icon-    folder.gif   Conference Reports for NATAP  
 
  18th CROI
Conference on Retroviruses
and Opportunistic Infections
Boston, MA
February 27 - March 2, 2011
Back grey_arrow_rt.gif
 
 
 
Diabetes Mellitus in Treated HIV-Infected Patients: Incidence over Ten Years in 1,046 patients from the ANRS CO8 APROCO-COPILOTE Cohort
 
 
  Reported by Jules Levin
CROI 2011 March 2 Boston
 
Jacqueline Capeau 1, Vincent Bouteloup 2, Franois Raffi 3, Jean-Philippe Bastard 1, Christian Michelet 4, Elisabeth Bouvet 5, Bruno Spire 6, Christine Katlama 7, Catherine Leport 8, Genevive Chne 2, and ANRS CO8 APROCO-COPILOTE Cohort Study Group 1 UniversitŽ Paris 6, Inserm, APHP Tenon, Paris, France, 2 Inserm U897, Bordeaux, France, 3 CHU, Nantes, France, 4 CHU, Rennes, France, 5 PHP Bichat, Paris, France, 6 Inserm, Marseille, France, 7 UniversitŽ Paris 6, Inserm, APHP PitiŽ-SalpŽtrire, Paris France, and 8 Inserm, UniversitŽ Paris 7, Paris France
 
AUTHOR CONCLUSIONS
 
Diabetes incidence in long-term treated HIV-infected patients is markedly higher (14.1 per 1,000 PYFU) than that reported both in the European uninfected and HIV-infected populations.
 
Diabetes incidence peaked in 1999/2000 and decreased thereafter.
 
In addition to age and markers of adiposity, high frequency of HIV-related lipodystrophy and frequent and early exposure to indinavir, stavudine and/or didanosine played a role in the occurrence of diabetes.
 
Increased cardio-vascular risk indicates the importance of screening, close surveillance and targeted intervention on glucose parameters in long-term treated HIV-infected patients.
 
ABSTRACT
 
Background: Incidence of diabetes in HIV-infected patients increased early after first generation PIs initiation. However, long-term estimations of incidence are scarce. We investigated diabetes incidence and risk factors during the 10y follow-up of a homogeneous cohort of HIV-infected adults initiated with first generation PI-containing regimens in 1997-1999.
 
Methods: 1,046 patients with available measurements were included. Diabetes was diagnosed on fasting and/or post-glucose-charge glycemia or reporting of antidiabetic treatment initiation. Risk factors were assessed by a Cox model, including time-updated metabolic parameters and antiretroviral exposure.
 
Results: Baseline characteristics were: 78.5% males, median age 36.7y, median body mass index (BMI) 22.1 kg/m2, 20% stage C, median CD4 280 cells/mm3, 44% ART-naive. During follow-up, 54% were exposed to indinavir, 56% to nelfinavir, 75% to stavudine, 65% to zidovudine and 52% to didanosine. After 7,846 person years of follow-up (PYFU), 111 patients experienced new-onset diabetes, incidence: 14.1 per 1000 PYFU (95% Confidence Interval, 11.6 to 17.0), similar in men and women. Incidence peaked in 1999-2000 (23.2 per 1000 PYFU, p<10-4) and decreased onwards. It was not associated with parameters related to viral infection, immunity, HCV co-infection, hypertension or familial history of diabetes.
 
In the multivariate analysis, a higher incidence of diabetes was associated with: age >40y (Hazard Ratio [HR]: 2.13, p<10-4 ), BMI >25 kg/m2 (HR: 1.98, p=0.003), waist-to-hip ratio ≥0.97 male/0.92 female (HR: 3.95, p<10-4 ), lipoatrophy (HR: 1.92, p=0.007) and exposure to indinavir (HR: 1.92, p=0.001), stavudine (HR: 2.77, p<10-4 ) or didanosine (HR: 1.51, p<0.05). Patients who developed diabetes presented at baseline a 10% 10-year risk of cardio-vascular disease (Framingham score) vs 3% in normoglycemic patients.
 
Conclusions: In a cohort of more than 1,000 patients followed for 10y after PI- introduction, diabetes incidence was markedly higher than that reported both in the European uninfected and HIV-infected populations. In addition to age and markers of adiposity, high frequency of HIV-related lipodystrophy and frequent and early exposure to indinavir, stavudine and/or didanosine may have played a role in the occurrence of diabetes. Increased cardio-vascular risk indicates the importance of screening, close surveillance and targeted intervention on glucose parameters in treated HIV-infected patients
.
 
BACKGROUND
 
General European population : overall incidence of type 2 diabetes mellitus (DM) ranges from 2-4 per 1,000 person-years of follow-up (PYFU) in subjects with age comparable to that of HIV-infected patients
 
In HIV-infected patients, the Swiss HIV Cohort Study (SHCS) and the Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) Study reported a diabetes incidence not markedly different from the European general population of the same age and gender (respectively 4.4 and 5.7 per 1,000 PYFU )
 
Long-time exposure to anti-retrovirals and HIV-related lipodystrophy may be linked to DM occurrence, in addition to other conventional individual risk factors
 
Objectives
 
To evaluate the incidence of new-onset DM over the long-term in patients treated with first generation PIs and NRTIs
 
To identify associated determinants
 
To compare the characteristics of patients according to different status of glucose tolerance.
 
METHODS
 
PATIENTS AND EXAMINATIONS

 
Patients enrolled in the ANRS CO8 APROCO-COPILOTE cohort between 1997 and 1999 : all started a PI regimen while PI-na•ve, exclusion of prevalent cases and patients with < 2 glycemia measures
 
Lipodystrophy, fasting glucose, lipid parameters, a 2h-oral glucose tolerance test (2h-OGTT) at the month 12 or 20 visit and annually thereafter.
 
DIABETES DEFINITION
 
Fasting or post-charge glycemia ≥7.0/11.1 mmol/l on at least two occasions, or glycemia gradually increased over these values, and/or treatment
 
GLYCEMIA STATUS
 
Patients were categorized according to their glycemia during their follow-up · normoglycemia at all occasions : fasting and post-charge glycemias always <5.6 and <7.8 mmol/l,
· at least one fasting or post-charge glycemia in the range [5.6- 7] or [7.8-11.1] mmol/l,
· one fasting or post-charge glycemia in the diabetic range
· diabetic patients (as defined above)
 
STATISTICAL ANALYSIS
 
Incidence rates estimated per 1,000 PYFU and their 95% confidence interval ([95%CI] ), using exact Poisson method
 
Cox proportional hazards model included baseline characteristics and time-updated variables (variables with p<0.25 in univariable included in multivariate model)
 
To deal with high colinearity between adiposity markers, we only kept variables which minimized the Aka•ke criterion of the multivariate model.
 
Comparisons over subgroups through Fisher's exact, Wilocoxon rank-sum or Kruskal-Wallis tests (α=0.05)

table2.gif

fig1.gif