| |
The Effect of HAART and HCV Infection on the Development of Hyperglycemia Among HIV-Infected Persons
|
| |
| |
JAIDS Journal of Acquired Immune Deficiency Syndromes 2003; 33(5):577-584
*Shruti H. Mehta; †Richard D. Moore; *†David L. Thomas; †Richard E. Chaisson; †Mark S. Sulkowski
This study finds HCV/HIV co-infected individuals are more likely to have elevated glucose (>200 mg/dl) than HIV infected patients (5.9% vs 3.3% in this study. After starting HAART, co-infected patients were more likely to develop elevated glucose (5.8% vs 2.8%). Patients receiving PI were more likely to develop elevated glucose than patients receiving NNRTI (5.0% vs 1.6%), 3.9% for patients receiving PI+NNRTI.
"......A total of 269 (22%) were prescribed an NNRTI-containing HAART regimen, 845 (69%) a PI-containing regimen, and 116 (9%) individuals were prescribed a regimen that contained both a PI and an NNRTI...... Of the 1230 persons studied, 579 (47%) were HCV negative and 651 (53%) were HCV positive...... the prevalence (before starting HAART) of hyperglycemia was significantly higher among persons with HCV coinfection (33 of 564, 5.9%) compared with those without HCV infection (21 of 639, 3.3%, P = 0.03)...... New-onset hyperglycemia (after starting HAART) was observed more frequently in HCV-coinfected persons (30/517, 5.8%) compared with HCV-uninfected persons (17/611, 2.8%, P = 0.01)...... Incident hyperglycemia was detected in 4 of 245 patients receiving an NNRTI (1.6%), 39 of 780 patients receiving a PI (5.0%), and 4 of 103 patients receiving both a PI and an NNRTI (3.9%) (P = 0.07)...... Incidence rates per 100 person-years were 1.5, 4.1, and 2.8, among NNRTI users, PI users, and those on both an NNRTI and PI, respectively. In univariate Cox regression analysis, persons prescribed PIs were significantly more likely to develop hyperglycemia compared with those prescribed NNRTIs (relative hazard [RH], 2.84; 95% CI, 1.01-7.86) ...... neither concurrent nor cumulative exposure to individual NRTIs was associated with incident hyperglycemia...... There were no statistically significant differences in the incidence of hyperglycemia for specific drug regimens (data not shown). However, no cases of hyperglycemia occurred among 113 persons receiving nevirapine only...... Only 1 case of incident hyperglycemia occurred among persons who were neither HCV infected nor receiving a PI, and those who were both HCV coinfected and receiving a PI had the highest incidence of hyperglycemia (6.4 cases per 100 persons)...... suggesting an additive effect of HCV infection and PI use on the development of hyperglycemia......
"......Among HCV-infected individuals, insulin resistance and glucose intolerance are highly correlated with the severity of liver disease, and in some studies, insulin resistance has been observed prior to the development of advanced liver disease (Konrad T, Zeuzem S, Toffolo G, et al. Severity of HCV-induced liver damage alters glucose homeostasis in noncirrhotic patients with chronic HCV infection. Digestion. 2000; 62:52-59)...... Similarly, insulin resistance has also been demonstrated in patients receiving PIs and may underlie the development of hyperglycemia in these individuals. Recent
evidence also suggests that among persons receiving HAART, impaired glucose transport and phosphorylation contribute to impaired glucose uptake and that skeletal muscle is the primary site of impaired glucose utilization. Reduced glucose uptake by the skeletal muscle has also been suggested as a mechanism for liver cirrhosis-induced insulin resistance (Petrides AS, DeFronzo RA. Glucose and insulin metabolism in cirrhosis. J Hepatol. 1989; 8:107-114). Moreover, HCV-related liver disease may be associated with decreased hepatic metabolism of PIs via the cytochrome P450 system, leading to increased exposure to these agents and thereby augmenting the effect of PIs on insulin resistance and glucose metabolism. Thus, the increased occurrence of hyperglycemia and diabetes among HCV-infected persons receiving PIs may reflect an additive or synergistic effect on the development or severity of insulin resistance. However, in the absence of liver histology and information regarding drug metabolism or insulin resistance, we were unable to further evaluate this hypothesis.
ABSTRACT/SUMMARY:
Objective: To examine the prevalence and incidence of hyperglycemia among HIV-infected patients by hepatitis C virus (HCV) infection and type of highly active antiretroviral therapy (HAART).
Design: Retrospective cohort analysis of 1230 persons on their first HAART regimen who had at least 1 random glucose measurement before and during antiretroviral therapy.
Methods: The prevalence of hyperglycemia and the incidence of hyperglycemia were compared among persons with and without HCV infection while on a protease inhibitor (PI)-containing HAART regimen, a nonnucleoside reverse transcriptase inhibitor (NNRTI)-containing regimen, or a regimen that contained both a PI and an NNRTI. Hyperglycemia was defined as either 2 random glucose levels > 11.1 mM (200 mg/dL) or documentation of the diagnosis of diabetes in themedical record.
Results: The prevalence of hyperglycemia was significantly higher in HCV-coinfected (5.9%) than HCV-uninfected persons (3.3%, P = 0.02). Among persons receiving HAART, both HCV coinfection (adjusted relative hazard [ARH], 2.28; 95% CI, 1.23-4.22) and PI use (ARH, 5.02; 95% CI, 1.39-18.16) were independent risk factors of developing hyperglycemia. The incidence of hyperglycemia was highest among HCV-coinfected persons receiving a PI (5.6 cases per 100-person years) and only 1 person who was neither HCV-infected nor receiving a PI developed hyperglycemia.
Conclusion: In this urban HIV cohort, the risk of hyperglycemia was increased in HCV-coinfected patients and those prescribed a PI.
Background: The use of highly active antiretroviral therapy (HAART) has substantially reduced the morbidity and mortality associated with HIV infection by decreasing HIV replication, increasing immune function, and reducing rates of opportunistic infections. However, several reports have documented an increased occurrence of metabolic abnormalities including hyperlipidemia,
hyperglycemia, insulin resistance, and overt diabetes mellitus among persons receiving HAART, particularly HIV-1 protease inhibitors (PIs), raising concerns about long-term use of these agents. Moreover, some studies have suggested that removal of the PI, or replacement with a nonnucleoside reverse transcriptase inhibitor (NNRTI) such as nevirapine can normalize glucose
levels and, in some cases, reverse diabetes mellitus.
Hepatitis C virus (HCV) infection has also been independently associated with diabetes mellitus in nearly 30 studies. Due to shared routes of transmission, hepatitis C coinfection is common among HIV-infected persons, occurring in approximately 15-30% of HIV-infected patients. Despite both the relatively high prevalence of HCV among HIV-infected patients and its association with diabetes, the role of HCV coinfection in the development of hyperglycemia and diabetes mellitus in patients receiving HAART remains uncertain. Accordingly, the objective of this study was to examine the prevalence and incidence of hyperglycemia among HIV-infected patients according to HCV infection and the type of HAART received.
Study Population
The study population included 1230 persons receiving their first HAART regimen from January 1996 through May 2002 at the Johns Hopkins Hospital HIV clinic. As previously described, all patients in this urban, university-based setting undergo an initial comprehensive evaluation and are longitudinally monitored for therapeutic interventions and disease outcomes.
Patients were classified as receiving HAART if a PI or an NNRTI was prescribed with a nucleoside reverse transcriptase inhibitor (NRTI) backbone. According to clinical practice guidelines, patients receiving HAART are routinely scheduled for clinic visits and laboratory evaluations 4 weeks after the initiation of therapy and every 12 weeks thereafter. At each visit, patients had a clinical
examination and laboratory evaluation that included a complete blood count, serum chemistries, including serum glucose level, and determination of CD4 cell count and plasma HIV RNA level (reverse transcription polymerase chain reaction). All laboratory assessments were performed by licensed clinical laboratories. HCV testing is routinely performed on all cohort members, according to the practice guidelines of the clinic and the United States Public Health Service, using a second- or third-generation enzyme immunoassay.18 Persons who were positive for HCV antibodies by 2 enzyme immunoassays were considered to be HCV antibody positive and are hereafter referred to as HCV infected.
Definition of Hyperglycemia
To maximize the sensitivity of detecting hyperglycemia, patients were eligible for this analysis if they had >1 pretreatment glucose level and >1 glucose level during treatment. Fasting glucose measurements were not available so a modified definition of the 2003 American Diabetes Association criteria was used to characterize hyperglycemia.19 Any single random glucose level >11.1 mM (200 mg/dL) was investigated further by physician review of patient medical records. Persons were considered to have prevalent hyperglycemia only if the elevated glucose level was substantiated by another random glucose > 11.1 mM or documentation of diabetes in the medical record prior to initiation of HAART. Those with prevalent hyperglycemia were excluded from the analysis of incident hyperglycemia. Persons were considered to have incident hyperglycemia if a single elevated glucose level during treatment was confirmed by a subsequent glucose level > 11.1mM or diabetes was documented in the medical record after the initiation of HAART. Medical charts were also reviewed for persons with incident hyperglycemia to ensure that diabetes or
hyperglycemia was not present prior to initiation of HAART. Persons with any glucose level >11.1 mM who did not have either a second elevated glucose level or a confirmatory diagnosis of diabetes in their medical record were excluded from analysis.
Additionally, to confirm the specificity of this definition, we examined medical records of a random sample of 145 individuals whose glucose levels were < 11.1 mM at all visits and 40 individuals with the highest glucose levels that were not diagnostic of hyperglycemia. The random sample of 145 individuals was selected to have 80% power to detect an incidence of at least 0.05% in this group. None of these individuals had a diagnosis of prevalent or incident hyperglycemia or diabetes recorded in their medical records nor evidence of diabetes in their progress notes, recorded blood sugars, and medication lists.
Statistical Analysis
The prevalence and incidence of hyperglycemia were examined according to 2 primary exposures: HCV infection and type of HAART. Type of HAART was categorized as PI alone, NNRTI alone, or PI and NNRTI (all 3 regimens were accompanied by an NRTI backbone) according to the first regimen that was prescribed after enrollment.
A total of 269 (22%) were prescribed an NNRTI-containing HAART regimen, 845 (69%) a PI-containing regimen, and 116 (9%) individuals were prescribed a regimen that contained both a PI and an NNRTI. Persons prescribed PIs were more often male and less frequently African-American than those on other regimens. At baseline, persons prescribed PIs also had lower glucose levels compared with those on other regimens.
RESULTS
Study Participants
Of the 1230 persons studied, 579 (47%) were HCV negative and 651 (53%) were HCV positive. Compared with those without HCV coinfection, persons with HCV coinfection were older, more often African-American, and more likely to use or have used injection drugs. At baseline, no significant differences were detected between HCV-coinfected and uninfected patients in the prevalence of chronic hepatitis B virus infection, baseline CD4 cell counts, HIV RNA levels, or
random glucose levels. Persons with HCV infection had significantly higher alanine aminotransferase (ALT) levels and aspartate aminotransferase (AST) levels compared with those without HCV infection.
Prevalent Hyperglycemia
Of the 1230 individuals eligible for analysis, the median number of pretreatment glucose measurements was 6 (interquartile range [IQR], 2-14). Fifty-four individuals (4.47%) had hyperglycemia at baseline, of whom 36 (66.7%) also had a diagnosis of diabetes in their medical record. There were an additional 27 individuals (2.2%) who had only 1 pretreatment glucose >11.1 mM and no documentation of diabetes in the medical record. These 27 individuals, of whom 15 (56%) were HCV positive, were excluded from subsequent analysis. Among the 1203 who were analyzed further, the prevalence of hyperglycemia was significantly higher among persons with HCV coinfection (33 of 564, 5.9%) compared with those without HCV infection (21 of 639, 3.3%, P = 0.03). However, the prevalence of hyperglycemia was not significantly different by
type of HAART selected, 5.0% (13 of 262) for those prescribed an NNRTI, 4.2% (35 of 828) for those prescribed a PI, and 5.3% (6 of 113) for persons prescribed both an NNRTI and PI (P =0.80). In addition to HCV coinfection and the type of HAART prescribed, multivariate analysis considered other known risk factors of hyperglycemia including age, gender, race, and weight. We also considered hepatitis B virus infection (as indicated by hepatitis B surface antigen), parameters of HIV disease at baseline including CD4 cell count and HIV RNA viral load, and baseline ALT and AST levels. Only HCV infection (odds ratio [OR], 1.83; 95% CI, 1.05-3.20), HBV infection (OR, 2.18; 95% CI, 1.02-4.64), and age >50 years (OR, 3.46; 95% CI, 1.74-6.89) were independently associated with prevalent hyperglycemia. The results did not change when only persons with chart-diagnosed diabetes were analyzed.
Incident Hyperglycemia
Among the 1203 persons analyzed above, 1149 did not have prevalent hyperglycemia at baseline and were thus included in the analysis of incident hyperglycemia. The median duration of follow-up after initiation of the first HAART regimen was 8.9 months (IQR, 3.9-20.7 months) and the median number of glucose measurements during treatment was 5 (IQR, 2-11). The median
number of months on the first prescribed HAART regimen was not different by the type of regimen prescribed: NNRTI (9.6; IQR, 4.2-18.2), PI (9.1; IQR, 4.1-21.2), and PI and NNRTI (10.1; IQR, 3.5-29.2; P = 0.73), nor did the median number of glucose measurements differ by HCV status or the type of HAART received. Twenty-one individuals (1.8%), 14 (67%) of whom were HCV infected, were excluded because they had >1 glucose during treatment > 11.1 mM that was not substantiated by a second glucose level > 11.1 mM or documentation of diabetes in the medical record. New-onset hyperglycemia was observed in 47 (4.2%) of the remaining 1128 HAART recipients, an incidence of 3.5 per 100 person-years (95% CI, 2.6-4.6). Of these 47 individuals, 37 (78.7%) also had a diagnosis of diabetes in their medical record.
New-onset hyperglycemia was observed more frequently in HCV-coinfected persons (30/517, 5.8%) compared with HCV-uninfected persons (17/611, 2.8%, P = 0.01). When time of exposure was considered, the incidence of hyperglycemia per 100 person-years was significantly higher in HCV-coinfected persons (4.9; 95% CI, 3.4-7.1) compared with HCV-uninfected persons (2.3;
95% CI, 1.4-3.7). Incident hyperglycemia was detected in 4 of 245 patients receiving an NNRTI (1.6%), 39 of 780 patients receiving a PI (5.0%), and 4 of 103 patients receiving both a PI and an NNRTI (3.9%) (P = 0.07). Incidence rates per 100 person-years were 1.5 (95% CI, 0.6-4.1), 4.1 (95% CI, 3.0-5.6), and 2.8 (95% CI, 1.0-7.4), among NNRTI users, PI users, and those on both an NNRTI and PI, respectively. In univariate Cox regression analysis, persons prescribed PIs were significantly more likely to develop hyperglycemia compared with those
prescribed NNRTIs (relative hazard [RH], 2.84; 95% CI, 1.01-7.86). However, the incidence of hyperglycemia among persons receiving both a PI and a NNRTI was not different from the other two groups.
In univariate Cox regression analysis, HCV infection, older age, injection drug use, baseline glucose level, baseline ALT >40 U/L, baseline AST >40 U/L, and baseline CD4 cell count <50 cells/mm3 were also associated with the development of hyperglycemia. Of note, neither concurrent nor cumulative exposure to individual NRTIs was associated with incident hyperglycemia. Accounting for other independent risk factors of hyperglycemia including age (RH, 1.04; 95% CI, 1.01-1.07) and baseline glucose level (RH, 1.25; 95% CI, 1.15-1.36) actually strengthened the relationship between PI use (RH, 5.02; 95% CI, 1.39-8.16) and HCV infection (RH, 2.28; 95% CI, 1.23-4.22) and hyperglycemia. The effects of HCV infection and PI use were not different when persons with diabetes diagnosed by medical chart were analyzed
separately. There were no statistically significant differences in the incidence of hyperglycemia for specific drug regimens (data not shown). However, no cases of hyperglycemia occurred among 113 persons receiving nevirapine only and only 1 case occurred among 94 persons receiving nevirapine in combination with a PI. Only 1 case of incident hyperglycemia occurred among persons who were neither HCV infected nor receiving a PI, and those who were both HCV
coinfected and receiving a PI had the highest incidence of hyperglycemia (6.4 cases per 100 persons). This observed incidence was higher than what would be expected based on the joint effects of HCV infection and PI use (5.1 cases per 100 persons) on an arithmetic scale, suggesting an additive effect of HCV infection and PI use on the development of hyperglycemia. However, the observed incidence did not support a multiplicative or synergistic effect of HCV infection and PI use.
DISCUSSION
Among HIV-infected adults receiving medical care in an urban clinic, we observed an increased prevalence of hyperglycemia among persons with HCV coinfection compared with those without HCV infection prior to the initiation of HAART. Moreover, both HCV coinfection and PI use appeared to increase the risk of new-onset hyperglycemia during HAART. Both of these effects were independent of other risk factors of hyperglycemia including age, race, and body weight.
The detection of an increased prevalence of hyperglycemia among HIV-infected persons with HCV coinfection compared with those without HCV infection is consistent with nearly 30 other reports that have suggested an association between HCV infection and diabetes. These reports have documented that the prevalence of diabetes among persons with HCV infection ranges from 25-50%. Our study extends these findings by detecting a similar positive association of HCV and hyperglycemia in another group of persons, HIV/HCV-coinfected persons. In a previous cross-sectional study, Duong et al. found that glucose levels were not elevated among HIV/HCV-coinfected patients compared with those with HIV alone. However, in this report, HCV coinfection was associated with evidence of insulin resistance. Since we did not have information
on insulin levels in our study, were unable to address the relationship of HCV or HAART and insulin resistance.
The observed prevalence of diabetes (based on medical chart diagnosis) in this population is comparable to what would be expected based on prevalence rates from the general United States population. Accounting for the age and racial distribution of these patients, 45 cases of diabetes would have been expected compared with the 36 that were actually observed (standardized prevalence ratio, 0.80). This is surprising considering that almost 50% of this population was HCV coinfected. However, the median body weight was considerably lower in this population compared with the general population (70 kg for men and 60 kg for women in this population compared with 80 kg for men and 66 kg for women in the general U.S. population). Additionally, traditional risk factors of diabetes (e.g., obesity and race) were not associated with prevalent diabetes. This suggests that although the overall prevalence is low, diabetes in this urban, HIV-infected population may be primarily attributable to other factors, including HCV coinfection.
On the other hand, the incidence of diabetes (based on medical chart diagnosis) among persons receiving HAART appeared higher than what would have been expected based on age-adjusted general population incidence estimates. Applying the incidence rates of the general population to the
age distribution of this population, only 27 cases would be expected during the observed person-time.22 However, we observed 37 cases (standardized incidence ratio, 1.37). The change in the epidemiology of diabetes in this population was probably driven by the introduction of HAART and potential synergism between HCV infection and PI use. Previous reports have consistently
suggested that insulin resistance is common among persons receiving HAART, but have been conflicting regarding the occurrence of overt diabetes mellitus, with some reports finding it to be a common occurrence4 and others suggesting it to be relatively uncommon. Interestingly in this study, the adjusted RH of hyperglycemia for PI users compared with NNRTI users was nearly double the unadjusted RH. This was primarily driven by the lower baseline glucose levels among PI users, suggesting that, if anything, the incidence of diabetes among PI users might be even higher than what was observed here.
It is difficult to directly compare the results of these prior studies to our findings because of differences in study design and control group selection. Many of the previously published studies have been cross-sectional and have included variable control groups, ranging from PI-naive patients (usually on NRTIs) to healthy individuals. In our analysis, prevalent cases of hyperglycemia were excluded from the analysis of incident hyperglycemia and the control group included persons receiving non-PI-containing HAART regimens (nevirapine or efavirenz). Compared with control groups that include only patients receiving NRTIs, patients receiving NNRTI-based HAART are more likely to be similar to PI users with respect to parameters of HIV infection (e.g., CD4 cell count and HIV-1 viral load) and other confounders that might affect the
occurrence of hyperglycemia. Additionally, most previous studies of hyperglycemia and diabetes among HIV-infected persons have not accounted for HCV coinfection. To our knowledge, only 1 other study examined the effect of HCV coinfection on the incidence of hyperglycemia among persons receiving HAART, and in that retrospective cohort analysis, HCV infection was not
associated with the development of hyperglycemia.
In this analysis, both HCV infection and PI use were associated with an increased risk of developing hyperglycemia. Our finding of a lower incidence of hyperglycemia among NNRTI users compared with PI users is consistent with a previous study by Martinez et al. Suggesting that replacement of a PI with nevirapine reversed hyperglycemia and supports the hypothesis that it
is PIs and not HIV itself or the viral suppression and immune reconstitution associated with therapy that are responsible for hyperglycemia. The incidence of hyperglycemia was highest among persons who were both HCV coinfected and receiving PI-containing HAART, and only 1 individual who was neither HCV infected nor receiving a PI developed new-onset hyperglycemia during the follow-up period. The magnitude of the joint effect of HCV infection and PI use was suggestive of an additive effect of HCV coinfection and PI use on hyperglycemia.
While our data support the association of hyperglycemia with HCV coinfection and PI use, the etiopathogenesis of hyperglycemia among HIV-infected adults is unknown. Among HCV-infected individuals, insulin resistance and glucose intolerance are highly correlated with the severity of liver disease, and in some studies, insulin resistance has been observed prior to the development of
advanced liver disease. Similarly, insulin resistance has also been demonstrated in patients receiving PIs and may underlie the development of hyperglycemia in these individuals. Recent evidence also suggests that among persons receiving HAART, impaired glucose transport and phosphorylation contribute to impaired glucose uptake and that skeletal muscle is the primary site of impaired glucose utilization. Reduced glucose uptake by the skeletal muscle has also been
suggested as a mechanism for liver cirrhosis-induced insulin resistance. Moreover, HCV-related liver disease may be associated with decreased hepatic metabolism of PIs via the cytochrome P450 system, leading to increased exposure to these agents and thereby augmenting the effect of PIs on insulin resistance and glucose metabolism. Thus, the increased occurrence of hyperglycemia and diabetes among HCV-infected persons receiving PIs may reflect an additive or synergistic effect on the development or severity of insulin resistance. However, in the absence of liver histology and information regarding drug metabolism or insulin resistance, we were unable to
further evaluate this hypothesis.
Our findings were limited by a lack of information on the severity of liver disease and persistent HCV infection (indicated by the presence of HCV RNA). Because 85-90% should have had detectable HCV RNA based on the results of prior studies in similar cohorts, few participants who had cleared HCV infection would have been included in this analysis. Moreover, this misclassification would have been nondifferential with respect to diabetes status and would have effectively driven the results toward the null hypothesis of no difference.
In addition, our findings were limited by the potential misclassification of prevalent and incident hyperglycemia. Glucose levels were not measured in the fasting state and were not obtained systematically for all members of this cohort but rather at the discretion of the clinician. Thus, individuals who were sicker may have had more frequent testing, leading to increased detection of
hyperglycemia among these individuals. However, accounting for the frequency of glucose monitoring in the final multivariate models could not fully explain the observed relationships. The associations between HCV infection, PI use, and hyperglycemia were only moderately attenuated after accounting for these differences (data not shown). Data on the use of medications to treat diabetes, such as insulin or oral hypoglycemic agents, were not available. It is possible that patients to whom PIs were prescribed were less frequently treated with hypoglycemic agents because of concerns about drug interactions or hepatotoxicity, thus leading to an increased detection of hyperglycemia in this group. However, there is no evidence to support these prescribing practices.
Despite these limitations in the diagnosis of hyperglycemia, most prevalent and incident cases of hyperglycemia were confirmed by a physician diagnosis of diabetes in the patient medical record, and persons with only one or more elevated glucose levels that were not confirmed by either a second elevated glucose level or medical record review were excluded from analysis because of
potential misclassification. We also reviewed charts of 40 individuals with the highest glucose levels that were not diagnostic of hyperglycemia and 145 randomly selected individuals with no elevated glucose measurements. None of these individuals had a confirmed diagnosis of hyperglycemia or diabetes, suggesting that our definition of hyperglycemia was also specific. In
any case, we would anticipate this misclassification of hyperglycemia to be nondifferential with respect to HCV infection and the type of HAART therapy received, thus potentially attenuating the effects of PI therapy and HCV infection on incident hyperglycemia.
In conclusion, we observed an increased prevalence and incidence of hyperglycemia among HIV-infected adults with HCV coinfection compared with those without HCV infection. PI use also appeared to increase the risk of hyperglycemia compared with persons receiving NNRTI-containing HAART regimens. Additional prospective studies are needed to better understand the pathogenesis of hyperglycemia among HIV-infected patients receiving HAART,
particularly those coinfected with hepatitis C without previously described risk factors such as obesity.
|
|
| |
| |
|
|
|
