UOG Journal Club: Prevention of perinatal death and adverse perinatal outcome using low-dose aspirin: a meta-analysis
UOG Journal Club: May 2013Journal Club slides prepared by Dr Leona Poon (UOGEditor for Trainees)Prevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisS. Roberge, K. H. Nicolaides, S. Demers, P. Villa and E. BujoldVolume 41, Issue 5, Date: May 2013, pages 491–499Opinion:Aspirin for pre-eclampsia: beware of subgroup meta-analysisS. Meher, Z. AlfirevicVolume 41, Issue 5, Date: May 2013, pages 479–485
UOG Journal Club: May 2013Prevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisS. Roberge, K. H. Nicolaides, S. Demers, P. Villa and E. BujoldVolume 41, Issue 5, Date: May 2013, pages 491–499
• Leading causes of perinatal death are preterm birth, fetal abnormalityand impaired placentation leading to PET and FGR• Meta-analyses of RCTs suggest that prophylactic use of low-doseaspirin initiated at <16 weeks is associated with significant reductionin prevalence of severe PET, FGR and preterm birth (Bujold et al. 2010;Roberge et al. 2012a; Roberge et al. 2012b)• Mechanism of action of aspirin remains unclear but it could include animprovement in transformation of uterine spiral arteries, which istypically incomplete in PET (Fraser et al. 2012; Vainio et al. 2005)• It is still uncertain whether intervention aimed at improvingplacentation could lead to a reduction of placental-related adversepregnancy outcomes and to a reduction of perinatal deathBujold E et al. Obstet Gynecol 2010;116;402-14.Fraser R et al. J Pathol. 2012;228:322-32.Roberge S et al. Am J Perinatol 2012a;29:551-6.Roberge S et al. Fetal Diagn Ther 2012b;31:141-6.Vainio M et al. Acta Obstet Gynecol Scand 2005;84:1062-7.
Prevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013Compare effect of early vs. late administration oflow-dose aspirin on the risk of perinatal death andadverse perinatal outcomeObjective
Methodology• RCTs that evaluated prophylacticuse of low-dose aspirin (50-150mg/day) during pregnancy• 1965 – October 2011• All studies that involved womenwho initiated low-dose aspirin at<16 weeks and at >16 weeksInclusion criteriaPerinatal mortality:Fetal death >16 weeks orNeonatal death <28 days of agePrimary outcomeSecondary outcomesPre-eclampsiaFetal growth restrictionPreterm birthPlacental abruptionBirth weightGA at deliveryPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013
8,377 potentially appropriate trials from electronicsearch1,104 potentially appropriate trials for inclusion inmeta-analysis7,273 studies excluded: personal communication,duplicate, not randomized study with aspirin, allocationconcealment inadequate, relevant outcome notprovided24 studies excluded: data not available for outcomeof interest, GA at treatment not available or overlap,other reason66 studies meeting inclusion criteria42 studies (27,222 women randomized) withrelevant information included in the systematicreview.1,038 studies excluded: personal communication,duplicate, letter, commentary, editorial, meta-analysis,review, paper retracted, other study design, othertreatment, quality inadequate, relevant outcome notprovided, other reasons
• Methodological quality of studies: Cochrane Handbook Criteria tool• Relative risks (RR) calculated for each study• Global RR calculated stratified according to GA at entry (≤16 weeks vs >16weeks):• Significant heterogeneity: DerSimonian and Laird random effect• Homogeneity: fixed effect• Heterogeneity between studies: Higgins’ I2test• Weighted difference between subgroups of GA at entry evaluated by mixedregression• Publication bias: distribution of RCTs examined using funnel plots• Robustness of findings and heterogeneity between studies assessed by sensitivityanalysisMethodology
Low-dose aspirin starting at ≤ 16 weeks is associated with a significantreduction in risk of perinatal deathAspirin for the prevention of perinatal death0.02 0.1 1 10 50Favors aspirin Favors controlNo. of study Events Total Events TotalAspirin ≤16 wks 0.41 (0.19-0.92)12 7 660 26 6480.93 (0.73-1.19)Aspirin >16 wks 20 125 4,737 143 2,820Aspirin Control0.87 (0.69-1.10)Total 132 5,397 169 5,46832Results
ResultsThere is a possibility ofpublication biasbecause small studieswith no beneficial effectwere missingFunnel plot of distribution of RR for perinatal deathassociated with aspirin treatment (black circles =aspirin started ≤16 weeks, white circles = aspirinstarted >16 weeksPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013
The majority of included studies were deemed low / unclear risk of bias, exceptfor one study with >20% of individuals lost to follow-up and two with a risk ofselective reporting.Assessment of risk of bias in studies included in the analysis(low , unclear , high ) following the Cochrane HandbookResultsPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013
ResultsPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013No statistical differenceobserved between any ofthe subgroups identified insensitivity analysisSensitivity analysis of RR for perinatal death for studieswhere aspirin was initiated ≤ 6 weeks (dashed verticalline = RR from random effects meta-analysis)
• Effect of low-dose aspirin started at ≤ 6 weeks on perinatal death wassignificant and homogeneous according to I2test and sensitivity analysis• Similar effects of early aspirin prophylaxis on perinatal outcomes wereobserved:• in women who received ≤ 80 mg daily and those who received ≥ 100mg daily.• in women who were selected using abnormal uterine artery Doppleras an inclusion criterion as compared with those selected usinganamnesis factors only.ResultsPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013
• The beneficial effect of aspirin is a consequence of an improvement intransformation of uterine spiral arteries. This statement is based on:1. Low-dose aspirin has a greater reduction of preterm and severePET, which are associated with poor placentation2. Abnormal uterine artery blood flow is present as early as 12weeks in women who will subsequently develop PET and aspirinimproves such blood flow between 1stand 2ndtrimesters• Low-dose aspirin should probably be initiated in the first trimester inhigh-risk womenDiscussionPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013
1. None of the included trials was designed to evaluate perinatal deathand few reported the reasons for perinatal death2. Small studies without effect were missing, raising the possibility ofpublication bias3. Definition of perinatal death was heterogeneous between studies4. Six largest trials that could have had the power to examine the impactof low-dose aspirin on perinatal death recruited women mostly >16weeks5. Findings were limited by the small size of the studies that recruited at<16 weeks: each of them individually was underpowered to addressperinatal deathLimitationsPrevention of perinatal death and adverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013
UOG Journal Club: May 2013Opinion:Aspirin for pre-eclampsia: beware of subgroup meta-analysisS. Meher, Z. AlfirevicVolume 41, Issue 5, Date: May 2013, pages 479–485
• The aim of systematic reviews of RCTs is to help achieve consensus abouteffects of interventions by:1. Summarizing the evidence2. Increasing power to detect differential effects3. Assessing consistency of findings4. Reducing risk of bias by using pre-specified, explicit methodology• However, systematic reviews have brought about as much controversy as theyhave consensus, when it comes to aspirin for prevention of PET and adverseperinatal outcomeOpinionAspirin for pre-eclampsia: beware of subgroup meta-analysisMeher and Alfirevic, UOG 2013What is known?
• Data are now available on >37,000 women recruited to >55 RCTs• The PARIS Collaboration performed the ‘gold standard’ individual patient data(IPD) meta-analysis, showing a significant 10% reduction in PET andcomposite serious adverse outcome• Estimated NNT is 67 high-risk women to prevent one case of seriousadverse pregnancy outcome• Recent focus is shifted towards trying to resolve uncertainties throughsubgroup analyses of available dataOpinionAspirin for pre-eclampsia: beware of subgroup meta-analysisMeher and Alfirevic, UOG 2013What is known?
• Smaller numbers and multiple analyses, which increase susceptibility to bias• Important to explore subgroup analyses if:1. Potential large differences between groups in the risk of a poor outcomewith or without treatment2. Potential heterogeneity of treatment effect in relation to pathophysiology3. Practical questions about when to treat4. Doubts about benefit in specific groups• It should be justified carefully and limited to a small number of researchquestions (high risk of +ve effects purely by chance)Subgroup analysesOpinionAspirin for pre-eclampsia: beware of subgroup meta-analysisMeher and Alfirevic, UOG 2013
• As a rule, reports of statistical significance in individual groups should beignored, because rates of false +ve and –ve are high, and the only reliablestatistical approach is to perform an interaction test for subgroup-treatmentinteraction effect• From the review by Roberge et al, the interaction test between subgroupswas statistically significant, suggesting that differences in the effect sizebetween subgroups were more than would be expected to occur bychance• Subgroup analyses should be seen as ‘hypothesis generating’; the best testfor validity of subgroup findings is confirmation in subsequent trialsOpinionAspirin for pre-eclampsia: beware of subgroup meta-analysisMeher and Alfirevic, UOG 2013Subgroup analyses
• The review by Roberge et al has potential important implications regarding whenwe should start aspirin in pregnancy• However, the findings cannot be taken as conclusive because:• Relatively small amount of data (6% of all data) and the data missing fromthe ≤16 week subgroupi. Chances of summary estimate changing significantly with theaddition of new data are highii. Small studies are more likely to overestimate treatment effects dueto publication bias: trials with significant findings are more likely to bepublishediii. Data on perinatal death from the three largest RCTs seem to havebeen available separately for inclusion in the >16 week subgroup butnot in ≤ 16 week subgroupOpinionAspirin for pre-eclampsia: beware of subgroup meta-analysisMeher and Alfirevic, UOG 2013Summary
• Potential impact of other factors such as variable outcome definitions andpossible systematic differences between the women in the two GA subgroupsi. Marked imbalance between baseline risk of PET in controls for <16week subgroup (18%) vs >16 week subgroup (8%)• Choice of 16 weeks as a cut-off is arbitrary• Although multiple subgroup analyses carry the risk of finding significant effectsby chance, the research question can potentially be confirmed by furtheranalyses of PARIS Collaboration IPD data• Future research must ensure that sample sizes are adequate, rather thanresorting to subgroup meta-analyses of small trialsSummaryOpinionAspirin for pre-eclampsia: beware of subgroup meta-analysisMeher and Alfirevic, UOG 2013
Discussion points• Should we be using low-dose aspirin to prevent perinatal death andadverse perinatal outcome ?• When is the optimal gestational age for starting low-dose aspirin?• How do we identify women at risk?• What proportion of women would fit our definition of high risk?• Should low-dose aspirin ever be started after 16 weeks?• What is the optimal dosage of aspirin for the prevention of perinataldeath perinatal death and adverse perinatal outcome?Prevention of perinatal death andadverse perinatal outcome usinglow-dose aspirin: a meta-analysisRoberge et al., UOG 2013OpinionAspirin for pre-eclampsia: beware ofsubgroup meta-analysisMeher and Alfirevic, UOG 2013