Abstract
Objective
Diabetes is a growing cause of morbidity and mortality in people living with HIV (PLHIV) receiving antiretroviral therapy (ART). We investigated the association between fasting plasma glucose (FPG) levels and mortality, and factors associated with FPG monitoring rates in Asia.
Methods
Patients from the TREAT Asia HIV Observational Database Low Intensity TransfEr (TAHOD-LITE) were included if they had initiated ART. Competing risk and Poisson regression were used to analyse the association between FPG and mortality, and assess risk factors of FPG monitoring rates, respectively. FPG was categorised as diabetes (FPG ≥7.0 mmol/L), prediabetes (FPG 5.6–6.9 mmol/L) and normal FPG (FPG <5.6 mmol/L).
Results
In total, 33,232 patients were included in the analysis. Throughout follow-up, 59% had no FPG test available. The incidence rate for diabetes was 13.7 per 1000 person-years in the 4649 patients with normal FPG at ART initiation. Prediabetes (sHR 1.32 95%CI 1.07–1.64) and diabetes (sHR 1.90 95%CI 1.52–2.38) were associated with mortality compared to those with normal FPG. FPG monitoring increased from 0.34–0.78 tests per person-year from 2012–2016 (p<0.001). Male sex (IRR 1.08, 95%CI 1.03–1.12), age >50 years (IRR 1.14, 95%CI 1.09–1.19) compared to ≤40 years, and CD4 ≥500 cells/uL (IRR 1.04, 95%CI 1.00–1.09) compared to <200 cells/uL were associated with increased FPG monitoring.
Conclusion
Diabetes and prediabetes were associated with mortality. FPG monitoring increased over time, however, less than half of our cohort had been tested. Greater resources should be allocated towards FPG monitoring for early diabetic treatment and intervention and to optimise survival.
Keywords: HIV, diabetes, mortality, glucose monitoring, Asia
Introduction
Due to effective treatment with combination antiretroviral therapy (ART), treatment outcomes have substantially improved in people living with HIV (PLHIV). However, due to the side-effects of treatment and aging of the HIV-positive population [1], metabolic co-morbidities have become more prevalent. Evidence shows that PLHIV receiving ART are at increased risk of diabetes. A meta-analysis found that PLHIV on ART had an almost 4-fold increased odds of having diabetes and especially those on ART for more than 18 months had significantly higher fasting plasma glucose (FPG) [2]. Some antiretroviral drugs in particular, specifically zidovudine (ZDV), stavudine (d4T), didanosine (ddI), indinavir (IDV), and lopinavir (LPV), are thought to contribute to increasing FPG levels [3].
Few studies in the Asia-Pacific region have reported on diabetes in PLHIV. Studies from Taiwan, Thailand, and Korea have shown increased FPG levels, indicating diabetes or prediabetes, in 31%, 21%, and 10% of their HIV-positive study populations, respectively [4–6]. Other available data show relatively high diabetes incidence rates of approximately 26, 13, and 5 per 1000 person-years (pys) after ART initiation in China, Taiwan, and Thailand, respectively [7–9]. These findings suggest that diabetes-associated morbidity and mortality may be a growing burden in PLHIV in the Asia-Pacific region, especially as PLHIV are living longer [10] and endure long-term exposure to ART.
In efforts to prevent and manage diabetes and its negative impacts in PLHIV, standards of practice have been developed for screening and monitoring of diabetes specifically in this population. The preferred screening method to diagnose diabetes in PLHIV is performing an FPG test and this diagnosis can be confirmed by repeat testing [11, 12]. The American Diabetes Association (ADA) guidelines recommend FPG monitoring in PLHIV as follows: every 6 to 12 months in those who have not initiated ART yet, 3 months after ART initiation or modification, and, thereafter, every year if previous test results indicated normal FPG levels [12]. This screening advice differs from that for the general population, in which screening is recommended every three years in case of normal FPG levels. The 2017 AIDS info guidelines of the U.S. Department of Health and Human Services additionally recommend testing of FPG levels every 3–6 months when a previous test indicated prediabetes [3]. In some regions in the Asia-Pacific, frequent monitoring of FPG may not be feasible due to a lack of health resources.
To adequately allocate health resources and improve long-term health outcomes of PLHIV in the Asia-Pacific region, more knowledge is needed on the prevalence and incidence of diabetes, its impacts, and the frequency of FPG monitoring in this population. This study conducted on a multi-country cohort of PLHIV on ART in the Asia-Pacific region aimed to investigate the incidence of diabetes, assess the association of FPG levels with all-cause mortality, and assess factors associated with FPG monitoring rates.
Methods
Study design and patients
The TREAT Asia HIV Observational Database Low Intensity TransfEr (TAHOD-LITE) is a sub-study of the TREAT Asia HIV Observational Database (TAHOD), a prospective observational cohort of the International Epidemiology Databases to Evaluate AIDS. TAHOD collects detailed data from a subset of HIV-positive adult patients at 20 clinical sites in 12 Asian countries and territories. TAHOD-LITE was initiated in 2014 to better represent the complete clinical population, and collects a more limited dataset (demographics, hepatitis serology, ART history, and HIV-related laboratory results) from all patients at participating sites. The 2017 TAHOD-LITE data transfer gathered additional data on plasma glucose and creatinine, and 10 of 20 TAHOD sites participated. Both studies have been described in more detail previously [13–15]. Institutional Review Board approvals were obtained at all participating sites, the data management and analysis centre (The Kirby Institute, UNSW Sydney, Australia), and the coordinating centre (TREAT Asia/amfAR, Bangkok, Thailand). We included adults (aged ≥18 years at first clinic visit) who started ART between January 1992-June 2017 and had at least one clinic visit after ART initiation.
Fasting plasma glucose and covariates
FPG was categorised according to ADA definitions for diabetes (FPG ≥7.0 mmol/L), prediabetes (FPG 5.6–6.9 mmol/L) and normal (FPG <5.6 mmol/L) [12], with an additional category for patients whose FPG was not tested/unavailable. Covariates were sex, age group (≤40, 41–50, and >50 years), HIV exposure category (heterosexual, men who have sex with men [MSM], injecting drug use, and other), HBsAg test result and anti-HCV test result (ever tested positive, yes or no), HIV viral load (<1000 and ≥1000 copies/mL), CD4 count (<200, 200–349, 350–499, and ≥500 cells/μL), and calendar year (continuous). Age, HIV viral load, CD4 count, and calendar year were time-updated variables. We also included specific antiretroviral drugs, i.e. ZDV, d4T, ddI, IDV, and LPV, as time-updated variables.
Statistical analysis
Baseline descriptive data of all variables were provided in numbers and percentages or median and interquartile range (IQR), as appropriate. The incidence rate (IR) per 1000 pys of prediabetes and diabetes was calculated for those patients who started ART with normal FPG levels and the percentage was given of patients with a confirmed diabetes test (i.e. a second FPG test ≥7.0 mmol/L). Furthermore, we conducted two analyses, one to assess the association between FPG levels and one to determine factors associated with glucose monitoring rates. Data management and statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA) and Stata software version 14.1 (StataCorp, College Station, TX, USA).
Analysis 1. Glucose levels and all-cause mortality
Fine and Gray competing risk regression [16] was used to assess the association between FPG levels and all-cause mortality, with lost to follow-up (LTFU) treated as a competing risk. Follow-up time was censored at death, LTFU (not seen at clinic in last 12 months), transfer to another clinic, or last clinic visit. Throughout follow-up, patients could contribute to multiple categories of FPG if their FPG levels increased over time. For example, if patients started with normal FPG but had a test indicating diabetes at a later time, they were considered to have diabetes from that time onwards and subsequent tests were ignored if they indicated lower levels of FPG. We provided the overall mortality IR per 1000 pys and we calculated sub-hazard ratios (sHR) for the association between FPG levels and all-cause mortality in various models. Model A adjusted for site and year, to account for potential differences in testing frequency at site level or over time. Model B adjusted for site, calendar year and covariates that were significant (p <0.05) in multivariable analysis which was conducted using a stepwise backwards selection procedure and included all covariates that were univariately associated with all-cause mortality at Wald’s test p <0.10. Specific antiretroviral drugs were not considered in this analysis. Three sensitivity analyses were conducted: in Model C we excluded all patients with diabetes at ART initiation; in Model D we additionally excluded all patients with prediabetes at ART initiation; and in Model E we excluded sites in which >80% of patients had unknown FPG throughout follow-up.
As a supplementary analysis, we also presented the univariate and multivariate models in which patients were only considered to have diabetes if they had a second FPG test ≥7.0 mmol/L.
Analysis 2. Factors associated with glucose monitoring rates
Poisson regression was used to investigate factors associated with glucose monitoring rates. Analysis time began from 2012 and was censored at death, LTFU, transfer to another clinic, last clinic visit or 31 December 2016, whichever came first. For three sites analysis time was left truncated at 2013 or 2014 because glucose testing data was not provided for earlier years. Annual crude glucose monitoring rates were calculated per person-year for 2012–2016. Furthermore, we used the stepwise backwards selection procedure (similar to Analysis 1) to create a multivariable model. In a separate analysis we included specific antiretroviral drugs in addition to the other covariates of the multivariable model. All analyses were adjusted for site and calendar year.
Results
Table 1 shows characteristics of the 33,232 included patients. The majority of patients were from India (61%), followed by Singapore (9%), Cambodia (9%), Vietnam (7%), Indonesia (5%), Hong Kong (4%), South Korea (2%), and Taiwan (2%). Median age at ART initiation was 36 years (IQR 30–42) and 23,059 (69%) patients were men. Median year of ART initiation was 2010 (IQR 2006–2012), median CD4 count at ART initiation was 154 cells/μL (IQR 59–268) and the median HIV viral load at ART initiation was log10 5.0 copies/mL (IQR 4.4–5.5). Positive HBsAg and anti-HCV were observed in 1555 (5%) and 1673 (5%) of patients, respectively. FPG at ART initiation was not available in 27,321 (82%) patients, but in the remaining patients the median FPG was 4.8 mmol/L (IQR 4.3–5.4). FPG assessments were not available throughout follow-up for 19,681 (59%) patients. Of the 4649 patients who started ART with normal FPG levels, 926 (20%) developed prediabetes (IR 43.0 per 1000 pys) and 332 (7%) developed diabetes during follow up (IR 13.7 per 1000 pys). Of these, 98% had confirmed diabetes as indicated by a second FPG test ≥7.0 mmol/L.
Table 1:
Patient characteristics (N=33,232)
| Total population | Patients who died | |
|---|---|---|
| N (%) | N (%) | |
| Total | 33232 (100.0) | 1518 (100.0) |
| Sex | ||
| Female | 10173 (30.6) | 301 (19.8) |
| Male | 23059 (69.4) | 1217 (80.2) |
| Age at ART initiation (years) | ||
| ≤40 | 23151 (69.7) | 919 (60.5) |
| 41–50 | 6839 (20.6) | 314 (20.7) |
| >50 | 3242 (9.8) | 285 (18.8) |
| Median (IQR) | 36 (30–42) | 38 (32–46) |
| HIV exposure category | ||
| Heterosexual | 19492 (58.7) | 1151 (75.8) |
| MSM | 3216 (9.7) | 99 (6.5) |
| IDU | 1214 (3.7) | 132 (8.7) |
| Other/unknown | 9310 (28.0) | 136 (9.0) |
| Pre-ART HIV viral load (copies/mL) | ||
| <1000 | 473 (1.4) | 11 (0.7) |
| ≥1000 | 6033 (18.2) | 324 (21.3) |
| Missing/not tested | 26726 (80.4) | 1183 (77.9) |
| Median log (IQR) | 5.0 (4.4–5.5) | 5.2 (4.8–5.7) |
| Pre-ART CD4 count (cells/μL) | ||
| <200 | 14182 (42.7) | 1028 (67.7) |
| 200–349 | 5933 (17.9) | 122 (8.0) |
| 350–499 | 1871 (5.6) | 24 (1.6) |
| ≥500 | 1254 (3.8) | 22 (1.4) |
| Missing/not tested | 9992 (30.1) | 322 (21.2) |
| Median (IQR) | 154 (59–268) | 65 (27–147) |
| ART initiation year | ||
| Median (IQR) | 2010 (2006–2012) | 2007 (2005–2011) |
| Positive HBsAg* | ||
| No | 17349 (52.2) | 785 (51.7) |
| Yes | 1555 (4.7) | 126 (8.3) |
| Missing/not tested | 14328 (43.1) | 607 (40.0) |
| Positive anti-HCV* | ||
| No | 12275 (36.9) | 660 (43.5) |
| Yes | 1673 (5.0) | 160 (10.5) |
| Missing/not tested | 19284 (58.0) | 698 (46.0) |
| Pre-ART FPG (mmol/L) | ||
| Normal | 4649 (14.0) | 231 (15.2) |
| Prediabetes | 919 (2.8) | 53 (3.5) |
| Diabetes | 344 (1.0) | 28 (1.8) |
| Missing/not tested | 27320 (82.2) | 1206 (79.4) |
| Median (IQR) | 4.8 (4.3–5.4) | 4.9 (4.4–5.6) |
ART, antiretroviral therapy; IQR, interquartile range; FPG, fasting plasma glucose; MSM, men who have sex with men; IDU, injecting drug use.
HBsAg and anti-HCV test on ever tested positive throughout follow-up.
Analysis 1. Glucose levels and all-cause mortality
The median follow-up time in our cohort was 4.2 years (IQR 1.3–7.5). In total, 1518 patients died (IR: 9.3 per 1000 pys) and 9689 were LTFU (IR: 59.6 per 1000 pys) during 162,668 pys at risk. Of 13,551 patients with at least one FPG assessment during follow-up, 3945 (29%) had a highest FPG ≥7.0 mmol/L, and 2037 (15%) had a highest FPG 5.6–6.9 mmol/L. Figure 1 shows the association between FPG levels and all-cause mortality. In univariate competing risk regression (Model A), those with prediabetes and diabetes had a 1.36- and 2.32-fold increased sub-hazard of all-cause mortality, respectively, compared to those with normal FPG levels. The association between diabetes (sHR 1.90 95%CI 1.52–2.38) and all-cause mortality decreased in multivariable analysis, while it remained similar for prediabetes (sHR 1.32 95%CI 1.07–1.64) (Model B). However, the sub-hazard for diabetes increased substantially in sensitivity analyses I in which we excluded patients with known diabetes at baseline (Model C), and sensitivity analysis II, in which we excluded patients with known prediabetes or diabetes at baseline (Model D). Results were similar across these sensitivity analyses, with the hazard of all-cause mortality in those with new onset diabetes being around 2.4-fold larger compared to those who maintained normal FPG levels. In sensitivity analysis III we excluded sites where >80% of patients did not have FPG data available (Model E). Associations in this analysis were attenuated to a 1.79-fold increased mortality hazard in those with diabetes. In all analyses, those who did not have any FPG assessment available throughout follow-up had similar mortality compared to those with normal FPG levels.
Figure 1:
Association between fasting plasma glucose and mortality A = Adjusted for site and calendar year. B = Adjusted for all significant variables in multivariate analysis (i.e. sex, age group, HIV exposure category, HBsAg test result, anti-HCV test result, CD4 count, calendar year, and site). C = Sensitivity analysis I: Excluding patients with diabetes at ART initiation. D = Sensitivity analysis II: Excluding patients with diabetes or prediabetes at ART initiation E = Sensitivity analysis III: Excluding sites where >80% of patients had unknown FPG throughout follow-up
In a supplementary analysis we repeated the univariate and multivariate models with diabetes being defined by a second FPG test ≥7.0 mmol/L. Only 8350 of the 33,232 (25%) patients had a second FPG tests after the first assessment. Median time between the first FPG test ≥7.0 mmol/L and the subsequent FPG test was 168 days (IQR 91–331). In the univariate model using the confirmed diabetes definition, diabetes was significantly associated with a 1.5-fold increased sub-hazard of all-cause mortality compared to those with normal FPG levels. In the multivariate model this association was decreased and no longer significant (sHR 1.26 95%CI 0.89–1.77).
Analysis 2. Factors associated with fasting plasma glucose monitoring rates
Crude FPG monitoring rates increased over time (p<0.001), from 0.34 tests per person-year in 2012 to 0.78 tests per person-year in 2016 (Figure 2). Table 2 shows the factors associated with FPG monitoring rates. Except for HIV viral load and positive HBsAg or anti-HCV, all factors considered were significantly associated with FPG monitoring rates in the univariate analysis, adjusting for site and calendar year. Overall, the results were somewhat attenuated in multivariate analysis. Male sex (IRR 1.10, 95%CI 1.06–1.14), older age (41–50 years: IRR 1.13, 95%CI 1.09–1.16, and >50 years: IRR 1.16, 95%CI 1.12–1.21, compared to ≤40 years) and higher CD4 count (200–349 cells/uL: IRR 1.07, 95%CI 1.02–1.13, 350–499 cells/uL: IRR 1.08, 95%CI 1.03–1.13, and ≥500 cells/uL: IRR 1.04, 95%CI 1.00–1.09, compared to <200 cells/uL) were independently associated with increased FPG monitoring. Compared to heterosexual HIV exposure, MSM HIV exposure category was associated with reduced FPG monitoring (IRR 0.95, 95%CI 0.91–0.99).
Figure 2:
Crude glucose monitoring rates 2012–2016
* Excluding 3 sites due to lack of FPG data;
^ Excluding 2 sites due to lack of FPG data
Table 2:
Factors associated with fasting plasma glucose monitoring rates
| Univariate# | Multivariate~ | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Person-years | Assessments (n) | Crude rate | IRR | 95%CI | P-value | IRR | 95%CI | P-value | |
| Sex | |||||||||
| Female | 19482.5 | 5229 | 0.27 | 1.00 | 1.00 | ||||
| Male | 43134.6 | 21541 | 0.50 | 1.10 | 1.06–1.14 | <0.001 | 1.10 | 1.06–1.14 | <0.001 |
| Age (years)* | <0.001 | <0.001 | |||||||
| ≤40 | 29237.8 | 10392 | 0.36 | 1.00 | 1.00 | ||||
| 41–50 | 21364.7 | 9265 | 0.43 | 1.15 | 1.12–1.19 | <0.001 | 1.13 | 1.09–1.16 | <0.001 |
| >50 | 12014.7 | 7113 | 0.59 | 1.19 | 1.15–1.24 | <0.001 | 1.16 | 1.12–1.21 | <0.001 |
| HIV exposure category | <0.001 | <0.001 | |||||||
| Heterosexual | 36942.1 | 11609 | 0.31 | 1.00 | 1.00 | ||||
| MSM | 8218.4 | 9066 | 1.10 | 0.93 | 0.89–0.98 | 0.002 | 0.95 | 0.91–0.99 | 0.024 |
| IDU | 2038.8 | 888 | 0.44 | 0.95 | 0.88–1.04 | 0.253 | 0.95 | 0.87–1.04 | 0.256 |
| Other/unknown | 15417.8 | 5207 | 0.34 | 0.77 | 0.73–0.82 | <0.001 | 0.79 | 0.75–0.83 | <0.001 |
| HIV viral load (copies/mL)* | |||||||||
| <1000 | 34719.5 | 22287 | 0.64 | 1.00 | 1.00 | ||||
| ≥1000 | 5223.1 | 1696 | 0.32 | 0.97 | 0.92–1.03 | 0.307 | 0.97 | 0.91–1.02 | 0.234 |
| Missing/not tested | 22674.5 | 2787 | 0.12 | - | - | - | - | - | - |
| CD4 count (cells/uL)* | 0.005 | 0.006 | |||||||
| <200 | 8405.6 | 2729 | 0.32 | 1.00 | 1.00 | ||||
| 200–349 | 12460.9 | 5220 | 0.42 | 1.08 | 1.02–1.13 | 0.004 | 1.07 | 1.02–1.13 | 0.004 |
| 350–499 | 14462.8 | 6815 | 0.47 | 1.08 | 1.03–1.13 | 0.003 | 1.08 | 1.03–1.13 | 0.002 |
| ≥500 | 25485.6 | 11909 | 0.47 | 1.04 | 0.99–1.09 | 0.105 | 1.04 | 1.00–1.09 | 0.080 |
| Missing/not tested | 1802.3 | 97 | 0.05 | - | - | - | - | - | - |
| Positive HBsAg | |||||||||
| No | 38102.0 | 20554 | 0.54 | 1.00 | 1.00 | ||||
| Yes | 3616.6 | 2159 | 0.60 | 0.97 | 0.92–1.03 | 0.321 | 0.96 | 0.90–1.01 | 0.119 |
| Missing/not tested | 20898.6 | 4057 | 0.19 | - | - | - | - | - | - |
| Positive anti-HCV | |||||||||
| No | 29863.7 | 18676 | 0.63 | 1.00 | 1.00 | ||||
| Yes | 3392.8 | 1951 | 0.58 | 0.97 | 0.92–.03 | 0.379 | 0.99 | 0.92–1.05 | 0.697 |
| Missing/not tested | 29360.6 | 6143 | 0.21 | - | - | - | - | - | - |
P-values for test for heterogeneity excluded missing or not-tested values. P-values in bold represent significant covariates in the final model.
adjusted for site and year.
additionally adjusted for all factors that were significant in the final model.
Variables were time-updated, i.e. each patient can contribute to more than one category.
IRR, incidence rate ratio; CI, confidence interval; ART, antiretroviral treatment; MSM, men who have sex with men; IDU, injecting drug use.
Next, we assessed the association between prescription of specific antiretroviral drugs known to increase the risk of diabetes, and monitoring rates (Table 3). Crude monitoring rates were somewhat lower in patients taking ZDV or d4T, whereas they were higher in those taking ddI, IDV, or LPV. When controlling for site and calendar year there was a slight to moderate increase in monitoring rates in patients using any of these specific antiretroviral drugs. However, in the multivariate analysis, only use of LPV remained significantly associated with FPG monitoring rates compared to those who did not have LPV in their regimen (IRR 1.43, 95%CI 1.37–1.49).
Table 3:
Association between prescription of specific antiretroviral drugs and fasting plasma glucose monitoring rates
| Univariate# | Multivariate~ | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Person-years | Assessments (n) | Crude rate | IRR | 95%CI | P-value | IRR | 95%CI | P-value | |
| ZDV | |||||||||
| No | 30526.5 | 14186 | 0.46 | 1.00 | 1.00 | ||||
| Yes | 32090.6 | 12584 | 0.39 | 1.05 | 1.02–1.08 | 0.002 | 1.02 | 0.99–1.05 | 0.191 |
| d4T | |||||||||
| No | 39671.5 | 19939 | 0.50 | 1.00 | 1.00 | ||||
| Yes | 22945.6 | 6831 | 0.30 | 1.05 | 1.02–1.09 | 0.004 | 1.01 | 0.97–1.04 | 0.671 |
| ddI | |||||||||
| No | 59955.7 | 24146 | 0.40 | 1.00 | 1.00 | ||||
| Yes | 2661.5 | 2624 | 0.99 | 1.10 | 1.04–1.17 | 0.001 | 1.05 | 1.00–1.11 | 0.065 |
| IDV | |||||||||
| No | 60939.5 | 24887 | 0.41 | 1.00 | 1.00 | ||||
| Yes | 1677.6 | 1883 | 1.12 | 1.07 | 1.00–1.15 | 0.036 | 0.99 | 0.93–1.06 | 0.842 |
| LPV | |||||||||
| No | 56315.9 | 20853 | 0.37 | 1.00 | 1.00 | ||||
| Yes | 6301.2 | 5917 | 0.94 | 1.46 | 1.40–1.53 | <0.001 | 1.43 | 1.37–1.49 | <0.001 |
adjusted for site and year.
additionally adjusted for all factors significant in multivariable model (Table 2), i.e. sex, age HIV exposure category, CD4 count.
Use of specific antiretroviral drugs was time updated, as such, i.e. patient can contribute to more than one category.
IRR, incidence rate ratio; CI, confidence interval; ZDV, zidovudine; d4T, stavudine; ddI, didanosine; IDV, indinavir; LPV, lopinavir
Discussion
In the TAHOD-LITE cohort, around 7% of patients with normal FPG at ART initiation developed diabetes during follow up (IR 13.7 per 1000 pys) and around 20% developed prediabetes (IR 43.0 per 1000 pys). Diabetes was independently associated with a 90% increased and prediabetes with a 32% increased sub-hazard of all-cause mortality. FPG monitoring rates increased over time and were higher in males, older patients, and those with higher CD4 count. FPG monitoring rates were 43% higher in those on LPV, but no difference was seen in monitoring rates of patients on any of the other antiretroviral drugs.
In our cohort of PLHIV, we found a diabetes incidence rate of 13.7 per 1000 pys. Notably, our findings are identical to a recent meta-analysis that found a global diabetes incidence in PLHIV of 13.7 per 1000 pys [17]. This falls within the range of those reported in other studies of PLHIV in the region [7–9]. The incidence of diabetes in our study is substantially higher than in the general population of countries represented in TAHOD-LITE. Based on 2016 data from the Global Burden of Disease, incidence was estimated to vary from around 2.2 per 1000 pys in Cambodia to almost 5.1 per 1000 pys in Taiwan [18]. This corroborates literature describing PLHIV having a 4-fold increased risk of developing diabetes compared to the general population.
Since diabetes and HIV both affect CD4 T cell immune response [19, 20], disease outcomes may be less favourable in those who are affected by both conditions. In the DAD cohort, all-cause mortality was 1.77 times more prevalent in those with diabetes compared with those without diabetes (relative rate 1.77, 95%CI 1.54–2.03) [21]. Supporting this association, we found a 90% increased mortality hazard in TAHOD-LITE patients who had comorbid diabetes. Associations were stronger in sensitivity analyses where we excluded patients with known increased FPG levels before ART initiation. It is possible that patients who started ART with elevated FPG levels, and their healthcare providers, were more aware of the associated risks and better able to adopt or support appropriate lifestyle changes or treatment plans to reduce and maintain FPG levels, and thus improve overall survival outcomes.
When assessing annual FPG testing in the last five years, we found improved monitoring rates over time. FPG was tested more frequently in men and older patients, factors previously associated with increased diabetes risk [22]. Testing was slightly less frequent in patients with CD4 counts <200 cells/μL, possibly because diabetes is not considered a priority health issue in those with more advanced HIV. Furthermore, mortality in those not tested on FPG was comparable to those with normal FPG levels, possibly indicating that clinicians decided not to perform testing when patients were expected to have no or low diabetes risk. Taken together, this suggests that clinical sites conduct some form of targeted testing in those at higher risk of diabetes when routine testing is not feasible. Higher monitoring rates were also seen in those on LPV, but not in those taking any of the other drugs that require more frequent FPG monitoring. This may be an additional indication that sites did not have enough resources to test all patients frequently. It should be noted that crude monitoring rates cannot be interpreted as annual monitoring rates since monitoring rates differed substantially between sites. Due to differences in available resources, sites in high-income countries may test patients more frequently compared to sites in lower-income countries. This might partly explain why LPV use was associated with FPG monitoring rates while other antiretroviral drugs were not. LPV use was particularly common in one high-income site and FPG monitoring rates in this site were considerably higher than the overall monitoring rate.
Regardless of type of antiretroviral drug used, both the guidelines of the U.S. Department of Health and Human Services and the ADA recommend FPG testing in PLHIV before ART initiation and at least annually once on ART [3, 12]. Testing from 2012 to 2016 in TAHOD-LITE was substantially lower than recommended by these guidelines. However, these general guidelines may deviate from recommendations provided by country specific and local guidelines. Almost 60% of our study population did not have a FPG assessment available throughout follow-up, signifying that they may never have been tested for diabetes. Without FPG monitoring, there is little opportunity to adequately intervene and prevent avoidable cases of diabetes and diabetes-related morbidity and mortality in Asian PLHIV.
Our study has several strengths and limitations. We had a large number of patients, representing the wider clinical population at various sites across the Asia-Pacific region, long-term follow-up, and availability of all clinical data on FPG and HIV related laboratory results. Furthermore, this was one of few large studies to assess the association between FPG levels and mortality. A limitation was that for our main models we used the broader categorization of FPG levels instead of confirmed diabetes diagnoses because repeat FPG testing to verify diabetes is not routinely performed at participating sites. Although we have also reported results based on diabetes as defined by a second FPG ≥7.0 mmol/L, these findings are likely to underestimate the association between diabetes and mortality due to the low number of repeated tests and high time interval in between the tests. Additionally, data on antidiabetic treatment and cause-specific mortality was not available in our cohort. As antidiabetic treatment can be protective of mortality [23, 24], incorporating treatment-related information would likely show an amplification of the association between untreated diabetes and mortality. Information on cause-specific mortality might have shed more light on the contribution of diabetes to mortality related to cardiovascular or kidney disease, which may be secondary to diabetes. Lastly, a high proportion of missing FPG data was from one large site in a lower-middle-income country, which contributed almost a third of the entire TAHOD-LITE study population. As such, FPG monitoring rate findings may be distorted and may actually be higher in some of the other sites. Nonetheless, in a sensitivity analysis which excluded sites with a high proportion of missing data the association between FPG levels and mortality remained significant, supporting the robustness of our findings.
In summary, our findings suggest substantial levels of diabetes and prediabetes among our adult PLHIV cohort, and show that higher FPG levels are associated with increased all-cause mortality. Although FPG monitoring rates have increased over time, the majority of PLHIV in our study did not receive an FPG assessment. Improved FPG monitoring in PLHIV in the Asia-Pacific region is essential to enable timely intervention and to avoid preventable cases of diabetes and related morbidity and mortality.
Acknowledgements
PS Ly and V Khol, National Center for HIV/AIDS, Dermatology & STDs, Phnom Penh, Cambodia; MP Lee, PCK Li, W Lam and YT Chan, Queen Elizabeth Hospital, Hong Kong SAR; N Kumarasamy, S Saghayam and C Ezhilarasi, Chennai Antiviral Research and Treatment Clinical Research Site (CART CRS), YRGCARE Medical Centre, VHS, Chennai, India; S Pujari, K Joshi, S Gaikwad and A Chitalikar, Institute of Infectious Diseases, Pune, India; TP Merati, DN Wirawan and F Yuliana, Faculty of Medicine Udayana University & Sanglah Hospital, Bali, Indonesia; OT Ng, PL Lim, LS Lee and Z Ferdous, Tan Tock Seng Hospital, Singapore; JY Choi, Na S and JM Kim, Division of Infectious Diseases, Department of Internal Medicine, Yonsei University College of Medicine, Seoul, South Korea; WW Wong, WW Ku and PC Wu, Taipei Veterans General Hospital, Taipei, Taiwan; CD Do, AV Ngo and LT Nguyen, Bach Mai Hospital, Hanoi, Vietnam; KV Nguyen, HV Bui, DTH Nguyen and DT Nguyen, National Hospital for Tropical Diseases, Hanoi, Vietnam; AH Sohn, JL Ross and B Petersen, TREAT Asia, amfAR - The Foundation for AIDS Research, Bangkok, Thailand; R Bijker, A Jiamsakul, D Rupasinghe and MG Law, The Kirby Institute, UNSW Sydney, Sydney, Australia.
Funding
The TREAT Asia HIV Observational Database Low-Intensity TransfEr study is an initiative of TREAT Asia, a program of amfAR, The Foundation for AIDS Research, with support from the U.S. National Institutes of Health’s National Institute of Allergy and Infectious Diseases, the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Cancer Institute, the National Institute of Mental Health, and the National Institute on Drug Abuse, as part of the International Epidemiology Databases to Evaluate AIDS (IeDEA; U01AI069907). The Kirby Institute is funded by the Australian Government Department of Health and Ageing, and is affiliated with the Faculty of Medicine, UNSW Sydney. The PhD of R Bijker has been supported through an Australian Government Research Training Program Scholarship. The content of this publication is solely the responsibility of the authors and does not necessarily represent the official views of any of the governments or institutions mentioned above.
Footnotes
Conflicts of interest
No conflicts of interest.
References
- 1.Deeks SG, Lewin SR, Havlir DV. The end of AIDS: HIV infection as a chronic disease. Lancet 2013; 382: 1525–1533. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Nduka CU, Stranges S, Kimani PK, Sarki AM, Uthman OA. Is there sufficient evidence for a causal association between antiretroviral therapy and diabetes in HIV‐infected patients? A meta‐analysis. Diabetes Metab Res Rev 2017; 33: e2902. [DOI] [PubMed] [Google Scholar]
- 3.Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the use of antiretroviral agents in adults and adolescents living with HIV. Updated Oct 2017. Department of Health and Human Services, February 13, 2018. Available at: http://www.aidsinfo.nih.gov/ContentFiles/AdultandAdolescentGL.pdf. [Google Scholar]
- 4.Wu PY, Hung CC, Liu WC, Hsieh CY, Sun HY, Lu CL, et al. Metabolic syndrome among HIV-infected Taiwanese patients in the era of highly active antiretroviral therapy: prevalence and associated factors. J Antimicrob Chemother 2012; 67: 1001–1009. [DOI] [PubMed] [Google Scholar]
- 5.Jantarapakde J, Phanuphak N, Chaturawit C, et al. Prevalence of metabolic syndrome among antiretroviral-naive and antiretroviral-experienced HIV-1 infected Thai adults. AIDS Patient Care STDS 2014; 28: 331–340. [DOI] [PubMed] [Google Scholar]
- 6.Oh DH, Ahn JY, Kim SI, Kim MJ, Woo JH, Kim WJ, et al. Metabolic complications among Korean patients with HIV infection: The Korea HIV/AIDS Cohort Study. J Korean Med Sci 2017; 32: 1268–1274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Riyaten P, Salvadori N, Traisathit P, Ngo-Giang-Huong N, Cressey TR, Leenasirimakul P, et al. New-onset diabetes and antiretroviral treatments in HIV-infected dults in Thailand. J Acquir Immune Defic Syndr 2015; 69: 453–459. [DOI] [PubMed] [Google Scholar]
- 8.Zhang C, Chow FC, Han Y, Xie J, Qiu Z, Guo F, et al. Multicenter cohort study of diabetes mellitus and impaired fasting glucose in HIV-infected patients in China. J Acquir Immune Defic Syndr 2015; 68: 298–303. [DOI] [PubMed] [Google Scholar]
- 9.Lo YC, Chen MY, Sheng WH, Hsieh SM, Sun HY, Liu WC, et al. Risk factors for incident diabetes mellitus among HIV-infected patients receiving combination antiretroviral therapy in Taiwan: a case-control study. HIV Med 2009; 10: 302–309. [DOI] [PubMed] [Google Scholar]
- 10.Puhr R, Kumarasamy N, Ly PS, et al. HIV and Aging: Demographic Change in the Asia-Pacific Region. J Acquir Immune Defic Syndr 2017; 74: e146–e148 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kim PS, Woods C, Georgoff P, Crum D, Rosenberg A, Smith M, et al. A1C underestimates glycemia in HIV infection. Diabetes Care 2009; 32: 1591–1593. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.American Diabetes Association. Standards of Medical Care in Diabetes 2018. Diabetes Care 2018; 41: S1–S159. [DOI] [PubMed] [Google Scholar]
- 13.De La Mata NL, Ahn MY, Kumarasamy N, Ly PS, Ng OT, Nguyen KV, et al. A pseudo-random patient sampling method evaluated. J Clin Epidemiol 2017; 81: 129–139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.De La Mata NL, Kumarasamy N, Khol V, Ng OT, Nguyen KV, Merati TP, et al. Improved survival in HIV treatment programmes in Asia. Antivir Ther 2016; 21: 517–527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.The TREAT Asia HIV Observational Database (TAHOD) Network. A decade of combination antiretroviral treatment in Asia: The TREAT Asia HIV Observational Database Cohort. AIDS Res Hum Retroviruses 2016; 32: 772–781. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Fine JP, Gray RJ. A proportional hazards model for the sub-distribution of a competing risk. J Am Stat Assoc 1999; 446: 496–509. [Google Scholar]
- 17.Nansseu JR, Bigna JJ, Kaze AD, Noubiap JJ. Incidence and risk factors for prediabetes and diabetes mellitus among HIV-infected adults on antiretroviral therapy: a systematic review and meta-analysis. Epidemiology 2018; 29: 431–441. [DOI] [PubMed] [Google Scholar]
- 18.Global Burden of Disease Study 2016. Global Burden of Disease Study 2016 (GBD 2016) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), June 6, 2018. Available at: http://ghdx.healthdata.org/gbd-results-tool. [Google Scholar]
- 19.Xia C, Rao X, Zhong J. Role of T lymphocytes in type 2 diabetes and diabetes-associated inflammation. J Diabetes Res 2017; epub 6494795. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Walker LS, von Herrath M. CD4 T cell differentiation in type 1 diabetes. Clin Exp Immunol 2016; 183: 16–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Data Collection on Adverse Events of Anti HIV drugs (D:A:D) Study Group. Factors associated with specific causes of death amongst HIV-positive individuals in the D:A:D Study. AIDS 2010; 24: 1537–1548. [DOI] [PubMed] [Google Scholar]
- 22.Kalra S, Agrawal N. Diabetes and HIV: current understanding and future perspectives. Curr Diab Rep 2013; 13: 419–427. [DOI] [PubMed] [Google Scholar]
- 23.Writing Group for the DCCT/EDIC Research Group. Association between 7 years of intensive treatment of type 1 diabetes and long-term mortality. JAMA 2015; 313: 45–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Liu L, Simon B, Shi J, Malhi AK, Eisen HJ. Impact of diabetes mellitus on risk of cardiovascular disease and all-cause mortality: evidence on health outcomes and antidiabetic treatment in United States adults. World J Diabetes 2016; 7: 449–461. [DOI] [PMC free article] [PubMed] [Google Scholar]