Determinants of epidermal nerve fibre density in antiretroviral-naïve HIV-infected individuals

Authors


Correspondence: Dr Cecilia M. Shikuma, Hawaii Center for AIDS, John A. Burns School of Medicine, University of Hawaii at Manoa, 3675 Kilauea Ave. Young Bldg 5th Floor, Honolulu, HI 96816, USA. Tel: (808) 737 2751; fax: (808) 735 7047; e-mail: shikuma@hawaii.edu

Abstract

Objectives

Distal leg epidermal nerve fibre density (ENFD) is a validated predictor of small unmyelinated nerve fibre damage and neuropathy risk in HIV infection. As pre-existing damage may increase the risk of neuropathy following antiretroviral (ARV) therapy, particularly when the regimen contains stavudine (d4T), we assessed the relationship between ENFD and various parameters including mitochondrial factors in HIV-infected Thai individuals naïve to ARV therapy.

Methods

Distal leg and proximal thigh ENFDs were quantified in HIV-infected Thai individuals without neuropathy prior to randomization to a HIV clinical trial that focused on mitochondrial toxicity issues. We assessed their association with various clinical and immunovirological parameters as well as with peripheral blood mononuclear cell (PBMC) mitochondrial (mt) DNA copies/cell, oxidative phosphorylation (OXPHOS) complex I (CI) and complex IV (CIV) enzyme activities, and mt 8-oxo-deoxyguanine (8-oxo-dG) break frequencies.

Results

In 132 subjects, the median (interquartile range) ENFD (fibres/mm) values were 21.0 (16.2–26.6) for the distal leg and 31.7 (26.2–40.0) for the proximal thigh. By linear regression, lower CD4 count (P < 0.01), older age (P < 0.01), increased body mass index (BMI) (P = 0.04), increased height (P = 0.02), and higher PBMC OXPHOS activity as measured by CIV activity (P = 0.02) were associated with lower distal leg ENFD.

Conclusions

Older age, increased height, higher BMI, poorer immunological status and higher PBMC OXPHOS activity are associated with lower distal leg ENFD in HIV-infected subjects free of neuropathy prior to initiation of first-time ARV therapy.

Introduction

HIV-associated sensory neuropathy (HIV-SN) is a common neurological complication of HIV infection characterized by bilateral lower extremity burning pain and numbness. The aetiologies for HIV-SN include HIV immunovirological factors as well as antiretroviral (ARV) toxic neuropathy associated with the use of potentially neurotoxic nucleoside reverse transcriptase inhibitors (NRTIs) including stavudine (d4T), a drug that continues to be used as a component of ARV therapy in some developing countries [1].

Distal leg epidermal nerve fibre density (ENFD) is a validated predictor of small unmyelinated nerve fibre damage and neuropathy risk in various diseases including HIV infection [2-4]. We determined ENFD in HIV-infected Thai individuals without signs or symptoms of neuropathy prior to the initiation of first-time ARV therapy to investigate which factors were associated with lower ENFD and therefore might increase neuropathy risk following initiation of d4T-containing regimens. In addition to epidemiological and HIV-specific factors such as CD4 cell count and plasma HIV RNA, we assessed mitochondrial parameters based on the known role of mitochondrial toxicity in the pathogenesis of neuropathy following the use of potentially neurotoxic NRTIs such as d4T.

Methods

Subjects

This analysis utilized baseline data on subjects who were enrolled in SEARCH (South East Asia Research Collaboration with Hawaii; www.searchthailand.org/) 003, a 150-patient, 72-week, two-site ARV clinical trial in ARV-naïve subjects conducted in Thailand at the Thai Red Cross AIDS Research Centre (TRCARC) in Bangkok and at the Queen Savang Vadhana Memorial Hospital in Chonburi, Thailand (www.clinicaltrials.gov identification NCT00669487). SEARCH 003 compared, in a randomized fashion, rates of anaemia, lipoatrophy and neuropathy among three ARV regimens differing by NRTI backbone. Specifically, a backbone of 24 weeks of stavudine (d4T) followed by a switch to zidovudine (ZDV) was compared with continuous ZDV and with continuous tenofovir (TDF) for the entire 72-week duration of the study. Skin punch biopsies and ENFD assessments were performed as elective procedures at baseline, week 24 and week 72 to allow an in-depth evaluation of neuropathy risk during ARV therapy. The baseline ENFD and its relationship to various parameters prior to initiation of ARV therapy are the topics of this report. The SEARCH 003 study was approved by the Chulalongkorn University Institutional Review Board (IRB) and the Queen Savang Vadhana Memorial Hospital IRB as primary IRBs of record and by the University of Hawaii Committee on Human Subjects and the University of California San Francisco Committee for Human Research as secondary IRBs. Informed consent was obtained from all subjects.

Entry criteria included documented HIV infection, age ≥18 years, CD4 lymphocyte count <350 cells/μL and ARV-naïve status except for women with past exposure to ARV associated with pregnancy who were allowed to enroll as long as the exposure was at least 3 months prior to entry. The study utilized an entry criterion of CD4 lymphocyte count <350 cells/μL to be consistent with Thai national guidelines for initiation of ARV therapy. Subjects were excluded for active AIDS-defining illness or other active illness, current use of immunomodulator therapy or any experimental therapy, pregnancy or breast feeding and the presence of active malignancies. Subjects underwent a neuropathy examination during the screening process utilizing the AIDS Clinical Trials Group (ACTG)/Neurology and Neurologic AIDS Research Consortium (NARC) methodology [3]. Subjects diagnosed with having any signs or symptoms of neuropathy (absent or diminished ankle reflex OR diminished vibratory, pin or temperature sensation OR contact allodynia) were excluded from the study because of the potential risk of randomization to the d4T-containing arm.

Baseline medical history and a general physical examination were performed. Routine safety laboratory measurements, CD4 cell count, HIV RNA and fasting metabolic blood work including glucose levels were obtained. Viable peripheral blood mononuclear cells (PBMCs) were obtained for mitochondrial (mt) DNA copies/cell, oxidative phosphorylation (OXPHOS) NADH dehydrogenase [complex I (CI)] and cytochrome c oxidase [complex IV (CIV)] enzyme activities, and mt 8-oxo-deoxyguanine (8-oxo-dG) break frequencies (BFs) as described below.

Epidermal Nerve Fibre Density (ENFD) assessment

Skin punch biopsies for ENFD were performed prior to initiation of ARV therapy using the skin punch biopsy technique and processing recommendations of the Cutaneous Nerve Laboratory at Johns Hopkins (www.hopkinsmedicine.org/neurology_neurosurgery/specialty_areas/cutaneous_nerve_lab/). Briefly, following a 1% lidocaine subcutaneous injection and utilizing sterile techniques, a 4-mm skin punch biopsy was performed on the distal leg at the level of the ankle with an additional skin punch biopsy of the upper lateral thigh. Skin specimens were processed on site and forwarded, via the University of Hawaii, to the Cutaneous Nerve Laboratory at Johns Hopkins for protein gene product (PGP9.5) immunostaining. Slides of 50 μM thick immunostained sections were examined to ensure acceptable specimen quality, and the number of unmyelinated nerve fibres per mm length of epidermis was assessed (Fig. 1).

Figure 1.

A distal leg skin biopsy section from an antiretroviral-naïve subject, immunostained with protein gene product (PGP9.5), showing numerous sensory nerve fibres (arrows) entering from the dermis into the epidermis. Scale bar, 50 μm.

Assessment of PBMC mitochondrial parameters

PBMC mtDNA copies/cell was assayed by absolute quantitative real-time polymerase chain reaction (PCR) as previously described [5]. Briefly, DNA was extracted from frozen PBMCs using a Qiagen DNA kit (Qiagen, Valencia, CA). Standardization of real-time PCR was performed using LightCycler FastStart DNA Master SYBR Green I with the Roche LightCycler instrument (Roche, Indianapolis, IN). A dilution series of the control plasmid containing the 90-bp mtDNA NADH dehydrogenase, subunit 2 and the 98-bp Fas ligand gene was prepared to set up the standard. Each sample and standard were run in duplicate and the results were analysed with Version 4.0 LightCycler software (Roche).

PBMC OXPHOS CI and CIV enzyme activities were measured in duplicate by thin-layer chromatography and immunoassays as described previously [6]. Each vial of viable PBMCs was thawed and washed in 0.5 mL of phosphate-buffered saline (PBS) twice before the addition of 0.5 mL of ice-cold extraction buffer [1.5% lauryl maltoside, 25 mM Hepes (pH 7.4) and 100 mM NaCl, plus protease inhibitors (P-8340; Sigma Chemical Co, St Louis, MO)]. Samples were mixed gently and kept on ice for 20 min, and were then spun in a microcentrifuge at 16 400 g at 4°C for 20 min to remove insoluble cell debris. The supernatant, an extract of detergent-solubilized cellular proteins, was then assayed with the OXPHOS immunoassays. All samples were loaded on the immunoassays with equal amounts of total cell protein (7.5 μg) using an amount previously established with control samples to generate signals within the linear range of the assay. Therefore, the resulting signal was directly proportional to the amount of OXPHOS enzyme activity in the sample. We quantified the signal by densitomeric scanning with a Hamamatsu ICA-1000 reader (Hamamatsu Corp., Bridgewater, NJ, USA). Activity was assessed as optical density (OD)/μg of protein × 103.

PBMC mt 8-oxo-dG damage was assessed using a gene-specific repair assay as previously described [7]. Ten micrograms of PBMC DNA was isolated with a DNeasy Blood and Tissue Kit (Qiagen). DNA was then digested with PvuII (New England BioLabs, Inc., Ipswich, MA) overnight to linearize mtDNA. The digested DNA was separated into two halves: 5 mg of DNA was treated with human 8-oxoguanine DNA glycosylase (hOGG1) for 1 h at 37°C in a reaction volume of 15 μL and then for 1 h at 65°C for enzyme deactivation, and the remaining 5 mg of DNA was left untreated and stored at 48°C. For analysis, 4 μL of 1X Alkaline Agarose Loading Dye (Boston Bioproducts, Boston, MA, USA) was added, and cleaved and noncleaved products were resolved on a 0.75% alkaline agarose gel. DNA was transferred to nylon (+) membranes using standard Southern blot methodology. Human mitochondrial probes specific for cytochrome b were labelled with digoxigenin-dUTP (Roche) by linear PCR amplification. Primer sequences were: DigFor, GCT ACC TTC ACGCCA A (14976–15001); and DigRev, CCG TTT CGT GCA AGAAT (15357–15341). Blots were hybridized overnight at 45°C and processed for chemiluminescent detection following Roche protocols. Finally, membranes were developed on a chemilumiimager (Roche) using LumiAnalyst software (Roche). Mitochondrial 8-oxo-dG damage was quantified by calculating BFs based on the Poisson distribution of DNA treated with the hOGG1 repair enzyme and untreated DNA.

Statistical methods

Correlations between ENFD values and various parameters were assessed by Pearson correlation. We evaluated various variables in terms of their association with, and relative impact on, ENFD values in ARV-naïve subjects by multiple regression analyses. ENFD and other selected independent variables were log-transformed to stabilize variance and to make the residuals more approximately normal. Parameters previously reported in the literature to be associated with ENFD (age, height, CD4 cell count and HIV RNA) were predictors of interest; inference was made after adjustment for confounding variables. Confounding variables were defined as variables that upon inclusion resulted in more than a 10% change in the regression coefficients of main effects. The relative contributions of variables that were highly correlated [i.e. gender and height; body mass index (BMI) and height] were evaluated in nested models. To examine the incremental effect of OXPHOS CI and CIV enzyme activity as well as that of mt 8-oxo-dG levels, each was then introduced individually into the previously constructed model. Model selection was based on adjusted R-square and Akaike's information criterion (AIC).

Results

Cohort characteristics

Of the 152 subjects enrolled in SEARCH 003, skin punch biopsies were obtained from 132 subjects who agreed to participate in the neuropathy substudy. All of these 132 ENFD specimens were judged by the Johns Hopkins Cutaneous Nerve Laboratory as evaluable, and are the focus of this report.

All subjects were Thai, with 56.1% recruited from the Thai Red Cross AIDS Research Centre and 43.9% from Queen Savang Vadhana Hospital (Table 1). The gender distribution of 44.7% male is consistent with the gender distribution of the HIV/AIDS epidemic in Thailand. Only a small percentage of subjects had other common aetiologies for neuropathy (history of isoniazid use, concomitant infection with hepatitis C or the presence of diabetes).

Table 1. Baseline patient characteristics
 Median (IQR); min, max (unless otherwise specified)
  1. BFs, break frequencies; HOMA-IR, homeostatic model assessment for insulin resistance; IQR, interquartile range; Mt 8-oxo-dG, mitochondrial 8-oxo-deoxyguanine; OD, optical density; OXPHOS CI, oxidative phosphorylation NADH dehydrogenase; OXPHOS CIV, oxidative phosphorylation cytochrome c oxidase.
  2. *Diabetes by entry fasting glucose >125 mg/dL.
  3. ** n = 101.
Number of subjects132
Age (years)34 (29–40); 20, 59
Gender (% male)44.7 %
Height (cm)162 (155–168); 145, 180
Body mass index21.5 (19.5–23.7); 16.3, 32.7
CD4 count (cells/μL)169 (71–232); 7, 342
Log10 HIV RNA (copies/mL)4.9 (4.5–5.3); 2.8, 6.5
Hepatitis C (% antibody positive)1.4 %
Isoniazid use (% current or past use)3.8 %
Diabetes* (%)0.8 %
HOMA-IR** 1.0 (0.6–1.4); 0.4, 4.2
Mitochondial DNA (copies/cell)183 (143–238); 32, 488
OXPHOS CI enzyme activity (OD/μg × 103)93 (75–113); 31, 213
OXPHOS CIV enzyme activity (OD/μg × 103)90 (73–113); 35, 194
Mt 8-oxo-dG (BFs)0.1 (0.0–0.2); 0.0, 1.0

ENFD and its correlates

The median (interquartile range) ENFD (fibres/mm) values prior to initiation of ARV therapy were 21.0 (16.2–26.6) for the distal leg and 31.7 (26.2–40.0) for the proximal thigh. Distal leg ENFD correlated positively with CD4 cell count, and negatively with age, height, log10 plasma HIV RNA, and OXPHOS CI and CIV activity levels (Table 2). The relationships between distal leg ENFD and height, CD4 cell count and OXPHOS CIV are shown graphically in Figure 2. No significant correlations were found with BMI, homeostatic model assessment for insulin resistance (HOMA-IR), fasting glucose, PBMC mtDNA or mt-specific 8-oxo-dG. Women had significantly higher distal leg natural log (ln) ENFD than men (mean ENFD: women, 24.2 fibres/mm; men, 19.5 fibres/mm; P < 0.01). Proximal thigh ENFD correlated positively with distal leg ENFD. Similar to distal leg ENFD, proximal thigh ENFD correlated positively with CD4 cell count and negatively with height, with no correlations with HOMA-IR, fasting glucose, PBMC mtDNA or mt-specific 8-oxo-dG. Proximal thigh ENFD, however, differed from distal leg ENFD in showing significant negative correlations with BMI and no correlations with PBMC OXPHOS CI or CIV activity levels. Women had slightly higher proximal thigh ln ENFD than men (mean ENFD: women, 36.0 fibres/mm; men, 31.6 fibres/mm; P = 0.03). Neither distal leg nor proximal thigh ENFD correlated with history of previous ARV medication use during pregnancy or with history of neurotoxic medical comorbidity/medication use (data not shown).

Figure 2.

Scatterplot of distal leg epidermal nerve fibre density and various parameters of interest. (a) Distal leg epidermal nerve fibre densities (fibres/mm epidermis) vs. patient height (cm), (b) distal leg epidermal nerve fibre densities vs. baseline CD4 counts (cells/μL), and (c) distal leg epidermal nerve fibre densities vs. oxidative phosphorylation complex IV activity (optical density/μg protein × 103). CIV, complex IV; ENFD, epidermal nerve fibre density; PBMC, peripheral blood mononuclear cell; OD, optical density; OXPHOS, oxidative phosphorylation.

Table 2. Pearson correlations between distal leg and proximal thigh epidermal nerve fibre densities and various parameters of interest
Pearson correlation coefficientsENFD distal legENFD proximal thigh
  1. ENFD, epidermal nerve fibre density; HOMA-IR, homeostatic model assessment for insulin resistance; Mt, mitochondrial; 8-oxo-dG, 8-oxo-deoxyguanine; OXPHOS CI, oxidative phosphorylation NADH dehydrogenase; OXPHOS CIV, oxidative phosphorylation cytochrome c oxidase; PBMC, peripheral blood mononuclear cell.
ENFD distal leg1 0.56 (P < 0.01)
Age−0.27 (P < 0.01)−0.11 (P < 0.19)
Height−0.31 (P < 0.01)−0.18 (P = 0.04)
Body mass index−0.15 (P = 0.09)−0.19 (P = 0.03)
HOMA-IR−0.10 (P = 0.30)−0.03 (P = 0.79)
Fasting glucose−0.01 (P = 0.88)0.07 (P = 0.43)
CD4 count0.31 (P < 0.01)0.21 (P = 0.02)
Log10 HIV RNA−0.27 (P < 0.01)−0.09 (P = 0.29)
PBMC mtDNA−0.06 (P = 0.47)−0.00 (P = 0.97)
PBMC OXPHOS CI enzyme activity−0.19 (P = 0.03)−0.11 (P = 0.20)
PBMC OXPHOS CIV enzyme activity−0.21 (P = 0.02)−0.09 (P = 0.31)
PBMC mt-specific 8-oxo-dG0.04 (P = 0.68)0.05 (P = 0.56)

The results of the multiple linear regression analyses are shown in Table 3. Simple linear regression analysis showed age, height, CD4 cell count and HIV RNA to each be significantly associated with distal leg ENFD (all P-values < 0.01). In a regression model that included all four variables, age, height, and CD4 cell count but not log10 HIV RNA were significant predictors of distal leg ENFD (adjusted R 2 = 23.02%). Both gender and height were strongly correlated with ENFD; however, when both were included in the model, height remained significant whereas gender was not significant at an alpha level of 0.10. A partial F-test on the additional effect of gender confirmed that gender could be dropped from the model. To examine the incremental effect of OXPHOS CI and CIV enzyme activity as well as of mt 8-oxo-dG levels, each was introduced individually into the previously constructed model. The association between distal leg ENFD and log PBMC CIV activity was significant (P = 0.04; incremental adjusted R 2 = 2%); that between distal leg ENFD and log PBMC CI activity was on the border of significance (P = 0.06; incremental adjusted R 2 = 1.58%). No significant association was observed between distal leg ENFD and PBMC mt 8-oxo-dG. BMI was included in the adjusted model for distal leg ENFD because of its confounding effect on the relationship between ENFD and HIV RNA. The final model revealed that age, CD4 cell count, height, BMI, and log10 PBMCCIV activity were significant predictors of distal leg ENFD (adjusted R 2 = 27.33%; Table 3).

Table 3. Multiple regression analysis of predictors of epidermal nerve fibre density levels in the distal leg and proximal thigh
 EstimateStandard error P-value
  1. ENFD, epidermal nerve fibre density; OXPHOS CIV, oxidative phosphorylation cytochrome c oxidase; PBMC, peripheral blood mononuclear cell.
Log distal leg ENFD   
Intercept6.51510.7015<0.01
CD4 count0.00110.0003<0.01
Log10 HIV RNA−0.07310.04970.14
Age−0.01350.0038<0.01
Height−0.00870.00370.02
Body mass index−0.01940.00870.03
Log10 PBMC OXPHOS CIV enzyme activity−0.22180.09430.02
Log proximal thigh ENFD   
Intercept4.30420.6163<0.01
CD4 count0.00080.00030.02
Log10 HIV RNA0.03440.04980.49
Age−0.00480.00390.22
Height−0.00590.00370.11

Similar analyses were performed to construct a final regression model for proximal thigh ENFD. Although Pearson correlation showed potential associations of proximal thigh ENFD with height and CD4 cell count, a model with all effects of interests (age, height, CD4 cell count, and log10HIV RNA) showed that only CD4 cell count was a significant predictor, explaining approximately 4.6% of the variability in proximal thigh ENFD.

Discussion

Our study found that older age, larger BMI, taller stature, lower CD4 cell count and higher PBMC OXPHOS CIV levels were risk factors for lower distal leg ENFD in ARV-naïve Thai subjects free of neuropathy. ENFD documents the extent of damage present in unmyelinated nerve fibres per mm length of epidermis. A distal ENFD of 10 fibres/mm or less in US HIV-infected individuals with either no neuropathy or asymptomatic disease has been reported to confer a 14-fold greater risk of developing symptomatic disease than ENFD > 10 fibres/mm [8]. Early data obtained from hospitalized patients in the US before ARV medications were available indicated that approximately one-third of HIV-infected patients had both clinical and electrophysiological evidence of neuropathy [9]. Neuropathy was primarily noted to be a complication of late-stage HIV disease associated with advanced immunosuppression [10]. However, while neuropathic symptoms frequently did not occur until the development of AIDS, electrophysiological evidence of peripheral nerve involvement was found in many patients with normal or near-normal CD4 cell counts [11].

In the post-highly active antiretroviral therapy (HAART) era, in ARV-treated individuals, the strong correlation of neuropathy with advanced immunosuppression is no longer seen [3, 12-14]. However, the impact attributable to HIV itself may still be important. The CNS HIV Anti-Retroviral Therapy Effects Research (CHARTER) study found a lower CD4 cell count nadir to be a significant risk factor for neuropathy [15]. In the HIV Outpatient Study (HOPS) cohort, low nadir CD4 cell count and high plasma HIV RNA at first visit were a nearly equivalent or stronger predictor of developing neuropathy than use of ARV medications, with the exception of higher dose d4T [16]. Our study results are consistent with pre-HAART data and demonstrate that peripheral nerve damage is seen at lower CD4 cell counts among individuals free of clinical neuropathy signs and/or symptoms. The findings of this study are also consistent with other reports that have identified age and height as significant predictors of neuropathy risk [15, 17, 18]. High ENFDs are seen in young people [19] and remodelling and ‘pruning’ of epidermal nerves may be part of the aging process.

The aetiology of HIV-SN even in the absence of ARV medications may include important contributions from mitochondrial dysfunction. Recently, mtDNA damage was demonstrated to be more pronounced in the distal mitochondria of long axons from post mortem patients who died with HIV-SN, consistent with the length-dependent nature of this neuropathy [20]. Furthermore, terminal nerve endings in skin from HIV-infected patients have demonstrated abnormal mitochondrial accumulations [21]. Depletion of PBMC mtDNA and OXPHOS enzyme activities has been observed in ARV-naïve subjects, suggesting that HIV in the absence of ARV medications causes mitochondrial dysfunction [22, 23]. Older potentially mitochondrial-toxic NRTIs [d4T and didanosine (ddI)] are still used as components of ARV regimens in developing countries. Pre-existing mitochondrial dysfunction related to HIV may increase the risk of neuropathy when such drugs are used.

Adequate energy production is necessary for normal metabolism; thus, our initial expectation was that lower ENFD, as a predictor of peripheral nerve damage, would be associated with lower OXPHOS enzyme activities. However, our study found that OXPHOS CI and CIV activity levels increased, rather than decreased, as ENFD decreased. We speculate that such increases in PBMC OXPHOS enzyme activities among ARV-naïve patients with lower ENFD may represent a general increase in cellular energy requirements secondary to increases in inflammatory tendencies mediated by HIV. In vitro, HIV infection has been reported to increase CIV activity in HIV-infected T cells [24]. In another tissue source, we have reported increases in ATP level within adipocytes from ARV-naïve subjects compared with HIV-seronegative controls [25].

Our study has several important limitations. Its applicability is limited to subjects of Thai descent and to those free of clinically defined neuropathy. Because this study recruited only individuals with a CD4 count <350 cells/μL, the findings may not be applicable to HIV-infected individuals with higher CD4 counts. Furthermore, mitochondrial OXPHOS and oxidative stress measurements in PBMCs may not necessarily reflect mitochondrial dysfunction in the dorsal root ganglion or sural nerves. Nevertheless, some important conclusions are possible. The correlation of ENFD to previously established risk factors for neuropathy, namely age and height, lends credibility to ENFD as a valid predictive marker of neuropathy risk. Lower CD4 cell counts and higher OXPHOS CIV activity levels are found in association with subclinical peripheral nerve damage in HIV-infected ARV-naïve individuals with moderate to severe HIV immunodeficiency. Whether HAART regimens with less mitochondrial toxicity can repair such damage has yet to be determined. Furthermore, pre-existing subclinical ENFD damage may have clinical consequences if it lowers the threshold for the development of clinical neuropathy upon exposure to d4T or other neurotoxic medications and conditions.

Acknowledgements

The authors wish to thank the patients for their participation in this study. Additionally, we would like to acknowledge the specific contributions to the study by Stephen J. Kerr, Patcharawee Rungrojrat, Somsong Teeratakulpisarn, and Tippawan Pankam from SEARCH/TRCARC and Daniel E. LiButti, Julia Choi and Heidi Fink from the University of Hawaii. The biostatistician for the study was Victor DeGruttola, Harvard School of Public Health, Boston, MA, USA. Funding was received from the Thai Government Pharmaceutical Organization, the National Institute of Health [R01NS063932 (CMS), R01AI074554 (MG), and P20RR011091 U54RR026136], Gilead Sciences, and MitoScience Inc. [P30MH075673 (JCM) and NS44807 (JCM)].

Ancillary