A Randomized, Double-Blind, Placebo-Controlled Trial of the Use of Prednisolone as an Adjunct to Treatment in HIV-1-Associated Pleural Tuberculosis

Page created by Dwight Morris
 
CONTINUE READING
MAJOR ARTICLE

A Randomized, Double-Blind, Placebo-Controlled
Trial of the Use of Prednisolone as an Adjunct
to Treatment in HIV-1–Associated Pleural Tuberculosis
Alison M. Elliott,1,3 Henry Luzze,2 Maria A. Quigley,3 Jessica S. Nakiyingi,1 Steven Kyaligonza,2
Proscovia B. Namujju,1 Constance Ducar,2 Jerrold J. Ellner,4 James A. G. Whitworth,1,3 Roy Mugerwa,2
John L. Johnson,2 and Alphonse Okwera2
1
 Uganda Virus Research Institute, Entebbe, and 2Uganda–Case Western Reserve University Research Collaboration, Mulago Hospital,
Kampala, Uganda; 3London School of Hygiene and Tropical Medicine, London, United Kingdom; 4University of Medicine and Dentistry
of New Jersey, Newark

                                                                                                                                                                    Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
   Background. Active tuberculosis may accelerate progression of human immunodeficiency virus (HIV) infection
by promoting viral replication in activated lymphocytes. Glucocorticoids are used in pleural tuberculosis to reduce
inflammation-induced pathology, and their use also might reduce progression of HIV by suppressing immune
activation. We examined the effect that prednisolone has on survival in HIV-1–associated pleural tuberculosis.
   Methods. We conducted a randomized, double-blind, placebo-controlled trial of prednisolone as an adjunct
to tuberculosis treatment, in adults with HIV-1–associated pleural tuberculosis. The primary outcome was death.
Analysis was by intention to treat.
   Results. Of 197 participants, 99 were assigned to the prednisolone group and 98 to the placebo group. The
mortality rate was 21 deaths/100 person-years (pyr) in the prednisolone group and 25 deaths/100 pyr in the
placebo group (age-, sex-, and initial CD4+ T cell count–adjusted mortality rate ratio, 0.99 [95% confidence
interval, 0.62–1.56] [P p .95 ]). Resolution of tuberculosis was faster in the prednisolone group, but recurrence
rates were slightly (though not significantly) higher, and use of prednisolone was associated with a significantly
higher incidence of Kaposi sarcoma (4.2 cases/100 pyr, compared with 0 cases/100 pyr [P p .02]).
   Conclusions. In view of the lack of survival benefit and the increased risk of Kaposi sarcoma, the use of
prednisolone in HIV-associated tuberculous pleurisy is not recommended.

Glucocorticoids, such as prednisolone, have potent sup-                               infection status [2, 3]. In pleural tuberculosis, the ben-
pressive effects on the immune response to tuberculosis.                              efits are less dramatic, but, in HIV-negative patients,
In the absence of specific treatment, this leads to in-                               the rate of clinical resolution is also more rapid [4–6].
creased susceptibility to tuberculosis [1] but, given with                            Glucocorticoids have also been recommended in forms
antituberculosis drugs, glucocorticoids have some ben-                                of tuberculosis in which inflammatory damage is of
efits. In patients with tuberculous pericarditis, the use                             particular concern (such as meningeal, ocular, and ure-
of prednisolone is associated with better survival and                                teric tuberculosis) and in patients who are seriously ill
more-rapid resolution of the effusion, regardless of HIV                              with any form of tuberculosis [7].
                                                                                         Several lines of evidence suggest that prednisolone
                                                                                      may have particular benefits in HIV-associated tuber-
   Received 10 October 2003; accepted 26 January 2004; electronically published       culosis. Observational studies in Zambia, where pred-
29 July 2004.
    Presented in part: 33rd International Union against Tuberculosis and Lung         nisolone has been used as an adjunct to tuberculosis
Disease World Conference on Lung Health, Montreal, Canada, October 2002; Third        treatment for conventional indications [8], have, after
National AIDS Conference, Kampala, Uganda, October 2002.
   Financial support: Wellcome Trust Career Development Fellowship (grant 049523      2 years of follow-up, suggested a survival benefit in
to A.M.E.); National Institutes of Health (contract NO1-AI45244/AI95383); British     patients with HIV-associated pleural tuberculosis (mor-
Medical Research Council Programme on AIDS in Uganda (to J.A.G.W. and J.S.N.).
   Reprints or correspondence: Dr. Alison Elliott, Uganda Virus Research Institute,   tality rate ratio, 0.3 [P p .03]; A.M.E., unpublished
PO Box 49, Entebbe, Uganda (alison.tom@infocom.co.ug).                                data). Laboratory studies have emphasized the high de-
The Journal of Infectious Diseases 2004; 190:869–78
 2004 by the Infectious Diseases Society of America. All rights reserved.
                                                                                      gree of immune activation associated with active tu-
0022-1899/2004/19005-0001$15.00                                                       berculosis [9] and the role that immune activation plays

                                                                                          Prednisolone in Pleural Tuberculosis • JID 2004:190 (1 September) • 869
in promoting replication of HIV. Cytokines such as tumor ne-         diagnostic purposes; if they were found to have empyema; if they
crosis factor (TNF)–a, induced in response to tuberculosis,          had a second, major HIV-related disease (Kaposi sarcoma, oral-
promote transcription of proviral DNA through activation of          esophageal thrush, or a positive serum cryptococcal antigen test);
host transcription factors, such as NF-kB, which have binding        if they had risk factors for serious steroid-related adverse events
sites on the long-terminal repeat sequence of the viral genome       (a history of diabetes or positive urine glucose, a history or
[10]. Furthermore, glucocorticoids induce transcription and          clinical finding of hypertension, or a history of peptic ulcer dis-
synthesis of the inhibitor of NF-kB (IkB), suggesting that this      ease or mental illness); if standard doses of antituberculosis drugs
might be an important mechanism for their immunosuppres-             could not be used (as in participants with concurrent liver dis-
sive effects [11, 12] and that glucocorticoids might block cy-       ease); or if they were seronegative for HIV.
tokine-induced increases in HIV transcription through this               Ethics approval for the study was given by the Uganda National
pathway. Such effects might be of considerable significance,         AIDS Research subcommittee, the science and ethics committee
because active tuberculosis is associated with increased viral       of the Uganda Virus Research Institute, the Uganda National
load in vivo [13] and with a subsequent increase in the mortality    Council for Science and Technology, and the ethics committee
rate in HIV-infected subjects [14]. The use of prednisolone in       of the London School of Hygiene and Tropical Medicine.
asymptomatic HIV-infected subjects is associated with reduced            Randomization and blinding.           The randomization se-
immune activation and a sustained increase in CD4+ T cell            quence was generated by a statistician who was not involved
counts [15].                                                         in the care of the patients, by use of STATA (version 5; Stata

                                                                                                                                            Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
   We therefore proposed the hypothesis that the use of pred-        Corporation). Randomization was done in blocks of 20. Pred-
nisolone as an adjunct to tuberculosis therapy in HIV-infected       nisolone and matching placebo tablets (Berk Pharmaceuticals)
patients with pleural tuberculosis would be associated with a        were packaged in identical plastic bags, which were labeled with
decrease in viral replication and with a long-term improvement       randomization code numbers by 2 people who were not in-
in survival. Prednisolone is widely available and is cheap; if we    volved in the study. Medical staff gave participants the next
could demonstrate a substantial survival benefit, we hoped that      number in the sequence in the order in which they were en-
the use of prednisolone for this indication could be widely          rolled. All participants and medical, laboratory, and statistical
recommended in sub-Saharan Africa. Set against this, we were
concerned that prednisolone-induced immunosuppression might
increase the risk of opportunistic diseases, including Kaposi
sarcoma, cryptococcal meningitis, and herpes zoster. We de-
signed the present study to address this hypothesis and to assess
the benefits and risks of the intervention.

PARTICIPANTS, MATERIALS, AND METHODS

Participants. Participants were recruited at the National Tu-
berculosis Treatment Centre, Mulago Hospital (Kampala, Ugan-
da), between November 1998 and January 2002. Follow-up for
all participants continued until July 2002. Participants were
screened for inclusion in the study if they (1) were ⭓18 years
old and they presented with clinical features suggesting pleu-
ral tuberculosis, with a pleural effusion occupying at least one-
third of 1 hemithorax (as determined by a radiograph); (2) had
not previously received treatment or prophylaxis for tuberculosis;
(3) had not recently received treatment with glucocorticoids; (4)
were not pregnant or breast-feeding; and (5) were residents of
Kampala. After providing written, informed consent for partic-
ipation in the study and for HIV testing, potential participants
received a full medical examination, provided urine and blood
samples, and underwent pleural aspiration and, if possible, bi-
opsy. All participants received pre– and post–HIV test counseling.
   Participants were excluded if they failed to complete the
screening procedures; if pleural fluid could not be obtained for                     Figure 1.    Trial profile. +, Positive.

870 • JID 2004:190 (1 September) • Elliott et al.
Table 1.     Comparison of placebo and prednisolone groups at enrollment.

                                                                                            Placebo          Prednisolone
                                                                                                                                  a
                    Category, variable                                                     (n p 98)            (n p 99)       P
                    Age, mean (SD), years                                                34 (8)              34 (9)           .70
                    Sex
                      Male                                                                60 (61)             54 (55)         .34
                      Female                                                              38 (39)             45 (45)         …
                    Weight, mean (SD), kgb                                               53 (8)              54 (9)          .21
                    Blood pressure, mean (SD), mm Hg
                      Systolic                                                          101 (10)             102 (13)          .29
                      Diastolic                                                          72 (11)              73 (11)          .81
                    Symptoms
                      Fever                                                               60   (61)           66   (67)        .43
                      Cough                                                               84   (86)           91   (92)        .17
                      Dyspnea                                                             86   (88)           83   (84)        .43
                      Chest painb                                                         82   (85)           84   (85)        .95
                      Anorexia                                                            77   (79)           72   (73)        .34
                      Weight loss                                                         83   (85)           86   (87)        .66
                    Signs

                                                                                                                                                     Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
                      Fever ⭓37.5C                                                       53   (54)           55   (56)        .84
                      Karnofsky score ⭓80%c                                               49   (51)           59   (60)        .17
                                  d
                      Oral thrush                                                          5   (5)             9   (9)         .28
                      Herpes zoster scars                                                 12   (12)           13   (13)        .85
                      Lymphadenopathyd                                                    11   (11)           12   (12)        .86
                    Laboratory findings
                      CD4+ T cell count, median (interquartile range), cells/mL           93 (58–219)        118   (57–211)    .57
                      Confirmed TB                                                        91 (93)             89   (90)        .46
                      Isoniazid resistancee                                                5 (6)               5   (6)        1.00
                      Pyrazinamide resistancee                                             0                   1   (1)         .32
                    Radiography findingsf
                      1 zone affected                                                     18 (19)             14 (15)         .75
                      2 zones affected                                                    46 (47)             49 (51)         …
                      ⭓3 zones affected                                                   33 (34)             33 (34)         …

                      NOTE.    Data are no. (%) of participants, unless otherwise noted. TB, tuberculosis.
                      a
                        For proportions, the x2 test was used; for means, the t test was used; for medians, the Wilcoxon rank sum
                    test was used.
                      b
                         One missing value.
                      c
                        Two missing values.
                      d
                         Three missing values.
                      e
                         Eighty-two participants in each group had complete drug sensitivity data.
                      f
                        One zone equals one-third of 1 hemithorax; enrollment radiographs were not analyzed for 4 participants.

staff remained blinded to the treatment allocation until all data             used for participants with predominantly pleural tuberculosis,
collection had been completed.                                                to make allowance for the effect that rifampicin has on steroid
   Treatment regimens.       Tuberculosis treatment comprised                 metabolism [8, 17].
ethambutol, isoniazid, rifampicin, and pyrazinamide taken on a                   Patient care and follow-up.         Participants either were
daily basis for 2 months, followed by isoniazid and rifampicin                admitted to the tuberculosis ward or (in exceptional circum-
taken on a daily basis for 4 months. Doses were adjusted ac-                  stances) attended the ward daily, for directly observed treatment
cording to each participant’s weight, using standard criteria [16].           for 1 week; during this period, blood pressure was also mon-
   Prednisolone was supplied as 5-mg tablets and was given                    itored daily. After 1 week, participants were taken home by the
concomitantly with tuberculous therapy at a dosage of 50 mg                   home health visitor, to fully identify their addresses. During
daily for 2 weeks, followed by 40 mg daily for 2 weeks, followed              the 8-week period of treatment with the study drug, participants
by 25 mg daily for 2 weeks, followed by 15 mg daily for 2                     collected their medicine from the clinic every 2 weeks; in ad-
weeks; prednisolone treatment was then stopped. This dosage                   dition, they were visited regularly by the home health visitor,
regimen was ∼2.5 times that used in the previous study in                     who counted the number of tablets remaining, for comparison
Zambia, during which rifampicin-containing regimens were not                  with the expected number. Thereafter, participants attended the

                                                                           Prednisolone in Pleural Tuberculosis • JID 2004:190 (1 September) • 871
Table 2.       Comparison of follow-up and adherence to treatment.

                                                                                                        Placebo        Prednisolone
                                                                                                                                                 a
            Category, variable                                                                         (n p 98)          (n p 99)            P
            Follow-up and attendance
              Losses to follow-up/withdrawals                                                         6 (6)                 3 (3)            .30
              Follow-up time, median (interquartile range), years                               1.48 (0.78–2.45)      1.65 (0.90–2.63)       .23
              Follow-up time, total person-years                                                      153                   172              …
              Scheduled visits, total                                                                 923                   983              …
              Unscheduled visits, total                                                               563                   561              …
                                                                                                                  b                 b
                                                                                                       (n p 97)          (n p 97)
            Treatment deviations from the study protocol
              Unscheduled removal of pleural fluid                                                      3   (3)          4   (4)             .70
              Temporary discontinuation of TB treatment due to TB drug reaction                         6   (6)          2   (2)             .15
              Change in TB treatment for TB drug resistance                                             3   (3)          1   (1)             .31
              Discontinuation of study drug                                                             9   (9)         13   (13)            .36
            Participants’ adherence to treatment with study drug
              No. of home visits/patient, mean                                                          5                 5                  .87
              Proportion of visits at which tablet counts were correct, meanc                           0.96              0.98               .07

                                                                                                                                                     Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
              Participants known to have missed 1 week to 1 monthd                                      8 (8)             4 (4)              .23
              Participants known to have missed ⭓1 monthd                                               4 (4)             5 (5)              .73
            Participants’ adherence to tuberculosis treatment
              Urine isoniazid tests, no. positive/performed (%)
                 1 month                                                                               82/88 (93)       85/88 (97)        .30
                 2 months                                                                              73/85 (86)       78/85 (92)        .22
                 6 months                                                                              60/76 (79)       60/80 (75)        .56
              Participants known to have missed 1 week to 1 monthd                                      6 (6)            4 (4)            .52
              Participants known to have missed ⭓1 monthd                                               8 (8)            8 (8)           1.00
            Completion of tuberculosis treatment
              Completede                                                                               81 (83)          85 (86)              .76
              Defaulted before completion                                                               6 (6)            4 (4)               …
              Died before completion                                                                   11 (11)          10 (10)              …

              NOTE.      Data are no. (%) of participants, unless otherwise noted. TB, tuberculosis.
              a
                  For proportions, the x2 test was used; for means, the t test was used; for medians, the Wilcoxon rank sum test was used.
              b
                  Excluding 1 participant in the placebo group and 2 participants in the prednisolone group who had no follow-up.
              c
                  Excluding 5 participants who were not visited at all.
              d
                  Including participants whose drugs were omitted for legitimate reasons as well as defaulters.
              e
                  Including 3 participants who received all drugs but were not seen at or after completion of treatment.

clinic monthly until the end of tuberculosis therapy (at 6                        therapy was monitored by use of urine tests for isoniazid me-
months) and then every 3 months afterward. In addition, par-                      tabolites (BBL Taxo INH strips; Becton Dickinson) at 1, 2, and
ticipants were encouraged to attend the clinic at unscheduled                     6 months. Blood and pleural fluid samples from the first 40
times for clinical care if they became sick, and investigations                   participants were analyzed for HIV load by use of the Amplicor
and treatment were provided free of charge.                                       HIV-1 Monitor test (version 1.5; Roche Diagnostic Systems);
   Clinical laboratory investigations. A diagnosis of HIV in-                     samples of pleural fluid were analyzed at enrollment and at 1
fection was made at enrollment on the basis of a positive rapid                   week; samples of plasma were analyzed at enrollment, at 1 week,
test for HIV (Determine HIV-1–2; Abbott Laboratories) and was                     and at 1, 2, and 6 months.
confirmed either by ELISA (Vironostika HIV-1 Microelisa; Or-                         Diagnosis of tuberculosis. The diagnosis of tuberculosis
ganon Teknika) or by a 3-level, rapid-test algorithm. Serum cryp-                 was made as described elsewhere [18]. Pleural tuberculosis was
tococcal antigen tests (Crypto LA; Wampole Laboratories) were                     considered to be confirmed if a patient had a positive culture
performed at enrollment and as clinically warranted. CD4+ T cell                  for Mycobacterium tuberculosis from pleural biopsy tissue, pleu-
counts (FACScount; Becton Dickinson) were measured at en-                         ral fluid, or sputum or if histopathologic analysis of pleural
rollment and at 1, 2, 6, and 18 months after treatment was begun.                 tissue was consistent with tuberculous pleurisy. Susceptibility
Serum glucose levels were measured at enrollment and at 1 week.                   testing against isoniazid, rifampicin, ethambutol, and pyrazin-
Urine was examined for glucose by dipstick testing at enrollment                  amide was performed by use of the BACTEC radiometric cul-
and at 1 week, 1 month, and 2 months. Adherence to tuberculosis                   ture system (Becton Dickinson).

872 • JID 2004:190 (1 September) • Elliott et al.
signed rank test for matched pairs. We calculated rates for death
                                                                        and for progression to the first episode of each HIV-related
                                                                        disease. Crude and adjusted rate ratios were estimated by use
                                                                        of Cox proportional hazards regression, and P values were es-
                                                                        timated by use of the likelihood ratio test. For HIV-related
                                                                        diseases that could occur more than once, we also calculated
                                                                        overall rates and estimated rate ratios (using a random-effects
                                                                        Poisson model) and P values (using the Wald test).
                                                                           Two interim analyses were conducted by a data monitoring
                                                                        committee not otherwise involved in the trial. The first interim
                                                                        analysis was for safety and included the first 40 participants,
                                                                        with follow-up to at least 1 week. The main outcome was viral
                                                                        load; the objective was to ensure that prednisolone was not
                                                                        associated with an unexpected increase in viral load. The second
                                                                        interim analysis was conducted after 150 participants had been
Figure 2. Kaplan-Meier survival estimates and data showing the num-     enrolled; the objective was to seek the committee’s advice on
ber of participants at risk at the beginning of each time interval.     whether additional funding should be sought for an extension

                                                                                                                                                Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
                                                                        of the trial.

   Diagnosis of major HIV-associated diseases. Kaposi sar-              RESULTS
coma was diagnosed on the basis of clinical findings and was
supported, if the patient consented, by histological examination        Three hundred seventy-seven participants were screened (figure
of a biopsy of a typical lesion. Cryptococcal meningitis was            1). Of these, 22 failed to complete the screening procedures.
diagnosed on the basis of clinical findings and a positive se-          Seventy-six participants were excluded, the majority (48) be-
rum or cerebrospinal fluid cryptococcal antigen test. Esopha-           cause pleural fluid could not be obtained; in many of these
geal candidiasis was a presumptive diagnosis, made on the basis         cases, fluid had already been removed at the referring ward or
of the presence of oral thrush with dysphagia. Diagnoses of             clinic. Eighty-two were HIV seronegative. One hundred ninety-
herpes zoster, herpes simplex, oral thrush, and gastroenteritis         seven HIV-1–seropositive participants entered the trial; 98 re-
were made on the basis of clinical grounds alone.                       ceived placebo and 99 received prednisolone.
   Definition of hypertension and hyperglycemia. Hyper-                    Clinical, radiological, and laboratory findings at enrollment
tension was defined as systolic blood pressure of 1160 mm Hg            in the placebo and prednisolone groups were similar (table 1).
or diastolic blood pressure of 190 mm Hg. Hyperglycemia was             The duration of follow-up, attendance at the clinic, and ad-
defined as a random blood glucose level of 1160 mg/dL.                  herence to protocol in the 2 groups also were similar (table 2).
   Statistical analysis. The primary outcome was mortality.             Three participants were not seen at any follow-up visit, 1 from
On the basis of the preliminary data from Zambia, we designed           the placebo group and 2 from the prednisolone group (1 of
the study to have 80% power to detect a 2-fold difference in            whom had died within 1 week, with clinical findings suggestive
mortality between the placebo and prednisolone groups, with             of deep vein thrombosis and pulmonary embolism). Therefore,
P ! .05. We calculated that, for mortality rates of 35 deaths/100       clinical findings during follow-up are reported for 97 partici-
person-years (pyr) and 17.5 deaths/100 pyr and allowing for             pants in each group. Participants’ adherence to treatment, as
15% loss to follow-up, this would require enrollment of 200             assessed by tablet counts and urine testing for isoniazid, in the
participants, to give a total follow-up of 134 pyr in each group.       2 groups was similar and appeared to be excellent, except at
The analysis was conducted by intention to treat. Secondary             the 6-month visit when the percentages of participants with
outcomes were adverse events attributable to the use of glu-            positive urine results for isoniazid were 75%–79%; however, at
cocorticoids, resolution of tuberculosis, clinical episodes of HIV-     the 6-month visit a proportion of participants had already com-
associated disease, and changes in CD4+ T cell count and viral          pleted the full treatment regimen before attending the clinic.
load (in a subset of cases).                                               Adverse events that were attributed to prednisolone and that
   Data were entered and analyzed by use of FoxPro for Win-             required discontinuation of the use of the study drug were more
dows (version 2.6; Microsoft) and STATA (version 7; Stata Cor-          frequent in the prednisolone group (9/97 [9%]) than in the
poration). Simple comparisons were made by use of x2 tests              placebo group (2/97 [2%]) (P p .03). In the prednisolone group,
and t tests. Comparisons of CD4+ T cell counts, which show              the study drug was discontinued for 2 participants with hyper-
a markedly skewed distribution, were made by use of the Wil-            glycemia, 3 with hypertension, 3 with episodes of active herpes
coxon rank sum test for unmatched groups and the Wilcoxon               zoster, and 1 with esophageal candidiasis; in the placebo group,

                                                                      Prednisolone in Pleural Tuberculosis • JID 2004:190 (1 September) • 873
Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015

Figure 3. Symptoms and signs of pleural tuberculosis in the study population. A, anorexia; B, weight gain; C, cough; D, pleural effusion obscuring
⭓1 hemidiaphragm on chest radiograph. *P ! .05 ; **P ! .01 ; ***P ! .001 . Superscript a denotes one missing value for weight at enrollment (participant
could not stand on scale).

the study drug was discontinued for 1 participant with hyper-                 but the median blood glucose level was 98 mg/dL in both groups
glycemia and for 1 with hypertension. After 1 week of treatment,              (P p .86). The mean systolic blood pressure was significantly
the proportion of participants with a random blood glucose level              higher in the prednisolone group than in the placebo group after
of 1160 g/dL was higher in the prednisolone group than in the                 1 week of treatment (111 mm Hg, compared with 104 mm Hg
placebo group (6/96 [6%], compared with 1/95 [1%] [P p .06]),                 [P p .002]) and after 1 month of treatment (again 111 mm Hg,

874 • JID 2004:190 (1 September) • Elliott et al.
Table 3.      Proportion of participants who developed HIV-related diseases and rates of progression to disease events.

                                                                                            Placebo      Prednisolone
                                                                                                                                                             a
    HIV-related disease, rate                                                              (n p 97)        (n p 97)         Rate ratio (95% CI)          P
                                                                                                                   b                                             c
    Kaposi sarcoma                                                                           0               6 (6)                     …                .03
      Rate to first event, per 100 pyr                                                       0               4.2d                      …                .02e
                                                                                                                   b
    Cryptococcal meningitis                                                                  5 (5)           3 (3)                     …                .47
                                                                                                                d
      Rate to first event, per 100 pyr                                                       3.8             2.0              0.55   (0.13–2.29)        .40
    Esophageal candidiasis                                                                  23 (24)         35 (36)b                   …                .06
                                                                                                                 d
      Rate to first event, per 100 pyr                                                      20.6            28.6              1.39   (0.82–2.36)        .22
      Rate allowing for multiple events      in the same participant, per 100 pyr           34.4            34.2f             1.07   (0.60–1.93)        .82
    Herpes zoster                                                                           19 (19)         22 (23)b                   …                .60
      Rate to first event, per 100 pyr                                                      16.7            17.7d             1.07   (0.58–1.98)        .83
    Oral or genital herpes simplex                                                          20 (21)         22 (23)b                   …                .73
      Rate to first event, per 100 pyr                                                      17.1            16.9d             0.98   (0.53–1.79)        .94
                                                                                                                 f
      Rate allowing for multiple events      in the same participant, per 100 pyr           18.3            20.8              1.19   (0.63–2.25)        .59
    Oral thrush                                                                             31 (32)         31 (32)b                   …               1.00
                                                                                                                 d
      Rate to first event, per 100 pyr                                                      28.3            25.0              0.90   (0.54–1.48)        .67
      Rate allowing for multiple events      in the same participant, per 100 pyr           32.1            38.3f             1.35   (0.75–2.43)        .31
                                                                                                                     b
    Gastroenteritis                                                                         28 (29)         34 (35)                    …                .36

                                                                                                                                                                     Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
                                                                                                                 d
      Rate to first event, per 100 pyr                                                      24.2            27.5              1.15   (0.70–1.90)        .58
      Rate allowing for multiple events      in the same participant, per 100 pyr           29.8            36.2f             1.21   (0.70–2.07)        .50

      NOTE.     CI, confidence interval; pyr, person-years.
      a
        P values given are for x2 tests for proportions. Rate ratios and P values for single events were estimated by use of Cox proportional hazards
    methods. Rate ratios and P values for multiple events (in italics) were estimated by use of a random effects model, to allow for clustering of events
    within individual participants.
      b
        Data are the number (%) of participants who had at least 1 episode (for a total of 194 subjects).
      c
        Fisher’s exact test.
      d
        Data are the rate per 100 pyr and rate ratio for progression to the first event (for a total of 194 subjects).
      e 2
        x test for unequal rates.
      f
        Data are the rate per 100 pyr and rate ratio allowing for multiple episodes in the same patient (for a total of 3030 clinic visits). Multiple episodes
    were defined as episodes separated by at least 2 weeks.

compared with 104 mm Hg [P p .001]), but not at later visits.                      was compared in participants with CD4+ T cell counts of ⭓200
Symptoms suggestive of gastritis or peptic ulcer disease occurred                  cells/mL (n p 54) and !200 cells/mL (n p 143). The mortality
in 18 (19%) of the 97 participants in the prednisolone group                       rate ratio for participants who received prednisolone, compared
and in 12 (12%) of the 97 participants in the placebo group                        with those who received placebo, was 0.47 (95% CI, 0.14–1.61
(P p .23); in both groups, edema occurred in 15 (15%) of the                       [P p .22]) in participants with CD4+ T cell counts of ⭓200/
97 participants (P p 1.00).                                                        mL, whereas it was 1.04 (95% CI, 0.64–1.71 [P p .87]) in par-
   The use of prednisolone had no effect on survival (figure 2).                   ticipants with CD4+ T cell counts of !200/mL. These 2 rates
There were 36 deaths in the prednisolone group and 39 in the                       were not statistically significantly different (P p .13 , for inter-
placebo group. The mortality rate was 21 deaths/100 pyr in the                     action). The second analysis was to determine whether the pred-
prednisolone group and 25 deaths/100 pyr in the placebo group,                     nisolone’s effect changed with time. The mortality rate ratio was
giving a crude mortality rate ratio of 0.84 (95% confidence in-                    0.94 (95% CI, 0.57–1.56 [P p .81]) during the first 1.5 years of
terval [CI], 0.53–1.32 [P p .44]). The mortality rate was higher                   follow-up and was 0.52 (95% CI, 0.18–1.47 [P p .21]) during
in participants with low initial CD4+ T cell counts, and this                      the follow-up period after 1.5 years. These 2 rates were not
was the main confounder. Mortality rates were also higher in                       statistically significantly different (P p .32, for interaction).
participants who were older and in men (compared with                                 The use of prednisolone was associated with more-rapid im-
women), but these 2 effects were not statistically significant.                    provement in all of the principal symptoms and signs of pleural
After adjusting for age, sex, and the natural logarithm of the                     tuberculosis. This effect was statistically significant, particularly
initial CD4+ T cell counts, the mortality rate ratio for the pred-                 during the first few weeks of treatment, for anorexia, weight
nisolone group, compared with the placebo group, was 0.99                          loss, and cough (figure 3A–C). Radiological resolution of the
(95% CI, 0.62–1.56 [P p .95]).                                                     pleural effusion was also more rapid and more complete in the
   Two additional, exploratory analyses of mortality were per-                     prednisolone group (figure 3D) than in the placebo group. Ten
formed. The first analysis was to determine whether predni-                        participants required retreatment for recurrent tuberculosis.
solone’s effect was different in participants with relatively pre-                 The recurrence rate was higher in the prednisolone group (4.5
served immunity. The effect that prednisolone has on survival                      cases/100 pyr) than in the placebo group (1.8 cases/100 pyr),

                                                                                 Prednisolone in Pleural Tuberculosis • JID 2004:190 (1 September) • 875
Table 4.       Effect that prednisolone has on CD4+ T cell count and viral load.

                                                                                       Placebo                         Prednisolone
                                                                             No. of                            No. of
                                                                                                                                                  a
                                                                            samples                           samples                           P (placebo vs.
Category, variable                                                          analyzed           Result         analyzed            Result         prednisolone)
Peripheral blood CD4+ T cell count, median (interquartile
     range), cells/mL
  Enrollment                                                                    98        93   (58–219)           99        118   (57–211)             .57
  4 weeks                                                                       88       150   (74–236)b          92        132   (65–266)b            .90
  8 weeks                                                                       87       153   (75–276)b          90        180   (98–272)b            .58
  24 weeks                                                                      77       176   (114–292)b         83        200   (104–349)b           .33
  72 weeks                                                                      46       174   (85–280)c          49        200   (112–294)
                                                                                                                                            c
                                                                                                                                                       .36
Peripheral blood viral load, mean (SD), log10 RNA copies/mL
  Enrollment                                                                    20          5.39   (0.52)         20          5.16   (0.78)            .29
  1 week                                                                        20          5.57   (0.55)c        19          5.47   (0.80)
                                                                                                                                           b
                                                                                                                                                       .62
  4 weeks                                                                       19          5.75   (0.48)b        19          5.53   (0.69)
                                                                                                                                           c
                                                                                                                                                       .26
                                                                                                         c                                 c
  24 weeks                                                                      15          5.79   (0.49)         17          5.44   (0.65)            .09
Pleural fluid viral load, mean (SD), log10 RNA copies/mL
  Enrollment                                                                    20          6.25 (0.74)           20          5.74 (1.52)              .19

                                                                                                                                                                  Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
  1 week                                                                        16          6.28 (0.60)           11          5.93 (1.14)              .30
  a
      For means, the t test was used; for medians, the rank sum test was used.
  b
      P ⭐ .001, for comparison of value at enrollment with value at given time point for individuals with data at both time points (Wilcoxon signed rank test).
  c
      P ⭐ .05, for comparison of value at enrollment with value at given time point for individuals with data at both time points (Wilcoxon signed rank test).

but this difference was not statistically significant (rate ratio,                   for the first week, and subsequent tablet counts suggested good
2.3 [95% CI, 0.6–9.0] [P p .20]).                                                    compliance with the prescribed regimen. The dose prescribed
   The use of prednisolone had little effect on HIV-associated                       was sufficient to achieve demonstrable pharmacological effects
opportunistic infections; the exception was Kaposi sarcoma,                          in the prednisolone group (i.e., increases in blood sugar and
which occurred only in the prednisolone group (table 3). The                         blood pressure). In addition, the dramatic effect that predni-
diagnosis was confirmed by biopsy in 5 of the 6 cases; the sixth                     solone had on radiological resolution of effusions in the pred-
patient had florid disease with typical lesions but refused to                       nisolone group indicated that the dose was adequate to achieve
undergo a biopsy. All but 1 of the cases occurred during or                          a significant anti-inflammatory effect. On the other hand, we
soon after treatment with prednisolone: the diagnoses were                           found no evidence that prednisolone has any effect on viral
made at 2, 4, 7, 9, 16, and 138 weeks from the start of treatment.                   replication—a quite remarkable result, given the potent effects
In contrast, other conditions, such as cryptococcal meningitis,                      that glucocorticoids have on gene transcription, and on im-
occurred in both groups and throughout the follow-up period.                         mune responses (such as TNF-a and Th1 and Th2 cytokine
   In both the placebo and prednisolone groups, CD4+ T cell                          production) that are believed to influence viral replication and
counts and viral loads (the latter analyzed in the first 40 partic-                  immunity to HIV infection [19].
ipants) increased following treatment (table 4). There were no                          Active pulmonary tuberculosis has been found to have a
statistically significant differences in CD4+ T cell counts or viral                 greater effect on mortality in people with CD4+ T cell counts of
loads at any time point, nor were there any statistically significant                1200 cells/mL and after follow-up of 11 year [14]. This finding
differences in the changes between the values at enrollment and                      suggests that the expected benefit of prednisolone might be
those at subsequent time points (data not shown).                                    greater in participants with higher CD4+ T cell counts and after
                                                                                     prolonged follow-up. However, the present study was not de-
DISCUSSION
                                                                                     signed for these subgroup analyses and no statistically significant
Our results showed that the use of prednisolone as an adjunct                        differences in effect were observed between participants with high
to treatment for tuberculosis did not improve survival in the                        and low CD4+ T cell counts, or by duration of follow-up.
HIV-1–infected adults with pleural tuberculosis who partici-                            There was a marked and sustained increase in CD4+ T cell
pated in our study. This result could have occurred because                          counts following treatment for tuberculosis, but, paradoxically,
the treatment was not taken or because the dose prescribed                           this was accompanied by an increase in viral load, suggest-
was insufficient to achieve the expected effects. These expla-                       ing that the increase in CD4+ T cell counts may not have
nations are unlikely, however. Treatment was directly observed                       been associated with an improvement in the immune response

876 • JID 2004:190 (1 September) • Elliott et al.
against HIV infection. A similar increase in CD4+ T cell counts           specifically enhances the risk of developing Kaposi sarcoma.
occurred in both the prednisolone and placebo groups, which               Current theories suggest that the production of Th1 cytokines
is in keeping with prednisolone’s lack of effect on mortality.            and extravasation of lymphocytes and monocytes are involved
   We did, however, observe a considerable benefit from the               in the development of Kaposi lesions [26]; surprisingly, these
use of prednisolone with respect to the rate of improvement in            processes are the opposite of the principal systemic effects of
symptoms and in radiological resolution of tuberculosis. This             glucocorticoids [27]. Thus, the mechanisms by which pred-
finding was in keeping with those of previous studies of HIV-             nisolone promotes the development of Kaposi sarcoma remain
seronegative patients [4–6]. The adverse pharmacological ef-              to be determined.
fects of prednisolone observed were relatively minor and could               Kaposi sarcoma is a disease that causes prolonged suffering,
be monitored quite easily. Given the neutral effect that pred-            is difficult and expensive to treat, and is usually fatal. We there-
nisolone has on mortality and viral load, these results might             fore conclude that, in view of the lack of survival benefit and
suggest that the use of prednisolone should be implemented.               the significant increase in incidence of Kaposi sarcoma, pred-
Despite limited diagnostic facilities, the identification of suitable     nisolone should not be used in the treatment of pleural tu-
patients would be relatively straightforward in sub-Saharan Af-           berculosis. This recommendation can be applied regardless of
rica, where, after the exclusion of such clinically obvious diag-         HIV status, because no long-term benefits have been reported
noses as empyema and cardiac, renal, or hepatic failure, the vast         with regards to the use of glucocorticoids in patients with pleu-
majority of patients presenting with lymphocytic pleural effusions        ral tuberculosis who are not infected with HIV [28]. This rec-

                                                                                                                                                        Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
are found to have tuberculosis [18, 20]. Of concern would be              ommendation cannot necessarily be generalized to apply to the
the inadvertent treatment of patients with a second opportunistic         use of prednisolone for other forms of tuberculosis or for other
infection, such as the 3 participants with cryptococcal disease in        indications in HIV-infected patients. In the case of pericardial
the present study, who were identified by serum cryptococcal              tuberculosis, the use of prednisolone has been shown to im-
antigen testing and would not have been excluded without the              prove survival regardless of HIV status, and its use should be
use of this expensive, and rarely available, test.                        continued [2, 3]. Similarly, the use of prednisolone is lifesaving,
   Of much greater concern was the striking association observed          and should continue, in severe Pneumocystis carinii pneumonia;
                                                                          however, in HIV-infected patients, prednisolone should be used
between the use of prednisolone and the incidence of Kaposi
                                                                          with caution for indications where no beneficial effect on im-
sarcoma. Given the relatively small number of cases and the fact
                                                                          mediate survival can be expected.
that similar associations were not seen for other opportunistic
infections, this finding might be dismissed as a chance associa-
tion. We believe that this would be inappropriate, for several
                                                                          Acknowledgments
reasons. First, the result supports the findings of the previous
study in Zambia in which cases of Kaposi sarcoma occurred                    We thank the staff of the Uganda–Case Western Reserve University Re-
                                                                          search Collaboration, Mulago Hospital (Kampala, Uganda); the Depart-
only in the prednisolone-treated group [8]. Second, in the pres-          ment of Medicine, New Mulago Hospital (Kampala, Uganda); and the
ent study, there was a marked temporal association between                AIDS Support Organization (Entebbe, Uganda), for referral and care of
the use of prednisolone and the date of diagnosis of Kaposi               the study participants. We are indebted to all of the patients who partic-
                                                                          ipated in this study and to the staff of the tuberculosis ward, for nursing
sarcoma in all but 1 of the cases. Third, there have been a               care. We thank the staff of the statistics unit and of the microbiology
considerable number of reports of Kaposi sarcoma occurring                laboratory of the Medical Research Programme on AIDS in Uganda; Moses
in patients not infected with HIV following the use of immu-              Joloba and the staff of the Joint Clinical Research Centre Mycobacteriology
                                                                          Laboratory; the Uganda Tuberculosis Investigations Bacteriological Unit
nosuppressive drugs—particularly of glucocorticoids—after                 (Wandegeya, Kampala); and Michael Odida, who performed the histolog-
transplantation, in treatment of autoimmune diseases, and in              ical investigations. We thank the Data Monitoring and Ethical Committee.
tuberculous pericarditis [21–23]. In some of these cases, Kaposi
sarcoma resolved spontaneously after immunosuppressive ther-
apy had been stopped [21, 23]. Exacerbation of Kaposi sarcoma             References
during the use of glucocorticoids has also been observed in                1. Fred L, Levin MH, Rivo JB, Barrett TF. Development of active pul-
HIV-infected patients and, in one of these patients, the lesions              monary tuberculosis during ACTH and cortisone therapy. JAMA 1951;
resolved spontaneously when immunosuppressive therapy was                     147:242–6.
                                                                           2. Strang JIG, Kakaza HHS, Gibson DG, et al. Controlled clinical trial of
discontinued [24]. Kaposi sarcoma has also been noted to im-                  complete open surgical drainage and of prednisolone in treatment of
prove after treatment of HIV with antiretroviral therapy [25].                tuberculous pericardial effusion in Transkei. Lancet 1988; 2:759–64.
Kaposi sarcoma is an angioproliferative disease and is associated          3. Hakim JG, Ternmouth I, Mushangi E, Siziya S, Robertson V, Malin A.
                                                                              Double blind randomised placebo controlled trial of adjunctive pred-
with infection with human herpesvirus–8. Taken together with                  nisolone in the treatment of effusive tuberculous pericarditis in HIV
earlier reports, our findings suggest that the use of prednisolone            seropositive patients. Heart 2000; 84:183–8.

                                                                        Prednisolone in Pleural Tuberculosis • JID 2004:190 (1 September) • 877
4. Lee C-H, Wang W-J, Lan R-S, Tsai Y-I, Chiang Y-C. Corticosteroids             18. Luzze H, Elliott AM, Joloba ML, et al. Evaluation of suspected tuber-
    in the treatment of tuberculous pleurisy. Chest 1988; 94:1256–9.                  culous pleurisy: clinical and diagnostic findings in HIV-1–positive and
 5. Galarza I, Canete C, Granados A, Estopa R, Manresa F. Randomised                  HIV-negative adults in Uganda. Int J Tuberc Lung Dis 2001; 5:746–53.
    trial of corticosteroids in the treatment of tubercouls pleurisy. Thorax      19. Sereti I, Lane HC. Immunopathogenesis of human immunodeficiency
    1995; 50:1305–7.                                                                  virus: implications for immune-based therapies. Clin Infect Dis 2001;
 6. Wyser C, Walzl G, Smedema JP, Swart F, van Schalkwyk EM, van de                   32:1738–55.
    Wal BW. Corticosteroids in the treatment of tuberculous pleurisy. Chest       20. Batungwanayo J, Taelman H, Allen S, Bogaerts J, Kagame A, van de
    1996; 110:333–8.                                                                  Perre P. Pleural effusion, tuberculosis and HIV-1 infection in Kigali,
 7. Ross JD, Horne NW, Crofton J. Modern drug treatment in tuberculosis.              Rwanda. AIDS 1993; 7:73–9.
    5th ed. London: The Chest Heart and Stroke Association, 1976.
                                                                                  21. Nagy S, Gyulai R, Kemeny L, Szenohradsky P, Dobozy A. Iatrogenic
 8. Elliott AM, Halwiindi B, Bagshawe A, et al. Use of prednisolone in the
                                                                                      Kaposi’s sarcoma: HHV8 positivity persists but the tumors regress
    treatment of HIV-positive tuberculosis patients. Q J Med 1992; 85:855–60.
                                                                                      almost completely without immunosuppressive therapy. Transplanta-
 9. Vanham G, Edmonds K, Qing L, et al. Generalized immune activation
                                                                                      tion 2000; 69:2230–1.
    in pulmonary tuberculosis: coactivation with HIV infection. Clin Exp
    Immunol 1996; 103:30–4.                                                       22. Sato-Matsumura KC, Matsumura T, Nabeshima M, Katano H, Sata T,
10. Zhang Y, Nakata K, Weiden M, Rom WN. Mycobacterium tuberculosis                   Koizumi H. Serological and immunohistochemical detection of human
    enhances human immunodeficiency virus–1 replication by transcriptional            herpesvirus 8 in Kaposi’s sarcoma after immunosuppressive therapy
    activation at the long terminal repeat. J Clin Invest 1995; 95:2324–31.           for bullous pemphigoid. Br J Dermatol 2001; 145:633–7.
11. Scheinman RI, Cogswell PC, Lofquist AK, Baldwin AS. Role of tran-             23. Sangiorgi G, Orlandi A, De Nardo D, Sangiorgi M, Spagnoli LG. Com-
    scriptional activation of IkBa in mediation of immunosuppression by               plete regression of iatrogenic Kaposi’s sarcoma due to corticosteroid
    glucocorticoids. Science 1995; 270:283–6.                                         treatment in a patient with tubercular pericarditis: a case report. Ann
12. Auphan N, DiDonato JA, Rosette C, Helmberg A, Karin M. Immuno-                    Ital Med Int 1993; 8:21–4.

                                                                                                                                                                    Downloaded from http://jid.oxfordjournals.org/ by guest on August 14, 2015
    suppression by glucocorticoids: inhibition of NF-kB activity through          24. Gill PS, Loureiro C, Bernstein-Singer M, Rarick MU, Sattler F, Levine
    induction of IkB synthesis. Science 1995; 270:286–90.                             AM. Clinical effects of glucocorticoids on Kaposi’s sarcoma related to
13. Goletti D, Weissman D, Jackson RW, et al. Effect of Mycobacterium                 the acquired immunodeficiency syndrome (AIDS). Ann Intern Med
    tuberculosis on HIV replication: role of immune activation. J Immunol             1989; 110:937–40.
    1996; 157:1271–8.                                                             25. Gill J, Bourboulia D, Wilkinson J, et al. Prospective study of the effects
14. Whalen CC, Nsubuga P, Okwera A, et al. Impact of pulmonary tu-                    of antiretroviral therapy on Kaposi sarcoma–associated herpesvirus in-
    berculosis on survival of HIV-infected adults: a prospective epidemi-             fection in patients with and without Kaposi sarcoma. J Acquir Immune
    ologic study in Uganda. AIDS 2000; 14:1219–28.                                    Defic Syndr 2002; 31:384–90.
15. Andrieu J-M, Lu W, Levy R. Sustained increases in CD4 cell counts in
                                                                                  26. Ensoli B, Sturzl M, Monini P. Reactivation and role of HHV-8 in
    asymptomatic human immunodeficiency virus type 1–seropositive pa-
                                                                                      Kaposi’s sarcoma initiation. Adv Cancer Res 2001; 81:161–200.
    tients treated with prednisolone for 1 year. J Infect Dis 1995; 171:523–30.
                                                                                  27. Elenkov IJ, Chrousos GP. Stress hormones, proinflammatory and anti-
16. American Thoracic Society. Treatment of tuberculosis and tuberculosis
    infection in adults and children. Am J Respir Crit Care Med 1994;                 inflammatory cytokines, and autoimmunity. Ann N Y Acad Sci 2002;
    149:1359–74.                                                                      966:290–303.
17. McAllister WA, Thompson PJ, Al-Habet SM, Rogers HJ. Rifampicin                28. Dooley DP, Carpenter JL, Rademacher S. Adjunctive corticosteroid ther-
    reduces effectiveness and bioavailability of prednisolone. Br Med J (Clin         apy for tuberculosis: a critical reappraisal of the literature. Clin Infect
    Res Ed) 1983; 286:923–5.                                                          Dis 1997; 25:872–87.

878 • JID 2004:190 (1 September) • Elliott et al.
You can also read