AVT-765 Law.qxp
15/3/06
12:26 pm
Page 179
Antiviral Therapy 11:179–186
Evaluation of the HIV lipodystrophy case definition
in a placebo-controlled, 144-week study in
antiretroviral-naive adults
Matthew Law 1, Rebekah Puls 1, Andrew K Cheng 2, David A Cooper 1 and Andrew Carr 3*
1
National Centre in HIV Epidemiology and Clinical Research, University of New South Wales, Sydney, Australia
Gilead Sciences, Foster City, CA, USA
3
St Vincent’s Hospital, Sydney, Australia
2
*Corresponding author: Tel: +61 2 8382 3359; Fax: +61 2 8382 3893; E-mail: [email protected]
Objective: To compare three versions of the objective
HIV-associated lipodystrophy (HIVLD) case definition
(LDCD) and derived severity scale to spontaneous clinical
LD assessment in adults initiating antiretroviral therapy.
Design and main outcome measures: The LDCD versions
were the ‘primary’ LDCD [which includes dual-energy
X-ray absorptiometry (DXA) and computerized tomography (CT)], a simpler ‘central’ LDCD that omits CT data,
and a simpler but probably less accurate ‘non-imaging’
LDCD. Physician LD assessments were passively reported.
Two of the 10 parameters in the primary LDCD were not
collected and were imputed.
Setting, participants and interventions: Retrospective
analysis of a randomized, placebo-controlled, 144-week
study of tenofovir DF or stavudine (d4T) in 600 antiretroviral-naive adults.
Results: Central LDCD and clinical assessment diagnosed
LD in 27% and 19% of d4T recipients at week 144,
<0.001), and 3% and 3% of tenofovir DF
respectively (P<
recipients, respectively (P=0.248). The central LDCD
performed at least as well as the primary LDCD; both
were more sensitive than the non-imaging model. There
was poor concordance between clinical and LDCD-based
diagnosis (kappa 0.02–0.20); most clinical cases did not
fulfill any LDCD. Using the central LDCD, most LD was
grade 1; 6% of d4T recipients and no tenofovir DF recipient had grade 3–4 LD at week 144 (P=0.007).
Independent risk factors for LD using the central LDCD
were d4T, increasing age, female sex and higher baseline
triglycerides, whereas clinical assessment consistently
identified only d4T. The LDCD score was more sensitive
than DXA for assessing LD severity.
Conclusions: In this prospective study of a first antiretroviral regimen, the LDCD was more sensitive for LD
diagnosis and identified more lipodystrophy risk factors
than spontaneous clinical assessment or DXA, and also
objectively quantified LD severity. The central LDCD
should make objective LD assessment cheaper and
simpler. Spontaneous clinical LD assessment of is of
limited value, even in placebo-controlled trials.
Introduction
Lipodystrophy is most common and severe in HIVinfected adults receiving stavudine (d4T) [1–4]. In
the double-blind, placebo-controlled Gilead 903
trial, significantly more lipodystrophy was observed
over 144 weeks in patients randomized to d4T
(19%) than in those randomized to tenofovir DF
(TDF) (3%), as well as significantly less limb fat
mass [5]. Clinical assessment of changes in lipodystrophy, however, correlates poorly with objective
body fat changes even when actively performed [6,
7]. Also, the rate of lipodystrophy observed with d4T
in Gilead 903 was less than in prospective cohort
studies (20–40% at 12–18 months) [8, 9], perhaps
because these cohorts actively recorded lipodystrophy [10]. As a result, the incidences observed and
© 2006 International Medical Press 1359-6535
the 16% difference between the d4T and TDF arms
in Gilead 903 may have been underestimates.
An objective case definition incorporating ten clinical, metabolic and body composition variables can
diagnose lipodystrophy with at least 80% accuracy
[11]. The subsequently developed lipodystrophy case
definition (LDCD) score appears to be the most accurate tool currently available for determining lipodystrophy severity, using either the absolute score or a
derived grading scale [12]. The LDCD and severity
score have been evaluated in only one prospective
study in which the LDCD score reduced significantly
in those that switched from d4T or zidovudine to
abacavir [6]. The ‘primary’ LDCD is complex,
however, given its need for both dual-energy X-ray
179
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 180
M Law et al.
absorptiometry (DXA) and computerized tomography
(CT)-derived data. To address this, a simpler, ‘nonimaging’ LDCD model was developed, but was
less sensitive and specific. Using centrally analysed
DXA data, however, a third model, which does not
include CT-derived data, has accuracy similar to the
primary model [13]; this ‘central’ LDCD has not been
evaluated prospectively.
Risk factors for lipodystrophy, particularly non-drug
factors, differ between lipodystrophy surveys, possibly
because of the subjective nature of lipodystrophy
assessment in these studies [1–4, 14–21]. The LDCD
provides an opportunity to objectively determine such
risk factors. Lastly, although the LDCD is better at
assessing lipodystrophy severity in a cross-sectional
setting than is spontaneous clinical reporting, this
comparison has not been assessed prospectively.
We retrospectively analysed data from a placebocontrolled, 144-week study (Gilead 903) in order to
describe the objective incidence and severity of lipodystrophy using the LDCD models, to compare lipodystrophy prevalence, severity and risk factors using the
LDCD and clinical lipodystrophy assessment, and to
compare the relative power of the LDCD and DXA for
the assessment of lipodystrophy.
Methods
HIV lipodystrophy case definitions
The HIV LDCD study evaluated 1,081 HIV-infected
adult outpatients without active AIDS at 32 sites globally [11]. A lipodystrophy-specific questionnaire and a
standardized, lipodystrophy-specific physical examination recorded lipoatrophy and/or diffuse fat accumulation in the face, neck, dorso-cervical spine, arms,
breasts, abdomen, buttocks and legs, as well as the
presence, site, and number of any lipomata. Subjects
with at least one moderate or severe lipodystrophic
feature (except isolated abdominal obesity) apparent to
both physician and patient were assigned as cases.
Subjects with no lipodystrophic feature of any severity
were controls. Logistic regression models including
only objective and locally measured clinical, fasting
metabolic and body composition data were constructed
in a training subset of randomly selected cases and
controls and validated in the remainder.
The primary LDCD includes age, gender, CDC clinical stage, duration of HIV infection, anion gap, highdensity lipoprotein (HDL) cholesterol, leg fat percent
and trunk:limb fat ratio (both using DXA), and
visceral:subcutaneous abdominal fat area (VAT:SAT)
ratio (using CT), and has 79% sensitivity and 80%
specificity (Table 1). An LDCD score was derived from
this model by adding the relevant individual score for
each parameter and then subtracting 43 (the constant).
A final score of at least zero defines the presence of
lipodystrophy. A grading scale (grades 0 to 4) was then
derived based on the LDCD score, with inter-grade
cut-off scores of 0, 10, 15 and 23 [12].
By the omission of all imaging data, a simpler ‘nonimaging’ model was derived, but this had less sensitivity (73%) and specificity (71%), even though it also
included low-density lipoprotein (LDL) cholesterol,
triglycerides and lactate. Subsequent to publication of
Table 1. Parameters included in each LDCD model
LDCD
Parameters
Clinical
Metabolic
DXA
CT
Sensitivity†
Specificity†
Primary
(DXA and CT)
Score
Non-imaging
(no DXA or CT)
Clinical only
Central
(centrally read DXA, no CT)
Female
Age >40 years
CDC stage (A/B/C)
HIV >4 years
Waist:hip ratio*
HDL cholesterol
Anion gap
Leg fat percent
Trunk:limb fat ratio
VAT:SAT ratio*
79%
80%
22
7
0/3/7
11
× 29
× –14
×1
–16 to 0
×5
0 to 13
LD ≥0
no LD <0
+
+
+
+
+
+ LDL-C/Tg
+ lactate
not used
not used
not used
73%
71%
+
+
+
∆ CD4 nadir
+
not used
not used
not used
not used
not used
75%
60%
+
+
+
+
not used
+
+
+
+
not used
76%
80%
*These parameters were not recorded in Gilead 903, so these values were imputed for use in the primary LDCD. †Sensitivity and specificity values are derived from
[11–13]. ∆ CD4 nadir, change from nadir CD4+ lymphocyte count; CT, computed tomography; DXA, dual-energy X-ray absorptiometry; HDL, high-density lipoprotein;
LDCD, lipodystrophy case definition; + LDL-C/Tg, non-imaging model includes low-density lipoprotein (LDL)-cholesterol and triglycerides in addition to HDL-cholesterol;
+ lactate, non-imaging model includes lactate in addition to anion gap; VAT:SAT, ratio of visceral adipose tissue (VAT) area to subcutaneous adipose tissue (SAT) area.
180
© 2006 International Medical Press
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 181
Longitudinal study of HIVLD case definition
the primary LDCD, a model was generated using
centrally analysed imaging data [13]. This central
model, in which VAT:SAT ratio and waist:hip ratio are
not statistically significant and so are omitted (and so
excludes all CT-derived data), and which includes no
new variable, has 76% sensitivity and 80% specificity.
The Gilead 903 study
Study 903 is a trial with an ongoing 96-week extension
phase (903e) and a completed 144-week, prospective,
randomized, double-blind phase designed to evaluate
TDF (n=299) or d4T (n=303) in combination with
lamivudine and efavirenz in treatment-naive HIVinfected adults, of which 74% were male [5]. A lipodystrophy substudy was initiated after the study
commenced in which whole-body DXA was performed
and centrally analysed at weeks 96 and 144. Two
primary LDCD variables were not collected: VAT:SAT
ratio and waist:hip ratio. Lactate, which is required for
the non-imaging LDCD, was not measured at all sites.
As in most cross-sectional and in all prospective,
randomised antiretroviral trials, lipodystrophy was
reported as an unprompted or serious adverse event; no
specific lipodystrophy questionnaire or case record
form was used. There was no requirement to report the
type of lipodystrophy (fat accumulation or lipoatrophy
or both). Lipodystrophy severity was determined by the
treating physician.
Analysis
We hypothesized that the prevalence and severity of
lipodystrophy in Gilead 903 using the LDCD would
be greater than that recorded passively, that the LDCD
is a valid method for assessing changes in LD prevalence and severity in ART-naive adults, that the LDCD
would be more able to identify multiple risk factors for
lipodystrophy, and that the LDCD would be more
sensitive than DXA alone for assessing changes
in lipodystrophy.
The three above-mentioned formulations of the
LDCD were assessed, namely, the primary, nonimaging and central models. For the primary LDCD,
waist:hip ratio and VAT:SAT ratio were imputed using
data from controls in the LDCD Study [11], using
linear regression based on age and sex. This imputation
would be expected to lead to underestimates of the
LDCD score (that is, the primary LDCD in this dataset
will have reduced sensitivity). No data were imputed
for the central or non-imaging LDCD models.
Rates of lipodystrophy according to different assessment methods were compared using methods appropriate to paired data (McNemar’s test, signed-rank test).
Rates were compared between treatment arms using
chi-square tests. Risk factors for lipodystrophy at weeks
96 and 144 were assessed using logistic regression.
Antiviral Therapy 11:2
Multivariate models were developed using forward
stepwise techniques. The measure of agreement for
concordance between clinical and LDCD assessment
was by kappa statistics. All P-values are 2-sided, with
no adjustment for multiple comparisons.
Results
Subjects
The primary LDCD (with imputed VAT:SAT and
waist:hip) and the central LDCD could both be used
for 255 (43%) and 229 (38%) subjects at weeks 96
and 144, respectively, and the non-imaging LDCD for
252 (42%) and 228 (38%) participants, respectively.
All patient subsets had similar baseline characteristics
to those subjects not included in each analysis (data
not shown).
Prevalence
The overall lipodystrophy prevalence at week 96 by
clinical assessment was 6% (Table 2), which was significantly less than with the central LDCD (19%;
P<0.001) and the non-imaging LDCD (14%; P<0.001),
but not the primary LDCD (9%; P=0.505). At week
144, the prevalence assessed clinically had increased to
11% (P<0.001), but did not change significantly with
any LDCD.
There was significant discordance between clinical
diagnosis and LDCD diagnosis at both timepoints
(kappa values ranged from 0.02 to 0.20). In particular,
more than 80% of patients diagnosed with lipodystrophy using the case definitions were not diagnosed
clinically (Table 3). Concordance between the primary
and central LDCDs was substantially higher (Table 4).
Although the central LDCD was more sensitive than
the primary LDCD, the central LDCD appeared to
have similar specificity, as it did not identify any subject
as having lipodystrophy who was assigned as not
having lipodystrophy by the primary LDCD.
Lipodystrophy was reported for more patients in the
d4T group than in the TDF group at both timepoints.
The prevalence in the d4T group at week 96 was 31%
with the central LDCD, as compared with 13% with
the primary LDCD (P<0.001), 18% with the nonimaging LDCD (P<0.001) and 12% by clinical assessment (P<0.001). At week 144, rates using clinical
assessment, the primary LDCD and the central LDCD
were similar, but all were significantly greater than with
the non-imaging LDCD.
The difference in prevalence between the d4T and
TDF groups at week 96 was 11% using clinical assessment, which was significantly less than with the central
LDCD (25%; P<0.001), but similar to the non-imaging
LDCD (7%; P=0.136) and primary LDCD (9%;
P=0.010). These differences were similar at week 144.
181
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 182
M Law et al.
Table 2. Prevalence and severity of lipodystrophy according to clinical and objective assessments
Lipodystrophy assessment
Clinical
Primary LDCD
(subjective) (DXA and CT)
n=600
Week 96
Prevalence
Overall
Tenofovir DF
Stavudine
Difference
P-value§
Severity¶
Grade 1
Grade 2
Grade 3/4
Week 144
Prevalence
Overall
TDF
d4T
Difference
P-value§
Severity¶
Grade 1
Grade 2
Grade 3/4
Central LDCD (no DXA)
n=255
Non-imaging LDCD (no DXA or CT)
n=255
n=252
†
P-value†
P-value‡
14 <0.001
11
0.001
18 0.041
7
0.136
0.001
0.005
0.103
0.819
0.034
0.132
11
2
1.4
0.013
0.473
%
%
P value*
%
P-value*
P-value
%
6
1
12
11
<0.001
9
4
13
9
0.010
0.505
0.257
0.853
19
6
31
25
<0.001
<0.001
0.058
<0.001
<0.001
<0.001
0.083
<0.001
<0.001
4
3
0.2
7
1
0.4
0.839
13
3
3
<0.001
<0.001
n=600
n=229
P-value*
0.001
n=229
%
%
P-value*
11
3
19
16
<0.001
8
0.004
0
0.001
17
0.077
19
<0.001
7
4
0.2
7
0.4
1
%
0.012
P-value*
n=228
†
P-value†
P-value‡
12 0.857
7
0.405
17 0.578
10
0.018
0.480
0.039
0.999
0.317
0.999
0.317
9
1
1.4
0.999
0.055
P-value
%
15
0.074
3
0.248
27
0.758
24
<0.001
<0.001
0.011
<0.001
8
4
3
<0.001
0.420
P-value*
0.174
*P-value for comparison with subjective report – McNemar’s test. †P-value for comparison with primary lipodystrophy case definition (LDCD) – McNemar’s test.
‡
P-value for comparison between non-imaging LDCD and central LDCD–McNemar’s test. §P-value for comparison between d4T and TDF – 2-sample chi-square
test. ¶P-value for comparison of severity by signed-rank test. d4T, stavudine; LDCD, lipodystrophy case definition; TDF, tenofovir DF.
Table 3. Comparison of lipodystrophy diagnosis: central and
modified primary case definitions compared with clinical
assessment
Table 4. Comparison of lipodystrophy diagnosis: central case
definition compared with modified primary case definition
Lipodystrophy
diagnosed by
central LDCD
Lipodystrophy diagnosis
Central LDCD
Lipodystrophy Week 96
diagnosed
clinically
(n=255)
Week 144
Yes
No
kappa
Yes
No
kappa
LDCD, lipodystrophy case definition.
182
Primary LDCD
Yes
No
Yes
No
7
41
11
196
2
20
0.02
8
11
0.20
16
217
0.12
9
26
28
166
0.11
29
181
Lipodystrophy
diagnosed
by primary
LDCD (n=255)
Week 96
Week 144
Yes
No
kappa
Yes
No
kappa
Yes
No
22
26
0
207
0.58
19
16
0
194
0.67
LDCD, lipodystrophy case definition.
© 2006 International Medical Press
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 183
Longitudinal study of HIVLD case definition
Lipodystrophy severity
Comparison of LDCD with DXA scanning
Most lipodystrophy was grade 1 or 2 by LDCD or clinical assessment. Significantly more grade 1 or 2 lipodystrophy was detected by the central and non-imaging
LDCDs at week 96, but not week 144. Using the
central LDCD, more grade 3 or 4 lipodystrophy
occurred with d4T than with TDF (Figure 1) at week
96 (5.4% versus none; P=0.014) and at week 144
(6.1% versus none; P=0.007). The prevalence of any
grade of lipodystrophy did not change appreciably
between weeks 96 and 144.
Mean LDCD severity scores were similar at weeks
96 and 144 (–14 ±15 and –16 ±17, respectively). These
scores were significantly higher at each timepoint in
those receiving d4T (–7 ±15 and –7 ±17, respectively)
than in those receiving TDF (–23 ±13 [P<0.001] and
–25 ±12 [P=0.03], respectively), as compared with
mean scores of 11, 11, 16 for adults with moderate or
severe lipodystrophy in the MITOX, rosiglitazone and
LDCD studies, respectively, and of –13 for non-lipodystrophic controls in the LDCD study [6, 11, 22].
To assess the sensitivity of the central LDCD score in
detecting differences between groups in comparison
with DXA, we compared the TDF and d4T groups in
terms of the central LDCD score, leg fat percent and leg
fat mass at weeks 96 and 144. At week 96, the mean
differences between TDF and d4T were –13.2, 9.4%
and 2.6 kg, respectively, and at week 144, –17.8,
12.2.% and 3.3 kg, respectively. The t-statistics, which
measure the strength of the treatment difference, were
7.7, 7.2 and 7.1, respectively, at week 96, and 9.3, 8.7,
and 8.1 at week 144. This indicates that in this study
the central score was slightly more powerful for
detecting treatment differences than DXA scanning of
the leg.
Risk factors for lipodystrophy
Using the central LDCD, increasing age, female sex,
assignment to d4T therapy and higher baseline triglycerides were independently associated with a greater
likelihood of having lipodystrophy at week 96 and
week 144 (Table 5). Lower baseline HDL cholesterol
levels were also associated with greater lipodystrophy
risk by the LDCD at week 144. In contrast, multivariate analysis based on clinical assessment identified
only d4T therapy as a risk factor for lipodystrophy at
both timepoints.
Figure 1. Lipodystrophy severity by randomized group
according to central LDCD grades
Lipodystrophy, %
25
20
TDF
d4T
15
10
5
0
Gd 1
Week
96
144
Gd 2
96
144
Gd 3
96
144
Gd 4
96
144
d4T, stavudine; Gd, grade; LDCD, lipodystrophy case definition; TDF, tenofovir.
Antiviral Therapy 11:2
Discussion
There are several key observations from our analysis.
In this prospective trial, the LDCD was more sensitive
and possibly more accurate for lipodystrophy diagnosis
than spontaneous clinical assessment or DXA, objectively quantified lipodystrophy severity, and more
completely identified lipodystrophy risk factors.
Spontaneous clinical assessment of lipodystrophy is of
limited value, even in placebo-controlled trials.
The central LDCD should make objective assessment of lipodystrophy cheaper and simpler. The
increased sensitivity of the LDCD would increase the
power of clinical trials. For example, using the prevalence rate of 19% at 3 years of lipodystrophy by clinical assessment in patients receiving d4T, a clinical trial
would need 480 patients to detect a decrease in the risk
of lipodystrophy of 50% with 80% power. Using the
central LDCD prevalence rate of 27%, the same study
would only require 333 patients to detect a 50% reduction in the risk with 80% power.
The LDCD was a more sensitive and accurate tool
for assessing lipodystrophy prevalence, severity and
risk factors than subjective, clinical reporting. This
finding is in keeping with the poor correlation between
clinical assessment and body composition-based assessment of lipodystrophy in other studies, including
studies that were placebo-controlled [6, 7, 22]. As such,
subjective and passive clinical assessment appears inadequate to report lipodystrophy in clinical trials.
Although this may seem self-evident, it is important to
recognize that almost all prospective antiretroviral
trials that have studied lipodystrophy have only studied
it subjectively, that is, by patient or physician assessment. The present data suggest the rates of lipodystrophy reported in such trials are underestimates.
Furthermore, the poor agreement between clinical
assessment and LDCD assignment in this placebocontrolled study also argues for active collection of
183
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 184
M Law et al.
Table 5. Risk factors for lipodystrophy by the central LDCD and by clinical assessment
Week 96
Central LDCD, n=229
Parameter
Nucleoside
analogue
OR
TDF
d4T
Age (per 1 year
increase)
Sex
1.0
8.68
1.08
M
F
1.0
3.71
Week 144
Clinical assessment, n=600
95% CI P
OR
95%CI P
1.0
10.27 2.17,
48.6
3.59,
21.0
<0.001
1.03,
1.13
0.001
1.67,
8.25
0.001
1.0
1.02
Central LDCD, n=229
OR
0.003
1.0
23.9
0.92
1.07
0.30,
3.45
0.98
1.0
5.56
0.997 0.94,
1.06
Clinical assessment, n=600
95% CI
P
OR
5.8,
99.1
<0.001
1.0
7.69
1.02,
1.14
0.01
1.03
1.71,
18.01
0.004
1.0
1.04
95% CI P
3.73,
15.86
<0.001
1.002,
1.061
0.03
0.56,
1.91
0.90
Baseline triglycerides
(per 1 mmol/l increase)
1.32
1.04,
1.66
0.02
1.39
1.06,
1.83
0.02
1.64
1.13,
2.37
0.009
1.02
0.89,
1.16
0.80
HDL cholesterol
(per 1 mmol/l increase)
0.43
0.09,
1.98
0.277
0.78
0.10,
6.30
0.812
0.02
0.00,
0.18
0.001
1.57
0.58,
4.21
0.37
0.53,
3.14
0.92,
7.30
0.58
1.0
1.75
0.55,
5.59
0.43,
7.67
0.35
1.0
1.81
0.59,
5.55
1.54,
24.0
0.30
1.0
1.97
1.08,
3.58
0.16,
1.03
0.03
CDC clinical stage
A
B
1.0
1.29
C
2.60
0.07
(0.083)
1.82
0.42
(0.31)
6.07
0.01
(0.011)
0.41
0.06
(0.32)
Results of multivariate analyses based on all parameters significant on univariate analysis. Multivariate models built using forward stepwise methods. Variables in the
final model are in bold. All other variables are presented adjusted for variables in the final model. P-values in parentheses from test for trend. Parameters non-significant on univariate analysis at both timepoints were baseline total cholesterol, anion gap, weight and height. d4T, stavudine; F, female; HDL, high-density lipoprotein;
LDCD, lipodystrophy case definition; M, male; TDF, tenofovir DF.
lipodystrophy as an adverse event in future antiretroviral studies. Nevertheless, the LDCD was superior to
actively collected clinical lipodystrophy assessments in
at least one study [7].
The adapted central LDCD was also more sensitive
than the primary LDCD, which used imputed VAT:SAT
and waist:hip ratios, but had equal specificity. This
greater sensitivity appeared more for the detection of
milder lipodystrophy. The necessity of imputing two of
the primary LDCD’s ten parameters in the present
study was very likely to have reduced the sensitivity of
the primary LDCD. As such, we cannot determine
whether the central or primary LDCD is preferable.
The central and primary LDCDs, therefore, deserve
further comparison where all parameters required for
both models are available. The central LDCD was also
more specific than the non-imaging LDCD, as originally suggested [11].
One unlikely explanation for the greater sensitivity of
the central LDCD than passive clinical assessment for
diagnosis of lipodystrophy is because the LDCD is in
fact overly sensitive. However, there are several reasons
184
we believe that the central LDCD is not over-sensitive.
Firstly, the 31% prevalence of lipodystrophy observed
in the d4T group with the LDCD by week 144 is in
keeping with the 40% lower limb fat mass in this group
and with the prevalence rates in large cross-sectional
studies [1–4, 14–21], as well as the 40% prevalence in
those receiving d4T in a recently published randomized
trial [23]. Secondly, in subjects receiving TDF, the
LDCD reported a very low and unchanging prevalence
of lipodystrophy (3%) and a very normal (negative)
mean LDCD score.
Lipodystrophy was passively reported in Gilead 903.
There was no requirement for specific reporting of
lipoatrophy or fat accumulation and most reports did
not specify whether the patient had experienced lipoatrophy or fat accumulation or both. One could speculate that most reports were for lipoatrophy, given the
use of d4T without a protease inhibitor, and the
predominance of lipoatrophy in the LDCD [11–13].
Most risk factors we identified for lipodystrophy
have been observed in some, but not all, prior lipodystrophy studies. The fact that four out of five risk
© 2006 International Medical Press
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 185
Longitudinal study of HIVLD case definition
factors were identified at both timepoints suggests that
the LDCD is a consistent screening tool. Of interest,
baseline triglyceride and HDL cholesterol levels were
both independent predictors of lipodystrophy. This
suggests that differences between patients in their
capacities to store or produce lipids contribute to the
pathogenesis of lipodystrophy. Our study is also the
first to objectively demonstrate that women are at
greater risk of lipodystrophy than men; whether this
relates to different body fat mass and distribution at
baseline, or to sex differences in adipocyte function
deserves investigation. Most risk factors identified by
the LDCDs are components of the LDCD. This raises
the possibility of circularity – that the risk factors were
identified merely because they are part of the definition. Nevertheless, passive clinical assessment did not
identify these risk factors.
Our study has limitations including the lack of baseline and week 48 DXA scans, an incomplete metabolic
dataset, missing data for the primary LDCD, the
passive collection of clinical lipodystrophy presence
and severity, as well as for the sites affected (lipoatrophy or fat accumulation). The central and primary
LDCDs deserve further comparison in antiretroviralnaive trials using other classes of therapy such as
protease inhibitors, and should be compared with clinical lipodystrophy assessment that is actively
performed. In addition, the LDCD will remain a
research tool in its current form; the primary LDCD
remains very complex whereas the simpler central
LDCD requires central DXA analysis. Any further
simplification would require more accurate imaging
techniques or development of newer metabolic tests
that are specific to lipodystrophy. Nevertheless, the
central LDCD should make objective assessment of
lipodystrophy cheaper and simpler in clinical trials.
The LDCD has considerable advantages. It can be
used to objectively measure incidence, severity (both
numerically and by grades) and risk factors, and can be
applied with equal validity in open-label trials. Also, it
is equally applicable to men and women, whereas use
of any other imaging tool such as DXA alone would
require separate imaging scales for men and women.
Acknowledgements
The National Centre in HIV Epidemiology and Clinical
Research is funded by the Commonwealth Department
of Health and Ageing.
Potential conflicts of interest
Matthew Law and Rebekah Puls declare no conflict of
interest. Andrew Cheng is an employee of Gilead
Sciences. David Cooper has received research support
Antiviral Therapy 11:2
from Boehringer-Ingelheim, Bristol-Myers Squibb,
Chiron, Gilead Sciences, GlaxoSmithKline, Merck and
Roche; honoraria from Boehringer-Ingelheim, BristolMyers
Squibb,
Chiron,
Gilead
Sciences,
GlaxoSmithKline, Merck, Pfizer and Roche; and
consultancies from Boehringer-Ingelheim, BristolMyers Squibb, Gilead Sciences, GlaxoSmithKline,
Merck, Pfizer and Roche. Andrew Carr has received
research grants or funding from Boehringer-Ingelheim
and Roche; consultancy fees from BoehringerIngelheim, Bristol-Myers Squibb and GlaxoSmithKline;
lecture sponsorships from Abbott, BoehringerIngelheim, Bristol-Myers Squibb and GlaxoSmithKline;
and has served on advisory boards for Bristol-Myers
Squibb, GlaxoSmithKline and Roche.
References
1.
Carr A. HIV lipodystrophy: risk factors, pathogenesis,
diagnosis and management. AIDS 2003, 17(Suppl
1):S141–148.
2.
Miller J, Carr A, Emery S, et al. HIV lipodystrophy: prevalence, severity and correlates of risk in Australia. HIV Med
2003, 4:293–301.
3.
Bernasconi E, Boubaker K, Junghans C, et al. Swiss HIV
Cohort Study. Abnormalities of body fat distribution in
HIV-infected persons treated with antiretroviral drugs: the
Swiss HIV Cohort Study. J AIDS 2002, 31:50–55.
4.
Lichtenstein KA, Ward DJ, Moorman AC, et al. HIV
Outpatient Study Investigators. Clinical assessment of HIVassociated lipodystrophy in an ambulatory population.
AIDS 2001, 15:1389–1398.
5.
Gallant JE, Staszewski S, Pozniak AL, et al. 903 Study
Group. Efficacy and safety of tenofovir DF vs stavudine in
combination therapy in antiretroviral-naive patients; a 3year randomized trial. JAMA 2004, 292:191–201.
6.
Carr A, Workman C, Smith D, et al. for the Mitochondrial
Toxicity (MITOX) Study Group. Abacavir substitution for
nucleoside analogs in patients with HIV lipoatrophy: a
randomized trial. JAMA 2002, 288:207–215.
7.
Carr A, Hudson J, Chuah J, et al. for the PIILR investigators. HIV protease inhibitor substitution in patients with
lipodystrophy: a randomised, multicentre, open-label study.
AIDS 2001, 15:1811–1822.
8.
Martinez E, Mocroft A, Garcia-Viejo MA, et al. Risk of
lipodystrophy in HIV-1-infected patients treated with
protease inhibitors: a prospective cohort study. Lancet
2001, 357:592–598.
9.
Mallon PWG, Miller J, Cooper DA, Carr A. Prospective
evaluation of the effects of antiretroviral therapy on body
composition in HIV-1-infected men starting therapy. AIDS
2003, 17:971–999.
10. Fellay J, Boubaker K, Ledergerber B, et al. Swiss HIV
Cohort Study. Prevalence of adverse events associated with
potent antiretroviral treatment. Lancet 2001,
358:1322–1327.
11. Carr A, Emery SE, Law M, Puls R, Lundgren JD, Powderly
WG. HIV Lipodystrophy Case Definition Study Group. An
objective case definition of lipodystrophy in HIV-infected
adults. Lancet 2003, 361:726–735.
12. Carr A, Law M. HIV Lipodystrophy Case Definition Study
Group. An objective lipodystrophy severity grading scale
using the lipodystrophy case definition score. J AIDS 2003,
33:571–576.
13. Puls RL, Law M, Freund J, et al. The effect of centralized
body composition data analysis on an objective case
185
AVT-765 Law.qxp
15/3/06
12:26 pm
Page 186
M Law et al.
definition of HIV lipodystrophy. Antivir Ther 2002,
7(Suppl 3):80.
14. Carr A, Samaras K, Burton S, et al. A syndrome of peripheral lipodystrophy, hyperlipidaemia and insulin resistance
in patients receiving HIV protease inhibitors. AIDS 1998,
12:F51–58.
15. Gervasoni C, Ridolfo AL, Trifirò G, et al. Redistribution of
body fat in HIV-infected women undergoing combined
antiretroviral therapy. AIDS 1999, 13:465–472.
16. Carr A, Samaras K, Thorisdottir A, Kaufmann G,
Chisholm DJ, Cooper DA. Diagnosis, prediction and
natural course of HIV protease inhibitor-associated lipodystrophy, hyperlipidaemia and diabetes mellitus. Lancet
1999, 353:2893–2899.
17. Saint-Marc T, Partisani M, Poizot-Martin I, et al. A
syndrome of peripheral fat wasting (lipodystrophy) in
patients receiving long-term nucleoside analogue therapy.
AIDS 1999, 13:1659–1667.
18. Carr A, Miller J, Law M, Cooper DA. A syndrome of
lipoatrophy, lactic acidaemia and liver dysfunction associated with HIV nucleoside analogue therapy: contribution to
protease inhibitor-related lipodystrophy syndrome. AIDS
2000, 14:F25–F32.
19. Thiebaut R, Daucourt V, Mercie P, et al. Lipodystrophy,
metabolic disorders, and human immunodeficiency virus
infection: Aquitaine Cohort, France, 1999. Clin Infect Dis
2000; 31:1482–1487.
20. Heath K, Hogg RS, Chan KJ, et al. Lipodystrophy-associated morphological, cholesterol and triglyceride
abnormalities in a population-based HIV/AIDS treatment
database. AIDS 2001, 15:231–240.
21. Lichtenstein KA, Delaney KM, Armon C, et al.; HIV
Outpatient Study Investigators. Incidence of and risk
factors for lipoatrophy (abnormal fat loss) in ambulatory
HIV-1-infected patients. J AIDS 2003, 32:48–56.
22. Carr A, Workman C, Carey D, et al.; Rosey investigators.
No effect of rosiglitazone for HIV-1 lipoatrophy: a
randomised, double-blind, placebo-controlled trial. Lancet
2004, 363:429–438.
23. Dubé MP, Parker RA, Tebas P, et al. Glucose metabolism,
lipid, and body fat changes in antiretroviral-naive subjects
randomized to nelfinavir or efavirenz plus dual nucleosides.
AIDS 2005; 19:1807–1818.
Received 15 June 2005, accepted 19 December 2005
186
© 2006 International Medical Press