Grading of follicular lymphoma: comparison of routine histology with immunohistochemistry

Recommendations on compiling test datasets for evaluating artificial intelligence solutions in pathology

Artificial intelligence (AI) solutions that automatically extract information from digital histology images have shown great promise for improving pathological diagnosis. Prior to routine use, it is important to evaluate their predictive performance and obtain regulatory approval. This assessment requires appropriate test datasets. However, compiling such datasets is challenging and specific recommendations are missing. A committee of various stakeholders, including commercial AI developers, pathologists, and researchers, discussed key aspects and conducted extensive literature reviews on test datasets in pathology. Here, we summarize the results and derive general recommendations on compiling test datasets. We address several questions: Which and how many images are needed? How to deal with low-prevalence subsets? How can potential bias be detected? How should datasets be reported? What are the regulatory requirements in different countries? The recommendations are intended to help AI developers demonstrate the utility of their products and to help pathologists and regulatory agencies verify reported performance measures. Further research is needed to formulate criteria for sufficiently representative test datasets so that AI solutions can operate with less user intervention and better support diagnostic workflows in the future.

Single-cell analysis of AIMP2 splice variants informs on drug sensitivity and prognosis in hematologic cancer

Aminoacyl-tRNA synthetase-interacting multifunctional protein 2 (AIMP2) is a non-enzymatic component required for the multi-tRNA synthetase complex. While exon 2 skipping alternatively spliced variant of AIMP2 (AIMP2-DX2) compromises AIMP2 activity and is associated with carcinogenesis, its clinical potential awaits further validation. Here, we found that AIMP2-DX2/AIMP2 expression ratio is strongly correlated with major cancer signaling pathways and poor prognosis, particularly in acute myeloid leukemia (AML). Analysis of a clinical patient cohort revealed that AIMP2-DX2 positive AML patients show decreased overall survival and progression-free survival. We also developed targeted RNA-sequencing and single-molecule RNA-FISH tools to quantitatively analyze AIMP2-DX2/AIMP2 ratios at the single-cell level. By subclassifying hematologic cancer cells based on their AIMP2-DX2/AIMP2 ratios, we found that downregulating AIMP2-DX2 sensitizes cells to anticancer drugs only for a subgroup of cells while it has adverse effects on others. Collectively, our study establishes AIMP2-DX2 as a potential biomarker and a therapeutic target for hematologic cancer.

Ku, Kim et al develop a method to analyse the ratio of the alternatively spliced variant of AIMP2 to full length AIMP via single-molecule RNA-FISH. They can subclassify hematologic cancer based on AIMP2-DX2/AIMP2 ratio and find that cells with high AIMP2-DX2 ratio can be sensitized to chemotherapy drugs by depleting AIMP2-DX2.

Divergent immunohistochemical expression of CD21 and CD23 by follicular dendritic cells with increasing grade of follicular lymphoma

BACKGROUND

Ultrastructural and immunohistochemical differences have been described in FDCs of primary and secondary follicles, illustrating the highly compartmentalized structure of lymph follicles. Differences in FDC immunophenotype in different grades of FL may reflect some parallelism between reactive and neoplastic conditions in terms of FDC-B cell interaction and may be used as a valuable additional tool for grading FL.

METHODS

A total of 60 paraffin blocks from patients with follicular lymphoma, 30 cases each of grade 1 and 3, were retrieved from our archive. Immunohistochemical analysis was carried out for CD21, CD23, cyclin A, and Ki-67.

RESULTS

Our study demonstrates that during evaluation, six patterns of FDC distribution were distinguished. The intensity of stain for CD21 was not statistically significant in grade 1 and grade 3 FL (p = 0.340). In contrast, grade 3 FLs exhibited a significant decrease of CD23 expression by the FDCs (p < 0.001). By CD21 stain, there was no significant difference in the distribution of pattern 1 in grades 1 and 3 (p = 0.098). In contrast, in grade 3, this pattern was significantly less observed by CD23 stain (p = 0.016). The same was observed for pattern 2 for CD21 (p = 0.940) and CD23 (p = 0.010) and pattern 4 for CD21 (p = 0.305) and CD23 (p = 0.005), respectively. Distribution of pattern 5 was significantly different between grades 1 and 3 both for CD21 (p = 0.005) and CD23 (p < 0.001). Distribution of patterns 2 and 6 was not significantly different between grades 1 and 3 for CD21 and CD23. The values of cyclin A and Mib-1 were also significantly different between grades 1 and 3 (p < 0.001).

CONCLUSIONS

The observed patterns of FDCs lead us to believe that similar to reactive lymph node follicles, neoplastic follicles in FL, at least in early stages, have an organized structure. Hypothetically, with CD21, CD23, and cyclin A immunohistochemistry, the sequence of events in FL progression may be traced.

Multicenter Analysis of Advanced Stage Grade 3A Follicular Lymphoma Outcomes by Frontline Treatment Regimen

BACKGROUND

Follicular lymphoma (FL) is a common form of non-Hodgkin lymphoma with a wide spectrum of presentation. While grade 1/2 FL is considered low grade and grade 3B FL is approached as an aggressive lymphoma, the management of grade 3A FL remains controversial.

PATIENTS AND METHODS

We performed a retrospective, multicenter analysis of patients aged ≥ 18 years with advanced stage 3/4 grade 3A FL diagnosed between January 2006 and July 2016. Patients were stratified by frontline chemotherapy regimen: anthracycline based (ATC), bendamustine (BD), and cyclophosphamide, vincristine, and prednisone (CVP). A total of 103 patients were identified from 6 contributing centers: 65 patients received ATC chemotherapy, 30 BD, and 8 CVP. The primary outcome was time to progression (TTP). Secondary outcomes included progression-free survival, overall survival, complete response rates, large cell transformation, and impact of standardized maximum uptake value on positron emission tomography/computed tomography with outcomes. Patient characteristics were similar among the 3 treatment groups.

RESULTS

For TTP at 24 months from initiation of treatment, 72% of ATC, 79% of BD, and 50% of CVP patients had not experienced disease progression (P = .01). Multivariate analysis demonstrated a TTP benefit for ATC compared to CVP (hazard ratio 3.22; 95% confidence interval, 1.26-8.25; P = .01) but no difference when compared to BD. Similar findings were seen with progression-free survival. While overall survival was similar among the 3 arms, there was a higher risk of large cell transformation following BD and CVP. Last, standardized maximum uptake value on positron emission tomography/computed tomography did not affect TTP when comparing BD- and ATC-treated patients.

CONCLUSION

Although ATC was superior to CVP, clinical outcomes (TTP, progression-free survival, and overall survival) were similar compared to BD chemotherapy for patients with grade 3A FL.

Ki-67 expression in mature B-cell neoplasms: a flow cytometry study

SUMMARY OBJECTIVE: Ki-67 is a nuclear protein associated with cellular proliferation in normal or leukemic conditions that can help identify more aggressive diseases and is usually evaluated with immunohistochemistry. The aim of this was to assess Ki-67 expression on mature B-cell neoplasms samples with flow cytometry immunophenotyping. METHOD: After surface staining with CD19 and CD45, intracellular staining for Ki-67 was performed in leukemic mature B-cells. Ki-67 expression was evaluated with flow cytometry. RESULTS: Ki-67 expression was higher in mantle cell lymphoma, Burkitt lymphoma, and diffuse large B-cell lymphoma cases. It was also associated with CD38 mean fluorescence intensity. CONCLUSIONS: Ki-67 expression evaluated by flow cytometry can be a useful tool in the diagnosis of mature B-cell neoplasms. More studies are needed to validate Ki-67 assessment with flow cytometry immunophenotyping.

Sensitivity of fine-needle aspiration biopsy for diagnosing and grading follicular lymphomas using a multiparameter approach in a cancer center

OBJECTIVES

The diagnosis and grading of follicular lymphomas (FLs) by fine-needle aspiration biopsy (FNAB) has not been systematically compared with core needle biopsy (CNB). We evaluated the sensitivity of FNAB in diagnosing and grading FLs using a multiparameter approach in a large cancer center.

METHODS

We retrospectively identified CNBs of lymph nodes diagnosed as FL that also had a concurrently acquired FNAB on the same site. The majority of cases had flow cytometric analysis and these results were available for interpretation of both the FNAB and CNB.

RESULTS

Out of 342 patients, CNB diagnoses included 291 (85%) low-grade (LG) FLs, 30 (9%) high-grade (HG) FLs, and 21 (6%) non-graded FLs/other. FNAB diagnoses included 194 (57%) LG FLs, 19 (6%) HG FLs, 93 (27%) non-graded FLs, 9 (3%) large B-cell lymphomas (LBCL) of follicle center origin, and 27 (7%) insufficient for diagnosis/other. Review of non-graded FLs showed 45% LG, 35% indeterminate due to polymorphous lymphoid cells with increased numbers of large cells, and 20% scant cellularity. Sensitivity of FNAB for diagnosing FL was 89%, and 66% for LG FL. The latter increased (94%), however, when grading was performed.

CONCLUSION

FNAB is highly sensitive for diagnosing FLs when cellular material for cytomorphology and flow cytometric analysis is obtained, and grading is feasible for most LG FLs. A subset of FLs composed of a polymorphous lymphoid population with increased numbers of large cells may be more difficult to grade, and HG FLs can be difficult to distinguish from CD10-positive diffuse LBCLs.

Grading of follicular lymphoma in cytological samples

OBJECTIVE

The treatment of follicular lymphoma (FL) depends on its grade. The current World Health Organization (WHO) 2008 Classification of Tumours of Haematopoietic and Lymphoid Tissues recommends the grading of FL on histological samples according to the Mann and Berard method, taking into consideration the number of centroblasts. There is no generally accepted method for the grading of FL in fine needle aspiration biopsy (FNAB) samples. The aim of the present study was to devise a grading system for FL in cytological samples.

METHODS

Flow cytometry (FC) was performed on 60 FNAB samples of patients with primary FL. We assumed that FL cells larger than reactive T lymphocytes on FC histograms corresponded to centroblasts. The percentage of large cells was calculated and compared with histological grade, proliferative activity and number of centroblasts per high-power field (HPF) on histological slides, and with survival.

RESULTS

The histological analysis of lymph nodes revealed 20 patients with high-grade and 40 patients with low-grade FL. The percentage of large cells in FNAB samples correlated significantly with histological grade (P = 0.02), MIB1 status (P < 0.001) and the number of centroblasts per HPF (P < 0.001). An age over 60 years and a percentage of large cells over 50% in FNAB samples were found to have a statistically significant impact on survival by univariate analysis (P = 0.001 and P = 0.006, respectively).

CONCLUSIONS

The percentage of large lymphoma cells in FNAB samples of FL determined by FC can be used as a reliable method for FL grading, as it is comparable with the histological grading system.

Frequent disruption of the RB pathway in indolent follicular lymphoma suggests a new combination therapy

Loss of cell cycle controls is a hallmark of cancer and has a well-established role in aggressive B cell malignancies. However, the role of such lesions in indolent follicular lymphoma (FL) is unclear and individual lesions have been observed with low frequency. By analyzing genomic data from two large cohorts of indolent FLs, we identify a pattern of mutually exclusive (P = 0.003) genomic lesions that impair the retinoblastoma (RB) pathway in nearly 50% of FLs. These alterations include homozygous and heterozygous deletions of the p16/CDKN2a/b (7%) and RB1 (12%) loci, and more frequent gains of chromosome 12 that include CDK4 (29%). These aberrations are associated with high-risk disease by the FL prognostic index (FLIPI), and studies in a murine FL model confirm their pathogenic role in indolent FL. Increased CDK4 kinase activity toward RB1 is readily measured in tumor samples and indicates an opportunity for CDK4 inhibition. We find that dual CDK4 and BCL2 inhibitor treatment is safe and effective against available models of FL. In summary, frequent RB pathway lesions in indolent, high-risk FLs indicate an untapped therapeutic opportunity.

Software-automated counting of Ki-67 proliferation index correlates with pathologic grade and disease progression of follicular lymphomas

OBJECTIVES

To examine the accuracy of software-assisted measurement of the Ki-67 proliferation index (PI) and its correlation with the grade and clinical progression of follicular lymphoma (FL).

METHODS

High-power field equivalents were extracted from H&E- and Ki-67-immunostained slides of FL, and a nuclear quantitation algorithm was used to calculate a PI. Representative fields were manually counted for validation with close agreement.

RESULTS

The PI was significantly higher in World Health Organization grade 3 FL than grade 1 to 2 FL. Disease progression, as defined by subsequent treatment with radiation or cytotoxic chemotherapy, was also significantly associated with elevated PI but not pathologic grade.

CONCLUSIONS

These data show that software-automated quantitation of Ki-67 can provide both a useful adjunct to pathologic grade in FL and improved prognostic information for patients.

Multispectral imaging and pathology: seeing and doing more

BACKGROUND

The current appreciation of the biological complexity of disease has led to increasing demands on pathologists to provide clinically relevant, quantitative morphological and molecular information while preserving cellular and tissue context. This can be technically challenging, especially when signals of interest are colocalized. With fluorescence-based methods, sensitivity and quantitative reliability may be compromised by spectral cross-talk between labels and by autofluorescence. In brightfield microscopy, overlapping chromogenic signals pose similar imaging difficulties.

APPROACH

These challenges can be addressed using commercially available multispectral imaging technologies attached to standard microscope platforms, or alternatively, integrated into whole-slide scanning instruments.

ASSESSMENT

Multispectral techniques, along with other developments in digital analysis, will allow pathologists to deliver appropriate quantitative and multiplexed analyses in a reproducible and timely manner. Caveats apply - as the complexity of the sample preparation and analysis components increases, commensurate attention must be paid to the use of appropriate controls for all stages of the process.

Imaging in follicular NHL

Imaging contributes to management of follicular lymphoma (FL) through guiding biopsy, determining disease stage and assessing therapeutic response. Molecular imaging with positron emission tomography (PET), especially when combined with computer tomography (PET/CT), is more accurate than conventional imaging and extends the role of imaging to lesion characterisation, including non-invasive assessment of high-grade transformation. There is strong data to support the use of FDG PET/CT for primary staging, resulting in significant management change. In patients with early stage follicular lymphoma (stage I or II), there is a clear role for PET/CT to avoid futile involved-field radiotherapy in patients with widespread disease and to optimise the treatment field in patients with confirmed localised disease. For restaging, use of PET/CT allows discrimination between scar tissue and viable tumour in residual masses. Molecular imaging is likely to play an increasing role in selection of patients for specific treatments and in prognostic stratification.

Computer-aided detection of centroblasts for follicular lymphoma grading using adaptive likelihood-based cell segmentation

Follicular lymphoma (FL) is one of the most common lymphoid malignancies in the western world. FL has a variable clinical course, and important clinical treatment decisions for FL patients are based on histological grading, which is done by manually counting the large malignant cells called centroblasts (CB) in ten standard microscopic high-power fields from H&E-stained tissue sections. This method is tedious and subjective; as a result, suffers from considerable inter and intrareader variability even when used by expert pathologists. In this paper, we present a computer-aided detection system for automated identification of CB cells from H&E-stained FL tissue samples. The proposed system uses a unitone conversion to obtain a single-channel image that has the highest contrast. From the resulting image, which has a bimodal distribution due to the H&E stain, a cell-likelihood image is generated. Finally, a two-step CB detection procedure is applied. In the first step, we identify evident nonCB cells based on size and shape. In the second step, the CB detection is further refined by learning and utilizing the texture distribution of nonCB cells. We evaluated the proposed approach on 100 region-of-interest images extracted from ten distinct tissue samples and obtained a promising 80.7% detection accuracy.

Survival in non-Hodgkin's lymphoma by histology and family history

PURPOSE

Although survival has been studied for various subtypes of non-Hodgkin's lymphoma (NHL), there have been few comprehensive studies to quantify the prognosis, including all specific histologies. The effect of family history on survival in NHL has not been examined.

METHODS

We used the Swedish Family-Cancer Database to estimate hazard ratios in NHL by histology and family history.

RESULTS

Using diffuse centroblastic lymphoma as reference (HR 1.0), patients with Waldenström's macroglobulinemia and hairy-cell leukemia had the best survival. Survival advantage was also noted among patients with lymphoplasmacytic lymphoma and different kinds of follicular lymphomas. For T-cell lymphoma, mycosis fungoides showed a favorable prognosis. As for survival by family history, a total of 98 familial cases were noted in our Database with a similar prognosis compared to sporadic cases in both parental and offspring generations. A non-significant familial concordance of either good or poor survival was noted among family members when probands' prognosis was stratified by survival time.

CONCLUSIONS

Our results provide quantitative prognosis data for patients with NHL according to specific histologies. Patients with a familial NHL had a similar prognosis compared to patients with sporadic disease. The data suggest familial concordance in either good or poor survival among family members.

Role and prognostic significance of the Ki-67 index in non-Hodgkin's lymphoma

Expression of Ki-67, a nuclear antigen protein present in all cycling cells, is used to determine the growth fraction of tumors. The aim of this study was to evaluate the role and prognostic significance of the Ki-67 proliferation index (PI) in non-Hodgkin's lymphoma. Ki-67 was assayed immunohistochemically in tissue samples of 319 patients with newly-diagnosed non-Hodgkin's lymphoma. In 268 patients, the Ki-67 PI was correlated with clinical course and outcome. The mean Ki-67 PI ranged from 26.6% in indolent lymphomas to 97.6% in very aggressive lymphomas (P < 0.001). The index was <45% in 82.8% of indolent lymphomas and >45% in 85% of aggressive lymphomas (AUC = 0.877, P < 0.001). In patients with diffuse large B-cell lymphoma (n = 141), a Ki-67 PI of 70% was found to significantly discriminate patients with good or bad prognosis (AUC = 0.65, P = 0.004). Three-year survival was 75% +/- 5.6% in patients with a low Ki-67 index compared with 55.9% +/- 6% in patients with a high index (P = 0.015). In patients with a low IPI (<or=2), 3-year survival was 94% +/- 4% in those with a Ki-67 index <or=70% and 64% +/- 8.1% in those with a higher index (P = 0.002); in patients with bulky disease (>10 cm), the corresponding 3-year survival by Ki-67 index was 100% and 25% +/- 12% (P = 0.012). Our results suggest that the mean Ki-67 PI differs by type of lymphoma. A cut-off value of 45% can help differentiate indolent from aggressive disease. In diffuse large B-cell lymphoma, a cut-off value of 70% can distinguish patients with a good and bad prognosis when combined with other prognostic factors of low IPI score and bulky disease.

Constitutive overexpression of a novel 21 kDa protein by Hodgkin lymphoma and aggressive non-Hodgkin lymphomas

Background

CD30, a 120 kDa surface phosphorylated protein is a member of tumour necrosis/nerve growth factor receptor (TNF/NGFR) family and constitutively expressed by Hodgkin and Reed-Sternberg (HRS) cells of Hodgkin lymphoma (HL) and the neoplastic cells of Anaplastic Large Cell Lymphoma (ALCL). A disease-specific protein marker is yet to be identified in Hodgkin lymphoma cells. In order to define HL-specific biomarkers, novel murine monoclonal antibodies were developed in our laboratory.

Results

Murine monoclonal antibodies (mabs) were raised against the B3 sub clone of HL-derived cell line KM-H2. Two of these mabs (clone R23.1 mab and clone R24.1 mab) are IgG₁ class antibodies that recognize a 21 kDa protein present at the cell membrane and in the cytoplasm in HL-derived cell lines. Clone R24.1 mab recognizes a formalin-resistant epitope and labels HRS cells in tissue samples from patients with HL of the classical type, ALCL, and subsets of T and B cell aggressive Non-Hodgkin Lymphomas (NHL). The antigen recognized by the clone R23.1 mab and clone R24.1 mab does not share epitopes with CD30 cluster regions A, B, or C, and, unlike CD30, is not expressed by phytohemagglutinin (PHA) activated T cells.

Conclusion

The 21 kDa protein detected by clone R23.1 and clone R24.1 mabs is a novel membrane-associated protein that may be a potential marker for the diagnosis and targeted therapy of HL and aggressive T and B cell NHL.