In the last post I discussed some interesting work from a group in Japan that suggested that the number of lymph nodes that are removed and positive for NSCLC may be a very important prognostic variable, potentially an even more important factor than location of the nodes, which is the basis for how we stage nodal involvement in NSCLC now. Following along the same theme was another interesting presentation I reviewed at the recent ASCO meeting that came from Rome, where investigators at the Regina Elena National Cancer Institute reviewed the detailed results of 415 consecutive patients who underwent surgery there for stage I to stage IIIA NSCLC (abstract here). They attempted to assess the prognostic value of many variables reported to be useful in prior work, such as age, patient sex, tumor size, tumor grade, and also focused on the variables of the total lymph nodes removed at the time of surgery (regardless of whether they were involved with cancer or not), and the proportion of positive nodes compared with the total number removed (related to the concept in the previously described Japanese abstract about the number of positive nodes as a number alone). From there, they also developed a new system of weighted variables that could assign patients as low, intermediate, or high risk of recurrence and death from lung cancer after surgery.

They performed what is known as a Cox regression multivariate analysis, which basically means that they did a very complex and detailed statistical analysis to see which variables were more important in an overlapping collection. For instance, if you have an elderly woman with a poorly differentiated, 6 cm squamous cancer and 2 of 15 lymph nodes positive, this statistical analysis looks at the total collection of data to discern what contribution of her outcome derives from her age, sex, tumor size, and all of the other associated variables. This analysis found that the most important variables were patient age (worse outcomes as patients get older, not surprisingly), patient sex (women do significantly better than men, all other variables being equal), stage (higher being worse, obviously), tumor grade (poor grade worse than well or moderately differentiated), and also total nodes resected (the more nodes removed at surgery, the better patients were likely to do), as well as proportion of nodes positive vs. total nodes (higher proportion positive associated with worse survival).

They also looked at the cut-off points that worked best for separating better outcomes from worse ones, when the results weren't just categorical (male vs. female), but are continuous (number of nodes removed, patient age). They found that age 67 was an effective cut-off, for instance, and for the lymph node variables, 10 or more nodes removed separated the better from worse outcome patients, as did 9% of resected lymph nodes being involved with the cancer. Here are the curves that show the differences based on total nodes removed and percent of positive nodes out of the total:

They then identified the most important variables that seemed to predict who would have the worst survival, identifying 4 high risk factors of stage (IIB and IIIA worse than earlier stages), node status (any nodes involved vs. none), age (67 or younger vs. older), and number of lymph nodes resected (10 or more vs. fewer). Beyond that were three additional intermediate variables that were prognostic but not as associated with survival differences: patient sex (male worse than female), tumor grade (poorly differentiated as a worse factor), and tumor histology (squamous designated as a worse factor). From that designation, they created a model that assigned patients as high, medium, or low risk based on the number of these high and intermediate risk factors that each patient had:

This assignment of risk classes did a great job of separating patients by their subsequent survival, as shown here:

Overall, my impression is that this work is very well done, and that there is likely to be a lot of truth and value in their observations. Several of the variables that they identified here are ones that have already been covered in other posts here as being relevant in the population with resected NSCLC. There are always limitations, however, in using cases from a single center, with the surgeries largely done by one or a few surgeons over several years, to generalize the conclusions to the rest of the world. Surgical practice may be different in Rome and/or Japan than in other places, the characteristics of the cancers may be different in different parts of the world (indeed, European NSCLC populations often have a higher proportion of squamous than adenocarcinoma, unlike North America and Asia; and in Japan there are much larger proportions of never-smoking women with NSCLC that is almost always adenocarcinoma/BAC, but this is likely very specific to Asia). Before any of the findings from a retrospective study can be truly accepted, they need to be validated in other settings.

But the key issue and major limitation that I see is that these results represent excellent lung surgery, and unfortunately that isn't something that we can presume is going to be widely available everywhere. It's certainly not as common as we need to see in the US. I'll cover that topic of US-based standards next.