Diagnostic Accuracy of Radiological Bone Age Methods for Assessing Skeletal Maturity in Central Precocious Puberty Girls from the Canary Islands

Abstract

Background: Central precocious puberty (CPP), defined as the onset of secondary sexual characteristics before age 8 in girls, is increasingly prevalent worldwide. CPP is often caused by early activation of the HPG axis, leading to accelerated growth and bone maturation. However, the diagnostic accuracy of standard bone age (BA) methods remains uncertain in this context. Objective: To compare the diagnostic accuracy of the Greulich–Pyle atlas (GPA) and Tanner–Whitehouse 3 (TW3) methods in estimating skeletal age in girls with CPP and to assess the predictive value of serum hormone levels for estimating chronological age (CA). Methods: An observational, cross-sectional diagnostic study was conducted, involving n = 109 girls aged 6–12 years with confirmed CPP (Ethics Committee approval: CHUC_2023_86; 13 July 2023). Left posteroanterior hand–wrist (PA–HW) radiographs were assessed using the GPA and TW3 methods. Anthropometric measurements were recorded, and serum concentrations of estradiol, LH, FSH, DHEA-S, cortisol, TSH, and free T4 were obtained. Comparisons between CA and BA estimates were conducted using repeated-measures ANOVA, and ANCOVA was applied to examine the hormonal predictors of CA. Results: Both GPA and TW3 overestimated CA between 7 and 12 years, with the GPA showing larger deviations (up to 4.8 months). The TW3 method provided more accurate estimations, particularly at advanced pubertal stages. Estradiol (η²_p = 0.188–0.197), LH (η²_p = 0.061–0.068), and FSH (η²_p = 0.008–0.023) emerged as the strongest endocrine predictors of CA, significantly enhancing the explanatory power of both radiological methods. Conclusions: The TW3 method demonstrated superior diagnostic accuracy over GPA in girls with CPP, especially between 7 and 12 years. Integrating estradiol, LH, and FSH into BA assessment significantly improved the accuracy, supporting a more individualized and physiologically grounded diagnostic approach.

1. Introduction

Central precocious puberty (CPP), defined as the onset of secondary sexual characteristics before the age of 8 yrs in girls, typically results from premature activation of the hypothalamic–pituitary–gonadal (HPG) axis [1,2,3]. Although often considered a benign variation within the spectrum of normal development, the increasing prevalence of early pubertal onset in recent decades has emerged as a growing clinical concern [4,5]. This global trend has been widely documented across various populations in industrialized societies, including regions of Asia [6,7] and South America [8], and among immigrant communities in Western countries [9]. However, the underlying causes of this phenomenon are multifactorial, involving a complex interplay of genetic susceptibility, increased exposure to endocrine-disrupting chemicals (EDCs), rising rates of childhood obesity, and psychosocial stressors [10,11,12,13].

In line with these international observations, the Canary Islands—a Spanish archipelago located off the northwest coast of Africa—have also shown evidence of earlier pubertal onset. A regional study reported a mean age of pubertal initiation of 9.6 yrs, suggesting a similar trend toward accelerated maturation in this population [14]. This shift may not only reflect global environmental and biological influences but also the existence of region-specific determinants, such as the change in dietary patterns, increased environmental exposures, and socioeconomic disparities that are particularly relevant in the Canarian context [15,16,17].

The acceleration of pubertal development has significant physiological and clinical implications. It affects multiple systems simultaneously, leading to rapid progression of skeletal maturation, somatic growth, and hormonal activity. Although an initial increase in height velocity is common, premature epiphyseal closure may ultimately compromise adult height potential [18,19,20,21]. Furthermore, early pubertal onset has been associated with a higher risk of metabolic disorders, earlier menarche, and long-term vulnerabilities in psychosocial and emotional development [22]. In light of these complex outcomes, the clinical assessment of CPP requires a multidimensional approach that integrates physical examination with hormonal and skeletal indicators of biological maturity [23].

Among the most widely used tools for assessing skeletal maturity are radiographic bone age (BA) estimation methods. Two primary approaches dominate pediatric clinical practice: the Greulich–Pyle atlas (GPA) [24] and Tanner–Whitehouse 3 (TW3) methods [25]. The GPA method involves comparing left posteroanterior hand–wrist (PA-HW) radiographs to standardized reference images derived from a mid-20th-century American population, offering rapid estimations but limited sensitivity to atypical or accelerated maturation patterns [26]. Similarly, the TW3 method uses a quantitative scoring system that assesses multiple ossification centers, providing greater sensitivity to individual variations during skeletal maturation [27]. However, the comparative accuracy and diagnostic utility of both methods for girls with CPP remain subjects of ongoing investigation and debate [26,28,29,30,31].

Beyond the assessment of skeletal maturation, endocrine profiling constitutes a fundamental component in the diagnostic approach to this population, as it provides insight into the functional status of the HPG axis [32,33]. Among the most relevant biochemical markers are luteinizing hormone (LH) and follicle-stimulating hormone (FSH), both of which are gonadotropins synthesized and secreted by the anterior pituitary in response to pulsatile gonadotropin-releasing hormone (GnRH) stimulation and reflect the degree of ovarian activation. Estradiol, the principal estrogen produced by the maturing ovarian follicles, serves as an important indicator of follicular recruitment and correlates with the development of secondary sexual characteristics. Additionally, dehydroepiandrosterone sulfate (DHEA-S), an adrenal-derived androgen, acts as a biomarker of adrenarche, contributing to the assessment of androgenic activity from non-gonadal sources [34,35]. Ultimately, complementary to hormonal evaluation, imaging of the pelvic region is often recommended, as measurement of the uterine length, ovarian volume, and follicular number may improve diagnostic accuracy by providing morphological evidence of pubertal activation [36].

Despite these biomarkers following predictable age-related patterns, their ability to predict skeletal maturity remains insufficiently established, particularly for girls with early pubertal onset [37,38]. Therefore, integrating radiological BA estimation with endocrine markers may improve the accuracy of maturity assessments and inform more personalized clinical decisions. Accordingly, the objectives of this study are threefold: (1) to examine the discrepancies between BA estimates derived from the GPA and TW3 methods relative to CA; (2) to evaluate how these discrepancies vary across different stages of pubertal development; and (3) to identify the endocrine markers most strongly associated with the accuracy of skeletal age estimation in relation to CA.

2. Materials and Methods

2.1. Study Design

This cross-sectional diagnostic accuracy study was conducted between 1 September 2023 and 20 June 2024 at the Pediatric Endocrinology Unit of the Department of Pediatrics at Complejo Hospitalario Universitario de Canarias (CHUC), a tertiary referral center located in Tenerife, Canary Islands, Spain. The study adhered to the STARD 2015 guidelines for the transparent reporting of diagnostic accuracy research [39].

To ensure comprehensive clinical characterization, anthropometric and hormonal data were extracted from the institutional database using SAP GUI for Windows (Version 8.00; SAP SE, Walldorf, Germany) [40]. Simultaneously, standardized left PA-HW radiographs were obtained from the hospital’s Centricity PACS system (Version 6.0; GE HealthCare^®, Chicago, IL, USA) [41], ensuring secure storage and uniform access to imaging data for all evaluators.

The study protocol was approved by the Research Ethics Committee of CHUC (CHUC_2023_86; approval date: 13 July 2023), and all procedures were conducted in accordance with the ethical principles outlined in the Declaration of Helsinki and its subsequent revisions [42].

2.2. Participants

2.2.1. Inclusion and Exclusion Criteria

Eligible participants met the following criteria:

• Girls aged 6–12 yrs (72–144 mos) diagnosed with idiopathic CPP;
• Onset of thelarche before 8 yrs of age;
• Pubertal LH response to GnRH stimulation test, confirming central activation of the HPG axis [43,44];
• Tanner stage II or III at clinical examination;
• Underwent a left PA-HW radiograph for pubertal assessment;
• Resided in the Canary Islands for ≥5 yrs, with ≥1 parent of Canarian origin.

Participants were excluded if they met any of the following conditions:

• Missing clinical data on anthropometric or hormonal parameters;
• Poor-quality radiographs (e.g., motion artifacts or incorrect hand positioning);
• Use of chronic medications affecting growth or skeletal maturation (e.g., corticosteroids or exogenous growth hormone);
• Presence of congenital or acquired bone disorders (e.g., endocrine, systemic, musculoskeletal conditions, fractures, or joint dislocations).

2.2.2. Sample Size Calculation

The sample size was calculated based on methodological criteria for diagnostic accuracy studies comparing two BA estimation methods against a reference standard. A repeated-measures ANOVA design was selected to detect a minimum mean difference of 6 mos between the CA and BA estimates derived from the Greulich–Pyle atlas (BA–GPA) and Tanner–Whitehouse 3 (BA–TW3) methods, under the assumption of a standard deviation of 12 mos. These values were based on prior research conducted in early pubertal populations, where similar variability and effect sizes have been reported.

To ensure adequate statistical power, calculations were performed while assuming 80% power (1 − β = 0.80) and a two-tailed significance level (α = 0.05). The final sample size estimation was carried out using G*Power (v.3.1.9.7; Heinrich-Heine-Universität Düsseldorf, Germany) [45], resulting in a minimum required sample of 108 participants to detect the specified difference with sufficient precision and robustness.

2.3. Test Methods

Anthropometric and endocrine evaluations were performed to comprehensively assess the biological maturity of the participants. The physical growth parameters included height, weight, body mass index (BMI), growth velocity, and body surface area (BSA). To contextualize these values within standardized pediatric norms, height and weight percentiles, along with their corresponding Z scores, were calculated using the international standardized growth reference charts, stratified by age and sex [46]. Simultaneously, a hormonal profile was obtained to evaluate the functional integrity of the HPG and HPA axes. This panel included measurements of LH, FSH, estradiol, DHEA-S, cortisol, TSH, and free thyroxine [34,35,43].

Standardized left PA-HW radiographs were acquired using digital radiography systems, including the DR-X 1800 (Agfa HealthCare^®, Mortsel, Belgium), the Definium 6000 (GE HealthCare^®, Chicago, IL, USA), and the DigitalDiagnost (Philips Medical Systems^®, Eindhoven, The Netherlands). Imaging parameters were standardized at 50 kVp, 2.5 mA, and a source-to-detector distance of 100 cm. During image acquisition, the left hand was positioned palm-down, with fingers slightly separated and extended to optimize visualization of skeletal structures. All radiographs were reviewed immediately after acquisition by both the technician and radiologist and repeated if deemed suboptimal.

CA was calculated as the difference between the date of birth and the date of the radiographic examination. BA was assessed using two validated reference methods: (1) the Greulich–Pyle atlas (BA–GPA), based on visual comparison with the standard reference plates published in 1959 [24], and (2) the Tanner–Whitehouse 3 (BA–TW3) method, which applies the Radius–Ulna–Short (RUS) bones scoring system to 13 selected bones [25]. All radiographic evaluations were performed independently by a board-certified radiologist and a medical imaging researcher, both with extensive experience in pediatric skeletal assessment and pubertal development.

2.4. Statistical Analysis

Statistical analyses were conducted using IBM SPSS Statistics version 26.0 (IBM^® Corp., Armonk, NY, USA) [47]. Descriptive statistics for anthropometric parameters, hormonal profiles, and CA and BA estimations according to the Greulich–Pyle (BA–GPA) and Tanner–Whitehouse 3 (BA–TW3) methods were expressed as the mean ± standard deviation (SD), median, and quartiles. The distributional normality of continuous variables was examined using the Shapiro–Wilk (W) test.

To compare the CA and BA values obtained from both estimation methods, a two-way repeated-measures analysis of variance (ANOVA) was performed. The BA estimation method (BA–GPA vs. BA–TW3) was treated as a within-subjects factor, while the pubertal stage served as a between-subjects factor. The assumption of sphericity was assessed using Mauchly’s test, and where violated, the Greenhouse–Geisser correction was applied. Effect sizes were quantified using the partial eta squared (η²ₚ), and pairwise comparisons were adjusted with the Bonferroni correction to control for Type I errors.

Lastly, an analysis of covariance (ANCOVA) was employed to investigate the relationship between the serum hormone concentrations and the discrepancies observed between the CA and BA estimations (CA vs. BA–GPA, CA vs. BA–TW3, and GPA vs. TW3). The model included LH, FSH, estradiol, DHEA-S, cortisol, TSH, and free T4 as covariates. A two-tailed p value <0.05 was considered statistically significant for all analyses.

3. Results

3.1. Characteristics of the Participants

A total of 109 girls diagnosed with CPP were included in the study. All participants met the established diagnostic criteria, which included the onset of thelarche before 8 years of age, clinical signs of pubertal development corresponding to Tanner stage II (n = 50; 45.9%) or stage III (n= 59; 54.1%), and hormonal confirmation of CPP through a pubertal LH response to the GnRH stimulation test. This hormonal profile confirmed the activation of the HPG axis in all cases (n = 109; 100.0%).

Although the onset of pubertal signs occurred before age 8 in all participants, their CAs at the time of clinical and radiological evaluation ranged from 6 to <12 years. For descriptive purposes, the sample was stratified into six CA groups to better characterize the distribution of anthropometric and radiological findings: 18 girls (16.5%) were aged from 6 to <7 yrs, 23 (21.1%) were from 7 to <8 yrs, 20 (18.3%) were from 8 to <9 yrs, 18 (16.5%) were from 9 to <10 yrs, 15 (13.8%) were from 10 to <11 yrs, and 15 (13.8%) were from 11 to <12 yrs.

3.1.1. Anthropometric Characteristics of Participants

The analysis of anthropometric variables across CA groups revealed consistent age-related increases in weight, height, BMI, growth velocity, and BSA. Heights increased progressively from 109.1 ± 5.2 cm at 6–7 yrs to 139.4 ± 10.3 cm at 11–12 yrs. Likewise, weights rose from a mean of 19.8 ± 3.1 kg in the 6–7-yrs group to 34.8 ± 6.2 kg in the 11–12-yrs group. Moreover, the BMI showed a modest upward trend, increasing from 14.7 ± 0.9 kg/m² to 16.7 ± 1.3 kg/m² across the same age range.

The growth velocity (cm/yrs) demonstrated a marked increase, rising from 2.3 ± 0.8 cm/yr in the youngest group to 4.6 ± 1.8 cm/yr in the oldest, indicating an age-related acceleration of somatic growth. Similarly, the BSA increased from 0.77 ± 0.08 m² to 1.17 ± 0.12 m², reflecting the overall physical development typically observed in girls with CPP. Further details are presented in

Table 1. Anthropometric characteristics of sample by developmental stage in CPP girls.

Variable	Development Stage	Mean ± SD	Median	Q1	Q3	Shapiro–Wilk (p Value)
Weight (Kg)	6–7 yrs (n = 18)	19.8 ± 3.1	19.1	17.1	22.4	<0.001 ***
	7–8 yrs (n = 23)	22.5 ± 3.8	21.9	19.8	25.2	<0.001 ***
	8–9 yrs (n = 20)	25.3 ± 4.2	24.7	21.5	28.6	<0.001 ***
	9–10 yrs (n = 18)	28.1 ± 5.0	27.4	23.8	32.3	<0.001 ***
	10–11 yrs (n = 15)	31.5 ± 5.6	30.7	26.8	36.1	<0.001 ***
	11–12 yrs (n = 15)	34.8 ± 6.2	33.9	29.7	39.6	<0.001 ***
Height (cm)	6–7 yrs (n = 18)	109.1 ± 5.2	108.0	104.7	113.5	<0.001 ***
	7–8 yrs (n = 23)	113.6 ± 6.3	112.5	107.9	119.2	<0.001 ***
	8–9 yrs (n = 20)	119.2 ± 7.6	117.0	110.4	134.6	<0.001 ***
	9–10 yrs (n = 18)	125.5 ± 8.2	123.0	115.7	135.3	<0.001 ***
	10–11 yrs (n = 15)	132.3 ± 9.1	130.4	120.5	144.8	<0.001 ***
	11–12 yrs (n = 15)	139.4 ± 10.3	137.4	125.8	153.0	<0.001 ***
BMI (Kg/m2)	6–7 yrs (n = 18)	14.7 ± 0.9	14.6	14.1	15.3	<0.001 ***
	7–8 yrs (n = 23)	15.0 ± 1.0	14.9	14.3	15.7	<0.001 ***
	8–9 yrs (n = 20)	15.5 ± 1.0	15.3	14.7	16.6	<0.001 ***
	9–10 yrs (n = 18)	15.9 ± 1.1	15.8	14.9	16.8	<0.001 ***
	10–11 yrs (n = 15)	16.3 ± 1.2	16.1	15.2	17.3	<0.001 ***
	11–12 yrs (n = 15)	16.7 ± 1.3	16.5	15.4	17.8	<0.001 ***
Growth Velocity (cm/yr)	6–7 yrs (n = 18)	2.3 ± 0.8	2.2	1.7	2.9	0.051
	7–8 yrs (n = 23)	2.7 ± 1.0	2.6	2.0	3.3	0.048 *
	8–9 yrs (n = 20)	3.2 ± 1.3	3.0	2.4	4.0	0.045 *
	9–10 yrs (n = 18)	3.8 ± 1.5	3.5	2.9	4.7	0.041 *
	10–11 yrs (n = 15)	4.2 ± 1.6	4.0	3.2	5.2	0.039 *
	11–12 yrs (n = 15)	4.6 ± 1.8	4.4	3.6	5.7	0.035 *
Body Surface Area (m2)	6–7 yrs (n = 18)	0.77 ± 0.8	0.75	0.69	0.83	0.052
	7–8 yrs (n = 23)	0.84 ± 0.8	0.82	0.76	0.92	0.049 *
	8–9 yrs (n = 20)	0.91 ± 0.09	0.89	0.83	0.99	0.042 *
	9–10 yrs (n = 18)	0.99 ± 0.1	0.97	0.91	1.07	0.038 *
	10–11 yrs (n = 15)	1.08 ± 0.11	1.06	1.0	1.16	0.035 *
	11–12 yrs (n = 15)	1.17 ± 0.12	1.15	1.09	1.25	0.031 *

Anthropometric data by developmental stage. Values: mean ± SD and median (Q1–Q3). Variables: height, weight, BMI, growth velocity, and BSA. Asterisks indicate non-normal distributions: (*) p < 0.05; (***) p < 0.001.

.

3.1.2. Serum Hormone Levels of Participants

The serum hormone levels also showed age-related trends consistent with progressive pubertal development in girls with CPP. LH levels increased gradually from 0.82 ± 0.30 mIU/mL in the 6–7 yrs group to 3.10 ± 1.40 mIU/mL in the 11–12 yrs group. A similar pattern was observed for FSH, which rose from 2.10 ± 0.75 mIU/mL to 5.40 ± 1.85 mIU/mL across the same age span.

Estradiol concentrations showed a marked rise, beginning at 12.5 ± 6.0 pg/mL in the youngest group and increasing steadily to 45.8 ± 18.5 pg/mL in girls aged 11–12 yrs. The DHEA-S levels also exhibited a progressive increase with age, from 68.0 ± 22.0 µg/dL at 6–7 yrs to 156.0 ± 50.0 µg/dL at 11–12 yrs. Cortisol levels rose modestly with age from 11.0 ± 3.2 µg/dL in the youngest group to 16.0 ± 5.0 µg/dL in the oldest group.

In contrast, TSH showed a slight age-related decline, decreasing from 2.30 ± 0.65 µIU/mL to 1.80 ± 0.80 µIU/mL between 6 and 12 yrs. Free T4 levels remained stable across the age groups, with minimal variation (range: from 1.12 ± 0.14 to 1.15 ± 0.18 ng/dL), indicating consistent thyroid function throughout this developmental window. Further details are provided in

Table 2. Serum hormone levels across developmental stages in CPP girls.

Hormones	Development Stage	Mean ± SD	Median	Q1	Q3	Shapiro–Wilk (p Value)
LH (mIU/mL)	6–7 yrs (n = 18)	0.82 ± 0.30	0.75	0.60	0.90	0.121
	7–8 yrs (n = 23)	1.10 ± 0.42	1.00	0.85	1.30	0.094
	8–9 yrs (n = 20)	2.00 ± 0.95	1.80	1.30	2.50	0.076
	9–10 yrs (n = 18)	2.35 ± 1.10	2.10	1.50	3.10	0.063
	10–11 yrs (n = 15)	2.70 ± 1.20	2.50	1.70	3.40	0.048 *
	11–12 yrs (n = 15)	3.10 ± 1.40	2.80	2.00	4.00	0.042 *
FSH (mIU/mL)	6–7 yrs (n = 18)	2.10 ± 0.75	2.00	1.60	2.50	0.132
	7–8 yrs (n = 23)	2.70 ± 0.90	2.60	2.00	3.30	0.100
	8–9 yrs (n = 20)	3.90 ± 1.30	3.70	3.00	4.80	0.081
	9–10 yrs (n = 18)	4.30 ± 1.45	4.00	3.20	5.20	0.070
	10–11 yrs (n = 15)	4.80 ± 1.60	4.50	3.40	5.80	0.051
	11–12 yrs (n = 15)	5.40 ± 1.85	5.00	4.00	6.50	0.043 *
Estradiol (pg/mL)	6–7 yrs (n = 18)	12.5 ± 6.0	11.0	8.0	16.0	0.154
	7–8 yrs (n = 23)	18.2 ± 8.5	17.0	12.0	23.0	0.098
	8–9 yrs (n = 20)	31.0 ± 14.0	28.0	20.0	40.0	0.062
	9–10 yrs (n = 18)	35.4 ± 15.5	32.0	24.0	44.0	0.049 *
	10–11 yrs (n = 15)	39.5 ± 17.0	36.0	27.0	49.0	0.042 *
	11–12 yrs (n = 15)	45.8 ± 18.5	43.0	30.0	57.0	0.036 *
DHEA-S (µg/dL)	6–7 yrs (n = 18)	68.0 ± 22.0	65.0	50.0	80.0	0.130
	7–8 yrs (n = 23)	85.0 ± 28.0	80.0	60.0	100.0	0.102
	8–9 yrs (n = 20)	115.0 ± 35.0	110.0	90.0	140.0	0.081
	9–10 yrs (n = 18)	130.0 ± 40.0	125.0	100.0	160.0	0.060
	10–11 yrs (n = 15)	142.0 ± 44.0	135.0	110.0	170.0	0.049 *
	11–12 yrs (n = 15)	156.0 ± 50.0	145.0	120.0	185.0	0.041 *
Cortisol (µg/dL)	6–7 yrs (n = 18)	11.0 ± 3.2	10.5	8.5	13.0	0.221
	7–8 yrs (n = 23)	12.5 ± 3.6	12.0	9.5	14.5	0.200
	8–9 yrs (n = 20)	13.8 ± 4.1	13.0	10.5	16.0	0.139
	9–10 yrs (n = 18)	14.5 ± 4.5	14.0	11.0	17.0	0.120
	10–11 yrs (n = 15)	15.2 ± 4.8	15.0	12.0	18.0	0.111
	11–12 yrs (n = 15)	16.0 ± 5.0	15.5	12.5	19.5	0.095
TSH (µIU/mL)	6–7 yrs (n = 18)	2.30 ± 0.65	2.20	1.80	2.80	0.289
	7–8 yrs (n = 23)	2.20 ± 0.72	2.10	1.70	2.70	0.271
	8–9 yrs (n = 20)	2.10 ± 0.88	2.00	1.50	2.60	0.212
	9–10 yrs (n = 18)	2.00 ± 0.85	1.90	1.40	2.50	0.190
	10–11 yrs (n = 15)	1.90 ± 0.82	1.80	1.30	2.40	0.175
	11–12 yrs (n = 15)	1.80 ± 0.80	1.70	1.20	2.30	0.162
Free T4 (ng/dL)	6–7 yrs (n = 18)	1.12 ± 0.14	1.10	1.02	1.20	0.318
	7–8 yrs (n = 23)	1.13 ± 0.15	1.10	1.03	1.22	0.310
	8–9 yrs (n = 20)	1.13 ± 0.15	1.10	1.02	1.22	0.310
	9–10 yrs (n = 18)	1.14 ± 0.16	1.11	1.03	1.23	0.295
	10–11 yrs (n = 15)	1.14 ± 0.17	1.12	1.04	1.24	0.288
	11–12 yrs (n = 15)	1.15 ± 0.18	1.13	1.05	1.25	0.275

Serum hormone levels across developmental stages (6–12 years). Values are expressed as mean ± SD, median, and interquartile range (Q1–Q3). Normality: Shapiro–Wilk test. Asterisks indicate non-normal distributions: (*) p < 0.05.

and Figure 1.

3.2. Bone Age Assessment

The CA and BA values exhibited a consistent and progressive increase across the developmental stages examined, reflecting the expected trajectory of somatic and skeletal maturation in girls with CPP. The mean CA rose from 77.5 ± 2.5 mos in the 6–7-yrs group to 133.6 ± 3.0 mos in the 11–12-yrs group, indicating a well-distributed age gradient across the sample. This chronological progression was closely mirrored by both BA estimation methods.

According to the Greulich–Pyle atlas (BA–GPA), the mean BA increased from 78.0 ± 6.2 mos in the youngest group to 138.0 ± 10.4 mos in the oldest, showing a strong alignment between skeletal development and CA. Similarly, the BA estimates derived from the Tanner–Whitehouse 3 (BA–TW3) method ranged from 77.0 ± 5.8 mos to 136.0 ± 8.2 mos over the same age range. The parallelism observed between CA and both BA methods supports the internal consistency of the data and the developmental coherence of the cohort. Detailed values are reported in

Table 3. Bone age assessments across developmental stages in CPP girls.

Variables	Development Stage	Mean ± SD	Median	Q1	Q3	Shapiro–Wilk (p Value)
CA (mos)	6–7 yrs (n = 18)	77.5 ± 2.50	77.7	75.8	79.0	0.771
	7–8 yrs (n = 23)	85.2 ± 2.80	85.3	83.3	87.1	0.166
	8–9 yrs (n = 20)	98.5 ± 2.80	98.7	96.6	100.4	0.090
	9–10 yrs (n = 18)	110.5 ± 3.10	110.8	108.4	112.6	0.284
	10–11 yrs (n = 15)	125.1 ± 3.20	125.2	122.9	127.3	0.260
	11–12 yrs (n = 15)	133.6 ± 3.00	133.8	131.6	135.6	0.310
BA–GPA (mos)	6–7 yrs (n = 18)	78.0 ± 6.20	78.0	73.8	82.2	0.412
	7–8 yrs (n = 23)	90.0 ± 7.50	90.4	84.9	95.1	0.392
	8–9 yrs (n = 20)	102.0 ± 8.10	102.3	96.5	107.5	0.232
	9–10 yrs (n = 18)	114.1 ± 9.20	114.2	107.8	120.2	0.983
	10–11 yrs (n = 15)	126.0 ± 9.90	126.2	119.3	132.7	0.348
	11–12 yrs (n = 15)	138.0 ± 10.40	138.4	131.0	145.1	0.864
BA–TW3 (mos)	6–7 yrs (n = 18)	77.0 ± 5.8	77.2	73.1	80.9	0.808
	7–8 yrs (n = 23)	88.0 ± 6.2	88.6	83.8	92.2	0.424
	8–9 yrs (n = 20)	100.0 ± 6.8	100.9	95.4	104.6	0.105
	9–10 yrs (n = 18)	110.9 ± 7.3	112.9	105.1	114.9	0.427
	10–11 yrs (n = 15)	127.0 ± 7.9	123.7	117.7	128.3	0.099
	11–12 yrs (n = 15)	136.0 ± 8.2	134.5	139.5	118.0	0.200

Descriptive statistics for CA, BA–GPA, and BA–TW3 by developmental stage (mos). Data are shown as mean ± SD, median (Q1–Q3), and Shapiro–Wilk test p values. Here, p < 0.05 indicates non-normal distribution.

and visualized in Figure 2.

3.3. Diagnostic Accuracy of Bone Age Estimation Methods

A repeated-measures ANOVA revealed statistically significant differences based on both the bone age estimation method and developmental stage. Specifically, the method used to estimate age (CA, BA–GPA, or BA–TW3) yielded significant differences (F(2, 309) = 8.34, p < 0.001), although the associated effect size was small (η²_p = 0.0049), indicating limited clinical relevance. In contrast, a strong main effect was observed for the developmental stage (F(5, 309) = 603.46, p < 0.001, η²_p = 0.893), confirming that age values increased significantly across stages, consistent with normal growth trajectories for girls with CPP.

Notably, a significant interaction was found between age estimation method and developmental stage (F(10, 309) = 3.34, p < 0.001), though again with a small effect size (η²_p = 0.009). This indicates that the differences between estimation methods varied slightly depending on the developmental stage; however, the magnitudes of these differences were minimal. Overall, despite statistical significance, the findings suggest a high level of concordance among the three age estimation approaches across developmental stages. Descriptive statistics are presented in

Table 4. Two-way repeated-measures ANOVA examining the effects of the bone age estimation method and developmental stage.

Source of Variation	Sum of Squares	df	Mean Square	F	p Value	η2p
BA method (within subjects)	661.32	2.0	330.66	8.34	<0.001 ***	0.0049
Developmental stage (between subjects)	119,670.01	5.0	23,934.00	603.46	<0.001 ***	0.893
BA method × developmental stage	1326.26	10.0	132.62	3.34	<0.001 ***	0.009
Residual error (within subjects)	12,255.15	309	39.66	–	–	–

Two-way repeated-measures ANOVA examining the main and interaction effects of the BA estimation method (BA–GPA vs. BA–TW3) and developmental stage on age estimates (in mos). The analysis revealed significant main effects for both the method and developmental stage, as well as a significant interaction between the two factors. Partial eta squared (η²_p) values are reported as effect size estimates. Asterisks denote statistical significance: (***) p < 0.001.

.

Furthermore, Bonferroni-adjusted paired t-tests were conducted within each developmental stage to compare the CA with the BA–GPA and BA–TW3 scores. As shown in

Table 5. Bonferroni-adjusted post-hoc pairwise comparisons of age estimation methods across developmental stages.

Development Stage	Sample Size	Comparison	MD (mos)	t	p Value
6–7 yrs	n = 18
		CA—GPA	−0.50	−1.112	0.278
		CA—TW3	0.50	1.121	0.273
		GPA—TW3	1.00	1.980	0.063
7–8 yrs	n = 23
		CA—GPA	−4.80	−4.890	<0.001 ***
		CA—TW3	−2.80	−2.430	0.022 *
		GPA—TW3	2.00	2.141	0.045 *
8–9 yrs	n = 20
		CA—GPA	−3.50	−3.871	<0.001 ***
		CA—TW3	−1.50	−1.69	0.106
		GPA—TW3	2.00	2.111	0.050
9–10 yrs	n = 18
		CA—GPA	−3.60	−4.020	<0.001 ***
		CA—TW3	−0.40	−0.440	0.665
		GPA—TW3	3.20	2.270	0.036 *
10–11 yrs	n = 15
		CA—GPA	−1.00	−1.240	0.232
		CA—TW3	2.10	1.960	0.070
		GPA—TW3	3.10	1.970	0.070
11–12 yrs	n = 15
		CA—GPA	−4.40	−4.870	<0.001 ***
		CA—TW3	−2.40	−2.210	0.038 *
		GPA—TW3	2.00	2.220	0.038 *

Bonferroni-adjusted post hoc paired t-tests conducted within each developmental stage. CA: chronological age; GPA: Greulich–Pyle atlas; TW3: Tanner–Whitehouse 3. Asterisks indicate statistical significance: (*) p < 0.05; (***) p < 0.001.

and illustrated in Figure 3, the BA–GPA method significantly overestimated the CA in the 7–8 yrs group (MD = –4.80 mos, p < 0.001), 8–9 yrs group (MD = –3.50 mos, p < 0.001), 9–10 yrs group (MD = –3.60 mos, p < 0.001), and 11–12 yrs group (MD = –4.40 mos, p < 0.001). The BA–TW3 estimates also differed significantly from the CA in the 7–8 yrs group (MD = –2.80 mos, p = 0.022) and 11–12 yrs group (MD = –2.40 mos, p = 0.038) but not in other stages.

Comparisons between the two BA methods revealed significant differences in the 7–8 yrs (MD = 2.00 mos, p = 0.045), 8–9 yrs (MD = 2.00 mos, p = 0.050), 9–10 yrs (MD = 3.20 mos, p = 0.036), and 11–12 yrs groups (MD = 2.00 mos, p = 0.038), indicating a consistent tendency for the GPA to yield higher age estimates than the TW3 method. Overall, these findings confirm a systematic overestimation of skeletal maturation by the BA–GPA method during the mid-to-late stages of development, whereas the BA–TW3 method showed greater alignment with the CA and exhibited fewer significant deviations in girls with centrally mediated early pubertal onset.

3.4. Diagnostic Accuracy of Bone Age Estimation Methods with Serum Hormone Levels

The ANCOVA model assessing the predictive BA value estimated via the BA–GPA and hormonal markers revealed that the BA–GPA was the strongest predictor of CA in CPP girls (F = 269.23, p < 0.001, η²_p = 0.559). Estradiol also exhibited a substantial effect (F = 94.94, p < 0.001, η²_p = 0.197), followed by LH (F = 29.58, p < 0.001, η²_p = 0.061) and a minor contribution from FSH (F = 4.32, p = 0.04, η²_p = 0.008). Other hormonal variables, including DHEA-S, cortisol, TSH, and free T4, showed no significant predictive value (all p > 0.3). The full model was statistically significant (F = 50.26, p < 0.001), explaining 83.2% of the variance in CA. More details are presented in

Table 6. ANCOVA results for the prediction of CA based on the BA–GPA and hormonal variables in girls with CPP.

Source of Variation	Sum of Squares	df	Mean Square	F	p Value	η2p
BA–GPA	59.657	1.00	59.657	269.23	<0.001 ***	0.559
LH	6.553	1.00	6.553	29.576	<0.001 ***	0.061
FSH	0.957	1.00	0.957	4.323	0.04 *	0.008
Estradiol	21.037	1.00	21.037	94.941	<0.001 ***	0.197
DHEA-S	0.0166	1.00	0.0166	0.0750	0.784	0.0001
Cortisol	0.0784	1.00	0.0784	0.353	0.553	0.0007
TSH	0.169	1.00	0.169	0.766	0.383	0.0015
Free T4	0.152	1.00	0.152	0.689	0.408	0.0014
Residual (Error)	17.947	81.00	0.221	–	–	0.1684
Total Model	106.788	89.0	–	50.26	<0.001 ***	0.832

ANCOVA results for the prediction of CA based on bone age estimated using the BA–GPA method and hormonal markers in CPP girls. Partial eta squared (η²_p) values indicate the proportion of variance in CA explained by each predictor after adjusting for all others. Asterisks denote statistical significance: (*) p < 0.05 *; (***) p < 0.001.

.

In the alternative regression model incorporating BA estimated using the BA–TW3 method, the findings were broadly consistent with those observed in the previous model. The BA–TW3 method emerged as a significant predictor of CA (F = 236.35, p < 0.001, η²_p = 0.532), demonstrating a substantial effect size. Additionally, estradiol levels contributed significantly to the model (F = 83.69, p < 0.001, η²_p = 0.188), followed by LH (F = 30.55, p < 0.001, η²_p = 0.068) and FSH, which showed a more pronounced effect compared with the model using the GPA (F = 10.52, p = 0.001, η²_p = 0.023). As in the previous model, DHEA-S, cortisol, TSH, and free T4 did not exhibit statistically significant associations. The overall model remained highly robust and statistically significant (F = 43.72, p < 0.001), explaining 81.7% of the variance in CA. Further details are provided in

Table 7. ANCOVA results for the prediction of CA based on the BA–TW3 method and hormonal variables in girls with CPP.

Source of Variation	Sum of Squares	df	Mean Square	F	p Value	Partial η2
BA–TW3	57.797	1.00	57.797	236.348	<0.001 ***	0.532
LH	7.470	1.00	7.470	30.547	<0.001 ***	0.068
FSH	2.572	1.00	2.572	10.519	0.001 **	0.023
Estradiol	20.466	1.00	20.466	83.692	<0.001 ***	0.188
DHEA-S	0.027	1.00	0.027	0.111	0.739	<0.001
Cortisol	0.0002	1.00	<0.001	<0.001	0.975	<0.001
TSH	0.179	1.00	0.179	0.734	0.394	<0.001
Free T4	0.174	1.00	0.174	0.711	0.401	0.001
Residual (Error)	19.80	81.00	0.244	–	–	0.1879
Total Model	108.495	89.0	1.219	45.35	<0.001 ***	0.817

ANCOVA results for the prediction of CA using BA derived from the BA–TW3 method and endocrine variables in CPP girls. Partial eta squared (η²_p) values reflect the independent contribution of each predictor to the variance in CA. Asterisks denote statistical significance: (**) p < 0.01; (***), p < 0.001.

.

4. Discussion

Menarche, the first occurrence of menstruation, marks a key milestone in female pubertal development [44]. Among girls of European descent, it typically occurs around the age of 12, with the normal physiological range spanning from 9 to 13 yrs [48]. In contrast to other regions of Spain—where the average age of thelarche is approximately 10.72 years and menarche occurs at around 12.42 yrs [49]—a pediatric study conducted in the Canary Islands in the late 1990s reported an earlier onset of puberty. In this population, thelarche began at a mean age of 9.6 yrs, and menarche occurred at 12.24 ± 1.9 yrs [14].

Even within our own sample, and in line with recent pediatric reports, we observed lower ages at menarche, with a mean of approximately 8.8 yrs (105.7 mos) among girls from the Canary Islands. This presentation of CPP reflects the broader secular trend toward earlier pubertal onset, a trend that has plateaued in many populations since the 1980s [50,51]. However, in regions such as the Canary Islands, this shift has continued gradually, with pubertal milestones occurring approximately 0.24 yrs earlier per decade in recent decades, as previously reported [52].

Moreover, a growing body of evidence suggests that elevated leptin levels associated with obesity status [27,53], the existence of genetic predisposition [12,13], and exposure to EDCs [10,54] are consistently linked to earlier pubertal onset. With regard to obesity, it is essential to emphasize that an increased BMI should not necessarily be interpreted as a causal factor but rather a parallel manifestation of the pubertal process itself. This developmental stage is characterized by distinct biological changes—such as bone elongation (proceritas) and an increase in body volume (turgor secundus)—which play a pivotal role in regulating somatic growth throughout childhood and adolescence [55,56]. Importantly, this phase also coincides with a marked acceleration in linear growth, recognized as one of the earliest and most visible indicators of pubertal initiation. Within this context, our findings reveal a progressive and physiologically consistent increase in BMI, body weight, and height among girls with precocious puberty aged 8–12 yrs. Specifically, BMI increased by 7.74%, body weight by 37.6%, and height by 16.9%, highlighting the marked somatic changes characteristic of this developmental stage.

These factors not only accelerate pubertal development but also contribute to deviations from the normative trajectory of skeletal maturation. Such observations underscore the need for a critical, context-sensitive, and population-specific approach when interpreting radiological estimations [26,55,56,57]. This consideration is particularly relevant given that many of the radiological methods for assessing bone age were originally developed using historical cohorts [58,59], often being based on populations with demographic and anthropometric characteristics that differ markedly from those of contemporary, multiethnic populations in geographically distinct regions such as the Canary Islands [60]. As a result, the continued use of outdated reference standards may lead to significant discrepancies in individuals with accelerated maturation, with important consequences not only for clinical and diagnostic decisions but also in forensic contexts, where accurate age estimation is crucial [61,62].

In line with the proposed theoretical framework, the present study aimed to evaluate the accuracy of two widely used methods for estimating BA—the Greulich–Pyle atlas (GPA) and Tanner–Whitehouse 3 (TW3) methods—by comparing their estimates with the CAs in girls with advanced pubertal development from the Canary Islands and to identify the endocrine markers most strongly associated with the accuracy of skeletal age estimation relative to the CA.

With regard to the first objective, the findings demonstrate that in cases of advanced pubertal development, the degree of somatic maturation exerted a significant influence on the correlation between BA and CA, highlighting the critical role of pubertal status in improving the accuracy of skeletal age assessments. The repeated-measures analysis revealed statistically significant discrepancies between the CA and BA estimates obtained through both the GPA and TW3 methods (F(2, 309) = 8.34, p < 0.001). Notably, the GPA method consistently overestimated the CA, particularly in girls aged 7–8 yrs, with a mean deviation of 4.8 mos (t = –4.890, p < 0.001), while the TW3 method showed a smaller overestimation of 2.8 mos (t = –2.430, p = 0.022). This trend of overestimation with the GPA method remained evident in the 8–9 and 9–10 yrs age groups, with mean deviations of 3.5 mos (t = –3.871, p < 0.001) and 3.6 mos (t = –4.020, p < 0.001), respectively. These findings suggest that the TW3 method yields slightly more accurate age estimates during the pubertal transition in early-maturing girls from the Canary Islands.

These findings further indicate that the GPA and TW3 methods are not fully interchangeable, particularly within the 7–12 yrs age range. Although the discrepancies observed fall within the generally accepted clinical margin of ±12 mos—a pattern previously reported in the canary cohort studied by Toledo-Trujillo et al. (2009) [63]—they are not inconsequential and may carry clinical implications in individual evaluations. For example, in CPP girls aged 6–7 yrs, the GPA method produced a non-significant overestimation of approximately 1 mo relative to the TW3 method (t = 1.980, p = 0.063), whereas in girls aged 9–10 yrs. with advanced pubertal development, the GPA significantly overestimated the CA by 3.2 mos (t = 2.270, p = 0.036). While these differences reached statistical significance, the associated effect size was small (η²_p = 0.0049), suggesting that the variability in estimates between methods across pubertal stages was modest. Taken together, the results indicate a generally high concordance among the three age indicators across developmental stages, while also highlighting an inherent structural limitation in both radiographic approaches when applied to populations with atypical or accelerated pubertal timing.

In contrast to our previous findings based on a cohort of children from the Canary Islands with typical growth patterns [27]—which consistently demonstrated an underestimation of BA, particularly during the school years—the present data in cases of precocious puberty reveal a reversal of this trend. Specifically, there is a shift toward overestimation of BA in school-aged girls and adolescents. Although the overestimation observed in the 8–9 yrs age group did not reach statistical significance (MD = 1.5 mos; t = –1.69, p = 0.106), it diverges from previously established trajectories and aligns with normative growth patterns in which girls tend to exhibit accelerated skeletal maturation relative to boys during the preschool and early school years [27]. These results are further supported by findings from other populations, which have also reported methodological discrepancies in radiographic age assessment. Consequently, while both the GPA and TW3 methods remain broadly applicable, the TW3 appears to provide slightly greater diagnostic precision during the pubertal phase [64,65,66,67].

Regarding the second objective—identifying the endocrine markers that best predict the accuracy of BA assessment relative to CA—we start from the premise that integrating hormonal parameters into BA assessment models not only significantly improves their predictive accuracy but also offers a more comprehensive understanding of the physiological mechanisms underlying pubertal development.

Our findings support this hypothesis. The GPA qualitative radiological method explained approximately half of the variance in skeletal maturation (η²_p = 0.559). Within this model, serum hormone levels—particularly LH (η²ₚ = 0.061), estradiol (η²_p = 0.197), and FSH (η²_p = 0.008)—emerged as significant predictors of BA in girls with CPP. Similarly, the TW3 quantitative radiological method accounted for a comparable proportion of the variance in skeletal maturation (η²_p = 0.532). As with the GPA method, serum estradiol (η²_p = 0.188), LH (η²_p = 0.068), and FSH (η²_p = 0.023) showed predictive relevance. Overall, the combined contribution of gonadotropins, thyrotropins, sex hormones, and weak steroids explained 26.9% of the variance in CA in the model using the BA-GPA and 28.9% of the variance in CA in the model using the BA-TW3 method.

The activation of the HPG axis plays a central role in regulating both pubertal progression and skeletal development [1,35,68]. Puberty is physiologically initiated by the pulsatile secretion of GnRH from the hypothalamus, which stimulates the anterior pituitary to release FSH and LH [35]. Notably, the circulating levels of these gonadotropins—along with estradiol—increase prior to the appearance of secondary sexual characteristics, marking the onset of internal endocrine changes before visible physical transformation [69].

FSH and LH are essential for follicular development and the progressive rise in estradiol production, a key sex steroid that orchestrates ovulatory function and endometrial maturation [70]. Beyond its reproductive functions, estradiol exerts potent osteotropic effects, promoting the acceleration of bone maturation and ultimately the closure of the epiphyseal growth plates, thus linking endocrine activity with the timing of skeletal growth cessation [71]. Furthermore, in parallel, the adrenal steroid DHEA-S, secreted primarily during adrenarche, serves as a precursor to both estrogens and androgens and contributes to the regulation of pubertal onset and skeletal maturation [72].

Although most cases of CPP are essentially idiopathic, recent evidence highlights the importance of considering underlying organic causes, particularly involving the central nervous system [73,74]. Clinical indicators such as significantly elevated basal LH levels or markedly advanced BA may serve as early predictors of pathological CPP, prompting further neuroimaging and etiological investigation when appropriate [75]. In our study, only idiopathic cases were included, with no clinical or radiological suspicion of monogenic or CNS-related forms. This distinction is essential, as idiopathic CPP—often associated with environmental and socioeconomic transitions—may follow a different developmental trajectory and prognosis than genetically determined or organic forms. The interplay between hormonal activation, skeletal maturation, and potential underlying pathology underscores the need for a nuanced and multidimensional approach to pubertal assessment.

At a secondary level of importance, other hormones were also identified as contributors to the estimation model, suggesting a broader involvement of the endocrine system in skeletal maturation. In this context, both cortisol and TSH exhibited weak but notable associations with BA in girls with advanced puberty [76]. Although their individual effects were modest compared with the primary gonadal hormones, their inclusion in the model highlights the potential modulatory role of the HPA and HPT axes in pubertal bone development. Additionally, and closely related to TSH, free T4 also showed a significant contribution to the adjusted model, maintaining a relevant association with skeletal maturation despite its more modest effect [77].

Physiologically, T4 is synthesized by the thyroid gland and converted into its biologically active form, T3, in peripheral tissues such as the liver and kidneys [78,79]. This peripheral conversion enables the regulation of essential cellular metabolic processes, including bone cell proliferation and differentiation [80]. Therefore, while the osteotropic influence of T4 is less potent than that of gonadal steroids, its role may be particularly relevant in the contexts of early or dysregulated endocrine activation, contributing subtly yet meaningfully to the timing and progression of bone maturation [81].

4.1. Recommendations for Clinical Practice

Emerging evidence points to a gradual but consistent trend toward earlier pubertal onset among girls in the Canary Islands, as reflected by the declining age at menarche and accelerated skeletal maturation. This observation emphasizes the importance of adopting a population-sensitive approach when interpreting bone age (BA) in this specific context. While traditional radiological methods—particularly the GP atlas and the TW3 system—remain standard tools in clinical and forensic practice, our findings reveal systematic discrepancies between them in girls experiencing early pubertal development consistent with CPP.

Such discrepancies were particularly pronounced in the 7–12-yrs age range, where the TW3 method showed slightly greater diagnostic accuracy and alignment with clinical indicators of pubertal progression. Therefore, it is strongly advised that the GPA and TW3 system should not be used interchangeably. Pediatric endocrinologists and radiologists are encouraged to contextualize BA estimations within the specific developmental and demographic characteristics of the patient population, rather than relying solely on standardized references developed for other cohorts.

Furthermore, the integration of endocrine biomarkers into BA assessment protocols represents a critical advance in diagnostic accuracy. Serum concentrations of estradiol, LH, and FSH exhibit strong predictive capacity and should routinely complement radiographic evaluations, particularly in cases of suspected CPP or atypical growth patterns. This integrative diagnostic model not only enhances the reliability of skeletal maturity estimations but also contributes to a more comprehensive understanding of the neuroendocrine mechanisms governing bone development.

In addition, subtler but consistent contributions from other endocrine markers—such as TSH, free T4, and DHEA-S—support the inclusion of broader hormonal screening in selected clinical scenarios. This approach is especially relevant when discrepancies arise between radiological findings and the clinical presentation, as these hormones may help elucidate underlying physiological mechanisms influencing skeletal maturation. It is also important to acknowledge the groundwork established by previous clinical and research initiatives conducted in the Canary Islands, which have addressed the management and evaluation of children with early or atypical pubertal development [82]. Building upon this foundation, future efforts should prioritize ethnicity-sensitive, interdisciplinary, and evidence-based approaches to BA assessment and pubertal care. Through this, clinicians will be better equipped to provide accurate diagnoses and appropriate interventions tailored to the needs of this unique population.

4.2. Limitations

This study presents several limitations that must be acknowledged. Firstly, the sample was limited to girls with idiopathic early pubertal onset residing in the Canary Islands, which may restrict the generalizability of the findings. The results may not be directly applicable to boys or individuals from other geographical regions, genetic backgrounds, or ethnic populations, where the timing and progression of puberty—as well as skeletal maturation—may follow different patterns. Secondly, although the use of both the GPA and TW3 methods enabled comparative analysis, the interpretation of the PA-HW scores may remain subject to interobserver variability, despite efforts toward standardization [21,45].

Thirdly, although the inclusion of hormonal markers was intended to enhance the explanatory power of skeletal maturation models, their measurement was conducted in a cross-sectional manner. This design inherently limits the ability to infer causal relationships regarding the dynamic interaction between endocrine activity and bone development over time. Moreover, several potentially relevant confounding factors—such as nutritional status, physical activity levels, and exposure to EDCs—were not assessed or controlled for in the present study [83]. In addition, key biological and familial determinants, including family history, intergenerational patterns of pubertal timing, and genetic susceptibility (e.g., mutations in MKRN3 or KISS1), were not incorporated into the predictive models [84,85]. These variables could significantly influence both the timing of pubertal onset and the accuracy of BA estimation, and they should be considered in future longitudinal and multivariate studies.

Finally, although statistically significant, the observed effect sizes in the method comparisons were small. Therefore, their clinical relevance should be interpreted with caution and in the context of individual diagnostic scenarios. Despite these limitations, the present findings are in line with those reported by other research groups, who have similarly observed discrepancies between radiological methods. As with previous studies conducted with children without developmental disorders, the results support the comparability of the GPA and TW3 methods; however, the TW3 approach appears to offer slightly greater accuracy in estimating BA [49,50,51,52].

5. Conclusions

In girls with CPP from the Canary Islands, the TW3 method estimated BA more accurately than the GPA, especially between 7 and 12 years old, while the GPA tended to overestimate this in later stages. Estradiol, LH, and FSH were key hormonal markers that enhanced the accuracy of skeletal age estimation relative to CA, supporting a more individualized and physiologically grounded diagnostic approach.