Table 1 Stepwise textual narrative synthesis

Step 1: Study grouping. Studies belonging to each of the sub-groups were identified. For example studies classified by relationship between participant and infant were:

a) Mothers;16 studies [20,28–41]

b) other family members;1 study [20]

c) health professional; 2 studies [28,42]

d) unrelated others; 2 studies [43–45]

There was overlap between sub-groups. For example a study of mothers 2 months after their infants were admitted to NICU would fall within 3 groups, determined by the 'participants being mothers', the 'age of the infant's and the fact that the infants were considered 'high risk'.

Step 2: Study commentaries produced. These commentaries summarised key aspects of the studies in relation to the sub-group within which they were included. For example a study of mothers' views:

A study by Baughcum and colleagues[28] reported on focus groups conducted with 14 mothers attending WIC clinics (Special Supplemental Nutrition Program for Women, Infants and Children) in USA with infants aged 12–36 months. The study focussed on maternal attitudes to feeding and proposed an association between these and overweight in their babies. The study design was judged adequate, although the bias introduced by sampling from WIC clinics was not discussed by study authors. Authors concluded that mothers are more concerned about under- than overweight; two supporting quotations stated that weight gain is always good, because it means children are eating.

Step 3: Sub-group synthesis produced. For example the views of mothers:

Most of the studies in this review (16/19) explored the views of mothers. The mothers, varied in terms of the age of their infants, the present and past health status of their children, their country of residence, their country of origin, income level, socio-economic status (SES), and number of children. North American Caucasians made up most of the sample. Background data for participants was often unknown, unreported or incomplete. Sampling strategies in the studies created difficulties in interpreting findings. For example, three studies explicitly set out to sample low income groups[28,29,37] using WIC clinics to achieve this. To be recruited to these studies, families needed to have a low income, but also needed to register for the WIC programme and attend clinics. This strategy is likely to selectively recruit participants[29]. Studies typically did not allow comparison between groups (for example those from different ethnic backgrounds) because findings were not reported separately.

Growth and size were concerns for mothers, particularly achieving average or normal growth. Mothers used a variety of sources of information to define norms, including growth charts, clothing and familial patterns of growth/size. There was evidence of concern for underweight, but the extent of concern about overweight was unclear.