Patterns of stromal invasion in ovarian serous tumors of low malignant potential (borderline tumors): a reevaluation of the concept of stromal microinvasion.

Stromal-epithelial patterns of invasion in serous tumors of the ovary have been subclassified as destructive and nondestructive. By definition, well-differentiated serous tumors featuring destructive stromal invasion are classified as low-grade serous carcinomas whereas those with either no stromal invasion or stromal microinvasion are classified as serous tumors of low malignant potential (S-LMP). The histologic features of stromal microinvasion in ovarian S-LMP have been addressed in a variety of studies, but controversy persists regarding diagnostic criteria and prognostic significance, particularly in patients with high-stage disease. In addition, a subset of otherwise typical S-LMP has patterns of invasion that are not classic destructive invasion and do not meet the current diagnostic criteria for stromal microinvasion because of either qualitative features or size restrictions. To further evaluate the full histologic spectrum of stromal-epithelial patterns of invasion in otherwise typical S-LMP, we examined a series of 60 ovarian S-LMP (34 FIGO stage I; 26 FIGO stages II, III, and IV) with stromal-epithelial alterations not meeting criteria for classic destructive invasion. This group of cases included those meeting the definition of microinvasion and a subset that would be excluded based on size measurements or unusual qualitative features, but did not exhibit significant stromal reaction. Five patterns of invasion were identified: individual eosinophilic cells and cell clusters, cribriform, simple and noncomplex branching papillae, inverted macropapillae, and micropapillae. Individual, discrete aggregates of invasive epithelium ranged from 1 to 12 mm (mean, 1.4 mm) in greatest linear dimension as measured by conventional methods. The number of discrete foci ranged from 1 to greater than 10; in 7 tumors (12%), the invasive foci were diffusely scattered throughout the stroma without discrete aggregates. These stromal-epithelial alterations were associated with disease progression and/or death due to disease in 9 of 50 (18%) patients with follow-up (mean, 92.5 mo) and were covariant with other adverse prognostic features (invasive implants, nodular lymph node aggregates, high stage, and unresectable disease). Disease progression was most strongly linked to the presence of micropapillae, but the majority of patients with adverse outcome had the more common, classic stromal-epithelial patterns associated with microinvasion (ie, individual cells, cell clusters, and simple papillae). Neither size of the largest contiguous aggregate nor extent of stromal involvement correlated with outcome. Classic microinvasion disproportionately occurred in patients presenting during pregnancy (P<0.0001), and was not associated with adverse outcome in that setting, but follow-up was limited. Based on the cumulative outcome data, the presence of stromal-epithelial patterns of invasion distinct from classic destructive invasion in otherwise typical S-LMP stratifies patients at long-term risk for disease progression, but does not warrant a diagnosis of carcinoma or a change in current management schemes. Maintaining classification as a serous tumor of low malignant potential (serous borderline tumor) with stromal invasion seems appropriate even in the presence of diffuse stromal involvement or discrete aggregates measuring greater than 3 (or 5) mm. As the stromal-epithelial alteration featuring micropapillae may represent a comparatively higher-risk lesion with a clinical course analogous to that of low-grade serous carcinoma, pathologists should identify this specific stromal-epithelial pattern in the diagnostic report until sufficient data is acquired to form more definitive conclusions regarding its prognosis.