Introducing the Node Reporting and Data System 1.0 (Node-RADS): a concept for standardized assessment of lymph nodes in cancer

Evaluation of lymph nodes for the likelihood of disease involvement is important in the context of cancer staging because nodal involvement is a powerful adverse prognostic indicator that often determines patient management, frequently distinguishing surgical candidates from those best suited for non-surgical management [1, 2]. In most cases, the incidence of nodal involvement increases with tumor bulk and stage and is dependent on tumor histological type and grade.

Even though numerous studies have evaluated morphologic criteria in computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound (US), practically there is still no consensus on which criteria should be used, although nodal size is generally accepted [3]. Paralleling approaches to assess size include short- and long-axis diameters, as well as volumetric measurements. However, lymph node size is a poor indicator for predicting the presence of secondary malignancy. In a study focusing on mesorectal lymph nodes in rectal cancer patients, Gina Brown and colleagues showed a broad size overlap between histologically benign (2–10 mm) and malignant (3–15 mm) lymph nodes [3]. Furthermore, using nodal size as a criterion for metastatic involvement can be confusing in the head and neck (HN) region as benign nodes have different sizes depending on patient age and on anatomic location; e.g., submandibular nodes are typically larger than lymph nodes in other neck groups (sometimes called stations/levels). The choice of the size cutoff will influence sensitivity and specificity for the detection of nodal metastases. In a HN study, Curtin and colleagues showed that a size cutoff of 10 mm in the largest axial nodal diameter results in a sensitivity and specificity of 88% and 39%, respectively, whereas a size cutoff of 15 mm yielded corresponding values of 56% and 84%, respectively [4]. Also, there is no consensus on whether lymph nodes should be measured in the axial or craniocaudal dimensions.

Criteria for lymph node configuration also exist with numerous descriptors, which can sometimes be helpful [5‐8]. There have been some promising approaches by combining size and configuration criteria to facilitate and standardize the diagnostic workup for lymph nodes at specific anatomic locations, such as in the mesorectum in a European Society of Gastrointestinal and Abdominal Radiology (ESGAR) consensus statement [9]. However, focusing on specific disease entities and body sites limits the practical application of the approach in clinical practice.

In addition to being highly dependent on the interpretative skills of radiologists, unstructured reports that use a variety of terms to describe the likelihood of disease involvement also cause difficulties for clinical referrers making treatment decisions. Recently, there have been attempts at standardizing the reporting of oncological scans with the increasing adoption of the Reporting and Data Systems (RADS) for many scenarios for the detection and characterization of lesions within specific organs. For example, BI-RADS is well established for reporting on the detection of breast cancer using x-ray mammography [10]. Recently, other systems incorporating the -RADS acronym have emerged including LI-RADS for the MRI evaluation of hepatocellular carcinoma in liver cirrhosis, NI-RADS for surveillance of head and neck cancer, and PI-RADS for the MRI detection of prostate cancer, finding acceptance among radiologists and referring clinicians alike [10‐13]. Structured reports can improve the reliability and validity of imaging assessments in the clinical routine but can also aid research studies and audits.

Generally, -RADS systems incorporate a Likert-type scale for the likelihood of the target disease being present. The use of scoring systems to assess the likelihood of malignant disease involvement has emerged, but is limited to specific body sites, cancer entities, time of evaluation, or imaging modalities [14‐17]. There is no generally applicable imaging system for the whole body that indicates the scaled likelihood of malignant nodal disease involvement.

Node-RADS is a concept that addresses the aforementioned limitations, aiming to enhance the reporting of regional and distant lymph nodes in cancer patients by (i) clearly defining imaging criteria to increase consensus among radiologists, (ii) facilitating standardized lymph node reporting, and (iii) having broad applicability to cancer types across multiple anatomic sites, being evaluated by CT and/or MRI. To serve these purposes, the Node-RADS scheme is graphically depicted, is logically arranged, and is intuitive, enabling its adoption into the clinically routine without the need for additional time expenditure.

The evaluation of a lymph node using the Node-RADS scheme results in an assessment category scored between 1 and 5, which reflects the level of suspicion for involvement by malignancy: “1—very low”; “2—low”; “3—equivocal”; “4—high”; “5—very high.” For this purpose, the interpreting radiologist is guided through a three-level flowchart (Fig. 1). Levels 1 and 2 address the two principle imaging criteria of “size” and “configuration.” Level 3 provides the resulting Node-RADS assessment score. Categories, definitions, and features of the respective criteria are also given in Fig. 1.

There is an ongoing debate about the setting of limits for physiologic and pathologic size of lymph nodes. Ultimately, any chosen size cutoffs depend on the desired sensitivity and specificity concerning the stage of the primary tumor and/or clinical status of patients [18, 19]. Many radiologists do not consider nodal size to be an absolute criterion for the assessment of likely malignant involvement. However, since the size of lymph nodes is a criterion that is widely used, and is incorporated into the RECIST (Response Evaluation Criteria in Solid Tumors) criteria for response assessment studies [20], Node-RADS divides nodal sizes into three categories: “normal,” “enlarged,” and “bulk.”

In general, a lymph node is defined to have a “normal” size, when its short-axis diameter is < 10 mm. Node-RADS defines exceptions for normal size for inguinal nodes which could measure < 15 mm in short-axis diameter; on the other hand, Node-RADS sets a lower cutoff for specific subregions, where the normal short-axis diameter should be < 5 mm: facial, parotid, retroauricular, occipital, retropharyngeal, anterior jugular, retrocrural, cardio-phrenic, mesenteric, obturator, and mesorectal lymph nodes [20]. A “bulk” is defined as a lymph node with the longest diameter of ≥ 30 mm measured in any dimension. Since there are no generally accepted measurements for the term “bulk,” especially outside of lymphomas, we were guided by the TNM classification for head and neck cancer, where a lymph node diameter of 30 mm is the threshold for N2 (with the exception of HPV-positive oropharyngeal, nasopharyngeal, and thyroid cancer) [1]. Lymph nodes of “enlarged” size do not meet either the definitions of the categories “normal” or “bulk” or are between 10 and 15 mm (short axis) for RECIST purposes, or show interval increases of ≥ 2 mm in short-axis diameter, if prior imaging datasets are available for comparison. The ≥ 2-mm threshold is based on a consideration regarding the spatial resolution of CT scans (512 matrix over a 40–50-cm field-of-view results in a pixel size of 0.78 to 0.98 mm). The requirement of two pixels for a reliable measurement translates into axial dimensions of ≥ 2 mm.

Node-RADS seeks not to overinterpret the size of lymph nodes, which in a specific case means that an “enlarged” lymph node by size criteria without accompanying abnormal morphologic features discussed under the “configuration” criteria (below) will receive a Node-RADS score of 2 (“low suspicion”).

After using the “size” criterion in the first-level evaluation, the second level considers other anatomic features under the criterion of “configuration” which have a critical importance for the final Node-RADS assessment score. The “configuration score” is made up of the summed numerical value from the three sub-categories of “texture,” “border,” and “shape.”

One feature is chosen from each sub-category, which translates into a minimum achievable “configuration score” of 0 points, and a maximum possible score of 5 points.

As indicated in the Node-RADS flowchart (Fig. 1), each “configuration score” consecutively translates into the respective final Node-RADS score.

The TNM system places regional nodal involvement in the N category, but nodal involvement at other than regional sites is classified as distant metastases (i.e., belongs in the M category). It is therefore important for radiologists to know where regional and distant metastatic sites reside for each primary tumor. These details can be found in staging manuals, preferably in the AJCC Cancer Staging Manual [1]. Sometimes, the same organ may have different regional nodal groups; thus, the retroperitoneum is a distant metastatic site for cervical cancer but defined as “regional” for endometrial cancer.

It is important to note that the TNM system emphasizes different aspects for nodal involvement depending on the primary tumor; thus, for the bladder and head and neck cancers, nodal size is part of the N category. For many adenocarcinomas, the presence or absence of microscopic metastatic disease, regardless of primary tumor burden, is emphasized, whereas nodal involvement sometimes does not alter staging category at all (e.g., for well-differentiated follicular/papillary thyroid cancers in patients less than 45 years of age).

The RECIST system is based on the measurement of target lesions, summing up lesion sizes to assess tumor response to therapy [20]. According to RECIST v1.1., lymph nodes should be measured along the short axis in the axial plane. To be considered pathologically enlarged and measurable, the short-axis diameter should be ≥ 15 mm. If the short-axis diameter is between 10 and 15 mm, they are considered pathologic, but cannot be used as target sites (non-measurable disease) [24].

In order to account for temporary nodal enlargement without clinical disease progression during immunotherapy, the RECIST system has been modified, so creating the iRECIST (immune-related RECIST) system [25].

The following general rules apply to Node-RADS regarding RECIST:

Node-RADS can be used and reported in addition to already existing RADS (e.g., LI-RADS or PI-RADS). An exception is NI-RADS in the surveillance of head and neck cancer, where a score is already assigned for the cervical lymph nodes (“neck”). The criteria and algorithm of Node-RADS and NI-RADS are not concordant.

Undoubtedly, independent prospective studies on reliability (i.e., inter- and intra-reader agreement) and validity are mandatory to make any adjustments to the Node-RADS system proposed here. Although the combination of CT or MRI with functional sequences (e.g., perfusion imaging) or PET is widely used, these are intentionally not included in Node-RADS at this time to facilitate the straightforward use of the system. Ultrasound is frequently used for evaluation of superficial lymph nodes applying criteria that differ distinctly from CT and MRI and therefore not included in Node-RADS. However, as in existing -RADS, changes could expand the scope of the application upon the availability of further evidence [12]. Node-RADS is likely in need to be periodically modified in response to independent prospective studies that evaluate its efficacy, which may include the inclusion of additional imaging features, e.g., clustering of lymph nodes, or adjustments on the size cutoff for bulky disease. Finally, minimum technical imaging parameters might be a subject for inclusion in the standard in future iterations of Node-RADS.