INT J TUBERC LUNG DIS 4(6):550–554

© 2000 IUATLD

A randomised controlled trial of lay health workers as

direct observers for treatment of tuberculosis

M. Zwarenstein,* J. H. Schoeman,* C. Vundule,*† C. J. Lombard,‡ M. Tatley§

* Health Systems Division, Medical Research Council, Tygerberg, † Department of Health, North West Province, ‡ Centre
for Epidemiological Research in Southern Africa, Medical Research Council, § Cape Metropolitan Council, South Africa

SUMMARY

S E T T I N G : Study conducted in a suburb of Cape Town,
South Africa.
O B J E C T I V E : Comparison of successful tuberculosis treat-
ment outcome rates between self supervision, supervi-
sion by lay health worker (LHW), and supervision by
clinic nurse.
M E T H O D S : Open, randomised, controlled trial with
intention-to-treat analysis.
R E S U L T S : All groups (n  156) achieved similar out-
comes (LHW vs. clinic nurse: risk difference 17.2%,
95% confidence interval [CI] 0.1–34.5; LHW vs. self
supervision 15%, 95%CI 3.7–33.6). New patients
benefit from LHW supervision (LHW vs clinic nurse:
risk difference 24.2%, 95%CI 6–42.5, LHW vs. self
supervision 39.1%, 95%CI 17.8–60.3) as do female
patients (LHW vs. clinic nurse 48.3%, 95%CI 22.8–
73.8, LHW vs. self supervision 32.6%, 95%CI 6.4–
58.7).
C O N C L U S I O N S : LHW supervision approaches statisti-
cally significant superiority, but fails to reach it most
likely due to the study’s limitation, the small sample size.
It is possible that subgroups (new and female patients)
do well under LHW supervision. LHW supervision
could be offered as one of several supervision options
within TB control programmes.
K E Y W O R D S : tuberculosis; directly observed therapy;
treatment outcome; randomised controlled trial

TUBERCULOSIS (TB) on its own, and even more in
the context of the human immunodeficiency virus (HIV)
and acquired immune-deficiency syndrome (AIDS) epi-
demic, is an increasingly serious global problem.1 In
1993, before the rise of HIV, TB was the most com-
mon reported infectious disease in South Africa, with
a national notification rate of 223 per 100 000 popu-
lation, and of 703/100 000 in the Western Cape.2
Directly observed treatment, short course (DOTS),
is a tuberculosis care strategy that includes regular
drug supply, improved diagnosis and programme
monitoring, and a 6-month multidrug rifampicin-
containing regimen taken under direct observation by
another person. Direct observation of treatment
(DOT) is the internationally recommended approach
to improving adherence and treatment outcome,1 and
is the basis of the South African National Tuberculo-
sis Control Programme.3,4
The person conducting DOT can be a member of
the clinic staff, an employer, teacher, family member
or a lay volunteer.5 The effectiveness of DOT by lay
health workers (LHWs) has not yet been demon-
strated in a randomised controlled trial.6
The aim of the present study was to compare, as
part of a randomised controlled trial,7 the treatment
outcomes of patients supervised daily in their com-
munities by volunteer LHWs with those of patients
on daily clinic nurse DOT and with those of self
supervised patients on the same drug regimen who
visited the clinic once weekly to collect their drugs.
METHODS
Definition of terms
DOT patients were expected to attend the clinic dur-
ing working hours to take their medication under
observation by clinic nurses. Attendance was expected
5 days per week for the first 8 weeks for new patients
(12 weeks for re-treatment patients), followed by 3
days per week for the continuation phase. A clinic-
held adherence card was signed and dated by a nurse
at each visit. Because most of the responsibility for
adherence lies with the staff, we called this ‘frequent
clinic nurse DOT’.
Self supervised patients visited the clinic once a
week, or sent a family member to collect their drugs.
As responsibility for adherence lay with the patient,
this was termed ‘self supervision’. Patients completed
their own adherence cards for each day of pill taking,
and a nurse recorded the weekly collection of drugs.

Correspondence to: Dr Merrick Zwarenstein, Health Systems Division, Medical Research Council, P O Box 19070, Tyger-
berg, South Africa 7505. Tel: (27) 21 938 0247. Fax: (27) 21 938 0483. e-mail: merrick.zwarenstein@mrc.ac.za
Article submitted 27 May 1999. Final version accepted 1 February 2000.

During the weekly visit to the clinic, patients handed
their adherence record to the nurse.
Patients on LHW supervision took their drugs sev-
eral times per week at their LHW’s home and under
the LHW’s direct supervision. The drugs, held by the
LHW, were replenished each month by the project,
which is managed by the South African National
Tuberculosis Association (SANTA). Patients visited
the clinic only for tests and in the case of complica-
tions. If a patient missed a day’s treatment the LHW
visited the patient’s home; if the LHW could not solve
the problem a SANTA staff member visited the
patient. This project has been described elsewhere.8
The principal outcome, ‘successful treatment’,
includes patients who were cured and those who
completed treatment. ‘Cure’ applied to patients with
pre-treatment sputum that was smear and/or culture
positive, who received 6 months of treatment, and
who had a negative sputum smear and/or culture
result at the end of treatment. ‘Treatment completed’
applied to patients who completed the full course of
treatment, but had neither pre- nor end-of-treatment
bacteriological results; or had negative pre-treatment
results (and had been placed on treatment for clinical
reasons); or had positive pre-treatment results, nega-
tive results at the end of 2 months and no end-of-
treatment result.
Patients with a positive smear or culture result at
the end of treatment were labelled ‘treatment fail-
ures’, while ‘treatment interrupters’ were patients who
stopped taking treatment for 8 or more weeks during
the course of the treatment period.
Other possible outcomes include transfer to another
health facility, and death, due to TB or other causes,
while on treatment.
Treatment regimen
New patients received weight-adjusted Rifater® (com-
bined rifampicin, isoniazid and pyrazinamide, Hoechst
Marion Roussel [HMR], Strasbourg, France), five
times per week during the intensive phase. If they
were smear-negative at 8 weeks, they received Rifi-
nah® (combined rifamipicin and isoniazid, HMR,
France) plus additional isoniazid three times per week
until the end of the 26-week treatment period. Retreat-
ment patients received weight-adjusted Rifater® plus
ethambutol for 12 weeks, and if they were then
smear-negative they received weight-adjusted Rifi-
nah® plus additional isoniazid and ethambutol three
times a week during the 22-week continuation phase.
Study sites
The study was conducted at the four Cape Metropol-
itan Council (CMC) clinics in Elsies River, which
have a high TB caseload (total 443 adults in 1993).9
Elsies River is a suburb comprised of small brick
houses, 20 km from central Cape Town. It was devel-
oped in the 1950s under the apartheid era racial leg-
Trial of lay health worker DOT 551
islation, whereby citizens were evicted from their
homes in areas declared white and moved here. The
community is mainly poor, and in terms of apartheid
era racial classifications, largely coloured.
Study population
The study population consisted of adult (aged over 15
years) pulmonary TB patients, who started a course
of TB treatment (new and retreatment). Diagnosis
was made by primary care doctors and based on
smear and/or culture results.
Exclusion criteria
Patients who were excluded were those who could
not be put on the study regimens for clinical reasons,
such as multiple drug resistance (MDR—defined as
resistance to at least isoniazid and rifampicin) or
severe illness, who were transferred to the study clin-
ics but had been on treatment elsewhere for more
than 2 weeks, who were to be supervised at school,
who were employed and wished to be supervised at
their workplace, and patients who knew they would
be leaving the area within a month. Patients could
only be included in the study once.
Study design and sampling
The study was planned as an unblinded, randomised
controlled trial. The proposed trial procedures were
discussed with nurses and doctors at each clinic.
The planned sample size, 103 in each arm, would
have detected a 20% difference in outcome between
groups, with power of 80% at a significance level of
5%. Based on information provided by the health ser-
vices, enough patients were thought to be available to
obtain the envisaged sample size.
Randomisation
At the first treatment visit after diagnosis, eligible
patients were approached by a clinic nurse, who
explained the purpose and voluntary nature of the trial
and asked for verbal consent. If given, the nurse drew
an envelope from the top of a box of consecutively
numbered opaque sealed envelopes, recording the allo-
cation in the trial book for that clinic. The randomisa-
tion sequence was generated by computer algorithm.
Researchers (CV and HS) regularly checked that trial
numbers were allocated in sequence, that changes in
option were explained, and that questionnaires were
complete.
All patients received daily clinic nurse DOT for the
first 2 weeks of treatment, and were only informed of
their supervision allocation after this. Anyone attend-
ing fewer than eight of these first 10 appointments was
shifted to standard therapy, frequent clinic nurse DOT.
Data collection
Trained lay interviewers interviewed patients in their
own language during the first 2 weeks of treatment.
552 The International Journal of Tuberculosis and Lung Disease

The trial was explained again, and written consent Table 2 Patient characteristics: Elsies River
was obtained. At the end of the trial, we collected
Clinic DOT LHW Self Total
information on sputum results and treatment out- n (%) n (%) n (%) n (%)
come from patient files and on excluded patients from
All patients 58 (37) 54 (35) 44 (28) 156 (100)
the CMC database.
Age (years)
35 28 (48) 25 (46) 23 (52) 76 (49)
Statistical analysis 35 30 (52) 29 (54) 21 (48) 80 (51)
Analysis was by intention to treat.10 Rates of success- Sex
ful treatment for each supervision approach were cal- Male 39 (67) 27 (50) 25 (57) 91 (58)
Female 19 (33) 27 (50) 19 (43) 65 (42)
culated for all patients, and for subgroups by sex and
Status
first occurrence vs. repeat TB. We report the differ- Single 39 (67) 40 (74) 29 (66) 108 (69)
ence in successful treatment outcome rates, with a Married 19 (33) 14 (26) 15 (34) 48 (31)
95% confidence interval (CI). Overall comparisons Housing
between the three groups are made using Fisher’s Formal 49 (85) 49 (91) 41 (93) 139 (89)
Informal 9 (15) 5 (9) 3 (7) 17 (11)
exact test.
Income
We used multiple logistic regression analysis in Employed 11 (19) 5 (9) 1 (2) 17 (11)
SAS to adjust for potential confounders, using the Other* 47 (81) 48 (91) 43 (98) 138 (88)
dichotomous variable, successful outcome (yes/no).11 Information missing 1
Education (years)
0–5 16 (28) 15 (28) 13 (29) 44 (28)
6–8 23 (40) 29 (54) 25 (56) 77 (49)
RESULTS 9 19 (33) 10 (19) 6 (14) 35 (22)
From May 1994 to September 1995, 174 patients Patient category
were recruited into the study (Table 1). None refused. New 36 (62) 33 (61) 27 (61) 96 (62)
Re-treatment 22 (38) 21 (39) 17 (39) 60 (38)
After exclusion of 12 MDR and six non-TB patients,
* Unemployed, supported by family or friends, pension or grant.
156 were included for analysis. Thirteen patients DOT  directly observed treatment; LHW  lay health worker.
from community and eight from self supervision
changed to clinic DOT, two from clinic DOT and one
from community supervision changed to self supervi- other sources of income fell into ‘other’, which makes
sion, and two from clinic DOT and one from commu- it difficult to interpret.
nity changed to workplace supervision. There were no statistically significant differences in
Demographic and other characteristics of the successful treatment rates across the three supervision
study patients and outcomes for the supervision options (P  0.136, Table 3). LHW supervision out-
methods are shown in Table 2. No significant differ- comes, over all patients, were not statistically signifi-
ences were observed between supervision methods cantly superior to either clinic DOT (17.2%, 95%CI
and these variables, except for the income variable (P  0.1–34.5) or to self supervision (15%, 95%CI 3.7–
0.026). This variable was problematic, as we applied 33.6). Clinic and self supervision outcomes were simi-
strict criteria for the category ‘employed’ and all lar to each other (2.2%, 95%CI 21.5–17.1).
A number of pre-planned subgroup analyses were
performed. For new patients, LHW supervision was
Table 1 Patient accounting
superior to both clinic (24.2%, 95%CI 6–42.5) and
self supervision (39.1%, 95%CI 17.8–60.3). For re-
No. of treatment patients LHW supervision was not statisti-
patients
cally significantly different from clinic supervision
CMC listing 536 (6.7%, 95%CI 22.9–36.3) or self supervision (23%,
Excluded 95%CI 53.47.5).
Transfer in 34
Transfer out 35
For female patients, LHW supervision was supe-
Not traced 4 rior to clinic supervision (48.3%, 95%CI 22.8–73.8)
Missed 15 and self supervision (32.6%, 95%CI 6.4–58.7), while
Retreatment 0
School supervised* 20
for male patients it was not (clinic supervision
Workplace supervised* 70 3.7%, 95%CI 27.2–19.8, self supervision 1%,
Working or seeking work† 184 95%CI 27.2–25.2).
Total excluded 362 Multiple logistic regression showed significant
Randomised 174 effects for supervision option (P  0.0405) and
MDR 12
Not TB 6
patient category (new or re-treatment, P  0.0215),
Analysed 156
but not for sex (P  0.3008). The model also showed
significant interactions between patient category and
* Estimated by staff.
† Estimated by subtraction from CMC total.
supervision option (P  0.0053) and borderline sig-
CMC  Cape Metropolitan Council; MDR  multidrug-resistant. nificant interactions between sex and supervision
 Trial of lay health worker DOT 553

Table 3 Analysis of treatment outcome by supervision than either clinic DOT (US$580) or self supervision
method (US$174)12 in this study area. This refers only to
Clinic DOT LHW Self Total direct service costs, ignoring other possibly substan-
n (%) n (%) n (%) n (%) tial advantages of LHW DOT.
All patients Although sample size limitations prevent us from
Cured 24 (41) 31 (57) 18 (41) 73 (47) unequivocally recommending LHW DOT, the findings
Treatment completed 9 (16) 9 (17) 8 (18) 26 (17) of this study are promising, and may well be relevant
Successful outcome 33 (57) 40 (74) 26 (59) 99 (64)
Treatment failure 1 (2) 3 (6) 2 (5) 6 (4) for other deprived urban communities. Multicentre
Treatment interrupter 15 (26) 8 (15) 11 (25) 34 (22) randomised trials of LHW DOT are required to reach
Transferred 9 (16) 1 (2) 4 (9) 14 (9) adequate scale and evaluate the approach under a
Died 0 (0) 2 (4) 1 (2) 3 (2)
Other outcome 35 (43) 14 (26) 18 (41) 57 (36) wider range of settings. In the context of such a trial,
Total 58 (37) 54 (35) 44 (28) 156 (100) LHW supervision should be offered as one of several
New patients supervision options within TB control programmes.
Successful outcome 24 (67) 30 (91) 14 (52) 68 (71)
Other outcome 12 (33) 3 (9) 13 (48) 28 (29)
Retreated patients Acknowledgements
Successful outcome 9 (41) 10 (48) 12 (71) 31 (52) We would like to thank the patients who participated in the study,
Other outcome 13 (59) 11 (52) 5 (29) 29 (48) clinic staff, Dr Jennifer Talent, interviewers Nosisi Dingani, Prin-
Male cess Toyi, Damien Buckton and Louen Kleinschmidt.
Successful outcome 26 (67) 17 (63) 16 (64) 59 (65) Funding was received from the Medical Research Council, the
Other outcome 13 (33) 10 (37) 9 (36) 32 (35) Department of Health and the Health Systems Trust.
Female
Successful outcome 7 (37) 23 (85) 10 (52) 40 (62)
Other outcome 12 (63) 4 (15) 9 (47) 25 (38)
References
DOT  directly observed treatment; LHW  lay health worker.
1 World Health Organization. WHO report on the tuberculosis
epidemic 1997. Geneva: WHO, 1997.
option (P  0.0643). Inclusion of these potentially 2 South African Department of National Health. Tuberculosis
update. Epidemiological Comments, Pretoria, South Africa,
confounding terms in the model did not render the
1995; 22: 13–17.
main effect, that of supervision option, non-significant. 3 Cameron N. Tuberculosis control—a new paradigm in South
Africa. [Letter] S Afr Med J 1996; 86: 271.
4 Department of Health. The South African tuberculosis control
DISCUSSION programme. Practical guidelines. Pretoria: Department of
This is the first randomised trial evidence of the effects Health, 1996.
5 Dick J, Schoeman J H, Mohammed A, Lombard C. Tuberculo-
on tuberculosis treatment outcome of DOT by lay sis in the community: 1. Evaluation of a volunteer health
health workers. As it was not our intention to include worker programme to enhance adherence to anti-tuberculosis
patients allocated to school or workplace supervision, treatment. Tubercle Lung Dis 1996; 77: 247–249.
our findings should not be generalised to these options, 6 Volmink J, Garner P. Promoting adherence to tuberculosis
nor to supervisors other than clinic nurses and lay treatment. In: Garner P, Gelband H, Olliaro P, Salinas R,
Volmink J, Wilkinson D, eds. Infectious Diseases Module of
health workers. The Cochrane Database of Systematic Reviews [updated 01
The sample size obtained was smaller than September 1997]. The Cochrane Collaboration; Issue 4. Ox-
expected, and this prevents us from concluding that ford: Update Software; 1997. Updated quarterly.
LHW supervision, over all patients, is superior to 7 Zwarenstein M, Schoeman, J H, Vundule C, Lombard C, Tat-
clinic or self supervision. With regard to our smaller ley M. Surveillance or support: a randomised controlled trial of
self supervised versus directly observed treatment of tuberculo-
sample size, we calculated that if a study is planned,
sis. Lancet 1998; 352: 1340–1343.
to detect a difference of 20%, with a 5% significance 8 Dick J, Schoeman J H. Tuberculosis in the community: 2. The
level and 55 patients per group, it will render a power perceptions of members of a tuberculosis health team towards
of 60%. However, if we were to combine self and a voluntary health worker programme. Tubercle Lung Dis
clinic supervised groups, as suggested by our findings 1996; 77: 380–383.
9 Western Cape Regional Services Council. Annual Report of the
of equivalence between these groups,7 and compare
Department of Health Services. Cape Town: Western Cape Re-
this combined group with LHW supervision, the lat- gional Services Council, 1993.
ter would be significantly superior to the combined 10 Newell D J. Intention-to-treat analysis: implications for quan-
group (16.2%, 95%CI 1.1–31.3). titative and qualitative research. Int J Epidemiol 1992; 21:
LHW supervision might be superior to clinic and 837–841.
self supervision for women and for new patients. These 11 SAS Institute Inc. SAS/STAT software: changes and enhance-
ments through release 6.12. Cary, NC: SAS Institute, 1997.
results need to be researched in further larger studies, 12 Dick J, Henchie S. A cost analysis of the tuberculosis control
as they may have important policy implications. programme in Elsies River, Cape Town. S Afr Med J 1998; 88:
LHW supervision, at US$116 per patient, is cheaper 380–383.
554 The International Journal of Tuberculosis and Lung Disease
C A D R E : Etude conduite dans un faubourg de Cape
Town, Afrique du Sud.
O B J E C T I F : Comparaison des taux de succès finaux du
traitement obtenus chez les patients tuberculeux laissés à
eux-mêmes ou supervisés par un aide sanitaire (LHW)
ou par une infirmière de polyclinique.
M É T H O D E S : Essai contrôlé randomisé ouvert suivi
d’une analyse selon l’intention de traiter.
R É S U L T A T S : Quand on considère l’ensemble des 156
patients, des résultats similaires ont été obtenus dans
tous les groupes : différence de succès entre LHW et
infirmières de policlinique 17,2% (intervalle de confi-
ance à 95% [IC95%] 0,1 à 34,5) ; entre LHW et auto-
supervision 15% ; (IC95% 3,7 à 33,6). Les nouveaux
patients bénéficient d’une supervision par les LHW : dif-
férence de succès entre LHW et infirmières de policlin-
M A R C O D E R E F E R E N C I A : Estudio efectuado en un sub-
urbio de la Ciudad del Cabo, Sud África.
O B J E T I V O : Comparación de los resultados de las tasas
de tratamiento exitoso entre los pacientes tuberculosos
auto-supervisados, supervisados por un agente sanitario
(LHW) y por enfermeras de dispensarios.
M É T O D O S : Ensayo controlado abierto aleatorio con
análisis intencional.
H A L L A Z G O S : Todos los grupos (n  156) alcanzaron
resultados similares (LHW versus enfermera : diferencia
de riesgo 17,2%, IC95% 0,1–34,5 ; LHW versus auto-
supervisión : 15%, IC95% 3,7–33,6). Los pacientes nue-
vos se benefician por la supervisión LHW (LHW vs
RÉSUMÉ
ique 24,2% (IC95% 6 à 42,5) ; entre LHW et auto-
supervision 39,1% (IC95% 17,8 à 60,3). Il en est de
même pour les patients de sexe féminin : LHW versus
infirmières de polyclinique 48,3% (IC95% 22,8 à 73,8) ;
LHW versus auto-supervision 32,6% (IC95% 6,4 à
58,7).
C O N C L U S I O N S : La supervision par les LHW approche
une supériorité statistiquement significative mais sans
l’atteindre, ce qui est le plus probablement dû à la limi-
tation que représente dans cette étude la faible taille de
l’échantillon. Il est possible que des sous-groupes (en
l’occurrence les nouveaux patients et les patients fémi-
nins) bénéficient d’une supervision par les LHW. Cette
supervision par les LHW pourrait être présentée comme
une des différentes options de supervision au sein des
programmes de lutte contre la TB.
RESUMEN
enfermera : diferencia de riesgo 24,2%, IC95% 6–
42,5%, LHW vs. auto-supervisión : 39,1%, IC95%
17,8-60,3) como tambien las pacientes mujeres (LHW
vs. enfermera : 48,3%, IC95% 22,8–73,8, LHW vs
auto-supervisión 32,6%, IC95% 6,4–58,7).
C O N C L U S I O N E S : La supervisión LHW aporta una supe-
rioridad estadísticamente significativa pero carece de
valor debido a lo limitado del estudio y a la muestra
pequeña. Es posible que algunos subgrupos (pacientes
nuevos y mujeres) obtengan beneficio con la supervisión
LHW. La supervisión LHW puede ser ofrecida como
una de las opciones de supervisión en los programas de
control de la TB.