Stavudine or indinavir-containing regimens are associated with an increased risk of diabetes mellitus in HIV-infected individuals

Diabetes mellitus was diagnosed in 16 out of 1011 HIV-positive patients over a median follow-up of 289 days (person-year incidence 2.06, 95% confidence interval 1.18-3.33). Significant risk factors for the onset of diabetes were older age and antiretroviral therapy with stavudine or indinavir. Older men with HIV infection should be considered at higher risk of diabetes, and caution may be warranted in the use of both indinavir and stavudine in these patients.

The association between highly active antiretroviral therapy (HAART) and metabolism disorders was described shortly after the introduction of HIV-1 protease inhibitors (PI) into routine clinical practice [¹]. Since then, the complex syndrome of hyperlipidemia, lipodystrophy or diabetes mellitus (DM) has been observed in a considerable proportion of patients [²], leading to an increased risk of morbidity and mortality as a result of metabolic disorders and coronary heart disease [³], increased diagnostic and therapeutic costs, and significant patient discomfort secondary to body image alterations.

It has been reported that hyperglycemia with or without DM occurs in 3-17% of patients receiving HAART, shortly after the start of PI or even after their prolonged use [⁴]. Whether this effect is a direct result of PI therapy or a PI-exacerbated primary HIV metabolic defect is still unclear [⁵]. At present, most data concerning the incidence of diabetes and its correlation with antiretroviral therapy (ART) come from cross-sectional studies of small numbers of patients.

In the present study, we have analysed retrospective data from a cohort of 1011 HIV-positive patients (68% men, median age 37 years, range 18-74, median duration of known HIV infection 84 months, range -3 to 210 months); clinical, biochemical, immunological, virological and therapeutic data were collected from an in-house database, started on 1 November 1999. Blood tests were obtained for each patient at regular intervals as a part of routine outpatient care.

Study patients met the following criteria: normal fasting plasma glucose levels at study entry, and no previous diagnosis of DM; at least two fasting plasma glucose determinations during follow-up; a follow-up of at least 3 months; a stable (without interruption) antiretroviral regimen (or no therapy) for at least one month before entry and throughout the follow-up period.

Follow-up was censored at the last available plasma glucose assay or at the diagnosis of DM; the whole exposure time to each drug since its start was also recorded. DM was diagnosed on the basis of the 1997 American Diabetes Association guidelines (fasting plasma glucose ≥ 126 mg/dl on two different occasions) [⁶].

Data are reported as median values (minimum, maximum). As the variables did not have a Gaussian distribution, the Mann-Whitney test was used to investigate between-group differences. Cox proportional hazard regression models were used to investigate the associations between DM and putative risk factors (age, sex, CD4 cell count, HIV-RNA level, antiretroviral drugs).

During the follow-up (median 289 days, 91-624 days) DM was diagnosed in 16 out of 1011 patients [person-year incidence 2.06, 95% confidence interval (CI) 1.18-3.33]. Table 1 summarizes the statistical analyses of putative risk factors for DM. Older age [hazard ratio (HR) 1.11, 95% CI 1.06-1.16, P < 0.001] was associated with a higher risk of developing diabetes, whereas male sex showed a trend towards it, without reaching conventional statistical significance (HR 6.4, 95% CI 0.83-48.8, P = 0.07).

Table 1

After multivariate analysis adjusted for age and sex, diabetes onset was unrelated to the CD4 cell count, HIV-RNA level, ART assumption as a whole, PI therapy versus non-PI-containing regimens. Of note is the fact that there was no statistically significant difference in the whole duration of ART between diabetic and non-diabetic patients, thus including the pre-enrollment time. Furthermore, age and sex-adjusted Cox regression models of the individual drugs showed that patients treated with stavudine or indinavir were at significantly higher risk of developing DM (stavudine: HR 16.0, 95% CI 3.03-83.8, P = 0.001; indinavir HR 4.0; 95% CI 1.26-12.7, P = 0.018) (Table 2). To address the issue of whether the onset of diabetes may be stochastic or dose dependent, we investigated whether exposure to stavudine or indinavir (in the whole cohort) was significantly longer compared with other drugs. We found a potential confounding bias in the case of abacavir, efavirenz or nelfinavir because these drugs were registered later in Italy. However, the median duration of exposure was similar among stavudine and other reverse transcriptase inhibitors, as it was among indinavir, saquinavir and ritonavir (data not shown). We therefore tried to understand a possible synergic role of stavudine and indinavir therapy in the onset of DM, since after the multivariate analysis in the subgroup of patients taking indinavir alone did not show a significant HR for the development of diabetes (HR 16, 95% CI 0.26-75.4, P = 0.3), whereas patients on stavudine alone (HR 16.5, 95% CI 2.23-121.6, P = 0.006) and those on stavudine and indinavir were at significantly higher risk (HR 16.5, 95% CI 5.0-314.9, P < 0.0001). Finally, fasting cholesterol and triglyceride levels in index cases were collected close to the time of diabetes diagnosis (Table 1).

Table 2

The present study has two limitations, the first being the small group of index cases and the second that cohort patients were already on antiretroviral drugs at study enrolment; in this regard our data should be confirmed in large cohorts of antiretroviral-naive patients [⁷]. Despite that, this is to our knowledge the first longitudinal study about diabetes incidence in such a large cohort of HIV-positive patients. Very few studies investigating the incidence of diabetes in the general population have been performed, even in Italy. One study in a population in northern Italy [⁸] estimated a yearly incidence of 2.2 per 1000 person-years. The population was older than the HIV cohort we have followed, therefore we cannot draw any conclusion from this, but we may speculate whether some antiretroviral drugs can trigger the onset of diabetes in predisposed individuals. Despite that, the results about the drug-related risk of diabetes were partly surprising. First, we clearly showed an increased risk of DM in patients on indinavir, thus confirming preliminary results from in-vitro and in-vivo studies, occasionally leading to some pathogenic models [^9,10].

As for stavudine, this is the first report suggesting that this drug may be an independent and strong risk factor for the onset of DM. At present, data are both anecdotal and controversial with regard to the causative role of single nucleoside reverse transcriptase inhibitors (didanosine and abacavir) in diabetes onset [^11,12], whereas data on lipoatrophy/lipodystrophy and hypertriglyceridemia after stavudine use are as yet inconclusive [^10-12]. As a result of our epidemiological results we cannot speculate about a pathogenic model for stavudine-related DM, but we are concerned about the risk of diabetes in patients on prolonged stavudine or indinavir therapy. In this regard, we propose that older men with HIV infection are considered at higher risk of DM, and that caution may be warranted in the use of both indinavir and stavudine in these patients.