This month's Archives of Pathology & Laboratory Medicine (February 2010) publishes abstracts from the meeting we had last June in Portland. This was one of the best meetings I have ever attended! The presentations were all current summaries of the state-of-the-art in a variety of lung and pleura topics. I want to call attention to 3 notable abstracts in Archives.
The first one is actually notable to avoid (in a way) but I want to comment on because I think it could be done better and be more helpful the pathologists who "daily sign out" lung cases. "Interobserver Reproducibility for the 2004 World Health Organization Classification of Non-Small Cell Lung Cancer." This study takes 12 "expert pathologists" (EP) and 12 "community pathologists" (CP) and measured the interobserver reproducibility (IOR) by the kappa statistic using H&E "virtual slides" from 96 lung tumors based on the 2004 WHO classification. The authors report comparisons between various major classes (SqCa versus non-SqCa, AdCa versus non-AdCa, EP versus CP) and categories accounting for "grade" (non-poorly differentiated SqCa versus non-SqCa, pdSqCa versus non-SqCa, non-pdAdCa versus non-AdCa, pdAdCa versus non-AdCa). Not surprisingly, EPs have a higher degree of agreement than CPs in every category, although even in the "basic" binary comparisons (SqCa vs. non-SqCa and AdCa vs. non-AdCa), the "best" agreement is relatively modest (for EPs kappa 0.64 for SqCa vs. non-SqCa, 0.69 for AdCa vs. non-AdCa) and slightly more than a coin-toss for all.
I realize that this is an abstract but, if this has been written for publication as a paper, I hope that the reviewers look past the reputations of the authors and consider (and push them) on some of these points. My overarching objection to these sorts of "studies" is that they are "fake." They do not re-create or represent what one encounters in daily practice. Yet it would be and is very valuable to appreciate the variability and reproducibility of diagnostic categories First, how were the specimens selected and by whom? A study of consecutive, unselected cases would most faithfully represent actual practice. Also, the type of specimens studied is crucial: are these CT-guided needle biopsies, bronchoscopic specimens, resected tumors, mediastinoscpic node specimens? It would be valuable to compare the diagnostic agreement both within and between these different types of specimens so that we could communicate the diagnostic uncertainty with better precision to clinicians address issues related to specimen adequacy regarding the most basic decision-point for NSCLC (SqCa vs. AdCa). The IOR doesn't surprise me if the sample includes small samples but I would expect more agreement for resected specimens that were adequately sampled. Also, the number and type of "virtual slides" needs to be specified. Does anyone make a diagnosis, virtual or real, on one slide? Again, the study does not reflect actual practice. Second, the idea of incorporating the idea of "grade" (i.e. "poorly-differentiated") into the diagnostic categories needs to be junked or radically de-emphasized. "Grading" of NSCLC is totally subjective and, unless the WHO folks or an expert panel or another ad hoc group wants to define and validate a grading system, the authors should just leave this alone. Again, since there is no defined, validated grading system for NSCLC (I'm thinking here of something like the Nottingham system for invasive ductal carcinoma of the breast where grading is related to survival), one CP's "poorly-differentiated (fill-in-blank) carcinoma" is a EPs "moderately-differentiated (fill-in-the-other-blank) carcinoma" (or vice versa!). Double-down the above point since this study is based just on apparently just one H&E slide--does anyone actually practice like this?! Finally, the conclusion that "IOR improvement may be possible (my emphasis) with reflex screening for mucin and type-specific immunophenotypes." May be possible?! Are you serious?! Doesn't everyone already reflexively do mucin and p63/TTF-1 for any case that doesn't show unequivocal squamous or glandular differentiation--especially in small biopsies? I consider this standard of care in 2010. Here is the study I think we need: a large unselected series of consecutive diagnostic NSCLC lung tumor surgical pathology biopsies (CT-guided, endo- or transbronchial biopsies, wedge resections, lobectomies) from statistically representative community hospitals and tertiary care hospitals (so it will be weighted toward community settings where the vast majority of lung cancers are diagnosed) using WHO 2004 classifications, digitized slides from all slides of the tumor, and mucin/p63/TTF-1 (on all cases? or all cases with exclusive solid pattern? or all cases not definitively showing squamous or adenocarcinoma features?) and compare the diagnoses between a different group of EPs and CPs not from these institutions. This would give us information on the diagnostic agreement and uncertainty based on present practice, on small biopsy versus resected diagnostic specimens, the effect of the presence of solid pattern on diagnostic agreement in adenocarcinoma, and info on the sensitivity, specificity, PPV and NPV of the special stains and a benchmark for other IHC stains in the future.
SInce this post is already way too long and a bit of a rant, I do want to briefly end with two other interesting abstracts that you might keep an eye out for subsequent papers:
"Stem Cell Marker Expression in Pulmonary Carcinoma Associated with Histologic Type and Differentiation" Stem cell marker expression is obviously a rapidly developing and exciting area right now not only for lung but other solid organ carcinomas as well. This study looks at Mushashi-1 as a marker which I've not seen elsewhere in the literature.
"Anaplastic Mesothelioma: A Nonrecognized Histologic Pattern." Not anymore. I like the conclusions to this abstract--simple, straight-forward--"The presence of high-grade or anaplastic features in either sarcomatoid or epithelioid tumors of the pleura should not prompy a diagnosis of sarcoma or carcinoma or discourage a diagnosis of malignant mesothelioma." Bam! I would love to see this published as a paper with good illustrations.
Recent Comments