Diabetes mellitus (DM) has been of increased concern in HIV-infected individuals since an association between protease inhibitors (PI) and hyperglycemia, impaired glucose tolerance, and insulin resistance was first reported [1–4]. Recent studies [5,6] suggesting an increased risk of premature cardiovascular disease in HIV-infected individuals have heightened the need to understand the relation of HIV infection with risk of DM, a primary risk factor for cardiovascular disease. Prior publications on the risk of DM to date have been limited by small numbers of incident DM cases , reliance on self-report of incident DM , or lack of an HIV-negative comparison group .
This study uses a large well-characterized prospective cohort of HIV-infected and HIV-uninfected women followed over 5½ years (October 2000 to March 2006) to examine the association of both the type and duration of antiretroviral therapy (ART) exposure on DM incidence determined, in part, using fasting glucose measurements.
The Women's Interagency HIV Study (WIHS) is a multicenter prospective cohort study that was established in 1994 to investigate the progression of HIV in women with and at risk for HIV. A total of 3766 women (2791 HIV-infected and 975 HIV uninfected) were enrolled in either 1994–5 (n = 2623) or 2001–2 (n = 1143) from six United States cities (Bronx, Brooklyn, Chicago, Los Angeles, San Francisco, and Washington DC). Baseline sociodemographic characteristics and HIV risk factors were similar between HIV-infected and HIV-uninfected women [10,11]. Participating institutions institutional review boards approved study protocols and consent forms, and each study participant gave written informed consent.
Every 6 months, participants complete a comprehensive physical examination, provide biological specimens for CD4 cell count and HIV RNA viral load determination, and complete an interviewer-administered questionnaire, which collected information on demographics, disease characteristics, and specific ART use.
Beginning in October 2000, fasting glucose was measured after participants had fasted for at least 8 h. Of the 2859 women with a study visit between October 2000 and March 2006, 2554 had at least one fasting glucose measurement; the first study visit with fasting glucose data available will be referred to as the index study visit. Of the 2554 women with at least one fasting glucose measurement, 67 were excluded because of either a positive (n = 60) or a missing (n = 7) report of pregnancy at the index visit. Of the remaining 2487 women, 280 (202 of the 1808 HIV-infected and 78 of the 679 HIV-uninfected women) were excluded owing to prevalent DM [fasting glucose ≥ 1.26 mg/l at index (n = 119), a history of self-report DM prior to or at index (n = 235), or antidiabetic medication use reported prior to or at index (n = 126)]. Of the remaining 2207 women, 110 had no follow-up visits with data to determine incident DM after the index visit, resulting in a total of 2097 women. Nine women were excluded because they were HIV uninfected at study entry and seroconverted during follow-up. The final study population consisted of 2088 women (1524 HIV-infected and 564 HIV-uninfected women). The 2088 women contributed a total of 15 156 study visits of follow-up (including the index visit). Fasting glucose data were available at 9537 (63%) of these study visits. The median number of study visits with glucose data for the 2088 women was 4 [interquartile range (IQR), 3–6].
Ascertainment of diabetes mellitus
Incident DM was considered to have occurred at the first follow-up visit after the index visit at which fasting glucose was ≥ 1.26 mg/l, antidiabetic medication use was reported, or DM was reported with subsequent confirmation by either a fasting glucose ≥ 1.26 mg/l or report of antidiabetic medication use.
Assessment of antiretroviral therapy use
At each semiannual study visit, participants were shown photo-medication cards and were asked the names of specific ART medications used since their prior study visit. The WIHS uses a standard definition of HAART  adapted from the Department of Health and Human Services/Kaiser Panel guidelines . All non-HAART combination therapy regimens were classified as combination therapy; reports of a single nucleoside reverse transcriptase inhibitor (NRTI), PI, or nonnucleoside reverse transcriptase inhibitor (NNRTI) were classified as monotherapy. In analyses, ART regimens were classified as no therapy, monotherapy or combination therapy, PI-based HAART, and non-PI-based HAART.
At each study visit from index through the end of follow-up, the cumulative drug-years of exposure to NRTI, PI, and NNRTI were determined. The time of exposure (in years) to each class of ART (i.e., NRTI, PI, NNRTI) at each study visit was defined as the product of the number of drugs which were reported in a given class since the last study visit and 0.5, since semiannual visits occurred approximately half a year apart. Drugs within a class were considered exchangeable and additive. Finally, the cumulative drug-years of exposure to the four most commonly used NRTI (i.e., zidovudine, abacavir, stavudine, and lamivudine) were determined in our cohort from index to the end of follow-up. For analyses, categories of cumulative drug exposure were defined as no exposure (reference category), exposed and less than or equal to the median, and exposed and greater than the median. In adjusted cumulative drug-years analyses, adjustment was also made for cumulative amount of ART use reported prior to the index visit.
Comparisons of continuous and categorical characteristics among HIV-infected and HIV-uninfected women at the index visit were made by Wilcoxon rank sum test or Pearson χ2 test, respectively. Time at risk for incident DM was calculated from the date of the index visit through either the date of DM (for those with incident DM) or the date of the last study visit (for those without incident DM). The crude incidence rate of DM for each ART exposure group was calculated by dividing the number of incident DM cases by the total person-time at risk for DM.
Cox proportional hazards models  were approximated using pooled logistic regression models . Relative hazards (RH) were used as a measure of association and 95% confidence intervals (CI) as a measure of precision. Regression models were used to adjust for the potentially confounding effects of variables measured at the index visit. Specifically, adjustment was made for age, race, body mass index, smoking status (current versus noncurrent), hepatitis C virus antibody status, family history of DM, menopause, CD4 cell count (set to zero for HIV-uninfected women so that the association between CD4 and incident DM was only assessed in HIV-infected women), history of ART use prior to the index visit, and enrollment cohort (1994–5 versus 2001–2). To explore changes in body size as a possible pathway through which ART may cause alterations in the rates of incident DM, separate analyses further adjusted for time-varying change in hip size.
The proportional hazards assumption was explored by estimating the interaction between ART exposure and time. There was no strong evidence of nonproportional hazards among the different exposure groups (P = 0.652); however, owing to the relatively few events, there was only modest statistical power to detect nonproportional hazards. All analyses were conducted using SAS version 9 (SAS Institute, Cary, North Carolina, USA).
At the index visit, HIV-infected women were older, more likely to be postmenopausal, and more likely to be positive for hepatitis C virus antibodies than HIV-uninfected women (Table 1). The racial distribution and the proportion with a family history of DM were similar. Body mass index and hip circumference were lower in HIV-infected women, and waist circumference was similar.
The 2088 women at risk for DM contributed a total of 6802 person-years (4962 HIV-infected, 1840-uninfected) of follow-up from index until either incident DM or the last study visit. The median fasting glucose at index was 830 mg/l (IQR, 780–910) and was the same for HIV-infected and HIV-uninfected women. Over the follow-up period, 152 (116 HIV-infected, 36 uninfected) women developed DM, 121 died (114 HIV-infected, 7 uninfected) without developing DM, and 105 (81 HIV-infected, 24 uninfected) were lost to follow up prior to April 2005 free of DM. The remaining 1710 (1213 HIV-infected, 497 uninfected) women completed follow-up between April 2005 and March 2006 without DM.
Of the 152 incident DM cases, 99 (72 HIV-infected and 27 uninfected) were defined as having a fasting glucose concentration ≥ 1.26 mg/l; 46 (39 HIV-infected and 7 uninfected) were based on reported antidiabetic medication use, and 7 (5 HIV-infected and 2 uninfected) were based on reported DM diagnosis subsequently confirmed by either a fasting glucose concentration ≥ 1.26 mg/l or reported antidiabetic medication use. Of the 46 classified by report of antidiabetic medications, 29 (26 HIV-infected and 3 HIV-uninfected) women reported antidiabetic medication specifically to lower blood sugar. For the remaining 17 (13 HIV-infected and 4 HIV-uninfected) women, it was not possible to determine whether the antidiabetic medications were used to treat DM or for other reasons such as to treat pre-DM or fat distribution changes associated with HIV.
Among the 116 HIV-infected women with incident DM: 25 reported using no ART immediately prior to the incident visit; 9 reported using ART (but not HAART) immediately prior to the incident visit; 41 reported a PI-containing HAART regimen; and 41 reported a non-PI containing HAART regimen. Of the nine women on a non-HAART regimen, three reported NRTI use only; four reported a single NRTI with at least one PI; one reported one NNRTI; and one reported an NNRTI and PI combination. Of the 41 women reporting a PI-containing HAART regimen, the PI used included nelfinavir (16; one also with ritonavir), lopinavir (14), indinavir (6; two also with ritonavir). Of the 41 women reporting a non-PI containing HAART regimen, 14 reported nevirapine, 13 reported efavirenz, and 14 reported either abacavir or tenofovir as a third NRTI without an NNRTI.
Compared with the DM incidence rate in HIV-uninfected women (1.96/100 person-years): HIV-infected women reporting no recent ART had a DM incidence rate of 1.53/100 person-years; those reporting monotherapy or combination therapy at the visit prior to the incident event had a DM incidence rate of 3.40/100 person-years; those reporting PI-containing HAART had a rate of 2.50/100 person-years; and those reporting non-PI containing HAART a rate of 2.89/100 person-years. In adjusted analyses (Table 2), none of the associations between treatment group and DM incidence was substantial nor were any associations precise enough to rule out chance.
Among HIV-infected women, after adjustment for potentially confounding factors measured at index (Table 3), longer cumulative exposure to NRTI was associated with an increased DM incidence compared with no NRTI exposure [RH, 1.81 (95% CI, 0.83–3.93) for 0 to 3 years NRTI exposure; RH, 2.64 (95% CI, 1.11–6.32) for > 3 years NRTI exposure]. Neither cumulative exposure to PI nor cumulative exposure to NNRTI was associated with DM incidence in adjusted analyses. Contrary to our mediation hypothesis, further adjustment for time-varying changes in hip size did not decrease the association between cumulative exposure to NRTI and incident DM (comparison of > 3 drug-years versus 0 drug-years: adjusted RH, 3.25; 95% CI, 1.16–9.11).
Because NRTI appeared most associated with increased incident DM, the rate of DM among HIV-infected women reporting the four most commonly used NRTI (zidovudine, abacavir, stavudine, and lamivudine) was also examined. No association between cumulative exposure to zidovudine, abacavir or stavudine and incident DM was observed in unadjusted or adjusted analyses (Table 4). However, cumulative exposure of > 1 year to lamivudine was associated with a nearly three-fold increase in the rate of DM incidence after adjustment for covariates including cumulative use of PI and NNRTI.
In the largest prospective study to date among HIV-infected women of the risk of DM using fasting glucose measurements, we found that longer cumulative exposure to NRTI was associated with increased DM incidence compared with no NRTI exposure. No association between cumulative exposure to PI or NNRTI and incident DM was observed.
Our findings of an association between longer cumulative NRTI exposure and DM incidence in HIV-infected women are consistent with those from a large study of HIV-infected men , where an association between cumulative NRTI exposure and insulin resistance was observed. In that study, the association between NRTI and incident DM was not studied owing to the small number of incident DM events. Recent findings of another large collaborative cohort study, which showed an increased risk of DM per year of NRTI use, also agree with our findings . When the individual NRTI drugs were studied in these cohorts, however, a variety of NRTI were implicated. In one study, stavudine and lamivudine were independently associated with insulin resistance . In another study, stavudine, zidovudine, and didanosine were associated with DM . Among the NRTI, we found an association between increased DM incidence and the cumulative exposure to lamivudine.
The mechanism by which NRTI may cause disorders in glucose metabolism is unclear. Some have postulated a role of NRTI-induced mitochondrial dysfunction and insulin resistance . NRTI drugs (especially stavudine) have been implicated as a cause of lipoatrophy [18,19], and lipoatrophy has been associated with insulin resistance . However, we observed that the association between NRTI and incident DM was not attenuated by further adjustment for changes in hip circumference (which we have previously demonstrated to be the most affected body site in this cohort ).
We did not find a notable association between PI use and incident DM despite prior studies reporting an association between this group and disorders of glucose metabolism. This may be partially explained by the changing patterns in PI use. Indinavir, commonly used early in the HAART era, has been associated with hyperglycemia and decreased insulin sensitivity in HIV-infected individuals [3,22] and in healthy HIV-uninfected volunteers [23,24]. An early prospective study of self-reported DM incidence from the WIHS cohort (median follow-up of 2.9 years) demonstrated a three-fold rate of DM incidence among HIV-infected women reporting PI use compared with HIV-uninfected women . Indinavir was the most frequently used PI during that time period (1994–8).
A recent study in HIV-negative volunteers given a single dose of lopinavir/ritonavir, a commonly used PI in our study population, demonstrated an acute decrease in insulin sensitivity  but found no change in insulin sensitivity from baseline after 4 weeks on therapy . In contrast, insulin sensitivity did not normalize after 4 weeks of indinavir use [23,24]. Therefore, the lack of an association between PI use and DM in our study may reflect the wider use of lopinavir/ritonavir, which may have less of an impact on insulin sensitivity than indinavir.
Prior observations indicated that recent PI exposure was associated with disorders in glucose metabolism; consequently, we initially examined the risk of recent exposure to different types of ART regimen on DM incidence. We found that any recent use of ART, regardless of being monotherapy, combination therapy, or PI-containing or non-PI containing HAART regimen, appeared to be associated with increased DM incidence when compared with HIV-uninfected women, although the associations were not statistically significant. The magnitude of the association was not as large as previously reported. One prospective study of men over a median follow up of 2.3 years from 1999 to 2003 found a four-fold increase in DM incidence (24 cases in 506 person-years) among those reporting HAART compared with 361 HIV-uninfected men (10 cases in 709 person-years) . Unlike the present study, that study did not adjust for key risk factors associated with DM, including family history of DM and hepatitis C infection, which have been shown to be independently associated with DM in several large epidemiological studies [27,28].
Among the limitations of our study is the definition of DM using a single fasting glucose measurement rather than confirmation on a subsequent day as recommended by the American Diabetes Association. In some cases, we also were not able to distinguish whether antidiabetic medications were used to treat DM or for other reasons such as to treat pre-DM or fat distribution changes associated with HIV. While we had many more incident DM cases than prior studies, the limited number of DM cases restricted our ability to explore possible synergistic effects of individual ART or interactions across ART classes. As with all observational studies, our findings are subject to possible unmeasured confounding. Finally, the design of this cohort study, which examines participants every 6 months, only allows us to coarsely definite cumulative exposure to specific ART, and, therefore, our results with respect to the association between specific ART drugs and DM should be interpreted with caution.
In summary, we conclude that cumulative exposure to NRTI, but not PI or NNRTI, was associated with increased DM incidence, using the definition stated in our methods. NRTI remain the backbone of effective ART, and so regular monitoring of fasting glucose levels in HIV-infected patients is warranted. Study of the biological mechanisms by which NRTI might induce disorders in glucose metabolism is a priority.
Data in this manuscript were collected by the WIHS Collaborative Study Group with centers (principal investigators) at New York City/Bronx Consortium (Kathryn Anastos); Brooklyn, New York (Howard Minkoff); Washington DC Metropolitan Consortium (Mary Young); the Connie Wofsy Study Consortium of Northern California (Ruth Greenblatt); Los Angeles County/Southern California Consortium (Alexandra Levine); Chicago Consortium (Mardge Cohen); Data Coordinating Center (Stephen J. Gange).
Sponsorship: The WIHS is funded by the National Institute of Allergy and Infectious Diseases, with additional supplemental funding from the National Cancer Institute, the National Institute of Child Health and Human Development, the National Institute on Drug Abuse, the Agency for Healthcare Policy and Research, the National Center for Research Resources, and the Centers for Disease Control and Prevention. U01-AI-35004, U01-AI-31834, U01-AI-34994, U01-AI-34994, U01-AI-34989, U01-HD-32632 (NICHD), U01-AI-34993, U01-AI-42590, M01-RR00079, and M01-RR00083. Dr Tien is supported by the National Institute of Allergy and Infectious Diseases through K23 AI 66943-01 and currently has a research grant from Gilead. The funding sources had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation and review of the manuscript. Dr Tien had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
1. Carr A, Samaras K, Burton S, Law M, Freund J, Chisholm DJ, et al
. A syndrome of peripheral lipodystrophy, hyperlipidaemia and insulin resistance in patients receiving HIV protease inhibitors. AIDS 1998; 12:F51–F58.
2. Behrens G, Dejam A, Schmidt H, Balks H, Brabant G, Körner T, et al
. Impaired glucose tolerance, beta cell function and lipid metabolism in HIV patients under treatment with protease inhibitors. AIDS 1999; 13:F63–F70.
3. Dube MP, Johnson DL, Currier JS, Leedom JM. Protease inhibitor-associated hyperglycaemia. Lancet 1997; 350:713–714.
4. Walli R, Goebel FD, Demant T. Impaired glucose tolerance and protease inhibitors. Ann Intern Med 1998; 129:837–838.
5. Mary-Krause M, Cotte L, Simon A, Partisani M, Costagliola D. Increased risk of myocardial infarction with duration of protease inhibitor therapy in HIV-infected men. AIDS 2003; 17:2479–2486.
6. Friis-Moller N, Sabin CA, Weber R, d'Arminio Monforte A, El-Sadr WM, Reiss P, et al
. Combination antiretroviral therapy and the risk of myocardial infarction. N Engl J Med 2003; 349:1993–2003.
7. Brown TT, Cole SR, Li X, Kingsley LA, Palella FJ, Riddler SA, et al
. Antiretroviral therapy and the prevalence and incidence of diabetes mellitus in the Multicenter AIDS Cohort Study. Arch Intern Med 2005; 165:1179–1184.
8. Justman JE, Benning L, Danoff A, Minkoff H, Levine A, Greenblatt RM, et al
. Protease inhibitor use and the incidence of diabetes mellitus in a large cohort of HIV-infected women. J Acquir Immune Defic Syndr 2003; 32:298–302.
9. de Wit S, Sabin C, Weber R, Worm SW, Reiss P, Thiebaut R, et al.
for the DAD Study Group. d4T and lipodystrophy raise diabetes risk in DAD cohort; nevirapine protects
. 8th International Congress on Drug Therapy in HIV infection
. November Glasgow, 2006.
10. Bacon MC, von Wyl V, Alden C, Sharp E, Robison N, Hessol S, et al
. The Women's Interagency HIV Study: an observational cohort brings clinical sciences to the bench. Clin Diagn Lab Immunol 2005; 12:1013–1019.
11. Barkan SE, Melnick SL, Preston-Martin S, Weber K, Kalish LA, Miotti P, et al
. The Women's Interagency HIV Study. WIHS Collaborative Study Group. Epidemiology 1998; 9:117–125.
12. Cole SR, Hernan MA, Robins JM, Anastos K, Chmiel J, Detels R, et al
. Effect of highly active antiretroviral therapy on time to acquired immunodeficiency syndrome or death using marginal structural models. Am J Epidemiol 2003; 158:687–694.
13. Panel on Clinic Practices for Treatment of HIV infection from the US Department of Health and Human Services and Henry J. Kaiser Family Foundation. Guidelines for the Use of Antiretroviral Agents in HIV-1 Infected Adults and Adolescents
. Bethesda, MD: National Institute of Health; 1998.
14. Cox DH, Hinkley DV. Theoretical Statistics
. London: Chapman & Hall; 1974.
15. D'Agostino RB, Lee ML, Belanger AJ, Cupples LA, Anderson K, Kannel WB. Relation of pooled logistic regression to time dependent Cox regression analysis: the Framingham Heart Study. Stat Med 1990; 9:1501–1515.
16. Brown TT, Li X, Cole SR, Kingsley LA, Palella FJ, Riddler SA, et al
. Cumulative exposure to nucleoside analogue reverse transcriptase inhibitors is associated with insulin resistance markers in the Multicenter AIDS Cohort Study. AIDS 2005; 19:1375–1383.
17. Shikuma CM, Day LJ, Gerschenson M. Insulin resistance in the HIV-infected population: the potential role of mitochondrial dysfunction. Curr Drug Targets Infect Disord 2005; 5:255–262.
18. FRAM Study Investigators. Fat distribution in men with HIV infection
. J Acquir Immune Defic Syndr
19. FRAM Study Investigators. Fat distribution in women with HIV infection
. J Acquir Immune Defic Syndr
20. Mynarcik DC, McNurlan MA, Steigbigel RT, Fuhrer J, Gelato MC. Association of severe insulin resistance with both loss of limb fat and elevated serum tumor necrosis factor receptor levels in HIV lipodystrophy. J Acquir Immune Defic Syndr 2000; 25:312–321.
21. Tien PC, Schneider MF, Cole SR, Justman JE, French AL, Young M, et al
. Relation of stavudine discontinuation to anthropometric changes among HIV-infected women. J Acquir Immune Defic Syndr 2007; 44:43–48.
22. Dube MP, Edmondson-Melancon H, Qian D, Aqeel R, Johnson D, Buchanan TA. Prospective evaluation of the effect of initiating indinavir-based therapy on insulin sensitivity and B-cell function in HIV-infected patients. J Acquir Immune Defic Syndr 2001; 27:130–134.
23. Noor MA, Lo JC, Mulligan K, Schwarz J-M, Halvorsen RA, Schambelan M, et al
. Metabolic effects of indinavir in healthy HIV-seronegative men. AIDS 2001; 15:F11–F18.
24. Noor MA, Seneviratne T, Aweeka FT, Lo JC, Schwarz JM, Mulligan K, et al
. Indinavir acutely inhibits insulin-stimulated glucose disposal in humans: a randomized, placebo-controlled study. AIDS 2002; 16:F1–F8.
25. Lee GA, Lo JC, Aweeka F, Schwarz J-M, Mulligan K, Schambelan M, et al
. Single-dose lopinavir-ritonavir acutely inhibits insulin-mediated glucose disposal in healthy volunteers. Clin Infect Dis 2006; 43:658–660.
26. Lee GA, Seneviratne T, Noor MA, Lo JC, Schwarz JM, Aweeka FT, et al
. The metabolic effects of lopinavir/ritonavir in HIV-negative men. AIDS 2004; 18:641–649.
27. Mehta SH, Brancati FL, Sulkowski MS, Strathdee SA, Szklo M, Thomas DL. Prevalence of type 2 diabetes mellitus among persons with hepatitis C virus infection in the United States. Ann Intern Med 2000; 133:592–599.
28. Mehta SH, Moore RD, Thomas DL, Chaisson RE, Sulkowski MS. The effect of HAART and HCV infection on the development of hyperglycemia among HIV-infected persons. J Acquir Immune Defic Syndr 2003; 33:577–584.