Lymphatic Mapping in Colon Cancer Depending on Injection Time and Tracing Agent: A Systematic Review and Meta-Analysis of Prospective Designed Studies

Simple Summary Lymphatic spreading is a main driver of metastasis and, thus, associated death in colon cancer. Therefore, resecting all metastatic lymph nodes is vital for cancer-free survival. Although resection within established resection lines provides a good lymph node yield, aberrant lymphatic drainage pathways may be missed. Lymphatic mapping can compensate for this shortcoming. Different methods for tracing lymphatic drainage exist, such as radiocolloid tracers, ink, and fluorescent tracers. Tracers can be applicated either during surgery or before surgery through colonoscopy, giving the tracer more time to travel through the lymphatic system and highlighting more distant tumor-draining lymph nodes. This review aims to assess which protocol best maps the lymphatic drainage pathway and thus enables an optimized, personalized approach for curative resection. Abstract An optimized lymph node yield leads to better survival in colon cancer, but extended lymphadenectomy is not associated with survival benefits. Lymphatic mapping shows several colon cancers feature aberrant drainage pathways inducing local recurrence when not resected. Currently, different protocols exist for lymphatic mapping procedures. This meta-analysis assessed which protocol has the best capacity to detect tumor-draining and possibly metastatic lymph nodes. A systematic review was conducted according to PRISMA guidelines, including prospective trials with in vivo tracer application. The risk of bias was evaluated using the QUADAS-2 tool. Traced lymph nodes, total resected lymph nodes, and aberrant drainage detection rate were analyzed. Fifty-eight studies met the inclusion criteria, of which 42 searched for aberrant drainage. While a preoperative tracer injection significantly increased the traced lymph node rates compared to intraoperative tracing (30.1% (15.4, 47.3) vs. 14.1% (11.9, 16.5), p = 0.03), no effect was shown for the tracer used (p = 0.740) or the application sites comparing submucosal and subserosal injection (22.9% (14.1, 33.1) vs. 14.3% (12.1, 16.8), p = 0.07). Preoperative tracer injection resulted in a significantly higher rate of detected aberrant lymph nodes compared to intraoperative injection (26.3% [95% CI 11.5, 44.0] vs. 2.5% [95% CI 0.8, 4.7], p < 0.001). Analyzing 112 individual patient datasets from eight studies revealed a significant impact on aberrant drainage detection for injection timing, favoring preoperative over intraoperative injection (OR 0.050 [95% CI 0.010–0.176], p < 0.001) while indocyanine green presented itself as the superior tracer (OR 0.127 [95% CI 0.018–0.528], p = 0.012). Optimized lymphatic mapping techniques result in significantly higher detection of aberrant lymphatic drainage patterns and thus enable a personalized approach to reducing local recurrence.


Introduction
In the curative treatment of colon cancer, long-term survival has increased significantly since optimized lymphadenectomy in complete mesocolic excision (CME) has been carried out [1][2][3]. The CME implies the sharp dissection of the visceral from the retroperitoneal plane, aiming to resect an intact package of the tumor with its main lymphatic drainage, thus maximizing lymph node harvest. However, tumor recurrence and metachronal metastasis affect a substantial proportion of formerly R0 resected patients [4], which raises the need for further improved diagnostics and therapy.
The sentinel technic was developed to highlight a tumor's first draining lymph node (LN). While precise for other tumor entities, the colonic drainage is not as linear as that of breast cancer or melanoma [5], and skip metastases are frequent findings [6,7]. In CC, sentinel LN mapping has been associated with low sensitivities [8][9][10][11][12], while skip metastases, and metastases in aberrant LNs, can occur. Skip metastases are tumor-positive LNs distal of tumor-negative LNs [13,14], while aberrant LN describes draining LN outside the standard resection margin [7,[15][16][17]. Identifying the first few draining LNs via sentinel technique could not produce reliable results in colon cancer with sensitivities of 63-73.7% [8][9][10][11][12] of predicting the nodal status of the disease. Moreover, recent studies detected aberrant drainage patterns outside the known lymphatic drainage routes of CC [18,19]. Possible pathomechanisms leading to this phenomenon include different congenital drainage, lymphangiogenesis, and lymphatic occlusion with consecutive rerouting [20].
Varying protocols have been proposed for lymphatic drainage tracing in CC, which differ in both the tracer used and the application timing and method. Historically common is the intraoperative, subserosal application of ink detected either in pathological assessment or intraoperatively. The same technique was studied with a radiocolloid tracer and the fluorescent tracer indocyanine green (ICG), relying on either a radioactivity detection system or a fluorescent camera system. However, intraoperative tracer application only allows a narrow time frame to visualize lymphatic drainage, but especially aberrant drainage paths with slower lymph flow might be missed. Alternatively, the tracer can be applicated through colonoscopy, injecting the tracer submucosally near the tumor. This can be done directly prior to surgery or a day or more in advance, giving the tracer time to travel through the lymphatic system, staining passed LNs tracer-positive.
Given recurrence rates range between 5 and 10% [21,22] in non-metastatic patients receiving R0 resection in CC and local recurrence originating mostly from lymphatic metastasis [6], there is potential for improvement. Visualizing the individual lymphatic flow might be a key to reducing tumor recurrence and precise staging in CC. This review aims to identify the best protocol to thoroughly visualize a patient's individual lymphatic drainage pattern in CC to improve oncological outcomes further.

Protocol and Guidance
A systematic review was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [23]. The protocol was prospectively PROSPERO-registered with the registration ID CRD42021258766 [24].

Data Sources and Search Strategy
A comprehensive database search was performed using Medline and Web of Science, including "forward cited search" [16,25,26], and Embase through OVID. Registers used include Cochrane and PROSPERO. A search string was conducted with the help of an experienced librarian from the University of Hamburg, validated by preliminarily finding already known studies, and translated via SR-accelerators polyglot search [27]. All searches were conducted on 20 July 2021; the original search string can be found in the Supplementary Data (Data S1). A second search using the same search strategies was conducted on 13 February 2023, which led to five newly published studies [28][29][30][31][32]. Reviews [8][9][10][11][12]33,34] concerning similar research topics were manually searched for possible missed publications. One study [35] was found through this process.

Eligibility Criteria
Prospectively designed studies including a minimum of five patients with in vivo tracer application intended for lymphatic mapping in colonic malignancies, followed by oncologic resection in patients aged above 18 years, were eligible. Exclusion criteria consisted of ex vivo mapping, other than standard pTNM staging used, doubled patient data through multiple publications of the same collective, number of nodal positive patients only given including patients upstaged via experimental immunohistochemistry (IHC) analysis of LNs, retrospective study design, other tracing detection used, publications in languages other than English, German, or French, and inclusion of rectal neoplasms without the option for extrapolation of CC patient data, this due to the different lymphatic drainage in the mesorectum and mesocolon. Conference abstracts [36,37] were included if sufficient data were provided.

Study Selection and Data Extraction
Two authors (K.L. and J.-K.G.) individually conducted title, abstract, and full-text screening using Covidence [38]. Conflicts were discussed, and an agreement was reached.
Data extracted included study design, patient number, tracing method and timing, application site, TNM stage, LN tracing procedures, and LN data such as total resected LN and total traced LN, data of aberrant drainage searched and found, number of aberrant LNs, if a change in resection lines was done, and number of changes of resection line. Supplementary Data were considered, and selected authors were contacted for additional information. Individual patient datasets (IPD) were extracted when reported. As tracers ICG, radiocolloids and ink (such as patent blue, isosulfan blue, lymphazurin, methylene blue, carbon particles, or blue dye V) were eligible. Detection methods for aberrant LNs were visibility in the mesentery for ink, usage of an ICG camera before and/or after oncologic resection, and usage of a gamma camera before and/or after the oncologic resection. The preoperative tracer injections took mostly a day, and up to three days prior to surgery place; intraoperative injection was defined as an injection during the surgical procedure after skin incision. Aberrant drainage was defined as LNs outside the standard resection margin. Studies describing a combination of tracers or application methods were included if data for each application or detection method could be extrapolated. In studies assessing upstaging, and thus reporting the pTNM stage with additional IHC staining for detecting micrometastases, which were in some studies considered tumor-positive nodes, the LN status assessed with standard examination techniques was extracted for comparability without bias.

Quality Assessment and Risk of Bias
Quality assessment was done individually by two reviewers (K.L. and J.-K.G.). Discrepancies were discussed, and an agreement was reached. The QUADAS-2 [39] tool with partly reviewed specifically tailored signaling questions. Questions and criteria of quality assessment are described in Table S1. Outcomes of interest were the proportion of traced LNs in all resected LNs according to injection timing, application site and tracer, aberrant nodal positivity, and detection of aberrant lymphatic drainage patterns. Analyzed were overall study results and individual patient data (IPD).

Data Synthesis and Analysis
These meta-analyses were implemented in Stata version 14.2 using the metaprop_one command. When data were available in at least three studies, heterogeneity was assessed by Cochran's Q test and the I 2 statistics. Results for each study were pooled using a singlearm meta-analysis of proportions models. Individual study results were analyzed using random-effects models based on the DerSimonian and Laird method, with heterogeneity estimated from the inverse variance model. Wilson score confidence intervals were used, and the Freeman-Tukey double-arcsine transformation was applied to stabilize the variances. To calculate outcomes, assumptions of mean values were made when data were given in median and range. Individual patient data were analyzed using traditional covariate-adjusted linear models and generalized linear models, adjusting for the tracer used, the timing of tracer application, tumor localization, and T-stage. In all instances, a p-value < 0.05 was considered statistically significant.

Study Characteristics and Quality Assessment
Fifty-eight studies could be included, analyzing 3393 patients. The majority of studies (89.7%) had a monocentric study design. Nineteen studies used ICG, 30 ink, and five radiocolloid as tracers. Four studies applied a combination of tracers in their cohort. While most studies performed a completely intraoperative tracer application (81.0%), labeling was done preoperatively in nine cohorts and intra-and preoperatively in two studies (Table 1).

Study Characteristics and Quality Assessment
Fifty-eight studies could be included, analyzing 3393 patients. The majority of studies (89.7%) had a monocentric study design. Nineteen studies used ICG, 30 ink, and five radiocolloid as tracers. Four studies applied a combination of tracers in their cohort. While most studies performed a completely intraoperative tracer application (81.0%), labeling was done preoperatively in nine cohorts and intra-and preoperatively in two studies (Table 1). Low risk of bias was generally present in the quality assessment using the QUADAS-2 tool. In studies with a high risk of bias for the reference standard, the number of tumorpositive LNs had to be extracted without the experimentally used staging techniques. An overlap of exclusion criteria and the QUADAS-2 suggested index test questions led to included studies showing a low risk of bias in patient selection, index test, and reference standard for the section of applicability concerns. Results can be found in Table 2.

Effectiveness of Lymph Node Mapping
Thirty-six studies qualified for analysis of LN mapping effectiveness. When studies reported multiple techniques, separate cohorts were considered. After intraoperative tracer injection, proportions of mapped lymph nodes ranged from 3.28 to 35.63% with a pooled rate of 14.1%. In contrast, preoperative LN mapping resulted in a significantly higher pooled rate of 30.1% traced LNs (p = 0.030), ranging from 15.58 to 49.12% (Figure 2a,b, and Table S2). Both analyses showed a significantly high level of heterogeneity.
From 35 studies, data on the tracers used could be extracted. Twenty-two studies used ink, six radiocolloids, and eight ICG. Lim [179] had to be excluded based on the summarized reporting of different mapping techniques. One study [19] reported results in two groups. The pooled estimate of the traced LNs proportion was stable regardless of the type of tracer: 14.2% in studies using ink, 15.2% in studies using radiocolloid, and 17.1% in studies using ICG with high heterogeneities in all analyses (p = 0.740; Figure 3a-c), and Table S3.
Data on the tracer application site were retrievable from 35 studies, while two cohorts [12,28] were distributed to both groups. Studies not specifying the application site were excluded from this analysis. The pooled estimate of the traced LNs proportion was 22.9% in studies injecting the tracer submucosal and 14.3% for subserosal injection, with high levels of heterogeneity in all analyses (p = 0.070, Figure 4a,b) and Table S4.  [12,16,19,25,26,28,32,35,157-159,162,163,165,168-1 176,178-183,187,188,190,192-195,197].

Abberrant Lymphatic Drainage Detection
All included studies were searched for the mention of aberrant drainage pathways. Of the 58 studies originally meeting the inclusion criteria, 42 mentioned searching for aberrant drainage patterns. Of those 42 studies, 24 found aberrant drainage. For the quantitative analysis, further studies were disqualified due to missing quantification of aberrant drainage [35,161,172,194]. Results are displayed in Table 3.

Individual Patient Data
All individual patient data (IPD) available from the included studies were further analyzed according to factors influencing the effectiveness of LN mapping and aberrant LN detection. Data on tumor characteristics, the timing, and the tracer application site from ten studies could be extracted and included in this analysis [12,16,17,28,29,31,32,159,174,190]. Studies neither searching nor resecting aberrant LNs or not reporting the mapped LN yield did not contribute to the respective analysis.
IPD sets of 210 patients could be extracted from ten studies. One hundred and sixteen patients had right-sided CC, including the caecum, ascending colon, hepatic flexure, and proximal transverse colon. At the same time, 74 suffered from CC located at the splenic flexure, the descending colon, or the sigmoid ( Table 4). The majority of tumors were staged pT3 (41.0%), pN0 (51.0%) and had no distant metastasis (31.4%), with a substantial number of 142 cases not reporting the M-stage. Most patients (52.9%) received an intraoperative tracer injection, while the tracer was applicated preoperatively in 82 patients (39.0%). In 17 patients (8.1%), the injection timing remains unclear. The included studies used ink as a tracer for 44 patients (21%) and ICG for 166 patients (79%). No IPD were found for the radiocolloid tracer. The tracer was injected evenly in 96 patients (45.7%) subserosally and in 97 patients (46.2%) submucosally, while data were missing for 17 patients (8.1%). The application site and timing of tracer injection were highly correlated within the IPD. All preoperative applications were performed submucosally, and the vast majority of intraoperative markings were performed subserosally. In only one study's subcohort, the tracer was applied intraoperatively submucosally in 15 patients [12]. Therefore, the application site of the tracer application was not further analyzed in the IPD.   Seven studies searched for aberrant drainage and reported injection timing, tracer, tumor location, and T-stage, involving 121 patients (Table 6). A generalized linear model for adjusted outcome analysis was used, adjusting for the independent parameter. While the tracing methods significantly impacted aberrant lymphatic drainage detection, neither tumor burden nor tumor location were associated with finding an aberrant drainage pattern. Intraoperative tracer application was used in 64 patients and had highly significantly lower odds of finding aberrant drainage compared to preoperative tracer injection, which was present in 57 patients (intraoperative vs. preoperative OR 0.050 [95% CI 0.010-0.176], p < 0.001). Furthermore, ink, applied in 44 patients, revealed significantly lower odds of aberrant drainage detection than ICG, and was used in 77 patients (ink vs. ICG OR 0.127 [95% CI 0.018-0.528], p = 0.012).

Discussion
Skip metastases [13,14] and aberrant lymphatic drainage patterns, hence lymph flow inconsistent with the sustentative blood vessels [7,[15][16][17], are frequently found in CC. Malignancies can induce not only neoangiogenesis but also lymphangiogenesis, thus surrounding a tumor, increased and newly formed lymphatic flow can exist [20], which differs from the original anatomy. Lymphatic mapping has the potential to unveil those drainage patterns and thereby improve surgical resection when carried out effectively.
This meta-analysis demonstrates that a preoperative tracer application and earlier tumor stages allow a higher mapped LN yield. In contrast, the tracer or application sites have no relevant impact on this ratio. However, for the detection of aberrant LNs, preoperative tracer application, usage of ICG, and a submucosal application demonstrated significantly better results.
Effective LN mapping provides an accurate picture of the tumor's lymphatic drainage, particularly of those LNs further away and connected via slow-draining lymphatic vessels, as usual following lymphangiogenesis [20]. The tracer application timing was revealed as the strongest predictor of this effectiveness. Preoperative tracer application resulted in a significantly increased rate of traceable LNs with a pooled rate of 30.1%, compared to intraoperative tracer application resulting in 14.1% traceable LNs. Moreover, the traced LN yield was significantly higher when marked preoperatively (p = < 0.001), which is in line with results of previous studies [12,119,121].
A broad range of tracers has been described in the literature. To analyze their effectiveness, we grouped tracers according to detection properties, such as staining tracers (ink, methylene blue, carbon particles, and patient blue), radiocolloid as a radioactive tracer, and ICG as a fluorescent tracer, since the detection methods of the respective tracer group decisively influence their tracing performance. No significant difference was observed in the rate, or the yield of LNs traced (pooled estimate of the proportion of traced LN: ink vs. radiocolloid vs. ICG: 14.2% vs. 15.2% vs. 17.1%; IPD traced LNs: ink vs. ICG, 1.0 ± 0.8 vs. 3.9 ± 4.2, p = 0.104), suggesting that all tracers have comparable abilities to travel through the lymphatic system. However, aberrant drainage was significantly more frequently detected using ICG (ink vs. radiocolloid vs. ICG: 2.5% (0.9, 4.7) vs. 0.0% (0.0, 1.2) vs. 18.1% (9.2, 28.7)), which could be confirmed in the IPD analysis (ink vs. ICG OR 0.127 [95% CI 0.018-0.528], p = 0.012). Aberrant LNs can easily be missed intraoperatively due to their embeddedness in the mesentery. ICG provides bright visibility even through fatty mesentery, which might enable more precise detection of LNs outside standard resection lines in the surgical site [199]. Further advantages of ICG as the used tracer are the minimal adverse effects reported and that patient and surgeon are not exposed to radioactivity.
The timing of the tracer application is methodically associated with the injection site. Intraoperative application, most commonly subserosal and preoperative tracer application, is performed before skin incision via colonoscopy, submucosally. Therefore, data for a structured investigation of the influence of tracer application sites are rare. This metaanalysis revealed no impact of the injection site on the effectiveness of LN tracing (pooled rate of traced LN: subserosal vs. submucosal 14.3% vs. 22.9%). Only two studies [12,169] used on-table colonoscopy for submucosal tracer application intraoperatively, while all other studies contributing to the submucosal group also performed preoperative tracer injection. Both studies revealed substandard traced lymph node rates, though Ankersmit et al. could prove a higher sensitivity for sentinel lymph node detection after submucosal injection in a comparative study design. Intestinal lymphatic drainage works as dual independent networks in the muscular and the mucosal layer of the intestine wall, which drain to a shared network of collecting ducts. While adenocarcinomas form in the mucosal layer, draining LNs of the mucosal layer and, thus, the tumor may be missed when the tracer is applied subserosally [200,201]. Significantly higher rates of aberrant drainage could be detected when a tracer is injected submucosally. However, this has to be interpreted cautiously due to the increased risk of interference with injection timing.
While the tumor site did not impact the traced LN yield, the tumor stage proved to have a significant association with the amount of traced LNs favoring earlier tumor stages (p = 0.020). LN metastases can occlude lymphatic pathways in advanced settings [20], which may lead to fewer traceable LNs. Consequently, LN metastases distant from others might not be traceable by LN mapping. On the other hand, neither tumor location nor tumor stage did influence aberrant lymphatic drainage detection. Lymphangiogenesis as a primary driver of aberrant lymphatic pathways is involved in early tumor stages and takes place in advanced settings as rerouting after lymphatic occlusion [20]. Previous studies postulated the relevance of lymphatic mapping, particular in earlier tumor stages, to correctly stage the disease and reduce recurrence rates. Ultrastaging of traced LNs was proposed to address this topic [73,129,160,163,181] with varying results.
Our data prove the feasibility of tracing a patient's individual lymphatic drainage, enabling an accurate picture of the lymphatic draining pathway, especially in these relevant earlier stages. This might allow for a tailored multimodal therapeutic approach by refined staging and reducing tumor recurrence. In contrast, neither tumor stage nor site influenced aberrant drainage detection, emphasizing the importance of intraoperative screening for aberrant LNs, which could affect all colon carcinomas equally.
Several aspects limit these findings: significant heterogeneity was present in all traced LN proportion analyses, so these results must be interpreted cautiously. While some studies excluded their learning curves, others did not comment on this or include all patients in which a mapping procedure was carried out. Moreover, the investigation of traced LN rates might be biased since a minimum of 12 pathologically assessed LNs is sufficient according to the ESMO guidelines [202], and it depends on the diligence of the pathologist as to how many LNs are examined beyond that. Unfortunately, data on the measures of dispersion of the absolute number of traced LNs were rare in published literature, so this probably more precise analysis could not be performed. Different standards of lymphatic resection throughout the studies are present, given that standardized CME was only introduced in 2009, and the ongoing debate concerning D2 versus D3 lymphadenectomy. However, this meta-analysis reflects all currently available evidence and approaches the role and methodology of tracing lymphatic drainage scientifically accurately.

Conclusions
LN mapping has the potential to improve tumor staging and reduce local recurrence by aberrant drainage detection when carried out systematically. Preoperative mapping by colonoscopy and usage of ICG provides the best capacity for accurate visualization of lymphatic drainage. To further investigate the influence of lymphatic mapping on the quality of oncological resections, prospective studies with large patient numbers should be conducted and a standardized protocol adopted for lymphatic mapping prior to surgery to assess whether recurrence rates can be lowered, and long-term survival can be increased.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers15123196/s1, Data S1: Search strategy. Table S1: Quality assessment according to QUADAS-2 with partly review-specific tailored questions. Table S2: Traced LNs of all LNs according to injection timing. Table S3: Traced LNs of all LNs according to tracer. Table S4: Traced LNs of all LNs according to application site of tracer injection.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.