Diabetes mellitus burden among people living with HIV from the Asia‐Pacific region

Abstract Introduction Comorbidities including diabetes mellitus (DM) among people living with HIV (PLHIV) are of increasing clinical concerns in combination antiretroviral therapy (cART) era. We aimed to determine the incidence and risk factors of new‐onset DM among PLHIV in Asian settings. Methods PLHIV from a regional observational cohort without DM prior to antiretroviral therapy (ART) initiation were included in the analysis. DM was defined as having a fasting blood glucose ≥126 mg/dL, glycated haemoglobin ≥6.5%, a two‐hour plasma glucose ≥200 mg/dL, or a random plasma glucose ≥200 mg/dL. A Cox regression model, stratified by site, was used to identify risk factors associated with DM. Results and discussion Of the 1927 participants included, 127 were diagnosed with DM after ART initiation. Median follow‐up time from ART initiation to DM diagnosis was 5.9 years (interquartile range (IQR): 2.8 to 8.9 years). The crude incidence rate of DM was 1.08 per 100 person‐years (100 PYS), 95% confidence interval (CI) (0.9 to 1.3). In the multivariate analysis, later years of follow‐up (2011 to 2013: HR = 2.34, 95% CI 1.14 to 4.79, p = 0.02; and 2014 to 2017: HR = 7.20, 95% CI 3.27 to 15.87, p < 0.001) compared to <2010, older age (41 to 50 years: HR = 2.46, 95% CI 1.39 to 4.36, p = 0.002; and >50 years: HR = 4.19, 95% CI 2.12 to 8.28, p < 0.001) compared to <30 years, body mass index (BMI) >30 kg/m2 (HR = 4.3, 95% CI 1.53 to 12.09, p = 0.006) compared to BMI <18.5 kg/m2, and high blood pressure (HR = 2.05, 95% CI 1.16 to 3.63, p = 0.013) compared to those without high blood pressure, were associated with developing DM. The hazard was reduced for females (HR = 0.47, 95% CI 0.28 to 0.80, p = 0.006). Conclusions Type 2 DM in HIV‐infected Asians was associated with later years of follow‐up, high blood pressure, obesity and older age. This highlights the importance of monitoring and routine screening for non‐communicable diseases including DM as PLHIV age.


| INTRODUCTION
People living with human immunodeficiency virus (PLHIV) have better prognosis and greater longevity because of the benefits of highly effective combination antiretroviral therapy (cART), more effective management strategies and improvements in patient monitoring [1][2][3][4][5]. With increasing survival, non-AIDS complications and comorbidities are important key factors influencing morbidity and mortality among PLHIV. Studies have pointed out metabolic disorders such as diabetes mellitus (DM) were common in PLHIV [6,7]. A study using nationally representative survey data from the U.S. showed that DM prevalence was 3.8% higher in HIV-infected individuals compared with the uninfected general population [6]. A report from the Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) study showed that the incidence of DM was 5.7 per 1000 person-years of follow-up [8].
Use of certain protease inhibitors (PI)-based regimen has been reported [9] to be associated with higher incidence of DM in the early antiretroviral therapy (ART) era, but the associations were less common with the newer classes of PI. Antiretrovirals (ARV) containing older classes of nucleoside reverse transcriptase inhibitors (NRTIs) such as stavudine or didanosine might also increase the risk of developing DM, probably due to insulin resistance caused by mitochondrial toxicities [10][11][12]. Moreover, DM is commonly associated with other comorbidities such as hypertension and dyslipidaemia, which can result in increased risk of developing cardiovascular diseases [10,13].
Non-communicable diseases including DM have been increased dramatically over the past few decades in Asia [14], of which more than half of the global DM population are located in this region [15]. However, DM prevalence data among PLHIV in Asia-Pacific region is still sparse. The incidence of DM varied among HIV population (0.5 to 1.31 cases per 100 persons-years of follow-up) in HIV population [8,10,16]. The incidence of DM in Asia varied from Western countries and the risk factors for the development of DM among PLHIV are understudied in the region. Hence, we assessed the incidence and risk factors of new-onset DM among PLHIV after cART initiation in a regional observational cohort in the Asia-Pacific region.

| Study design and participants
This study was a longitudinal analysis exploring the incidence of new-onset DM after cART initiation. The study participants were PLHIV enrolled in the TREAT Asia HIV Observational Database (TAHOD) between 2003 and 2017. The cohort and its methods have previously been characterized [17][18][19]. The TAHOD is a collaborative observational cohort study that involves 20 sites in the Asia and Pacific region. The participating countries are Cambodia, China and Hong Kong SAR, India, Indonesia, Japan, Malaysia, the Philippines, Singapore, South Korea, Taiwan, Thailand and Vietnam. The recruitment began in 2003. As of March 2017, there were 9160 participants enrolled. Data transfer occurs every six months in March and September. TAHOD does not mandate regular visit schedule and all tests/interventions are performed according to the site's local practices. Participants were included in this analysis if they have been on cART for more than six months, did not have evidence of DM prior to start of cART, and had at least one of the following measurements after cART initiation: fasting blood glucose (FBG), glycated haemoglobin (HbA1C), two-hour plasma glucose after 75 g oral glucose tolerance test (OGTT), or a random plasma glucose (RPG). Participants without DM screening prior to cART initiation were excluded from the study. Participant consent was deferred according to the individual participating sites and their institutional review boards, and is not required for all participants.

| Outcomes
DM was defined as having a single measurement showing FBG ≥126 mg/dL, HbA1C ≥6.5%, a two-hour plasma glucose level after OGTT ≥200 mg/dL, or a RPG ≥200 mg/dL, modified from the standard criteria for DM diagnosis from American Diabetes Association [20]. However, we did not include data of hyperglycaemic symptoms for RPG ≥200 mg/dL. Also, we did not use secondary confirmation testing of FBG as our median FBG testing frequency was once per patient per year (interquartile range (IQR) 1 to 2).

| Covariates
Time-fixed covariates included age, sex, mode of HIV exposure, initial cART regimen, any exposure to stavudine or didanosine in their first-line ART regimen, hepatitis B and C co-infection, prior AIDS diagnosis, smoking and alcohol status. Time-updated covariates included calendar year of follow-up, viral load, CD4, body mass index (BMI), high blood pressure and dyslipidaemia. Dyslipidaemia was defined as a single laboratory result of a fasting cholesterol >200 mg/dL or triglycerides >150 mg/dL. High blood pressure was defined as having at least one measurement of systolic blood pressure >140 mmHg or diastolic blood pressure >90 mmHg. All variables were categorical in the regression analysis.

| Statistical analysis
Factors associated with DM diagnosis after cART initiation were analysed using a Cox regression model stratified by site, to account for clustering within each site. Risk time for DM started from cART initiation and ended on date of first DM diagnosis, among participants who have been on cART for at least six months. Participants without DM events were censored on the date of last measurement for DM markers. Since the study was an intention-to-treat analysis, we included cART regimen and individual ART drugs as time-fixed covariates. Pre-ART VL and CD4 cell count were defined as measurements taken within six months prior to start of cART. Prior AIDS diagnosis was defined as having a CDC disease stage C prior to cART initiation. All variables measured were entirely observational according to site's local practices. Covariates from the univariate analysis with p < 0.10 were fitted in the multivariate model using backward stepwise selection process. Covariates with p < 0.05 in the multivariate model were considered significant.
Ethics approval were obtained from respective local ethics committees of all TAHOD-participating sites, the Kirby Institute (data management and statistical analysis centre), and TREAT Asia/amfAR (coordinating centre). All data management and statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA) and Stata software version 14.2 (Stata Corp., College Station, TX, USA).

| RESULTS AND DISCUSSION
A total of 1927 PLHIV receiving cART with no evidence of prior DM were included from 20 sites from the participating countries in TAHOD cohort. Of participants with glucose parameters available prior to cART initiation, there were 228 participants with previous diagnosis of DM who were excluded from the study. Sites contributed a median of 61 participants (IQR 20 to 163 participants) in the analysis. The median age was 35 years (IQR 30 to 41) and the median CD4 cell count at ART initiation was 162 cells/lL (IQR: 59 to 248) ( Table 1). The majority were males (74%) and acquired HIV via heterosexual route (57%). Hepatitis C virus co-infection occurred in 11% of total PLHIV. 1497 (78%) and 364 (19%) of the participants were on NRTIs plus NNRTI (nonnucleoside reverse transcriptase inhibitor) and NRTIs plus PI as initial cART regimen respectively. There were 127 PLHIV (7%) who had DM after cART with an incidence rate of 1.08 per 100 person-years (/100PYS) under a median follow-up time of 5.9 years (IQR: 2.8 to 8.9 years). Of the 127 participants, 117 met the FBG criteria, 9 met the HbA1C criteria and 1 met the OGTT criteria for DM. The incidence rate for DM for those with HCV co-infection was 1.18/100PYS which was higher than 1.05/100PYS for HCV negative participants, however the HR was not statistically significant when included in the univariate Cox regression analysis (p = 0.219). Factors associated with development of DM after ART initiation are shown in Table 2. Calendar year of follow-up (p < 0.001), age (p < 0.001), sex (p = 0.004), BMI (p = 0.009), high blood pressure (p = 0.001) and dyslipidaemia (p = 0.059) were significant in the univariate analysis and included in the multivariate model.
In Overall incidence of new-onset DM is 1.08 per 100PYS in a median follow-up time of nearly six years after cART initiation among PLHIV from a cohort in the Asia-Pacific region. Risk factors associated with new-onset DM in this cohort includes older age, higher BMI and high blood pressure. The DM incidence is similar to previous studies [10,16] in the Asia-Pacific region, with a total of 11,798 person-years of follow-up among 1927 PLHIV in our multicenter cohort in the Asia-Pacific, which includes 20 sites from 12 different countries and territories. The incidence rate of DM was higher among males (1.22/100PYS) compared to females (0.63/100PYS). Traditional risks factors such as age, BMI and high blood pressure were found to be predictors of DM, reflecting the emergence of non-communicable disease comorbidities. In addition, we also found that the later years of follow-up was associated with higher DM incidence, compared to follow-up years before 2010. This could be partly due to the collection of OGTT, HbA1C and RPG results into the cohort after 2015, although the routine FBG median FBG testing frequency was the same as once per patient per year. We did not observe an association between DM and the use of baseline ARVs such as stavudine or didanosine in the initial regimen or the use of PIs-or NNRTIs-based regimen as baseline ART.
Incident DM was more common in males, participants with increasing age and higher BMI, which are consistent with the risk factors among the general population. Moreover, there was no significant difference between HIV exposure risks (heterosexual, men who have sex with men (MSM) and injectable drug users) and incident DM in our study. DM incidence among PLHIV has varied by geographic region and country income level. The US Multicenter AIDS Cohort Study (MACS) [21] and Women's Interagency HIV Study (WIHS) [22], where the majority of the participants were African-American and Hispanic/Latino, have reported higher incidence rates of DM, with 4.7/100PYS and 3.4/100PYS among HIV-infected participants on ART in MACS and WIHS respectively. Our findings were similar to the incidence rates from the previous studies [10,16,23]. However, lower rates have been reported from other cohorts such as Swiss HIV Cohort [24] and D:A:D study [8], with incidence rates of 0.44/100PYS and 0.57/100PYS respectively. Differences in DM incidence rates among the cohorts are possibly due to different ethnic backgrounds since Asians are more prone to have more visceral adipose tissue accumulation than European population, which could lead to higher chances of insulin resistance [25,26]. Additionally, the differences in DM incidence among cohorts may also be contributed by the differences in the length of follow-up time and the definition of DM we used in the study which did not require a confirmation test of fasting blood sugar which can result in having lower specificity for detecting DM.
With regard to HIV-and ART-related factors, we did not observe associations between DM and time-updated CD4 cell counts, HIV viral load or prior AIDS-defining events. Previous reports have suggested that persistent inflammation due to chronic infection may have an impact on the pathogenesis of DM [27]. In addition to the effect of chronic inflammation of HIV infection on insulin resistance, past studies have shown an association of ART with DM development. Most of our participants (80%) who developed DM were on NNRTIs-based regimen as initial cART. In this analysis, we did not find the link of either stavudine or didanosine with incident DM, even though the use of stavudine as initial cART regimen was high (28%) in our cohort. It is interesting to note that stavudine was suggested to have been associated with lipodystrophy [28] but not DM in the same population. Furthermore, certain PIs may contribute to the inhibition of glucose transporter (glucose transporter type 4 isoform, GLUT4) and the reduction in insulin sensitivity, which could lead to decreased glucose uptake in peripheral adipose tissues and consequently result in the development of insulin resistance [29]. Contrary to the previous studies [9,24], we found no association between DM and PI-versus NNRTI-based initial cART regimen.
It is noteworthy that high blood pressure was also associated with DM incidence in our cohort. This is consistent with the findings from the general population without HIV infection [30]. HIV-infected participants with high blood pressure and DM may also have risks for multiple comorbidities and polypharmacy. This will increase the pill burden and have a high possibility to have drug-to-drug interactions. Therefore, proper treatment and risk reduction strategies such as diet and exercise should be implemented in those who have higher BMI and abnormal blood pressure. Our findings showed that traditional risk factors were associated with DM development among Asian HIV-infected individuals, suggesting the needs for the importance of clinicians to timely diagnose and properly manage DM in HIV-infected individuals. Even though we did not observe the use of ARVs to be an additional risk for the occurrence of DM, however it remains crucial to monitor and evaluate the potential toxicities from different ARVs used.  The limitations of the study include the unavailability of other indicators for central obesity such as abdominal fat and waist-hip circumference ratio. We therefore were not able to evaluate the effects of these indicators on the development of DM in our analysis. Also, another limitation of this study is that we did not include HIV-negative controls to compare the incidence rates of DM between the groups. Due to the limitations in DM testing data, such as FBG, we did not use a second confirmatory testing for DM, which could possibly lead to inaccurate estimation of our cohort's DM incidence rate. As our cohort recruit participants based on the likelihood of remaining in care, the study population may not represent patients typically seen at the clinical sites. Finally, our study has limitations for inability to adjust all the unobserved confounding factors due to the observational nature of the cohort.

| CONCLUSIONS
Our analysis shows DM in PLHIV is common in settings from Asian countries. Traditional risk factors such as age, sex, high blood pressure and BMI were found to be associated with the development of DM in our cohort. Careful assessment and routine screening for DM and other co-existing comorbid conditions, especially among older and obese PLHIV, are essential.

C O M P E T I N G I N T E R E S T S
The authors do not have any competing interests to declare.