Clinical research must determine whether treatments enhance lives, make little difference, cause significant harm, or do several of these things. This is well illustrated by the epidemic of blindness due to retrolental fibroplasia that affected thousands of preterm babies in the 1950s.1 Although oxygen was accepted as lifesaving in severe respiratory distress syndrome, randomised controlled trials showed that its unrestricted use could also cause permanent visual impairment. The risk is minimised with modern oxygen therapy, which is strictly controlled. The lesson is that new treatments need to be tested with randomised trials that are large enough and with follow ups long enough to provide robust data on all clinically important endpoints.1,2 Dexamethasone for chronic lung disease in preterm infants may be a similar case where we need better data from larger trials with longer follow up.
In the past two decades dexamethasone has gained wide acceptance in routine practice for the postnatal treatment or prevention of chronic lung disease in preterm infants. Although dexamethasone is commonly associated with transient side effects, several randomised trials have shown that it rapidly reduces oxygen requirements and shortens the duration of ventilation.3 However, the longer term impact of postnatal dexamethasone on mortality and morbidity in survivors is less clear.
The table shows hospital mortality in all randomised studies of postnatal dexamethasone which have been identified and critically appraised according to the methods of the Cochrane Collaboration. These are arranged in groups, prespecified by the Cochrane reviewers, by onset of treatment: (a) early postnatal (<96 hours),4 (b) moderately early postnatal (7-14 days),5 and (c) delayed (>3 weeks).6 There was a trend to lower mortality after moderately early dexamethasone, but not after early or delayed therapy. Of concern, however, is the suggestion of a risk of cerebral palsy.7–9 The table summarises the best currently available evidence, including follow up data not yet available in the Cochrane Library.
Two trials suggest an increased risk of morbidity after 12 months, defined as cerebral palsy—one of them after early dexamethasone and one of them after delayed dexamethasone. Two further trials of delayed dexamethasone show a trend to increased risk and one of moderately early treatment shows no increased risk, though two of these trials were very small. As there is no evidence that early or delayed dexamethasone increases survival, any excess in cerebral palsy could not be attributed to increased salvage of prenatally impaired infants. Furthermore, the table may underestimate the true benefits and risks of dexamethasone, as in some studies control infants “crossed over” to receive open label dexamethasone, potentially blunting differences in outcome. The picture may be further complicated by publication bias, as so far only five studies have described neurodevelopmental outcome. We do not yet know whether these studies have been preferentially published because of the disquieting implications.
The paucity of follow up data is a major problem, however, and highlights the need to invest in better national and regional organisation of routine follow up, so perinatal studies can report follow up data as promptly as possible.10 Only then can we determine whether or not dexamethasone is a lifesaving therapy with potentially serious complications, like oxygen.
A definitive randomised controlled trial of dexamethasone, started in infants at high risk between 7 and 14 days at a lower dose than usual, now merits special priority, since these are the infants whom dexamethasone seems most likely to benefit. Until more follow up data are available clinicians who use dexamethasone liberally or in infants without life threatening lung disease would be wise to review their practice against the best currently available evidence. In any trial clinicians should discuss the need for longer term follow up when seeking initial consent and forge a continuing partnership with the parents to achieve this.
The “short termism” that is satisfied with surrogate measures as a substitute for longer term outcomes seems increasingly difficult to justify. Clinicians may be pleased that dexamethasone allows babies to come off supplemental oxygen more quickly, but from the parents' perspective this is a minor issue. For parents the key question is, “Will this treatment increase the chances of survival, and its quality?” Clinical trials should assess all clinically relevant outcomes. We need randomised trials which report longer term follow up to be sure that patients are not paying a long term price for some relatively minor short term gains.
Table.
Hospital mortality, and morbidity after 12 months, in randomised studies of postnatal dexamethasone for preterm infants at risk of chronic lung disease
| Onset of dexamethasone therapy | Hospital mortality in treated infants | Hospital mortality in control infants | Relative risk* of hospital mortality (95% CI) | Morbidity in treated survivors | Morbidity in control survivors | Relative risk* of morbidity (95% CI) |
|---|---|---|---|---|---|---|
| Early (<96 hours) | ||||||
| Baden 1972 | 7/22 | 8/22 | 0.88 (0.39 to 1.97) | - | - | - |
| Rastogi 1996 | 4/36 | 2/34 | 1.89 (0.43 to 8.47) | - | - | - |
| Sanders 1994 | 2/19 | 7/21 | 0.32 (0.08 to 1.14) | - | - | - |
| Shinwell 1996 | 31/132 | 22/116 | 1.24 (0.77 to 2.01) | - | - | - |
| Sinkin 1998 | 40/189 | 33/195 | 1.25 (0.83 to 1.89) | - | - | - |
| Subhedar 1997 | 9/21 | 8/21 | 1.13 (0.54 to 2.36) | - | - | - |
| Suske 1996 | 1/14 | 1/12 | 0.86 (0.10 to 7.75) | - | - | - |
| Tapia 1998 | 17/55 | 18/54 | 0.93 (0.54 to 1.60) | - | - | - |
| Yeh 1990 | 3/28 | 8/29 | 0.39 (0.12 to 1.20) | - | - | - |
| Yeh 19987 | 46/132 | 42/130 | 1.08 (0.77 to 1.52) | 25/63 | 12/70 | 2.31 (1.30 to 4.22) |
| Total | 160/648 | 149/634 | 1.05 (0.87 to 1.28) | 25/63 | 12/70 | 2.31 (1.30 to 4.22) |
| 7-14 days | ||||||
| Brozanski 1995 | 4/39 | 9/39 | 0.44 (0.15 to 1.24) | - | - | - |
| Cummings 1989 | 7/25 | 6/11 | 0.51 (0.23 to 1.21) | 5/18 | 2/ 5 | 0.69 (0.19 to 2.57) |
| Durand 1995 | 2/23 | 4/20 | 0.43 (0.10 to 1.84) | - | - | - |
| Kari 1993 | 1/17 | 3/24 | 0.47 (0.07 to 2.97) | - | - | - |
| Kovacs 1998 | 8/30 | 5/30 | 1.60 (0.62 to 4.24) | - | - | - |
| Total | 22/134 | 27/124 | 0.66 (0.40 to 1.09) | 5/18 | 2/ 5 | 0.69 (0.19 to 2.57) |
| Delayed | ||||||
| Ariagno 1987 | 3/10 | 2/11 | 1.65 (0.39 to 7.21) | - | - | - |
| Avery 1985 | 3/8 | 2/8 | 1.5 (0.38 to 6.32) | - | - | - |
| Jones 1995† | 25/143 | 25/142 | 0.99 (0.60 to 1.64) | 20/100 | 18/109 | 1.21 (0.69 to 2.14) |
| Harkavy 1989 | 1/9 | 2/12 | 0.67 (0.09 to 4.41) | - | - | - |
| Kazzi 1990 | 2/12 | 0/11 | 4.62 (0.25 to 86.72) | - | - | - |
| Ohlsson 1992 | 1/12 | 0/13 | 3.23 (0.14 to 72.46) | - | - | - |
| Vincer 1998 | 2/11 | 1/9 | 1.64 (0.25 to 11.59) | 4/9 | 2/8 | 1.78 (0.50 to 7.08) |
| O'Shea 19999 | 7/57 | 16/61 | 0.47 (0.21 to 1.02) | 12/48 | 3/45 | 3.75 (1.24 to 11.87) |
| Total | 44/262 | 48/267 | 0.94 (0.65 to 1.36) | 36/157 | 23/162 | 1.61 (1.00 to 2.61) |
Relative risk is the ratio of mortality in treated infants divided by that in control infants (calculated using Arcus Quickstat with Haldane approximation if a cell had no cases).
40% of control infants received dexamethasone, potentially blunting differences in outcome.
All studies started with a dose of 0.5 mg/kg/day, reducing and completed in 3-42 days.
Citations for all studies except references 7 and 9 are available in reference 3.
References
- 1.Silverman WA. Retrolental fibroplasia: a modern parable. New York: Grune and Stratton; 1980. [Google Scholar]
- 2.Johnson A. Randomised controlled trials in perinatal medicine: 3. Identifying and measuring endpoints in randomised controlled trials. Br J Obstet Gynaecol. 1997;104:768–771. doi: 10.1111/j.1471-0528.1997.tb12017.x. [DOI] [PubMed] [Google Scholar]
- 3.Halliday HL. Clinical trials of postnatal corticosteroids: inhaled and systemic. Biol Neonate. 1999;76 (suppl 1):29–40. doi: 10.1159/000047044. http://BioMedNet.com/karger .( http://BioMedNet.com/karger) ) [DOI] [PubMed] [Google Scholar]
- 4.Halliday HL, Ehrenkranz RA Cochrane Collaboration, editors. Cochrane Library. Issue 4. Oxford: Update Software; 1999. Early postnatal (< 96 hours) corticosteroids for preventing chronic lung disease in preterm infants. [Google Scholar]
- 5.Halliday HL, Ehrenkranz RA Cochrane Collaboration, editors. Cochrane Library. Issue 4. Oxford: Update software; 1999. Moderately early (7-14 days) postnatal corticosteroids for preventing chronic lung disease in preterm infants. [Google Scholar]
- 6.Halliday HL, Ehrenkranz RA Cochrane Collaboration, editors. Cochrane Library. Issue 4. Oxford: Update Software; 1999. Delayed (> 3 weeks) postnatal corticosteroids for chronic lung disease in preterm infants. [Google Scholar]
- 7.Yeh TF, Lin YJ, Huang CC, Chen YJ, Lin CH, Lin HC, et al. Early dexamethasone therapy in preterm infants: a follow-up study. Pediatrics. 1998;101(5):E7. doi: 10.1542/peds.101.5.e7. www.pediatrics.org/cgi/content/full/101/5/e7 ( www.pediatrics.org/cgi/content/full/101/5/e7) ) [DOI] [PubMed] [Google Scholar]
- 8.Greenough A. Gains and losses from dexamethasone for neonatal chronic lung disease. Lancet. 1998;352:835–836. doi: 10.1016/S0140-6736(05)60002-5. [DOI] [PubMed] [Google Scholar]
- 9.O'Shea TM, Kothadia JM, Klinepeter KL, Goldstein DJ, Jackson BG, Weaver RG, 3rd, et al. 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: outcome of study participants at 1-year adjusted age. Pediatrics 1999;10415-21. [DOI] [PubMed]
- 10.Field D. Evidence in perinatal medicine: enough of trial and error? Arch Dis Child Fetal Neonatal Ed. 1999;81:F161. doi: 10.1136/fn.81.3.f161. [DOI] [PMC free article] [PubMed] [Google Scholar]