Anastrozole vs Letrozole to Augment Height in Pubertal Males With Idiopathic Short Stature: A 3-Year Randomized Trial

Abstract

Context

Insufficient efficacy and safety data for off-label use of aromatase inhibitors to augment height in boys with short stature.

Objective

To compare anastrozole and letrozole in treatment of idiopathic short stature in pubertal boys.

Design

Open-label trial with 2 treatment arms.

Setting

Pediatric Endocrine Clinic at Stanford.

Participants

A total of 79 pubertal males ≥10 years with bone age (BA) ≤ 14 years, predicted adult height (PAH) < 5th percentile or >10 cm below mid-parental height.

Intervention

Anastrozole 1.0 mg or letrozole 2.5 mg daily for up to 3 years.

Main Outcome Measures

Annual hormone levels and growth parameters during treatment and a year posttherapy; annual BA and PAH (primary outcome measure); spine x-rays and dual energy X-ray absorptiometry at baseline and 2 years.

Results

Compared with anastrozole (n = 35), letrozole (n = 30) resulted in higher testosterone levels, lower estradiol and IGF-1 levels, and slower growth velocity and BA advance. The PAH increase observed at year 1 in both groups did not persist at years 2 and 3. Change in PAH from baseline was not different between treatment groups. In groups combined, PAH gain over 3 years vs baseline was +1.3 cm (P = .043) in linear mixed models.

Conclusion

Letrozole caused greater deviations than anastrozole in hormone levels, growth velocity, and BA advancement, but no group differences in PAH or side effects were found. Change in PAH after 2 to 3 years of treatment was minimal. The efficacy of AI as monotherapy for height augmentation in pubertal boys with idiopathic short stature may be limited, and safety remains an issue.

Evaluation and treatment for short stature is a common reason for referral to pediatric endocrinology clinics. Although a pathologic cause for short stature is not often identified, reduced height continues to cause significant psychological and social stress for both parents and children [1]. Human GH (hGH) can be used in the treatment of short stature [2, 3]. However, it is an expensive medication with at most modest efficacy in adding to final height in children who are not GH deficient [4]. Of children with short stature, approximately 60% to 80% of cases can be classified as having idiopathic short stature (ISS) where no pathology or alternative diagnosis can be found [3, 5].

Aromatase inhibitors (AIs) have been proposed as an alternative to hGH or in combination with hGH to augment height in short pubertal males with or without classic GH deficiency (GHD) [6]. Aromatase is a cytochrome P450 enzyme that catalyzes the formation of C18 estrogens (estrone and estradiol [E2]) from C19 androgens (androstenedione and testosterone) [7]. AIs suppress the conversion of testosterone to E2, the hormone that regulates the maturation and closure of the growth plates during puberty. In theory, a reduction in estrogen production during the pubertal growth period would permit a longer interval for growth and result in net gains in final adult height. Anastrozole and letrozole are third-generation nonsteroidal AIs that bind to the heme group at the active site of the aromatase enzyme and block estrogen formation. AIs are Food and Drug Administration-approved for adjuvant use in postmenopausal women with breast cancer [8-11] and have been used off-label to treat hypogonadism in males, as they have been shown to increase testosterone concentrations through the elevation of gonadotropins because of estrogen suppression [12, 13].

In the pediatric population, AIs have been used to suppress estrogen concentrations in McCune-Albright syndrome, familial male-limited gonadotropin-independent precocious puberty (testotoxicosis), aromatase excess syndrome, and gynecomastia, among other conditions [6], and to initiate or accelerate puberty in adolescent males with constitutional delay of growth and puberty [14]. Given the potential for AIs to slow maturation of the epiphyses and allow more time for linear growth, they have been used off-label to treat short stature in adolescent males [15-17]. One of the advantages of using AIs to reduce estrogen levels is their ability to do so without reducing testosterone levels, allowing for normal progression of puberty in males. Other potential treatments to prolong the pubertal growth period in males, such as GnRH analog therapy, suppress both testosterone and estrogen production, resulting in an undesirable delay in masculinization [18].

Multiple prospective trials and retrospective chart reviews have evaluated the efficacy of AIs to increase height in pubertal boys with various classifications of short stature [19-28]. In general, studies have reported an increase in predicted adult height (PAH) during therapy, or at end of therapy, compared with baseline measures or control groups. However, demonstrated gains in measured final or near-final height have been reported in a limited number of trials, and the literature is essentially lacking in definitive reports. A 2015 Cochrane review concluded there were insufficient data to determine either the efficacy or the safety of AIs in the treatment of ISS [29], and studies since then have done little to define the efficacy of AIs in this context. The review found no consensus on the optimal drug or when to initiate and terminate therapy and noted that further investigation is needed to determine both the short- and long-term adverse effects of AIs on gonadal function and bone health, as assessed by vertebral morphology and bone mineral density. Despite these shortcomings in the literature, clinical treatment with AIs appears to be common.

To date, no randomized controlled trials (RCTs) have directly compared the more potent AI letrozole with less potent anastrozole, both commonly used off-label for height augmentation. The aims of this 3-year open-label randomized treatment trial were to compare the efficacy and safety of anastrozole and letrozole in pubertal boys with ISS. We hypothesized that both AIs would increase PAH (and ultimately final height) compared with pretreatment PAH. We wanted to determine if the less potent anastrozole was as effective as letrozole but possibly safer. First-year data from 39 subjects were previously published by our group [28] and suggested a gain in PAH from anastrozole therapy only. In this paper, we report the completed 3 years of treatment data on the full cohort with ISS who received anastrozole or letrozole for 1 to 3 years as monotherapy, plus data from a 1-year posttreatment follow-up. Our report here is subject to some of the flaws in the existing literature—no untreated control group, potential selective loss from subject dropout, and lack of true final height data—but we believe the trial is informative nonetheless. Notable features of this prospective randomized trial include comprehensive head-to-head comparison of anastrozole and letrozole, no confounding therapies (eg, hGH and testosterone), and posttreatment data.

Materials and Methods

Study Design

This was an open-label, randomized trial in otherwise healthy pubertal boys with ISS. Study subjects were randomized into 1 of 2 treatment arms (anastrozole 1 mg by mouth daily or letrozole 2.5 mg by mouth daily). Randomization was 1:1 using block randomization and variable block sizes. Treatment duration was arbitrarily limited to 3 years. Requests to continue therapy beyond 3 years were refused. Subjects and families were encouraged to remain in the trial for at least 2 years. Criteria to discontinue therapy in the first 2 years included significant and/or persistent side effects, reduction in growth velocity to less than 3 cm/y while bone age (BA) was still mid-pubertal (< 15 y), or reduction in growth velocity to < 2 cm/y and advance in BA indicating the impending termination of pubertal growth.

The primary outcome measure was change in PAH from baseline PAH for each AI. Power was calculated at 80% with a sample size of 40 and an effect size of 0.65. Secondary outcome measures were hormone levels, height, growth velocity, BA advance, dual energy X-ray absorptiometry (DXA) Z-scores, and incidence of spine x-ray anomalies. Subjects were seen every 6 months throughout therapy for laboratory evaluation and physical examination, and then at a year posttherapy. BAs were obtained annually during therapy. Spine x-rays and whole-body and lumbar DXAs were obtained at baseline and 2 years.

Participants and Setting

Short pubertal males seen in the Pediatric Endocrine Clinic at Lucile Packard Children's Hospital at Stanford were approached to participate. Growth criteria for inclusion were males aged 10 years or older with a BA of 14 years or younger, and PAH (based on BA) less than the fifth percentile or more than 10 cm below mid-parental height (MPH). The primary purpose of the trial was to treat non-GHD pubertal boys who had a predicted final height more than 10 cm below MPH. Because of the subjectivity of BA interpretation and inaccuracy of final height prediction, an additional criterion of current height < 5th percentile was used. This definition of ISS is different from the conventional 1.2 percentile height threshold proposed by Cohen et al (3). Boys with PAH above the 25th percentile were prohibited from participating in the trial even if PAH was >10 cm below MPH. Most subjects were routine short stature referrals to the pediatric endocrine clinic or were follow-up patients for short stature who had entered puberty. Some subjects were specifically referred by local pediatricians or by other pediatric endocrinologists familiar with the trial and its inclusion criteria. All potential subjects were examined and screened by the senior investigator (E.K.N.).

Pubertal criteria for enrollment required all of the following: genital Tanner stage ≥ 2, testicular volume ≥ 4 mL, LH > 0.3 mIU/mL, and testosterone (T) > 15 ng/dL. ISS was defined as normal IGF-1 for age and Tanner stage, normal growth velocity for age and Tanner stage, and no other causes of short stature. Boys with known causes of short stature, including skeletal dysplasias and other genetic disorders associated with short stature, were excluded from the study. Boys with clinical evidence of gonadal failure were excluded from the study. Any subject with FSH > 20 mIU/mL was automatically excluded, and additional clinical context was provided by inhibin B, testosterone, LH, and testicular volume. Subjects were screened for hypothyroidism and celiac disease by referring physicians or by our clinic, and, as warranted, for other potential causes of short stature.

Boys with subnormal IGF-1 level for age and Tanner stage were excluded from the trial. Normal ranges for serum IGF-1 were reported by the reference laboratory (Esoterix) for a combination of age and Tanner stage, so no single cutoff for IGF-1 was used. Growth hormone stimulation tests were not performed. Boys who were already undergoing hGH therapy for ISS initially were allowed to enroll, but after the trial started, we decided to exclude from analysis any subjects who had received hGH for ISS. hGH was not initiated in any subjects during therapy. Enrollment began in 2010 and ended in 2015. Parents of all participants signed the informed consent approved by the Stanford University Human Subjects Committee, and all boys signed an assent. On reaching age 18 years, subjects were asked to sign the informed consent.

Laboratory Evaluation

Study laboratory blood values were drawn in the morning and were performed at Esoterix Inc (Calabasas Hills, CA) (with the exception of inhibin B). Serum T, DHT, androstenedione, E2, and estrone (E1) were assayed by high-pressure liquid chromatography followed by tandem mass spectrometry, with a lower limit of detection of 3 ng/dL (0.1 pmol/L) for T, 1 pg/mL (3.67 pmol/L) for E2, and 2.5 pg/mL for E1. LH and FSH were assayed by electrochemiluminescence with lower limits of 0.02 mIU/mL. IGF binding protein 3 and IGF-1 tests were performed by radioimmunoassay. Inhibin B was assayed by manual immunoenzymatic assay (Beckman Coulter Inc.) at the Mayo Clinic. In a subset of subjects, we obtained a prostate-specific antigen level at the year 2 visit (results not included). Vitamin D25 was measured at baseline (results not included). (Intervention was initiated for levels <20 ng/mL.) All participants were encouraged to take recommended daily allowances of vitamin D and calcium. Mean levels of T, E2, and IGF-1 are also stated in SI units in the Results section.

Imaging

A BA radiograph was obtained at baseline and annually thereafter during therapy and interpreted according to the method of Greulich and Pyle [30]. Predicted adult height was calculated by the Bayley-Pinneau method [31]. For study inclusion, BA readings performed by hospital radiologists were accepted. For data analysis and calculation of PAH, a single pediatric endocrinologist (S.R.), blinded to the randomization, reread all x-rays, including baseline films, to determine BAs.

Bone mineral density (BMD) by DXA (Hologic) and lateral thoracolumbar spine radiographs were obtained at baseline and 2 years. Whole-body and lumbar images were reported as bone mineral concentration and age-adjusted BMD as performed by the manufacturer’s software. Spine images were evaluated for the presence of vertebral anomalies and compression fractures.

Data Analysis

Descriptive laboratory measures over time are presented as means with SD, with key outcome measures (total T, E2, IGF-1) additionally displayed as boxplots by arm. Laboratory measures during AI therapy were analyzed if subjects were currently treated. Descriptive auxologic measures were similarly presented as means with SD, with key outcomes height, growth velocity, BA, and PAH displayed as boxplots by arm. Auxologic measures were analyzed regardless of current treatment status and adherence.

The per-protocol analysis excluded subjects failing to complete the first year of therapy. Mixed effects linear regression was used to model each key outcome as a function of treatment arm, indicators for follow-up timepoints, and their interaction for assessing effect modification of annual change by arm. Within-subject correlation was accounted for by a random effect. Estimated changes since baseline, with 95% CIs, were presented by arm if effect modification was deemed significant at the 0.05 alpha level, and as an overall effect otherwise. All analyses were performed in the R statistical computing framework, version 4.0 [32].

Baseline auxologic and hormonal measures were summarized by treatment arm and compared by t-test for continuous characteristics. DXA Z-scores over time were visualized as means with SD and analyzed (year 2 vs baseline) by linear mixed model. Comparisons between anastrozole and letrozole treatment groups at the posttreatment follow-up visit were performed by 2-tailed unpaired t-test. Adverse events were listed by arm. Spine film was assessed qualitatively.

Results

Study Flow

Seventy-nine boys were consented for the trial and randomized to anastrozole (A, n = 40) or letrozole (L, n = 39). Four subjects, all in the L group, failed to start therapy because of medication not being covered by insurance. (We subsequently arranged with the hospital pharmacy to provide medications at reduced cost.) Four subjects (1 A, 3 L) were nonadherent during the first year (assessed by subject reports and preliminary laboratory confirmation) and were excluded from further study monitoring and from final analysis. We excluded 6 boys (4 A, 2 L) from analysis because of concomitant use of hGH.

Sixty-five subjects (35 A, 30 L) with 1 to 3 years of treatment and adequate adherence became the cohort for analysis. Of the 65 study subjects, 4 withdrew during the second year of treatment (1 A, 3 L) because of concerns about side effects possibly related to treatment (see Safety section). None of the subjects met our established growth velocity criteria for discontinuation in the first 2 years. Sixty-one boys (34 A, 27 L) completed year 2.

An additional 12 subjects (8 A, 4 L) discontinued treatment during year 3, primarily because of subject disinterest/lack of compliance or for family concerns about slowing growth or limited growth potential. One subject in the A group withdrew following acute sacroiliitis. The remaining 49 subjects (26 A, 23 L) completed 3 years of treatment. Subjects returned for a posttreatment follow-up 1 year after drug discontinuation (average interval, 1.1 years; range, 6-18 months). Hormonal analysis and physical examinations were performed posttreatment, but not bone ages.

Baseline Characteristics

Boys were mid-pubertal by multiple parameters. Overall, mean age was 13.9 years, BA 13.0 years, testicular volume 7.7 mL, growth velocity 6.6 cm/y, and mean height 147.5 cm with mean PAH 167.2 cm and mid-parental height 174.0 cm. Mean serum testosterone was 185.7 (range, 21-537) ng/dL (6.44 pmol/L), E2 4.8 (range, 1.0-20) pg/mL (17.62 pmol/L), LH 2.0 (range, 0.3-5.8) mIU/mL, FSH 2.9 (range, 0.7-8.6) mIU/mL, and inhibin B 210 (range, 96-398) pg/mL). Mean IGF-1 was 306.6 ng/mL (40.07 nmol/L). IGF-1 ranges at baseline were 125 to 398 ng/mL for Tanner 2 boys, 161 to 552 ng/mL for Tanner 3, and 399 to 638 ng/mL for Tanner 4. There were no significant differences at baseline between groups for any measure.

Auxologic Findings During Treatment

Heights were obtained every 6 months and BAs every year during the 3-year treatment period. Descriptive auxologic data at baseline and 1, 2, and 3 years after initiation of therapy are summarized in Table 1. Graphic representations of height, growth velocity, BA, and PAH are shown in Figs. 1 and 2.

Growth in the first year slightly exceeded the baseline growth rate but then slowed markedly in the second and third years in both treatment groups (Fig. 1 A-B). Mean cumulative changes in height were 13.1 ± 4.3 cm A vs 12.5 ± 3.2 cm L at the end of 2 years and 18.2 ± 4.2 cm A vs 16.0 ± 3.4 cm L at the end of 3 years, but the differences did not reach statistical significance in linear mixed models (P interaction .077). Mean growth velocity stayed above 7 cm/y in the first year of therapy in both groups but declined to 4.9 ± 2.4 (A) vs 4.0 ± 1.4 cm/y (L) in year 2 and 3.5 ± 1.7 (A) vs 3.4 ± 1.3 cm/y (L) in year 3 (P < .001 vs baseline growth velocity [GV]). Growth velocity was significantly slower during letrozole treatment compared with anastrozole over the entire treatment duration (P interaction .039).

BA advanced more slowly in the letrozole group as well (Fig. 2A). BA advances in years 1, 2, and 3 for anastrozole and letrozole, respectively, were 1.0, 1.1, and .7 years vs 0.9, 0.8, and 0.5 years (P < .001 A vs L). PAHs were calculated from the yearly BAs (Fig. 2B and Table 1). For year 1, mean PAH showed an increase to 169.5 ± 8.1 cm for the anastrozole group and 168.7 ± 5.5 cm for the letrozole group. This increase in PAH was statistically significant in both groups when compared to baseline (P = .001 A and P = .05 L). At year 2, mean PAH was 168.0 ± 7.0 cm (A) and 167.8 ± 5.6 cm (L), neither significantly higher than baseline (P = .292 and P = .752, respectively). At year 3 of the study, mean PAH was 169.4 ± 5.1 cm for the anastrozole group and 167.8 cm (±5.8) for the letrozole group, neither statistically greater than baseline PAHs in linear mixed models, although the increase associated with anastrozole neared significance (P = .062 and P = .344, respectively). Overall, net gains in PAH at the end of 3 years of treatment vs baseline were +1.7 and +0.8 cm for the anastrozole and letrozole groups, respectively, in the linear analysis. Change in PAH at year 3 reached statistical significance only if the groups were combined (+1.3 cm, P = .043).

Because of the number of subjects discontinuing in the third year and the potential for bias from selective loss, we performed an additional analysis combining the year 3 PAHs with the year 2 PAHs when the latter was the final treatment BA obtained. The mean PAH based in this manner on the last BA for each subject was 168.2 cm (A) vs 168.1 (L). Mean cumulative change in PAH at the end of treatment (2 or 3 years) was +1.0 cm for anastrozole, + 0.5 cm for letrozole, and +0.7 cm in the groups combined (n = 61).

Hormonal Analysis During Treatment

Descriptive laboratory data at baseline and 1, 2, and 3 years after initiation of therapy are shown in Table 2. Graphic representations of key hormones testosterone, E2, and IGF-1 are shown in Fig. 3. Treatment with AIs resulted in an increase in androgens, with testosterone rising more than 3-fold from baseline to year 1 in both groups (P < .001). Testosterone levels were higher in the letrozole group compared to the anastrozole group throughout all 3 years of treatment (P interaction <.001). Mean testosterone at year 1 on anastrozole was 552 ng/dL (19.14 nmol/L), at the upper end of the normal range for Tanner 4 males (200-620 ng/dL), whereas the mean on letrozole was 982 ng/dL (34.08 nmol/L), well above the mean and range for Tanner 4 and at the high end of normal for Tanner 5 (350-970 ng/dL). Only 1 boy on anastrozole had T > 1000 at the year 1 visit, whereas 14 boys on letrozole (48%) had T > 1000 at year 1, 4 of them with a T level >1200 ng/dL.

The anticipated suppression in serum E2 concentrations was observed in both treatment groups (Table 2 and Fig. 3B), but in the anastrozole group, the suppression was only relative to the rise that would have been expected during a normal (untreated) pubertal progression. In this regard, the 2 treatment groups were clearly different (A vs L, P < .001). E2 levels decreased in the letrozole group from 4.9 pg/mL at baseline to 2.6 at year 1 (17.99 to 13.22 pmol/L) (P < .001) and remained similarly low throughout therapy. E1 concentrations were lower during all 3 years of treatment in both groups vs baseline. Gonadotropins (LH and FSH) rose during treatment and were higher in the letrozole group. The T/E2 ratio rose approximately 2- to 3-fold in the A group but more than 10-fold in the L group because of the combination of relatively higher T concentrations and lower E2 concentrations associated with letrozole. Inhibin B levels, which were used as an indirect measure of testicular function, were modestly higher during treatment compared with baseline in both groups, as would be expected with pubertal progression.

Growth factors obtained during treatment showed a decrease in IGF-1 in the letrozole group (Table 2 and Fig. 3C). The decrease in IGF-1 levels from baseline was significant in all 3 years of treatment for the letrozole group (P < .001 in years 1 and 2 and P = .002 for year 3). In contrast, IGF-1 remained unchanged during therapy with anastrozole. IGF-1 levels were consistently higher during anastrozole treatment compared with letrozole (A vs L, P < .001). Nevertheless, mean IGF-1 values in both groups remained well within the reported normal range for Tanner 4-5 males (161-467 ng/mL, mean 290). The lowest IGF-1 levels during therapy (lowest was 125 ng/mL, 16.34 nmol/L) occurred predominantly but not exclusively in the letrozole cohort. No subjects were started on hGH for declining IGF-1.

Posttreatment Analysis

Heights in each group at follow-up approximately 1 year after cessation of therapy were 164.3 ± 6.3 cm (A) and 163.0 ± 5.6 cm (L), not significantly different. Bone age was not obtained at that visit. Given that the mean BAs were in excess of 15 years 1 year earlier, these follow-up heights can be construed as near-final heights. Growth velocity in the posttreatment period (average interval, 1.1 year) was not different than GV during the third year of treatment (Fig. 1B). GV was 2.1 ± 1.0 cm/y in the anastrozole group and 3.2 ± 2.1 cm/y in the letrozole group, not statistically different.

Posttreatment laboratory data were collected (12 A and 11 L) at the follow-up visit. The results (Fig. 3 and Table 2) generally demonstrated posttreatment normalization of hormone concentrations: decreased androgens and gonadotropins, increased estrogens, and increased growth factor levels compared with treatment levels. From the year 3 visit to the posttreatment visit, mean testosterone declined from 680.1 to 428.5 ng/dL (23.58 to 14.86 nmol/L) in the A group and from 882.3 to 370.6 ng/dL (30.59 to 12.85 nmol/L) in the L cohort. Although the means were within normal ranges (Tanner 4 200-620 ng/dL, Tanner 5 350-970 ng/dL), 9 of the 23 boys (4 A, 5 L) exhibited levels <350 ng/dL (12.13 nmol/L) at posttreatment follow-up. Four (17%) of those subjects had T levels <200 (6.93 nmol/L) (2 A, 2 L), the lowest being 173 ng/dL (6.00 nmol/L). LH in these 4 boys ranged from 1.1 to 3.3 mIU/mL, within the normal range, and FSH and inhibin B levels were normal as well. Mean FSH in the letrozole cohort trended higher than in the anastrozole cohort (4.4 vs 2.5 mIU/mL).

E2 rose from 6.2 to 17.0 pg/mL (22.76 to 62.41 pmol/L) in the A group and from 2.7 to 12.3 pg/mL (9.1 to 45.15 pmol/L) in the L group, levels within the normal range for late pubertal males. With the drop in T and rise in E2, T/E2 ratio decreased markedly from 1219 to 255.1 (A) and from 5140 to 334.8 (L). Estrone similarly rose at posttreatment follow-up, gonadotropins declined, and inhibin B reverted to baseline levels. Mean IGF-1 in the A group was 298.7 ng/mL (39.04 nmol/L) at year 3 and 330.8 (43.24 nmol/L) at follow-up. IGF-1 rose more noticeably after discontinuation of letrozole, from 226.1 to 362.3 ng/mL (29.55 to 47.35 nmol/L). Mean IGF-1 posttreatment was within the stated normal range for late pubertal males (161-467 ng/mL).

Safety

During treatment, all subjects were screened for potential side effects and adverse events. The adverse events are summarized in Table 3. Given the more pronounced increase in androgens than expected in normal puberty, the boys and parents were specifically asked questions regarding perceived changes in behavior, hair loss, and acne. Twelve boys overall reported possible side effects in these categories, but it is unclear whether the frequency was greater than in normal puberty. Joint pain and fractures were assessed at every visit. Joint pain related to sports was common. Five subjects had upper extremity fractures (elbow and wrist) related to sports, and 1 (on letrozole) incurred a hip avulsion fracture playing soccer. No boys reported vertebral fractures or persistent back pain. Scoliosis was reported in 3 subjects. Acute sacroiliitis was diagnosed in a subject on anastrozole. One boy, also on anastrozole, reported a transient neurologic event of slurred speech and difficulty swallowing, without any etiology identified. Anastrozole was discontinued in both of these subjects.

Lateral x-rays of the thoracolumbar spine were performed at both baseline and 2 years of treatment in 39 subjects (21 A, 18 L). None reported lower back pain at the time of examination. No compression fractures were observed in any of the subjects. Vertebral abnormalities were found at both baseline and 2-year follow-up in 3 subjects: biconcave vertebral body (A), spondylolysis (A), and irregular endplates (L). No de novo findings were reported at the 2-year examination in any subjects, although 2 boys (1 A, 1 L) were found to have anterior wedging on films performed off-protocol more than 4 years after treatment initiation.

DXA BMD Z-scores were obtained at both baseline and year 2 in 24 subjects (15 A, 9 L); Fig. 4 shows the whole-body and lumbar spine Z-scores. Mean lumbar spine Z-scores for anastrozole were −1.33 ± 0.81 at baseline and −1.54 ± 0.91 at year 2, whereas lumbar spine Z-scores for the letrozole group were −1.37 ± 0.89 at baseline and −1.94 ± 1.16 at year 2. The decrease in whole-body and lumbar Z-scores from baseline to 2 years was significant in both groups. Although the magnitude of BMD decline appeared greater in the L group, it was not statistically different from the A group. Hip Z-scores were not available because there were no established standards.

Discussion

In the absence of an untreated control group, our study was designed to compare outcomes from anastrozole and letrozole treatment and to assess treated vs pretreatment PAH. Our prospective randomized comparison of the 2 agents appears to be the most detailed to date. Treatment with anastrozole and letrozole in our trial resulted in clear, predictable, and reversible perturbations of male pubertal hormone levels. Both agents resulted in markedly diminished estrogen levels compared to normal puberty, a rise in gonadotropins, and augmented androgens. Letrozole is the more potent agent and caused notably exaggerated changes of hormone levels compared with anastrozole. E2 levels on letrozole were half those on anastrozole and T levels were nearly double the levels on anastrozole. IGF-1 remained at baseline levels during anastrozole therapy but diminished by one quarter or more on letrozole, aligning with the degree of estrogen suppression. Mean T, E2, and IGF-1 levels all returned to late pubertal normal ranges after discontinuation.

The lower E2 and IGF-1 concentrations accompanying letrozole therapy are presumably the cause of the reduced growth compared with anastrozole. The premise behind AI therapy is that a reduction in BA advancement from reduced estrogen activity might outweigh the potential reduction in GV. We did document a marked slowing of BA advancement, particularly with letrozole and particularly later in the treatment period. BA increased by 2.6 and 2.1 years over 3 years of A and L therapy, respectively. Growth was comparatively slower with letrozole; thus, the treatment groups were indistinguishable in terms of height or PAH as the study progressed. More importantly, the BA slowing seen during therapy did not translate into substantive increases in PAH from baseline with either AI. At the end of treatment, PAH remained at about 168 cm in both groups vs MPH of 174 cm. Both methods of PAH analysis (regression and descriptive) suggested that anastrozole therapy might result in a slightly greater PAH compared with letrozole, but without statistical significance. The appearance of gain in PAH after 3 years disappeared in the descriptive analysis when it was adjusted to eliminate potential bias introduced by selective dropout. The groups combined did demonstrate a small, barely significant gain in PAH at 3 years in the regression modeling.

The clinical consequence of AI therapy is delayed bone maturation in tandem with delayed growth (ie, the degree of growth retardation and reduction in bone maturation is linked). This truism applies to any manipulation of estrogen in childhood or adolescence [33], as in GnRH analog therapy for precocious puberty. There simply is no therapy-induced BA slowing without accompanying growth retardation during induced estrogen reduction in puberty. Whether concomitant hGH therapy could alter the equation and compensate for the growth inhibition caused by lower estrogen is unclear.

In regard to the main outcome measure of end-of-treatment PAH compared to pretreatment PAH, ours is a fundamentally negative study using either agent. Although it is possible that greater power from larger subject numbers might identify a more credible gain in PAH during therapy, or that treatment for a briefer duration or earlier in puberty might be more successful, the magnitude we observed in our 3-year protocol appears to be clinically insubstantial, even when the groups are combined. This result informs and somewhat contradicts the existing literature, which, with exceptions, has suggested a significant increase in PAH, near-final height, or final height [19-28, 34-39]. One differentiating factor may be in compensating for the potential problem of selective bias. We used both regression analysis and final treatment PAH (whether at 2 or 3 years), and both methods resulted in nonexistent to minimal gain in PAH.

The literature regarding AI for height in pubertal boys is inconsistent. The most influential reports are from the early, relatively small Finnish RCTs of letrozole [19-22] (in which final height data rendered some prior findings moot) and from the RCTs of anastrozole with hGH from Mauras et al in GHD and ISS [24, 25]. The latter studies yield limited conclusiveness regarding the benefit of AI alone but do argue for a benefit from the combination of hGH and AI. The Finnish studies imply that letrozole, which reduces IGF-1 to lower levels than anastrozole, as we show, is effective alone. It might be reasoned instead that treatment with letrozole rather than anastrozole would require the boost from hGH combination therapy to maintain growth. In any case, our study comes to a different conclusion altogether, showing that AIs are relatively ineffective in improving PAH or near-final height, even as IGF-1 levels remain normal.

We are aware of the limitations of our trial, which indeed are similar to design and implementation flaws in other AI trials. Most importantly, there was no untreated control group, which we considered but decided would be a serious impediment to recruiting or keeping subjects. Lack of an untreated control group is a common issue in the literature. One approach to this is delivering an alternative growth therapy to all subjects (eg, testosterone [19, 20] or hGH [24, 25]), as just described. We preferred to test AIs as monotherapy to avoid the confounding influence. (Furthermore, testosterone would provide little additional benefit to subjects already progressing through puberty, and hGH would be expensive, inconvenient, and likely not warranted in boys with normal IGF-1 levels.) We felt that the comparison of anastrozole and letrozole and comparison of treatment PAH with pretreatment PAH would provide sufficient, albeit indirect results. We also have not followed the boys to final height, although one could consider our year 3 visit (with mean BA of 15.3 years) as near final height, or furthermore the 1-year posttreatment follow-up (but no BA obtained).

Regarding safety issues, our primary focus during the trial was gonadal function and bone accrual. We were surprised by the paucity of comments from subjects or families regarding aggressive behavior, hair loss, or acne. There was no indication that any of those was more common during treatment with letrozole or in subjects with elevated testosterone levels. We expected differential subject reporting during A vs L therapy but discerned no such relationship. There was no apparent difference in significant side effects or in overall elective dropouts with A vs L. Interestingly, dropouts seemed to be more frequent in the letrozole arm in the first 2 years but more common in the anastrozole arm in year 3. Although letrozole results in more extreme hormonal distortion during treatment, both agents were associated with cautionary safety data both during and after therapy. We were surprised that letrozole was not associated with more side effects than anastrozole. (We did not monitor complete blood counts during therapy for erythrocytosis, which has been reported from markedly elevated testosterone levels [37].)

Based on animal data [6], there has been concern that dramatically altering the ratio of testosterone to E2 during this key period of gonadal development may be damaging to the reproductive system (see our prior report [28] for further discussion). Short-term effects of aromatase inhibition in men [40] have included diminished sexual function. However, there is no evidence of human testicular dysfunction caused by AIs [41]. In fact, AIs are widely used in male infertility regimens to augment the T/E2 ratio and improve fertility [42]. Nevertheless, long-term follow-up of adolescents treated with AIs is lacking. In our trial, testicular volume during and after therapy was comparable in the 2 groups. We used inhibin B as an indirect marker of testicular function, seeing the expected increase during therapy, presumably associated with the rise in gonadotropins, but there was no difference in inhibin B levels between groups during treatment (despite the major between-group differences in T and E2 levels) or after discontinuation. We also focused on posttreatment testosterone levels. Means were in the normal range, but several boys, on both therapies, exhibited borderline levels (slightly below 200 ng/dL). None had gonadotropin levels indicating either hyper- or hypogonadism. Long-term data should be gathered in these subjects and from other trials as the subjects move into adulthood.

Other long-term safety concerns for the use of AIs in adolescent males include possible adverse effects on bone density and morphometry because estrogen is an important modulator of bone accrual in both sexes. AI therapy in breast cancer is clearly associated with reductions in BMD [43]. In addition, loss of BMD occurs during AI therapy in elderly hypogonadal men [44]. In those conditions, unlike in adolescent males, posttherapy “catch-up” bone accrual is not possible. Longitudinal bone density studies by DXA after AI therapy in boys have been reassuring [24, 25, 45, 46]. A subset of our subjects had a DXA scan at baseline and 24 months into treatment. The relatively low scores at baseline can be attributed to small stature and bone size as well as to an element of pubertal delay. Although the magnitude of decline in lumbar BMD Z-score appeared greater in the letrozole group, the group difference was not significant. The decline during AI therapy is concerning but not unexpected, given the further lag in growth and bone accrual induced by the estrogen-lowering medication. Long-term DXA data are still warranted. Although some of our subjects reported fractures during treatment, they were sports-related injuries mostly of upper extremities, the hip avulsion during soccer being the exception. This age group is when most fractures occur in boys [47], and we believe these events were unlikely to be related to treatment. Because of reported abnormal spine morphology following prepubertal treatment with letrozole [23], we obtained lateral thoracolumbar x-rays at baseline and 24 months. We found no evidence of change in vertebral morphology, and there were no subject reports suggesting compression fractures.

In summary, we believe the trial has been helpful in comparing anastrozole and letrozole in more detail, in demonstrating the at-best limited efficacy of therapy using either medication, and in further elucidating some of the safety issues. No randomized AI trials have carried substantive numbers of subjects to final height. Although several studies have suggested gains in PAH or near-final height, the literature regarding AI therapy for height augmentation remains inconclusive. The current study corrects our own optimistic preliminary data and brings into question the magnitude of therapeutic benefit. In light of our demonstration of marginal treatment efficacy using either AI as monotherapy and the continued safety concerns, we are currently unenthusiastic about routine clinical use of aromatase inhibitors for height augmentation in pubertal boys.

Acknowledgments

The authors thank Victoria Ding, Nikta Forghani, Sydney Payne, Rajiv Kumar, Diane Stafford, and Laura Bachrach for their assistance. The authors are very grateful to all the referring physicians and especially to the study participants and their families.

Funding

W.Z. received a Stanford Endocrinology T32 NIH Training Grant and was the Pete and Arline Harman Fellow and the Rosa A. Wann and Marjorie Shannon Fellow of the Stanford Maternal and Child Health Research Institute.

Data Availability

Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Disclosures

Nothing to disclose.

Clinical Trial Information

Clinical trial number: NCT02137538.

References

Abbreviations

 
• A

  anastrozole

 
• AI

  aromatase inhibitor

 
• BA

  bone age

 
• BMD

  bone mineral density

 
• DXA

  dual energy X-ray absorptiometry

 
• E1

  estrone

 
• E2

  estradiol

 
• GHD

  GH deficiency

 
• GV

  growth velocity

 
• hGH

  human GH

 
• ISS

  idiopathic short stature

 
• L

  letrozole

 
• MPH

  mid-parental height

 
• PAH

  predicted adult height

 
• RCT

  randomized controlled trial

 
• T

  testosterone