Quantifying the number of lymph nodes for examination in breast cancer

Abstract

Objective

Examining the correct number of lymph nodes when diagnosing breast cancer invasion is still a problem. This work aimed to develop a qualification model that estimates the possibility of missing nodes and the number of lymph nodes that need to be examined.

Methods

By analyzing lymph node invasion of 303,760 breast cancer samples with primary tumor stage and the number of examined and positive lymph nodes from the Surveillance, Epidemiology and End Results database using a beta-binomial model, the number of nodes that should be examined was quantified in different stages.

Results

In general, to reduce the possibility of missing positive nodes to less than 10%, 21 lymph nodes should be examined; thus, the current median of dissected nodes (12) is not adequate. The number of nodes needed to be dissected for stages T1, T2, and T3 are 8, 37, and 87, respectively. Currently, the median number of node dissections for these stages were 12, 13, and 14, respectively. The clinical significance of the nodal staging score was validated with survival information.

Conclusion

Currently, the number of lymph nodes dissected in breast cancer are excessive for T1 but insufficient for T2 and T3.

Introduction

Breast cancer is the most prevalent cancer among women, with 272,400 new cases and 70,700 deaths in China in 2015.¹ As an important adjuvant therapy, radiation is currently recommended for breast cancer subtypes,² and use of radiation significantly reduces recurrence, metastasis, and prolongs the survival of breast cancer patients.^3,4 However, drawbacks of radiation have also been widely reported, including impaired fertility,⁵ subsequent malignancies,^6,7 and lung metastasis.⁸ Although some genome-based radiation assessments have been reported and used, currently, use of radiation is largely dependent on detecting lymph node invasion.⁹ It has been reported that radiation efficacy is significantly associated with lymph node status.¹⁰

As lymph node invasion status is the most important clinical indicator for radiation and other adjuvant therapies after surgery, adequate removal of lymph nodes is needed for more accurate diagnosis. However, the retrieving of nodes has significant long-term complications, including lymphedema, sensory morbidity, and restriction of shoulder function.¹¹ Thus, a balance between diagnostic accuracy and quality of life should be considered. The optimal number of nodes to examine has not been quantitatively defined.

In this study, using a straightforward beta-binominal model and the number of examined lymph nodes and invaded nodes in the Surveillance, Epidemiology and End Results (SEER) database (N=303,760), the relationship between possibly missing positive nodes and primary tumor stage was estimated. According to the results, the minimum number of nodes that should be examined for stages pT1 to pT2 were 8 and 37, respectively; the effect of nodal dissection in pT3 was unclear and it was not recommended for pT4 stage breast cancer.

Methods and materials

Data collection

The SEER database includes 17 registries covering 26% of cancer patients in the United States. Breast cancer was selected, and only breast cancer as the primary malignancy was considered to exclude metastases of other cancer types. Patients without records of primary stage, nodal status, regional nodes examined, or positive nodes were excluded. Furthermore, insufficient records or those missing data were also excluded. The stages T1a-c, T2a-c, T3a-c, and T4a-c were converted to T1, T2, T3, and T4, respectively. In total, 303,760 patients were enrolled in this study.

Hypotheses and calculation processes

A mathematical model called beta-binomial distribution¹² was used to fit SEER data and to calculate the possibility of missing invasion-positive lymph nodes according to the number dissected and examined. In the following text, true positive means that a lymph node was invaded, and the results showed breast cancer cells had invaded at least one lymph node. True negative refers to when both the true status and result were invasion-negative. False negative is defined as when the true status was invaded, while the results showed no invasion.

To simplify the model, the following hypotheses were employed:

1. All examination results of lymph nodes were correct.

2. The distribution of lymph nodes is equal, i.e., the lymph nodes were exchangeable. Biologically, this assumption is incorrect; however, if a large cohort is employed, the assumption is reasonable, as stated by most previous studies.

3. The sensitivity of true positive is equal to false negative. Sensitivities can only be calculated in node-positive samples.

Model development and validation steps:

1. The number of invasion-positive lymph nodes (non-N0 stage) and the number of total nodes dissected and examined was used to estimate the parameters of the beta-binomial (α and β) distribution. In this procedure, only samples in which at least two lymph nodes were examined were used, because by definition, the node-invasive rate would reach 100% if only one lymph node was examined. Breast cancers were categorized into stages T1 to T4.

2. False-negative rates were estimated according to the model and estimated coefficients in the overall datasets and sub-datasets (different primary tumor size), and the observed and corrected prevalence was calculated using the following functions:

   $$F{N}_{\mathit{adj},i}=\frac{\left[1-F{N}_i\right]\ast T{P}_{\mathit{adj},i}}{F{N}_i}$$

   $$Prev{(pT)}_{\mathit{adj}}=\frac{{{\displaystyle \sum }}^{\text{}}(F{N}_i+T{P}_i)}{{{\displaystyle \sum }}^{\text{}}(F{N}_i+T{P}_i+T{N}_i)}$$

   $$\text{NSS}=\frac{1-\mathit{Pre}(T)}{1-\mathit{Prev}\left(T\right)+[\mathit{Prev}\left(T\right)\ast F{N}_i]}$$

where FN_adj,k indicates adjusted false negative values; FN is observed false negatives; TP_adj,k is the number of true positives; T indicates primary tumor stage; and FN, TP, and TN represent false negatives, true positives, and true negatives, respectively.

• 3. Overall survival information is not related to lymph nodes dissected or nodal staging score, so it was used for validation. In this step, tumors with no positive lymph nodes were used. Samples in each category (T1, T2, T3, and T4) were divided based on the quartile of the nodal staging score (NSS), and then survival was compared among groups.

Software

All analyses in this study were performed using R software (v3.1.0) and R packages. The VGAM (v1.0-3) and bbmle (v1.0.18) packages were used to evaluate α and β parameters of the beta-binomial model. The survival analysis was performed with R package “survival” using the Kaplan–Meier method.

Results

Data profile

After removing cases with unavailable data from the SEER database, 303,760 samples were incorporated into this study. Detailed information of primary tumor stage, nodal invasion, and nodes examined are shown in Table 1. The primary tumor stages for most patients were T1 and T2, which accounted for 93.24% of all samples. The mean nodes examined in T1 to T4 ranged from 12.1 to 15.1, and the median nodes examined ranged from 12 to 14.

Table 1.

General information of included samples.

Stage	Patients		Nodes examined
	No.	%	Mean	SD	Median	IQR
T stage (all patients)
1	192453	63.35693	12.1	7.5	12	6–17
2	90766	29.88083	13.9	7.5	13	9–18
3	13111	4.316237	15.1	7.9	14	10–19
4	7430	2.44601	13.7	8	13	8–18
T stage (N0 patients)
1	147272	48.48301	11.4	7.4	11	5–16
2	44148	14.53384	12.7	7.5	12	7–17
3	3929	1.293455	13.3	7.5	13	8–18
4	1403	0.461878	12.2	7.3	11	7–16
N stage
0	196752	64.77219	11.8	7.5	11	6–16
1	101939	33.55906	14.6	7.4	14	10–19
2	4720	1.553858	16.1	8.5	15	10–21
3	349	0.114893	16.3	9.7	15	10-21

False negative diagnosis in overall data

Using our model, the two parameters for the beta-binominal model, α and β, were estimated to be 0.933 (95% confidence interval: 0.924–0.942) and 1.917 (95% confidence interval: 1.895–1.939), respectively. As shown in Figure 1, the overall probability of missing nodal invasion (1-sensitivity) as a function of the number of examined nodes was calculated. As expected, the probability of missing nodal invasion decreased as the number of examined nodes increased. When only one lymph node was examined, the probability of missing nodal invasion was 67.25%. The minimum number of nodes that were needed to be examined was nine to reach a less than 20% chance of missing invasion, and 21/45 for 10%/5%, respectively. The current median value of examined nodes was 12, and the corresponding probability of missing nodal invasion was 15.52%, suggesting that the current number of nodes examined is inadequate.

Figure 1.

Sensitivity of nodal diagnosis according to the number of examined nodes. The x- and y-axes indicate the number of nodes examined and probability of missing positive nodes, respectively.

False negative diagnosis in different primary stages

The stage of the primary tumor is the most important indicator for nodal examination (Figure 2). The probability of missing nodal invasion was also estimated in different primary tumor TNM stages (T1–T4). To minimize the probability of missing nodal invasion to less than 20%, the least number of nodes to be examined are 3, 15, 36 and 65 for T1, T2, T3, and T4, respectively. While to reach 10%, the number of nodes were 8, 37, and 87 for T1, T2, and T3, respectively. When the number of examined nodes is 12 (the current median value of examined nodes), the probabilities of missing nodal invasion were 7.08%, 21.49%, 38.40%, and 452.00% for T1, T2, T3, and T4, respectively. From the SEER database, the median number of examined nodes in T1, T2, T3, and T4 were 12, 13, 14 and 13 (Table 1), which had corresponding probabilities of missing nodal invasion of 7.08%, 20.42%, 35.47%, and 50.38%, respectively.

Figure 2.

Sensitivity of nodal diagnosis (y-axis) in different primary tumor sizes according to the number of nodes examined (x-axis).

Corrected rates were calculated by combining the current rates of positive nodes in different primary tumor stages and our results (Table 2). The corrected node-positive rates were higher than observed rates due to false negatives. These results indicated that the node counts of currently implemented methods are adequate for pT1, but too low for pT2.

Table 2.

Observed and corrected prevalence of node-positive rates.

	T1	T2	T3	T4
Observed prevalence	23.38%	51.20%	69.52%	80.19%
Corrected prevalence	32.93%	63.82%	80.06%	87.46%

Nodal staging score and survival

Follow-up information was not used in our model and was independent from our nodal staging score. The nodal staging score in samples diagnosed as N0 stage at primary diagnosis was divided by quantiles, and survival differences were compared. Nodal staging score was significantly associated with survival in pT1N0 patients and the pT2N0 group, while not significant in pT3N0 patients and aversely associated with nodal staging score in the pT4N0 group (Figure 3).

Figure 3.

Patient survival in different primary tumor sizes according to the nodal staging score. Q1 to Q4 indicate the lowest to highest nodal staging scores.

Discussion

Lymph node invasion is highly associated with distant recurrence and relapse in cancers.^13,14 For breast cancer, it is critical for therapeutic decisions, including radiation therapy, chemotherapy, and follow up.^15,16 While the number of lymph nodes examined is controversial for patients: inadequate nodal examination may lead to missing nodal invasion, while excessive nodal examination reduces immune activity and has long term complications.¹¹ Thus, balancing diagnostic accuracy with the drawbacks of lymph node removal is critical for improving survival times and the life quality of breast cancer patients.

To investigate whether the current number of nodes removed is reasonable, we fitted the beta-binominal model on the SEER database (303,760 samples). To our knowledge, this is the first study of the number of nodes required to make an accurate diagnosis of nodal invasion status. In this model, a few hypotheses are employed. First, the beta-binominal distribution of nodal invasion was used. According to previous reports, no issues with this hypothesis have been reported.^12,17–19 Another assumption in this study is that all examined nodes were correct, i.e., no false negatives or false positives were detected. This is reasonable, as all diagnoses were made by experienced pathologist, according to the SEER database.

Our results showed that the probability of missing nodal invasion using the current median examined nodes (12) was 15.52%. To reach an accuracy of 90%, 21 nodes are needed for breast cancer patients when primary tumor stage was not considered. A less than 10% probability of missing nodal invasion demands eight, 37 and 87 lymph nodes be examined in pT1, pT2, and pT3 disease, respectively. Thus, the current median number of examined nodes is sufficient for pT1, but the probability of missing nodal invasion is >20% for pT2 to pT4. This explains why the missing nodal status of pT1 was much less than pT2 and pT3.

It is noteworthy that 69.52% and 80.19% of pT3 and pT4 stage breast cancer (corrected rates: 80.06% and 87.46%), respectively, have observed lymph node invasion, indicating that lymph node invasion is common in pT3 and pT4. Additionally, the drawbacks of removing and examining lymph nodes are exaggerated, while the necessity of diagnosing nodal status is rapidly reduced in these patients. This explains why the nodal staging score was not significantly associated with overall survival in pN0T3 patients and inversely correlated with the survival of pN0T4 patients. Thus, nodal examination is recommended for pT1 and pT2 patients, but not for pT3 and pT4 patients, based on there being no benefit of nodal examination in the latter populations.

There were some limitation of this study. Although this was a multicenter study, it was also retrospective in nature. Some important clinical variables were not available for this study due to the lack of enrollment control, and some bias may have existed in patient selection. For example, the distribution of primary tumor stages was unbalanced. Most patients were pT1 and pT2 (N=283,219, 93.24%), while the samples size of the pT4 cohort was 7430 (0.024%).

Conclusion

Our study investigated the number of lymph nodes needed to be removed for optimal diagnosis and found that the current number (12 nodes) is sufficient for pT1 patients (eight recommended), but inadequate for pT2 patients (37 recommended). Nodal examination is not recommended for stages pT3 and pT4 based on the prevalence of nodal invasion.

Declaration of conflicting interest

The authors declare that there is no conflict of interest.

Funding

National Key R&D Program of China (2018YFB1307700) to Liping Sun.

Ethics and consent

All data used in this study are fully publicly accessible, and the authors did not participate in data collection and generation. Thus, written consent is not necessary.