Protease inhibitors are considered a crucial component of antiretroviral combination therapy for HIV-1 infection . These drugs interfere with post-translational processing of viral precursor proteins, and thereby prevent the formation of new infectious virus particles [2–4]. As false substrates, they block the active site of the highly specific HIV aspartate protease. A variety of side-effects have been documented in patients treated with protease inhibitors. Some of these are drug-specific, such as indinavir-associated crystalluria and nephrolithiasis [5, 6], or nelfinavir-associated diarrhoea . However, the new onset of diabetes mellitus [7–9], lipodystrophy [10, 11] and hyperlipidaemia [10, 12] has been observed in patients treated with all currently approved protease inhibitors (indinavir, nelfinavir, ritonavir, saquinavir). This suggests that class-specific mechanisms associated with the use of protease inhibitors are involved in the induction of these metabolic alterations. To date, it is unknown how protease inhibitors might interfere with metabolic regulatory processes, and a causative role of protease inhibitors remains to be proven. Peripheral insulin resistance has been suggested to contribute to the described changes in lipid and glucose metabolism [8, 10], because some of the findings resemble those seen in patients with non-insulin-dependent diabetes mellitus (NIDDM) . Insulin sensitivity can be assessed by performing an intravenous insulin tolerance test, a well-established and feasible alternative to the euglycaemic insulin clamp technique . The aim of this study was to evaluate the effect of treatment with protease inhibitors on insulin sensitivity, oral glucose tolerance and serum lipids in HIV-infected patients.
Between August 1997 and April 1998, 67 HIV-infected patients on antiretroviral combination therapy including protease inhibitors, 13 therapy-naive patients (i.e., patients not treated with protease inhibitors, reverse transcriptase inhibitors or any other antiretroviral drugs) and 18 control subjects (HIV-negative medical and non-medical hospital staff) were enrolled in the study. Data on pretreatment serum lipids and glucose was obtained from medical files. Blood glucose levels recorded prior to treatment were normal in 57 of the 67 patients, in all 13 therapy-naive patients, and in all 18 control subjects. There was no data on blood glucose or serum lipid levels available for 10 of the treated patients. All patients and control subjects underwent intravenous insulin tolerance testing, and oral glucose tolerance was assessed in 24 of the 67 treated patients. The first four of the 67 patients were intentionally selected on the basis of having developed diabetes mellitus under treatment with indinavir. The next 20 patients were selected for oral glucose tolerance testing regardless of prior blood glucose levels.
The patients reported to our clinic between 0800 and 1000 h, after overnight fasting and administration of the regular morning medication. Concurrent medication of the patients did not include drugs with a known diabetogenic potential. Three of the nine diabetic patients were using glibenclamide, which was not administered on the morning of testing. The remaining six diabetic patients were not treated with oral antidiabetics or insulin.
In the control group, four of the 18 subjects were women. In the group of therapy-naive patients, two of the 13 subjects were women. Four subjects had asymptomatic HIV infection [Centers for Disease Control and Prevention (CDC) clinical stage A], seven had symptomatic HIV infection (CDC stage B), and two subjects had previous AIDS-defining illnesses [CDC stage C; both cases of Pneumocystis carinii pneumonia (PCP)]. The duration of known HIV infection ranged from 3 to 97 months (median, 47 months). In the group of patients on protease inhibitors, four of the 67 subjects were women. Sixteen patients had asymptomatic HIV infection, 28 had symptomatic HIV infection, and 23 patients had previous AIDS-defining illnesses (15 PCP, three cerebral toxoplasmosis, four cytomegalovirus infection, two non-Hodgkin's lymphoma, three oesophageal candidiasis). Duration of known HIV infection ranged from 5 to 127 months (median, 53 months). There was no statistically significant difference in the duration of known HIV infection between therapy-naive and treated patients. All patients and controls gave informed consent prior to enrolment into the study.
Intravenous insulin tolerance test
A modified intravenous tolerance test  was performed. After overnight fasting (> 10 h), an antecubal 18 g intravenous cannula was positioned. The forearm was maintained on a hot-water bottle to arterialize the venous blood. The patients were given a bolus of 0.05 IU/kg of human insulin (Insulin Actrapid HM, Novo Nordisk, Mainz, Germany) intravenously. Blood samples were taken before (0 min), and 5, 8, 9, 10, 11, 12, 13, 14 and 15 min after administration of insulin for analysis.
Evaluation of insulin sensitivity
Peripheral insulin sensitivity (expressed in μmol/l/min) was determined by linear regression analysis of the decline of blood glucose levels over time (5–15 min after administration of insulin) . No normal range of insulin sensitivity has been defined. The distribution of insulin sensitivity in our control group (mean ± SEM, 174 ± 10 μmol/l/min) was comparable to that documented for control groups in previous studies (171 ± 10, 179 ± 11, 167 ± 11, 183 ± 7 μmol/l/min, respectively [14–16]). Values below the mean insulin sensitivity measured in the 18 HIV-negative control subjects minus 2 SD (92 μmol/l/min) were classified as consistent with pathological insulin sensitivity (i.e., peripheral insulin resistance).
Oral glucose tolerance test
After overnight fasting (>10 h), patients were challenged with 100 g glucose. Blood samples were taken before (0 min) and 30, 60, 90 and 120 min after administration of glucose for analysis. Oral glucose tolerance was considered as normal (whole blood glucose level <120 mg/dl at baseline and <120 mg/dl after 120 min), impaired (< 120 mg/dl at baseline and 120–180 mg/dl after 120 min) or diabetic (> 120 mg/dl at baseline or >180 mg/dl after 120 min) according to the 1985 World Health Organization definition .
Determination of blood glucose, serum lipids and basal insulin
Glucose levels were determined in whole blood samples using a commercial glucose hexokinase assay (Boehringer-Mannheim, Mannheim, Germany). Postprandial triglycerides and total cholesterol in serum were determined by a commercial photoenzymatic assay (Boehringer-Mannheim). Basal insulin level were determined using a commercial radioimmunoassay (Boehringer-Mannheim).
The exact unpaired Mann—Whitney U-test  was used to assess the statistical significance of differences in age, body mass index, duration of treatment with protease inhibitors, insulin sensitivity and other parameters. The effect of treatment with protease inhibitors on serum lipids was studied with an exact paired Mann—Whitney U-test . Statistical significance was assumed for P <0.05. The correlation of insulin sensitivity, changes in plasma lipids and the duration of treatment was assessed using Pearson correlation analysis .
Comparison of therapy-naive patients and HIV-negative controls
There were no statistically significant differences in median age, body mass index or basal blood glucose levels between the group of therapy-naive patients and the control subjects (Table 1). More importantly, there was no significant difference in median insulin sensitivity when comparing untreated patients and controls (156 and 177 μmol/l/min, respectively; P = 0.144). None of the 13 therapy-naive patients or 18 control subjects had an insulin sensitivity consistent with peripheral insulin resistance. Therefore, HIV-1 infection in untreated patients is not associated with decreased insulin sensitivity. Blood was not analysed for serum lipids and basal insulin in the control subjects.
Influence of treatment with protease inhibitors on insulin sensitivity
The 13 therapy-naive patients did not differ significantly from the 67 patients treated with protease inhibitors in median age, body mass index and CD4 cell count (Table 1). As expected, median plasma viral load was significantly lower in the treated patients (below level of detection, <2.70 log10 copies/ml) than in therapy-naive patients (4.39 log10 copies/ml; P <0.001). The most striking difference between the two patient groups was the significantly lower median insulin sensitivity in treated patients (75 μmol/l/min; P = 0.001; Fig. 1). Moreover, 61% of the treated patients showed peripheral insulin resistance. In addition, basal glucose and basal insulin levels were significantly higher in patients on protease inhibitors (P = 0.014 and 0.021, respectively), both findings consistent with decreased insulin sensitivity.
Influence of treatment with protease inhibitors on oral glucose tolerance
A subgroup (n = 24) of the 67 patients treated with protease inhibitors underwent an oral glucose tolerance test (Table 2). Oral glucose tolerance was normal in 11 (46%) of the patients, and a pathological result was obtained for the remaining 13 patients (nine were diabetic and four had impaired oral glucose tolerance). There was a significant difference in median insulin sensitivity between treated patients with normal and those with pathological oral glucose tolerance (121 and 55 μmol/l/min, respectively; P < 0.001; Fig. 2a). Interestingly, all 13 patients with pathological oral glucose tolerance had a peripheral insulin resistance, whereas four (36%) of the treated patients with normal oral glucose had an insulin sensitivity consistent with peripheral insulin resistance and seven (64%) had normal insulin sensitivity. The median duration of treatment with protease inhibitors was insignificantly longer in the 11 patients with normal oral glucose tolerance (median, 14.0 and 10.0 months, respectively; P = 0.241).
Influence of treatment with protease inhibitors on serum lipids
In the 67 treated patients, serum total triglycerides were significantly higher than in the therapy-naive patients (median, 294 versus 138 mg/dl; P = 0.007), as were total cholesterol levels (median, 203 versus 177 mg/dl; P = 0.021). Serum triglycerides were significantly higher in the 13 treated patients with pathological oral glucose tolerance than in the 11 treated patients with normal oral glucose tolerance (441 and 191 mg/dl, respectively; P <0.001). Total cholesterol did not differ between the two groups (191 and 199 mg/dl, respectively; P = 0.386).
The 67 patients treated with protease inhibitors showed a marked increase in total triglycerides (median, 180 versus 294 mg/dl; median increase, 113 mg/dl; P <0.001) and also a significant increase in total cholesterol (median, 175 versus 203 mg/dl; median increase, 37 mg/dl; P <0.001) when comparing levels prior to treatment with protease inhibitors to those at time of insulin tolerance testing (Fig. 3). Duration of treatment correlated with the increases in total cholesterol, but not with the increase in triglycerides. In contrast, insulin sensitivity was correlated with the increase in total triglycerides, but not with the increase in total cholesterol (Pearson's correlation coefficient, 0.014 and 0.633, respectively).
Differences between protease inhibitors
Fifty of the 67 treated patients were on their first protease inhibitor at the time of testing (29 on indinavir, nine on nelfinavir, two on ritonavir, 10 on saquinavir). The other 17 patients had changed protease inhibitors at least once. Six of these 17 patients were treated with a combination of two protease inhibitors. Insulin sensitivity was lower in the 29 patients treated with indinavir (median, 59 μmol/l/min) than in those treated with nelfinavir (median, 75 μmol/l/min) or saquinavir (median, 105 μmol/l/min). Median levels for serum triglycerides and total cholesterol was lowest in patients treated with saquinavir, and median total cholesterol was highest in patients on nelfinavir. The median duration of treatment with protease inhibitors was longer in the saquinavir subgroup (18 months) than in the indinavir and nelfinavir subgroups (12 and 6 months, respectively). None of these differences were statistically significant (Table 3). The highest total triglyceride level was seen in a patient treated with ritonavir (1176 mg/dl).
The most striking finding of this study was the significantly lower insulin sensitivity in the patients treated with protease inhibitors compared with therapy-naive patients. In the randomly selected subgroup of the treated patients, all patients with impaired and diabetic oral glucose tolerance had peripheral insulin resistance. In contrast, the patients with normal oral glucose tolerance showed both normal and pathological insulin sensitivity, and all untreated patients and control subjects had normal insulin sensitivity. Thus, impairment of insulin sensitivity, associated with the use of protease inhibitors, differs largely in severity. We therefore suggest a stepwise model of impairment of glucose homeostasis in patients treated with protease inhibitors. Minor decreases in insulin sensitivity can be compensated, resulting in normal oral glucose tolerance. However, more pronounced impairment can also effect oral glucose tolerance, and in some cases even lead to the manifestation of diabetes mellitus.
There are several mechanisms that are known to lead to peripheral insulin resistance . These include reduced binding of insulin to the insulin receptor, alterations in intracellular signalling pathways, and defective cellular uptake of glucose. In treated patients with pathological oral glucose tolerance, decreased insulin sensitivity was associated with increased basal glucose and insulin levels. This constellation is also seen in patients with NIDDM , a condition in which peripheral insulin resistance is common. We cannot rule out that other factors, such as interference with insulin secretion or an imbalance of insulinergic and anti-insulinergic regulators, might also contribute to the onset of protease inhibitor-associated metabolic side-effects. We cannot prove that the decreased insulin sensitivity in the treated patients was caused by protease inhibitors alone, and not partially by reverse transcriptase inhibitors as well. However, prior to the introduction of protease inhibitors, there were no reports on a higher frequency of new onset hyperglycaemia or diabetes mellitus after starting antiretroviral therapy. Therefore, we believe that the reported effects were linked to protease inhibitors, and not to reverse transcriptase inhibitors.
The second major finding of the study was the influence of treatment with protease inhibitors on serum lipids. Both total triglycerides and cholesterol were higher in treated patients than in therapy-naive patients. In addition, serum levels of both triglycerides and cholesterol increased significantly after starting treatment with protease inhibitors. These effects could be related to the observed peripheral insulin resistance, but other mechanisms are possible. There was no correlation between increases in serum lipids and the duration of protease inhibitor therapy or weight gain. In fact, even in patients with apparent treatment failure (i.e., rising viral load) and subsequent weight loss, the initial increase in serum lipids was not reversed. In six of the patients on protease inhibitors with peripheral insulin resistance, serum lipoprotein differentiation revealed a decrease in both high and low density lipoproteins, as well as an increase in very low density lipoproteins (VLDL) and total triglycerides (data not shown). The accumulation of triglyceride-rich lipoproteins may be caused by over-production of VLDL by the liver and low lipoprotein lipase activity. Such a phenomenon has been reported in untreated or poorly controlled diabetes . A decreased activity of lipoprotein lipase has been reported by Jimenez-Expósito et al.  in a patient with hypertriglyceridaemia and consecutive pancreatitis treated with ritonavir. Lipodystrophy, the third major metabolic side-effect related to treatment with protease inhibitors, was not evaluated in this study.
Due to the large variation of parameters within the three subgroups, and the small number of patients on nelfinavir (n = 9) and saquinavir (n = 10), caution is warranted in the interpretation of differences in the potential of the substances to induce the above-mentioned side-effects. There was a trend towards lower insulin sensitivity in patients treated with indinavir. In contrast, saquinavir seems to be the least diabetogenic protease inhibitor. Interestingly, the metabolic situation of two patients who developed diabetes mellitus after starting treatment with indinavir improved within a couple of months after replacing indinavir by nelfinavir. Both patients were able to discontinue treatment with glibenclamide. The incidence of hyperglycaemia has been reported to be 0.7–2.7% [8, 9]. In our outpatient clinic, 15 (6.5%) out of 234 patients developed transient or prolonged postprandial hyperglycaemia (> 180 mg/dl) after starting treatment with a protease inhibitor. Pretreatment postprandial blood glucose levels, when available, were normal (< 120 mg/dl) in all patients.
The prevalence of diabetes amongst our study population of treated patients (13%; nine out of 67) was skewed, because the first four patients were intentionally selected on the basis of having developed diabetes under indinavir or nelfinavir. Serum lipids are apparently least effected by treatment with saquinavir. Indinavir and nelfinavir do not differ in their potential to lead to increases in total triglycerides. There is a slight trend towards higher levels of total cholesterol in patients on nelfinavir. Severe hypertriglyceridaemia and associated pancreatitis are most frequently seen in patients treated with ritonavir. Absence of hepatic lipase activity has been reported as a possible mechanism leading to dyslipidaemia . In a report by Carr et al. , the prevalence of lipodystrophy in patients treated with protease inhibitors was 74%, with no apparent preferential association with specific drugs. Structural similarity of the enzymatic active site of HIV protease and a functional domain of lipoprotein receptor binding protein, involved in a scavenger pathway in chylomicron remmant metabolism, was suggested as a possible factor involved in the pathogenesis of lipodystrophy. The authors also reported a significant increase in basal glucose and insulin, and therefore hypothesized that peripheral insulin resistance might also be involved. However, insulin sensitivity was not evaluated in the study by Carr et al. .
Overall, the number of patients treated with the different protease inhibitors and the differences in duration of therapy reflect the general practice of prescribing protease inhibitors in our outpatient clinic. Besides the discussed differences in clinical appearance of metabolic side-effects and the potential reversibility of changes in glucose tolerance, the lacking correlation with the duration of treatment reflects the complex nature of these protease inhibitor-associated side-effects. We observed lipodystrophy as early as 6 weeks after starting nelfinavir, and manifestation of diabetes mellitus occurred after 2 months of treatment with cohort saquinavir in one patient. However, there are patients in our study cohort who have been treated for as long as 25 months with protease inhibitors in whom insulin sensitivity and serum lipids are normal, and who do not show any sign of lipodystrophy.
In summary, our study indicated that treatment with protease inhibitors is associated with the new onset of impaired oral glucose tolerance, diabetes mellitus and hyperlipidaemia. This seems to be a class-specific characteristic of protease inhibitors, although there is a large variation in the clinical presentation of metabolic changes. For the first time, we demonstrate that protease inhibitor treatment is associated with peripheral insulin resistance, a possible factor involved in the induction of metabolic side-effects. Better understanding of the described metabolic alterations is necessary, since the prospect of serious complications (e.g., artherosclerosis and coronary heart disease, pancreatitis) will influence the physicians decision of whether and when to use protease inhibitors for antiretroviral combination therapy in the future.
1. Carpenter CC, Fischl MA, Hammer SM, et al.
: Antiretroviral therapy for HIV infection in 1997: updated recommendations of the International AIDS Society—USA panel
2. Moyle G, Gazzard B: Current knowledge and future prospects for the use of HIV protease inhibitors
3. McDonald CK, Kuritzkes DR: Human immunodeficiency virus type 1 protease inhibitors
. Arch Intern Med
4. Peng C, Ho BK, Chang TW, Chang NT: Role of human immunodeficiency virus type 1-specific protease in core protein maturation and viral infectivity
. J Virol
5. Michl G, Herfort O, Walli R, Anders HJ, Jacob K, Goebel FD: Incidence and risk factors of crystalluria in patients treated with indinavir
. Sixth European Conference on Clinical Aspects and Treatment of HIV Infection
. Hamburg, October 1997 [abstract 442].
6. Kopp JB, Miller KD, Mican JA, et al.
: Crystalluria and urinary tract abnormalities associated with indinavir
. Ann Intern Med
7. Ault A: FDA warns of potential protease-inhibitor link to hyperglycaemia [news]
8. Dubé MP, Johnson DL, Currier JS, Leedom JM: Protease inhibitor-associated hyperglycaemia [letter]
9. Eastone JA, Decker CF: New-onset diabetes mellitus associated with use of protease inhibitor
. Ann Intern Med
10. Carr A, Samaras K, Burton S, Freund J, Chisholm DJ, Cooper DA: A syndrome of peripheral lipodystrophy, hyperlipidemia and insulin resistance in patients receiving HIV protease inhibitors
11. Hengel RL, Watts NB, Lennox JL: Benign symmetric lipomatosis associated with protease inhibitors [letter]
12. Jimenez-Expósito MJ, Paul A, Laville A, Masana L: Non activity of hepatic lipase during treatment with ritonavir
. Sixth European Conference on Clinical Aspects and Treatment of HIV Infection
. Hamburg, October 1997 [abstract 733].
13. Keen H, Barnes DJ: The diagnosis and classification of diabetes mellitus and impaired glucose tolerance
. In Textbook of Diabetes
. Edited by Pickup JC, Williams G. London: Blackwell Science; 1996.
14. Gelding SV, Robinson S, Lowe S, Niththyananthan R, Johnston DG: Validation of the low dose short insulin tolerance test for evaluation of insulin sensitivity
. Clin Endocrinol (Oxf)
15. Gelding SV, Niththyananthan R, Chan SP, et al.
: Insulin sensitivity in non-diabetic relatives of patients with noninsulin-dependent diabetes from two ethnic groups
. Clin Endocrinol (Oxf)
16. McKeigue PM, Pierpoint T, Ferrie JE, Marmot MG: Relationship of glucose intolerance and hyperinsulinaemia to body fat pattern in south Asians and Europeans
17. WHO: Diabetes mellitus. Report of a WHO Study Group
. World Health Organ Tech Rep Ser
18. Rosner B: Fundamentals of Biostatistics
. 3rd Edn
. Boston: PWS-Kent Publishing; 1990.
19. Syvanne M, Taskinen M: Lipids and lipoproteins as coronary risk factors in non-insulin-dependant diabetes mellitus