Introduction

Rectal cancer metastasizes to the perirectal lymph nodes contained within the mesorectum and along the iliac arteries [1]. Assessment of lymph node involvement is a strong predictor of recurrence-free survival and overall survival in patients with rectal cancer [2]. Currently, total mesorectal excision constitutes the gold standard for removing perirectal lymph nodes [3]. However, the therapeutic strategy regarding lymph nodes, notably those located along the iliac arteries, may include neoadjuvant radiochemotherapy, as performed in Western countries [3], or lateral lymph node dissection (LLND), as performed in Japan when the lower border of the tumour is located under the reflection of the peritoneum and when it has passed the muscularis propria [4].

However, to date, only two randomized controlled trials (RCTs) [5, 6] have compared the outcomes of the two therapeutic strategies in terms of survival, looking at local and distant recurrences and complications. Furthermore, the choice of the best strategy, either LLND or radiochemotherapy or the combination of the two treatments, is difficult to identify, as the prevalence of metastatic lateral lymph nodes in patients with rectal cancer is poorly documented. Indeed, before comparing two therapeutic strategies and their outcomes, it is necessary and more relevant to know if it is legitimate to propose them as a treatment; in other words, if they would even correctly target metastatic lateral lymph nodes (LLNs).

Therefore, our objective was to perform a systematic review and meta-analysis reporting on the prevalence of metastatic lateral lymph nodes in patients with rectal cancer with the hypothesis that this meta-analysis will elucidate the best therapeutic strategy in the current absence of reliable assessment of tumour aggressiveness.

Materials and methods

The present methodology is in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist (Table S1).

Literature search and study selection

A literature search was conducted in MEDLINE, EMBASE and COCHRANE from inception (DATE??) until November 19, 2018. Keyword combinations are reported in Table S2. Additional records were identified by manually searching the reference lists of the included publications. To be included, studies had to be written in English or French and to report the prevalence of metastatic lateral lymph nodes among patients with rectal cancer who underwent surgery with LLND. LLND was defined by the dissection of nodes included in areas such as the common iliac (IC), internal iliac (II), external iliac (EI), obturator (O), middle sacral (MS) and aorta bifurcation (Ao).

We excluded case series, conference abstracts, letters to the editor and secondary analyses of previously published papers.

Data extraction

Two independent reviewers (NC and JM) independently selected articles for inclusion and extracted the data according to a pre-established data collection form. Discrepancies were resolved by reaching a consensus with the senior authors (NCB and FR). The following data were extracted: first author; publication year; country where the study took place; study period, after and before 2010 due to modifications of the Japanese Guidelines for LLND; study design; number of patients who underwent LLND, prophylactic: dissection and removal of nodes that seem invaded (enlarged) based on pre- and peri-operative examinations, versus therapeutic/curative: removal of all nodes in the “lateral lymph area”; sex; number of patients with metastatic lateral lymph nodes; number of patients who underwent pre-operative radio- and/or chemotherapy; type of neoadjuvant treatment in those patients; and oncological stages of included patients.

Statistical analysis

Models with random effects (DerSimonian and Laird’s approach [7]) were used to combine the prevalence of metastatic lateral lymph nodes across the studies. A logit transformation was applied to prevalence before statistical pooling, and the pooled logit of prevalence was then transformed back. Heterogeneity was assessed by using the I2 statistic, and a leave-one-out sensitivity analysis was conducted to check the robustness of the pooled prevalence. Potential sources of heterogeneity were investigated by comparing the pooled prevalence between subgroups of studies. A sensitivity analysis was also conducted excluding stage 4 patients from the denominator and from the numerator of the prevalence because stage 4 patients were assumed to have metastatic lateral lymph nodes (model with random effects). In addition, a meta-regression analysis was conducted to assess the relationship between the sample size of studies and the prevalence with the restricted maximum likelihood method. [8] All analyses were performed with the Meta and Metafor packages for R version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Literature search and study characteristics

Two hundred and sixty-five publications were identified in MEDLINE, EMBASE and COCHRANE. Three publications were identified from other sources. Fifteen duplicates were removed. Of the 250 publications that were identified as eligible, 155 were excluded after title/abstract screening, and 64 were excluded after full-text screening (44 publications did not report the number of patients with metastatic lateral lymph nodes and 20 publications did not distinguish patients with metastatic lateral lymph nodes from those with metastatic lymph nodes from other areas). Ultimately, 31 publications [5, 6, 9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37] were included in the quantitative synthesis, all coming from Japan except one from China (Fig. 1, Table 1).

Fig. 1
figure 1

Flow chart showing the selection of publications for quantitative review. Forest plot of the prevalence of metastatic lateral lymph nodes in patients who underwent lateral lymph node dissection. Each horizontal bar summarizes a study. The bars represent 95% confidence intervals. The grey squares represent each of the studies’ weights in the meta-analysis. The diamond in the lower part of the graph depicts the pooled estimate along with 95% confidence intervals. Events = number of patients with metastatic lateral lymph nodes, total = number of patients who underwent lateral lymph node dissection for rectal cancer

Full size image
Table 1 Characteristics of included studies
Full size table

Among the included publications, 28 were retrospective cohort studies [9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35, 37], one was a prospective cohort study [36] and two were RCTs [5, 6]. Thirty studies were performed in Japan [5, 6, 9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33, 35,36,37] and one in China [34]. Eleven studies only included patients with TNM cancer stages 2 and 3 [5, 9,10,11,12,13, 15, 31, 32, 35, 38]. In 24 studies [5, 6, 9,10,11, 15, 17,18,19,20,21,22,23,24,25,26,27,28, 31,32,33,34,35, 37], systematic LLND was performed in all included patients, whereas in six studies [12,13,14, 16, 29, 30, 36], LLND was performed only in patients with enlarged lateral lymph nodes on pre-operative imaging, which distinguished these patients from others who had confirmation by histopathological results. In one study, the type of LLND was not described [28]. In two studies, patients underwent systematic bilateral dissection [22, 32]; in 27 studies, uni- or bilateral dissection was defined according to clinical and/or imaging findings [6, 12,13,14,15,16,17,18,19,20,21, 23, 25, 26, 28,29,30,31, 37,38,39] and in four studies, lateralization of dissection was not documented [24, 34,35,36]. Nine studies included patients who received neoadjuvant treatment [6, 11, 13,14,15,16,17, 30, 31]. Indications for neoadjuvant treatment were not documented in four studies [11, 15,16,17] and depended on tumour stage in five studies [13, 14, 16, 30, 31]. The characteristics of the included studies are reported in Tables 2, 3, S3, S4. It is worth noting that the study of Yano et al. has been classified above as a study with “curative LLND” because pre-imaging with enlarged nodes was considered the main decision-making criterion for LLND. However, in fact, after the flow chart analysis, dissection was also performed in a systematic way (Table S4). Thus, for statistical analysis, we classified it as a “prophylactic study”.

Table 2 Demographics of included studies
Full size table
Table 3 Characteristics of studies according to the type of LLND surgery and neoadjuvant treatment
Full size table

Prevalence of metastatic lateral lymph nodes

We obtained a pooled prevalence of metastatic lateral lymph nodes of 17.3% (95% CI: 14.6–20.5) (31 studies, 7599 patients) (Fig. 2). The inter-study variability (heterogeneity) was noteworthy, with I2 = 89%. In other words, the variation in study outcome (prevalence) between studies was high. Furthermore, the prediction interval was 6.5 to 38.6%. This means that if a new study is conducted, the assessed prevalence is likely, with a probability of 95%, to fall between 6.5 and 38.6%. However, the pooled prevalence was robust according to leave-one-out sensitivity analysis. The pooled prevalence varied from 16.6% (14.0 to 19.5%), when the study by Yano et al. [36] was omitted, to 17.9% (15.1 to 21.1%), when the study by Min et al. [16] was omitted (Fig. S1). The heterogeneity, however, remained stable, as the I2 statistic for heterogeneity varied from 85%, when the study by Sato et al. [18] was omitted, to 89.1%, when one of the following studies was omitted [6, 10, 13, 15, 21, 23, 25, 26, 32, 34, 35].

Fig. 2
figure 2

Meta-analysis of the prevalence of metastatic lateral lymph nodes

Full size image

To determine whether neoadjuvant treatment could alter the prevalence of metastatic lateral lymph nodes, we performed subgroup analyses by separately pooling studies reporting patients who received neoadjuvant treatment from others (Fig. S2). The difference in pooled prevalence of metastatic lateral lymph nodes between studies with versus without neoadjuvant treatment was 16.9% (95% CI: 14.1–20.2), but the inter-study variability (heterogeneity) was high important, with I2 = 88%.

Then, we explored whether the heterogeneity could be explained by changes in the therapeutic strategies across the years. The pooled prevalence of metastatic lateral lymph nodes between studies published before and after 2010 was 17.3% (95% CI: 14.6–20.4), but the inter-study variability (heterogeneity) was noteworthy, with I2 = 89% (Fig. S3).

Furthermore, we excluded patients with TNM stage IV (in six studies). After exclusion of these patients, the pooled prevalence of metastatic lateral lymph nodes was 15.8% (95% CI 13.1 to 18.9) (Fig. S4).

Then, we separately pooled studies according to the type of LLND performed (systematic versus curative). The difference in the pooled prevalence of metastatic lateral lymph nodes between studies with systematic or curative dissection was not statistically significant (p = 0.7111) (Fig. S5).

However, the pooled prevalence of metastatic lateral lymph nodes appeared to be associated with the sample size of studies (p = 0.005): the prevalence decreased when the sample size increased (Fig. S6). Meta-regression showed that, on average, a study with a sample size that was 10 times larger than that of some other study had an estimated odds ratio for prevalence that was reduced by approximately 50% (Fig. S7).

Discussion

We performed a systematic review and meta-analysis estimating the prevalence of metastatic lateral lymph nodes in patients with rectal cancer. By pooling 31 studies of Asian origin (30 from Japan and one from China) totalling 7599 patients, we found a prevalence of metastatic lateral lymph nodes of 17.3%, thereby underlining the importance of further examination of the treatment of these node areas to avoid cancer recurrence.

Six studies included patients with TNM stage IV rectal cancer. According to the most recent Japanese guidelines, metastatic disease is not the most important stage to indicate LLND [4]. However, within this stage, some patients may present both primary tumour and resectable distant metastases, thus proceeding to the possibility of undergoing LLND. As a result, to understand the prevalence of metastatic lymph nodes within the most common indications, we excluded patients with TNM stage IV disease by considering them to have metastatic lateral lymph nodes. After exclusion, the pooled prevalence of metastatic lateral lymph nodes remained stable (15.8%).

In the present systematic review and meta-analysis, we were not able to show any improvement in neoadjuvant treatment in terms of lateral lymph node metastasis. However, it should be noted that the power of this subgroup analysis might be insufficient. Additionally, Asian protocols differed from Western protocols, and only a small proportion of patients among studies with neoadjuvant treatment benefited from that modality. Therefore, it is not possible to draw any conclusions about the efficacy of neoadjuvant treatment in terms of the prevalence of metastatic lateral lymph nodes based on our results. Despite this, it is worth noting that neoadjuvant treatments between the studies were quite similar using 5-fluorouracil, oxaliplatin and irinotecan (Table S3).

Some centres performed systematic LLND, whereas others performed LLND in patients with enlarged lateral lymph nodes diagnosed by pre-operative imaging. We pooled these studies separately and found no difference in terms of metastatic lateral lymph nodes. After 2010, as Japanese guidelines changed and recommended limiting LLND for all cT3-T4 tumours with a lower edge below the peritoneal reflection, we reported a subgroup analysis of the pooled prevalence of metastatic LLN according to the publication date. Due to the high heterogeneity, we cannot interpret this result as significant.

Furthermore, we note that meta-regression showed that the pooled prevalence of metastatic lymph nodes was associated with the sample size of studies, as the prevalence of metastatic lymph nodes decreased when the sample size increased. We think that this might be explained by more selective indications for LLND in studies including fewer patients with LLND. Thus, even if the number of studies between curative and systematic LLND intention was not the same, the tendency of the sample size of studies with curative LLND was lower. This indirectly shows that smaller studies were more likely to have a selected population. As a consequence, studies with larger sample sizes (in other words, studies with prophylactic LLND) were more likely to represent the “real” proportion of patients with rectal cancer with LLN metastasis.

Our meta-analysis is the first to examine the prevalence of metastatic lateral lymph nodes, which seems to be an essential prerequisite before considering any treatment. Indeed, recent literature [40,41,42] directly questions the oncological and functional results of LLND without even first investigating whether it is necessary.

Advantages of the study

The strengths of our systematic review and meta-analysis are as follows: 1) the inclusion of a large number of studies reporting the prevalence of metastatic lateral lymph nodes based on histopathological analysis and to clarify, for the first time to our knowledge, its subsequent effects in patients operated on for rectal cancer; and 2) the highlighting of the necessity of systematic treatment of lateral lymph nodes due to the high prevalence (17.3%) of LLN metastasis. Indeed, despite the heterogeneity of the studies, which could have impacted the aim of our work, we were able to make some sound calculations due to a specific statistical analysis methodology.

Limitations of the study

The limitations of the present study are as follows: 1) the high heterogeneity of the calculated pooled prevalence of metastatic lateral lymph nodes, resulting from the quality and heterogeneity of the included publications; and 2) the final analysis included 31 papers, most of which were from Japan and were retrospective studies. Therefore, some patients might have overlapped because these papers were published from some limited leading hospitals. Moreover, it is worth noting that by focussing on patients who have benefited from LLND, a surgical procedure that is most common in Asian countries, selection has been carried out. As a result, it is expected that the selected studies tend to oversample patients with indications for LLND, with these patients being more likely to have advanced disease. Thus, a bias could have been introduced. The percentage of 17.3% of metastatic LLNs could have been overestimated. This element can explain the counter-intuitive (but non-significant) results of a higher prevalence of metastasis in the prophylactic LLND group (17.3%) versus the curative group (15.9%).

Conclusion

In conclusion, this systematic review and meta-analysis of Asian studies indicates that the prevalence of metastatic lateral lymph nodes in patients with low rectal cancer who underwent surgery is 17.3%. This prevalence is of clinical importance, as it may result in cancer recurrence and calls for the systematic treatment of these lymph node areas. Future randomized controlled trials should determine which therapeutic strategy, whether neoadjuvant radiochemotherapy versus systematic or imaging-guided lateral lymph node dissection, offers the best improvement in overall and recurrence-free survival.