Serum cortisol in HIV-infected patients with and without highly active antiretroviral therapy

Abstract

Cortisol serum levels were evaluated in clinically stable HIV-infected patients. Patients with metabolic disturbances had higher cortisol concentrations than those without (P= 0.007). Cortisol levels were associated with non-nucleoside reverse transcriptase inhibitors (NNRTI) (P< 0.0001) and protease inhibitors (P= 0.006), but not with nucleoside reverse transcriptase inhibitors only (P= 0.96). Untreated patients who started therapy experienced an increase in serum cortisol (P= 0.01). Only NNRTI (P= 0.01) and metabolic disturbances (P= 0.02) were significantly associated with cortisol levels in multivariate analysis. Although many studies have analysed the adrenal function in HIV-infected patients [1–7], to our knowledge only two small studies have been performed in patients receiving highly active antiretroviral therapy (HAART) with contradictory results [8,9]. We prospectively studied a large number of stable patients, who underwent sequential measurements of serum cortisol to ascertain the prevalence of adrenal dysfunction, to analyse the factors that could be associated with cortisol levels, and in particular, to examine the possible influence of antiretroviral therapy. The study group was composed of 197 adult patients (188 men, mean age 36.8 years), who underwent a total of 351 determinations of cortisol, as well as of other clinical and laboratory parameters, each separated by a mean interval of 28.1 weeks. The patients were seen at our outpatient clinic and none of them had identifiable severe, acute diseases at the time of sampling. All laboratory determinations were performed in the same blood samples collected between 08:00 and 08.30 a.m. because of the circadian rhythms. Cortisol was determined by immunochemoluminescence (automatic analyser ADVIA: CENTAUR; Bayer Diagnostics, Tarrytown, NY, USA). The range of normality was 166–773 nmol/l. The distribution of cortisol values was Gaussian. Therefore, the Pearson's correlation coefficient, Student's t-test, one-way analysis of variance, and paired t-test were used as appropriate. The independent influence of diverse variables on cortisol concentrations was evaluated by a stepwise multiple linear regression analysis. The mean serum cortisol for all determinations was 577 nmol/l (95% confidence interval, 555–599 nmol/l). Increased serum cortisol was found in 45 determinations (12.8%), and decreased concentrations in 10 determinations (2.8%). None of these patients had symptoms of hyper- or hypocortisolism. There was no significant correlation between cortisol and age (P = 0.9), duration of HAART (P = 0.06), or CD4 cell counts (P = 0.9). Patients who had a past diagnosis of AIDS had significantly higher cortisol concentrations than patients without such a diagnosis (mean 599 versus 549 nmol/l, P = 0.02), and patients with an undetectable viral load had higher cortisol levels than patients with a detectable viral load (mean 612 versus 519 nmol/l, P < 0.0001). Patients with metabolic disturbances had higher cortisol concentrations than patients without (mean 607 versus 546 nmol/l, P = 0.007). Among patients receiving HAART, those who had lipodystrophy as the only metabolic disturbance had lower cortisol levels than those with other metabolic disturbances (diabetes or hyperlipidemia) that did not include lipodystrophy (mean 579 versus 679 nmol/l, P = 0.01), and similar concentrations to patients who did not develop metabolic disturbances (560 nmol/l, P = 0.6). Patients on HAART had higher cortisol values than untreated patients (mean 596 versus 486 nmol/l, P < 0.0001). Fig. 1 shows the mean values of cortisol according to different treatment groups. Patients on efavirenz had significantly higher cortisol concentrations than those on nevirapine (mean 668 versus 596 nmol/l, P = 0.026), and than those treated with protease inhibitors (PI) (mean 571 nmol/l, P = 0.003). Patients treated with nucleoside reverse transcriptase inhibitors (NRTI) plus PI who switched to NRTI plus non-nucleoside reverse transcriptase inhibitors (NNRTI) experienced an increase in their cortisol levels with respect to patients who continued with NRTI plus PI (mean 668 versus 541 nmol/l, P = 0.049). Finally, cortisol values clearly increased after the onset of HAART (from a mean of 455 to 651 nmol/l, P = 0.01) in 15 patients who had determinations before and after the initiation of therapy.Fig. 1.: Mean cortisol values in the five treatment groups. NNRTI, Non-nucleoside reverse transcriptase inhibitors; NO, no antiretroviral drug; NRTI, nucleoside reverse transcriptase inhibitors; PI, protease inhibitors. Vertical bars represent the standard error.There was a trend towards higher cortisol concentrations over time (P = 0.056), which reached statistical significance between the first and the third determination (P = 0.02). A multiple regression analysis of all parameters evaluated revealed that only treatment with NNRTI (P= 0.01) and the presence of metabolic disturbances (P = 0.02) were significantly associated with cortisol levels. This series constitutes, to our knowledge, the largest study approaching the adrenal function in stable, HAART-treated patients. Our results suggest that HAART increases cortisol levels, and this was confirmed by the subset of patients who started treatment during the study period. A direct effect of antiretroviral drugs on the cortisol metabolic pathways seems the most likely explanation for these findings, as all PI inhibit to different degrees the CYP3A4 isozyme [10]. In addition, PI have been found to inhibit the metabolism of other sterols, such as testosterone [11]. Similarly to others [12], we did not find significant differences in cortisol levels between patients treated with NRTI only and those untreated. However, the NNRTI-treated patients had higher cortisol values than the group taking PI. This difference was more marked in efavirenz-treated patients, a drug that both induces and inhibits CYP3A4. Efavirenz increases the area under the curve of ethynyl estradiol, a synthetic estrogen [13], and the serum levels of 17β-estradiol [14], but, to our knowledge, no effects on cortisol or other corticosteroids have been reported. Why patients treated with nevirapine, a pure CYP3A4 inducer, had similar levels to those treated with PI is unclear, but this observation suggests that the mechanism is not mediated through CYP3A4 inhibition. Two previous studies failed to demonstrate any significant difference in cortisolemia between lipodystrophic patients and either non-lipodystrophic patients [8] or healthy HIV-negative individuals [9]. Our study also demonstrates significantly higher cortisol levels in patients with laboratory abnormalities as the sole metabolic side-effect, than in those with lipodystrophy not accompanied by biochemical abnormalities. We found progressively increasing cortisol levels over time. This increase cannot be attributed to a worsening in the condition of the patients and the subsequent stress, because they were clinically stable and, in fact, their immunological and virological parameters improved over time (data not shown). Probably, the reason for this increase was related to the higher frequency of antiretroviral treatment in subsequent determinations, which was related to increasing metabolic disturbances. In fact, both antiretroviral therapy and metabolic disturbances were associated with higher cortisol values in both the univariate and multivariate analyses. We conclude that hypercortisolemia is commonly observed in stable HIV infection in the absence of clinical manifestations, whereas the prevalence of hypocortisolemia is substantially lower. Metabolic disturbances and antiretroviral treatment with drugs other than NRTI, particularly NNRTI, are associated with higher cortisol concentrations. Although alterations in the metabolic pathways of cortisol by these drugs could account for our findings, the intimate mechanism responsible for th these interactions remains to be elucidated.