Growth pattern-based grading of pulmonary adenocarcinoma-Analysis of 534 cases with comparison between observers and survival analysis.

OBJECTIVES

The 2015 WHO classification of pulmonary adenocarcinoma recommends recording observed architectural growth patterns in 5% increments for resected tumors, and determining the predominant growth pattern, which seems to be prognostic. There is debate over the best way to implement pattern-based grading, and whether such systems are reproducible.

MATERIALS AND METHODS

534 resected adenocarcinomas were reviewed by 2 pulmonary pathologists to determine predominant pattern and percentages of all observed patterns. Three different grading schemes were applied based on predicted prognosis scores: score 1 (lepidic), score 2 (acinar/papillary), and score 3 (solid/micropapillary/cribriform). Mucinous tumors were separately evaluated as both scores 2 and 3 since their prognosis is more ambiguous. The first grading scheme used the score of the predominant pattern; the second used the worst observed pattern score; and the third scored tumors with ≥80% lepidic growth as 1, tumors with ≥20% of any score 3 pattern(s) as 3, and all remaining as 2.

RESULTS

The predominant pattern assigned by each observer was an exact match in 51.7% of cases, a "close match" in 27.3% (same prognosis score), and a mismatch in 21%. Predominant pattern determined by both observers showed significant stratification of overall and progression-free survival (OS and PFS, respectively). All 3 grading schemes showed a significant difference in OS and PFS determined by both observers; but the worst score scheme provided suboptimal results, likely due to a very small score 1 group, and this scheme did not maintain significance on multivariable analysis. Survival differences for all grading schemes maintained significance whether mucinous was considered score 2 or 3, but mucinous tumors trended towards poor survival.

CONCLUSION

Pattern-based grading has prognostic significance in pulmonary adenocarcinoma. Interobserver variation is present, but two observers were able to predict significant differences in OS and PFS using various pattern-based grading schemes.