This article is mentioned in:

Introduction

Lymph node (LN) status is a key prognostic factor in colon cancer (1). Staging accuracy and prognosis improve with increased numbers of dissected and examined LNs, both in stage II and III colon cancer (2). At present, the most widely used LN staging system is number of metastatic LNs in pathological assessment (pN) from the 8th edition of the TNM American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) staging system, which takes into account the number of metastatic LNs (3). To evaluate LN status, the AJCC/UICC guidelines recommend resection and assessment of ≥12 LNs. Nevertheless, as reported by numerous studies, the LN yield from a colectomy varies and may not be sufficient to meet these recommendations (4,5); this is because LN yield depends on various factors, such as surgical technique and differences in pathological procedure for obtaining LNs from the specimen (5,6).

Evaluation of LN status based only on the number of metastatic LNs and omitting the total number of dissected LNs (as in pN) may be insufficient and lead to understaging. For that reason, two novel systems analysing not only the number of positive LNs, but also number of dissected LNs (NDLNs) were developed: LN ratio (LNR) (7) and log odds of positive LNs (LODDS) (8). LNR is the ratio of positive LNs divided by the total number of dissected LNs, which reflects the probability of metastatic LNs in the LN yield. LNR has been reported to be a strong independent prognostic factor in rectal and colon cancer (9,10), as well as multiple types of solid malignant tumors, such as breast, pancreatic, gastric, lung and bladder cancer (11–14). LODDS is another LN staging system that has been proved to be a strong prognostic factor in colon and rectal cancer (8,15–17). This staging system is particularly useful for patients without metastases to regional LNs. In patients with node-negative cancer, LNR is zero and does not provide any additional information in comparison with pN0 from TNM classification, whereas LODDS may stratify patients according to prognosis in this group (15–17).

The aim of the present study was to compare three LN staging systems in patients with colon cancer who underwent elective tumour resection.

Materials and methods

Patients

The present study recruited 298 patients who were operated on between September 2006 and May 2014 in the Department of Oncological Surgery, Gdynia Centre of Oncology, Pomeranian Hospitals, Poland. The inclusion criteria were as follows: Patients aged >18 years; histologically proven adenocarcinoma of the colon; curative surgical tumor resection and minimal follow-up period of 65 months or until death. Patients with in situ disease, other malignancy, lack of complete information about LN status or follow-up and neoadjuvant treatment were excluded. The present study was approved by the Independent Ethics Committee of the Regional Medical Chamber in Gdańsk (approval no. KB-2/20). Because of the retrospective design of the study based on data analysis, the requirement for informed consent was waived. The patient data included age at diagnosis, sex, primary tumor site, details of surgical procedure, pathological grade (G), postoperative tumor staging (pT stage), presence of distant metastasis (M stage), number of retrieved LNs, number of metastatic LNs (pN stage) and follow-up. The follow-up of all investigated cases proceeded until death or October 2019. The caecum, ascending colon, hepatic flexure and proximal two-thirds of the transverse colon were defined as the right colon. The distal one-third of the transverse colon, splenic flexure, descending colon, sigmoid and rectosigmoid junction were defined as the left colon. In some cases, groups were combined due to minor abundance (e.g., pT1 + pT2, pT3 + pT4, G1 + G2 and G3 + G4). The clinical and histopathological characteristics of the study population are presented in Table I.

Table I. Univariate survival analysis.

Table I.

Univariate survival analysis.

Parameter	n	5-year OS probability	P-value
Age, years (median, 71 years)			0.0005
≤71	156	0.75
>71	142	0.57
Sex
Female	143	0.69	0.5000
Male	155	0.65
Location			0.4000
Right colon	138	0.68
Left colon	160	0.65
Depth of invasion, pT			0.0200
1 + 2	50	0.80
3 + 4a + 4b	248	0.64
pN			<0.0001
pN0	181	0.80
pN1a	27	0.66
pN1b	49	0.46
pN2a	17	0.35
pN2b	24	0.29
LNR			<0.0001
LNR0	182	0.80
LNR1	51	0.62
LNR2	32	0.43
LNR3	21	0.33
LNR4	12	0.08
LODDS			<0.0001
LODDS1	190	0.79
LODDS2	78	0.55
LODDS3	30	0.23
Distant metastasis			<0.0001
M0	274	0.71
M1a + 1b + 1c	24	0.20
Number of nodes retrieved by lymphadenectomy			0.1000
<12	123	0.62
≥12	175	0.70
World Health Organization histological grade			0.0200
G1 + G2	264	0.69
G3 + G4	34	0.50

[i] pN, pN from pTNM; LODDS, Log odds of positive lymph nodes; LNR, lymph node ratio.

Definition of LN classifications

LN status according to pN from TNM staging system is defined by the number of metastatic LNs (Table II) (3).

Table II. Classification of LN staging systems.

Table II.

Classification of LN staging systems.

A, pN
Stage	Value (number of metastatic LN)
pN0	0
pN1a	1
pN1b	2-3
pN1c	No metastatic LN but there are tumor deposits
pN2a	4-6
pN2b	≥7
B, LNR
Stage	Values
LNR0	0.00
LNR1	0.01-0.17
LNR2	0.18-0.41
LNR3	0.42-0.69
LNR4	≥0.70
C, LODDS
Stage	Values
LODDS1	<-2.18
LODDS2	-(2.18-0.23)
LODDS3	>-0.23

[i] The pN classification was obtained from 8th edition of the TNM American Joint Committee on Cancer/Union for International Cancer Control. LNR classification was obtained from Rosenberg et al (9) LODDS classification was obtained from Zhang et al (20) LN, lymph node; LNR, LN ratio; LODDS, log odds of positive LNs.

LNR is defined as the ratio of metastatic to examined LNs. Previous studies used distinct methods to determine LNR cut-off values to discriminate patients by their prognosis, with only a few studies using statistical methods (9,16,18). The majority of previous studies used quartiles, mean values or arbitrary classification. The present study used cut-off values developed by Rosenberg et al (9), which were established using statistical methods (classification and regression trees technique) and evaluated using a large population of patients with colorectal cancer (n=17,309) (19). This classification divides patients into five subgroups according to LNR (Table II).

LODDS is the log of the ratio between the numbers of positive and negative LNs: loge [(pN + 0.5)/(nN + 0.5)], where pN is the number of positive LNs and nN is the number of negative LNs and 0.5 is added to both the numerator and denominator to avoid an infinite value. In the present study, cut-off values determined statistically by Zhang et al (20) using a large group of patients (n=240,898) were used to stratify the patients. The prognostic cut-off values were determined as −2.18 and −0.23 and subgroups are presented in Table II.

Statistical analysis

Statistical analysis was performed using Statistica (version 13; TIBCO Software, Inc.). Pearson's χ2 test was performed to evaluate the association between clinical and histopathological parameters and investigate the LN staging systems. Univariate analysis of survival was performed using the Kaplan-Meier method and differences in survival rates between subgroups were compared using log-rank test. The end point of the present study was 5-year overall survival (OS). A multivariate analysis was conducted using the Cox proportional hazard model. The three-step multivariate analysis was applied to assess the prognostic discriminating power of different LN staging systems. In step one, all relevant factors from the univariate analysis were encompassed, including pN, but excluding LNR and LODDS. In step 2, LODDS was added, but not LNR. In step 3, all three LN classifications were included. Additionally, Akaike Information Criterion (AIC) was used to assess which model fit best. Principally, the predictive model with the lowest AIC displayed the best fit. The correlation between different LN classifications was analysed using Pearson correlation coefficient. P<0.05 was considered to indicate a statistically significant difference.

Results

The results of the univariate analysis of survival and 5-year OS rates are presented in Table I. Age (>71 years), pT, M stage and pathological grade were significantly negatively associated with patient prognosis. Furthermore, the analysis demonstrated a significant association between increasing number of metastatic LN and worse prognosis (Fig. 1); this was also observed for increasing LNR (Fig. 2) and increasing LODDS (Fig. 3). Multivariate analysis was performed in three steps. In step 1 (Model 1), pN was identified as an independent prognostic factor. In step 2 (Model 2), when LODDS was added, pN was displaced by LODDS, which became a significant independent prognostic factor in relation to prognosis. In step 3 (Model 3), when all three classifications were included, pN and LODDS were replaced by LNR. The results of the multivariate analysis are shown in Table III. Additionally, the AIC confirmed that the LNR staging system displayed the best performance in relation to prognosis among all three classifications (AIC for Model 1, 2 and 3, 1012.022, 1010.827 and 1004.578, respectively). The scatter plots of the correlation between different LN classifications are presented in Figs. 4 and 5. Patients with the same number of metastatic LNs can be assigned to different LNR values (r=0.77, P<0.001) (Fig. 4). LODDS and LNR were closely correlated (r=0.9, P<0.001), despite the situation when LNR is close to 0 or 1. In this case the value of LODDS is still heterogeneous (Fig. 5).

Figure 1. Kaplan-Meier curves of 5-year overall survival for patients stratified by pN, according to pTNM 8th edition American Joint Committee on Cancer/Union for International Cancer Control staging system.

Figure 2. Kaplan-Meier curves of 5-year overall survival for patients stratified by LNR. LNR, lymph node ratio.

Figure 3. Kaplan-Meier curves of 5-year overall survival for patients stratified by LODDS. LODDS, log odds of positive LNs.

Figure 4. Scatter plot of the distribution of LNR vs. number of metastatic LNs (r=0.77, P<0.001). LN, lymph node; LNR, LN ratio.

Figure 5. Scatter plot of the distribution of LNR vs. LODDS (r=0.9, P<0.001). LNR, lymph node ratio; LODDS, log odds of positive lymph nodes.

Table III. Three-step multivariate analysis (Cox proportional hazards model).

Table III.

Three-step multivariate analysis (Cox proportional hazards model).

	Model 1			Model 2			Model 3
Parameter	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
Median age, years	0.45	−1.220–0.380	0.0001	0.40	−1.330–0.480	<0.0001	0.37	−1.410–0.550	<0.0001
World Health	1.57	−0.080-0.980	0.0900	1.55	−0.090-0.970	0.1000	1.67	−0.010-1.050	0.0500
Organization histological grade
Depth of invasion, pT	1.28	−0.420-0.920	0.4600	1.52	−0.260-1.100	0.2000	1.52	−0.250-1.100	0.2000
Distant metastasis, M	4.54	0.960-2.060	<0.0001	3.84	0.770-1.910	<0.0001	3.41	0.660-1.800	<0.0001
pN from pTNM	1.01	0.005-0.010	<0.0001	1.00	−0.002-0.009	0.3000	0.99	−0.007-0.006	0.8000
LODDS	-	-	-	1.87	0.200-1.040	0.0030	0.89	−0.740-0.530	0.7000
LNR	-	-	-	-	-	-	1.88	0.200-1.050	0.0030

[i] LODDS, Log odds of positive lymph nodes; LNR, lymph node ratio.

Further analysis was performed in subpopulations of patients with NDLN ≥12 and <12. In univariate analysis in both subpopulations, all three LN classifications were significant prognostic factors. In the multivariate analysis of patients with NDLN≥12, the best LN staging system was LNR. In the multivariate analysis of patients with NDLN<12, when all three LN classifications were included, none of them displayed significant differences between levels of staging. However, when 5-year OS (according to pN) was directly compared in subgroups pN1b and pN2a, the prognosis was significantly worse in patients with inadequate lymphadenectomy (NDLN<12) compared with that in patients with adequate lymphadenectomy (NDLN≥12; Table IV). By contrast, when comparing LNR and LODDS subgroups, prognosis was similar in patients with inadequate and adequate lymph node dissection (Table IV).

Table IV. Univariate analysis of 5-year OS depending on extent of lymphadenectomy.

Table IV.

Univariate analysis of 5-year OS depending on extent of lymphadenectomy.

A, pN
	5-year OS probability
Parameter	NDLN<12	NDLN≥12	P-value
pN0	0.77	0.83	0.40
pN1a	0.62	0.68	0.80
pN1b	0.30	0.60	0.02
pN2a	0.11	0.62	0.02
pN2b	0.40	0.26	0.70
B, LNR
	5-year OS probability
Parameter	NDLN<12	NDLN≥12	P-value
LNR0	0.76	0.83	0.30
LNR1	0.57	0.63	0.70
LNR2	0.31	0.61	0.10
LNR3	0.42	0.28	0.80
LNR4	0.12	<0.001	0.10
C, LODDS
	5-year OS probability
Parameter	NDLN<12	NDLN≥12	P-value
LODDS1	0.75	0.81	0.4
LODDS2	0.51	0.58	0.5
LODDS3	0.21	0.25	0.6

[i] LNR, lymph node ratio; LODDS, log odds of positive lymph nodes; OS, overall survival.

Discussion

The prognostic value of LNR and LODDS have been investigated and proven by numerous researchers (7–10,15–21). pN from TNM is still the most widely used classification despite only considering the number of metastatic LNs. A potential limitation of this classification is that its prognostic power is highly affected by the number of examined LNs. Dissection of ≥12 LNs is enough to evaluate LN status (3). However, inadequate lymphadenectomy during colon cancer surgery is commonly observed (4,5). Here, LNR and LODDS were shown to be more accurate LN staging systems compared with pN, potentially due to decreased dependence on the number of dissected LNs.

Numerous studies have investigated the prognostic value of each LN staging system assessed in the present study, but few surveys that have directly compared the three staging systems (16,18,21). As their results have not reached a consensus, the differences may be an outcome of different statistical methods, cut-off values and populations of patients. Song et al (18) (n=1,297) concluded that among all three LN staging systems, LNR was superior to the other two classifications. Moreover, for patients in each LNR subgroup, prognosis was highly homologous between those in different pN or LODDS subgroups. However, for patients in particular pN and LODDS subgroups, significant differences in survival were identified between patients in different LNR subgroups.

Fang et al (16) performed multivariate analysis in a smaller sample of patients (n=192), which showed that LODDS was a better prognostic factor compared with LNR and pN and that LODDS was more efficient in differentiating patients with different outcomes, particularly when the ratio of metastatic LNs was close to 0 or 1.

Pei et al (21) compared all three LN staging systems using a large group of patients with colorectal cancer (n=56,747), analysing pN, LNR and LODDS as both continuous and categorical variables. When investigated as a categorical variable, cut-off values from Berger et al (7) and Rosenberg et al (9) were used for the LNR staging system, whereas cut-off values from Wang et al (8) and Persiani et al (15) were implemented for the LODDS staging system. The analysis showed that, when considered as a categorical variable, LNR using Rosenberg's cut-off values exhibited the best prognostic performance in the whole population of patients. This was also true for the subpopulation with NDLN<12, whereas for the subpopulation with NDLN≥12, pN was the best prognostic model. When investigated as a continuous variable, the LODDS staging system was superior to the others. However, according to Pei et al (21) LODDS as a continuous variable would be impractical to apply in clinical practice; thus, it should be changed into a categorical variable by calculating the optimal cut-off values to make it applicable for clinical use.

As aforementioned, the most suitable cut-off values for LNR and LODDS are still under discussion. In the present study, cut-off values developed by Rosenberg et al (9) were used for LNR, whereas cut-off points determined by Zhang et al (20) were used for LODDS. These cut-off values were selected because they were established statistically using large populations of patients with colorectal cancer. The results of the present study confirmed the usefulness of these cut-off values and, in our opinion, they should be tested in consecutive studies and then considered for use in wide clinical practice. In the present study, LNR was identified as the best LN staging system, both for the whole population of patients and for the subpopulation with NDLN≥12. In multivariate analysis of the subpopulation with NDLN <12, none of three LN classifications provided significantly different prognostic results in the context of 5-year OS. This may be due to the small number of patients in subgroups. However, there were significant differences between pN1b and pN2a subgroups in 5-year OS for patients with adequate and inadequate lymphadenectomy. These differences were not observed for LNR and LODDS, which suggested that LNR and LODDS were more independent of number of dissected LNs, as previously hypothesized (15,22).

A scatter plot demonstrated an association between LODDS and LNR. LODDS value was heterogeneous only when LNR was close to 0 or 1. This indicated that LODDS was particularly valuable in patients without metastatic LNs because when LNR and pN for all patients was 0, LODDS classification still divided patients into subgroups with different prognoses.

The present study has certain limitations because it was single centre retrospective study that included a relatively limited number of patients. Therefore, multicentre studies on larger populations of patients should be performed to verify the present conclusions. Patients without metastatic LNs could not be used for full analysis of prognosis according to LODDS due to small sample size of LODDS subgroups. Therefore, further studies on larger populations of patients are required to test the prognostic value of LOODS in pN0 patients operated on for colon adenocarcinoma.

In conclusion, the present study indicated that LNR is the most accurate LN staging system for predicting prognosis for patients with colon cancer who have undergone surgical resection, especially those with metastatic LNs subjected to adequate lymphadenectomy. Due to the limitations of LNR in pN0 patients and the promising prognostic results for LODDS, further studies are required in this group. Subsequent studies should also be performed to investigate patients with NDLN <12, as LNR or LODDS might be more suitable in estimating prognosis in this group than LN staging using pN from TNM.

Acknowledgements

Not applicable.

Funding

Funding: No funding was received.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Authors' contributions

AMM conceptualized and designed the present study, collected and interpreted data and drafted the manuscript. MS analysed the data, generated figures and tables and drafted the manuscript. WJK conceived the study and revised the manuscript. AMM and MS confirm the authenticity of all the raw data. All authors have read and approved the final manuscript.

Ethics approval and consent to participate

The present study was approved by the Independent Ethics Committee of the Regional Medical Chamber in Gdańsk (approval no. KB-2/20). Due to the retrospective design of the study based on data analysis, the requirement for informed consent was waived.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Glossary

Abbreviations

Abbreviations:

References