Background: Diabetes mellitus (DM) is more prevalent among patients with HIV infection. Besides protease inhibitors (PIs), other factors may contribute to the development of DM.
Objective: To assess characteristics associated with the development of DM in HIV-infected persons.
Methods: We conducted a case-control study in an urban HIV clinic among patients with incident DM (49 cases) matched to 2 controls (n = 98) on age ±5 years, race, sex, and length of clinic follow-up. There was a second set of unmatched controls (n = 196).
Results: Compared with matched controls, case patients had higher mean body mass index (BMI; 30.0 vs. 25.3 kg/m2, matched odds ratio [OR] = 1.20; P < 0.001), higher alanine aminotransferase (ALT; 66 vs. 44 U/L, OR = 1.12 per 10 U/L; P = 0.013), and stronger family history of DM (50% vs. 29%, OR = 3.30; P = 0.009). Hepatitic C virus coinfection and PI use were not significant factors. In unmatched controls, there was no significant difference in age, sex, or ethnicity. In multivariate analyses, BMI (OR = 1.13 per kg/m2; P = 0.012), family history (OR = 5.55; P = 0.014), and ALT (OR = 1.16; P = 0.012) were associated with DM.
Conclusion: These findings suggest a complex interaction among genetic factors, body composition, and liver injury in the pathogenesis of DM in HIV-infected patients.
From the *Department of Medicine, Weill Medical College, Cornell University, New York, NY; and †Department of Statistics and Institute for Health, Health Care Policy and Aging Research, Rutgers University, Piscataway, NJ.
Received for publication February 4, 2004;
accepted June 16, 2004.
Supported by AI 51966 (K24 Investigator Grant to Roy M. Gulick).
Presented in part at the 9th Conference on Retroviruses and Opportunistic Infections, Seattle, February 24-28, 2002.
Reprints: Cecilia Yoon, Center for Special Studies, 119 West 24th Street, Ground Floor, New York, NY 10011 (e-mail: firstname.lastname@example.org).
Disorders of glucose metabolism have been reported in individuals infected with HIV.1-4 Cross-sectional studies have reported a prevalence of diabetes of 2% to 7% among HIV-infected patients receiving protease inhibitors (PIs)3,5,6 and an additional 16% having impaired glucose tolerance.3 The incidence of diabetes mellitus (DM) in HIV-infected patients has been estimated to range from 1% to 10% in various studies.7-9 An analysis of a California Medicaid database found that the age-specific relative risk for diabetes in persons with HIV compared with those without was indeed higher in all age groups, peaking at 7.74 among those 18 to 24 years of age.4
Clinical and in vitro data support a direct causative role of certain PIs in the pathogenesis of insulin resistance and DM in some patients with HIV infection.10 Other data have linked insulin resistance and diabetes to the lipodystrophy syndrome that is prevalent in HIV-infected patients and whose pathogenesis remains poorly understood.11-15 The relation of risk factors for DM that are well established in the general population such as family history, obesity, race/ethnicity, age, and dyslipidemia16 is poorly understood in HIV-infected patients, however. Furthermore, few data exist on the potential relations between DM and liver disease in HIV-infected patients, which may be of importance given emerging data associating DM with hepatitis C virus (HCV) infection in the general population.17-20 Drugs such as megestrol acetate and corticosteroids seem to be responsible for severe hyperglycemia in HIV-infected persons.21 To identify potential risk factors for diabetes in HIV-infected persons, we performed a retrospective case-control study of patients attending an urban HIV clinic.
This study was reviewed and approved by the Institutional Review Board at the Weill Medical College of Cornell University. The Center for Special Studies of the New York Presbyterian Hospital-Weill Cornell Center is composed of 2 clinics in Manhattan that provided care to approximately 5400 HIV-infected patients from May 1991 to December 2000. We ascertained cases of DM by searching the electronic medical records of these patients for a diagnosis of DM, prescriptions of drugs to treat diabetes, and elevated glucose values. Cases were defined as patients who met the World Health Organization (WHO) criteria for the diagnosis of DM: random glucose ≥200 mg/dL on 2 different occasions, fasting glucose ≥126 mg/dL on 2 different occasions, and/or receiving oral hypoglycemic agents and/or insulin.16 Glucose measurements were considered to be fasting only when documented as such in the medical record. To establish the temporal relation between potential risk factors and the onset of disease, we restricted our definition of cases to those with incident diagnoses of DM. Because height measurements and family history data were not routinely available in the medical records but were covariates of interest, we required that cases (and controls) be patients who were alive and had at least 1 clinic visit within 6 months at the time of this study (December 2000) so that these data could be collected prospectively as needed. We excluded patients whose DM was attributed to medications other than PIs known to affect glucose metabolism (ie, didanosine, corticosteroids, megestrol acetate, growth hormone, intravenous pentamidine), DM attributed to acute or chronic pancreatitis, and preexisting DM before the patient’s first clinic visit.
A set of “matched” controls was generated by randomly selecting 2 controls per case matched on age (±5 years), race, sex, and length of clinic follow-up from the initial visit to diagnosis of DM in cases or for an equivalent time period for controls (±24 weeks). Because cases and controls were matched on age, sex, and race on this first set of controls, we randomly selected a second set of 4 controls to each case matched only on length of clinic follow-up (thereafter known as an “unmatched” control group). This was done to determine specifically if age, sex, or race was associated with diabetes. We reviewed the medical records of all cases and controls to extract demographic and clinical data. Patients without documented family histories or heights had these data obtained prospectively at clinic visits. The mean values of laboratory tests such as alanine aminotransferase (ALT), cholesterol, and triglycerides (TGs, fasting and/or nonfasting) were calculated on the serial measurements available from the time of the initial visit to the time of diagnosis of diabetes in cases or for the equivalent calendar time in the matched controls. Other characteristics whose point of reference was the time of diagnosis with diabetes in cases were obtained at similar calendar time points in the matched controls. For HCV testing, the enzyme immunoassay (EIA) for the qualitative detection of antibody to HCV in serum or plasma was used. From June 1992, the EIA-2 was used, whereas before this date, the EIA-1 was used.
Matched univariate odds ratios (ORs) with 95% confidence intervals (CIs) were calculated by conditional logistic regression using Stata 7 software (Stata Corporation, College Station, TX). HIV RNA levels below the level of detection (<400 copies/mL or <50 copies/mL) were assigned values of 400 and 50 copies/mL, respectively, before log transformation. Wilcoxon signed-rank tests were performed on continuous variables with skewed distributions. Multivariate analysis was done using conditional logistic regression with forward stepwise selection.
Cases were selected by a stepwise process of exclusion as illustrated in Figure 1. Over 8.5 years, we identified 62 incident cases of DM among actively followed patients. Twelve (19.4%) case patients were excluded because of growth hormone or megestrol acetate use as the cause of DM, however. One case patient was excluded because no matched control was found, thus leaving 49 incident cases of diabetes matched with 98 randomly selected matched control patients without DM. Two female case patients could not be matched on gender or race, and 1 male case patient could not be matched on gender. An additional male case patient could be matched on time of initial visit with 1 of his controls only by extending the window for the date of the initial visit to −52 weeks.
Median time of clinic follow-up from initial visit to a diagnosis of diabetes was 21 months (interquartile range: 6-43 months). Table 1 summarizes the demographic characteristics of the 49 case patients with 196 unmatched control patients. There were no statistically significant differences in age, sex, and race, although case patients tended to be older and more commonly female and Hispanic (see Table 1).
In univariate analyses (Table 2) with matched control patients, case patients had a higher mean body mass index (BMI; 30.0 vs. 25.3 kg/m2; matched OR = 1.20 per kg/m2, 95% CI: 1.08-1.27; P < 0.001), family history of diabetes (50% vs. 29%, OR = 3.30, 95% CI: 1.34-8.09; P = 0.009), and mean ALT (66 vs. 44 U/L, OR = 1.12 per 10 U/L, 95% CI: 1.02-1.22; P = 0.013). Although not statistically significant, HCV coinfection was more common in case patients than in controls (51% vs. 37%, OR = 2.10, 95% CI: 0.95-4.82; P = 0.066). There was also greater PI use among case patients than controls (71% vs. 58%; OR = 2.30, 95% CI: 0.93-5.81; P = 0.072). Mean and median cholesterol levels were also similar, but median TG levels were higher in case patients than in controls (262 vs. 145 mg/dL; P = 0.001), although approximately half of the patients did not have TG levels available. There were no statistical differences in median CD4 cell count and mean HIV viral load at the time of diagnosis of diabetes or in nadir CD4 cell counts. There was greater tendency of prior injection drug use as the primary risk factor for HIV acquisition in the case patients (47% vs. 32%, OR = 2.13; 95% CI: 0.98-4.64; P = 0.057; see Table 2).
Table 3 displays the results of multivariate analyses using the matched controls. Of the variables considered in the model generated by forward, stepwise, conditional logistic regression, only BMI (OR = 1.13/kg/m2, 95% CI: 1.03-1.23; P = 0.012), family history (OR = 5.55, 95% CI: 1.41-21.85; P = 0.014), and ALT (OR = 1.16 per 10 U/L, 95% CI: 1.03-1.30; P = 0.012) were associated with DM.
In this case-control study, we found that several traditional risk factors for DM in the general population were associated with its development in HIV-infected patients. Specifically, a family history of diabetes in a first-degree relative and BMI were strong predictors. Also, level of ALT but not HCV serostatus was associated with DM.
More data have become available on the relative contributions of risk factors for DM in HIV-infected patients. Justman et al9 prospectively found that PI use was associated with a 3-fold increased risk of incident self-reported DM compared with non-HIV-infected women and HIV-infected women on no or non-PI-based therapy. A small case-control study in Spain identified baseline obesity, duration of PI use, and patient lipodystrophy as being associated with the development of DM.15
Various PIs have been shown to cause insulin resistance and DM.10,22,23 Administration of indinavir to healthy HIV-uninfected volunteers has caused reduced insulin sensitivity with as little as a single dose.10 PIs directly inhibit the Glut4 isoform, which mediates the transport of glucose, thus decreasing glucose disposal.22-24 This reduces uptake of glucose in muscle and adipocytes and increases intracellular lipolysis, subsequently increasing the release of unesterified fatty acids. This, in turn, stimulates hepatic glucose production and insulin secretion, thereby worsening hyperinsulinemia. In our study, there was a trend for an association between PI use and the development of DM that was not statistically significant. Several factors may account for this finding, including limited statistical power and possible heterogeneity in the propensity of specific PIs to cause diabetes.
Insulin resistance, impaired glucose tolerance, and, less commonly, DM are components of the lipodystrophy syndrome associated with HIV infection and its therapy.3,6,22 It appears that increased visceral fat and lipoatrophy may contribute to disordered glucose homeostasis.11-14,25-28 We were unable to assess the impact of lipodystrophy on the development of DM, given the retrospective nature of our study and the frequent lack of documentation.
HCV infection has been found to be associated with DM.17-20,29 The pathogenesis is unclear, but it has been suggested that hepatic steatosis, iron deposition in the liver, and progressive HCV-induced liver damage may induce insulin resistance and predispose to DM.19,29 Hepatic steatosis has been associated with lipid abnormalities, obesity, alcohol use, HCV, nucleoside analogue therapy, and insulin resistance and DM.30 Increased hepatic fat content has been strongly correlated with elevated fasting insulin levels and is associated with impaired glucose tolerance in HIV-infected persons.31 In a study of HIV-infected patients with lipodystrophy, Chung et al32 showed that ALT was a strong predictor of insulin resistance independent of viral hepatitis status after adjustment for age, BMI, and PI use. In our study, the mean ALT in cases was above the normal limit, implying underlying liver pathologic change that may be a marker for or predisposing factor for DM.
Similar to our findings in non-HIV-infected persons, Howard et al20 showed in a cross-sectional study that HCV was strongly associated with DM (OR = 2.9) after controlling for age, race, and BMI in HIV-infected drug users. Our study did not find a significant association between DM and HCV infection, although there was a trend toward a higher seroprevalence of HCV among the diabetic patients. It is possible that ALT was a more accurate marker of active HCV infection than HCV serology in this study, because tests for HCV viremia were generally not available to confirm the diagnosis of HCV in seropositive patients. Furthermore, the sensitivity of HCV serology may be reduced in patients with low CD4 cell counts,33,34 and the median nadir CD4 cell count in cases and controls was <200 cells/mm3. Lastly, patients known to have liver disease may have undergone more frequent monitoring of serum chemistries, allowing more opportunity for the diagnosis of hyperglycemia and diabetes. Although we did find that serum ALT was measured slightly more frequently in cases than controls (median: 10 vs. 8 times, respectively), we think that this is an unlikely explanation for our findings.
In univariate analysis, TG levels tended to be higher in diabetic patients compared with controls, although this was not confirmed in the multivariate analysis after adjustment for BMI and other covariates. This is consistent with findings from the general population that insulin resistance correlates closely with abdominal obesity and hypertriglyceridemia.35 Hypertriglyceridemia has been found to precede the development of type II DM in the general population.36-38 Although data in the HIV-infected population are limited, studies of HIV-infected subjects with lipodystrophy suggest that increased basal rates of lipolysis and free fatty acid concentrations may contribute to insulin resistance.39,40
Our study has a number of noteworthy limitations. Because we matched on age and ethnicity, we were unable to evaluate these potential associations in our study, although comparisons with a second set of unmatched controls found no significant differences in these demographic factors. Because of the retrospective nature of this study, patients were not monitored with laboratory tests in a consistent fashion. TG levels were available in only half of the patients. There was no uniform time interval for checking laboratory tests. We were unable to determine if the cholesterol and TG levels were obtained in the fasting state in most patients. Similarly, because we relied mostly on random rather than fasting glucose measurements, we may have excluded milder cases of diabetes that would have been detected only by fasting glucose determinations or oral glucose tolerance tests. To obtain height measurements and family histories on those with missing data in the medical record, we limited our study to patients who were actively followed at the clinic. The exclusion of patients who were no longer in follow-up poses several potential problems. Because we evaluated patients before and after the introduction of potent antiretroviral therapy, there may have been a significant number of patients who died of AIDS before 1996, thereby leaving a cohort of healthier individuals who were still being followed. The association of HCV infection with DM might be underappreciated because patients with HIV/HCV coinfection have increased mortality rates,41,42 thus potentially reducing the length of time in follow-up for DM to develop. Lastly, we were unable to assess the contribution of lipodystrophy to the development of DM.
The development of overt hyperglycemia seems to be the result of a complex metabolic cascade beginning with insulin resistance. Our data suggest that several steps can be taken in the care of HIV-infected persons to assess and potentially reduce the risk of diabetes. A family history to ascertain first-degree relatives with DM should be taken routinely. Obese patients should be counseled to lose weight. Patients with lipodystrophy, HCV coinfection, hepatic steatosis, and hypertriglyceridemia, independently or in combination, should probably be more closely monitored for development of DM. Ideally, patients with these conditions should avoid PI-based therapy or use PIs with a more favorable metabolic profile. The oral glucose tolerance test may be a useful tool to screen for subclinical DM for those patients at risk.
In summary, we found that a family history of DM, elevated BMI, and serum ALT levels are associated with the development of DM in HIV-infected patients. Our findings suggest a complex interrelation among genetic host factors, treatment-related metabolic changes, and liver injury in the pathogenesis of DM. Larger prospective studies are needed to delineate the relative contribution of other factors to the development of diabetes, such as specific PIs, liver disease, lipodystrophy, and other metabolic disorders.
The authors thank the following physicians, all from the Center for Special Studies, for their assistance in conducting this study: Jonathan Jacobs, Samuel Merrick, Susan Ball, Alice Barton, Brian Boyle, Lois Estok, Simon Paul, Duane Smith, Paul Smith, and Bruce Stewart.
1. Dube MP, Johnson DL, Currier JS, et al. Protease inhibitor associated hyperglycemia. Lancet
2. Dever LL, Oruwari PA, Figueroa WE, et al. Hyperglycemia associated with protease inhibitors in an urban HIV-infected minority patient population. Ann Pharmacother
3. Carr A, Samaras K, Thorisdottir A, et al. Diagnosis, prediction, and natural course of HIV-1 protease-inhibitor-associated lipodystrophy, hyperlipidemia, and diabetes mellitus; a cohort study. Lancet
4. Currier J, Boyd F, Kawabata H, et al. Diabetes mellitus in HIV infected individuals [abstract 677-T]. Presented at the 9th Conference on Retrovirus and Opportunistic Infections, Seattle, February 2002.
5. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cell counts of 200 per cubic millimeter of less. N Engl J Med
6. Vigouroux C, Gharakhanian S, Salhi Y, et al. Diabetes, insulin resistance and dyslipidaemia in lipodystrophic HIV-infected patients on highly active antiretroviral therapy (HAART). Diabetes Metab
7. Dube MP. Disorders of glucose metabolism in patients infected with human immunodeficiency virus. Clin Infect Dis
8. Hardy H, Esch LD, Morse GD. Glucose disorders associated with HIV and its drug therapy. Ann Pharmacother
9. Justman JE, Benning L, Danoff A, et al. Protease inhibitor use and the incidence of diabetes mellitus in a large cohort of HIV-infected women. J Acquir Immune Defic Syndr
10. Noor MA, Lo JC, Mulligan K, et al. Metabolic effects of indinavir in healthy HIV-seronegative men. AIDS
11. Carr A. HIV protease inhibitor-related lipodystrophy syndrome. Clin Infect Dis
. 2000;30(Suppl 2):S135-S142.
12. van der Valk M, Bisschop PH, Romijn JA, et al. Lipodystrophy in HIV-1 positive patients is associated with insulin resistance in multiple metabolic pathways. AIDS
13. Grinspoon S. Insulin resistance in the HIV-lipodystrophy syndrome. Trends Endocrinol Metab
14. Hadigan C, Meigs JB, Corcoran C, et al. Metabolic abnormalities and cardiovascular disease risk factors in adults with human immunodeficiency virus infection and lipodystrophy. Clin Infect Dis
15. Palacios R, Santos J, Ruiz J, et al. Factors associated with the development of diabetes mellitus in HIV-infected patients on antiretroviral therapy: a case-control study. AIDS
16. American Diabetes Association. Supplement 1. Screening for type 2 diabetes. Diabetes Care
. 2000;23(Suppl 1):S20-S23.
17. Mehta SH, Brancati FL, Sulkowski MS, et al. Prevalence of type 2 diabetes mellitus among persons with hepatitis C virus infection in the United States. Ann Intern Med
18. Mason AL, Lau JYN, Hoang N, et al. Association of diabetes mellitus and chronic hepatitis C virus infection. Hepatology
19. Garrido A, Guerrero FJ, Lepe JA, et al. Hepatitis C and diabetes mellitus: what is the connection? Presented at the 37th Annual Meeting of European Association for the Study of the Liver; Madrid, April 2002.
20. Howard AA, Klein RS, Schoenbaum EE. Association of hepatitis C infection and antiretroviral use with diabetes mellitus in drug users. Clin Infect Dis
21. Kilby JM, Tabereaux PB. Severe hyperglycemia in an HIV clinic: preexisting versus drug-associated diabetes mellitus. J Acquir Immune Defic Syndr Hum Retrovirol
22. Noor MA, Seneviratne T, Aweeka FT, et al. Indinavir acutely inhibits insulin-stimulated glucose disposal in humans: a randomized, placebo-controlled study. AIDS
23. Nolte LA, Yarasheski KE, Kawanaka K, et al. The HIV protease inhibitor indinavir decreases insulin-and contraction-stimulated glucose transport in skeletal muscle. Diabetes
24. Murata H, Hruz PW, Mueckler M. Indinavir inhibits the glucose transporter isoform Glut4 at physiologic concentrations. AIDS
25. Thiebaut R, Daucourt V, Mercie P, et al. Lipodystrophy, metabolic disorders, and human immunodeficiency virus infection: Aquitaine Cohort, France, 1999. Groupe d’Epidemiologie Clinique du Syndrome d’Immunodeficience Acquise en Aquitaine. Clin Infect Dis
26. Mynarcik DC, McNurlan MA, Steigbigel RT, et al. 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
27. Hadigan C, Miller K, Corcoran C, et al. Fasting hyperinsulinemia and changes in regional body composition in human immunodeficiency virus-infected women. J Clin Endocrinol Metab
28. Hadigan C, Corcoran C, Stanley T, et al. Fasting hyperinsulinemia in human immunodeficiency virus-infected men: relationship to body composition, gonadal function, and protease inhibitor use. J Clin Endocrinol Metab
29. Duong M, Petit JM, Piroth L, et al. Association between insulin resistance and hepatitis C virus chronic infection in HIV-hepatitis C virus-coinfected patients undergoing antiretroviral therapy. J Acquir Immune Defic Syndr
30. Angulo P. Nonalcoholic fatty liver disease. N Engl J Med
31. Sutinen J, Hakkinen AM, Westerbacka J, et al. Increased fat accumulation in the liver in HIV-infected patients with antiretroviral therapy-associated lipodystrophy. AIDS
32. Chung RT, Casson DR, Murray G, et al. Alanine aminotransferase levels predict insulin resistance in HIV lipodystrophy. J Acquir Immune Defic Syndr
33. Quan CM, Krajden M, Grigoriew GA, et al. Hepatitis C virus infection in patients infected with the human immunodeficiency virus. Clin Infect Dis
34. Marcellin P, Martinot-Peignoux M, Elias A, et al. Hepatitis C virus (HCV) viremia in human immunodeficiency virus-seronegative and -seropositive patients with indeterminate HCV recombinant immunoblot assay. J Infect Dis
35. Best JD, O’Neal DN. Diabetic dyslipidaemia: current treatment recommendations. Drugs
36. Haffner SM, Stern MP, Hazuda HP, et al. Cardiovascular risk factors in confirmed pre-diabetic individuals. Does the clock start ticking before the onset of clinical diabetes? JAMA
37. Mykkanen L, Kuusisto J, Pyorala K, et al. Cardiovascular disease risk factors as predictors of type 2 (non-insulin-dependent) diabetes mellitus in elderly subjects. Diabetologia
38. Defronzo RA, Bonadonna RC, Ferrannini E. Pathogenesis of NIDDM. A balanced overview. Diabetes Care
39. Reeds DN, Mittendorfer B, Patterson BW, et al. Alterations in lipid kinetics in men with HIV-dyslipidemia. Am J Physiol Endocrinol Metab
40. Meininger G, Hadigan C, Laposata M, et al. Elevated concentrations of free fatty acids are associated with increased insulin response to standard glucose challenge in human immunodeficiency virus-infected subjects with fat redistribution. Metabolism
41. Poles MA, Dieterich DT. Hepatitis C virus/human immunodeficiency virus coinfection: clinical management issues. Clin Infect Dis
42. Sulkowski MS, Mast EE, Seeff LB, et al. Hepatitis C virus infection as an opportunistic disease in persons infected with human immunodeficiency virus. Clin Infect Dis