Rationale and design of the Investigation of the
Management of Pericarditis (IMPI) trial: A 2 × 2 factorial
randomized double-blind multicenter trial of adjunctive
prednisolone and Mycobacterium w immunotherapy in
Bongani M. Mayosi, DPhil,aMpiko Ntsekhe, PhD,aJackie Bosch, Msc,bJanice Pogue, PhD,bFreedom Gumedze, PhD,c
Motasim Badri, PhD,a,vHyejung Jung, Msc,bShaheen Pandie, FCP(SA),aMarek Smieja, PhD,bLehana Thabane, PhD,b
Veronica Francis, RN,aKandithal M. Thomas, MBBS,dBaby Thomas, MBBS,dAbolade A. Awotedu, MBBS, FWACP,d
NombuleloP.Magula, FCP(SA),eDatshanaP. Naidoo,MD,eAlbertinoDamasceno,PhD,fAlfredChitsa Banda,MBChB,g
Arthur Mutyaba, MBChB,hBasil Brown, FCP(SA),hPatrick Ntuli, FCP(SA),iPhindile Mntla, FCP(SA),i
Lucas Ntyintyane, PhD,jRohan Ramjee, FCP(SA),jPravin Manga, FCP(SA),jBruce Kirenga, MBChB,k
Charles Mondo, PhD,kJames B. W. Russell, MBChB,lJacob M. Tsitsi, FCP(SA),mFerande Peters, FCP(SA),m
Mohammed R. Essop, FCP(SA),mAyub Felix Barasa, MBChB,nMuhammad S. Mijinyawa, MBChB,o
Mahmoud U. Sani, MBChB,oTaiwo Olunuga, MBChB,pOkechukwu Ogah, MBChB,pAdewole Adebiyi, FWACP,q
Akinyemi Aje, FMCP,qVictor Ansa, MBBS, FWACP,rDike Ojji, MBChB,sSolomon Danbauchi, MBChB,t
Pretoria, Gauteng, and Soweto, South Africa; Ontario, Canada; Maputo, Mozambique; Lilongwe, Malawi; Kampala,
and Harare, Zimbabwe
Background In spite of antituberculosis chemotherapy, tuberculous (TB) pericarditis causes death or disability in nearly
half of those affected. Attenuation of the inflammatory response in TB pericarditis may improve outcome by reducing cardiac
tamponade and pericardial constriction, but there is uncertainty as to whether adjunctive immunomodulation with
corticosteroids and Mycobacterium w (M. w) can safely reduce mortality and morbidity.
Objectives The primary objective of the IMPI Trial is to assess the effectiveness and safety of prednisolone and M. w
immunotherapy in reducing the composite outcome of death, constriction, or cardiac tamponade requiring pericardial
drainage in 1,400 patients with TB pericardial effusion.
Design The IMPI trial is a multicenter international randomized double-blind placebo-controlled 2 × 2 factorial study.
Eligible patients are randomly assigned to receive oral prednisolone or placebo for 6 weeks and M. w injection or placebo for
3 months. Patients are followed up at weeks 2, 4, and 6 and months 3 and 6 during the intervention period and 6-monthly
thereafter for up to 4 years. The primary outcome is the first occurrence of death, pericardial constriction, or cardiac
tamponade requiring pericardiocentesis. The secondary outcome is safety of immunomodulatory treatment measured by effect
on opportunistic infections (eg, herpes zoster) and malignancy (eg, Kaposi sarcoma) and impact on measures of
immunosuppression and the incidence of immune reconstitution disease.
FromaThe Cardiac Clinic, Department of Medicine, Groote Schuur Hospital and University
of Cape Town, Cape Town, South Africa,bMcMaster University and the Population Health
Research Institute, Hamilton Health Sciences, Hamilton, Ontario, Canada,cDepartment of
Statistical Sciences, University of Cape Town, Cape Town, South Africa,dNelson Mandela
Academic Hospital, Mthatha, South Africa,eKing Edward VIII Hospital, Durban, South
Services, Lilongwe, Malawi,hProvincial and Livingstone Hospitals, Port Elizabeth, South
Johannesburg Academic Hospital, Gauteng, South Africa,
Kampala, Uganda,lConnaught Hospital, Freetown, Sierra Leone,mChris Hani Baragwa-
nath Hospital, Soweto, South Africa,
Teaching Hospital, Kano, Nigeria,
Nigeria,qUniversity College Hospital, Ibadan, Oyo State, Nigeria,rUniversity of Calabar
fHospital Central de Maputo, Maputo, Mozambique,gMalawi Military Health
iDr George Mukhari Hospital, Pretoria, South Africa,
kNew Mulago Hospital,
nMoi Hospital, Eldoret, Kenya,
pFederal Medical Centre, Abeokuta, Ogun State,
Teaching Hospital,Calabar, Nigeria,sAbuja TeachingHospital, Abuja, Nigeria,tAhmadu
Bello Teaching Hospital, Zaria, Kaduna State, Nigeria and,
vPresent address: College of Medicine, King Saud Bin Abdelaziz University of Medical
Sciences, Riyadh, Kingdom of Saudi Arabia.
RCT reg no. NCT100810849.
Submitted February 4, 2012; accepted August 22, 2012.
Reprint requests: Bongani M. Mayosi, Department of Medicine, J Floor Old Groote Schuur
Hospital, Observatory, Cape Town, 7925, South Africa.
0002-8703/$ - see front matter
© 2012, Mosby, Inc. All rights reserved.
Conclusions IMPI is the largest trial yet conducted comparing adjunctive immunotherapy in pericarditis. Its results will
define the role of adjunctive corticosteroids and M. w immunotherapy in patients with TB pericardial effusion. (Am Heart J
In 2010, there were approximately 8.8 million new
cases of tuberculosis (TB) and 1.1 million deaths from TB
among HIV-negative people and an additional 0.35
million deaths from HIV-associated TB in the world.1
HIV infection is associated with a higher risk of
extrapulmonary forms of TB, such as TB pericarditis.2
If pericardial disease occurs in 1% to 2% of all cases of TB,3
then an estimated 440,000 to 880,000 individuals will
be affected with TB pericarditis worldwide over the
next 5 years.
Tuberculous pericarditis is a significant clinical
problem in developing countries where it is a common
cause of heart failure and the most frequent cause of
constrictive pericarditis.4,5In the Western Cape Prov-
ince of South Africa, TB pericarditis accounted for 70%
of cases with large pericardial effusion.6By contrast, TB
pericarditis accounts for only 4% of cases of pericarditis
in developed regions.7In sub-Saharan Africa, at least half
the patients with large pericardial effusions are infected
Current therapy for TB pericarditis consists of medical
treatment with rifampicin, isoniazid, pyrazinamide, and
ethambutol for 6 months, pericardial drainage for cardiac
tamponade, and pericardiectomy for pericardial constric-
tion.9Mortality and morbidity, however, remain high in
spiteofanti-TB chemotherapy. DatafromourInvestigation
mp-ee” for warriors in Zulu) Registry indicate that TB
pericarditis has a case fatality rate of 26% at 6 months,
which rises to 40% in people with AIDS.10Pericardial
constriction, which occurs in 10% to 30% of cases of TB
pericardial effusion, is a major cause of heart failure and,
without surgical resection, has a high fatality rate.11
The attenuation of the inflammatory response in TB
pericarditis may improve outcome by reducing cardiac
tamponade, pericardial constriction, and death. A meta-
(16 deaths/167 active group vs 27 deaths/183 control
were too small to provide a reliable assessment of steroids
on survival.12Although the data are promising, concerns
remain that corticosteroids might increase opportunistic
addition, guidelines from major organizations support the
perception of uncertainty about the role of corticoste-
roids in this condition. European Society of Cardiology
guidelines state that the use of steroids in TB pericarditis
“remains controversial.”15This European Society of
Cardiology guideline highlights the conflicting evidence
and divergent opinion about the effectiveness of steroids
in TB pericarditis, such that the use of adjunctive steroids
in this condition is regarded as “less well established by
evidence” (ie, class IIb). The uncertainty concerning the
place of adjunctive corticosteroids in TB pericarditis is
reflected by the marked variation in the use of
corticosteroids by practitioners, with nearly half of
patients not receiving this intervention.16Taken together
with the statistical uncertainty regarding the effective-
ness of adjunctive steroids in TB pericarditis, these
observations probably reflect a state of clinical equipoise
among practitioners who care for patients with TB
pericarditis. These data provide a justification for the
establishment of adequately powered randomized clini-
cal trials to assess the effectiveness of adjunctive
corticosteroids in TB pericarditis.
In addition to the promising but inconclusive evidence
on adjunctive steroids, there is preliminary evidence
suggesting that repeated doses of intradermal heat-killed
Mycobacterium w (M. w) immunotherapy may reduce
T-cell count in people infected with HIV.17Mycobacter-
ium w is a nonpathogenic, saprophytic, rapidly growing
atypical mycobacterium species, which has shown
clinical benefit, when administered as a heat-killed
intradermal formulation, in the treatment of leprosy18,19
and may have benefit in treating pulmonary TB,17HIV
infection,20and anogenital warts.21Mycobacterium
w shares B- and T-cell antigen epitopes with Mycobac-
terium leprae and with Mycobacterium tuberculosis.22
The data for immunotherapy with intradermal heat-killed
M. w in human TB are suggestive of faster sputum
clearance and lower relapse but require a large, adequate-
ly powered trial to examine whether it reduces clinically
important events. Given the widespread use of M. w for
leprosy, its safety record, and minimal side effects and
promise in pulmonary TB, it could be effective in
extrapulmonary TB, for example, TB pericarditis.
The IMPI Trial
The primary objective of the IMPI Trial is to assess the
effectiveness and safety of adjunctive oral prednisolone
and parenteral M. w immunotherapy in reducing the
composite outcome of death, constriction, or cardiac
tamponade requiring pericardial drainage in patients with
presumed or proven TB pericardial effusion over a
median follow-up period of 2 years.
2 Mayosi et al
American Heart Journal
IMPI is a prospective multicenter international double-
blind placebo-controlled, randomized trial with a 2 × 2
factorial design, evaluating the effects of prednisolone
and M. w immunotherapy (and their combination) on a
composite of clinical outcomes of pericarditis and safety
(Figure). In this factorial design involving 1,400 patients,
the main assessment of the effect of adjunctive steroids
involves a comparison of outcomes among 700 patients
allocated to the adjunctive steroid group versus outcomes
among all 700 patients allocated to the placebo adjunc-
tive steroid group. Similarly, the assessment of the effect
of M. w immunotherapy involves comparison of out-
comes between 700 patients in each of the experimental
and control groups. If significant interactions occur, the
effects of each therapy alone and in combination can also
Patients with suspected TB pericarditis with the
following criteria are eligible if they have (1) confirmed
pericardial effusion on echocardiography (ie, echo-free
space of at least 1 cm at any location around the heart and
at any phase of the cardiac cycle), (2) evidence of definite
or probable TB pericarditis23(Table I), (3) within 1 week
of starting of anti-TB treatment, (4) indicated their
willingness to participate for the full duration of the
trial, and (5) provided written informed consent.
Exclusion criteria are (1) presence of an alternative
cause of pericardial disease, for example, penetrating
IMPI Trial study design.
Table I. Diagnostic criteria for TB pericarditis in a TB-endemic country or community23
Definite TB pericarditis
• Tubercle bacilli are found in stained smear or culture of pericardial fluid; or
• Tubercle bacilli or caseating granulomata are found on histologic examination of pericardium
• Evidence of pericarditis in a patient with TB demonstrated elsewhere in the body; or
• Lymphocytic pericardial exudate with elevated adenine deaminase activity; or
• A Tygerberg TB Pericarditis Diagnostic Index Score ≥6 in patients in whom pericardiocentesis
is not feasible, and other causes of pericarditis have been excluded
Probable TB pericarditis
Mayosi et al 3
American Heart Journal
Volume 0, Number 0
chest trauma in the previous 12 months or malignancy;
(2) use of corticosteroids within the previous month; (3)
hypersensitivity or allergy to the M. w vaccine; (4)
pregnancy; and/or (5) age b18 years.
Before the full scale phase of the trial, an internal pilot
phase was conducted to ensure that the centers had the
ability to (1) enroll patients, (2) adhere to the protocol,
and (3) follow up patients successfully. This pilot phase
allowed us to measure patient adherence with the
treatment protocols and the preliminary safety of the
coadministration of prednisolone and M. w. To increase
the likelihood that the targets of the trial will be met, a
number of measures have been taken. These included (1)
Registry8,10,16; (2) including centers with a documented
large number of patients with TB pericarditis over the
previous 2 years; (3) ensuring each center is able to
perform echocardiograms, routinely perform pericardio-
centesis, and is able to perform surgical pericardiectomy;
(4) using uniform diagnostic criteria for TB; and (5)
preparing a larger formulation of prednisolone tablets (40
and 30 mg) in addition to the 5 mg tablet to increase
adherence to study medication. The anti-TB treatment
course programs, which are the cornerstone of the World
Health Organization STOP TB initiative.24All patients
with HIV were treated according to international guide-
lines developed for African countries. Patients were
provided with mobile cell phones to improve follow-up.
The internal pilot study of first 200 patients confirmed
the feasibility of conducting a large scale multicenter
clinical trial in patientswith TBpericarditis in sub-Saharan
Africa and also provided preliminary safety data, which
were reviewed by the independent trial monitoring
committee. The minimum recruitment rate of 1 partici-
pant per center per month was achieved, with 50% of
screened patients being enrolled in the trial. Ninety-five
percent of the patients are being followed up, and the
missing 5% are being intensely tracked by the Project
Coordinating Office, with the expectation that most of
these individuals will be “recovered.” Completeness of
case report forms (ie, complete and error free) is 96%.
Adherence to allocated medications is N80% (which is the
same as the target stated in the protocol).
Patients are randomized to experimental and control
arms using central concealed randomization, which is
stratified by site and uses a random block size. Key
clinical data are provided by telephone/fax to a
centralized randomization service located at the IMPI
Project Coordinating Office, Cape Town, South Africa.
Participants assigned to the prednisolone experimen-
tal arm receive a 6-week tapering course of predniso-
lone, and those assigned to the prednisolone control
arm receive the same number of placebo tablets, given
concomitantly with standard anti-TB treatment. The
duration of 6 weeks of adjunctive corticosteroids was
based on previous experience of effectiveness of a
similar period of use of additional steroids in TB
meningitis.25Prednisolone and placebo are supplied
as identical tablets (5, 30, and 40 mg) and given at a
dosage of 120 mg/d in the first week, followed by 90
mg/d in the second week, 60 mg/d in the third week,
30 mg/d in the fourth week, 15 mg/d in the fifth week,
and 5 mg/d in the sixth week.
Patients enrolled in the M. w experimental arm
receive 5 doses of 0.1 mL of the vaccine intradermally
(on enrollment; at 2, 4, and 6 weeks; and 3 months).
Patients in the control arm of the M. w comparison
receive a similar regimen of placebo injections of
normal saline in identically packaged vials.
Patients are screened after undergoing echocardiogra-
phy with the permission of the treating physician and
then randomized after informed consent if eligibility
criteria are met. The attending physician is expected to
make every effort to search for TB through testing of
sputum, gastric lavage, urine and blood culture, and
lymph node aspiration/biopsy, whenever feasible, and to
test the patient for HIV, white cell count, serum
creatinine, and globulin. In cases where pericardiocent-
esis is feasible, the pericardial fluid is sent for differential
white cell count, adenine deaminase level, microscopy
for acid fast bacilli, and culture for M tuberculosis.
There are several quality control measures that are
implemented in the echocardiographic study of IMPI.
First, all participating centers are required to have an
echocardiographer with a standard of training that is
equivalent to at least level 1 of the American Society of
Echocardiography (ie, a minimum of 6 months of
training) as a condition for participating in the study.
Second, site initiation training includes an inspection of
echocardiographic equipment and verification of the
ability of trial echocardiographers to identify the
specific outcomes of the study based on a training set
of echocardiograms that illustrate typical examples of
pericardial effusion, pericardial effusion with features of
tamponade, and constrictive pericarditis. Third, the
case report form requires the physician to indicate the
criteria for diagnosis of pericardial effusion (ie, size of
echo-free space) and the features of tamponade and
constriction that are defined in Table II. Finally, centers
are required to store all echocardiographic images of
4 Mayosi et al
American Heart Journal
participants as source data (as a picture, video, or on
DVD). A random selection of enrollment and outcomes-
related images are verified by the Outcomes Adjudica-
The primary outcome is the first occurrence of death,
pericardial constriction, or cardiac tamponade requiring
pericardiocentesis. The secondary outcome is safety of
immunomodulatory treatment measured by effect on
opportunistic infections (including herpes zoster, genital
herpes, cryptococcal meningitis, esophageal candidiasis,
Kaposi sarcoma and lymphoma) and impact on measuresof
immunosuppression (eg, CD4+ T-lymphocyte cell count)
and the incidence of immune reconstitution disease.26
The study outcome definitions are listed in Table II.
Central events adjudication
A committee of clinicians blinded to treatment alloca-
tion adjudicates all primary and secondary outcomes of
the trial (the Outcomes Adjudication Committee).
The trial proposes to enroll 1,400 patients. This sample
size is determined for the effect of each intervention on
the composite outcome of death, clinically important
on a log-rank test assuming recruitment over 39 months
and the last participant being followed for 6 months (total
of 45 months with mean follow-up of approximately 24
months). The sample size calculation was based on an
assumed control composite event rate of 35%, a 30% to
40% (midpoint of 35%) relative risk reduction in the
composite end point with either steroid therapy or M. w,
a loss-to-follow-up rate of 6% over 2 years, rates of
nonadherence of 10% in the active treatment arm, a 1:1
allocation ratio, a power N80%, and a 95% level of
significance. The sample size calculation also accounted
for the potential of subadditive interaction between
steroids and M. w on the primary outcome. Much of
these inputswere informed bydata derivedfromthe IMPI
There are at least 2 potential limitations of our study. The
first relates to the potential for interaction between
adjunctive prednisolone and M. w immunotherapy in
patients on active interventions resulting either synergistic
or subadditive interactions. The immunostimulatory M.
interaction).27A synergistic interaction increases power,
whereas a subadditive interaction will decrease power.
We found 1 experimental study that raises the potential
for subadditivity as well as a study on surrogate markers
that indicate a potential for synergistic effects. The
potential for subadditive interactions is suggested in an
Table II. Definition of study outcomes
Death will be classified as all-cause mortality.
Cardiac tamponade can be diagnosed if the following clinical and echo
criteria are met:
1) Heart rate N90
2) Systolic blood pressure b100 mm Hg
3) Pulsus paradoxus N10 mm Hg
4) Jugular venous pressure N4 cm
Echo confirmed circumferential pericardial effusion measuring N1 cm
anterior to the right ventricle in the subcostal view and any 1 of
(a) Swinging heart
(b) Right or left ventricular diastolic collapse
A diagnosis of constrictive pericarditis can be made if the following
and echo criteria are met.
1) Pericardial thickening N3 mm in the absence of pericardial effusion
2) A left ventricular ejection fraction ≥50%
plus reciprocal effect on tricuspid flow and/or expiratory increase of
hepatic vein diastolic flow reveral
Plus any 3 of the following
(a) Pulsus paradoxus N10 mm Hg
(c) Jugular venous pressure N4 cm
(c) Pericardial knock or early third heart sound
(d) Sudden instantaneous split of the second heart sound on inspiration
(e) Peripheral edema, hepatomegaly, and/or ascites
B. Safety outcomes
1. Opportunistic infections: Major: Pneumocystic carinii pneumonia,
cryptococcal meningitis, esophageal candidiasis, disseminated TB,
cerebral toxoplasmosis, cytomegalovirus, recurrent pneumonia.
Minor: oral thrush, herpes zoster, genital herpes.
2. Opportunistic HIV/AIDS associated malignancies: Kaposi sarcoma
3. Immune reconstitution disease:
Any 1 of the following:
1) New or worsening enlarged lymph nodes
2) New or worsening radiological features of TB
3) Breakthrough tuberculous meningitis
4) New or worsening serositis (pleural effusion, ascites,
Within 3 m of initiating HAART; evidence of good compliance and an
initial clinical response to anti-TB treatment; a N1 log decrease in
HIV RNA; no alternative explanation for the clinical deterioration.
4. Impact on measures of immunosuppression
Measured CD4 counts
Mayosi et al 5
American Heart Journal
Volume 0, Number 0
animal study in which steroids reduced the immune
response to Bacillus Calmette-Guerin (BCG) vaccination
but not to a recombinant form of the BCG.28These data
others may not. The potential for synergistic effects is
derived from studies, which indicate that both predniso-
loneand M.w may increasethe peripheralbloodCD4− T-
lymphocyte counts in patients with HIV infection, and so,
the 2 interventions may act synergistically to improve the
immune status of patients who are immunosuppressed by
HIV.20,29Therefore, the available data on the interaction
betweenadjunctive corticosteroids and M.w immunother-
apy are sparse and contradictory. The only way to resolve
this question is to use a factorial design with an adequate
sample size and power. Our trial is well powered to detect
plausible and clinically relevant differences in treatment
effects and interactions (see online Appendix for details of
sample size estimation).
Second, the use of the clinical case definition of TB may
important because without specific treatment, the mean
survival rate is 3.7 months (mortality rate 85% at 6
months).3Therefore, in TB-endemic countries, there is a
critical need to start empiric therapy in suspected TB
pericarditis cases before microbiological confirmation.23
Although some overascertainment of TB cases may occur,
especially in HIV-uninfected patients where pericardial
effusions may be due to causes other than TB, 34% of our
population is HIV negative, among whom 25% will not
have M tuberculosis in the pericardial fluid, although the
absence of M tuberculosis does not rule out that the
effusion is due to TB.30By contrast, in HIV-infected
patients, it is unlikely that overascertainment will occur
the earlypresentingfeaturesofHIV infectionandis dueto
TB in N95% of cases; 66% of participants in the IMPI trial
are HIV infected.31Therefore, we expect that N90% of
patients in IMPI will either have confirmed TB or be HIV
positive (with almost all having TB).
Patients recruited in the initial pilot study phase have
been retained in the trial, and the trial is being expanded
to enroll a total of 1,400 subjects. The pilot phase was
initiated on January 1, 2009, and the full study is expected
to be completed in June 30, 2014. As of December 30,
2011, 815 participants were recruited from 23 sites in
8 countries: 11 sites in South Africa; 6 in Nigeria; and 1
each in Kenya, Malawi, Mozambique, Sierra Leone,
Uganda, and Zimbabwe. The first 815 patients recruited
into the study are similar to those who were enrolled in
the registry in terms of age and sex (Table III). The
median age of the 815 randomized patients is 35 years
and 56% are men. There is, however, a greater proportion
of cases who are HIV seropositive (75% vs 57%). There is
also a higher proportion of cases with proven TB (25%)
compared with the patients enrolled in the registry (3%).
IMPI is the first large international multicenter clinical
trial of patients with TB pericarditis that is adequately
powered to determine effects of adjunctive immunother-
apy on major effectiveness and safety outcomes. The trial
addresses a very serious complication of TB, commonly
occurring in the young, and will create the capacity and
network to address similar important health problems in
these resource poor settings.
We are grateful to Drs J. Parswani, B. Khamar, A. Maseeh,
V. Vaishnavi, and B. Mody of Cadila Pharmaceuticals for
their generous support with part of the funding, the
of investigational products to the study sites. The assistance
and support of Ms Margaret Koufie and Ms Stephanie Hall
(at the Population Health Research Institute) and that of Ms
Lavinia Petersen, Ms Roxi Virgotine, Mr Simphiwe Nkepu,
Ms Carolina Lemmer, and other members of the Project
Coordinating Office is recognized with gratitude.
IMPI is registered with the ClinicalTrials.gov
(NCT100810849). Funding for the IMPI Trial was
provided by the Canadian Network and Centre for Trials
Internationally of the Canadian Institutes for Health
Research, the Canadian Institutes for Health Research
Infection and Immunity Operating Grant, the Population
Health Research Institute, the Medical Research Council
of South Africa, Lily and Ernst Hausmann Trust, and Cadila
Pharmaceuticals (the manufacturers of M. w). The
authors are solely responsible for the design and conduct
of this study, all study analyses, the drafting and editing of
the manuscript, and its final contents.
The authors have no conflicts of interest.
Table III. Clinical characteristics of the 815 patients enrolled in
the IMPI trial on December 30, 2011, compared with the 185
patients enrolled in the IMPI registry
(n = 815)
(n = 185)
Median age in years (range)
Proportion of tested patients
with positive HIV serology
Number with pericardiocentesis
Cases with definite TB on
microscopy or culture
6 Mayosi et al
American Heart Journal
1. Global tuberculosis control: WHO report 2011. Geneva: World
Health Organization; 2011.
2. Cegielski JP, Lallinger GJ, Ramaiya K, et al. Pericardial disease and
human immunodeficiency virus in Dar es Salaam, Tanzania. Lancet
3. Syed FF, Mayosi BM. A modern approach to tuberculous pericarditis.
Prog Cardiovasc Dis 2007;50:218-36.
4. Mayosi BM. Contemporary trends in the epidemiology and
management of cardiomyopathy and pericarditis in sub-Saharan
Africa. Heart 2007;93(10):1176-83.
5. Dwivedi SK, Rastogi P, Saran RK, et al. Antituberculous treatment
does not prevent constriction in chronic pericardial effusion of
undetermined aetiology. Indian Heart J 1997;49:411-4.
6. Reuter H, Burgess LJ, Doubell AF. Epidemiology of pericardial
effusions at a large academic hospital in South Africa. Epidemiol
7. Imazio M, Mayosi BM, Brucato A, et al. Pericardial effusion triage. Int
J Cardiol 2010;145(2):403-4.
8. Mayosi BM, Wiysonge CS, Ntsekhe M, et al. Clinical characteristics
and initial management of patients with tuberculous pericarditis in the
HIV era: the Investigation of the Management of Pericarditis in Africa
(IMPI Africa) registry. BMC Infect Dis 2006;6:2.
9. Mayosi BM, Ntsekhe M, Volmink JA, et al. Interventions for treating
tuberculous pericarditis. Cochrane Database Syst Rev 2002;4(1469-
493X (Electronic) IP - 4 DP - 2002):CD000526.
10. Mayosi BM, Wiysonge CS, Ntsekhe M, et al. Mortality in patients
treated for tuberculous pericarditis in sub-Saharan Africa. S Afr Med
11. Schwefer M, Aschenbach R, Heidemann J, et al. Constrictive
pericarditis, still a diagnostic challenge: comprehensive review of
clinical management. Eur J Cardiothorac Surg 2009;36(3):502-10.
12. Ntsekhe M, Wiysonge C, Volmink JA, et al. Adjuvant corticosteroids
for tuberculous pericarditis: promising, but not proven. QJM 2003;
13. Elliott AM, Halwiindi B, Bagshawe A, et al. Use of prednisolone in the
treatment of HIV-positive tuberculosis patients. Q J Med 1992;85:
14. Elliott AM, Luzze H, Quigley MA, et al. A randomized, double-blind,
placebo-controlled trial of the use of prednisolone as an adjunct to
treatment in HIV-1–associated pleural tuberculosis. J Infect Dis 2004;
15. Maisch B, Seferovic PM, Ristic AD, et al. Guidelines on the diagnosis
and management of pericardial diseases executive summary: The
Task Force on the Diagnosis and Management of Pericardial Diseases
of the European Society of Cardiology. Eur Heart J 2004;25:
16. Wiysonge CS, Ntsekhe M, Gumedze F, et al. Contemporary use of
adjunctive steroids in tuberculous pericarditis. Int J Cardiol 2007.
17. Mathur ML. Potential utility of Mycobacterium w vaccine in control of
tuberculosis. Curr Resp Med Rev 2006;2(2):183-8.
18. Talwar GP. An immunotherapeutic vaccine for multibacillary leprosy.
Int Rev Immunol 1999;18(3):229-49.
19. Nath I. A vaccine for leprosy. Nat Med 1998;4:548-50.
20. Kharkar R. Immune recovery in HIV with Mycobacterium W. J Indian
Med Assoc 2002;100(9):578-9.
21. Gupta S, Malhotra AK, Verma KK, et al. Intralesional immunotherapy
with killed Mycobacterium w vaccine for the treatment of ano-genital
warts: an open label pilot study. J Eur Acad Dermatol Venereol 2008;
22. Singh IG, Mukherjee R, Talwar GP. Resistance to intravenous
inoculation of Mycobacterium tuberculosis H37Rv in mice of different
inbred strains following immunization with a leprosy vaccine based
on Mycobacterium w. Vaccine 1991;9(1):10-4.
23. Mayosi BM, Burgess LJ, Doubell AF. Tuberculous pericarditis.
24. Volmink J, Matchaba P, Garner P. Directly observed therapy and
treatment adherence. Lancet 2000;355(9212):1345-50.
25. Thwaites GE, Bang ND, Dung NH, et al. Dexamethasone for the
treatment of tuberculous meningitis in adolescents and adults. N Engl
J Med 2004;351(17):1741-51.
26. Meintjes G, Lawn SD, Scano F, et al. Tuberculosis-associated
immune reconstitution inflammatory syndrome: case definitions for
use in resource-limited settings. Lancet Infect Dis 2008;8(8):
27. Fleiss JL. Multicentre clinical trials: Bradford Hill's contributions and
some subsequent developments. Stat Med 1982;1:353-9.
28. Young SL, O'Donnell MA, Buchan GS. IL-2–secreting recombinant
bacillus Calmette Guerin can overcome a type 2 immune response
and corticosteroid-induced immunosuppression to elicit a type 1
immune response. Int Immunol 2002;14(7):793-800.
29. Mayanja-Kizza H, Jones-Lopez E, Okwera A, et al. Immunoadjuvant
prednisolone therapy for HIV-associated tuberculosis: a phase 2
clinical trial in Uganda. J Infect Dis 2005;191:856-65.
30. Strang G, Latouf S, Commerford P, et al. Bedside culture to
confirm tuberculous pericarditis. Lancet 1991;338(8782–8783):
31. Reuter H, Burgess L, van Vuuren W, et al. Diagnosing tuberculous
pericarditis. QJM 2006;99:827-39.
Mayosi et al 7
American Heart Journal
Volume 0, Number 0
1. Effect of benefit of 1 medication on the margin event
Margin event rate⁎
with 30% RRR
Margin event rate⁎
with 35% RRR
RRR, Relative risk reduction.
⁎The margin event rates are derived after adjusting for half of the control group
receiving another effective treament, which is effective by 30% or 35% RRR. This will be
the observed event rate in the 700 people who will form the control group for each of
the 2 comparisons.
†Event rate observed in the IMPI registry. All other event rates below in the same
column are conservative estimates for the double placebo arm of the trial.
‡Margin event rates of 30% and 35% are considered in subsequent tables below for
our mid case calculations, and these are close to the margin event rates of 29% and
37% in the above tables.
2. Detectable RRR by n with 10% noncompliance and
6% of patients lost to follow-up (2α = .05)
Assuming that one of the treatments is effective
(RRR of 30% or 35%), this will then reduce the event
rate for the corresponding control group of the other
part of this factorial design. The detectable reductions
in hazard ratio by margin control event rate, at 80%
and 90% power, and for various sample size are
shown in the table below. These calculations assume
comparison using a Cox regression model, with a 42-
month recruitment period and the last participant
followed for 6 months (51 months in total). A total of
6% lost to follow-up by the end of trial is assumed,
but we will make every attempt to minimize this.
However, to be conservative, we have also calculated
power using a 10% rate of loss to follow-up. There is
very little change in the RRR detectable at a specific
power with a 10% rate of loss to follow-up compared
with 6%. For example, with 1400 people and a
margin event rate of 30%, the RRR detectable at 80%
power changes from 20.1%, with 6% lost to follow-
up, to 20.2% with 10% lost to follow-up. For all
estimates of RRR in the subsequent table, the impact
of a 10% rate of lost to follow-up compared with the
6% rate of loss to follow-up is minimal (usually b0.2%)
and, therefore, not shown. All comparisons are 2-
tailed test (P = .05).
n = 1200n = 1400n = 1600
Margin event rates in the range of 30% to 36% are
considered to be plausible in the control group.
With 1,400 patients, there will be 80% power to
detect a 19% to 20% RRR across a range of event
rates (30%-36%) and 90% power to detect RRR of
21% to 23%. Therefore, a trial of 1,400 patients is
well powered to detect moderate differences (note
the meta-analysis suggested a 30% to 35% RRR with
steroids). So even with a 50% subadditivity, which
will correspondto an observed
of 23% to 27% RRR (if the effects of any treatment
are 30%-35% RRR), the study of 1,400 patients will
have high power.
3. Effect of subadditivity of various magnitude (0%-
100%) on the observed effect size if the “true” RRR is
30%. This assumes that the true effect of B
observed in the absence of A is 30%
RRR of the
of A and B
given level of subadditivity
Conclusions: Subadditivity of up to 75% would
reduce a 30% RRR to 21%. A trial of 1,400 patients
will have N80% power to detect this difference. If the
subadditivity is more extreme, then the power to
detect a difference declines. For example, if subaddi-
tivity is 100% (ie, treatment B has no effect in the
presence of A or vice versa), then the observed RRR
will be only 18%, for which the comparison of 700 vs
700 patients will have only 38% power. In such cases,
greater power is obtained by comparing the 350
patients on double placebo vs 350 patients on only A
or only B (see below).
Effect of subadditivity of various magnitudes (0%-
100%) on the observed effect size if the “true” RRR
RRR of the
of A and B
Effect at margin of med B for
a given level of subadditivity
7.e1 Mayosi et al
American Heart Journal
Conclusions: Subadditivity of up to 75% would
reduce a 35% RRR to 25% for which a trial of 1,400
patients has high power. However, even with 100%
subadditivity, the apparent RRR would be 21%, which
can be detected with N80% power (see table in section
8.2 above). Nevertheless, this treatment estimate is
“biased” and so the comparison of the 350 receiving
double placebo versus a single treatment alone will
provide the best estimate of the effect of A or B (see
below for power).
4. Detectable RRR comparisons between the cell with
double placebo versus each individual drug (2-tailed
α = .0264, 10% drop-out and 6% lost to follow-up)
n = 300
n = 350
n = 400
Conclusions: Comparing the cell with double placebo
(n = 350) versus group A or B (each with 350 subjects)
indicates that these comparisons have high power to
detect differences ranging from 26% RRR to 32% RRR.
Therefore, if a statistically significant interaction oc-
curred (especially N75% subadditivity), then the most
sensitive analysis is done using individual cells, which
will also provide the least biased estimate of the
treatment effects. Please note that we have used a
more stringent P value of .0264 (which represents a
correction for multiplicity after taking the overlap of
the common control group, ie, double placebo) for the
comparison of each active intervention versus the
double placebo group.
5. Power to detect an interaction at various levels of
subadditivity with 35% event rate and 35% RRR of each
medication separately (based on Fleiss, 1982).52
Conclusions: There is high power to statistically detect
subadditivity of N50% using tests of interaction.
By enrolling 1,400 patients and using a 2 × 2 factorial
design for the IMPI trial:
1.We will havehigh power to detect the main effects of
each intervention (steroids or M. w vaccine) if the real
effect sizes are moderate (20%-25% or greater RRR), in the
absence of any major interactions.
2. If the real effect of steroids is closer to the 30% to 35%
RRR observed in the meta-analysis of small trials
evaluating steroids, this will be detected easily.
3. If the real effects are 30% to 35% of either
intervention when given alone, but there is an important
level of subadditivity (N50% or more), then the trial will
still detect this. Lesser degrees of subadditivity may not be
detected (but is clinically less important) and will not
materially reduce the power for the overall analyses. If
there is synergy, this too will be detected, but the overall
power for the comparisons at the “margins” will increase.
4. Also, such a large trial (which is 3 times larger than
the total number of participants in all previous trials of
this condition to date) will also provide the best estimate
of safety of each of the interventions given alone and in
combination. Therefore, the 2 × 2 factorial design with
1,400 patients provides the optimal power to detect both
main effects and interactions.
5. Please note that the above arguments regarding
power apply in exploring variability/consistency of
effects in key subgroups as long as they are of reasonable
size (at least 40% of the sample).
Conclusions: Therefore, our overall strategy is efficient
under a range of scenarios.
Plans to ensure high retention of participants in the
Mayosi et al 7.e2
American Heart Journal
Volume 0, Number 0
Action Download full-text
- Simple visit procedures that are not
- Regular visits during initial study phase
to ensure potential issues with
interventions are dealt with rapidly
(ie, address both medication and
- Randomization: Centers must consider
whether the participant will adhere to
visits as well as medication before
entering the study and not randomize
those that they are concerned will not
be able to comply.
- Each visit: Date of next visit is collected
to ensure that local staff and the IMPI
Study Team know when the participant
- Contact telephone/cell phone numbers
are verified by local site staff (by calling)
at each visit. If any are not working,
the center asks the participant for
- Each participant will be provided with a
cell phone and air time, if necessary.
- Local site staff must fax data as soon
as visit is completed.
- Programs are written to send the IMPI
Study Team an email notification if a visit
is missed by 1 d.
- Reports indicating missed visits are
reviewed at monthly teleconferences
between the IMPI Study Team and PHRI.
- All missed visits are discussed to
ensure appropriate action has been taken.
- If a center is concerned that a particular
participant may be at high risk for not
returning, they will contact the IMPI Study
Team to discuss possible solutions and
strategies, which include sending a social
worker or private investigator to the address
of the participant and contact persons.
- IMPI Study Team
- Local site staff
- Local site staff
- Local site staff
- IMPI Study
- IMPI Study Team
7.e3 Mayosi et al
American Heart Journal