Literature research
A total of 1039 records were searched in Medline, PubMed, Embase, Cochrane and Web of Science. After excluding duplicates, 619 records remained. Next, we screened the titles and abstracts of the 619 papers, and only 42 papers were included for further full-text review. Among these papers, 20 papers were excluded because they did not provide relevant data in estimating TILs in DCIS, and another 5 were conference abstracts that displayed the same data as other included papers. Next, 2 papers were excluded because they focused on infiltrating macrophage cells and TIL-Bs in DCIS. Ultimately, 15 articles including 17 sets of studies were included for quality assessment. Through quality assessment, 2 articles were excluded because their NOS score are fewer than 7. Finally,13 articles including 15 sets of studies were used for following meta-analysis. Among these 13 studies, 12 studies including 14 studies investigated the prognostic role of total TILs in DCIS, and 6 articles containing 10 studies explored the value of TIL subsets in DCIS (Figure 1).
Included studies’ characteristics
Detailed characteristics of the included articles are listed in Table 1. These articles were conducted in the United States (2), China (1), Europe (1), Australia (2), the United Kingdom (3), Singapore (1), the Netherlands (1), Italy (1) and Belgium (1), including approximately 4335 participations. All 13 articles were retrospective cohort studies, 1 of the 13 was a conference abstract, and the others were full-reported articles. Among these, 12 articles, which included 14 sets of studies, evaluated the relationship between TILs and DCIS recurrence, and 6 articles investigated the prognostic value of TIL subtypes (CD4+, CD8+, FOXP3+, PD-L1) in DCIS.
Study quality and risk of bias
After full-text review, we performed critical assessment for each study by NOS, and the quality of each study is summarized in Figure 2. Most of the studies exhibit excellent quality with more than six stars. Two studies scored fewer than 7 stars due to missing data, unclear TIL assessment method and univariate analysis. Studies with fewer than 7 stars were excluded from our meta-analysis; therefore, 15 studies were eligible for inclusion in the meta-analysis.
The value of total TILs in the recurrence of DCIS
A total of 14 studies in 12 sets of articles containing 4335 patients were included in our meta-analysis to evaluate the prognostic value of total TILs in DCIS. The results showed that dense TILs in DCIS indicates a higher recurrence risk. The pooled HR was 2.11 (95% CI, 1.35-3.28) for the total TIL level (dense vs. sparse), with statistically significant heterogeneity (I2 = 78.3%, P = 0.000) (Figure 3A).
To further investigate the heterogeneity in our meta-analysis, we performed meta-regression and subgroup analyses (Table 2). Through meta-regression analysis, we identified “TIL assessment method” as the main cause of heterogeneity (P>|t| = 0.005). Furthermore, in subgroup analysis, we also observed that the assessment method of TILs in different studies may affect the prognostic value of TILs in DCIS. The pooled HR for 11 studies using TIL ratio classification was 1.49 (1.11-1.99), with no obvious heterogeneity between the results of the studies (I2 = 28.4%, P = 0.175). In the other 3 studies that employed touching-TIL classification, the pooled HR was 4.73 (2.28-9.8), with slight heterogeneity (I2 = 72.4%, P = 0.027). The pooled HR indicated that stromal touching TILs in DCIS were associated with recurrence more closely than the stromal TIL ratio (Figure 3B). In addition, we observed that the therapeutic strategy is also significant for the evaluation of the prognostic value of TILs. The pooled HRs for those patients who underwent surgery only or surgery accompanied by radiotherapy were 2.77 (1.26-6.07) and 2.26 (1.29-3.95), respectively. TILs in those patients who experienced comprehensive adjuvant therapy displayed no predictive effect on recurrence risk, with an HR of 1.16 (0.62-2.18, I2 = 28.3%, P = 0.233) (Figure 3C). The prognostic value of TILs in patients receiving diverse therapies varies differently, which indicates that the value of TILs in predicting DCIS recurrence is more suitable for patients diagnosed with DCIS who receive surgery only or surgery accompanied by radiotherapy.
Different subtypes of TILs play different roles in the recurrence of DCIS
Aside from total TILs, we also investigated the prognostic role of CD4+, CD8+, FOXP3+ and PD-L1+ TILs in DCIS. The pooled HRs of CD4+ and FOXP3+ TILs were estimated to be 1.98, 1.44-3.44 and 1.83, 1.23-2.70, respectively, with no considerable heterogeneity between studies (CD4+: I2 = 45.8%, P = 0.174; FOXP3+: I2 = 0%, P = 0.382). This indicates that dense CD4+, FOXP3+ TILs in DCIS are associated with a higher recurrence risk. In addition to CD4+ and FOXP3+ TILs, we also evaluated the prognostic value of PD-L1+ TILs in DCIS. The pooled HR for stromal PD-L1 TILs was 6.21 (4.26-9.06). Considering that some studies observed positive expression of PD-L1 in intraductal cancer cells in DICS, we further investigated the association between PD-L1+ tumor cells and the recurrence risk of DCIS. The pooled HR for PD-L1+ tumor cells was 3.33 (0.65-17.21), without apparent heterogeneity (I2 = 36.8%, P = 0.206). Through the above integrated analysis, we observed that both PD-L1+ TILs and PD-L1+ tumor cells are associated with the recurrence risk of DCIS. Regarding CD8+ TILs, an insufficient number of studies provided data to perform a meta-analysis, and the only study with such data provided showed no significant association between CD8+ TILs and patient recurrence (HR 0.90 (0.47-1.71)) (Figure 4). Taken together, CD4+, FOXP3+, PD-L1 TILs and PD-L1 tumor cells possess the potential to predict the recurrence risk of DCIS, and the stromal PD-L1 is more valuable than the others in evaluating DCIS recurrence risk. Thus, in clinical practice, dense stromal PD-L1+ TILs may act as a favorable marker in predicting higher recurrence risk and poor outcome of DCIS.
Sensitivity analysis
Sensitivity analysis was performed in our study to estimate the effect of each study on pooled HR by consecutive deletion of each study. The results show that no eligible study exhibited a significant influence on the pooled HR of total TILs. (Figure 5).
Publication bias
Funnel plots display symmetrical distribution and did not indicate any obvious publication bias affecting the HR for cancer recurrence in the included studies (PBegg = 0.228, PEgger = 0.931). (Figure 6A-6B).