Grading system for lung adenocarcinoma still elusive
Posted Sep 07 2010 2:33pm
The August 2010 issue of American Journal of Surgical Pathology features an article from the Memorial Sloan Kettering Cancer Center group proposing a grading system for lung adenocarcinoma in stage 1 cancer.
Sica G, Yoshizawa A, Sima CS, et al. A grading system of lung adenocarcinoma based on histologic pattern is predictive of disease recurrence in stage I tumors. Am J Surg Pathol 2010;34:1155-1162.
The authors articulate the frustration pathologists experience in dealing with the histological heterogeneity of lung adenocarcinoma (ACa) since the vast majority are mixed subtype as well as the lack of an objective, clinically relevant grading system for lung ACa. Recent literature has shown that mixed subtype tumors with solid or micropapillary components appear to have worse outcomes compared with those that do not, while tumors with a predominantly bronchioloalveolar (BAC) component appear to have a better prognosis.
An important part of this study is the idea of "comprehensive histologic subtyping" per the mothod outlined by Motoi et al (Am J Surg Pathol, 2008). It is only by essentially submitting the entire tumor for histology that one can accurately assess the histologic heterogeneity of a tumor and semiquantitatively identify the various subtype components. Another interesting feature of this paper is the use of a "concordance probability estimate" in their statistical analysis. Although the term seems like more complex statistical jargon, this measure seems useful to me in that it provides an estimate of the ability of each grading system to predict the outcome for an individual case.
To identify patterns of dissemination of different histological subtypes, the authors retrospectively studied 73 stage IIA-IV primary lung ACa and corresponding metastatic tumors. The majority of these tumors were composed of predominantly acinar subtype in 48%, solid subtype in 19%, and papillary in 19%. While tumors at each of the metastatic sites also showed a mixture of patterns, the predominant subtype pattern was more likely to be found at the metastatic site. Of note, even when the solid or micropapillary patterns were a small percentage of the primary tumor, these patterns are disproportionately more often found in the metastatic site.
Based on this work the authors devised a 3-tier grading system based on the mix of histological subtypes: grade 1, predominantly BAC pattern; grade 2, acinar and papillary, and grade 3, solid and micropapillary patterns. The second part of this study was to test this grading system and a number of different scoring systems in 366 stage 1 ACa for the ability to predict disease-free survival (DFS).
In the stage 1 ACa group, predominantly acinar subtype was seen in 49%, papillary subtype in 26%, solid in 13%, BAC in 9% and micropapillary in 3%--a slightly different distribution (but perhaps expectedly so) compared to the high stage group. But here is where this study, imho, jumps off the rails. I refer the reader to Table 2 on page 1159 for more details but I'll try to summarize.
The authors show some value to reporting the most predominant grade which showed nice separation of DFS but only had a "concordance probability estimate" of 0.61--that is, in 2 randomly selected patients, this system would order them correctly with respect to DFS in 61% of cases. The authors preferred scoring system based on the two most prominent grades only performed marginally better with a concordance probability estimate of 0.65. In looking at the data, I would have preferred a scoring system of predominant grade + highest grade because this showed the best separation between the three tiers if you look at the 95% CI of the hazard ratios and yet still had a concordance probability estimate of 0.62. The reality is that none of these systems actually perform well enough to warrent their clinical use. The authors' own preferred system seems like it would be less reproducible than the others that were evaluated because it completely relies on semiquantitative estimates of proportions of patterns. I think it could also underestimate the DFS for tumors which have small but significant proportions of solid and micropapillary patterns.
Anyhow, it is a start and much further work is needed before we get something that approaches, for example, the Nottingham scoring system for breast cancer.