Open access Protocol BMJ Open: first published as 10.1136/bmjopen-2019-035968 on 23 July 2020. Downloaded from Infant formula composition and educational performance: a protocol to extend follow- up for a set of randomised controlled trials using linked administrative education records Maximiliane Verfürden ,1 Katie Harron ,1 John Jerrim,2 Mary Fewtrell,1 Ruth Gilbert1 To cite: Verfürden M, Harron K, ABSTRACT Strengths and limitations of this study Jerrim J, et al. Infant formula Introduction The effect of infant nutrition on long- term composition and educational cognition is important for parents and policy makers. ► This study uses seven infant formula trials con- performance: a protocol to However, most clinical trials typically have short follow-up extend follow- up for a set of ducted in England to determine whether a range of periods, when measures of cognition are poorly predictive of randomised controlled trials nutritional interventions during infancy affects edu- later function. The few trials with longer- term follow- up have using linked administrative cational performance. high levels of attrition, which can lead to selection bias, and education records. BMJ Open ► We showcase the research potential of linking ex- in turn to erroneous interpretation of long- term harms and 2020;10:e035968. doi:10.1136/ tant trials to administrative data, which can offer a benefits of infant nutrition. We address the need for unbiased, bmjopen-2019-035968 low cost way to extend follow- up of early nutrition long- term follow- up, by linking measures of educational Prepublication history and trials, maintain high rates of follow-up and safe- ► performance from administrative education records. additional material for this guard confidentiality. Educational performance is a meaningful marker of cognitive paper are available online. To ► There is no agreement on what defines educational function in children and it is strongly correlated with IQ. We view these files, please visit performance and at what age it is best measured aim to evaluate educational performance for children who, the journal online (http:// dx. doi. to determine intervention effects. We therefore rely org/ 10. 1136/ bmjopen- 2019- as infants, were part of a series of trials that randomised on data from two objectively assessed, compulsory http://bmjopen.bmj.com/ 035968). participants to either nutritionally modified infant formula or examinations sat by the majority of state school stu- standard formula. Most trialists anticipated positive effects of dents in England. Received 25 November 2019 these interventions on later cognitive function. Revised 03 June 2020 Methods and analysis Using data from 1923 participants Accepted 10 June 2020 of seven randomised infant formula trials linked to the English National Pupil Database (NPD), this study will provide cognitive function to inform infant feeding new insights into the effect of nutrient intake in infancy on recommendations and guide parents. school achievement. Our primary outcome will be the mean However, the majority of randomised differences in z- scores between intervention and control controlled trials (RCTs) report only short- on September 24, 2021 by guest. Protected copyright. groups for a compulsory Mathematics exam sat at age 16. term effects of formula composition on proxy Secondary outcomes will be z- scores for a compulsory endpoints for cognition such as the Bayley English exam at age 16 and z- scores for compulsory Scales of Infant Development before 2 years Mathematics and English exams at age 11. We will also 1 evaluate intervention effects on the likelihood of receiving of age. Such early measures are crude, highly special educational needs (SEN) support. All analyses will be observer dependent and have poor predictive performed separately by trial. properties for later academic and employ- 1 © Author(s) (or their Ethics and dissemination Research ethics approval, and ment outcomes. Better measures such as IQ employer(s)) 2020. Re- use approval from the Health Research Authority Confidentiality scores are available but limited by method- permitted under CC BY. Advisory Group, has been obtained for this study. The ological issues such as small sample sizes, high Published by BMJ. results of this study will be disseminated to scientific, attrition and no intention- to- treat analyses.2 3 1Great Ormond Street Institute of practitioner, and lay audiences, submitted for publication in Advances in the availability and quality of Child Health, University College peer- reviewed journals, and will contribute towards a PhD administrative datasets have created new London, London, UK dissertation. 2Institute of Education, possibilities for obtaining long-term outcomes University College London, on children’s educational performance, London, UK INTRODUCTION which is a good predictor of future academic 4 Correspondence to Rationale and employment opportunities. Extending Maximiliane Verfürden; There is a need for high quality evidence on extant trial cohorts with educational perfor- m. verfuerden@ ucl. ac. uk the long- term effects of infant formula on mance data is achievable where trial data and Verfürden M, et al. BMJ Open 2020;10:e035968. doi:10.1136/bmjopen-2019-035968 1 Open access BMJ Open: first published as 10.1136/bmjopen-2019-035968 on 23 July 2020. Downloaded from participant identifiers have been retained and governance interventions (sn-2 Palmitate and Nucleotides) were arrangements allow secure linkage without the need for assumed to have an effect on long-term cognitive func- participants to re-engage. Educational performance is tion. All analyses will be stratified by trial as effect mecha- highly correlated with IQ scores5 but socially more rele- nisms might differ. vant than IQ scores as it determines future academic and employment opportunities.6 Linkage to administrative Outcomes records is available for a fraction of the cost compared Primary outcome with setting up new RCTs or using traditional follow-up At age 16 years (during Key Stage 4), English pupils methods and could offer a more rapid, complete and take their General Certificate of Secondary Education therefore less biased way to determine long- term inter- (GCSE). The primary outcome for this study will be the vention effects.7 8 mean difference in GCSE mathematics exam z-score We sought to determine long-term effects of modified between control and intervention groups. GCSE math- infant formulas by linking RCTs to educational perfor- ematics exams are compulsory, nationally administered mance data for seven existing trial cohorts that compared and are graded from 58 points to 0 point (table 1). modified infant formula interventions with standard We chose Mathematics over English, which is also infant formula. compulsory and nationally administered, because exam results for mathematics are commonly considered to be Research objective 1 less subjectively graded.11 We chose the primary endpoint To determine the effect of nutritionally modified infant at age 16 rather than age 11 as it is a more relevant formula on educational performance in standardised predictor of future education and employment opportu- Mathematics and English language tests at ages 11 and nities. Children who have not yet attended Key Stage 4 at 16. the time of linkage will not be included in the primary outcome. Research objective 2 To assess whether infant formula modifications have an Secondary outcomes effect on the risk of receiving special educational needs As secondary outcomes, we will investigate intervention support during school. effects of modified infant formula versus control formula on: Research objective 3 ► Mean GCSE English language exam z- scores. To explore whether the effect of infant formula modifi- ► Mean Mathematics and English reading exam z- scores cations on educational performance differs for boys and at age 11 (Key Stage 2, final year of primary school). girls, by maternal smoking status during pregnancy, birth ► Probability of receiving five or more GCSEs with weight or maturity at birth. grades A* to C, which includes Mathematics and http://bmjopen.bmj.com/ English. (This is a critical accountability measure for schools and is used in school performance tables). METHODS ► Probability of receiving special educational needs Data sources and linkage (SEN) support. We will follow 1923 participants from seven separate All mentioned Mathematics and English exams are infant formula trials that randomised infants between compulsory and nationally administered. GCSE English 1993 and 2002 and were conducted in England (figure 1). language scores are graded like GCSE Mathematics scores. Full details on the individual trial cohorts as well as previ- Mathematics exams at age 11 are graded from 0 to 100, on September 24, 2021 by guest. Protected copyright. ously reported outcomes can be found elsewhere (online with 100 being highest possible score. English reading supplementary table 1). exams at age 11 are graded from 0 to 50, with 50 being Outcome data for this follow-up will be retrieved from highest. Receiving five or more GCSEs with grades A* to the National Pupil Database (NPD), an administrative C is a commonly reported measure of academic perfor- data resource containing pupil- level and school- level data mance as this measure feeds into entry requirements for a on all pupils in state schools in England and held by the 9 10 large number of sixth form colleges and therefore deter- Department for Education. The Fischer Family