One of the more interesting websites that has some interesting tools for making cancer-related predictions is nomograms.org, part of the website for Memorial Sloan Kettering Cancer Center. A nomogram is similar to the venerable slide rule, the old calculating device, but the important thing to know is that it allows for people to either use a pencil and paper to assign points for each of various risk factors that can be added together in a weighted way to produce a master "predicted risk" for a problem. Folks at Memorial Sloan Kettering and other centers have produced some very helpful tools, particularly in other cancer settings. Their predictive calculators for risks in prostate cancer can be very helpful in predicting outcomes after various treatments for prostate cancer based on multiple very relevant clinical variables.

There is only a single nomogram relevant to the field of lung cancer, and it relates not to treatment outcomes but to risk among current or former smokers of developing lung cancer in the next 10 years. This could be very helpful in determining who has a risk high enough to warrant screening CT scans, for instance (it can also demonstrate the difference in risk by quitting smoking vs. not). This tool is based on some very complex computerized modeling work done on a collection of more than 18,000 people enrolled on a trial of lung cancer prevention (looking at beta carotene and a form of vitamin A, and it didn't work). From this modeling, a tool was developed that can predict risk in a limited group of people who were represented by the data on the trial (full article on the development of this tool here): it can only provide predictions for current or former smokers who quit within the last 20 years, smoked between 10 and 60 cigarettes per day for between 25 and 55 years, and are currently between the ages of 55 and 75. These are the situations for which the large trial data were available, so the calculation only applies for this population. The tool looks like this:

(Click to enlarge)

and the website where it's available is here.

It would be great if we had more tools like this, based on large databases of cases with multiple variables and outcomes data. In a way, work I've described on serum proteins predicting survival on EGFR inhibitors (prior post here) and tumor gene signatures (prior post here) have a similar strategy, just looking at molecular instead of clinical variables. I hope we see a lot more of these, calculating based on molecular or clinical variables or both, in the next few years to help us refine our assessments of risk vs. benefit and make better predictions of what treatments will be most helpful.