METHODS

Study area and subjects

The study was done in Lungwena, Mangochi District, in southern Malawi. A government health centre served an approximately 100 km² rural area with some 17 000 people in 23 villages. Most of the inhabitants were Muslims of the Yao tribe. The literacy rate was very low and subsistence farming (mainly for maize) and fishing formed the main occupation. The cultivated land areas were often small and many farmers had no alternative sources of income, leading to poor food security in the area. The climate was monomodal and the staple food, maize, was harvested in March to April.

The study participants were 767 live born babies from a community based cohort of 795 pregnant women. Because of a very high enrolment rate, the study cohort comprised approximately 95% of all newborn children in the area during the time of enrolment. Details of recruitment, collection of background data, and follow up have been described previously.¹⁴ The research protocol was reviewed and approved by the Malawi National Health Science Research Committee and informed consent was obtained verbally from each pregnant woman before enrolment.

Anthropometric measurements

Eight trained research assistants made home visits to collect anthropometric measurements at monthly intervals from birth up to the age of 18 months and at three month intervals thereafter. Weight was measured in light clothes with a spring scale (reading increments 10 g) until the age of 12 months and thereafter with a battery operated digital bathroom scale (100 g increments). Length/height was measured with locally constructed length/height boards having reading increments of 5 mm. Children were measured supine until they were able to stand, and erect thereafter. The scales and length/height boards were checked and calibrated regularly with internal standards. To verify data accuracy, a random sample of 5–10% of the measurements was taken twice by two independent research assistants. The same research assistants carried out all the anthropometric measurements throughout the study.

Data management

Data entry and analysis were done with Microsoft Excel 7.0 and SPSS 10.1 computer programs, respectively. The comparisons of growth were made using both the recently published growth reference from the United States CDC,¹² and the traditional NCHS reference, promoted by the WHO (NCHS/WHO).³

Construction of growth curves

The construction of the attained size growth curves was done using the LMS method described by Cole.⁵ In brief, the data were treated cross sectionally in terms of computing the mean and standard deviation for each anthropometric measurement at each age group. One month age groups were used from birth to 18 months of age, and three month age groups thereafter. All visits done within three days (up to 18 months of age) and seven days (up to 36 months of age) of the expected date of examination were included in the analysis (age grouping). The distributions of weight and height were examined for normality, before and after transformation, by calculating skewness, kurtosis, and checking normality plots for the distribution at each age group. Data points suggesting a significant deviation from examination of individual growth trend line were considered questionable and discarded from the analysis. For weight, deviations over 2 kg or 15% of the interpolated value on the individual growth trend line were excluded, whereas height values over 4 cm from the individual growth trend line were similarly excluded. The estimated centile lines were plotted and smoothed using the LMS computer program.¹⁶

RESULTS

Subjects

Table 1 presents the background characteristics of the participants and their mothers. Of the 767 live born babies who started the follow up, 128 had died and 62 absconded by 36 months of age. Measurements collected before their death or losses to follow up were, however, included in the analysis. For all participating children, the median (range) follow up time was 27 (1–36) months.

In total, anthropometric data were collected at 13 728 independent home visits, of which 961 (7.0%) were excluded from analysis because of age grouping and 38 (0.3%) home visits were excluded because of questionable data quality. After birth, the mean (range) monthly number of measurements included in the analysis was 532 (504–562). Because of the large number of home deliveries, birth measurements (within 48 hours of birth) were available from only 197 children (26 %).

Growth in Lungwena in comparison to the reference populations

Figure 1 illustrates the growth pattern of the Lungwena children compared to the 2000 CDC growth reference. At birth, newborns in Lungwena were on average 510 g lighter and 2.5 cm shorter (table 2) than the reference babies. During the first three months, the median weight in Lungwena increased rapidly and approached the reference median, whereas the median height faltered towards the 3rd centile of the reference population. Between 3 and 12 months of age, both the median weight and the median height in Lungwena deviated significantly downwards from the reference. After 12 months, the median weight in Lungwena consistently paralleled the 3rd weight centile of the reference population. For height, the Lungwena curves showed persistent, though slight, deviation from the reference even during the second and third year of life. At 36 months of age, Lungwena children were approximately 2.3 kg lighter and 10.5 cm shorter than the reference children.

Figure 2 shows the comparison of Lungwena children to the NCHS/WHO reference. As shown, the general findings on weight development were comparable, regardless of the reference used. The only difference was in height development after 12 months, when the Lungwena children deviated slightly less from the NCHS/WHO than from the 2000 CDC reference.

Although not shown separately, boys in Lungwena were slightly heavier and longer than girls throughout the first three years of life.

DISCUSSION

Several factors have been associated with poor childhood growth in low income countries. Of these, the most commonly identified explanatory variables include suboptimal weaning from breast feeding, enteral and other infections, and inadequate dietary intakes.⁴ The first two—complementary feeding and a concomitant increase in the incidence of enteral infections—have received special attention, because their occurrence presumably coincides in time with the typical age of onset for growth faltering. However, strong conclusions have been hampered by the fact that the commonly used NCHS/WHO growth reference is based on mostly formula fed infants, who are known to grow differently from their exclusively breast fed counterparts.¹⁰ Therefore, the age of onset of true growth retardation in low income countries has remained debatable.¹⁷

In this community based cohort of rural Malawian children using both the new CDC as well as the traditional NCHS/WHO reference, babies appeared short at birth and linear growth faltering continued throughout the first three years of life. Weight faltering, on the other hand, was largely restricted to the period between 3 and 12 months. During the first three months, babies in Malawi actually gained more weight than the reference children, and after 12 months the populations were comparable in weight increments. The initial weight spurt in Lungwena could be explained by the fact that, although not exclusively breast fed by the WHO definition, the babies were predominantly breast fed during the first three months of life. At this age, exclusively breast fed infants are known to gain more weight than formula fed babies.¹¹

Besides the known growth retarding effects of common infections and inappropriate complementary feeding,⁴ there are at least three potential explanations to the small birth lengths and poor linear growth in Lungwena. First, the prevalence of preterm deliveries was very high (22%), resulting in smaller than expected newborns.¹⁸ Additionally, because of the frequency of malaria and the low maternal intakes of energy during pregnancy,¹⁹ many babies may have suffered from intrauterine growth retardation. The fact that lengths were more compromised than birth weights is consistent with the hypothesis that fetal growth was affected before the third trimester. In such conditions, not only lipid accumulation but also cellular synthesis is compromised, leading to short newborns who do not express much catch up growth postnatally.²⁰

A third possible explanation for the linear growth faltering is perinatal HIV infection, which was a widespread problem in the current study area (18% seroprevalence among the pregnant women). Previous studies from Malawi and elsewhere have shown that babies born to HIV positive women are lighter than those whose mothers are not infected, and they more often exhibit faltering of both weight gain and linear growth during infancy.^21,22 Undoubtedly, there was also a familial component in the length gain of children in Lungwena. Mean adult female height in this population was only 155 cm. Male adult heights were not available, but in a comparable Malawian population the mean was 162 cm.²³ Thus, a part of the apparent childhood growth retardation could have been a “physiological” shift across centiles to follow a predetermined familial path. However, this trend is unlikely to be genetic in origin, since under 5 year old children from more affluent families in Malawi and elsewhere in low income countries grow close to the NCHS/WHO reference.^3,24

Until now, the community prevention and management of childhood malnutrition has largely focused on weight faltering and the period between 6 and 24 months after birth.²⁵ Relatively little attention has been put on height gain faltering, although in many countries, especially in sub-Saharan Africa, stunting is much more prevalent than wasting.⁵ Although not often directly stated, inherent in this approach is the assumption that weight and height faltering have similar aetiologies and that the same interventions may correct both problems. In the light of the present results and those of others,^8,9 it is not surprising that few interventions have been successful in correcting or preventing linear growth faltering. If linear growth is slowed down at birth (as in the present study) and remains so throughout the first three years of life, interventions focusing on a narrow age range do not address the whole problem. In areas where stunting is a major problem, intervention programmes targeting the infancy period coupled with early enough antenatal period interventions may thus prove to be more successful.

One issue needs to be considered when interpreting the current results. Because of the high number of home deliveries, anthropometric data at birth were available only from every fourth newborn. Theoretically, this might have biased our results on low birth lengths. In practice, this was not a problem since the difference to the reference populations remained identical at 1 month of age, when length data were available from almost every study subject. Some studies have, however, not documented low birth lengths compared to the international reference standard.⁹ Thus, some caution is required when applying the current results to other conditions. However, because of the prospective community based design and high enrolment with very low dropout rate, the conclusion that weight and height faltering started at different periods is likely to be a true reflection of the situation in the study area. Similar results have been observed before in some other low income countries.^8,9

Taken together, we have documented different periods of greatest risk for weight and height faltering among a cohort of rural Malawian children. Compared to two separate reference populations, weight gain faltering was generally restricted to the ages between 3 and 12 months of age, whereas height faltering occurred from birth and continued throughout the first three years of life. Because both weight and height faltering have important health consequences for growing children, both forms of growth failure need to be considered in a malnutrition prevention and management strategy. Traditional programmes encouraging breast feeding, correct weaning practices, and prevention of enteric infections should be complemented with appropriate antenatal and postnatal interventions against stunting. Further studies are however needed to clarify the details of such interventions.