Loss to Followup in HIV-Infected Patients from Asia-Pacific Region: Results from TAHOD.

Abstract

This study examined characteristics of HIV-infected patients in the TREAT Asia HIV Observational Database who were lost to follow-up (LTFU) from treatment and care. Time from last clinic visit to 31 March 2009 was analysed to determine the interval that best classified LTFU. Patients defined as LTFU were then categorised into permanently LTFU (never returned) and temporary LTFU (re-entered later), and these groups compared. A total of 3626 patients were included (71% male). No clinic visits for 180 days was the best-performing LTFU definition (sensitivity 90.6%, specificity 92.3%). During 7697 person-years of follow-up, 1648 episodes of LFTU were recorded (21.4 per 100-person-years). Patients LFTU were younger (P = 0.002), had HIV viral load ≥500 copies/mL or missing (P = 0.021), had shorter history of HIV infection (P = 0.048), and received no, single- or double-antiretroviral therapy, or a triple-drug regimen containing a protease inhibitor (P < 0.001). 48% of patients LTFU never returned. These patients were more likely to have low or missing haemoglobin (P < 0.001), missing recent HIV viral load (P < 0.001), negative hepatitis C test (P = 0.025), and previous temporary LTFU episodes (P < 0.001). Our analyses suggest that patients not seen at a clinic for 180 days are at high risk of permanent LTFU, and should be aggressively traced.

Full text

Page 1

Hindawi Publishing Corporation
AIDS Research and Treatment
Volume 2012, Article ID 375217, 10 pages
doi:10.1155/2012/375217
Clinical Study
Lossto FollowupinHIV-Infected Patients from
Asia-Pacific Region:Results fromTAHOD
JialunZhou,1Junko Tanuma,2Romanee Chaiwarith,3Christopher K. C.Lee,4
MatthewG.Law,1NagalingeswaranKumarasamy,5PraphanPhanuphak,6Yi-MingA.Chen,7
SasisopinKiertiburanakul,8FujieZhang,9Saphonn Vonthanak,10RossanaDitangco,11
Sanjay Pujari,12Jun Yong Choi,13Tuti Parwati Merati,14Evy Yunihastuti,15Patrick C. K. Li,16
Adeeba Kamarulzaman,17Van KinhNguyen,18Thi ThanhThuy Pham,19andPohLianLim20
1The Kirby Institute, The University of New South Wales, Sydney, NSW 2034, Australia
2National Center for Global Health and Medicine, Tokyo 162-8655, Japan
3Research Institute for Health Sciences, Chiang Mai 50200, Thailand
4Department of Medicine, Hospital Sungai Buloh, 47000 Kuala Lumpur, Malaysia
5YRG Centre for AIDS Research and Education, Chennai 600113, India
6HIV-NAT/Thai Red Cross AIDS Research Centre, Bangkok 10330, Thailand
7Taipei Veterans General Hospital and AIDS Prevention and Research Centre, National Yang-Ming University, Taipei 112, Taiwan
8Ramathibodi Hospital, Faculty of Medicine, Mahidol University, Bangkok 10400, Thailand
9Beijing Ditan Hospital, Capital Medical University, Beijing 100050, China
10National Center for HIV/AIDS, Dermatology and STDs, Phnom Penh, Cambodia
11Research Institute for Tropical Medicine, 1781 Manila, Philippines
12Institute of Infectious Diseases, Pune 411037, India
13Division of Infectious Diseases, Department of Internal Medicine, Yonsei University College of Medicine,
Seoul 120-752, Republic of Korea
14Faculty of Medicine, Udayana University and Sanglah Hospital, Bali 80233, Indonesia
15Working Group on AIDS, Faculty of Medicine, University of Indonesia/Ciptomangunkusumo Hospital, Jakarta 10430, Indonesia
16Department of Medicine, Queen Elizabeth Hospital, Hong Kong
17Faculty of Medicine, University of Malaya Medical Centre, 59100 Kuala Lumpur, Malaysia
18National Hospital for Tropical Diseases, Hanoi, Vietnam
19Department of Infectious Diseases, Bach Mai Hospital, Hanoi, Vietnam
20Department of Infectious Diseases, Tan Tock Seng Hospital, Singapore 308433
Correspondence should be addressed to
Matthew G. Law, mlaw@kirby.unsw.edu.au
Received 26 October 2011; Accepted 14 December 2011
Academic Editor: Anthony Harries
Copyright © 2012 Jialun Zhou et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This study examined characteristics of HIV-infected patients in the TREAT Asia HIV Observational Database who were lost to
follow-up (LTFU) from treatment and care. Time from last clinic visit to 31 March 2009 was analysed to determine the interval
that best classified LTFU. Patients defined as LTFU were then categorised into permanently LTFU (never returned) and temporary
LTFU (re-entered later), and these groups compared. A total of 3626 patients were included (71% male). No clinic visits for
180 days was the best-performing LTFU definition (sensitivity 90.6%, specificity 92.3%). During 7697 person-years of follow-up,
1648 episodes of LFTU were recorded (21.4 per 100-person-years). Patients LFTU were younger (P = 0.002), had HIV viral load
≥500copies/mL or missing (P = 0.021), had shorter history of HIV infection (P = 0.048), and received no, single- or double-
antiretroviral therapy, or a triple-drug regimen containing a protease inhibitor (P < 0.001). 48% of patients LTFU never returned.
These patients were more likely to have low or missing haemoglobin (P < 0.001), missing recent HIV viral load (P < 0.001),
negative hepatitis C test (P = 0.025), and previous temporary LTFU episodes (P < 0.001). Our analyses suggest that patients not
seen at a clinic for 180 days are at high risk of permanent LTFU, and should be aggressively traced.

Page 2

2AIDS Research and Treatment
1.Introduction
Loss to followup (LTFU) in patients receiving antiretroviral
therapy can cause serious consequences such as discontinua-
tion of treatment and increased risk of death [1–3]. At a pro-
gram level, LTFU can make it difficult to evaluate outcomes
of treatment and care [4, 5]. In resource-limited settings,
where treatment has become rapidly available following the
rollout of antiretroviral therapy, LTFU presents even more
challenging obstacles that require special consideration and
approaches [6, 7].
One of the key questions in patient followup is how to
defineapatientasLTFU.Thishasvariedinstudiesconducted
in different settings [8–10]. Defining LTFU using a very early
threshold,forexample,apatientwithnoclinicvisitinthelast
three months, may result in many patients being considered
as LTFU who would return to clinic naturally at a later date.
Defining LTFU with a long threshold, for example, one year,
maymeandelayingtoolongbeforeanyeffortismadetotrack
patients potentially at risk of LTFU.
The majority of research into LTFU in HIV-infected
patients receiving antiretroviral treatment in resource-
limited settings has been conducted in the sub-Saharan
Africa region [3, 10–13]. A few studies have been conducted
among Asian, mostly female, patients [14–16]. Using
data from the TREAT Asia HIV Observational Database
(TAHOD), this study was carried out to find the best-
performing definition of LTFU and examine the characteris-
tics of HIV-infected patients from the Asia-Pacific who were
LTFU from treatment and care.
2.Methods
Established in 2003, TAHOD is a collaborative observational
cohort study involving 18 sites in the Asia-Pacific region (see
Acknowledgement). Detailed methods have been published
previously [17]. Briefly, each site recruited approximately
200–300 HIV-infected patients, with recruitment based on
a consecutive series of patients regularly attending a given
site from a particular start-up time. Ethical approval for the
study was obtained from the University of New South Wales
Ethics Committee, Western Institutional Review Board,
and respective local ethics committee from each TAHOD
participating site.
The following data were collected: patient demographics
and baseline data, CD4 and CD8 count, HIV viral load,
prior and new AIDS defining illness (ADI), date and cause of
death, prior and current prescribed antiretroviral treatment
(ART), and reason for treatment change. Data were collected
according to a common protocol. Upon recruitment, all
available data prior to entry to TAHOD (considered as
retrospective data) were extracted from patient case notes.
Prospectivedatawereupdatedsix-monthlyateachclinic and
transferred to the data management centre for aggregation
and analyses in March and September each year. TAHOD
sites were encouraged to contact patients who have not been
seen in the clinics in the previous 12 months.
TAHOD data submitted at March 2009 and March 2010
were used to find the best-performing definition of LTFU.
TAHOD patients who had no followup after recruitment
were not included in this analysis. Patients who were not
seen in clinic for more than 12 months prior to the March
2010 data submission (i.e., last clinic visit prior to March
2009) were considered to be truly LTFU. The days between
the last clinical visit and 31 March 2009 in the March 2009
data transfer were then used to find the interval that best
classified a true LTFU in the following way. A series of
cutoffs were considered, from ten to 365 days, to define
patients as potentially LTFU. Each of these definitions of
potential LTFU was compared with the gold standard of true
LTFU, defined as no patient followup in the 12 months prior
to 31 March 2010. The sensitivity and specificity of each
cutoff in identifying true LTFU were calculated, and the best
performingcutoffidentifiedusingtheareaunderthereceiver
operator characteristic (ROC) curve. The optimal definition
of LTFU identified in terms of maximising the sensitivity
and specificity of true LTFU was found to be 180 days (see
Results). This definition was then used in the risk factor
analyses that follow.
Followup started from the last clinic visit at the March
2007 data submission. Patients who were considered LTFU
before March 2007 (i.e., had no clinic visits 180 days before
31March2007)wereexcludedfromtheanalysis.Forpatients
enrolled after March 2007, the followup started at the time of
enrolment. In terms of calculating person-years of followup,
the end of followup for patients who had no clinic visit for
180 days and so were considered as LTFU was defined as 90
days after their last clinic visit. For patients not considered
LTFU, the end of followup was also defined as 90 days after
their last clinic visit. If a patient died, the followup was
censored on the date of death if the date was within 180
days of their last clinic visit. Patients who died after March
2007wereconsideredtohavecompletefollowup.Itshouldbe
noted that patients who were considered LTFU could return
to clinic and reenter followup. The start of this reentry to
followup was defined as 3 months prior to the first clinic
visit that reinitiated followup. The patients that reentered
followup could also be re-LTFU if the patient subsequently
did not attend clinic for more than 180 days. The definitions
we adopted were consistent with those in a previous study
[18].
The rates of LTFU were calculated by the number of
total LTFU periods divided by the total duration of followup
contributed by the patients included in the analysis [18].
Because of the reentering and re-LTFU, patients could
contribute more than one episode of LTFU in this analysis.
Therateswerefurthercalculatedindifferentstrata,including
age, sex, exposure category, hepatitis B and C infection, year
since HIV infection, calendar year, the latest CD4 count
and viral load, antiretroviral treatment status, CDC disease
stages, prophylaxis (coded as receiving or not), and haemo-
globin level, all taken at the start of each episode.
Factors associated with LTFU were assessed by multivari-
ate Poisson regression models, using generalised estimating
equations, to allow for multiple events of LTFU in the same
patients.CD4count,HIVviralload,antiretroviraltreatment,
AIDS diagnosis, and haemoglobin tests were included as

Page 3

AIDS Research and Treatment3
Table 1: Receiver operating characteristic (ROC) analysis for the best-performing definition for loss to followup.
Cutoff
(days)
10
20
30
40
50
60
70
80
90
100
120
150
Sensitivity (%)Specificity (%)
Area under
ROC
58.32
61.67
64.68
68.36
72.93
76.48
80.40
83.19
86.16
87.57
89.71
90.85
Cutoff (days)Sensitivity (%)Specificity (%)
Area under
ROC
90.87
91.04
91.34
91.41
91.38
91.17
90.98
90.96
90.26
89.99
89.52
88.36
99.67
99.02
98.05
96.82
96.34
95.77
95.28
95.11
94.71
94.22
93.24
91.53
16.97
24.32
31.31
39.90
49.52
57.20
65.52
71.26
77.62
80.91
86.18
90.17
160
170
175
180
185
190
200
210
240
270
300
365
90.96
90.64
90.64
90.55
90.23
89.33
88.52
87.79
85.26
83.55
82.00
78.99
90.77
91.44
92.05
92.26
92.53
93.01
93.44
94.13
95.25
96.43
97.04
97.73
True LTFU defined as no patient followup in the 12 month prior to 31 March 2010. Each cutoff used as a potential definition of LTFU was the days between
last clinical visit and 31 March 2009 in the March 2009 data transfer. The sensitivity and specificity of each cutoff in identifying true LTFU were calculated,
and the optimal cutoff identified based on ROC analysis.
time-dependent variables and updated at the time the new
measurement or diagnosis was available.
Patients who had at least one episode of LTFU were
then categorised into two groups: those who had no more
clinical visits in the database (permanently LTFU) and those
who later reentered followup (temporary LTFU). Multivari-
ate logistic regression models were used to compare the
characteristics in patients who were considered permanently
LTFU with those who were temporary LTFU. All covariates
were taken at the end of the episode in patients with truly
LTFU or at the end of the first episode in patients considered
temporary LTFU.
Multivariate models were built using a forward-stepwise
approach. The final model included covariates that remained
significantattheP < 0.05level.Nonsignificantvariableswere
alsopresentedandadjustedforthefinalmultivariatemodels.
Data management and statistical analyses were performed
using SAS for Windows (SAS Institute Inc., Cary, NC, USA)
and Stata (StataCorp, STATA 10.1 for Windows, College
Station, TX, USA).
3.Results
InMarch2007,therewere2565patientsinthedatabase.1061
patients were subsequently enrolled in TAHOD up to March
2010. A total of 3626 patients from TAHOD who had follow-
up visits in the clinic were included in this analysis. During
the study period (from March 2007 to March 2010), there
were 54 patients who died and considered to have complete
followup.
Using days between last clinic visit and 31 March 2009 in
the March 2009 data transfer, we identified the interval that
best classifies a true LTFU (i.e., no clinic visit after 31 March
2009). An interval of 180 days was determined as the best-
performing definition (Table 1, sensitivity 90.6%, specificity
92.3%). Using 180 days as the LTFU cutoff, during 7697
person-years of followup, a total of 1648 episodes of LTFU
from 1298 patients were identified, giving a crude LTFU rate
of 21.4 per 100 person-years (95% confidence interval, CI,
20.4to22.5).Ofthose1648episodesofLTFUidentifiedusing
180 days as the cutoff, 48% were considered permanently
LTFU (i.e., the patient did not return to clinic before 31
March 2010), corresponding to 45% of the 1298 patients.
The patient characteristics are summarised in Table 2.
The majority of patients were male (71%), aged between 36
and 45 years (40%), and reported heterosexual transmission
(64%). Chinese (27%), Thai (26%), and Indian (11%) were
the main ethnic groups. At recruitment, approximately 12%
didnothaveaCD4counttest,andofthosetested,themajor-
ity had a CD4 count more than 200cells/µL. Nearly half
(45%)didnothaveanHIVviralloadtest,andofthosetested,
the majority were below 500copies/mL. Close to half of the
patients (46%) were diagnosed with an AIDS defining illness
atrecruitment,withtuberculosisbeingthemainillness.Most
patients (63%) had been reported to be diagnosed with HIV
for less than 6 years when recruited to TAHOD (measured
as the time from first reported positive HIV test). Less than
10% of the patients were coinfected with either hepatitis B
or hepatitis C. At recruitment, the majority of patients had
normal haemoglobin level. At the start of study followup,
most of the patients were on antiretroviral therapy including
three or more drugs in combination including at least one
nucleoside reverse transcriptase inhibitor (NRTI) and one
nonnucleoside reverse transcriptase inhibitor. Over 20% of
patients were in a combination with at least one NRTI and a
proteaseinhibitor(PI).Allpatientswerereceiving,orstarted,
antiretroviral therapy during followup.
Table 3 summarises univariate and multivariate analyses
of factors associated with LTFU using 180 days as cut-off. In
univariate analyses, the rate of LTFU was significantly lower
in patients with a current CD4 counts above 200cells/µL
comparedtopatientswithaCD4countlessthan100cells/µL,
but this was not significant in the final multivariate model.
In the final multivariate model (Table 3), factors associated

Page 4

4 AIDS Research and Treatment
Table 2: Patient characteristics.
Total
Characteristics
Sex
Male
Female
Current age (years)
≤35
36–45
46+
Reported exposure
Heterosexual contact
Homosexual contact
Injecting drug use
Other/unknown
Ethnicity
Chinese
Indian
Thai
Other/unknown
Baseline CD4 count (cells/µl.)
≤100
101–200
201+
Missing
Baseline HIV RNA (copies/ml)
≤500
501+
Missing
CDC disease stage at baseline
Stage A
Stage B
Stage C
Tuberculosis diagnosis at baseline
No
Yes
Time since HIV infection (years)
≤5
6+
Missing
Hepatitis B infection
No
Yes
Not tested
Hepatitis C infection
No
Yes
Not tested
Anemia at baseline
No
Yes
Haemoglobin not tested
3626
Number%
2567
1059
71
29
1383
1449
794
38
40
22
2337
749
263
277
64
21
7
8
989
390
933
1314
27
11
26
36
239
406
2531
450
7
11
70
12
1482
379
1765
41
10
49
1621
321
1684
45
9
46
2758
868
76
24
2295
1246
85
63
34
2
2297
257
1072
63
7
30
2007
324
1295
55
9
36
2480
597
567
68
16
16
Table 2: Continued.
Total
Characteristics
Antiretroviral treatment at baseline
3 + (NRTI + NNRTI)
3 + (NRTI + PI)
No/mono/double drug
3 + (other combination)
3626
Number%
2224
744
583
75
61
21
16
2
Anemia: haemoglobin <13g/dl (male), <11g/dl (female); NRTI: nucleoside
reverse transcriptase inhibitor; NNRTI: nonnucleoside reverse transcriptase
inhibitor; PI: protease inhibitor.
with LTFU included age (younger patients had higher rate
of LTFU), current HIV viral load (either patients with
HIV viral load ≥500 copies/mL or no tests in recent 180
days had higher rate of LTFU), history of HIV infection
(patients with shorter history of HIV infection had higher
rate of LTFU), hepatitis C infection (patients with positive
hepatitis C antibody had higher rate of LTFU), and, finally,
current combination of antiretroviral treatment (compared
topatientsontriple-drugregimenwithatleastoneNRTIand
one NNRTI, patients receiving no-, single-, or double-drug
antiretroviral therapy, or a triple-drug regimen containing at
least one NRTI and one PI, had higher rate of LTFU).
Table 4 shows factors that predict permanent LTFU
among patients who had no clinic visit for 180 days and so
met our optimal definition of LTFU. In the final multivariate
model, patients permanently LTFU were more likely to be
older, have not been anemic, have no recent HIV viral load
test, have tested negative for hepatitis C infection or have
never tested for hepatitis C, and have had more than one
episode of previous temporary LTFU.
4.Discussion
We found that an interval of 180 days between clinic
visits was the best-performing definition of LTFU based
on sensitivity and specificity in identifying true LTFU. By
this definition, we observed that approximately one in five
patients in our cohort would miss clinic visits for more
than 180 days and so become defined as LTFU. Among
these patients in our cohort close to half eventually returned
to followup, with half becoming truly lost to HIV-related
treatment and care.
The 180-day cutoff has been used by other studies as
a working definition of LTFU [10, 19–21]. Other intervals
have also been proposed as measurements of classifications
of LTFU, such as 90 days [8] and 365 days [9]. Regional- and
cohort-dependent characteristics, such as scheduled clinic
visits, patient burden, and drug availability could result in
specific intervals that best categorise patients at risk of LTFU.
Nevertheless, a 180-day (or 6-month) cutoff is an appealing
and easy-to-apply definition that could be used in different
clinical settings in the Asia-Pacific region to flag patients at
risk of being permanently lost to treatment and care. Our
analyses suggest patients with no clinic visits for six months

Page 5

AIDS Research and Treatment5
Table 3: Factors associated with permanent or temporary LTFU, defined as no clinic visit for 180 days, among all patients under followup.
Person-
years
Number
LTFU
Crude Rate1
Adjusted
95% CIIRR2
95% CI
P
value
IRR2
95% CI
P
value
Sex
Male
Female
Current age (years)
≤35
36∼45
46+
Reported exposure
Heterosexual
contact
Homosexual
contact
Injecting drug use
Other/unknown
Current CD4 count (cells/µl.)
≤100
101–200
201+
Missing
Current HIV RNA (copies/ml)
≤500
501+
Missing
CDC disease stage
Stage A
Stage B
Stage C
Tuberculosis diagnosis
Yes
No
Time since HIV infection (years)
≤5
6+
Missing
Hepatitis B infection
Yes
No
N/A
Hepatitis C infection
Yes
No
N/A
Current anemia (male < 13g/dl, female < 11g/dl)
Yes
No
N/A
5468.1
2229.2
1206
442
22.06
19.83
(20.85, 23.34) 1.00
(18.06, 21.77) 1.10
1.00
1.04(0.98, 1.24)0.090 (0.93, 1.17)0.446
2210.4
3320.2
2166.6
575
718
355
26.01
21.62
16.39
(23.97, 28.23) 1.00
(20.10, 23.27) 0.82
(14.77, 18.18) 0.69
1.00
0.89
0.0023
0.050
<0.001
(0.74, 0.92)
(0.60, 0.79)
0.001
<0.001 0.76
(0.79, 1.00)
(0.66, 0.88)
5144.598519.15 (17.99, 20.38) 1.00 1.00
1707.2 344 20.15(18.13, 22.40) 1.10 (0.93, 1.29)0.2751.05 (0.89, 1.25)0.540
344.3
501.3
125
194
36.31
38.70
(30.47, 43.27) 1.21
(33.62, 44.55) 1.64
(0.97, 1.51)
(1.37, 1.98)
0.098
<0.001 1.56
1.10(0.86, 1.40)
(1.29, 1.88)
0.437
<0.001
233.7
635.7
6327.6
500.3
69
136
1181
262
29.52
21.40
18.66
52.37
(23.32, 37.38) 1.00
(18.09, 25.31) 0.92
(17.63, 19.76) 0.75
(46.40, 59.11) 1.18
1.00
0.96
0.79
0.99
(0.68, 1.22)
(0.58, 0.96)
(0.90, 1.55)
0.551
0.023
0.235
(0.72, 1.29)
(0.61, 1.02)
(0.74, 1.31)
0.800
0.071
0.922
4213.7
537.1
2946.4
679
158
811
16.11
29.42
27.52
(14.95, 17.37) 1.00
(25.17, 34.38) 1.71
(25.69, 29.49) 1.75
1.000.0213
0.026
<0.001
(1.43, 2.04)
(1.55, 1.98)
<0.001 1.24
<0.001 1.64
(1.03, 1.51)
(1.45, 1.86)
3205.1
801.6
3690.5
828
118
702
25.83
14.72
19.02
(24.13, 27.65) 1.00
(12.29, 17.63) 0.93
(17.67, 20.48) 0.84
1.00
0.95
0.92
(0.76, 1.14)
(0.75, 0.93)
0.507
0.001
(0.77, 1.17)
(0.82, 1.02)
0.623
0.125
1806.7
5890.6
372
1276
20.59
21.66
(18.60, 22.79) 1.00
(20.51, 22.88) 1.04
1.00
0.98 (0.92, 1.18) 0.537(0.87, 1.12) 0.801
3477.2
4115.7
104.3
785
844
19
22.58
20.51
18.21
(21.05, 24.21) 1.00
(19.17, 21.94) 0.84
(11.61, 28.55) 0.58
1.00
0.89
0.49
0.0053
0.048
0.004
(0.75, 0.94)
(0.36, 0.94)
0.002
0.027
(0.79, 1.00)
(0.30, 0.79)
584.5
5101.9
2010.8
112
883
653
19.16
17.31
32.48
(15.92, 23.06) 1.00
(16.20, 18.49) 0.93
(30.08, 35.06) 0.98
1.00
0.90
1.07
(0.76, 1.13)
(0.80, 1.21)
0.474
0.859
(0.74, 1.10)
(0.85, 1.35)
0.319
0.548
541.4
4692.8
2463.0
149
796
703
27.52
16.96
28.54
(23.44, 32.31) 1.00
(15.82, 18.18) 0.81
(26.51, 30.73) 0.75
1.00
0.81
0.77
0.0303
0.034
0.008
(0.67, 0.98)
(0.62, 0.91)
0.029
0.004
(0.67, 0.98)
(0.63, 0.93)
1021.1
5771.6
904.5
155
1157
336
15.18
20.05
37.15
(12.97, 17.77) 1.00
(18.92, 21.24) 1.09
(33.38, 41.34) 1.31
1.00
1.11
1.09
(0.92, 1.30)
(1.07, 1.59)
0.302
0.008
(0.94, 1.32)
(0.89, 1.34)
0.227
0.382

Page 6

6 AIDS Research and Treatment
Table 3: Continued.
Person-
years
Number
LTFU
Crude Rate1
Adjusted
95% CI IRR2
95% CI
P
value
IRR2
95% CI
P
value
Current ART4
3 + (NRTNRTI)
3 + (NRTI + PI)
No/mono/double
ARV
3 + (other
combination)
4830.8
1898.3
942
377
19.50
19.86
(18.29, 20.79) 1.00
(17.95, 21.97) 1.21
1.00
1.22
0.0013
0.003 (1.06, 1.38)0.005(1.07, 1.39)
762.7300 39.33 (35.12, 44.05) 2.18 (1.90,2.50)
<0.001 1.92 (1.66, 2.22)
<0.001
205.42914.12(9.81, 20.32)0.95(0.65, 1.38)0.7861.01(0.69, 1.47)0.975
(1) Crude rate, per 100 person-years.
(2) Stratified by TAHOD sites.
(3) Overall for test for trend (ordinal categorical covariates) or for homogeneity (nominal categorical covariates).
(4) ART: NRTI: nucleoside reverse transcriptase inhibitor; NNRTI: nonnucleoside reverse transcriptase inhibitor; PI: protease inhibitor.
are at high risk of being permanently lost and should be
aggressively traced.
Chi et al. also found that a cutoff of 180 days was
optimal to define LTFU after analysing data from the Africa,
Asia, and Latin America regions of the IeDEA collaboration
(including data from our cohort) [22]. There are some
methodological differences between our analyses, principally
regarding minimum numbers of patients for site inclusion.
Chi et al. found quite extensive heterogeneity between sites,
something we also found to a lesser extent. However, it
is nevertheless reassuring that we found a similar optimal
cutoff of 180 days without clinic visits to define LTFU. With
rapid scaling up of antiretroviral treatment taking place
globally, there is a need to adopt a universal consistent
definition of LTFU, or a general algorithm to define cutoffs,
to evaluate HIV treatment programs in different regions
[6, 7, 19].
Over one in five patients in our cohort failed to come
to clinic for more than 180 days in a given year. Similar
rates have also been found in patients from Africa [3, 11].
However, the LTFU rate was lower in EuroSIDA [23], a large
prospective cohort study with HIV-infected patients mainly
from Europe (using one year as a cutoff). Approximately half
ofthepatientswhoexperiencedLTFUinourstudylatercame
back to clinic, and patients who had a previous episode of
LTFU were more likely to prove to be true LTFU, similar to
previous findings [18].
We found that younger patients, patients infected with
hepatitis C, and patients with detectable or unmeasured viral
load were more likely to experience LTFU. These findings
are all consistent with previous study findings [10, 11, 24–
26]. Patients with undetectable viral load are likely to be
motivated and adherent to antiretroviral treatment and thus
remain in care. Among those patients who experienced
LTFU, we found that those who tested negative for hepatitis
C infection or were never tested for hepatitis C were more
likely to be permanently LTFU. This finding seems coun-
terintuitive, but it might be that patients who have tested
positive for hepatitis C receive more medical attention from
their clinicians and thus prove less likely to be permanently
LTFU. Among patients identified as LTFU, anemic patients
were also more likely to be permanently lost to treatment
and care. Anemia has been shown to be a strong prognostic
marker for HIV disease progression and survival [27], which
could, at least in part, explain these patients failing to return
to followup.
Compared to patients on NNRTI-based regimen,
patients receiving no-, single-, or double-drug antiretroviral
therapy or a triple-drug regimen containing PI were more
likely to experience LTFU. The reasons for this are not clear.
Thegreaterlosstofollowupmaybeassociatedwithincreased
drug toxicity, either resulting in a patient receiving mono- or
dual therapy or from receiving a PI. Patients receiving PI-
based regimens are also those who are more likely to be on
a second line regimen, a regimen that may be substantially
more expensive than first line. In the Asia Pacific region,
out-of-pocket expenses are needed to pay for treatment in
some clinics. Hence, the lost to followup may be associated
with drug availability or affordability. It is worth noting that
patients receiving mono- or dual therapy, or a PI based
regimen, were also associated with being less likely to be
permanently lost to followup, that is to say more likely to
return to clinic (albeit not quite statistically significantly
so). This possibly supports the idea of these regimens being
associated with short-term drug availability or affordability
issues. Unfortunately, data are not available to address this
issue in any greater detail.
It has been shown that, in resource-limited settings,
predominantlyinAfrica,patientswhoareLTFUhaveamuch
poorer prognosis than patients who remain in followup
[5]. In part, this is due to a proportion of patients who
die not having vital status information updated at their
treatment site. The extent to which this occurs in TAHOD
is uncertain. While it seems likely that at least some patients
who are LTFU have died without this information reaching
the site, the lack of association between key measures of HIV
disease progression, such as CD4 count and AIDS defining
illnesses, and LFTU suggests it may be lower than in African
settings. However, this association between LTFU and poorer
prognosis underpins the need for consistent definitions of

Page 7

AIDS Research and Treatment7
Table 4: Factors that predict permanent LTFU in patients without a clinic visit for 180 days.
Number True loss%OR1
95% CI
P value Adjusted OR1
95% CI
P value
Sex
Male
Female
Current age (years)
≤35
36∼45
46+
Reported exposure
Heterosexual contact
Homosexual contact
Injecting drug use
Other/unknown
Current CD4 count (cells/µl.)
≤100
101–200
201+
Missing
Current HIV RNA (copies/mL)
≤500
501+
Missing
CDC disease stage
Stage A
Stage B
Stage C
Tuberculosis diagnosis
Yes
No
Time since HIV infection (years)
≤5
6+
Missing
Hepatitis B infection
Yes
No
N/A
Hepatitis C infection
Yes
No
N/A
Current anemia (male < 13g/dL, female < 11g/dL)
Yes
No
N/A
Current ART∗∗
3 + (NRTI + NNRTI)
3 + (NRTI + PI)
No/mono/double ARV
3 + (other combination)
1206
442
584
209
48.4 1.00
47.3 0.89 (0.69, 1.15)
1.00
0.80 0.359 (0.61, 1.05)0.104
568
717
363
278
340
175
48.9 1.00
47.4 1.33 (1.03, 1.71)
48.2 1.27 (0.94, 1.72)
1.00
1.31
1.28
0.0972
0.050
0.128
0.031
0.118
(1.00, 1.72)
(0.93, 1.77)
985
344
125
194
443
199
55
96
45.0 1.00
57.8 1.12 (0.78, 1.60)
44.0 1.01 (0.59, 1.73)
49.5 1.07 (0.69, 1.64)
1.00
1.24
1.32
1.22
0.532
0.969
0.773
(0.85,1.81)
(0.72, 2.41)
(0.78, 1.93)
0.262
0.364
0.382
58
129
1068
393
36
66
465
226
62.1 1.00
51.2 0.76 (0.36, 1.60)
43.5 0.62 (0.33, 1.18)
57.5 1.50 (0.77, 2.93)
1.00
0.99
0.82
1.18
0.471
0.144
0.238
(0.47, 2.13)
(0.42, 1.59)
(0.58, 2.42)
0.989
0.551
0.649
598
153
897
230
78
485
38.5 1.00
51.0 1.02 (0.68, 1.52)
54.1 2.13 (1.63, 2.80) <0.001
1.00
0.94
1.54
0.0112
0.767
0.006
0.924 (0.62, 1.42)
(1.13, 2.09)
828
121
699
413
54
326
49.9 1.00
44.6 0.77 (0.48, 1.22)
46.6 1.00 (0.78, 1.27)
1.00
0.70
1.05
0.258
0.975
(0.43, 1.14)
(0.81, 1.36)
0.154
0.702
361
1287
186
607
51.5 1.00
47.2 0.87 (0.66, 1.16)
1.00
0.85 0.342(0.63, 1.15) 0.297
771
858
19
400
389
4
51.9 1.00
45.3 1.25 (0.98, 1.60)
21.1 0.37 (0.12, 1.17)
1.00
1.03
0.43
0.076
0.091
(0.79, 1.34)
(0.13, 1.43)
0.835
0.170
112
883
653
47
431
315
42.0 1.00
48.8 1.30 (0.84, 2.03)
48.2 1.31 (0.82, 2.09)
1.00
1.35
1.03
0.243
0.253
(0.84, 2.16)
(0.60,1.76)
0.222
0.908
149
796
703
66
376
351
44.3 1.00
47.2 1.57 (1.01, 2.45)
49.9 1.96 (1.26, 3.05)
1.00
1.66
2.16
0.0042
0.034
0.001
0.046
0.003
(1.04, 2.66)
(1.35, 3.46)
141
1065
442
87
456
250
61.7 1.00
42.8 0.53 (0.35, 0.81)
56.6 1.15 (0.73, 1.81)
1.00
0.50
0.78
<0.0012
0.001
0.310
0.003
0.549
(0.32, 0.76)
(0.49, 1.26)
911
356
352
29
404
167
209
13
44.3 1.00
46.9 0.76 (0.57, 1.02)
59.4 0.93 (0.69, 1.26)
44.8 0.89 (0.40, 1.98)
1.00
0.74
0.78
0.85
0.072
0.644
0.770
(0.54,1.01)
(0.57, 1.08)
(0.38, 1.94)
0.057
0.137
0.707

Page 8

8 AIDS Research and Treatment
Table 4: Continued.
Number True loss%OR1
95% CI
P value Adjusted OR1
95% CI
P value
Previous episode of temporary LTFU
None
Once
Twice
1298
296
54
589
158
46
45.4
53.4
85.2 31.76 (13.91, 72.52) <0.001
1.00
2.79
1.00
2.71
27.75
<0.0012
<0.001(2.05, 3.80)
<0.001(1.97, 3.72)
(12.03, 64.01) <0.001
(1) Stratified by TAHOD sites.
(2) Overall for test for trend (ordinal categorical covariates) or for homogeneity (nominal categorical covariates).
(3) ART: NRTI: nucleoside reverse transcriptase inhibitor; NNRTI: nonnucleoside reverse transcriptase inhibitor; PI: protease inhibitor.
LTFUinresearchcohortstudies,andwheretherearepossible
active patient tracing strategies or at least sampling-based
approaches [28] to ensure comparability of results across
studies and settings.
Several limitations should be considered in interpreting
the results in this paper. First, TAHOD participating sites are
generally urban referral centres, and the patients recruited in
TAHOD were those regularly attending a given TAHOD site.
Hence, TAHOD patients are not representative of all HIV-
infected patients in the Asia and Pacific region. The overall
rate of LTFU we saw in our study is therefore likely to be
an underestimate of rates across the region. However, the
effect of these sampling biases on the optimal definition of
LTFU and on the covariate analyses is arguably less strong.
It is reassuring that our estimate of the optimal definition
of LTFU is consistent with that seen across Africa and Latin
America [22]. Second, since antiretroviral treatment has
become more decentralised and available in distant or rural
communities with rapid scale-up programs, patients might
choose to receive treatment and care locally rather than at
tertiary and referral centres [29, 30]. Consequently, patients
may have been retained in care but not necessarily in the
clinics involved in this study. Information on referral to
other health facility was only recently included in the data
collection, so we could not further verify if patients were
retained in care or truly loss to health services. Third, we
do not collect data on the measures TAHOD sites undertake
to routinely trace patients who are LTFU. These measures
differ across sites according to local practices and conditions.
Effective patient tracking and recording are essential to
program evaluation and maintenance of treatment and care
[1, 18]. What patient tracking measures are effective in
retaining patients in treatment and care in the Asia-Pacific
region is an area that deserves further research. We also do
not have data on transportation [31], social and economic
status [32], pregnancy for women [10], and community
support [33], all of which have been found to be important
determinants of LTFU. Lastly, the patients included in this
study were all receiving, or started, antiretroviral treatment
and had clinical assessments. Consequently, the results
cannot be extrapolated to patients not yet initiated on
antiretroviral therapy. Research into followup among HIV-
infected patients not receiving antiretroviral treatment in
the Asia-Pacific region needs to be considered [34–36],
particularly in the context of the move to start treatment
earlier.
5.Conclusion
With rapid scaleup of antiretroviral treatment, it is essential
to study factors that predict loss to followup and identify
patients at risk of loss to treatment and care, particularly
in resource-limited settings. At the treatment and care level,
this can maintain efficacy of antiretroviral therapy and avoid
adverse events. At the program evaluation level, the impact
of loss to followup on overall treatment outcome, disease
progression, and survival can then be accounted for with
appropriate statistical adjustments. Collaboration with HIV
treatment programs in other regions in studies on LTFU and
inparticularstandardisationofLTFUdefinitionsareessential
for reporting and program evaluation.
Acknowledgments
The TREAT Asia HIV Observational Database and the
Australian HIV Observational Database are part of the Asia
Pacific HIV Observational Database and are initiatives of
TREAT Asia, a program of amfAR,The Foundation for AIDS
Research, with support from the following institutes of the
US National Institutes of Health (NIH): National Institute of
Allergy and Infectious Diseases (NIAID), National Institute
of Child Health and Human Development (NICHD), the
Office of the Director (OD), and the National Cancer
Institute (NCI), as part of the International Epidemio-
logic Databases to Evaluate AIDS (IeDEA) (Grant no.
U01AI069907). Additional support is provided by the Dutch
Ministry of Foreign Affairs through a partnership with
Stichting Aids Fonds and from the Austrian AIDS Life Asso-
ciation (AALA). The National Centre in HIV Epidemiology
and Clinical Research is funded by the Australian Govern-
ment Department of Health and Ageing and is affiliated with
the Faculty of Medicine, The University of New South Wales.
The content of this publication is solely the responsibility
of the authors and does not necessarily represent the official
views of any of the institutions mentioned above.
References
[1] R. P. Dalal, C. MacPhail, M. Mqhayi et al., “Characteristics
and outcomes of adult patients lost to follow-up at an
antiretroviral treatment clinic in Johannesburg, South Africa,”
Journal of Acquired Immune Deficiency Syndromes, vol. 47, no.
1, pp. 101–107, 2008.

Page 9

AIDS Research and Treatment9
[2] A. T. Brennan, M. Maskew, I. Sanne, and M. P. Fox, “The
importance of clinic attendance in the first six months on
antiretroviral treatment: a retrospective analysis at a large
public sector HIV clinic in South Africa,” Journal of the
International AIDS Society, vol. 13, no. 1, article 49, 2010.
[3] H. Bygrave, K. Kranzer, K. Hilderbrand et al., “Trends in loss
tofollow-upamongmigrantworkersonantiretroviraltherapy
in a community cohort in Lesotho,” PLoS ONE, vol. 5, no. 10,
Article ID e13198, 2010.
[4] M. W. G. Brinkhof, B. D. Spycher, C. Yiannoutsos et al.,
“Adjusting mortality for loss to follow-up: analysis of five art
programmes in sub-saharan africa,” PLoS ONE, vol. 5, no. 11,
Article ID e14149, 2010.
[5] M. W. G. Brinkhof, M. Pujades-Rodriguez, and M. Egger,
“Mortality of patients lost to follow-up in antiretroviral treat-
ment programmes in resource-limited settings: systematic
review and meta-analysis,” PLoS ONE, vol. 4, no. 6, Article ID
e5790, 2009.
[6] B. H. Chi, R. A. Cantrell, A. Mwango et al., “An empirical
approach to defining loss to follow-up among patients
enrolled in antiretroviral treatment programs,” American
Journal of Epidemiology, vol. 171, no. 8, pp. 924–931, 2010.
[7] M.Egger,B.D.Spycher,J.Sidleetal.,“Correctingmortalityfor
loss to follow-up: a nomogram applied to antiretroviral treat-
mentprogrammesinsub-SaharanAfrica,” PLoSMedicine,vol.
8, no. 1, article e1000390, 2011.
[8] K.Wools-Kaloustian,S.Kimaiyo,L.Dieroetal.,“Viabilityand
effectiveness of large-scale HIV treatment initiatives in sub-
Saharan Africa: experience from western Kenya,” AIDS, vol.
20, no. 1, pp. 41–48, 2006.
[9] P. Braitstein, M. W. Brinkhof, F. Dabis et al., “Mortality
of HIV-1-infected patients in the first year of antiretroviral
therapy: comparison between low-income and high-income
countries,” The Lancet, vol. 367, no. 9513, pp. 817–824, 2006.
[10] B. Wang, E. Losina, R. Stark et al., “Loss to follow-up in a
community clinic in South Africa-roles of gender, pregnancy
and CD4 count,” South African Medical Journal, vol. 101, no.
4, pp. 253–257, 2011.
[11] V. Ochieng-Ooko, D. Ochieng, J. E. Sidle et al., “Influence
of gender on loss to follow-up in a large HIV treatment
programme in western kenya,” Bulletin of the World Health
Organization, vol. 88, no. 9, pp. 681–688, 2010.
[12] R. Weigel, M. Hochgesang, M. W.G. Brinkhof et al., “Out-
comes and associated risk factors of patients traced after being
lost to follow-up from antiretroviral treatment in Lilongwe,
Malawi,” BMC Infectious Diseases, vol. 11, article 31, 2011.
[13] O. Keiser, B. H. Chi, T. Gsponer et al., “Outcomes of
antiretroviral treatment in programmes with and without
routine viral load monitoring in southern Africa,” AIDS, vol.
25, no. 14, pp. 1761–1769, 2011.
[14] S. Thai, O. Koole, P. Un et al., “Five-year experience with
scaling-up access to antiretroviral treatment in an HIV care
programme in Cambodia,” Tropical Medicine and Interna-
tional Health, vol. 14, no. 9, pp. 1048–1058, 2009.
[15] P. L. Toro, M. Katyal, R. J. Carter et al., “Initiation of
antiretroviral therapy among pregnant women in resource-
limited countries: CD4+ cell count response and program
retention,” AIDS, vol. 24, no. 4, pp. 515–524, 2010.
[16] M. Panditrao, S. Darak, V. Kulkarni, S. Kulkarni, and R.
Parchure, “Socio-demographic factors associated with loss to
follow-up of HIV-infected women attending a private sector
PMTCT program in Maharashtra, India,” AIDS Care, vol. 23,
no. 5, pp. 593–600, 2011.
[17] J. Zhou, N. Kumarasamy, F. Zhang et al., “Predicting short-
term disease progression among HIV-infected patients in Asia
and the Pacific region: preliminary results from the TREAT
Asia HIV Observational Database (TAHOD),” HIV Medicine,
vol. 6, no. 3, pp. 216–223, 2005.
[18] T. Hill, L. Bansi, C. Sabin et al., “Data linkage reduces loss to
follow-up in an observational HIV cohort study,” Journal of
Clinical Epidemiology, vol. 63, no. 10, pp. 1101–1109, 2010.
[19] M. W. G. Brinkhof, F. Dabis, L. Myer et al., “Early loss of HIV-
infectedpatientsonpotentantiretroviraltherapyprogrammes
in lower-income countries,” Bulletin of the World Health
Organization, vol. 86, no. 7, pp. 559–567, 2008.
[20] E. H. Geng, N. Emenyonu, M. B. Bwana, D. V. Glidden,
and J. N. Martin, “Sampling-based approach to determining
outcomesofpatientslosttofollow-upinantiretroviraltherapy
scale-up programs in Africa,” Journal of the American Medical
Association, vol. 300, no. 5, pp. 506–507, 2008.
[21] C. Cesar, B. E. Shepherd, A. J. Krolewiecki et al., “Rates
and reasons for early change of first HAART in HIV-1-
infectedpatientsin7sitesthroughouttheCaribbeanandLatin
America,” PLoS ONE, vol. 5, no. 6, Article ID e10490, 2010.
[22] B. H. Chi, C. T. Yiannoutsos, A. O. Westfall et al., “Universal
definition of loss to follow-up in HIV treatment programs: a
statistical analysis of 111 facilities in Africa, Asia, and Latin
America,”PLoSMedicine,vol.8,no.10,articlee1001111,2011.
[23] A. Mocroft, O. Kirk, P. Aldins et al., “Loss to follow-up in an
international,multicentreobservationalstudy,”HIVMedicine,
vol. 9, no. 5, pp. 261–269, 2008.
[24] R. Zachariah, K. Tayler-Smith, M. Manzi et al., “Retention
and attrition during the preparation phase and after start
of antiretroviral treatment in Thyolo, Malawi, and Kibera,
Kenya: implications for programmes?” Transactions of the
Royal Society of Tropical Medicine and Hygiene, vol. 105, no.
8, pp. 421–430, 2011.
[25] T. Hill, L. Bansi, C. Sabin et al., “Data linkage reduces loss to
follow-up in an observational HIV cohort study,” Journal of
Clinical Epidemiology, vol. 63, no. 10, pp. 1101–1109, 2010.
[26] A. Mocroft, O. Kirk, P. Aldins et al., “Loss to follow-up in an
international,multicentreobservationalstudy,”HIVMedicine,
vol. 9, no. 5, pp. 261–269, 2008.
[27] J. D. Lundgren and A. Mocroft, “Anemia and survival in
human immunodeficiency virus,” Clinical Infectious Diseases,
vol. 37, no. 4, pp. s297–s303, 2003.
[28] C.T.Yiannoutsos,M.W.An,C.E.Frangakisetal.,“Sampling-
based approaches to improve estimation of mortality among
patient dropouts: experience from a large PEPFAR-funded
program in Western Kenya,” PLoS ONE, vol. 3, no. 12, Article
ID e3843, 2008.
[29] M. Bedelu, N. Ford, K. Hilderbrand, and H. Reuter, “Imple-
menting antiretroviral therapy in rural communities: the
Lusikisiki model of decentralized HIV/AIDS care,” Journal of
Infectious Diseases, vol. 196, no. 3, pp. S464–S468, 2007.
[30] A. K. Chan, G. Mateyu, A. Jahn et al., “Outcome assessment
of decentralization of antiretroviral therapy provision in a
rural district of Malawi using an integrated primary care
model,” Tropical Medicine and International Health, vol. 15,
supplement 1, pp. 90–97, 2010.
[31] E. H. Geng, D. R. Bangsberg, N. Musinguzi et al., “Under-
standing reasons for and outcomes of patients lost to follow-
up in antiretroviral therapy programs in Africa through
a sampling-based approach,” Journal of Acquired Immune
Deficiency Syndromes, vol. 53, no. 3, pp. 405–411, 2010.
[32] M. Maskew, P. MacPhail, C. Menezes, and D. Rubel, “Lost
to follow up—contributing factors and challenges in South

Page 10

10AIDS Research and Treatment
African patients on antiretroviral therapy,” South African
Medical Journal, vol. 97, no. 9, pp. 853–857, 2007.
[33] N. C. Ware, J. Idoko, S. Kaaya et al., “Explaining adherence
success in sub-Saharan Africa: an ethnographic study,” PLoS
Medicine, vol. 6, no. 1, Article ID e1000011, pp. 0039–0047,
2009.
[34] E. H. Geng, D. Nash, A. Kambugu et al., “Retention in care
among HIV-infected patients in resource-limited settings:
emerging insights and new directions,” Current HIV/AIDS
Reports, vol. 7, no. 4, pp. 234–244, 2010.
[35] B. Amuron, G. Namara, J. Birungi et al., “Mortality and
loss-to-follow-up during the pre-treatment period in an
antiretroviral therapy programme under normal health ser-
vice conditions in Uganda,” BMC Public Health, vol. 9, article
290, 2009.
[36] T. Togun, I. Peterson, S. Jaffar et al., “Pre-treatment mortality
and loss-to-follow-up in HIV-1, HIV-2 and HIV-1/HIV-2
dually infected patients eligible for antiretroviral therapy in
The Gambia, West Africa,” AIDS Research and Therapy, vol.
8, no. 1, p. 24, 2011.