Role of breastfeeding cessation in mediating
the relationship between maternal HIV disease
stage and increased child mortality among
HIV-exposed uninfected children
Matthew P Fox,1* Daniel R Brooks,2Louise Kuhn,3Grace Aldrovandi,4Moses Sinkala,5
Chipepo Kankasa,6Robert Horsburgh2and Donald M Thea1
28 October 2008
Background Maternal CD4 count predicts child mortality in HIV-uninfected
children born to HIV-infected women.
To explore the mediating role of breastfeeding cessation in this
relationship, we compared marginal structural models of maternal
CD4 count on child death with and without adjustment for
In crude analyses, children of mothers with CD4<200 during
pregnancy were 3.2 times more likely to die by 18 months
(CI 1.3–8.1) as children whose mothers had CD44500. Earlier
breastfeeding cessation was also associated with low CD4 (HR 1.8;
CI 1.2–2.7). After adjusting for breastfeeding and low birth weight
using a marginal structural model, the low CD4 count-child mortal-
ity association through 18 months was reduced 17%. The change
was overestimated using a traditional Cox proportional hazards
model (35% reduction in HR from 3.4 to 2.5).
Conclusions Our analysis suggests that only a small part of the effect of low vs
high CD4 count on child mortality through 18 months is mediated
through breastfeeding cessation. Our results must be taken into
account when deciding whether or not to recommend breastfeeding
for infants of HIV-infected mothers.
HIV, infant mortality, direct and indirect effects, marginal structural
model, Africa, breastfeeding
Most infants born to HIV-infected mothers escape
HIV infection during childhood. Evidence suggests
that HIV-exposed (i.e. born to an HIV-infected
woman) but uninfected infants have higher morbidity
and mortality than infants born to HIV-uninfected
mothers.1–5Some hypothesize that the increased risk
reflects behavioural factors affecting a mother’s ability
to care for her sick child, though there is currently
little evidence to support this. Several studies have
suggested an HIV-infected mother’s degree of immu-
nosuppression adversely affects survival in her unin-
fected infant.6–8This was established more firmly in
a cohort of HIV-exposed but confirmed uninfected
infants who were significantly more likely to die by
* Corresponding author. Center for International Health and
Development, 85 East Concord St 5th floor, Boston, MA
02118, USA. E-mail: email@example.com
1Center for International Health and Development, Boston
University, Boston, MA, USA.
2Department of Epidemiology, Boston University, Boston, MA,
3Sergievsky Center and Department of Epidemiology, Mailman
School of Public Health, Columbia University, NY, USA.
4Childrens Hospital of Los Angeles, Los Angeles, CA, USA.
5Lusaka District Health Management Team, Lusaka, Zambia.
6University Teaching Hospital, University of Zambia, Lusaka,
Published by Oxford University Press on behalf of the International Epidemiological Association
? The Author 2008; all rights reserved. Advance Access publication 30 November 2008
International Journal of Epidemiology 2009;38:569–576
4 months of age if their mother’s CD4 count near
birth was <350.9Given the apparent significance of
maternal health in the survival of her uninfected
infant, a logical next step is to investigate potential
causal mechanisms through which maternal CD4
count could affect infant survival in order to seek
ways to mitigate the harmful impact on the infant.
One possible explanation for increased mortality
among HIV-exposed infants is that the general debility
associated with advanced maternal AIDS leads to a
decrease or complete cessation of breastfeeding.10,11
Numerous studies have established the immunological
and nutritive benefits of breast milk, including the
reduction of both major morbidity12,13and mortality.14
While likely underestimating the total effect in
developing countries, odds ratios for mortality compar-
ing breastfeeding to non-breastfeeding ranged from
4.2 during the first months of life to 1.4 in the 9th
though 11th months of life in a pooled analysis.14
Given the link between breastfeeding duration and
infant health, it is possible that increased infant mor-
tality associated with a lower maternal CD4 count may
be mediated through breastfeeding.
This analysis uses longitudinal data collected from
a well-defined cohort of pregnant women and their
infants enrolled in a breastfeeding intervention trial
in Zambia to evaluate the mediating role of breast-
maternal CD4 count and child mortality.
Materials and methods
Study design and population
The Zambia Exclusive Breastfeeding Study (ZEBS) was
a randomized controlled trial designed to assess the
impact of abrupt weaning at 4 months on overall HIV-
free survival. Details of the study have been described
September 2004, 1435 HIV-infected pregnant women,
were enrolled into a randomized controlled trial of a
breastfeeding intervention to prevent mother-to-child
transmission of HIV at two clinics in Lusaka, Zambia
at antenatal clinic visits. At the time of this study
highly active antiretroviral therapy (HAART) was not
yet available in Zambia. ZEBS was approved by insti-
tutional review boards at Boston University, Childrens
Hospital of Los Angeles, Columbia University and the
University of Zambia Research Ethics Committee.
All enrolled women were intensively counselled
to exclusively breastfeed through 4 months of age. At
1 month post-delivery, women were randomized to
either: (i) abrupt breastfeeding cessation at 4 months;
or (ii) exclusive breastfeeding through 6 months
with weaning as they normally would. Mothers
and infants were followed for 2 years. The current
analysis includes all live born, singleton infants in
ZEBS randomized to usual breastfeeding behaviour
(i.e. the control group who were counselled to stop
breastfeeding when they felt it was appropriate) who
did not test HIV-positive at any time during follow-up.
Thus for all mothers included in the current analysis,
breastfeeding behaviour was not under the control of
the investigator and was not influenced by the study
team. This analysis is limited to the first 18 months of
follow-up as there were only two child deaths after
At baseline all mothers had blood drawn for CD4þ
T-cell lymphocyte counts (FACSCount system, BD
Immunocytometry Systems, San Jose, CA). Maternal
disease stage as indicated by CD4 count during
pregnancy at enrollment was categorized into three
groups (<200, 200–499, 4500).
All mothers were scheduled for clinic visits at 1
week, then monthly through 6 months, then every
3 months. For any child who died, the date and
circumstances of the death were elicited from the
mother or family members and hospital records when
available. Infant follow-up time was defined as the
period from birth to the date of the earliest of:
(i) completion of 18 months of follow-up; (ii) infant
death; and (iii) last visit if lost to follow-up.
We defined breastfeeding cessation to have occurred
at the child’s actual as reported by the mother on her
first report of stopping all breastfeeding. We assumed
children who died were breastfed up to their date of
death unless records specifically reported that breast-
feeding stopped prior to the start of the illness
immediately preceding the child’s death. Breastfeed-
ing time was defined as the period from birth to the
date of the earliest of: (i) first visit a mother reported
stopping all breastfeeding; (ii) infant or maternal
death; (iii) completion of 18 months; and (iv) last
visit before loss to follow-up.
To assess the mediating role of breastfeeding in the
we first examined the relationship between maternal
CD4 count and child mortality. Next, we explored the
association between maternal CD4 count and breast-
feeding cessation. Finally, we created a model with
child mortality as the outcome and maternal CD4
count as the exposure and then added breastfeeding
cessation as a covariate to look for changes in the
CD4-child mortality relationship when adjusting for
We calculated crude rates of child mortality and
breastfeeding cessation (per 100 person-years) and
incidence rate ratios within each CD4 category. To
identify confounders for multivariable proportional
hazards models, we used a two step process. We first
conducted univariate comparisons by child vital status
of each variable that could plausibly confound the
CD4-child mortality relationship. For those variables
with a chi-square test of independence P<0.20, we
then used a modified forward step-wise procedure18
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
to identify covariates that changed the hazard ratio
(HR) for maternal CD4 count. The covariate produ-
cing the largest change410% was retained and a new
percentage change was calculated with each remain-
ing variable until no remaining variable resulted in
a change 410%. We did not adjust for any factor
that was affected by maternal CD4 count, such as
We hypothesize that some of the effect of maternal
CD4 count on infant mortality is mediated through
breastfeeding cessation (Figure 1); that is, mothers
with lower CD4 counts near birth have a reduced
ability to adequately maintain breastfeeding, which in
turn leads to increased child mortality. To test this, we
(p. 184). We created two distinct but similar regression
models, both of which had maternal CD4 count as
the independent variable and child mortality as the
dependent variable. Model 1 also included confounders
of the CD4 count-child mortality relationship. If there
were no bias or random error in our study, model 1
would reflect the total causal effect of maternal CD4
count on infant mortality (some of which we believe
is mediated through breastfeeding cessation).
Model 2 included the covariates in model 1, but also
included breastfeeding cessation as a time dependent
covariate and any confounders of the breastfeeding–
contains a variable affected by the exposure (i.e. we
hypothesize women with lower CD4 counts breastfeed
for shorter duration), it represents the effect of
maternal CD4 count on child mortality remaining
after controlling for breastfeeding. What remained
would be the effect of maternal CD4 count on child
mortality through other pathways (e.g. low birth
Figure 1 shows the proposed causal pathways for
the effect of maternal CD4 count on child survival.
The dotted line represents the potential indirect effect
of maternal CD4 count on infant survival mediated
through duration of breastfeeding. Similarly, an
indirect effect is likely mediated through LBW (e.g.
mothers with low CD4 counts are more likely to have
a LBW child, which likely, in turn affects the child’s
risk of death). Because birth weight is affected by
maternal CD4 count, including LBW as a covariate in
a standard regression model would provide biased
results. However, we believe LBW is also a confounder
of the breastfeeding cessation–child mortality rela-
tionship.20LBW may cause changes in breastfeeding
duration and independently increase infant mortality.
Thus, any model intended to identify the indirect
effect of breastfeeding cessation that did not adjust
for LBW would also be biased.
To adjust for LBW, we employed a proportional
hazards marginal structural model (MSM).21,22MSMs
weight each subject in the dataset proportional to the
inverse of the probability of receiving the exposure
they actually received given their covariates. In this
analysis this would be the probability of being
breastfed the duration a child was actually breastfed
given their birth weight and other covariates. Because
we weight subjects we use robust standard errors to
prevent unduly small confidence intervals (CI).
Changes in the estimate of effect of CD4 count on
infant mortality from models 1 to 2 are expressed as
where HRxis the HR from model x. Any decrease in
HR from models 1 to 2 with adjustment for breast-
feeding would be consistent with a portion of the
effect of maternal disease stage on child mortality
being mediated through breastfeeding cessation.
Of the 1435 ZEBS children, 357 met the eligibility
criteria for this analysis. Two children were excluded
because we had no data on the mother’s enrollment
CD4 count. Maternal CD4 counts largely fell in the
200–500 range (54.6%, N¼194) with fewer in the
4500 category (27.6% N¼98) or <200 CD4 range
(17.7% N¼63). There were 14 maternal deaths (3.9%)
during follow-up and 45 infant deaths (12.7%) of
which 17 were in the first 6 months. The median
follow-up time was 548 days. By Kaplan-Meier esti-
mate the median duration of breastfeeding was 487
days, but was considerably shorter in the low CD4
group (372 days) than either the intermediate or high
groups (487 and 517 days, respectively).
Crude analyses revealed few independent predictors
score predicted lower risk of child death (RR 0.55;
Figure 1 Causal diagram of the effect of maternal
CD4 count on child mortality through breastfeeding
cessation. U represents other unknown variables
THE MEDIATING ROLE OF BREASTFEEDING CESSATION
95% CI 0.30–1.0), while food insecurity in the prior
month (RR 1.9; 95% CI 1.1–3.3), LBW (RR 3.1;
95% CI 1.7–5.6) and maternal death (RR 3.0; 95%
CI 1.4–6.5) increased the likelihood of child death.
Table 2 shows incidence rates of infant mortality by
time period stratified by CD4 group. Children whose
mothers had low CD4 counts were at markedly
increased risk of death compared with those whose
mothers had high CD4 count through 18 months
(HR 3.2; 95% CI 1.3–8.1). Over 18 months, earlier
breastfeeding cessation was clearly associated with
advanced maternal disease stage (HR 1.8; 95% CI
1.2–2.7) (Table 2). Little difference was seen between
the intermediate and high CD4 groups.
The proportional hazards marginal structural model
regression (Table 3) shows the effect of adjustment for
SES on child mortality up to 18 months. Model 1 is not
adjusted for cessation of breastfeeding or LBW and
represents the estimate of the total effect of CD4 count
on child mortality. After adjusting for SES, infants of
mothers with CD4 counts below 200 were at substan-
tially increased risk of death through 18 months
compared with those with mothers with CD4 above
500 (HR 3.4; 95% CI 1.3–8.4). Little difference was seen
between the intermediate and high CD4 groups.
The second column of Table 3 (model 2) shows the
analysis further adjusted for breastfeeding cessation
and LBW. This model estimates the effects of CD4
count on child mortality not mediated through breast-
feeding cessation. In this model breastfeeding cessa-
18 months (HR 2.0; 95% CI 1.0–3.7).
The percentage change in the hazard ratio from
models 1 to 2 is shown in column 3. This value rep-
resents an estimate of the indirect effect of maternal
CD4 count on mortality mediated through breastfeed-
ing cessation. After adjusting for breastfeeding and
LBW (model 2), the association between low CD4
count (<200) and child mortality through 18 months
Table 1 Crude relative risks for predictors of child mortality through 18 months among 355 HIV-uninfected infants born to
HIV-infected mothers in Zambia between May 2001 and September 2004
risk (95% CIa)
Yes5 (35.7)9 (64.3)143.0 (1.4–6.5)
Number of previous02 (4.7)41 (95.3)43 1.0
pregnancies1–3 33 (13.7)208 (86.3)241 2.9 (0.73–11.8)
10 (14.1)61 (85.9)71 3.0 (0.70–13.2)
Mother’s age4 (11.8)30 (88.2)341.0
20–29.933 (13.9)205 (86.1)2381.2 (0.45–3.1)
8 (9.6)75 (90.4) 830.82 (0.26–2.5)
11 (20.4)43 (79.6)54 1.0
Yes34 (11.3) 267 (88.7)3010.55 (0.30–1.0)
Has a water tap in houseNo43 (12.8) 292 (87.2) 3351.0
Yes2 (10.0) 18 (90.0)200.78 (0.20–3.0)
Electricity in house No29 (13.4) 187 (86.6)2161.0
Yes 16 (11.5) 123 (88.5)1390.86 (0.48–1.5)
Ran out of food in previous 30 daysNo29 (10.6)245 (89.4)2741.0
Yes16 (19.8)65 (80.2)811.9 (1.1–3.3)
Mother employedNo41 (12.3)292 (87.7)3331.0
Low birth weightb
Yes 4 (18.2)18 (81.8) 221.5 (0.58–3.7)
No34 (10.8)281 (89.2)3151.0
Yes10 (33.3)20 (66.7)303.1 (1.7–5.6)
Child is male No25 (14.5)148 (85.5)1731.0
Yes20 (11.0)161 (89.0)1810.76 (0.44–1.3)
Child born at home No36 (11.6)275 (88.4)3111.0
Yes9 (20.5)35 (79.5)441.8 (0.91–3.4)
aCI, confidence interval; SES, socio-economic status. SES defined as sum of: owns a fridge, owns electricity, completed48 years of
education, employed and had food security for the previous month.
bLow birth weight defined as <2500g.
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
was reduced compared with model 1 (from 3.4
to 3.0), a reduction in estimate of effect of 17%.
This suggests that only a small part of the effect of
low vs high CD4 count on child mortality through
18 months is mediated through breastfeeding cessa-
tion. For comparison, we note that adjustment for
breastfeeding cessation using a conventional Cox
proportional hazards model showed an even greater
effect. The high-low CD4 comparison was reduced by
35% (from 3.4 to 2.5); however this approach is
simultaneously adjusting for the intermediate effects
of both breastfeeding cessation and low birth weight
and is therefore likely biased.
Our results build upon the earlier finding in this
cohort (9) which showed that uninfected infants of
HIV-infected mothers with CD4 counts <350 were
2.9 times more likely to die by 4 months than infants
of mothers with CD4 count4350. The current analysis
in the same population found that after adjusting
for socioeconomic status, children born to mothers
with CD4 count <200 had 3.4 times the risk of death
through 18 months as did infants born to mothers
with high CD4 counts. A similar finding was observed
in a cohort of uninfected children of HIV-infected
mothers in Kenya, where children of mothers with a
CD4 count <200 had 2.8 times the risk of death
compared with all other CD4 counts.7Our results
differ from a pooled analysis which found a small,
non-significant effect of a mother’s CD4 count <200
compared with a CD4 count 4500.23They found a
strong effect of maternal death on infant mortality
and these variables likely both measure maternal
disease stage; additionally, their results were adjusted
for breastfeeding cessation which, according to our
biologic model, would adjust away some of the effect
of CD4 count.
We also found that breastfeeding cessation occurred
earlier among immunocompromised mothers com-
pared with healthier mothers near the time of
the birth of their child. Our results agree with an
earlier finding by Sedgh et al.24in Tanzania where
women with a CD4 count <200 breastfed on average
3.8 months less than women with a CD4 count 5500
near the birth of her child. They are also consistent
with data from Botswana which showed an increased
risk of morbidity in children of HIV-infected mothers
if the mother had stopped breastfeeding.5
Unlike previous studies, we assessed the mediating
role of breastfeeding in the CD4 count–child mortality
relationship. Our analysis supports the hypothesis
Table 3 Marginal structural hazards ratios of the relationship between maternal CD4 count and child mortality through
18 months among 355 HIV-exposed but uninfected children in Zambia between May 2001 and September 2004a
Maternal CD4 200–500
Models 1 to 2c
2.0 (0.87–4.6)1.9 (0.80–4.3)10.0
3.4 (1.3–8.4)3.0 (1.2–7.9)16.7
SES 1–20.43 (0.21–0.89)0.43 (0.21–0.89)0.0
0.57 (0.24–1.3)0.49 (0.21–1.2)
N/A 2.0 (1.0–3.7)
aModel adjusted for low birth weight (defined as <2500g) using marginal structural model.
bBoth models 1 and 2 show hazard ratios of infant mortality and 95% CIs adjusted for all other variables in the model.
cPercent change from models 1 to 2 defined as (HR2?HR1)/(HR1?1).
dSES, Socio-economic status; SES defined as sum of: owns a fridge, owns electricity, completed 48 years of education, employed
and had food security for the previous month.
Table 2 Crude incidence rate ratios of the association between maternal CD4 count and child mortality and breastfeeding
cessation through 18 months among 355 HIV-uninfected infants born to HIV-infected mothers in Zambia between May
2001 and September 2004
Child death Breastfeeding cessation
PY IRR 95% CI
194 25233.410.7 1.9 (0.81–4.4)113187.4 60.31.3 (0.94–1.8)
<200 63 1370.918.33.2 (1.3–8.1) 39 47.681.91.8 (1.2–2.7)
PY, Person-years; IRR, Incidence Rate Ratio.
THE MEDIATING ROLE OF BREASTFEEDING CESSATION
that breastfeeding cessation accounts for some of
the effect of maternal CD4 count on child death
among HIV-exposed but uninfected infants. However,
most of the CD4 count–child mortality association
remained after adjusting for breastfeeding, suggesting
other factors also explain the observed association.
These may include increased exposure to infectious
diseases, low birth weight and poor growth, or a
general inability to care for the child.
The current study employed a marginal structural
model to adjust for confounding by LBW as it is likely
both a confounder and causal intermediate. When
using a conventional Cox proportional hazards model,
we found a substantially larger change in the estimate
of effect of CD4 count on child mortality through
6 months when adjusting for LBW. This suggests that
some of the effect of maternal CD4 count may be
mediated through birth weight and that conventional
approaches to adjusting for LBW will be biased.
While we identified no other study of breastfeeding
cessation as a mediator of the CD4 count–infant
mortality relationship, the reduction in effect of
maternal CD4 count and child survival seen in one
previous study after adjusting for breastfeeding cessa-
tion is consistent with breastfeeding cessation being
a mediator.23In their crude analyses, CD4 <200 was
associated with 2.5-fold increased risk of child death
compared with children of mothers with CD4 count
4500. After adjusting for breastfeeding and other
covariates their odds ratio was reduced to 1.7
(95% CI 0.9–3.2); this is in agreement with our
finding that some of the effect is being mediated
through breastfeeding cessation. However, as their
analysis was not designed to assess the mediating role
of breastfeeding it is not clear whether this change
reflects adjustment for the mediator or adjustment for
other confounders in their model.
Considerable debate still exists regarding the possi-
bility and optimal approach for separating direct and
indirect effects. The parsing of direct and indirect
must be no residual confounding between the expo-
sure and the outcome, or between the causal inter-
mediate and the outcome, and there must be no
interaction between the exposure and the intermedi-
ate. While our study was observational and therefore
subject to confounding, we identified few confound-
ers of the CD4 count–child mortality relationship and
we adjusted for LBW, the only confounder identified
Some feel these assumptions are overly restric-
tive28,29while others feel the conditions for decon-
structing total effects rarely occur.27Some27argue
that the no interaction requirement is unlikely to
be satisfied. However, others25note that, ‘in the
presence of interaction, it is the potentially esti-
effect that could be eliminated by control of the
intermediate—that will usually be the parameter of
public health interest’. Thus, even if our assumption
of no interaction between CD4 count and breast-
feeding has been violated, our results still imply that
a portion of the relative increase in child deaths
through 6 months attributable to advanced maternal
This study should be considered in light of several
limitations. First, there could have been residual con-
founding in our analysis. However, because maternal
CD4 count was measured only during pregnancy, and
most of our data was collected after the exposure,
we had few candidate variables for adjustment. While
it is possible that important residual confounding
remained, as we could identify no confounders in this
analysis other than low birth weight (either statis-
tically or through substantive knowledge) we consider
this scenario unlikely.
Second, time of breastfeeding cessation could have
been misclassified. Because the misclassification is
of the intermediate variable, if it is non-differential
with respect to the exposure and the outcome, it would
be expected to reduce the ability to adjust for the
intermediate. This implies that any reductions seen in
the CD4 child–mortality relationship after adjusting for
breastfeeding cessation would not completely control
for the intermediate, so that the indirect effect
observed would be expected to be an underestimate.30
Thus our results likely represent an underestimation
of breastfeeding as a causal intermediate.
Finally, we did not have data on cause of death for
the current analysis, and therefore we cannot exclude
the possibility that some of the child deaths were
unrelated to breastfeeding (e.g. deaths due to injury).
The impact of including these deaths is unclear, and
future research on the topic should seek to refine the
In conclusion, we found that uninfected children
born to HIV-infected mothers with more advanced
immunosuppression were breastfed for shorter dura-
tion and had higher mortality though 18 months
of life compared with children of mothers who were
not immunocompromised. If confirmed, our results
must be taken into account when deciding whether or
not to recommend breastfeeding for infants of HIV-
infected mothers. Our results suggest it is particularly
important for mothers with low CD4 counts near
delivery to persist in providing breast milk to their
uninfected infants, though this risk must be balanced
with the increased risk of HIV transmission with
longer duration of breastfeeding. The more wide-
spread availability of highly
therapy should also allow breastfeeding mothers the
opportunity to continue to prolong breastfeeding
without substantial increased risk of HIV transmis-
sion and therefore provide their uninfected infants
the added protection and benefits of breast milk
during the critical first months of life.31
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
This research was supported by the National Institute
for Child Health and Human Development, Bethesda,
Maryland (R01 HD39611 and R01 HD 40777) for the
Zambia Exclusive Breastfeeding Study. This publica-
tion was also made possible through support provided
by the [G/PHN/HN/CS, Global Bureau], U.S. Agency
for International Development, under the terms of
Cooperative Agreement GHS-A-00-03-00020-00. The
opinions expressed herein are those of the authors
and do not necessarily reflect the views of the U.S.
Agency for International Development.
The authors wish to thank Katherine Semrau, Nancy
Scott and Don Decker for their help with data and
Jonathon Simon, Deirdre Pierotti, Elwyn Chomba,
Marc Bulterys and Jeff Stringer for their support of
Conflict of interest: None declared.
1Spira R, Lepage P, Msellati P et al. Natural history of
human immunodeficiency virus type 1 infection in
children: a five-year prospective study in Rwanda.
2Schim Van Der Loeff M, Hansmann A, Awasana AA et al.
Survival of HIV-1 and HIV-2 perinatally infected children
in The Gambia. AIDS 2003;17:2389–94.
3Brahmbhatt H, Kigozi G, Wabwire-Mangen F et al.
Mortality in HIV-infected and uninfected children of
HIV-infected and uninfected mothers in rural Uganda.
J Acquir Immune Defic Syndr 2006;41:504–8.
4Thea DM, St Louis ME, Atido U et al. A prospective study
of diarrhea and HIV-1 infection among 429 Zairian
infants. N Engl J Med 1993;329:1696–1702.
5Shapiro RL, Lockman S, Kim S et al. Infant morbidity,
mortality, and breast milk immunologic profiles among
breast-feeding HIV-infected and HIV-uninfected women
in Botswana. J Infect Dis 2007;196:562–69.
6Newell ML. Mortality in children born to HIV infected
mothers in Africa: Infants, HIV and Mortality in Africa
(IHMA) Study Group [abstract]. Proceedings of the 13th
International Conference on AIDS and STIs in Africa, Nairobi,
Kenya, September 26, 2003.
7Otieno P, John-Stewart G, Gichuhi C et al. Predictors of
mortality in HIV-1 uninfected infants born to HIV-1
seropositive women. [abstract]. Proceedings of the 13th
International Conference on AIDS and STIs in Africa, Nairobi,
Kenya, September 26, 2003.
8Chatterjee A, Bosch RJ, Hunter DJ et al. Maternal disease
stage and child undernutrition in relation to mortality
among children born to HIV-infected women in Tanzania.
J Acquir Immune Defic Syndr 2007;46:599–606.
9Kuhn L, Kasonde P, Sinkala M et al. Does severity of HIV
disease in HIV-infected mothers affect mortality and
morbidity among their uninfected infants? Clin Infect Dis
10Hanson LA, Hahn-Zoric M, Berndes M et al. Breast
feeding: overview and breast milk immunology. Acta
Paediatr Jpn 1994;36:557–61.
11Orlando S. The immunologic significance of breast milk.
J Obstet Gynecol Neonatal Nurs 1995;24:678–83.
12Feachem RG, Koblinsky MA. Interventions for the control
of diarrhoeal diseases among young children: promotion
of breast-feeding. Bull WHO 1984;62:271–91.
13Victora CG. Infection and disease: the impact of early
weaning. Food Nutr Bull 1996;17:390–96.
Breastfeeding on the Prevention of Infant Mortality.
Effect of breastfeeding on infant and child mortality
due to infectious diseases in less developed countries: a
pooled analysis. Lancet 2000;355:451–55.
15Thea DM, Vwalika C, Kasonde P et al. Issues in the design
of a clinical trial with a behavioral intervention–the
Zambia exclusive breast-feeding study. Control Clin Trials
16Kuhn L, Aldrovandi G, Sinkala M et al. for the Zambia
Exclusive Breastfeeding Study. Effects of Early, Abrupt
Weaning on HIV-free Survival of Children in Zambia. N
Engl J Med 2008;359:130–41.
17Szklo M, Nieto FJ. Epidemiology Beyond the Basics. Sudbury,
MA: Jones and Bartlett, 2004.
18Greenland S. Introduction to regression modeling. In:
Rothman KJ, Greenland S (eds). Modern Epidemiology,
19Cole S, Hernan M. Fallibility in estimating direct effects.
Int J Epidemiol 2002;31:163–65.
20Fawzi WW, Herrera MG, Nestel P et al. A longitudinal
study of prolonged breastfeeding in relation to child
undernutrition. Int J Epidemiol 1998;27:255–60.
21Robins JM, Hernan MA, Brumback B. Marginal structural
models and causal inference in epidemiology. Epidemiology
22Hernan MA, Brumback B, Robins JM. Marginal structural
models to estimate the causal effect of zidovudine on the
survival of HIV-positive men. Epidemiology 2000;11:561–70.
23Newell ML, Coovadia H, Cortina-Borja M et al. Mortality
infected mothers in Africa: a pooled analysis. Lancet
24Sedgh G, Spiegelman D, Larsen U et al. Breastfeeding and
maternal HIV-1 disease progression and mortality. AIDS
25Robins JM, Greenland S. Identifiability and exchange-
ability for direct and indirect effects. Epidemiology 1992;3:
26Kaufman S, Kaufman JS, Maclehose RF et al. Improved
estimation of controlled direct effects in the presence of
unmeasured confounding of intermediate variables. Stat
27Kaufman JS, Maclehose RF, Kaufman S. A further
critique of the analytic strategy of adjusting for covariates
to identify biologic mediation. Epidemiol Perspect Innov
THE MEDIATING ROLE OF BREASTFEEDING CESSATION
28Petersen ML, Sinisi SE, van der Laan MJ. Estimation of
direct causal effects. Epidemiology 2006;17:276–84.
29Blakely T. Commentary: estimating direct and indirect
effects-fallible in theory, but in the real world? Int J
30Greenland S, Robins J. Confounding and misclassifica-
tion. Am J Epidemiol 1985;122:495–506.
31Kuhn L, Thea DM, Aldrovandi GM. Bystander effects:
children who escape infection but not harm. J Acquir
Immune Defic Syndr 2007;46:517–18.
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY