Melanoma Tumor Location: A Utility-Driven Approach to Common Data Elements


The National Institute of Health (NIH) Common Data Elements (CDEs) are a collection of data definition standards used in clinical trials, typically consisting of a question and set of permissible responses. For example, the variable “Ear” has two permissible values, “Left” and “Right”.  NIH has largely relied on the investigative community to create and enhance CDEs. Although helpful, the primary risks and limitations of this approach are:

  • Research Tail Wagging the Dog

  • Theory-driven new or more precise empirical targets guide CDE changes and enhancements, which are often nuanced and years ahead of clinical significance and/or utility.

  • An Un-Crossable Chasm

  • Poor adoption outside the research community due to limited clinical utility.

This article is intended as an example approach to advancing CDEs in a manner that addresses these two limitations.  A case study of CDE modification is presented using melanoma ‘Tumor Location’ (Public ID: 4320475). In the pursuit of data definition standards, broad adoption is best stimulated through an approach that is guided by both theoretical and applied clinical concerns, and that also provides tools, materials and ideas that leverage the CDE’s use. Broad adoption, in turn, is key to large scale data aggregation and analysis initiatives, providing the basis of further standards development. It is hoped that others will use these tools, materials and ideas presented here in their work. The following topics are covered:

  • Clinical Utility: A parallel driver of CDEs

  • CDE Changes

  • Standardized Image and Pixel Coordinates

  • Javascript Tumor Plotting Functions

Clinical Utility: A parallel driver of CDEs

Providing clinical utility is one of the most effective ways to ensure widespread adoption of data standards, as well as accurate and timely data capture. Thus, changes to the tumor location CDE and mapping to a standardized image was only one component of a broader strategy to leverage patient registry data. Indeed, data capture alone, even if entirely standards driven, is of little benefit within a clinical practice if not effectively leveraged in a timely fashion. Consider the overwhelming number of forms and variable values collected in the example case below (see image). Transforming these data into actionable information was a critical aim.

As a part of this broader context, the components of an example patient report is presented below (note: different cases to demonstrate each component). Although a detailed rationale is not pursed in this article, color and spatial relationships were thought key considerations for the effective and efficient use of these data.

Basic demographics, diagnosis and staging information (partial list).

A standardized image with example indicator images (i.e., red circle) over tumor areas.

Tumor number within each procedure site, conditionally colored red if tumor at site

Vitals Over Time

Number of data collections for each event type

Indicator of what tumor specific information is available in the registry

Example 1: Treatment Timeline table (see ‘Event Table’ for color coding)

Example 2: Treatment Timeline table (see ‘Event Table’ for color coding)

Timeline details table (see ‘Event Table’ for color coding)

Most recent tumor specific information

Toxicity list.

CDE Changes

Melanoma tumor location is an important prognostic factor (1-3) and is an existing CDE in its first iteration (see Table 1).

Table 1: Melanoma Anatomic Site

Table 2 compares the proposed and the existing response set.  Modifications targeted new sites, increased granularity (e.g., face), and/or lateralization (i.e., left/right/mid-line/other).  Since the rationale for modifications may vary considerably across settings, it is important to note the intended target was an academic surgical registry. Medical Subject Headings (MeSH) Tree Numbers and IDs are presented for the proposed sites, if available.

Table 2: Melanoma Anatomic Site: Existing and Proposed Values

Note: * = Proposed new site. Laterality – L = Left, R = Right, M = Midline, O = Other.

Standardized Image and Pixel Coordinates

The image below captures all possible tumor locations. An indicator image (e.g., red circle) can be overlay-ed on the image for each tumor.

Tumor Location Pixel Coordinates

The X, Y coordinates that centers the indicator image are provided below. Note X, Y coordinates have .5 added merely due to the programming code allowing for a variable height and width of the indicator image (height = width in this case [i.e., a circle]), and the best size was thought to be 15 pixels. Since half the height/width was used for centering the image (i.e., 7.5), the .5 was added to the x,y coordinates to ensure an integer result (see javascript section further below).

Skin and Subcutaneous

Javascript Tumor Plotting Functions

The javascript program puts all tumor site locations into an array, and then passes the information to the “drawCirlce” function below. Several parameters were added to the function to allow for enhancement of the indicator image as needed. Note, the ‘count’ parameter indicates how many tumors are in a given location. If more than one, the total number is displayed within the indicator image.

Having these additional parameters sets the stage for a number of exciting possibilities in pursuit of quick and efficient consumption of data. For example, varying the size of the indicator image based on the Clark level. Look for more articles coming soon in epilepsy and pancreatic cancer that will leverage these parameters. As Studytrax is used in numerous disease populations and therapeutic areas, would be great to hear about the reader’s use or modifications of this function (

At the location on the webpage where the image is to be shown, insert the following code block.

Contact for any questions or materials related to this article.


1. Chakera AH, Quinn MJ, Lo S, Drummond M, Haydu LE, Bond JS, Stretch JR, Saw RPM, Lee KJ, McCarthy WH, Scolyer RA, Thompson JF. Subungual Melanoma of the Hand. Ann Surg Oncol. 2018 Dec 18. 

2. Beaudoux O, Riffaud L, Barbe C, Grange FEur. Prognostic factors and incidence of primary mucosal melanoma: a population-based study in France.  J Dermatol. 2018 Oct 1;28(5):654-660.

3. Farahi, J. M, Fazzari, M., Braunberger, T., Caravaglio, J. V, Kretowicz, A., Wells, K., et al. (2018). Gender differences in melanoma prognostic factors. Dermatology Online Journal, 24(4).

© Sciencetrax, LLC

  • LinkedIn