Naidoo et al.

BMC Public Health 2013, 13:396

http://www.biomedcentral.com/1471-2458/13/396

RESEARCH ARTICLE Open Access

Predictors of tuberculosis (TB) and antiretroviral

(ARV) medication non-adherence in public
primary care patients in South Africa: a cross
sectional study
Pamela Naidoo1,2*, Karl Peltzer1,3,4, Julia Louw1, Gladys Matseke1, Gugu Mchunu5 and Bomkazi Tutshana1

Abstract
Background: Despite the downward trend in the absolute number of tuberculosis (TB) cases since 2006 and the
fall in the incidence rates since 2001, the burden of disease caused by TB remains a global health challenge. The
co-infection between TB and HIV adds to this disease burden. TB is completely curable through the intake of a
strict anti-TB drug treatment regimen which requires an extremely high and consistent level of adherence.The aim
of this study was to investigate factors associated with adherence to anti-TB and HIV treatment drugs.
Methods: A cross-sectional survey method was used. Three study districts (14 primary health care facilities in each)
were selected on the basis of the highest TB caseload per clinic. All new TB and new TB retreatment patients were
consecutively screened within one month of anti-tuberculosis treatment. The sample comprised of 3107 TB patients
who had been on treatment for at least three weeks and a sub-sample of the total sample were on both anti-TB
treatment and anti-retro-viral therapy(ART) (N = 757). Data collection tools included: a Socio-Demographic
Questionnaire; a Post-Traumatic-Stress-Disorder (PTSD) Screen; a Psychological Distress Scale; the Alcohol Use
Disorder Identification Test (AUDIT); and self-report measures of tobacco use, perceived health status and
adherence to anti-TB drugs and ART.
Results: The majority of the participants (N = 3107) were new TB cases with a 55.9% HIV co-infection rate in this
adult male and female sample 18 years and older. Significant predictors of non-adherence common to both anti-TB
drugs and to dual therapy (ART and anti-TB drugs) included poverty, having one or more co-morbid health
condition, being a high risk for alcohol mis-use and a partner who is HIV positive. An additional predictor for
non-adherence to anti-TB drugs was tobacco use.
Conclusions: A comprehensive treatment programme addressing poverty, alcohol mis-use, tobacco use and
psycho-social counseling is indicated for TB patients (with and without HIV). The treatment care package needs to
involve not only the health sector but other relevant government sectors, such as social development.
Keywords: Adult TB patients, Bio-Psycho-Social factors, Anti-TB treatment, ART, High burden country, Adherence
to anti-TB treatment and ART

* Correspondence: pnaidoo@hsrc.ac.za
1
Population Health, Health Systems and Innovation (PHHSI)/HIV/STIs and TB
(HAST) Research Programmes, Human Sciences Research Council, Pretoria,
Durban and Cape Town, South Africa, Private Bag X 9182, Cape Town 8000,
South Africa
2
Department of Psychology, University of the Western Cape, Cape Town,
South Africa
Full list of author information is available at the end of the article

© 2013 Naidoo et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Naidoo et al. BMC Public Health 2013, 13:396 Page 2 of 10
http://www.biomedcentral.com/1471-2458/13/396

Background TB; quality of health care received; use of the traditional

Despite the downward trend in the absolute number of healing system; and the patient’s HIV status [5,7]. Quan-
TB cases since 2006 and the fall in the incidence rates titative studies examining the factors that influence
since 2001, the burden of tuberculosis (TB) disease re- adherence to anti-TB treatment regimens found the fol-
mains a global health challenge [1]. TB is completely cur- lowing to be important: a good patient–practitioner rela-
able through the intake of a strict drug treatment tionship, ability of the patient to disclose medication use
regimen. The Directly Observed Treatments Short- to members of their social network and regular clinic
Course Strategy (DOTS) introduced by the World Health visits [8,9].
Organization (WHO) and subsequently the Stop TB Numerous studies have also been conducted in Africa
Strategy is an inexpensive strategy that could prevent and globally on factors influencing adherence to anti-TB
millions of TB cases and deaths. TB/HIV co-infection treatment and adherence to the dual treatment approach
and multi-drug resistant (MDR)/ extensive drug-resistant for TB and HIV [2,10,11]. A systematic review of quali-
(XDR)TB are given focal attention in the Stop TB Strat- tative research, exploring patient adherence to tubercu-
egy and one of its primary TB targets is to reduce by half losis treatment, conducted by Munro et al found that
the TB prevalence rates by 2015 relative to 1990 [1]. the following four factors interact to affect adherence to
The HIV and AIDS pandemics have exacerbated the TB treatment: structural factors, including poverty and
public health dimensions of TB due to the fact that gender discrimination; the social context; health service
many individuals are co-infected. TB is the leading cause factors and personal factors [5]. An African-based quali-
of death among people who are HIV positive. In the tative study which explored barriers and facilitators of
African region, which accounted for 82% of the new TB adherence to anti-TB treatment and concomitant TB
cases that were also HIV positive, an estimated 900 000 and HIV treatment in Ethiopia, found that adherence to
(39%) of the 2.3 million people who developed TB were TB treatment was positively influenced by beliefs in the
living with HIV [1]. curability of TB, beliefs in the severity of TB in the pres-
ence of HIV infection and support from family members
Adherence to anti-TB treatment and to treatment for and health professionals [11].
HIV/TB co-infection A systematic review done on studies conducted in the
Poor adherence to the prescribed anti-TB treatment US and Canada on adherence to treatment for latent tu-
programmes, such as those falling under the DOTS strat- berculosis infection found that there was a “sub-optimal”
egy, is one of the factors associated with low cure rates for level of adherence indicated across all studies [12]. The
TB. In addition, inconsistent adherence to the anti-TB drug factors influencing adherence were clinic facilities, treat-
regimen may lead to multiple and extensive drug resistance ment characteristics and patient factors although the as-
(MDR-TB and XDR-TB respectively) making it difficult to sociation between adherence and these factors was found
achieve high cure rates. In individuals with TB/HIV dual to be inconsistent [12]. Adherence to ARVs as compared
infection receiving treatment, correct adherence to anti-TB to anti-TB drugs is, however, reported to be high the
drugs and anti-retroviral therapy (ART) are essential for world over and Corless et al based a quantitative study
good treatment outcomes [2]. Poor and inconsistent adher- conducted in clinics in Durban, SA, on this premise [10].
ence to ART can also lead to drug resistance and even They found that in fact adherence to ARVs in their study
death, an outcome which is similar to non-adherence to was also high [10]. Amuha et al found that the most im-
anti-TB drugs [2,3]. portant factor associated with non-adherence to anti-TB
The categories of factors influencing adherence to medication in individuals co-infected with HIV was being
drug treatments for most health-related conditions, in- on a continuous treatment regimen phase as compared
clude: practitioner’s negative assumptions about their to the intensive treatment phase. Confounding factors in-
patients; psychological attributes of the patient; environ- fluencing the association between the stage of the TB
mental, social and cultural factors; treatment character- regimen and non-adherence were alcohol consumption,
istics; and the doctor–patient relationship [3,4]. Many being on ARVs and smoking [13]. The exact nature of
quantitative and qualitative studies conducted in South these confounding factors and its influence on non-
Africa (SA) have identified factors known to influence adherence has not been interrogated.
adherence to anti-TB treatment [5-7]. Qualitative studies Relatively few studies have looked specifically at the
exploring adult TB patient’s adherence to anti-TB treat- sub-group of individuals on anti-TB treatment who also
ment at public health sites in SA found that the factors meet the criteria for alcohol use disorders in South
that influenced patient co-operation included: social and Africa (SA). This study forms part of a larger study en-
economic resources; causal attributions assigned to TB; titled: Screening and brief interventions for hazardous
the social, cultural, economic, disease-related and psy- and harmful alcohol use among patients with active tu-
chological challenges faced as a consequence of having berculosis attending primary public care clinics in SA.
Naidoo et al. BMC Public Health 2013, 13:396 Page 3 of 10
http://www.biomedcentral.com/1471-2458/13/396

There is adequate evidence in the international litera- of the DOTS programme which required them to at-
ture to confirm that TB medication non-adherence is tend the clinic during the week and take their anti-TB
associated with the use and mis-use of alcohol [14,15]. drugs dose at their homes over the week-end.
In SA, however, the association between alcohol (mis) Within the larger sample there was a sub-sample of
use and non-adherence to anti-TB treatment and dual patients who were HIV positive (N = 1729) and a pro-
therapy in those individuals co-infected with HIV has portion of the HIV positive patients (44%) who were on
been not been adequately studied. In addition, the way ART (N = 757).
in which individual mental health risk factors, such as A screening interview was conducted by trained re-
distress and post-traumatic stress, mediate TB and TB/ search assistants over a period of six (6) months in 2011.
HIV health outcomes for individuals mis-using alcohol, A health care provider who identified a new TB treat-
has been neglected. ment or retreatment patient 18 years and above informed
The value of this study is that it used a different meth- the patient about the study and referred the patient for
odology to most other TB treatment adherence studies. participation if they were willing. A consenting procedure
The study uses different questionnaires (such as the was adopted prior to the start of the screening interview.
measures of distress and PTSD) and different outcome Ethical approval was received from the Human Sciences
measures (namely, adherence to anti-TB treatment, and Research Council Research Ethics Committee (Protocol
adherence to ART and anti TB treatment (dual therapy) REC No.1/16/02/11) and the Department of Health
as compared to other adherence studies. Consequently, (DoH) in SA.
the primary aim of this study, using base-line data was
to investigate the factors that are associated with non- Data collection tools
adherence to anti-TB drugs for those with active TB and The data collection tools specified in this section was
to ART and anti-TB drugs taken by individuals who have administered to all the participants in this study.
dual infection. The secondary aim, which also addresses
a gap in research, was to specifically investigate the sig- Socio-demographic questionnaire
nificance of alcohol misuse as a predictive factor for A researcher-designed questionnaire was used to record
non-adherence to anti-TB drugs and to dual therapy information on participants’ age, gender, educational
(namely, anti-TB drugs and ART). level, marital status, income, employment status, dwell-
ing characteristics and residential status. Using a previ-
Methods ously designed questionnaire, which is also extensively
Study design, sample and procedure used in the Human Sciences Research Council (HSRC_SA)
This study is a cross-sectional survey. Fourteen (14) national health surveys, poverty was assessed by five (5)
public primary health care clinics in only one district pertinent items on the questionnaire by asking about the
in each of three provinces, namely, Northern Cape, availability or non-availability of shelter, fuel or electricity,
Eastern Cape and Kwa-Zulu Natal in SA with the clean water, food and cash income in the past week
highest TB caseload were included in the study. All [17]. Response options ranged from 1 = “Not one day”
new TB and new retreatment patients were consecu- to 4 = “Every day of the week”. Participants ranked high on
tively screened within one month of anti-TB treat- poverty if they had higher scores on non-availability of es-
ment. The public primary health care clinics that sential items. The total poverty score ranged from 5 to 20
were utilized in this study followed the SA Depart- with the following categories: 5 indicating low poverty; 6 to
ment of Health’s (DoH) Guidelines for TB treatment 12 indicating a medium level of poverty; and 13 to 20 indi-
who in turn are guided by the WHOs TB treatment cating high levels of poverty. The Cronbach alpha for the
guidelines framed by the Stop TB strategy which poverty index in this study was fairly good (0.89).
strongly recommends the DOT programme [16]. The
DOT programme requires intensive involvement and The Kessler Psychological Distress Scale (K-10)
monitoring by the clinic staff but the treatment suc- The Kessler Psychological Distress Scale (K-10) was
cess is also dependent on the patient adhering strictly used to measure global psychological distress [18,19].
to the treatment guidelines. In this study a total of The K-10 measures the following symptoms over the
3129 potential participants were screened and 22 preceding 30 days: nervousness, hopelessness, restless-
(0.7%) refused to participate. It was not necessary to ness, depression, worthlessness and tiredness. The fre-
perform an attrition analysis due to the good response quency with which each of these items was experienced
rate. In the data analysis the total sample size used was recorded using a five-point Likert scale ranging from
was 3107 participants (N = 3107). All the participants’ “none of the time” to “all the time”. Increasing total
in this study were taking anti-TB treatment for at scores reflect an increasing degree of psychological dis-
least three weeks, which fell into the intensive phase tress. The K-10 has been shown to capture variability
Naidoo et al. BMC Public Health 2013, 13:396 Page 4 of 10
http://www.biomedcentral.com/1471-2458/13/396

related to non-specific depression, anxiety and substance the second question asked about the frequency of
abuse [19]. This scale serves to identify individuals who use over the past month.
are likely to meet formal definitions for anxiety and/or (b) Non-adherence to anti-TB treatment, non-adherence
depressive disorders, as well as to identify individuals to ART, and HIV testing: Adherence was assessed by
with sub-clinical illness who may not meet formal defini- self-report. Whether a participant tested for HIV
tions for a specific disorder [18]. This scale is increasingly infection was also assessed by self-report. Anti-TB
used in population based mental health research and has medication adherence was assessed with the following
been validated in multiple settings [20,21] including HIV question: “In your tuberculosis treatment in the past
positive individuals in SA [22]. The internal reliability co- 3–4 weeks how many percent (%) of your anti-
efficient for the K-10 in this study was alpha = 0.92 which tuberculosis medication did you take?” TB medication
is fairly high. non-adherence was defined as having taken less than
90% of the anti-TB drugs. Self-reporting adherence
Alcohol consumption behaviour over the past 3–4 week period for anti-TB
The 10-item Alcohol Use Disorder Identification Test drugs was asked due to the fact that the participants
(AUDIT) [23] assesses alcohol consumption level, symp- in this study were recruited only if they had started
toms of alcohol dependence and problems associated with their TB treatment at least three weeks before being
alcohol use. The AUDIT largely measures heavy episodic enrolled into the study.
drinking with only one (1) item measuring binge drinking. ART adherence was assessed with the question:
Heavy episodic drinking is defined as the consumption of “How many percent of your HIV medication did
six standard drinks or more on a single occasion. A stand- you take in the past 4 weeks?” ART non-adherence
ard drink in this instance is 10 g of alcohol. In SA a stand- was defined as having taken less than 90% of ART.
ard drink is 12 g of alcohol and not 10 g. The AUDIT is In each instance, that is for both ART and anti-TB
reported to be less sensitive at identifying risk drinking in treatment non-adherence, participants were
women [24] so it was recommended that the cut-off point required to mark the percentage adherence on an
for the binge drinking measure is reduced by one (1) unit illustrated scale indicating progression from 0% to
among women in SA and the overall heavy episodic drink- 100% adherence.
ing measure is also reduced due to the fact that a standard (c) Co-morbidity with other chronic medical conditions
drink is 12 g of alcohol (and not 10 g as stipulated by the including hypertension, diabetes, depression, stomach
AUDIT). Responses to items on the AUDIT are rated on a ulcer, migraine headache, cancer, arthritis, asthma,
four-point-Likert scale from 0 to 4, for a maximum score diabetes, cholesterol were also ascertained.
of 40 points. A cut off score of 8 indicates a tendency to
problematic drinking. The AUDIT was developed by the Data analysis
WHO as an effective screening instrument for alcohol Data were analyzed using the Statistical Package for the So-
use problems among patients seeking primary care for cial Sciences (SPSS-version 19). Frequencies, means, and
other medical problems in international settings including standard deviations, were calculated to describe the sam-
African countries (Kenya and Zimbabwe) [24-27]. The ple. Data were checked for normality distribution and out-
AUDIT has been validated in HIV patients in SA showing liers. For non-normal distribution non-parametric tests
excellent sensitivity and specificity in detecting alcohol de- were used. Associations between TB medication and ART
pendence and alcohol abuse as defined on the Mini- non-adherence were identified using logistics regression
International Neuropsychiatric Interview (MINI) [26]. The analyses. Following each univariate regression, multivari-
MINI is an internationally recognized diagnostic tool in able logistic regression models were constructed. A total of
the form of a psychiatric interview. 574 participants were used in the multivariate analysis. In-
The MINI used in a study of TB and HIV patients in pri- dependent variables from the univariate analyses were en-
mary care in Zambia also demonstrated good discrimin- tered into the multivariable model if significant at P < 0.05
atory ability in detecting MINI-defined current Alcohol level. For each model, the R2 are presented to describe
Use Disorders (AUDs) (AUDIT = 0.98 for women and 0.75 the amount of variance explained by the multivariable
for men) [27]. Cronbach alpha for the AUDIT in this sam- model. Probability below 0.05 was regarded as statistically
ple was 0.92, indicating excellent reliability. significant.

Additional self-report measures Results

Characteristics of the final sample
(a) Tobacco use: Two researcher generated questions Table 1 indicates that the sample included 671 (21.8%;
were asked about the use of tobacco products. The N = 3107) retreatment cases, 2408 (78.2%; N = 3107)
first question asked about current tobacco use and new TB cases and 55.6% HIV infected cases. A little
Naidoo et al. BMC Public Health 2013, 13:396 Page 5 of 10
http://www.biomedcentral.com/1471-2458/13/396

Table 1 Characteristics of the total sample of TB patients Table 1 Characteristics of the total sample of TB patients
and sub-sample of TB patients on ART and sub-sample of TB patients on ART (Continued)
Socio-demographics Total TB patients Total TB-ART Partner HIV negative/ 2061 (71.6) 436 (61.8)
(N = 3107) patients (N = 757) unknown
N (%) N (%) Partner HIV positive 818 (28.4) 269 (38.2)
Age 18–24 416 (13.5) 65 (8.7)
Unknown/missing N 228 (7.3) 52 (6.9)
Age 25–34 1160 (37.7) 283 (37.8)
Sex partner on ART 300 (11.8) 355 (48.8)
Age 35–44 861 (28.0) 252 (33.7)
TB non-adherence 812 (26.1) 315 (41.6)
Age 45 and older 639 (20.8) 148 (19.8)
ART non adherence 321 (42.4)
Missing N 31 9
Female 1427 (46.5) 387 (51.7)
Male 1639 (53.5) 362 (48.3) below half the sample was 35 years and older. Half the
Missing N 41 8 sample reported medium levels of poverty, about a third
Grade 7 or less 774 (25.2) 234 (31.5) (35.3%) good perceived health status with more partici-
pants (46.1%) reporting poor perceived health status.
Grade 8–11 1415 (16.1) 330 (44.4)
Other characteristics of significance include the follow-
Grade 12 or more 881 (28.7) 179 (24.1)
ing: 776 (25%) met the criteria for severe psychological
Missing N 37 14 distress and 16.4% for the category of medium risk for
Poverty low 1052 (35.5) 173 (24.4) alcohol misuse on the AUDIT. Finally, 26.1% were non-
Poverty medium 1484 (50.1) 377 (53.1) adherent to the anti-TB treatment (min = 0, max = 100,
Poverty high 427 (14.4) 160 (22.5) mean = 77.0, median = 90.0, SD = 43.9).
A sub-sample (N = 757) of the total sample was on
Missing N 144 47
dual anti-TB treatment and ART. The characteristics of
Health variables
this group were as follows: the majority of participants
Perceived health status were between the ages of 25 and 44 years, half the sam-
Excellent/Very good 573 (18.6) 217 (28.8) ple reported medium levels of poverty, two-fifths (40.5%)
Good 1089 (35.3) 231 (30.2) reported poor perceived health status, 15.9% were at the
Fair/Poor 1423 (46.1) 305 (40.5) medium risk category for alcohol misuse on the AUDIT,
38.2% had a partner who was HIV positive, and 42.4%
Unknown N 22 4
were non-adherent to ART (min = 0, max = 100, mean =
TB retreatment patient 671 (21.8) 224 (30.0)
64.8, median = 100.0, SD = 43.9). Of a total of 268 partic-
New TB patient 2408 (78.2) 523 (70.0) ipants 83.8% were not adherent to both anti-TB and
Unknown N 28 10 ART medication (r = 0.71).
HIV positive 1729 (55.9)
HIV negative 1073 (34.5) Predictors of non-adherence to TB treatment
Univariate analysis (see Table 2) shows that the follow-
HIV unknown status 305 (9.8)
ing factors were more likely to be associated with
Chronic conditions
non-adherence to TB treatment: being male [OR: 1.26
Zero 1983 (72.5) 422 (65.7) (1.07–1.48), p < 0.05], medium poverty [OR: 1.97 (1.62–
One 443 (16.2) 112 (17.4) 2.41), p < 0.001], high levels of poverty [OR: 4.01 (3.12–
Two 194 (7.1) 60 (9.3) 5.10), p < 0.001], having one (1) chronic condition [OR:
Three or more 117 (4.3) 48 (7.5) 1.54 (1.23–1.93), p < 0.001], having two (2) chronic
conditions [OR: 1.84 (1.34–2.51), p < 0.001], having
Unknown N 370 115
three or more chronic conditions [OR: 1.98 (1.34–
Severe psychological distress 776 (25.0) 208 (27.5)
(K ≥ 30)
2.92), p < 0.001], severe psychological distress [OR: 1.31
(1.09–1.57), p < 0.01], medium risk for alcohol misuse
Alcohol: low risk (AUDIT 2392 (78.0) 599 (80.0)
0–7) [OR: 1.74 (1.42–2.19), p < 0.001], high risk for alcohol
misuse [OR: 2.42 (1.76–3.32), p < 0.001], tobacco use
Medium (AUDIT 8–19) 504 (16.4) 119 (15.9)
[OR: 2.09 (1.76–2.49), p < 0.001] and having a partner
High (AUDIT 20–40) 172 (5.6) 31 (4.1)
who is HIV positive [OR: 1.43 (1.10–1.84), p < 0.01].
Unknown N 39 8 Univariate analysis (see Table 2) shows that the follow-
Current tobacco use 799 (26.2) 203 (27.4) ing factors were more likely to be associated with adher-
ence to TB treatment: being in the 25–34 year age group
Naidoo et al. BMC Public Health 2013, 13:396 Page 6 of 10
http://www.biomedcentral.com/1471-2458/13/396

Table 2 Predictors of non-adherence to TB medication [OR: 0.40 (0.34–0.48), p < 0.001], and being on ART [OR:
(N = 3107) 0.70 (0.56–0.86), p < 0.001].
Socio-demographics Crude OR Adjusted OR Multivariable analysis (see Table 2) indicates that the
(95% CI) (95% CI)a,b following factors are more likely to predict TB treat-
Age 18–24 1.00 1.00 ment non-adherence: medium levels of poverty [OR:
Age 25–34 0.76 (0.59–0.97)* 0.93 (0.65–1.32) 1.73 (1.34–2.24), p < 0.001], high levels of poverty [OR:
Age 35–44 0.77 (0.60–1.00) 1.00 (0.68–1.45) 1.65 (1.14–2.39), having one (1) chronic condition
Age 45 and older 0.80 (0.61–1.05) 0.65 (0.43–1.00)
[OR: 1.86 (1.41–2.46), p < 0.001], having two (2) chronic
conditions [OR: 2.44 (1.68–3.56), p < 0.001], having
Female 1.00 1.00
three or more chronic conditions [OR: 2.37 (1.45–3.88),
Male 1.26 (1.07–1.48)** 1.09 (0.86–1.38) p < 0.001], medium risk for alcohol misuse [OR: 1.65
Grade 7 or less 1.00 1.00 (1.23–2.29), p < 0.001], high risk for alcohol misuse
Grade 8–11 0.88 (0.72–1.06) 1.07 (0.80–1.43) [OR: 3.06 (1.94–4.81), p < 0.001], tobacco use [OR: 1.35
Grade 12 or more 0.68 (0.55–0.85)*** 0.76 (0.54–1.07) (1.04–1.75), p < 0.05], and having a partner who is HIV
Poverty low 1.00 1.00
positive [OR: 1.43 (1.10–1.84), p < 0.01].
Multivariable analysis (see Table 2) indicates that the
Poverty medium 1.97 (1.62–2.410*** 1.73 (1.34–2.24)***
following factors are more likely to predict adherence to
Poverty high 4.01 (3.12–5.10)*** 1.65 (1.14–2.39)** TB treatment: perceiving health status as being ‘poor’
Health variables [OR: 0.44 (0.32–0.60), p < 0.001], perceiving health status
Perceived health status to be ‘good’ [OR: 0.50 (0.37–0.67), p < 0.001], and being
Excellent/Very good 1.00 1.00 HIV negative [OR: 0.44 (0.33–0.59), p < 0.001].
Good 0.36 (0.29–0.45)*** 0.50 (0.37–0.67)***
Predictors of non-adherence to dual therapy (anti-TB
Fair/Poor 0.24 (0.20–0.30)*** 0.44 (0.32–0.60)***
treatment and ART)
TB retreatment patient 1.00 – Univariate analysis (see Table 3) indicates that the variables
New TB patient 0.82 (0.67–1.00) significantly associated with non-adherence to dual therapy
HIV positive 1.00 1.00 were: being male [OR: 1.61 (1.20–2.16), p < 0.001], being
HIV negative 0.40 (0.34–0.48)*** 0.44 (0.33–0.59)*** in the medium poverty category [OR: 3.14 (2.05–4.79),
Chronic conditions
p < 0.001], being in the high levels of poverty category
[OR: 7.13 (4.36–11.66)], having three or more chronic
Zero 1.00 1.00
conditions [OR: 2.45 (1.33–4.5), p < 0.01], medium risk
One 1.54 (1.23–1.93)*** 1.86 (1.41–2.46)*** for alcohol misuse [OR: 1.65 (1.11–2.45), p < 0.05], high
Two 1.84 (1.34–2.51)*** 2.44 (1.68–3.56)*** risk for alcohol misuse [OR: 2.76 (1.30–5.86), p < 0.01], to-
Three or more 1.98 (1.34–2.92)*** 2.37 (1.45–3.88)*** bacco use [OR: 2.30 (1.65–3.19), p < 0.001], having a part-
Severe psychological distress 1.31 (1.09–1.57)** 0.94 (0.73–1.22) ner who is HIV positive [OR: 2.41 (1.75–3.32), p < 0.01],
(K ≥ 30) and having sex in the past three months [OR: 1.44 (1.07–
Alcohol: low risk (AUDIT 0–7) 1.00 1.00 1.93), p < 0.05].
Medium (AUDIT 8–19) 1.74 (1.42–2.19)*** 1.65 (1.23–2.29)*** The factors that are more likely to be associated with
adherence to dual therapy as indicated by the univariate
High (AUDIT 20–40) 2.42 (1.76–3.32)*** 3.06 (1.94–4.81***
analysis are: being in the 35–44 year age group [OR:
Current tobacco use 2.09 (1.76–2.49)*** 1.35 (1.04–1.75)*
0.68 (0.49–0.95), p < 0.05], having a grade 8 to 11 educa-
On ART 0.70 (0.56–0.86)*** 0.78 (0.60–1.03) tion [OR: 0.66 (0.47–0.93), p < 0.05], perceiving health
Partner HIV negative/unknown 1.00 1.00 status as being ‘poor’ [OR: 0.20 (0.14–0.29), p < 0.001],
Partner HIV positive 1.41 (1.16–1.70)*** 1.43 (1.10–1.84)** perceiving health status to be ‘good’ [OR: 0.41 (0.28–
*P < 0.05; ** P < 0.01; *** P < 0.001. 0.59), p < 0.001], being a new TB patient [OR: 0.51
a
Using “enter” LR selection of variables. (0.37–0.71), p < 0.001], and having a sex partner on ART
b
Hosmer and Lemeshow Chi-square 7.68, df8, 0.466; Cox and Snell R2 0.11;
Nagelkerke R2 0.16.
[OR: 0.51 (0.35–0.74), p < 0.001].
The results of the multivariate analysis as indicated
in Table 3 highlights the following significant predict-
[OR: 0.76 (0.59–0.97), p < 0.05], having a grade12 or ive factors for non-adherence to dual therapy in the
more education [OR: 0.68 (0.55–0.85), p < 0.001], per- sub-sample of TB patients co-infected with HIV were:
ceiving health status as being ‘poor’ [OR: 0.24 (0.20– medium poverty [OR: 2.60 (1.46–4.65), p < 0.001], high
0.30), p < 0.001], perceiving health status to be ‘good’ levels of poverty [OR: 3.89 (1.87–8.12), p < 0.001],
[OR: 0.36 (0.29–0.45), p < 0.001], being HIV negative having one (1) chronic condition [OR: 2.73 (1.31–5.65),
Naidoo et al. BMC Public Health 2013, 13:396 Page 7 of 10
http://www.biomedcentral.com/1471-2458/13/396

Table 3 Predictors of non-adherence to ART and anti-TB being ‘poor’ [OR: 0.20 (0.14–0.29), p < 0.001] and having a
medication (N = 757) sex partner on ART [OR: 0.50 (0.25–0.97), p < 0.05].
Socio-demographics Crude OR Adjusted OR
(95% CI) (95% CI)a,b Discussion
Age 18–34 1.00 1.00 The sub-sample of HIV positive patients on dual ther-
Age 35–44 0.68 (0.49–0.95)* 0.63 (0.38–1.03) apy for TB and HIV infection, in this study, shared a
Age 45 and older 1.12 (0.76–1.64) 0.90 (0.49–1.67) similar economic and alcohol misuse profile to the total
Female 1.00 1.00
sample of TB infected patients. In addition, nearly 40%
of patients on dual therapy reported having a partner
Male 1.61 (1.20–2.16)*** 1.49 (0.95–2.32)
who is HIV positive. Finally, the self-report of non-
Grade 7 or less 1.00 1.00 adherence to dual therapy is high in this sub-sample
Grade 8–11 0.66 (0.47–0.93)* 1.13 (0.67–1.93) (42.4%) but comparable to the figures reported in the
Grade 12 or more 0.79 (0.53–1.17) 1.19 (0.65–2.18) literature [2]. Non-adherence to both ART and anti-TB
Poverty low 1.00 1.00 drugs are of concern in this sub-group. However, given
Poverty medium 3.14 (2.05–4.79)*** 2.60 (1.46–4.65)***
the fact that this sub-group has a dual infection re-
quires that they take two sets of drugs, which could
Poverty high 7.13 (4.36–11.66)*** 3.89 (1.87–8.12)***
prove to be quite burdensome. The use of a multiple
Health variables drug regimen may also lead to drug reactions causing
Perceived health status unmanageable side-effects ultimately leading to treat-
Very good/good 1.00 1.00 ment default [28-30]. The complexity of the drug treat-
Good 0.41 (0.28–0.59)*** 0.69 (0.40–1.17) ment regimen and the impact on the daily lives of
Fair/Poor 0.20 (0.14–0.29)*** 0.28 (0.17–0.49)***
these patients are also factors that are associated with
poor adherence in these patients [29,30]. Patients on
TB retreatment patient 1.00 1.00
strict drug treatment programmes also have daily com-
New TB patient 0.51 (0.37–0.71)*** 0.67 (0.42–1.07) peting “life demands” associated with work and family.
Chronic conditions The four common predictive factors independently
Zero 1.00 1.00 associated with non-adherence to anti-TB drugs and to
One 1.39 (0.91–2.12) 1.72 (1.01–2.94)* dual therapy (i.e. ART and anti-TB drugs) were poverty,
Two 1.50 (0.87–2.58) 2.73 (1.31–5.65)**
having co-morbid disease conditions, being at risk for
alcohol misuse and having a HIV positive partner. This
Three or more 2.45 (1.33–0.45)** 5.33 (2.27–12.55)***
finding is supported by existing literature on the rela-
Severe psychological distress 1.00 (0.72–1.37) – tionship between the social determinants of health and
(K ≥ 30)
health outcome and/or the quality of life of individuals
Alcohol: low risk (AUDIT 0–7) 1.00 1.00
with one or more disease condition [31,32]. In this
Medium (AUDIT 8–19) 1.65 (1.11–2.45)* 1.08 (0.56–2.09) study non-adherent behavior for both anti-TB drugs
High (AUDIT 20–40) 2.76 (1.30–5.86)** 13.09 (2.96–57.99)* and ART was associated with a lack of economic re-
Current tobacco use 2.30 (1.65–3.19)*** 1.44 (0.85–2.43) sources (poverty) and negative personal circumstances
Partner HIV negative/ 1.00 1.00 (co-morbid conditions and a HIV positive partner).
unknown Clearly the lack of social, personal and economic re-
Partner HIV positive 2.41 (1.75–3.32)*** 3.12 (1.84–5.29)*** sources is a barrier to adherence. Unfortunately, the lack
Had sex in past 3 months 1.44 (1.07–1.93)* 1.48 (0.96–2.28)
of resources, in particular poverty, has historically been
associated with TB onset [33]. It stands to reason, there-
Sex partner on ART 0.51 (0.35–0.74)*** 0.50 (0.25–0.97)*
fore, that being poor is a barrier to health promoting be-
*P < 0.05; ** P < 0.01; *** P < 0.001.
a
Using “enter” LR selection of variables.
havior, such as adhering to a treatment programme,
b
Hosmer and Lemeshow Chi-square 13.14df8, 0.107; Cox and Snell R2 0.24; because individuals faced with economic restraints do
Nagelkerke R2 0.33. not have an enabling environment that facilitates behav-
ior that will lead to better health outcomes. Being poor,
p < 0.01], having three (3) or more chronic conditions having negative personal circumstances and engaging in
[OR: 5.33 (2.27–12.55), p < 0.001], high risk for alcohol risky behavior such as alcohol misuse is likely to lead to
misuse [OR: 13.09 (2.96–57.990), p < 0.05] and having poor health outcomes as a consequence of treatment de-
a partner who is HIV positive [OR: 3.12 (1.84–5.29), fault [13,34]. It is plausible that patient mis-use of alco-
p < 0.001]. hol, in particular, may have led to poor and impaired
Finally, the following predictive factors were associated judgement [35,36] when making health decisions, such
with adherence to dual therapy: perceiving health status as as not adhering to their anti-TB medication.
Naidoo et al. BMC Public Health 2013, 13:396 Page 8 of 10
http://www.biomedcentral.com/1471-2458/13/396

Of particular concern in this study is the fact that HIV treatment group and the dual treatment group. In the anti-
negative participants’ who were non-adherent to anti-TB TB treatment group health status perceived to be good is
drugs and in a relationship with a HIV positive partner, associated with adherence to treatment but this is not the
placed not only themselves but their partners at risk for case in the smaller group receiving dual treatment.
poor health outcomes. Inconsistent adherence to anti- Finally, the additional predictive factor for adherence
TB drugs may not only lead to MDR and XDR-TB but to dual therapy (i.e. anti-TB treatment and ART) was
may increase the risk of transmission of TB to others liv- having a sex partner on ART. Having a sex partner on
ing and working in close proximity. The HIV positive ART is in some way a protective factor for an individual
patients, in this study, who were not adhering to dual who is also on ART because of the sero-concordant na-
therapy, placed themselves at risk for drug resistance ture of the relationship. Both partners in this relation-
which could in turn reduce their life-span. ART, taken in ship are able to support each other in an open manner
accordance with treatment guidelines is known to pro- without fear of being stigmatized for being HIV positive
long an HIV infected person’s life and the availability of [38,39]. Non-disclosure of certain medical conditions,
ART has led health professionals to re-conceptualize such as having HIV or AIDS, is known to fuel the epi-
HIV and AIDS as chronic medical conditions. demic [40]. Consequently, one may assume that in
One additional factor, namely tobacco use, was inde- households where there is non-disclosure of HIV status,
pendently associated with non-adherence to TB treat- the adherence to ART will be non-existent, poor or in-
ment drugs. Individuals who smoke tobacco may be consistent for fear of one’s HIV positive status being re-
conceived of as having a personality with an “increased vealed [41]. In essence social support has a positive
risk profile” which makes them prone to engaging in un- association with medical treatment adherence [42].
desirable behaviours regardless of the consequences
[35,36]. It is not surprising, therefore, that participant’s Conclusion
adherence to anti-TB drugs in this study, was mediated A major limitation of this study is that it is cross-sectional.
by tobacco smoking. Ideally, the adherence patterns of TB patients and those
The common predictive factor independently associ- dually infected with HIV should be tracked over time in
ated with adherence to anti-TB drugs and to dual therapy order to better understand how the disease course(s) influ-
(i.e. ART and ant-TB drugs) was: perceiving health status ences the patient’s willingness to co-operate with the pre-
to be poor. This finding may seem counter-intuitive at scribed treatment. Despite this limitation, the results of
first glance. It is reasonable, however, that the partici- this study provide insight into the factors that are associ-
pants who perceived that their health status was poor on ated with non-adherence to treatment among TB patients
the basis that they had a “double-disease” burden of be- and those co-infected with HIV. Given the factors found to
ing infected with TB and HIV were more motivated to be associated with non-adherence in this study, a compre-
take the medication which they know to be an effective hensive treatment programme using a patient-centred
form of treatment. Knowledge of medical treatment effi- model is indicated for TB patients attending public health
cacy for a specific disease condition is known to influ- care clinics in SA [5]. This comprehensive treatment
ence the adherence behavior of the recipients of care [5]. programme being advocated should address structural fac-
Additional predictive factors for adherence to TB tors (such as poverty); risky behaviours (such as alcohol
treatment were: perceiving health status to be good and mis-use and tobacco use); and psycho-social well-being
being HIV negative. This finding is supported by the lit- (such as social relationship counseling on engaging with a
erature that reports on factors found to be influencing sero-discordant partner is indicated for TB patients [with
adherence to medical treatment regimens for commu- and without HIV]) and counseling related to perceived
nicable and non-communicable diseases. Individuals health status and living with co-morbid health conditions).
with a transient medical condition, that can be cured, The treatment care package needs to involve not only the
equipped with the knowledge that heeding to health care health sector but other relevant government sectors, such
practitioner’s suggested treatment regimens for cure are as social development.
more likely to adhere [37]. Of course there are numer-
ous other factors that may influence adherence to medi- Competing interests
cation for a curable disease, such as TB. These factors The authors declare that they have no competing interests.
include knowing someone who was cured, the need to
Authors’ contributions
maintain a good health status, and having a high level of PN and KP were primarily responsible for the conceptualization of the study,
health literacy (such as knowledge of disease transmis- for strategic guidance on the project and the first draft of this manuscript.
sion as in the case of TB) [7,32]. They were assisted by GM during the grant application phase and application
for ethical approval of the study. JL, GM, GM and BT were involved in the
The results indicate that there is a trend between operational aspects of the research process. All authors read and approved
perceived health status and adherence in the anti-TB the final manuscript.
Naidoo et al. BMC Public Health 2013, 13:396 Page 9 of 10
http://www.biomedcentral.com/1471-2458/13/396

Acknowledgements consumption among men and women with tuberculosis in Tomsk,

The Department of Health in South Africa funded this study through a Russia. Cent Eur J Public Health 2010, 18:132–138.
tender “NDOH: 21/2010 Implementation and monitoring of Screening and 16. World Health Organization (WHO): Treatment of Tuberculosis Guidelines. 2010.
Brief Intervention for alcohol use disorders among Tuberculosis patients” that Geneva, Switzerland: WHO; 2010.
was awarded to the HSRC. 17. South African National AIDS Council (SANAC) report: The HIV epidemic in
South Africa: What do we know and how has it changed? South Africa; 2011.
Author details Available from: www.hsrc.ac.za. Accessed August 2011.
1
Population Health, Health Systems and Innovation (PHHSI)/HIV/STIs and TB 18. Kessler R, Andrews G, Colpe LJ, Hiripi E, Mroczek DK, Normand S-LT, Walters EE,
(HAST) Research Programmes, Human Sciences Research Council, Pretoria, Zaslavsky AM: Short screening scales to monitor population prevalences and
Durban and Cape Town, South Africa, Private Bag X 9182, Cape Town 8000, trends in nonspecific psychological distress. Psychol Med 2002, 32:959. e976.
South Africa. 2Department of Psychology, University of the Western Cape, 19. Kessler RC, Barker PR, Colpe LJ, Epstein JF, Gfroerer JC, Hiripi E, Howes MJ,
Cape Town, South Africa. 3Department of Psychology, University of Limpopo, Normand S-LT, Manderscheid RW, Walters EE, Zaslavsky AM: Screening for
Turfloop, South Africa. 4ASEAN, Institute for Health Development, Mahidol serious mental illness in the general population. Arch Gen Psychiatry 2003,
University, Salaya, Thailand. 5Department of Nursing, University of KwaZulu 60(2):184.
Natal, Durban, South Africa. 20. Brooks RT, Beard J, Steel Z: Factor structure and interpretation of the K10.
Psychol Assess 2006, 18(1):62e70.
Received: 5 September 2012 Accepted: 23 April 2013 21. Andrews G, Slade T: Interpreting scores on the Kessler Psychological
Published: 26 April 2013 Distress Scale (K10). Aust N Z J Public Health 2001, 25:494–497.
22. Spies G, Kader K, Kidd M, Smit J, Myer L, Stein DJ, Seedat S: Validity of the
K-10 in detecting DSM-IV-defined depression and anxiety disorders
References
among HIV-infected individuals. AIDS Care 2009, 21(9):1163–1168.
1. World Health Organization: (WHO): Global Tuberculosis Control. WHO report
23. Babor TF, Higgins-Biddle JC: Brief Intervention for Hazardous and Harmful
2011. Geneva, Switzerland: WHO; 2011. http://www.who.int/tb/publications/
Drinking a Manual for Use in Primary Care. Geneva, Switzerland: World
global_report/2011/gtbr11_full.pdf. Accessed March 2012.
Health Organization; 2001.
2. Thompson MA, Mugavero MJ, Rivet Amico K, Cargill VA, Chang LW, Gross R,
Orrell C, Altice FL, Bangsberg DR, Bartlett JG, Beckwith CG, Dowshen N, 24. Freeborn DK, Polen MR, Hollis JF, Senft RA: Screening and brief intervention for
Gordon CM, Horn T, Kumar P, Scott JD, Stirratt MJ, Remien RH, Simoni JM, hazardous drinking in an HMO: effects on medical care utilization. J Behav
Nachega JB: Guidelines for improving entry into and retention in care Health Serv Res 2000, 27(4):446–453.
and antiretroviral adherence for persons with HIV: evidence-based 25. Saunders JB, Aasland OG, Amundsen A, Grant M: Alcohol consumption and
recommendations from an International Association of Physicians in AIDS related problems among primary health care patients: WHO collaborative
Care Panel. Ann Intern Med 2012, 156(11):817–833. http://www.annals.org. project on early detection of persons with harmful alcohol consumption.
Accessed March 2012. Addiction 1993, 88(3):349–362.
3. Gurung RAR: Health Psychology: A Cultural Approach. Belmont, CA: Wadsworth; 26. Myer L, Smit J, Roux LL, Parker S, Stein DJ, Seedat S: Common mental
2010. disorders among HIV-infected individuals in South Africa: prevalence,
4. Friedman HS: Health Psychology. New Jersey: Prentice Hall; 2002. predictors, and validation of brief psychiatric rating scales. AIDS Patient
5. Munro SA, Lewin SA, Smith HJ, Mark EE, Fretheim A, Volmink J: Patient Care STDS 2008, 22(2):147–158.
adherence to tuberculosis treatment: a systematic review of qualitative 27. Chishinga N, Kinyanda E, Weiss HA, Patel V, Ayles H, Seedat S: Validation of
research. Plos Med 2007, 4(7):e238. brief screening tools for depressive and alcohol use disorders among TB
6. McInerney PA, Nicholas PK, Wantland D, Corless IB, Ncama B, Bhengu B, and HIV patients in primary care in Zambia. BMC Psychiatry 2011, 11:75.
McGibbon CA, Davis SM, Gallagher DM: Characteristics of anti-tuberculosis 28. Kwara A, Flanigan TP, Carter EJ: Highly active antiretroviral therapy
medication adherence in South Africa. Appl Nurs Res 2007, 20(4):164–170. (HAART) in adults with tuberculosis: current status. Int J Tuberc Lung Dis
7. Naidoo P, Dick J, Cooper D: Exploring tuberculosis patient’s adherence to 2005, 9:248–257.
treatment regimens and prevention programs at a public health site. 29. Claxton AJ, Cramer J, Piece C: A systematic review of the associations
Qual Heal Res 2009, 19(1):55–70. between dose regimens and medication compliance. Clin Ther 2001,
8. Mindachew M, Deribew A, Tessema F, Biadgilign S: Predictors of 23:1296–1310.
adherence to isoniazid preventive therapy among HIV positive adults in 30. Ingersoll KS, Cohen J: The impact of medication regimen factors on
Addis Ababa, Ethiopia. BMC Public Health 2011, 11:916. adherence to chronic treatment: a review of the literature. J Behav Med
9. Peltzer K: Factors at follow-up associated with adherence with directly 2008, 31:213–224.
observed therapy (DOT) for tuberculosis patients in South Africa. 31. Nevarro V: What we mean by the social determinants of health. Int J Heal
J Psychol Afr 2001, 11(2):165–185. Serv 2009, 39(3):423–441. doi:10.2190/HS.39.3.a.
10. Corless IB, Wantland D, Bhengu B, McInerney P, Ncama B, Nicholas PK, 32. Marks DF, Murray M, Evans B, Willig C, Woodall C, Sykes CM: Health Psychology:
McGibbon C, Wong E, Davis SM: HIV and tuberculosis in Durban, Theory, Research & Practice. London: Sage; 2006.
South Africa: adherence to two medication regimens. AIDS Care 2009, 33. Farmer P: Infections and Inequalities. Berkeley: Univ of California Press; 1999.
21(9):1106–1113. 34. Muture BN, Keraka MN, Kimuu PK, Kabiru EW, Ombeka VO, Oguya F: Factors
11. Gebremariam MK, Bjune GA, Frich JC: Barriers and facilitators of adherence associated with default from treatment among tuberculosis patients in
to TB treatment in patients on concomitant TB & HIV treatment: a Nairobi province, Kenya: a case control study. BMC Publ Health 2011,
qualitative study. BMC Publ Health 2010, 10:651. doi:10.1186/1471-2458- 11:696.
10-651. 35. Morgenstern J, Langenbucher J, Labouvie E, Miller KJ: The co-morbidity of
12. Hirsch-Moverman Y, Daftary A, Franks J, Colson PW: Adherence to alcoholism and personality disorders in a clinical population: prevalence
treatment for latent tuberculosis infection: systematic review of studies rates and relation to alcohol typology variables. J Abnorm Psychol 1997,
in the US and Canada. Int J Tuberc Lung Dis 2008, 2:1235–1254. 106:74–84.
13. Amuha MG, Kutyabami P, Kitutu FE, Odoi-Adome R, Kalyango JN: Non- 36. Zuckerman M: Sensation Seeking: Beyond the Optimal Level of Arousal. New
adherence to anti-TB drugs among TB/HIV co-infected patients in Jersey: Lawrence Erlbaum; 1979.
Mbarara Hospital Uganda: prevalence and associated factors. Afr Heal Sci 37. Sluijs E, van Dulmen S, van Dijk L, de Ridder D, Heerdink R, Bensing J:
2009, 9(1):8–15. Patient adherence to medical treatment: a meta review. http://www.novel.nl
14. Santha T, Garg R, Frieden TR, Chandrasekaran V, Subramani R, Gopi PG, 2206. Accessed February 2013.
Selvakumar N, Ganapathy S, Charles N, Rajamma J, et al: Risk factors 38. Li L, Lee S-J, Thammawijaya P, Jiraphongsa C, Rotheram-Borus M: Stigma,
associated with default, failure and death among tuberculosis patients social support, and depression among people living with HIV in
treated in a DOTS programme in Tiruvallur District, South India. Int J Thailand. AIDS Care 2009, 21(8):1007–1013. doi:10.1080/09540120802614358.
Tuberc Lung Dis 2000, 6:780–788. 39. Kalichman SC, Simbayi LC, Cloete C, Ginindza T, Mthembu P, Nkambule T,
15. Shinn SS, Mathew TA, Yanova GV, Fitzmaurice GM, Livchits V, Yanov SA, Cherry C, Cain D: Measuring AIDS stigmas in people living with HIV/AIDS:
Strelis AK, Mishustin SP, Bokhan NA, Lastimoso CS, et al: Alcohol the internalized AIDS-related stigma scale. AIDS Care 2009, 21:87–93.
Naidoo et al. BMC Public Health 2013, 13:396 Page 10 of 10
http://www.biomedcentral.com/1471-2458/13/396

40. UNAIDS: UNAIDS report on the global AIDS epidemic 2012. Joint United
Nations Programme on HIV/AIDS (UNAIDS). www.unaids.org. Accessed
October 2012.
41. Ross E, Deverell A: Psychosocial Approaches to Health, Illness and Disability: A
Reader for Health Care Professionals. Pretoria: Van Schaik Publishers; 2004.
42. Corbett EL, Marston B, Churchyard GJ, De Cock KM: Tuberculosis in Sub-
Saharan Africa: opportunities, challenges and changes in the era of
antiretroviral treatment. Lancet 2006, 367:926–937.

doi:10.1186/1471-2458-13-396
Cite this article as: Naidoo et al.: Predictors of tuberculosis (TB) and
antiretroviral (ARV) medication non-adherence in public primary care
patients in South Africa: a cross sectional study. BMC Public Health 2013
13:396.

Submit your next manuscript to BioMed Central

and take full advantage of:

• Convenient online submission

• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution

Submit your manuscript at

www.biomedcentral.com/submit
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.