Among children with the largest glands, in the highest decile, 92.2 percent did. The study, published in Frontiers in Endocrinology, reframes a routine anatomical measurement as a physiological checkpoint while simultaneously delivering a sobering message about how much that measurement can actually tell clinicians. Central precocious puberty results from premature activation of the hypothalamic-pituitary-gonadal axis and affects approximately one in 5,000 to 10,000 children worldwide, with a marked female predominance.

Early axis activation accelerates epiphyseal maturation, the process by which growth plates harden and close, ultimately compromising adult height and carrying psychosocial and metabolic consequences. The GnRH stimulation test remains the reference standard for confirming axis activation, but it requires serial venipuncture over 60 to 120 minutes, is costly, and carries a burden of anxiety for children and families. Basal luteinizing hormone and other biochemical surrogates have been explored, but their diagnostic accuracy varies considerably across assays and populations, leaving a reliable non-invasive triage tool an unmet clinical need.

The biological rationale for imaging the pituitary is well established. The adolescent pituitary gland undergoes dramatic physiological hypertrophy during normal puberty, a phenomenon first demonstrated on MRI by Elster and colleagues in 1990 and extended in large normative series by Kato and colleagues in 2002. Under hypothalamic stimulation, gonadotrophs, the specialized pituitary cells that produce luteinizing hormone, hypertrophy and proliferate, swelling the gland progressively with age and pubertal stage.

Magnetic resonance imaging provides reproducible, radiation-free quantification of pituitary volume, and several studies have reported larger volumes in children with central precocious puberty than in prepubertal controls. But previous diagnostic studies left three critical questions unanswered: how much of the apparent diagnostic signal merely reflects the fact that older children have both bigger glands and more mature axes, whether volume carries information beyond what age and insulin-like growth factor 1 already provide, and whether gland size functions as a threshold below which activation is physiologically improbable. The Wuhan team, whose investigators included Jianjian Cai, Ling Zhang, Weiyin Vivian Liu, Qin Liu, Jun Lu, and Dong Liu, retrospectively enrolled 896 children aged 3 to 12 who had undergone both diagnostic-quality pituitary MRI and complete GnRH stimulation testing between January 2018 and December 2023.

Children with organic central nervous system lesions, peripheral precocious puberty, hypopituitarism, or medications affecting the axis were excluded. Two blinded pediatric neuroradiologists, with 8 and 12 years of experience, manually delineated the anterior pituitary on 3D-CUBE T1 sequences, achieving an inter-reader reliability intraclass correlation above 0.90. The cohort of 575 girls and 321 boys was heavily skewed toward activation, with 553 children, or 61.7 percent, meeting the biochemical criterion.

Activated children were older, with a median age of 9.1 versus 7.3 years, and had substantially larger glands, with a median volume of 225.8 versus 165.1 cubic millimeters. The authors emphasize that axis activation, defined by the biochemical response to stimulation, is a physiological endpoint and not synonymous with the clinical diagnosis of precocious puberty; in boys, the gold-standard criterion for that diagnosis is the appearance of secondary sexual characteristics before 9 years of age, with testicular enlargement as the primary sign. The threshold pattern emerged most clearly at the extremes of the volume distribution.

In the lowest decile, children with glands of 115 cubic millimeters or less, activation prevalence was just 28.9 percent. In the highest decile, glands of 335 cubic millimeters or more, it reached 92.2 percent, a more than threefold difference. The gradient rose monotonically across quartiles, from 37.1 percent in the lowest to 84.8 percent in the highest, and the multivariable model translated it into predicted probabilities climbing from 38.7 percent at the 10th volume percentile to 73.4 percent at the 90th, even with all other predictors held at their means.

The odds ratio between the extreme deciles was 4.38, and each interquartile increase of roughly 106 cubic millimeters doubled the activation odds, with an odds ratio of 2.04. The asymmetry is telling: among non-activated children, 71.7 percent had below-median gland volumes, suggesting a small gland effectively vetoes activation, while among activated children, 63.1 percent had above-median volumes, indicating a large gland permits but does not compel it. Crucially, the researchers insist this is a permissive threshold, not a strict logical necessity.

Nearly 29 percent of children with very small glands still showed activation, proving that the axis can switch on without pronounced pituitary enlargement. The team therefore operationalizes the concept as a strongly permissive range-boundary effect: below a critical volume, activation is physiologically improbable enough to be clinically actionable, while above it, activation is permitted but not guaranteed. Physiologically, the interpretation aligns with the classic morphometric literature, since a gland that has not undergone pubertal gonadotroph expansion is anatomically unprepared to sustain axis activation.

Clinically, it means a borderline hormone value in a child with a very small pituitary carries a different meaning than the same value in a child with a large gland. To test whether pituitary volume is genuinely informative or merely a redundant proxy for age, the team deployed three complementary decomposition frameworks. Shapley value decomposition, which averages each predictor's marginal contribution across all possible orderings, ranked volume second behind age, attributing 22.2 percent of the model's total AUC gain to it, slightly more than insulin-like growth factor 1 at 21.4 percent.

Permutation importance confirmed the hierarchy, and Lindeman-Merenda-Gold variance partitioning assigned 20.4 percent of the model's pseudo-R-squared to volume, far exceeding IGF-1's 3.7 percent. Most persuasively, the component of volume orthogonal to all other predictors, isolated through residual analysis, remained significantly associated with activation, with a beta of 0.295 and a P value of 0.001. Pituitary size therefore contributes unique anatomical information that age, sex, and the biochemical surrogates do not already capture.

The mediation analysis added a mechanistic wrinkle. Insulin-like growth factor 1 statistically mediated 37.4 percent of the association between gland volume and axis activation, with 35.1 percent by an alternative ratio and the remaining majority appearing direct. The authors stress this is hypothesis-generating, associational evidence from cross-sectional data that cannot establish causality, but it nominates a plausible pathway in which pituitary enlargement reflects coordinated somatotroph and gonadotroph activity.

Meanwhile, the simple bedside arithmetic proved surprisingly usable: a volume tertile score alone achieved an AUC of 0.693, and combined with an IGF-1 tertile score, 0.748, approaching the full multivariable model's internally validated out-of-fold AUC of 0.801. Yet the study's most honest finding concerns what pituitary volume cannot do. Adding it to a clinical model already built on age, sex, IGF-1, and IGF-binding protein 3 raised the AUC by only 0.003, with a confidence interval of −0.007 to +0.013 that excludes any clinically meaningful improvement.

A formal power analysis revealed the study had roughly 5 percent power to detect an increment that small; confirming it would require approximately 25,360 children. The unadjusted AUC of 0.730 collapsed to 0.592 after age and sex adjustment, showing that most of the apparent standalone discrimination reflects developmental confounding. Performance was weakest in children under 7, where discrimination approached chance, and strongest in boys, at 0.890.

The incremental signal was largest exactly where diagnostic uncertainty peaks, at +0.027 in the 7-to-9-year age group and +0.006 in borderline IGF-1 cases. Decision curve analysis showed the direction of net benefit was positive at the clinically most relevant thresholds of 0.30 to 0.40 and 0.60, though all confidence intervals included zero. The authors conclude that pituitary volume must never replace the GnRH stimulation test, that MRI should not be ordered solely to measure the gland, and that the proposed thresholds of 115, 203.9, and 335 cubic millimeters require external validation before adoption.

At the Youden-optimal cutpoint of 203.9 cubic millimeters, sensitivity was 62.0 percent, specificity 74.6 percent, the positive likelihood ratio 2.44, the negative likelihood ratio 0.51, the diagnostic odds ratio 4.78, and the number needed to screen 2.73. Where a guideline-indicated MRI already exists, however, the volume measurement adds less than two minutes to the radiology read and functions as a complementary morphological checkpoint, contextualizing hormone interpretation at negligible cost. Current guidelines recommend cranial MRI for all boys with central precocious puberty, all girls younger than 6 years with the condition, and girls aged 6 years or older with rapidly progressive sexual development or neurological abnormalities, to exclude intracranial lesions; pituitary imaging is not mandatory for the diagnosis of precocious puberty per se.

A sensitivity analysis restricted to children within the Chinese diagnostic age window, girls younger than 10 and boys younger than 12, preserved the extreme-decile gradient, the second-ranked Shapley importance, and the small incremental value. Because the thresholds derive from a single-center cohort, the authors caution they must be recalibrated locally across scanner vendors, field strengths, and assay conventions before clinical use.

Subject of Research: Pituitary volume as a permissive morphological threshold for hypothalamic-pituitary-gonadal axis activation in children: a comprehensive diagnostic prediction study with relative importance decomposition

Article Title: Pituitary volume as a permissive morphological threshold for hypothalamic-pituitary-gonadal axis activation in children: a comprehensive diagnostic prediction study with relative importance decomposition

Article References: Cai, J., Zhang, L., Liu, W. V., Liu, Q., Lu, J., & Liu, D. (2026). Pituitary volume as a permissive morphological threshold for hypothalamic-pituitary-gonadal axis activation in children: a comprehensive diagnostic prediction study with relative importance decomposition. Frontiers in Endocrinology, 17, Article 1981603. https://doi.org/10.3389/fendo.2026.1981603

Image Credits: AI Generated

DOI: 10.3389/fendo.2026.1981603

Keywords: Pituitary, volume, permissive, morphological, threshold, hypothalamic-pituitary-gonadal, axis, activation, children, comprehensive, diagnostic, prediction