DOI: 10.1111/j.1471-0528.2010.02686.x www.bjog.

org

General obstetrics

Caesarean section surgical techniques: a randomised factorial trial (CAESAR)*

The CAESAR study collaborative group
Correspondence: Professor P Brocklehurst, National Perinatal Epidemiology Unit, University of Oxford, Headington, Oxford OX3 7LF, UK. Email peter.brocklehurst@npeu.ox.ac.uk

Accepted 27 June 2010.

Objective In women undergoing delivery by caesarean section, do the following alternative surgical techniques affect the risk of adverse outcomes: single- versus double-layer closure of the uterine incision; closure versus nonclosure of the pelvic peritoneum; liberal versus restricted use of a subrectus sheath drain? Design Pragmatic, 2 2 2 factorial randomised controlled trial. Setting Hospitals in the UK and Italy providing intrapartum care. Population Women undergoing their rst caesarean section. Methods The interventions were alternative approaches to the

Results A total of 3033 women were recruited. Overall, the risk of

maternal infectious morbidity was 17%. For each pair of interventions, there were no differences between the arms of the trial for the primary outcome: single- versus double-layer closure of the uterine incision [relative risk (RR) = 1.00, 95% condence interval (95% CI) = 0.851.18]; closure versus nonclosure of the pelvic peritoneum (RR = 0.92, 95% CI = 0.781.08); liberal versus restricted use of a subrectus sheath drain (RR = 0.92, 95% CI = 0.781.09). There were no differences in any of the secondary morbidity outcomes and no signicant adverse effects of any of the techniques used.
Conclusions These results have implications for clinical practice, particularly in relation to current guidance on the closure of the peritoneum, which suggests that nonclosure is preferable. The potential effects of these different surgical techniques on longer term outcomes, including the functional integrity of the uterine scar during subsequent pregnancies, are now becoming increasingly important for guiding clinical practice.
Keywords Caesarean section, randomised controlled trial.

three aspects of the caesarean section operation. A telephone randomisation service was used. Surgeons could not be masked to allocation, but women were unaware of which allocations had been used. The analysis was by intention-to-treat, with a prespecied subgroup analysis for women in labour or not in labour at the time of caesarean section.
Main outcome measures Maternal infectious morbidity.

Please cite this paper as: The CAESAR study collaborative group. Caesarean section surgical techniques: a randomised factorial trial (CAESAR). BJOG 2010;117:13661376.

Introduction
Caesarean section is one of the most common operations undertaken worldwide. In Latin America, the median rate of caesarean delivery is 33%.1 The global rate of caesarean section is not known, but, if it is 10%, 13 million caesarean sections are performed each year, equivalent to 24 each minute. The surgical techniques used at caesarean section vary between surgeons, and few of these techniques have been
* Trial registration number: ISRCTN 11849611. The Caesar study collaborative group are listed in the Acknowledgements section.

evaluated in randomised controlled trials. The existing trials have been reviewed and, of 44 technical aspects of the operation, six had good evidence and three had fair evidence to make clear recommendations about their use.2 A survey of practice in the UK in 1999 asked obstetricians to nominate aspects of the caesarean section operation that they would like to see evaluated in a randomised trial.3 This report contains the outcome of this trial.

Methods
The CAESAR study was a pragmatic, 2 2 2 factorial, multicentre, randomised controlled trial, designed to evaluate two alternative approaches to three aspects of the technique of caesarean section.

1366

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

Caesarean section surgical techniques: a randomised factorial trial

The trial assessed the following null hypotheses. In women undergoing delivery by caesarean section, no differences would be detected with respect to maternal infectious morbidity when comparing the following three pairs of alternative surgical techniques: single- versus double-layer closure of the uterine incision; closure versus nonclosure of the pelvic peritoneum; liberal versus restricted use of a subrectus sheath drain. Women were eligible if they were undergoing delivery by their rst caesarean section, this was planned to be performed through the lower uterine segment and there was no clear indication for any particular technique to be used. The only contraindication to participation was an age of <16 years. All women provided written informed consent. A telephone randomisation service was employed to allocate the interventions using a minimisation algorithm to ensure comparability between women with respect to three prognostic factors: participating centre; in labour or not in labour; single or multiple pregnancy. Participants were enrolled by the clinicians providing care for women about to undergo their caesarean section. The allocation provided by the randomisation service was made available to the operating surgeon prior to the onset of surgery. Details of the clinical management used are described in the CAESAR study protocol (www.npeu.ox.ac.uk/trials/ CAESAR). In summary, the trial interventions were as follows. 1 Single- versus double-layer uterine closure. Single-layer closure involves bringing both edges of the uterine incision together with a single layer of sutures. In double-layer closure, the uterine incision is closed with two layers of sutures. The rst layer opposes the endometrial aspect of the uterine muscle layer and the second brings together the serosal layer. 2 Closure of the peritoneum. This refers to closure of the pelvic peritoneum (the layer of peritoneum overlying the uterine incision). 3 Liberal versus restricted use of a subsheath drain. We recognised that clinicians would feel uncomfortable using a subsheath drain if the abdominal wound had excellent haemostasis. Similarly, if the wound continued to ooze following appropriate haemostatis procedures, clinicians would use a drain. Hence, the intervention was to liberal or restricted use of the drain. All nonallocated surgical elements and all other aspects of the caesarean section procedure were performed at the discretion of the surgeon. In particular, there were no restrictions on the type of suture material that could be used, and standard measures to effect haemostasis were employed regardless of the allocated intervention. Each participating centre initiated and maintained a training and

accreditation programme which ensured that all personnel involved in undertaking caesarean sections were competent in the techniques being compared before they were able to recruit women into the trial. During the training, it was made clear that obstetricians should not vary other aspects of their surgical approach depending on the allocation. For example, regardless of whether the woman was allocated single- or double-layer uterine closure, the suture material used should be the same. The following outcomes were collected from all women participating in the trial up to the time of the 6-week follow-up assessment after the caesarean section. The primary outcome was maternal infectious morbidity, dened as having one or more of the following: (i) antibiotic use for maternal febrile morbidity during the postnatal hospital stay; (ii) endometritis; (iii) wound infection treated with antibiotics. Although not all the trial interventions might be expected to affect all components of the primary outcome equally, factorial trials have largely used a single primary outcome for all of the factorial components. It was anticipated that each of the trial interventions would be expected to contribute to a substantial part of the composite primary outcome, making it justiable to have a composite consisting of these aspects of infectious morbidity. For example, the use of a subsheath drain may be expected to make a greater difference than single- versus double-layer closure of the uterus to the risk of wound infection, but single- versus double-layer closure of the uterus may be expected to make a greater difference to the risk of endometritis. Secondary outcomes included the individual components of the primary outcome, as well as further operative procedures on the wound, pain, blood transfusion, breastfeeding at hospital discharge and at 6 weeks, and other severe or unexpected maternal morbidity.

Sample size
In a meta-analysis of 68 trials of antibiotic prophylaxis at caesarean section involving over 10 000 women, the incidence of febrile morbidity/endometritis in the antibiotic prophylaxis group was 13%.4 Prior to the start of the trial, we therefore anticipated that the incidence of antibiotic use for febrile morbidity would be approximately 12%. This combined primary outcome of febrile morbidity/endometritis was dened as either any antibiotics prescribed as treatment for maternal fever (temperature of >39C on any occasion or >38C on two or more successive days. Antibiotics prescribed for prophylaxis, such as those given for Group B streptococcus carriage, were not included), or any clinical diagnosis of endometritis made in hospital, which was treated with antibiotics within 6 weeks of the caesarean section. A sample size of 3500 women was therefore necessary to demonstrate a change in the incidence of the primary

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

1367

CAESAR study collaborative group

outcome between any pair of arms from 12% to 9%, with 80% power and a two-sided signicance level of 5%. The sample size estimation was based on the detection of differences in the main effects of the interventions and not the interactions between them.

Data analysis
Data were collected from the hospital notes up to the time of hospital discharge. Women were then sent a questionnaire to complete 6 weeks after the date of caesarean section. In this questionnaire, they were asked to record whether antibiotics or additional painkillers had been prescribed during the postnatal period and, if so, to give the reason. Responses were coded by two clinicians (PB and EA), independently, without knowledge of the trial allocation. Patients were analysed in the groups to which they were assigned, regardless of deviation from the protocol or treatment received. Comparative statistical analysis entailed the calculation of the relative risk (RR) plus the 95% condence interval (95% CI) for the primary outcome and 99% CI for the secondary outcomes to take account of multiple comparisons. Pairwise interactions between the different interventions were examined. Most of the results are presented using the numbers allocated to each pair of inter-

ventions including all those allocated, irrespective of allocation to the other pairs of interventions (i.e. the six margins of the 2 2 2 table).5,6 In addition, three prespecied subgroup analyses were intended, based on the primary and key secondary outcomes, stratied by in labour or not in labour caesarean section, single and multiple pregnancy and compliance with the subsheath drain comparison by quartile of compliance. An independent Data Monitoring and Ethics Committee (DMEC) monitored the effectiveness and safety on a regular basis (at least annually). DMEC employed the HaybittlePeto approach for interim analyses using three standard errors as the cut-off for early cessation, preserving the type I error rate across the trial.7

Research ethics committee approval

The trial sponsor was the University of Oxford. Multicentre Research Ethics Committee approval for the trial was given by the Eastern MREC (99/5/62), and all participants gave informed consent to participate. The trial was registered (ISRCTN number 11849611).

Role of the funding source

The trial was funded by the NHS South East Region Research and Development Ofce. The funding source had

Figure 1. Randomisation and follow-up of study participants. CS, caesarean section; IQR, interquartile range.

1368

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

Caesarean section surgical techniques: a randomised factorial trial

no role in the study design, the collection and interpretation of the data, writing of the report, or decision to submit the paper for publication.

Results
Between November 2000 and June 2006, 3033 women were randomised into the study in 45 UK and two Italian centres. Two centres elected to participate in two rather than all three of the pairs of alternative interventions; hence, the denominator for the pairs of interventions varies slightly.

The total population recruited was less than originally anticipated. The recruitment period lasted for 5 years and 7 months. This lower than anticipated recruitment rate was a consequence of the training and accreditation programme. The recruitment period was 32 months longer than planned and the combination of a lack of funding and trial fatigue led the Trial Steering Committee (TSC) to recommend premature closure at approximately 3000 women. Assuming that the primary outcome would be observed in 12% of women, a trial of 3000 women would allow us to detect an RR reduction of 27% with 80%

Table 1. Description of women at trial entry* Uterine closure Single layer n (%) 1483 (100) Mean age [SD] 30.6 [5.9] Single/multiple pregnancy Singleton 1404 (95) Twin 78 (5) Triplet or higher order 1 (0) Gestation at entry 37 weeks 1334 (90) 35 to <37 weeks 100 (7) <34 weeks 35 (2) Mean [SD] 39.0 [2.0] Median {IQR} 39.0 {38.340.1} Parity Primiparous 989 (67) Multiparous 480 (32) Caesarean section carried out In labour 490 (33) Not in labour 993 (67) Ruptured membranes 495 (33) Unsure 3 (0) Duration of labour <6 hours 149 (10) 612 hours 152 (10) >12 hours 163 (11) No labour 993 (67) Indication for caesarean section Breech 538 (36) Failure to progress 236 (16) AN or IP compromise 196 (13) AN complications 142 (10) Past obstetric history 100 (7) Past medical history 54 (4) Maternal request 77 (5) Abnormal presentation 7 (0) Other 118 (8) AN, Antenatal; IP, Intrapartum. *Less than 1% of data missing for each item. Double layer n (%) 1496 (100) 30.6 [5.9] 1420 (95) 75 (5) 1 (0) 1364 88 29 39.1 39.0 (91) (6) (2) [1.9] {38.340.4} Closure of pelvic peritoneum Closed n (%) 1496 (100) 30.7 [5.9] 1420 (95) 75 (5) 1 (0) 1355 92 36 39.0 39.0 (91) (6) (2) [2.0] {38.340.3} Not closed n (%) 1499 (100) 30.6 [6.0] 1420 (95) 78 (5) 1 (0) 1358 97 28 39.1 39.0 (91) (6) (2) [1.9] {38.340.4} Subrectus sheath drain Liberal n (%) 1398 (100) 30.7 [6.0] 1326 (95) 71 (5) 1 (0) 1279 83 27 39.1 39.0 (91) (6) (2) [1.9] {38.340.3} Restricted n (%) 1398 (100) 30.5 [5.8] 1325 (95) 73 (5) 0 (0) 1259 96 27 39.0 39.0 (90) (7) (2) [1.9] {38.340.3}

1027 (69) 460 (31) 511 985 514 8 166 175 160 985 554 250 193 135 105 52 79 10 108 (34) (66) (34) (1) (11) (12) (11) (66) (37) (17) (13) (9) (7) (3) (5) (1) (7)

1005 (67) 480 (32) 496 1000 510 5 153 160 166 1000 543 240 203 160 102 45 72 9 112 (33) (67) (34) (0) (10) (11) (11) (67) (36) (16) (14) (11) (7) (3) (5) (1) (7)

1026 (68) 461 (31) 506 993 502 6 162 169 156 993 561 246 188 117 103 61 85 8 115 (34) (66) (33) (0) (11) (11) (10) (66) (37) (16) (13) (8) (7) (4) (6) (1) (8)

957 (68) 435 (31) 455 943 456 4 133 139 169 943 511 235 172 135 101 52 73 10 102 (33) (68) (33) (0) (10) (10) (12) (67) (37) (17) (12) (10) (7) (4) (5) (1) (7)

939 (67) 446 (32) 452 946 464 7 151 157 129 939 521 200 185 128 96 49 79 6 119 (32) (68) (33) (1) (11) (11) (9) (67) (37) (14) (13) (9) (7) (4) (6) (0) (9)

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

1369

CAESAR study collaborative group

power (rather than an RR reduction of 25% as originally planned). Following the rst meeting of DMEC when data from 600 women were reviewed, DMEC informed TSC that, instead of a primary event rate of around 12%, the event rate was closer to 1%. TSC recommended changing the primary outcome from the original composite of febrile morbidity/ endometritis to include all infectious morbidity into a combined primary outcome, i.e. the original primary outcome of febrile morbidity/endometritis plus wound infection. This would produce a revised event rate of approximately 12%. This amendment to the trial protocol was agreed by the funding source, the sponsor and the research ethics committee. After randomisation, but prior to caesarean section, two women withdrew consent and 30 women had a vaginal delivery. These women were evenly distributed amongst the trial arms, and these 32 women (1%) were excluded from the analysis. Trial entry and discharge data were available on over 99% of women and 6-week data on up to 90% of women (Figure 1).8 The characteristics of the women at trial entry are broadly similar (Table 1). The mean age of the women was 31 years, two-thirds were primiparous, about one-third underwent an in labour caesarean section and about 95% had a singleton pregnancy. Compliance with the single- versus double-layer uterine closure and closure versus nonclosure of the pelvic peritoneum was high (at least 93%). Compliance was poorer for the liberal use of a subrectus sheath drain, for which a drain was used in 63% of women allocated to liberal use (Table 2). The main reasons for noncompliance with the

liberal use of a subrectus sheath drain were clinical, mainly because there was good haemostasis (n = 374, 75% of those noncompliant). The nonallocated clinical aspects of the caesarean section were compared between the arms of the study in order to assess whether the allocated interventions had an effect on other aspects of the surgical procedures used. There was a difference in relation to closure of the parietal peritoneum within the pelvic peritoneum comparison. For women allocated to closure of the pelvic peritoneum, more operators also closed the parietal peritoneum (23%) than if they were allocated to nonclosure of the pelvic peritoneum (7%) (Table 3). The risk of maternal infectious morbidity varied only slightly (between 16% and 18%) according to the trial arm. For each pair of interventions, there were no differences between the arms of the trial for the primary outcome or any of the secondary outcomes (Table 4). There was evidence of a strong positive (multiplicative) interaction (P = 0.006) between closure of the peritoneum and use of a subrectus sheath drain on the primary outcome, such that, in women allocated liberal use of a subsheath drain, there was a higher risk of the primary outcome associated with nonclosure of the pelvic peritoneum (20.8% versus 14.8%), whereas, in women allocated to restricted use of a subsheath drain, there was a lower risk of the primary outcome associated with nonclosure of the pelvic peritoneum (15.6% versus 17.7%). This interaction was also apparent for the outcome of wound infection treated with antibiotics (P = 0.01) and duration of the caesarean section (P = 0.005) (see Tables S1 and S2). When the effect of varying levels of compliance with the allocation of subsheath drain between centres was explored,

Table 2. Compliance with allocated treatment* Procedure allocated Uterine closure Single layer n (%) 1483 (100) Option received Single layer Double layer Option received Closed Not closed Option received Drain used Drain not used Double layer n (%) 1496 (100) Closure of pelvic peritoneum Closed n (%) 1496 (100) Not closed n (%) 1499 (100) Subrectus sheath drain Liberal n (%) 1398 (100) Restricted n (%) 1398 (100)

1377 (93) 96 (6)

11 (1) 1477 (99) 1398 (93) 88 (6) 23 (2) 1459 (97) 883 (63) 502 (36) 82 (6) 1302 (93)

*Less than 1% of data missing for each item.

1370

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

Caesarean section surgical techniques: a randomised factorial trial

Table 3. Clinical details of caesarean section* Uterine closure Closure of pelvic peritoneum Closed n (%) 1496 (100) Not closed n (%) 1499 (100) Subrectus sheath drain

Single layer n (%) 1483 (100) Abdominal entry Pfannenstiel JoelCohen Abdominal packs Used Not used Uterine entry Blunt dissection Sharp dissection Removal of placenta CCT Manual Uterine repair Single Double Uterine suture material Vicryl Dexon Other First-layer uterine suturing technique Continuous nonlocking Continuous locking Interrupted Second-layer uterine suturing technique Continuous nonlocking Continuous locking Interrupted Pelvic peritoneum Closed Not closed Pelvic peritoneum suture material Vicryl Dexon Other Pelvic peritoneum suture technique Continuous nonlocking Continuous locking Interrupted Parietal peritoneum Closed Not closed Parietal peritoneum suture material Vicryl Plain catgut Dexon Other Parietal peritoneum suturing technique Continuous nonlocking Continuous locking Interrupted

Double layer n (%) 1496 100)

Liberal n (%) 1398 (100)

Restricted n (%) 1398 (100)

1103 (74) 355 (24) 118 (8) 1347 (91) 1265 (85) 191 (13) 1227 (83) 242 (16) 1377 (93) 96 (6) 1369 (92) 24 (2) 77 (5) 446 1002 2 n = 96 80 6 6 699 767 n = 699 649 10 38 n = 699 657 20 5 211 1252 n = 211 175 0 1 31 n = 211 188 11 2 (30) (68) (0) (83) (6) (6) (47) (52) (93) (1) (5) (94) (3) (1) (14) (84) (83) (0) (0) (15) (89) (5) (1)

1147 (77) 321 (21) 111 (7) 1370 (92) 1304 (87) 168 (11) 1250 (84) 236 (16) 1477 (99) 11 (1) 1373 (92) 21 (1) 92 (6) 464 1007 2 n = 1477 1389 69 5 712 772 n = 712 657 12 39 n = 712 674 23 5 228 1249 n = 228 194 1 1 30 n = 228 209 11 4 (31) (67) (0) (94) (5) (0) (48) (52) (92) (2) (5) (95) (3) (1) (15) (83) (85) (0) (0) (13) (92) (5) (2)

1148 (77) 320 (21) 109 (7) 1375 (92) 1291 (86) 183 (12) 1238 (83) 248 (17) 700 (47) 788 (53) 1375 (92) 25 (2) 86 (6) 468 997 3 n = 788 738 36 7 1398 88 n = 1398 1296 22 75 n = 1398 1323 41 10 343 1134 n = 343 299 0 2 37 n = 343 313 14 4 (31) (67) (0) (94) (5) (1) (93) (6) (93) (2) (5) (95) (3) (1) (23) (76) (87) (0) (1) (11) (91) (4) (1)

1111 (74) 357 (24) 121 (8) 1357 (91) 1289 (86) 182 (12) 1251 (83) 234 (16) 686 (46) 803 (54) 1383 (92) 20 (1) 83 (6) 442 1028 1 n = 803 746 42 4 23 1459 n = 23 20 0 2 n = 23 18 2 0 99 1381 n = 99 73 1 0 24 n = 99 87 8 2 (29) (69) (0) (93) (5) (0) (2) (97) (87) (0) (9) (78) (9) (0) (7) (92) (74) (1) (0) (24) (88) (8) (2)

1053 (75) 317 (23) 99 (7) 1289 (92) 1208 (86) 172 (12) 1174 (84) 219 (16) 649 (46) 744 (53) 1284 (92) 18 (1) 86 (6) 443 930 3 n = 744 695 35 6 668 720 n = 668 613 6 45 n = 668 626 23 5 201 1181 n = 201 165 0 0 32 n = 201 183 12 1 (32) (67) (0) (93) (5) (1) (48) (52) (92) (1) (7) (94) (3) (1) (14) (84) (82) (0) (0) (16) (91) (6) (0)

1062 (76) 309 (22) 101 (7) 1277 (91) 1183 (85) 185 (13) 1153 (82) 229 (16) 645 (46) 743 (53) 1281 (92) 27 (2) 80 (6) 435 931 1 n = 743 699 30 4 654 728 n = 654 604 16 32 n = 654 621 15 5 208 1170 n = 208 176 1 2 29 n = 208 190 9 3 (31) (67) (0) (94) (4) (1) (47) (52) (92) (2) (5) (95) (2) (1) (15) (84) (85) (0) (1) (14) (91) (4) (1)

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

1371

CAESAR study collaborative group

Table 3. (Continued) Uterine closure Closure of pelvic peritoneum Closed n (%) 1496 (100) Not closed n (%) 1499 (100) Subrectus sheath drain

Single layer n (%) 1483 (100) Supercial fat layer Closed Not closed Fat layer suture material Vicryl Plain catgut Dexon Other Fat layer suturing technique Continuous nonlocking Continuous locking Interrupted Skin closure Subcuticular Interrupted Clips/staples Other

Double layer n (%) 1496 100)

Liberal n(%) 1398 (100)

Restricted n (%) 1398 (100)

507 960 n = 506 382 4 59 57 n = 506 311 12 171 1266 144 57 1

(34) (65) (75) (1) (12) (11) (61) (2) (34) (85) (10) (4) (0)

549 931 n = 549 417 2 68 57 n = 549 333 27 175 1272 152 55 1

(37) (62) (76) (0) (12) (10) (61) (5) (32) (85) (10) (4) (0)

517 965 n = 516 396 5 55 55 n = 516 336 20 149 1277 147 59 1

(35) (65) (77) (1) (11) (11) (65) (4) (29) (85) (10) (4) (0)

537 943 n = 537 401 1 72 59 n = 537 307 19 196 1279 149 52 1

(36) (63) (75) (0) (13) (11) (57) (4) (36) (85) (10) (3) (0)

457 928 n = 457 328 3 60 62 n = 457 286 16 147 1200 130 57 1

(33) (66) (72) (1) (13) (14) (63) (4) (32) (86) (9) (4) (0)

475 907 n = 474 353 3 66 47 n = 474 300 13 144 1194 130 55 1

(34) (65) (74) (1) (14) (10) (63) (3) (30) (85) (9) (4) (0)

CCT, Controlled cord traction. *Less than 2% of data missing for each item, except parietal peritoneum suturing technique (34% of data missing) and fat layer suturing technique (3% of data missing).

there was no evidence that increasing compliance increased the difference between the two groups. For the pelvic peritoneum comparison, the effect of differential closure of the parietal peritoneum on the outcomes was adjusted for in the analysis, and the MantelHaensel risk ratio for the primary outcome was 0.95 (95% CI = 0.78 1.15), which is similar to the crude risk ratio of 0.92 (95% CI = 0.761.10). Similarly, this adjusted analysis produced no differences in the estimate of effect for any of the secondary outcomes (Table 4). Twelve serious adverse events were reported, and these occurred equally in all arms. These included three women who required hysterectomy to control bleeding and three additional women who required blood transfusions of >4 units. When the study outcomes were analysed according to the eight randomised groups, there were no signicant differences between the groups for any outcomes (data not shown). A prespecied subgroup analysis was performed for the comparison of liberal versus restricted use of a subsheath drain. In this analysis, centres were divided into quartiles of compliance (065.5%, >65.5% to <83.3%, 83.3% to <94.4%, >94.4%). There was no evidence of any effect on the primary outcome or the secondary outcomes by compliance. Finally,

the effect of each intervention on the study outcomes did not vary according to whether the caesarean was in labour or not in labour, and there was no evidence of an interaction. A prespecied subgroup analysis by singleton versus multiple pregnancy was not performed owing to the small number of multiple pregnancies (155, 5%).

Discussion
CAESAR is the largest randomised trial of caesarean section surgical techniques undertaken to date, and provides important evidence to guide clinical practice. Mounting such a large surgical trial presented a number of challenges, including the need to ensure adequate competency of the operators in surgical techniques which may have been unfamiliar, particularly single-layer uterine closure. This step was important to limit concerns that the comparisons between groups may have been affected by operator experience or competence,9 but it did introduce a substantial delay in the recruitment of operators, and therefore the recruitment of women. Despite lower than anticipated recruitment, the event rate for the primary outcome was higher than that used in the estimation of the sample size, and so the power for

1372

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

Caesarean section surgical techniques: a randomised factorial trial

Table 4. Outcome measures by three comparisons* Uterine closure Closure of pelvic peritoneum Closed n (%) 1496 (100) Not closed n (%) 1499 (100) Subrectus sheath drain

Single layer n (%) 1483 (100) Primary outcome** Maternal infectious morbidity RR (95% CI) RR (95% CI) adjusted for parietal peritoneum Secondary outcomes Temperature 39C on any occasion RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Temperature 38C on 2 days RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Antibiotics given for febrile morbidity RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Known to have had endometritis RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Known to have had wound infection treated with antibiotics RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Known to have had operative procedures on wound RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Additional analgesia on day 3 RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Pain VAS at hospital discharge Median {IQR} P value for rank sum test Pain VAS at 6 weeks post-partum Median {IQR} P value for rank sum test Blood transfusion RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Severe maternal morbidity RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Readmitted to hospital within 6 weeks of caesarean section RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Breastfeeding at hospital discharge RR (99% CI) RR (99% CI) adjusted for parietal peritoneum Duration of caesarean (minutes) Mean [SD]

Double layer n (%) 1496 (100)

Liberal n (%) 1398 (100)

Restricted n (%) 1398 (100)

247 (16.9) 249 (16.9) 1.00 (0.851.17)

240 (16.3) 262 (17.7) 0.92 (0.781.08) 0.95 (0.801.12) 8 (0.5) 2 (0.1) 4.02 (0.8518.9) 3.03 (0.7812.0) 13 (0.9) 9 (0.6) 1.45 (0.623.39) 1.36 (0.543.46) 14 (1.0) 9 (0.6) 1.56 (0.683.60) 1.39 (0.583.30) 59 (4.0) 66 (4.5) 0.90 (0.641.27) 0.92 (0.651.30) 182 (12.3) 200 (13.5) 0.92 (0.761.10) 0.95 (0.781.15) 9 (0.6) 9 (0.6) 1.01 (0.402.53) 1.05 (0.452.44) 506 (36) 514 (37) 0.99 (0.871.13) 1.00 (0.911.11) 20 {936} 0.08 19 {835}

247 (17.9) 228 (16.6) 1.08 (0.921.27)

5 (0.3) 5 (0.3) 1.01 (0.205.13) 12 (0.8) 10 (0.7) 1.21 (0.403.64) 12 (0.8) 11 (0.7) 1.09 (0.383.19) 63 (4.3) 62 (4.2) 1.02 (0.651.61) 188 (12.8) 188 (12.7)

3 (0.2) 4 (0.3) 0.75 (0.105.33) 11 (0.8) 10 (0.7) 1.10 (0.363.36) 10 (0.7) 10 (0.7) 0.99 (0.323.13) 65 (4.7) 53 (3.9) 1.21 (0.761.94) 186 (13.4) 178 (12.9)

1.01 (0.791.29) 9 (0.6) 9 (0.6) 1.01 (0.303.39) 526 (38) 493 (35) 1.08 (0.951.23)

1.04 (0.811.33) 12 (0.9) 5 (0.4) 2.40 (0.619.42) 483 (37) 501 (38) 0.96 (0.851.10)

20 {835} 0.2

20 {936}

20 {937} 0.2

19 {834}

2 {010} 2 {011} 0.7 54 (3.7) 59 (4.0) 0.93 (0.571.49) 7 (0.5) 11 (0.7) 0.65 (0.192.23) 34 (2.6) 37 (2.7)

2 {011} 2 {011} 0.3 61 (4.1) 52 (3.5) 1.15 (0.801.66) 1.12 (0.771.63) 12 (0.8) 6 (0.4) 2.00 (0.755.32) 1.96 (0.765.06) 39 (2.9) 32 (2.4) 1.22 (0.771.93) 1.31 (0.812.12) 894 (64.2) 896 (64.0) 1.00 (0.951.06) 0.99 (0.931.05) 38.5 [12.0] 36.1 [11.5]

3 {011} 2 {010} 0.1 53 (3.8) 52 (3.7) 1.02 (0.621.66) 13 (0.9) 3 (0.2) 4.32 (0.8322.4) 33 (2.6) 34 (2.7)

0.93 (0.511.71) 857 (62.5) 920 (65.4) 1.05 (0.891.03)

0.98 (0.521.82) 816 (62.8) 836 (64.2) 0.98 (0.911.06)

36.2 [11.6]

38.3 [11.8]

37.7 [11.4]

37.3 [12.5]

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

1373

CAESAR study collaborative group

Table 4. (Continued) Uterine closure Closure of pelvic peritoneum Closed n (%) 1496 (100) Not closed n (%) 1499 (100) Subrectus sheath drain

Single layer n (%) 1483 (100) Difference in means (99% CI) Difference in means (99% CI) adjusted for parietal peritoneum Duration of postnatal hospital stay (days) Mean [SD] Median {IQR} Difference in means (99% CI) Difference in means (99% CI) adjusted for parietal peritoneum 2.1 (1.0, 3.2)

Double layer n (%) 1496 (100)

Liberal n (%) 1398 (100) 0.3 (0.8, 1.5)

Restricted n (%) 1398 (100)

)2.4 ()3.6, )1.3) )2.3 ()3.4, )1.2)

4.7 [2.0] 4.0 {45} 0.0 (0.2, 0.2)

4.7 [2.0] 4.0 {45}

4.8 [1.9] 4.7 [2.0] 4.0 {45} 4.0 {45} )0.1 ()0.3, 0.1) )0.1 ()0.3, 0.1)

4.8 [1.9] 4.0 {45} 0.1 (0.1, 0.3)

4.7 [1.8] 4.0 {45}

CI, condence interval; IQR, interquartile range; RR, relative risk; VAS, visual analogue scale. *Data missing for <2% of items, except for whether the woman was readmitted to hospital within 6 weeks of caesarean section (missing for 10% of women), additional analgesia on day 3 after caesarean section (missing for 7% of women) and breastfeeding at hospital discharge (missing for 7% of women). **Primary outcome comprises antibiotics given for pyrexia (temperature of >39C on any occasion or >38C on two or more successive days) prior to discharge from hospital, or wound infection treated with antibiotics within 6 weeks or endometritis treated with antibiotics within 6 weeks of caesarean section.

detecting the main effects of the arms of the trial should not have been adversely affected. For most of the comparisons, there was no statistically signicant interaction between the trial arms, although it should be noted that the trial had relatively low power to detect interactions. Although there were no differences detected between the randomised interventions, the results of CAESAR do have implications for clinical practice, particularly in relation to current guidance about closure of the peritoneum at the time of caesarean section repair. Guidance in the UK10 recommends that both layers of the peritoneum should not be closed, as this is associated with a lower risk of morbidity based on existing evidence.11 The data from CAESAR demonstrate that there are no differences in any of the shortterm morbidity outcomes associated with closure versus nonclosure of the pelvic peritoneum at caesarean section, although there is a difference in the duration of surgery (mean difference, 2.4 minutes; 95% CI, 1.33.6 minutes), favouring nonclosure. However, the duration of surgery is a poor surrogate for morbidity, as there will be a lower threshold below which duration is associated with greater morbidity. As there were no apparent differences in any of the short-term outcomes, the potential effects of these surgical techniques on longer term outcomes, such as the functional integrity of the uterine scar during subsequent pregnancies and effects such as chronic pelvic pain, infertility, peritoneal adhesions and obstruction, become of increasing

importance for guiding clinical practice. Long-term outcomes, even if uncommon, are likely to have a substantial impact on morbidity, and can occasionally result in lifethreatening events or even death. There have been three follow-up studies of existing trials of caesarean section techniques.1214 The sample sizes of these studies are small, and included only 18% of the women randomised into the original trials for two studies,12,13 and 51% for the third.14 In the only study comparing single- versus double-layer uterine closure, no differences were found in interpregnancy interval, vaginal delivery, length of hospital stay, preterm delivery, amnionitis, postpartum endometritis, placental abruption, postpartum haemorrhage, blood transfusion or uterine dehiscence. This was based on a sample size of 145 women of the 906 randomised in the original trial.13 From the two follow-up studies of trials comparing closure versus nonclosure of the peritoneum (209 women in total), no differences were observed in terms of intra-abdominal adhesions, abdominal pain, dyspareunia, constipation, urinary symptoms and infertility.14 Given the relatively low incidence of uterine rupture (0.40.6% for women undergoing a trial of labour versus 0.2% for those having elective repeat caesarean) and dehiscence (1.1%),1517 attempts to detect differences in these outcomes between different surgical techniques with adequate power is challenging. One large observational study has looked at this outcome and found an association with previous single-layer uterine closure.18

1374

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

Caesarean section surgical techniques: a randomised factorial trial

This lack of adequate long-term follow-up and the absence of short-term effects of these techniques raise important issues about current clinical guidance. The lack of differences between the interventions in the short term may suggest that any one of the three pairs of approaches is acceptable. However, there have been suggestions that nonclosure of the peritoneum may be harmful in the longer term. It is possible, therefore, that clinical guidance may dictate19 a course of action, such as continued recommendation for nonclosure of the peritoneum, which may be harmful in the longer term, and this information will not be available until follow-up of trials, such as CAESAR, are completed. In the meantime, there are difcult decisions to be made if clinical guidance is to be provided to clinicians about the surgical techniques used at caesarean section.

Bowler, Peter Brocklehurst, Ruth Davis, Madalena Gallagher, Lucy Tully, Simon Gates, Juan Manuel Nardin, Maria Quigley, Sally Tyndel. CAESAR Trial Steering Committee: Jo Anthony (Northampton NHS Trust), Felicity Ashworth (Chair) (Buckinghamshire NHS Trust), Ursula Bowler (National Perinatal Epidemiology Unit), Peter Brocklehurst (National Perinatal Epidemiology Unit), Ali Chevassut (Oxford Radcliffe Hospital Trust), Debbie Chippington Derrick (National Childbirth Trust), Simon Gates (until April 2005) (National Perinatal Epidemiology Unit), Pauline Hurley (Oxford Radcliffe Hospital Trust), Maria Quigley (from April 2006) (National Perinatal Epidemiology Unit). CAESAR Data Monitoring and Ethics Committee: Zarko Alrevic, Susan Bewley, Janet Darbyshire (Chair), Jon Deeks, Ed Juszczak. CAESAR Trial Participating Hospitals: See Appendix S1 for a full list of participating hospitals.

Conclusions
These results have implications for clinical practice, particularly in relation to current guidance about closure of the peritoneum, which suggests that nonclosure is preferable. The potential effects of these different surgical techniques on longer term outcomes, including the functional integrity of the uterine scar during subsequent pregnancies, are becoming increasingly important for guiding clinical practice.

Supporting information
The following supplementary materials are available for this article: Table S1. P values for test for pairwise interactions between trial arms. Table S2. Stratum-specic results where there is evidence of interaction between trial arms. Appendix S1. CAESAR trial participating hospitals. Additional Supporting Information may be found in the online version of this article. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author. j

Disclosure of interests
All authors declare that they have no relevant interests to declare.

Contribution to authorship
PB designed the study and wrote the proposal with contributions from SA. MQ and SA analysed the data, with input from EJ and PB. PB wrote the nal report, with all of the writing committee contributing to the editing.

References
1 Villar J, Valladares E, Wojdyla D, Zavaleta N, Carroli G, Vela A, et al. Caesarean delivery rates and pregnancy outcomes: the 2005 WHO global survey on maternal and perinatal health in Latin America. Lancet 2006;367:181929. 2 Berghella V, Baxter JK, Chauhan SP. Evidence-based surgery for cesarean delivery. Am J Obstet Gynecol 2005;193:160717. 3 Tully L, Gates S, Brocklehurst P, McKenzie-McHarg K, Ayers S. Surgical techniques used during caesarean section operations: results of a national survey of practice in the UK. Eur J Obstet Gynecol Reprod Biol 2002;102:1206. 4 Smaill F. Prophylactic antibiotics in Caesarean section (all trials). In: Enkin MW, Keirse MJNC, Renfrew MJ, Neilson JP, editors. Pregnancy and Childbirth Module. Cochrane Database of Systematic Reviews, Review No. 03690. Oxford: Update Software; 1994. Disk Issue 1. 5 McAlister FA, Straus SE, Sackett DL, Altman DG. Analysis and reporting of factorial trials: a systematic review. J Am Med Assoc 2003;289:254553. 6 Montgomery AA, Peters TJ, Little P. Design, analysis and presentation of factorial randomised controlled trials. BMC Med Res Methodol 2003;3:26.

Details of ethics approval

Multicentre Research Ethics Committee approval for the trial was given by the Eastern MREC (99/5/62), and all participants gave informed consent to participate.

Funding
The trial was funded by the NHS South East Region Research and Development Ofce. The funding source had no role in the study design, the collection and interpretation of the data, writing of the report or decision to submit the paper for publication.

Acknowledgements
Writing Committee: Peter Brocklehurst, Maria Quigley, Sarah Ayers, Ed Juszczak, National Perinatal Epidemiology Unit. CAESAR Co-ordinating Centre at the National Perinatal Epidemiology Unit: Elizabeth Anderson, Sarah Ayers, Ursula

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

1375

CAESAR study collaborative group

7 Schulz KF, Grimes DA. Multiplicity in randomised trials II: subgroup and interim analyses. Lancet 2005;365:165761. 8 Bourton I, Moher D, Altman D, Schulz K, Ravaud P.; for the CONSORT group. Extending the CONSORT statement to randomized trials of nonpharmacologic treatment: explanation and elaboration. Ann Intern Med 2008;148:295309. 9 Devereaux PJ, Bhandari M, Clarke M, Montori VM, Cook DJ, Yusuf S, et al. Need for expertise based randomised controlled trials. Br Med J 2005;330:88. 10 National Institute for Health and Clinical Excellence (NICE). Caesarean Section Guideline CG13. London: NICE; 2004. 11 Bamigboye AA, Hofmeyr GJ. Closure versus non-closure of the peritoneum at caesarean section. Cochrane Database Syst Rev 2003;4: CD000163. 12 Weerawetwat W, Buranawanich S, Kanawong M. Closure vs nonclosure of the visceral and parietal peritoneum at cesarean delivery: 16 year study. J Med Assoc Thai 2004;87:100711. 13 Chapman SJ, Owen J, Hauth JC. One versus two layer closure of a low transverse cesarean: the next pregnancy. Obstet Gynecol 1997; 89:168.

14 Roset E, Boulvain M, Irion O. Nonclosure of the peritoneum during caesarean section: long-term follow-up of a randomised controlled trial. Eur J Obstet Gynaecol Reprod Biol 2002;108:404. 15 Mozurkewich EL, Hutton EK. Elective repeat cesarean delivery versus trial of labor: a meta-analysis of the literature from 1989 to 1999. Am J Obstet Gynecol 2000;183:118797. 16 Chauhan SP, Martin JN Jr, Henrichs CE, Morrison JC, Magann EF. Maternal and perinatal complications with uterine rupture in 142,075 patients who attempted vaginal birth after cesarean delivery: a review of the literature. Am J Obstet Gynecol 2003;189: 40817. 17 Lydon-Rochelle M, Holt VL, Easterling TR, Martin DP. Risk of uterine rupture during labor among women with a prior cesarean delivery. N Engl J Med 2001;345:38. 18 Bujold E, Bujold C, Hamilton EF, Harel F, Gauthier RJ. The impact of a single-layer or double-layer closure on uterine rupture. Am J Obstet Gynecol 2002;186:132630. 19 Lyell D, Caughey A, Hu E, Daniels K. Peritoneal closure at primary caesarean delivery and adhesions. Obstet Gynecol 2005;106: 27580.

Journal Club Discussion points

1. Background: Discuss the existing recommendations with regard to the surgical techniques for caesarean sections, for example by the National Institute for Clinical Excellence (NICE).1 Debate with your colleagues aspects of the technique that you prefer, and discuss why. 2. Methods: Compare the methods in this study with those used by a previous study on a similar subject.2 Comment on the advantages and disadvantages of the two different approaches. Debate the choice, denition and ascertainment of the outcome measures used in this study. Discuss the strengths and weaknesses of factorial (e.g. 2 2) designs for randomised controlled trials. Discuss the practical, ethical, clinical governance and scientic issues around premature trial discontinuation. Illustrate with examples from this study as well as others of which you might be aware. 3. Results and implications: Discuss the event rates for the individual outcomes in this study. Are you surprised by the low rate (1%) of infection at the interim analysis/rst meeting of the steering committee? Explain how the results of this study, combined with other recent ones, may affect your current practice. Which outcome measures would you want follow-up studies to address? j

D Siassakos University of Bristol & Southmead Hospital, Bristol, UK Email jsiasakos@gmail.com

References
1 National Collaborating Centre for Womens and Childrens Health. Caesarean Section. London: National Institute for Clinical Excellence (NICE), 2004. 2 Blumenfeld Y, Caughey A, El-Sayed Y, Daniels K, Lyell D. Single- versus double-layer hysterotomy closure at primary caesarean delivery and bladder adhesions. BJOG 2010;117:6904.

1376

2010 The Authors Journal compilation RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology