You are on page 1of 9

MAJOR ARTICLE

HIV/AIDS

Postpartum Tuberculosis Incidence and Mortality


among HIV-Infected Women and Their Infants in
Pune, India, 20022005
Amita Gupta,1 Uma Nayak,2 Malathi Ram,2 Ramesh Bhosale,3 Sandesh Patil,3 Anita Basavraj,3 Arjun Kakrani,3
Sheeja Philip,4 Dipali Desai,3 Jayagowri Sastry,4 and Robert C. Bollinger,1,2 for the Byramjee Jeejeebhoy Medical
CollegeJohns Hopkins University Study Group
1

(See the editorial commentary by Mofenson and Laughon on pages 2503)

Background. In contrast with many other countries, isoniazid preventative therapy is not recommended in
clinical care guidelines for human immunodeficiency virus (HIV)infected persons with latent tuberculosis (TB)
in India.
Methods. Seven hundred fifteen HIV-infected mothers and their infants were prospectively followed up for 1
year after delivery at a public hospital in Pune, India. Women were evaluated for active TB during regular clinic
visits, and tuberculin skin tests were performed. World Health Organization definitions for confirmed, probable,
and presumed TB were used. Poisson regression was performed to determine correlates of incident TB, and
adjusted probabilities of mortality were calculated.
Results. Twenty-four of 715 HIV-infected women who were followed up for 480 postpartum person-years
developed TB, yielding a TB incidence of 5.0 cases per 100 person-years (95% confidence interval [CI], 3.27.4
cases per 100 person-years). Predictors of incident TB included a baseline CD4 cell count !200 cells/mm3 (adjusted
incident rate ratio [IRR], 7.58; 95% CI, 3.0718.71), an HIV load 150,000 copies/mL (adjusted IRR, 3.92; 95%
CI, 1.699.11), and a positive tuberculin skin test result (adjusted IRR, 3.08; 95% CI, 1.277.47). Three (12.5%)
of 24 women with TB died, compared with 7 (1.0%) of 691 women without TB (IRR, 12.2; 95% CI, 2.0353.33).
Among 23 viable infants with mothers with TB, 2 received a diagnosis of TB. Four infants with mothers with TB
died, compared with 28 infants with mothers without TB (IRR, 4.71; 95% CI, 1.1913.57). Women with incident
TB and their infants had a 2.2- and 3.4-fold increased probability of death, respectively, compared with women
without active TB and their infants, controlling for factors independently associated with mortality (adjusted IRR,
2.2 [95% CI, 0.63.8] and 3.4 [95% CI, 1.2210.59], respectively).
Conclusions. Among Indian HIV-infected women, we found a high incidence of postpartum TB and associated
postpartum maternal and infant death. Active screening and targeted use of isoniazid preventative therapy among
HIV-infected women in India should be considered to prevent postpartum maternal TB and associated motherto-child morbidity and mortality.
Tuberculosis (TB) is the most common HIV-related
opportunistic infection and the most important cause
of morbidity and mortality in HIV-infected adults in

Received 14 January 2007; accepted 13 March 2007; electronically published


4 June 2007.
Reprints or correspondence: Dr. Amita Gupta, Centre for Clinical Global Health
Education, Division of Infectious Diseases, 600 N. Wolfe St., Phipps 540B,
Baltimore, MD 21287 (agupta25@jhmi.edu).
Clinical Infectious Diseases 2007; 45:2419
 2007 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2007/4502-0016$15.00
DOI: 10.1086/518974

the developing world [1]. In areas with a high prevalence of HIV infection and TB, active case finding for
TB has revealed that TB is also a concern during pregnancy and at the time of delivery of a newborn [2, 3].
Data suggest that both TB and HIV infection have been
identified as independent factors for maternal mortality
and perinatal outcomes; in combination, TB and HIV
infection have a greater detrimental impact than their

The views expressed in this manuscript do not necessarily represent the views
of the National Institutes of Health, Fogarty International Centre, the Johns Hopkins
University, or Byramjee Jeejeebhoy Medical College.

HIV/AIDS CID 2007:45 (15 July) 241

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

Infectious Diseases, Johns Hopkins University School of Medicine, and 2Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland;
and 3Byramjee Jeejeebhoy Medical College and 4Byramjee Jeejeebhoy Medical CollegeJohns Hopkins University Maternal Infant Transmission
Study, Pune, India

METHODS
From 16 August 2002 through 31 December 2005, HIV-infected
mother and infant pairs were enrolled and prospectively followed up as part of an ongoing National Institutes of Health
funded phase III randomized clinical trial to assess the role of
nevirapine therapy administered to infants during breastfeeding. Although the primary outcome of HIV transmission is
pending completion of the trial, this study is an analysis of the
data specific to TB that were collected as part of the secondary
outcome of maternal and infant morbidity and mortality.
HIV-infected women at a large urban, public teaching hospital in Pune, Maharashtra, India, were enrolled during their
third trimester, at delivery, or within 1 week after delivery.
Mother and infant pairs were followed up prospectively for up
to 12 months after delivery. At each scheduled visit (at weeks
1, 2, 3, 4, 5, 6, 10, and 14 and at 6, 9, and 12 months), women
and their infants underwent clinical examination and select
242 CID 2007:45 (15 July) HIV/AIDS

laboratory investigations. All infants were given bacille Calmette-Guerin vaccination at birth, according to Indian pediatric guidelines. HIV infection in infants was confirmed by HIV
DNA PCR testing and quantitative HIV load measurements.
The maternal baseline hemoglobin concentration was obtained
at enrollment in the study, and the CD4 cell count and HIV
load measurement that were obtained closest to delivery were
used in this analysis.
Informed consent was obtained from all participants in the
study. Human experimentation guidelines of the US Department of Health and Human Services and the participating institutions were observed in the conduct of this research. Informed consent procedures and this research were reviewed
and approved by independent ethical committees in Pune, India, and in the United States.
TB screening and diagnosis. At delivery, women were offered tuberculin skin tests (TSTs; also known as Mantoux tests),
according to standard methods [16], and were evaluated during
regular clinic visits for active TB. If the TST results were positive
(defined as an induration 5 mm in diameter for HIV-infected
persons) [17] or symptoms suggested active TB (e.g., persistent
cough, fever, and weight loss), chest radiographs were performed and clinical and laboratory information were obtained
using a standardized data collection form. TB was confirmed
when Mycobacterium tuberculosis was cultured from the mother,
and TB was probable when (1) acid-fast bacilli were detected
in maternal sputum smears by microscopic examination, (2)
histological features suggested TB, or (3) congenital TB was
confirmed after the culture of M. tuberculosis from the neonate.
TB was suspected when the mother had only clinical and
radiological evidence suggesting TB and showed a response to
anti-TB therapy. Some microbiological data were missing. Culture is not a routine practice in India and, according to national
TB guidelines, is only recommended when there is treatment
failure or lack of response to treatment; for 7 cases, either the
clinician did not send appropriate samples for culture or appropriate media was not available, and for 4 cases, sputum
samples for acid-fast bacilli testing were not adequately obtained. A diagnosis of TB in infants was determined according
to standard methods [18]. Because of the difficulty of confirming cases, most infant TB cases were determined using clinical
and radiological criteria (including response to treatment), as
well as history of close contact to persons with active TB.
Analyses. Data were analyzed using Stata statistical software, version 9.0 (Stata). Students t test for means were used
when continuous variables were normally distributed. Nonparametric Mann-Whitney U tests or Fisher exact tests were
used when continuous variables were not normally distributed
or when cell size was 5 observations, respectively; x2 tests
were used for discrete variables. Log-linear (Poisson) regression
analyses were used with TB incidence as the outcome variable.

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

individual effects [4, 5]. In the pre-HIV era, !6% of maternal


mortality was shown to be caused by TB [6]. In the HIV era,
this statistic has dramatically changed. Coinfection with TB and
HIV has been shown to account for 15% of maternal deaths
in teaching hospitals in southern Africa [5, 79]. Furthermore,
newborns exposed to an adult with active TB may be at high
risk for TB morbidity and mortality, but the risk has been
poorly quantified [10, 11].
India has the highest TB burden globally, accounting for
120% of the 8.8 million global cases each year, and up to 50%
of Indias population is estimated to have latent TB [12, 13].
TB kills 370,000 Indian adults annually; most of these persons
are aged 1554 years, and TB is a major cause of death in
women of reproductive age [12]. India is also estimated to be
one of the countries with the largest number of HIV-infected
persons in the world (5.25.7 million) [14], and TB is by far
the most important cause of morbidity and mortality in Indian
HIV-infected persons, accounting for 50% of deaths.
Although there is a high burden of TB and high risk for
active TB among HIV-infected persons in India, Indias national
HIV and TB guidelines do not promote isoniazid preventive
therapy (IPT) for HIV-infected persons [15]. There is legitimate
concern that widespread and inappropriate use of IPT could
increase the community burden of isoniazid-resistant TB infection and limit the success of the national TB-control program. The objective of our study was to determine the incidence
of active TB infections among postpartum Indian women and
to assess the impact of maternal HIV and TB coinfection on
maternal and infant outcomes in the absence of IPT. Such data
are needed, because delivery and the postpartum period represent important entry points for women and their infants into
the health care system and, therefore, ideal times to consider
an intervention that includes TB prevention and management.

RESULTS
A total of 715 HIV-infected women were followed up for 480.6
person-years (PY; median duration of follow-up, 0.96 PY; interquartile range [IQR], 0.271.0 PY). The characteristics of
the cohort are shown in table 1. Of 715 women, 688 (96.2%)
had a TST administered around the time of delivery and appropriately read within 72 h of delivery; 145 of these women
(21.1%) had positive TST results. Twenty-four postpartum incident TB infections were detected, yielding an incidence estimate of 5.0 cases per 100 PY (95% CI, 3.2-7.4 cases per 100
PY), with a median time after delivery to incident TB of 90
days (range, 4333 days). Seven cases were confirmed within
the first 2 weeks after delivery, suggesting that the women may
have had subclinical TB at the time of delivery. Fifteen women
(62.5%) had a baseline CD4 cell count 1200 cells/mm3, 15
(62.5%) had a baseline HIV load 150,000 copies/mL, and 9
(37.5%) had positive TST results at the delivery screening. Sixteen women received a diagnosis of pulmonary TB, 6 received
a diagnosis of extra-pulmonary TB, and 2 had both of these
types of TB. Six cases were confirmed, 12 were probable,
and 6 were suspected (12 [60%] of 20 women were smearpositive for TB, and 6 [35.3%] of 17 were culture-positive for
TB). All but 3 women initiated TB treatment (1 was lost to
follow-up, and the other 2 died before therapy was initiated).

Seven women completed treatment and were cured, 10 had


treatment ongoing at the time of analysis, and 4 discontinued
treatment. Three women (11.5%) died at 5, 6, and 11 months
after delivery; 2 of these deaths were due to TB, and 1 was due
to bacterial sepsis.
Compared with women who did not develop TB during the
study period, women with incident TB were more likely to be
unemployed (41.7% vs. 19.8 %; P p .009, by x2 test), to have
a lower median baseline CD4 cell count (333 cells/mm3 vs. 472
cells/mm3; P p .001), and to have a higher median baseline
HIV load (69,208 copies/mL vs. 11,430 copies/mL) (table 1).
Women who developed TB were also more likely than women
who did not develop TB to have positive TST results (37.5%
vs. 20.5%; P p .04, by x2 test) and to have at least moderate
anemia (hemoglobin concentration, 9 g/dL; 37.5% vs. 19.1%;
P p .03, by x2 test).
Correlates for incident maternal TB. In univariate Poisson
regression, the following baseline characteristics were associated
with TB incidence: older age, being employed, a CD4 cell count
200 cells/mm3, an HIV load 150,000 copies/mL, positive TST
results, and a hemoglobin concentration 9 g/dL (table 1).
In multivariate Poisson regression, a CD4 cell count 200
cells/mm3 was the strongest independent predictor and was
associated with a 7.58-fold increased incidence of TB (adjusted
incidence rate ratio [IRR], 7.58; 95% CI, 3.0718.71). An HIV
load 150,000 copies/mL (adjusted IRR, 3.92; 95% CI, 1.69
9.11), being employed (IRR, 3.00; 95% CI, 1.326.79), and
positive TST results (adjusted IRR, 3.08; 95% CI, 1.277.47)
were also independently associated with incident maternal TB
infection. Within 12 months after delivery of an infant, 3
(12.5%) of 24 mothers with TB died, compared with 7 (1.0%)
of 691 mothers without TB (IRR, 12.2; 95% CI, 2.0353.33).
Adjusting for CD4 cell count, HIV load, hemoglobin concentration, education, and age, the probability of maternal mortality within the first year after delivery was 3-fold higher among
women with postpartum incident TB than among those who
did not develop incident TB (0.9 cases per 100 PY vs. 0.4 cases
per 100 PY; adjusted IRR, 2.2; 95% CI, 0.63.8) (figure 1).
Infant characteristics and outcomes. Infants born to
women who subsequently developed TB did not differ from
infants born to women who did not develop incident TB with
regard to median birth weight (2500 g vs. 2600 g; P p .66),
median gestational age (38 weeks vs. 38 weeks), and median
duration of breastfeeding (105.5 days vs. 100 days) (table 2).
Infants with mothers with incident TB, however, were more
likely to have HIV infection (determined by infant HIV DNA
PCR) by 1 year of age than were infants with mothers who did
not have incident TB (37.5% vs. 9.1%; P ! .001). Thirteen infants received TB prophylaxis after their mothers developed TB
with either a combination of isoniazid and rifampin (administered to 12 infants) or isoniazid alone (administered to 1
HIV/AIDS CID 2007:45 (15 July) 243

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

Continuous variables were categorized into clinically meaningful groups (i.e., CD4 cell count !200 cells/mm3 or 200 cells/
mm3). Age, marital status, employment status, family type, education, TST status, CD4 cell count, HIV load, and maternal
hemoglobin concentration were covariates. Collinearity was
checked by calculating the variance inflation factors, which were
all !2.5. Variables that were significantly associated in the univariate analysis were analyzed using multivariate Poisson regression. Forward and backward stepwise selection of covariates
was performed using a significance level of P .05. Interactions
between age and other covariates were explored. Model fit was
assessed by using the Pearsons goodness-of-fit test, as well as
Akaikes information criterion [19]. Overdisperson was not evident in our model. Both forward and backward selection arrived at the same model as did Akaikes information criterion,
with the exception of 1 variable (maternal hemoglobin concentration). This variable was forced into the model, because
it has been identified as an independent predictor in published
studies. Lastly, adjusted probabilities for maternal and infant
mortality by maternal TB status were calculated, controlling
for all other statistically significant covariates using Poisson
regression. All continuous variables were centered. To calculate
infant mortality probabilities, we added key available covariates
for infant mortality, such as birth weight (low birth weight,
!2500 g), HIV PCR status, gestational age (preterm, !38
weeks), and infant mode of feeding.

Table 1. Characteristics of patients and unadjusted and adjusted incidence rate ratios (IRRs) of maternal
tuberculosis (TB) in a cohort of HIV-infected women in Pune, India, 20022005.

Characteristic

All patients

Patients
with incident
active TB
(n p 24)

Age, median years (IQR)

23 (1835)

25 (2127)

Patients
without TB
(n p 691)

Unadjusted IRR
(95% CI)

23 (2125)

1.14 (1.01.24)

Adjusted IRR
(95% CI)

Religion
Hindu

549 (76.8)

14 (58.3)

535 (77.4)

1.00 (0.891.14)

Other

166 (23.2)

10 (41.7)

156 (22.6)

Ref

686 (95.9)

23 (95.8)

663 (95.9)

0.69 (0.095.12)

29 (4.1)

1 (4.2)

28 (5.1)

Ref

Marital status
Married
Other
Educational status
None or 4th grade

274 (38.6)

12 (50.0)

262 (38.2)

1.88 (0.854.20)

14th grade

435 (61.4)

12 (50.0)

423 (61.8)

Ref

Joint

334 (47.3)

17 (70.8)

317 (46.5)

1.02 (0.911.13)

Nuclear

371 (52.7)

7 (29.2)

364 (53.5)

Ref

Employed

147

10 (41.7)

137 (19.8)

2.67 (1.196.02)

Unemployed

568

14 (58.3)

554 (80.2)

Ref

Family type

3.00 (1.326.79)

Ref

Parity
Median (range)

1 (05)

1.0 (03)

1.0 (05)

512 (71.6)

18 (75.0)

494 (71.5)

1.14 (0.452.87)

11

203 (28.4)

6 (25.0)

197 (28.5)

Ref

465 (318675)

333 (141401)

472 (322680)

50 (7.1)

9 (37.5)

41 (6.0)

8.80 (3.8520.11)

656 (92.9)

15 (62.5)

641 (94.0)

Ref

CD4 cell count


Median cells/mm3 (IQR)
200 cells/mm3
3

1200 cells/mm

7.58 (3.0718.71)

Ref

HIV load
Median copies/mL (IQR)

11,692 (230248,187)

69,208

11,430

150,000 copies/mL

169 (24.1)

15 (62.5)

154 (22.7)

5.67 (2.4812.96)

50,000 copies/mL

533 (75.9)

9 (37.5)

524 (77.3)

Ref

Mean mm  SD

3.5  6.6

6.2  8.5

3.4  6.5

5 mm

145 (21.1)

9 (37.5)

136 (20.5)

2.47 (1.085.65)

!5 mm

543 (78.9)

15 (62.5)

528 (79.5)

Ref

3.92 (1.699.11)

Ref

TST diameter

3.08 (1.277.47)

Ref

Hemoglobin concentration
Median g/dL (IQR)

11.2 (9.612.4)

10.2 (8.312.2)

11.2 (9.712.4)

9 g/dL

127 (19.8)

9 (37.5)

118 (19.1)

2.81 (1.236.41)

19 g/dL

514 (80.2)

15 (62.5)

499 (80.9)

Ref

2.05 (0.884.78)
Ref

NOTE. Data are no. (%) of patients, unless otherwise indicated. Baseline characteristics were assessed at the time of delivery
(i.e., CD4 cell count, HIV load, TST diameter, and hemoglobin concentration). Poisson regression was used to calculate unadjusted and
adjusted IRRs. IQR, interquartile range; Ref, reference; TST, tuberculin skin test.
a
b
c

P p .05.
P ! .01.
P ! .001.

infant). Three infants did not complete a full course of prophylaxis. The remaining infants did not receive prophylaxis
because parents refused treatment (1 infant), the mother had
extrapulmonary TB (4 infants) or was smear-negative for TB
(2 infants), the infant was lost to follow-up (1 infant), or the
infant died before prophylaxis was initiated (3 infants). Two
infants were suspected to have developed active TB (a third
suspected case occurred in an infant whose father, but not
mother, received a diagnosis of TB; data not shown). Four
244 CID 2007:45 (15 July) HIV/AIDS

(16.7%) of 24 infants of mothers with TB died, compared with


28 (4.1%) of 691 infants of mothers who did not have TB
(P ! .001, by x2 test). The unadjusted infant mortality rates were
25.8 cases per 100 PY for infants with mothers with TB, compared with 5.9 cases per 100 PY for infants with mothers without TB (IRR, 4.71; 95% CI, 1.1913.57). Adjusting for infant
gestational age, low birth weight, infant mode of feeding, and
infant HIV status, as well as for maternal factors, such as CD4
cell count, hemoglobin concentration, and HIV load, the prob-

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

Employment status

ability of infant mortality was 8.5 cases per 100 PY among


infants with mothers who developed incident TB, compared
with 2.5 cases per 100 PY among infants whose mothers did
not develop TB (adjusted IRR, 3.4; 95% CI, 1.2210.59) (figure 1).
DISCUSSION
In a cohort of postpartum HIV-infected women in India, we
identified a high maternal TB incidence of 5 cases per 100 PY
in the absence of IPT and found that many cases developed
early in the postpartum period. Our incidence rate is consistent
with the 3%10% annual incidence estimated by other studies,
including 1 study in India [2022]. We also confirmed several
findings involving postpartum HIV-infected women that have
been described in other cohorts involving nonpregnant adult
HIV-infected women. A CD4 cell count !200 cells/mm3 was
the strongest predictor of incident TB, associated with an 8fold increased incidence of TB, compared with a CD4 cell count
200 cells/mm3 [23]. However, nearly two-thirds of TB cases
occurred in women with a CD4 cell count 1200 cells/mm3,
which is above the threshold recommended for HAART initiation in India. Therefore, many postpartum Indian women
who are at high risk for TB would not benefit from the reduction in TB incidence that has been demonstrated with a
combination of HAART and IPT or HAART alone [24, 25].

We also determined that a baseline HIV load 150,000 copies/


mL was independently associated with a 3-fold increased risk
of incident TB [26, 27]. Finally, an important finding was the
impact that maternal postpartum TB had on overall infant
mortality within the first year of life. Infants with mothers who
developed incident TB had a 3-fold increased mortality rate,
compared with infants whose mothers did not have TB, controlling for infant HIV status and several other important potential confounders.
There are limited published data about postpartum TB in
the HIV era. A study of 120 South African women with positive
TST results (screening performed after delivery) found that
11% of the women had active TB [3]. In our study, 7 cases
were confirmed within the first 2 weeks after delivery; less than
one-half of the women in whom these cases occurred had a
positive TST result. This finding is concerning, because it suggests that these women may have had subclinical active TB at
the time of delivery. Some researchers have also suggested that
the early postpartum period may be a time of higher risk because of the nutritional stress of lactation, lack of sleep, rapid
hormonal changes, and depression of cell-mediated immunity
[28, 29]. Our finding that the median onset of maternal TB
was 3 months after delivery lends some credence to the early
postpartum period being a time of higher risk for developing
TB, but larger studies are needed to definitively address this
HIV/AIDS CID 2007:45 (15 July) 245

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

Figure 1. Adjusted maternal and infant mortality incidence rate ratios (IRRs) within the first year after delivery, by maternal tuberculosis (TB) status,
among a cohort of HIV-infected women and their infants in Pune, India (20022005). The unadjusted IRR for mothers was 12.2 (95% CI, 2.053.3;
P p .004), and the unadjusted IRR for infants was 4.7 (95% CI, 1.213.6; P p .017 ). The maternal IRR was adjusted for CD4 cell count, log HIV load,
hemoglobin concentration, age, and educational status at the time of delivery of the infant. Infant mortality was adjusted for HIV PCR status during
the first year of life, preterm birth (!38 weeks), low birth weight (!2500 g), infant mode of feeding, and the aforementioned maternal factors. Mortality
IRRs and 95% CIs were calculated using Poisson regression. aIRR, adjusted incidence rate ratio; PY, person-years.

Table 2. Characteristics and outcomes of infants born to HIV-infected women, by maternal incident tuberculosis
(TB) status, in a cohort in Pune, India, 20022005.
No. of infants
for whom data
were available

Incident active TB

Birth weight, median g (IQR)


Gestational age, median weeks (IQR)

704
706

2500 (23003000)
38 (3739)

Positive PCR results for HIV


Still birth

709
715

9/24 (37.5)
1/24 (4.17)

Breastfed
Duration of breastfeeding, median days (IQR)

715
715

Infant characteristic

Maternal TB status

21/24 (87.5)
105.5 (32.5199.5)

No TB
2600 (24002950)
38 (3838)
62/685 (9.10)
5/691 (0.72)
624/691 (90.3)
100 (97183)

P
.66
.44
!.001

.19
.65
.86

NOTE. Data are no. (%) of infants, unless otherwise indicated. A 2-sample Wilcoxon rank sum (Mann-Whitney U) test was used to
compare medians. A x2 test was used for comparison of discrete categories. Fishers exact test was used to calculate P values when the
cell size was 5 mm. PCR for infant HIV infection was assessed at 11 time points during the first year of life and confirmed with infant
HIV load measurements. Mode of infant feeding was also assessed at these 11 time points. IQR, interquartile range.

246 CID 2007:45 (15 July) HIV/AIDS

positive TST result was an independent predictor of incident


postpartum maternal TB, but the test had poor sensitivity
(37.5%) and positive predictive value (6.2%), although 92.9%
of women had a CD4 cell count 1200 cells/mm3. Our findings
are consistent with previous studies demonstrating poor predictive value of TST in regions where TB is highly endemic
[34]. However, it is possible that many of these women were
recently infected with HIV. Because our follow-up of these
women was limited to 12 months after delivery, our study did
not assess the longer-term predictive value of TST screening
among this population. Despite its overall low sensitivity, TST
may still have a role in identifying who is most likely to benefit
from IPT.
The optimal screening strategy and timing of IPT initiation
in HIV-infected pregnant women is not known but likely varies
in different regions, depending on access to prenatal care, TB
risk and incidence, and relative willingness of the patient to
receive IPT. Current US guidelines recommend delaying IPT
until after delivery, largely because of concern that pregnancy
may be a risk factor for isoniazid-induced hepatitis [1]. This
concern is largely based on a retrospective study of Hispanic
women, in which a 2.5-fold increased risk of isoniazid-induced
hepatitis and a 4-fold increased mortality rate occurred, but
both of these effects were not statistically significant [35]. In
the absence of controlled trials to determine the optimal time
to initiate IPT, a Markov decision analysis of HIV-uninfected
women found that IPT was most effective when initiated during
the antepartum period (at 20 weeks gestation) and that the
benefits outweighed the risks if the case-fatality rate was 10.45%
[36]. The benefit would be expected to be higher for HIVinfected women. Because many cases in our study were diagnosed around the time of delivery and during the early postpartum period, a strategy of antepartum IPT initiation would
likely be most effective. Assuming published efficacies for IPT
and our results, irrespective of CD4 cell count or TST results,
56 women would have to receive IPT during the antepartum

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

controversy [29, 30]. Small studies have suggested that TB and


HIV coinfectionassociated deaths have accounted for 15%
of maternal mortality within the first year after delivery in
teaching hospitals in southern Africa [7, 8]. We found that 20%
of HIV-associated maternal deaths were due to complications
of TB.
Maternal illness has an important influence on infant health,
but few studies have examined the impact of maternal TB on
infant outcomes in the HIV era [31]. Although our sample was
small, our study highlights that infants with mothers with incident TB have higher all-cause mortality. Although some of
the infant deaths are likely to be attributable to TB, it is biologically plausible that infants with mothers with poor nutritional status or illness due to active TB would have an increased
risk of death [11].
Recent data reveal that providing IPT to HIV-infected children is beneficial for preventing mortality in areas where the
prevalence of TB is high [32]. Prophylaxis is generally safe, and
isoniazid for 6 months or isoniazid plus rifampin for 3 months
are typically administered to TB-exposed infants in India. In
our study, 72% of HIV-exposed infants who met criteria for
prophylaxis received it. The rest of the infants did not have
therapy initiated, because they were lost to follow-up, their
parents refused treatment, or the infants died before therapy
could be initiated. Furthermore, 23% of the infants did not
complete therapy, suggesting that this strategy of preventing
infant TB and mortality is suboptimal. Thus, consideration of
the role of IPT for mothers as a strategy for reducing TB incidence in mothers, as well as morbidity and mortality in infants, may be useful. Currently, many guidelines recommend
screening for TB during pregnancy using TSTs.
In India, the TST is not used as a routine screening test, and
its use is limited to the diagnosis of TB in children and to
epidemiological surveys. A 15-year follow-up of 280,000 HIVuninfected individuals revealed that TST response is associated
with development of active TB [33]. We also found that a

25

20
10

Women with CD4 cell counts !200 cells/mm3


9

15

24
9

31.59

89.40

449.80

480.0
92.40

29.42

6.71

3.34

5.00
9.74

7.06

23

2.66

0.38 (0.250.57) [37]


0.24 (0.110.53) [25]

91

1.92

0.64 (0.510.81) [37]

56
16

3.20
3.80

No. of
Event rate
patients necessary
associated with
to treat to
administration of
avert 1 case of
IPT
incident TB
0.64 (0.510.81) [37]
0.38 (0.250.57) [37]

Reduction
in risk
associated with
administration of
IPT (95% CI) [ref]

We assumed that, for women with CD4 cell counts !200 cells/mm3, a combination of HAART and IPT would be initiated, because Golub et al. [25] found a more improved risk reduction in TB incidence with the
combination therapy than with HAART alone.

NOTE. Risk reduction estimates were taken from the references listed. The Cochrane meta-analysis [37] did not find any statistically significant difference for IPT by CD4 cell count and AIDS strata; thus, we
assumed the same risk reduction estimates for a CD4 cell count !200 cells/mm3 and for a CD4 cell count 200 cells/mm3. IPT, isoniazid preventive therapy; IR, incidence rate; PY, person-years; ref, reference; TST,
tuberculin skin test.

38

63

90

Women with CD4 cell counts 200 cells/mm3


Women with CD4 cell counts 200 cells/mm3
and with positive TST results

100
38

100
21

All women
Women with positive TST results

Targeted patient group

No. of women
Percentage of Percentage of with incident TB Duration of
antenatal
women
who did not
follow-up,
IR,
women
with incident TB
receive IPT
PY
cases/100 PY

Table 3. Expected reduction in postpartum incident tuberculosis (TB) cases with isoniazid preventive therapy (IPT) initiated during the antepartum period, using different criteria
for targeted intervention in an Indian antenatal clinic population.

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

Acknowledgments
We thank Shuchi Anand, Joline Choi, and all of the Maternal Infant
Transmission Study participants and clinical staff.
Financial support. The US National Institutes of Health (NIH; R01
AI45462) and the Fogarty International Center/USNIH (2 D 43 TW00001019-AITRP). This study was undertaken in collaboration with Byramjee
Jeejeebhoy Medical College in Pune, India.
Potential conflicts of interest. All authors: no conflicts.

6.
7.

8.

9.

10.

11.

12.

13.

14.

15.
16.
17.

18.
19.

20.

21.

22.

23.

References
1. World Health Organization. HIV/TB clinical manual. 2nd ed. Geneva,
Switzerland: World Health Organization, 2004.
2. Kali PB, Gray GE, Violari A, Chaisson RE, McIntyre JA, Martinson
NA. Combining PMTCT with active case finding for tuberculosis. J
Acquir Immune Defic Syndr 2006; 42:37981.
3. Nachega J, Coetzee J, Adendorff T, et al. Tuberculosis active casefinding in a mother-to-child HIV transmission prevention programme
in Soweto, South Africa. AIDS 2003; 17:1398400.
4. Mofenson LM, Rodriguez EM, Hershow R, et al. Mycobacterium tuberculosis infection in pregnant and nonpregnant women infected with
HIV in the Women and Infants Transmission Study. Arch Intern Med
1995; 155:106672.
5. Pillay T, Khan M, Moodley J, Adhikari M, Coovadia H. Perinatal

248 CID 2007:45 (15 July) HIV/AIDS

24.

25.

26.
27.

tuberculosis and HIV-1: considerations for resource-limited settings.


Lancet Infect Dis 2004; 4:15565.
Juneja Y, Goel U, Sood M. Changing trends in maternal mortality over
a decade. Int J Gynaecol Obstet 1994; 46:2659.
Ahmed Y, Mwaba P, Chintu C, Grange JM, Ustianowski A, Zumla A.
A study of maternal mortality at the University Teaching Hospital,
Lusaka, Zambia: the emergence of tuberculosis as a major non-obstetric
cause of maternal death. Int J Tuberc Lung Dis 1999; 3:67580.
Khan M, Pillay T, Moodley J, Connolly C. Maternal mortality associated
with tuberculosis-HIV coinfection in Durban, South Africa. Ann N Y
Acad Sci 2000; 918:3679.
Pillay T, Khan M, Moodley J, et al. The increasing burden of tuberculosis in pregnant women, newborns and infants under 6 months of
age in Durban, KwaZulu-Natal. S Afr Med J 2001; 91:9837.
Marais BJ, Gie RP, Schaaf HS, et al. The natural history of childhood
intra-thoracic tuberculosis: a critical review of literature from the prechemotherapy era. Int J Tuberc Lung Dis 2004; 8:392402.
Marais BJ, Gie RP, Schaaf HS, et al. The clinical epidemiology of childhood pulmonary tuberculosis: a critical review of literature from the
pre-chemotherapy era. Int J Tuberc Lung Dis 2004; 8:27885.
TB Control of India. TB India 2006 RNTCP status reportDOTS
for all: all for DOTS. Available at: http://www.tbcindia.org/
documents.asp#. Accessed 28 June 2006.
World Health Organization (WHO). Global tuberculosis controlsurveillance, planning, financing: WHO report 2006. Geneva, Switzerland:
WHO, 2006.
United Nations AIDS Program. 2006 Report on the global AIDS epidemic. 2006. Available at: http://www.unaids.org/en/HIV_data/
2006GlobalReport/default.asp. Accessed 12 July 2006.
Padmapriyadarsini C, Swaminathan S. Preventive therapy for tuberculosis in HIV-infected individuals. Indian J Med Res 2005; 121:41523.
Huebner RE, Schein MF, Bass JB Jr. The tuberculin skin test. Clin
Infect Dis 1993; 17:96875.
Targeted tuberculin testing and treatment of latent tuberculosis infection. American Thoracic Society. MMWR Recomm Rep 2000; 49(RR6):151.
Enarson PM, Enarson DA, Gie R. Management of tuberculosis in children in low-income countries. Int J Tuberc Lung Dis 2005; 9:1299304.
Burnham K, Anderson D. Model selection and multimodel inference:
a practical informationtheoretic approach. New York: Springer-Verlag,
2002.
Selwyn PA, Hartel D, Lewis VA, et al. A prospective study of the risk
of tuberculosis among intravenous drug users with human immunodeficiency virus infection. N Engl J Med 1989; 320:54550.
Whalen CC, Nsubuga P, Okwera A, et al. Impact of pulmonary tuberculosis on survival of HIV-infected adults: a prospective epidemiologic study in Uganda. AIDS 2000; 14:121928.
Swaminathan S, Ramachandran R, Baskaran G, et al. Risk of development of tuberculosis in HIV-infected patients. Int J Tuberc Lung
Dis 2000; 4:83944.
Antonucci G, Girardi E, Raviglione MC, Ippolito G. Risk factors for
tuberculosis in HIV-infected persons: a prospective cohort study. The
Gruppo Italiano di Studio Tubercolosi e AIDS (GISTA). JAMA
1995; 274:1438.
Lawn SD, Badri M, Wood R. Tuberculosis among HIV-infected patients
receiving HAART: long term incidence and risk factors in a South
African cohort. AIDS 2005; 19:210916.
Golub JE, Saraceni V, Cavalcante S, et al. The impact of antiretroviral
therapy and isoniazid preventative therapy on tuberculosis incidence
in HIV-infected patients in Rio de Janeiro, Brazil. AIDS (in press).
Day JH, Grant AD, Fielding KL, et al. Does tuberculosis increase HIV
load? J Infect Dis 2004; 190:167784.
Wolday D, Hailu B, Girma M, Hailu E, Sanders E, Fontanet AL. Low
CD4+ T-cell count and high HIV viral load precede the development
of tuberculosis disease in a cohort of HIV-positive ethiopians. Int J
Tuberc Lung Dis 2003; 7:1106.

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

period to avert 1 case of TB (table 3) [24, 25, 37]. Using a


more targeted approach that involves administering IPT to
women with positive TST results and CD4 cell counts 200
cells/mm3 and administering HAART plus IPT to women with
CD4 cell counts !200 cells/mm3, 523 women would need to
be given IPT to avert 1 case of TB. Thus, 30% of antepartum
women, who account for 63% of patients with incident TB,
would be targeted. With any strategy, stringent efforts to rule
out active TB would be necessary (e.g., using sensitive symptom
screening, including cough, fever, and weight loss; chest radiography; and, ideally, collection of sputum samples for acidfast bacilli testing).
Even in the HAART era, TB will remain an important cause
of morbidity and mortality [24, 38]. In the absence of preventive TB vaccines, the United Nations AIDS Program and World
Health Organization, as well as the Centers for Disease Control
and Prevention guidelines, currently recommend IPT for HIVinfected adults with latent TB [17, 39]. Targeted IPT has been
shown to reduce the risk of active TB by up to 60% both in
regions where TB is endemic and in regions where it is not
endemic [37, 39, 40]. Despite Indias large HIV and TB burden,
IPT is not currently recommended. Indian HIV-infected
women have a high risk of TB during the postpartum period,
regardless of CD4 cell count, which is associated with the need
for infant TB prophylaxis and high rates of postpartum maternal and infant death. Therefore, although widespread use of
IPT in India may be problematic, targeted administration of
IPT to antepartum HIV-infected women is likely to have great
benefit for preventing postpartum maternal TB and to reduce
the associated high incidence of morbidity and mortality caused
by mother-to-child transmission of infection.

28. Yancey MK. Host defenses and bacterial resistance. Obstet Gynecol
Clin North Am 1992; 19:41334.
29. Saraiya UB, Bhalerao SA. Pregnancy and tuberculosis. In: Krishna U,
ed. Pregnancy at risk: current concepts. 4th ed. New Delhi, India: The
Federation of Obstetric and Gynecological Societies of India, Jaypee
Brothers Medical Publishers, 2001: 1327.
30. Robinson CA, Rose NC. Tuberculosis: current implications and management in obstetrics. Obstet Gynecol Surv 1996; 51:11524.
31. Newell ML, Coovadia H, Cortina-Borja M, Rollins N, Gaillard P, Dabis
F. Mortality of infected and uninfected infants born to HIV-infected
mothers in Africa: a pooled analysis. Lancet 2004; 364:123643.
32. Zar HJ, Cotton MF, Strauss S, et al. Effect of isoniazid prophylaxis on
mortality and incidence of tuberculosis in children with HIV: randomised controlled trial. BMJ 2007; 334:136.
33. Radhakrishna S, Frieden TR, Subramani R. Association of initial tuberculin sensitivity, age and sex with the incidence of tuberculosis in
south India: a 15-year follow-up. Int J Tuberc Lung Dis 2003; 7:
108391.
34. Pai M, Gokhale K, Joshi R, et al. Mycobacterium tuberculosis infection
in health care workers in rural India: comparison of a whole-blood

35.

36.

37.
38.

39.

40.

interferon gamma assay with tuberculin skin testing. JAMA 2005; 293:
274655.
Franks AL, Binkin NJ, Snider DE Jr, Rokaw WM, Becker S. Isoniazid
hepatitis among pregnant and postpartum Hispanic patients. Public
Health Rep 1989; 104:1515.
Boggess KA, Myers ER, Hamilton CD. Antepartum or postpartum
isoniazid treatment of latent tuberculosis infection. Obstet Gynecol
2000; 96:75762.
Woldehanna S, Volmink J. Treatment of latent tuberculosis infection
in HIV infected persons. Cochrane Database Syst Rev 2004:CD000171.
Lawn SD, Myer L, Bekker LG, Wood R. Burden of tuberculosis in an
antiretroviral treatment programme in sub-Saharan Africa: impact on
treatment outcomes and implications for tuberculosis control. AIDS
2006; 20:160512.
World Health Organization Global Tuberculosis Programme. Policy
statement on preventive therapy against tuberculosis in people living
with HIV. Geneva, Switzerland: World Health Organization, 1998.
Grant AD, Charalambous S, Fielding KL, et al. Effect of routine isoniazid preventive therapy on tuberculosis incidence among HIV-infected men in South Africa: a novel randomized incremental recruitment study. JAMA 2005; 293:271925.

Downloaded from http://cid.oxfordjournals.org/ by guest on June 25, 2016

HIV/AIDS CID 2007:45 (15 July) 249