Community and International Nutrition

Pregnancy Increases BMI in Adolescents of a Population-Based Birth

Cohort1
Denise P. Gigante,*†2 Kathleen M. Rasmussen,† and Cesar G. Victora*
*Post-Graduate Program in Epidemiology, Universidade Federal de Pelotas, CP 464, 96001–970, Pelotas,
RS, Brazil and †Division of Nutritional Sciences, Cornell University, Ithaca, NY 14853

ABSTRACT Evidence from developed countries suggests that adolescents grow while pregnant and that their
growth is associated with increased weight gain and fat storage, but this has never been examined in girls from
developing countries. Adolescents born in 1982 in Pelotas, Brazil, are being followed in a birth cohort study.
Information on social and biological determinants of nutritional status was collected in early life. Both in 1997 and
in 2001, 464 girls were located through household visits, 16% of whom had had a pregnancy in this period.
Changes in height, weight, and BMI between 1997 and 2001 were analyzed in relation to the occurrence of
pregnancy, after adjustment for previous anthropometric status, as well as social and biological characteristics.
The average gains were 2.0 ⫾ 2.0 cm in height, 3.1 ⫾ 5.9 kg in weight, and 0.7 ⫾ 2.2 kg/m2 in BMI. Each pregnancy

Downloaded from jn.nutrition.org by on October 17, 2008

was associated with a reduction of 0.46 cm on height gain from 1997 to 2001 (P ⫽ 0.02). Girls who became
pregnant gained 2.25 kg more than all others (P ⫽ 0.004). There was a clear association between pregnancy and
BMI change. A single pregnancy was associated with an increase of 0.81 kg/m2 (P ⫽ 0.01) and 2 or more
pregnancies were associated with an increase of 1.58 kg/m2 (P ⫽ 0.02). Teenage pregnancy was associated with
an important increase in BMI. Given the growing epidemic of obesity in low- and middle-income countries,
particularly among women, efforts to reduce teenage pregnancy may also contribute to preventing overweight. J.
Nutr. 135: 74 – 80, 2005.

KEY WORDS: ● adolescence ● pregnancy ● nutritional status ● BMI ● cohort studies

The prevalence of obesity has increased worldwide as a
result of the rapid nutritional transition observed in many
countries (1). Among adolescents, the age group between 10
and 19 y old, increased weight and height have been docu-
mented even in less-developed regions (2). Data from devel-
oped countries suggest that the diet of adolescents puts them
at risk for chronic diseases, such as cardiovascular disease,
diabetes, cancer, and osteoporosis (3). From the perspective of
prevention, the identification of groups of people at risk of
becoming obese has been considered an important concern in
research. In a systematic review that included studies of child-
hood predictors of adult obesity, Parsons and colleagues (4)
concluded that the lack of longitudinal data from childhood to
adulthood is the major research gap.
Likewise, many developing countries are experiencing an
increase in the rate of adolescent pregnancy. In Brazil, there
were 8.0 births per 1000 adolescents in 1980, which increased
14% to 9.1 in 2000. Among all births, the increase in the
proportion of adolescent mothers has been even sharper. In
1980, 9.1% of all births were among adolescent mothers, and
this proportion more than doubled, to 19.4% in 2000 (5).
1
This study was supported by the Fundação de Amparo à Pesquisa no Rio
Grande do Sul (FAPERGS), by the Coordenação de Aperfeiçoamento de Pessoal
de Nı́vel Superior (CAPES) of Brazil, and by WHO.
2
To whom correspondence should be addressed.
E-mail: denise@epidemio-ufpel.org.br.
The consequences of teenage pregnancy for adult nutri-
tional status have been little studied in developing countries as
a result of the lack of longitudinal studies. Evidence from
developed countries suggests that adolescents grow while preg-
nant and that their growth is associated with increased weight
gain and fat storage. Scholl and colleagues (6), in the Camden
study, showed that the amount of postpartum retained weight
was significantly greater in still-growing gravidas than in other
pregnant women. Growth in stature continues after menarche
(7,8) and during teenage pregnancy. In the Camden Study,
growth in stature was detected on the basis of a knee height
measuring device to eliminate the effect of shrinkage in stature
during pregnancy. Adolescents had significantly positive in-
crements compared with mature gravidas (9).
In Pelotas, a city in southern Brazil, girls have been studied
since their birth in 1982. In 2001, the risk factors for child-
bearing during adolescence were studied in this cohort (10). A
total of 16% of the girls who belonged to the cohort study gave
birth up to 2001. Anthropometric variables were measured
since birth.
The purpose of the research reported here was to study the
effects of teenage pregnancy on nutritional status at age 19 in
a subsample of female adolescents that was followed from 1982
to 2001. The Pelotas Birth Cohort Study allows examining
these associations, while accounting for some of the possible
confounding factors that have not been considered in many
studies.

0022-3166/05 $8.00 © 2005 American Society for Nutritional Sciences.

Manuscript received 16 August 2004. Initial review completed 1 September 2004. Revision accepted 11 October 2004.

74
 TEENAGE PREGNANCY AND NUTRITIONAL STATUS
METHODS
Study population and sample. In 1982, births of all women
living in the urban area of Pelotas were included in a population-
based, birth-cohort study. Information was obtained on 5914 live-
born infants, representing 99% of births in the year. This cohort has
been followed several times, and the methods used in earlier phases
have been described elsewhere (11–14).
The research described here involved female adolescents who
belonged to the Cohort Study and who were examined in both 1997
and 2001. In 1997, a systematic sample of 27% of all households in
the city was visited in search of adolescents who were born in 1982.
It resulted in the identification of 515 female adolescents, who were
also visited in 2001. In this latest follow-up, 502 girls were located,
representing 97.5% of those located in 1997. Of the subcohort of
female adolescents who were seen in both 1997 and 2001, 38 were
excluded because they were pregnant at the time of the 2001 inter-
view (n ⫽ 24), had a child before 1997 (n ⫽ 4), had a child in the
last 6 mo (n ⫽ 5), or had missing information in anthropometric
variables (n ⫽ 5). All analyses were based on the remaining 464
young women.
Data collection. In all phases of this cohort study, a group of field
workers was trained in interview techniques and anthropometry. The
enrollment of the Birth Cohort Study took place in the maternity
hospital, and follow-ups were conducted in the girls’ households.
Informed consent was obtained in all phases of the study, and the
adolescent and her mother gave their consent in this phase of the
cohort study. The study was approved by the maternity hospitals’
ethical committees and by the university ethical committee.
Information concerning the mother and her pregnancy was ob-
tained at enrollment in 1982 and was collected at the maternity
hospitals up to 3 d after birth. Information concerning the children
was obtained through standardized questionnaires given to their
mothers in 1984 and in 1986 at mean ages of 2 and 4 y, respectively.
Mothers of the study subjects were weighed and measured in the
maternity hospitals. Their weight at the beginning of pregnancy was
obtained by recall or, if available, collected from the antenatal card.
Children were weighed in the maternity hospital and in all follow-up
contacts. Length or height (as appropriate) was measured at each
follow-up. In the 1997 and 2001 follow-up contacts, questionnaires
were applied to both the adolescents and their mothers. Mean ages
were 15 and 19 y in 1997 and 2001, respectively.
In all phases, a fieldwork supervisor repeated 5% of the interviews.
These findings were compared with the original, ensuring the quality
control.
Variables. This analysis included variables from different phases
of this cohort study. The variables collected in 1982 were family
income (the sum of monthly incomes of all working people living in
the household in minimal wage, categorized as ⱕ1 minimal wage, 1.1
to 3, 3.1 to 6, and ⬎6 minimal wages), maternal skin color (white or
nonwhite), maternal age (categorized as ⬍20; 20 to 29; and ⱖ30 y),
pregestational weight (from the antenatal care register or by asking
the mother about her weight before pregnancy), maternal height
(measured soon after admission to the maternity hospital), and birth
weight (measured in grams immediately after birth by trained inter-
viewers, using regularly calibrated pediatric scales). Children were
weighed using portable scales (CMS Weighing Equipment) and had
their supine length (1984) or height (1986) measured using boards
manufactured locally according to international specifications (AHR-
TAG). BMI at ages 2 and 4 y were calculated by dividing weight by
the square of height measured in 1984 and 1986, respectively. To
define overweight in childhood and adolescence, dichotomous vari-
ables were used according to the recommended cutoff values by age
(15). In 1984, the age of children varied from 12 to 27 mo and in
1986, from 36 to 51 mo. In 1997 and 2001, these ranges were 19 and
20 mo, respectively.
The variables collected during adolescence included BMI based
on weight and height measured at ages 15 and 19 y, skin color
reported by adolescent in 1997, and age at menarche (continuous
variable, collected in 1997 and confirmed in 2001, divided into 4
categories: ⬍12, 12, 13, and ⱖ14 y). Whether there was a pregnancy
between 1997 and 2001 was obtained by asking the adolescent in the
interviewer-applied questionnaire and in a confidential question-
75
naire. In this later questionnaire, information about abortion could be
more precise. The higher number of pregnancies was used in the
analyses.
The outcomes variables were changes in height, weight, and BMI
between 1997 and 2001. Seven girls were excluded from the analyses
because their change in height or weight was 3 SD or more above or
below the mean change for the whole sample; these were likely to
represent major measurement errors (16).
Statistical methods. To describe the sample, frequency distribu-
tions were used to compare main variables collected in earlier phases
of the cohort study for all girls and for those who were included in this
analysis. The ␹2 test was used to compare these proportions. To
examine associations, ANOVA and F test were used to compare
mean values of anthropometric variables across different categories of
explanatory variables. Multiple linear regression analyses were carried
out for changes in height, weight, and BMI between 1997 and 2001.
Potential confounding factors, including anthropometric measure-
ments in childhood and age at menarche, were investigated and those
that showed an association (P ⬍ 0.2) were taken to the multivariable
analyses. Variables were included as continuous in the linear regres-
sion models. Age at menarche and teenage pregnancy were also
assessed both as categorical variable (⬍12, 12, 13, or 14 y and none
pregnancy, 1 or ⱖ2 pregnancies, respectively) and as dichotomous
variables (ⱖ12 vs. ⬍12 y and yes vs. no, respectively). Interactions
between socioeconomic and nutritional variables included in these
Downloaded from jn.nutrition.org by on October 17, 2008
models were tested jointly through the product of all variables. The
same criterion of significance for confounding factors (P ⬍ 0.2) was
used for interactions. Graphical analyses of the residuals were devel-
oped for the linear regression assumptions, and colinearity was tested
through variance inflation factors analysis (17) and the STATA
package (Intercooled Stata 8.0 for Windows, Stata) was used in the
analyses. The regression model included all potential confounding
variables, and results are shown for age at menarche and the occur-
rence of teenage pregnancy.
RESULTS
The comparison between characteristics of girls included in
the analysis and those identified in the original cohort shows
that adolescents from the lowest income group, those born
with a low birth weight, whose mothers were ⬍30 y old were
less likely to have been identified in the recent follow-up and
included in this analysis (Table 1).
The incidence of pregnancy between 1997 and 2001 was
inversely related to family income in 1982 (Table 2). There
was a 10-fold difference between girls belonging to families in
the lowest and the highest income groups. An inverse associ-
ation was also observed between birth weight and pregnancy.
In addition, pregnancy between 1997 and 2001 was less fre-
quent among taller girls, with a 4-fold difference between the
extreme height groups (Table 2).
Means and standard deviations of heights and weights were
80.7 ⫾ 5.1 cm and 11.0 ⫾ 1.6 kg at age 2; 97.4 ⫾ 5.2 cm and
15.5 ⫾ 2.4 kg at age 4; 159.5 ⫾ 6.3 cm and 54.7 ⫾ 10.2 kg at
age 15; and 161.5 ⫾ 6.3 cm and 58.3 ⫾ 12.9 kg at age 19.
Height and weight at ages 2 and 4 were negatively associated
with age at menarche—taller and fatter girls had earlier men-
arche (data not shown). Inverse associations were also ob-
served between age at menarche and attained weights at ages
15 and 19.
BMI values were 16.8 ⫾ 1.4 kg/m2, 16.2 ⫾ 1.5 kg/m2, 21.5
⫾ 3.5 kg/m2, and 22.3 ⫾ 4.5 kg/m2 at ages 2, 4, 15, and 19,
respectively. Higher BMI values at ages 4, 15, and 19 were
observed among girls who had early menarche [P ⬍ 0.001
(data not shown)].
Family income, birth weight, and overweight in childhood
were positively associated with attained height in 2001 (Table
3). An inverse association was observed between teenage
76 GIGANTE ET AL.

TABLE 1 In the adjusted analysis, height changes were significantly

larger among girls with later menarche (Table 4). Each year
Characteristics of the original 1982 Pelotas birth cohort and for which menarche was delayed resulted in an increase of 0.47
adolescents included in the analyses. Pelotas, 1982–2001 cm in height (P ⬍ 0.001). When age at menarche was treated
as a categorical variable, the effect was restricted to ages 13 y
Included
Original in
or greater. The dichotomous variable was also significant (P
cohort analysis ⫽ 0.001). The continuous variable expressing the number of
Variable (n ⫽ 2876) (n ⫽ 464) P-value pregnancies showed that adolescents who had been pregnant
were shorter than those who had never been pregnant (P
% ⫽ 0.02). Girls who were never pregnant were 0.40 cm (P
⫽ 0.11) taller than those with one or more pregnancies.
Family income1 (minimum wage) 0.04
ⱕ1 21.7 15.8 Finally, categorical analysis showed that, compared with girls
1.1–3 46.4 49.0 who were never pregnant, those who had 1 pregnancy were
3.1–6 19.1 21.5 0.18 cm shorter (P ⫽ 0.5) and those with 2 or more pregnan-
⬎6 12.8 13.7 cies were 1.60 cm shorter (P ⫽ 0.006). Therefore, the effect of
Maternal skin color 0.2 pregnancy was restricted to those with more than 1 pregnancy.
White 82.5 84.7
Nonwhite 17.5 15.3 Weight changes were also larger among girls with later
Maternal age, y 0.03 menarche (Table 5). Each year for which menarche was
⬍20 15.4 12.5 delayed, adolescents weighed 0.79 kg more than the others (P
20–29 57.3 54.7 ⬍ 0.001). When age at menarche was treated as a categorical
ⱖ30 27.4 32.8
Birth weight, g 0.05
⬍2500 10.1 6.7

Downloaded from jn.nutrition.org by on October 17, 2008

2500–2999 26.7 25.4 TABLE 2
3000–3499 37.9 37.3
3500–3999 21.3 25.2 Incidence of pregnancy between 1997 and 2001 according
ⱖ4000 4.0 5.4
Overweight in 1984 0.6 to selected covariates included in the analysis.
No 83.2 84.5 Pelotas, 1982–2001
Yes 16.8 15.5
Overweight in 1986 0.8 Variable n % Pregnant P-value1
No 81.4 82.4
Yes 18.6 17.6 Family income2 in 1982 (minimum
wage) ⬍0.001
1 Income expressed as multiple of the minimum wage. ⱕ1 73 31.5 ⬍0.0013
1.1–3 226 19.9
3.1–6 99 9.1
pregnancy and attained height (P ⫽ 0.002). Skin color, age at ⬎6 63 3.2
menarche, and maternal age were not associated with height. Maternal age in 1982, y 0.07
⬍20 58 12.1 0.53
Maternal age, birth weight, and overweight in childhood 20–29 254 20.9
were positively associated with attained weight in 2001, but ⱖ30 152 13.2
there was an inverse association with age at menarche (Table Birth weight, g 0.06
3). BMI in 2001 was positively associated with age of the ⬍2500 31 16.1 0.023
adolescent’s mother, with overweight during childhood, and 2500–2999 118 24.6
with early menarche. There was a trend (P ⫽ 0.07) toward 3000–3499 173 17.3
3500–3999 117 12.8
greater BMI among nonwhite adolescents. Height change be- ⱖ4000 25 4.0
tween 1997 and 2001 was 2.0 ⫾ 2.0 cm. Of the girls, 11% had Skin color 0.1
negative changes in height of up to 3 cm, which were likely White 373 15.8
due to measurement error; these values were retained in the Nonwhite 91 23.1
analyses, but the results were unchanged when they were Age at menarche, y 0.8
excluded. The change in weight was 3.1 ⫾ 5.9 kg, and for BMI ⬍12 119 20.2 0.43
12 130 16.2
it was 0.7 ⫾ 2.2 kg/m2. Changes in height were smaller among 13 120 16.7
girls who were overweight in 1986 and for white adolescents ⱖ14 95 15.8
(Table 3). A positive association was observed between height Z score weight for age in 1997 0.5
change and age at menarche (P ⬍ 0.001). Height change was ⬍⫺1 64 17.2 0.43
similar among adolescents who had zero or one pregnancy, but ⫺1 to 1 326 18.4
was smaller among those with 2 or more pregnancies. ⬎1 72 12.5
Z score height for age in 1997 0.05
Change in weight was positively associated with age of the ⬍⫺1 109 22.9 0.023
adolescent’s mother. Weight change was also greater among ⫺1 to 1 316 16.8
adolescents who were not overweight in 1986 (P ⫽ 0.01) and ⬎1 37 5.4
among those with later menarche (P ⬍ 0.001). The effect of BMI in 1997, kg/m2 0.4
pregnancy on weight change was not linear; weight gain was ⬍18.5 77 13.0 0.93
most pronounced among adolescents who had a single preg- 18.5–24.9 320 19.4
25–29.9 51 11.8
nancy (Table 3). ⱖ30 15 13.3
BMI change between 1997 and 2001 was inversely related
to family income and directly related to age at menarche. 1 Pearson ␹2 test.
Teenage pregnancy was positively associated with BMI change 2 Income expressed as number of times the minimum wage.
(P ⫽ 0.001), with a significant linear trend (Table 3). 3 Linear trend.
 TEENAGE PREGNANCY AND NUTRITIONAL STATUS 77

TABLE 3
Height, weight, and BMI at age 19 and height, weight, and BMI change between ages from 15 to 19 according to selected
covariates included in the analysis. Pelotas, 1982–2001

Variable n Height1 Weight BMI Height change Weight change BMI change

cm kg kg/m2 cm kg kg/m2

Family income2 (minimum

wage)
ⱕ1 73 159.3 ⫾ 6.8 56.1 ⫾ 9.8 22.1 ⫾ 3.8 2.2 ⫾ 2.1 3.7 ⫾ 6.3 1.0 ⫾ 2.5
1.1–3 226 161.2 ⫾ 5.9 59.3 ⫾ 15.1 22.8 ⫾ 5.3 2.0 ⫾ 2.1 3.5 ⫾ 6.1 0.8 ⫾ 2.3
3.1–6 99 162.5 ⫾ 6.6 58.7 ⫾ 11.4 22.2 ⫾ 4.0 2.2 ⫾ 1.9 2.4 ⫾ 6.0 0.4 ⫾ 2.3
⬎6 63 163.8 ⫾ 6.1 57.0 ⫾ 8.4 21.2 ⫾ 2.5 1.9 ⫾ 1.9 2.5 ⫾ 4.7 0.4 ⫾ 1.7
t test ⬍0.001 0.2 0.10 0.8 0.3 0.2
Linear trend ⬍0.001 0.9 0.10 0.8 0.09 0.05
Maternal age, y
⬍20 58 160.2 ⫾ 5.8 56.0 ⫾ 10.0 21.8 ⫾ 3.8 1.9 ⫾ 2.0 1.9 ⫾ 5.3 0.3 ⫾ 1.8
20–29 254 161.5 ⫾ 6.5 57.5 ⫾ 12.6 22.0 ⫾ 4.5 2.0 ⫾ 2.1 3.0 ⫾ 6.1 0.7 ⫾ 2.4
ⱖ30 152 162.0 ⫾ 6.2 60.4 ⫾ 14.0 23.0 ⫾ 4.8 2.2 ⫾ 1.9 3.8 ⫾ 5.9 0.9 ⫾ 2.3
t test 0.19 0.03 0.09 0.5 0.11 0.3
Linear trend 0.09 0.01 0.04 0.2 0.04 0.11
Birth weight, g
⬍2500 31 157.5 ⫾ 6.2 53.6 ⫾ 9.6 21.6 ⫾ 4.4 2.2 ⫾ 1.8 2.9 ⫾ 5.4 0.6 ⫾ 2.2
2500–2999 118 159.8 ⫾ 5.8 57.9 ⫾ 13.1 22.6 ⫾ 4.7 2.1 ⫾ 2.0 4.0 ⫾ 7.2 0.9 ⫾ 2.8

Downloaded from jn.nutrition.org by on October 17, 2008

3000–3499 173 161.6 ⫾ 6.1 57.5 ⫾ 11.9 22.0 ⫾ 4.5 1.9 ⫾ 2.0 3.0 ⫾ 5.1 0.6 ⫾ 1.9
3500–3999 117 163.1 ⫾ 6.2 60.5 ⫾ 14.1 22.7 ⫾ 4.6 2.1 ⫾ 2.1 3.1 ⫾ 5.8 0.6 ⫾ 2.1
ⱖ4000 25 166.1 ⫾ 6.1 61.4 ⫾ 11.4 22.2 ⫾ 4.0 2.2 ⫾ 2.1 0.8 ⫾ 11.4 ⫺0.2 ⫾ 1.9
t test ⬍0.001 0.04 0.6 0.9 0.18 0.2
Linear trend ⬍0.001 0.006 0.6 0.9 0.11 0.13
Overweight at age 2 y
No 366 161.2 ⫾ 6.3 57.3 ⫾ 12.8 22.0 ⫾ 4.6 2.0 ⫾ 2.1 2.9 ⫾ 5.9 0.6 ⫾ 2.2
Yes 67 163.4 ⫾ 6.7 63.3 ⫾ 11.9 23.7 ⫾ 4.2 2.2 ⫾ 1.8 3.4 ⫾ 5.6 0.7 ⫾ 2.2
t test 0.007 ⬍0.001 0.005 0.6 0.5 0.8
Overweight at age 4 y
No 356 161.4 ⫾ 6.3 57.3 ⫾ 13.2 22.0 ⫾ 4.7 2.2 ⫾ 2.1 3.5 ⫾ 5.9 0.8 ⫾ 2.2
Yes 76 163.2 ⫾ 6.4 63.9 ⫾ 10.9 24.0 ⫾ 4.1 1.6 ⫾ 1.6 1.6 ⫾ 6.2 0.2 ⫾ 2.4
t test 0.02 ⬍0.001 ⬍0.001 0.02 0.01 0.07
Adolescent skin color
White 373 161.6 ⫾ 6.3 57.8 ⫾ 12.2 22.1 ⫾ 4.3 2.0 ⫾ 1.9 2.9 ⫾ 6.0 0.6 ⫾ 2.3
Nonwhite 91 161.2 ⫾ 6.4 60.1 ⫾ 15.3 23.1 ⫾ 5.4 2.5 ⫾ 2.4 4.1 ⫾ 5.7 0.9 ⫾ 2.2
t test 0.6 0.13 0.07 0.03 0.10 0.2
Age at menarche, y
⬍12 119 160.9 ⫾ 6.8 62.5 ⫾ 16.1 24.0 ⫾ 5.4 1.5 ⫾ 1.7 2.5 ⫾ 6.8 0.6 ⫾ 2.6
12 130 161.3 ⫾ 6.0 57.0 ⫾ 12.3 21.9 ⫾ 4.7 1.7 ⫾ 1.9 1.7 ⫾ 5.0 0.2 ⫾ 2.0
13 120 161.9 ⫾ 6.3 57.6 ⫾ 10.8 21.9 ⫾ 3.6 2.1 ⫾ 1.9 3.9 ⫾ 5.9 1.0 ⫾ 2.2
ⱖ14 95 162.0 ⫾ 6.3 55.7 ⫾ 10.4 21.2 ⫾ 3.7 3.2 ⫾ 2.2 4.9 ⫾ 5.6 0.9 ⫾ 2.0
t test 0.5 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 0.03
Linear trend 0.14 ⬍0.001 ⬍0.001 ⬍0.001 ⬍0.001 0.05
Teenage pregnancy
None 384 161.8 ⫾ 6.4 58.2 ⫾ 13.1 22.2 ⫾ 4.5 2.1 ⫾ 2.0 2.7 ⫾ 5.7 0.5 ⫾ 2.1
1 68 160.7 ⫾ 5.5 58.8 ⫾ 12.0 22.8 ⫾ 4.3 2.2 ⫾ 2.0 5.3 ⫾ 6.8 1.4 ⫾ 2.5
2 or 3 12 155.6 ⫾ 5.9 56.9 ⫾ 12.8 23.6 ⫾ 5.9 0.0 ⫾ 1.4 3.9 ⫾ 7.7 1.7 ⫾ 3.1
t test 0.002 0.9 0.4 0.002 0.005 0.004
Linear trend 0.002 0.9 0.17 0.04 0.005 0.001

1 Values are means ⫾ SD.

2 Income expressed as multiple of the minimum wage.

variable, the association was restricted to ages 14 y or more.
When expressed as a continuous variable, weight was higher
by 1.56 kg per pregnancy (P ⫽ 0.01). When expressed as a
dichotomous variable, girls who were never pregnant weighed
2.25 kg less than those who had become pregnant (P ⫽ 0.004).
Compared with girls who were never pregnant, those who had
1 pregnancy weighed 2.24 kg more (P ⫽ 0.007). Those with 2
or more pregnancies were 2.31 kg heavier (P ⫽ 0.2). There-
fore, the effect of pregnancy was restricted to those with 1
pregnancy (Table 5).
BMI change and age at menarche were not significantly
associated in either the crude or the adjusted analyses (Table
6). There was a clear association between pregnancy and BMI
change. Girls who had ever been pregnant were 0.92 kg/m2
fatter than all others (P ⫽ 0.002). For each pregnancy, BMI
increased by 0.71 kg/m2 (P ⫽ 0.002). Finally, a single preg-
nancy was associated with an increase of 0.81 kg/m2 (P
⫽ 0.01), and 2 or more pregnancies were associated with an
increase of 1.58 kg/m2 (P ⫽ 0.02).
No interactions were observed in the models, and graphic
analyses of the residuals indicated that the assumptions of
linear regression were not violated. Colinearity was found
when 2 measurements of weight (1984 and 1986) were in-
cluded in the multiple linear regression analyses. The results
were similar when just 1 or 2 measurements were included in
the analyses.
78 GIGANTE ET AL.

TABLE 4
Multiple linear regression analysis of the effects of menarche and pregnancy on height change. Pelotas, 1982–20011

Crude Adjusted2

␤ SE P-value ␤ SE P-value

Age at menarche, y
Continuous 0.48 0.06 ⬍0.001 0.47 0.07 ⬍0.001
ⱖ12 vs. ⬍12 0.79 0.21 ⬍0.001 0.78 0.22 0.001
⬍12 0.00 — — 0.00 — —
12 0.22 0.24 0.4 0.24 0.25 0.4
13 0.66 0.25 0.008 0.70 0.26 0.007
ⱖ14 1.76 0.26 ⬍0.001 1.70 0.28 ⬍0.001
Pregnancy
Continuous ⫺0.44 0.19 0.02 ⫺0.46 0.19 0.02
Yes vs. no ⫺0.25 0.25 0.3 ⫺0.40 0.25 0.11
None 0.00 — — 0.00 — —
1 0.08 0.26 0.8 ⫺0.18 0.27 0.5
ⱖ2 ⫺2.07 0.58 ⬍0.001 ⫺1.60 0.57 0.006

1 n ⫽ 457.
2 Adjusted for skin color, maternal age, weight, and length at 2 y, and for weight and height at 4 y. Pregnancy was also adjusted for age at
menarche as a continuous variable.

Downloaded from jn.nutrition.org by on October 17, 2008

DISCUSSION
In this study, we investigated the occurrence of pregnancy
from 1997 to 2001 and the changes on nutritional status
during this period in a sample of girls in the Pelotas Birth
Cohort Study who have been followed since their birth in
1982. Longitudinal data allowed us to assess the association of
teenage pregnancy with nutritional status of female adoles-
cents. Nutritional status was assessed several times since birth,
and this analysis included girls who were measured in 1997 and
2001. Very few studies from developing countries permit this
type of analysis (18). In addition, information bias was reduced
by collecting information close to the time at which events
occurred and by collecting weight and height during child-
hood and adolescence.
This study included all girls belonging to the cohort
study who were identified in a cluster sample of 27% of all
households in the city in 1997. We estimate that these
represent 72% of the cohort adolescents who should have
located in this sample of the city, in the absence of out-
migration. Losses to follow-up were more common among
the poorest families, when the adolescent herself had been
born to an adolescent mother, and for low-birth-weight
girls, which may be partly explained by survival bias. How-
ever, differences between those located and those lost to
follow-up were not marked. A more detailed analysis of
losses to follow-up is available elsewhere (14). Among
female adolescents interviewed in 1997, 96.5% were seen
again in 2001. Of all girls included in this study, only 4
adolescents had a term delivery before 1997. These teenage
mothers were excluded in this analysis because pregesta-
tional nutritional status was unknown.
According to the official information systems for live and
stillbirth in Pelotas, childbearing during adolescence was re-
ported for 16.2% of all girls belonging to the original cohort.

TABLE 5
Multiple linear regression analysis of the effects of menarche and pregnancy on weight change. Pelotas, 1982–20011

Crude Adjusted2

␤ SE P-value ␤ SE P-value

Age at menarche, y
Continuous 0.66 0.20 0.001 0.79 0.21 ⬍0.001
ⱖ12 vs. ⬍12 0.85 0.64 0.18 0.99 0.68 0.14
⬍12 0.00 — — 0.00 — —
12 ⫺0.79 0.75 0.3 ⫺0.66 0.79 0.4
13 1.39 0.76 0.07 1.47 0.80 0.07
ⱖ14 2.35 0.81 0.004 2.77 0.87 0.002
Pregnancy
Continuous 1.46 0.57 0.01 1.56 0.62 0.01
Yes vs. no 2.34 0.72 0.001 2.25 0.78 0.004
None 0.00 — — 0.00 — —
1 2.54 0.78 0.001 2.24 0.83 0.007
ⱖ2 1.17 1.73 0.5 2.31 1.88 0.2

1 n ⫽ 460.
2 Adjusted for family income, skin color, birth weight, weight and length at 2 y, and for weight and height at 4 y. Pregnancy was also adjusted for
age at menarche as a continuous variable.
 TEENAGE PREGNANCY AND NUTRITIONAL STATUS 79

TABLE 6
Multiple linear regression analysis of the effects of menarche and pregnancy on BMI change. Pelotas, 1982–20011

Crude Adjusted2

␤ SE P-value ␤ SE P-value

Age at menarche, y
Continuous 0.11 0.07 0.13 0.14 0.08 0.07
ⱖ12 vs. ⬍12 0.10 0.24 0.7 0.13 0.25 0.6
⬍12 0.00 — — 0.00 — —
12 ⫺0.36 0.28 0.2 ⫺0.33 0.30 0.3
13 0.39 0.29 0.17 0.37 0.30 0.2
ⱖ14 0.36 0.31 0.2 0.47 0.32 0.15
Pregnancy
Continuous 0.55 0.32 0.09 0.71 0.23 0.002
Yes vs. no 0.91 0.27 0.001 0.92 0.30 0.002
None 0.00 — — 0.00 — —
1 0.86 0.29 0.004 0.81 0.31 0.01
ⱖ2 1.18 0.64 0.07 1.58 0.70 0.02

1 n ⫽ 454.
2 Adjusted for family income, maternal age, birth weight, weight at age 2 y, and for weight at 4 y. Pregnancy was also adjusted for age at menarche
as a continuous variable.

Downloaded from jn.nutrition.org by on October 17, 2008

Among girls included in the present analyses, 16.9% delivered
a baby between 1997 and 2001.
Teenage pregnancy is a time of nutritional risk (19). Ado-
lescents most likely to become pregnant are often those with
inadequate nutritional status and unfavorable socioeconomic
background (20,21). Scholl and Hediger (22) showed that
there is a competition for nutrients between the growing
pregnant adolescent and her fetus. The present results con-
firmed a strong inverse association between pregnancy and
change of height, especially among those who had 2 or more
pregnancies.
Growth in stature continues after menarche and during
teenage pregnancy (7,8,23). In this study, mean attained
heights at the end of adolescence were similar for girls with
zero or one pregnancy, but markedly shorter for those with 2 or
more gestations. This suggests that there was a cumulative
effect on growth in height that only became apparent for girls
with 2 pregnancies.
Researchers have found that there is considerable weight
gain and increase in subcutaneous fatness at central sites when
pregnancy occurs during the final phase of adolescent growth
(6,23,24). Some authors have documented that adolescent
pregnancy is associated with larger gestational weight gains
and increased risk for weight retention (25). The present
results showed that having one or more pregnancies during
adolescence was associated with weighing 2 kg more at age
19 y than not having any pregnancies.
Changes in BMI were a result of the combination of rela-
tive gains in weight and height. Girls with a single pregnancy
gained more weight but did not gain more height than those
who did not get pregnant. On the other hand, those with 2 or
more pregnancies gained about the same amount of weight as
those with a single pregnancy but gained considerably less
height. Thus, BMI change was linearly related to the number
of pregnancies.
In adult women, the findings are inconclusive with respect
to excess weight gain and risk of becoming overweight asso-
ciated with pregnancy. In adolescents, maternal growth ac-
counts for some of the weight increase during pregnancy (26).
Although most of the pubertal gains in stature in girls are
around the time of peak height velocity, and, prior to men-
arche, growth continues during late adolescence, small and
positive increments in stature were observed in adolescents
during pregnancy (7,8,23). Hediger and colleagues (8) have
documented that adolescent pregnancy is associated with
larger gestational weight gains. They found that, after week 28,
growing adolescents on average failed to lose fat and tended to
continue to accrue fat in their upper arm fat area. In contrast,
mature women and the nongrowing adolescents both lost fat
from their upper arms and back. Thus, continued subcutaneous
fat accrual after week 28, at a time when reductions in the
subcutaneous fat mass are expected, appears to be character-
istic of maternal growth during pregnancy. In the present
study, linear association between BMI gain and pregnancy was
related to both reduced height and increased weight. In an-
other Brazilian study, Sichieri and co-workers (27) used 1996
Demographic Health Survey data on 2297 women aged 20 to
45 y and showed that short stature was associated with an
increased risk of weight gain with pregnancy in the developed
areas of Brazil.
Some limitations must be considered when interpreting
these data. Only 12 adolescents had had 2 or more pregnan-
cies, and the possibility of lack of statistical power must be
considered, particularly for the analyses of weight change.
Also, the 1997 follow-up was restricted to 27% of the urban
area, and 28% of the original cohort was lost to follow-up. If
we had been able to include all of the female adolescents who
were born in 1982 (2876 girls) in this follow-up, the number
of adolescents who were pregnant and the power of the study
would have been greater, which may have permitted us to
identify additional associations with weight change.
In summary, the negative effect of pregnancy on height
change was restricted to adolescents who had 2 or more
pregnancies, while greater gains in weight and BMI were
observed among all of those who became pregnant. Both by
reducing adult height and increasing overweight, teenage
pregnancy may contribute to the prevalence of obesity and the
incidence of chronic diseases from adolescence to adulthood
(28,29). Using existing data on the distribution of BMI among
adult Pelotas women (30), it is possible to estimate the impact
of teenage pregnancy. Currently, 54.2% of Pelotas women are
overweight. If one assumes a normal distribution of BMI, a
shift of 1.58 kg/m2 in a group of this population and, conse-
quently, a shift of 0.29 Z score, would increase the prevalence
80 GIGANTE ET AL.
of overweight in this group to 65.5%. So, efforts to reduce
teenage pregnancy may also contribute to prevent overweight,
taking into account the growing epidemic of obesity in low-
and middle-income countries, particularly among women.
LITERATURE CITED
1. Popkin, B. M. & Doak, C. M. (1998) The obesity epidemic is a world-
wide phenomenon. Nutr. Rev. 56: 106 –114.
2. Schneider, D. (2000) International trends in adolescent nutrition. Soc.
Sci. Med. 51: 955–967.
3. Lytle, L. A. (2002) Nutritional issues for adolescents. J. Am. Diet.
Assoc. 102: S8 –12.
4. Parsons, T. J., Power, C., Logan, S. & Summerbell, C. D. (1999) Child-
hood predictors of adult obesity: a systematic review. Int. J. Obes. Relat. Metab.
Disord. 23: S1–S107.
5. IBGE. Censo Demografico 2000: fecundidade e mortalidade infantil. Re-
sultados preliminares. IBGE, Rio de Janeiro, Brazil.
6. Scholl, T. O., Hediger, M. L., Schall, J. I., Khoo, C. S. & Fisher, R. L.
(1994) Maternal growth during pregnancy and the competition for nutrients.
Am. J. Clin. Nutr. 60: 183–188.
7. Frisancho, A. R. (1997) Reduction of birth weight among infants born
to adolescents: maternal-fetal growth competition. Ann. N.Y. Acad. Sci. 817:
272–280.
8. Hediger, M. L., Scholl, T. O. & Schall, J. I. (1997) Implications of the
Camden Study of adolescent pregnancy: interactions among maternal growth,
nutritional status, and body composition. Ann. N.Y. Acad. Sci. 817: 281–291.
9. Scholl, T. O., Hediger, M. L. & Schall, J. I. (1997) Maternal growth and
fetal growth: pregnancy course and outcome in the Camden Study. Ann. N.Y.
Acad. Sci. 817: 292–301.
10. Gigante, D. P., Victora, C. G., Gonçalves, H., Lima, R. C., Barros, F. C. &
Rasmussen, K. M. (2004) Risk factors of childbearing during adolescence in a
population-based birth cohort in southern Brazil. Rev. Panam. Salud Publica 16:
1–10.
11. Victora, C. G., Barros, F. C. & Vaughan, J. P. (1989) . Epidemiologia da
Desigualdade. Hucitec, Sao Paulo, Brazil.
12. Barros, F. C., Victora, C. G. & Vaughan, J. P. (1990) The Pelotas birth
cohort study, 1982–1987. Strategies for following up 6,000 children in a devel-
oping country. Paediatr. Perinat. Epidemiol. 4: 267–282.
13. Barros, F. C., Victora, C. G., Vaughan, J. P., Tomasi, E., Horta, B. L.,
Cesar, J. A., Menezes, A.M.B., Halpern, R., Post, C. L. & Garcia, M. M. (2001)
The epidemiologic transition in maternal and child health in a Brazilian city,
1982–1993: a comparison of two population-based cohorts. Paediatr. Perinat.
Epidemiol. 15: 4 –11.
14. Victora, C. G, Barros, F. C., Lima, R. C., Behague, D., Gonçalves, H.,
Horta, B. L., Gigante, D. P. & Vaughan, J. P. (2003) The Pelotas (Brazil) Birth
Cohort Study, 1982–2001. Cad. Saude Publica 19: 1241–1256.
15. Cole, T. J., Bellizzi, M. C., Flegal, K. M. & Dietz, W. H. (2000) Estab-
lishing a standard definition for child overweight and obesity worldwide: interna-
tional survey. Br. Med. J. 320: 1240 –1243.
16. [WHO] World Health Organization (1995) Physical status: the use and
interpretation of anthropometry. Technical Report Series 854. WHO, Geneva,
Switzerland.
17. Kleinbaum, D. G., Kupper, L. L., Muleer, K. E. & Nizam, A. (1998)
Applied Regression Analysis and Other Multivariate Methods. 3rd ed. Duxbury
Press, Pacific Grove, CA.
18. Harpham, T., Huttly, S., Wilson, I. & De Wet, T. (2003) Linking public
issues with private troubles: panel studies in developing countries. J. Int. Dev. 15:
1–11.
19. Lederman, S. A. (1997) Nutritional support for the pregnant adoles-
cent. Ann. N.Y. Acad. Sci. 817: 304 –312.
20. Lawlor, D. A. & Shaw, M. (2002) Too much too young? Teenage
pregnancy is not a public health problem. Int. J. Epidemiol. 31: 552–554.
21. Smith, S. (2002) Too much too young? In Nepal more a case of too
little, too young. Int. J. Epidemiol. 31: 557–558.
22. Scholl, T. O. & Hediger, M. L. (1993) A review of the epidemiology of
nutrition and adolescent pregnancy: maternal growth during pregnancy and its
effect on the fetus. J. Am. Coll. Nutr. 12: 101–107.
23. Garn, S. M., Lavelle, M., Pesick, S. D. & Ridella, S. A. (1984) Are
pregnant teenagers still in rapid growth? Am. J. Dis. Child. 138: 32–34.
24. Scholl, T. O., Hediger, M. L., Schall, J. I., Mead, J. P. & Fischer, R. L.
(1995) Maternal growth during adolescent pregnancy. J. Am. Med. Assoc. 274:
26 –27.
25. Olson, C. M., Strawderman, M. S., Hinton, P. S. & Pearson, T. A. (2003)
Downloaded from jn.nutrition.org by on October 17, 2008
Gestational weight gain and postpartum behaviors associated with weight
change from early pregnancy to 1 y postpartum. Int. J. Obes. Relat. Metab.
Disord. 27: 117–127.
26. Gunderson, E. P. & Abrams, B. (2000) Epidemiology of gestational
weight gain and body weight changes after pregnancy. Epidemiol. Rev. 22:
261–274.
27. Sichieri, R., Silva, C.V.C. & Moura, A. S. (2003) Combined effect of
short stature and socioeconomic status on body mass index and weight gain
during reproductive age in Brazilian women. Braz. J. Med. Biol. Res. 36: 1319 –
1325.
28. Allebeck, P. & Bergh, C. (1992) Height, body mass index and mortality:
do social factors explain the association? Public Health 106: 375–382.
29. Must, A., Jacques, P. F., Dallal, G. E., Bajema, C. J. & Dietz, W. H. (1992)
Long-term morbidity and mortality of overweight adolescents. A follow-up of the
Harvard Growth Study of 1922 to 1935. N. Engl. J. Med. 327: 1350 –1355.
30. Gigante, D. P., Barros, F. C., Post, C. L. & Olinto, M. T. (1997) Prev-
alence and risk factors of obesity in adults. Rev. Saude Publica 31: 236 –246.