The timing of tuberculosis after isoniazid
preventive therapy among gold miners in
South Africa: a prospective cohort study.
Sabine Hermans1,2,3, Alison Grant1,4, Violet Chihota4,5, James Lewis1,
Emilia Vynnycky1,6, Gavin Churchyard1,4,5, Katherine Fielding1,4
1TB
Centre, London School of Hygiene & Tropical Medicine, London, United Kingdom
of Global Health, Academic Medical Center, University of Amsterdam, Amsterdam Institute for Global Health
and Development, the Netherlands
3Desmond Tutu HIV Centre, Institute for Infectious Diseases and Molecular Medicine, University of Cape Town, South Africa
4The School of Public Health, University of the Witwatersrand, Johannesburg, South Africa
5Aurum Institute, Johannesburg, South Africa
6Public Health England, London, United Kingdom
2Department
Hermans et al, BMC Medicine 2016 Mar 23;14(1):45
Background
• Upsurge in tuberculosis (TB) since advent of HIV epidemic in
South Africa
• Isoniazid preventive therapy (IPT) reduces TB incidence in
both HIV-positive and negative population
• Durability of protection is limited in high-prevalence settings;
mechanism is not well understood (reactivation of persistent
latent TB or reinfection)
• Limited power of previous studies to investigate timing and
change in associated risk factors
Akolo et al, Cochrance Rev 2010; Smieja et al, Cochrane Rev 2000
Martinson et al, NEJM 2011; Churchyard et al, NEJM 2014
Rangaka et al, Lancet 2014; Samandari et al, Lancet 2011 and AIDS 2015.
The Thibela TB study
• Cluster-randomised trial to investigate impact of communitywide IPT on TB incidence and prevalence compared to
standard of care among 15 clusters of gold mines and their
workforce in South Africa
• High TB risk population; HIV prevalence estimated at 29%
• Incident TB case definition:
– positive sputum smear or culture, or
– clinical TB diagnosis assigned by a senior study clinician
Churchyard et al, NEJM 2014
Lewis et al. Am J Resp Crit Care Med. 2009
The Thibela TB study
• Cluster-randomised trial to investigate impact of communitywide IPT on TB incidence and prevalence compared to
standard of care among 15 clusters of gold mines and their
workforce in South Africa
• High TB risk population; HIV prevalence estimated at 29%
• Incident TB case definition:
– positive sputum smear or culture, or
– clinical TB diagnosis assigned by a senior study clinician
• No impact on TB incidence (IRR 1.00 [95% CI 0.75-1.34]) or
prevalence (PR 0.98 [95%CI 0.65-1.48])
Churchyard et al, NEJM 2014
Lewis et al. Am J Resp Crit Care Med. 2009
Study objectives
1. To estimate the timing of TB incidence during and after IPT
2. To investigate risk factors associated with post-IPT TB
incidence and whether these differed over time after the end
of IPT
3. To compare the observed TB incidence rate in first year after
IPT with a crude modelled estimate of TB incidence rate
attributable to reinfection during that period
Methods I
• Post-hoc analysis of IPT arm of Thibela TB study (8
intervention clusters)
• We included all participants prescribed at least one of
intended 9 months of IPT
• Analysis:
– Crude TB incidence rates during and after IPT, overall and stratified by
follow-up time after end of IPT (per 6 months)
– Rate ratio for TB after versus during IPT
– Risk factors for TB after IPT
– Effect modification of risk factors by follow-up time; <12 versus ≥12
months after stopping IPT
Methods II
• All analyses were done intention to treat (ITT) and as treated
(AT):
Intention to treat
IPT (intended period of 9 months)
Time 0 (objective 1)
As treated
IPT (actual period)
Follow-up
Time 0 (objective 2+3)
Follow-up
Time 0 (objective 1) Time 0 (objective 2+3)
• Estimates from published mathematical model of the Thibela
study data were used to calculate average TB incidence rate
in first year after IPT attributable to reinfection, and compared
to observed incidence during that period (ITT)
Vynnycky et al, Am J of Epid 2015
Patient characteristics (n=18,520)
Characteristic
Gender
Age (years)
Country of origin
Years in workforce
Type of work
Type of housing
Previous TB
Male
Mean (SD)
South Africa
Total (n [col%])
17763 (95.9)
41 (9)
10501 (56.7)
Lesotho
5178 (28.0)
Mozambique
1854 (10.0)
Other
Mean (SD)
Underground
Hostel
979 (5.3)
17 (10)
16821 (91.2)
10913 (58.9)
2212 (12.0)
TB incidence rate over time
• 708 TB cases in 37,321 person-years; 541 (76%) after IPT
(ITT analysis)
Rate during IPT
(/100 pyrs [95% CI])
Rate after IPT
(/100 pyrs [95% CI])
Rate ratio
(95% CI)
Intention to treat
1.3 (1.0-1.6)
2.3 (1.9-2.7)
1.8 (1.4-2.3)
As treated
0.7 (0.5-1.0)
2.3 (1.9-2.7)
3.0 (2.3-4.3)
TB incidence rate over time
• 708 TB cases in 37,321 person-years; 541 (76%) after IPT
(ITT analysis)
Rate during IPT
(/100 pyrs [95% CI])
Rate after IPT
(/100 pyrs [95% CI])
Rate ratio
(95% CI)
Intention to treat
1.3 (1.0-1.6)
2.3 (1.9-2.7)
1.8 (1.4-2.3)
As treated
0.7 (0.5-1.0)
2.3 (1.9-2.7)
3.0 (2.3-4.3)
Intention to treat
As treated
Risk factors for TB after IPT
Total
Gender
Age (years)
Country of origin
Type of work
Type of housing
Previous TB
Number of
IPT prescriptions
aHR (95% CI)
Male
1
Female
1.09 (0.67-1.80)
<=29
1
30-39
2.18 (1.46-3.25)
40-49
2.84 (1.92-4.18)
50+
2.77 (1.81-4.23)
South Africa 1
Lesotho
1.15 (0.95-1.40)
Mozambique 0.79 (0.57-1.11)
Other
0.82 (0.54-1.24)
Surface
1
Underground 1.75 (1.20-2.54)
Hostel
1
Other
0.74 (0.61-0.91)
No
1
Yes
1.89 (1.52-2.35)
1-2
1
3-5
0.81 (0.59-1.09)
6+
0.63 (0.51-0.77)
P-value
0.73
<0.001
0.09
0.002
0.004
<0.001
<0.001
No evidence for
effect modification
by follow-up time
(<12 versus ≥12
months after IPT)
Risk factors for TB after IPT
Total
Gender
Age (years)
Country of origin
Type of work
Type of housing
Previous TB
Number of
IPT prescriptions
aHR (95% CI)
Male
1
Female
1.09 (0.67-1.80)
<=29
1
30-39
2.18 (1.46-3.25)
40-49
2.84 (1.92-4.18)
50+
2.77 (1.81-4.23)
South Africa 1
Lesotho
1.15 (0.95-1.40)
Mozambique 0.79 (0.57-1.11)
Other
0.82 (0.54-1.24)
Surface
1
Underground 1.75 (1.20-2.54)
Hostel
1
Other
0.74 (0.61-0.91)
No
1
Yes
1.89 (1.52-2.35)
1-2
1
3-5
0.81 (0.59-1.09)
6+
0.63 (0.51-0.77)
P-value
0.73
<0.001
0.09
0.002
0.004
<0.001
<0.001
No evidence for
effect modification
by follow-up time
(<12 versus ≥12
months after IPT)
Modelled versus observed rate
In first year after IPT:
• Modelled estimate of average crude TB incidence rate
attributable to reinfection: 1.3/100 pyrs
• Observed TB incidence rate: 2.2/100 pyrs
Conclusions & Interpretation
• Durability of protection by IPT was lost within 6-12 months in
this setting of high HIV prevalence and high annual risk of M.
tuberculosis infection
• No evidence for different risk factors in first year compared to
later years after IPT
• Observed rate was higher than modelled rate, suggesting a
role for reactivation of persistent latent infection in rapid return
to baseline TB incidence
• Composite of both mechanisms: reactivation disease
gradually replaced by reinfection disease, eventually reaching
steady state
• Rifamycins have higher sterilizing potential with possible
longer duration of protection – need for further studies
Acknowledgements
The Thibela TB Study team and participants
Amsterdam Institute for Global Health and Development
Joep Lange
Frank Cobelens
Funders
Bill & Melinda Gates Foundation
Swiss National Science Foundation
European Commission
UK Department of Health
MRC UK
DFID
Assumptions in mathematical model
•
•
•
•
•
•
•
•
Annual risk of infection 20%
Reduction in annual risk of infection after intervention 11-20%
HIV prevalence 30%
Proportion with CD4 count <200/ul 25%, all on ART
Initial pre-treatment loss to follow-up of 40%
Average time to detection 1 year
Cluster-specific silicosis prevalence
Reinfection could take place during IPT, but no progression to
disease
Vynnycky et al, Am J of Epid 2015
Sensitivity analysis: optimal adherence
(>= 6 months of IPT)