Included Studies

The study retrieved 4302 articles from the PUBMED, Medline, Web of Science, CBM, and CNKI databases. After 1336 duplicate studies were removed, 2966 studies were retained pending title and abstract screening. Thereafter, 183 records were carefully reviewed to ensure compliance with the inclusion criteria. Ultimately, 17 studies [10–12, 20,21,22,23,24,25,26,27,28,29,30,31,32,33], comprising 10,525 patients who underwent VATS, were included in this study (Fig. 1).

Fig. 1

Flow diagram of identified and selected studies

Study Characteristics

The main characteristics of the included studies are summarized in Table 1.

Table 1 Characteristics of the included studies

Of the 17 included studies, 3 were prospective and 14 were retrospective. Fourteen of the studies were conducted in China; the rest were conducted in South Korea, Italy, and Israel. Of the 10,525 patients who underwent VATS, 2526 had CPSP and 7999 did not have CPSP. The primary diagnosis of the included patients was most commonly lung cancer, accompanied by pectus excavatum and spontaneous pneumothorax. The follow-up period was usually 3 months or more; only three studies included a 2-month follow-up period, owing to historical differences in the definition of CPSP [4, 34].

Methodological Quality

The methodological quality of the included studies varied, with NOS scores ranging from 5 to 8 out of 9 (median 7) (see Table S1 in the electronic Supplementary Material). The included studies that scored low in the selection and outcome/exposure categories. The most common sources of risk included the lack of a prospective design, inadequate matching for important confounders, and insufficient information on participants lost to follow-up. However, the selection of study populations, predictors, and measurement of outcome measures were reliable and valid. This suggested that the included studies had an acceptable risk of methodological bias.

Prevalence of CPSP After VATS

Based on the 17 included studies, the overall incidence rate of CPSP after VATS was 35.3% (95% CI 27.1–43.5%) when pooled using a random-effects model. The results indicated significant heterogeneity between the studies (I2 = 99.13%). Given that discrepancies in the definition of CPSP may be a source of heterogeneity, we performed subgroup analyses of different pain scores. Pain scores > 3 were based on numerical rating scales, whereas pain scores > 1 indicated the presence of CPSP. Subgroup analysis showed a 41.0% (95% CI 34.6–47.3%) incidence rate of pain score ≥ 1 for CPSP after VATS, although only a 10.5% (95% CI 5.5–15.6%) incidence rate of pain score ≥ 3 after VATS. Figure 2 shows the results of the 17 studies that reported the occurrence of CPSP after VATS.

Fig. 2

Results of the subgroup analyses of the prevalence of chronic postsurgical pain after video-assisted thoracoscopic surgery. ES effect size, CI confidence interval

Predictors of CPSP After VATSQualitative Synthesis

Predictors of CPSP after VATS were confounded, and predictor estimates were contradictory; therefore, the predictors were categorized and pooled according to methodological quality and number of studies (Table 2).

Table 2 Estimates of the association between unique predictors and chronic postsurgical pain

All studies had 30 factors reported in the multivariate analysis at least once, of which 19 were independent factors with positive results. Female sex, age, and acute postsurgical pain (APSP) were identified as definite independent predictors of CPSP after VATS. In addition, there were four predictors with at least two high-quality studies revealing positive results; thus, they were classified as likely independent predictors of CPSP. These predictors were the number of ports, operation time, duration of drainage, and insufficient analgesia. Of the non-predictors, pathological stage was not considered an independent predictor of CPSP because two studies with a low risk of bias indicated negative results, and no study offered positive results. For the remaining predictors, no clear judgment could be made from the adjusted results owing to the small number or insufficient quality of the studies.

Quantitative Synthesis

We performed a meta-analysis of predictors evaluated in two or more studies. The unadjusted OR (uOR) was merged with the adjusted OR (aOR) to make the estimates of each predictor more thorough [18], as this permits the inclusion of studies with negative outcomes. Continuous variables from the baseline data were pooled to determine the predictive effect of these factors.

Female Sex

The association of female sex with the occurrence of CPSP was investigated in 17 studies that included 10,525 patients who underwent VATS. Six uORs were employed to supersede absent values because of their nonsignificance. Owing to the high heterogeneity between studies (I2 = 84.2%), the 17 studies were pooled using a random-effects model. The results revealed that female patients who underwent VATS had a significantly higher risk of CPSP than that of male patients (OR 1.58; 95% CI 1.20–1.96) (Fig. 3I).

Fig. 3

Forest plot of predictors of chronic postoperative pain after video-assisted thoracoscopic surgery. I Female sex, II OR used as the effect estimate of age, III SMD used as the effect estimate of age. CI confidence interval, OR odds ratio, SMD standardized mean difference

Advanced Age

Nine studies were included in the analysis of the influence of age on the occurrence of CPSP after VATS, including six multivariate analyses and three univariate analyses. The combined effects showed that older patients had a lower risk of CPSP after VATS than younger patients (OR 0.92; 95% CI 0.85–0.99) (Fig. 3II). A meta-analysis of continuous variables across the eight studies revealed that younger patients were more likely to develop CPSP after VATS (SMD =  − 0.16; 95% CI − 0.24 to 0.07) (Fig. 3III). Random-effects models were used to pool the effects because of considerable heterogeneity between studies (I2 = 76.4% and 92.5%, respectively).

Acute Postsurgical Pain

Nine multifactorial studies indicated an association between APSP and CPSP events after VATS. Our analysis indicated that patients with APSP after VATS were more likely to develop CPSP than their counterparts (OR 1.84; 95% CI 1.57–2.12) (Fig. 4I). Moreover, a pooled analysis of APSP scores based on patient characteristic data revealed that patients with CPSP had higher APSP scores than their counterparts (SMD 0.67; 95% CI 0.08–1.26) (Fig. 4II). Owing to the small heterogeneity between the aforementioned studies (I2 = 49.1% and 0%), a fixed-effects model was utilized to assess their pooled effects.

Fig. 4

Forest plot of predictors of chronic postoperative pain after video-assisted thoracoscopic surgery. I OR used as the effect estimate of acute postoperative pain, and II SMD used as the effect estimate of acute postoperative pain, III postoperative analgesia, IV OR used as the effect estimate of operative time. CI confidence interval, OR odds ratio, SMD standardized mean difference

Postoperative Analgesia

The aORs of five studies and the uORs of one study were pooled to estimate the effect of postoperative analgesia on the development of CPSP after VATS. A random-effects model was used to assess the results because of the notable heterogeneity between studies (I2 = 88.0%). The pooled results showed that postoperative analgesia significantly reduced the risk of CPSP after VATS (OR 0.54; 95% CI 0.17–0.91) (Fig. 4III).

Port Number

Multivariate analysis of two studies examined the effect of port number on CPSP after VATS. There was insignificant heterogeneity between the studies (I2 = 30.4%); therefore, a fixed-effects model was used to assess the pooled results. The analysis revealed that three-port thoracoscopic surgery increased the risk of CPSP (OR 4.32; 95% CI 3.51–5.13) (Fig. 4IV).

Operative Time

The aORs of three multivariate analyses were combined to determine whether the operative time was a predictor of CPSP after VATS. On the basis of the slight heterogeneity between studies (I2 = 40.5%), a fixed-effects model was used and revealed that the risk of CPSP increased with a prolonged operation time (OR 1.007; 95% CI 1.003–10.010) (Fig. 5I). However, the results of the random-effects model for a continuous variable with significant heterogeneity (I2 = 76.8%) indicated that the difference in operative time between the CPSP and non-CPSP groups was not statistically significant (SMD 0.13; 95% CI − 0.04 to 0.29) (Fig. 5II).

Fig. 5

Forest plot of predictors of chronic postoperative pain after video-assisted thoracoscopic surgery. I SMD used as the effect estimate of operative time, II port number, III OR used as the effect estimate of duration of drainage, and IV SMD used as the effect estimate of duration of drainage. CI confidence interval, OR odds ratio, SMD standardized mean difference

Duration of Drainage

Six studies evaluated drainage duration as a predictor using multivariate analysis. The results of the meta-analysis showed that a longer duration of postoperative drainage increased the risk of CPSP after VATS (OR 1.08; 95% CI 1.05–1.12) (Fig. 5III). This resulted from fixed-effects model pooling based on the low heterogeneity (I2 = 34.1%). However, analysis derived from baseline data showed no significant difference between the CPSP and non-CPSP groups regarding drainage time (SMD 0.15; 95% CI − 0.25 to 0.55) (Fig. 5IV). This was based on the results of a random-effects model owing to the apparent heterogeneity (I2 = 93.7%).

Postoperative Chemotherapy

After one missing variable was added that was nonsignificant in the univariate analysis, three studies involving the effect of postoperative chemotherapy on the occurrence of CPSP after VATS were included in the meta-analysis. Our analysis showed that postoperative chemotherapy significantly increased the risk of CPSP occurrence after VATS (OR 1.58; 95% CI 1.07–2.09) (Fig. 6I). No heterogeneity existed between the studies (I2 = 0.0%); thus, a fixed-effects model was employed to evaluate the pooled results.

Fig. 6

Forest plot of predictors of chronic postoperative pain after video-assisted thoracoscopic surgery. I Postoperative chemotherapy, II educational level less than junior school, III smoking history, and IV drinking history. 95% CI 95% confidence interval, OR odds ratio

Educational Level Less Than Junior School

Only three studies evaluated the effect of educational level on the occurrence of CPSP after VATS. The included studies had low heterogeneity (I2 = 0%), and a fixed-effects model was used to evaluate the results. The pooled effect showed that an educational level less than junior school was more likely to lead to CPSP (OR 1.27; 95% CI 1.05–1.48) (Fig. 6II).

Smoking and Drinking Histories

Our meta-analysis results showed that patients undergoing VATS with smoking or drinking histories had a lower risk of CPSP than their counterparts (OR 0.80; 95% CI 0.65–0.95 and OR 0.77; 95% CI 0.60–0.94, respectively) (Fig. 6III, IV). Although the analysis was based on a fixed-effects model because the heterogeneity between the studies was low (I2 = 19.2%; I2 = 0.0%), the results should be interpreted with caution because they did not adjust for the effect of sex.

Negative Results

Postoperative rescue analgesic use, American Society of Anesthesiologists grade, body mass index, weight, tumor nature and stage, intraoperative fentanyl and remifentanil use, intraoperative nerve block use, blood loss, hypertension, and diabetes mellitus showed negative results in this meta-analysis (see Figs. S1–S10 in the electronic Supplementary Material). On the basis of the pooled effects, these factors were not associated with the occurrence of CPSP after VATS.

Publication Bias and Sensitivity Analysis

In our meta-analysis, funnel plots showed publication bias for three factors: female sex, age, and APSP. The results revealed incomplete symmetry between the left and right sides of the funnel plot, suggesting the existence of some publication bias (see Fig. S11 in the electronic Supplementary Material). Nevertheless, quantification using Egger’s linear regression revealed no significant bias for female sex (P = 0.92), age (P = 0.25), and APSP (P = 0.11) (see Fig. S12 in the electronic Supplementary Material). Furthermore, the effect of publication bias on the stability of the results was detected using the trim-and-fill method. After the trim-and-fill analysis, the results did not change; therefore, the pooled results were considered robust (see Fig. S13 in the electronic Supplementary Material). This study also performed a sensitivity analysis of the defined predictors of CPSP after VATS. The sensitivity analysis indicated that the meta-analysis results for each exposure factor were consistent with low sensitivity, suggesting that the results of this study were stable and reliable (Fig. S14 in the electronic Supplementary Material).