To ensure the quality of the research, every step of the research strictly followed the standard Preferred Reporting Items for Systematic Reviews and Meta-Analyses (36). To obtain potential literature relevant to PJK in AIS surgery, we implemented a systematic search based on the following abstract databases including PubMed, Web of Science, and Embase December 12th, 2023. The following were the search terms: (proximal junctional kyphosis) AND (scoliosis) AND (adolescent OR pediatric).

Inclusion criteria: (1) case-control, retrospective, or cohort study designs based on AIS patients; (2) posterior spinal fusion was performed for each case; (3) the study mainly reported the incidence and risk factors of PJK;(4) the diagnostic criteria of PJK was that the kyphotic angle between the lower endplate of UIV and the upper endplate of UIV + 2 was greater than 10°, which is more than 10° higher than that before surgery (16); (5) the literature was written in English. Exclusion criteria: (1) the study focused on other types of scoliosis; (2) anterior spinal fusion was performed; (3) the diagnostic criteria of PJK did not meet that the kyphotic angle between the lower endplate of UIV and the upper endplate of UIV + 2 was greater than 10°, which is more than 10° higher than that before surgery (16); (4) the literature was not written in English. According to the inclusion criteria, the two authors independently judge the inclusion or exclusion of the literature. If there is a dispute about the inclusion of a certain literature, the research group will decide whether to include it in this meta-analysis through discussion and voting. The Newcastle-Ottawa quality assessment scale (NOQA) was used to assess the quality for each paper.

The following items were extracted from papers meeting the inclusion criterion: (1) the author's name, publication year, study design, follow-up period, age, and gender distribution; (3) radiographic parameters: thoracic kyphosis (TK), global thoracic kyphosis (GTK) lumbar lordosis (LL), pelvic tilt (PT), pelvic incidence (PI), pelvic incidence minus lumbar lordosis (PI-LL), sagittal vertical axis (SVA), proximal junctional angle (PJA), RCA(rod contouring angle); (4) Surgery details: the fusion segments, screws at UIV, UIV hooks at UIV.

The Begg's test was conducted and the funnel plot was employed to evaluate the publication bias.

Sensitivity analyses were performed by omitting each included paper in sequence.

The statistics were pooled using the Review Manager Version 5.3. For the enumeration data, the odds ratio (OR) with the 95% confidence interval (95%CI) was calculated, while the weight mean difference (WMD) with the 95%CI was employed to assess the difference of measurement data between groups. That P < 0.05 meant the statistical significance between PJK and Non-PJK groups. That I² ≥ 50% meant that the random effect model should be used, while the fixed effect model would be used when I² < 50%.

Through systematic retrieval of the databases, 542 articles were retrieved. Among them, 24 articles met the inclusion criteria (21–29, 31–35, 37–46). Four studies groups were based on Lenke 1/2 AIS patients (26–28, 44), and 8 studies were based on Lenke 5 AIS patients (21–25, 34, 35, 46), while the remaining studies did not clearly distinguish different types of AIS when investigate the incidence of PJK and its influencing factors (29, 31–33, 37–43, 45) (Table 1). Figure 1 demonstrates the details of the literature screening procedures. A total of 4,063 AIS cases were included in this meta-analysis, and the incidence of PJK was 17.67% for the whole population (PJK = 718, non-PJK = 3,885). However, the corresponding rate of PJK inclined to 23.88% in Lenke type 5 AIS patients (PJK = 166, non-PJK = 529).

There was no difference in age at surgery (p = 0.47), gender distribution (p = 0.1), BMI (p = 0.52), Risser sign (p = 0.37), or follow-up time (p = 0.89) between groups. However, the pooled results showed more fusion segments (WMD = 0.49, 95% = 0.29∼0.68, p < 0.001) in PJK groups. Compared with the use of screws at UIV, the use of hooks (OR = 0.48, 95% = 0.31∼0.47, p = 0.001) could prevent PJK (Table 2).

Before surgery, the larger TK (WMD = 8.49, 95%CI = 7.69∼.30, p < 0.001), and GTK (WMD = 10.1, 95%CI = 5.55∼14.66, p < 0.001) were observed in PJK group, and a lager LL (WMD = 4.11, 95%CI = 3.40∼4.82, p < 0.001) was also presented in PJK group (Figure 2). For the pelvic parameters, a smaller PT (WMD = −1.02, 95%CI = −1.85∼−0.18, p = 0.02) was observed in the PJK group. However, there was no difference in PI (p = 0.16), SS (p = 0.38), PI-LL (p = 0.14), SVA (p = 0.65), RCA (p = 0.39), and PJA (p = 0.94).

Immediately postoperatively, the pooled results demonstrated higher values in the PJK group for the following parameters: TK (WMD = 5.09, 95%CI = 2.01∼8.17, p = 0.001), GTK (WMD = 10.07, 95%CI = 7.05∼13.09, p < 0.001), LL (WMD = 3.53, 95% CI = 0.21∼6.85, p = 0.04), PJA (WMD = 5.00, 95%CI = 4.03∼5.94, p < 0.001), and PJA-RCA (WMD = 3.89, 95%CI = 1.53∼6.25, p = 0.001) (Figure 3). However, a smaller PI-LL (WMD = −7.52, 95%CI = −14.79∼−0.25, p = 0.04) was observed in the PJK group, and no difference was demonstrated in PI (p = 0.34), PT (p = 0.15), SS (p = 0.80) and SVA (p = 0.06).

At the final follow-up, the following parameters were higher in the PJK group: TK (WMD = 5.13, 95%CI = 4.09∼6.16, p < 0.001), GTK (WMD = 12.30, 95%CI = 8.16∼15.99, p < 0.001), LL (WMD = 3.33, 95%CI = 2.09∼4.56, P < 0.001), and PJA (WMD = 13.20, 95%CI = 11.06∼15.34, P < 0.001). However, a smaller PI-LL (WMD = −11.65, 95%CI = −17.48∼−5.81, p < 0.001) was observed in the PJK group, and no difference was demonstrated in PI (p = 0.19), PT (p = 0.05), SS (p = 0.77) and SVA (p = 0.97) (Table 3).

Before surgery, the following parameters were larger in the PJK group: TK (WMD = 8.39, 95%CI = 6.32∼10.46, p < 0.001), LL (WMD = 5.14, 95%CI = 2.88∼7.39, p = 0.005), and PJA (WMD = 0.79, 95%=0.07∼1.50, p = 0.03). For the pelvic parameters, a smaller PT (WMD = −2.10, 95%CI = −3.47∼−0.73, p = 0.003) was observed in the PJK group. However, there was no difference in PI (p = 0.05), SS (p = 0.43), PI-LL (p = 0.18), and SVA (p = 0.58).

Immediately post-operatively, the pooled results demonstrated higher values in the PJK group for the following parameters: TK (WMD = 7.04, 95%CI = 3.63∼10.46, p = 0.02), LL (WMD = 3.06, 95%CI = 0.90–5.22∼−0.90, p = 0.005), and PJA (WMD = 5.54, 95%CI = 3.57∼7.52, p < 0.001). However, a smaller PI-LL (WMD = −9.76, 95%CI = −17.90 ∼−1.63, p = 0.02) was observed in the PJK group, and no difference was demonstrated in PI (p = 0.33), PT(p = 0.09), SS (p = 0.80) and SVA (p = 0.58).

At final follow-up, the following parameters were higher in the PJK group: TK (WMD = 9.51, 95%CI = 5.03∼13.99, p < 0.001), LL (WMD = 4.75, 95%CI = 2.57∼6.93, P < 0.001), and PJA (WMD = 11.47, 95%CI = 8.21∼14.74, P < 0.001). However, a smaller PI-LL (WMD = −13.23, 95%CI = −19.70∼−6.75, p < 0.001) and a smaller PT (WMD = −3.70, 95%CI = −6.75∼−0.66, p = 0.02) were observed in PJK group, and no difference was demonstrated in PI (p = 0.12), SS (p = 0.90) and SVA (p = 0.56) (Table 4).

There was no significant difference in the preoperative (p = 0.18) and postoperative SRS-22 scores (p = 0.46) between the PJK group and the non-PJK group.

Based on the preoperative TK Begg's test was performed, and a symmetrical funnel graph was obtained. So, there was no significant publication bias in this meta-analysis (Figure 4).

No single paper resulted in huge fluctuations in the pooled results.

The incidence of PJK reported in previous studies fluctuates greatly. Based on the diagnostic criteria of PJK by Glattes (16), the highest PJK rate varied from 3.33% to 42.86% (23, 26). While the majority of papers reported that the rate of PJK was 15∼30% (32, 33, 40–42). The current meta-analysis detected an incidence of PJK to be 16.50% for the whole population, while the corresponding figure inclined to 23.88% in Lenke type 5 AIS patients. Previously, several studies also reported that Lenke type 5 AIS patients were more susceptible to PJK after PSF with the rates varying from 18.57% to 42.86% (21–25, 34, 35). Therefore, the PJK phenomenon in AIS should be paid enough attention.

Previous studies have investigated various influencing factors that might be related to PJK in AIS patients. Among all the factors, the most reported ones are radiographic factors, and the current research mainly investigated the influence of sagittal spinopelvic parameters on PJK. It is commonly accepted that the UIV always locates in the thoracic segments, which means that the TK is always composed of PJA. Before surgery, Boeckenfoerde et al. (38) reported that the PJK group had significantly larger T4–T12 kyphosis (31.1° ± 13.93° vs. 23.3° ± 14.93°, p = 0.016). Kim et al. (30) reported that the preoperative TK > 40° was one of the risk factors for PJK. Currently, the pooled results also demonstrated the larger TK (WMD = 8.49, 95%CI = 7.69∼.30, p < 0.001), and larger GTK (WMD = 10.1, 95%CI = 5.55∼14.66, p < 0.001) in PJK group. Correspondingly, other studies also detected a more kyphotic thoracic spine in the PJK group before surgery (32, 33, 38). The subgroup analysis also detected a larger TK in the PJK group in Lenke type 5 AIS patients, which was also consistent with several previous reports (21, 22, 25, 34). Generally, the greater the thoracic kyphosis, the greater the lumbar lordosis. Ferrero et al. (28) and Clément et al. (39) reported that the preoperative LL was significantly greater in the PJK group. However, several reports did not detect differences in preoperative LL between groups (22–24, 32, 34, 38). The pooled result detected a larger preoperative LL (WMD = 3.53, 95%CI = 0.21∼6.85, p = 0.04) in the PJK group, and a similar result was also presented in subgroup analysis LL (p = 0.005). So, it is necessary to take appropriate measures to prevent PJK for patients with large TK before surgery, especially for those with TK greater than 40 degrees (30).

The pooled results detected that both TK and LL in the PJK group were greater in the non-PJK group immediately after operation and at the last follow-up. After PSF, the patient's thoracic kyphosis and lumbar lordosis are not completely developed and shaped. It was commonly accepted that the formation of physiological curvature of the thoracic kyphosis and lumbar lordosis involved the whole spinal segments. Correction surgery fixed several segments which were relatively stable after surgery. This was bound to interfere with the formation of thoracic or lumbar curvature. So, the proximal unfused segments may increase kyphosis in compensation for the development of thoracic kyphosis, which may be the potential mechanism of the increasing PJA after surgery. However, it was reported that the increase in PJA might mainly compensate for the SVA increase in adult spinal deformity patients, while the decompensation led to PJK (47). Therefore, there may be some differences in the main mechanism of PJK in different age groups after surgery.

In terms of pelvic parameters, only a smaller preoperative PT was detected in the PJK group, which was consistent with the reports by Wang et al. (22) and Chen et al. (23). At the last follow-up, subgroup analysis also detected a smaller PT in the PJK group. On the contrary, one of the main influencing factors of PJK is the larger preoperative PT in adult spinal deformity, and the follow-up PT was larger in the PJK group. This again showed that there were differences in the main influencing factors and evolution process of PJK between AIS and adult spinal deformity patients. PJK-related theory, which is applicable to explain adult degenerative spinal deformity, cannot be used to clarify the PJK phenomenon in AIS patients. In adult spinal deformity, correction surgery mainly reconstructed the SVA, LL, and PT, while these parameters kept deteriorating with aging. The fixed segments failed to compensate for the sagittal spinal misalignment, and the increased PJA could compensate for the trunk forward (47). However, in AIS patients the proper sagittal balance is always maintained both before and after surgery and the PJA increase might be a manifestation of thoracic kyphosis. Furthermore, the larger PJA-RCA was detected in the PJK group, which also reflected that the PJA was mainly influenced by the adjacent segments rather than the whole sagittal alignment (27).

In order to prevent PJK after surgery for adult spinal deformity, surgeons tried many surgical strategies, such as vertebral augmentation, ligament strengthening, semi-rigid structure at the proximal construct, and transverse process hook at UIV (47). However, there are not so many preventive measures in the AIS population. The most reported PJK prevention measure after PSF in AIS is to use lamina hook or transverse claws for UIV, and the reported results are inconsistent (28, 31, 40). The pooled results showed that the use of hooks for UIV (OR = 0.48, 95% = 0.31∼0.47, p = 0.001) could prevent PJK. In addition to proximal implants, other surgical-related PJK influencing factors have also been reported, such as disruption of junctional ligaments (25), exposure of adjacent joint (48), and thoracoplasty (42).

One of the main reasons why the phenomenon of PJK is getting more and more attention is that PJK can lead to poor clinical outcomes, which has been widely confirmed in adult degenerative spine deformity patients (49). However, for AIS, the researchers did not detect that PJK would have a significant impact on its clinical outcome (25, 35). These pooled results also did not find any difference in the preoperative and postoperative SRS-22 scores between the PJK group and the non-PJK group. The reason for this result may be that the current follow-up time is too short to show the effect of osteoporosis and muscle degeneration on the accelerated deterioration of PJK. We cannot be blindly optimistic that PJK patients after AIS still have no clinical symptoms after aging.

Even though the current meta-analysis retrieved papers on PJK after PSF in AIS patients, the following shortcomings should be taken into consideration. Firstly, the sample sizes were small for some papers included in this meta, which might lead to heterogeneity. Secondly, the current meta-analysis could not perform the subgroup analysis based on different curve types. Finally, only retrospective studies were included, and further prospective studies were needed to detect the primary risk factor of PJK in AIS.