Association of Prenatal and Postnatal Exposure to Lopinavir-Ritonavir and Adrenal Dysfunction Among Uninfected Infants of HIV-Infected Mothers FREE

The human immunodeficiency virus 1 (HIV-1) transmission rate to newborns is now less than 1% for women treated during pregnancy.¹ For pregnant women not optimally treated, as in cases of HIV diagnosis late during pregnancy or persistent viral replication at delivery, several guidelines, observational reports, and the results of a recent controlled study suggest reinforcing the postnatal phase of treatment with a combination of antiretrovirals, as a “postexposure prophylaxis.”^2- 7 The protease inhibitor lopinavir, with its pharmacological booster ritonavir (lopinavir-ritonavir, Kaletra; Abbott Laboratories, Abbott Park, Illinois), is now the ritonavir-boosted protease inhibitor most widely prescribed in children.⁸ Lopinavir-ritonavir is licensed in the United States for HIV-infected newborns older than 14 days and in Europe for children older than 2 years. However, published data concerning its use in newborns are scarce and limited to a few pharmacokinetic studies.^8- 12 The emergence of toxicity reports in newborns is evidence of its increasing use for postnatal prophylaxis. A possible cardiac toxicity has been described in 2 sets of premature twins in the United States and Europe.^13- 14 A recent Food and Drug Administration (FDA) warning¹⁵ confirms these findings, reporting other cases of life-threatening conditions in premature newborns. The mechanism of this toxicity is unknown, but possible yet unproven factors may include lopinavir accumulation due to liver immaturity in very young children, possible effects of the propylene glycol and ethanol used as excipients in the lopinavir-ritonavir oral solution, or both.

In April 2010, one of the centers of the French national screening program for congenital adrenal hyperplasia (CAH) identified a transient increase of 17-hydroxyprogesterone (17OHP) in dried blood spots from 2 children treated at birth with lopinavir-ritonavir. This unusual finding prompted us to cross-check the French CAH screening database and the French Perinatal Cohort of infants born to HIV-infected mothers. The aim of the study was to assess whether immediate postnatal exposure to lopinavir-ritonavir was associated with changes in adrenal function compared with standard prophylactic zidovudine treatment.

The French Perinatal Cohort is a nationwide prospective multicenter study following up children born to HIV-infected mothers. The goals and design have been described elsewhere.¹⁶ Between September 24, 1984, and January 1, 2011, 15 376 mother-child couples have been included at 97 sites. The present analysis was restricted to 32 centers in the Paris area (région Ile de France) at which 2816 couples were included during the study period (December 2004 to September 2008). All HIV-infected pregnant mothers giving informed consent (overall rate of refusal, 4%) and their newborns are anonymously included no later than delivery.

Care and prophylaxis strategies are the responsibility of the attending clinicians, who are encouraged to follow national guidelines. In brief, these guidelines—like those in all other Western countries—recommend a combination of antiretroviral treatment during pregnancy, an intravenous infusion of zidovudine during delivery, and zidovudine monotherapy for the newborn lasting 4 to 6 weeks. The guidelines include the possibility of reinforcing postnatal prophylaxis for every case in which the mother's treatment is considered to be suboptimal.^2- 4 In French guidelines,² one of the options recommended for reinforcement of postnatal prophylaxis includes combination therapy based on lopinavir-ritonavir alone with monitoring of plasma concentrations.

The French national screening for CAH is based on assaying 17OHP using dried blood collected on filter paper.¹⁷ The recommended timing of the test is at age 3 days, and samples are generally collected between days 2 and 5. A 17OHP value of 16.5 ng/mL or greater for full-term infants and of 23.1 ng/mL or greater for preterm infants (<37 weeks of gestation) indicates possible 21 hydroxylase deficiency.¹⁷ Children with a positive screening result are immediately recalled for clinical evaluation and serum 17OHP determination. During the study period, 692 200 screening tests (mean, 180 253 per year) were performed in the Paris area. The coverage rate of this screening is near 100%, with fewer than 50 refusals annually.

Written informed consent and written agreement of 1 of the parents are required for neonatal CAH screening. Participation in the French Perinatal Cohort is based on informed consent. All collected data are sent anonymously to the coordinating center, and only investigators in clinical centers could validate the possible cross identification between the 2 databases. The cross-checking process and the use of frozen samples for additional hormonal investigations were both expressly authorized for this study by an institutional review board (Comité de Protection des Personnes, Paris).

Children treated with postnatal lopinavir-ritonavir prophylaxis were eligible if they were in the Paris area between December 2004 and September 2008, were not HIV-infected, and received lopinavir-ritonavir for at least 3 days or until the screening test for CAH was performed, if not done at age 3 days. Sixty-four children at 15 of 32 sites fulfilled these criteria (lopinavir-ritonavir–treated group). All non–HIV-infected children who received a postnatal prophylactic regimen free of protease inhibitors were eligible for the control group if they were born during the same period and at the same 15 sites as eligible lopinavir-ritonavir–treated children, to limit site and period effects (n = 1402). To facilitate the complex process of confidential cross-checking with CAH, we restricted this process to a subgroup of children whose identity had already been well established on site—in strict respect of confidentiality—during a previous probabilistic cross-matching with French registers of pediatric cancers¹⁸ (eFigure).

Overall, 17OHP screening results were identified with certainty (and the children thus included) for 50 of 64 eligible newborns treated with lopinavir-ritonavir, and 108 of 142 eligible control newborns. Characteristics of included and nonincluded eligible children in each group were similar, except for a lower birth weight for included children of the group treated with lopinavir-ritonavir (eTable 1), and a higher proportion of recently born children, children of primiparous mothers, and those with late gestational age when pregnancy follow-up began in the control group (eTable 2). When compared with all remaining children of the eligible control group (n = 1294), included control children had a higher proportion of recently born children, a higher proportion of undetectable maternal viral load, and a higher birth weight (eTable 3). Matching failure for children entered into the process of cross-checking with CAH was mainly due to discrepancy between the name and surname given at birth and the names used later. Rate of matching failure was similar in treated and control groups.

Characteristics of mothers and children were available from the French Perinatal Cohort database. Clinical symptoms of children were retrospectively reassessed with the investigating clinician in light of possible adrenal dysfunction. The results of 17OHP assays on dried blood collected on filter paper were available from the CAH database. Plasma samples from frozen plasma banks at the virology laboratories responsible for neonatal HIV diagnosis were available for dehydroepiandrosterone-sulfate (DHEA-S) assays in a subgroup of children and additional 17OHP assays for cases with initially elevated results.

Testing for 17OHP levels using dried blood spots on filter paper collected at birth was centralized and performed using a solid-phase time-resolved fluoroimmunoassay method (AutoDELFIA Neonatal 17OHP Kit; Wallac Oy, Turku, Finland). Plasma 17OHP levels and DHEA-S levels were determined by radioimmunoassay as previously described (CISbio International, Gif-sur-Yvette, France, and Immunotech, Marseille, France).^19- 20 Plasma lopinavir assays involved high-pressure liquid chromatography.¹⁰

Statistical analysis was restricted to children born at term. We used a Spearman rank correlation coefficient (r) to test for correlation between levels of DHEA-S and 17OHP. χ² or Fisher exact tests were used as appropriate for categorical variables and t test or Wilcoxon or Kruskal-Wallis tests for continuous variables. P < .05 was the threshold for 2-sided statistical significance. SAS version 9.2 (SAS Institute, Cary, North Carolina) was used for analyses. The pharmacological data were analyzed using the nonlinear mixed-effect modeling program NONMEM (version VII, release 1) to calculate the analysis of variance and area under the curve for 0 through 24 hours (AUC_0-24).²¹

Between December 2004 and September 2008, 50 HIV-1 uninfected children received lopinavir-ritonavir just after birth, and 108 received standard prophylaxis: zidovudine alone (n = 100), zidovudine and lamivudine (n = 6), or zidovudine and nevirapine (n = 2). The median duration of postnatal lopinavir-ritonavir treatment was 30 days (interquartile range [IQR], 25-33 days), and the median daily dose was 43.8 mg/kg/d (IQR, 14.9-91.7 mg/kg/d). Lopinavir-ritonavir was administered with lamivudine, zidovudine, or both in 26 cases and was administered alone in 24 cases. The reasons for the particular choice of prophylactic regimen are not routinely collected. Retrospective analysis revealed inadequate treatment of mother (n = 29) or zidovudine resistance of the maternal virus (n = 7). The reasons were not provided or were unavailable for 14 cases.

Fewer mothers of lopinavir-ritonavir–treated children than control children originated from sub-Saharan Africa (68.3% vs 85%; P = .03), and more mothers began follow-up later in their maternity (18% began follow-up at 28 gestational weeks vs 3.6%; P = .007); as expected, fewer received antiretroviral therapy during pregnancy (90.5% vs 100%; P = .009) and had an undetectable viral load near delivery (57.9% vs 79.6%; P = .03) (Table 1). The proportion of mothers who received ritonavir-boosted protease inhibitor as the last antiretroviral therapy during pregnancy was higher in the treated group than in the control group (83.3% vs 62.4%; P = .01). In most cases (77.4%), the ritonavir-boosted protease inhibitor received by the mothers was lopinavir-ritonavir.

Among the 50 neonates treated with lopinavir-ritonavir, 7 had abnormally high 17OHP results from dried blood spots (>16.5 ng/mL at term or >23.1 ng/mL preterm) vs 0 of 108 controls (P < .001). For children born at term, 5 of 42 newborns treated with lopinavir-ritonavir vs 0 of 93 controls had values greater than 16.5 ng/mL (P = .003). Only children born at term who were also exposed to ritonavir-boosted protease inhibitor in utero had values greater than 16.5 ng/mL (14.3%, 5/35). The median 17OHP value for term newborns treated with lopinavir-ritonavir was 9.9 ng/mL (IQR, 3.9-14.1 ng/mL) vs 3.7 ng/mL (IQR, 2.6-5.3 ng/mL) in controls (P < .001) (Figure 1, Table 2, and Table 3). The difference observed in median 17OHP values between treated newborns and controls was higher in children also exposed in utero (11.5 ng/mL vs 3.7 ng/mL; P < .001) than not exposed in utero (6.9 ng/mL vs 3.3 ng/mL; P = .03) (Table 3). The 17OHP level was also associated with maternal origin and gestational age when follow-up began and at delivery (Table 2).

Frozen samples for DHEA-S determination were available from 20 newborns (18 at term) treated with lopinavir-ritonavir and 23 (17 at term) of the control group. The characteristics of children tested were no different from those of untested children. Age at sampling was similar in both groups (4 days of life; range, 2-6 days). The median DHEA-S values for children born at term were 9242 ng/mL (IQR, 1347-25 986 ng/mL) for the treated group vs 484 ng/mL (IQR, 218-1308 ng/mL) for the controls (mean [SD] values for 2-6 days of life: girls at term, 680 [650] ng/mL; boys at term, 820 [640] ng/mL²⁰) (Figure 2 and Table 3). DHEA-S and 17OHP values for individual children were strongly positively correlated (n = 35, r = 0.53; 95% confidence interval [CI], 0.23-0.73; P = .001). Consistent with the findings for 17OHP, the DHEA-S values were significantly higher only in cases also exposed in utero to ritonavir-boosted protease inhibitor.

For 6 of 7 newborns with abnormal screening results, the second test either with dried blood (n = 2) or plasma (n = 4) on day 9 (n = 4) or day 32 (n = 1) while still receiving lopinavir-ritonavir treatment indicated normal 17OHP concentrations. One child was lost to follow-up and could not be retested.

In addition to the screening control procedure, hormonal assays (17OHP and DHEA-S) were performed retrospectively using frozen samples from the virology centers. Serum 17OHP was assayed in samples collected at 1 month (1 patient), 6 months (2 patients), and 7 months (1 patient) of life from full-term newborns treated with lopinavir-ritonavir (and thus after completion of lopinavir-ritonavir treatment). In all 4 cases, the 17OHP levels were normal despite a high 17OHP value at birth for 1 of them. DHEA-S levels after lopinavir-ritonavir completion were normal in 6 cases; in a seventh case, the DHEA-S level was extremely high on day 4 (41 times the normal value), and substantially lower at 1 month, although it remained above normal (2.3 times the normal value).

The median prescribed dose of lopinavir (under the responsibility of individual clinicians) was 43.8 mg/kg per day. Two children (premature twins) received 91.7 mg/kg/d for 3 days after birth (this dose is almost 3 times the recommended dose of 30 mg/kg/d for newborns). Plasma concentrations of lopinavir (but not ritonavir) were available for 36 newborns, at a mean (SD) age of 12.8 (12.2) days. The mean (SD) delay between drug intake and lopinavir-ritonavir sampling was 6.2 (4.4) hours. No relationship was found between 17OHP level and the plasma concentration of lopinavir: tests for correlation between 17OHP levels and AUC_0-24 (r = 0.047, P = .78) and the analysis of variance between normal or abnormal levels of 17OHP and AUC_0-24 (P = .50) were not significant.

Retrospective review of all clinical symptoms reported in children treated with lopinavir-ritonavir revealed 17 hospitalizations for 17 children during the neonatal period for suspected maternal-fetal bacterial infection (n = 4), prematurity (n = 4), transient respiratory distress (n = 3), acute fetal distress syndrome (n = 1), poor social context precluding safe neonatal care and optimal antiretroviral treatment at home (n = 2), or various reasons linked to birth (n = 3). Three of these children presented with clinical and biological symptoms consistent with adrenal insufficiency (eTable 4). All 3 were premature (30, 34, and 34 weeks of gestation; 2 of them twins) and were exposed in utero to a lopinavir-ritonavir–based combination: lopinavir-ritonavir was administered from birth in all 3 cases. Lopinavir-ritonavir was associated with zidovudine and lamivudine for 1 child and was given alone for 2. For the twins, the dose of lopinavir given was above recommendations (3 times higher), with a residual plasma concentration that was double that expected.

Severe metabolic abnormalities were detected 24 hours, 72 hours, and 96 hours, respectively, after the start of treatment. All 3 children had profound hyponatremia with salt loss and hyperkalemia. Electrolytic disorders were severe enough to induce bradyarrhythmia and cardiogenic shock in 1 child. No hypoglycemia was detected.

In all cases, treatment was required to normalize the electrolytic disturbances: sodium chloride supplementation was administered in all cases, with 9α-fludrocortisone and hydrocortisone succinate in 1 case and sodium polystyrene sulfonate in 1 case. In all 3 cases, hormonal disturbances were discovered retrospectively: 17OHP levels in dried blood spots were in the normal range for 2 newborns in contrast with very high DHEA-S levels. 17OHP level was clearly in the pathological range in the third newborn. In all cases, the metabolic perturbations resolved within a few days of the completion of lopinavir-ritonavir treatment.

In this cohort of HIV-1–uninfected newborns of HIV-1 infected mothers, we found the postnatal treatment with lopinavir-ritonavir was associated with transiently increased 17OHP levels. 17OHP was assayed as part of the national screening program for CAH using dried blood spots collected between 2 and 5 days after birth. Complementary retrospective assays of DHEA-S, using frozen plasma samples, revealed extremely high values, up to 64 times normal levels, during the treatment period (<28 days). The 17OHP and DHEA-S values were strongly positively correlated, although a few children with normal 17OHP results had abnormally high DHEA-S concentrations. Most newborns were asymptomatic during the short period of treatment (4 weeks). However, we retrospectively identified clinical and biological features compatible with a transient but potentially life-threatening adrenal insufficiency in 3 preterm newborns.

The incidence of transient abnormal 17OHP levels, as assessed from the screening program in the general French population, is around 0.03%.²² In contrast, 7 of 50 newborns (14%) receiving lopinavir-ritonavir in our cohort had abnormal 17OHP values. Moreover, the difference relative to a control group of infants born to HIV-1–infected mothers receiving a standard postnatal zidovudine treatment was statistically significant. Despite reportedly weak transplacental transfer of lopinavir and ritonavir,^23- 24 we found that 17OHP levels were highest among infants exposed to lopinavir-ritonavir both during pregnancy and postnatally. Nevertheless, the small number of women not receiving ritonavir-boosted protease inhibitor limits the power of this study to detect a possible small association with postnatal exposure alone. Although these findings cannot establish a causal relationship, further studies are needed to test the hypothesis of whether lopinavir-ritonavir may act as an inhibitor of adrenal steroid synthesis in fetuses and newborns.

Drug-induced adrenal insufficiency is a well-known phenomenon in adults²⁵ but to our knowledge has not previously been described in neonates. Several drugs act on adrenal enzymes, primarily but not only those belonging to the p450 cytochrome (CYP) family. Concerning adrenal cytochrome enzymes, drugs can either inhibit cortisol synthesis (CYP19A1, CYP11A1, CYP11B1) or activate cortisol metabolism (CYP2B1, CYP2B2, CYP3A4). The fixed-dose combination lopinavir-ritonavir used in this cohort includes ritonavir, a potent inducer or inhibitor of several p450 cytochromes.⁸ This secondary effect led to its use only as a pharmacological enhancer to “boost” the antiviral effects of other protease inhibitors, notably lopinavir, a substantially less potent cytochrome disruptor. Interaction between ritonavir and adrenal enzymes has been described,²⁶ but with the opposite effect to what we observed here: the inhibition of CYP3A4 involved in glucocorticoid metabolism leads to deleterious interactions between ritonavir and synthetic steroids, with iatrogenic Cushing syndrome.²⁶

The absence of any reported adrenal insufficiency associated with long-term treatment with lopinavir-ritonavir—either in adults or children—suggests the possibility of a specific susceptibility during the neonatal period. Severe clinical symptoms were only observed in preterm newborns, reinforcing this notion. A relative adrenal insufficiency during the neonatal period transiently worsened by cytochrome inhibition is thus the main hypothesis.^27- 30 Abnormally high 17OHP and DHEA-S titers are compatible with a 21 hydroxylase deficiency.³¹ This enzyme (CYP21) is a cytochrome and therefore a potential target for ritonavir inhibition. However, the enzymatic activity—induction, inhibition, or both—of lopinavir-ritonavir on the various newborn adrenal cytochromes remains to be evaluated precisely in vitro, as does the potential role of geographical gene polymorphisms.³²

The use of lopinavir-ritonavir in newborns is recent but increasing.¹⁵ Although data concerning newborn tolerance are still limited, asymptomatic grade 3 serum hyponatremia and hyperkalemia disturbances in 2 newborns were noted in the first pharmacology study.⁹ These unusual biological disturbances in children born to HIV-infected mothers may have been linked to adrenal dysfunction. Electrolyte analyses were not recorded in the routine follow-up of the newborns in our study, and therefore, we cannot exclude occult electrolyte disequilibrium. The cardiac rhythm abnormalities and cardiogenic shock observed in 1 premature infant were considered probably related to electrolyte disturbances rather than direct cardiac toxicity.

The triad of “shock—hyponatremia with salt loss and renal insufficiency—hyperkalemia” is the classical presentation of the early and severe form of adrenal congenital deficiency.³³ Specific cardiac toxicity of lopinavir has been discussed in case reports,^13- 14 but potassium determinations were not reported. Cardiac dysfunction was also noted in 6 of the 10 cases included in the recent FDA warning.¹⁵ The authors considered possible specific lopinavir toxicity but also the possibility of a toxic effect of the propylene glycol and ethanol excipients. Four of those 10 children had hyperkalemia, and it is possible that some of them also had cardiac dysfunction induced by hyperkalemia and salt loss–induced dehydration.

There is not necessarily a single mechanism of the apparent toxicity of lopinavir-ritonavir in newborns. Although the drug dose administered to some newborns in our study was higher than recommended, there was no clear link between circulating concentrations of 17OHP and lopinavir. An evaluation of ritonavir titers would be interesting, although cytochrome inhibitory or induction effects can be observed at very low concentrations. Further studies are needed to investigate in more detail both the consequences of lopinavir and its booster ritonavir on adrenal function and the mechanism of cardiac toxicity in premature newborns.

In summary, our findings of the association between lopinavir-ritonavir and transient adrenal dysfunction in HIV-1 uninfected newborns suggest that lopinavir-ritonavir and more generally ritonavir boosting should be used with caution, if at all, in premature infants, and if this drug regimen is administered to full-term infants, it should be used under electrolyte monitoring. Whether more prolonged exposure of HIV-1-infected or uninfected infants via breast milk³⁴ is associated with endocrine disruption should be carefully investigated, and the apparent risk associated with prenatal ritonavir exposure also merits further evaluation.