Participants include girls referred to the pediatric endocrinology departments due to appearance of secondary sexual characteristics before the age of 8 years old. The indicators were ultrasonography measurements on female pelvic ultrasonography. The GnRH stimulation test was considered the comparator (gold standard) to confirm PP diagnosis, after taking clinical manifestations and radiological assessment into account. Regarding the outcomes of the first aim, we performed a comparative meta-analysis to identify standardized mean differences (SMDs) between the PP and PT groups with respect to each selected parameter. For the second aim, we performed a diagnostic accuracy meta-analysis to calculate the pooled sensitivity and specificity.

All types of studies, except for case reports and review articles, were included. Those whose organic conditions might cause PP were excluded. These criteria were pre-outlined in the selected articles. The following parameters were included in the comparative analysis due to sufficient data: ovarian volume, fundal-cervical ratio (FCR), uterine length, uterine cross-sectional area (CSA), and uterine volume. Two independent reviewers completed the process of searching for, screening, reviewing, and extracting data. In case of disagreements between the reviewers, a third reviewer was consulted to reach a final decision. We used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) to verify the transparent reporting.

This systematic review and meta-analysis was registered in the PROSPERO International Prospective Register of Systematic Reviews (ID: CRD42021232427).

We systematically searched the PubMed/MEDLINE, EMBASE, Scopus, and Cochrane databases for relevant articles up to 31/03/2021; the search did not include restrictions on language or publication year. The following keywords were used in the search ( Table S1 ): “precocious puberty,” “premature thelarche,” “ultrasound,” “sonography,” and “echography”. Furthermore, we identified additional relevant articles by manually searching the references of the articles found.

The mean, standard deviation of each parameter measurement, the number of observations in each group, and other demographic variables were documented. True positives, true negatives, false positives, and false negatives were directly extracted from the papers or indirectly calculated from sensitivity and specificity when appropriate. If the required data were not sufficiently furnished in an article, the corresponding author of that article was contacted through e-mail to request for the missing statistics.

The risk of bias of included studies in the comparative meta-analysis was assessed using the Newcastle–Ottawa Scale (15). On the other hand, we adopted a revised version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool to examine the risk of bias and applicability concerns (16).

Forest plots were constructed to display the estimated SMDs from each included study and the overall calculations. Heterogeneity among studies was tested using Cochran’s Q test and I². A random-effects model would be adopted when heterogeneity was observed between studies, as confirmed by a Cochran’s Q test p value of <0.1 or an I² of >50%. Otherwise, a fixed-effects model was preferred. Meta-regression and subgroup analysis were performed to explore sources of heterogeneity if indicated. Publication bias for each parameter was examined using Egger’s test.

A bivariate model was used to calculate the pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR), along with their corresponding 95% confidence intervals (CIs). The performance of pelvic ultrasonography parameters was represented by a summary receiver operating characteristics (SROC) curve and the area under the SROC curve (AUSROC). The closer the SROC curve is to the left corner and the higher AUSROC is, the higher discrimination ability of the test. The cutoff values of pooled sensitivity and specificity were defined using multiple thresholds modeling (17, 18).

An asymmetry test based on Deeks’ funnel plot was performed to validate the asymmetry assumption, with a p value of <0.1 signifying publication bias (19). Fagan’s nomogram was used to determine the posttest probability of PP after pelvic ultrasonography. A large-scale epidemiologic study in Korea reported that the overall prevalence of PP in girls was 0.4% (20). In the present study, we used this prevalence as the pretest probability of PP, and it was located on the left axis of the nomogram, whereas the PLR and NLR of pelvic ultrasonography parameters were located on the middle axis. These variables were used to project the posttest probability of PP, which was located on the right axis of the nomogram.

A two-sided p value of <0.05 was considered statistically significant. All analyses were performed using R software (version 4.0.2; R Foundation for Statistical Computing; Vienna, Austria).

A total of 3,273 articles were identified from the mentioned databases as shown in the PRISMA flowchart ( Figure 1 ). After performing deduplication, we observed that 2,674 articles remained and screened their titles and abstracts. Of these articles, 2,638 were excluded due to full text unavailability (n = 20), duplication (n = 120), and irrelevancy (n = 2,494); thus 36 reports remained for full text review and eligibility assessment. Six of them were then excluded because they reported a combined group of premature thelarche and other puberty variants, while 17 others could not provide the outcomes of interest, even after we contacted the corresponding authors. No additional articles were found through the manual search. Finally, 13 studies were included in this systematic review and meta-analysis. They were all deemed suitable for the comparative analysis, and seven were deemed appropriate for the diagnostic accuracy analysis.

A total of 1,977 subjects were available for analysis ( Table 1 ). Among the selected studies in this review, 3 were retrospective studies (10–12), 8 were prospective studies (13, 21, 23–26, 28, 29), and 2 were cross-sectional studies (22, 27). All patients had been referred to outpatient clinics to evaluate early breast development and any other pubertal progression signs. In the GnRH stimulation test, participants with a peak LH value of >5 UI/L were considered to have PP (PP group), and those with a peak LH value of <5 IU/L were considered to have PT (PT group) (14). Pelvic ultrasonography was performed using a conventional full-bladder technique with 3–13.5 MHz transducers and was interpreted by skilled and trained physicians. During the computation of ovarian and uterine volumes, both the ovaries and the uteri were considered ellipses, as demonstrated by the following formula: V = longitudinal diameter (length) × transverse diameter × fundal anterior–posterior diameter × 0.5233 (29).

The Newcastle–Ottawa assessment results revealed that all of the studies were rated as “Good” or “Fair” ( Table S2 ). Moreover, seven studies included in the diagnostic test accuracy meta-analysis yielded acceptable risks of bias using the QUADAS-2 tool ( Table S3 ).

This systematic review and meta-analysis confirmed that pelvic ultrasonography was an appropriate diagnostic tool to differentiate PP from PT. All investigated ultrasonography parameters were significantly greater in the PP group than in the PT group. The early increases in uterine and ovarian sizes represent the estrogenic effects of HPG axis activation on internal female genitalia, which indicates PP. In our meta-analysis, uterine length, CSA, and volume were determined to be valuable markers for differentiating PP from PT. The uterine length of 3.2 cm exhibited satisfactory diagnostic accuracy as indicated by sensitivity and specificity levels of 81.8% and 82.0%, respectively (AUSROC 0.82). It could be readily interpreted from our Fagan’s nomogram that for suspected cases referred to clinics due to breast development before the age of 8 years, a girl who has a uterine length of >3.2 cm would confer an approximately 17-time greater risk of PP than one having a uterine length of ≤3.2 cm. It is thus reasonably recommended that the ultrasonography should be performed during the initial evaluation of PP to help clinicians recognized those with high probability of PP.

The FCR was also determined to be a valuable indicator of puberty. In mid-childhood, the anteroposterior diameter of the uterine fundus and cervix are nearly the same, resulting in an FCR of ≤1. After puberty onset, the fundus widens under hormonal effects relative to the cervix, increasing the FCR to >1. However, previous studies have yielded inconsistent results; for example, some have reported a significantly higher FCR in PP (22, 24, 27), whereas others have not (12, 28). After pooling these studies, we found a meaningful difference in FCR between the PP and PT groups, suggesting that this parameter could successfully differentiate PP from PT. Nevertheless, the FCR might not be reliable in patients aged >7 years because Herter et al. reported that the FCR could not differentiate between different forms of early puberty in this age group (30).

Ovarian parameters are generally inferior PP markers compared with uterine parameters, as confirmed by several studies (12, 23, 25). This is partly because the shape of the ovaries is asymmetrical instead of perfectly oval; therefore pelvic ultrasonography’s results, particularly through the transabdominal approach, are challenging to interpret. Furthermore, the ovarian volume remains relatively constant from birth to puberty; this engenders considerable challenges in differentially diagnosing PT right before the pubertal onset, a period during which the volume in individuals with PP may overlap with that in those with PT (aged approximately seven years) (31). Finally, the ovaries begin to increase in size, in addition to exhibiting other pubertal signs, approximately two years later than the uterus does (31, 32). Thus, this explains the lower sensitivity of ovarian parameters in the early identification of PP compared with uterine parameters.

Numerous pelvic ultrasonography parameters had been suggested to help diagnose PP, including ovarian morphology, quantity of large follicles, maximum follicular diameter, uterine endometrial echogenicity, endometrial thickness, uterine arterial impedance, and vaginal wall thickness. However, none of these parameters have been proven to be reliable indicators of correct pubertal stages and HPG axis activity. Moreover, data on these parameters are insufficient for a meta-analysis because various studies have adopted different definitions and classifications of the parameters. Accordingly, additional studies are warranted to confirm the diagnostic values of these parameters.

According to our literature review, this study is the first to establish an explicit pooled cutoff for uterine length for differentiating PP from PT. Our findings are expected to provide clinicians with a more comprehensive perspective that can help them enhance PP diagnostic accuracy in equivocal cases and determine which patients need treatment. Furthermore, our findings reveal the potential application of several ultrasonography parameters other than uterine length. Accordingly, pelvic ultrasonography could become a complementary diagnostic tool to the GnRH stimulation test. The trim-and-fill result implies that even in the presence of publication bias due to missing studies, our pooled standardized mean differences still reflected true effect size.

Despite covering an appealing topic, our study has some limitations. First, we could include only observational studies rather than randomized controlled trials. However, a large number of girls with early pubertal development were analyzed in this multicenter review, and all of the studies were rated as “Good” or “Fair” in the risk-of-bias assessment and produced an acceptable result when being combined. Second, we found a high degree of heterogeneity among the studies, which could be attributed to unrecognized confounders. Nevertheless, this finding reflects real-world scenarios upon which most pediatric endocrinologists must rely. In these circumstances, the ultrasonography parameters vary by chronological age, abdominal fat mass, degree of bladder fullness, presence of dilated intestinal loops, uterine position, and ultrasonographic equipment resolution. Although studies using low frequency probes tended to have higher SMDs that was not as expected from our general knowledge, they had relatively smaller sample sizes and wider 95% CIs. Therefore, we documented this result but also further suggested that it should be interpreted cautiously. Regarding publication year, ultrasonography equipment with higher probe frequency and higher resolution has been advancing over time. Therefore, the significant result of publication year resulted from meta-regression might be attributable to the difference in probe frequency. Chronological age was negatively associated with effect sizes in meta-regression of uterine length and uterine volume, suggesting that the later these participants were referred to clinics, the smaller differences in uterine parameters were found between PP and PT groups. This was because PT girls tended to approach their normal puberty after following up, thus narrowing the gap between the two groups. In our analysis, chronological age at referral was comparable between two groups in most of the studies. However, even in the presence of age difference between two individuals, these parameters could still be useful in differentiating PP from PT. In more detail, there was a limited progression in ultrasonography parameters from birth to puberty. In other words, ultrasonography measurements are stable or only increase modestly during childhood until the HPG axis activation exerts estrogenic effects on genitalia. It can be seen that such an age difference will not affect the ultrasonography results much unless one actually suffers from PP. Therefore, our meta-analysis findings are relevant to the routine clinical contexts and could be used as a reference during the diagnostic workup for children with suspected PP or PT.

To conclude, girls with PP had significantly greater uterine and ovarian measurements as determined by pelvic ultrasonography than did those with PT. Furthermore, uterine length represented a reliable marker to differentiate PP from PT, thereby reducing the possibility of misdiagnosing PP. Therefore, pelvic ultrasonography emphasizing these measurements should be considered as an adjunct to clinical examination, bone radiography, and laboratory tests to enhance diagnostic precision.