Journal de Gyn´ ecologie Obst´ etrique et Biologie de la Reproduction (2015) 44, 393—397
Disponible en ligne sur
ScienceDirect www.sciencedirect.com
LATEST DEVELOPMENTS
To what extent do English language RCT meta-analysis justify induction of low-risk pregnancy for postdates? Dans quelle mesure les méta-analyses d’essais randomisés contrôlés de la littérature anglo-saxonne peuvent-elles justifier l’induction du travail pour grossesse prolongée à bas risque ? J.S. Cohain Alon Shvut 37, 90433, Israel Received 31 July 2014 ; received in revised form 11 November 2014; accepted 24 December 2014 Available online 24 February 2015
KEYWORDS Induction for postdates; Antepartum stillbirth; Perinatal mortality; Perinatal morbidity; Maternal mortality; Maternal morbidity
MOTS CLÉS Induction du travail ; Mort-né ante-partum ;
Summary Induction for postdates in low-risk pregnancy was adopted with the intent to prevent post-term antepartum stillbirth, the most common cause of perinatal death, based on evidence derived in English language RCT meta-analysis. Systematic English language meta-analysis of RCT studies of induction for postdates in low-risk pregnancy report perinatal mortality rates (PMRs) for low-risk pregnancy ranging from 2.6 to 7.6/1000, based on 2—5 stillbirths among 13—16 perinatal deaths, including diabetic pregnancies as well as other high-risk pregnancies irrelevant to the study question. Baseline PMR ≥ 41 weeks in large international databases for high and low risk pregnancies before routine induction 1998—2003 range from 0.9 to 2.4/1000 or about 300% lower than the reported PMR rates for postdate pregnancies in the expectant management arm in English language RCT meta-analysis. Deaths in the first week far exceed stillbirths in the RCT meta-analysis, the opposite of what is expected. These 2 implausible results bring into question the evidence used to justify induction for postdates ≥ 41 weeks. © 2015 Elsevier Masson SAS. All rights reserved.
Résumé L’induction médicale du travail pour grossesse prolongée à bas risque est une pratique courante dont l’objectif est de prévenir la mortalité anténatale à terme dépassé, la cause la plus fréquente de mortalité périnatale. Cette pratique est acceptée comme valide au vu des résultats de méta-analyses de publications de la littérature anglo-saxonne. Selon ces méta-analyses
E-mail address:
[email protected] http://dx.doi.org/10.1016/j.jgyn.2014.12.020 0368-2315/© 2015 Elsevier Masson SAS. All rights reserved.
394
Mortalité périnatale ; Morbidité périnatale ; Mortalité maternelle ; Morbidité maternelle
J.S. Cohain sur l’induction du travail pour grossesse prolongée à bas risque, le taux de mortalité varie entre 2,6 et 7,6 pour 1000 naissances, basé sur 2 à 5 mort-nés parmi 13 à 16 décès périnataux, y compris les grossesses diabétiques et autres grossesses à haut risque, sans rapport avec l’objet de l’étude. Si l’on regarde les grandes bases de données internationales de 1998 à 2003, la mortalité périnatale à ≥ 41 semaines est de 0,9 à 2,4 pour 1000 naissances, c’est-à-dire 300 % plus basse que la mortalité périnatale des grossesses prolongées dans les bras « wait-and-see » des études anglo-saxonnes prises en compte dans les méta-analyses. Dans ces méta-analyses, les décès de la première semaine dépassent de beaucoup les mort-nés, le contraire de ce à quoi l’on pourrait s’attendre. Ces deux résultats peu plausibles mettent en question la validité du choix d’une induction du travail pour les grossesses à ≥ 41 semaines. © 2015 Elsevier Masson SAS. Tous droits réservés.
Introduction Previous to 1999, induction for low-risk pregnancy was carried out where the foetus showed signs of distress [1]. The potentially life-threatening risks which can result from artificially inducing labor, such as increased chorioamnionitis, cord prolapse, postpartum haemorrhage, tachysystole, failed induction, caesarean delivery, and accidental delivery of preterm foetus [2] were only taken when they outweighed the risks of continuing the pregnancy. The protocol to induce for postdates was initially fuelled by the 1999 observational study [3] tenuously associating two variables that might have no relationship: a decrease of 1.9/1000 in antepartum stillbirth from 2.8/1000 to 0.9/1000 in Canada reported for all births between the years 1980 and 1995, with an increased induction rate reported at 2 Canadian hospitals and at the health departments of 6 provinces for births more than or equal to 41 weeks at various intervals between 1980 and 1995. This observational study did not control for any of the variables known to decrease stillbirth in all risk women, such as increased quality and availability of prenatal care, birth control and induced abortions; better controlled diabetes; decreased smoking, violence, and car accidents; increased intake of micronutrients and use of seat belts, to name a few. Another impetus fuelling the acceptance of the protocol to induce for postdates was the publication of stillbirth rates after 41 weeks in large databases 1998—2003. These articles showed that about 1/1000 antepartum stillbirths occur after 40 weeks and about 2/1000 stillbirths after 41 weeks [4—6]. After the excess post-term stillbirths were documented, it was optimistically hoped that the excess 1/1000 stillbirths after 41 weeks among low-risk pregnancies could be lessened by induction of labor. It was not only hoped that induction could prevent some or all of the excess 1/1000 stillbirths after 41 weeks, but also hoped or assumed that it would do so without causing excess long-term harm to mother and child. It was also assumed that the extra 1/1000 born alive, saved by induction, would not die in the first month. Systematic meta-analyses were set out to test the first of these optimistic hopes. The PMR rates in the meta-analyses for expectant management arm for low-risk pregnancy (not inducing for postdates) would be expected to be similar to expectant management rates of low-risk pregnancy, and lower than PMR rates of all risk pregnancies reported in large databases. PMR ≥ 41 weeks was defined as stillbirths after 41 weeks
plus perinatal deaths in the first week after live birth/1000 births. The aim of this study is to compare the PMR reported in the expectant management arms of the English language RCT meta-analyses 2009—2012 to baseline PMR rates before induction for postdates routines were established.
Results The 3 English language meta-analyses of induction for postdates published after 2008 are Gulmezoglu et al. [7], Hussain et al. [8] and Wennerholm et al. [9]. The 3 systematic reviews combine the results of more or less the same 3000—3700 pregnancies delegated to expectant management in about 17 RCT studies, and report on the same 13—16 babies who reportedly either died in the uterus late in the pregnancy (n = 2—5) or died in the first week of life (n = 8—14). The reported PMR rates of the expectant management arm in the English language studies were: • Gulmezoglu: 3.5/1000, (13/3700), (C.I 0.09—0.99); • Hussain: 4.9/1000, (16/3282), (C.I 0.11—0.88); • Wennerholm: 2.6/1000, (8/3097), (C.I 0.10—1.09). About half of all births analysed in all 3 RCT metaanalyses derived from one study [10], which had the lowest rate of adherence to protocols, reporting that 50% of those randomised to expectant management were induced. Since Hannah’s 1992 observational study [3] is biased in favor of induction and the Hannah study had a 50% rate of noncompliance to protocol, it was logical to extract the PMR rates without the Hannah study. When Hannah study [10] is excluded from the analysis, the perinatal mortality rates for the RCT meta-analysis expectant management arm are: • Gulmezoglu: 4.3/1000; • Hussain: 5.7/1000; • Wennerhold: 7.6/1000. These rates are compared to baseline rates reported in the literature in Table 1 and Fig. 1. Gulmezoglu et al. is the most careful about itemizing each instance of perinatal mortality. Four perinatal deaths occurred ≥ 41 weeks in the expectant management arm (2 stillbirths and 2 deaths in first week) or 4/3700 or PMR = 1/1000, which is a credible PMR for low-risk pregnancy, and equivalent to the PMR for inductions, which
Induction for postdates
395
Table 1 PMR per 1000 extracted from studies. Mortalité périnatale pour 1000 naissances ; données des études anglo-saxonnes. MacDorman et al., 2009 (US) [11] Based on 2005 US Birth Statistics Smith et al., 2003(Scotland) [5] 1980—1998 Heimstad et al. (2008) [12] Data for Norway 1999—2005
PMR including congenital anomalies ≥ 41 weeks = 0.91/1000
0.91
PMR including congenital anomalies 42 + weeks = 1.66/1000
1.66
Unexplained stillbirths ≥ 41 weeks on second births, no previous CS
0.3
Unexplained stillbirths ≥ 41 weeks on second births, after previous CS
1
Stillbirths excluding foetal malformations ≥ 41 weeks 147/98,559 = 1.5/1000
1.9
Perinatal death excluding foetal malformations in first week ≥ 41 weeks = 0.4/1000 1.5 + 0.4 = PMR 1.9 Hilder et al., 1998 (UK) [4] Years 1989—1991
Total Stillbirths ≥ 41 weeks incl. all births and anomalies = 1.7/1000
2.4
Neonatal deaths in first week ≥ 41 weeks all births and anomalies = 0.7 1.7 + 0.7 . . . PMR ≥ 41 weeks = 2.4 Olesen et al., 2003 (Denmark) [6] during the years 1978—1993 Years 1978—1993
No data for ≥ 41 weeks:
Stillbirths no exclusions 37—41 + 6 weeks Perinatal deaths no exclusions 37—41 + 6 Total PMR 37—41 + 6 weeks = 2.7 Stillbirths no exclusions > 42 weeks Perinatal deaths no exclusions > 42 weeks Total PMR after 42 weeks = 3.66 Deaths first month 37—41 + 6 weeks Deaths first month > 42 weeks Total PMR incl. first month 37—41 + 6 weeks = 4 Total PMR incl. first month 42 weeks = 5.5 Gulmezoglu et al., 2012 [7] Cochrane 22 RCT trials published 1969—2007 ‘‘low risk’’
Hussain et al., 2011 [8] 14 studies Published 1969—2007 ‘‘Low risk’’
Wennerholm et al., 2009 [9] ‘‘Low risk’’ 1987—2007
1.8 0.9 2.2 1.46 1.3 1.85
Stillbirths + first week deaths from 37 weeks, no exclusions 13/3700
3.5
Stillbirths excl. cong anomalies > 41 weeks: 2/3700: @ 41 + 5, @ 42 weeks Perinatal deaths excl. con. anom. > 41 weeks: 2/3700: @ 42 + 0, @ 43 + 4 without Hannah
0.5 0.5
Stillbirths (37 + weeks) 5/3282
1.5
Perinatal deaths (37 + weeks) first week 11/3282 PMR stillbirths plus perinatal deaths
3.4 4.9
Stillbirths-13 studies (excl Hannah study) 3/1576 Perinatal deaths 13 studies (excl Hannah study) 9/1576 PMR (excl. Hannah study) stillbirths plus perinatal deaths first week
1.9 5.7 7.6
Stillbirths excl. cong anomalies 3/3099
0.97
Perinatal deaths excl. cong anomalies 3/3097 Total PMR 6/3097 Total PMR 8/3097 Stillbirths excl. Hannah study 1/1395 Perinatal deaths excl. Hannah study 3/1397 Total PMR excl. Hannah with anomalies 6/1397
0.97 1.9 2.6 0.7 2.15 4.3
396
J.S. Cohain
Figure 1
Perinatal mortality per 1000 births ≥ 41 weeks.
Mortalité périnatale pour 1000 naissances à ≥ 41 semaines.
means induction and expectant management had the same outcomes. However, instead of this rate, the rate the authors report for expectant management includes congenital abnormalities and high risk. Wennerholm et al. attempted to improve on Gulmezoglu and Hussain meta-analyses. Wennerholm et al. discarded the poor quality studies, high-risk births, births before 41 weeks and enumerated births with anomalies, but curiously instead of reporting the PMR rate of expectant management without high risk and congenital malformations, report PMR 2.6, including the deaths due to congenital malformations. Excluding the births with congenital malformations, the PMR is a plausible 1.9 among low-risk women. Without the Hannah study [10], the PMR rises to 4.3 with congenital malformations or 2.85 without. The 3 RCT studies appear to assume a priori that induction has the ability to prevent stillbirth. The texts typically include statements, such as: ‘‘we aimed to estimate numbers needed to treat to avoid 1 foetal or neonatal death’’ [12].
Discussion It was concluded by the 3 recent English language RCT meta-analyses [7—9] that induction for postdates in straightforward pregnancies results in a significant decrease in PMR [7,8] or a non-statistically significant decrease in PMR [9] (8 deaths in the expectant management arm vs. 1 death in the induction arm). Sentilhes et al. [13], a 2011 French language metaanalysis of the same RCT studies, conclude that there is insufficient data to support induction of labor at 41(+0) weeks. Surprisingly, stillbirth rates were not the end point in any of these studies. The 3 meta-analyses report PMR rates for low-risk pregnancy after 41 weeks that are inexplicably higher than published baselines for all risk (high and low risk) births. A higher rate of deaths in the first week compared to stillbirths plagues all the studies. Stillbirth is much more common than perinatal death in the first week overall, with > 90% of perinatal deaths at 41 weeks deriving from stillbirth and
less than 10% deriving from newborn deaths in the first week [11], yet, all meta-analyses found a much higher rate of death (n = 8—16) in the first week in the expectant management arm than stillbirths (n = 2—5). It is not explained why the deaths in the first week were included in the analysis. There is not reason to report deaths of live births in the expectant management arm. Deaths in the induction arm might be reported to assure the reader that a decrease in stillbirths in the induction arm was not followed by an increased death rate in the first week among the live births. Foetuses with congenital abnormalities incompatible with life do not benefit from induction, and should not have been included in the final analyses. As stated in all the RCT’s conclusions, no trial in any of the meta-analyses was of high validity or precise methodology. Three trials were of fair quality. Fourteen trials were of poor quality. For examples, in Heimstead et al. [12], out of 254 randomised to induction, 39/254 (15%) were not induced. In Hannah et al. (1992) [10], 50% of those randomised to expectant management were induced. While much lip service is attributed to the desire to decrease stillbirths, it bears mention that efforts to reduce stillbirths are inconsistent. Starting in 2003, evidence has accumulated showing third trimester antepartum stillbirths increase in second pregnancies after one caesarean [5]. These third trimester stillbirths could be prevented by lowering caesarean rates, particularly primary caesarean rates, however, few efforts have been made to decrease stillbirth by lowering primary or general CS rates. In settings where induction is associated with increased cesarean rates, induction for postdates increases, not decreases, the stillbirth rate among women who go on to have subsequent pregnancies. Another curious contradiction is the willingness to enact a protocol, such as induction for postdates without any evidence of its long-term safety to mother and child.
Limitations It could be argued that 2 perinatal deaths (not stillbirth) derived from the countries: Turkey [14] and Tunisia [15] where baseline PMR may have been higher than the
Induction for postdates 1—2/1000 baselines. Pregnancies under the egis of published RCT studies are likely to receive adequate medical attention no matter where they take place. Low-risk pregnancy by definition, regardless of country, has very low rates of complications. The overall PMR for low-risk pregnancy hovers around 0.6/1000—1/1000 as compared to the overall PMR, for example, of Type 1 and 2 uncontrolled diabetics which is 35/1000. [16] These 3 meta-analyses were largely to determine if induction could prevent unexplained stillbirth, perhaps 1/1000, in low-risk pregnancy ≥ 41 weeks. A theory is lacking for how induction would save low-risk babies dying in the first week after birth.
Interpretation RCT studies do not support induction for postdates among straightforward pregnancies. Future studies to test whether 1/1000 antepartum stillbirths can be prevented by induction for postdates in straightforward postdate pregnancy need to be designed with:
397
[4]
[5]
[6]
[7]
[8]
[9]
[10]
• • • • • •
adherence to study protocol; inclusion only of postdate pregnancy; exclude high-risk pregnancies; exclude congenital malformations incompatible with life; compare outcome to known baselines; include long-term follow-up on the implications on the mother and newborn of induction.
[11] [12]
[13]
Disclosure of interests The author declares that she has no conflicts of interest concerning this article.
[14]
References [1] Sulik SM, Greenwald JL. Evaluation and management of postdate pregnancy. Am Fam Physician 1994;49:1177—86 [1191—2]. [2] Chauhan SP, Ananth CV. Induction of labor in the United States: a critical appraisal of appropriateness and reducibility. Semin Perinatol 2012;36:336—43. [3] Sue-A-Quan AK, Hannah ME, Cohen MM, Foster GA, Liston RM. Effect of labor induction on rates of stillbirth and
[15]
[16]
cesarean section in post-term pregnancies. CMAJ 1999;160: 1145—9. Hilder L, Costeloe K, Thilaganathan B. Prolonged pregnancy: evaluating gestation-specific risks of fetal and infant mortality. Br J Obstet Gynaecol 1998;105:169—73. Smith GC, Pell JP, Dobbie R. Caesarean section and risk of unexplained stillbirth in subsequent pregnancy. Lancet 2003;362:1779—84. Olesen AW, Westergaard JG, Olsen J. Perinatal and maternal complications related to post-term delivery: a national register-based study, 1978—1993. Am J Obstet Gynecol 2003;189:222—7. Gülmezoglu AM, Crowther CA, Middleton P, Heatley E. Induction of labor for improving birth outcomes for women at or beyond term. Cochrane Database of Syst Rev 2012;6:CD004945. Hussain AA, Yakoob MY, Imdad A, Bhutta ZA. Elective induction for pregnancies at or beyond 41 weeks of gestation and its impact on stillbirths: a systematic review with meta-analysis. BMC Public Health 2011;11:S5. Wennerholm UB, Hagberg H, Brorsson B, Bergh C. Induction of labor versus expectant management for postdate pregnancy: is there sufficient evidence for a change in clinical practice? Acta Obstet Gynecol 2009;88:6—17. Hannah ME, Hannah WJ, Hellmann J, Hewson S, Milner R, Willan A. Induction of labor as compared with serial antenatal monitoring in post-term pregnancy. A randomized controlled trial. The Canadian multicenter post-term pregnancy trial group. N Engl J Med 1992;326:1587—92. MacDorman MF, Kirmeyer S. Fetal and perinatal mortality, United States, 2005. Natl Vital Stat Rep 2009;57:1—19. Heimstad R, Romundstad PR, Salvesen KA. Induction of labor for post-term pregnancy and risk estimates for intrauterine and perinatal death. Acta Obstet Gynecol Scand 2008;87:247—9. Sentilhes L, Bouet PE, Mezzadri M, Combaud V, Madzou S, Biquard F, et al. [Assessment of the benefit-harm balance depending on gestationnal age to induce delivery for post-term pregnancies]. J Gynecol Obstet Biol Reprod (Paris) 2011;40:747—66 [Article in French]. Gelisen O, Caliskan E, Dilbaz S, Ozdas E, Dilbaz B, Ozdas E, et al. Induction of labor with three different techniques at 41 weeks of gestation or spontaneous follow-up until 42 weeks in women with definitely unfavorable cervical scores. Eur J Obstet Gynecol Reprod Biol 2005;120:164—9. Sahraoui W, Hajji S, Bibi M, Nouira M, Essaidi H, Khairi H. [Management of pregnancies beyond forty-one week’s gestation with an unfavorable cervix]. J Gynecol Obstet Biol Reprod 2005;34:454—62 [Article in French]. Mathiesen ER, Ringholm L, Damm P. Pregnancy management of women with pregestational diabetes. Endocrinol Metab Clin North Am 2011;40:727—38.
