk of developing CLD. Stratification by gestational age would ensure that the most at risk group is targeted. Importantly, further studies to define the optimal dose and duration of dexamethasone are needed to reduce further the risk of side effects. Finally, long term neurodevelopmental and pulmonary outcomes of infants who received dexamethasone in optimal doses and not crossed over would need to be undertaken before dexamethasone could be recommended for routine administration to all at risk populations. Thus follow up should be inbuilt in all future primary studies.