Research

Research

The impact of the roll-out of rapid molecular diagnostic testing for

tuberculosis on empirical treatment in Cape Town, South Africa
SabineHermans,a JudyCaldwell,b RichardKaplan,a FrankCobelensc & RobinWooda

Objective To investigate the impact of introducing a rapid test as the first-line diagnostic test for drug-sensitive tuberculosis in Cape Town,
South Africa.
Methods Xpert MTB/RIF (Xpert), an automated polymerase-chain-reaction-based assay, was rolled out between 2011 and 2013. Data
were available on 102007 adults treated for pulmonary tuberculosis between 2010 and 2014. Tuberculosis notification rates per 100000
population were calculated for each calendar year and for each year relative to the test roll-out locally, overall and by bacteriological
confirmation. Empirical treatment was defined as treatment given without bacteriological confirmation by Xpert, sputum smear microscopy
or sputum culture.
Findings Between 2010 and 2014, the proportion of human immunodeficiency virus (HIV)-negative patients treated empirically for
tuberculosis declined from 23% (2445/10643) to 11% (1149/10089); in HIV-positive patients, it declined from 42% (4229/9985) to 27%
(2364/8823). The overall tuberculosis notification rate decreased by 12% and 19% among HIV-negative and HIV-positive patients, respectively;
the rate of bacteriologically confirmed cases increased by 1% and 3%, respectively; and the rate of empirical treatment decreased by 56%
and 49%, respectively. These changes occurred gradually following the tests introduction and stabilized after 3years.
Conclusion Roll-out of the rapid test in a setting with a high prevalence of pulmonary tuberculosis and HIV infection was associated
with a halving of empirical treatment that occurred gradually after the tests introduction, possibly reflecting the time needed for full
implementation. More than a quarter of HIV-positive patients with tuberculosis were still treated empirically, highlighting the diagnostic
challenge in these patients.

and mortality.79 However, the four large randomized trials

Introduction published to date failed to document these reductions.1013
Sputum smear microscopy is traditionally the first-line di- This failure may have been due to empirical treatment being
agnostic test for tuberculosis in countries without routine replaced by bacteriologically confirmed treatment rather than
access to the gold standard: sputum culture. This approach to more patients being identified.14 Subsequently, when one of
is limited by low sensitivity, particularly among patients who the original modelling papers was modified to align its results
test positive for the human immunodeficiency virus (HIV), with one of the trials, the predicted decline in tuberculosis
and is associated with diagnostic delays, underdiagnosis incidence decreased from 6% to 1.6%.10,15 Modelling the effect
and empirical treatment.1 The Xpert MTB/RIF (Xpert) test of the test in India produced similar results.16
(Cepheid, Sunnyvale, United States of America) is an auto- There is a need for more data on the impact of the Xpert
mated, cartridge-based, rapid molecular diagnostic test for test in practice. In South Africa, roll-out of this test as the
Mycobacterium tuberculosis and its resistance to rifampicin.2 primary test in a new tuberculosis diagnostic algorithm started
The test detects the rpoB gene of M. tuberculosis, including in March 2011 it was completed in Cape Town in February
mutations that encode rifampicin resistance, using a real- 2013.17 In this study, we evaluated the impact of the test roll-out
time polymerase chain reaction and takes less than 2hours. in Cape Town on the diagnosis of patients with drug-sensitive
Because the test has higher sensitivity than smear microscopy tuberculosis, stratified by HIV status. We also analysed as-
(88% versus 65%, respectively),3 is rapid and has the ability sociated changes in the proportion of notified tuberculosis
to detect rifampicin resistance immediately, the World Health cases that were confirmed bacteriologically and examined
Organization (WHO) endorsed its use in resource-constrained risk factors for empirical treatment. Finally, we determined
settings in December 2010.4 By the end of 2014, concessionary whether the changes observed increased with time following
pricing had led to widespread roll-out of the test in lower- the introduction of the test.
income countries.5
In 2013, WHO identified understanding the impact of the
Xpert test on individual and public health outcomes as one of
Methods
the top 10 research areas in tuberculosis.6 Modelling studies The estimated population of Cape Town in 2011 was 3.7mil-
indicated the test would increase tuberculosis case-finding and lion.18 The diagnosis and treatment of tuberculosis in the
that the resulting earlier treatment would improve outcomes, city was provided by 101 government primary-care clinics;
leading eventually to reductions in tuberculosis incidence tuberculosis treatment in private clinics was infrequent.19 The

a
The Desmond Tutu HIV Centre, Institute for Infectious Diseases and Molecular Medicine, University of Cape Town Faculty of Health Sciences, Anzio Road, Observatory,
Cape Town, 7925, South Africa.
b
City of Cape Town Health Directorate, Cape Town, South Africa.
c
Amsterdam Institute for Global Health and Development, Amsterdam, Netherlands.
Correspondence to Sabine Hermans (email: s.hermans@aighd.org).
(Submitted: 19 September 2016 Revised version received: 1 February 2017 Accepted: 20 March 2017 Published online: 28 April 2017)

554 Bull World Health Organ 2017;95:554563 | doi: http://dx.doi.org/10.2471/BLT.16.185314

 Research
Sabine Hermans et al. Rapid diagnostic test roll-out, South Africa

diagnosis of pulmonary tuberculosis was

Fig. 1. Calendar year and year relative to rapid diagnostic test roll-out, Cape Town,
generally based on sputum smear mi- South Africa, 20102014
croscopy, with sputum culture reserved
for patients who remained symptomatic
despite negative microscopy findings Year relative to rapid molecular diagnostic test roll-out
or who were being retreated. At all First clinic where test Year 0 Year 1 Year 2 Year 3 Year 4
clinics, chest X-ray facilities were avail- was introduced
able, either on-site or through referral.
Although the empirical treatment of tu- Last clinic where test Year 0 Year 1 Year 2 Year 3 Year 4
berculosis based on symptoms and chest was introduced
X-ray findings alone was discouraged, rapid
molecular
it was an accepted practice for patients diagnostic
who remained symptomatic despite test
negative microbiological findings. roll-out
The Xpert test machines were in-
2010 2011 2012 2013 2014
stalled in all laboratories in Cape Town
Calendar year
between August 2011 and February
2013 and use of the test as the primary Notes: Patients who initiated tuberculosis treatment in the year before the Xpert MTB/RIF test (Cepheid,
diagnostic test in the tuberculosis di- Sunnyvale, United States of America) roll-out date for their clinic were assigned to year0, those who
agnostic algorithm for Western Cape started treatment in the first year after roll-out for their clinic were assigned to year1 and so on up to
year4. The Xpert MTB/RIF test was introduced gradually as the first-line diagnostic test at 101 clinics
Province was endorsed in a circular to between August 2011 and February 2013.
all primary health-care clinics in January
2013.17 Two sputum samples were col-
were entered into a separate register and assigned to year4; other patients were
lected and submitted simultaneously to
were not included in our evaluation. assigned to intermediate years accord-
a laboratory one was for the rapid test.
We defined the primary method ingly (Fig.1). The patient was defined as
If a rifampicin-sensitive M. tuberculosis
of diagnosis as either: (i)the rapid HIV-positive at the time of tuberculosis
strain was detected, the second sample
test; (ii)sputum smear microscopy; or treatment if the tuberculosis register
was used to determine the sputum
(iii)sputum culture if more than one recorded: (i)a positive HIV serological
smear status pretreatment and, thereby,
test was positive, the primary method test result; (ii)treatment with antiret-
helps identify smear conversion dur-
was the first test in this sequence. We roviral therapy or co-trimoxazole; or
ing follow-up. If a rifampicin-resistant
defined bacteriological confirmation (iii)a CD4+ T-lymphocyte (CD4+ cell)
strain was detected, the second sample
of infection as a positive result to one count. The patient was HIV-negative if
was used for culture and for testing
of these tests. Empirical treatment was a negative HIV serological test result
drug sensitivity. If the first sample from
defined as treatment given when no test had been recorded. All other patients
an HIV-infected patient tested nega-
was positive or no test was performed. In were regarded as having an unknown
tive, the second was used for culture. If
addition, patients who tested positive on HIV status.
the first sample from an HIV-negative
sputum culture and negative on, or did
patient tested negative, the second was Statistical analysis
not undergo, other tests were regarded
discarded because of the tests higher
as having started empirical treatment if We used descriptive statistics to pres-
sensitivity in these patients. When the
their sputum sample was sent for culture ent data on patients demographic and
test was introduced, the treatment regi-
after treatment initiation or up to 6days clinical characteristics in each calendar
men for previously treated patients was
before initiation (assuming that 7days year and on the primary method of
changed to that for new patients because
was the minimum time required for a diagnosis in each calendar year and in
rifampicin resistance could then be
positive sputum culture result).20 Treat- each year relative to the test roll-out. We
identified before treatment.17
ment followed national guidelines and calculated annual population disease
Our population-based study cov-
did not differ by method of diagnosis.21 rates by dividing the total number of
ered 2010 to 2014: 2010 was the last full
Two definitions of time were used: the bacteriologically confirmed and em-
calendar year before the test roll-out
calendar year and the year relative to pirically treated pulmonary tuberculosis
began in Cape Town and 2014 was the
the time when the test was introduced patients aged 15years or older in a year
first full calendar year after roll-out had
in the diagnosing clinic. For the latter by the mid-year estimate of the adult
been completed. We included all pulmo-
year, we identified the date on which population in the study area.22 Rates
nary tuberculosis patients aged 15years
the first test result was recorded for each were also stratified by bacteriological
or older who started treatment during
clinic and used this date as the roll-out confirmation and HIV status. The size
that period. To avoid duplication, we
date for that clinic. Patients were then of the HIV-negative and HIV-positive
excluded patients transferred between
assigned to a year depending on when adult population was derived using an-
subdistricts. We used anonymized data
their treatment started relative to the test nual HIV prevalence estimates from the
from the City of Cape Town electronic
roll-out in their clinic: patients who ini- Actuarial Society of South Africa West-
tuberculosis register on the patients
tiated treatment in the 365days before ern Cape AIDS demographic model.23
characteristics, microbiological test
the roll-out date were assigned to year0 We also calculated population disease
results, chest X-ray results, treatment
and those who started treatment 1096 rates relative to the year of test roll-out,
initiation dates and treatment outcomes.
to 1460days after the roll-out date were overall and stratified by bacteriological
Patients with drug-resistant tuberculosis

Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314 555

 Research
Rapid diagnostic test roll-out, South Africa Sabine Hermans et al.

confirmation. We estimated the popu- Table 1. Patients characteristics, impact of the rapid diagnostic test roll-out on
lation size for each year relative to test tuberculosis diagnosis, Cape Town, South Africa, 20102014
roll-out as follows, taking year2 as an
example: we calculated the proportion Characteristic Number of patients (%)a
of all patients in each calendar year who
were in year2 of roll-out (Fig.1) and Calendar year
multiplied this proportion by the esti- 2010 2011 2012 2013 2014
mated population for the corresponding
All 21255 20828 20657 20093 19174
calendar year. We then summed these
Female 9602 (45) 9214 (44) 8883 (43) 8621 (43) 8023 (42)
estimates for all calendar years, which
gave us the total estimated population Age in years
for year2 of roll-out. 1524 3991 (19) 3728 (18) 3654 (18) 3487 (17) 3287 (17)
Factors associated with empirical 2534 7099 (33) 6756 (32) 6574 (32) 6472 (32) 6046 (32)
tuberculosis treatment, both overall and 3544 5271 (25) 5461 (26) 5425 (26) 5315 (27) 4991 (26)
stratified by HIV status, were identified 4554 3202 (15) 3171 (15) 3236 (16) 3053 (15) 3130 (16)
by multivariable logistic regression 5564 1222 (6) 1235 (6) 1286 (6) 1249 (6) 1233 (6)
analysis. A priori risk factors included 65 470 (2) 477 (2) 482 (2) 517 (3) 487 (3)
age, sex, HIV status, CD4+ cell count at Previously treated 6626 (31) 6588 (32) 6714 (33) 5738 (29) 4864 (25)
the start of tuberculosis treatment, his- for tuberculosis
tory of tuberculosis treatment, calendar HIV-positiveb 9985 (48) 9922 (49) 9650 (48) 9162 (46) 8823 (47)
year and year relative to test roll-out. CD4+ 167 181 185 179 173
Age, calendar year and year relative T-lymphocyte (77296) (83316) (80329) (75330) (73339)
to test roll-out were included as either count per mm3,
continuous or categorical variables median (IQR)c
based on the results of tests for departure HIV: human immunodeficiency virus; IQR: interquartile range; SD: standard deviation.
from linearity. Because of the collinear-
a
All values in the table represent absolute numbers and percentages unless otherwise stated.
b
The number of patients whose HIV status was unknown was 627 in 2010, 413 in 2011, 412 in 2012, 272 in
ity between our two-time variables, we 2013 and 262 in 2014.
used two separate multivariable logistic c
The number of patients whose CD4+ T-lymphocyte count was unknown was 358 in 2010, 259 in 2011,
regression models one included the 310 in 2012, 287 in 2013 and 584 in 2014.
calendar year and the other included Note: The roll-out of the automated Xpert MTB/RIF test (Cepheid, Sunnyvale, United States of America)
the year relative to test roll-out. We ac- occurred between August 2011 and February 2013.
counted for clustering at the clinic level
by calculating robust standard errors. reported in 98% (100021/102007) of the main reason was a positive sputum
In addition, a sensitivity analysis was the patients. The only characteristic that smear result in 67% (7100/10643)
performed using random effects models changed substantially over time was the of HIV-negative patients and in 41%
that adjusted for clustering at the clinic proportion of patients previously treated (4082/9985) of HIV-positive patients;
level. We tested for changes over time for tuberculosis, which was lower in later in 2014, the main reason was a positive
in the odds of empirical treatment in years. By March 2012, 62% (63/101) of Xpert test result in 84% (8431/10089)
years2, 3 and 4 of test roll-out using a tuberculosis clinics had access to the and 67% (5947/8823) of these patient
model that included only those years. test. There was no difference in patients groups, respectively (Table2). Between
All analyses were performed using demographic or clinical characteristics 2010 and 2014, the proportion treated
Stata/IC version 13.0 (StataCorp. LP, by either year relative to test roll-out empirically decreased by 12percentage
College Station, USA) and Excel 2013 or calendar year. Details of the patients points among HIV-negative patients and
(Microsoft Corporation, Redmond, covered by rapid testing in each year and by 15percentage points among HIV-
USA). The study was approved by the their characteristics are available from positive patients. After excluding those
Human Research Ethics Committee at the corresponding author on request. for whom a positive sputum culture
the University of Cape Town and by the The pattern of microbiological result became available after treatment
City of Cape Town Health Department. testing in patients with pulmonary initiation, the decrease in empirical
tuberculosis changed during the test treatment was 8percentage points in
roll-out: use of this test increased to both groups: among HIV-negative pa-
Results 88% (16892/19174) of patients in 2014, tients, the proportion decreased from
In 2010, 21255 patients with pulmo- with 14551 of the 16892 (86%) testing 19% (2009/10643) in 2010 to 11%
nary tuberculosis aged 15years or positive. Correspondingly, utilization of (1115/10089) in 2014; and, among HIV-
older were treated in Cape Town. The sputum smear microscopy and sputum positive patients, it decreased from 34%
number declined annually to 19174 in culture decreased. Fig.2 shows that use (3440/9985) to 26% (2293/8823). The
2014, the year after the test roll-out was of the test stabilized after the first 2years proportion of patients with pulmonary
completed. Table1 shows the patients of roll-out. Findings were similar among tuberculosis diagnosed using the Xpert
demographic and clinical characteris- HIV-negative and HIV-positive patients test increased continuously during roll-
tics in each calendar year: their mean (details available from the correspond- out up to year3 and stabilized thereafter
age was 35years, 57% (57664/102007) ing author on request). (Table3). The principal change underly-
were male and 48% (47542/100021) The reasons for starting tuberculo- ing the decrease in empirical treatment
were HIV-positive. The HIV status was sis treatment changed over time: in 2010, during the study period was that fewer

556 Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314

 Research
Sabine Hermans et al. Rapid diagnostic test roll-out, South Africa

Fig. 2. Diagnostic tests for pulmonary tuberculosis, by year relative to rapid diagnostic time definitions (details available from
test roll-out, Cape Town, South Africa, 20102014 the corresponding author on request).
The tuberculosis notification rate
in the adult population decreased over
100 the 5-year study period: by 12% among
90 HIV-negative individuals and by 19%
80 among HIV-positive individuals (Fig.3
70 and Fig.4, respectively). The rate of
60 bacteriologically confirmed tuberculosis
50 increased by 1% and 3% in these two
40 groups, respectively, and the rate of
30 empirical treatment decreased by 56%
20 and 49%, respectively. A slightly differ-
10 ent pattern was seen when the data were
0 analysed by year relative to test roll-out:
Year 0 Year 1 Year 2 Year 3 Year 4 the rate of bacteriologically confirmed
Year relative to rapid molecular diagnostic test roll-out tuberculosis increased between year0
Rapid molecular test and smear test Rapid molecular test, no smear test and year4 by 7% in HIV-negative in-
Smear test, no rapid molecular test No smear test, no rapid molecular test dividuals (Fig.5) and by 17% in HIV-
Notes: Patients who initiated tuberculosis treatment in the year before the Xpert MTB/RIF test (Cepheid, positive individuals (Fig.6), and the rate
Sunnyvale, United States of America) roll-out date for their clinic were assigned to year0, those who of empirical treatment decreased by 47%
started treatment in the first year after roll-out for their clinic were assigned to year1 and so on up and 37%, respectively. These changes
to year4 (Fig.1). The proportion of patients for whom only a sputum culture result was available was stabilized after year3.
negligible.
Multivariable logistic regression
analysis showed that the odds of em-
Table 2. Reason for tuberculosis treatment before, during and after the roll-out of a pirical tuberculosis treatment were
rapid diagnostic test for tuberculosis, Cape Town, South Africa, 20102014 2.75-fold higher in HIV-positive than
HIV-negative patients (adjusted odds
Reason for starting Number of patients (%) ratio: 2.75; 95% confidence interval:
tuberculosis treatment Calendar year 2.552.98). Other factors associated
with empirical treatment were all pa-
2010 2011 2012 2013 2014 tients older than 45years and female
HIV-negative sex in HIV-infected patients. After
patients adjusting for these factors, the odds of
Totala 10643 10493 10595 10659 10089 empirical treatment decreased with time
Positive rapid test 0 631 (6) 3954 (37) 7975 (75) 8431 (84) relative to test roll-out (Table4). There
result was no evidence to support a further
Positive sputum smear 7100 (67) 6623 (63) 4040 (38) 1012 (9) 452 (4) reduction in odds between years2, 3
Positive sputum culture 1098 (10) 994 (9) 514 (5) 144 (1) 57 (1) and 4 (P=0.22). When the analysis was
Empirical treatmentb 2445 (23) 2245 (21) 2087 (20) 1528 (14) 1149 (11) performed separately in HIV-positive
HIV-positive patients and HIV-negative patients, the same risk
Totala 9985 9922 9650 9162 8823 factors were identified (Table4). Among
Positive rapid test 1 (0) 400 (4) 2906 (30) 5386 (59) 5947 (67) HIV-positive patients, every 50-cells/
result mm3 increase in CD4+ cell count at tu-
Positive sputum smear 4082 (41) 3693 (37) 2290 (24) 658 (7) 248 (3) berculosis diagnosis was associated with
Positive sputum culture 1673 (17) 1643 (17) 832 (9) 320 (3) 264 (3) 4% lower odds of empirical treatment.
These results were found to be robust
Empirical treatmentb 4229 (42) 4186 (42) 3622 (38) 2798 (31) 2364 (27)
in the sensitivity analysis performed
HIV: human immunodeficiency virus.
using a random effects model (details
a
The number of patients whose HIV status was unknown was 627 in 2010, 413 in 2011, 412 in 2012, 272 in
2013 and 262 in 2014. available from the corresponding author
b
Treatment was empirical when no test gave a positive result or no test was performed. on request).
Note: The roll-out of the automated Xpert MTB/RIF test (Cepheid, Sunnyvale, United States of America)
occurred between August 2011 and February 2013.
Discussion
patients with a negative smear result trast, the proportion treated despite a We found that the introduction of the
were treated: among HIV-negative negative Xpert test result did not change Xpert test as the first-line diagnostic
patients, 8% (800/10643) were treated substantially over time relative to test test for tuberculosis in a large population
despite a negative smear result in 2010 roll-out and the proportion treated be- cohort led to this test becoming the pri-
compared with 1% (36/10089) in 2014; cause of abnormal chest X-ray findings mary method of diagnosis in three quar-
among HIV-positive patients, the corre- alone decreased slightly among HIV- ters of adults treated for drug-sensitive
sponding figures were 15% (1544/9985) negative patients but did not change pulmonary tuberculosis. In addition, the
and 1% (84/8823), respectively. In con- among HIV-positive patients, by both rate of bacteriologically confirmed dis-

Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314 557

 Research
Rapid diagnostic test roll-out, South Africa Sabine Hermans et al.

ease increased following the introduc-

Fig. 3. Tuberculosis notification rates in HIV-negative patients, by calendar year, Cape
Town, South Africa, 20102014 tion of the test and the rate of empirical
treatment decreased, resulting in a net
decline in the total notification rate for
500 pulmonary tuberculosis. These changes
occurred cumulatively with test roll-out
Number of patients with pulmonary tuberculosis

450
and stabilized after 3years.
per 100 000 HIV-negative individuals

400 Few evaluations of the impact of

350 routine Xpert testing in programmatic
settings have been published and most
300 documented difficulties with roll-out
250 and implementation. 2429 Increased
proportions of bacteriological confir-
200 mation and less empirical treatment
150 were reported in Nepal and India but
an increase in the case notification rate
100 was reported only in India.30,31 More-
50 over, the four large randomized clinical
trials performed to date all reported
0 increased proportions of bacteriological
2010 2011 2012 2013 2014
confirmation but only one, performed
Calendar year
in Cape Town,11 found that the number
All patients Patients with bacteriologically confirmed tuberculosis
of patients diagnosed with tuberculosis
Patients who received empirical treatment
increased.1013
HIV: human immunodeficiency virus. Our data from a programmatic set-
Note: Treatment was empirical when no test gave a positive result or no test was performed. ting in Cape Town are consistent with
previous findings: routine use of the
rapid test did not lead to an increase in
the tuberculosis notification rate but was
temporally associated with an increased
bacteriological confirmation rate and
a decrease in empirical treatment. The
Table 3. Reason for tuberculosis treatment, by year relative to the rapid diagnostic test net effect was an apparent decline in
roll-out, Cape Town, South Africa, 20102014 the total tuberculosis notification rate.
We previously reported that notifica-
Reason for starting Number of patients (%) tion rates have decreased since 2010 in
tuberculosis treatment Year relative to rapid test roll-outa both HIV-negative and HIV-positive
individuals.32 There are two potential,
0 1 2 3 4 complementary explanations: (i)the
HIV-negative incidence of tuberculosis decreased
patients (assuming access to diagnosis did not
Totalb 10553 10827 10610 7207 918 change); and (ii)empirical treatment
Positive rapid test 0 6045 (56) 8141 (77) 5976 (83) 766 (83) decreased. It was not possible to separate
result the contributions of these factors using
Positive sputum smear 7189 (68) 2532 (23) 888 (8) 323 (4) 29 (3) our data. However, the observation that
Positive sputum culture 1002 (9) 372 (3) 124 (1) 37 (1) 8 (1) the rate of bacteriologically confirmed
Empirical treatmentc 2362 (22) 1878 (17) 1457 (14) 871 (12) 115 (13) tuberculosis remained stable with the
HIV-positive patients increasing use of a more sensitive test
Totalb 9848 9821 9138 6463 936 suggests that the incidence of tubercu-
Positive rapid test 0 4177 (43) 5494 (60) 4297 (66) 630 (67) losis may have decreased. A possible
result underlying mechanism could be greater
Positive sputum smear 4036 (41) 1619 (16) 617 (7) 165 (3) 18 (2) use of antiretroviral therapy in the
Positive sputum culture 1636 (17) 726 (7) 310 (3) 165 (3) 22 (2) HIV-infected population coverage in-
Empirical treatmentc 4176 (42) 3299 (34) 2717 (30) 1836 (28) 266 (28) creased from 0% in 2004 to 63% in 2013
in Cape Town.32 However, the decline
HIV: human immunodeficiency virus.
a
Patients who initiated tuberculosis treatment in the year before the Xpert MTB/RIF test roll-out date for in tuberculosis notification rates we
their clinic were assigned to year0, those who started treatment in the first year after roll-out for their observed was not affected by HIV status,
clinic were assigned to year1 and so on up to year4 (Fig.1). which does not support this explana-
b
The number of patients whose HIV status was unknown was 627 in 2010, 413 in 2011, 412 in 2012, 272 in tion. In contrast, the possibility that
2013 and 262 in 2014.
the empirical treatment rate decreased
c
Treatment was empirical when no test gave a positive result or no test was performed.
Note: The roll-out of the automated Xpert MTB/RIF test (Cepheid, Sunnyvale, United States of America) with the test roll-out is supported by our
occurred between August 2011 and February 2013. observation that the decline in this rate

558 Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314

 Research
Sabine Hermans et al. Rapid diagnostic test roll-out, South Africa

Fig. 4. Tuberculosis notification rates in HIV-positive patients, by calendar year, Cape probably attributable to another factor,
Town, South Africa, 20102014 such as declining incidence.12
The main limitation of our study
was its inability to determine whether
7000 the decline in empirical treatment
represented a decline in false-positive
Number of patients with pulmonary tuberculosis

6000 or true-positive diagnoses. The latter

per 100 000 HIV-positive individuals

could have occurred because clinicians

5000 overestimated the negative predictive
value of the rapid test. Interestingly, the
4000 proportional decline in empirical treat-
ment was smaller among HIV-infected
3000 patients, in whom the test is less sensi-
tive.3 Clinicians may have been reluctant
to miss active tuberculosis disease in
2000
this vulnerable population, thereby
increasing the number of false-positive
1000
diagnoses. Our lack of data on presump-
tive tuberculosis patients precluded an
0 evaluation of whether roll-out of the test
2010 2011 2012 2013 2014
led to the identification and treatment
Calendar year
of patients who would otherwise not
All patients Patients with bacteriologically confirmed tuberculosis
have been treated. Moreover, we were
Patients who received empirical treatment
not able to investigate the impact of
HIV: human immunodeficiency virus. symptomatology on the likelihood of
Note: Treatment was empirical when no test gave a positive result or no test was performed. empirical treatment (no data) or of the
time to treatment initiation (incomplete
Fig. 5. Tuberculosis notification rates in HIV-negative patients, by year relative to rapid recording of sputum collection dates).
diagnostic test roll-out, Cape Town, South Africa, 20102014 The apparent decrease in previously
treated patients was probably due to
misclassification following abolition of
500 the distinct retreatment regimen.
Number of patients with pulmonary tuberculosis

450 In conclusion, routine use of the

Xpert test in a setting with a high preva-
per 100 000 HIV-negative individuals

400
lence of tuberculosis and HIV infection
350 was associated with a halving of the
empirical treatment rate. This reduc-
300
tion occurred gradually following the
250 introduction of the test, probably due to
the time needed for full implementation
200
of a new diagnostic algorithm. More
150 than a quarter of HIV-infected patients
were still treated empirically, which
100
highlights the difficulty of diagnosing
50 tuberculosis in this group.
0
Year 0 Year 1 Year 2 Year 3 Year 4 Acknowledgements
Year relative to rapid molecular diagnostic test roll-out We thank Carl Morrow, Desmond Tutu
All patients Patients with bacteriologically confirmed tuberculosis
HIV Centre, Institute for Infectious
Patients who received empirical treatment
Diseases and Molecular Medicine, Uni-
versity of Cape Town, and Gareth Bow-
HIV: human immunodeficiency virus. ers, Cape Town. Sabine Hermans is also
Notes: Patients who initiated tuberculosis treatment in the year before the Xpert MTB/RIF test (Cepheid,
Sunnyvale, United States of America) roll-out date for their clinic were assigned to year0, those who
affiliated with the Amsterdam Institute
started treatment in the first year after roll-out for their clinic were assigned to year1 and so on up to for Global Health and Development and
year4 (Fig.1). Treatment was empirical when no test gave a positive result or no test was performed. the Department of Internal Medicine,
Makerere University College of Health
slowed gradually during roll-out and were mainly attributable to a reduction Sciences, Kampala, Uganda. Frank Co-
then stabilized. This may reflect the time in empirical treatment, we would ex- belens is also affiliated with the KNCV
needed to fully implement the new test pect the decline to stabilize within 3 to Tuberculosis Foundation in The Hague.
and, possibly, to apply the new diagnos- 4years of the test being introduced at Robin Wood is also affiliated with the
tic algorithm. If the decrease in the tu- the last clinic (i.e. by the end of 2017). Department of Medicine, University
berculosis notification rate we observed If the decline continues thereafter, it is of Cape Town and the Department of

Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314 559

 Research
Rapid diagnostic test roll-out, South Africa Sabine Hermans et al.

Clinical Research, Faculty of Infectious

Fig. 6. Tuberculosis notification rates in HIV-positive patients, by year relative to rapid
diagnostic test roll-out, Cape Town, South Africa, 20102014 and Tropical Diseases, London School of
Hygiene & Tropical Medicine.

7000 Funding: This work was supported by the

European Union (Marie Curie Interna-
Number of patients with pulmonary tuberculosis

6000 tional Outgoing Fellowship for Career

per 100 000 HIV-positive individuals

Development PIOF-GA-2012332311 to
5000 SH), the South African Medical Research
Council (MRC-RFAUFSP-012013/
4000 CCAMP to RW), the National Insti-
tutes of Health (R01AI058736 and
3000 R01AI093269 to RW) and the Bill &
Melinda Gates Foundation (OPP1116641
2000
to RW).

Competing interests: None declared.

1000

0
Year 0 Year 1 Year 2 Year 3 Year 4
Year relative to rapid molecular diagnostic test roll-out
All patients Patients with bacteriologically confirmed tuberculosis
Patients who received empirical treatment

HIV: human immunodeficiency virus.

Notes: Patients who initiated tuberculosis treatment in the year before the Xpert MTB/RIF test (Cepheid,
Sunnyvale, United States of America) roll-out date for their clinic were assigned to year0, those who
started treatment in the first year after roll-out for their clinic were assigned to year1 and so on up to
year4 (Fig.1). Treatment was empirical when no test gave a positive result or no test was performed.

) 10,089/1149(
42 )HIV( .
.)8823/2364( 27 ) 9985/4229(
(Xpert MTB/RIF
19 12
)Xpert
.2013 2011
102,007
3 1 .2014 2010
. 49 56 100,000
. 3

.

) HIV( Xpert
.
. 2014 2010

) HIV(
. 11 ) 10,643/2445( 23

100000
Xpert MTB/RIF (Xpert) Xpert
2011 2013
2010 2014 2010 2014
102007 (HIV)-

560 Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314

 Research
Sabine Hermans et al. Rapid diagnostic test roll-out, South Africa

Table 4. Risk factors associated with empirical tuberculosis treatment, Cape Town, South Africa, 20102014

Risk factor Risk of empirical treatment,a by multivariable logistic regression analysis

HIV-negative patients HIV-positive patients All patients
aOR (95% CI) P-value aOR (95% CI) P-value aOR (95% CI) P-value
Sex 0.92 <0.001 0.15
Female Reference Reference Reference
Male 1.00 (0.921.08) 0.92 (0.880.97) 0.97 (0.931.01)
Age in years <0.001 <0.001 <0.001
1524 Reference Reference Reference
2534 0.93 (0.831.04) 1.13 (1.041.22) 1.04 (0.961.12)
3544 1.02 (0.921.14) 1.22 (1.121.34) 1.12 (1.031.22)
4554 1.32 (1.181.47) 1.37 (1.221.54) 1.31 (1.191.44)
5564 1.77 (1.512.07) 1.44 (1.231.68) 1.64 (1.451.87)
65 2.91 (2.563.31) 2.33 (1.653.29) 2.89 (2.553.27)
CD4+ T-lymphocyte N/A N/A 0.96 (0.960.97) <0.001 N/A NA
count, per 50-cells/mm3
increase
Prior tuberculosis 0.54 <0.001 0.02
treatment
No Reference Reference Reference
Yes 0.97 (0.871.07) 1.20 (1.111.30) 1.09 (1.021.18)
Year relative to rapid <0.001c <0.001c <0.001c
test roll-outb
0 Reference Reference Reference
1 0.73 (0.640.82) 0.69 (0.630.77) 0.70 (0.640.77)
2 0.55 (0.480.61) 0.58 (0.510.65) 0.56 (0.510.62)
3 0.47 (0.410.54) 0.55 (0.470.64) 0.51 (0.450.58)
4 0.48 (0.320.71) 0.54 (0.410.72) 0.52 (0.390.68)
aOR: adjusted odds ratio; CI: confidence interval; HIV: human immunodeficiency virus; N/A: not applicable.
a
Treatment was empirical when no test gave a positive result or no test was performed.
b
Patients who initiated tuberculosis treatment in the year before the Xpert MTB/RIF test (Cepheid, Sunnyvale, United States of America) roll-out date for their clinic
were assigned to year0, those who started treatment in the first year after roll-out for their clinic were assigned to year1 and so on up to year4 (Fig.1).
c
P-values calculated using the Wald test were also <0.001 for each roll-out year relative to year0.

23%(2445/10643) 11%(1149/10089) 3
HIV- 42%(4229/9985) HIV
27%(2364/8823) HIV-
HIV-
12% 19% HIV-
1% 3%
56%49%

Rsum
Impact du dploiement de tests rapides de diagnostic molculaire de la tuberculose sur les traitements empiriques au Cap, en
Afrique du Sud
Objectif Analyser limpact de lintroduction dun test rapide comme bactriologique par le test Xpert, examen microscopique de frottis
test initial de diagnostic de la tuberculose pharmacosensible au Cap, dexpectorations ou culture dexpectorations.
en Afrique du Sud. Rsultats Entre 2010 et 2014, la proportion de patients srongatifs
Mthodes Le test Xpert MTB/RIF (Xpert), un test automatis qui au virus de limmunodficience humaine (VIH) traits empiriquement
repose sur le principe de lamplification en chane par polymrase, contre la tuberculose est passe de 23% (2445/10643) 11%
a t dploy entre 2011 et 2013. Des donnes taient disponibles (1149/10089); chez les patients sropositifs, elle est passe de 42%
au sujet de 102007adultes traits contre la tuberculose pulmonaire (4229/9985) 27% (2364/8823). Le taux global de signalement
entre 2010 et 2014. Le taux de signalement de cas de tuberculose des cas de tuberculose a baiss respectivement de 12% et de 19%
pour 100000habitants a t calcul pour chaque anne civile et chez les patients srongatifs et sropositifs au VIH; le taux confirm
pour chaque anne de dploiement du test lchelon local, global et bactriologiquement a augment respectivement de 1% et de 3%; et le
avec confirmation bactriologique. Nous avons dfini les traitements taux dadministration de traitements empiriques a baiss respectivement
empiriques comme les traitements donns sans confirmation de 56% et de 49%. Ces changements sont intervenus progressivement

Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314 561

 Research
Rapid diagnostic test roll-out, South Africa Sabine Hermans et al.

aprs lintroduction du test et se sont stabiliss au bout de 3ans. le temps ncessaire sa mise en uvre intgrale. Plus dun quart des
Conclusion Le dploiement du test rapide dans une zone ayant une patients sropositifs au VIH atteints de tuberculose taient encore
forte prvalence de tuberculose pulmonaire et de VIH a t associ traits empiriquement, ce qui montre les enjeux du diagnostic pour
une rduction de moiti des traitements empiriques, qui sest faite ces patients.
progressivement aprs lintroduction du test -ce qui reflte peut-tre

-
,
- , ,
23%(2445/10643) 11%(1149/10089),
, -
. 42%(4229/9985) 27%(2364/8823).
Xpert MTB/RIF(Xpert), 12 19%
, - -
2011 2013. , ,
102007 , 1 3% ,
2010 2014. 56
( 49% .
100000) 3.
, -
-
.
, , ,
Xpert, , -, ,
. . -
2010 2014 - - ,
( ) .

Resumen
El impacto de la puesta en marcha de una prueba rpida de diagnstico molecular para la tuberculosis en un tratamiento
emprico en Ciudad del Cabo, Sudfrica
Objetivo Investigar el impacto de la introduccin de una prueba rpida en pacientes positivos en VIH, cay de un 42% (4229/9985) a un
como la prueba de diagnstico de primera lnea para la tuberculosis 27% (2364/8823). En general, la tasa de notificacin de tuberculosis
sensible a los medicamentos en Ciudad del Cabo, Sudfrica. cay un 12% y un 19% entre pacientes con VIH negativo y positivo,
Mtodos Entre 2011 y 2013 se implement Xpert MTB/RIF (Xpert), respectivamente; la tasa de casos confirmados a nivel bacteriolgico
una prueba automatizada basada en una reaccin en cadena de aument un 1% y un 3%, respectivamente; y la tasa de tratamiento
polimerasa. Se dispuso informacin sobre 102007 adultos tratados emprico se redujo un 56% y un 49%, respectivamente. Estos cambios
de tuberculosis pulmonar entre 2010 y 2014. Se calcularon las tasas se produjeron de forma gradual tras la implementacin de la prueba y
de notificacin de tuberculosis por cada 100000 habitantes por ao se estabilizaron pasados 3 aos.
civil y por ao en relacin con la puesta en marcha de la prueba a nivel Conclusin La puesta en marcha de la prueba rpida en un lugar con
local, general y por confirmacin bacteriolgica. El tratamiento emprico una alta prevalencia de tuberculosis pulmonar e infeccin por VIH se
se defini como un tratamiento proporcionado sin confirmacin asoci con una reduccin a la mitad del tratamiento emprico que se
bacteriolgica por parte de Xpert, microscopia de frotis de esputo o aplic de forma gradual tras la introduccin de la prueba, reflejando
cultivo de esputo. posiblemente el momento necesario para la completa implementacin.
Resultados Entre 2010 y 2014, el porcentaje de pacientes negativos en el Aun as, ms de un cuarto de los pacientes positivos en VIH y con
virus de la inmunodeficiencia humana (VIH) tratados de forma emprica tuberculosis se sometieron a un tratamiento emprico, lo que destaca
de tuberculosis cay de un 23% (2445/10643) a un 11% (1149/10089); el problema del diagnstico en estos pacientes.

References
1. Boehme CC, Nicol MP, Nabeta P, Michael JS, Gotuzzo E, Tahirli R, et al. NEJMoa0907847 PMID: 20825313
Feasibility, diagnostic accuracy, and effectiveness of decentralised use of the 3. Steingart KR, Schiller I, Horne DJ, Pai M, Boehme CC, Dendukuri N. Xpert
Xpert MTB/RIF test for diagnosis of tuberculosis and multidrug resistance: MTB/RIF assay for pulmonary tuberculosis and rifampicin resistance in
a multicentre implementation study. Lancet. 2011 Apr 30;377(9776):1495 adults. Cochrane Database Syst Rev. 2014 01 21;1(1):CD009593. PMID:
505. doi: http://dx.doi.org/10.1016/S0140-6736(11)60438-8 PMID: 24448973
21507477 4. WHO endorses new rapid tuberculosis test: a major milestone for global TB
2. Boehme CC, Nabeta P, Hillemann D, Nicol MP, Shenai S, Krapp F, et al. diagnosis and care. News release. Geneva: World Health Organization; 2010.
Rapid molecular detection of tuberculosis and rifampin resistance. N Available from: http://www.who.int/mediacentre/news/releases/2010/
Engl J Med. 2010 Sep 9;363(11):100515. doi: http://dx.doi.org/10.1056/ tb_test_20101208/en/ [cited 2014 Dec 15].

562 Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314

 Research
Sabine Hermans et al. Rapid diagnostic test roll-out, South Africa

5. Albert H, Nathavitharana RR, Isaacs C, Pai M, Denkinger CM, Boehme CC. 19. Stevens M, Sinanovic E, Regensberg L, Hislop M. HIV and AIDS, STI and TB
Development, roll-out and impact of Xpert MTB/RIF for tuberculosis: what in the private sector. Chapter 14. In: South African Health Review 2007.
lessons have we learnt and how can we do better? Eur Respir J. 2016 Durban: Health Systems Trust; 2007. Available from: http://www.hst.org.za/
Aug;48(2):51625. doi: http://dx.doi.org/10.1183/13993003.00543-2016 uploads/files/chap14_07.pdf [cited 2017 Jan 30].
PMID: 27418550 20. Whitelaw A, Sturm W. Microbiological testing for Mycobacterium
6. Priorities for tuberculosis research: a report of the disease reference tuberculosis. In: Schaaf H, Zumla A, editors. Tuberculosis: a comprehensive
group report on TB, leprosy and Buruli ulcer. Geneva: World Health reference. London: Saunders; 2009.
Organization; 2013. Available from: http://apps.who.int/iris/ 21. National tuberculosis management guidelines 2009. Pretoria: National
bitstream/10665/85888/1/9789241505970_eng.pdf. [cited 2017 Apr 18]. Department of Health, Republic of South Africa; 2009. Available from:
7. Menzies NA, Cohen T, Lin HH, Murray M, Salomon JA. Population health http://familymedicine.ukzn.ac.za/Libraries/Guidelines_Protocols/
impact and cost-effectiveness of tuberculosis diagnosis with Xpert TB_Guidelines_2009.sflb.ashx [cited 2013 Nov 04].
MTB/RIF: a dynamic simulation and economic evaluation. PLoS Med. 22. District council projections 20022016 [spreadsheet]. Pretoria: Statistics
2012;9(11):e1001347. doi: http://dx.doi.org/10.1371/journal.pmed.1001347 South Africa, National Department of Health, Health Information Systems
PMID: 23185139 Programme; 2010. Available from: http://www.statssa.gov.za/publications/
8. Meyer-Rath G, Schnippel K, Long L, MacLeod W, Sanne I, Stevens W, et al. P0302/District_Council_projection_by_sex_and_age_(2002-2016).xlsx
The impact and cost of scaling up GeneXpert MTB/RIF in South Africa. PLoS [cited 2017 Apr 18].
One. 2012;7(5):e36966. doi: http://dx.doi.org/10.1371/journal.pone.0036966 23. AIDS models [Internet]. Cape Town: Actuarial Society of South Africa;
PMID: 22693561 2008. Available from: http://www.actuarialsociety.org.za/Societyactivities/
9. Dowdy DW, Davis JL, den Boon S, Walter ND, Katamba A, Cattamanchi A. CommitteeActivities/DemographyEpidemiologyCommittee/Models.aspx
Population-level impact of same-day microscopy and Xpert MTB/RIF for [cited 2012 Jul 17].
tuberculosis diagnosis in Africa. PLoS One. 2013 08 12;8(8):e70485. doi: 24. Creswell J, Codlin AJ, Andre E, Micek MA, Bedru A, Carter EJ, et al. Results
http://dx.doi.org/10.1371/journal.pone.0070485 PMID: 23950942 from early programmatic implementation of Xpert MTB/RIF testing
10. Theron G, Zijenah L, Chanda D, Clowes P, Rachow A, Lesosky M, et al.; in nine countries. BMC Infect Dis. 2014 01 2;14(1):2. doi: http://dx.doi.
TB-NEAT team. Feasibility, accuracy, and clinical effect of point-of-care org/10.1186/1471-2334-14-2 PMID: 24383553
Xpert MTB/RIF testing for tuberculosis in primary-care settings in 25. Durovni B, Saraceni V, Cordeiro-Santos M, Cavalcante S, Soares E, Loureno
Africa: a multicentre, randomised, controlled trial. Lancet. 2014 Feb C, et al. Operational lessons drawn from pilot implementation of Xpert
1;383(9915):42435. doi: http://dx.doi.org/10.1016/S0140-6736(13)62073-5 MTB/Rif in Brazil. Bull World Health Organ. 2014 Aug 1;92(8):6137. doi:
PMID: 24176144 http://dx.doi.org/10.2471/BLT.13.131409 PMID: 25177076
11. Cox HS, Mbhele S, Mohess N, Whitelaw A, Muller O, Zemanay W, et al. 26. Cowan J, Michel C, Manhia I, Monivo C, Saize D, Creswell J, et al.
Impact of Xpert MTB/RIF for TB diagnosis in a primary care clinic with high Implementing rapid testing for tuberculosis in Mozambique. Bull World
TB and HIV prevalence in South Africa: a pragmatic randomised trial. PLoS Health Organ. 2015 Feb 1;93(2):12530. doi: http://dx.doi.org/10.2471/
Med. 2014 11 25;11(11):e1001760. doi: http://dx.doi.org/10.1371/journal. BLT.14.138560 PMID: 25883406
pmed.1001760 PMID: 25423041 27. Sikhondze W, Dlamini T, Khumalo D, Maphalala G, Dlamini S, Zikalala T, et
12. Durovni B, Saraceni V, van den Hof S, Trajman A, Cordeiro-Santos M, al. Countrywide roll-out of Xpert() MTB/RIF in Swaziland: the first three
Cavalcante S, et al. Impact of replacing smear microscopy with Xpert years of implementation. Public Health Action. 2015 Jun 21;5(2):1406. doi:
MTB/RIF for diagnosing tuberculosis in Brazil: a stepped-wedge cluster- http://dx.doi.org/10.5588/pha.15.0001 PMID: 26400386
randomized trial. PLoS Med. 2014 Dec 9;11(12):e1001766. doi: http://dx.doi. 28. Molapo S, Berrie L, Marokane P, Magida V, Scott L, Stevens W. GeneXpert
org/10.1371/journal.pmed.1001766 PMID: 25490549 module failures: South Africas Xpert MTB/RIF national programme
13. Churchyard GJ, Stevens WS, Mametja LD, McCarthy KM, Chihota V, Nicol experience and impact on costs. Int J Tuberc Lung Dis. 2014 Nov;18(11)
MP, et al. Xpert MTB/RIF versus sputum microscopy as the initial diagnostic suppl 1:s395. Available from: http://barcelona.worldlunghealth.org/
test for tuberculosis: a cluster-randomised trial embedded in South African programme/body/Abstract_Book_2014-Web.pdf [cited 2017 Apr 18].
roll-out of Xpert MTB/RIF. Lancet Glob Health. 2015 Aug;3(8):e4507. doi: 29. Hanrahan CF, Haguma P, Ochom E, Kinera I, Cobelens F, Cattamanchi A,
http://dx.doi.org/10.1016/S2214-109X(15)00100-X PMID: 26187490 et al. Implementation of Xpert MTB/RIF in Uganda: missed opportunities
14. Theron G, Peter J, Dowdy D, Langley I, Squire SB, Dheda K. Do high rates to improve diagnosis of tuberculosis. Open Forum Infect Dis. 2016 05
of empirical treatment undermine the potential effect of new diagnostic 12;3(2):ofw068. doi: http://dx.doi.org/10.1093/ofid/ofw068 PMID: 27186589
tests for tuberculosis in high-burden settings? Lancet Infect Dis. 2014 30. Creswell J, Rai B, Wali R, Sudrungrot S, Adhikari LM, Pant R, et al. Introducing
Jun;14(6):52732. doi: http://dx.doi.org/10.1016/S1473-3099(13)70360-8 new tuberculosis diagnostics: the impact of Xpert() MTB/RIF testing on
PMID: 24438820 case notifications in Nepal. Int J Tuberc Lung Dis. 2015 May;19(5):54551.
15. Menzies NA, Cohen T, Murray M, Salomon JA. Effect of empirical treatment doi: http://dx.doi.org/10.5588/ijtld.14.0775 PMID: 25868022
on outcomes of clinical trials of diagnostic assays for tuberculosis. Lancet 31. Sachdeva KS, Raizada N, Sreenivas A, Vant Hoog AH, van den Hof S, Dewan
Infect Dis. 2015 Jan;15(1):167. doi: http://dx.doi.org/10.1016/S1473- PK, et al. Use of Xpert MTB/RIF in decentralized public health settings and
3099(14)71026-6 PMID: 25541164 its effect on pulmonary TB and DR-TB case finding in India. PLoS One. 2015
16. Sun AY, Denkinger CM, Dowdy DW. The impact of novel tests for 05 21;10(5):e0126065. doi: http://dx.doi.org/10.1371/journal.pone.0126065
tuberculosis depends on the diagnostic cascade. Eur Respir J. 2014 PMID: 25996389
Nov;44(5):13669. doi: http://dx.doi.org/10.1183/09031936.00111014 32. Hermans S, Boulle A, Caldwell J, Pienaar D, Wood R. Temporal trends in TB
PMID: 25186263 notification rates during ART scale-up in Cape Town: an ecological analysis.
17. Management and treatment of TB in adults and children older than 8 years. J Int AIDS Soc. 2015 09 25;18(1):20240. doi: http://dx.doi.org/10.7448/
Circular No: H22/2013. Cape Town: Western Cape Department of Health; 2013. IAS.18.1.20240 PMID: 26411694
18. City of Cape Town 2011 Census Cape Town. Cape Town: Strategic
Development Information and GIS Department, City of Cape Town;
2011. Available from: http://resource.capetown.gov.za/documentcentre/
Documents/Maps%20and%20statistics/2011_Census_Cape_Town_Profile.
pdf [cited 2017 Apr 18].

Bull World Health Organ 2017;95:554563| doi: http://dx.doi.org/10.2471/BLT.16.185314 563