Metabolic complications of HIV therapy in children.

Abstract

Introduction The implementation of potent antiretroviral therapy (ART) in combination regimens has profoundly decreased mortality and disease progression in patients infected with HIV [1]. Concomitantly, morbidity from the long-term effects of ART has grown in importance. Among all these complications, lipodystrophy, dyslipidemia, insulin resistance, and osteopenia are the most concerning side effects of prolonged ART. The medical literature contains numerous studies in adults related to these complications; however, the pediatric literature is relatively sparse. Consequently, comprehensive reviews of metabolic derangements associated with highly active antiretroviral therapy (HAART) in children are rare [2]. The virological successes of HAART certainly promise a longer lifespan for HIV-infected children, but the long-term metabolic consequences of this therapy are of serious concern. This review summarizes the epidemiology, clinical presentation, and management of lipodystrophy, dyslipidemia, insulin resistance, hyperlactatemia, and decreased bone mineral density (BMD) in HIV-infected children. In addition to describing the available pediatric data, our aim is to summarize and put in perspective for pediatricians the reported data on these complications from studies of HIV-infected adults. Each of the aforementioned morphological and metabolic phenomena of HAART-associated lipodystrophy syndrome is discussed separately, with attention to the epidemiology, clinical presentation, pathophysiology, and management of these disorders in HIV-infected children. Lipodystrophy syndrome and cardiovascular disease Soon after reports of morphological and metabolic abnormalities in HIV-infected adults, lipodystrophy syndrome became increasingly recognized in HIV-infected children [3–14]. Estimates of the prevalence of lipodystrophy range from 2 to 84% in HIV-infected adults [15] and from 1 to 43% in HIV-infected children [7]. The lipodystrophy syndrome encompasses changes in fat distribution typically manifesting as lipoatrophy with or without central adiposity, and it is frequently associated with alterations in lipid regulation and glucose homeostasis. Although affected individuals demonstrate differing patterns and severity of fat maldistribution, lipoatrophy is more specific to HIV infection and constitutes a key component of the lipodystrophy syndrome [16]. Regardless of the phenotypical presentation of these fat abnormalities, objective measurements are usually needed to establish the diagnosis unless they are severe enough to be visually obvious to patients and physicians. Various objective techniques to assess body fat content and distribution have been utilized. These are summarized in Table 1 and include bioelectrical impedance [17], anthropometric measurements, [10, 15], dual energy X-ray absorptiometry (DEXA) [15], and abdominal computed tomography or magnetic resonance imaging (MRI) [18,19]. Bioelectrical impedance provides useful information about lean body mass and total body fat but not about regional fat distribution. Anthropometric measurements assess only subcutaneous fat and require significant standardization in order to obtain reproducible measurements. DEXA scanning provides information about regional fat distribution, except for the face. Only computed tomography or MRI discriminates between subcutaneous fat stores and visceral fat.Table 1: Available modalities for assessing fat maldistribution.Patients with lipodystrophy often have significant central obesity; however, their subcutaneous abdominal fat is significantly lacking and visceral stores are markedly increased. Increased visceral fat has independently been associated with increased risk of cardiovascular disease as well as contributing to other known risk factors, including insulin resistance and lipid disorders, in patients without HIV infection [20–26]. The Framingham Offspring Study [25,27–30] has shown an increased 10-year coronary heart disease risk for HIV-infected adults with and without lipodystrophy, compared with healthy control subjects. The impact of fat maldistribution and its association with insulin resistance and dyslipidemia on HIV- infected adults is of obvious concern. Although Bozzette et al. [31] have published a large retrospective study showing no increase in cardiovascular or cerebrovascular hospitalizations in HIV-infected patients treated at Veterans Affairs hospitals between 1993 and 2001, several other studies have reported increased cardiovascular diseases in this population. Among these, the Data Collection on Adverse Events of Anti-HIV Drugs, a large prospective observational study, showed that exposure to combination ART was independently associated with a 26% relative increase in the rate of myocardial infarction per year of exposure during the first 4 to 6 years of use [32,33]. Dyslipidemia and insulin resistance, known metabolic associations with lipodystrophy, were also associated with increased risk of myocardial infarction in these patients. Similarly, Mary-Krause and co-workers [34] reported an increased rate of myocardial infarction in HIV-infected adults specifically related to duration of protease inhibitor (PI) therapy. The independent contribution of the fat abnormalities alone in these patients remains unclear and is currently under study. The overall impact of fat maldistribution on the care of HIV-infected patients has been very noteworthy, with a potential disincentive to therapy [35], low self-esteem, depression, and sexual difficulty [36]. These issues are especially problematic in adolescent patients, who are generally quite sensitive to body image, vulnerable to depression, and prone to non-compliance [37]. The determination of the etiology of fat maldistribution is further complicated by the phenotypical variability: lipohypertrophy, lipoatrophy, and mixed syndromes. Recent studies have proposed that lipoatrophy and lipohypertrophy should be considered as frequently comorbid, but distinct, entities when considered in HIV-infected patients. Lipoatrophy has been recognized as the more specific feature of fat abnormalities in HIV [16,38,39]. Multiple etiologies are hypothesized, with some evidence supporting each of these potential causes of fat changes. Table 2 summarizes the possible causes of lipoatrophy. Initially, PI therapy was implicated as the most likely cause of fat maldistribution [40], but more recent studies have shifted the focus to nucleoside reverse transcriptase inhibitors (NRTI) as primarily responsible for these fat abnormalities, particularly the lipoatrophy component [17,41,42].Table 2: Possible mechanisms of lipoatrophy and summary of available dataLipoatrophy is uncommon in patients who are treated with NRTI-sparing regimens [43,44]. Several clinical trials have demonstrated that an individual's risk to develop lipoatrophy depends on the choice of NRTI and have identified stavudine as having the greatest risk [17,41,42,45–49]. Further evidence for the culpability of NRTI, but not PI, therapy in lipoatrophy comes from the numerous therapy-switch studies. The replacement of a PI with either a non-nucleoside reverse transcriptase inhibitor (NNRTI) or abacavir in a virologically successful regimen did not lead to improvement of the fat abnormalities despite improvements in metabolic derangements in both adult and pediatric studies [50–53]. In contrast, therapy-switch studies to NRTI-sparing regimens [54] and changing certain NRTI (e.g. replacing stavudine with abacavir) did lead to an improvement of lipoatrophy [52,55,56]. Although the NRTI have a major role in the pathogenesis of lipoatrophy, other host factors also contribute. Older age, male gender, and low body fat prior to the initiation of ART are independent risk factors for lipoatrophy [57–59]. The body changes of lipodystrophy are difficult to examine in HIV-infected children because of the normal, dynamic alterations in body composition that occur during childhood and adolescence. Both DEXA [60,61] and MRI [18,62] have been validated to assess fat distribution in children. The entire spectrum of morphological derangements reported in adults has also been described in pediatric HIV-infected patients [3,5–7,10,13,37]. One study demonstrated increased ratios of visceral to total fat in HIV-infected children without overt signs of fat maldistribution when they were compared with age-, pubertal stage- and body mass index-matched controls [7]. As in adults, central obesity, hyperlipidemia, and insulin resistance are also felt to be risk factors in children for the development of cardiovascular disease [62–65]. This increased risk is concerning because dyslipidemia and, to a lesser extent, insulin resistance are associated with fat maldistribution in HIV-infected pediatric patients. Given the suggestion that the risk of cardiovascular disease is related to the duration of HAART therapy in adults [32–34,65] and that fat maldistribution in children is proportional to the exposure to HAART [13], one might surmise that HIV-infected children treated with HAART would be at exceptionally high risk for premature cardiovascular morbidity and mortality. In this regard, longitudinal data are being collected, and the first generation of HAART-treated children will soon enter their second decade. Aside from these therapy-switch studies, few therapeutic options exist for improving fat maldistribution in HIV-infected adults. Having found an association between oxidative markers and lipoatrophy [66], McComsey et al. [67] performed a pilot trial of antioxidants in HIV-infected adults. A significant decrease was reported in waist-to-hip ratio but not in other anthropometric measurements. A small study showed a potential increase in fat gain with the use of rosiglitazone in HIV-associated lipodystrophy [68], but this effect could not be confirmed in randomized, controlled trials [69,70]. Safety data for this agent are not available in children. In adults, visceral adiposity has been associated with impaired growth hormone secretion [71]. Kotler et al. [72] suggested that maintenance therapy with low-dose recombinant human growth hormone reduced visceral adipose tissue without creating insulin resistance in a group of HIV-infected adults without baseline glucose intolerance; however, several patients did develop insulin resistance in the initial higher-dose regimen [73]. Vigano et al. [74] have reported that HIV-infected adolescents with excessive accumulation of intra- abdominal adipose tissue had substantially impaired growth hormone secretion when stimulated with arginine and growth hormone-releasing hormone. These findings suggest a possible role for recombinant human growth hormone in this subpopulation, but studies are required on the benefit of decreased central adiposity versus the potential hazard of increased insulin resistance. The role and safety of growth hormone in prepubertal HIV-infected children with fat maldistribution is not known. Dyslipidemias Disorders in lipid metabolism have been commonly reported in HIV-infected adults as part of a lipodystrophy syndrome or as isolated abnormalities. Even before the availability of ART, low levels of high density lipoprotein cholesterol (HDL) and elevated triglycerides had been associated with HIV infection [75]. With the advent of PI therapy, dyslipidemia became more prevalent and more pronounced; patients demonstrated substantial increases in low density lipoprotein cholesterol (LDL) and even more pronounced elevations of triglycerides [76]. PI drugs have undergone the greatest scrutiny for their role in HIV-associated dyslipidemia. Several studies show an increased likelihood of dyslipidemia in patients who receive PI therapy, although the incidence depends on the type of PI used [15,40,77–82]. The best supporting evidence for a direct effect of PI on lipid metabolism comes from studies of HIV-seronegative volunteers, who developed dyslipidemia after a short course of treatment with a PI [83,84]. In one of the largest adult cohort studies, 49% of 581 HIV-infected patients had dyslipidemia, as determined by fasting lipids [15]; 33% had cholesterol > 212 mg/dl; 12% had HDL < 35 mg/dl; 25% had triglycerides > 195 mg/dl; and 16% had elevations in both cholesterol and triglycerides. Disorders in lipid metabolism have been reported in children and adolescents infected with HIV, but the literature is relatively sparse compared with that published on adult patients. Several small cross-sectional studies have shown elevations in total cholesterol in HIV-infected children, with a prevalence ranging from 13 to 75% [3,4,7,9,10,12]. In the largest published series of lipid assessment in HIV-infected children, Farley et al. [9] prospectively followed nearly 2000 perinatally HIV-infected children between the ages of 4 and 19 years. Approximately 13% had total cholesterol levels higher than the 95th percentile of the gender-, race-, and age-specific standards, as defined by NHANES III [85]. As in adults, PI therapy had the greatest association with dyslipidemia in the pediatric population [5,6,8,9,11,12,86,87]. Again, there has been a wide range in the reported prevalence of PI-induced dyslipidemia, ranging from 20% of children on a single PI to > 90% of children being treated with regimens containing two PI [3,87,88]. It is important to remember that prevalence does not necessarily imply causality. After multivariate analysis, Farley and associates [9] reported a 3.6 times risk of having a total cholesterol greater than the 95th percentile for gender, race, and age for PI-treated children and a 72% increase in risk for hypercholesterolemia for any additional PI added to a child's regimen. The most compelling data for the direct contribution of PI therapy to dyslipidemia in children come from a phase I/II study of ritonavir in children by Mueller et al. [87]. In this study, 12 weeks of ritonavir monotherapy in 48 children (age 6 months to 14 years) led to statistically significant increases in their total serum cholesterol, from 121 to 184 mg/dl. Cheseaux et al. [8] published the results of a retrospective review of lipid levels in 66 Swiss children before and after the addition of PI therapy to a regimen containing two NRTI drugs: 29 children were treated with ritonavir and 37 with nelfinavir. Although both groups showed significant increases in their total cholesterol after PI therapy, the ritonavir group had substantially higher cholesterol values than the nelfinavir group. Furthermore, the ritonavir group, but not the nelfinavir group, also had significant increases in their triglycerides. These two drugs have been the most commonly used PI in HIV-infected children and, therefore, are the best studied so far. Although longitudinal data for the consequences of dyslipidemia are lacking in the pediatric population, the hypothesized consequence would be the development of premature atherosclerotic disease. Cheseaux et al. [8] concluded that HIV-infected children treated with HAART have elevations in their cholesterol similar to those seen in patients heterozygous for familial hypercholesterolemia and, therefore, have a similar risk for premature atherosclerotic disease. In the Bogalusa Heart Study [63], 93 autopsies were performed on subjects aged 2–39 years, most of whom died from trauma and for whom antemortem data about cardiovascular risk factors were available. The aorta and coronary arteries were examined for the presence of fatty streaks or fibrous plaques. High body mass index, systolic and diastolic blood pressure, elevated serum concentrations of total cholesterol, triglycerides, and LDL-cholesterol significantly correlated with the extent of lesions in the coronary arteries and the aorta. Although the mean age of these subjects was 19 years, their autopsies already demonstrated atherosclerotic disease, which is a concerning observation for a pediatrician dealing with HIV-infected children with dyslipidemias and other risk factors of early atherosclerosis. The independent contribution of PI therapy to risk beyond elevated cholesterol levels is unclear. For instance, Depairon et al. [89] and Mercie et al. [90] have separately found that increased thickness of the carotid artery media, a surrogate for atherosclerosis, in HIV-infected adults was associated with conventional cardiovascular risk factors and not with ART. HIV-infected children with dyslipidemia would likely be at risk for accelerated atherosclerotic disease because of their elevations in total serum cholesterol, LDL, and triglycerides, plus, as is discussed elsewhere, their potential comorbid conditions (including increased waist-to-hip ratios, insulin resistance, and mitochondrial toxicity). The true risk will likely not be known until this first generation of HAART-treated children have been followed into adulthood. Management of lipid disorders in HIV-infected adults has proven difficult. There are no published data regarding the pharmacological treatment of HIV- infected children with hyperlipidemia. The Cardiovascular Subcommittee of the AIDS Clinical Trials Group (ACTG) has suggested following the National Cholesterol Education Program guidelines for the evaluation and treatment of dyslipidemia in HIV-infected patients [91,92]. This recommends improvements in diet and exercise as the first intervention. The standard lipid-lowering agents, 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins), must be used cautiously because several of them are metabolized by the P-450 enzyme CYP3A4 and, therefore, can affect serum concentrations of PI, leading to myalgias, myopathy, and rhabdomyolysis [93,94]. Few data exist about statin use in pediatric patients and these mostly reflect studies involving children with familial hypercholesterolemia [95]. In 2002, lovastatin became the first of this class of drugs to be approved for pediatric use: in adolescents with familial hypercholesterolemia. No long-term safety data in this population exist. The American Academy of Pediatrics recommends drug treatment in children only if older than 10 years and the presence after 6–12 months of dietary modification of serum LDL ≥ 190 mg/dl, or ≥ 160 mg/dl with a positive family history of coronary artery disease or in the presence of two other cardiac risk factors [96]. They recommend colestyramine and colestipol, two acid-binding resins that can significantly lower LDL cholesterol. These agents, however, may lead to further increases in serum triglycerides in patients taking a PI [94]. Another problem with these agents is their potential interference with the absorption of concurrently administered drugs, including antiretroviral drugs, which has the potential of subsequent virological failure. Fibric acid derivatives such as gemfibrozil and fenofibrate lower serum triglycerides levels and potentially raise serum HDL. They are not metabolized by CYPI but may interact with statins to cause rhabdomyolysis and hepatitis. The ACTG group recommends reserving these drugs for patients with triglyceride levels > 500 mg/dl [91]. These drugs are metabolized by a set of glucoronosyl transferases that are induced by the two most commonly used PI in pediatric HIV-infected patients, ritonavir and nelfinavir. Therefore, they may be only minimally efficacious in PI-associated dyslipidemia [94]. Another strategy for managing dyslipidemia in HIV-infected adults is to switch from a PI or to a PI-sparing regimens, strategies that have been shown to ameliorate the lipid abnormalities [52,97,98]. Few data, however, exist about the efficacy of these strategies in children. McComsey and colleagues [50] have published the only pediatric PI therapy switch study. Seventeen children were changed from a PI-containing regimen to efavirenz. After 48 weeks, the switch to efavirenz resulted in significant improvements in total cholesterol, LDL, and triglycerides while maintaining excellent virological control. Since no data exist to support therapy interruption in HIV-infected children, the development of PI-sparing HAART regimens or PI drugs that do not disturb lipid metabolism will be crucial in the prevention of treatment morbidity. Insulin resistance Since the initial US Food and Drug Administration report in 1997 of an association between PI therapy and diabetes mellitus [99], it has become increasingly recognized that insulin resistance and frank diabetes should be added to the list of potential long-term complications of HIV therapy. The lack of a standardized definition of insulin resistance has led to inconsistencies in the reporting of its prevalence. The diagnostic standard for insulin resistance is the euglycemic clamp, a complex procedure that is not feasible in most clinical settings. Therefore, the majority of HIV studies have used less-optimal diagnostic strategies, including the measurement of fasting glucose, fasting insulin, C-peptide, oral glucose tolerance tests, and the derivation of various indices generated from these values [15,17,100]. As for dyslipidemia, PI use has been the most frequently associated variable with disorders in glucose metabolism; however, these derangement do occur in PI-naive adults [76,77,80,101]. Walli et al. [80] reported that approximately 61% of PI-treated subjects have some degree of insulin resistance. In the French cohort study [15], 20% of the 581 patients had disorders in glucose metabolism. In the PI-treated patients, the point prevalence for any disturbance was 25%, significantly higher than in the non-PI-treated patients (P = 0.001); however, 17% of the patients treated with a NRTI drug also had disorders in glucose metabolism. This study is one of the first attempting to correlate metabolic disorders of lipid and glucose metabolism with fat maldistribution. Of their patients with body shape changes, 63% had at least one metabolic problem, compared with 58% of patients without such changes (P = 0.001). This difference remained significant (P = 0.012) when adjusted for PI therapy. Although there are case reports of diabetes mellitus developing in such patients, the long-term consequences of lower-grade insulin resistance are unknown; however, it is thought to pose an increased risk for cardiovascular disease [20,23,24,102]. The mechanism of insulin resistance in HAART-treated patients is not entirely understood and is likely multifactorial. As mentioned earlier, PI therapy has received the most attention, but studies have shown hyperinsulinemia with NRTI agents [103,104]. Some studies have suggested an independent contribution of increased visceral-to-subcutaneous fat ratio in HIV-infected patients to the development of insulin resistance [105,106]. Several PI agents can have direct effects on insulin resistance, through inhibition of insulin-stimulated glucose transport by the isoform transporter GLUT-4 [107,108]. Noor et al. [83] recently investigated the effect on insulin resistance of the administration for 6 days of atazanavir, lopinavir/ritonavir, or placebo to a group of healthy volunteers. Using the euglycemic clamp technique, they showed that, in contrast to lopinavir/ritonavir, atazanavir did not reduce insulin sensitivity, decrease glucose disposal, or decrease the glycogen storage rate. Larger clinical studies will be necessary to determine the long-term clinical significance of this finding. Hyperglycemia and insulin resistance have also been reported in HIV-infected children [3,10]. In a pediatric cohort of 39 patients, Jaquet et al. [10] reported that all patients had normal oral glucose tolerance tests and normal fasting serum glucose levels. However, fasting insulin values in patients with body changes were more than double those in patients of normal appearance. The children with body changes also demonstrated a degree of insulin resistance, as measured by fasting insulin-to-glucose ratios. Both observations trended toward significance (P = 0.07) even in this small study group. These trends are particularly disturbing in the context of the well-described decreased insulin sensitivity that occurs naturally during puberty [109,110]. Unlike the disturbances in glucose homeostasis reported in HIV-infected adults, insulin resistance in HIV-infected children is not clearly associated with PI therapy [5,10,11,14]. There are several reports of an association between fat maldistribution in HIV-infected children and insulin resistance, but the presence of confounding variables and small numbers preclude definitive linkage of these conditions [3,10,74]. In the largest published series to date on glucose homeostasis in HIV-infected children, Bitnun et al. [5] reported that only advancing Tanner stage (marker of pubertal development) correlated with insulin resistance. The authors speculated that the PI-associated alterations in glucose homeostasis were not seen in prepubertal children because of inherently greater insulin sensitivity and a tendency to have increased ratios of subcutaneous to visceral fat. The clinical difference in PI toxicity with regards to glucose homeostasis in children compared with adults is clear; however, the questions remain as to how to monitor the development of disorders in glucose metabolism in developing children. The American Heart Association and the American Diabetes Association have cited insulin resistance in children as a major cardiovascular risk factor and urged ‘aggressive’ therapy [65]. Unfortunately, there are fewer data about the management of insulin resistance in HIV-infected patients than for dyslipidemia. In non-HIV-infected children with insulin resistance, as for those with dyslipidemia, the first-line approach is dietary modification and weight control [65]. Unfortunately, data are sparse concerning the impact of these interventions in HIV-infected patients. Hadigan et al. [111], measured the insulin area under the curve after an oral glucose tolerance test in a cross-sectional study of 85 HIV-infected adults with fat redistribution; they reported a statistically significant correlation with the polyunsaturated-to saturated dietary fat intake and an inverse association with dietary fiber that were independent of PI use. To date, no large study examining the impact of dietary modification on insulin resistance in HIV-infected adults has been published. In a small French study, light aerobic exercise was shown to improve lipid profiles and fat maldistribution in HIV-infected adults, but it had no effect on derangements in glucose homeostasis [112]. Several small studies have demonstrated a positive effect of metformin on glucose homeostasis in HIV-infected adults [113–115], but lactic acidosis has been reported in non-HIV-infected adult patients with renal dysfunction. The drug is licensed in the United States for the treatment of children older than 10 years with type II diabetes. The thiazolidenediones, which decrease hepatic insulin resistance and act as promoter of adipogenesis to increase peripheral glucose uptake [116], have been shown to have therapeutic efficacy in adult patients with type II diabetes [117]. In HIV-infected subjects, rosiglitazone, a thizaoidenedione, has been shown to improve insulin resistance. Unfortunately, rosiglitazone is not licensed for pediatric usage. Switching to a PI-sparing regimen in HIV-infected adults has been shown to improve insulin resistance [51,97,118]. There are currently no formal recommendations to employ switch strategies in adults with HIV-associated metabolic disorders. Efavirenz substitution for a PI resulted in no significant changes in C-peptide or insulin levels during the 48-week study period [58]. The literature is minimal on the role of switch therapy in pediatric patients and more studies are needed. Bone disease Decreased bone mineral density Derangements in metabolism of bone are generally subclinical in HIV-infected children but have potentially serious consequences. In men and women, bone mass increases during childhood and adolescence, peaks in young adult life, and declines at a rate of approximately 0.5–1% per year [119]. The reported prevalence of decreased BMD in HIV-infected adults has been 22–50% of adult patients with osteopenia and 3–21% of those with osteoporosis [120–123]. Most HIV-infected patients with decreased BMD are asymptomatic, although fragility fractures in these patients are being increasingly reported [124–126]. Premature bone loss is particularly relevant in vertically infected children because lifelong administration of ART may be necessary and because of the risk of never achieving a physiological peak bone mass. Several studies have already shown a high prevalence of decreased BMD in HIV-infected children. [12,127–132]. Similarly to the situation in adults, the contribution of combination ART to the finding of decreased BMD in pediatric patients is not clear. This uncertainty arises from the cross-sectional nature of the majority of available pediatric studies, the limited number of children included in these studies, and the fact that the vast majority of these children have already been on combination therapy at the time of initial bone assessment. In this regard, the first reported study of bone assessment of HIV-infected adults before and after initiation of ART revealed that osteopenia was already more common in these HIV-infected treatment-naive subjects at baseline (before ART) compared with prevalence in the HIV-uninfected population [133]. After the initiation of therapy, and regardless of the choice of ART, BMD sharply decreased and then slightly plateaued. The end result was a substantial decrease in BMD of both the lumbar spine and hip when compared with pretherapy levels. This decrease was seen in both treatment arms, although it was significantly worse in the tenofovir-containing arm [134]. This study as well as others [133,135–137] imply that HIV itself m