Iron-Deficiency Anemia: Reexamining the Nature and

Magnitude of the Public Health Problem

An Analysis of Anemia and Pregnancy-Related Maternal Mortality1,2

Bernard J. Brabin,3 Mohammad Hakimi* and David Pelletier†
Liverpool School of Tropical Medicine, Liverpool, England and University of Amsterdam, Emma
Kinderziekenhuis, Academic Medical Centre, Amsterdam, Netherlands; *Gadjah Mada University, Yogyakarta,
Indonesia; and †Division of Nutritional Sciences, Cornell University, Ithaca, NY 14853

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

ABSTRACT The relationship of anemia as a risk factor for maternal mortality was analyzed by using cross-
sectional, longitudinal and case-control studies because randomized trials were not available for analysis. The
following six methods of estimation of mortality risk were adopted: 1) the correlation of maternal mortality rates with
maternal anemia prevalence derived from national statistics; 2) the proportion of maternal deaths attributable to
anemia; 3) the proportion of anemic women who die; 4) population-attributable risk of maternal mortality due to
anemia; 5) adolescence as a risk factor for anemia-related mortality; and 6) causes of anemia associated with
maternal mortality. The average estimates for all-cause anemia attributable mortality (both direct and indirect) were
6.37, 7.26 and 3.0% for Africa, Asia and Latin America, respectively. Case fatality rates, mainly for hospital studies,
varied from ⬍1% to ⬎50%. The relative risk of mortality associated with moderate anemia (hemoglobin 40 – 80 g/L)
was 1.35 [95% confidence interval (CI): 0.92–2.00] and for severe anemia (⬍47 g/L) was 3.51 (95% CI: 2.05– 6.00).
Population-attributable risk estimates can be defended on the basis of the strong association between severe
anemia and maternal mortality but not for mild or moderate anemia. In holoendemic malarious areas with a 5%
severe anemia prevalence (hemoglobin ⬍70 g/L), it was estimated that in primigravidae, there would be 9
severe-malaria anemia-related deaths and 41 nonmalarial anemia-related deaths (mostly nutritional) per 100,000
live births. The iron deficiency component of these is unknown. J. Nutr. 131: 604S– 615S, 2001.

KEY WORDS: ● pregnancy ● anemia ● mortality ● malaria ● iron deficiency

Maternal mortality continues to be a major health problem
in the developing world. Nearly 600,000 women die each year
as a result of complications of pregnancy and childbirth; most
of these deaths could be prevented with attainable resources
and skills (WHO 1996). The worldwide maternal mortality
ratio (annual number of deaths of women from pregnancy-
related causes per 100,000 live births) is estimated to be 390
per 100,00 live births (Abousahr and Royston 1991). Most of
these occur in developing countries, where women have a risk
of dying in pregnancy and childbirth that is 50 –100 times
greater than that of women in the developed world (Starrs
1987). In the developing world, rates are as high as 700 per
100,000 live births in many parts of Africa and in some
countries in south Asia. These large differences in risk are
related primarily to differences in available obstetric care for
women living in areas with inadequate antenatal and delivery
1
Presented at the Belmont Meeting on Iron Deficiency Anemia: Reexamining
the Nature and Magnitude of the Public Health Problem, held May 21–24, 2000 in
Belmont, MD. The proceedings of this conference are published as a supplement
to The Journal of Nutrition. Supplement guest editors were John Beard, The
Pennsylvania State University, University Park, PA and Rebecca Stoltzfus, Johns
Hopkins School of Public Health, Baltimore, MD.
2
This article was commissioned by the World Health Organization (WHO).
The views expressed are those of the authors alone and do not necessarily reflect
those of WHO.
3
To whom correspondence and reprint requests should be addressed. E-
mail: l.j.taylor@liverpool.ac.uk.
care facilities. Harrison (1989) has championed the arguments
for developing improved pregnancy care to reduce maternal
mortality in developing countries. In reports from Nigeria, he
has highlighted the importance of maternal anemia as a con-
tributory factor to maternal death (Harrison 1975, Harrison
and Rossiter 1985). In 1987, international agencies and lead-
ers from 45 countries established the Safe Motherhood initia-
tive with the goal of reducing half of maternal deaths by the
year 2000 (World Bank 1993). A key component of Safe
Motherhood is the eradication of anemia during pregnancy.
The WHO has produced estimates of the global burden of
deaths attributable to anemia (all forms) in women of repro-
ductive age (Murray and Lopez 1994). These are summarized
in Table 1. The total estimate is a minimum of 16,800 and
maximum of ⬃28,000 annually with a greater risk of anemia-
related death in younger women.
The relationship of anemia as a risk factor for mortality is
derived mainly from cross-sectional studies and can be con-
founded for several reasons. Most studies report hospital data,
often for moribund women, and there is limited attention to
factors such as pregnancy hemodilution, hemoglobin rise in
late pregnancy, concurrent infection, hemorrhage, prior treat-
ment or poor maternal nutritional status. In young women
living under endemic malaria conditions, especially in urban
areas in which adults may have poor malaria immunity, severe
malarial anemia and cerebral malaria may occur and can

0022-3166/01 $3.00 © 2001 American Society for Nutritional Sciences.

604S
 ANEMIA AND MATERNAL MORTALITY
TABLE 1
Estimated anemia deaths (in thousands) in women of
reproductive age1
Region 15–29 y 30–44 y
Developed (all) —2 — and number of studies available. Hemoglobin midpoint values were
Developing (all) 9.0 7.1
EME3 — —
FSE — —
India 3.5 1.0
China 1.5 3.0
OAI 1.6 0 Diseases, which defines a maternal death as the death of a woman
SSA — —
LAC — —
MEC — —
World 9.3 7.5
1 Source: World Bank (1993).
2 A dash (—) indicates ⬍1000 deaths.
3 Abbreviations: EME, established market economies; FSE, formerly
socialist economies; OAI; other Asia and islands; SSA, sub-Saharan
Africa; LAC, Latin America and Caribbean; MEC, Middle Eastern cres-
cent.
rapidly lead to death (Granje et al. 1998). For these reasons,
most studies form an inadequate basis for determining how
anemia relates causally to maternal survival in communities,
and extrapolation from hospital delivery data must be consid-
ered an approximation that may be misleading.
Intervention studies with maternal mortality as an outcome
measure are required to determine causality, but these are very
difficult to conduct for both ethical and logistic reasons. For
example, there are very few studies that did not use transfusion
as an emergency procedure in severely anemic women at term
(Fullerton and Turner 1962). If transfusions are taken into
account, then near-miss fatality could be an alternative out-
come measured, but the true risk in such cases remains uncer-
tain. In view of these difficulties, a number of alternative
approaches that independently assess this risk must be
adopted. Consistency between analyses of severe anemia and
poor survival would add credence to the strength of a causal
relationship. Several issues are related to estimating attribut-
able risk for specific causes of anemia and in quantifying risk
for moderately anemic women because less anemia may still
contribute to death from other causes. Such information
would be helpful for intervention decisions.
METHODS
Identification of published studies. Published studies on the
relationship between anemia (defined by severity) and maternal mor-
tality were identified using Medline, references in published papers,
Cochrane Review issues and personal communications. Unpublished
data from Nigeria available in a detailed hospital report by Lawson
and Lister were reanalyzed and included in a separate summary of
Nigerian data. Studies that included postnatal deaths up to 40 d were
included, although in practice few studies reported follow-up data
beyond delivery.
Selection of studies for inclusion in the analyses. Studies in-
cluded in the review were limited to cross-sectional, longitudinal and
case-control studies because no randomized controlled trials were
available for analysis. Attention was given to the assessment of
possible biases in studies of mixed validity. Studies identified were
reviewed with regard to the following factors: maternal age, parity,
anemia severity, clinical presentation, gestational age, use of blood
transfusion, length of follow-up, etiological diagnosis, laboratory es-
605S
timation of hemoglobin (Hb)4 or hematocrit, and analytical methods.
Hematocrit was converted to a Hb value by dividing by 3 and
multiplying by 10. Studies that listed anemia as a direct cause of death
were of particular value, permitting the estimate of the total number
of maternal deaths attributed to anemia. Data from the WHO com-
pilation of maternal mortality was reviewed and categorized by source
(hospital or community), direct or indirect cause of anemia, region
calculated when the range was available. For other studies, anemia
cut-off points were used below which proportional groups of women
with anemia were defined.
Analyses. The definition of maternal death used in this review
was based on the 10th revision of the International Classification of
while pregnant or within 42 d of termination of pregnancy, regardless
Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019
of the duration and site of the pregnancy, from any cause related to
or aggravated by the pregnancy or its management but not from
accidental or incidental causes (WHO 1992a).
Maternal deaths were also divided into two groups as follows: 1)
direct obstetric deaths, resulting from obstetric complications of the
pregnant state (pregnancy, labor and the puerperium), interventions,
omissions or incorrect treatment, or a chain of events resulting from
any of the above; and 2) indirect obstetric deaths, resulting from
previously existing disease or disease that developed during pregnancy
and was not due to direct obstetric causes but was aggravated by the
physiological effects of pregnancy.
For each of the studies selected, estimates of the relative risks and
their 95% confidence intervals were calculated using established
methods. These were used with prevalence estimates to obtain pop-
ulation-attributable risk (PAR) of anemia-related maternal mortality.
Several case fatality studies could not be used in the risk analysis
because they did not present mortality data for the less-anemic
subjects in their study population. The formula for PAR is as follows:
PAR ⫽ deaths related to anemia/total deaths
⫽ 关Prev ⫻ 共RR ⫺ 1兲兴/兵1 ⫹ 关Prev ⫻ 共RR ⫺ 1兲兴其
where Prev is the prevalence of anemia of a given severity and RR is
the ratio of mortality in the anemic to mortality in the less anemic
(referent group).
Methods of estimation. The following six methods of estimation
were adopted: 1) The correlation of maternal mortality rates with
maternal anemia prevalence derived from national statistics in the
WHO compilation on anemia in the world. 2) The proportion of
maternal deaths attributable to anemia. 3) The proportion of anemic
women who die (i.e., case fatality estimates) and how this risk varies
with anemia severity. 4) PAR of maternal mortality due to anemia. 5)
Adolescence as a risk factor for anemia related mortality. 6) Causes of
anemia associated with maternal mortality.
Definitions. Mild anemia was defined as Hb ⬍110 g/L, moderate
anemia as ⬍70 g/L and severe anemia as ⬍50 g/L. The 110 g/L cut-off
value is based on international convention, whereas the other two
cut-off values are commonly used in the literature. The 50 g/L cut-off
value is related in part to functional consequences associated with
cardiac decompensation.
RESULTS
Maternal mortality and anemia prevalence. A detailed
compilation of anemia prevalence in women published by
WHO includes estimates of maternal mortality from anemia
for nine selected countries (WHO 1992b). These estimates
range from 27 per 100,000 live births in India to 194 per
100,000 live births in a hospital-based study in Pakistan to 42
of 44 maternal deaths in Somalian refugee camps. The cut-off
values for defining anemia vary for these studies as does anemia
prevalence in the communities in which these women live
4
Abbreviations: CI, confidence interval; Hb, hemoglobin; HIV, human immu-
nodeficiency virus; PAR, population-attributable risk.
606S SUPPLEMENT

population prevalence estimates for all-cause anemia and ma-

ternal mortality ratios, and how does this differ between areas
with high and low maternal mortality?
The graph shown in Figure 1 uses data on anemia preva-
lence from the WHO tabulation of available information on
nutritional anemia in women (WHO 1992b), and maternal
mortality ratios reported by United Nations Children’s Fund
(1999) for the years 1990 –1997. Anemia prevalence values for
individual countries were selected by using the following cri-
teria: national data if available, altitude ⬍ 2000 m, not a
refugee population, survey completed after 1980, largest avail-
able sample size and actual (not estimated) prevalence avail-
able (Table 2). Anemia refers to Hb values ⬍110 g/L. The

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

FIGURE 1 Correlation of the prevalence of anemia in women and
the maternal mortality ratios per 100,000 live births for 44 countries.
Log Y ⫽ 1.50 ⫹ 0.019X; R2 ⫽ 41.1%; P ⬍ 0.0001. CI: confidence
interval; PI: predicted interval.
(WHO 1992b). The WHO tabulation adopts the interna-
tional definition for anemia for pregnant women of ⬍110 g/L.
The percentage below this value identifies the anemic popu-
lation, although no single value will separate all anemic from
all nonanemic women. What is the relationship between these
correlation between these two variables is highly significant
(Pearson correlation coefficient 0.561, P ⬍ 0.001). For eval-
uations of the goodness of fit for three models, i.e., linear,
quadratic and exponential, the coefficients of determination
were 0.315, 0.424 and 0.411 with F-values 19.3, 15., and 29.3,
respectively, indicating that the exponential (logarithmic)
model fits the data well.
Considerable caution is required in interpreting the associ-
ation shown in Figure 1 because confounding and correlates
that are based on country-level information may not neces-
sarily reflect the relationship at an individual level. There is a
strong possibility that the association may be driven by a large
number of data points in one quadrant of the scatterplot from
one region (sub-Saharan Africa) that has high rates of mater-
nal mortality, anemia and malaria, poor obstetric care and
other confounders. Nevertheless the association is of interest;
it has not been described previously and may provide insight at

TABLE 2
Maternal mortality ratios (MMR) and maternal anemia prevalence

MMR (per Anemia Sample MMR (per Anemia Sample

Country 100,000) prevalence size Country 100,000) prevalence size

% n % n

Africa Asia/Middle East

Burundi 1300 80 Random Indonesia 650 74 3300
Kenya 650 33 291 Philippines 280 48 158
Mozambique 1100 58 748 China 95 13 645
Tanzania 770 86 1317 Israel 7 16 559
Zambia 650 34 1911 Bangladesh 850 62 178
Algeria 220 42 222 India 570 88 3829
Sudan 660 36 121 Iran 120 8 169
Tunisia 170 38 — Pakistan 340 51 1571
Benin 990 55 126 Pacific
B. Fasso 930 24 30 Papua New Guinea 930 94 294
Gambia 1100 61 81 Vanuatu 280 41 1135
Liberia 560 78 547 Europe/USA
Niger 1200 47 360 United Kingdom 9 19 49
Nigeria 1000 47 95 Czechoslovakia Italy 15 23 664
Cameroon 550 8 90 Belgium 12 10 65
Chad 1500 37 112 France 10 0 83
S. Africa 230 33 229 Germany 15 15 233
Caribbean/S. America Netherlands 22 17 5000
Cuba 95 10 225 USA 12 26 796
Jamaica 120 62 National 12 28 Large
St Lucia 30 22 National
St Vincent 43 20 National
St Kitts/Nevis 130 43 3170
Bolivia 390 25 200
Brazil 220 32 1615
Guyana 190 63 19062
Peru 280 53 75
 ANEMIA AND MATERNAL MORTALITY
Maternal deaths and days of life lost due to severe anemia
All cause maternal Maternal deaths attributed
Region mortality ratio1 (A) to anemia (%)2 (B)
Africa 640
Asia 420
Latin America 270
1Per 100,000 live births.
2Derived from tabulations in World Health Organization 1991.
3Proportion of maternal deaths attributable to anemia ⫻ all cause maternal mortality ratio per 100,000.
4Days of maternal life lost per 100,000 live births per year due to death from anemia; (A ⫻ B ⫻ 38.7 [expectation of life in years at age 25 for a
West African country] ⫻ 365.25 [average days in a year]).
national levels for programs designed to reduced maternal
mortality.
The prevalence values are related to all-cause anemia and
no conclusions can be inferred in relation to iron-deficiency
anemia. The WHO compilation separately lists a smaller num-
ber of studies that report serum iron concentrations and give
values below the norm (⬍9 ␮mol/L). Applying the same
criteria for selection as for the anemia surveys, 17 studies were
available for analysis along with the maternal mortality ratio.
A positive correlation was observed that was not significant
(Pearson correlation 0.415, P ⬎ 0.098).
Few community studies are available that report anemia
prevalence in women and the maternal mortality ratio for
large samples from the same cohort of women. A prospective
rural community study in Malawi, in a malarious area, esti-
mated the maternal mortality ratio as 398 (per 100,000 live
births) and found an anemia prevalence (hematocrit ⬍0.25)
in pregnancy of 6.2% for the same cohort (McDermott et al.
1996).
The proportion of maternal deaths attributable to anemia.
A detailed compilation of reports on the causes of maternal
deaths attributable to anemia is published by WHO (1991).
This lists 62 reports from 33 countries for which a proportion
is provided for maternal deaths attributable to anemia. Ane-
mia is listed as a direct cause of death in 26% of these reports
and as an indirect cause in the remainder. The definitions of
anemia vary substantially between studies and many are based
on clinical assessment alone; most (88.5%) are hospital based,
with a high proportion of complicated deliveries.
Anemia was given as a direct cause of between 1 and 46%
(mean 10.0%) of maternal deaths in 23 studies. Many reports
did not include anemia as a cause of death; most were from
Latin America, but 52 studies were from Africa and 45 from
Asia. No study lists anemia both as a direct cause for severe
cases and indirect cause for others, suggesting that the criteria
for attribution depend on the obstetrician’s perception of the
relative importance of anemia, with many listing anemia only
as an indirect cause. There is little documentation for the
criteria used in these clinical judgments.
The average estimates for all-cause anemia-attributable
mortality (i.e., both direct and indirect) from these reports are
6.37, 7.26 and 3.00% for Africa, Asia, and Latin America,
respectively. These regional estimates average considerable
variation among countries. They correspond reasonably well
with three community-based studies from Africa (mean 7.3%)
and four community-based studies from Asia (mean 9.4%)
(WHO 1991). Crude maternal mortality ratios from anemia
can be calculated by using these values and regional estimates
for the maternal mortality ratio. These estimates are given in
Table 3, which shows the maternal mortality ratio from all-
607S
TABLE 3
Maternal mortality from Days of life lost from
anemia (A ⫻ B)3 maternal anemia4
6.37 40.77 576,291
7.26 30.49 430,981
3.00 8.10 114,495
Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019
cause anemia and the days of life lost from maternal anemia. In
Africa, this mortality is fivefold higher than for Latin America.
Within these regions, maternal mortality from anemia varies
greatly among countries. For example, in community studies in
Asia, values vary (per 100,000 live births) from 27 in India and
54 in Bangladesh to 194 in Pakistan and in Africa from 35 in
Senegal to 82 in Kenya (WHO 1991).
Within-country variation is also considerable. In the WHO
(1991) Global Factbook, only three countries worldwide listed
three or more studies giving values for maternal deaths attrib-
utable to anemia. These are Nigeria, Tanzania and India.
Table 4 shows the summary of data for these countries and
illustrates the widely inconsistent estimates of anemia-attrib-
utable mortality among studies in these selected countries.
There are a number of reasons for this. First, only two of the
Indian, one of the Tanzanian and none of the Nigerian studies
were community based. Second, estimates will vary according
to whether maternal deaths reflect differences in outcome
between large tertiary referral hospitals and smaller district
hospitals. Third, risk of death will alter depending on the
prevalence of hemoglobinopathies, malaria and nutritional
deficiency among populations within a country. Four of the
Nigerian studies, for example, selectively reported on maternal
deaths attributed mainly to hemoglobinopathies.
The proportion of anemic women who die. The relation-
ship of anemia and its correlates can best be examined in
individuals. Acute onset of anemia during pregnancy will
greatly increase the risk of death because this can lead to rapid
cardiac decompensation. When the Hb concentration is ⬍ 80
g/L, compensatory mechanisms fail, lactic acid accumulates
and patients become breathless at rest. Cardiac failure may
occur when Hb is ⬍ 40 g/L, especially with twin pregnancies
or splenomegaly (Fleming 1989b), and when anemia is not the
primary cause of death, it may frequently be a contributory
TABLE 4
Anemia as a cause of maternal death1
Percentage of deaths attributable
to anemia
Number of
Country studies Mean Median Range
n
Nigeria 19 8.0 4 1–46
Tanzania 9 9.1 9 1–15
India 11 13.3 13 7–22
1 Data derived from World Health Organization 1991.
608S SUPPLEMENT

TABLE 5
Pregnancy hematocrit levels and case fatality in Nigerian studies

Year Location Hematocrit Deaths Survivors Case fatality Reference

% n n %

1953–54 Ibadan1 4.5–8.0 6 17 26.09 Lawson and

9.0–13.0 7 54 13.72 Lister 1954
14.0–18.0 2 59 3.28
19.0–22.0 0 46 0.00
1957–58 Ibadan 5–72 5 4 55.56 Fullerton and
8–10 10 24 29.41 Turner 1962
11–13 3 36 6.12
⬍14

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

1957–68 Ahmadu Bello 7 158 4.24 Harrison 1975
14–18 3 153 1.92
19–26 1 244 0.41
⬍14 (transfused) 5 292 1.68
⬍14 (not transfused) 6 16 27.27
1961 Ibadan 5–8 4 12 25.0 Johnson and
9–12 5 68 6.85 Ojo 1967
13–16 1 59 1.67
17–20 0 57 0.00
21–24 0 28 0.00
Ahmadu Bello ⬍14 5 20 20.00 Harrison 1982
15–25 3 87 3.33
25–29 0 143 0.00
Ahmadu Bello3 ⬍14 11 111 9.02 Harrison and
15–20 10 175 5.42 Rossiter
20–25 20 221 8.30 1985
25–29 14 317 4.22
30–34 19 327 5.49
35–39 24 322 6.93
40–44 15 155 8.82
⬎45 8 28 22.22
1961–62 Ibadan ⬍15.5 4 132 2.94 Oja 1965

1 Deaths due to acute blood loss excluded.

2 Not treated by exchange transfusion.
3 Unbooked women.

factor. The distinction between anemia as a primary or con-
tributory factor for death is related to its acute and chronic
pattern of onset. Severe acute anemia can be a primary and
rapid cause of death, (e.g., in Nigeria) related to the acute
hemolysis of sickle cell disease (Lawson 1962), whereas
chronic anemias are considered to be frequent contributory
factors, especially to the consequences of hemorrhage and
infection. Iron-deficiency anemias may contribute to increased
morbidity and mortality by increasing maternal susceptibility
to infection (Brock 1999). Because there is good documenta-
tion that pregnant women are more susceptible to several
infections (Brabin 1985), further information is required to
determine how increased susceptibility to injection is related
to nutritional anemia. Increased infection risk could provide a
plausible biological mechanism for increased mortality risk in
moderately anemic women.
How can acute and chronic influences on mortality risk in
anemic women be distinguished, and is there a threshold effect
for anemia severity at which maternal mortality greatly in-
creases? Tables 5 and 6 summarize available data on case
fatality in relation to pregnancy hematocrit or Hb values.
Nearly all of these studies are hospital based and report women
dying mainly in the perinatal period. Several provide no
information on exclusions or duration of postpartum follow-
up. The proportion of women treated by transfusion is unclear
except for five studies (Cheng-Chi et al. 1981, Fullerton and
Turner 1962, Harrison 1975, Harrison and Rossiter 1985, Isah
et al. 1985). Differences in available obstetric care and blood
transfusion greatly influence mortality risk in severely anemic
women, and disparity among findings for individual countries
could primarily reflect these differences. In this context, it is of
value that there are seven studies for comparison from Nigeria
alone, three of which are reports by Harrison and his col-
leagues (Harrison 1975 and 1982, Harrison and Rossiter
1985). Case fatality fell with transfusion from 27.3 to 1.7% in
women with hematocrit values ⬍0.14. The Nigerian studies
are especially valuable because they allow assumed midpoints
to be calculated for each hematocrit category, and the results
represent findings from large teaching hospitals that are ter-
tiary referral centers in which adequate obstetric care facilities
should be available. Also at the time these were undertaken,
maternal human immunodeficiency virus (HIV) infection was
not a confounder. A single report from India from a tertiary
facility also presents data that allow a midpoint to be calcu-
lated (Table 6) (Sarin 1995). The data listed in Table 6 for the
non-Nigerian studies mostly do not allow estimation of Hb
midpoints nor provide case fatality estimates for very severe
anemia (Hb ⬍50 g/L).
Figure 2 shows the plot of maternal case fatality against Hb
level for studies from Tables 5 and 6 for which Hb midpoints
were available (Hb equals hematocrit divided by 3 and mul-
tiplied by 100). Case fatality ranges from ⬍1% to ⬎50% and
mortality increases with extremely low Hb levels (⬍30 g/L).
This result is driven by four data points from Ibadan, Nigeria,
in mid-century, with Hb levels ⬍25 g/L. If these four points
are excluded, there is no apparent relationship between Hb
 ANEMIA AND MATERNAL MORTALITY 609S

TABLE 6
Pregnancy hemoglobin (Hb) levels and case fatality in non-Nigerian studies

Case
Year Location Hb Deaths Survivors fatality Reference

g/L n n %

1935 Ceylon Hookworm ⫹ve 30 691 4.16 Wickramasuriya 1937

Marked pallor
1953–58 Malaysia (Kuala Lumpur) ⬍651 21 1205 1.71 Tasker 1958
1953–62 Malaysia (Kuala Lumpur) ⬍652 35 2215 1.58 Llewellyn-Jones 1965
1977–80 Indonesia (urban and rural) ⬍100 — — 0.72 Cheng-Chi et al. 1981
1964–73 India (Calcutta) ⬍70 87 2378 3.53 Konar et al. 1980
⬍70

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

1982–94 India (Punjab) 117 8231 1.40 Sarin 1995
70–109 184 19462 0.94
⬎110 38 10533 0.36
1985–86 Zambia (Ndola) ⬍70 0 11 0.0 Fleming 1989a
1985–87 Papua New Guinea ⬍80 1 64 1.54 Brabin et al. 1990
1989–91 Gambia ⬍75 2 199 1.0 Foord et al. 1992
1989–90 Guinea (Konakry) Symptomatic 59 128 31.55 Thonneau et al. 1992
1991–93 Guinea (Konakry) ⬍110 12 1396 0.85 Diallo et al. 1995
1990–91 Kenya ⬍60 6 67 8.21 Zucker et al. 1994
1993–94 Malawi ⬍803 2 362 5.49 Verhoeff et al. 1999,
unpublished

1 Transfusion rarely used; Hb distribution mean ⫽ 47.2 g/L.

2 Transfusion rarely used; Hb distribution mean ⫽ 49.4 g/L.
3 Hb distribution mean 70 g/L.

levels and case fatality rates among the remaining data points.
For the studies listed in Tables 5 and 6, few details are
provided on the etiology of anemia, the relative contribution
of acute or chronic disease, coexisting conditions, exclusions,
percentage transfused and other aspects of obstetric care.
These factors can create both positive and negative confound-
ing. No details on iron-deficiency anemia are provided, al-
though Llewellyn-Jones (1965) stated that aggressive paren-
teral iron was their main form of therapy. Fullerton and Turner
(1962) in Nigeria mention the importance of hookworm coin-
fection and Wickramasuriya (1937) in Ceylon stratified case
fatality by the presence or absence of hookworm infection and
showed significantly higher risk of death in infected women
who presumably had chronic iron-deficiency anemia, [relative
risk 2.1; 95% confidence interval (CI): 1.3–3.4]. Most reports
were from malarious areas, and malaria is an important con-
tributor to pregnancy anemia, especially in primigravidae
FIGURE 2 Case fatality in relation to maternal hemoglobin (Hb,
g/L).
(Brabin 1983). However, in a recent study in Malawi, the
attributable risk of anemia in pregnancy was greater for iron
deficiency than malaria (Verhoeff et al. 1999).
Population-attributable risk of maternal mortality due to
anemia. Attributable risk can be a useful summary statistic
for describing the effect of a risk factor on mortality at the
population level. However, the more severe anemia becomes,
the more likely it is to have multiple causes and not be due to
iron or nutritional deficiency alone. This creates difficulties in
establishing attributable risk, particularly across populations
whose epidemiological background and disease exposure may
be very different. This problem was addressed by Pelletier and
colleagues (1993) in discussing the epidemiological evidence
for a potentiating effect of malnutrition on child mortality.
Causality should be inferred only in the light of the con-
sistency of the epidemiological evidence, and in the present
discussion, terms such as PAR are meant to refer only to
statistical associations. Rush (2000) estimated relative risks for
anemia-attributable maternal mortality and discussed in detail
the limitations of several of the studies cited in Tables 5 and
6. On the basis of evidence available, he considered it a
reasonable working assumption that maternal mortality is
greatly increased with severe anemia, and the strength of the
relationship made it appropriate to assume a causal association
with severe anemia but that the association with moderate
anemia was less clear.
By way of deriving the most reliable estimates of the effects
of moderate anemia, the relative risks from five of the studies
that had adequate data were calculated using only internal
reference values and mutually exclusive categories of Hb con-
centrations. These estimates are shown in Tables 7 and 8. For
the moderate Hb range (40 – 80 g/L), there is no consistency in
the relative risk estimates among the five studies although all
are from one country (Nigeria). The table also highlights the
small sample size for most of these analyses, suggesting caution
in drawing inferences from these individual values. When the
data from all five studies are pooled, the relative risk of
610S SUPPLEMENT

TABLE 7
Relative risk of maternal mortality for moderate anemia using five Nigerian studies with adequate data

Moderate Relative risk

Study reference hemoglobin range Deaths Survivors (95% CI)

g/L n n

Ibadan (Johnson and Ojo 1967) 40–69 2 117 —

ⱖ 80 (reference) 0 27
Ahmadu Bello (Harrison and Rossiter 1985) 47–80 30 426 1.16
ⱖ 80 (reference)1 72 1193 (0.77–1.75)
Ahmadu Bello (Isah et al. 1985) 45–60 2 59 —
ⱖ 60 (reference) 0 45

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

Ahmadu Bello (Harrison 1975) 47–63 3 156 4.62
ⱖ 63 (reference) 1 244 (0.49–44.1)
Ahmadu Bello (Harrison 1982) 47–80 3 90 4.61
ⱖ 80 (reference) 1 142 (0.49–43.7)
Mantel Haenszel weighted relative risk 1.35
(0.92–2.00)

1 Hemoglobin ⬎150 g/L excluded.

mortality associated with moderate anemia was estimated to be
1.35 (95% CI: 0.92–2.00). The lack of a significant association
arises in part because mortality risk in the referent groups was
not low and none of these groups were nonanemic. The
relative risk of maternal mortality for severe anemia (⬍47 g/L)
for the same five studies was significantly increased at 3.51
(95% CI: 2.05– 6.00) (Table 8).
Estimates of PAR derived from these data are shown in
Table 9. The PAR value of 31% reported by Zucker et al.
(1994) for a group of women with a 6% severe anemia prev-
alence (Hb ⬍60 g/L) is higher than the estimated value for
severe anemia at this prevalence from Table 9 (⬃13%). A best
estimate of the actual prevalence of severe anemia in many
developing countries is likely to be ⱕ5%. Pending further
studies, the only PAR estimates that could be defended would
be based on the strong association between severe anemia and
maternal mortality.
Adolescence as a risk factor for anemia-related mortality.
Over half of the world’s population is ⬍25 y old and ⬎80% of
the world’s youth live in developing countries. In the mid-
1990s, the global teenage population was estimated at 513
million. In this group of adolescents (10 –19 y), the WHO has
estimated that anemia prevalence (Hb ⬍110 g/L) is 16% in
less-developed countries but 45% in Africa (DeMaeyer and
Adiels-Tegman 1985). The risk of anemia is high in teenage
primigravidae in developing (Arkutu 1979, Barr et al. 1998,
Fazio-Tirrozo et al. 1998) and developed countries (Beard
1994, Osbourne et al. 1981). Maternal deaths in a community
study using verbal autopsy in Tanzania showed no association
with maternal age (Macleod and Rhode 1998). These authors
did not examine whether maternal deaths related to anemia
were more common in adolescents. In a large hospital-based
study in Northern Nigeria, a higher maternal mortality from
severe anemia (43%) was compared in very young (⬍15 y)
adolescent, older adolescent and nonadolescent pregnant
women (⬍10%) (Harrison 1989). Lawson and Lister (1954) in
an early Nigerian study of 188 moderately anemic women (Hb
⬍70 g/L) observed a case fatality of 1.89% in adolescent
pregnancies compared with 8.89% in nonadolescent women
(␹2 ⫽ 2.9, P ⬍ 0.1). Only 3 of the 53 adolescents were ⬍16
y old.
In an early study from Guyana of the pattern of mortality

TABLE 8
Relative risk of maternal mortality for severe anemia using five Nigerian studies with adequate data

Severe hemoglobin Relative risk

Study reference range Deaths Survivors (95% CI)

g/L n n

Ibadan (Johnson and Ojo 1967) 27–39 4 73 —

ⱖ 80 (reference) 0 27
Ahmadu Bello (Harrison and Rossiter 1985) ⱕ47 11 111 1.58
ⱖ80 (reference)1 72 1193 (0.86–2.91)
Ahmadu Bello (Isah et al. 1985) ⬍45 13 68 —
ⱖ 60 (reference) 0 45
Ahmadu Bello (Harrison 1975) ⬍47 7 158 10.39
ⱖ 63 (reference) 1 244 (1.3–83.7)
Ahmadu Bello (Harrison 1982) ⬍47 5 20 28.60
ⱖ 80 (reference) 1 142 (3.5–235)
Mantel Haenszel weighted relative risk 3.51
(2.05–6.00)

1 Hemoglobin ⬎150 g/L excluded.

 ANEMIA AND MATERNAL MORTALITY 611S

TABLE 9 reasonable because in areas of high transmission, a large num-

ber of studies have confirmed that P. falciparum malaria and
Population attributable risk of maternal mortality for moderate anemia are more frequent in primigravidae (Brabin 1983).
and severe maternal anemia Figure 3 shows the relative risk for anemia in first compared
with later pregnancies at different Hb cut-off values using data
Anemia Moderate anemia Severe anemia derived from studies in malarious areas of Africa and Papua
prevalence hemoglobin hemoglobin
New Guinea. The figure is derived from a previous estimate of
% 40–80 g/L 27–47 g/L this excess risk (Brabin and Rogerson 2001) but includes
additional studies (Isah et al. 1985, Lawson and Lister 1964)
5 0.017 0.111 not identified at the time of the earlier analysis. The goodness
10 0.034 0.201 of fit shows a highly significant association for a quadratic
15 0.050 0.273 model (R2 ⫽ 0.996; P ⫽ 0.0041). This model indicates that, in
20 0.065 0.334 malarious areas, there is only a small excess of mild anemia in
25 0.080 0.386
primigravidae compared with multigravidae. A larger excess is

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

30 0.095 0.430
35 0.109 0.468 observed with moderate and severe anemia (Hb ⬍80 g/L:
40 0.123 0.499 relative risk, 1.55, 95% CI: 1.4 –1.7; Hb ⬍7: relative risk 1.86,
95% CI: 1.6 –2.1). The PAR values of anemia due to malaria
in primigravidae derived from this method are given in Table
10, which shows that ⬃1 in 6 cases of severe anemia (Hb ⬍70
after the eradication of hyperendemic malaria (Giglioli 1972), g/L) and 1 in 25 cases of mild anemia (⬍110 g/L) can be
100 deaths were recorded for pregnant women in 1937–1966. attributed to malaria in primigravidae. Table 10 also shows
Of these women, 24% were ⬍20 y old and none was ⬎40 y PAR values derived using a second method based on the
old. There was a marked reduction in the incidence of such presence or absence of P. falciparum parasitemia. There is
deaths in successive periods of improved malaria control. Ane- reasonable agreement between PAR calculations using these
mia related to hookworm infection was given as the primary two different methods. These are consistent with results from
cause in 4 of these deaths. No information was provided on the a randomized controlled trial of antimalarial drugs in Kenya
incidence of severe malarial anemia. (Shulman et al. 1999).
There is a scarcity of data on adolescent mortality and If 5–10% of severe anemias (Hb ⬍70 g/L) in primigravidae
severity of anemia in developing countries. Presumably, onset are assumed to be due to sickle cell disease in sub-Saharan
of nutritional anemia at an early age results in chronic anemia Africa and 18% due to malaria (from Table 10), then the
that perpetuates any risk of anemia-related mortality through remaining 75% of cases would be attributable mainly to iron,
subsequent pregnancies. Effective antenatal care may reduce folate, vitamin A and vitamin B-12 deficiencies or HIV infec-
these risks because more frequent antenatal care visits for tion. In the Nigerian studies summarized in Table 5, HIV was
pregnant adolescents in Malawi correlated with a significant not a contributing factor because all of these surveys were
reduction in the prevalence of severe anemia (Brabin et al. completed before 1962. Nutrient deficiencies were therefore
1998). the major contributors.
Causes of anemia associated with maternal mortality. Cause-specific mortality in primigravidae related to severity
Anemia in pregnancy in women in developing countries is of anemia can be calculated using the following formulas:
multifactorial in etiology. Iron- and folate-deficiency anemias
are common. The former are related to nutritional deficiency P ⫻ 共PAR m) ⫻ (CFR)
and intestinal helminthic infections and the latter to poor ⫽ maternal mortality from malaria anemia
intake and chronic hemolytic states. Hemolytic anemia, to a
greater or lesser degree, is commonly seen during pregnancy in
malarious areas of developing countries. The observation that
severe anemia is greatly reduced in patients who have received
regular malaria prophylaxis during pregnancy (Fleming et al.
1986, Garner and Brabin 1994, Shulman et al. 1999) indicates
that it is related to chronic infection with Plasmodium falcipa-
rum malaria. It is therefore not surprising to find that the
number of patients admitted with severe anemia is highest
during the months after the rainy season (Fleming 1970,
Verhoeff et al. 1999).
Hemolysis as a factor in the development of megaloblastosis
in folate-deficiency anemia has been demonstrated by Chana-
rin et al. (1959) and P. falciparum infection is an important
cause in holoendemic malarious areas (Fleming et al. 1986). A
further group of patients who contribute to these severe he-
molytic anemias are those with sickle cell disease. This group
accounted for ⬍10% of all cases in Ibadan, Nigeria (Fullerton
and Watson-Williams 1962). What proportion of the remain-
der of severe anemias can be attributed to either malaria or
iron deficiency or both?
One approach to estimating the malaria-attributable ane-
mia component is to calculate this anemia excess in primi- FIGURE 3 Relative risk for different severities of anemia in primi-
gravidae compared with multigravidae and attribute this ex- gravidae compared with multigravidae. Y ⫽ 7.48 ⫺ 1.2X ⫹ 0.056X2, R2
cess to their greater exposure to malaria. This assumption is ⫽ 99.6%; P ⫽ 0.004. Hb, hemoglobin; CI: confidence interval.
612S SUPPLEMENT

TABLE 10 TABLE 12
Population-attributable risk (PAR) estimates (%) due to Deaths attributable to iron deficiency anemia (all forms) in
malarial anemia in primigravidae using two different methods women 15– 44 y old1,2
of calculation
Deaths 1990 Deaths 2000 (projected)
Method for calculation of PAR
Number Rate (per Number Rate (per
Anemia risk in parasitemic Region (thousands) 100,000) (thousands) 100,000)
Anemia risk in
Hemoglobin primigravidae compared to non-parasitemic
cut-off value, compared with EME 0 0.2 0 0.1
g/L multigravidae1 Papua New Guinea2 Malawi3 FSE 0 0.2 0 0.1
India 5 2.8 2 0.9
China 4 1.3 1 0.4
⬍110 3.9 3.3 1.8
OAI 4 2.2 2 0.8
⬍100 3.8 — —

Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019

SSA 2 2.2 1 0.9
⬍90 6.1 — —
LAC 3 2.6 1 1.2
⬍80 12.1 14.52 14.63
MEC 2 1.9 1 0.8
⬍70 18.0 — —
World 20 1.7 8 0.7
1 From Figure 3. Assumes the proportion of primigravidae among
1 Source: Global Health Statistics by Murray and Lopez (World Bank
pregnant women is 25%.
2 From Brabin et al. (1990). 1993).
2 Abbreviations: EME, established market economies; FSE, formerly
3 From Verhoeff et al. (1999).
socialist economies; OAI, other Asia and islands; SSA, sub-Saharan
Africa; LAC, Latin American countries; MEC, Middle Eastern crescent.

P ⫻ 共1 ⫺ PAR m) ⫻ (CFR)
nant women in the denominator population, whereas the
⫽ maternal mortality from nonmalarious anemia estimate in this analysis is per 100,000 live births in primi-
where P is the prevalence of severe anemia, PARm and (1- gravidae. The difference between these estimates highlights
PARm) are the PAR estimates, respectively, for malarial and the fact that the risks for anemia-related mortality are greater
nonmalarial severe anemia in primigravidae, and CFR is the for the pregnant population and include several nutritional
case fatality rate (taken as 1.0% from Fig. 2). Through the use factors other than iron deficiency.
of this formula, then, in a holoendemic malarious area with a These calculations suggest that nutritional deficiency is a
5% severe anemic prevalence (Hb ⬍70 g/L), there would be 9 major component of severe anemia deaths even in malarious
severe malaria anemia-related deaths per 100,000 live births to areas. The calculations were based on primigravidae, but this
primigravidae and 41 nonmalarial anemia-related deaths (Ta- conclusion should apply to multigravidae, who are less suscep-
ble 11). tible to malarial infection and may have a higher prevalence of
Table 12 summarizes deaths attributable to iron-deficiency nutritional deficits and iron-deficiency anemia than primigrav-
anemia (all forms) in women 15– 44 y old and published by idae (Isah et al. 1985).
WHO (1993) as part of their Global Burden of Disease Statistical
Reports. The mortality rates per 100,000 attributable to iron- DISCUSSION AND CONCLUSIONS
deficiency anemia were ⬍ 2.8 per 100,000 population for the The more severe the anemia, the more likely it is to have
regional estimates in 1990, and projected deaths were lower for multiple causes and not be related solely to iron deficiency.
the year 2000. The estimate for sub-Saharan Africa (2.2 per This creates difficulties in establishing attributable risk. Be-
100,000 population) is much lower than the value derived cause several factors contribute to the prevalence and severity
above for nonmalarial pregnancy-related anemia mortality (41 of anemia, it cannot be assumed that distinct epidemiological
per 100,000 live births). The difference is influenced by the parameters predict the effect of anemia on maternal mortality.
method of calculation, which, for the global burden of disease This is a difficulty in an analysis that aims to identify specific
estimate, includes the total number of pregnant and nonpreg- components of attributable risk. The specific nonmalarial
components (mainly nutritional) of this attributable risk can
be estimated, but the proportion of these related specifically to
TABLE 11 iron-deficiency anemia, while uncertain, could be substantial.
Because moderate anemias are common and less strongly
Malaria and nonmalarial factors contributing to severe anemia
associated with malaria, nutritional deficiency anemias would
mortality in primigravidae living in malarious areas1 comprise the larger component of anemia-attributable mater-
nal mortality. This result highlights the need to determine
Severe anemia (hemoglobin ⬍70 g/L)
prevalence as proportion Malarial Nonmalarial mechanisms by which nutritional deficiency anemia, espe-
cially iron deficiency, could increase maternal mortality. Nu-
0.05 9.02 412 tritional deficiency may impair immune responsiveness, and in
0.06 10.1 49 nonpregnant women, iron-deficiency anemia has been associ-
0.07 12.6 57 ated with increased risk of death from circulatory disease
0.08 14.4 66 (Elwood et al. 1974). Iron deficiency is likely to be a major
0.09 16.2 74
0.10 18.0 82
contributory cause, although vitamin A deficiency could also
be important. Routine supplementation with vitamin A in a
1 PARm ⫽ 0.18 (taken from Table 10); 1-PARm ⫽ 0.82; case fatality large trial in Nepal reduced maternal mortality, but the mech-
rate taken as 1.0%. anisms were poorly defined and not obviously attributable to
2 Per 100,000 live births in primigravidae. anemia reduction (West et al. 1999). Folate deficiency may
 ANEMIA AND MATERNAL MORTALITY
also be important (Baily 1995). HIV infection, which is com-
mon in some pregnant populations in Africa and in some
studies has been associated with lower Hb levels, could en-
hance the effect of nutritional deficits on mortality risk.
Figure 2 showed that high Hb values (⬎130 g/L) were
associated with slightly increased mortality risk. This result
was obtained through the inclusion of the data of Harrison and
Rossiter (1985), which showed a marked increase in mortality
risk in women with hematocrits ⬎0.45. The explanation for
this is not known but could be related in part to dehydration
and hemoconcentration in emergencies. Mortality in nonpreg-
nant Caucasian women with high hematocrits was attributed
to higher cholesterol and blood viscosities in such subjects and
was related in part to cardiovascular disease (Elwood et al.
1974). Similar mechanisms may apply in women from devel-
oping countries, but some caution is required in interpreting
this observation because the result is from a single study.
There is almost no evidence that the treatment of anemia
other than with exchange transfusion (Fullerton and Turner
1962) or judicious use of blood transfusion (Lawson and Lister
1954), or treatment of acute severe malarial anemia (Gilles et
al. 1969) lowers risk of maternal mortality. A controlled in-
tervention trial would be a stronger approach, but this would
require a very large sample size and may not be ethically
acceptable. Thus, indirect methods of analysis are of particular
relevance in demonstrating the strength of associations of
anemia with maternal mortality. There are several limitations
to this approach that have been mentioned previously, not
least that the methods of Hb measurement vary (methods
include Sahli, Talquist, hematocrit, hemacue, Coulter counter
techniques and use of optical spectrophotometers). However,
this analysis has identified a large number of reports and the
strength of statistical associations can be adequately tested.
Estimates of PAR can be defended on the basis of the strong
association between severe anemia and maternal mortality,
but not for mild or moderate anemia. The policy implications
of this are, first, that some reduction in maternal mortality
should be achievable in developing countries through reduc-
tion in severe maternal anemia, with the greatest effect result-
ing from reductions in both malaria and nutritional anemias.
This conclusion contrasts with the situation in Western coun-
tries, where neither historical review nor review of obstetric
literature identified a plausible contribution of nutritional
factors to the decline in maternal mortality (Ronsmans et al.
1999). The size of this effect is likely to be small unless there
is a very high prevalence of severe anemia in the population.
However, the evidence is insufficient for or against treatment
of iron-deficiency anemia as a preventive measure for maternal
mortality. Second, with good antenatal and obstetric care,
most anemia-related deaths are preventable, and policies to
reduce anemia prevalence should not be divorced from efforts
to provide adequate antenatal and delivery facilities for
women in developing countries. Putting into operation nutri-
tion interventions as the magic pill approach will have to
compete with budgets allocated for essential obstetric care.
Finally, iron deficiency and malarial anemia should be treated
differently from other categories of risk in maternal health
such as height, weight, age, parity, previous history and use of
antenatal care services. Iron-deficiency anemia, like malarial
anemia, is in fact a complication, a medical condition that
requires treatment. The broad use of terminology, which clus-
ters together such unrelated criteria, could be detrimental to
effective health care strategies (Rhode 1995).
John Lawson, in his classic annual report in 1954, con-
cluded that it was hoped that maternal (and fetal) loss from
anemia would show a steady decline in the future. In his view,
613S
the declining level of Hb in some patients meant that they
reached a point of no return and would die however they were
treated. Fifty years later, maternal and fetal losses are still
unacceptably high, although today we have better ways of
preventing women from reaching that point of no return.
ACKNOWLEDGMENTS
We thank James Bunn for finding the early report by Professor
John Lawson and U. Lister (1953–1954), Jean Taylor for expert
secretarial assistance and several colleagues for kindly helping with
data sources and references.
Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019
LITERATURE CITED
Abouzahr, C. & Royston, E. (1991) Maternal mortality: a global factbook. World
Health Organization, Geneva, Switzerland.
Arkutu, A. A. (1979) Pregnancy and labour in Tanzanian primigravidae aged 15
years and under. Int. J. Gynaecol. Obstet. 16: 128 –131.
Bailey, L. B., ed. (1995) Folate in Health and Disease. Marcel Dekker, New
York, NY.
Barr, F., Brabin, L., Agbaje, S., Buseri, F., Ikimalo, J. & Briggs, N. (1998)
Reducing iron deficiency anaemia due to heavy menstrual blood loss in
Nigerian rural adolescents. Public Health Nutr. 1: 249 –257.
Beard, J. L. (1994) Iron deficiency: assessment during pregnancy and its
importance in pregnant adolescents. Am. J. Clin. Nutr. 59 (suppl.): 5025–
5105.
Brabin, B. J. (1983) An analysis of malaria in pregnancy in Africa. Bull. WHO
61: 1005–1016.
Brabin, B. J. (1985) Epidemiology of infection in pregnancy. Rev. Infect. Dis. 7:
579 – 603.
Brabin, B. J., Ginny, M., Sapau, J., Galme, K. & Paino, J. (1990) Consequences
of maternal anaemia on outcome of pregnancy in a malaria endemic area of
Papua New Guinea. Ann. Trop. Med. Parasitol. 84: 11–24.
Brabin, B. J. & Rogerson, S. (2000) The epidemiology and outcome of mater-
nal malaria. In: Malaria in Pregnancy: Deadly Parasite Susceptible Host (Duffy,
P. & Fried, M., eds.). Harwood Academic Publishers, Newark, NJ (2001).
Brabin, L., Verhoeff, F. H., Kazembe, P., Brabin, B. J., Chimsuku, L. & Broadhead,
R. (1998) Improving antenatal care for pregnant adolescents in southern
Malawi. Acta Obstet. Gynaecol. Scand. 77: 402– 409.
Brock, J. H. (1999) Iron and the immune system. In: Iron and Infection, 2nd ed.
(Bullen, J. J. & Griffiths, E., eds.). John Wiley and Sons, New York, NY.
Chanarin, I., Dacie, J. V. & Mollin, D. L. (1959) Folic acid deficiency in haemo-
lytic anaemia. Br. J. Haematol. 5: 245–256.
Cheng-Chi, I., Agoestina, T. & Harbin, J. (1981) Maternal mortality at twelve
teaching hospitals in Indonesia—an epidemiologic analysis. Int. J. Gynaecol.
Obstet. 19: 259 –266.
DeMaeyer, E. & Adiels-Tegman, M. (1985) The prevalence of anaemia in the
world. World Health Stat. Q. Rep. 38: 302–316.
Diallo, M. S., Diallo, T. S., Diallo, F. B., Diallo, Y., Camara, A. Y. & Onivogui, G.
(1995) Anémie et grossesse. Rev. Fr. Gynecol. Obstet. 90: 138 –141.
Elwood, P. C., Waters, W. E., Bentjamin, I. T. & Sweetnam, P. M. (1974)
Mortality and anaemia in women. Lancet 1: 891– 894.
Fazio-Tirrozo, G., Brabin, L., Brabin, B., Agbaje, O., Harper, G. & Broadhead, R.
(1998) A community based study of vitamin A and vitamin E status of
adolescent girls living in the Shire Valley, Southern Malawi. Eur. J. Clin. Nutr.
52: 637– 642.
Fleming, A. F. (1970) Seasonal incidence of anaemia in pregnancy in Ibadan.
Am. J. Clin. Nutr. 23: 224 –230.
Fleming, A. F. (1989a) The aetiology of severe anaemia in pregnancy in Ndola,
Zambia. Ann. Trop. Med. Parasitol. 83: 37– 49.
Fleming, A. F. (1989b) Anaemia in pregnancy in tropical Africa. Trans. R. Soc.
Trop. Med. Hyg. 83: 441– 449.
Fleming, A. F., Ghatoura, G.B.S., Harrison, K. A., Briggs, N. D. & Dunn, D. T.
(1986) The prevention of anaemia in pregnancy in primigravidae in the
guinea savanna of Nigeria. Ann. Trop. Med. Parasitol. 80: 211–233.
Foord, F. A., Fox-Rushby, J. & Weaver, L. T. (1992) The effect of a midwifery
outreach service on maternal anaemia and mortality in rural Gambia. Unpub-
lished report, pp 1–20. Dunn Nutrition Group, MRC Laboratories, Gambia.
Fullerton, W. T. & Turner, A. G. (1962) Exchange transfusion in treatment of
severe anaemia in pregnancy. Lancet 282: 75–78.
Fullerton, W. T. & Watson-Williams, E. J. (1962) Haemoglobin SC disease and
megaloblastic anaemia in pregnancy. J. Obstet. Gynaecol. 69: 729 –735.
Garner, P. & Brabin, B. J. (1994) A review of randomised controlled trials of
routine antimalarial drug prophylaxis during pregnancy in endemic malarious
areas. Bull. WHO 72: 89 –99.
Giglioli, G. (1972) Changes in the pattern of mortality following eradication of
hyperendemic malaria from a highly susceptible community. Bull. WHO 46:
181–202.
Gilles, H. M., Lawson, J. B., Sibelas, M., Voller, A. & Allan, N. (1969) Malaria,
anaemia and pregnancy. Ann. Trop. Med. Parasitol. 63: 245–263.
Granje, A. C., Machungo, F., Comes, A., Bergstrom, S. & Brabin, B. (1998)
614S SUPPLEMENT
Malaria related mortality in urban pregnant women in Mozambique. Ann. Trop.
Med. Parasitol. 92: 257–263.
Harrison, K. A. (1975) Maternal mortality in anaemia in pregnancy. West Afr.
Med. (April): 27–31.
Harrison, K. A. (1982) Anaemia, malaria and sickle cell disease. Clin. Obstet.
Gynaecol. 9: 445– 447.
Harrison, K. A. (1989) Tropical obstetrics and gynaecology. 2. Maternal mor-
tality. Trans. R. Soc. Trop. Med. Hyg. 83: 449 – 453.
Harrison, K. A. & Rossiter, C. E. (1985) Maternal mortality. Br. J. Obstet.
Gynaecol. 92 (suppl. 5): 100 –115.
Isah, H. S., Fleming, A. F., Ujah, I.A.O. & Ekwempu, C. C. (1985) Anaemia and
iron status of pregnant and non-pregnant women in the guinea savanna of
Nigeria. Ann. Trop. Med. Parasitol. 79: 485– 493.
Johnson, J.W.C. & Ojo, O. A. (1967) Amniotic fluid oxygen tensions in severe
maternal anemia. Am. J .Obstet. Gynecol. 97: 499 –506.
Konar, M., Sikdar, K., Basak, S. & Lahiri, D. (1980) Maternal mortality. Ten
years survey in Eden Hospital. J. Indian Med. Assoc. 75: 45–51.
Lawson, J. B. (1962) Maternal mortality in West Africa. Ghana Med. J. (De-
cember): 31–36.
Lawson, J. & Lister, U. G. (1954) Clinical Report of the Department of Obstet-
rics, University College, Ibadan, Nigeria. April 1st 1953–December 31st, 1954.
Vail and Company, Ltd., London, UK.
Llewellyn-Jones, D. (1965) Severe anaemia in pregnancy as seen in Kuala
Lumpur, Malaysia. Aust. N. Z. J. Obstet. Gynaecol. 5: 191–197.
Macleod, J. & Rhode, R. (1998) Retrospective follow-up of maternal deaths
and their associated risk factors in a rural district of Tanzania. Trop. Med. Int.
Health 3: 130 –137.
McDermott, J., Slutsker, L., Steketee, R. W., Wirima, J. J., Breman, J. G. &
Heymann, D. L. (1996) Prospective assessment of mortality among a
cohort of pregnant women in rural Malawi. Am. J. Trop. Med. Hyg. 55: 66 –70.
Murray, C.J.L. & Lopez, A. D. (1994) Global and regional causes of death
patterns in 1990. In: Global Comparative Assessments in the Health Sector—
Disease Burden, Expenditures and Intervention Packages, pp. 21–54. WHO,
Geneva, Switzerland.
Oja, O. A. (1965) The pattern of anaemia in Western Nigeria. J. Trop. Med. Hyg.
68: 32–36.
Osbourne, G. K., Howat, R. C. & Jordan, M. M. (1981) The obstetric outcome
of teenage pregnancy. Br. J. Obstet. Gynaecol. 88: 215–221.
Pelletier, D., Frongillo, E. A. & Habicht, J.-P. (1993) Epidemiologic evidence for
a potentiating effect of malnutrition on child mortality. Am. J. Public Health 83:
1130 –1133.
Rhode, J. E. (1995) Removing risk from safe motherhood. Int. J. Gynaecol.
Obstet. 50 (suppl. 2): S3–S10.
Ronsmans, C., Campbell, O. & Collumbien, M. (1999) Slight modifications in
definitions could alter interpretation of results. Br. Med. J. 319: 1201.
Rush, D. (2000) Nutrition and maternal mortality in the developing world.
Am. J. Clin. Nutr. (in press).
Sarin, A. R. (1995) Severe anaemia of pregnancy, recent experience. Int. J.
Gynecol. Obstet. 50 (suppl. 1): S45–S49.
Shulman, C. E., Dorman, E. K., Cutts, F., Kawuondo, K., Bulmer, J. N., Peshu, N.
& Marsh, K. (1999) Intermittent sulphadoxine-pyrimethamine to prevent
severe anaemia secondary to malaria in pregnancy: a randomised placebo-
controlled trial. Lancet 353: 632– 636.
Starrs, A. (1987) Preventing the tragedy of maternal deaths. A report of the
International Safe Motherhood Conference, Nairobi, Kenya. World Bank,
Washington, DC.
Tasker, P.W.G. (1958) Anaemia in pregnancy. A five year appraisal. Med. J.
Malaya 8: 3– 8.
Thonneau, P., Toure, B., Cantrelle, P., Barry, T. M. & Papiernik, E. (1992) Risk
factors for maternal mortality: results of a case-control study conducted in
Conakry (Guinea). Int. J. Gynaecol. Obstet. 39: 87–92.
United Nations Children’s Fund (1999) State of the World’s Children. United
Nations Children’s Fund, New York, NY
Verhoeff, F. H., Brabin, B. J., Chimsuku, L., Kazembe, P. & Broadhead R. (1999)
An analysis of the determinants of anaemia in pregnant women in rural
Malawi—a basis for action. Ann. Trop. Med. Parasitol. 93: 119 –133.
West, K. P., Katz, J., Khatry, S. K., LeClerq, S. C., Pradhan, E. K., Shrestha, S.,
Connor, P., Dali, S., Christian, P., Pokhrel, R. & Sommer, A. (1999) Double
blind, cluster randomised trial of low dose supplementation with vitamin A or
␤ carotene on mortality related to pregnancy in Nepal. Br. Med. J. 318:
570 –575.
Wickramasuriya, G.A.W. (1937) Malaria and Ankylostomiasis in the Pregnant
Woman. Their More Serious Complications and Sequelae. Oxford University
Press, London, UK.
World Bank (1993) World Development Report 1993: Investing in Health.
World Bank, Washington, DC.
World Health Organization (1991) Maternal Mortality—a Global Factbook.
WHO, Geneva, Switzerland.
World Health Organization (1992a) International Statistical Classification of
Diseases and Relative Health Problems. 10th rev. ed., vols. 1 and 2. WHO,
Geneva, Switzerland.
World Health Organization (1992b) The prevalence of anaemia in women: a
tabulation of available information. Maternal Health and Safe Motherhood
Programme, Geneva, Switzerland.
World Health Organization (1996) Revised 1990 estimate of maternal mortality.
A new approach by WHO and UNICEF. WHO, Geneva, Switzerland.
Zucker, J. R., Lackritz, E. M. & Ruebush, T. K. (1994) Anaemia, blood trans-
fusion practices, HIV and mortality among women of reproductive age in
Western Kenya. Trans. R. Soc. Trop. Med. Hyg. 88: 173–176.
DISCUSSION
Participants: Pelletier, Beard, Brabin, Allen, Rasmussen,
Habicht, Tielsch, Premji, Oppenheimer, Stoltzfus, Horton
Dr. Pelletier: Several comments on the Nigerian studies,
which report the lowest hemoglobin values. They are all from
around 1960, all from one country, and all with a certain level
and type of obstetric care and they are clearly pulling the risk
curve up. So, if you fit various models to that, it turns out the
best fit is exponential. I am trying to zero in on the mild and
moderate range, independent of any sort of very powerful data
Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019
points that are pulling it up. There does not seem to be any
relationship if you exclude those four studies. So, if we want to
look for a dose-response relationship, we really have to see
evidence of it over the entire distribution and be wary of
influential data points, especially because those seem excep-
tional.
Dr. Beard: Do we have any documentation of the kind or
type of obstetric care that was given then relative to what is
available in that part of the world now?
Dr. Brabin: I think the obstetrics has improved greatly. I do
not think we can ignore those four points. These old Nigerian
studies document the situation with relatively little interfer-
ence, where women have desperately low hemoglobin and are
dying. These are the only data that exist in the world.
Dr. Pelletier: I am not suggesting we ignore them. It would
be, probably, the causal effect of severe anemia.
Dr. Beard: Are you willing to allow those four studies to
stay in the analysis?
Dr. Pelletier: Yes, for the purposes of making inferences
about severe anemia, but if we start fitting curves they will
begin to have a distorting influence on our judgment. Imagine
that it was absolutely horizontal and then it goes up with
severe anemia. If you fit an exponential curve to that, it is
going to fit very nicely.
Participant: Are there other sub-Saharan Africa studies of
the mid-1960s in your pile of 28 studies? I am just trying to
figure out what the situation would be if you took them all out.
Dr. Pelletier: Actually, I am looking at that curve again
and Brabin is right. It is not just those four. It turns out that
eight data points are up there. Seven of the eight are from
Nigeria. One is from Guinea. So, they are all from sub-Saharan
Africa. Some of the less extreme points are also from Africa.
Dr. Allen: Were they all the same investigators?
Dr. Brabin: No. There were three different groups of in-
vestigators.
Dr. Beard: One of the things that generally concerns me
about hospital-based data in resource-poor environments is
what gets you admitted to a hospital. What gets you admitted
to a hospital if you show up with a hemoglobin of 30 or 40 g/L
is going to be very different from what gets you admitted if you
show up with a hemoglobin of 60 or 80 g/L. Right? So, it seems
to me that the people who are showing up with hemoglobins
of 60 or 80 g/L are being admitted primarily for completely
other reasons.
Participant: That would tend to diminish the relationship.
Dr. Beard: That would tend to inflate the mortality risk
among the moderately anemic because they are selected for a
higher risk profile. I think the question is in the mild-to-
moderate range of anemia, what is it that hospital-based data
can tell us in this kind of environment, and how much is
selection bias influencing our assessment of the relationship.
Dr. Habicht: At least down to 60 g/L or so, I do not see any
 ANEMIA AND MATERNAL MORTALITY
admissions because of hemoglobin. They are all there for other
reasons. Now, is there any reason to believe that those other
reasons would be different across the hemoglobin range? Prob-
ably not.
Dr. Brabin: Any woman who comes to the hospital what-
ever her hemoglobin is admitted so that she can deliver her
baby.
Dr. Habicht: I think we need to divide the conversation
into different parts. First, do we believe that that excess risk
below 50 g/L is really there? It seems to me that everybody
believes that. So, the second question is whether there is any
excess risk above 50 g/L? From these data, if you just took the
fitted lines away so you were not being prejudiced, you would
not see a relationship above 50 g/L. This is an underestimate
of the true relationship. If it is a flat line, it is an underestimate
because those people are being selected into the hospital
sample because they are likely to die.
Dr. Tielsch: So, you think the comorbidity profile of
women with hemoglobin 60 g/L at admission for birth is the
same as for women who have 100 g/L at admission. I suggest
that is not probably true, in fact, because we know that anemia
is related to poverty and poor health. So that women who get
admitted—who are coming to the hospital to deliver—and
have got an admission hemoglobin of 100 g/L are likely to be
healthier.
Dr. Habicht: Then your conclusion is very clear. Taking
that into account—rather than this apparent flat relationship
between hemoglobin and mortality—you then have a positive
relationship between hemoglobin and case fatality above 60
g/L, going up to the right-hand side.
Dr. Tielsch: I cannot figure out what the true relationship
is.
Dr. Premji: I wanted to ask Brabin whether he has any clue
about the association between malaria and mortality.
Dr. Brabin: I cannot enlighten you. We have done a
retrospective analysis of a very large data set from the north
coast of Papua New Guinea and an equally large data set from
the highlands of New Guinea. In the malaria-endemic north
coast, for the same level of hemoglobin in the mother at
delivery there was a significantly increased risk of postpartum
hemorrhage. This is just a hint that malaria is in some way
615S
related to the risk, because postpartum hemorrhage is associ-
ated with mortality. I do not know the mechanism.
Dr. Tielsch: This is outside primagravida—independent of
that?
Dr. Brabin: Independent.
Dr. Oppenheimer: I remember seeing a review about ma-
ternal mortality in Nigeria in the 1960s and they had a real
problem with anemia and heart failure because they did not
have effective rapid-acting diuretics. If they were transfused,
their heart failure got worse. In fact, they were trying to use
exchange transfusion to cope with this problem. So, there was
a particular problem of management of severe anemia and
heart failure.
Dr. Brabin: The Nigerian studies do give clinical reasons
Downloaded from https://academic.oup.com/jn/article-abstract/131/2/604S/4686842 by guest on 03 March 2019
for death, and heart failure is mentioned as one cause of death.
It has been shown in Nigeria that exchange transfusion dra-
matically reduced the risk of death in these severely anemic
women.
Dr. Beard: Some of us may recall Henry J. Whipple, who
won the Nobel Prize in Medicine for looking at the effects of
severe anemia on cardiovascular adaptation and cardiac fail-
ure. So, this question of severe anemia, oxygen transport and
cardiovascular adaptation has been around for a really, really
long time.
Dr. Stoltzfus: It is remarkable that this hemoglobin mor-
tality risk curve is flat across the wide hemoglobin range about
60 g/L, given all we expect from other anemia survival curves
in nonpregnant adults. I think that reverse causality is part of
all these anemia-survival associations, but the fact is that they
are there, even in well-cared-for populations, even in surgical
patients, who are not necessarily suffering from an infectious
disease that is causing their surgery. The fact that this occurs
in British data makes it astounding to me that that is absent in
African data. I do not know what to conclude from that.
Dr. Pelletier: Bear in mind that these data points are
assembled from 12 different studies. So, the picture is a bit
deceiving. We are used to having a reference group and several
groups of increasing severity, and then you would expect to see
something like that. However, this is a meta-analysis. There is
lots of stuff going on between these data points besides differ-
ent degrees of anemia.