Background: Reports from controlled studies have estimated that 7%–15% of HIV⁺ patients (pts) undergoing PI therapy will develop diabetes mellitus (DM). Lack of investigation in a large clinical population leaves unclear if the occurrence of DM in HIV⁺ pts is greater than what would be expected in the non-infected (HIV^-) population.

Methods: We compared incidence and prevalence of DM between the HIV⁺ and HIV^- members of the N.CA. Kaiser Permanente Health Plan (NC-KPHP) between the years 1994– 2000, and examined whether observed differences could be attributed to PI use. This study reports DM among the largest HIV⁺ pt population examined thus far.

Results: Using epidemiological methods, among 2,600 (yearly average) NC-KPHP HIV⁺ members the age-adjusted incidence rate was 97.1 (95% CI = 38.3–155.9) DM cases per 10,000 members in 1994, rising to 188.2 in 1997, declining to 88.4 in 2000. Among the 2 million NC-KPHP HIV^- members the age-adjusted incidence rate was 63.1 (95% CI = 62.0–64.3) in 1994, 82.3 in 1998, and 83.4 in 2000. The age-adjusted prevalence rate of DM among those HIV⁺ was 335.7 (95% CI = 290.5– 451.0) cases in 1994, rising to 617.5 in 2000. HIV^- members had an age-adjusted prevalence rate of 408.7 (95% CI = 405.9–411.5) in 1994, rising to 596.5 in 2000. The standardized morbidity ratio (SMR), depicting ratio of observed to expected incident DM cases, was 1.58 in 1994, rising to 1.93 in 1996 and declining to 1.17 in 2000. Among HIV⁺ members who had never received PI therapy, the age-adjusted incidence rate was 97.1 (95% CI = 38.3–155.9) in 1994, rising to 181.0 in 1998, declining to 63.0 in 2000. Among HIV⁺ members who had ever taken a PI, the incidence rate was 319.0 (95% CI = 35.7–602.3) in 1996 (1st year PI’s prescribed), declining to 100.8 in 2000. Multivariate Poisson regression modeling, adjusting for demographic factors, revealed no significant association between DM and PI use except when models were stratified by year: in 1996, RR for PI vs no PI = 2.49, 95% CI = 1.04-5.96.

Conclusion: These results suggest that the elevated incidence of DM in the NC-KPHP HIV⁺ population may be attributable to the introduction of PIs in 1996, but that iatrogenic effect diminished in subsequent years, most likely due to a change in PI prescribing patterns.