American Journal of Epidemiology
Copyright C 1997 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 145, No. 4
Printed in U.S.A
ORIGINAL CONTRIBUTIONS
Progression of Human Immunodeficiency Virus Infection in Patients with
Tuberculosis Disease
A Cohort Study in Bordeaux, France, 1988-1994
Valeriane Leroy,1 L Rachid Salmi,1 Michel Dupon,2 Angelique Sentilhes,2 Jeannette Texier-Maugein,2
Laurence Dequae,1 Frangois Dabis,1 and Roger Salamon1 for the Groupe d'Epidemiologie Clinique du Sida
en Aquitaine (GECSA)3
To assess the role of Mycobacterium tuberculosis disease in human immunodeficiency virus (HIV) infection,
the authors compared survival of tuberculosis patients and controls matched on year of HIV diagnosis and
CD4+ lymphocyte count. Patients were selected in the Aquitaine Cohort, which follows, since 1985, all
patients infected with HIV, aged more than 13 years, in five hospitals. Time of inclusion of controls was the date
of diagnosis of tuberculosis for the corresponding tuberculosis patient. Patients who had received primary
prophylaxis against mycobacteria other than tuberculosis were excluded. As of June 30, 1994, 104 tuberculosis patients and 620 controls were selected; they were similar, except for history of intravenous drug use
(tuberculosis patients, 5 1 % , vs. controls, 31%) and AIDS-defining opportunistic infection (40 vs. 29%).
Survival was shorter in tuberculosis patients than in controls (risk ratio 1.5, 95% confidence interval 1.2-2.1)
even after controlling for differences at entry. The risk of AIDS-defining opportunistic infection or a decrease
to fewer than 50 CD4+ cells/mm3 was slightly but not statistically greater in tuberculosis patients than in
controls. Tuberculosis disease affected survival but not occurrence of subsequent opportunistic infections or
rate of CD4+ count decline. Tuberculosis may be a marker of advanced HIV and may accelerate its course of
infection. Am J Epidemiol 1997;145:293-300.
HIV infections; natural history; tuberculosis
The incidence of reported Mycobacterium tuberculosis disease has been increasing worldwide since the
beginning of the 1980s (1, 2). Studies have suggested
that this trend was closely related to the human immunodeficiency virus (HIV) pandemic (3, 4). By mid1994, the World Health Organization estimated that
5.6 million individuals were infected with both HIV
and M. tuberculosis worldwide (2). The effect of HIV
infection on the reactivation of tuberculosis is now
well recognized (3-5). Consequently, since January
1993, the Centers for Disease Control has expanded
the AIDS case definition to pulmonary tuberculosis
disease (6). Tuberculosis is now a major cause of
death among HIV-infected patients (7-10).
Nevertheless, data are sparse on the effect of tuberculosis disease on the natural history of HIV infection
(11, 12). Several authors have suggested that tuberculosis disease may be an independent marker of ad-
Received for publication December 27, 1995, and accepted for
publication October 2, 1996.
Abbreviations: AIDS, acquired immunodeficiency syndrome; Cl,
confidence interval; GECSA, Groupe d'Epidemiologie Clinique du
Sida en Aquitaine; HIV, human immunodeficiency virus; RR, risk
ratio.
1
Universite de Bordeaux II, INSERM 330, France.
2
Centre Hospitalier Universitaire de Bordeaux, France
3
Groupe d'Epidemiologie Clinique du Sida en Aquitaine of the
Centre Hospitalier Universitaire de Bordeaux: Organizational and
epidemiologic coordination—Profs. F. Dabis and R. Salamon, Dr. G.
Chene; Medical coordination: Drs. J. Constans, M. Dupon, D.
Lacoste, and P. Moriat; Profs. J-F. Moreau, J-L Pellegrin, and J-M.
Ragnaud; Participating hospital departments: Bordeaux Hospital—
Prof. J Aubertin (Prof. J-M. Ragnaud, Drs. M. Buisson and S. Sire),
Prof. J. Beyiot (Drs. N. Bernard, D. Lacoste, and P. Moriat), Prof. C.
Beyiot (Prof. M-S. Doutre), Prof. C. Conri (Dr. J. Constans), Profs. P.
Couzigou and M. Geniaux (Mrs. A. Simon), Prof. J-Y. Lacut (Drs. M.
Dupon and M-C. Paty), Prof. B. Leng (Prof. J-L Pellegrin, Drs. C.
Desforges-Lasseur and P. Rispaf), Prof. M. Le Bras (Drs. E. Monlun
and J-P. Pivetaud), Prof. F. Moreau (Dr. C. Mestre), Prof. J. Paccalin
(Dr. H. Dabadie), Profs. C. Series and A. Taytard (Dr. J-M.
Vemejoux); Dax Hospital—Dr. P. Loste (Dr. I. Blanchard); Bayonne
Hospital—Ors. M. Ferrand and F. Bonnal (Drs. Y. Blanchard and S.
Farbos); Liboume Hospital—Drs. N. Carde and J. Ceccaldi (Dr. X.
Jacquelin); Villeneuve-sur-Lot Hospital—Dr. G. Brossard; Data management and analysis: Mrs. J. Caie, M. Decoin, L Dequae,
M. Errecart-Barbotin, D. Touchard, D. Belougne, and Drs. D.
Commenges, J-B. Hubert, C. Marimoutou, and V. Leroy.
Reprint requests to Dr. Valeriane Leroy, Universite de Bordeaux
II, Unite INSERM 330,146 rue L6o Saignat, 33076 Bordeaux Cedex,
France.
293
294
Leroy et al.
vanced immunosuppression in HIV-infected patients,
but that it could also be a cofactor in accelerating the
course of HIV infection (8, 13-15). The containment
of tubercle bacilli within a granuloma requires the
generation of cytokines, including tumor necrosis factor a. Tumor necrosis factor a may enhance HIV
replication, which in turn may increase the viral burden, leading to a faster evolution to clinical acquired
immunodeficiency syndrome (AIDS) and death (1618). However, only one epidemiologic study (19),
controlling for the degree of immunodepression, suggested a faster evolution of HFV infection in tuberculosis patients than in patients without tuberculosis
disease.
Several issues that are important from a clinical
viewpoint for improving care of HIV-infected patients
remain unanswered (12, 20). Whereas systematic prophylaxis against tuberculosis is recommended for all
purified protein derivative-positive HIV-infected persons (21), this recommendation is not routinely applied in other countries. In countries of high HIV and
tuberculosis prevalence, indications of systematic preventive therapy against tuberculosis disease should be
based on cost effectiveness evaluations, taking into
account both the risk of developing resistance strains
after preventive therapy and the number of active
tuberculosis actually avoided in the long term (11, 12,
22, 23). If tuberculosis accelerates HIV progression,
then an argument exists for a more active prevention
of tuberculosis.
Answering this kind of question requires a carefully
designed observational study, based on a large and
representative cohort with a sufficient follow-up. The
Aquitaine Cohort, followed since 1985 by the Groupe
d'Epide"miologie Clinique du Sida en Aquitaine
(GECSA), is based on a hospital-based surveillance
system of HTV infection, set up in the Bordeaux University Hospital and four public hospitals of Aquitaine
(24, 25). The purpose of our study was to assess the
effect of tuberculosis disease on the course of HTV
infection in the Aquitaine Cohort.
MATERIALS AND METHODS
Aquitaine is a region in southwestern France, with
2,750,000 inhabitants (1990 census); 65 percent of the
population lives in urban areas (26). As of December
31, 1994, the rate of notified AIDS was 57.8 per
100,000 inhabitants (27). Subjects in our study were
part of the Aquitaine Cohort in which inclusion was
retrospective before October 1987 and prospective
after that time. Inclusion criteria were 1) HIV-1 infection confirmed by Western blot analysis, regardless of
the clinical stage; 2) age more than 13 years; and 3)
informed consent. According to mandatory reporting
of AIDS cases to the French Ministry of Health, more
than 75 percent of AIDS cases in Aquitaine are reported by physicians participating in the GECSA (25).
Participating clinicians have been enrolling and following their patients using standardized computerized
epidemiologic, medical, and biologic records.
Follow-up is repeated at least every 6 months, and a
repeated search for patients lost to follow-up is carried
out on a yearly basis.
For the present study, we performed an historical
analysis comparing patients who had received diagnoses of tuberculosis disease with controls unexposed
to tuberculosis disease. To identify cases of pulmonary
and extrapulmonary tuberculosis disease among all
patients in the cohort, we used records completed by
clinicians during follow-up. Tuberculosis disease was
defined as a proven or definite case of tuberculosis
disease according to Centers for Disease Control criteria (28), i.e., presence of symptoms consistent with
active tuberculosis, associated with a positively confirmed M. tuberculosis culture. Probable or possible
cases of tuberculosis disease were excluded from the
study. Organ involvement, biologic features, and
histologic results, when available, were documented
and investigated. Moreover, to make an exhaustive
study of proven tuberculosis disease cases in HTVseropositive patients, we reviewed all records from the
Department of Microbiology of the Bordeaux University Hospital during the study period. This laboratory
receives mycobacterial strains for identification from
other public and private microbiology laboratories in
the Aquitaine region. Laboratory files of patients in
M. tuberculosis isolation were cross-checked with
computerized files of patients included in the cohort.
All of our tuberculosis patients were treated as recommended in France, beginning with 2 months of a daily
combination of rifampicin, isoniazid, ethambutol, and
pyrazinamide, followed by at least 4 months of a daily
combination of isoniazid and rifampicin.
Although tuberculosis appeared on the death certificates, we checked the causes of death reported on
death certificates against the medical records extracted
from the GECSA database corresponding to the last
clinical follow-up of the patient. Autopsy data were
unavailable. Death with tuberculosis was evoked for
all patients with a proven or definite case of tuberculosis disease for whom the death certificate and the last
medical record reported they had tuberculosis. Classification of deaths related to tuberculosis was made as
follows: A definite tuberculosis death was defined for
patients without a history of other ATDS-related disease (inaugural AIDS event) and without another
known cause of death; a probable tuberculosis death
was defined for patients with a history of an ADDSAm J Epidemiol
Vol. 145, No. 4, 1997
Tuberculosis Disease and Human Immunodeficiency Virus Infection
related disease but without another known current
AIDS-related disease at time of death; a possible tuberculosis death was defined for patients with other
known current AIDS-related disease at time of death;
otherwise, tuberculosis death was excluded.
For each tuberculosis case, we selected in the cohort
database all possible controls who had 1) the same
year of HTV diagnosis (±6 months), 2) the same time
interval between HIV diagnosis and inclusion in our
study (±3 months), and 3) the same level of CD4+
lymphocyte count (±25/mm 3 ) at inclusion in our
study. CD4+ lymphocyte counts were taken into account at the time of tuberculosis. Time of inclusion of
a tuberculosis patient was the date of diagnosis of
tuberculosis disease for the tuberculosis patient and
the time of CD4+ lymphocytes count for his/her controls. Exclusion criteria were either no follow-up after
the inclusion date or primary prophylaxis against
M. tuberculosis or mycobacteria other than tuberculosis, and these criteria were intended to exclude patients
with a greater risk of resistance to tuberculosis treatment.
We extracted from the GECSA database sociodemographic characteristics, dates of occurrence of opportunistic infections defined according to the 1986
Centers for Disease Control classification system for
HTV infection (29), CD4+ lymphocyte counts, treatments received, and dates and causes of death reported
on the death certificates. With death as the primary
outcome of this study, the incidence of a new ATDSdefming opportunistic infection and of a decrease to
fewer than 50 CD4+ cells/mm3 were secondary outcomes for the subgroup of patients who were free of
these events at inclusion in our study.
We compared baseline characteristics (sociodemographic and risk behavior, history of opportunistic
infections, treatment of HIV infection, and CD4+
lymphocyte count) in tuberculosis patients and controls using Chi-square and Student's t tests. Incidence
density rates of outcomes were defined as the number
of new events per 100 person-years of follow-up (30).
Follow-up for a patient was defined as the period from
inclusion to June 30, 1994, or to the date of the last
visit before this study endpoint or the patient's death.
The potential follow-up for assessing event-free probabilities was the maximum theoretical participation
time of all patients until June 30, 1994, minus the
period between the occurrence of an outcome (death,
opportunistic infection, decrease to fewer than 50
CD4+ cells/mm3) and June 30, 1994. The rate of
follow-up was calculated by dividing the actual participation time of all individuals by the potential
follow-up period (31). Event-free cumulative probabilities for survival, opportunistic infections, and a
Am J Epidemiol
Vol. 145, No. 4, 1997
295
decrease to fewer than 50 CD4+ cells/mm3 over time
were assessed using the Kaplan-Meier method and
were reported with their 95 percent confidence intervals (32). Comparisons of survival curves were assessed with the generalized log rank test. Ninety-five
percent confidence intervals around the risk ratio were
calculated using Rothman's approximate method (30).
To take into account the effect on mortality of differences between baseline characteristics not controlled
for by the matching method and the effect of known
prognostic factors of HTV disease progression, a proportional hazard model regression was performed
(33). Variables considered in-the model were tuberculosis disease, age, intravenous drug use, history of
opportunistic infection, use of preventive and antiretroviral treatments, and variables of the matching procedure (year of HTV diagnosis, interval between HTV
diagnosis and inclusion in the study, and baseline
CD4+ lymphocyte count).
RESULTS
As of June 30, 1994, 3,697 patients have been
enrolled in the Aquitaine Cohort. Tuberculosis disease
was reported in 193 patients (5.2 percent), of whom
104 (53.8 percent) were definite cases meeting inclusion criteria. Tuberculosis was pulmonary in 65 of
these cases (62.5 percent), extrapulmonary in 19 (18.3
percent), and disseminated in 20 (19.2 percent). A
total of 620 patients met the criteria for inclusion as
controls. Tuberculosis patients and controls had a similar median year of HTV diagnosis (1988) and mean
interval between HIV diagnosis and inclusion in our
study (32.4 months; standard deviation 27.6 months).
Mean age was 37.0 years (standard deviation 10.1
years) in tuberculosis patients and 37.5 years (standard
deviation 11.1 years) in controls (p = 0.26). Tuberculosis patients were more likely than controls to have
a history of intravenous drug use or opportunistic
infection (table 1).
Follow-up was comparable in the two groups (table
2). Of the 307 patients (42.4 percent) who died, 57
were tuberculosis patients. Sixty-two tuberculosis patients and 441 controls were at risk of developing an
AIDS-defining opportunistic infection; 70 tuberculosis
patients and 405 controls were at risk of a decreased
count of fewer than 50 CD4+ cells/mm3.
Mortality rates were higher in tuberculosis patients
than in controls (40.7/100 vs. 26.4/100 person-years;
risk ratio (RR) = 1.5, 95 percent confidence interval
(CI) 1.2-2.1). Median survival was shorter in tuberculosis patients than in controls (17.3 vs. 30.9 months;
p = 0.002). The Kaplan-Meier probabilities of survival in tuberculosis patients and controls were 60.1
percent (95 percent CI 49.8-70.3) and 75.9 percent
296
Leroy et al.
TABLE 1. Comparison of baseline characteristics among 104 patients with Mycobacterlum
tuberculosis disease (cases) and 620 controls'at inclusion in the study, AquKaine Cohort, France,
1985-1994
Cases
Controls
Variables
Men
Transmission group
Homosexuals
Intravenous drug users
Heterosexuals
Others or unknown
History of AIDSf-defining opportunistic
infection
Prevention of Pneumocystis carinii
Antiretroviral treatment
CD4+ lymphocyte count (/mm3)
<50
50-199
200-349
£350
No.
%
No.
%
78
75.0
472
76.1
23
53
16
12
22.1
51.0
15.4
11.5
235
194
138
53
37.9
31.3
22.3
8.5
42
39
60
40.4
37.5
57.7
179
272
377
28.8
43.9
60.8
34
42
20
8
32.6
40.4
19.2
215
249
129
27
34.8
40.3
P
value*
0.80
0.0002
7.8
0.01
0.22
0.54
0.52
20.8
4.3
• Chl-square test.
t AIDS, acquired immunodeficiency syndrome.
TABLE 2. Follow-up characteristics of 104 patients with Mycobacterlum tuberculosis disease (cases)
and 620 controls over the study period, Aqultalne Cohort, France, 1985-1994
Cases
Complete until June 30, 1994
Incomplete because of
Death
Loss to follow-up*
Person-years for survival analysist
Person-years for analysis of AIDS^-definlng
opportunistic infections
Person-years for analysis of CD4+ decrease§
Controls
No.
%
No.
%
26
25.0
268
43.2
57
21
140
54.8
20.2
86.4
250
102
947
40.3
16.5
86.9
82
89
80.0
74.2
667
620
86.6
81.0
• Last follow-up before June 30, 1994.
t Percentages are proportions of maximal theoretical follow-up after exclusion of time between death and June
30, 1994.
$ AIDS, acquired immunodeficiency syndrome.
§ Percentages are proportions of maximal theoretical follow-up after exclusion of time between death or event
and June 30, 1994.
(95 percent CI 72.2-79.6) at 12 months, 38.5 percent
(95 percent CI 26.8-50.2) and 42.5 percent (95 percent CI 31.7-47.9) at 36 months, and 17.8 percent (95
percent CI 5.7-29.9) and 30.8 percent (95 percent CI
24.2-37.5) at 60 months after inclusion.
In non-intravenous drug users, mortality rates were
higher in the 51 tuberculosis patients than in the 426
controls (RR = 1.9, 95 percent CI 1.3-2.7; p =
0.0008) (figure 1). For intravenous drug users, mortality was not different for the 53 tuberculosis patients
and 194 controls (RR = 1.3, 95 percent CI: 0.9-2.1;
p = 0.14). In patients without a history of AIDSdefining opportunistic infection at inclusion, mortality
was higher in the 62 tuberculosis patients than in the
441 controls (RR = 1.5, 95 percent CI 1.1-2.3; p =
0.04) (figure 1). In patients with a history of opportu-
nistic infection at inclusion, mortality for the 42 tuberculosis patients was higher, but not significant, than
for the 179 controls (RR = 1.4, 95 percent CI 0.9-2.0;
p = 0.15). Survival was shorter in patients with baseline CD4+ lymphocyte count of fewer than 50 cells/
mm3 (RR = 1.7, 95 percent CI 1.1-2.6;/? = 0.003) or
between 50 and 200 cells/mm3 (RR = 2.2, 95 percent
CI 1.4-3.6; p = 0.02). This difference was not observed in patients with CD4+ lymphocyte counts of
more than 200 cells/mm3 (RR = 1.9, 95 percent CI
0.9-4.0; p = 0.11).
The increased mortality in tuberculosis patients was
observed even when controlling for matching variables, differences in baseline characteristics, and
known prognostic variables (proportional hazard regression: RR = 1.59; p = 0.007). A diagnosis of
Am J Epidemiol
Vol. 145, No. 4, 1997
Tuberculosis Disease and Human Immunodeficiency Virus Infection
Non IV drug users
297
IV drug users
1
ro 0.8 '•-.-. N.
|
0.8
0.6
I•c 0.4
"•"'••-.
^ - — ~ ^ _ _
e 02
a.
0
0.4
^^~N_^
'"":
o
Q.
.
1
.
2
:
L
0.2
.
.
.
3
4
5
.
6
o
2
7
3
4
5
6
7
Patients with Ol at inclusion
Patients free of Ol at inclusion
1
0.8
06
0.4
0.2
2
3
4
0
0
Time (years)
2
3
4
Time (years)
FIGURE 1. Survival probabilities, according to transmission group and history of AIDS-defining opportunistic infection (Ol) in patients
exposed to Mycobacterium tuberculosis disease (
) and controls (
), Aquitaine Cohort, France (1985-1994). IV, intravenous.
tuberculosis disease, a CD4+ lymphocyte count of
fewer than 200 cells/mm3, aging, a history of at least
one AIDS-defining opportunistic infection, and absence of prolonged treatment after mycobacteria other
than tuberculosis were factors strongly associated with
mortality in the multivariate analysis (table 3).
Among the 57 tuberculosis patients who died, the
cause of death on the death certificates could be related to tuberculosis for 19.3 percent (11 patients), to
AIDS-defining opportunistic infections other than tuberculosis for 33.4 percent, to AIDS-defining events
other than opportunistic infections for 36.8 percent,
and to other causes for 10.5 percent. Twenty-four
percent of tuberculosis patients died within the first
month and 64.9 percent within the first year of tuberculosis disease diagnosis. Among the 11 patients for
whom the cause of death could be related to tuberculosis disease, five were classified as definite cases, one
as a probable case, and five as possible cases according to our case definition. The median survival for
these 11 patients was 21 days.
Incidence rates of opportunistic infections were not
significantly higher in tuberculosis patients than in
controls (28.1/100 vs. 22.2/100 person-years, RR =
1.3, 95 percent CI 0.8-2.0; p = 0.28). Incidence rates
of a decrease to fewer than 50 CD4+ cells/mm3 were
Am J Epidemiol
Vol. 145, No. 4, 1997
33.7/100 person-years in tuberculosis patients and
23.4/100 person-years in controls (RR = 1.4, 95 percent CI 1.0-2.1; p = 0.09).
DISCUSSION
This historical cohort study comparing 104 HTVinfected patients exposed to tuberculosis disease
matched with 620 HIV-infected patients unexposed to
tuberculosis disease allowed us to assess the effect of
tuberculosis disease on the natural history of HIV
infection. The main finding was a significant overall
reduced survival in patients exposed to tuberculosis
disease compared with controls, even when adjusting
for variables known to be associated with survival.
This difference was particularly marked in patients
with a baseline CD4+ lymphocyte count of fewer than
200 cells/mm3. In addition, tuberculosis patients had a
higher rate of AIDS-defining opportunistic infections
and a marked decline of CD4+ lymphocyte count than
controls, but these differences were not statistically
significant. Thus, we conclude that tuberculosis disease affected survival but not occurrence of subsequent opportunistic infections or rate of CD4+ count
decline.
298
Leroy et al.
TABLE 3. Proportional hazard regression analysis controlling for matching variables, differences in
baseline characteristics, and known predictors of mortality in 724 patients with human
Immunodeficiency virus (HIV), Aquitaine Cohort, France, 1985-1994
Unlvariate model
V/nrfahJoQ
ValklDIBo
Mycobacterium tuberculosis disease
Year of HIV diagnosis
Time from HIV diagnosis to Inclusion (month)
Baseline CD4+ cell count (cells/mm*)
(reference: 200/mm3)
50-200
<50
Intravenous drug users
Age (gain of 10 years)
History of AIDSt-defining opportunistic
Infections
Prophylaxis against Pneumocystis carinii
Prolonged treatment after MOTTf
AnBretroviral treatment
MuWvarlale model*
RR
95% Clf
RR
95%Clf
1.56
0.99
1.00
1.17-2.09
0.94-1.04
0.99-1.01
1.59
0.92
0.99
1.18-2.13
0.84-1.02
0.99-1.00
2.01
8.11
0.92
1.19
1.42-2.84
5.74-11.45
0.73-1.17
1.09-1.30
1.63
6.30
1.27
1.29
1.14-2.35
4.32-9.18
0.97-1.67
1.16-1.43
3.81
1.63
1.04
1.78
3.03-4.80
1.28-2.08
0.68-1.58
1.40-2.26
2.18
0.97
0.57
1.29
1.68-2.84
0.70-1.35
0.37-0.89
0.97-1.72
• Each of the risk ratios is adjusted to all of the other variables In the table.
t RR, risk ratio; Cl, confidence interval; AIDS, acquired immunodeficiency syndrome; MOTT, mycobacteria
other than tuberculosis.
To assess the role of tuberculosis disease on the
course of HTV infection, the ideal study design would
compare HTV disease progression of tuberculosis patients and patients without tuberculosis in an incident
cohort followed prospectively from HIV seroconversion. Although our study was a prevalent cohort for
HIV infection, we have compared HIV-infected patients with a similar degree of immunodepression
(baseline CD4+ lymphocyte count) and duration of
known HIV infection (i.e., same year of HTV diagnosis). We also believe that care was comparable between the two groups because of the homogeneity of
practices among the clinicians of the Aquitaine
Cohort.
In our sample, the fact that tuberculosis patients had
more frequent histories of AIDS-defining opportunistic infections at inclusion than controls, despite similar
CD4+ lymphocyte counts, suggests a more clinically
advanced HTV infection in tuberculosis patients. This
discrepancy between biology and clinic confirms the
fact, previously reported, that one measure of CD4+
lymphocyte count is insufficient to evaluate the progression of HIV infection (34); this is particularly true
if tuberculosis infection tends to induce by itself a loss
of CD4+ lymphocytes (16). This might have masked
a different evolution between the two groups. However, adjustment by proportional hazard regression
including clinical and biologic determinants of HTV
disease progression allowed us to take into account
this potentially confounding factor. If the CD4+ lymphocyte counts were temporarily reduced at the time
of tuberculosis diagnosis, a better baseline would have
been to take into account the subsequent higher CD4+
count in tuberculosis patients. Doing this would only
have increased the survival difference between the two
groups.
The diagnosis of tuberculosis disease was confirmed
in patients with clinical suspicion of tuberculosis,
but this systematic investigation could not be made
in controls. However, we guaranteed a minimal
follow-up in all patients, and we have cross-checked
all tuberculosis cases declared to the Department of
Microbiology of the Bordeaux University Hospital
with the Aquitaine Cohort data file. Thus, this selection problem was probably not important in our cohort
and is unlikely to mask a difference between the two
groups.
The reduced survival in tuberculosis patients was a
finding consistent over the different subgroup analyses, except in intravenous drug users and in patients
with a history of AIDS-defining opportunistic infection. These two characteristics were precisely those
altering the comparability of our groups at inclusion.
These discrepancies suggest that intravenous drug users might die before the diagnosis of tuberculosis
disease because of a different access to medical care.
Similarly, patients with a history of opportunistic infection might die before the diagnosis of tuberculosis
disease because of a more advanced HIV infection.
Nevertheless, the history of opportunistic infection
and the low CD4+ lymphocyte count level remained
more strongly associated with mortality than with the
occurrence of tuberculosis disease in the multivariate
analysis.
The excess mortality in tuberculosis patients may be
explained by deaths related to tuberculosis disease or
Am J Epidemiol
Vol. 145, No. 4, 1997
Tuberculosis Disease and Human Immunodeficiency Virus Infection
to other complications of HIV infection. This high
mortality may result from emergence of multidrug
resistance, but we have reported elsewhere (35) that
multidrug resistance was low in our cohort (1.7 percent of 59 tuberculosis patients). In addition, all patients diagnosed in our study had been treated for
tuberculosis disease, except when they died early. All
tuberculosis patients were treated in hospitals and
monitored by experienced practitioners; but unfortunately, evaluation of compliance was not available in
our study. Thus, we suggest that potential therapeutic
problems could not explain all the excess mortality in
tuberculosis patients, and we conclude that most of
them died from associated HIV infection. Nevertheless, a late diagnosis of tuberculosis disease, which in
turn would induce a delay in the specific treatment,
may explain in part this excess mortality (13, 15, 36).
With regard to interaction between tuberculosis disease and HTV infection, HIV-infected patients with
tuberculosis disease appear to have a shorter survival
than HTV-negative patients, despite an adequate antituberculosis therapy (8, 37). In addition, our study has
shown that tuberculosis disease was a strong factor of
mortality among HIV-infected patients. This finding is
consistent with previous studies that suggested the
same effect of tuberculosis disease on mortality among
HIV-infected patients (8, 15). Indirectly, Pape et al.
(14), in Haiti, have shown that preventive therapy of
tuberculosis disease with isoniazid appeared to delay
the onset of HIV-related opportunistic infections and
to prolong survival. However, their findings might
also be explained by a positive influence of tuberculosis therapy on CD4+ lymphocyte counts (38) and by
the lack of control for HIV progression. Recently,
Whalen et al. (19), in a study controlled for HTV
progression, have demonstrated the same effect as in
our study.
Because of the high early mortality rate in HIVinfected patients with tuberculosis disease, care providers following HIV-infected patients must maintain
a high level of suspicion for tuberculosis disease,
particularly in patients with increased risk of developing an active tuberculosis disease, like intravenous
drug users (3, 13). Unfortunately, intravenous drug
users are difficult to follow; assuring compliance with
primary tuberculosis prophylaxis is challenging in
such patients.
We conclude that tuberculosis disease might be a
marker of advanced immunosuppression in HIVinfected patients and a strong factor of mortality, particularly in patients with CD4+ lymphocyte counts of
fewer than 200 cells/mm3. These findings provide an
additional argument for a greater use of tuberculosis
preventive therapy in HIV-infected patients (11). ConAw J Epidemiol
Vol. 145, No. 4, 1997
299
sidering the occurrence of opportunistic infections, our
study did not show an accelerated progression of HTV
infection in relation to tuberculosis disease, in contrast
to the findings of Whalen et al. (19). With a similar
study design, they have shown a greater occurrence of
opportunistic infections in tuberculosis patients than in
controls. This may be subject to detection bias and
may affect their findings because tuberculosis patients
may be followed more closely than controls. Unfortunately, they did not provide data about the follow-up
CD4+ rate decline to support their results. Our results
were congruent regarding the CD4+ count decline and
the occurrence of opportunistic infections. However,
in our study, death may have eliminated the persons
whose CD4+ counts were falling fastest and whose
risk of a subsequent opportunistic infection was the
highest.
Considering these contrasting findings, additional
studies are needed to explore the interaction between
tuberculosis disease and HTV infection. With this objective, studies that also use measurements of the viral
load would be very informative (39).
ACKNOWLEDGMENTS
The authors acknowledge the assistance of the clinicians
participating in the GECSA for data collection as well as
Xavier Anglaret and Luc Letenneur from INSERM 330 of
the Bordeaux University for their helpful comments.
This study was supported in part by the Agence Nationale
de Recherches sur le Sida, France. The work was presented
in part at the Conference on Global Lung Health and the
1995 annual meeting of the International Union Against
Tuberculosis and Lung Disease, Paris, France, September
9-12, 1995 (abstract 272-PA12).
REFERENCES
1. Dolin PJ, Raviglione MC, Kochi A. Global tuberculosis incidence and mortality during 1990-2000. Bull World Health
Organ 1994;72:213-20.
2. Raviglione MC, Snider DE, Kochi A. Global epidemiology of
tuberculosis—morbidity and mortality of a worldwide epidemic. JAMA 1995;273:220-6.
3. Selwyn PA, Hartel D, Lewis VA, et al. A prospective study of
the risk of tuberculosis among intravenous drug users with
immunodeficiency virus infection. N Engl J Med 1989;320:
545-50.
4. De Cock KM, Gnaore E, Adjorlolo G, et al. Risk of tuberculosis in patients with HIV-I and HIV-II infections in Abidjan,
Ivory Coast. BMJ 1991;302:496-9.
5. Di Perri G, Micciolo R, Vento S, et al. Risk of reactivation of
tuberculosis in the course of human immunodeficiency virus
infection. Eur J Med 1993;2:264-8.
6. Centers for Disease Control. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR 1992;
41:1-19.
300
Leroy et al.
7. Chaisson RE, Schecter GF, Theuer CP, et al. Tuberculosis in
patients with the acquired immunodeficiency syndrome: clinical features, responses to therapy, and survival. Am Rev
RespirDis 1987;136:570-4.
8. Braun MM, Badi N, Ryder RW. A retrospective cohort study
of the risk of tuberculosis among women of childbearing age
with HIV infection in Zaire. Am Rev Rcspir Dis 1991 ;143:
501-4.
9. Lucas SB, Hounnou A, Peacock C, The mortality and pathology of HIV infection in a West African city. AIDS 1993;7:
1569-79.
10. Leroy, V, Msellati P, Lepage P. Four years of natural history
of HTV-1 infection in African women of reproductive age: a
prospective cohort study in Kigali, Rwanda, 1988-1993. J
Acquir Immune Defic Syndr Hum Retrovirol 1995;9:415-21.
11. De Cock KM, Grant A, Porter JDH. Preventive therapy for
tuberculosis in HlV-infected persons: international recommendations, research, and practice. Lancet 1995;345:833-6.
12. Shafer R, Edlin B. Tuberculosis in patients infected with
human immmunodeficiency virus: perspective on the past
decade. Clin Infect Dis 1996;22:683-704.
13. Small PM, Schecter GF, Goodman PC, et al. Treatment of
tuberculosis in patients with advanced human immunodeficiency virus infection. N Engl J Med 1991;324:289-94.
14. Pape JW, Jean SS, Ho JL, et al. Effect of isoniazid prophylaxis
on incidence of active tuberculosis and progression of HIV
infection. Lancet 1993;342:268-72.
15. Perneger TV, Sudre P, Lundgren JD, et al. Does the onset of
tuberculosis in AIDS predict shorter survival? Results of a
cohort study in 17 European countries over 13 years. BMJ
1995;311:1468-71.
16. Wallis RS, Vjecha M, Amir-Tahmasseb M, et al. Influence of
tuberculosis on HTV-1: enhanced cytokine expression and
elevated b2 microglobulin in HIV-1 associated tuberculosis.
J Infect Dis 1993;167:43-8.
17. Zhang M, Gong J, Vlyer D, et al. T cell cytokine responses in
persons with tuberculosis and HIV-infection. J Clin Invest
1994;94:2435-42.
18. Grange JM, Stanford JL, Rook GAW. Tuberculosis and
cancer: parallels in host responses and therapeutic approaches? Lancet 1995;345:1350-2.
19. Whalen C, Horsburgh CR, Horn D, et al. Accelerated course
of human immunodeficiency virus after tuberculosis. Am J
Respir Crit Care Med 1995; 151:129-35.
20. Castro KG. Tuberculosis as an opportunistic disease in persons infected with human immunodeficiency virus. Clin Infect
Dis 1995;21(suppl 1):S66-71.
21. American Thoracic Society and Centers for Disease Control
and Prevention. Treatment of tuberculosis and tuberculosis
infection in adults and children. Am J Respir Dis Crit Care
Med 1994;149:1359-74.
22. Anglaret X, Taelman H, Ladner J, et al. Tuberculosis and
chemoprophylaxis in Africa. Lancet 1995;345:867-8.
23. Perriens JH, SaintLouis ME, Mukadi YB, et al. Pulmonary
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
tuberculosis in HIV-infected patients in Zaire—a controlled
trial of treatment for either 6 or 12 months. N Engl J Med
1995;332:779-84.
Dabis F, ChSne G, Salamon R, et al. Hospital-based surveillance of HIV infection: Bordeaux, France, 1983-1990. AIDS
1991;5:774-5.
Chfine G, Dabis F, Salamon R, et al. Hospital-based surveillance of HIV infection: epidemiological trends in Bordeaux,
France, 1983-1989. Eur J Epidemiol 1992;8:58-66.
Institut National de la Statistique et des Etudes Economiques.
Recensement ge'ne'ral de la population. (In French). Paris:
INSEE, 1991:256.
Direction Ge'neYale de la SanUL Surveillance du Sida en
France. (In French). Bull Epidemiol Hebdomadaire 1995;8:
31-7.
Centers for Disease Control. Case definition for public health
surveillance: recommendations and report. MMWR Morb
Mortal Wkly Rep 1990;39:39-40.
Centers for Disease Control. Classification system for human
T-lymphotropic virus type III/lymphadenopathy-associated
virus infections. MMWR Morb Mortal Wkly Rep 1986;35:
334-9.
Rothman KJ. Modern epidemiology. Boston: Little, Brown,
1986.
Chen R, Wei L, Huang H. Methods for calculation of
follow-up rate in a cohort study. (Letter). Int J Epidemiol
1993;22:950-2.
Kaplan EL, Meier P. Nonparametric estimation from incomplete samples. J Am Stat Assoc 1958;53:457-81.
Cox DR, Oakes D. Analysis of survival data. London:
Chapman & Hall, 1984.
Taylor JMG, Fahey JL, Detels R, et al. CD4 percentage, CD4
number and CD4:CD8 ratio in HIV infection: which to use
and how to use. J Acquir Immune Defic Syndr Hum Retrovirol 1989;2:114-24.
Dupon M, Texier-Maugein J, Leroy V, et al. Tuberculosis and
HIV infection: a cohort study of incidence and susceptibility
to antituberculous drugs, Bordeaux, 1985-1993. AIDS 1995;
9:577-83.
Dupon M, Ragnaud JM, Groupe des Infectiolologues du Sud
de la France. Tuberculosis in patients infected with human
immunodeficiency virus: a retrospective multicentre study of
123 cases in France. Q J Med 1992;306:719-30.
Ackah AN, Coulibaly D, Digbeu H, et al. Response to treatment, mortality, and CD4 lymphocyte counts in HIV-infected
persons with tuberculosis in Abidjan, C6te d'lvoire. Lancet
1995;345:607-10.
Martin DJ, Sim JGM, Sole GJ, et al. CD4+ lymphocyte count
in African patients co-infected with HIV and tuberculosis. J
Acquir Immune Defic Syndr Hum Retrovirol 1995;8:386-91.
Mellors J, Kingsley LA, Rinaldo CR, et al. Quantification of
HIV-1 RNA in plasma predicts outcome after seroconversion.
Ann Intern Med 1995;122:573-9.
Am J Epidemiol
Vol. 145, No. 4, 1997