Influence of Hepatitis C Coinfection and Treatment on Risk of Diabetes Mellitus in HIV-Positive Persons

Abstract Background The role of hepatitis C virus (HCV) coinfection and HCV-RNA in the development of diabetes mellitus (DM) in HIV-positive persons remains unclear. Methods Poisson regression was used to compare incidence rates of DM (blood glucose >11.1 mmol/L, HbA1C >6.5% or >48 mmol/mol, starting antidiabetic medicine or physician reported date of DM onset) between current HIV/HCV groups (anti-HCV-negative, spontaneously cleared HCV, chronic untreated HCV, successfully treated HCV, HCV-RNA-positive after HCV treatment). Results A total of 16 099 persons were included; at baseline 10 091 (62.7%) were HCV-Ab-negative, 722 (4.5%) were spontaneous clearers, 3614 (22.4%) were chronically infected, 912 (5.7%) had been successfully treated, and 760 (4.7%) were HCV-RNA-positive after treatment. During 136 084 person-years of follow-up (PYFU; median [interquartile range], 6.9 [3.6–13.2]), 1108 (6.9%) developed DM (crude incidence rate, 8.1/1000 PYFU; 95% CI, 7.7–8.6). After adjustment, there was no difference between the 5 HCV strata in incidence of DM (global P = .33). Hypertension (22.2%; 95% CI, 17.5%–26.2%) and body mass index >25 (22.0%; 95% CI, 10.4%–29.7%) had the largest population-attributable fractions for DM. Conclusions HCV coinfection and HCV cure were not associated with DM in this large study. The biggest modifiable risk factors were hypertension and obesity, and continued efforts to manage such comorbidities should be prioritized.

Hepatitis C (HCV) monoinfection has been associated with an increased risk of a wide range of extrahepatic comorbidities [1], including cardiovascular disease and chronic kidney disease [2,3], and has also been associated with higher mortality from cardiovascular disease, cancer, and renal disease [4,5]. Most studies have reported an increased risk of diabetes mellitus (DM) in HCV-positive vs HCV-negative individuals, with a recent meta-analysis reporting a 1.6-fold increased odds [6]. However, a large French cohort study among HIV/ HCV-coinfected individuals found that DM was associated with cirrhosis but not HCV infection per se [7], while a large US study of HCV-monoinfected persons found that DM was associated with elevated Alanine Amino Transferase (ALT) and gamma glutamyl transpeptidase (GT), but not HCV itself [8]. The introduction of direct-acting antivirals (DAAs) for the treatment of HCV has had a major impact on HCV treatment [9], with cure rates in excess of 90% in persons coinfected with both HIV and HCV [10]. Data from the pre-DAA era of interferon-based HCV therapy have indicated that sustained virologic response (SVR) could also have indirect beneficial effects in terms of improvements in lifestyle factors, a so-called epiphany effect [11], which could also potentially impact the risk for diabetes. In studies of persons coinfected with HCV and HIV, persons with SVR had a significantly lower incidence of DM compared with those treated for HCV without SVR [12]. A lower incidence of DM among HCV treatment responders vs nonresponders was also found in a Spanish cohort of HIV/ HCV-coinfected persons [13]. Both previous studies were based on relatively few cases of DM and were conducted before the introduction of direct-acting antivirals, and therefore larger studies with substantial follow-up and well-defined end points are required to further understand the role of HCV on the development of DM in HIV-positive persons.
The aim of this study was therefore to investigate the incidence of DM in a large pan-European cohort study according to HCV status in HIV-coinfected persons across 5 strata: anti-HCV-negative individuals, spontaneous HCV-RNA clearers, those with chronic untreated HCV infection, those with cured HCV, or those who have been treated but remain HCV-RNA-positive.

The EuroSIDA Study
Persons were included from the EuroSIDA study, a large prospective observational cohort of almost 23 000 HIV-1-positive patients followed in 100 hospitals in 35 European countries plus Israel and Argentina. Individuals were enrolled into 10 cohorts from 1994 onward. In cohort 10, all HIV-positive patients were also required to be positive for anti-HCV antibodies (HCV-RNA-positive, -negative, or unknown status). At recruitment, in addition to demographic and clinical data, a complete antiretroviral therapy history was obtained, together with the most recent CD4 cell counts and HIV-RNA measurements, as well as all HCV tests, HCV-RNA, HCV genotype, hepatitis B surface antigen (HBsAg), and hepatitis B virus (HBV) DNA. Data are collected prospectively at clinical sites and sent to the coordinating center at yearly intervals. At each follow-up visit, all CD4 cell counts, HIV-RNA, HCV tests, HCV-RNA, genotype, and HBsAg results measured since last follow-up are collected, together with start and stop dates for antiretroviral drugs and HCV and HBV drugs. Detailed information about data collected in EuroSIDA can be found at http://www.chip. dk/Ongoing-Studies/EuroSIDA/About.

Patient Consent Statement
Informed patient consent was obtained according to local and/or national ethics committee requirements; consent was obtained from each particpant before any study-related procedure was performed and in accordance with the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) -Good Clinical Practice Guidelines. Further information is available at https:// www.chip.dk/Portals/0/files/Eurosida/EuroSIDA/EuroSIDA_ Protocol_v4_2019JULI05.pdf ?ver=2019-10-02-145631-730 (ClinicalTrials.gov Identifier: NCT02699736).

Definitions
DM was defined according to laboratory values (blood glucose levels >11.1 mmol/L or HbA1C >6.5%/48 mmol/L) and/or use of antidiabetic medication. We used a blood glucose >11.1 mmol/L to be conservative, reflecting missing information on whether glucose was measured in a fasting state or not. According to the EuroSIDA Manual of Operations, DM can be defined by sending the laboratory data, by the use of antidiabetic medication, or the site can report the clinical diagnosis based on these criteria without sending laboratory data (https://www.chip.dk/Portals/0/files/RESPOND/ RESPOND%20Manual%20of%20Operations%20MOOP__ Version%201.6.pdf?ver=2019-11-05-124535-643). Baseline was defined as the earliest date after cohort enrollment or January 1, 2001 (when collection of prospective information on DM began), with known HCV serostatus and, for those who were anti-HCV-positive, known HCV-RNA status. Persons aged <18 years at baseline or without a CD4 count and HIV viral load in the 12 months before or 1 month after baseline were excluded, as were persons with DM before baseline.
These 5 strata were defined on the basis of latest anti-HCV test and HCV-RNA measurement. Those treated and HCV-RNA-positive included persons who did not achieve SVR, persons who were HCV-RNA-positive who had started treatment more recently, those reinfected with HCV, and persons without an HCV-RNA measurement after treatment completion and lacking an end-of-treatment response. Persons were followed until their last visit (median June 2018), date of death, or DM, whichever occurred first. Person-years of follow-up (PYFU) and DM events accrued according to current HCV stratum using the last observation carried forward, and persons could contribute PYFU to >1 stratum.

Statistical Analysis
Characteristics of patients were compared across strata using chi-square tests for categorical variables and the Kruskal-Wallis test for continuous data. Incidence rates of DM per 1000 PYFU were calculated within HCV groups, and Poisson regression was used to compare these rates with persons cured as the reference group. In addition to univariate analyses, we adjusted for all factors fixed at baseline (except HCV status) and all factors updated. Factors included were gender, HIV exposure group, region of Europe (North, Central West, South, Central East, East, and Argentina) [15], development of chronic kidney disease [14], HIV viral load, prior AIDS, cumulative exposure to zidovudine, stavudine, and didanosine [16], cardiovascular disease, non-AIDS-defining malignancies (NADMs), end-stage liver disease (ESLD; ascites, hepatorenal syndrome, grade III/ IV hepatic encephalopathy, unspecified liver decompensation, esophageal variceal bleeding, spontaneous bacterial peritonitis, liver transplantation, and hepatocellular carcinoma; further information about these events is available at https:// www.chip.dk/Studies/EuroSIDA/Study-documents), smoking status (never smoked, current smoker, past smoker, unknown smoking status), hypertension [17], body mass index, CD4, nadir CD4, age, and baseline date. Liver fibrosis was defined according to our previous study [18]. In brief, data were available on all liver biopsy and Fibroscan test results performed at participating centers. Information on aspartate transaminase (AST) and platelet counts was used to calculate the AST-toplatelet ratio (APRI). Hyaluronic acid was available for a small subset. The most recent fibrosis marker measured (before baseline or time-updated) was used to define fibrosis stage, and where >1 marker was measured, priority was given to biopsy, Fibroscan, APRI, and hyaluronic acid. A priori, we investigated the interaction between age and HCV strata.
We performed a wide range of sensitivity analyses to investigate the robustness of our results to different assumptions. These included carrying the last HCV-RNA measurement forward for a maximum of 12 months, excluding persons with stage F3/F4 liver fibrosis at baseline, excluding men who have sex with men (MSM) with a more recent HCV infection, and an analysis limited to post-2014, when DAAs became more widely available for persons included in the EuroSIDA study [19]. In addition, we investigated excluding centers in EuroSIDA where the 95% CI for the center did not include the population point estimate, as incidence of DM could depend on frequency of laboratory testing within centers.
To address which risk factors have the strongest impact on the development of DM, we calculated the population-attributable risk fraction (PAF) for the key identified risk factors. PAF expresses the proportion of events that could have been avoided had that risk factor not been present and considers both the strengths of the associations and the incidence rate of the risk factor.
All analyses were performed in SAS, version 9.4 (SAS Institute, Cary, NC, USA).

RESULTS
Of 22 828 persons enrolled in the EuroSIDA study, 3948 were excluded as their HCV status and/or HCV-RNA status was unknown, 754 were excluded with DM before the study baseline, 1439 were excluded with no prospective follow-up, and 588 were excluded with missing CD4 and/or HIV viral load at baseline. Compared with the 16 099 included in these analyses, the 2029 excluded due to no follow-up or missing CD4/viral loads were more likely to be females, be younger, have a prior AIDS diagnosis, have fibrosis stage F3/F4 at baseline, be from Central Eastern Europe, have an earlier baseline date, and be HBsAg-positive.
Characteristics of the 16 099 included individuals are shown in Table 1, stratified by HCV strata at baseline. The majority of those included were in group 1 at baseline (HCV-Ab negative; 10 091; 62.7%), followed by those in group 3 (chronically infected; n = 3614; 22.4%), with smaller numbers in group 2 (spontaneous clearers; n = 722; 4.5%), group 4 (successfully treated; n = 912; 5.7%), and group 5 (treated, HCV-RNA positive; n = 760; 4.7%). The HCV strata were heterogeneous at baseline, with a lower burden of fibrosis and of comorbidities in group 1 (HCV-Ab negative) compared with other groups. The median follow-up (IQR) was 6.9 (3.6-13.2) years per person. During 136 084 PYFU, 1108 developed DM, for an incidence rate of 8.14/1000 PYFU (95% CI, 7.66-8.62). Of 1108 persons with DM, 354 cases were defined using HbA1c, 628 using glucose levels, 524 with a clinical diagnosis, and 521 based on use of antidiabetic medication. These figures include 225, 176, and 114 persons who were diagnosed with DM using 2, 3, or 4 methods, respectively. The crude incidence of DM stratified by HCV strata is shown in Table 2. The crude incidence of DM was highest in persons who are HCV-Ab negative (8.66/1000 PYFU; 95% CI, 8.08-9.25) and lowest in spontaneous clearers (5.62/1000 PYFU; 95% CI, 3.61-7.63). The incidence of DM in individuals who were chronically infected, successfully treated, and treated and HCV-RNA-positive was quite similar, at around 7/1000 PYFU. In univariate analyses, there was no evidence of a difference in incidence of DM across the HCV strata compared with those who were successfully treated.

Association Between HCV Strata and Development of DM
After adjustment, there was no difference between the HCV strata and incidence of DM, as shown in Figure 1. In a model that allowed cofactors to be updated over time (Figure 1), those who were HCV-Ab-negative (adjusted incidence rate ratio [aIRR], 1.21; 95% CI, 0.88-1.66) and chronically infected (aIRR, 1.29; 95% CI, 0.94-1.77) had the highest incidence rates of DM compared with persons who had been cured, although the increased incidence was not statistically significant, and there was no evidence of an overall difference between strata (global P = .33). Highly consistent results were seen in a model where all cofactors except HCV strata were fixed at baseline, with no differences between HCV strata (global P = .68). Other factors associated with developing DM are summarized in Figure 1 and were consistent in both multivariate models.     The strongest factors associated with the development of DM were increasing age, development of cardiovascular disease (CVD), NADM, ESLD, hypertension, a high body mass index (BMI), and stage 4 liver fibrosis. There was a nonsignificantly increased incidence of DM associated with stage 2 (aIRR, 1.24; 95% CI, 0.86-1.78) or stage 3 liver fibrosis (aIRR, 1.32; 95% CI, 0.85-2.06) compared with stage 0/1 after adjustment. Our results were consistent across all our sensitivity analyses, including carrying the last HCV-RNA measurement forward for a maximum of 12 months, adjusting for starting integrase inhibitors, excluding MSM from analyses with more recent HCV infection, excluding persons with stage F3/F4 liver fibrosis at baseline, an analysis limited to post-2014, and excluding centers in EuroSIDA where the 95% CI for the center did not include the population point estimate (data not shown). The association between HCV strata and DM differed for those aged ≤40 and those aged >40 (P interaction = .0060). This is summarized in Figure 2. In persons aged ≤40 years, those who were chronically infected had a similar incidence of DM compared with those who had been cured (aIRR, 0.97; 95% CI, 0.59-1.60), while in persons aged >40, there was a significantly increased incidence of DM in those who were chronically infected compared with individuals who had been cured (aIRR, 1.51; 95% CI, 1.01-2.26).

Population-Attributable Fractions for Modifiable and Nonmodifiable Risk Factors Associated With DM
The PAFs for key factors associated with DM are shown in Figure 3. The largest PAF was associated with development of hypertension, where an estimated 22.2% (95% CI, 17.5%-26.2%) of DM diagnoses could have been avoided if persons were not hypertensive. A BMI of >25 contributed 22.0% of DM cases (95% CI, 10.4%-29.7%). Comorbidities, including AIDS, ESLD, NADM, CVD, and stage F4 liver fibrosis, all contributed small but significant PAFs of ~2%-5%. Aging, a nonmodifiable risk factor, also contributed large PAFs, particularly among those aged 40-60 years. PAFs for HCV strata were not calculated, as there was no evidence of an association between HCV strata and development of DM.
Restricting the PAF analysis to those with chronic HCV infection and those treated for HCV (groups 3-5), the PAF for cirrhosis was 10.1% (95% CI, 6.2%-12.8%). The PAFs for the other factors were similar to the overall analysis except for a decreased contribution due to age >50 years (Supplementary Figure 1).

DISCUSSION
This large cohort study is one of the largest to date of HIV/ HCV-coinfected persons including >16 000 persons and 136 000 PYFU. We found no difference in the rates of DM

Non-AIDS malignancies a (yes vs no)
End-stage liver disease a (yes vs no)

Body mass index a (>25 vs ≤18)
Adjusted incidence rate ratio of DM

Figure 1.
Factors associated with incidence of diabetes mellitus. Adjusted for factors shown and gender, HIV transmission risk group, nadir CD4, baseline date, glucose and HbA1C levels (as fixed values at baseline), HBsAg, cumulative exposure to zidovudine, stavudine, and didanosine, HIV viral load, CD4, smoking, chronic kidney disease as time-updated. a Included as time-updated variables. DM was defined as either blood glucose >11.1 mmol/L, HbA1C >6.5%, or >48 mmol/mol, starting antidiabetic medicine, or physician-reported date of DM onset. Abbreviations: DM, diabetes mellitus; HCV, hepatitis C virus. Aged ≤ 40 Aged > 40

Figure 2.
Multivariate incidence rate ratios of diabetes mellitus: Stratification by age. Adjusted for factors shown and gender, ethnic origin, HIV transmission risk group, region, nadir CD4, baseline date, glucose and HbA1C levels (as fixed values at baseline), HBsAg, cumulative exposure to zidovudine, stavudine, and didanosine, HIV viral load, CD4, smoking, AIDS, cardiovascular disease, non-AIDS-defining malignancies, end-stage liver disease, hypertension, liver fibrosis, chronic kidney disease, and body mass index as time-updated. DM was defined as either blood glucose >11.1 mmol/L, HbA1C >6.5%, or >48 mmol/mol, starting antidiabetic medicine, or physician-reported date of DM onset. Abbreviations: DM, diabetes mellitus; HCV, hepatitis C virus.
between well-defined HCV strata including chronic and cured HCV. Traditional DM risk factors, such as age, hypertension, and obesity were much stronger predictive factors of DM than HCV status and whether someone with chronic HCV infection had been successfully treated.
We found no overall differences in the incidence of DM regarding the presence of replicating HCV infection across the HCV strata we included, and this was consistent across the different sensitivity analyses. Data from a smaller cohort study including 95 DM cases in persons with and without SVR from Berenguer et al. found a ~50% reduction in DM in those with SVR compared with those without SVR after interferon (IFN) therapy [13]. Data from the Swiss HIV Cohort Study including ~200 patients with DM found no increased incidence of DM comparing HCV-Ab-positive with HCV-Ab-negative individuals, while a comparison of HCV-Ab-positive individuals with and without SVR after IFN therapy suggested a lower incidence of DM in those with SVR, albeit with very wide confidence intervals [12]. Possible reasons for the discrepancies include differences in the populations included, that DM may develop over months or years with active HCV replication, and that a much longer follow-up period is required to observe differences, as well as differences in a study's ability to adjust for important risk factors for DM, such as body mass index and hypertension, as we were able to in this analysis.
In a preplanned analysis, we found that the association between HCV strata and DM differed in those aged ≤40 compared with those aged >40 years. In persons aged >40 years, where the greatest burden from DM typically occurs, those with chronic untreated HCV had a significantly higher incidence of DM after adjustment compared with those who had been treated, regardless of whether the individual had cleared HCV-RNA or remained HCV-RNA-positive. Modifiable risk factors for DM, such as increasing weight and hypertension, should be closely monitored as persons age, and interventions should be targeted at those at greatest risk, such as those aged >40 years. The higher rate of DM in chronically infected persons aged >40 years may reflect a longer duration of HCV infection in older persons, which we were unable to adjust for. Our findings of lower rates of DM in individuals treated for HCV aged ≥40 years, regardless of response to treatment, could partly be explained by confounding by indication or more speculatively by improvement in lifestyle factors among those treated, described as either the "epiphany" effect [11] or the "Hawthorne" effect [20]. Further, DAA treatment in EuroSIDA began to increase most notably around 2015 [19]; before this, it is likely that the healthiest persons with lower rates of advanced fibrosis/cirrhosis and those most likely to respond to treatment were selected for interferon and ribavirin treatment with its known toxicities and limited treatment responses [21]. This channeling bias might partly explain lower rates in those treated with pegylated interferon/ ribavirin than those with chronic untreated infection. Further studies focused on persons exposed to DAAs are needed to explore this association further.
Aging, hypertension, and obesity were all associated with the development of DM in our study, as previously reported and summarized by others [16,[22][23][24]. Smoking was not associated with development of DM, as previously reported by the D:A:D study [16]. We also found that HIV-associated comorbidities, including AIDS, non-AIDS-defining malignancies, and cardiovascular disease, were all associated with an increased incidence of DM. The association between incidence of DM and both malignancies and cardiovascular disease is likely to be explained by sharing risk factors. We have adjusted for some, such as BMI, hypertension, and aging, but there are likely to be other risk factors in common, such as physical inactivity and poor diet, that we were not able to adjust for [25]. While each of these HIVassociated comorbidities was significantly associated with DM, the PAFs were small, indicating that although they were risk factors, their underlying prevalence was low.
Whereas we did not find an association between incidence of DM and HCV infection per se, both liver cirrhosis and ESLD were associated with increased incidence of DM. These findings indicate an indirect role of HCV in increasing the risk of DM and are in agreement with findings from another recent large French cohort study of HIV/HCV-coinfected persons [7]. The liver plays a central role in the homeostasis of blood glucose levels through glycogenolysis and gluconeogenesis, although an impairment severe enough to translate into diabetes could probably only be seen in advanced or late-stage HCV-induced liver disease. The exact mechanisms through which liver cirrhosis can lead to DM are not well defined, but there is growing evidence that both insulin resistance and β-cell dysfunction contribute [26]. Even with a positive association with liver fibrosis and ESLD, the PAF was small due to the comparatively low prevalence of F4 fibrosis and ESLD among our population. In an analysis restricted to those with either chronic HCV infection or treated HCV infection, the PAF due to cirrhosis was 10%, with lifestyle factors such as increased BMI and hypertension being more important. Despite the comparatively low PAF due to cirrhosis, it remains important to diagnose and treat HCV early to reduce the risk of DM and other complications, as well as to achieve the global elimination of HCV [27]. Among the modifiable risk factors, hypertension and obesity were associated with the largest PAFs in our population. Traditional interventions used to reduce such risk factors in the general population may need modification to improve outcomes in adults with HIV in order to have a substantial impact [28]. Our results do, however, highlight the role of traditional risk factors for DM and the importance of screening and managing such risk factors.
The limitations of this study should be noted. We are not able to distinguish between type 1 and type 2 DM and used a variety of measures to define DM, reflecting heterogeneity in data collected across Europe. We focused our analyses on DM and have no data on insulin resistance, which may be associated with HCV and precede DM [29], possibly due to impaired pancreatic beta cells in HCV-positive individuals [30]. We have no information on duration of HCV infection and used the last HCV-RNA carried forward in our analysis. Where this was negative after HCV treatment we assumed SVR, consistent with our previous work [14]. As with many observational studies, our inclusion and exclusion criteria for these analyses excluded a significant proportion of persons, and there was variability in completeness of data. Using time-updated analyses and last observation carried forward includes data measured after baseline; for example, the proportion with missing data on fibrosis stage declined to 12.7% over time, of which the majority of missing data (88.8%) was among those who were anti-HCV-negative. Despite these limitations, our results were consistent across all our sensitivity analyses. We were not able to specifically address the role of DAAs in our HCV-treated groups due to limited follow-up after DAA initiation. The strength of our study is that it is one of the largest of coinfected persons reported to date, with an extensive quality assurance and data monitoring program.
In conclusion, HCV coinfection and cure of chronic HCV, mainly with DAAs, were not associated with the development of DM in this large cohort study, and there were few differences in DM across the 5 well-defined HCV strata included in this large study. The biggest modifiable risk factors of DM were hypertension and obesity, and continued efforts to manage such comorbidities should be prioritized.