Comparative analysis of patterns of survival by season of birth in

IJE vol.33 no.1 © International Epidemiological Association 2004; all rights reserved.
International Journal of Epidemiology 2004;33:137–143
DOI: 10.1093/ije/dyh007
Comparative analysis of patterns of survival by
season of birth in rural Bangladeshi and
Gambian populations
Sophie E Moore,1 Anthony JC Fulford,1 P Kim Streatfield,2 Lars Åke Persson2,3 and Andrew M Prentice1
Accepted
29 July 2003
Background Analysis of data from rural Gambia has previously shown that being born
during the annual hungry season strongly influences susceptibility to
mortality from infectious disease in young adulthood, possibly through an
influence on immune function. In rural Bangladesh pregnancies are exposed to
similar seasonality. The current paper uses data from a large demographic survey
in the Matlab region of Bangladesh to retest the Gambian-derived hypothesis that
early life exposures correlated with season of birth predict later patterns of
mortality.
Methods
Since 1966, a continuous demographic surveillance system has been in operation
in the rural Matlab region of Bangladesh. The current analysis is based on
172 228 births and 24 697 deaths between 1974 and 2000. Season of birth was
defined as ‘harvest’ (January–June) and ‘hungry’ (July–December), based on
monthly variations in rates of conception and neonatal mortality within the same
dataset.
Results
Birth during the hungry season resulted in excess mortality during the first year
of life. However, for adult mortality (deaths 15 years), there was no excess in
individuals born during the annual hungry season: ratio of hazard
July–December versus January–June 1.12; 95% CI: 0.87, 1.45.
Conclusions The current study found no excess mortality in young adults born during the
‘hungry’ season in rural Bangladesh. This differing pattern in survival when
compared with The Gambia may be a consequence of the greatly reduced
incidence of young adult deaths in Bangladesh (0.1%) compared with The
Gambia (3%). Under such conditions possible differences in immune function
may not be detectable with early adult death as the outcome. However, it also
remains possible that our Gambian observation could be a highly discrete
phenomenon localized in either time or place, and as such, will not be replicated
in other populations.
Keywords
Mortality, seasonality, Bangladesh, infections, maternal nutrition, intrauterine
growth retardation
1 MRC International Nutrition Group, Public Health Nutrition Unit, London
School of Hygiene and Tropical Medicine, 49–51 Bedford Square, London,
WC1B 3DP, UK and MRC Keneba, The Gambia.
2 Health and Demographic Surveillance Programme, Public Health Science
Division, ICDDR,B Centre for Health and Population Research, Dhaka,
Bangladesh.
3 Current address: Department of Women’s and Children’s Health, International
Maternal and Child Health (IMCH), University Hospital, SE-751 85, Uppsala,
Sweden.
Correspondence: Dr Sophie E Moore, MRC International Nutrition Group,
Public Health Nutrition Unit, London School of Hygiene and Tropical
Medicine, 49–51 Bedford Square, London, WC1B 3DP, UK. E-mail:
[email protected]
In 1999 we published in this journal the finding that, in rural
Gambia, birth during or shortly after the annual hungry season
(July–December) predicts excess mortality in young
adulthood.1 The hazard ratio for early death in the hungry
season births rose from 3.7 for deaths over 14.5 years (P 0.000013) to 10.3 for deaths over the age of 25 years (P 0.00002) when compared with the harvest season
(January–June) (Figure 1). Cause of death data indicated that
the majority of these deaths were from infectious diseases, or
had a likely infectious aetiology. None of the deaths over this
period were from chronic degenerative diseases.
137
138
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Figure 1 Kaplan–Meier survival plots by season of birth in rural
Gambia
In rural Gambia, there is a variety of early-life exposures
correlated with season of birth that could influence later mortality.
These include seasonal infections in pregnancy (especially
malaria),2 pre- and post-natal exposure to environmental toxins
(e.g. aflatoxins),3,4 and postnatal exposure to seasonal infections
(particularly diarrhoeal disease).5 However, a review of the
evidence suggests that nutritionally mediated intrauterine growth
retardation during the annual hungry season seems the most
likely cause.6 This suggestion that immune function and
susceptibility to infectious disease may be programmed during
early life adds a novel dimension to the fetal and infant origins
of adult disease hypothesis.7,8
In order to test whether the Gambian finding could be
replicated elsewhere, we searched for long-term demographic
datasets from other seasonally exposed regions of the world.
The Demographic Surveillance System (DSS) from the rural
area of Matlab in Bangladesh met the necessary criteria of
meticulously recorded births and deaths within an area of high
nutritional seasonality. This paper reports patterns of mortality
analysed according to season of birth and presented alongside
our previous Gambian analysis.
Methods
Study populations
Bangladesh
Since 1966, the International Centre for Diarrhoeal Disease
Research, Bangladesh (ICDDR,B), formerly the Cholera Research
Laboratory, has been conducting a demographic surveillance and
health-related research programme in Matlab, a rural but densely
populated riverine area of Bangladesh (Figure 2). The Matlab
field research area forms part of the low-lying deltaic plain of
Bangladesh, situated about 70 km southeast of Dhaka.
A DSS was initiated in Matlab in 1966 for the registration of
vital events (initially births, deaths, in- and out-migration, but
later extended to include additional information on marriage and
divorce). In addition, periodic censuses have been carried out in
the DSS area (1974, 1982, and 1996) to collect demographic and
socioeconomic information. Data on childhood illnesses,
nutritional status, health status of women of reproductive age,
contraceptive use, and use of health care have also been
collected. In 1966, the surveillance population was 112 711 and
this had increased to a total population of 219 884 by 2001.
Prior to 1974, the village Dais (traditional birth attendants)
were responsible for visiting households and recording all vital
events (births, deaths, migration in or out, marriage or divorce).
This information was then passed to the male Health Assistants
who would visit the households to collect all the details of each
event on to specific health forms. From 1974, Community Health
Workers took over the responsibility of recording and reporting
all events, and visited each household every 2 weeks. In addition,
the Health Assistants continued to visit each household every
month to complete the detailed health forms. This system
continued for the duration of the period used in the current
analysis (1974–2000). The type of identification used prior to
1974 was a VTS (vaccine trial number) which changed with any
change of residence. As a consequence, this has made it very
difficult to follow individuals beyond 1974, and for this reason
the current analysis runs from 1974 only. From 1974 onwards,
there have been no interruptions in the data collection process.
This rigorous data collection protocol coupled with multiple levels
of supervision ensures that no important events are missed, and
hence the demographic data can be considered as complete.
In rural Bangladesh, three seasons define its subtropical
climate; the monsoon season (June–September), the cool-dry
‘winter’ season (October–February), and the hot-dry season
(March–May). Traditionally, the major agricultural crop has been
deep-water aman rice grown during the monsoon season, with
lesser acreage during the winter season devoted to boro rice.9 Since
agriculture provides the livelihood for most families in this region,
their existence is highly dependent on this seasonal pattern and,
as in The Gambia, it has consequences on the nutritional status of
both adults and children. Figure 3 shows the seasonal pattern in
maternal weight in Matlab. This data is taken from a longitudinal
study of over 2000 women between 1975 and 1978.10 As
illustrated, weight fluctuated throughout the study period, with
lowest mean weights observed during the months of September
and October. These weight fluctuations were observed to
correspond with seasonal food shortages. Previous work in Matlab
has also demonstrated a seasonal pattern in the proximate
determinants of infant and child mortality and malnutrition,11
with the period of worst nutritional status coinciding with the
highest rice prices and the highest risk of mortality.9
In Bangladesh, the average birthweight is low, with between
25% and 48% of babies estimated to have a birthweight below
2.5 kg.12 Analysis of birthweights of 1772 singleton babies born
at Kumundini Hospital in rural Bangladesh over two
consecutive years showed a consistent variation correlated with
the seasonal availability of food, with the highest birthweights
occurring in the period March–May, and the lowest in
September–November.12 The well-documented effects of the
seasonal pattern in nutritional status of pregnant women and
their infants in rural Bangladesh, together with the large-scale
long-term demographic database provides the opportunity to
retest the hypothesis that season of birth predicts adult survival.
The Gambia
The Gambian data used in this comparative study come from
the analysis, by season of birth, of all deaths between 1949 and
1998 in three isolated rural villages (Keneba, Kanton Kunda,
and Manduar) in the West Kiang region of The Gambia. In this
COMPARATIVE ANALYSIS OF SURVIVAL BY SEASON OF BIRTH
139
Figure 2 Map of Bangladesh
Figure 3 Seasonality of maternal weights in rural Bangladesh,
1976–1977. (Adapted from ref. 10)
region, the subsistence farming existence is heavily influenced
by an annual rainy season (July–October). The annual rains
coincide with a ‘hungry’ season when food crops from the
previous year’s harvest become depleted, and adults are engaged
in heavy agricultural labour prior to the next harvest.13 As a
consequence, a chronic negative energy balance is observed in
all adults, including pregnant women. Birthweights are around
200 g lower during the hungry season, a deficit that can be
reversed by maternal dietary supplementation.14,15 Most
maternal and infant diseases also peak during the hungry
season, especially malaria and diarrhoea.5,16 Season of birth can
therefore be used as an indicator of fetal and early infant
exposure to malnutrition and infectious diseases.
Since 1949, a demographic record of all births and deaths has
been continuously maintained in this region. This record was
established by Professor Sir Ian McGregor to provide demographic
data for his early studies of parasitology and nutrition.17 The
system instituted consisted of village recorders whose task it was
to accurately record all births and deaths in each village. At
registration, each villager was assigned a unique identification
number, and this system has been maintained continuously to
date. Although rates of migration have not been tracked using
this recorder system, the fate of each individual in the survey at
the time of follow-up was established through discussion with
family members.
Using a database of 3162 individuals aged up to 48 years, for
whom month of birth and current fate (dead or alive) were
140
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
known with certainty, a survival analysis by season of birth
found a profound bias in adult mortality in individuals born
during the annual hungry season.1 In the current paper, the
same data will be analysed in conjunction with the data from
Bangladesh.
Statistical analysis
The seasonal patterns observed in neonatal mortality rate and
birth rate (taken as a proxy measure of maternal energy status
9 months previously) in Matlab, demonstrate that the first 6
months of the year (January–June) appear considerably better
than the latter 6 months (Figures 4a and b). Therefore for
the purpose of analysis, month of birth has been divided into two
seasons; January–June and July–December. By coincidence, the
same division of the year was found to be appropriate in
the original analysis of data from The Gambia. The similarity in
the seasonality of fertility rates between Matlab and Keneba
is illustrated in Figure 4b. In order to check whether postadolescent mortality might follow another pattern with month of
birth, four other seasonal divisions of the year were defined: two
with 6-month-long seasons (March–August/September–February
and May–October/November–April) and two with 4-month-long
seasons (January–April/May–August/September–December and
March–June/July–October/November–February).
Cox proportional hazards models were fitted and checked using
Stata 7 (Stata Corporation, Texas, USA). Estimates of the hazard
ratio and its standard error for the seasonal effect were derived
from separate analyses of the two datasets. In order to compare
the seasonal effects between the populations, the datasets were
pooled and the population—season interaction fitted. In this last
model allowance was made for the possibility that the underlying
pattern of hazard with age differed between the populations or
sexes by fitting a separate baseline hazard function for each group.
Analysis of the Schoenfeld residuals revealed no evidence for
deviation from the proportional hazards assumption in any of the
four sex—site groups after the age of 15 years.
Results
The current analysis is based on 172 228 births and 24 697 deaths
in the Matlab population between June 1974 and December
2000 and the subset of 59 834 births and 252 deaths followed
beyond the age of 15 years. These are compared with the 1842
Keneba births followed beyond the age of 15 years among whom
58 deaths were subsequently observed. Figure 5 shows survival
according to season of birth in both populations. Mortality among
young adults in rural Bangladesh is considerably lower than
in The Gambia, with a cumulative mortality of only 0.1%
between the ages of 15 and 25 years compared with
approximately 3% for the same age span in The Gambia.
In Matlab birth during the hungry season resulted in excess
mortality during the first year of life. However, following those
known to survive beyond 15 years of age, there was no excess
mortality in individuals born during the annual hungry season:
ratio of hazard during the second half of the year compared
with the first half 1.12 (95% CI: 0.87, 1.45). Furthermore, no
other stratification of the year into two or three equal-length
seasons correlated significantly with the post-adolescent
mortality rates. Stratifying the analysis by sex made little
difference to this result and produced no evidence for a gender
difference in the effect of season of birth on adult deaths. In
contrast, the hazard ratio for season among over 15 year olds in
Keneba is 3.80 (95% CI: 1.97, 7.34). In an analysis of the
pooled datasets, the term for the interaction between season
and country was highly significant (P 0.001) indicating that
the seasonal effect is genuinely smaller in Matlab.
Discussion
Figure 4 Seasonal variation in neonatal mortality in Matlab. (a) The
neonatal mortality curve plots the percentage dying in the first month
of life against the month of birth. Two cycles are plotted. (b) Seasonal
variation in fertility in Matlab and Keneba. The fertility curves plot
the percentage of births born versus the month of conception. Two
cycles are plotted
The possibility that human immune function may be
programmed by early-life exposure to nutritional or other
insults could have major implications for our understanding of
disease processes in relation to infections, auto-immune disease,
and cancer immuno-surveillance. Our original Gambian
observation is remarkable in terms of the size of the effect on
mortality in young adulthood.1 The quality of the data
collection, together with the high level of statistical significance,
leaves little room to doubt its validity. Subsequently we have
demonstrated evidence of seasonal effects on infant thymic
COMPARATIVE ANALYSIS OF SURVIVAL BY SEASON OF BIRTH
development in this population18 and on T-cell subset
profiles.19 Furthermore, a study of adolescents participating in
an ongoing longitudinal study in the Philippines has shown that
prenatal undernutrition is significantly associated with reduced
thymopoietin production, and growth in length during the first
year of life was shown to be positively associated with
adolescent thymopoietin production.20 In the same cohort, the
predicted probability of mounting an adequate antibody
response to a typhoid vaccine was lower among adolescents
who were currently undernourished and born with a low
birthweight than those who were currently undernourished but
born with an adequate birthweight (0.32 versus 0.70).21
However, in spite of these findings, it remains possible that
our Gambian observation could be a highly discrete
phenomenon localized in either time or place. We are therefore
collaborating with a number of other studies worldwide to test
whether the Gambian mortality pattern occurs in other
seasonally stressed populations.
The Matlab cohort has many strengths. The population data
are collected using a highly intensive and rigorously audited
DSS that captures all births and deaths. We are unaware of any
systematic source of error that might lead to a bias in relation to
survival analysis based on season of birth. The Matlab cohort is
much larger than our Gambian cohort although, due to the
much lower adult mortality rate, the Bangladesh analysis covers
only 252 deaths beyond the age of 15 years. Most importantly
there is good evidence for nutritional seasonality. In common
with The Gambia, life in rural Bangladesh revolves around the
annual rains through the impact that this has on agriculture,
rice prices, agricultural wage rates, household food stocks, and
in turn, on the nutritional status of both adults and children.9
Although the total weight loss in Matlab is less than the amount
observed during the annual hungry season in The Gambia, the
loss in these women who start with such low weights would be
as critical in terms of the effect on reproductive success.
Analyses of neonatal mortality rates and birth rates clearly show
a seasonal swing at least as severe as that observed in The
Gambia. Therefore season of birth in this environment can also
act as a proxy for early-life exposures. In spite of these
strengths, the current analysis has been unable to demonstrate
any effect of season of birth on adult survival in this region of
rural Bangladesh.
The fact that the season of birth division in Matlab was
classified according to the same monthly definition as used in the
analysis of data from The Gambia (January–June versus
July–December) is purely coincidental, and was based on
internal evidence relating to neonatal mortality and conception
patterns from within the Matlab dataset. The question of
whether the profound seasonality in conception rates represents
a true biological phenomenon, mediated by the Frisch fertility
hypothesis,22 requires scrutiny as there could be an alternative
sociological explanation. In rural Bangladesh, marriage is most
common in the first months of the year, and could theoretically
account for the peak in births during September. However, the
seasonal pattern of fertility is independent of birth order, thus
implying that the seasonality is most unlikely to reflect the
timing of weddings.
We therefore believe that our pre hoc selection of seasonal
boundaries is robust and optimal. Nonetheless, we repeated the
mortality analysis using other divisions of the year into two or
141
three equal periods to search for any possible effect. No other
post hoc division (either 6- or 4-monthly) gave a larger (or
significant) difference in mortality rates among young adults
than the January–June versus July–December split.
There are several possible explanations for the discordant
results between Gambia and Bangladesh. Although the
Bangladesh results ostensibly represent a negative finding, a
careful examination of the source of the discordance may help
in interpreting the Gambian data and in informing future
studies into the biological mechanisms involved.
Reference to Figure 5 suggests that one explanation for the
lack of an association between season of birth and adult survival
in Matlab lies in the fact that, once the age of 5 years has been
reached, survival rates in this population are very high, with
only 0.1% mortality between the age of 15 and 25 years. In the
Gambian cohort, mortality in the harvest season births was
20-fold higher at approximately 2% between these ages,
and in the hungry season births this increased to approximately
50-fold higher. Under such relatively benign circumstances it is
possible that underlying differences in immune function could
exist without becoming apparent in relation to mortality.
Unfortunately morbidity data have not been collected on a
Figure 5 Kaplan-Meier survival estimates by country and season.
(a) Survival from birth, (b) Survival from 15 years of age
142
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
systematic basis in Matlab and therefore an analysis of less
severe end-points is not possible. Future studies might include
morbidity surveillance or measures of in vivo immune function
(possibly using vaccine challenges as has been successfully
employed elsewhere).21 Longer term follow-up of the Matlab
cohort might eventually reveal differences in mortality as the
normal immunosenescence associated with ageing erodes the
function reserve of immune defences in the whole population.
However, it must be acknowledged that by the age of 30 years
there is no hint of an incipient divergence (Figure 5).
Another possible explanation is that the original seasonal
exposure responsible for impaired adult immunity in The
Gambia does not relate to fetal undernutrition. Various studies
are in progress in The Gambia and prospectively in Matlab to
investigate possible associations between size at birth and later
immune function. Those completed so far have found limited
evidence in relation to early-life measures of immunity in
infancy19 and at age 8 years.23 A large study in young men aged
18–24 years for whom birthweight is known is currently in
progress in rural Gambia. This will represent a more definitive
test since it relates to an age when the survival curves have
started to separate.
Other exposures need to be considered. The most notable
difference in environmental exposures between The Gambia
and Bangladesh is in the exposure to malaria. In The Gambia,
malaria is endemic, with exposure peaking during and shortly
after the annual rains, and mortality is significantly increased
during this time.16 In Matlab, there is no malaria. Malaria can
have a profound effect on placental function during pregnancy
which might influence the early ontogeny of immunity.
However, we have argued elsewhere1,6 that the timing of peak
malaria transmission does not fit well with the observed pattern
of affected individuals, and also that malaria infection is most
virulent in primiparous mothers. This would predict an excess
of risk in first-born offspring that has not been observed. Other
seasonal infections might also influence either the pre-natal or
post-natal development of immunity if there are critical
windows in which immunity can be strongly influenced by
infectious stimuli. Evidence in favour of this comes from the
observation that the precise timing of neonatal BCG vaccination
(a strong Th1 booster) has a profound effect on the incidence of
atopy.24 It may be that such infections occurred in The Gambia
and not in Bangladesh.
Finally, it is possible that the phenomenon observed in The
Gambia is confined to individuals born during the 1950s and
1960s and is not observed in younger cohorts. Indeed a similar
analysis from rural Senegal has found that individuals born
during the hungry season showed no greater mortality than
those born during the harvest season.25 Whilst environmental
conditions in rural Senegal are almost parallel to those observed
in The Gambia, the main primary difference in the two analyses
is the age of the Gambian dataset compared with that from
Senegal. This would provide further evidence that the effect
observed in The Gambia might be a cohort effect, and is thus
not observed in more recent populations. Possible candidates for
a time-dependent seasonal exposure which has now diminished
in intensity might be the consumption of certain bush foods to
ward off famine in the hungry season, or the use of early forms
of insecticide on stored crops. We continue to try and
investigate such possibilities.
KEY MESSAGES
•
Analysis of data from rural Gambia has shown that being born during the annual ‘hungry’ season strongly
influences susceptibility to mortality from infectious diseases in young adulthood. It was hypothesized that this
effect could be mediated by an early-life insult to the developing immune system.
•
The current study compared survival by season of birth in a large cohort of subjects from the rural Matlab region
of Bangladesh.
•
No excess mortality was found in young adults born during the ‘hungry’ season in rural Bangladesh, despite a
strong seasonality in indicators of maternal and infant health.
•
It is possible that the Gambian finding is a highly discrete phenomenon that will not be replicated in other
populations.
References
1 Moore SE, Cole TJ, Collinson AC, Poskitt EME, McGregor IA, Prentice AM.
Prenatal or early postnatal events predict infectious deaths in young
adulthood in rural Africa. Int J Epidemiol 1999;28:1088–95.
2 McGregor IA. Epidemiology, malaria and pregnancy. Am J Trop Med Hyg
1984;33:517–25.
3 Wild CP, Rasheed FN, Jawla MFB, Hall A J, Jansen LAM, Montesano R.
In-utero exposure to aflatoxin in West Africa. Lancet 1991;337:
1602.
4 Wild CP, Hudson GJ, Sabbioni G et al. Dietary intake of aflatoxins
and the level of albumin-bound aflatoxin in peripheral blood in
The Gambia, West Africa. Cancer Epidemiol Biomarkers Prev
1992;1:229–34.
5 Brewster DR, Greenwood BM. Seasonal variation of paediatric diseases
in The Gambia, West Africa. Ann Trop Paediatr 1993;13:133–46.
6 Prentice AM, Cole TJ, Moore SE, Collinson AC. Programming the adult
immune system. In: O’Brien PMS, Wheeler T, Barker DJP (eds). Fetal
Programming: Influence on Development and Disease in Later Life. Proceedings
of the 36th RCOG Study Group, London: John Libbey & Son: 1999, pp.
399–423.
7 Barker DJP. Early growth and cardiovascular disease. Arch Dis Child
1999;80:305–10.
8 Barker DJP. The fetal origins of Type 2 Diabetes Mellitus. Ann Int Med
1999;130:322–24.
9 Chen LC, Chowdhury AKMA, Huffman SL. Seasonal dimensions
of energy protein malnutrition in rural Bangladesh: the role of
COMPARATIVE ANALYSIS OF SURVIVAL BY SEASON OF BIRTH
143
agriculture, dietary practices, and infection. Ecol Food Nutr
1979;8:175–87.
17 Billewicz WZ, McGregor IA. The demography of two West African
10 Huffman SL, Wolff M, Lowell S. Nutrition and fertility in Bangladesh:
18 Collinson AC, Moore SE, Cole TJ, Prentice AM. Environmental and
nutritional status of nonpregnant women. Am J Clin Nutr 1985;
42:725–38.
nutritional factors influencing the tracking of thymic size in rural
Gambian children. Acta Paediatr 2002;(in press).
11 Brown KH, Black RE, Becker S. Seasonal changes in nutritional status
19 Collinson AC. Early Nutritional and Environmental Influences on Immune
and the prevalence of malnutrition in a longitudinal study of young
children in rural Bangladesh. Am J Clin Nutr 1982;36:303–13.
12 Hort KP. Seasonal variation of birthweight in Bangladesh. Ann Trop
Paediatr 1987;7:66–71.
(Gambian) villages, 1951–75. J Biosoc Sci 1981;13:219–40.
Function in Rural Gambian Infants. 2002, University of Bristol.
20 McDade TW, Beck MA, Kuzawa CW, Adair LS. Prenatal undernutrition
and postnatal growth are associated with adolescent thymic function.
J Nutr 2001;131:1225–31.
13 Singh J, Prentice AM, Diaz E et al. Energy expenditure of Gambian
21 McDade TW, Beck MA, Kuzawa C, Adair LS. Prenatal undernutrition,
women during peak agriculatural activity measured by the doublylabelled water method. Br J Nutr 1989;62:315–29.
postnatal environments, and antibody response to vaccination in
adolescence. Am J Clin Nutr 2001;74:543–48.
14 Prentice AM, Cole TJ, Foord FA, Lamb WH, Whitehead RG. Increased
22 Frisch RE. Malnutrition and fertility. Science 1982;215:1272–74.
23 Moore SE, Collinson AC, Prentice AM. Immune function in rural
birthweight after prenatal dietary supplementation of rural African
women. Am J Clin Nutr 1987;46:912–25.
15 Ceesay SM, Prentice AM, Cole TJ et al. Effects on birthweight and
Gambian children is not related to season of birth, birth size, or
maternal supplementation status. Am J Clin Nutr 2001;74:840–47.
perinatal mortality of maternal dietary supplementation in a primary
health care setting in rural Gambia. BMJ 1997;315:786–90.
24 Aaby P, Shaheen SO, Heyes CB et al. Early BCG vaccination and
16 Greenwood BM, Bradley AK, Greenwood AM et al. Mortality and
morbidity from malaria among children in a rural area of The Gambia,
West Africa. Trans R Soc Trop Med Hyg 1987;81:478–86.
reduction in atopy in Guinea-Bissau. Clin Exp Allergy 2000;30:644–50.
25 Simondon KB, Elguero E, Marra A, Diallo A, Aaby P, Simondon F.
Season of birth is not associated with risk of early adult death in rural
Senegal. Int J Epidemiol 2004;33:130–36.
Matlab newborn with parents and grandmother.
Photograph: Asem Ansari