ry outcomes for infants delivered greater than seven days after 1996: Liggins 1972b; Nelson 1985: Morales 1989)that reported treatment with antenatal corticosteroids. In fact, birthweight is this outcome in the subgroup of women with premature rupture reduced in this subgroup. This lack of benefit is not a new finding, of membranes at time of the first glucocorticoid dose. However, and in the past has lead to the practice of repeating courses of in one study(Liggins 1972b)birthweight was reduced in those antenatal corticosteroid weekly if women remained undelivered. neonates exposed to corticosteroids who experienced rupture of Eight of the included studies in this review used treatment pro- membranes for greater than 24 or greater than 48 hours. The clin- tocols that included repeated weekly courses. These studies were ical significance of this finding remains unclear and it may reflect ncluded in the review as they examined corticosteroid treatment a type one error. us no corticosteroid treatment, but they were analysed sepa- rately, post hoc, as a sensitivity analysis to determine if they biased Currently, there is not enough evidence to support the use of an the overall results. This does not appear to be the case. However, tenatal corticosteroids in multiple pregnancies. Although data for it would be misleading to draw conclusions from this subgroup most primary es were available from the two largest stud- analysis concerning the risks or benefits of repeat courses of ante- ies( Collaborative 1981: Liggins 1972b)the numbers of multiple natal corticosteroids. Information concerning the number of re- pregnancies included in this review remained small(n= 252 in- peat courses used in individual studies was not provided and there fants). A further 10 studies( Dexiprom 1999; Doran 1980; Fekih may have been few repeat courses. It would be meaningful to per- 2002; Gamsu 198%; Garite 1992; Kari 1994: Schutte 1980; Sil- form an individual patient data analysis to look at the relationship ver 1996; Taeusch 1979: Teramo 1980)included in this review between the interval from first dose to delivery and outcome, and had recruited an additional 252 infants from multiple pregnan- how this was influenced by factors such as whether corticosteroids cies. Analysis of these data may help clarify the risks and benefits were given and he ny doses each individual got. The effect of corticosteroids in multiple pregnancies without the need for of repeated courses of antenatal corticosteroids is the subject of further trials. w( Crowther 2000), which suggests that although repeated courses reduce the severity of neonatal lung disease, there No randomised studies have directly compared the two common are insufficient data to exclude other beneficial or harmful effects types of corticosteroid used in clinical practice, betamethasone and o the mother or infant. The recommendation of those authors is dexamethasone. Although this review suggests that betamethasone to await the outcome of trials looking at the long-term effects of treatment causes a larger reduction in RDS than dexamethasone the reasons for this may be a different background prevalence of RDS in the different study populations examined and not due ticosteroids in the subgroup of women with hypertension syn- to greater efficacy of the betamethasone. A large non-randomised dromes. In such women antenatal corticosteroids reduce the risk retrospective study has suggested that infants exposed to antenatal of neonatal death, RDS and cerebroventricular haemorrhage betamethasone have less neonatal cystic periventricular leukoma- their offspring. In the previous review, fetal death was increased lacia(which is strongly associated with later cerebral palsy)than infants exposed to antenatal dexamethasone(Baud 1999). How- reroids. However, since this review, an additional study has con- ever there is no evidence from this review of a difference in inci- dence of later cerebral palsy in infants exposed to either antenatal tributed data(Amorim 1999). Furthermore, in this new review, in betamethasone or dexamethasone. Further research is required dividual participant data were available for the one study that had determine the optimal dose and drug for use in this situation. ntributed to the previous result( Liggins 1972b). This study had never been completely analysed in full or by intention to treat A it is responsible for approximately 30% of all women and infants date because we recognise that clinicians will want to see this infor- randomised to corticosteroids, its inclusion in this new manner mation for its practical implications and also because it has been increases the validity of the review's conclusions. This new analysis the subject of much conjecture following the first review. Caution, of the Liggins study resulted in further cases of fetal death Deng must however, be expressed in the interpretation of the subgroup assigned to in the control arm of the study analyses conducted in this review. There is the possibility of a type one error due to the number of analyses conducted Furthermo In this new review, antenatal corticosteroids are shown to be ben. the subgroups of gestational age at delivery, length of premature eficial in the subgroup of infants whose mothers have premature rupture of membranes and entry to delivery interval, involve pos- pture of membranes. Neonatal death, RDS, cerebroventricular trandomisation variables Conducting subgroup analysis based rrhage, necrotising enterocolit duration of neonatal postrandomisation variables is liable to considerable bias as the respiratory support are all significantly reduced by corticosteroid variable on which the subgroup is based may be affected by the treatment in this subgroup without an increase in either maternal rvention that occurs at randomisation. The clinician should or neonatal infection. Birthweight was not significantly altered by therefore not draw too many conclusions from the results of the orticosteroid treatment in the five studies(Dexiprom 1999; Lewis subgroup analyses Antenatal corticosteroids for accelerating fetal lung maturation for women at risk of preterm birth( Rev Copyright @2006 The Cochrane Collaboration. Published by John wiley& Sons, Ltdmary outcomes for infants delivered greater than seven days after treatment with antenatal corticosteroids. In fact, birthweight is reduced in this subgroup. This lack of benefit is not a new finding, and in the past has lead to the practice of repeating courses of antenatal corticosteroid weekly if women remained undelivered. Eight of the included studies in this review used treatment pro￾tocols that included repeated weekly courses. These studies were included in the review as they examined corticosteroid treatment versus no corticosteroid treatment, but they were analysed sepa￾rately, post hoc, as a sensitivity analysis to determine if they biased the overall results. This does not appear to be the case. However, it would be misleading to draw conclusions from this subgroup analysis concerning the risks or benefits of repeat courses of ante￾natal corticosteroids. Information concerning the number of re￾peat courses used in individual studies was not provided and there may have been few repeat courses. It would be meaningful to per￾form an individual patient data analysis to look at the relationship between the interval from first dose to delivery and outcome, and how this was influenced by factors such as whether corticosteroids were given and how many doses each individual got. The effect of repeated courses of antenatal corticosteroids is the subject of a separate review (Crowther 2000), which suggests that although repeated courses reduce theseverity of neonatal lung disease, there are insufficient data to exclude other beneficial or harmful effects to the mother or infant. The recommendation of those authors is to await the outcome of trials looking at the long-term effects of repeated courses of antenatal corticosteroids. This review should dispelconcerns about the use of antenatal cor￾ticosteroids in the subgroup of women with hypertension syn￾dromes. In such women antenatal corticosteroids reduce the risk of neonatal death, RDS and cerebroventricular haemorrhage in their offspring. In the previous review, fetal death was increased amongst offspring of such women treated with antenatal corticos￾teroids. However, since this review, an additional study has con￾tributed data(Amorim 1999). Furthermore, in this newreview, in￾dividual participant data were available for the one study that had contributed to the previous result (Liggins 1972b). This study had never been completely analysed in full or by intention to treat. As it is responsible for approximately 30% of all women and infants randomised to corticosteroids, its inclusion in this new manner increases the validity of thereview’sconclusions. This new analysis of the Liggins study resulted in further cases of fetal death being assigned to women with hypertension syndromes in the control arm of the study. In this new review, antenatal corticosteroids are shown to be ben￾eficial in the subgroup of infants whose mothers have premature rupture of membranes. Neonatal death, RDS, cerebroventricular haemorrhage, necrotising enterocolitis and duration of neonatal respiratory support are all significantly reduced by corticosteroid treatment in this subgroup without an increase in either maternal or neonatal infection. Birthweight was not significantly altered by corticosteroid treatment in the fivestudies (Dexiprom 1999; Lewis 1996; Liggins 1972b; Nelson 1985; Morales 1989) that reported this outcome in the subgroup of women with premature rupture of membranes at time of the first glucocorticoid dose. However, in one study (Liggins 1972b) birthweight was reduced in those neonates exposed to corticosteroids who experienced rupture of membranes for greater than 24 or greater than 48 hours. The clin￾ical significance of this finding remains unclear and it may reflect a type one error. Currently, there is not enough evidence to support the use of an￾tenatal corticosteroids in multiple pregnancies. Although data for most primary outcomes were available from the two largest stud￾ies (Collaborative 1981; Liggins 1972b) the numbers of multiple pregnancies included in this review remained small (n = 252 in￾fants). A further 10 studies (Dexiprom 1999; Doran 1980; Fekih 2002; Gamsu 1989; Garite 1992; Kari 1994; Schutte 1980; Sil￾ver 1996; Taeusch 1979; Teramo 1980) included in this review had recruited an additional 252 infants from multiple pregnan￾cies. Analysis of these data may help clarify the risks and benefits of corticosteroids in multiple pregnancies without the need for further trials. No randomised studies have directly compared the two common types ofcorticosteroid used in clinical practice, betamethasoneand dexamethasone. Although thisreviewsuggeststhat betamethasone treatment causes a larger reduction in RDS than dexamethasone, the reasons for this may be a different background prevalence of RDS in the different study populations examined and not due to greater efficacy of the betamethasone. A large non-randomised retrospective study has suggested that infants exposed to antenatal betamethasone have less neonatal cystic periventricular leukoma￾lacia (which is strongly associated with later cerebral palsy) than infants exposed to antenatal dexamethasone (Baud 1999). How￾ever there is no evidence from this review of a difference in inci￾dence of later cerebral palsy in infants exposed to either antenatal betamethasone or dexamethasone. Further research is required to determine the optimal dose and drug for use in this situation. We have included the results of the subgroup analyses in this up￾date because werecognisethatclinicians will want to seethis infor￾mation for its practical implications and also because it has been thesubject of much conjecturefollowing the first review. Caution, must however, be expressed in the interpretation of the subgroup analyses conducted in this review. Thereis the possibility of atype one error due to the number of analyses conducted. Furthermore, the subgroups of gestational age at delivery, length of premature rupture of membranes and entry to delivery interval, involve pos￾trandomisation variables. Conducting subgroup analysis based on postrandomisation variables is liable to considerable bias as the variable on which the subgroup is based may be affected by the intervention that occurs at randomisation. The clinician should therefore not draw too many conclusions from the results of the subgroup analyses. Antenatal corticosteroids for accelerating fetal lung maturation for women at risk of preterm birth (Review) 13 Copyright © 2006 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd 第 107 页