The detailed study selection process is documented in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart. The search strategy is illustrated in Fig. 2. The initial systematic literature search yielded 323 publications. The full texts of 43 publications were examined, and 19 investigations were discarded.

5 papers were ineligible for the following reasons: 1 paper did not provide complete data for this meta-analysis, 1 paper without a control group, 1 paper with no explicit grouping, and 2 papers for other reasons. 19 studies that satisfied the screening requirements were selected for this meta-analysis (Fig. 1).

Literature screening flow chart and results

A total of 550 patients with AIS had PJK after undergoing correction surgery. The overall pooled incidence of PJK was 19% (95% CI 13–25%) based on the 19 studies (Fig. 2). Our results showed that age (WMD − 0.22, 95% CI (− 0.44, 0.00), P = 0.05) (Fig. 3) and body mass index (WMD 0.27, 95% CI (− 0.31, 0.86), P = 0.36) (Fig. 4) were not significantly associated with PJK. Sex (OR 1.40, 95% CI (1.08, 1.83), P = 0.01) (Fig. 5) is significantly associated with PJK.

Pooled incidence of proximal junctional kyphosis

Forest plot of age between the proximal junctional kyphosis (PJK) group and the non-PJK

Forest plot of BMI between the proximal junctional kyphosis (PJK) group and the non-PJK

Forest plot of proximal junctional kyphosis between the male and female groups

Regarding radiographic parameters, meta-analysis results indicated that larger preoperative TK (WMD 6.82, 95% CI (5.48, 8.16), P < 0.00001) (Fig. 6), larger follow-up TK (WMD 8.96, 95% CI (5.62, 12.30), P < 0.00001) (Fig. 6), larger postoperative LL (WMD 2.31, 95% CI (0.91, 3.71), P = 0.001) (Fig. 7), larger follow-up LL (WMD 2.51, 95% CI (1.19, 3.84), P = 0.0002) (Fig. 7), great change in LL (WMD − 2.72, 95% CI (− 4.69, − 0.76), P = 0.006) (Fig. 7), larger postoperative PJA (WMD 4.94, 95% CI (3.62, 6.26), P < 0.00001) (Fig. 8), larger follow-up PJA (WMD 13.39, 95% CI (11.09, 15.69), P < 0.00001) (Fig. 8), larger postoperative PI–LL (WMD − 9.57, 95% CI (− 17.42, − 1.71), P = 0.02) (Fig. 9), larger follow-up PI–LL (WMD − 12.62, 95% CI (− 17.62, − 7.62), P < 0.00001) (Fig. 9), larger preoperative SVA (WMD 0.73, 95% CI (0.26, 1.19), P = 0.002) (Fig. 10), larger preoperative SS (WMD − 3.43, 95% CI (− 4.71, − 2.14), P < 0.00001) (Fig. 11), RCA (WMD 1.66, 95% CI (0.48, 2.84), P = 0.006) (Fig. 12) were identified as risk factors for PJK in patients with AIS.

Forest plot of TK between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of LL between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of PJA between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of PI–LL between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of SVA between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of SS between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of RCA between proximal junctional kyphosis (PJK) and non-PJK groups

However, no significant associations were discerned between postoperative TK (WMD 4.46, 95% CI (− 0.47, 9.39), P = 0.08) (Fig. 6), change in TK (WMD − 3.00, 95% CI (− 7.47, 1.46), P = 0.19) (Fig. 6), preoperative LL (WMD 1.01, 95% CI (− 0.26, 2.28), P = 0.12) (Fig. 7), preoperative PJA (WMD 1.48, 95% CI (− 1.79, 4.75), P = 0.38) (Fig. 8), preoperative SVA (WMD 0.05, 95% CI (− 0.84, 0.93), P = 0.92) (Fig. 10), follow-up SVA (WMD 0.24, 95% CI (− 0.67, 1.14), P = 0.61) (Fig. 10), preoperative PI (− 3.46 1.01, 95% CI (− 6.89, − 0.02), P = 0.05) (Fig. 13), postoperative PI (WMD − 2.82, 95% CI (− 7.44, 1.80), P = 0.23) (Fig. 13), follow-up PI (WMD − 2.17, 95% CI (− 6.42, 2.08), P = 0.32) (Fig. 13), preoperative PT (WMD 0.61, 95% CI (− 2.72, 3.94), P = 0.72) (Fig. 14), postoperative PT (WMD − 2.61, 95% CI (− 5.16, − 0.05), P = 0.05) (Fig. 14), follow-up PT (WMD − 1.87, 95% CI (− 4.05, 0.30), P = 0.09) (Fig. 14), preoperative PI–LL (WMD − 4.96, 95% CI (− 12.07, 2.15), P = 0.17) (Fig. 9), postoperative SS (WMD − 0.21, 95% CI (− 1.87, 1.45), P = 0..80) (Fig. 11), follow-up SS (WMD 0.22, 95% CI (− 1.07, 1.51), P = 0.74) (Fig. 11), postoperative PJA-RCA (WMD 1.27, 95% CI (− 1.05, 3.60), P = 0.28) (Fig. 15), UIV (WMD 0.69, 95% CI (0.18, 2.68), P = 0.59) (Fig. 16) and occurrence of PJK.

Forest plot of PI between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of PT between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of postoperative PJA-RCA between proximal junctional kyphosis (PJK) and non-PJK groups

Forest plot of UIV between proximal junctional kyphosis (PJK) and non-PJK groups

According to the subgroup analysis of AIS classification, it was found that the probability of occurrence of PJK in Lenke 5 type (25%, 95% CI 21–29%) (Fig. 17) was significantly higher than that in other types. Sex in the subgroup (Fig. 18) was not a risk factor for PJK after Lenke 5 AIS. Age (WMD − 0.37, 95% CI (− 0.81, 0.07), P = 0.10) (Additional file 1: Figure S1) was not a risk factor for postoperative PJK.

Subgroup analysis of pooled incidence of proximal junctional kyphosis

Subgroup analysis of forest plot of proximal junctional kyphosis between the male and female groups

Regarding radiographic parameters, meta-analysis results indicated that larger preoperative TK (WMD 7.85, 95% CI (5.69, 10.00), P < 0.00001) (Additional file 1: Figure S2), larger postoperative TK (WMD 9.66, 95% CI (1.06, 18.26), P = 0.03) (Additional file 1: Figure S2), larger follow-up TK (WMD 11.92, 95% CI (6.99, 16.86), P < 0.00001) (Additional file 1: Figure S2), larger preoperative PJA (WMD 0.72, 95% CI (0.03, 1.41), P = 0.04) (Additional file 1: Figure S4), larger postoperative PJA (WMD 5.54, 95% CI (3.57, 7.52), P < 0.00001) (Additional file 1: Figure S4), larger follow-up PJA (WMD 12.42, 95% CI 9.24, 15.60), P < 0.00001) (Additional file 1: Figure S4), larger follow-up SVA (WMD 0.07, 95% CI (− 0.46, 0.60), P = 0.04) (Additional file 1: Figure S5), larger preoperative PT (WMD − 3.04, 95% CI (− 5.27, − 0.81), P = 0.008) (Additional file 1: Figure S7), larger follow-up PT (WMD − 3.69, 95% CI (− 6.66, − 0.72), P = 0.02) (Additional file 1: Figure S7) were identified as risk factors for PJK in patients with Lenke 5 AIS.

However, no significant associations were discerned between larger preoperative LL (WMD –11.72, 95% CI (− 36.09, 12.64), P = 0.35) (Additional file 1: Figure S3), larger postoperative LL (WMD 2.25, 95% CI (− 1.40, 5.90), P = 0.23) (Additional file 1: Figure S3), larger follow-up LL (WMD 3.14, 95% CI (− 1.46, 77.74), P = 0.18) (Additional file 1: Figure S3), preoperative SVA (WMD − 0.41, 95% CI (− 1.05, 0.23), P = 0.21) (Additional file 1: Figure S5), follow-up SVA (WMD 0.07, 95% CI (− 0.46, 0.60), P = 0.79) (Additional file 1: Figure S5), preoperative PI (WMD − 5.62, 95% CI (− 11.80, 0.56), P = 0.07) (Additional file 1: Figure S6), postoperative PI (WMD − 5.66, 95% CI (− 14.60, 3.28), P = 0.21) (Additional file 1: Figure S6), follow-up PI (WMD − 5.89, 95% CI (− 14.69, 2.92), P = 0.19) (Additional file 1: Figure S6), postoperative PT (WMD − 3.95, 95% CI (− 8.43, 0.53), P = 0.08) (Additional file 1: Figure S7), preoperative SS (WMD − 0.49, 95% CI (− 2.14, 1.16), P = 0.56) (Additional file 1: Figure S8), postoperative SS (WMD − 0.21, 95% CI (− 1.87, 1.45), P = 0.80) (Additional file 1: Figure S8), follow-up SS (WMD 0.47, 95% CI (− 1.42, 2.37), P = 0.62) (Additional file 1: Figure S8) and occurrence of PJK.

Sensitivity analysis was carried out by individually calculating and subtracting each study from the meta-analysis in order to ascertain the impact of each one. Publication bias was screened using funnel plots. A P < 0.05 was considered statistically significant. An example is indicated by sensitivity analysis showing the funnel plot of age reported in this meta-analysis for PJK and non-PJK groups (Fig. 19). Any study could be excluded after the heterogeneity test without significantly changing the overall statistical significance, showing that the findings of this meta-analysis were stable. Additionally, the funnel plot's shape was symmetrical, indicating that our study did not contain publication bias.

Risk of publication bias in the included literature