A Risk-Factor Guided Approach to Reducing Lactic
Acidosis and Hyperlactatemia in Patients on
Lynn T. Matthews1*, Janet Giddy2, Musie Ghebremichael3, Jane Hampton2, Anthony J. Guarino4, Aba
Ewusi5, Emma Carver6, Karen Axten3, Meghan C. Geary7, Rajesh T. Gandhi3,8, David R. Bangsberg3,9,10
1Division of Infectious Disease, Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States of America, 2HIV Program, McCord Hospital, Durban, South
Africa, 3Ragon Institute of Massachusetts General Hospital, Massachusetts Institute of Technology, and Harvard University, Boston, Massachusetts, United States of
America, 4Institute of Health Professions, Massachusetts General Hospital, Boston, Massachusetts, United States of America, 5Division of Internal Medicine, Harvard
Vanguard Medical Associates, Boston, Massachusetts, United States of America, 6Department of Emergency Medicine, University Hospital of Wales, Cardiff, Wales,
7Albany Medical College, Albany, New York, United States of America, 8Division of Infectious Disease, Massachusetts General Hospital, Boston, Massachusetts, United
States of America, 9Mbarara University of Science and Technology, Mbarara, Uganda, 10Center for Global Health, Massachusetts General Hospital, Boston, Massachusetts,
United States of America
Background: Stavudine continues to be used in antiretroviral treatment (ART) regimens in many resource-limited settings.
The use of zidovudine instead of stavudine in higher-risk patients to reduce the likelihood of lactic acidosis and
hyperlactatemia (LAHL) has not been examined.
Methods: Antiretroviral-naı ¨ve, HIV-infected adults initiating ART between 2004 and 2007 were divided into cohorts of those
initiated on stavudine- or zidovudine-containing therapy. We evaluated stavudine or zidovudine use, age, sex, body mass
index (BMI), baseline CD4 cell count, creatinine, hemoglobin, alanine aminotransferase, and albumin as predictors of time to
LAHL with Cox Proportional Hazards (PH) regression models.
Results: Among 2062 patients contributing 2747 patient years (PY), the combined incidence of LAHL was 3.2/100 PY in
those initiating stavudine- and 0.34/100 PY in those initiating zidovudine-containing ART (RR 9.26, 95% CI: 1.28–66.93). In
multivariable Cox PH analysis, stavudine exposure (HR 14.31, 95% CI: 5.79–35.30), female sex (HR 3.41, 95% CI: 1.89–6.19),
higher BMI (HR 3.21, 95% CI: 2.16–4.77), higher creatinine (1.63, 95% CI: 1.12–2.36), higher albumin (HR 1.04, 95% CI: 1.01–
1.07), and lower CD4 cell count (HR 0.96, 95% CI: 0.92–1.0) at baseline were associated with higher LAHL rates. Among
participants who started on stavudine, switching to zidovudine was associated with lower LAHL rates (HR 0.15, 95% CI: 0.06–
0.35). Subgroup analysis limited to women with higher BMI$25 kg/m2 initiated on stavudine also showed that switch to
zidovudine was protective when controlling for other risk factors (HR 0.21, 95% CI .07–0.64).
Conclusions: Stavudine exposure, female sex, and higher BMI are strong, independent predictors for developing LAHL.
Patients with risk factors for lactic acidosis have less LAHL while on zidovudine- rather than stavudine-containing ART.
Switching patients from stavudine to zidovudine is protective. Countries continuing to use stavudine should avoid this drug
in women and patients with higher BMI.
Citation: Matthews LT, Giddy J, Ghebremichael M, Hampton J, Guarino AJ, et al. (2011) A Risk-Factor Guided Approach to Reducing Lactic Acidosis and
Hyperlactatemia in Patients on Antiretroviral Therapy. PLoS ONE 6(4): e18736. doi:10.1371/journal.pone.0018736
Editor: Landon Myer, University of Cape Town, South Africa
Received November 30, 2010; Accepted March 12, 2011; Published April 11, 2011
Copyright: ? 2011 Matthews et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Dr. Matthews’ work was supported by the Mark and Lisa Schwartz Family Foundation and by a postdoctoral fellowship in tropical infectious diseases
from the Burroughs Wellcome Fund/American Society for Tropical Medicine and Hygiene. Dr. Gandhi is supported by NIH R01 AI066992-04A1 and NIH
G08LM008830-01 and by grants to the AIDS Clinical Trials Group (NIH U01 AI 694722) and the Harvard University Center for AIDS Research (NIH 2P30 AI060354-
06). Dr. Bangsberg was supported by NIH grant MH K-24 87227. The authors are also grateful to Dr. Heather Ribaudo of the Harvard University Center for AIDS
Research Biostatistical Core (NIH #AI060354). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: firstname.lastname@example.org
Lactic acidosis is a potentially fatal side effect of nucleoside
analog reverse transcriptase inhibitors (NRTIs) [1,2], which are
commonly used in combination antiretroviral therapy (ART).
This complication is related to NRTI-induced mitochondrial
toxicity possibly dueto structuralsimilarities between
mitochondrial DNA polymerase and HIV-reverse transcriptase
(the target of NRTIs) . The incidence of lactic acidosis among
patients on ART ranges from 1–4 per 100 patient years in
resource-rich settings and is as high as 10 per 100 patient years
in sub-Saharan African cohorts [4,5,6,7,8,9,10,11]. The lactic
acidosis case-fatality rate in resource-limited settings can be as
high as 60% .
PLoS ONE | www.plosone.org1April 2011 | Volume 6 | Issue 4 | e18736
Of the NRTIs, the dideoxynucleosides (stavudine and didano-
sine) confer the highest risk of lactic acidosis [1,2,5]. While
stavudine is rarely used in resource-rich settings and is no longer
recommended by the World Health Organization for initial
treatment of HIV-1 infection , it remains an important
component of standard ART regimens in many resource-limited
countries, largely due to cost [14,15]. In South Africa where
stavudine is no longer recommended for use in first-line therapy,
patients receiving stavudine-containing ART are only switched if
there is evidence of toxicity, again because of financial constraints.
In settings where stavudine is widely prescribed, lactic acidosis is a
frequent cause of morbidity and mortality [1,2,4,5,6,7,8,9,10,16]
and is associated with high losses to follow-up and treatment
Observational studies suggest that specific risk factors associated
with the development of hyperlactatemia include female sex
[1,4,7,11,16,17,18], elevated weight or body-mass index (BMI)
[1,11,16,17,18], older age (.40 years) [1,11], and lower CD4 cell
counts . Where financial constraints prevent comprehensive
adoption of less-toxic agents, a risk factor-guided approach to
choosing an initial regimen may reduce the incidence of lactic
acidosis. Studies have shown that after resolution of lactic acidosis
it is safe to treat patients with zidovudine (an alternative thymidine
analog NRTI which is widely used in resource-limited settings)
[10,19], but none have examined whether avoiding stavudine in
patients with lactic acidosis risk factors reduces incidence of lactic
acidosis or hyperlactatemia.
Until April 2010, first-line therapy in South Africa included
stavudine, lamivudine, and either efavirenz or nevirapine. Based
on observational findings from a site-specific study that identified a
high incidence of lactic acidosis in women with BMI$28 kg/m2,
in August 2005 the HIV Clinic at McCord Hospital in Durban,
South Africa substituted zidovudine for stavudine in initial ART
for patients with these two risk factors . The policy continued
until March 2007, when the clinic was accredited as a Department
of Health site and required to follow Department of Health
guidelines for ART, including the use of stavudine as part of initial
To evaluate the impact of risk factor-guided selection of initial
therapy, we compared the combined incidence of lactic acidosis
and hyperlactatemia among treatment-naive patients initiating
stavudine-containing therapy with those starting zidovudine-
containing therapy. We hypothesized that risk-factor-guided
ART (initiating women with BMI$28 kg/m2 on zidovudine
rather than stavudine) would be associated with decreased
incidence of lactic acidosis and hyperlactatemia. We also assessed
predictors of lactic acidosis and hyperlactatemia.
Ethics approvals were obtained from the McCord Hospital
Medical Ethics Research Committee and from the Partners
Healthcare Institutional Review Board (Boston, MA). Given the
nature of the study (retrospective chart review), the requirement
for informed consent was waived by the ethics committees.
Study design and population
Patient data were collected from the outpatient HIV clinic at
McCord Hospital in Durban, South Africa which has initiated
over 8000 patients on ART. During the study period, initial ART
included two NRTIs and one NNRTI: stavudine (30 mg twice
daily; 40 mg twice daily if weight .60 kg) or zidovudine plus
lamivudine and either efavirenz or nevirapine.
The study population included antiretroviral (ARV)-naı ¨ve,
HIV-infected adults (age $18 years) with baseline laboratory data
and at least one follow-up visit after ART initiation. Two
retrospective cohorts were identified. The first cohort included
patients who initiated stavudine-containing therapy between July
2004 and March 2007. The second cohort included patients who
initiated zidovudine-containing ART between July 2004 and
March 2007. Both cohorts included patients who initiated ART
between August 2005 and March 2007 when the clinic made
women with BMI$28 kg/m2 eligible for initiation of zidovudine -
containing therapy or for regimen switch from stavudine to
Outcomes and their measurement
The primary outcome was event-free survival defined as the
time from treatment initiation to development of lactic acidosis
(symptomatic or asymptomatic) or hyperlactatemia (symptomatic
or asymptomatic) (Table 1). Lactic acidosis and hyperlactatemia
were defined based on AIDS Clinical Trials Group criteria .
Lactic acidosis is defined as having a lactate level above the upper
limit of normal (4.4 mmol/L) along with evidence of acidosis
(bicarbonate level ,20 mmol/L or pH,7.35). Hyperlactatemia is
defined as a lactate level greater than the upper limit of normal
without evidence of acidosis. Cases of symptomatic lactic acidosis
or hyperlactatemia met the above criteria and had new, otherwise
unexplained symptoms of nausea, vomiting; abdominal pain,
discomfort, or distention; increased hepatic transaminases; fatigue;
dyspnea; weight loss ($5%); or muscle weakness. Because these
patients were ambulatory and often did not have repeat
measurements, confirmed elevation of lactate levels was not
required if at least two symptoms were present.
Blood was drawn for lactate levels without use of a tourniquet
and specimens were transported on ice and processed within four
hours (Beckman Coulter, Synchron systems, California, USA). A
handheld lactate detection device, a reliable proxy for serum
samples, was introduced in 2006 (Accutrend model #3012522)
[21,22] and was used for initial screening in addition to serum
Outcomes were classified from a review of the medical records
of patients initiating ART during the study period. This review
was facilitated by the requirement that clinicians record the reason
for any change or discontinuation in ARV regimen from an
electronic pull down menu. For patients who had a regimen
change noted in the electronic record, paper charts were reviewed
for the following: 1) documentation of a regimen change due to
lactic acidosis; 2) documentation of a regimen change and signs or
symptoms that could be consistent with lactic acidosis or
hyperlactatemia (nausea, vomiting, abdominal discomfort, bloat-
ing, increased hepatic transaminases, fatigue, dyspnea, weight loss,
muscle weakness); 3) documentation of a regimen change without
specific reason listed; 4) death. Serum lactate test results for all
study patients were reviewed. A lactate value above 3 mmol/L
prompted review of the medical record for symptoms of
hyperlactatemia, all available lactate values, other possible causes
for symptoms or elevated lactate levels, and clinical outcome. Data
were abstracted using standardized abstraction forms (LM, AE,
JH). For a subset of patients (n=20), two physicians carried out the
abstractions with 100% agreement on outcome classification (JH,
LM). Cases with unclear outcomes were adjudicated by a senior
We also identified patients for whom clinicians had changed
ART due to peripheral neuropathy, lipodystrophy, high BMI, and
drug resistance, as indicated in the electronic medical record. We
identified patients with regimen switch for clinical suspicion of LA
Avoiding ART-Related Lactic Aacidosis
PLoS ONE | www.plosone.org2 April 2011 | Volume 6 | Issue 4 | e18736
or HL but who did not meet criteria. These subjects were not
censored at change in regimen but followed out to a total two years
of follow-up from treatment initiation. In addition, we identified
subjects who changed clinic site, stopped ART, died or were lost to
Covariates were obtained from paper chart abstractions and
included weight at treatment initiation (within 3 months) and
height. Weights obtained during pregnancy were excluded. BMI
was calculated (kg/m2) for all subjects in whom height and
baseline weight were available. Sex, date of birth and baseline (the
last value prior to ART initiation, or within 2 weeks) CD4 count,
creatinine, hemoglobin, alanine aminotransferase, and albumin
were extracted from the electronic record or the paper chart. All
specimens were processed using standardized methods at labora-
tories in Durban.
Time to event or censor
The primary outcome was 2-year event-free survival (EFS)
defined as the time from treatment initiation up to development of
lactic acidosis or hyperlactatemia. Patients were also censored for
loss to follow-up, change in clinic site, termination of treatment,
death, or at study end. All others were followed for two years or
until the primary outcome. Time on stavudine and zidovudine was
calculated from start and stop dates entered by clinicians in the
We calculated crude incidence rates for the combined primary
outcome (LAHL), the combined incidence of peripheral neurop-
athy and lipodystrophy, death, and loss-to-follow-up. Confidence
intervals for event rates based on initial therapy with stavudine or
zidovudine were estimated using methods for exact binomial
confidence intervals and compared using Chi-square tests .
Kaplan-Meier curves were plotted for event-free survival based on
initial treatment and rates were compared using the log-rank test
statistic. Univariate and multivariate analyses using Cox propor-
tional hazards (PH) regression models were utilized to assess the
effect of treatment on time to event . We evaluated time on
zidovudine or stavudine as a time-varying covariate to account for
variable time on drug among patients whose regimens were
switched in the absence of the outcome of interest (e.g. switch for
peripheral neuropathy or increased BMI). Covariates for multi-
variate analysis were selected based on significance (p value,0.05)
in univariate analysis and significant covariates in the literature.
CD4 count was modeled as a continuous variable with the effect
size reported per 10-cell increment. BMI was modeled on a
natural logarithmic scale with effect size reported per 30% shift. In
the full model, BMI deviated from the proportional hazards
assumption and was modeled with a time-dependent association
for early (within the first year) and late (after one year) failure. In
subgroup analysis, BMI followed proportional hazards. All
statistical analyses were carried out using SAS version 9.2 for
Baseline patient and disease characteristics
Two-thousand-sixty-two patients contributing 2747 person
years of follow-up were included in the study. The median age
was 34.7 years (IQR 29.8, 40.6) and 60% were women. Eighty-
nine percent initiated therapy with a stavudine-containing
regimen. One-hundred sixty one (77%) of those who were
initiated on a zidovudine-containing regimen were started because
of higher BMI or other perceived lactic acidosis risk factors. The
remaining patients were initiated on zidovudine because of pre-
existing lipodystrophy (,1%), peripheral neuropathy (,1%),
pregnancy (10%), or unknown reason (10%).
Median CD4 count at entry was 80 cells/mm3 (IQR 29–142).
Median BMI for subjects with complete data (88% had
documented weight at entry, 76% had documented height) was
22 kg/m2 (IQR 20, 26). Compared with those initiated on a
stavudine-containing regimen, patients started on zidovudine were
older, more likely to be female, had a higher BMI, higher CD4 cell
count, higher albumin and higher hemoglobin. Other character-
istics are described in Table 2.
Outcomes for full cohort
In intention to treat analysis, combined incidence of LAHL was
3.2/100 PY in the stavudine- and 0.34/100 PY in the zidovudine-
initiated group (RR 9.26, 95% CI 1.28–66.93, p=.007). There
were 36 lactic acidosis and 43 hyperlactatemia events in the
stavudine group. In contrast, there was 1 lactic acidosis event in
the zidovudine group: this occurred in a woman who initiated
zidovudine-based therapy because of high BMI (31 kg/m2); one
year later, she was switched to stavudine because of anemia; after
eight months on stavudine-containing ART, she was diagnosed
with lactic acidosis. Mortality due to causes other than LAHL was
8.3% and 2.8% in stavudine- and zidovudine-initiated patients,
respectively (RR=2.89, 95% CI:1.45–5.78, p=0.001). The
combined incidence of physician-reported peripheral neuropathy
and lipodystrophy was 16.8/100 PY in stavudine- and 0.34/100
PY in zidovudine-initiated groups (RR=59.84, 95% CI: 8.36–
Table 1. Criteria for lactic acidosis and hyperlactatemia outcomes1.
Abnormal values required
1Based on AACTG criteria .
2Bicarbonate ,20 mmol/L or pH,7.35.
3New or otherwise unexplained symptoms of nausea or vomiting, abdominal pain or discomfort, abdominal distention, increased hepatic transaminases, unexplained
fatigue, dyspnea, weight loss ($5%), or muscle weakness.
Avoiding ART-Related Lactic Aacidosis
PLoS ONE | www.plosone.org3 April 2011 | Volume 6 | Issue 4 | e18736
428.12, p,0.001). Loss to follow-up was equivalent between the
two groups (RR=1.42, 95% CI: 0.68–2.96, p=0.35). (Table 3)
In univariate Cox proportional hazards analysis, stavudine in
the initial treatment regimen, female sex, higher BMI, and higher
baseline albumin were each associated with increased risk of
LAHL (Table 4). The Kaplan Meier curve for time to LAHL
based on initial treatment regimen is shown in Figure 1 (p=.006).
In multivariable Cox PH regression to assess predictors of event-
free survival, hemoglobin and ALT were removed but age and
CD4 cell count were included because prior data and a priori
knowledge suggested an association with lactic acidosis .
Creatinine was into the full model when it was found to be
significant in subgroup analysis. The adjusted hazards of
(HR=14.31, 95% CI 5.79–35.30), women (HR=3.41, 95% CI:
1.89–6.19), subjects with higher BMI in the first year (HR=3.21,
95% CI: 2.16–4.77), higher albumin (HR=1.04, 95% CI:1.01–
1.07), higher creatinine (HR=1.63, 95% CI 1.12–2.36), or lower
baseline CD4 cell count (HR=0.96, 95% CI: 0.92–1.00) at
baseline (Table 5). Among those initiated on stavudine, the
hazards of experiencing LAHL was lower for those who were
switched to zidovudine during follow-up (HR 0.15, 95% CI
for those onstavudine
Outcomes for women with higher BMI
Women with BMI greater than or equal to 25 kg/m2
comprised 326 patients with 434 years of follow-up. The 194
women initiated on stavudine were younger; with lower BMI,
baseline CD4 cell count, creatinine, albumin, and hemoglobin
compared with 132 women initiated on zidovudine (Table 5).
Obese women initiated on stavudine had 22 LAHL events (8.0/
100 woman years), compared with 1 (0.53/100 woman years)
among those initiated on zidovudine (RR=9.94, 95% CI, 1.46–
67.91, p=.0002). When controlling for BMI, CD4 cell count,
albumin, creatinine, and age, stavudine use was associated with a
13-fold increase in hazards of LAHL (HR 13.37, 95% CI 4.31–
41.53) (Table 6). For women in this subgroup who initiated on
stavudine-containing therapy, switching to zidovudine was pro-
tective (HR 0.21, 95% CI 0.07–0.64, p=0.006).
Of the 194 women with higher BMI who initiated stavudine-
inclusive therapy, 137 were switched to zidovudine for reasons
other than LAHL. Baseline characteristics (age, BMI, CD4 cell
count, creatinine albumin, ALT, hemoglobin) were not signifi-
cantly different from women with higher BMI initiated on
stavudine-treatment who did not switch treatment arms. Women
were switched for high BMI (79, 56%); lipodystrophy, peripheral
neuropathy, or these plus elevated BMI (47, 34%); lab values and/
or symptoms suggestive of hyperlactatemia that did not meet
criteria for LAHL (7, 5%); and the remainder were switched for
anemia, pregnancy, rash or other reasons. These participants
subsequently contributed an additional 131.7 woman-years of
follow-up (mean 1.1 years 60.5) during which there were 5 LAHL
events (3.8/100 woman years). All but one event occurred within
2–8 weeks of switching off stavudine after an average of 0.660.4
years on stavudine, suggesting that the recent and cumulative
stavudine exposure contributed to the toxicity. When controlling
for other LAHL risk factors, switch to zidovudine conferred 80%
lower hazards of LAHL for this subgroup (HR 0.21, 95% CI 0.07–
0.64, p=.006). The remainder of women in this subgroup of obese
women, initiated on stavudine and switched to zidovudine,
included two who subsequently had anemia and two who died;
the rest were followed until the end of the study, change in service
provider, or a maximum of two years of follow-up without adverse
In our study of 2062 HIV-positive patients who initiated ART,
stavudine use confers a fourteen-fold increased risk of developing
hyperlactatemia or lactic acidosis when controlling for other risk
factors (HR 14.31, 95% CI 5.79–35.30). Other risk factors for the
primary outcome of LAHL were female sex, higher baseline BMI,
Table 2. Patient characteristics at study entry by treatment
Initial ART includes:
Number (patient years follow-up) 1853 (2460) 209 (287)
Age, years Mean (SD) 35.7 (8.3)37.8 (9.6)
Patient years of follow up
1.3 (0.7)1.4 (0.6)
Female n (%) 1078 (58.2)188 (90)
BMI (kg/m2) Median (IQR)22 (19, 24)30 (28, 33)
CD4 (cells/mm3) Median (IQR)75 (27, 138)129 (61, 172)
Creatinine (mg/dL) Median (IQR)1.0 (0.4)0.9 (0.3) .21
ALT (IU/L) Median (IQR) 24 (18, 35)23 (17, 32).27
Albumin (g/L) Median (IQR) 31.2 (7.3)34.9 (5.3)
Hemoglobin (g/dL) Median (IQR) 10.8 (2.1) 11.6 (1.3)
1Chi-square test was used for categorical variables, T-test for continuous where
mean and standard deviation reported, and Wilcoxon rank sum where median
and IQR reported.
Table 3. Incidence of mitochondrial toxicity, death and loss to follow-up by initial treatment.
Initial ART includes:
Ratio [95% CI]
Lactic acidosis or hyperlactatemia1
79 (3.2) 1 (0.3)9.26 [1.28, 66.93].007
Mortality due to cause other than LAHL 205 (8.3)8(2.8) 2.89 [1.45, 5.78] .001
Peripheral neuropathy or lipodystrophy2
414 (16.8)1 (0.3) 59.84 [8.36, 428.12]
Loss to follow-up 99 (4.0)8 (2.8)1.42 [0.68, 2.96].35
1Primary endpoint: 37 lactic acidosis, 43 hyperlactatemia.
2As indicated by clinician report in the medical record.
Avoiding ART-Related Lactic Aacidosis
PLoS ONE | www.plosone.org4 April 2011 | Volume 6 | Issue 4 | e18736
higher baseline creatinine or albumin, and lower initial CD4 cell
count. For patients who started a stavudine-containing regimen,
switching to zidovudine was associated with 85% lower hazards of
developing LAHL (HR 0.15, 95% CI 0.06–0.35). For the high-risk
subgroup of women with BMI$25 kg/m2 who initiated therapy
on stavudine-containing ART, switch to zidovudine was also
protective when controlling for other risk factors (HR 0.21, 95%
Our study adds to the literature by demonstrating that female
sex is a strong independent risk factor for developing LAHL
[1,4,7,11,16,17,18]. Higher weight has been associated with these
outcomes in prior studies, but this is the first to confirm a
relationship with BMI and LAHL when controlling for other
covariates [1,11,16,17,18]. For every 30% change in BMI (i.e. 18
to 23 kg/m2or 24 to 31 kg/m2), we observed a three-fold
increase in the LAHL rate (HR=3.21, 95% CI: 2.16–4.77). For
the full dataset, the effect was only significant in the first year of
follow-up which may reflect increased risk earlier in treatment or
insufficient power to detect an association after the first year.
During the study, patients with weight .60 kg received 80 mg of
stavudine daily, which has been linked to worse mitochondrial
toxicities compared to use of 60 mg . We were unable to
control for stavudine dose; thus, the high incidence of LAHL in
patients with higher BMI might be related to higher stavudine
dose. This possibility is supported by observations that patients on
higher dose stavudine (40 mg twice daily) have a higher incidence
of elevated lactate than those who receive lower doses (20 or
30 mg twice daily) [16,25]. However, given that multiple studies
involving varying stavudine dose have found an association
between higher weight or BMI and lactic acidosis [7,11,17] while,
in some cases, controlling for dose [16,18], it is unlikely that drug
dosing explains the entire effect. Furthermore, a three-fold
increase in hazards of LAHL was observed in our subgroup
analysis of women with BMI$25 kg/m2 (who likely received
uniform stavudine dosing).
Higher creatinine was associated with increased hazards of
LAHL, about 25% per 1 mg/dL unit increase in creatinine. This
risk factor has not been reported in prior univariate analyses and
has not been included in studies that control for other risk factors,
but is not unexpected given the kidney’s role in lactate metabolism
. We also found that higher albumin is associated with an
increased risk of LAHL and a small protective effect of higher
CD4 cell count at treatment initiation (4% decrease in hazards for
each 10-point increase in baseline CD4 cell count). Two other
Figure 1. Kaplan Meier curves for lactic acidosis/hyperlactatemia-free survival for patients initiated on either stavudine- or
zidovudine-containing antiretroviral therapy (p=.006).
Table 4. Univariate and multivariate Cox Regression Analysis for time to lactic acidosis or hyperlactatemia.
VariableHazards Ratio [95% CI]p-valueAdjusted Hazards Ratio [95% CI] p-value
Age (years)1.02 [0.99, 1.04] .14–.22
Female sex2.22 [1.31, 3.75].0033.42 [1.89, 6.19]
BMI in first year (30% change kg/m2) 1.53 [1.10, 2.12] .013.21 [2.16, 4.77]
BMI after first year (30% change kg/m2) 0.76 [0.44, 1.32].33–.55
Stavudine use5.81 [2.52, 13.43]
,.000114.31 [5.79, 35.30]
Initial CD4 count (10 cells/mm3) 0.99 [0.96, 1.03] .80 0.96 [0.92, 1.00].04
Initial Albumin (g/L) 1.04 [1.01, 1.07].004 1.04 [1.01, 1.07].004
Initial Creatinine (mg/dL) 1.00 [0.99, 1.01].091.63 [1.12, 2.36].010
Initial ALT (IU/L) 1.00 [0.99, 1.01] .38––
Hemoglobin (g/dL)1.07 [0.95, 1.19].26––
Multivariate model with 80 events, 1546 subjects with complete data for all variables.
Avoiding ART-Related Lactic Aacidosis
PLoS ONE | www.plosone.org5April 2011 | Volume 6 | Issue 4 | e18736
studies have also found an association between CD4 cell count and
lactic acidosis , . Each of these associations (CD4 cell count,
creatinine and albumin) was small with confidence intervals close
The Lactic Acidosis International Study group showed an
association of older age (age .40 years) with the development of
LAHL. This was not seen in our cohort nor in other studies based
in Southern Africa [16,17,18]. The majority of their subjects were
from Europe and the Americas with an older age distribution than
in our study (mean of 42 years for cases vs. 35 years).
For the subgroup of women with higher BMI, stavudine use,
when controlling for other risk factors, remained associated with a
significant increase in risk of LAHL. Switching these women to
zidovudine conferred an 80% reduction in hazards of LAHL.
These data suggest that higher-risk individuals should be switched
off stavudine-based therapy in order to reduce adverse events.
There are several limitations to this study. The two treatment
groups were quite different as demonstrated in Table 1. In our
model we were able to control for the variables in the model, but
not for unmeasured confounders (e.g. HIV clinical stage).
Clinicians may have been more likely to test for hyperlactatemia
in patients on stavudine resulting in a detection bias. However,
14% of subjects initiated on stavudine and 18% of subjects
initiated on zidovudine had at least one serum lactate level
checked during the study period, suggesting that serum lactate
testing was not biased towards subjects on stavudine. We do not
think deployment of the handheld lactate machine for screening
(introduced in 2006) differentially affected case finding between
the two groups. Any patient with a positive handheld device test
required confirmatory serological testing; as above, testing rates
were not higher for subjects initiated on stavudine. Our data, in
combination with prior data evaluating risk factors for lactic
acidosis and hyperlactatemia, strongly suggest that women and
patients with higher BMI treated with stavudine are at high risk for
developing LAHL. In addition, our data demonstrate that using
zidovudine rather than stavudine, even among patients at highest
risk for mitochondrial toxicities, dramatically reduces the risk of
developing lactic acidosis. Further, for patients initiated on
stavudine-based therapy, switching to zidovudine is protective.
Although a recent Cochrane Review concluded there is no
difference in treatment outcomes (toxicity, death, disease progres-
sion) for stavudine- compared with zidovudine-based ART, the
randomized-controlled trials on which their analyses were based
included patients from North America, the Caribbean, Australia
and China . However, most studies observing high rates of
mitochondrial toxicities include patients from sub-Saharan Africa.
As of April 2010, first-line ART in South Africa includes
tenofovir, lamivudine, and either efavirenz or nevirapine .
However, because of drug shortages many clinics remain unable to
initiate all patients on tenofovir-based therapy and are not able to
routinely switch patients to tenofovir-containing regimens. Thus,
in South Africa (and other countries still using stavudine) these
findings will be helpful in identifying patients who are at highest
risk for stavudine-induced complications. While all patients will
benefit from using alternatives to stavudine, avoiding this drug in
women and patients with higher BMI may offer an effective and
practical strategy for reducing the incidence of lactic acidosis and
hyperlactatemia until countries can completely eliminate use of
We would like to acknowledge the patients, clinicians and monitoring and
evaluation staff at Sinikithemba. We would also like to acknowledge data
capturers and McCord volunteers Lisa Bevilacqu, Anthony Sawyer, Winn
Seay, Mary Gallo, Dr. Hannah Willoughby, and Dr. Eileen Scully. We are
also grateful to Dr. Heather Ribaudo of the Harvard University Center for
AIDS Research Biostatistical Core and Dr. Roger Davis for statistical
Conceived and designed the experiments: LTM JG JH RTG DRB.
Performed the experiments: LTM JH AE EC KA MCG. Analyzed the
data: LTM MG AJG. Wrote the paper: LTM JG RTG DRB. Chart
abstraction and data entry: LTM JH AE EC KA MCG.
Table 5. Patient characteristics at study entry by treatment
arm, limited to women with BMI$25 kg/m2.
Initial ART includes:
194 (274)132 (190)
Age (years) Mean
36 (7) 38 (9).03
Patient years of
follow up Mean
1.4 (0.6) 1.4 (0.6) .64
27 (26, 30) 30 (29, 34)
99 (64)122 (58).0003
0.85 (0.77, 0.94)0.88 (0.80, 0.98) .02
ALT (IU/L) Median
22 (17, 32) 22 (17, 30).97
32 (6) 36 (4)
11.1 (1.7) 11.7 (1.1).0001
1T-test for continuous where mean and standard deviation reported, and
Wilcoxon rank sum where median and IQR reported.
Table 6. Multivariate Cox Regression Analysis for time to
lactic acidosis or hyperlactatemia limited to women with
Adjusted Hazards Ratio
[95% CI] p-value
BMI (30% change kg/m2)3.15 [1.39, 7.17].005
Stavudine use 13.37 [4.31, 41.53]
Initial CD4 (10 cells/mm3)– 0.99
Initial Albumin (g/L) 1.01 [0.94, 1.08]0.83
Initial Creatinine (mg/dL) 1.75 [1.17, 2.62].006
Multivariate model with 20 events, 298 subjects with complete data for all
Avoiding ART-Related Lactic Aacidosis
PLoS ONE | www.plosone.org6 April 2011 | Volume 6 | Issue 4 | e18736
1. Lactic Acidosis International Study Group (2007) Risk factors for lactic acidosis
and severe hyperlactataemia in HIV-1-infected adults exposed to antiretroviral
therapy. AIDS 21: 2455–2464.
2. Wohl DA, McComsey G, Tebas P, Brown TT, Glesby MJ, et al. (2006) Current
concepts in the diagnosis and management of metabolic complications of HIV
infection and its therapy. Clin Infect Dis 43: 645–653.
3. Shibuyama S, Gevorkyan A, Yoo U, Tim S, Dzhangiryan K, et al. (2006)
Understanding and avoiding antiretroviral adverse events. Current Pharmaceu-
tical Design 12: 1075–1090.
4. Bolhaar MG, Karstaedt AS (2007) A high incidence of lactic acidosis and
symptomatic hyperlactatemia in women receiving highly active antiretroviral
therapy in Soweto, South Africa. Clin Infect Dis 45: 254–260.
5. Boubaker KMF, Sudre P, Furrer H, Haensel A, Hirschel B, Boggian K,
Chave J.-P, Bernasconi E, Egger M, Opravil M, Rickenbach M, Francioli P,
Telenti A (2001) Hyperlactatemia and antiretroviral therapy: The Swiss HIV
cohort study. Clin Infect Dis 33: 1931.
6. Fabian J, Venter WD, Mkhabela L, Levin JB, Baker L, et al. (2008)
Symptomatic hyperlactataemia in adults on antiretroviral therapy: a single-
centre experience. S Afr Med J 98: 795–800.
7. Geddes R, Knight S, Moosa MY, Reddi A, Uebel K, et al. (2006) A high
incidence of nucleoside reverse transcriptase inhibitor (NRTI)-induced lactic
acidosis in HIV-infected patients in a South African context. S Afr Med J 96:
8. Kumarasamy N, Venkatesh KK, Cecelia AJ, Devaleenol B, Saghayam S, et al.
(2008) Gender-based differences in treatment and outcome among HIV patients
in South India. J Womens Health (Larchmt) 17: 1471–1475.
9. Moyle GJ, Datta D, Mandalia S, Morlese J, Asboe D, et al. (2002)
Hyperlactataemia and lactic acidosis during antiretroviral therapy: relevance,
reproducibility and possible risk factors. Aids 16: 1341–1349.
10. Stead D, Osler M, Boulle A, Rebe K, Meintjes G (2008) Severe hyperlacta-
taemia complicating stavudine first-line antiretroviral therapy in South Africa.
Antivir Ther 13: 937–943.
11. Wester CW, Okezie OA, Thomas AM, Bussmann H, Moyo S, et al. (2007)
Higher-than-expected rates of lactic acidosis among highly active antiretroviral
therapy-treated women in Botswana: preliminary results from a large
randomized clinical trial. J Acquir Immune Defic Syndr 46: 318–322.
12. Stenzel MS, Carpenter CC (2000) The management of the clinical complica-
tions of antiretroviral therapy. Infect Dis Clin North Am 14: 851–878, vi.
13. World Health Organization (2010) Antiretroviral therapy for HIV infection in
adults and adolescents: recommendations for a public health approach-2010
revision. Geneva: World Health Organization.
14. Murphy RA, Sunpath H, Kuritzkes DR, Venter F, Gandhi RT (2007)
Antiretroviral therapy-associated toxicities in the resource-poor world: the
challenge of a limited formulary. J Infect Dis 196 Suppl 3: S449–456.
15. Rosen S, Long L, Fox M, Sanne I (2008) Cost and cost-effectiveness of switching
from stavudine to tenofovir in first-line antiretroviral regimens in South Africa.
J Acquir Immune Defic Syndr 48: 334–344.
16. van Griensven J, Zachariah R, Rasschaert F, Mugabo J, Atte EF, et al.
Stavudine- and nevirapine-related drug toxicity while on generic fixed-dose
antiretroviral treatment: incidence, timing and risk factors in a three-year cohort
in Kigali, Rwanda. Trans R Soc Trop Med Hyg 104: 148–153.
17. Boulle A, Orrel C, Kaplan R, Van Cutsem G, McNally M, et al. (2007)
Substitutions due to antiretroviral toxicity or contraindication in the first 3 years
of antiretroviral therapy in a large South African cohort. Antivir Ther 12:
18. Osler M, Stead D, Rebe K, Meintjes G, Boulle A (2009) Risk factors for and
clinical characteristics of severe hyperlactataemia in patients receiving
antiretroviral therapy: a case-control study. HIV Med 11: 121–129.
19. Lonergan JT, Barber RE, Mathews WC (2003) Safety and efficacy of switching
to alternative nucleoside analogues following symptomatic hyperlactatemia and
lactic acidosis. AIDS 17: 2495–2499.
20. Adult AIDS Clinical Trial Group (2003) AACTG toxicity evaluation group,
Chair Rob Murphy.
21. Ivers LC, Mukherjee JS (2006) Point of care testing for antiretroviral therapy-
related lactic acidosis in resource-poor settings. AIDS 20: 779–780.
22. Kiragga AK, Ocama P, Reynolds SJ, Kambugu A, Ojiambo H, et al. (2008)
Validation of a portable hand-held lactate analyzer for determination of blood
lactate in patients on antiretroviral therapy in Uganda. J Acquir Immune Defic
Syndr 49: 564–566.
23. Cox D, Snell E (1989) Analysis of Binary Data, 2nd Edition. London: Chapman
24. Harrell FE, Jr. (2001) Regression modeling strategies: With applications to linear
models, logistic regression, and survival analysis. New York: Springer-Verlag.
25. Hill A, Ruxrungtham K, Hanvanich M, Katlama C, Wolf E, et al. (2007)
Systematic review of clinical trials evaluating low doses of stavudine as part of
antiretroviral treatment. Expert Opinion in Phamacotherapy 8.
26. Bellomo R (2002) Bench-to-bedside review: lactate and the kidney. Crit Care 6:
27. Bonnet F, Bonarek M, Morlat P, Mercie P, Dupon M, et al. (2003) Risk factors
for lactic acidosis in HIV-infected patients treated with nucleoside reverse-
transcriptase inhibitors: a case-control study. Clin Infect Dis 36: 1324–1328.
28. Spaulding A, Rutherford GW, Siegfried N Stavudine or zidovudine in three-
drug combination therapy for initial treatment of HIV infection in antiretroviral-
naive individuals. Cochrane Database Syst Rev CD008651.
29. South African National AIDS Council (2010) The South African Antiretroviral
Treatment Guidelines Department of Health; Republic of South Africa. 8 p.
Avoiding ART-Related Lactic Aacidosis
PLoS ONE | www.plosone.org7April 2011 | Volume 6 | Issue 4 | e18736