HIV-associated neurocognitive disorder in HIV-infected Koreans: the Korean NeuroAIDS Project

Authors


Abstract

Objectives

HIV-associated neurocognitive disorder (HAND) is an independent predictor of early mortality and is associated with many difficulties in activities of daily living. We sought to determine the prevalence of and risk factors for HAND in HIV-infected Koreans. In addition, we investigated the performance of screening tools and components of neuropsychological (NP) tests for diagnosing HAND.

Methods

HIV-infected patients were enrolled consecutively from two different urban teaching hospitals in Seoul, South Korea between March 2012 and September 2012. Participants completed a detailed NP assessment of six cognitive domains commonly affected by HIV. The Frascati criteria were used for diagnosing HAND. Four key questions, the International HIV Dementia Scale (IHDS) and Montreal Cognitive Assessment (MoCA)-K were also assessed as potential tools for screening for HAND.

Results

Among the 194 participants, the prevalence of HAND was 26.3%. Asymptomatic neurocognitive impairment and minor neurocognitive disorder accounted for 52.9 and 47.1% of the patients with HAND, respectively. In multivariate analysis, haemoglobin (Hb) level ≤ 13 g/dL (P = 0.046) and current use of a protease inhibitor-based regimen (P = 0.031) were independent risk factors for HAND. The sensitivity and specificity of the IHDS were 72.6 and 60.8%, and those of MoCA-K were 52.9 and 73.4%, respectively. The IHDS (P < 0.001) and MoCA-K (P < 0.001) were both useful for screening for HAND. Among NP tests, the sensitivity and specificity of the Grooved Pegboard Test were 90.2 and 72.0%, and those of the Wisconsin Card Sorting Test were 61.2 and 84.4%, respectively.

Conclusions

HAND is a prevalent comorbidity in HIV-infected Koreans. Active screening and diagnosis with effective tools, such as the IHDS, MoCA-K and Grooved Pegboard Test, could be used to identify this important complication.

Introduction

HIV may enter the central nervous system (CNS) early after infection [1] and, although it does not directly infect neurons, it is frequently associated with structural and functional brain abnormalities [2-5]. This CNS infection can lead to neurocognitive impairment. This HIV-associated neurocognitive disorder (HAND) strongly predicts a wide variety of difficulties in activities of daily living, such as employment, automobile driving, medication adherence, financial management, shopping, cooking and use of public transportation [6]. Moreover, viral levels in the cerebrospinal fluid (CSF) have been associated with HAND in HIV-unsuppressed patients with AIDS [7], while suppression of CSF HIV can be associated with improvement in cognition [8]. However, in the era of combination antiretroviral therapy (cART), HAND in Western countries is still common (being found in 16−52% of HIV-infected individuals), even when cART is successful with suppression of viral load [9-14]. The extent to which cART leads to a reduction in the incidence and prevalence of HIV-related cognitive impairment remains unclear. Additionally, most neurocognitive research in HIV-infected individuals has been conducted in Western populations, and the burden of HAND in Korean HIV-infected individuals is so far unknown. Furthermore, there are few studies that have investigated how to screen HIV-infected individuals for HAND and the diagnostic validation of neuropsychological (NP) tests for diagnosing HAND among HIV-infected Asian persons.

Methods

Study design

Two hundred HIV-infected patients who were 18 years old or older were recruited consecutively in Severance Hospital and Korea University Guro Hospital between March 2012 and September 2012 in Seoul, Korea. All subjects provided informed consent and received standardized neurological, NP and functional assessments at two study sites. This study was approved by the Institutional Review Board of the hospital (IRB #4-2011-0630).

The exclusion criteria were (1) recent and/or significant traumatic brain injury; (2) a neurological disorder not related to HIV infection; (3) infections that can affect the CNS; (4) a significant CNS opportunistic infection based upon history and/or neuromedical examination; (5) a current or past psychotic disorder; (6) significant substance use (more than three alcoholic drinks per day over the last month or recreational drug use more than once per week during the last month); (7) symptoms of a current, active infection, a body temperature of > 38.5°C at the time of recruitment or current treatment for a serious, systemic infection within 3 months; (8) colour blindness; (9) a hearing deficit that appears to affect auditory comprehension [15].

Six of the two hundred subjects were excluded because of confounding comorbidity (five subjects) or because they withdrew early from the study (one subject). Enrolled participants underwent demographic, clinical and NP assessments. At baseline visits, a neuromedical history was obtained and a standardized examination, neurobehavioural testing, and structured evaluation of selected psychiatric variables were performed. The following variables were also assessed: age at first visit, gender, body mass index (BMI), haemoglobin (Hb) level, level of education, duration of HIV infection, Centers for Disease Control and Prevention (CDC) classification, reported mode of transmission, prior AIDS diagnosis, hepatitis B or C virus coinfection, antiretroviral therapy regimen [two nucleoside reverse transcriptase inhibitors (NRTIs) plus a nonnucleoside reverse transcriptase inhibitor (NNRTI), two NRTIs plus a protease inhibitor (PI), or another combination], CNS penetration effectiveness (CPE) score [16], initial CD4 T-cell count, pre-cART CD4 T-cell count, current CD4 T-cell count, lowest CD4 T-cell count, initial viral load (VL), pre-cART VL, current VL, and highest VL. Additionally, a clinician administered the depression questionnaire [17] and interview for assessing depression and substance use history. The Karnofsky Performance Status Scale [18] was used to assess functional impairments. Also, eight questions in Korean designed by the authors based on the suggestions of Antinori et al. [15] were used to assess functional activities; these translate as: ‘Is it hard to take medication in the correct dosages at the correct time?’; ‘Is it hard to manage financial matters independently (budgeting, writing cheques, paying rent and bills, going to the bank)?’; ‘Is it hard to perform household tasks alone or with occasional assistance?’; ‘Do you have trouble managing your daily schedule?’; ‘Do you make more mistakes in your working?’; ‘Do you need more time than before to do the same amount of work?’; ‘Do you find it more difficult than before to carry out tasks successfully?’; ‘Are you less able to produce your best work?’.

Neuromedical evaluation

Neuromedical assessments were standardized between study sites and included multiple, linked evaluations. These evaluations included the following elements.

  • (1) Medical history. Standardized forms were used to record medical and neurological symptoms. Medical conditions were classified using International Classification of Diseases 9th Revision, Clinical Modification (ICD-9-CM) codes, and HIV disease stage was classified according to 1993 CDC guidelines [19]. Staff also administered a brief questionnaire to collect information about HIV risk behaviour.
  • (2) Current medications, medication history, and adherence. A clinician recorded prescribed medications taken for longer than 1 month. Detailed information about past and present ART use was obtained. Each site used the Adherence Scale to Anti-HIV Medications form to collect ART adherence data [20].
  • (3) Neurological and general physical examinations. The research clinician performed a standardized neurological, medical examination that included assessment of vital signs, height and weight, mental status, cranial nerves, motor, sensory and cerebellar function, reflexes and gait. The medical staff also performed a general physical examination and noted health-related functioning.

Assessment of the screening tool

Four key questions proposed by the Asia, Australia, Africa and Middle East (AAAME) HAND Advisory Board as a screening tool (unpublished data) to evaluate early potential signs of neurocognitive impairment and depression were put to participants: ‘Are you slower in your thinking processes?’; ‘Are you more forgetful?’; ‘Is it harder to organize things?’ and ‘Are you less able to find pleasure in things you used to enjoy?’ Also, the International HIV Dementia Scale (IHDS) [21] and Montreal Cognitive Assessment (MoCA)-K [22] were evaluated as screening tools for diagnosing HAND. The most useful cut-off values for each screening test were also investigated.

Neuropsychological tests

In order to establish rates of HAND, we used NP test norms based on results for HIV-negative subjects[23-27]. The neurobehavioural evaluation assessed six ability domains (Supplementary Table S1). These are well-established measures that have been used in numerous studies of HIV/AIDS. In addition to providing individual NP test results, the battery facilitates clinical determination of the level of impairment within each domain, as well as the global level of impairment based on population-specific normative standards. This approach has shown good inter-rater reliability, even across raters at different institutions [28].

Diagnosis of HAND

Subjects completed a detailed NP assessment measuring their functioning in six cognitive domains known to be commonly affected by HIV infection. The Frascati criteria were used to diagnose HAND, which was classified as asymptomatic neurocognitive impairment (ANI), minor neurocognitive disorder (MND) and HIV-associated dementia (HAD) [15].

Statistical analyses

An independent t-test or χ2 test was used to assess differences in each variable between neurocognitively impaired and nonimpaired subjects. To identify independent factors associated with HAND, a multivariate logistic regression analysis was performed which included the variables that were significantly associated with HAND in the univariate analysis (P < 0.05) and well-known as a risk factor of HAND. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were then calculated in comparison with the reference diagnosis, which was based on cases of HAND. In addition, we conducted a receiver operating characteristic (ROC) curve analysis to compare the predictive accuracies of screening tools and the area under the curve (AUC) was calculated. The 95% confidence interval (CI) was calculated using the Wilson score method. All P-values are two-tailed and P < 0.05 was considered statistically significant. All analyses were performed using spss for Windows 12.0 (SPSS, Chicago, IL).

Results

General characteristics of study subjects

Of the 194 enrolled subjects, 93.8% were male, and the mean age (range) was 45.12 (21–72) years. The most common exposure category was men who have sex with men (MSM) (52.6%), followed by heterosexual contact (27.3%). The mean [± standard deviation (SD)] duration of education was 13.4 ± 3.3 years, and 31.9 and 27.9% of the subjects were at clinical CDC stages B and C of HIV disease, respectively. The mean (± SD) current CD4 T-cell count was 481.4 ± 236.0 cells/μL and the mean (± SD) current viral load was 2.0 ± 1.4 log10 HIV-1 RNA copies/mL (Table 1).

Table 1. Baseline characteristics and factors associated with HIV-associated neurocognitive disorder (HAND) in HIV-infected individuals
VariableTotal (n = 194)Patients with HAND (n = 51)Patients without HAND (n = 143)P-valueMultivariate analysis
Odds ratio (95% CI)P-value
  1. BMI, body mass index; cART, combination antiretroviral therapy; CI, confidence interval; IQR, interquartile range; VL, viral load; PI, protease inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; II, integrase inhibitor; CPE, central nervous system penetration effectiveness; HCV, hepatitis C virus; HBV, hepatitis B virus; SD, standard deviation.
Age (years) [mean (range)]45.12 (21–72)44.39 (21–70)45.38 (25–72)0.581  
Male gender [n (%)]182 (93.8)49 (96.1)133 (93.0)0.735  
BMI (kg/m2) (mean ± SD)23.0 ± 4.522.1 ± 2.923.3 ± 4.90.105  
Haemoglobin ≤ 13.0 g/dL [n (%)] 15 (29.4)22 (15.4)0.0292.287 (1.014–5.162)0.046
Education (years) (mean ± SD)13.4 ± 3.312.8 ± 3.413.6 ± 3.30.188  
Mode of transmission [n (%)]      
Homosexual contact102 (52.6)24 (47.1)78 (54.5)0.358  
Heterosexual contact53 (27.3)17 (33.3)36 (25.2)0.262  
Unknown and others39 (20.1)10 (19.6)29 (20.3)   
Duration of HIV infection (years) (mean ± SD)77.7 ± 70.173.4 ± 52.579.3 ± 75.50.606  
Previous AIDS diagnosis [n (%)]55 (28.4)18 (35.3)37 (25.9)0.200  
AIDS at diagnosis [n (%)]      
No7617 (33.3)59 (41.3)0.320  
Yes11834 (66.7)84 (58.7)0.320  
Initial CD4 count (cells/μL) (mean ± SD)243.8 ± 188.3233.8 ± 220.7247.3 ± 176.00.668  
Initial VL (log10 copies/mL) (mean ± SD)4.8 ± 1.24.7 ± 1.34.8 ± 176.00.797  
Pre-cART CD4 count (cells/μL) (mean ± SD)186.4 ± 136.0176.5 ± 167.5189.8 ± 123.80.601  
Pre-cART VL (log10 copies/mL) (mean ± SD)5.0 ± 0.95.0 ± 0.95.0 ± 0.90.965  
Current CD4 count (cells/μL) (mean ± SD)481.4 ± 236.0444.5 ± 259.2494.6 ± 226.60.1940.999 (0.997–1.001)0.225
Current VL (log10 copies/mL) (mean ± SD)2.0 ± 1.42.4 ± 1.71.9 ± 1.20.069  
Nadir CD4 count (cells/μL) (mean ± SD)187.0 ± 138.0171.8 ± 167.2192.4 ± 126.30.365  
Highest VL (log10 copies/mL) (mean ± SD)5.0 ± 0.95.0 ± 0.94.9 ± 1.00.566  
Current cART [n (%)]      
No treatment20 (10.3)7 (13.7)13 (9.1)0.350  
2 NRTIs + PI104 (59.8)32 (72.7)72 (55.4)0.0432.323 (1.078–5.005)0.031
2 NRTIs + NNRTI49 (28.2)7 (15.9)42 (32.3)0.037  
2 NRTIs + II12 (6.9)3 (6.8)9 (6.9)1.000  
3 NRTIs3 (1.7)0 (0.0)3 (2.3)0.572  
Other6 (3.4)2 (4.5)4 (3.1)   
CPE rank (mean ± SD) 6.7 ± 1.36.7 ± 1.10.961  
HCV coinfection [n (%)]6 (3.7)3 (6.7)3 (2.6)0.350  
HBV coinfection [n (%)]9 (5.3)2 (4.3)7 (5.6)1.000  

Prevalence and clinical characteristics of HAND

The overall prevalence of HAND in the study cohort was 26.3%. Of the 51 participants with HAND, 52.9 and 47.1%, respectively, had ANI and MND. No individuals were diagnosed with HAD. Of those with HAND, the mean age (range) was 44.39 (21–70) years, and 96.1% of them were male. The most common exposure category was MSM (47.1%), followed by heterosexual contact (33.3%). The mean (± SD) duration of education was 12.8 ± 3.4 years. The mean (± SD) Hb level was 13.6 ± 1.8 g/dL, with 15% having Hb < 13.0 g/dL. Initial and current mean (± SD) CD4 T-cell counts were 233.8 ± 220.7 and 444.5 ± 259.2 cells/μL, respectively. Initial and current mean (± SD) viral loads were 4.7 ± 1.3 and 2.4 ± 1.7 log10 copies/mL, respectively, and 71.7% of subjects were on a cART regimen of two NRTIs plus a PI (Table 1). Also, 82% of study participants with HAND receiving cART in our cohort had suppressed viral loads (< 50 copies/ml). In the individuals with HAND, among the six cognitive functioning domains, impairment of sensory perceptual/motor skills was the most common (90.2%), followed by abstraction/executive impairment (78.0%) and memory (learning and recall) impairment (48.0%) (Table 2).

Table 2. Impairment in six domains of cognitive functioning in 51 Korean HIV-infected individuals with HIV-associated neurocognitive disorder (HAND)
DomainPatients with HAND who showed impairment [n (%)]
Verbal/language2 (3.9)
Attention/working memory3 (5.9)
Abstraction/executive39 (78.0)
Memory (learning and recall)24 (48.0)
Speed of information processing1 (2.0)
Sensory perceptual/motor skills46 (90.2)

In univariate analysis, Hb levels ≤ 13 g/dL (P = 0.029) and current use of a PI-based regimen (P = 0.043) were the only factors significantly associated with HAND. The significance remained for both Hb level (P = 0.046) and current use of a PI-based regimen in multivariate analysis (P = 0.031) (Table 1).

The performance of screening tests for HAND

The sensitivity and specificity of the IHDS were 72.6 and 60.8%, and those of MoCA-K were 52.9 and 73.4%, respectively (Table 3). The MoCA-K and IHDS screening test scores were significantly correlated with HAND, and cut-off values of ≤ 25 and ≤ 10 were the most useful in diagnosing HAND, respectively (Table 3 and Tables S2 and S3). To compare the predictive accuracies of MoCA-K and the IHDS as screening tools, a ROC analysis was performed (Fig. 1), which demonstrated that the AUCs of the IHDS and MoCA-K for diagnosing HAND were 0.678 (P < 0.001) and 0.666 (P < 0.001), respectively. However, the ‘four key questions’ devised for use as a screening tool did not significantly correlate with the diagnosis of HAND by standard methods (Table S4).

Figure 1.

Areas under the curve (AUCs) for the International HIV Dementia Scale (IHDS) and Montreal Cognitive Assessment (MoCA)-K for diagnosing HIV-associated neurocognitive disorder (HAND). ROC, receiver operating characteristic.

Table 3. Diagnostic performance of the International HIV Dementia Scale (IHDS) and Montreal Cognitive Assessment (MoCA)-K
NP testSensitivity % (95% CI)Specificity % (95% CI)Accuracy % (95% CI)
  1. CI, confidence interval; NP, neuropsychological.
IHDS ≤ 1072.55 (60.30−84.80)60.84 (52.84–68.84)63.92 (57.16–70.68)
MoCA-K ≤ 2552.94 (39.24–66.64)73.43 (66.19–80.67)68.04 (61.48–74.60)
p-value for comparison0.02440.00710.3134

NP tests for diagnosing HAND

Diagnostic characteristics for each NP test for diagnosing HAND in HIV-infected Koreans are provided in Table 4. Among NP tests, the Grooved Pegboard Test had the highest sensitivity and modest specificity (90.2 and 72.0%, respectively). The next highest scoring NP test was the Wisconsin Card Sorting Test, which had a sensitivity and specificity of 61.2 and 84.4%, respectively. The other NP tests had much lower sensitivity but higher specificity.

Table 4. Diagnostic value of neuropsychological tests for diagnosing HIV-associated neurocognitive disorder (HAND) in Korean HIV-infected individuals
Cognitive domain and neuropsychological testSensitivity % (95% CI)Specificity % (95% CI)PPV % (95% CI)NPV % (95% CI)Accuracy % (95% CI)
  1. CI, confidence interval; K-WAIS, K-Wechsler Adult Intelligence Scale; NPV, negative predictive value; PPV, positive predictive value.
Speed of information processing     
K-WAIS Digit Symbol Subtest1.96 (0.00–5.77)99.30 (97.93–100.00)50.00 (0.00–100.00)73.96 (67.75–80.17)73.71 (67.52–79.91)
Trail Making Test, Part A0.00 (0.00–0.00)100.00 (100.00–100.00)73.71 (67.52–79.91)73.71 (67.52–79.91)
Memory (learning and recall)     
K-Auditory Verbal Learning Test (K-AVLT)6.00 (0.00–12.58)99.30 (97.93–100.00)75.00 (32.56–100.00)75.13 (68.97–81.29)75.13 (69.03–81.23)
K-Complex Figure Test (K-CFT)9.80 (1.64–17.97)99.30 (97.93–100.00)83.33 (53.51–100.00)75.53 (69.39–81.68)75.77 (69.74–81.80)
Abstraction/executive     
Wisconsin Card Sorting Test (128-item)61.22 (47.58–74.87)84.40 (78.41–90.39)57.69 (44.26–71.12)86.23 (80.48–91.98)78.42 (72.57–84.27)
K-WAIS Similarity Subtest4.00 (0.00–9.43)100.00 (100.00–100.00)100.00 (100.00–100.00)74.74 (68.56–80.92)75.00 (68.88–81.13)
Trail Making Test, Part B32.00 (19.07–44.93)95.10 (91.57–98.64)69.57 (50.76–88.37)80.00 (73.99–86.01)78.76 (72.99–84.53)
Attention/working memory     
K-WAIS Digit Span Subtest5.88 (0.00–12.34)99.30 (97.93–100.00)75.00 (32.56–100.00)74.74 (68.56–80.92)74.74 (68.63–80.86)
Sensory perceptual/motor skills     
Grooved Pegboard Test (Dominant and Non-dominant Test)90.20 (82.03–98.36)72.03 (64.67–79.39)53.49 (42.95–64.03)95.37 (91.41–99.33)76.80 (70.86–82.74)
Verbal/language     
K-WAIS Vocabulary Subtest3.92 (0.00–9.25)100.00 (100.00–100.00)100.00 (100.00–100.00)74.48 (68.31–80.65)74.74 (68.63–80.86)

Discussion

This is the first study of the prevalence of HAND in HIV-infected patients in Korea. We found that approximately a quarter of our patients had HAND (26.3%). Of these individuals, 52.9% had ANI and 47.1% had MND. Of those with HAND, 82% were receiving cART with suppressed viral loads, suggesting that suppressing HIV replication might be not sufficient to treat HAND. These results are similar to reports from Western countries, with the reported prevalence of HAND being 16−38% in the era of cART [9-13].

Our study also found a lower Hb level and use of a current PI-based cART regimen to be independent risk factors for HAND. Although low Hb levels may represent underlying chronic disease, the specific mechanism of the effect of low Hb on the occurrence of HAND is unclear, although previous studies have also noted a low Hb level as a risk factor for HAND [29] and HAD [30]. The reason for the observed significant correlation between the current use of a PI-based cART regimen and HAND is unclear, but it may be an effect of the clinician's efforts to treat HAND rather than indicating that use of a PI-based regimen is an independent risk factor for HAND. In other words, we speculate that cART might have been switched to a PI-based regimen because of memory impairment in HIV-infected patients. Such a switch may have been made because NNRTIs such as efavirenz have been reported to be risk factors for HAND [31] and most PIs have high CPE scores [32]. Thus, in our study, HIV-infected patients with HAND seemed to be receiving a PI-based regimen more frequently than those without HAND. Another reason for the association between PIs and HAND may be that clinicians may tend to prescribe PI-based cART regimens rather than NNRTI-based regimens for patients who are more ill.

Some studies of Western cohorts have reported that low nadir CD4 T-cell counts were an independent risk factor for HAND [33-35], but this was not observed in our cohort. In our study, the interquartile range of nadir CD4 T-cell counts of all participants was 69−277 cells/μL and, in particular, the 25% quartile was higher than that of Ellis et al. [33]. This may be because there was a tendency to start cART earlier in Korea; thus, we speculate that nadir CD4 T-cell count was not significantly associated with HAND because of the overall higher CD4 counts in our population. As the prevalence of HAND in our population was similar to that in the Western reports, this finding may suggest that nadir CD4 count may in fact be a confounder with and not a predictor of HAND.

In our study, the IHDS and MoCA-K measures used as screening tools for HAND showed better performance than the ‘four key questions’. As the IHDS is widely used as a screening tool for HAD [21], it has the potential to be a useful screening instrument for HAND. In our study, participants who were classified as having neurocognitive impairment on the IHDS also performed significantly less well on other tests of cognitive function, especially those of processing speed and verbal learning/memory [36]. MoCA is a brief cognitive screening tool with high sensitivity and specificity for detecting mild cognitive impairment or dementia [22, 37], and in our study, a cut-off value of 26 (scores of ≤ 25 indicate impairment) yielded the best balance between sensitivity and specificity for HAND. As clinicians often depend on patients to self-report memory complaints, especially in the early stages of cognitive impairment, we attempted to develop a screening tool with four questions, but this tool was not found to be reliable. We speculate that this may have been because patients might not have answered the questions frankly.

Beyond screening tools, we also investigated which NP tests had the best sensitivity and specificity for diagnosing HAND. We found that the Grooved Pegboard Test and Wisconsin Card Sorting Test had the best sensitivity and specificity in our study population. In particular, the Grooved Pegboard Test is a commonly used measure of psychomotor slowing in HIV-positive populations [38], and it has been reported that it can significantly differentiate HIV-infected patients with and without dementia [39]. Therefore, the Grooved Pegboard Test may be useful in diagnosing HAND in HIV-infected Koreans.

Our study had some limitations. First, although there were 194 well-characterized participants in the study, the sample size of those with HAND was only 51. The small sample size has the possibility to influence results due to an unequal distribution of impairment. Secondly, there were no patients with HAD in this study. This may represent a selection bias, i.e. those with HAD may not have been able to consent to participate in the study. Thirdly, our results are not applicable to women with HIV infection because 93.8% of the enrolled patients were men; however, the study sample was representative of the Korean HIV-infected population [40]. Fourthly, the study was conducted in Korean among HIV-infected patients in an urban out-patient clinic, which may not be representative of HIV-infected individuals in a community or rural setting or of the general Korean HIV-infected population. Lastly, the possibility of the presence of peripheral neuropathy in patients with HAND influencing NP tests, especially the Grooved Pegboard Test, was not excluded by neurological examination.

In conclusion, HAND is a prevalent comorbidity among HIV-infected Koreans. Active screening and diagnosis with effective tools such as the IHDS, MoCA-K and Grooved Pegboard Test should probably be performed in order not to overlook this important complication.

Acknowledgements

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (NRF-2013R1A1A2005412 and NRF-2011-220-E00015), a Chronic Infectious Disease Cohort grant (4800-4859-304-260) from the Korea Centers for Disease Control and Prevention, and a grant from the Korean Society for AIDS. This research was also supported by the US National Institutes of Health (NIH) (grants AI100665, AI36214, MH097520, MH83552 and MH62512). Statistical advice was provided by the statistician Hye Sun Lee, M.S., at the Biostatistics Collaboration Unit, Yonsei University College of Medicine, Seoul, Korea.

Conflicts of interest: The authors have no conflicts of interest to declare.

Ancillary