| dc.description.abstract | Evidence synthesis is considered a corner stone of modern health care and clinical practice. Systematic reviews of randomised trials, when undertaken with meta-analysis provide summary estimates on the effectiveness of interventions. However, the findings of meta-analysis are often limited by the selective reporting of primary studies, and the variations in population, intervention and outcomes. Furthermore, difficulties in disentangling the study and individual level associations in meta-analysis make them susceptible to ecological fallacy, and may lead to incorrect conclusions.
Meta-analysis using Individual Participant Data (IPD) has the potential to overcome many of the above limitations, by using raw trial data. Access to IPD minimises problems from incomplete or incorrect reporting of trial outcomes, by verifying reported results, and by standardising the definition of outcomes where possible. Importantly, this allows detecting any variation in the effects of interventions according to characteristics of the participants. Amalgamated individual datasets assembled to address the effectiveness question, can be further used to explore secondary objectives such as the relationship between surrogate and clinical outcomes. This maximises the use of available clinical data, and addresses the problem of research waste.
In this thesis, I evaluated the effects of diet and physical activity based interventions in pregnancy on maternal and offspring outcomes using both study-level and IPD meta-analyses, and assessed the differential effects of interventions on outcomes according to mother’s BMI pre or in early pregnancy. I reviewed the variation in outcomes reported in this field, and developed composite outcomes for IPD meta-analysis. I also evaluated the relationship between weight gain in pregnancy and clinical outcomes in pregnancy using the IPD meta-analysis methodology.
Aims
The aim of this thesis was to evaluate the effects of diet and physical activity based interventions in pregnancy on clinical outcomes using standard and advance methods of evidence synthesis; assess the variation in outcomes and their clinical importance in a trial with those interventions and examine the relationship between gestational weight gain and important clinical outcomes.
Methods
Delphi methodology, systematic reviews of literature, and meta-analyses using study-level and individual participant data of randomised controlled trials (RCTs).
Results
Composite outcomes
Developed composite outcomes comprise of four maternal (gestational diabetes, hypertensive disorders in pregnancy, preterm birth, caesarean section) and four offspring outcomes (stillbirth, small for gestational age, large for gestational age, and admission to neonatal intensive care unit). The components to assess maternal composite outcome were available in two-thirds (66.7%, 24/36) and for offspring composite in half (50%, 18/36) of the studies in the IPD meta-analysis. The effect of interventions was not statistically significant neither on the maternal nor on the offspring composite – Odds Ratio (OR) 0.90 (95% CI 0.79, 1.03) and OR 0.94 (95% CI 0.83, 1.08), respectively. The direction of the pooled effect was consistent between the composite and its components for the maternal composite and variable for the offspring outcomes.
6
Effects of diet and physical activity based interventions
The IPD meta-analysis of 36 RCTs (>12 500 women) showed a significant effect of diet and physical activity based interventions in pregnancy in reducing gestational weight gain (Mean Difference -0.70 kg, 95% CI -0.92, -0.48) and chance of caesarean section delivery (OR 0.91, 95% CI 0.83, 0.99) in comparison to routine antenatal care. There was no effect of the interventions on any of the offspring complications. Incorporation of outcome data unavailable on study-level returned more modest magnitude of the summary estimates in comparison to effects obtained using study-level data of trials that shared IPD. The addition of study-level data from non-IPD trials changed the magnitude and the statistical significance of the summary effects on GDM – from OR 0.89 with only IPD (95% CI 0.72, 1.10; 27 studies, 9 427 women) to OR 0.76 (95% CI 0.65, 0.89; 59 studies, 16 885 women). It has also changed the funnel plot structure in the meta-analysis for gestational weight gain (Egger’s test p = 0.04 with only IPD to p= 0.61).
The IPD meta-analysis shows that the effects of diet and physical activity based interventions on the maternal and the offspring outcomes did not differ by women’s BMI status. While the study-level meta-regression indicated that the interventions might reduce gestational weight gain stronger for the obese women – coefficient -0.22 (95% CI -0.33, -0.11) for each 10% change in the proportion of women in the obese class.
Outcomes in trials with diet and physical activity based interventions
66 primary publications from trials with diet and physical activity based interventions in pregnancy reported 142 outcomes. Half of those outcomes appeared in the publications once (72/142). ‘Critically important’ outcomes are reported less often in comparison to ‘non-critical’ ones (15.5%, 22/142 vs 68.3%, 97/142). The overall quality of outcome reporting varied between trials with the least frequently provided information on the methods to improve the quality of outcome measures (33.3%, 22/66 publications).
7
Gestational weight gain and pregnancy outcomes
IPD from 4 429 pregnant women randomised to the control arms of RCTs with diet and physical activity based interventions were available for the analysis. Women who most often exceeded the IOM recommendation belonged to the overweight (51.5%, 641/ 1 245 women) and the obese groups (44.5%, 695/ 1 562 women) while women with normal BMI most often gained below the recommended amounts (40%, 649/1 622 women). Each kilogram of gestational weight gain within the IOM ranges was not link with a change in the chances of preterm birth, caesarean section, or birth of LGA and SGA infant. Not achieving of the recommended weight was associated with the decreasing chance of giving birth to LGA infant with each kilogram below the lower limit among the obese women (OR 0.80, 95% CI 0.65, 0.99). Each kilogram of weight gain above the upper limit was associated with an increase in the chance of caesarean section (adjusted 1.04, 95% CI 1.01, 1.08) and delivering LGA infant (adjusted 1.08, 95% CI 1.05, 1.12) regardless on women’s BMI status.
Conclusions
Diet and physical activity based interventions in pregnancy moderately reduced gestational weight gain and decrease the odds of caesarean delivery. Overall, IPD meta-analysis improved the robustness of the evidence synthesis of RCTs with diet and physical activity based interventions. However, more attention is needed for the data-related issues in IPD meta-analysis as the purported benefits of the method are not always practically realised. The use of the composite outcomes was hampered by the variable availability of important clinical outcomes. The introduction of minimal core outcome set would facilitate the comparison of the wide range of the evaluated interventions and improve implementation of the composite outcomes. Gestational weight gain was found to be associated with the odds of delivering LGA infant and caesarean section. Future research should aim to collect and report a minimal set of outcomes, and ensure better reporting of study conduct and its findings. | en_US |
