Effect of radical lymphadenectomy in colorectal cancer with para-aortic lymph node metastasis: a systematic review and meta-analysis

A comprehensive literature search of PubMed, Embase and Cochrane Library irrespective of languages was conducted for research related to the management of CRC with PALNM up to 31 October 2021. We conceived a strategy that combined exploded medical subject heading (MeSH) terms and entry terms, and the terms were as follows: “Colorectal Neoplasms”, “Colorectal Neoplasm”, “Neoplasm, Colorectal”, “Colorectal Carcinoma”, “Carcinoma, Colorectal”, “Carcinomas, Colorectal”, “Colorectal Carcinomas”, “Colorectal Cancer”, “Cancer, Colorectal”, “Cancers, Colorectal”, “Colorectal Cancers”, “Colorectal Tumors”, “Colorectal Tumor”, “Tumor, Colorectal”, “Tumors, Colorectal”, “Neoplasms, Colorectal”, “Lymphatic Metastases”, “Lymph Node Metastasis”, “Lymph Node Metastases”, “Metastasis, Lymph Node”, “para-aortic” and “paraaortic”. Meanwhile, we identified and included some studies by screening the reference lists of similar reviews or systematic reviews. The current study was conducted in conformity to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statements [10].

The inclusion criteria were as follows: (1) the studies comparing the efficacy and safety of radical lymphadenectomy with those not undergoing lymphadenectomy or radical lymphadenectomy among adult CRC patients with PALNM; (2) all included patients in the study must have clear preoperative imaging data and postoperative pathology reports to confirm PALNM.

The exclusion criteria included: (1) CRC patients with distant organs metastases; (2) CRC patients with other non-regional lymph nodes metastases rather than PALNM; (3) the studies had no control groups; (4) ongoing clinical trials; (5) lack of sufficient information or without follow-up.

Two reviewers independently screened and identified all potentially included studies. In the process of inclusion and exclusion, titles and abstracts were first checked, any conflicts between two reviewers were resolved by the third reviewer to achieve consensus. When the screening process of titles and abstracts was finished, full-text was subsequently assessed to determine its eligibility.

The data from all eligible studies was independently extracted by reviewers with a standardized and predesigned table. The characteristics of eligible studies including first author, publication year, study type, total number of enrolled patients, intervention and comparison, primary outcomes etc. were recorded. Moreover, the inconsistency of extracted data was resolved by discussion or consulting another reviewer until a consensus was achieved.

We also assessed the risk of bias of each eligible study according to its study type. The quality of observational study (cohort study and case–control study) was judged using the Risk of Bias in Non-randomized Studies—of Interventions (ROBINS-I) tool. The ROBINS-I tool assesses bias across seven domains including: confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes and selection of reported results. For each domain an outcome of low, moderate, serious, critical and no information for risk of bias is recorded. An overall risk of bias judgement is then determined through combination of the above seven domains [11].

Considering the generalization of overall survival (OS) in determining the prognosis of cancer patients, we chose OS as primary endpoint regardless of the follow-up time. For research purposes, 3-year or 5-year OS was acceptable. Besides, OS was preferentially reported by majority of included studies. The secondary endpoint was the rate of adverse reactions, which could reflect the safety of treatment. In addition, we performed a sensitivity analysis that excluded every study, each at a given time, to prove whether our conclusions were stable [12].

Review Manager (RevMan 5.3, Copenhagen: the Nordic Cochrane Center, the Cochrane Collaboration, 2014) was used to conduct the current Meta-analysis. Odds ratios (ORs) were applied for dichotomous outcomes, and pooled proportions were calculated with 95% confidence intervals (95% CI). The heterogeneity of each outcome was evaluated by calculating the I² statistic. The I² > 50% indicated a significant heterogeneity and random effects model was applied, otherwise fixed effects model was used accordingly. In addition, we constructed the Funnel plot that could be visually inspected to assess the publication bias. Meanwhile, both Begg’s and Egger’s tests were conducted and a two-sided P-value < 0.05 was considered as statistically significance. In order to seek potential factors that could influence the heterogeneity, we performed subgroup analysis according to various subgroup standards, which was conductive to validating the consistency and robustness of our finding. To be specific, all eligible clinical studies were stratified by type of study (cohort study or case–control study), management (dissection or R0 resection), timing of metastasis (synchronous or metachronous), publication country (Korea or Japan) and location of primary tumors (left-sided or all).

The database search initially identified 393 potentially relevant records (238 from PubMed, 149 from Embase and 6 from Cochrane Library). Through checking titles and abstracts, 354 literature was excluded for the reason of case reports, duplication, mixing with other sources of cancer, mixing with other lymph nodes metastasis, reviews or meta-analyses, non-English, animal or cell experiments and wrong intervention. The remaining 39 records were further determined for eligibility by reviewing the full text, and 32 of them were eliminated on account of no comparison, irrelevant outcomes, wrong intervention, duplication etc. Eventually, 7 clinical studies were included in our final meta-analysis. It’s worth noting that all eligible studies were retrospective cohort or case control series. The detail screening process was shown in Fig. 1.

Table 1 presented the characteristics of all eligible studies in detail. 7 retrospective clinical studies (5 case control and 2 cohort studies) including 327 patients were enrolled in current research. Among all included studies, the treatment of the experimental group was dissection, R0 resection and radical resection. Three studies adopted the management of non-dissection for PALNM in control groups, and the rest 4 implemented R1 or R2 resection. In addition, the majority of studies chose OS as the primary outcome, while Ichikawa’s selected 3-year recurrence‐free survival (RFS) as the first endpoint. In terms of the metastasis time, there were 2 and 1 studies on simultaneous and metachronous metastasis, respectively. Three studies covered patients with both types of metastases, and one study did not clearly specify the status of metastasis.

All the included studies, except Ichikawa’ s, chose OS as the primary clinical outcome and were followed up for 5 years. Compared to those who did not receive lymphadenectomy or underwent non-R0 resection, CRC patients with PALNM that received radical lymphadenectomy had significant survival time benefit (OR: 6.80, 95% CI: 3.46–13.38, P < 0.01; I² = 0%) (Fig. 2). Through performing a sensitivity analysis that excluded every study, each at a time, we found that the conclusion was proven to be stable.

Of all the articles included in this study, four of them compared the incidence of postoperative complications between patients who underwent radical lymphadenectomy and those did not. As presented in Fig. 3, there was no significant difference in the rate of postoperative adverse reactions between the two groups (OR: 0.71, 95% CI: 0.35–1.44, P = 0.48; I² = 0%). Similarly, this conclusion was quite stable after conducting the sensitivity analysis.

Considering that all the studies included in this paper were retrospective studies, it was necessary to conduct subgroup analysis according to different characteristics of studies. We first conducted subgroup analysis according to the study type, namely CS or CCS studies, and the results showed that the inconsistency of study types did not change the conclusions (OR: 6.64; 95% CI:3.39–13.00; P = 0.40; I² = 0%), indicating that heterogeneity among the included studies was insignificant (Fig. 4). Then we performed subgroup analysis according to the management strategy. Three studies were grouped since the patients in these studies were divided according to whether PALNM resection was performed, while the remaining 4 studies were grouped for the patients were divided according to whether R0 resection was applied. Similarly, no significant heterogeneity was found between groups (OR: 7.33; 95% CI:3.62–14.87; P = 0.36; I² = 0%), showing that our conclusions were relatively stable (Fig. 5). As shown in Fig. 6, the same conclusions were obtained when subgroup analysis was applied according to the time of tumor metastasis, indicating that the patients with synchronous or metachronous metastasis both benefited from radical lymphadenectomy (OR: 5.55; 95% CI:2.72–11.30; P = 0.99; I² = 0%). Considering the inconformity of the staging of CRC patients with PALNM in different countries or regions, we performed subgroup analysis according to the nation. As a result, we found that the classification by country did not overturn the conclusion we have drawn before (OR: 6.80; 95% CI:3.46–13.38; P = 0.65; I² = 0%) (Fig. 7). Finally, we conducted a subgroup analysis based on the location of primary tumor, and 3 studies involving 111 patients were accordingly grouped because the patients in these studies all suffered left-sided colon tumors. Although there was a certain heterogeneity within the group, it did not affect the conclusion (OR: 6.80; 95% CI:3.46–13.38; P = 0.30; I² = 0%) (Fig. 8). The detailed information of subgroup analyses is presented in Table 2.

According to ROBINS-I scoring tool, all the included studies in current meta-analysis were judged to be of low or moderate risk of bias (Additional file 1: Table S1 and Additional file 1: Table S2). Moreover, a funnel plot was constructed to assess the possible publication bias of primary outcome (Fig. 9). The results indicated that there appeared to be no publication bias by visual inspection.

However, several limitations may exist in this systematic review and meta-analysis. First, all the studies included in the analysis were retrospective studies, and no relevant RCT_S had been published by the time of screening. As the study type with the highest level of evidence in evidence-based medicine, the lack of RCTs will inevitably affect the reliability of our research conclusions. Second, the final eligible studies were all conducted in Japan or South Korea. As a global disease, CRC urgently needs joint efforts of scholars worldwide to provide more research data and strive to improve the prognosis of CRC patients with PALNM. Next, the managements in the control groups were also uneven, ranging from surveillance, systematic chemotherapy to R2 or R3 surgical resection, and all of which might become a confounding factor affecting the robustness of our conclusions. Moreover, several essential information like whether patients received neoadjuvant chemotherapy, detailed postoperative chemotherapy regimens and cycles were not mentioned in some studies. Finally, the clinical outcome indicators of included studies were relatively few. In addition to the primary outcome overall survival and incidence of postoperative adverse reactions, other secondary endpoints reflecting long-term survival of patients, such as tumor-free survival, relapse-free survival and tumor recurrence rate, were not recorded. Comprehensive analysis of these outcome indicators may be more accurate and reasonable for evaluating the survival benefit of patients.