Characteristics of participants
In total, the information of 1493 potentially eligible girls was extracted from the electronic medical databases. According to the inclusion and exclusion criteria, 338 girls with CPP or ENP were preliminarily included, among which 156 girls with missing data of FAH were excluded. A total of 182 patients were finally enrolled in this study, including 54 girls in the CG, 80 girls in the MG, and 48 girls in the NG (Fig. 1).
Flow diagram of participants with CPP and ENP. Notes: GnRH: gonadotropin-releasing hormone; CPP: central precocious puberty; ENP: early normal puberty; FAH: final adult height.
Before the initiation of treatment, there were no differences in the age of puberty onset, disease type, height, weight, BMI, Tanner stage, BA advancement, PAH below the normal range, PAH below the parental target range and PAH below both among the three groups (P > 0.05). The age of treatment initiation of CG (9.61 [0.86]) was larger than MG (9.25 [0.73]) and NG (9.12 [0.63]) (P = 0.003). The HSDS of NG (0.78 [1.00]) was larger than CG (0.02 [0.82]) and MG (0.45 [0.91]) (P < 0.001). The TH of NG (157.39 [4.24]) was larger than CG (155.04 [3.50]) and MG (156.73 [3.97]) (P = 0.007). The BA of CG (11.08 [0.60]) was larger than MG (10.68 [0.60]) and NG (10.58 [0.68]) (P < 0.001). And the PAH of NG (160.86 [5.92]) was larger than CG (156.23 [4.76]) and MG (158.78 [5.56]) (P = 0.007). In addition, the treatment duration of GnRHa was similar between the CG and the MG. The characteristics of the included girls are shown in Table 1.
Final adult height
The unadjusted analysis showed that the mean (SD) of FAH of the CG, MG, and NG was 160.31(5.11) cm, 160.61(5.03) cm, and 160.35(5.06) cm, respectively, and no difference was observed among the three groups (P = 0.935). The FAHSDS was also similar among the CG (0.24[0.87]), MG (0.31[0.93]), and NG (0.24[0.92]) (P = 0.858). Besides, the height gain (CG: (4.08 [4.22])cm, MG: (1.83 [3.61])cm, NG(− 0.51 [3.81]), P1[CG VS MG] = 0.001, P2[CG VS NG] < 0.001, P3[MG VS NG] = 0.001;P < 0.001), genetic height gain (CG: 5.27 [4.51] , MG: 3.88[4.37], NG: 2.97[4.98] , P1 = 0.086, P2 = 0.012, P3 = 0.278, P = 0.039), and HSDS gain (CG: 0.21[0.72], MG: − 0.14[0.62], NG: − 0.54[0.78], P1 = 0.004, P2 < 0.001, P3 = 0.002, P < 0.001) were all different among the three groups.
There are six potential covariates included in the multivariable linear regression analysis model, of which age of treatment initiation, TH, H0SDS, BA0, and PAH0 are selected based on P < 0.05 in the univariate analysis of baseline characteristics, and age of puberty onset is from clinical experience and expert consultation. After adjustment, the results of multivariable linear regression revealed that all of the height outcomes were different among the three groups (P < 0.001). The adjusted mean FAH of the CG, MG, and NG were 162.58 (0.46) cm, 160.25 (0.35) cm, and 158.39 (0.47) cm, respectively. The height gain of the three groups was 4.00 (0.46) cm, 1.68 (0.36) cm, and − 0.18 (0.47) cm, and the genetic height gain of the three groups was 6.18 (0.46) cm, 3.85 (0.35) cm, and 1.99 (0.47) cm. The FAHSDS of the CG, MG, and NG were 0.66 (0.08), 0.24 (0.06), and − 0.13 (0.08), and the HSDS gain of the three groups was 0.26 (0.08), − 0.17 (0.06), and − 0.54 (0.08), respectively (Table 2).
Adverse events
A total of 45 girls (83.30%) had at least one adverse event of any type in the CG, 12 girls (15.00%) in the MG, and 8 girls (16.70%) in the NG, which were different among the three groups (P < 0.001). There were 11 kinds of adverse events involved in this study. Among them, the rate of elevated fasting insulin (CG: 55.6% [30/54], MG: 1.25% [1/80], NG: 2.08% [1/48], P < 0.001) and hypothyroidism (CG: 44.4% [24/80], MG: 0.00% [0/80], NG: 0.00% [0/48], P < 0.001) in the CG were higher than in the MG and the NG. The rates of other adverse effects are listed in Table 3.
The reported adverse events were all mild and no serious adverse event had happened in this study. In addition to 6 patients with elevated fasting insulin and 3 patients with hypothyroidism who failed to recover to normal levels after the discontinuation of medication in the CG, the remaining adverse events that occurred during medication all got back to normal levels after withdrawal.
Subgroup analysis
Among the girls whose puberty onset before 8 years of age, who were also defined as fast progressive CPP, the FAH of the CG and MG was higher than that of NG (5.76 [3.65, 7.88], P < 0.001, 2.30[0.59, 4.02], P = 0.009). When puberty was onset after 8 years of age, who were also defined as ENP, the FAH of the CG was higher than that of NG (3.19 [1.55, 4.83], P < 0.001), and difference was presented between the MG and the NG (1.46 [− 0.01, 2.93], P = 0.051), suggesting that the monotherapy may not improve the FAH in girls who began the puberty after the age of 8 years (Fig. 2).
Subgroup analysis of final height among girls with CPP and ENP. Notes: FAH: final adult height; MD: mean difference; CI: confidence interval; GnRH: gonadotropin-releasing hormone; rhGH: recombinant human growth hormone.
For the girls who initiated the treatment at 6–8 years of age, the difference in FAH was observed between the MG and the NG (2.16 [− 1.61, 5.03], P = 0.262), while the FAH of the CG was higher than NG (6.67 [2.77, 10.56], P = 0.001). When the treatment initiation was at 8–10 years of age, the FAH of the CG and MG was higher than NG (4.81 [3.40, 6.23], P < 0.001, 1.80 [0.66, 2.93], P = 0.002). However, the pharmacotherapy may not improve the FAH in girls who initiated the treatment at 10–12 years of age, neither combination therapy nor monotherapy (1.41 [− 0.76, 3.58], P = 0.204, 1.70 [− 0.77, 4.16], P = 0.178) (Fig. 2).
There was a difference in FAH presented between the MG and the NG when the treatment duration of GnRHa was more than 1 year (1–2 years: 2.38 [1.13, 3.62], P < 0.001; > 2 years: 2.19 [0.31, 4.08], P = 0.023), and a difference was not observed between the two groups when the treatment duration of GnRHa was less than 1 year (0.30 [− 1.34, 1.94], P = 0.718). Besides, a difference was also presented between the CG and the MG when the treatment duration of rhGH was more than 1 year (1–2 years: 2.21 [0.74, 3.68], P = 0.003; > 2 years: 4.29 [2.20, 6.38], P < 0.001), but similar efficacy was observed between the two groups when the treatment duration of rhGH was less than 1 year (1.48 [− 0.09, 3.05], P = 0.064) (Fig. 2).
Sensitivity analysis
After excluding the girls with missing data of baseline variables, the results were consistent with those of the main analysis, suggesting that the results of multivariable linear regression were robust (Appendix S1). The baseline characteristics of enrolled patients with available data of FAH in each group were compared with those who were excluded due to the lack of FAH on account of the loss to follow-up, and the results showed that there was no difference in the baseline variables between the follow-up population and the loss to follow-up population in each group, indicating that the possibility of selection bias caused by the loss to follow-up was relatively low and had little effect on the results (Appendix S2).