Weight faltering is a common paediatric problem that causes much parental anxiety. Recent population based studies show that infants whose weight gain is poor tend to be developmentally delayed.1^–3 They also have a substantially higher prevalence of feeding problems.3^–7 Their weight gain can be increased by a largely dietary intervention,8 but it has yet to be shown that their intellectual development can also be improved. Weight faltering in the first few months after birth is associated with particularly adverse intellectual effects,2 9 10 suggesting that early intervention could be particularly important.

Early intervention would depend on the early identification of slow weight gain, which would be most practical using weights collected at birth and at the 6–8-week check, as these are already routinely available in the UK for almost all infants. Sensitive measures are needed to identify slow weight gain over this period, which take into account the infant’s birth weight and exact age. The conditional weight gain or “thrive index” methods are suitable for this purpose.11 12 A thrive index is a z score for weight gain that is conditional on the infant’s initial weight (birth weight in this context) and so provides a measure of relative weight gain that is not confounded with birth weight.

The aim of this study was to systematically identify infants whose weight gain was slow from birth to the 6–8-week check so that their characteristics could be compared with those of control families. All infants whose weight gain was below the fifth centile from birth to the 6–8-week check were identified in a 2-year birth cohort, using data already recorded in routine clinical practice. Early feeding problems in the infants, their social circumstances and their subsequent weight gain and behavioural development were investigated.

METHODS

The study was conducted in 18 family practices in the Easington area in North East England from 1 April 2001 to 31 March 2003. Approximately 2000 births were anticipated over this period. Screening of 2000 births was expected to give about 100 cases. A group of 68 cases and an equal number of controls gives 80% power to detect a difference of 0.5 SD on the Bayley Scales,13 allowing a reasonable margin for loss to follow-up.

Each practice was visited monthly and new births were recorded from birth registers kept by the health visitors. The infant’s sex, birth weight and gestational age at birth, and their weights at the 6–8-week check and at 9 months were extracted from health visitors’ records as they became available. All 65 balances used for weighing infants in clinics and maternity units were calibrated before the study started. Two were seriously inaccurate, weighing light by an average of 259 g and 623 g, and were corrected.

Data were entered directly onto an ACCESS database on a laptop computer. Error flags detected data values outside reasonable ranges, which were checked. Weights were converted to SD scores based on the 1995 British Growth Reference14 15 and the thrive index (z score for weight gain, conditional on birth weight, age and sex) was calculated using normative data derived from a large population based study.16 The index was calculated as in Wright et al11 except that the infant’s exact age was also taken into account. Infants below the fifth centile for weight gain were then flagged. When the later weights became available, infants below the fifth centile for weight gain from birth to 9 months were also flagged. Thus the numbers showing weight faltering over this longer period could be determined in the case sample, whose weight gain was below the fifth centile from birth to the 6–8-week check, and in the screened population as a whole.

The case sample comprised all infants with weight gain below the fifth centile over the first 6–8 weeks. Controls were the infant nearest in birth date to each case on the same health visitor’s list. Adjustment for covariates, when necessary, used regression methods. Mothers were invited to take part in the study by their own health visitor, and visited at home when the infant reached 4 months of age. They were given a structured questionnaire, focussing on family details, feeding and the infant’s health (see McDougall17 for the complete questionnaire). This was administered orally to ensure that literacy problems did not hinder any responses. There were no problems with understanding of English in this sample. The significance of associations was tested using χ² statistics (tests for a linear association with 1 df if three ordered response categories were used). Overall, n = 160 except where an occasional question was not answered (minimum n = 158), and for the question concerning the refusal of breast milk (asked only of still breastfed infants, n = 41). The infants were weighed and their development assessed blind to their group using the Bayley Scales. This was repeated when the infants reached 9 months. Comparison for these and other continuous variables were made using t statistics. As major health conditions can sometimes only be identified retrospectively, these were identified over the course of the first year in both cases and controls and taken into account in the analysis.

The project was approved by Durham, Hartlepool and Sunderland Local Research Ethics Committees.

RESULTS

Weight data and recruitment

A record of both a birth weight and a weight at 6–8 weeks was available for 1880 infants of the 1966 born (96%). At 9 months weights were available for 1631 infants.

Table 1 shows that weights were slightly below the average (zero) for the UK 1995 growth reference at birth and 6–8 weeks and slightly above it at 9 months, although the differences were small.

Figure 1 shows weight SD scores at birth and at the 6–8-week check with the slowly gaining infants identified. Across the whole range of birth weights, the graphs shows that infants identified as cases had lower weights at 6–8 weeks than controls of the same birth weight. Small overlaps between the distributions is due to slight variation in the age at which the 6–8-week weights were recorded, which was taken into account in identifying cases but cannot be shown on the graph. For any birth weight, the criterion is identifying the infants with the lowest 6–8-week weight, that is, it is identifying slow weight gain rather than a low attained weight. As expected, there was no correlation between the thrive index and birth weight (r = −0.025, p = 0.3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Weights at the 6–8-week check plotted against weight at birth, showing the “cases” identified as having weight gain below the fifth centile. Full circles indicate cases and empty circles controls. SDS, SD scores.

In total, 121 infants (61 females and 60 males) met the slow weight gain criterion from birth to the 6–8-week check. This is 6.4% of the cohort of 1880 screened infants. Among the 1729 infants born at term, 6.1% met the criterion. Infants born preterm, or from multiple pregnancies, were excluded, leaving 102 case infants (51 males and 51 females) for the follow-up. An equal number of controls were selected and both groups (n = 204) were invited to take part in the study. Of the 204, 42 infants were not available for study, mostly because the mother declined to take part. In all, 88 controls and 74 cases were recruited (79% of the 204). A smaller number (65% of the 204) were successfully followed up at 9 months. Two control infants were subsequently excluded. One, although recorded in notes as a male, was in fact a female, and one did not have a 6-week weight. Two cases and one control were identified with major syndromes (Down syndrome and myotonic dystrophy in cases; phenylketonuria in a control).

Case–control comparisons

Table 2 shows that there was no significant difference between case and control mothers in age, smoking or full-time education after 16. Using the Higher Education Statistics Agency (HESA) scale for 2001, the median educational attainment of the mothers in both groups was the same (HESA code 55, corresponding to GCSE O level). As shown in table 2, there were no significant differences in any of the economic indicators or in the heights of the mother or father.

Table 3 shows that there was also no significant difference in the proportion breast fed. However, the infants who met the criterion for failure to thrive were more frequently described as feeding slowly, as taking in small quantities of milk, with weak sucking, and as refusing breast milk more. The mean (SD) age at which solids were introduced was 15.4 (2.8) weeks in cases and 13.7 (2.8) weeks in controls. The difference is significant (t = 3.8 with 155 df, p = 0.001). The mothers were also asked the global question “Did you feel that you ever had difficulties feeding your baby” with possible answers “Yes, great difficulty”, “Yes, some of the time” or “No, not at all”. Of the case mothers, 36% reported having difficulties compared with 9% of control mothers (p<0.001).

In the whole cohort (excluding infants born preterm, or from multiple pregnancies), of the 1338 infants who did not meet the criterion for failure to thrive over the period from birth to 6 weeks, 38 (3%) met it over the longer period from birth to 9 months. Of the 96 infants who did, 24% met it over the longer period (relative risk 8; 95% CI 6 to 12). Table 4 shows some evidence for developmental delay in the cases at 4 months, and the difference remains after exclusion of cases with major syndromes. There is no clear evidence at 9 months.

There were small but significant correlations between birth weight and Mental Developmental Index (MDI) and Psychomotor Developmental Index (PDI) scores both at 4 months (r = 0.21, p = 0.008 and r = 0.19, p = 0.02, respectively) and at 9 months (r = 0.17, p = 0.06 and r = 0.21, p = 0.02, respectively).

DISCUSSION

We identified all infants in the cohort whose weight gain was below the fifth centile over the first 6–8 weeks. The proportion identified (6.1%) was slightly higher than the expected proportion of 5%. Figure 1 shows that the identification procedure worked effectively, identifying infants whose postnatal weight gain was poor irrespective of their birth weight. The lack of correlation between birth weight and weight gain after birth shows that there is no confounding between the two. Although both result in a lower than expected attained weight, the causes of slow weight gain before and after birth are quite different so it is important to distinguish between them, as we did here. Selection biases in referred cases is generally reflected in a worse outcome, for example greater intellectual impairment18 and a higher prevalence of feeding problems,19 but in this study there was no referral bias. A reasonably large number of early weight faltering children was studied, and detailed information collected on them at the time of weight faltering rather than retrospectively.

Slow weight gain over the first 6–8 weeks was not associated with the age, education or economic circumstances of the mother. This result may seem surprising, but it is consistent with results from other recent population based studies which have examined poor weight gain over longer periods in infants in the UK.4 16 20 21

Slow weight gain was, however, clearly associated with slow feeding, weak sucking, the consumption of small quantities of milk and the refusal of breast milk (but not with the mode of feeding). Previous population based studies of children whose weight gain was poor over longer periods of infancy have found similar associations.3^–5 10 21 Solids were introduced significantly later in the cases but still earlier than is currently recommended. The difference here may simply reflect a lower appetite in the cases, since the introduction of solids by a mother is partly a response to the infant’s own behaviour.22 23

There was also some evidence that slow weight gain over the first 2 months was associated with developmental delay. Large effects are not to be expected very early in life, but this is consistent with data from a much larger birth cohort using mothers’ reports9 and from an earlier but smaller case–control study using direct measurement.2 Recent evidence in a sample of over 5000 children found that intellectual ability at school age is strongly associated with weight gain in the first 2 months.6 Indeed, it was only slow weight gain in the first 2 months, and not later, that was associated with lower IQ at school age.

In this study, 24% of the cases identified at 6–8 weeks met criteria for failure to thrive over the period to 9 months, compared with 3% of the remainder of the cohort. If the procedure we used were to be conceptualised as a screening procedure at 6–8 weeks for failure to thrive as more commonly identified by prolonged weight faltering to 9 months, then this would imply a low sensitivity (0.19), a high specificity (0.95) and a reasonable positive predictive value. But that would be to assume that prolonged weight faltering provides the “gold standard”, while the evidence cited above shows that it is poor weight gain over the early months after birth, and not poor weight gain later, that is most strongly associated with developmental delay and with a lower IQ at school age. Thus, poor weight gain in the early months is a risk factor in its own right and not simply because of its association with later weight faltering. We now have appropriate means to identify infants who show this early weight faltering. Whether intervening with such infants would improve later outcomes is not clear and should be the focus of future research.