Lancet 2005, 365, 749-751

Correspondence (authors' reply at bottom)

Co-trimoxazole prophylaxis in African children with HIV-1

Although we appreciate the potential of C Chinto and colleagues' study
(Nov 20, p 1865)1 to promote wider use of prophylaxis in African
HIV-1-infected children, we are concerned about the ethics of a placebo
group and disagree with the authors' interpretation and explanation of
their findings.

We are disturbed that severely immunocompromised children (CD4% <15)
were randomised to placebo in breach of widely accepted 1995 Centers
for Disease Control (CDC) guidelines2 and subsequent UNAIDS
recommendations3 on co-trimoxazole prophylaxis for HIV-1-infected
children in developing countries. These recommendations advocate
prophylaxis for any child born to an HIV-1-infected woman; any
HIV-1-infected infant; and children older than 15 months who have had
Pneumocystis jiroveci pneumonia (PCP), have symptomatic HIV disease,
have an AIDS-defining illness, or have a CD4% of less than 15. It is
therefore unacceptable that severely immunocompromised children, who
accounted for 69% of children overall and 71% of those on placebo, were
denied prophylaxis and offered placebo instead.

Co-trimoxazole is cheap and widely available in Africa. Although
co-trimoxazole prophylaxis may not be the current standard of care in
Zambia, we believe that the good clinical practice principle of
providing the minimum standard of recommended care to all study
participants was breached. Such practice would not have occurred in any
developed country. The enormity of this lapse is compounded if one
considers the findings of an earlier post-mortem study from the same
group, in which PCP accounted for 29% of deaths in HIV-1-infected
children with respiratory illnesses, and was an important cause of
mortality in those older than 12 months.4

The data do not adequately support the authors' conclusion that
co-trimoxazole prophylaxis would benefit all HIV-1-infected children,
irrespective of their immunological status. Although the test of
heterogeneity did not find any difference in effect on mortality for
different CD4 categories, the subanalysis merely reaffirms the benefit
of co-trimoxazole prophylaxis in reducing mortality in severely
immunocompromised children (hazard ratio 0·51, 95% CI 0·34-0·77;
p=0·0005]), while failing to show any significant effect in children
whose baseline CD4% was greater than 15 (0·62, 0·26-1·47; p=0·44).
Further, what the CD4% of children with baseline levels greater than 15
were at the time of their death is not clear. Many of these children's
imunosuppression could have worsened during the study period, such that
CD4% were less than 15 at the time of death.

We note the authors' difficulty in adequately explaining the effect of
co-trimoxazole, and their ascribing its benefit to a possible reduction
in mortality from bacterial pneumonia rather than PCP. The authors
admit that nasopharyngeal aspirates are an insensitive method by which
to identify P jiroveci.5 Further, in only 73 (35%) of 206 hospital
admissions were investigations for PCP done, and no attempt was made to
determine whether PCP was the cause of death. Consequently, the study
cannot satisfactorily exclude the possibility that co-trimoxazole was
primarily effective against PCP rather than bacterial pneumonia.

Although infants and severely immunosuppressed children clearly benefit
from co-trimoxazole prophylaxis, we remain to be convinced that
children older than 1 year, with mild or moderate immunosuppression,
can derive substantial benefit from the intervention.

We declare that we have no conflict of interest.

*Shabir A Madhi, Heather J Zar, Haroon Saloojee, Glenda E Gray

madhis@hivsa.com

*PO Bertsham, Chris Hani-Baragwanath Hospital, Old Nurses Home, 1st
Floor West Wing, Bertsham, Gauteng 2013, South Africa (SAM); School of
Child and Adolescent Health, Red Cross Children's Hospital, University
of Cape Town, South Africa (HJZ); Department of Community Paediatrics
(HS) and Perinatal HIV Research Unit (GEG), University of the
Witwatersrand, Johannesburg, South Africa

1 Chintu C, Bhat GJ, Walker AS, et al. Co-trimoxazole as prophylaxis
against opportunistic infections in HIV-infected Zambian children
(CHAP): a double-blind randomised placebo-controlled trial.  Lancet
2004; 364: 1865-71. [Text]

2 National Pediatric and Family HIV Resource Center and National Center
for Infectious Diseases, Centers for Disease Control and Prevention.
1995 revised guidelines for prophylaxis against Pneumocystis carinii
pneumonia for children infected with or perinatally exposed to human
immunodeficiency virus.  MMWR Recomm Rep 1995; 44 :(RR-4) 1-11.
[PubMed]

3 World Health Organization, UNAIDS. Provisional WHO/UNAIDS secretariat
recommendations on the use of cotrimoxazole prophylaxis in adults and
children living with HIV/AIDS in Africa.
http://www.unaids.org/html/pub/publications/irc-pub04/
recommendation_en_pdf.pdf (accessed Nov 29, 2004).

4 Chintu C, Mudenda V, Lucas S, et al. Lung diseases at necropsy in
African children dying from respiratory illnesses: a descriptive
necropsy study.  Lancet 2002; 360: 985-90. [Text]

5 Ruffini DD, Madhi SA. The high burden of Pneumocystis carinii
pneumonia in African HIV-1-infected children hospitalized for severe
pneumonia.  AIDS 2002; 16: 105-12. [PubMed]
------------

Authors' reply

We are pleased that Shabir Madhi and colleagues recognise the
importance of the results of the CHAP trial in promoting evidence-based
use of co-trimoxazole prophylaxis in HIV-1-infected children in Africa.
However, we believe that their concerns about the ethics of randomising
children with CD4% of less than 15, based on 1995 US guidelines, are
misguided.

Co-trimoxazole prophylaxis for children older than 1 year is
recommended for prevention of PCP in 1995 US guidelines, on the basis
of evidence extrapolated from US trials in HIV-infected adults.
Although it is established that PCP is common in African infants,1,2
data from several studies (including ours) suggest that this pathogen
occurs relatively infrequently in older children in resource-limited
settings and that bacterial pneumonias and tuberculosis predominate.1,2
The Zambian postmortem study quoted by Madhi and colleagues as having a

PCP prevalence of 29% included mainly infants (median age 9 months);
only 7% of the total pathology among children older than 1 year was due
to PCP, and this is a higher proportion than that of other post-mortem
studies from the region.1

A cohort study of HIV-infected adults and children older than 2 years
with low CD4 counts in Malawi showed PCP only rarely (0·3 per 100
person-years) despite use of sensitive induced sputum techniques.2
Madhi's own paediatric study from South Africa (median age 4 months) is
notable for reporting a higher prevalence of PCP in children after
infancy in Africa,3 but the specificity of PCP diagnosis could have
been reduced by the methods of diagnosis and patient selection used.4
Although we agree that the sensitivity of nasopharyngeal aspirate is
lower than that of induced sputum, our finding of only one positive
among 73, and none by PCR strongly suggests that PCP is uncommon in
older children.

In west Africa in 1999, trials in HIV-infected adults reported the
efficacy of co-trimoxazole prophylaxis in reducing mortality and
hospital admissions caused predominantly by bacterial pneumonia, not by
PCP. The major issue, raised by the authors of those papers and since
hotly debated, was whether co-trimoxazole would prevent bacterial
infections in countries such as Zambia where the prevalence of
resistance is high. The CHAP trial is the first randomised evidence to
show efficacy in a high resistance setting, as well as being the first
trial in children worldwide. CHAP and results from a recently reported
cohort study in HIV-infected Ugandan adults5 have already prompted a
revision of national guidelines in several African countries.
WHO/UNAIDS guidelines for children (which previously called for more
research in older children) have also been revised.

Madhi and colleagues are wrong in interpreting lack of evidence for a
treatment effect within a subgroup of children (with baseline CD4% >15)
as evidence of "no treatment effect" in such a subgroup. By definition,
a subgroup will invariably have inadequate power to detect treatment
differences in that subgroup alone because trials are powered on the
total randomised population. The most appropriate test for subgroup
effects in randomised controlled trials is a test for heterogeneity, as

in our analysis. Comparison of the observed hazard ratios of 0·62 and
0·51, for a reduction in mortality in children with CD4% greater and
less than 15, respectively, suggests that there is little if any
evidence, albeit with large confidence intervals, of a difference in
efficacy of co-trimoxazole between the two groups. Such a result would
not be unexpected if the main cause of mortality is bacterial
infections rather than PCP. This, coupled with lack of toxic effects,
suggests that provision of co-trimoxazole prophylaxis to all
symptomatic children with HIV-1 would be a reasonable policy, and
remove the need to do expensive CD4 tests.

We declare that we have no conflict of interest.

*D M Gibb, C Chintu, G J Bhat, A S Walker, A J Nunn, on behalf of the
CHAP trial team

di.gibb@ctu.mrc.ac.uk

*Medical Research Council Clinical Trials Unit, London, UK (DMG, ASW,
AJN); University Teaching Hospital, Lusaka, Zambia (CC, GJB)

1 Graham SM. Prophylaxis against Pneumocystis carinii pneumonia for
HIV-exposed infants in Africa.  Lancet 2002; 360: 1966-68. [Text]

2 van Oosterhout JJG, Laufer MK, Graham SM, et al. Low incidence of
Pneumocystis pneumonia in HIV infected persons in Blantyre, Malawi. XV
International AIDS Conference, Bangkok 2004 [MoPeB3198].

3 Madhi SA, Cutland C, Ismail K, et al. Ineffectiveness of
trimethoprim-sulfamethoxazole prophylaxis and the importance of
bacterial and viral coinfections in African children with Pneumocystis
carinii pneumonia.  Clin Infect Dis 2002; 35: 1120-26. [PubMed]

4 Graham SM. HIV and respiratory infections in children.  Curr Opin
Pulm Med 2003; 9: 215-20. [PubMed]

5 Mermin J, Lule J, Ekwaru JP, et al. Effect of co-trimoxazole
prophylaxis on morbidity, mortality, CD4 cell count, and viral load in
HIV infection in rural Uganda.  Lancet 2004; 364: 1428-34. [Text]
Top