Contents - Previous - Next

3. Factors determining early and late faltering

3.1. Differences at birth

3.1.1. Perinatal complications It has been proposed by a number of workers (e.g. Mitchell, Gorrell & Greenburg, 1980) that full-term infants who subsequently fail to thrive have an excess of perinatal complications. Table I shows examples of the key variables that were analysed. There were few differences between the groups, and the overall mean scores were similar. The main contribution to the high proportion with some degree of fetal distress was a transient bradycardia (< 120) at some point during delivery.

3.1.2. Anthropometry at birth Virtually all subjects had been born in a teaching hospital where it was a routine procedure to record birth weight, length and occipito-frontal head circumference. Data were available on all subjects for birthweight, on 83% for length' and 96% for head circumference at birth. It was possible to standardise birthweights for maternal stature, mid-pregnancy weight, gestation, ordinal position of the child (first or other born) and gender according to the method of Tanner & Thomson (1970). These corrected birthweights have been converted into standard deviation scores. A ponderal index was computed according to the method of Miller & Hassanein (1971), based on the formula: [birthweight in g/(crown-heel length cm)³] x 100. Values were then corrected for gestational age. Birth length standards from Kitchen et al. (1981) and head circumference from Yudkin et al. (1987) were also corrected for gender and gestational age and both were converted into standard scores.

Table 1. Examples of individual risk factors used to compute total antenatal and perinatal risk score for mothers of 'early' and 'late' growth faltering subjects

* There are no significant differences between the subgroups on any of the above variables.

The possibility of differences between the groups in the mean values of these three variables was investigated by one-way ANOVA, with gestation as a control variable where appropriate. No significant differences were found.

However, a different picture emerged when we came to compare the groups in terms of standardised birthweights (corrected according to the method of Tanner & Thomson (1970)). The mean values for the early and late FTT children were -0.42 ±0.78 and -1.05 ±0.5 (P = 0.002). A test for the non-normality of both distributions did not reach conventional levels of significance. No child had a standardised birthweight which was more than 2 SD below the population mean (Tanner & Thomson, 1970).

3.1.3. Socioeconomic status A wide range of variables measuring aspects of families' socioeconomic status was obtained based upon the factors used by Osborn (1987) to compute his composite social index (Osborn & Morris, 1979).

There were no differences between early and late FTT in the father's occupation and educational status; in the proportion of single parent households; in the type of accommodation or number of people per room; in parental height and weight (as far as the evidence went), or in distribution between different ethnic groups.

However, there was a highly significant difference in responses to the question "How many times in the past year have you had no money in the house?". Despite similar levels of mean weekly income, 64% of families with late growth faltering infants said this had happened to them at least once, whereas this was true for only 27% of 'early FTT' families. Similar proportions of families in both groups (82% and 83% respectively) said their first spending priority was food.

3.1.4. Mothers' cognitive abilities It is unusual for studies that are examining patterns of growth faltering in infancy to have reliable measures of the cognitive abilities of the mothers. All mothers of children in this investigation were tested with the Wechsler Scales of Adult Intelligence (WAIS-R, Wechsler, 1974). All the measures were made by a senior psychologist who was blind to case-comparison group status. Proration of the scores on block design and picture completion yielded a nonverbal performance score; vocabulary and information a verbal score. Proration is a procedure by which a full scale IQ score can be approximated by computation from a subset of all the subscales which would normally be administered, thereby reducing the time necessary to give the test. Where the native language of the mother was not English, and she was not fluent in the language, no verbal score was computed. Substantial differences were found between the early and late growth faltering infants' mothers, with those in the latter subgroup having lower IQs by a margin that amounted to about 0.75 SD (Table 2).

3.1.5. Mothers' mental state In order to investigate the mental state of the mothers they were asked to fill in a variety of questionnaires. Where the mother's first language was not English, a professional translator assisted her. All instruments were rated during the course of the home visit and were taken away for subsequent scoring. They included a 28-item version of the General Health Questionnaire (GHQ) (Goldberg & Hillier, 1979) which is a scaled version of the parent instrument, with four subscales derived by factor analysis. These include 'somatic symptoms', 'anxiety and insomnia', 'social dysfunction' and 'severe depression'. The validity of the subscales is discussed in Goldberg & Williams (1988, p. 39). These subscales represent dimensions of symptomatology, more symptoms resulting in a higher score, and do not necessarily correspond to any psychiatric diagnosis. There is no absolute consensus about what 'threshold' score best discriminates between cases of probable psychiatric disorder and others, but the modal value has been 4/5 out of a total possible score of 28 (Goldberg & Williams, 1988, p. 64). At this level, the positive predictive value of a high score (i.e. 5 or more), where there is a prevalence of 30% of 'true' disorders in the population being tested, is 0.67.

Table 2. The cognitive abilities of mothers of 'early' and 'late' growth faltering subjects, measured by the Wechsler Adult Intelligence Scales (WAIS)

Both the overall scores on the GHQ -28 and the scores on certain of the subscales distinguished between the mental state of the mothers of early and late growth faltering infants. The overall mean scores of the late growth faltering mothers were significantly higher than those in the early growth faltering group (5.8 ±5.6 and 2.9 ±3.9 respectively; P = 0.039). Taking the cut-off point of 4/5 - as recommended - the proportions of mothers who were 'probable cases' of psychiatric disorder (i.e. had scores of ³ 5) in the two groups differed substantially (56% and 27% respectively), but the magnitude of differences did not reach a conventional level of statistical significance, once a continuity correction had been applied (P = 0.09).

On an alternative measure of anxiety (the Multiple Affect Adjective Checklist; Zuckerman & Lubin, 1965) the mothers of later growth faltering children obtained higher scores (3.8 ±3.2 vs 1.6 ±1.7; P = 0.007). They were also more disturbed in terms of social dysfunction (a GHQ -28 subscale that asks about everyday activities) (mean scores 7.8 ±2.9 and 5.8 ±2.8; P = 0.02). However, their most striking form of psychiatric disturbance was on the GHQ -28 subscale for severe depression (2.7 ±2.5; 0.9 ±1.2; P = 0.003).

The broader question of how our findings for the mental health of these mothers compares with other inner city surveys is hard to answer, for none of the mental health scales we used has published norms for a general population sample in this country which is exactly comparable to the subjects of our survey. Perhaps the closest set of comparison data is from Brooke et al. (1989), who studied a consecutive series of 1860 white women booking for delivery at a district general hospital in inner London. The 28-item version of the General Health Questionnaire, administered at 17 weeks gestation, gave the following results (see Table 3). It is clear that the mothers of early growth faltering infants were comparable in their depression scores to those in this pregnant sample, but the scores of the mothers of late growth faltering infants were rather higher, implying more of them were suffering from some depressive symptoms.

Table 3. The mental state (severe depression) of mothers of 'early' and 'late' growth faltering subjects, compared with population sample

* Pregnant women, 17 weeks gestation, inner London population.
^a 'Severe depression' subscale of GHQ (Goldberg & Hillier, 1979).

3.1.6. Mothers' social support It has been suggested that mothers of failing to thrive infants are exceptionally socially isolated (e.g. Kotelchuck, 1980) and that they lack social support (Bithoney & Rathbun, 1983; Kotelchuck & Newberger, 1983). This issue was addressed with the aid of a scale devised specifically for the purpose of our own investigation, based upon the work of Tietjen & Bradley (1985). The scale was filled in by the mother, with guidance from the research worker if necessary. Typical questions ask about whether there is anybody to whom the mother could turn in an emergency, or if she had a personal problem, how often there is no-one willing to talk to her, how satisfied she is with her neighbourhood, and so on.

Significant differences were found, with the mothers of the early failing to thrive babies having a higher social support score, on average, than mothers in the late failure to thrive group (31.0 ±5.1 vs 25.7 ±5.4; P = 0.001).

3.1. 7. Marital relationship Disharmony in the intimate relationships of the mothers we surveyed may have contributed to parenting difficulties. Whilst this is true for parenting in general (Rutter, 1985), it has been argued specifically that the quality of a father's relationship with his spouse and his family may have an important longer term influence upon the mental health outcome of failing to thrive children (e.g. Drotar, 1985).

Mothers' relationship with their spouses were assessed with the Dyadic Adjustment Scale (Spanier, 1976; Spanier & Thompson, 1982; Antill & Cotton, 1982). Only the Dyadic Satisfaction subscale was rated. There was only a small and non-significant difference between the reports of mothers of early growth faltering infants (mean 39.2 ±5.5) and others (mean 35.7 ±8.8). Both scores were similar to the values for a wider sample of women (mean 38.7 ±6.1) given by Antill & Cotton (1982).

3.1.8. Familial patterns of growth failure The distinct psychosocial characteristics of families whose children had early or late growth faltering led to the hypothesis that similar patterns of growth would tend to be found in the siblings of affected individuals; if the mechanisms producing the growth trajectories were different in the groups there might be a tendency for affected siblings to have similar profiles of growth faltering in the first year. Accordingly, a search was made for data on sibling growth patterns in infancy. All information came from child health or 'well baby' clinics, not from paediatricians or other hospital records.

In the group with early failure to thrive just 35% of siblings could be said with certainty to have grown normally in the first year. Thirty percent had a similar pattern of 'early' growth failure but this was not necessarily sustained for as long a period; however, these siblings did have weights for age which fell clearly below the 3rd population centile at some point in the first six postnatal months. Just one of the 20 siblings had growth faltering after six months. An additional 30% had some degree of growth faltering at some point in the first year; in a few cases this was severe but the onset was uncertain. In others it was not sufficiently severe (i.e. weight for age fell to below the 10th but not the 3rd population centile) to meet diagnostic criteria.

In the group with late growth faltering 57% of the siblings grew normally through the first year. Surprisingly just 10% had a pattern of failure to thrive which commenced in the latter part of that year, and in 20% the onset was 'early', in the first six months. There were an additional 14% of siblings whose degree of growth faltering did not quite meet criteria, or for whom the onset was uncertain.

Cross tabulating growth trajectory by the timing of the growth failure onset among siblings revealed no significant differences between the groups. However, it is important to note that overall 54% of the siblings of our subjects had had some degree of growth faltering within the first postnatal year, and in 32% it was relatively severe although not necessarily sustained for as long a period as in the probands.

3.2. Differences during the first year

3.2.1. Medical history Children with serious medical illness, which might have caused growth faltering, were excluded from the samples of infants discussed here. There were reports of only relatively minor illnesses during the first year of life, such as episodes of upper respiratory tract infection, diarrhoea and otitis media. A weighted score was computed for 17 such variables, which took a maximum value of 25. Neither individual indices of risk nor the total score distinguished the early from the late growth falterers (the mean value being 8.5 ±3.7 and 8.0 ±3.4, respectively), suggesting that the frequency of illness was not a relevant factor in causing their disorder. However, because the data were collected retrospectively it was not possible to specify the timing of most of their illnesses within the first year; the data would have been too unreliable.

3.2.2. Feeding histories Histories of how the children were fed as infants were obtained from mothers by interview at follow-up, when the infants were about 15 months of age. The information was therefore retrospective and for that reason cannot be regarded as wholly reliable. Twenty percent of the 'early FTT' group and 13.6% of the 'late FTT' group were exclusively breast fed. Twenty-four percent of the 'early' and 32% of the 'late' group were exclusively bottle fed; the corresponding figure for those given a mixture of breast and bottle was 56% and 54.5%. None of these differences reached statistical significance. The mean age at which solids were introduced was virtually identical in the 'early' and 'late' groups too, being 17 + 6.7 and 17 + 6 weeks respectively. There did not seem to be any major feature of their early feeding histories which distinguished the groups, although it might have been expected that the 'later' FTT group began to show growth faltering around the time of weaning, possibly because of difficulties in getting the children to accept solid foods. We did attempt to 'map onto' standardised charts of the children's growth trajectories the times at which weaning occurred for individual subjects, according to mothers' reports, but the findings were inconclusive.

A substantial proportion of the infants in both groups were reported as having slept through feeds at some time in the first year. That is to say, when it was time for the infant to have a meal the child was asleep, or alternatively fell asleep during a feed. This was the case for 68% of infants with early growth faltering and 64% of the infants with growth faltering after six months. Although we attempted to discover whether there were systematic differences between the groups according to the timing and the duration of this behaviour, no differences could be detected.

It is, of course, of great interest to know how mothers responded to this pattern of behaviour. Of those who slept through feeds at some time, in the 'early FTT' group 40% were woken by their mother, according to their retrospective reports. In the 'late FTT' group 62% were woken by their mother. Although these differences are relatively small, they are at least suggestive that perhaps infants who were sleeping through feeds at a time when their growth rate was at its most rapid, and who were not yet weaned, were being chronically undernourished because of missed meals. The interview material certainly raised the suspicion that overnight feeds were frequently missed in the early days if the infant was a 'good' baby and slept through the night without demanding to be fed.

However, similar proportions of infants in both groups were said to sleep through the night before 6 weeks of age, 'early' 24%; 'late' 25%.

3.2.3. Growth of sleepy infants If these mothers' reports of their babies lack of demandingness were correct, and they were indeed missing feeds because they were asleep, one would expect this to be reflected in their growth trajectories-although it could not of course explain the difference in the growth trajectories of the 'early' and 'late' groups. Accordingly, the groups were combined and then categorised according to whether their mothers reported them as having ever slept through feeds, or not. Remarkably, the weight gain trajectories of those who were in the 'sleepy baby' group fell below the remainder, and a significant discrepancy persisted from 4 weeks (the earliest standardised measurement) right up until 12 months of age (Fig. 4).

Their birthweights were not significantly different from one another, whether measured in absolute terms (sleepy infants mean 2938 ±344, others mean 3150 ±380) or in standard deviation scores (Tanner & Thomson, 1970). Nor were their ponderal indices or occipito-frontal circumferences, but the sleepy infants were shorter (mean 49.6 cm SD 12.5 vs mean 51.3 cm SD 2.7; P = 0.06). Where the birth length was standardised for gestation and gender a much greater magnitude of differences emerged (-0.31 ±1.1 vs 0.66 ±1.1; P = 0.01). It is worth noting that, despite a difference in standardised length of about 1 SD, both mean values are well within the normal range.

In order to explore this matter further, a series of analyses of variance was undertaken with weights (standardised) at 4 weeks, 6 weeks, 3 months, 6 months, 9 months and 12 months as the dependent variables, and a set of predictor variables which included whether the child had ever slept through feeds, whether mother reported that she regularly woke the baby for feeds or not, whether the child was breast, bottle or combination fed, and the gender and ordinal position of the child. Other potential explanatory variables included mother's IQ, various measures of her mental state and the summary score for the child's oral motor difficulties. Potential interactions between these variables were investigated but none of any significance was found. A series of multiple regression equations was then calculated for each of the above mentioned ages, in which only those predictor variables that had been found to be of relevance in the former exploratory analyses were entered. The results are shown in Table 4.

Differences between groups at 4 weeks 3 months 6 months 12 months c 0.01 and at 6 weeks 9 months < 0.02. No significant differences at birth or 15 months

As can be seen, the variable 'slept through feeds' was significant at every age, and children who did so had lower weights for age. Interestingly, the variable 'mother woke baby' had no significant effect and was not included in the final models. Neither did mother's mental state, even when this was entered as a variable relating to the duration of postnatal depression. The ordinal position of the child had no significant influence once the other variables had been entered. Mother's IQ had a substantial impact on the child's weight at 12 months only, children of mothers with higher IQs having a greater weight for age, presumably reflecting their better feeding practices.

3.3. Differences at outcome

3.3.1. Anthropometry The children were examined as close to 15 months of age as possible. A variety of anthropometric indices are presented in Table 5. The computation of the anthropometric indices for arm muscle and fat areas are from Sann et al. (1980). Norms for mid and upper arm circumference are available from Rees et al. (1987). There are no significant differences in any of the indices, whether the groups are compared on the basis of mean values or on the basis of their centile distributions, suggesting that by the age of 15 months or so, despite contrasting trajectories of early growth, on average their anthropometric outcome is identical.

Mean values of weight for length are given in Table 5. All infants, in both groups, were to some degree underweight for their length. Nearly half (45%) of those with early growth faltering were at or below the 3rd population centile; the figure was similar for the late growth faltering infants (40%).

Case criteria demanded a substantial fall in weight, relative to population norms, over the first postnatal year. Overall, the mean decrement in standardised weight for age was -1.33 (±0.81). This was associated with an equivalent fall in length (-1.33 ±1.3), suggesting that the children were becoming stunted. However, taking 90% of NCHS median lengths as the cutoff point for a child to be classified as stunted, at the time of follow-up only 13% of our sample met this criterion. We do not have serial measures of length through the first postnatal year as these are not routinely measured in child health clinics, so it has not been possible to look at changes in length velocity in any meaningful way. Head circumference was below norms for gender and gestation at birth for 89% of the sample. There was a small degree of catch-up in this variable for two out of five subjects, such that at follow-up 28% had values at or above the 50th population centile.

Table 4. Multiple regression of postnatal standardised weights on explanatory variables

Table 5. Anthropometry of 'Early' and 'Late' growth faltering infants

Values are given as mean Z scores ±1 standard deviation. There are no significant differences between groups on any variables, except where indicated.

3.3.2. Blood tests A 1 ml (minimum) sample of capillary blood was taken at 15 months from all subjects for a full blood count, haemoglobin electrophoresis and ferritin estimation. Analyses were also made for red blood cell folate and serum B₁₂ The mean haemoglobin levels of late faltering subjects were significantly lower than those of early faltering subjects (11.46 ±1.13; 12.11 ±0.92; P = 0.03). Taking the WHO (1972) recommended cut-off of 11 g/dl between 6 months and 6 years as an indicator of anaemia, we found 14% of the early FTT group and 36% of the later FTT group had haemoglobin at or below this level, but the difference failed to reach statistical significance. Both mean corpuscular haemoglobin (26.0 ±1.8; 24.8 ±2.4) and mean corpuscular volume (80.1 ±3.2; 76.9 ±6.4) were significantly lower in the 'late' failure to thrive children (P < 0.05) suggesting that the anaemia was for the most part a hypochromic microcytic one and that iron deficiency was most likely to have been responsible. We had screened for possible inherited anaemias such as thalassaemia or sickle cell anaemia but none of the children was found to be suffering from these conditions. Serum ferritin levels were very similar in the two groups: the mean values for early FTT were 20.3 ±15.5 ng/ml; for late FTT 19.3 ±13.1 ng/ml and the proportions with values < 7 ng/ml (Slimes, Addiego & Dallman, 1974) were 14% and 24% respectively. Values of serum B₁₂ and red cell folate were all within the normal range for infancy.

3.3.3. Neurodevelopmental attainments There is a risk that children with early severe growth faltering may suffer detriment to their neurological development. Studies of children who have been hospitalised with failure to thrive are reported to show a wide range of 'neurodevelopmental deficits', including low scores on standardised developmental assessments (Leonard, Rhymes & Solnit, 1966; Ramey et al., 1975, Field, 1984; Singer & Fagan III, 1984; Powell & Low, 1983). However, the relationship between early growth faltering and persistent developmental impairment is complicated by the fact that nutritional deprivation is usually accompanied by other forms of socioeconomic deprivation too (e.g. Casey, Bradley & Wortham, 1984) which can independently influence attainments.

Frank (1985) discussed the possibility that early malnutrition during a critical period of brain growth (Dobbing, 1990) may produce structural deficits which have functional consequences. There is evidence from studies in animals that the cerebellum, which contains large populations of microneurons, might be especially vulnerable to early undernutrition (e.g. Rodier, 1980). However, it is difficult to distinguish cause and effect here. Children born with subtle neurodevelopmental problems may themselves be at risk of undernourishment (for example, because of poor oral-motor skills; see Mathisen et al., 1989), and consequently of a suboptimal rate of growth. Subsequently, persistent malnutrition could of course independently contribute to neurodevelopmental disorders (see for example, Reyes et al., 1990).

The role of congenital determinants of behaviour and cognition is relatively easily demonstrated in clearly recognised medical syndromes such as the Down or Prader Willi syndromes. Nevertheless, a wide variety of minor physical stigmata is found within populations of children with no recognised medical disorder. Waldrop, Pederson & Bell (1968) have suggested that an inverse relationship exists between intellectual functioning and the number of anomalies in young children. Accordingly, physical anomalies that could be entered into the computation of a congenital anomalies score were sought in the course of physical examinations. The total possible anomaly score was 24. The mean scores for the two groups were very similar: early FTT children 1.8 ±1.8 and late FTT children 2.2 ±2.9. There was a highly significant correlation with head circumference at birth, standardised for gestation and gender (-0.42; P = 0.002).

An assessment of neurological functioning and maturity was also undertaken, based upon the work of Touwen (1976) and Amiel-Tison & Grenier (1986). A composite score of gross motor skills was derived from the ten variables for which there was the least missing data; this allowed the comparison of 18 cases of early FTT and 24 cases of later FTT: Skills such as the ability to walk or sit unsupported, visual following while sitting, optical placing, reaction of hands, were scored according to the weighting system recommended by Touwen (1976).

The total mean scores were very similar in respect of both gross motor skills (20.7 ±4.8 and 21.2 ±4.0) and fine motor skills (6.8 ±2.6 and 7.9 ±1.4) for early FTT and late FTT group respectively. A composite score comprising five variables that related to the appearance of the children-whether skin or nails were clean or dirty, whether burns or scars were visible, and whether extremities were red, puffy or shiny in appearance-did not distinguish between the groups either (0.81 ±1.1 and 0.88 ±1.1).

3.3.4. Cognitive abilities There is ample evidence that children in the developing world who are stunted, on account of chronic undernutrition and related adversities, have impaired mental abilities. This evidence has accumulated over many years (e.g. Hertzig, Birch & Richardson, 1972; Galler, 1987). Developmental undernutrition during early postnatal life may have greater consequences than were it to occur later. Dobbing (1990) has emphasised the importance of taking into account the duration, severity and above all the timing of any such insult when making extrapolation about the likely impact upon development.

Accordingly, it was of considerable interest to see whether the cognitive abilities of the early and late growth faltering infants were significantly different when measured early in the second postnatal year. The instrument used was the Bayley Scales of Mental Development (Bayley, 1969). The mental scale contains language items and many types of problem solving tasks. The psychomotor scale mainly addresses gross motor development. The tests were administered blind to the status of the infants. The results of an ANOVA, with both the psychomotor (PDI) and mental development index (MDI) of the Bayley Scales as dependent variables, and with mother's IQ as a covariate, are shown in Tables 6 and 7.

Table 6. Analysis of variance for effects of postnatal weight gain trajectory on mental development index

Multiple classification analysis

The data show that the growth trajectory did indeed have a substantial and highly significant influence upon the cognitive and psychomotor abilities of the infants at 15 months, with the early growth faltering children doing much worse than those whose faltering began after 3-6 months of age. Mother's IQ does have a significant influence, but upon the MDI only. Both MDI and PDI are depressed to an equivalent extent by the early growth faltering.

Further analyses of these data have shown that, even when other potentially confounding psychosocial variables, such as the cognitive stimulation the child receives at home, are taken into account a statistical model can be constructed that enables the timing, duration and severity of growth faltering to be used as predictors of mental functioning. Up to 37% of the variance in cognitive and psychomotor outcome at 15 months can be explained by the model. The first six postnatal months appear to constitute a 'sensitive period' for growth and mental development (Skuse et al., 1993).

Table 7. Analysis of variance for effects of postnatal weight gain trajectory on psychomotor development index

Multiple classification analysis

Contents - Previous - Next