only those studies where the women were randomised to either boxes supplied to the study centres. This was to be opened"or in an emergency". There was no information available in the study We calculated heterogeneity between trial results using an 1? test. manuscripts or from the authors as to how many times this en In multiple pregnancies, the number of babies was used as the velope was opened. Thus this study was given C, inadequate, for denominator for fetal and neonatal outcomes allocation concealment. Performance bias is unlikely to have oc- curred in the studies included in this review but ifit did it was most likely to have occurred in those where allocation concealment was DESCRIPTION OF STUDIES Thirteen of the included studies were placebo controlled(3255 Twenty-one studies met our inclusion criteria, with data available women and 3626 infants), with the majority of these studies using for 3885 women and 4269 infants(see Characteristics of included normal saline, or the vehicle of the corticosteroid preparation, as studies table). Six new studies have been included since the previ- the placebo. The remainder of the included studies used expectant ous review involving 802 women and 819 infants(Amorim 1999; management in the control arm Dexiprom 1999: Fekih 2002; Lewis 1996; Nelson 1985; Qublan Eight of the included studies allowed weekly repeat courses of study medication in their study protocols(821 women and 848 Six of the included studies used dexamethasone as the corticos- infants). These studies were included in the review. As stated above teroid in the treatment arm(1391 women and 1514 infants), while separate analysis of primary outcomes for those studies allowing 14 studies used betamethasone (2476 women and 2737 infants) a single course of study medication and those studies allowing and one study did not specify the corticosteroid used( Cararach weekly repeat courses of study medication was conducted post and infants The included studies were conducted over a wide range of gesta- In only six studies was evidence available to suggest that sample- tional ages, including those of extreme prematurity; obstetric in- size calculations had been performed prospectively(Amorim 1999; dications for recruitment were premature rupture of membranes, Collaborative 1981; Dexiprom 1999: Kari 1994; Silver 1996; spontaneous preterm labour and planned preterm delivery. Taeusch 1979). Intention-to-treat analysis was possible from study The included studies came from a range of healthcare systems and dara in only nine of the studies included in the review (Cararach treatment eras. Ten of the studies were conducted in the USA. 1991; Doran 1980: Gamsu 1989: Kari 1994; Liggins 1972b; Nel- with two studies conducted in Finland and one study from each son 1985; Parsons 1988: Qublan 2001; Teramo 1980).However, of the following countries; Brazil, Spain, South Africa, Canada, the remaining studies losses to fol Tunisia, UK, New Zealand, Jordan, and The Netherlands. Six of and less than 5%. There is no evidence to suggest that these ex- the included studies completed recruitment mainly in the 197 clusions occurred preferentially in one arm or the other of the (1753 women and 1994 infants), six of the included studies com- studies. The four studies( Collaborative 1981; Kari 1994; Liggins ed recruitment mainly in the 1980s(1100 women and 1173 2b: Schutte 1980)that reported long-term follow up after th neonatal period had their follow-up data included regardless of the infants),and nine of the included studies completed recruitment follow-up rate unless there was evidence of bias in follow-up rates mainly in the 1990s(1032 women and 1102 infan between the treatment and control groups; this was not found to METHODOLOGICAL QUALITY Three studies that were included in the previous review have been excluded from this update. Two(Papageorgiou 1979; Schmidt The methods of randomisation used in the included studies are 1984)were excluded because of greater than 20% postrandomisa- summarised in the Characteristics of included studies table. Eight tion exclusions. The third(Morales 1986)was excluded as it was studies used computer-generated or random number-generated quasi-randomised randomisation sequences with either coded drug boxes/vials or sealed envelopes used in order to conceal the randomisation se quence or study treatment. These studies were coded A for alloca- RESULTS tion concealment. Twelve studies either did not state the method of randomisation, or it was unclear, or the method of allocation Twenty-one studies involving 3885 women and 4269 infants were ncealment was not stated or unclear, and no further informa. included tion was available from the authors. These studies were coded B for allocation concealment. In the remaining study( Collaborative 1. Antenatal corticosteroids versus placebo or no treatment 1984). p n taining the trial allocation to the coded drug Primary outcomes hajor potential for bias was introduced by attaching (all included studies) Antenatal corticosteroids for accelerating fetal lung maturation for women at risk of preterm birth( Review) Copyright @2006 The Cochrane Collaboration. Published by John wiley& Sons, Ltdonly those studies where the women were randomised to either single or multiple doses. We calculated heterogeneity between trial results using an I² test. In multiple pregnancies, the number of babies was used as the denominator for fetal and neonatal outcomes. D E S C R I P T I O N O F S T U D I E S Twenty-one studies met our inclusion criteria, with data available for 3885 women and 4269 infants (see ’Characteristics of included studies’ table). Six new studies have been included since the previ￾ous review involving 802 women and 819 infants (Amorim 1999; Dexiprom 1999; Fekih 2002; Lewis 1996; Nelson 1985; Qublan 2001). Six of the included studies used dexamethasone as the corticos￾teroid in thetreatmentarm (1391women and 1514 infants),while 14 studies used betamethasone (2476 women and 2737 infants) and one study did not specify the corticosteroid used (Cararach 1991; 18 women and infants). The included studies were conducted over a wide range of gesta￾tional ages, including those of extreme prematurity; obstetric in￾dications for recruitment were premature rupture of membranes, spontaneous preterm labour and planned preterm delivery. The included studies came from a range of healthcaresystems and treatment eras. Ten of the studies were conducted in the USA, with two studies conducted in Finland and one study from each of the following countries; Brazil, Spain, South Africa, Canada, Tunisia, UK, New Zealand, Jordan, and The Netherlands. Six of the included studies completed recruitment mainly in the 1970s (1753 women and 1994 infants), six of the included studies com￾pleted recruitment mainly in the 1980s (1100 women and 1173 infants), and nine of the included studies completed recruitment mainly in the 1990s (1032 women and 1102 infants). M E T H O D O L O G I C A L Q U A L I T Y The methods of randomisation used in the included studies are summarised in the’Characteristics of included studies’ table. Eight studies used computer-generated or random number-generated randomisation sequences with either coded drug boxes/vials or sealed envelopes used in order to conceal the randomisation se￾quence or study treatment. These studies werecoded A for alloca￾tion concealment. Twelve studies either did not state the method of randomisation, or it was unclear, or the method of allocation concealment was not stated, or unclear, and no further informa￾tion was available from the authors. These studies were coded B forallocation concealment. In theremaining study (Collaborative 1984), a major potential for bias was introduced by attaching a sealed envelope containing the trial allocation to the coded drug boxes supplied to the study centres. This was to be opened “only in an emergency”. There was no information availablein thestudy manuscripts or from the authors as to how many times this en￾velope was opened. Thus this study was given C, inadequate, for allocation concealment. Performance bias is unlikely to have oc￾curred in thestudies included in thisreview but if it did it was most likely to have occurred in those where allocation concealment was inadequate. Thirteen of the included studies were placebo controlled (3255 women and 3626 infants), with the majority of thesestudies using normal saline, or the vehicle of the corticosteroid preparation, as the placebo. Theremainder of theincluded studies used expectant management in the control arm. Eight of the included studies allowed weekly repeat courses of study medication in their study protocols (821 women and 848 infants).Thesestudieswereincluded in thereview.Asstated above, separate analysis of primary outcomes for those studies allowing a single course of study medication and those studies allowing weekly repeat courses of study medication was conducted post hoc. In only six studies was evidence available to suggest that sample￾sizecalculations had been performed prospectively (Amorim1999; Collaborative 1981; Dexiprom 1999; Kari 1994; Silver 1996; Taeusch 1979). Intention-to-treatanalysis was possiblefrom study data in only nine of the studies included in the review (Cararach 1991; Doran 1980; Gamsu 1989; Kari 1994; Liggins 1972b; Nel￾son 1985; Parsons 1988; Qublan 2001; Teramo 1980). However, in the remaining studies losses to follow up were generally small and less than 5%. There is no evidence to suggest that these ex￾clusions occurred preferentially in one arm or the other of the studies. The four studies (Collaborative 1981; Kari 1994; Liggins 1972b; Schutte 1980) that reported long-term follow up after the neonatal period had their follow-up dataincluded regardless of the follow-up rate unless there was evidence of bias in follow-up rates between the treatment and control groups; this was not found to be the case. Three studies that were included in the previous review have been excluded from this update. Two (Papageorgiou 1979; Schmidt 1984) were excluded because of greater than 20% postrandomisa￾tion exclusions. The third (Morales 1986) was excluded as it was quasi-randomised. R E S U L T S Twenty-onestudies involving 3885 women and 4269 infants were included. 1. Antenatal corticosteroids versus placebo or no treatment (all included studies) Primary outcomes Antenatal corticosteroids for accelerating fetal lung maturation for women at risk of preterm birth (Review) 7 Copyright © 2006 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd 第 101 页