This meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [20]. We conducted a literature search in PubMed, Ovid MEDLINE, and MEDLINE databases to cover all articles published until June 2021, with the following search terms in their titles, abstracts or keyword lists: ‘PTK6 or Protein tyrosine kinase 6 or Brk or Breast tumor kinase’ and ‘cancer or carcinoma or malignancy or tumor or neoplasm’. Furthermore, we inspected the reference lists of the relevant articles to identify additional articles. We evaluated all potentially relevant articles by examining their titles and abstracts. Subsequently, the full text of eligible articles was screened to retrieve all articles that met the inclusion criteria.

Studies that discussed the following points were included in this meta-analysis: (i) evaluation of PTK6 expression by immunohistochemical (IHC) staining; (ii) relationships between PTK6 expression level and clinicopathological features of the patients or sufficient published data to estimate the hazard ratio (HR) for survival; and (iii) articles written in English.

The collected characteristics of the included studies were as follows: first author, year of publication, number of patients, clinicopathological features (lymph node [LN] involvement, stage, T classification, and pathological grade), and survival outcomes (overall survival [OS] and disease-free survival [DFS]) along with their HRs with 95% confidence intervals (CIs). When HR values of both univariate and multivariate analyses were available, the data from the multivariate analysis were selected preferentially.

Two investigators independently extracted data from each eligible study using a standard data collection form and assessed the risk of bias. Discrepancies were resolved through discussion.

Pooled odds ratios (ORs) with 95% CIs were calculated to assess the association between PTK6 expression level and clinicopathological features such as LN metastasis, advanced stage, T classification, and pathological grade. HRs with 95% CIs were estimated using statistical values directly extracted from the original articles. If HRs with their 95% CIs were not provided, they were estimated from the Kaplan-Meier curves using the plotdigitizer software. The significance of the OR and HR was determined using the Z-test. Statistical significance was set at p 0.05.

The traditional Q-test and I${}^{\mathrm{2}}$ statistic were used to evaluate heterogeneity. A p value $\ge$ 0.10 for the Q-test or I${}^{\mathrm{2}}$ $\le$ 50% indicates that there was no significant heterogeneity among the studies, and the fixed-effect model (Mantel–Haenszel method) was used. However, the random-effects model (DerSimonian-Laird method) was adopted when significant heterogeneity was observed (p 0.10, I${}^{\mathrm{2}}$ $>$ 50%).

Publication bias was assessed using Begg’s test and Egger’s linear regression test [21, 22]. Statistical significance was determined by t-test, and a p value of 0.05, which was representative of significant publication bias.

Fig. 1 shows a flow diagram illustrating the literature search process. Altogether, 596 potentially relevant articles were initially identified, of which 364 duplicates were removed. Of the remaining articles, 179 were excluded by screening the titles and abstracts. The full text of the remaining 53 potentially eligible studies was reviewed, and 45 articles were further excluded: 27 articles did not present appropriate data for meta-analysis; 13 articles were biological studies and the data of one study [8] was duplicated in another article by the same authors [9]. Finally, eight studies were included in the meta-analysis [9, 11, 12, 13, 14, 15, 16, 17].

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flow diagram showing the search process of studies included in this meta-analysis.

The main characteristics and clinicopathological findings of the selected studies are presented in Table 1 (Ref. [9, 11, 12, 13, 14, 15, 16, 17]). Most of the studies were retrospective. All 1475 patients from the eight studies were included in the meta-analysis. All patients were surgically treated, except for nasopharyngeal cancer patients [12] and none of the patients received neoadjuvant treatment. All studies used immunohistochemistry (IHC) to assess PTK6 expression status and provided the criteria for PTK6 expression status, although slightly different cutoff values were adopted to define PTK6 expression. IHC staining was performed on non-tumor tissues as a reference in all the studies. Breast cancer patients received adjuvant hormonal therapy, radiotherapy, and/or chemotherapy when indicated [9] and curative radiotherapy with or without chemotherapy was administered to patients with nasopharyngeal cancer [12].

The rates of high PTK6 expression varied, ranging from 29.1% to 68.8%. Non-tumor tissues from the larynx and esophagus showed high expression of cytoplasmic PTK6 [11, 15], while other non-tumor tissues were stained less than tumor tissues.

As mentioned above, non-tumor tissues showed heterogeneity in PTK6 expression. Therefore, the included studies were divided into two subgroups according to PTK expression in non-tumor tissues: the low expression subgroup (LESG) or high expression subgroup (HESG), and subgroup analysis was thus performed.

From seven studies [11, 12, 13, 14, 15, 16, 17], 1049 patients were included in the meta-analysis of ORs with 95% CIs for LN metastasis and T classification. There was no significant association between PTK6 expression and LN metastasis (OR = 1.10 [95% CI, 0.59–2.06], p = 0.76) (Supplementary Fig. 1), but a positive correlation between high PTK expression and advanced T classification was observed in the whole group (OR = 1.85 [95% CI, 1.23–2.77], p = 0.003) and LESG (OR = 2.35 [95% CI, 1.69–3.28], p 0.00001) (Supplementary Fig. 2).

From four studies [11, 12, 13, 14, 15], 748 patients were included in the meta-analysis of ORs with 95% CIs for tumor stage (I, II vs. III, IV). Compared with low PTK6 expression, tumors with high PTK6 expression showed significantly lower rates of advanced stages in HESG (OR = 0.60 [95% CI, 0.38–0.95], p = 0.03), while no significant association was observed in LESG (OR = 2.23 [95% CI, 0.85–5.85], p = 0.10) (Supplementary Fig. 3).

For pathological grade, tumors with high PTK6 expression showed a significantly lower rate of advanced pathological grade in HESG (OR = 0.13 [95% CI, 0.04–0.44], p = 0.001) (Supplementary Fig. 4). The results are summarized in Table 2.

From five studies [11, 12, 15, 16, 17], 823 patients were included in the meta-analysis of HRs with 95% CIs for OS. There was no correlation between PTK expression and overall survival (HR = 1.41 [95% CI, 0.61–3.25], p = 0.42); however, it became apparent when the analysis was restricted to subgroup: patients with high PTK expression showed significantly worse OS in LESG (HR = 2.53 [95% CI, 1.68–3.83], p 0.0001) (Fig. 2A), but significantly better OS in HESG (HR = 0.56 [95% CI, 0.40–0.78], p = 0.0006) (Fig. 2B).

Fig. 2.

Forest plots of hazard ratio for overall survival. (A) Low expression subgroup; (B) High expression subgroup.

Meta-analysis of HRs with 95% CIs for DFS was performed in only two studies of 636 patients. Compared to low PTK6 expression, high PTK6 expression in cancer tissues was associated with improved DFS (HR = 0.61 [95% CI, 0.47–0.79], p = 0.0002) (Fig. 3).

Fig. 3.

Forest plot of hazard ratio for disease-free survival.

Begg’s test and Egger’s test were conducted for all ORs and HRs, and none of the p values of either test were less than 0.05, indicating that there were no substantial publication biases.