Mberi et al.

BMC Health Services Research (2015) 15:259

DOI 10.1186/s12913-015-0912-2

RESEARCH ARTICLE Open Access

Determinants of loss to follow-up in patients

on antiretroviral treatment, South Africa,
2004–2012: a cohort study
Mazvita Naome Mberi1,2,3,4*, Lazarus Rugare Kuonza1,2, Nomathemba Michelle Dube1, Cornelius Nattey5,6,
Samuel Manda7,8 and Robert Summers3,9

Abstract
Background: The number of Human Immunodeficiency Virus (HIV) infected people eligible for initiation on
antiretroviral Therapy (ART) is increasing. ART programmatic success requires that patients who are taking ART
remain on treatment and are followed up regularly. This study investigated factors associated with being lost to
follow-up, in a cohort of patients enrolled in a pharmacovigilance study in South Africa.
Methods: This was a retrospective observational cohort study performed at one of the Medunsa National
Pharmacovigilance Centre’s (MNPC) ART sentinel surveillance sites. Loss to Follow-up (LTFU) was defined as “a patient
who had been followed up at the sentinel site, who had not had contact with the health facility for 180 days or more
since their last recorded expected date of return or if there were 180 days or more between the expected date of
return and the next clinic visit”.
Results: Out of 595 patients, 65.5 % (n = 390) were female and 23.4 % (n = 139) were LTFU. The median time
on ART before LTFU was 21.5 months (interquartile range: 12.9 – 34.7 months). The incidence rate of LTFU
was 103 per 1000 person-years in the first year on ART and increased to 405 per 1000 person-years in the
eighth year of taking ART. Factors associated with becoming LTFU included not having a committed partner
(Adjusted Hazard Ratio (aHR): 2.9, 95 % Confidence Interval (CI):1.19-6.97, p = 0.019), being self-employed (aHR: 13.9,
95 % CI:2.81 - 69.06, p = 0.001), baseline CD4 count > 200 cells/ml (aHR: 3.8, 95 % CI: 1.85-7.85, p < 0.001), detectable last
known Viral Load (VL) (aHR: 3.6, 95 % CI:1.98 - 6.52, p < 0.001) and a last known World Health Organisation clinical stage
three or four (aHR: 2.0, 95 % CI:1.22-3.27, p = 0.006). Patients that previously had an ART adverse event had a lower risk
(aHR: 0.6, 95 % CI: 0.38 - 0.99, p = 0.044) of becoming LTFU than those that had not.
Conclusion: The incidence rate of LTFU increases with additional years on ART. Intensified measures to improve
patient retention on ART must be prioritised with increasing patient time on ART and in patients that are at increased
risk of becoming lost to follow-up.
Keywords: Loss to follow-up, Surveillance cohort, Medunsa National Pharmacovigilance Centre, Antiretroviral therapy

* Correspondence: [email protected]
1
South African Field Epidemiology Training Programme, National Institute for
Communicable Diseases of the National Health Laboratory Services, 1
Modderfontein Road, Monument Park 0105, Post net suite 179, P/bag
X27923 Sandringham, South Africa
2
School of Health Systems and Public Health, Faculty of Health Sciences,
University of Pretoria, Pretoria, Dr Savage Road 0084 Pinshof 349, Pretoria,
South Africa
Full list of author information is available at the end of the article

© 2015 Mberi et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://
creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Mberi et al. BMC Health Services Research (2015) 15:259 Page 2 of 11

Background impact of LTFU definitions recommended that the

Human immunodeficiency Virus (HIV) is a public health definition of LTFU should be based on the study out-
challenge worldwide. In 2013, the United Nations pro- come of interest, available encounter data and the cohort
gram on HIV/AIDS (UNAIDS) reported that there were visit schedule [22].
an estimated 35.3 million people living with HIV globally The objectives of this study were to determine the
[1]. The most affected region is Sub Saharan Africa, incidence of, and factors associated with, LTFU in a
which accounted for close to 70 % of all new HIV infec- cohort of patients enrolled in a Pharmacovigilance (PV)
tions in 2013 [2]. South Africa has the highest number sentinel site in Gauteng province, South Africa.
of people living with HIV (6.3 million) [3], and also has
the largest HIV treatment programme, with 1.8 million
persons reported to have been initiated on antiretroviral Methods
treatment (ART) since the commencement of the treat- Study setting
ment programme [4]. Though there has been an increase Tshepang clinic is an ART clinic located at Dr. George
in the scale up of access to ART in South Africa, there is Mukhari Hospital, one of the largest academic hospitals in
also an increasing number of people that are becoming Gauteng province. The clinic is one of three functional
eligible for ART as new ART eligibility criteria are being pilot sites of the Medunsa National Pharmacovigilance
introduced [4] and newly acquired HIV infections con- Centre’s (MNPC) ART surveillance cohort study. The
tinue to be reported in the population [1]. MNPC was opened in September 2004 as a National
ART, which continues to be the mainstay of HIV man- Department of Health (NDoH) initiative at the same time
agement, acts to prevent HIV multiplication and reduces when ART became available in the South African public
the viral load in the blood, resulting in improved im- health sector. The Centre operates a structured ART
mune function of an HIV-infected person and a decrease surveillance program which monitors and assesses the
in the risk of transmitting the virus [2, 5]. The acceler- impact and safety profile of ART medicines in a cohort
ated scale up of access to ART has led to a decline in of patients [23].
HIV related morbidity, mortality and new HIV infections One of the Centre’s main objectives is to determine
globally [1, 6]. As sustainability is becoming a priority in and reduce morbidity and mortality related to ART drug
the response to the AIDS epidemic, high loss to follow- use in HIV-infected patients [24]. From 2004, more than
up (LTFU) rates have the potential to reverse gains that 1000 patients were enrolled into the Medunsa National
have been achieved thus far. Pharmacovigilance cohort in Tshepang clinic. In a report
There is evidence that interrupting ART leads to inferior submitted to the NDoH in 2011, the MNPC noted that
clinical outcomes and higher risks of opportunistic compli- a large number of patients were LTFU at its sentinel
cations and death in patients [7]. However, studies in sites, but the incidence and risk factors of LTFU were
sub-Saharan Africa have shown that less than two-thirds of not known.
patients initiated on ART are on treatment 2 years after
ART initiation [8], and a systemic review showed a 70 %
24 month retention rate and a 65 % 36 month patient Study population and sampling
retention rate [9]. Furthermore, it has been shown that HIV-infected patients who were more than 14 years of
there is an increasing cumulative incidence of LTFU with age and were on ART were selected to participate in this
each year of taking ART in South Africa [10]. The South study using a systematic random sampling method at
African National Strategic Plan on HIV, Tuberculosis the PV sentinel site. Patients that signed informed
and Sexually Transmitted Infections (2012–2016) has a consent were enrolled into the PV study. A retrospective
goal not only to initiate at least 80 % of eligible patients case report form (CRF) was completed and submitted
on ART, but for at least 70 % of patients that are initi- for data capturing after which patients were followed up
ated on ART to be alive and on treatment 5 years after prospectively with each clinic visit.
initiation [11].
Though investigations have been done around factors
that possibly minimise the rates of LTFU [8, 12–16], few Study design
studies have been done to determine factors associated This was a retrospective observational cohort study.
with LTFU in South Africa [17–21]. Coupled with the Secondary data analysis of patients enrolled into the
challenge of limited research around LTFU, there are Medunsa PV cohort and being followed up at Tshepang
also several definitions of LTFU which vary across differ- clinic was performed. Chart reviews were done to
ent programs and studies. This means research findings include other variables that were not included on the
around LTFU may not, necessarily, remain the same if a standard PV CRF as well as to gather any missing infor-
different definition of LTFU is applied. A study on the mation in the PV database.
Mberi et al. BMC Health Services Research (2015) 15:259 Page 3 of 11

Definitions often outside of the sub-district in which they were

The working definition of LTFU in South Africa is being followed up. They are given a transfer letter for
“when 3 calendar months have passed without the further follow-up at their requested facility irrespective
patient having drugs in hand” [25]. LTFU, as defined by of whether they are clinically or virologically, well or
Chi et al. (2011), is when a patient has had 180 days or unwell [25].
more since the last clinic visit. This definition was found
to minimise misclassifications of patients and was rec- Statistical analyses
ommended for use to allow for comparison of programs Statistical analysis was done in STATA Corp version 12
[26]. In our study, a patient was classified as LTFU if the [28]. Patients that were LTFU, transferred out, who had
patient had been followed up at least once after ART died and those who are still in care at Tshepang clinic
initiation, but had not had contact with the clinic for were identified and described. Further analysis was then
180 days or more since their last recorded expected re- focused on the LTFU outcome. Patients’ data was cen-
turn date, or if there were 180 days or more between the sored on the 30th June 2012 and patients who were trans-
expected date of return and the next clinic visit. Patients ferred out were considered to be still in care at the time of
that were not LTFU were referred to as patients that are censoring. Patients who had missing files, an unknown
“in care” in this article. A maximum of 3 months initiation date or unknown last visit date or whose out-
(90 days) of treatment can be dispensed to a patient at come could not be determined were not included in the
one time in Tshepang clinic. Therefore, according to our analysis.
definition, patients that were classified as LTFU had not Medians, means and proportions were used to describe
had treatment for 3 (if three months of treatment had demographic characteristics and baseline characteristics.
been dispensed at the last clinic visit) to under 6 months We compared the differences in variables measured at
(if less than 1 month of treatment had been dispensed to baseline and after ART initiation using chi2 for pro-
the patient at the last visit). portions and non-parametric statistics for data that was
Virological suppression refers to an individual who has continuous and not normally distributed. Univariate com-
a fully suppressed viral load (VL), as evidenced by a VL parisons between proportions of patients that were classi-
level that is undetectable or a VL that is below 50 fied as LTFU and those not classified as LTFU was done
copies/ml. using Kaplan-Meier survival curves with a log-rank test
A person who has a partner includes people who are for equality of survival curves. Sex, age and factors that
married or are cohabiting with a life partner. A person had a p-value less than 0.20 in bivariate analysis were
without a partner includes people who are single, sepa- included in the multivariable cox regression that was
rated, divorced or who are a widow or widower and not performed to determine significant predictors of LTFU.
living with or married to a partner. The sex and age variables were included into the model as
An adverse event is “any untoward medical occurrence known priori. A p-value of 0.05 or less was considered to
that may appear during treatment with a pharmaceutical be statistically significant.
product, but which does not necessarily have a causal
relationship with the treatment” [20]. Ethical and legal considerations
WHO Stages three and four in adults and adolescents All the participants in the MNPC cohort signed
(ages of 15 years or more) are stages that indicate informed consent to be included in the PV surveillance
advanced HIV symptoms and Severe advanced HIV study. Participants that were less than 18 years of age
symptoms, respectively, in people who have been found were required to provide verbal assent and their legal
to be HIV infected, as illustrated by the presence of any representative or guardian gave written informed con-
clinical conditions or symptoms that are classified as sent for participation in the study. Ethical approval for
WHO stage three or WHO stage four conditions [27]. this study was received from the University of Pretoria’s
Patients were classified as dead using the South African Faculty of Health Sciences Research Ethics Committee.
Home affairs website which identifies death status at the
date and time of the enquiry. Retrospective information Results
cannot be retrieved on the website therefore date of death Description of the Tshepang pharmacovigilance cohort
could not be ascertained. Overall, 1117 patients were enrolled into the MNPC PV
Patients that are clinically stable on ART and virolo- cohort at Tshepang sentinel site, from September 2004
gically suppressed are “down referred” to primary care to June 2012. Fifty-three percent (595/1117) of the
facilities that are nearest to them, usually within the patients were included in the current study. The other
same sub-district, for patient follow-up [25]. 522 (47 %) patients were excluded either because the
Transferred patients are those patients that for any patient file could not be retrieved from the clinic records
reason request to be followed up at a different clinic, (n = 491) or there was either an unknown date of ART
Mberi et al. BMC Health Services Research (2015) 15:259 Page 4 of 11

initiation or unknown date of last clinic visit (n = 31) level of education (60.6 %), were unemployed (75.5 %),
(Fig. 1). There were no significant differences in the and had no committed partner (80.3 %). Baseline
distribution of sex (p = 0.334) and age (p = 0.891) vari- demographic characteristics of patients that were LTFU
ables in patients included in the study when compared versus those that remained in care were similar except
to those excluded from the analysis. with regards to employment status. A lower proportion of
Among the patients included in the analysis, 15.3 % the patients that remained in care were unemployed
(91/595) were still in care at Tshepang clinic, 61.3 % (73.6 %) compared to the patients that were LTFU
(365/595) had been down-referred or transferred to (81.0 %) (p = 0.028) (Table 1).
another clinic and 23.4 % (139/595) were LTFU. Of the
patients who were LTFU, 11.5 % (16/139) returned to Incidence of LTFU
care after missing more than 180 days of clinic appoint- The data of the 595 patients that we included in the study
ments and 10.6 % (63/139) were known to have died (Fig. 1). represented 1280 person-years of follow up time. The
The proportion of patients that died was higher in patients overall incidence rate of LTFU in the cohort was esti-
that never returned to care (48.8 %) compared to those that mated to be 109 per 1000 person-years (95 % Confidence
returned to care (18.8 %), (p = 0.023). (Refer to Fig. 1). The Interval (CI): 92–128). Among the patients LTFU, the
median time in care for patients that were down referred median duration on follow up was 21.5 months (inter-
or transferred was 614 days (Interquartile Range (IQR): quartile range (IQR): 12.9 - 34.7). Patient retention in care
416 to 843 days). The median time in care for patients was 81.8 % at 2 years and 54.7 % at 5 years in the cohort
that are in care was 1205 days (IQR: 851 to 1604 days) as illustrated in Fig. 2.
and median time in care for those patients that were Approximately 40 % (57/139) of the patients that
LTFU was 482 days (IQR: 251 to 1046 days). became lost to follow up were lost within the first year
of being initiated on ART. The absolute number of pa-
Baseline demographic characteristics of study participants tients that were LTFU with each year on ART appeared
Table 1 shows the baseline demographic and clinical to decrease over time. However, on further analysis, we
characteristics of the 595 patients that were included in found that the incidence rate of LTFU was 103 per 1000
the study. The median age at ART initiation was 35.9 years person-years in the first year on ART and increased to
(IQR: 30.1- 43.5), and 65.5 % were women. At ART initi- 405 per 1000 person-years in the eighth year of taking
ation the majority of the patients had achieved secondary ART (Fig. 3).

Patients enrolled into the Tshepang sentinel site

Pharmacovigilance cohort (N=1117)

Patient files not found (n=491)

Unknown date of ART initiation or last

clinic visit (n=31)

Patients included in study (N=595)

Transferred/down referred to another

clinic (n=365)
Still in care at Tshepang (n=91)

Loss to follow-up (n=139)

Returned to care (n=16) Never returned to care (n=123)

Alive (n=13) Dead (n=3) Alive (n=63) Dead (n=60)

Fig. 1 A flow diagram describing the patients enrolled into the Tshepang Pharmacovigilance cohort and their various outcomes, 2004–2012
Mberi et al. BMC Health Services Research (2015) 15:259 Page 5 of 11

Table 1 Baseline demographic characteristics of patients included in study, Tshepang Pharmacovigilance cohort, 2004-2012
Variable Total LTFU In care χ2
N % n (%) n % p-value
Sex Female 390 (65.5) 87 (62.6) 303 (66.5) 0.402
Age at ART initiation (years) ≥ 45 years 129 (21.7) 30 (21.6) 99 (21.7) 0.113
≥ 30 and < 45 320 (53.8) 66 (47.5) 254 (55.7)
< 30 146 (24.5) 43 (30.9) 103 (22.6)
Racial grouping Black 512 (99.4) 100 (99.0) 412 (99.5) 0.548
(80 missing) Other 3 (0.6) 1 (1.0) 2 (0.5)
Education level None 34 (6.2) 7 (5.4) 27 (6.5) 0.110
(50 missing) Primary 153 (28.1) 34 (26.2) 119 (28.7)
Secondary 330 (60.6) 77 (59.2) 253 (60.9)
Tertiary 28 (5.1) 12 (9.2) 16 (3.9)
Employment status Employed 141 (24.2) 24 (17.5) 117 (26.2) 0.028
(14 missing) Unemployed 439 (75.3) 111 (81.0) 328 (73.6)
Self employed 3 (0.5) 2 (1.5) 1 (0.2)
Live-in partner Present 115 (19.7) 19 (14.0) 96 (21.4) 0.057
(10 missing) Absent 472 (80.3) 117 (86.0) 353 (78.6)

Clinical characteristics of the participants Baseline CD4 cell counts were lower among patients
The baseline patient weight was lower (median = 55.6 kg) who later became LTFU (median = 79 cells/ml, IQR:
among patients who became LTFU, compared the baseline 30–146), compared to cell counts for patients who
weight (median = 57.2 kg) of patients who remained in remained in care (median = 87cells/ml, IQR: 36–150)
care (p < 0.001). The median patient weight increased (p < 0.001) (Table 2). This was the same for last known
from 55.6 kg at baseline to 58.5 kg at the last recorded CD4 count where median CD4 count was significantly
weight among patients LTFU (p < 0.001), and from 57.2 kg lower in patients that were LTFU (median = 197.5, IQR:
at baseline to 65.2 kg at the last clinic visit among patients 113.8-347.0 cells/ml) than in those that were not LTFU
who remained in care (p < 0.001). Median of the last (median = 275.0, IQR: 177.5-389.2cells/ml). The median
known weight was lower in patients that were LTFU CD4 counts increased from 79 cells/ml at baseline to
(58.5 kg) than those that were not (65.2 kg) (p < 0.001) 198 cells/ml at the last measured count among patients
(Table 2). LTFU (p < 0.001), and increased from 87 cells/ml at

Kaplan-Meier survival estimate

1.00
Probability of remaining in care
0.75
0.50
0.25
0.00

0 1 2 3 4 5 6 7 8
Time after ART initiation (years)
Fig. 2 Kaplan Meier graph showing the probability of remaining in care with time, Tshepang Pharmacovigilance cohort, 2004–2012
Mberi et al. BMC Health Services Research (2015) 15:259 Page 6 of 11

Period incidence ( per 1000 person-years)

60 450
400
50

Number of patients LTFU

350
40 300
250
30
200
20 150
100
10
50
0 0
1 2 3 4 5 6 7 8
Time period after ART initiation (years)
Number of patients LTFU

Period incidence of LTFU (per 1000 person-years)

Fig. 3 Number of patients and the period incidence of LTFU with each year of taking ART, Tshepang Pharmacovigilance cohort, 2004–2012

baseline to 275 cells/ml at last measured count among recorded VL measurement, 43.8 % (50/114) of the pa-
patients who remained in care (p < 0.001) (Table 2). tients that were LTFU had an undetectable VL compared
There was a lower proportion (121/135, 90 %) of to 85.8 % (376/438) of patients that remained in care
patients with CD4 count <200 cells/ml at baseline mea- (p < 0.001). The findings show a decrease in the percent-
surement in patients that became LTFU than in those age of patients that had undetectable VL levels among
that were not LTFU (427/452, 94.4 %) (p = 0.048). There patients that were LTFU (p = 0.012), while revealing no
was, however, a higher proportion (60/118, 50.8 %) significant changes in the percentage of patients with
of patients that had a last known CD4 count <200 undetectable VL among patients who remained in care
(p < 0.001) in patients that were LTFU (60/119, 50.4 %) (p = 0.114).
when compared patients that were not LTFU (43/441,
32.4 %) (Fig. 4).
Most (62 %) patients were in WHO clinical stages Treatment and related characteristics
three or four prior to ART initiation. At the initiation of Most patients in the cohort (581/591, 98.3 %) were initi-
ART, 63 % (80/127) of the patients that became LTFU ated on stavudine (d4t) containing regimens. Almost
were in WHO clinical stages 3 or 4, compared to 18 % (100/568) of patients reported that they drank alco-
61 % (255/418) of the patients that remained in care hol at the time of ART initiation. A total of 316/567
(p = 0.190). However, at last clinic visit before outcome, (55.7 %) patients in the cohort have experienced at least
53.6 % (67/125) of patients that were LTFU had a WHO one ART related adverse event. There were no differ-
clinical stage 3 or 4, compared to 26.0 % (100/384) of ences in the proportions of patients who were initiated
the patients that remained in care (p < 0.001) (Fig. 4). on a stavudine containing regimen (p = 0.466), drank
Sixty-one percent (61/100) of the patients that were alcohol at baseline (p = 0.852) or experienced an ART
LTFU had an undetectable VL when it was first mea- related adverse event during treatment (p = 0.057), when
sured after initiating ART compared to 81.9 % (353/431) comparing patients that were LTFU to those that
of patients that remained in care (p < 0.001). At last remained in care (Table 3).

Table 2 Characteristics of patients that were LTFU compared to those that were not LTFU, Tshepang Pharmacovigilance cohort, 2004-2012
Variable LTFU In care Rank sum p-value
Median IQR Median IQR
Baseline weight (kg) 55.6 46.8 - 62.6 57.2 51.0 - 66.0 <0.001
Last known weight (kg) 58.5 48.5 - 66.7 65.2 56.8 - 73.3 <0.001
Baseline CD4 count (cells/ml) 79.0 30.0 - 146.0 87.0 36.0 - 150.0 <0.001
Last known CD4 count (cells/ml) 197.5 113.8 - 347.0 275.0 177.5 - 389.2 <0.001
Mberi et al. BMC Health Services Research (2015) 15:259 Page 7 of 11

Fig. 4 Comparing CD4 counts, viral load levels and clinical staging in patients that were LTFU and that are in care, Tshepang Pharmacovigilance
cohort, 2004–2012

Close to a quarter (74/316, 23.4 %) of patients that (aHR = 13.92, 95 % CI: 2.81-69.06), having a baseline
had adverse events experienced more than one adverse CD4 count > 200 cells/ml (aHR = 3.81, 95 % CI: 1.85-7.85),
event. The most commonly reported adverse events having a history of an ART related adverse event prior to
included peripheral neuropathy (165/316, 52.2 %), lipo- outcome (aHR = 0.61, 95 % CI: 0.38-0.99), having a
dystrophy (121/316, 38.3 %) and hyperlactaemia (83/316, last known viral load that was detectable (aHR = 3.60,
26.3 %). Other ART related adverse events that were 95 % CI: 1.98-6.52), and having a last known WHO clinical
experienced included anaemia, various dermatological stage three or four (aHR = 2.00, 95 % CI: 1.22-3.27),
adverse events, central nervous system adverse events, (Table 5). Education level and alcohol intake status at ART
diarrhoea and vomiting, lipid abnormalities and hepa- initiation, regimens on which patients were initiated and
titis. The proportion of LTFU patients who experienced baseline WHO clinical staging were not associated with be-
peripheral neuropathy was 28.1 % (39/139), along with coming LTFU on univariate analysis and were not included
7.2 % (11/139) who experienced lipodystrophy, 9.4 % in the multivariable regression model (Table 5).
(13/139) experienced hyperlactataemia, and 31.6 %
(6/139) experienced other adverse events (Table 4). Discussion
We found a progressive increase in the incidence of
Factors associated with LTFU LTFU patients with each year after initiation of ART: the
In the multivariable Cox regression model, the risk fac- patient retention rate was 82 % after 2 years and 55 %
tors that were independently associated (p < 0.05) with after 5 years. This finding is similar to other studies
becoming LTFU were not having a committed partner done in sub-Saharan Africa [8–10]. It is, however,
(aHR = 2.88, 95 % CI 1.19-6.97), being self-employed important to keep in mind that different definitions of

Table 3 ART and adverse event related characteristics of patients that were LTFU compared to those that remained in care, Tshepang
Pharmacovigilance cohort, 2004-2012
Variable Total LTFU In care χ2
N % n (%) n % p - value
Initiating regimen d4t containing* 581 98.3 443 98.0 138 99.3 0.466
4missing Other 10 1.7 9 2.00 139 0.7
Alcohol No 468 82.4 107 82.95 361 82.2 0.852
27 missing Yes 100 17.6 22 17.05 78 7.8
Adverse events No 251 44.3 67 51.5 184 42.1 0.057
28 missing Yes 316 55.7 63 48.5 253 57.9
*d4t – stavudine
Mberi et al. BMC Health Services Research (2015) 15:259 Page 8 of 11

Table 4 ART related adverse events experienced by patients and proportions that had adverse events in patients that were LTFU
versus those still in care, Tshepang Pharmacovigilance cohort, 2004-2012
Adverse event Total LTFU In Care
N % n (%) n %
Peripheral Neuropathy 165 52.2 39 28.1 126 21.2
Lipodystrophy 121 38.3 11 7.2 110 24.1
Hyperlactateaemia 83 26.3 13 9.4 70 15.4
*Others 19 6.0 6 4.3 13 2.9
*Others include anaemia, various dermatological adverse events, Central nervous system adverse events, diarrhoea and vomiting, lipid abnormalities and
hepatititis

LTFU may have been used in these studies. Due to retaining patients in care at 2 years after ART initiation
limited research, we compare our study findings to the when compared to other sub-Saharan programs, we are
available research literature, bearing in mind that some not achieving this 70 % retention target at 5 years after
of these study findings may or may not have been diffe- ART initiation, and more effort is needed to improve
rent, if our definition of LTFU would have been applied patient retention in care at 5 years after ART initiation.
to their analysis. Our retention rate at 2 years gives the Our study, like other studies that have shown that be-
impression that our retention rates are at higher end of ing single is a risk factor for becoming LTFU [20], found
the rates that have been reported in some sub-Saharan that patients that did not have a committed partner had
ART programs (range from 24 to 77 % at 2 years) [8]. a higher risk of becoming LTFU. This suggests that
South Africa has a goal that at least 70 % of patients having a committed primary relationship is protective
that are initiated on ART are alive and on treatment against being LTFU. In our study we found that 80 % of
5 years after ART initiation [11]. Though we appear to patients did not have committed primary relationships.
have comparatively better programmatic success in It is therefore important that other systems and

Table 5 Cox Regression analysis to determine factors associated with LTFU, Tshepang Pharmacovigilance cohort, 2004-2012
Characteristic Univariate analysis Multivariate Analysis
HR 95 % CI P value aHR 95 % CI P value
Age at ART ≥ 45 Ref* Ref*
Initiation ≥ 30 and < 45 0.77 0.50 - 1.19 0.240 0.71 0.37 - 1.37 0.310
< 30 0.94 0.59 - 1.51 0.810 0.58 0.28 - 1.21 0.144
Partner status Present Ref* Ref*
Absent 1.75 1.08 - 2.84 0.024 2.88 1.19 - 6.97 0.019
Employment Employed Ref* Ref*
Status Self employed 7.18 1.69 - 30.54 0.008 13.92 2.81 - 69.06 0.001
Unemployed 1.57 1.01 - 2.44 0.045 1.75 0.86 - 3.56 0.123
Adverse events No Ref* Ref*
Yes 0.49 0.34 - 0.70 <0.001 0.61 0.38 - 0.99 0.044
Baseline CD4 CD4 ≤ 200 Ref* Ref*
Count CD4 > 200 2.47 1.42 - 4.31 0.001 3.81 1.85 - 7.85 < 0.001
First VL level Undetectable VL Ref* Ref*
after ART Detectable VL 1.81 1.21 - 2.72 0.004 1.00 0.58 - 1.73 0.998
Last known CD4 CD4 ≤ 200 Ref* Ref*
Count CD4 > 200 0.42 0.29 - 0.60 < 0.001 0.88 0.51 - 1.53 0.651
Last known VL Undetectable VL Ref* Ref*
Level Detectable VL 3.20 2.20 - 4.67 < 0.001 3.60 1.98 - 6.52 < 0.001
Last known WHO Stage 1 or 2 Ref* Ref*
clinical stage Stage 3 or4 3.20 2.24 - 4.57 < 0.001 2.00 1.22 - 3.27 0.006
*Ref - Reference category
Mberi et al. BMC Health Services Research (2015) 15:259 Page 9 of 11

programs that lower the rates of LTFU be employed. In- status (CD4 count) or actual disease progress, especially
tegrated patient care, community based adherence sup- where diagnostic tests for AIDS defining conditions are
port programs, active outreach to patients, the not readily available.
availability of adherence support services as well as facil- In our study we were able to demonstrate that patients
ity and community based patient adherence groups on that were not virologically suppressed (VL > 50 copies/ml)
ART have been found to lead to lower rates of LTFU at last clinic visit had a three times higher risk of
[12, 14–16, 19, 20]. These findings are well in support of becoming LTFU than those that had a suppressed VL.
the South African government’s call for integrated pa- Similarly, patients who had advanced clinical disease at
tient care and the nationwide rolling out of adherence last clinic visit (WHO clinical stage three or four) were
clubs/groups. It is also important for health facilities to twice as likely to become LTFU. Similar findings have
investigate facility specific factors that can impact LTFU been reported in other studies [17, 18]. It is important for
rates and implement programs that are tailored specific- measures to be put in place that improve retention rates
ally to reduce LTFU in their unique settings [19]. Though among patients whose viral loads remain high and/or who
other studies have also shown that young age, being male continue to display advanced clinical stages of HIV
and illiteracy are risk factors for becoming LTFU, an asso- disease, despite taking ART.
ciation between age, sex or level of education and becom- There is limited literature on the association between
ing LTFU was not found our study [9, 17, 20]. ART adverse events and LTFU. Just over half (56 %) of
The benefits of taking ART were clearly identified: we our cohort had a history of at least one ART related
found that there were significant increases in weight as adverse event since ART initiation. Less than a quarter
well as in immune function (evidenced by increasing of the patients who reported experiencing each of the
CD4 cell counts) from the time of ART initiation till common ART related adverse event became LTFU. We
the last known measurements both in patients that would have expected that patients experiencing adverse
were LTFU and in those that remained in care. Our events would likely be discouraged from taking their
findings showed that patients that became LTFU had ART. However we found that these patients had a lower
significantly lower weight and CD4 cell count measure- risk of becoming LTFU when compared to patients that
ments both at ART initiation and at last known meas- had not had ART related adverse events. This result may
urement. However, patient weight was not associated possibly be due to the perception that patients have of
with becoming LTFU. their health status, but further investigation is necessary.
There have been varying findings with regards to the The MNPC is the only structured surveillance cohort
association between baseline CD4 cell count and LTFU, that has been set up by the South African National
with some studies showing that patients that had CD4 Department of Health to follow up patients that have
counts between 301 cells/ml and 350 cells/ml had a been initiated on ART. As such, the cohort is the
lower risk of becoming LTFU than those with CD4 evidence base which provides insight into different as-
counts < 200 cells/ml [13], while others showed that pects of the National ART program. Our study did have
higher baseline CD4 cell count was associated with a limitations that should be taken into account when
higher risk of LTFU [12, 20, 21]. Our findings showed interpreting the findings. Firstly, we only analysed about
that patients that had a CD4 cell count of > 200 cells/ml 53 % of the patients that were enrolled into the PV
at ART initiation were at a higher risk of becoming surveillance cohort during the period of the study. This
LTFU. This finding could possibly be explained by the could have possibly introduced some selection bias into
fact that patients with a high CD4 cell count generally the study and could have led to an over- or under-
do not have the typical clinical symptoms of advanced estimation of the true incidence of LTFU in our study.
HIV disease and thus, may not perceive the benefits of We, however, found no significant differences in demo-
the strict patient follow-up schedule. However, we did graphic characteristics when we compared the patients
not find any association between baseline WHO clinical that we included in the study and those that were
staging and LTFU to further support this explanation. excluded. Secondly, we were unable to obtain the date of
That we did not find any significant association between death for some of the patients that died after being
baseline WHO staging and LTFU was unlike other studies LTFU, and thus were not able to confirm whether the
[17]. This could be explained by the fact that eligibility patients died 180 days or more after the last clinic
criteria for ART initiation may be based only on WHO appointment. It is therefore possible that some of these
staging in resource constrained countries, and may patients were misclassified as LTFU, resulting in an
account for the variable study findings regarding the over-estimation of the true incidence of LTFU. Thirdly,
association between LTFU and WHO clinical staging. we assumed that all the patients that were transferred
The use of WHO clinical staging as an ART eligibility from Tshepang clinic to other ART clinics were still in
criteria, may not necessarily correlate with immune care at the time of censoring. It is possible that some of
Mberi et al. BMC Health Services Research (2015) 15:259 Page 10 of 11

these patients became LTFU after being transferred, developing the data abstraction tool, data analysis and development of the
leading to an under-estimation of the rate of LTFU in article. RS is a Professor in pharmacology who was involved in formulating
the research protocol, developing the data abstraction tool, abstracting and
our study. Lastly, it is also likely that some of the analysing the data and development of the article. All authors read and
patients that we classified as LTFU could have self- approved the final manuscript.
transferred and continued ART at other sites without
notifying Tshepang clinic. This could have led to an Acknowledgements
We would like to acknowledge the Medunsa National Pharmacovigilance
over-estimation of the incidence rate of LTFU in our Centre, who permitted us to utilise the pharmacovigilance database and
study. Despite these limitations, we believe that our patient information for this study. The Centre also provided stationery and
study findings provide an important insight into the other resources required for communication and data management. The
University of Pretoria School of Health systems and Public Health provided
magnitude and the factors related to LTFU among funding for data abstraction. We also thank the South African Field
patients on ART. Epidemiology and Laboratory training program for providing funding for
transport and communication. Finally, we acknowledge Dr. Seymour
Williams, Rachel Eidex, Dorothy Southern and the staff of the Centre for
Conclusions Disease Control South Africa, for technical support in the development of
Our study shows that LFTU after ART initiation is the article.
high in South Africa, although patient retention in care
Author details
at 2 years after ART initiation appears to be higher than 1
South African Field Epidemiology Training Programme, National Institute for
in some ART programs in sub-Saharan Africa. The inci- Communicable Diseases of the National Health Laboratory Services, 1
dence rate of LTFU increases with additional years on Modderfontein Road, Monument Park 0105, Post net suite 179, P/bag
X27923 Sandringham, South Africa. 2School of Health Systems and Public
ART, and patient retention at 5 years is much lower than Health, Faculty of Health Sciences, University of Pretoria, Pretoria, Dr Savage
the target of at least 70 % that is in the South African Road 0084 Pinshof 349, Pretoria, South Africa. 3Medunsa National
NSP on HIV, STI’s and TB (2012–2016). Our findings Pharmacovigilance Centre, Medunsa Campus, University of Limpopo,
Pretoria, South Africa. 4Wits Reproductive Health and HIV Institute, 22 Esselen
will help ART clinicians to recognize patients that Street, Hillbrow, South Africa. 5National Institute for Occupational Health,
require additional support to remain in care, such as Smit Street, Braamfontein, South Africa. 6Global Partnership Initiated
patients with higher CD4 cell counts (> 200 cells/ml) Academia for the control of health threats, Bernard Nocht Institute for
Tropical Medicine, Bernad-Nocht Street 74, 20259 Hamburg, Germany.
at baseline, patients who do not have committed 7
Biostatistics Research unit, South African Medical Research Council, 1
partners and patients who continue to have detectable Soutpansberg Road, Pretoria, South Africa. 8Division of Epidemiology and
VL (> 50 copies/ml) or advance clinical disease (WHO biostatistics, School of Public Health, University of the Witwatersrand, 27 St
Andrews Road Parktown, Johannesburg, South Africa. 9Department of
clinical stages 3 or 4) after ART initiation. These Pharmacy, Faculty of Health Sciences, MEDUNSA Campus, University of
patients require additional support such as community Limpopo, Pretoria, South Africa.
based adherence support programs and the use of
Received: 4 September 2014 Accepted: 8 June 2015
reminder tools, which have been shown to reduce LTFU
in similar settings. There is also a need to intensify
surveillance activities in order to allow the country to References
accurately monitor rates of, and reasons for, LTFU in 1. UNAIDS. Global Health report: UNAIDS report on global AIDS epidemic.
2013. http://www.unaids.org/sites/default/files/media_asset/
patients that are on ART. UNAIDS_Global_Report_2013_en_1.pdf. Accessed 13 Jun 2014.
2. WHO. WHO fact sheet No 360. 2012. http://www.who.int/mediacentre/
Abbreviations factsheets/fs360/en/index.html. Accessed 14 Oct 2013.
ART: Antiretroviral Therapy; CI: Confidence Interval; CRF: Case Report Form; 3. UNAIDS: AIDS info Data. http://www.unaids.org/en/dataanalysis/datatools/
d4t : Stavudine; HIV: Human Immunodeficiency Virus; HR: Hazard Ratio; aidsinfo/. Accessed 12 May 2015.
aHR: Adjusted Hazards Ratio; IQR: Interquartile Range; LTFU: Loss To Follow-Up; 4. Republic of South Africa. Global AIDS Response: Progress Report 2012.
MNPC: Medunsa National Pharmacovigilance Centre; NDoH: National http://www.unaids.org/sites/default/files/country/documents//ce_ZA_
Department of Health; PV: Pharmacovigilance; UNAIDS: United Nations Narrative_Report.pdf. Accessed 5 Nov 2013.
program on HIV/AIDS; VL: Viral Load; WHO: World Health Organisation. 5. Wilson D, Cotton M, Bekker L, Meyers T, Venter F, Maartens G. Handbook of
HIV Medicine. 2nd ed. Capetown: Oxford university Press Southern Africa
Competing interests (Pty) Ltd; 2008.
The authors declare that they have no competing interests. 6. UNAIDS. World AIDS Day Report 2011: How to get to zero: Faster, Smarter,
Better. 2011. http://www.unaids.org/sites/default/files/en/media/unaids/
Authors’ contributions contassets/documents/unaidspublication/2011/
MNM is an HIV clinician and field epidemiologist who was involved in JC2216_WorldAIDSday_report_2011_en.pdf. Accessed 14 Oct 2013.
formulating the research protocol, extracting and cleaning 7. Kauffman GR, Elzi L, Weber R, et al. Interruption of cART limits CD4 T-cell
pharmacovigilance data, developing the data abstraction tool and recovery and increases the risk of opportunistic complications and death.
abstracting data, data analysis and development of the article. LRK is a AIDS. 2011;25(4):441–51.
medical epidemiologist was involved in formulating the research protocol, 8. Rosen S, Fox MP, Gill CJ. Patient retention in Antiretroviral programs in
developing the data abstraction tool, data analysis and development of the sub-Saharan Africa: a systematic review. PLoS Med. 2007;4(10):1691–701.
article. NMD is a field epidemiologist who was involved in formulating the 9. Rosen S, Fox MP. Patient retention in antiretroviral therapy programs up to
research protocol, extracting and cleaning pharmacovigilance data, 3 years on treatment in sub-Sahara Africa, 2007–2009: systematic review.
developing the data abstraction tool, data analysis and development of the Trop Med Int Health. 2010;15(1):1–15.
article. CN is a statistician who was involved in data analysis planning, data 10. Cornell M, Grimsrud A, Fairall A, et al. Temporal changes in programme
analysis, revision and review of the article. SM is a Professor in statistics, and outcomes among adult patients initiating antiretroviral therapy across SA,
was involved in formulating the research protocol, data analysis planning, 2002–2007. AIDS. 2010;24(14):2263–70.
Mberi et al. BMC Health Services Research (2015) 15:259 Page 11 of 11

11. South African National Aids Council. National strategic plan on HIV, STI’s
and TB, 2012–2016. http://www.thepresidency.gov.za/MediaLib/Downloads/
Home/Publications/SANACCallforNominations/A5summary12-12.pdf.
Accessed 13 Oct 2013.
12. Greig J, O’brien DP, Ford N, Spelman T, Sabapathy K, Shanks L. Similar
mortality and reduced loss to follow-up in integrated compared with
vertical programs providing antiretroviral treatment in sub-Saharan Africa.
J Acquir Immune Defic Syndr. 2012;59(5):e92–8. http://www.ncbi.nlm.nih.gov/
pubmed/22134144. Accessed 23 Sept 2013.
13. Clouse K, Pettifor A, Maskew M, et al. Initiating antiretroviral therapy when
presenting with higher CD4 cell counts resulted in reduced loss to follow-up
in a resource limited setting. AIDS. 2013;27:645–50.
14. Luque-Fernandez MA, Van Cutsem G, Goemaere E, et al. Effectiveness of
patient adherence groups as a model of care for stable patients on
antiretroviral therapy in Khayelitsha, Cape town, South Africa. PLoS One.
2013;8(2), e56088.
15. Fatti G, Meintjies G, Shea J, Eley B, Grimwood A. Improved survival and
antiretroviral treatment outcomes in adults receiving community-based
adherence support: 5 year results from a multicentre cohort in South Africa.
J Acquir Immune Defic Syndr. 2012;61:e50–8.
16. Lamb MR, El-Sadr WM, Geng E, Nash D. Association of Adherence support and
outreach services with total attrition, Loss to Follow-Up, and death among ART
patients in Sub Saharan Africa. PLoS One. 2012;7(6), e38443.
17. Nakiwogga-Muwanga A, Alamo-Talisuna S, Musaazi J, Kambugu A,
Ssekawungu P, Katabira E. Inadequate monitoring and advanced stges of
disease with lack of supportive counselling increases attrition among
patients on antiretroviral treatment at a large urban clinic in Uganda. J Int
Assoc Provid AIDS Care. 2013;13(6):547–54. doi:10.1177/2325957413501719.
18. Orrel C, Kaplan R, Wood R, Bekker L. Virological breakthrough: a risk factor
for loss to follow-up in a large community based cohort on antiretroviral
therapy. AIDS Research and Treatment. 2011;2011:469127.
doi:10.1155/2011/469127.
19. Wandeler G, Keiser O, Pfeifer K, Pestilli S, Fritz C, Labhardt ND. Outcomes of
antiretroviral treatment programs in rural Southern Africa. J Acquir Immune
Defic Syndr. 2012;59(2):e9–16. doi:10.1097/qa1.0b013e31823edb6a.
20. Alverex-Uria G, Naik PK, Pakam R, Midde M. Factors associated with attrition,
mortality, and loss to follow up after antiretroviral therapy initiation: data
from an HIV cohort study in India. Glob Health Action. 2013;6:21682.
21. Van Cutsem G, Ford N, Hildebrand K, Goemaere E, Mathee S, Abrahams S.
Correcting for mortality among patients lost to follow up on antiretroviral
therapy in South Africa: a cohort analysis. PLoS One. 2012;6(2):e1468. doi:
http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.
pone.0014684.
22. Shepherd BE, Blevins M, Vaz LME, et al. Impact of definitions of lost to
follow-up on estimates of retention, disease progression, and mortality:
application to an HIV program in Mozambique. Am J Epidemiol.
2013;178(5):819–28.
23. Dube NM. Database analysis of the pharmacovigilance surveillance study at
the Medunsa National Pharmacovigilance Centre: October 2006-March 2011.
Analysis report. South Africa: University of Pretoria, School of health systems
and public health; 2010.
24. Mkele G, Mayayise G, Summers R, O’Connor N, Gous H, Oosthuizen A.
Pharmacovigilance and its role in patient safety: a short course for health
care providers. Pretoria: University of Limpopo; 2010.
25. University of Capetown: Centre for Infectious diseases, Epidemiology and
Research. Capetown: Tier.Net reports; 2012. Accessed 10 Jan 2015.
26. Chi BH, Yiannoutsos CT, Westfall AO, Newman JE, Zhou J, Cesar C, et al., on
behalf of the International Epidemiologic databases to Evaluate AIDS
Collaboration. Universal Definition of Loss to follow-up in HIV treatment Submit your next manuscript to BioMed Central
programs: a statistical analysis of 111 facilities in Africa, Asia and Latin and take full advantage of:
America. Plos Medicine. 2011; doi:10.1371/ journal.pmed.100111.
http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal. • Convenient online submission
pmed.1001111. Accessed 18 May 2013
• Thorough peer review
27. WHO HIV/AIDS programme: Strengthening health systems to fight HIV/
AIDS. WHO case definitions of HIV for surveillance and revised clinical • No space constraints or color figure charges
staging and immunological classification of HIV-related disease in adults • Immediate publication on acceptance
and children. http://www.who.int/hiv/pub/guidelines/HIVstaging150307.pdf.
Accessed 23 Jan 2015. • Inclusion in PubMed, CAS, Scopus and Google Scholar
28. StataCorp 2011. Stata statistical software: release 12. TX; StataCorp LP: • Research which is freely available for redistribution
College Station; 2011.
Submit your manuscript at
www.biomedcentral.com/submit