perinatal outcomes in singletons following in vitro fertilization:...

REVIEW

Perinatal Outcomes in Singletons Following In VitroFertilization: A Meta-Analysis

Rebecca A. Jackson, MD, Kimberly A. Gibson, MD, MPH, Yvonne W. Wu, MD, MPH, andMary S. Croughan, PhD

OBJECTIVE: To estimate whether singleton pregnancies fol-lowing in vitro fertilization (IVF) are at higher risk ofperinatal mortality, preterm delivery, small for gestationalage, and low or very low birth weight compared withspontaneous conceptions in studies that adjusted for ageand parity.

DATA SOURCES: We searched MEDLINE, BIOSIS, Doc-toral Dissertations On-Line, bibliographies, and confer-ence proceedings for studies from 1978–2002 using theterms “in vitro fertilization,” “female infertility therapy,”and “reproductive techniques” combined with “fetaldeath,” “mortality,” “fetal growth restriction,” “small forgestational age,” “birth weight,” “premature labor,” “pre-term delivery,” “infant,” “obstetric,” “perinatal,” and“neonatal.”

METHODS OF STUDY SELECTION: Inclusion criteria weresingleton pregnancies following IVF compared with spon-taneous conceptions, control for maternal age and parity; 1of the above outcomes; and risk ratios or data to determinethem. Study selection and data abstraction were performedin duplicate after removing identifying information.

TABULATION, INTEGRATION, AND RESULTS: Fifteen studiescomprising 12,283 IVF and 1.9 million spontaneously con-ceived singletons were identified. Random-effects meta-analysis was performed. Compared with spontaneous con-ceptions, IVF singleton pregnancies were associated withsignificantly higher odds of each of the perinatal out-comes examined: perinatal mortality (odds ratio [OR] 2.2;95% confidence interval [CI] 1.6, 3.0), preterm delivery(OR 2.0; 95% CI 1.7, 2.2), low birth weight (OR 1.8; 95% CI1.4, 2.2), very low birth weight (OR 2.7; 95% CI 2.3, 3.1),

and small for gestational age (OR 1.6; 95% CI 1.3, 2.0).Statistical heterogeneity was noted only for preterm deliv-ery and low birth weight. Sensitivity analyses revealed nosignificant changes in results. Early preterm delivery, spon-taneous preterm delivery, placenta previa, gestational dia-betes, preeclampsia, and neonatal intensive care admissionwere also significantly more prevalent in the IVF group.

CONCLUSION: In vitro fertilization patients should be ad-vised of the increased risk for adverse perinatal outcomes.Obstetricians should not only manage these pregnancies ashigh risk but also avoid iatrogenic harm caused by electivepreterm labor induction or cesarean. (Obstet Gynecol2004;103:551–63. © 2004 by The American College ofObstetricians and Gynecologists.)

Since the birth of Louise Brown in 1978, in vitro fertili-zation (IVF) has become the standard treatment formany types of infertility, despite an initial absence ofresearch examining possible adverse effects on themother or child. Delayed child bearing and increasedaccess to infertility treatment have resulted in a dramaticincrease in demand for IVF. In several European coun-tries, assisted reproductive technology pregnancies rep-resent 2–3% of all births,1 whereas 0.7% of all U.S. birthsare the result of assisted reproductive technology.2 Untilrecently, research has focused primarily on the efficacyof various assisted reproductive technology methodsand the rates of early pregnancy loss or multiple gesta-tions. Now some researchers have begun to question thesafety of assisted reproductive technology in terms of itseffects on the patient’s health, her pregnancy, and herinfant.

Recent studies have specifically addressed perinataloutcomes following IVF compared with spontaneousconception after controlling for maternal age, parity,multiple gestations, and other factors. Most studiesfound increases in preterm birth, low birth weight(LBW), or small for gestational age (SGA),3–16 althougha few did not.17–21 Most individual studies had insuffi-cient statistical power to detect significant differences inperinatal mortality.6,9,13,16,20 Because of different study

From the Departments of Obstetrics, Gynecology and Reproductive Sciences andEpidemiology and Biostatistics, University of California, San Francisco; Depart-ment of Obstetrics and Gynecology, Kaiser Permanente, San Francisco; Depart-ments of Neurology and Pediatrics, University of California, San Francisco; andDepartments of Family and Community Medicine, University of California, SanFrancisco, California.

Supported by a grant from the National Institute of Child Health and HumanDevelopment (1 PO1 HD 37074–05).

Presented in an earlier version at the meeting of the American Society forReproductive Medicine, October 20–25, 2001, Orlando, Florida.

VOL. 103, NO. 3, MARCH 2004551© 2004 by The American College of Obstetricians and Gynecologists. 0029-7844/04/$30.00

Published by Lippincott Williams & Wilkins. doi:10.1097/01.AOG.0000114989.84822.51

designs, populations, and reported outcomes, it is diffi-cult to determine whether IVF pregnancies are at higherrisk of adverse outcomes. Reviews on this topic havebeen neither systematic nor quantitative and have cometo different conclusions.22–26

We conducted a meta-analysis to evaluate whethersingleton pregnancies conceived with IVF are at higherrisk of perinatal mortality, preterm delivery, LBW, orSGA when compared with naturally conceived singletonpregnancies. Given the influence of maternal age andparity on obstetric outcome,27,28 only studies that con-trolled for maternal age and parity in either the design oranalysis were included.

SOURCES

We searched MEDLINE and BIOSIS using a searchstrategy combining the keywords and subject terms “invitro fertilization,” “female infertility therapy,” and “repro-ductive techniques” with “mortality,” “fetal death,” “fetalgrowth restriction,” “small for gestational age,” “pretermdelivery,” “premature labor,” “birth weight,” “infant,” “ob-stetric,” “perinatal,” and “neonatal” for articles publishedbetween 1978 (the date of the first IVF birth) and October2002. Animal studies and case reports were excluded.Conference proceedings available online and DoctoralDissertations On-Line were also searched as were ab-stracts from the 1999 to 2001 meetings of the AmericanSociety of Reproductive Medicine, Society for MaternalFetal Medicine, European Society for Human Reproduc-tion and Embryology, and International Federation ofGynecology and Obstetrics. Bibliographies were re-viewed to identify other relevant studies. Studies pub-lished in languages other than English were considered ifan English abstract was provided.

STUDY SELECTION

Inclusion criteria were applied in 2 stages. First, weidentified studies for blinded review that included acomparison of IVF singleton pregnancies to spontaneousconceptions and reported rates of perinatal mortality,preterm birth, LBW, very low birth weight (VLBW), orSGA in both groups. When the type of infertility treat-ment was unclear or the study group received a varietyof treatments in addition to IVF, the study was submittedfor further review. Two authors (R.J., K.G.) indepen-dently reviewed manuscripts meeting the above criteriaafter removing authors’ names, journal titles, and fund-ing sources. Final inclusion criteria were then applied: 1)comparison of IVF to spontaneous conception; 2) morethan 50% in the IVF group had received standard IVF;3) more than 50% in the spontaneous conception group

were fertile; 4) control for at least maternal age andparity; 5) singleton gestations reported and analyzedseparately from multiple gestations; 6) outcomes explic-itly defined and including 1 of the above; and 7) riskratios with 95% confidence intervals (CIs) provided orsufficient data to enable calculation. Standard IVF wasdefined as ovulation induction, egg retrieval, IVF andintrauterine embryo transfer of a fresh embryo. A thirdauthor (M.C.), blinded to study author and journal,settled discrepancies regarding inclusion criteria anddata abstraction.

We excluded studies that compared a series of IVFbirths with population rates unadjusted for age or parity,studies where more than 50% of subjects received varia-tions of standard IVF such as gamete intrafallopiantransfer, IVF with intracytoplasmic sperm injection orcryopreserved or donor embryos, and studies of sponta-neous conception in infertile women. When multiplepublications reported data for the same study subjects,the most recent publication was selected. Study authorswere asked to provide further information when it wasunclear whether a study met our inclusion criteria.

Two authors (K.G., R.J.) abstracted data about studydesign, population, data collection methods, and rawdata onto standardized forms. Study designs were clas-sified as 1) traditional cohort: IVF births (exposed) com-prised a subpopulation of a larger birth cohort (unex-posed), and outcomes were compared between the 2groups in their entirety; 2) matched cohort: IVF birthscomprised a subpopulation of a larger birth cohort, butonly a matched group of the unexposed cohort was usedfor comparison; and 3) external comparison cohort(double cohort): IVF births were compared with a groupof spontaneous conceptions from a different population.

Primary outcomes were defined as follows: SGA �birth weight less than tenth percentile; preterm deliv-ery � delivery less than 37 completed weeks of gesta-tion; LBW and VLBW � weight less than 2,500 g andless than 1,500 g respectively; and perinatal mortality �stillbirths plus early neonatal deaths (7 days or less). Thedefinition of stillbirth ranged from more than 20 weeksto more than 28 weeks depending on the definition usedin the author’s locale; we used the author’s definition ofstillbirth in our analyses. Meta-analysis of secondaryobstetric and neonatal outcomes was performed when 3or more studies provided data for a given outcome.These included early preterm birth (less than 32–33weeks), type of labor, delivery method, malpresentation,gestational diabetes, pregnancy-induced hypertension,vaginal bleeding, placenta previa, and neonatal intensivecare unit admission. Definitions of secondary outcomeswere taken directly from the included reports and insome cases differed between studies. Malformation rates

552 Jackson et al IVF Meta-Analysis OBSTETRICS & GYNECOLOGY

were abstracted but are not analyzed because of extremevariability between studies in the definition and ascer-tainment of malformations.

Studies were assigned a quality score for use in strati-fied analysis. The score of 0 to 10 was based on 5 studydesign characteristics awarded 0 to 2 points as follows:traditional cohort � 2 points, matched cohort � 1,external comparison cohort � 0; enrollment begun on orafter 1990 � 2, before 1990 � 0; adjustment for age andparity only � 1, for additional factors � 2; delivery ofboth groups at the same hospital(s) � 2, delivery of bothgroups region-wide � 1; delivery of IVF group at 1–3hospitals and spontaneous group region-wide � 0; andmore than 500 subjects in the IVF group � 2, more than300 � 1, less than 300 � 0.

We used adjusted odds ratios (ORs) from reports thatprovided them. For reports that provided multiple ORsfor a given outcome, we chose the one that adjusted forage, parity, and delivery date. For studies that providedprimary data and matched subjects for at least age andparity, we calculated ORs and 95% CIs using the methodof Woolf.29 When a zero cell was encountered, a value of0.5 was added to each cell in the table. When a publishedOR was inconsistent with the raw data provided by thestudy, the effect measure was recalculated using the avail-able data. When data were insufficient to calculate an OR,we attempted to contact the original authors.

Summary ORs were calculated by taking a weightedaverage of individual study results using a general vari-ance-based, random-effects model, weighing individualstudy results by the inverse of their variance.30 Theweight for each study was the inverse of the sum of 2terms: the study variance and a term accounting for thebetween-study variability.30 We chose the random-effects model because it is considered more conservativethan a fixed effects model. In sensitivity analysis, wecompared results of random and fixed effects models.We did not use the quality score in the calculation ofweights of the individual studies. An OR less than 1.0indicates a better outcome in the IVF group, whereas anOR greater than 1.0 favors the spontaneous conceptiongroup.

Heterogeneity was tested by using the general vari-ance-based method30 in which a conservative value ofP � .10 was used to classify study results as heteroge-neous. We attempted to identify sources of heterogene-ity and bias by performing stratified analyses. We com-pared ORs in subgroups with differing study designs,delivery sites, study dates, sample sizes, and quality.Differences in the stratified summary estimates wereevaluated using a z score.29

A sensitivity analysis was performed to evaluate thestability of the overall risk estimate. Sensitivity analyses

were performed separately for each of the 5 primaryoutcomes. Potential publication bias was investigatedvisually using funnel plots and mathematically usingEgger’s regression asymmetry test31 and Kendall rankcorrelation test.32 The Egger test evaluated whether theintercept deviated significantly from zero in a regressionof standardized effect estimates against their precision.The Kendall test examined the significance of the Ken-dall rank correlation between the standardized effectsizes and their variances. All calculations were per-formed using STATA (StataCorp, College Station, TX).

RESULTS

Our initial search produced 1,452 citations. Of these, 1,415did not meet initial screening criteria. These included casereports or series; unadjusted population-based IVF registryreports; reports of pregnancy rates, specific IVF methods,selective fetal reduction, and multiple gestation; and studiesthat did not report perinatal outcomes or that involvedgroups other than IVF and spontaneous conception. Theremaining 37 articles were submitted for blinded evalua-tion. Of these, 13 were excluded because they did notmatch or adjust for maternal age and parity,21,33–44 4had fewer than 50% IVF in the infertile group,10,11,45,46 2did not analyze singletons separately from twins,47,48 1 useda composite neonatal morbidity outcome,49 1 included thesame subjects in a more recent study,19 and 1 did not matchthe standard IVF group to the spontaneous conceptiongroup.50 The remaining 15 studies are summarized inTables 1 and 2. The author of a study published in Norwe-gian7 provided English translation.

The 15 studies include 12,283 IVF singleton pregnan-cies and 1.9 million spontaneously conceived singletons.Individual sample sizes ranged from 5420 to 3,3053 in theIVF group. All studies were retrospective and used varia-tions of a cohort design: 3 were traditional cohort stud-ies,3,5,7 8 were matched cohort studies,4,6,9,14–18 and 4 werecohorts with an external comparison group.8,12,13,20 Noneof the studies were blinded in that obstetricians caring forthe patients could have known that patients had receivedIVF. The majority of studies were performed in Europe,with 1 performed in the United States.20

Study subjects were identified using hospital deliverylogs, IVF clinic records, IVF registries, and regional ornational birth registries. Two studies specifically ex-cluded infertile women or women treated for infertilityfrom the spontaneous conception group.4,16 Many stud-ies included small proportions with non-IVF treatmentsin the IVF group.3,4,8,9,13,14,18 In addition to age, parity,and delivery date, about half of the studies controlled forother factors such as ethnicity, insurance, smoking, body

553VOL. 103, NO. 3, MARCH 2004 Jackson et al IVF Meta-Analysis

mass index, or obstetric history.5,6,14–18,20. Only 4 con-trolled for delivery site.6,16–18 In the others, both groupsdelivered at hospitals throughout the region3–5,7,9,14,15 orthe IVF group delivered throughout the region and thespontaneous group delivered at 1 to 3 local hospi-tals.8,12,13,20

Individual study data and meta-analytic summaryORs for each primary outcome are shown in Figures1–5. The number of eligible studies and subjects for eachoutcome ranged from 7 studies with 1,889 IVF pregnan-cies for the outcome of SGA (Figure 5), to 14 studies with12,114 IVF pregnancies for the outcome of pretermdelivery (Figure 2). The overall incidence of each out-come in the IVF group was perinatal mortality 19.6/1,000, preterm delivery 11.5%, LBW 9.5%, VLBW2.5%, and SGA 14.6%. Meta-analysis of each outcomerevealed that IVF singletons had significantly elevatedodds of each adverse outcome as compared with spon-taneously conceived singletons: perinatal mortality (OR2.19; 95% CI 1.61, 2.98), preterm delivery (OR 1.95;95% CI 1.73, 2.20), LBW (OR 1.77; 95% CI 1.40, 2.22),VLBW (OR 2.70; 95% CI 2.31, 3.14), and SGA (OR1.60; 95% CI 1.25, 2.04).

Statistical heterogeneity (P � .10) was not detected forperinatal mortality, VLBW, or SGA but was found forpreterm delivery and LBW. Sources of heterogeneityfor these outcomes were investigated by stratifying forstudy design, sample size, years of study, delivery site,and study quality (Figure 6). The majority of subgroups

had ORs that were similar to the overall summary ORand remained statistically significant. Heterogeneity thatwas present in the overall meta-analysis of preterm de-livery and LBW was partially explained with stratifica-tion by study design features. For example, for bothpreterm delivery and LBW, smaller studies, those per-formed after 1990, with region-wide delivery of IVFbirths but local delivery of spontaneous conceptions, orwith external comparison groups were not statisticallyheterogeneous (P � .10). Furthermore, higher-qualitystudies showed significantly higher odds of preterm birthand LBW than did lower-quality studies.

Sensitivity analyses were performed for each of the 5primary outcomes. Sequential removal of each studyfrom the meta-analyses resulted in no significant changesin the overall summary ORs. Fixed effects models pro-duced similar results as random-effects models. A num-ber of studies that were originally excluded were addedin the sensitivity analyses.10,11,21,41,46,50 Addition ofthese did not change any of the summary ORs. Namely,when we included Schieve,10 the largest study excluded,the OR for LBW changed from 1.77 to 1.76 and forVLBW from 2.70 to 2.51. For SGA, we added studieswith slightly different definitions of SGA10,12,13 with nochange in the summary OR. Two studies that met ourinclusion criteria were excluded in sensitivity analysis.Von During7 reported adjusted relative risks instead ofORs. Tanbo13 controlled for parity by including onlynulliparous women in the spontaneous group. Exclusion

Table 1. Characteristics of Included Studies Comparing Perinatal Outcomes in IVF Versus Spontaneous Conceptions

AuthorStudyperiod Country

Qualityscore*

Sample size (n) IVF group

IVF SpontaneousMeanage (y)

Nulliparous(%)

Intracytoplasmicsperm injection

(%)

Traditional cohort (single population, entire cohort examined)Bergh3 1982–1995 Sweden 5 3,305 1,490,667† . . . . . . 7Gissler5 1991–1993 Finland 9 746 188,381 . . . 77 0Von During7 1988–1991 Norway 5 545 233,905 . . . . . . 0

Matched cohort (single population, matched subgroup of cohort examined)Dhont4 1992–1997 Belgium 6 3,057 3,057 31.6 75 10Koivurova15 1990–1995 Finland 5 153 287 . . . . . . 0.6Koudstaal16 ?–1992‡ Netherlands 5 307 307 32.8 72 0Maman17 1989–1994 Israel 5 169 469 34.1 60 0Reubinoff18 1983–1993 Israel 5 260 260 32.7 . . . 0Verlaenen6 1979–1986 Belgium 5 140 140 31.7 84 0Wang14 1986–1998 Australia 6 1,019 1,019 32.5 68 20Westergaard9 1994–1999 Denmark 6 1,298 1,298 33.1 . . . 8

External comparison cohort (2 independent populations)Howe20 Not stated United States 2 54 54 33 57 0Tan12 1982–1989 England 3 494 978 34.2 74 0Tanbo13 1978–1987 Norway 3 355 643 33 79 0Wang8 1982–1989 Australia 3 465 21,547 . . . 73 0

* 0 to 2 points each for study date, design, delivery site, sample size, and adjustment for confounders† Exact number not given; estimate made by subtracting multiple gestations from total births‡ Start date not given.


Table 2. Design of Included Studies Comparing Perinatal Outcomes in IVF Versus Spontaneous Conceptions

Author

Study population Additionalvariables

controlled for*Method of

control Delivery siteSource of

outcome dataIVF Spontaneous

Traditional cohortBergh3 All assisted reproductive

technology pregnanciesin Sweden obtained viaquery of the 14 IVFclinics and linked to thenational birth registry;99% successfully linked;7% intracytoplasmicsperm injection, 8%cryopreserved embryos

All births during thesame period fromthe national birthregistry

. . . Stratifiedanalysis

ThroughoutSweden

National birthand deathregistries

Gissler5 All births in the Finnishbirth registry coded asIVF; approximately 20%of IVF birthsmisclassified asspontaneous conceptions


County, smoking,marital status

Multiplelogisticregression

ThroughoutFinland

National birthregistry

Von During7 All births in the Norwegianbirth registry coded asIVF; 11% excluded dueto insufficient data in thebirth registry


. . . Stratifiedanalysis

ThroughoutNorway


Matched CohortDhont4 All assisted reproductive

technology births from asingle Dutch region;method of classifyingbirths as assistedreproductive technologynot explicitly stated;approximately 10%intracytoplasmic sperminjection

Matched sample ofbirths in the sameregion from theregional birthregistry; birthsfrom ovulationinductionexcluded

Fetal sex 1:1 matching ThroughoutDutch-speakingBelgium

IVF: notstated;spontaneous:regionalbirth registry

Koivurova15 All IVF births in northernFinland identified fromIVF clinic records

Matched sample ofbirths in the sameregion from thenational birthregistry

Fetal sex, socialclass, region

1:2 matching ThroughoutnorthernFinland


Koudstaal16 All IVF singletons thatreceived IVF, prenatalcare and delivery at thesame hospital; excludedcryopreserved embryosand embryo reduction;excluded 19% because noappropriate match couldbe found

Matched singletonswith prenatal careand delivery at thesame hospital asthe IVF subject;excluded infertilitytreatment andunknowngestational length

Delivery site,height, weight,smoking, race,medical andobstetric history

1:1 matching 4 universityhospitals

Hospital charts

Maman17 All IVF singleton births toJewish women at a singlehospital; IVF received atmultiple sites

Matched births toJewish women atthe same hospital

Gestational age,delivery site

1:3 matching Single universityhospital

Computerizedhospitalcharts

Reubinoff18 All IVF pregnancies from 1center, 14%cryopreserved embryos,6% ovum donation

Matched sampledelivering at thesame hospital

Ethnicity 1:1 matching “Most” IVF andall spontaneousat 1 universityhospital

Hospitalrecords

Verlaenen6 All IVF pregnancies from 1center who also receivedprenatal care anddelivered at the affiliatedhospital; excludedembryo reduction

Matched sample ofpublicly insuredpatients withprenatal care anddelivery at thesame hospital

Height, weight,delivery site

1:1 matching Single universityhospital

Hospitalrecords

Wang14 All assisted reproductive tech-nology singletons from 1center that delivered inSouth Australia and hadbirth information availablein regional birth database;11% GIFT, 20% intracyto-plasmic sperm injection

Matched samplefrom the regionalbirth registry

Obstetric history,fetal sex,delivery type,malformations

1:1 matching ThroughoutSouth Australia

Regional birthregistry

(continued )


Figure 1. Individual study dataand meta-analytic summary oddsratios for perinatal mortality out-comes. CI � confidence interval;IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol2004.

Table 2. Design of Included Studies Comparing Perinatal Outcomes in IVF Versus Spontaneous Conceptions (continued)

Author

Study population Additionalvariables

controlled for*Method of

control Delivery siteSource of

outcome dataIVF Spontaneous

Westergaard9 All assisted reproductivetechnology births inDenmark identified viamandatory assistedreproductive technologyregistry linked to nationalbirth registry; 8.5%intracytoplasmic sperminjection, 5%cryopreserved and 2%ovum donation

Matched samplefrom nationalbirth registry

1:1 matching ThroughoutDenmark

National birthand deathregistries

External comparison cohortHowe20 First 100 IVF pregnancies

� 20 weeks gestationfrom 1 center withavailable records

Matched samplefrom an affiliateduniversity hospital

Race, medicalproblems,diethylstilbestrol,insurance

1:1 matching IVF: 32 hospitals;spontaneous: 1universityhospital.

IVF: query ofobstetrician;spontaneous:hospitalrecords

Tan12 All IVF pregnancies from 1center, identified via theBritish IVF registry;British residents only;16% excluded becauseobstetric outcomes notavailable

All primiparouspatients from1988–89 from thedatabase of 1university and 1communityhospital

Not matched forparity butcontrols allprimiparous

Stratummatching

IVF: throughoutEngland;spontaneous: 2hospitals

IVF: query ofobstetrician;spontaneous:hospitalrecords

Tanbo3 All assisted reproductivetechnology pregnancies�20 weeks gestationsfrom 1 center;Scandinavian descent;vanishing twinsexcluded; 11%intrauterine insemination,6.5% GIFT

Matched sample ofhealthyScandinavianpatients deliveringat 1 hospital; 13%infertile butconceivedspontaneously

1:2 matching IVF: throughoutNorway;spontaneous: 1communityhospital

Hospitalrecords

Wang8 All IVF and GIFTsingletons from 1 center

All singletons fromobstetric databaseof a largeuniversity hospital

Gestational age Logisticregression

IVF: throughoutAustralia;spontaneous: 1universityhospital

IVF: databaseat IVFcenter;spontaneous:labor anddeliverydatabase

IVF � in vitro fertilization; GIFT � gamete intrafallopian transfer.* All studies controlled for at least maternal age, parity and delivery date.


of these studies did not change summary ORs for theassociated outcomes. Details of these analyses are avail-able on request. No evidence of publication bias wasobserved using funnel plots, Egger’s regression asymme-try test,31 or the Kendall rank correlation test32 (dataavailable on request).

Meta-analyses of secondary outcomes revealed thatthe IVF group had significantly higher ORs for stillbirth,early preterm delivery, spontaneous preterm birth, ges-tational diabetes, preeclampsia, placenta previa, vaginalbleeding, labor induction, elective and emergent cesar-ean, neonatal death, and neonatal intensive care unit

admissions (Figure 7). Data were insufficient to examinepremature rupture of membranes, antepartum hospital-ization, fetal distress, or Apgar scores.

DISCUSSION

Our meta-analysis suggests that singleton IVF pregnan-cies are associated with numerous adverse perinatal out-comes, including perinatal mortality, preterm delivery,LBW, and SGA, even after controlling for maternal ageand parity. The findings were remarkably consistent:significantly increased ORs ranging from 1.6 to 2.7 were

Figure 2. Individual study dataand meta-analytic summary oddsratios for preterm delivery out-comes. CI � confidence interval;IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol2004.

Figure 3. Individual study dataand meta-analytic summary oddsratios for low birth weight out-comes. CI � confidence interval;IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol2004.


observed for all of the major perinatal outcomes andnearly all of the secondary outcomes. This consistencywas observed despite varying study designs, patientpopulations, and IVF and obstetric protocols. Manyhave ascribed the higher rates of adverse outcomes to theeffects of multiple gestations and the increased age andnulliparity of women who obtain IVF.3,51–54 Our resultsrefute this assertion in that we observed increased ad-verse outcomes even in studies of singleton gestationsthat controlled for 2 major confounders: maternal ageand parity.

Absolute risks and risk differences are generally moreuseful in counseling patients than relative risks. How-ever, summary absolute risk differences could not beestimated using these studies given that several studieseither did not provide raw incidence data in the sponta-neous group3,14 or provided incidence rates unadjustedfor age and parity.5,7 The absolute risks in the IVF groupas a whole were clearly elevated as evidenced by theincreased incidence of each outcome, ranging from 2.0%

for perinatal mortality to 14.6% for SGA. These num-bers can be used to approximate the absolute risk in IVFsingletons as a group but are simple arithmetic averagesand cannot be applied to individual patients. Further-more, individualized risks based on age and parity wouldbe clinically useful but were also not estimable from thedata given.

A limitation of any meta-analysis, especially one basedupon observational studies, is that biases in individualstudies will be reflected in the summary statistics. Themost likely source of bias in our meta-analysis is relatedto altered management of IVF pregnancies. Because IVFpregnancies are highly valued by patients and theirdoctors, these patients may be more likely to be hospi-talized or to undergo labor induction or cesarean forminor complications, thus leading to iatrogenic increasesin preterm delivery and LBW. Indeed, most studiesreported higher rates of induced labor6,13,16 and electivecesarean6,9,13,16,18 in the IVF group. Could the adverseoutcomes we observed be due to treatment bias? Out-

Figure 4. Individual study dataand meta-analytic summary oddsratios for very low birth weightoutcomes. CI � confidence inter-val; IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol2004.

Figure 5. Individual study dataand meta-analytic summary oddsratios for small for gestational ageoutcomes. CI � confidence inter-val; IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol2004.


comes that are influenced by the obstetrician are cer-tainly subject to treatment bias. However, treatment biascannot account for outcomes that are less directly influ-enced by obstetric management such as perinatal mor-tality, early preterm birth, VLBW, SGA or placentaprevia. For each of these “hard” outcomes, our meta-analysis showed significantly increased odds in the IVFgroup. Furthermore, some researchers eliminated theeffect of labor induction and elective cesarean by exam-ining preterm deliveries that occurred after spontaneouslabor.6,12,16,18 The summary OR was increased in theIVF pregnancies. Finally, a few studies attempted tocontrol for treatment bias by including only IVF andspontaneous deliveries occurring at the same site andmanaged using similar obstetric protocols.6,16–18 Again,the summary ORs were increased in the IVF group. Ifour results were due to treatment bias alone, one wouldexpect lower ORs in the studies that controlled fordelivery site and spontaneous labor.

Examination of the potential sources of heterogeneityamong studies is one of the goals of meta-analysis. In ourstudy, for outcomes that were relatively independent oftreatment decisions such as SGA, very early pretermbirth, VLBW and placenta previa, no heterogeneity wasnoted. However, for outcomes dependent on subjectivemanagement decisions like preterm delivery, cesarean,and LBW, significant heterogeneity was detected. Thisheterogeneity reflects the differing styles of managementat each study center. Stratification by design featureseliminated some of the heterogeneity in that recent,smaller, higher quality studies were more likely to lackheterogeneity. In fact, studies with higher quality scoresnot only showed less heterogeneity, they also showedsignificantly higher rates of preterm birth, perinatal mor-tality, and LBW than lower quality studies, furthersupporting the robustness of our results.

Results relating to the secondary outcomes should beinterpreted with caution. We did not specifically search

Figure 6. Sources of heterogeneity for pretermdelivery and low birth weight. CI � confidenceinterval; IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol 2004.


for studies with these outcomes. Our search was limitedto studies reporting on preterm delivery, LBW, VLBW,SGA or perinatal mortality. Second, for some of theoutcomes, definitions varied from study to study. Forexample, definitions of vaginal bleeding ranged from anyself-reported bleeding to bleeding severe enough to war-rant hospital admission.

The causes of the increased risks we observed areunknown but could be due to the IVF procedure itself, tocomponents of the IVF procedure, or to infertility per se.Recent studies by McElrath45 and Draper41 observedincreased odds for VLBW and perinatal mortality inwomen with untreated infertility. To definitively addresswhether adverse outcomes are associated with infertilitytreatment as opposed to infertility would be ethicallyimpossible, because it would require a randomized clin-ical trial comparing assisted conception with naturalconception in fertile women. Clues can be obtainedabout the risks of assisted reproductive technology bycomparing different subgroups of treated women.Wang14 compared “low technology” treatments (intra-uterine insemination, donor insemination), assisted re-productive technology, and spontaneous conceptions

and found 50% increased odds for preterm birth in thelow technology group and a 2-fold increase in the as-sisted reproductive technology group, indicating thatboth infertility itself and high technology treatments maybe associated with increased preterm birth. Bergh3 com-pared intracytoplasmic sperm injection , primarily per-formed for male factor infertility, with standard IVF,controlling for age and parity and found no statisticallysignificant difference in preterm delivery or LBW, im-plying that fertile and infertile women undergoing IVFhave similar outcomes. However, sample sizes were notgiven, so we were unable to determine whether therewas adequate power. Several smaller studies have alsocompared intracytoplasmic sperm injection with IVFand generally found no difference in perinatal out-comes,55–59 but it is unclear how many women in theintracytoplasmic sperm injection group were known tobe fertile, and most studies did not report singletongestations separate from multiples or control for mater-nal age and parity. Comparing stimulated with unstimu-lated cycles would isolate the effect of ovulation induc-tion agents. Bergh3 found no difference in outcomes forstandard IVF compared with unstimulated IVF con-

Figure 7. Results of meta-analyses of secondary outcomes. Data were insufficient to examine premature rupture ofmembranes, antepartum hospitalization, fetal distress, or Apgar scores. CI � confidence interval; IVF � in vitro fertilization.Jackson. IVF Meta-Analysis. Obstet Gynecol 2004.


trolled for age and parity. Wennerholm50 found a lowerpreterm delivery rate in unstimulated IVF with cryopre-served embryos compared with standard IVF. Olive-ness36 found no difference in preterm birth in thosereceiving only ovulation induction compared with IVF.

Future research should be conducted to further delin-eate the causes of the adverse outcomes observed in IVFsingletons and attempt to better control for treatmentbiases. Given our findings, we recommend that informedconsent for women undergoing IVF should include adiscussion of possible perinatal risks. Furthermore, al-though obstetricians caring for these patients shouldconsider IVF a risk factor for adverse perinatal out-comes, they should also be aware of the increased ratesof labor induction and elective cesarean and attempt toavoid iatrogenic harm caused by preterm labor induc-tion and cesarean.

REFERENCES1. Nygren KG, Andersen AN. Assisted reproductive technol-

ogy in Europe, 1997. Results generated from Europeanregisters by ESHRE. European IVF-Monitoring Pro-gramme (EIM), for the European Society of HumanReproduction and Embryology (ESHRE). Hum Reprod2001;16(2):384–91.

2. Assisted reproductive technology in the United States:1998 results generated from the American Society forReproductive Medicine/Society for Assisted ReproductiveTechnology Registry. Fertil Steril 2002;77(1):18–31.

3. Bergh T, Ericson A, Hillensjeo T, Nygren KG, Wenner-holm UB. Deliveries and children born after in-vitro fertil-isation in Sweden 1982–95: a retrospective cohort study.Lancet 1999;354:1579–85.

4. Dhont M, De Sutter P, Ruyssinck G, Martens G, BekaertA. Perinatal outcome of pregnancies after assisted repro-duction: a case-control study. Am J Obstet Gynecol 1999;181:688–95.

5. Gissler M, Malin Silverio M, Hemminki E. In-vitro fertil-ization pregnancies and perinatal health in Finland1991–1993. Hum Reprod 1995;10:1856–61.

6. Verlaenen H, Cammu H, Derde MP, Amy JJ. Singletonpregnancy after in vitro fertilization: expectations andoutcome. Obstet Gynecol 1995;86:906–10.

7. von During V, Maltau JM, Forsdahl F, Abyholm T, Kol-vik R, Ertzeid G, et al. [Pregnancy, births and infants afterin-vitro-fertilization in Norway, 1988–1991]. Tidsskr NorLaegeforen 1995;115:2054–60.

8. Wang JX, Clark AM, Kirby CA, Philipson G, Petrucco O,Anderson G, et al. The obstetric outcome of singletonpregnancies following in-vitro fertilization/gamete intra-fallopian transfer. Hum Reprod 1994;9:141–6.

9. Westergaard HB, Johansen AM, Erb K, Andersen AN.

Danish National In-Vitro Fertilization Registry 1994 and1995: a controlled study of births, malformations andcytogenetic findings. Hum Reprod 1999;14:1896–902.

10. Schieve LA, Meikle SF, Ferre C, Peterson HB, Jeng G,Wilcox LS. Low and very low birth weight in infantsconceived with use of assisted reproductive technology.N Engl J Med 2002;346:731–7.

11. Perri T, Chen R, Yoeli R, Merlob P, Orvieto R, Shalev Y,et al. Are singleton assisted reproductive technology preg-nancies at risk of prematurity? J Assist Reprod Genet2001;18:245–9.

12. Tan SL, Doyle P, Campbell S, Beral V, Rizk B, BrinsdenP, et al. Obstetric outcome of in vitro fertilization pregnan-cies compared with normally conceived pregnancies. Am JObstet Gynecol 1992;167:778–84.

13. Tanbo T, Dale PO, Lunde O, Moe N, Abyholm T.Obstetric outcome in singleton pregnancies after assistedreproduction. Obstet Gynecol 1995;86:188–92.

14. Wang JX, Norman RJ, Kristiansson P. The effect of vari-ous infertility treatments on the risk of preterm birth. HumReprod 2002;17:945–9.

15. Koivurova S, Hartikainen AL, Gissler M, Hemminki E,Sovio U, Jarvelin MR. Neonatal outcome and congenitalmalformations in children born after in-vitro fertilization.Hum Reprod 2002;17:1391–8.

16. Koudstaal J, Braat DD, Bruinse HW, Naaktgeboren N,Vermeiden JP, Visser GH. Obstetric outcome of singletonpregnancies after IVF: a matched control study in fourDutch university hospitals. Hum Reprod 2000;15:1819–25.

17. Maman E, Lunenfeld E, Levy A, Vardi H, Potashnik G.Obstetric outcome of singleton pregnancies conceived byin vitro fertilization and ovulation induction comparedwith those conceived spontaneously. Fertil Steril 1998;70:240–5.

18. Reubinoff BE, Samueloff A, Ben-Haim M, Friedler S,Schenker JG, Lewin A. Is the obstetric outcome of in vitrofertilized singleton gestations different from natural ones?A controlled study. Fertil Steril 1997;67:1077–83.

19. Dhont M, De Neubourg F, Van der Elst J, De Sutter P.Perinatal outcome of pregnancies after assisted reproduc-tion: a case-control study. J Assist Reprod Genet 1997;14:575–80.

20. Howe RS, Sayegh RA, Durinzi KL, Tureck RW. Perinataloutcome of singleton pregnancies conceived by in vitrofertilization: a controlled study. J Perinatol 1990;10:261–6.

21. Tallo CP, Vohr B, Oh W, Rubin LP, Seifer DB, HaningRVJr. Maternal and neonatal morbidity associated with invitro fertilization. J Pediatr 1995;127:794–800.

22. Tarlatzis BC, Grimbizis G. Pregnancy and child outcomeafter assisted reproduction techniques. Hum Reprod 1999;14(Suppl 1):231–42.

23. Corabian P, Hailey D. The efficacy and adverse effects ofin vitro fertilization and embryo transfer. Int J TechnolAssess Health Care 1999;15:66–85.


24. Buitendijk SE. Children after in vitro fertilization. Anoverview of the literature. Int J Technol Assess HealthCare 1999;15:52–65.

25. Dawood MY. In vitro fertilization, gamete intrafallopiantransfer, and superovulation with intrauterine insemina-tion: efficacy and potential health hazards on babies deliv-ered. Am J Obstet Gynecol 1996;174:1208–17.

26. Schenker JG, Ezra Y. Complications of assisted reproduc-tive techniques. Fertil Steril 1994;61:411–22.

27. Bianco A, Stone J, Lynch L, Lapinski R, Berkowitz G,Berkowitz RL. Pregnancy outcome at age 40 and older.Obstet Gynecol 1996;87:917–22.

28. Prysak M, Kisly A. Age greater than thirty-four years is anindependent pregnancy risk factor in nulliparous women. JPerinatol 1997;17:296–300.

29. Kahn HA, Sempos CT. Statistical methods in epidemiol-ogy. New York (NY): Oxford University Press; 1989.

30. Petitti DB. Meta-analysis, decision analysis and cost-effec-tiveness analysis: methods for quantitative synthesis inmedicine. 2nd ed. New York (NY): Oxford UniversityPress; 2000.

31. Egger M, Smith GD. Bias in location and selection ofstudies. BMJ 1998;316:61–6.

32. Begg CB, Mazumdar M. Operating characteristics of arank correlation test for publication bias. Biometrics 1994;50:1088–101.

33. D’Souza SW, Rivlin E, Cadman J, Richards B, Buck P,Lieberman BA. Children conceived by in vitro fertilisationafter fresh embryo transfer. Arch Dis Child Fetal NeonatalEd 1997;76:F70–4.

34. Frydman R, Belaisch-Allart J, Fries N, Hazout A, GlissantA, Testart J. An obstetric assessment of the first 100 birthsfrom the in vitro fertilization program at Clamart, France.Am J Obstet Gynecol 1986;154:550–5.

35. Leslie GI, Bowen JR, Arnold JD, Saunders DM. In-vitrofertilisation and neonatal ventilator use in a tertiary peri-natal centre. Med J Aust 1992;157:165–7.

36. Olivennes F, Rufat P, Andre B, Pourade A, Quiros MC,Frydman R. The increased risk of complication observedin singleton pregnancies resulting from in-vitro fertilization(IVF) does not seem to be related to the IVF method itself.Hum Reprod 1993;8:1297–300.

37. Petersen K, Hornnes PJ, Ellingsen S, Jensen F, Brocks V,Starup J, et al. Perinatal outcome after in vitro fertilisation.Acta Obstet Gynecol Scand 1995;74:129–31.

38. Saunders K, Spensley J, Munro J, Halasz G. Growth andphysical outcome of children conceived by in vitro fertili-zation. Pediatrics 1996;97:688–92.

39. Tough SC, Greene CA, Svenson LW, Belik J. Effects of invitro fertilization on low birth weight, preterm delivery,and multiple birth. J Pediatr 2000;136:618–22.

40. Hill GA, Bryan S, Herbert CMd, Shah DM, Wentz AC.

Complications of pregnancy in infertile couples: routinetreatment versus assisted reproduction. Obstet Gynecol1990;75:790–4.

41. Draper ES, Kurinczuk JJ, Abrams KR, Clarke M. Assess-ment of separate contributions to perinatal mortality ofinfertility history and treatment: a case-control analysis.Lancet 1999;353:1746–9.

42. Cadman J, Richards B, D’Souza S, Lieberman BA, Buck P,Rivlin E. A computerised study on the results of in-vitro-fertilisation. Stud Health Technol Inform 1999;68:343–6.

43. Silverman RK, Wojtowycz M. The effects of in vitrofertilization on subsequent pregnancy outcome. Prim CareUpdate Ob Gyns 1998;5:169.

44. Johnson MR, Irvine R, Hills F, Bolton VN, Abbas AA,Brooks AA, et al. Superovulation, IGFBP-1 and birthweight. Eur J Obstet Gynecol Reprod Biol 1995;59:193–5.

45. McElrath TF, Wise PH. Fertility therapy and the risk ofvery low birth weight. Obstet Gynecol 1997;90(4 Pt1):600–5.

46. Sheiner E, Shoham-Vardi I, Hershkovitz R, Katz M,Mazor M. Infertility treatment is an independent riskfactor for cesarean section among nulliparous women aged40 and above. Am J Obstet Gynecol 2001;185:888–92.

47. Venn A, Lumley J. Births after a period of infertility inVictorian women 1982–1990. Aust N Z J Obstet Gynaecol1993;33:379–84.

48. Yeh J, Leipzig S, Friedman EA, Seibel MM. Results of invitro fertilization pregnancies: experience at Boston’s BethIsrael Hospital. Int J Fertil 1990;35:116–9.

49. Addor V, Santos-Eggimann B, Fawer CL, Paccaud F,Calame A. Impact of infertility treatments on the health ofnewborns. Fertil Steril 1998;69:210–5.

50. Wennerholm UB, Hamberger L, Nilsson L, WennergrenM, Wikland M, Bergh C. Obstetric and perinatal outcomeof children conceived from cryopreserved embryos. HumReprod 1997;12:1819–25.

51. Beral V, Doyle P, Tan SL, Mason BA, Campbell S.Outcome of pregnancies resulting from assisted concep-tion. Br Med Bull 1990;46:753–68.

52. Wennerholm UB, Janson PO, Wennergren M, Kjellmer I.Pregnancy complications and short-term follow-up ofinfants born after in vitro fertilization and embryo transfer(IVF/ET). Acta Obstet Gynecol Scand 1991;70:565–73.

53. Tarin JJ, Cano A. Effects of manipulations in vitro ofhuman oocytes and embryos on the birthweight of theresultant babies. Hum Reprod 1995;10:1322–4.

54. MRC. Births in Great Britain resulting from assisted con-ception, 1978–87. MRC Working Party on ChildrenConceived by In Vitro Fertilisation. BMJ 1990;300:1229–33.

55. Bonduelle M, Legein J, Derde MP, Buysse A, SchietecatteJ, Wisanto A, et al. Comparative follow-up study of 130children born after intracytoplasmic sperm injection and130 children born after in-vitro fertilization. Hum Reprod1995;10:3327–31.


56. Aytoz A, Van den Abbeel E, Bonduelle M, Camus M, JorisH, Van Steirteghem A, et al. Obstetric outcome of preg-nancies after the transfer of cryopreserved and freshembryos obtained by conventional in-vitro fertilizationand intracytoplasmic sperm injection. Hum Reprod 1999;14:2619–24.

57. Bonduelle M, Liebaers I, Deketelaere V, Derde MP,Camus M, Devroey P, et al. Neonatal data on a cohort of2889 infants born after ICSI (1991–1999) and of 2995infants born after IVF (1983–1999). Hum Reprod 2002;17:671–94.

58. Palermo GD, Colombero LT, Schattman GL, Davis OK,Rosenwaks Z. Evolution of pregnancies and initial fol-low-up of newborns delivered after intracytoplasmicsperm injection. JAMA 1996;276:1893–7.

59. Bowen JR, Gibson FL, Leslie GI, Saunders DM. Medicaland developmental outcome at 1 year for children con-ceived by intracytoplasmic sperm injection. Lancet 1998;351:1529–34.

Address reprint requests to: Rebecca Jackson, MD, Depart-ment of Obstetrics and Gynecology, #6D, San Francisco Gen-eral Hospital, 1001 Potrero Avenue, San Francisco, CA 94110;e-mail: [email protected].

Received September 6, 2003. Received in revised form November 20,2003. Accepted December 4, 2003.


perinatal outcomes in singletons following in vitro fertilization:...

Documents