Portal de Periódicos da CAPES

Core Data for Nutrition Trials in Infants: A Discussion Document—A Commentary by the ESPGHAN Committee on Nutrition

2003; Lippincott Williams & Wilkins; Volume: 36; Issue: 3 Linguagem: Inglês

10.1097/00005176-200303000-00007

1536-4801

Peter Aggett, Carlo Agostoni, Irené Axelsson, Olivier Goulet, Olle Hernell, Berthold Koletzko, Harry N. Lafeber, Kim F. Michaelsen, Ruth Morley, J Rigó, Hania Szajewska, Lawrence T. Weaver,

Infant Nutrition and Health

Modifications of infant feeding regimens and dietary products need to be evaluated in clinical trials meeting accepted standards of scientific methodology (1). Presentation of the study design and results in a standardized format is now required by many journals and has led to improvement in the reporting of all clinical trials (2). Recommendations of good clinical practice, including those relating to data monitoring, should also be followed by researchers undertaking clinical trials (3). Consensus still needs to be achieved on the core data that should be recorded in nutrition trials and included in all reports. Data presented in published reports of clinical trials of infant nutrition can vary considerably. This inconsistency can make it difficult to compare and fully interpret results from different studies, and it limits the potential for performing meta-analyses. Therefore, in this commentary, the ESPGHAN Committee on Nutrition proposes core data that should be collected for all randomized nutrition trials and suggests principles for the identification of data required for studies with different outcome measures. Our recommendations are based on current understanding and practice and are intended to promote discussion. We have considered core data for trials generally and, by way of example, for those in which growth or neurodevelopment are outcome measures. Trials with different outcome measures (e.g., body composition, markers of atopic and allergic disorders, and gut function and disease) will likely need to record specific additional data. WHICH DATA ARE NEEDED? 1. Confidential Identification Data Subject anonymity and secure storage of confidential data are essential, in accordance with general ethical principles and local ethics committee requirements. These are not strictly core data but are clearly needed to keep in touch with parents and caregivers through the study. They are additionally important to facilitate contact with health professionals if there is need and to facilitate later follow-up evaluation. Trials of infants should be regarded as potentially long term, even if the primary outcome measure(s) specified in the protocol is short term. If intergroup differences in efficacy or safety measures are identified, the long-term impact of these can only be addressed if researchers are able to trace study participants. Researchers should aim to obtain consent for future contact with study participants and identify strategies that will facilitate later tracing of subjects. It is suggested that researchers collect locally useful information (e.g., health service, identity, social security or insurance numbers, and addresses of parents' employers) and names, addresses, and telephone numbers of a few close relatives or friends. Suggested identification data are shown in Table 1.TABLE 1: Suggested identification data2. Data to Characterize the Cohort Participants need to be well characterized, so that: It is possible to determine whether participants are representative of the eligible population; and Trial data can be more effectively compared (and, if relevant, aggregated) with data and results from other studies. We suggest that the core data shown in Table 2 be recorded in all studies, and additionally those shown in Table 3 be recorded in studies of preterm infants.TABLE 2: Core data for all studies of infantsTABLE 3: Additional core data for studies of infants born preterm3. Data on Withdrawals From the Study If parents or guardians ask for their child to be withdrawn from studies and request no further contact, investigators must respect their wishes. In all trials, details of subjects who withdraw and the reasons for withdrawal should be carefully recorded and reported so potential bias can be assessed. Subjects should not be withdrawn from the study by the investigators because of nonconcordance (noncompliance), and nonconcordant participants should be followed in exactly the same way as compliant subjects (4,5). 4. Compliance or Concordance Data Data indicating compliance or concordance must be recorded in intervention trials because they are important for interpretation of trial results, although primary analyses should usually be based on intention to treat (5). Compliance can be recorded in several ways. By asking the person who cares for the child to report his or her compliance with a nutrition protocol. When professionals administer the intervention, this can be a highly reliable method. In a study where, for example, large numbers of infants are randomized to different milks over a long period, this may be the only feasible method. By providing nutritional products, and recording number of unused doses or weighing residual products in preweighed individual containers (e.g., ready-to-feed milks). By measuring blood or urine levels of a biochemical marker of compliance. However, collection of blood samples could make recruitment more difficult and lead to later refusal to continue participation in a trial. 5. Efficacy and Safety Measures The choice of measures of efficacy and safety will be determined by the nature of the study and the hypothesis being tested. Many different efficacy issues are studied; the following are those studied most commonly in nutritional trials. Growth; Markers of body composition (e.g., dual x-ray absorptiometry [DEXA], total body electrical conductivity [TOBEC], bioelectrical impedance analysis [BIA]); Neurodevelopment measures; Markers of allergic and atopic disorders; Markers of gut function and disease; and Markers of immune function. We suggest that growth measures be recorded for all nutrition trials of infants and children, as safety measures if not outcome measures. We also suggest that researchers use outcome measures that are widely considered useful, including quality control and assurance information on the study methods and the outcomes, to facilitate comparative analysis. 6. Baseline Factors That Need to Be Reported Outcomes of interest can be influenced by many factors unrelated to the intervention. Information is needed on such factors so that: The study cohort can be well characterized; and Checks can be made for baseline imbalance in these factors between randomized treatment groups. Adjustment for such factors in secondary statistical analyses should be prespecified in the research protocol (6,7). Such factors need to be considered for each outcome measure. Here, we give examples of such factors that may affect growth (8–12) and neurodevelopment. For other outcomes, further factors may need to be considered, and a careful literature search and consultation with experts in the field of the outcome measures can help researchers identify relevant factors. Factors Associated With Growth or Body Composition Birth weight for gestation (g and SD score for sex and gestation); Sex of the infant; Parental heights and weights; Maternal smoking; and Maternal alcohol, drug, or substance abuse. Factors Associated With Neurodevelopment or Its Measurement Socioeconomic status (see below); Maternal and paternal education; Birth order of child; Neurosensory impairment (especially in preterm infants); Neonatal illness (especially in preterm infants, e.g., cerebral hemorrhagic or ischemic lesions); and Conditions of neurodevelopmental assessment (e.g., identity of assessor, assessment in an office or the child's home). Socioeconomic Status of the Family Socioeconomic status of the child's family, usually based on parental occupation(s) or family income, is important information and is associated with several outcomes of nutrition trials, including neurodevelopment (13,14). Coding systems for socioeconomic status are generally based on occupation because this captures differences in lifestyle and attitudes that are independent of income. Furthermore, parents are likely to find questions about occupation less intrusive than questions about income. The system for coding parental occupation differs between countries and even between investigators within countries, and there is no consensus on how to record this information in multinational collaborative trials. The following coding system, for developed countries, is suggested for discussion. Coding should be in accordance with the principles of the classification, and in some cases a degree of subjective judgment will be necessary to code specific occupations. It is recommended that coding is based on the consensus of two or more experienced investigators and that original information on occupation is retained for future reference. Professional occupations: Require a university degree or appropriate higher qualification, e.g., lawyers, health professionals, management consultants, scientists (including social or behavioral), professional engineers, architects, surveyors, educators, and technologists. Intermediate occupations: Nonmanual occupations not always requiring qualifications of university degree standard, e.g., performing or visual arts, technicians, estimators, appraisers, inspectors, writers, and therapists. Skilled occupations: Nonmanual or manual tasks applying a body of trade- or industry-specific technical knowledge, e.g., retailers, dealers, proprietors and managers of accommodation, sports persons, entertainers, skilled clerks or secretaries, and all skilled craft and trades persons. Partly skilled occupations: Tasks requiring limited skills or knowledge, including persons working in many manufacturing and service industries. Unskilled occupations: e.g., cleaners, porters, laborers. Not classified: e.g., people who are unemployed or unable to work because of age or disability. Even when one has decided on a common classification, there may be difficulties. For example, if a young mother is living with and supported financially by her parents, then the coding could be based on her occupation or income, or on that of the child's father or grandparents. If both parents are full-time paid employees, different researchers adopt different approaches to coding. The options are 1) code on the highest earner; 2) code on the child's father; or 3) calculate the mean code for the two parents. Discussion on these points is welcomed. CONCLUSIONS We believe it is important that, when possible, randomized nutrition trials of infants are conducted in a way that facilitates comparison with other studies and meta-analysis. We suggest that: All studies of infants should collect and report on the core data outlined in Table 2. Studies of nutrition in preterm infants should additionally collect and report on data outlined in Table 3. Outcome measures should, when possible, include those that are widely believed to be important, although additional or more detailed measures will be needed in many studies. One should work toward agreement on factors that need to be recorded and reported for each outcome measure. We provide examples for growth and neurodevelopment. There is need for agreement on a common coding of socioeconomic status, even if local measures are also reported. The proposed core data set may deserve future modifications, taking into account scientific developments and the practicalities of performing and reporting clinical studies. Therefore, we encourage the submission of comments and suggestions for enhancement of this core data set by e-mail ([email protected]) for publication on the ESPGHAN web site (http://www.espghan.org). Acknowledgments: The authors thank the participants of the Scientific Workshop on Characterisation of Infant Food Modifications in the European Union held June 2002 representing academia, the infant food industry, consumer organizations, the Commission of the European Union, and national regulatory bodies (19) for reviewing the suggested set of core data presented here, and for providing valuable suggestions that have been incorporated. TABLETABLETABLE 4: Necrotizing enterocolitis (NEC). Coding based on severity of illness, according Bell staging criteria 15 and modifications of Kliegman and Walsh 16TABLE 5: Cerebral lesions