Keywords
case-controlled studies - developing countries - low birth weight - risk factors -
term birth
Palavras-chave
estudos de caso-controle - países em desenvolvimento - baixo peso ao nascer - fatores
de risco - termo de nascimento
Introduction
The World Health Organization (WHO) defines low birth weight as weight at birth lower
than 2,500 g.[1] However, birth weight is determined by two critical considerations: gestational
age at delivery and the rate of fetal growth.[2] Consequently, low weight determining factors may differ between a preterm and a
term neonate because, in the former, low weight is usually explained by prematurity,
while in the latter, it is result of intrinsic and/or extrinsic factors that impact
on developmental potential.[1]
[2] Therefore, it comes as no surprise that low birth weight neonates have a worse prognosis
in terms of survival and neural development.[3]
According to The United Nations International Children's Emergency Fund (UNICEF),
more than 20 million infants are born with low weight in the world, accounting for
15% of all births. Of these, more than 95% are born in middle- and low-income countries,
with those in Asia, Africa and Latin America being the most frequently affected.[4] Prevalence in these countries is twice as high as the ones observed in developed
countries, reflecting the inequities faced by pregnant women in those regions of the
world where adverse conditions, such as malnutrition, poor weight gain, anemia or
pregnancy-related disorders, like hypertension, possibly explain the observed frequencies.[1]
[5]
Preventing low birth weight is a public health priority, and one of the goals for
the new millennium.[4] Reducing the frequency of low birth weight pregnancies could have a positive impact
on infant mortality in middle- and low-income countries, which are the ones with the
largest shortage of resources required for the care of these neonates.[4] Consequently, efforts aimed at identifying risk factors associated with this condition
are mandatory.[6] Knowledge of the factors that influence low birth weight will contribute to the
timely identification and early intervention in pregnant women at risk.[7] Therefore, the objective of this study was to identify maternal factors associated
with the presence of low birth weight in term neonates born in 2015 and 2016 in a
high complex institution in the city of Neiva, Colombia.
Methods
This was a retrospective, analytical case-controlled study conducted at Hospital Universitario
Hernando Moncaleano Perdomo, a high complexity institution located in the city of
Neiva, serving the population of southwestern Colombia.
The subjects included were women with term gestations and singleton live fetuses,
who were treated at the participating institution between January 2015 and October
2016. Patients with any existing diseases before pregnancy (such as hypertensive vascular
disease, nephropathy, diabetes mellitus, thrombophilia, heart, neurodegenerative,
cancer or autoimmune diseases), coming from a different geographic area, with a gestation
resulting from assisted reproduction, or with a confirmed diagnosis of major fetal
abnormality or aneuploidy were excluded.
Cases were defined as live neonates with a weight at birth under 2,500 g and a gestational
age of 37 weeks of gestation or more. Gestational age was estimated based on the last
menstrual period or ultrasound. Term neonates weighing 2,500 g or more were considered
controls. The cases were identified using entries in the database of the epidemiological
surveillance system corresponding to low birth weight, and the information was compared
with the newborn vital statistics registry. In Colombia, the birth of a neonate with
low weight is a public health event and reporting is mandatory. Controls were selected
from the vital statistics database of the participating institution and the quality
of the data was verified by means of a clinical record review. The cases and controls
were selected by random sampling until the required sample size was completed.
Sample size estimation was based on the approach suggested by Freeman, consisting
of the use of the event of interest by variable. In this way, the sample size for
a non-conditioned logistic regression was determined to be 10 * (k + 1), in which
k is the number of study variables.[8] According to this principle, and given that for this study, a priori, research into the potential association between low birth weight and maternal age,
weight gain, educational level, socioeconomic and marital status, social security,
gestational age at the start of prenatal care, and the presence of maternal disease
during gestation had been proposed, at least 90 cases and an equal number of controls
were required (n = 180). However, it was decided to select 2 controls for every case to increase the
power. Thus, the required sample size for the study was 270 participants.
The data were analyzed using the Stata software package, version 15 (StataCorp. College
Station, TX, USA). Descriptive statistics were applied to clinical and sociodemographic
variables. The central trend and scatter were estimated for continuous data, and proportions
and frequency measurements were used for qualitative data. The frequency and distribution
were examined for categorical variables, and a normal distribution and variance homogeneity
were analyzed for continuous variables. For the categorical variables, differences
between cases and controls were tested using the Fisher exact test or the Pearson
Chi-squared test. Continuous variables were compared using the Mann-Whitney test,
given the absence of a normal distribution. A univariate logistic regression was applied
to assess the association between candidate variables and low birth weight.
A multivariate logistic model was built to incorporate the clinical and sociodemographic
variables mentioned above. A non-conditional logistic regression was performed to
adjust for the presence of potential confounding factors, the predictive ability of
the model was estimated, and the goodness of the adjustment was assessed using McFadden
R2. However, to identify the most parsimonious model, a second analysis was performed
using the stepwise backward approach, to which the Bonferroni correction was applied,
thus adjusting the significance level.[9] For the second model, the independent variables associated with the outcome of interest
were preserved, with a significance level lower than 0.005.[9] Confidence intervals were estimated, and adjusted and non-adjusted odds ratios (aOR
and OR) are presented as the association measure. The study protocol was approved
by the Ethics Committee of the participating institution (Reference 009–006).
Results
There were 4,882 term deliveries in the participating institution during the study
period. Of this total, 3.0% were neonates with low birth weight for gestational age.
Once the target population was identified, the cases and controls were selected by
random sampling until the required sample size was completed. For each selected case
and control, we verify the inclusion and exclusion criteria before their incorporation
into study. Finally, the study population was then assembled and consisted of 90 cases
and 180 controls.
In terms of the characteristics of the population analyzed, the mean age was 23 years,
and there was a predominance of secondary education level, low socioeconomic bracket,
free union as marital status, affiliation with the subsidized healthcare regime, and
urban place of residence. Regarding clinical characteristics, 54.0% of the women were
multiparous and 61.4% had attended 5 or more prenatal care visits; 62.9% were vaginal
deliveries, the mean gestational age at the time of delivery was 38.4 weeks, and the
mean birth weight was 2,965 g (SD ± 552 g). [Table 1] summarizes the baseline characteristics of the population by group (cases or controls).
Table 1
Description of the sociodemographic and clinical characteristics of cases and controls
|
Variable Mean
|
Cases
(n = 90)
|
Controls
(n = 180)
|
p value
|
|
Mean
|
Range
|
Mean
|
Range
|
|
Age
|
22.0
|
(14–42)
|
24.1
|
(13–45)
|
p = 0.00[a]
|
|
Variable measured
|
n (%)
|
n (%)
|
p value
|
|
Educational level
|
p = 0.29[b]
|
|
None
|
0 (0.00)
|
1(0.56)
|
|
Primary
|
20 (22.22)
|
30 (16.67)
|
|
Secondary
|
61 (67.78)
|
138 (76.6)
|
|
Technical/University
|
9 (10.00)
|
11 (6.11)
|
|
Socioeconomic status
|
p = 0.00[b]
|
|
Low
|
58 (64.44)
|
139 (77.22)
|
|
Medium
|
27 (30.00)
|
40 (22.22)
|
|
High
|
5 (5.56)
|
1 (0.56)
|
|
Marital status
|
p = 0.89[c]
|
|
Single
|
39 (43.33)
|
79 (43.89)
|
|
Free union
|
43 (47.78)
|
88 (48.89)
|
|
Married
|
8 (8.89)
|
13 (7.22)
|
|
Social security
|
p = 0.52[b]
|
|
Subsidized
|
83 (92.22)
|
154 (85.56)
|
|
Contributive/Special
|
6 (6.67)
|
23 (12.78)
|
|
No payment capability
|
1 (1.11)
|
3 (1.67)
|
|
Place of residence
|
p = 0.37[c]
|
|
Urban
|
64 (71.11)
|
137 (76.11)
|
|
Rural
|
26 (28.89)
|
43 (23.89)
|
|
Gestational age at the start of prenatal care
|
p = 0.00[c]
|
|
First trimester
|
30 (33.33)
|
86 (47.78)
|
|
Second trimester
|
16 (17.78)
|
48 (26.67)
|
|
Third trimester
|
20 (22.22)
|
39 (21.67)
|
|
No prenatal care
|
24 (26.67)
|
7 (3.89)
|
|
Disease during pregnancy
|
p = 0.00[c]
|
|
No
|
58 (64.44)
|
143 (79.44)
|
|
Yes
|
32 (35.56)
|
37 (20.56)
|
|
Sex of the newborn
|
p = 0.79[c]
|
|
Female
|
48 (53.33)
|
93 (51.67)
|
|
Male
|
42 (46.67)
|
87 (48.33)
|
a Mann-Whitney test for mean differences.
b Fisher exact test.
c Pearson Chi-squared test.
The two groups were similar, except in terms of socioeconomic status, gestational
age at the start of prenatal care, and the presence of disease during pregnancy. [Table 2] summarizes the frequency and type of disease by group (cases or controls). Maternal
age was significantly different between the groups, although the difference observed
was not clinically relevant. For the neonates (data not shown), the mean gestational
age at the time of birth was 37.9 weeks for the cases and 39.0 weeks for the controls,
while the average weight was 2,328 g (SD ± 166 g) for the cases and 3,282 g (SD ± 371
g) for the controls. Of the low-weight neonates, 78.2% were admitted to the kangaroo
program and 3.2% to the neonatal intensive care unit.
Table 2
Frequency and type of disease during pregnancy by group (cases or controls)
|
Disease during pregnancy
|
Cases (n = 90)
|
Controls (n = 180)
|
p-value
|
|
n (%)
|
n (%)
|
|
No
|
58 (64.4)
|
143 (79.4)
|
p
= 0.00
[a]
|
|
Yes
|
32 (35.5)
|
37 (20.5)
|
|
By type
|
|
|
|
Hypertensive disorder
|
19 (21.1)
|
17 (9.4)
|
|
Perinatal infection
|
6 (6.6)
|
7 (3.8)
|
|
Endocrine disease
|
2 (2.2)
|
5 (2.7)
|
|
Placental-amniotic fluid disease
|
4 (4.4)
|
2 (1.1)
|
|
Anemia
|
0 (0.0)
|
4 (2.2)
|
|
Neurological condition
|
1 (1.1)
|
1 (0.05)
|
|
Pulmonary disease
|
0 (0.0)
|
1 (0.05)
|
a Pearson Chi-squared test.
The univariate logistic regression revealed that maternal age (OR 0.94, 95% CI 0.90–0.98)
and weight gain during pregnancy (OR 0.77, 95% CI 0.70–0.84) behaved as protective
factors. The risk was also significantly lower depending on the educational level
of the mother (technical/university education OR 0.10, 95% CI 0.01–0.83). On the other
hand, the absence of prenatal care (OR 9.82, 95% CI 3.84–25.13) and the presence of
maternal disease during pregnancy (OR 2.13, 95% CI 1.21–3.74) behaved as risk factors.
[Table 3] shows the results of the logistic regression, together with their respective aOR
and CI.
Table 3
Multivariate analysis of maternal factors associated with low birth weight
|
Variables
|
First model[a]
aOR (95% CI)[b]
|
Second model[a]
aOR (95% CI)
|
|
Maternal age
|
0.91 (0.86–.96)
|
3[c]
|
|
Weight gain
|
0.77 (0.69–0.86)
|
0.77 (0.70–0.85)
|
|
Absence of prenatal care
|
10.67 (3.49–32.64)
|
8.20 (3.22–20.87)
|
Abbreviations: aOR, adjusted odds ratio; 95% CI, 95% confidence interval.
a Adjusted for variables: level of schooling, socioeconomic bracket, marital status,
social security and the presence of maternal disease.
b Adjusted odds ratios together with the 95% confidence intervals;
c Variable removed from the model: p of 0.01 but greater than 0.005.
Following this exploration, a multivariate analysis was performed with the main goal
of identifying the clinical or sociodemographic characteristics that could be linked
to the presence of low birth weight. An initial model was built, ratifying the role
played by maternal age, low weight gain and absence of prenatal care. When the goodness
of the adjustment was assessed, the McFadden R2 was 0.27, with a predictive capacity of 80.3%, reflecting a high percentage of correctness.[10] However, when the second analysis was performed, only the absence of prenatal care
and of adequate weight gain continued to be statistically significant. This time,
the McFadden R2 was 0.17, with a predictive capacity of 72.4%, showing an acceptable percentage of
correctness.[10]
Discussion
Low birth weight is one of the most important determinants of infant morbidity and
mortality because it increases the frequency of adverse perinatal outcomes.[11] Identification of factors that may have an impact on potential fetal growth is an
excellent opportunity to have an impact on the health conditions of the population.[12]
There were 4,882 term live births during the observation period, of which 3.0% were
low birth weight neonates. The population consisted mainly of young, single mothers
with low socioeconomic and education levels, living in urban areas. Only 43% of the
women started prenatal care during the first trimester of gestation. In our study,
the prevalence of low birth weight was similar to that reported in other Latin American
countries, but substantially lower than the one described for countries in Southeast
Asia.[13]
[14]
[15]
[16]
[17]
[18]
[19] These differences could be explained, at least in part, by the rate of fetal growth,
genetic factors and the presence of other extrinsic circumstances not related to pregnancy.[7]
[20]
On the other hand, regarding the multivariate analysis, it revealed that maternal
weight gain (aOR 0.77, 95% CI 0.70–0.85) behaves as a protective factor, while the
absence of prenatal care (aOR 8.20, 95% CI 3.22–20.87) increases the probability of
an unfavorable outcome. Our findings are similar to those documented in the literature.
Observational studies have shown the association between maternal weight gain and
neonatal birth weight.[7] For example, it is known that infants born to mothers with poor weight gain are
at a higher risk of being small for gestational age, while those born to mothers with
substantial weight gain have a higher probability of being large.[21] This association is consistent in low, middle and high-income countries alike.[22]
Regarding poor prenatal care or absence thereof, the observed association emerges
in populations from middle and low-income countries, and it is not completely clear
in developed countries.[7]
[23]
[24]
[25]
[26]
[27] The potential explanation is that early and adequate prenatal care could be of greater
benefit in women with less favorable conditions, where the implementation of this
intervention could help address harmful behaviors that affect fetal growth (for example,
smoking), early detection and treatment of diseases affecting gestation (such as anemia,
malnutrition), and promote healthy lifestyle habits that can have a positive impact
on the fetal environment.[20]
[28]
Finally, although maternal age was eliminated as a variable in the second model, the
observed association seems plausible, is clinically relevant, and has been documented
in other studies.[29] This association could be explained by the conditions of inequity and disadvantage
faced by teen mothers when compared with older women. Therefore, not surprisingly,
the risk of low birth weight decreases as a function of older maternal age, given
that older women may have better socioeconomic conditions and find themselves less
at a social disadvantage.[30] It is no secret that racial segregation and deprivation are associated with low
birth weight.[31] Notwithstanding, it needs to be said that this conclusion must be interpreted cautiously,
given that the finding could be the result of multiple comparisons.
This study has some strengths, starting with the appropriate and widely accepted definition
of the cases, which are considered representative of the study population because
all the eligible neonates with the outcome of interest were included. In Colombia,
reporting of low birth weight events is mandatory, hence there is a low probability
of having missed a case.[32]
[33] On the other hand, given matched and independent data recording, there is confidence
regarding the reliability of the data and a low probability of error in the definition
of cases and controls.
Case-controlled comparability was achieved by means of the design and the analysis.[34] In the design, stringent inclusion and exclusion criteria were used in an attempt
to arrive at a relatively homogenous population. In the analysis, a mathematical model
was used to adjust for the presence of confounding factors. Although it is true that
excluding the presence of residual confounding factors is not feasible, given the
nature of the design, the importance and number of the variables considered allows,
in some way, for acceptable case-controlled comparability.[35] Finally, the other strength of this study is that exposure was proven through entry
in the clinical record and in the vital statistics database of the Health Secretariat,
reducing the risk of poor classification.[35]
This study also has weaknesses. The subjects used as controls came from the same institution
and not from the community, making the study prone to selection bias (Berkson fallacy)
because, given the origin of the controls, they could be more prone to having certain
factors associated with low birth weight, leading to potential distortions for some
exposure-disease associations.[35] Sample size is yet another weakness of the study. Although a design was developed
a priori for estimating sample size, the critical assessment of the confidence intervals points
to some degree of inaccuracy.[36]
Despite its limitations, this study has many practical implications. First, it highlights
the need to ensure adequate weight gain during pregnancy. This is important because
adequate weight gain during gestation behaves as an indirect indicator of good nutritional
status of the pregnant woman. Secondly, the association observed between poor prenatal
care and low birth weight should prompt timely and adequate access to medical care
during gestation, especially for vulnerable populations. Medical care during the reproductive
period is a valuable opportunity to have a positive impact on health conditions by
means of education regarding healthy lifestyle and to ensure early detection and treatment
of diseases affecting gestation. Finally, the role of an unfavorable socioeconomic
environment could play out in the association observed between maternal age and low
birth weight. Hence the need for highlighting the relevance of providing timely and
equitable access to quality healthcare systems.
Conclusion
Based on the findings of this study, low maternal weight gain and untimely initiation
of prenatal care are some of the factors known to be associated with the presence
of low birth weight in term neonates. On the other hand, the role of maternal age
could also be relevant, considering that this association reflects, at least in part,
the conditions of poverty and deprivation of the study population. This study, despite
its many strengths, has limitations. Therefore, further studies are required to undertake
a more extensive evaluation of the maternal factors associated with the presence of
low birth weight among term neonates.
