Archives of pathology & laboratory medicine最新文献

Context: Automated prostate cancer detection using machine learning technology has led to speculation that pathologists will soon be replaced by algorithms. This review covers the development of machine learning algorithms and their reported effectiveness specific to prostate cancer detection and Gleason grading.

Objective: To examine current algorithms regarding their accuracy and classification abilities. We provide a general explanation of the technology and how it is being used in clinical practice. The challenges to the application of machine learning algorithms in clinical practice are also discussed.

Data sources: The literature for this review was identified and collected using a systematic search. Criteria were established prior to the sorting process to effectively direct the selection of studies. A 4-point system was implemented to rank the papers according to their relevancy. For papers accepted as relevant to our metrics, all cited and citing studies were also reviewed. Studies were then categorized based on whether they implemented binary or multi-class classification methods. Data were extracted from papers that contained accuracy, area under the curve (AUC), or κ values in the context of prostate cancer detection. The results were visually summarized to present accuracy trends between classification abilities.

Conclusions: It is more difficult to achieve high accuracy metrics for multiclassification tasks than for binary tasks. The clinical implementation of an algorithm that can assign a Gleason grade to clinical whole slide images (WSIs) remains elusive. Machine learning technology is currently not able to replace pathologists but can serve as an important safeguard against misdiagnosis.

Context: Antinuclear antibodies (ANAs) against certain antigens are useful for identifying autoimmune disorders. Although new solid phase-based immunoassays have been developed for evaluating ANAs, the conventional line immunoassay (LIA) is commonly used in clinical practice.

Objective: To compare the clinical performance of 2 newly developed methods for detecting specific ANAs with LIA.

Design: Six hundred ninety-six serum samples were collected from 559 patients with autoimmune disease (AID) and 137 controls. The samples were screened by using the LIA, digital liquid chip method (DLCM), and chemiluminescent immunoassay (CLIA) for specific ANAs. The agreement across assays and the clinical performance of each assay in diagnosing ANA-associated rheumatic diseases (AARDs) were evaluated.

Results: Almost perfect agreement was observed among all assays for anti-centromere protein B (κ = 0.85-0.97), anti-ribosome P (κ = 0.85-0.88), anti-SSA 52 (κ = 0.86-0.89), and anti-SSA 60 (κ = 0.89-0.91); moderate to substantial agreement was detected for the autoantibodies against Sm, Jo-1, ribonucleoprotein, Scl-70, and SSB (κ = 0.55-0.80). LIA exhibited better sensitivity for diagnosing AARDs, while DLCM and CLIA demonstrated higher specificity. In the subset of AIDs, especially systemic lupus erythematosus, antibody positive percentages varied greatly between assays.

Conclusions: The 3 assays showed comparable qualitative agreement; however, the standardization of testing for ANAs remains challenging owing to intermanufacturer variations. Moreover, DLCM and CLIA exhibited better specificity in distinguishing non-AID individuals, whereas LIA was more sensitive in diagnosing AARDs.

Context: In 2021 the World Health Organization distributed a new classification of central nervous system tumors that incorporated modern testing modalities in the diagnosis. Although universally accepted as a scientifically superior system, this schema has created controversy because its deployment globally is challenging in the best of circumstances and impossible in resource-poor health care ecosystems. Compounding this problem is the significant challenge that neuropathologists with expertise in central nervous system tumors are rare.

Objective: To demonstrate diagnostic use of simple unsupervised machine learning techniques using publicly available data sets. I also discuss some potential solutions to the deployment of neuropathology classification in health care ecosystems burdened by this classification schema.

Data sources: The Cancer Genome Atlas RNA sequencing data from low-grade and high-grade gliomas.

Conclusions: Methylation-based classification will be unable to solve all diagnostic problems in neuropathology. Information theory quantifications generate focused workflows in pathology, resulting in prevention of ordering unnecessary tests and identifying biomarkers that facilitate diagnosis.

Context: Loose tumor cells and tumor cell clusters can be recognized in the lumen of intratumoral pulmonary arteries of resected non-small cell lung cancer specimens. It is unclear whether these should be considered tumor-emboli, and as such could predict a worsened prognosis.

Objective: To investigate the nature and prognostic impact of pulmonary artery intraluminal tumor cells.

Design: This multicenter study involved an exploratory pilot study and a validation study from 3 institutions. For the exploratory pilot study, a retrospective pulmonary resection cohort of primary adenocarcinomas, diagnosed between November 2007 and November 2010, were scored for the presence of tumor cells, as well as potentially other cells in the intravascular spaces, using hematoxylin-eosin and cytokeratin 7 (CK7) stains. In the validation part, 2 retrospective cohorts of resected pulmonary adenocarcinomas, between January 2011 and December 2016, were included. Recurrence-free survival (RFS) and overall survival (OS) data were collected.

Results: In the pilot study, CK7+ intravascular cells, mainly tumor cells, were present in 23 of 33 patients (69.7%). The 5-year OS for patients with intravascular tumor cells was 61%, compared with 40% for patients without intravascular tumor cells (P = .19). In the validation study, CK7+ intravascular tumor cells were present in 41 of 70 patients (58.6%). The 5-year RFS for patients with intravascular tumor cells was 80.0%, compared with 80.6% in patients without intravascular tumor cells (P = .52). The 5-year OS rates were, respectively, 82.8% and 71.6% (P = .16).

Conclusions: Loose tumor cells in pulmonary arterial lumina were found in most non-small cell lung cancer resection specimens and were not associated with a worse RFS or OS. Therefore, most probably they represent an artifact.

Context: Recent molecular discoveries have led to improved understanding of tumor biology and the development of new diagnostic assays.

Objective: To review primarily 3 examples of liver tumors and to briefly illustrate how recent molecular discoveries have altered clinical liver pathology practice.

Data sources: First, we will discuss fibrolamellar carcinoma, which will be the main focus of discussion, as an example for new diagnostic tests that have been developed as a result of molecular discoveries. Additional information on the role of molecular diagnostics in hepatocellular adenoma and hepatocellular carcinoma will be provided. Second, we will use the example of epithelioid hemangioendothelioma as an example of how new diagnostic tools, based on molecular discoveries, may support improved prognostication. Finally, we will use the example of intrahepatic cholangiocarcinoma as an example of a liver tumor where new molecular discoveries have identified tractable therapeutic targets and led to new effective therapies. This portion of the manuscript will also include a description of the anatomic and molecular differences between intrahepatic, hilar, and extrahepatic cholangiocarcinoma.

Conclusions: Fueled by molecular discoveries, new and better diagnostic tests and therapeutic targets have improved clinical care in patients affected by liver tumors.

Context: PathElective.com was created in response to the pandemic's restrictions on interactions with trainees, and since has been incorporated into many training programs worldwide, serving as a unique means of delivering high-quality pathology and laboratory medical education at multiple levels of training.

Objective: To analyze student usage, performance, and satisfaction to provide insight into the effectiveness of virtual education to guide curricular evolution.

Design: Squarespace (Squarespace, Inc) was used for website development and to collect website analytics. Students were assessed before and after course participation using a dual-form crossover quiz design. Quiz data were anonymous and analyzed with a paired t test to account for varying student backgrounds. A novel analysis was performed aimed at examining the attrition rate of students across multiple modules.

Results: During the study period (May 1, 2020 to October 31, 2021), PathElective.com received 577 483 page views, 126 180 visits, 59 928 unique visitors, and 10 278 registered users who earned 15 305 certificates. A total of 7338 premodule and postmodule quiz pairs were analyzed. The overall average increase in score was 13.83% (P = .02). All but 5 of the 56 courses experienced a statistically significant increase in score. All courses received median scores of Very Satisfied/Satisfied in all 6 assessment domains. Aggregate attrition data revealed a unique, negative polynomial relationship (R2 = 0.656).

Conclusions: PathElective.com is a free, effective means of enhancing anatomic/clinical pathology training in medical education. These analyses offer a unique perspective on the online user experience and could guide the development of future online medical education resources.

Context: To provide high-quality, safe training during the COVID-19 pandemic, our anatomic pathology fellowship program implemented a hybrid virtual/in-person training model with supplemental digital material.

Objective: To evaluate the impact of this model.

Design: We examined Accreditation Council for Graduate Medical Education survey results and board pass rates for fellows before the pandemic (group 1); during the pandemic peak (group 2); and early and late after the pandemic peak (groups 3 and 4). Additionally, we distributed an online survey, including questions related to performance as attending physicians and fellowship experience, to recent graduates.

Results: Information loss during handover, supervision and teaching by faculty, and having at least 4 free days a month exhibited the greatest score declines between group 1 and groups 2, 3, and 4 on the Accreditation Council for Graduate Medical Education surveys. No differences were seen in board passing rates between groups. The groups did not differ in responses regarding preparation for role as attending, confidence in role as attending, or overall impression of the fellowship program. The pandemic-affected groups responded more positively on the perceived utility of supplemental digital material, impact of digital pathology on quality of education, and impact of supplemental digital material on familiarity with digital pathology. The difference was particularly large between group 1 and combined groups 3 and 4.

Conclusions: Despite the limitations noted, the hybrid training model was effective and successfully prepared fellows for their role as attending physicians. Similar studies can be informative for the implementation of similar programs or for the meaningful integration of digital pathology into training curricula.

Context: Recent data suggest mesenteric tumor deposits (MTDs) indicate poor prognosis in small bowel well-differentiated neuroendocrine tumors (SB-NETs), including compared to positive lymph nodes, making their distinction crucial.

Objective: To study interobserver agreement in distinguishing SB-NET MTDs from positive nodes.

Design: Virtual slides from 36 locally metastatic SB-NET foci were shared among 7 gastrointestinal pathologists, who interpreted each as an MTD or a positive node. Observers ranked their 5 preferred choices among a supplied list of potentially useful histologic features, for both options. Diagnostic opinions were compared using Fleiss multirater and Cohen weighted κ analyses.

Results: Preferred criteria for MTD included irregular shape (n = 7, top choice for 5), perineural invasion/nerve entrapment (n = 7, top choice for 2), encased thick-walled vessels (n = 7), and prominent fibrosis (n = 6). Preferred criteria for positive nodes included peripheral lymphoid follicles (n = 6, top choice for 4), round shape (n = 7, top choice for 2), peripheral lymphocyte rim (n = 7, top choice for 1), subcapsular sinuses (n = 7), and a capsule (n = 6). Among 36 foci, 10 (28%) each were unanimously diagnosed as MTD or positive node. For 13 foci (36%), there was a diagnosis favored by most observers (5 or 6 of 7): positive node in 8, MTD in 5. Only 3 cases (8%) had a near-even (4:3) split. Overall agreement was substantial (κ = .64, P < .001).

Conclusions: Substantial interobserver agreement exists for distinguishing SB-NET MTDs from lymph node metastases. Favored histologic criteria in making the distinction include irregular shape and nerve/vessel entrapment for MTD, and peripheral lymphocytes/lymphoid follicles and round shape for positive nodes.

Context: The pathologic diagnosis of pT1 substage in conventional transurethral resection of bladder tumor specimens is inaccurate and has low interobserver reproducibility owing to fragmentation and cauterization of the specimens. En bloc resection of bladder tumor is a novel surgical procedure that improves diagnostic feasibility and accuracy in the pathologic diagnosis of bladder cancer, including depth and extent of invasion.

Objective: To examine the prognostic value of multiple pT1 subclassification methods, using only en bloc resection specimens.

Design: We examined 106 patients with T1 bladder cancer who underwent en bloc resection. The pT1 substages were assigned by 3 different subclassification methods by using the muscularis mucosae or stalk of the papillary lesion as diagnostic landmarks or millimetric depth of invasion. Intergroup differences in progression-free survival and recurrence-free survival rates were analyzed. The prognostic values of clinicopathologic factors for progression/recurrence were analyzed by using multivariate analysis.

Results: The pT1 substage was evaluable in all cases. Tumors with invasion into/beyond the muscularis mucosae and those beyond the stalk of the papillary lesion were associated with worse progression-free survival (P = .002 and P = .01, respectively). Notably, no patient with invasion confined to the stalk had disease progression during the 23-month median follow-up period. Only the pT1 subclassification method using the muscularis mucosae was an independent prognosticator of progression in multivariate analysis (P = .03).

Conclusions: Precise pathologic subclassification of invasion using en bloc resection specimens may enable accurate prognosis and assessment in patients with bladder cancer with suspicious shallow invasion.

Context.—: Two major categories of laboratories performing nonwaived testing under the Clinical Laboratory Improvement Amendments of 1988 (CLIA) are the Certificate of Accreditation (CoA) and Certificate of Compliance (CoC) laboratories. Accreditation organizations collect more detailed laboratory personnel information than the Centers for Medicare & Medicaid Services (CMS) Quality Improvement and Evaluation System (QIES).

Objective.—: To estimate total numbers of testing personnel and testing volumes in CoA and CoC laboratories, by laboratory type and state.

Design.—: We developed a statistical inference method by using the respective correlations between testing personnel counts and test volume by laboratory type.

Results.—: QIES reported 33 033 active CoA and CoC laboratories in July 2021. We estimated testing personnel to be 328 000 (95% CI, 309 000-348 000), which is supported by the count of 318 780 reported by the US Bureau of Labor Statistics. There were twice as many testing personnel in hospital laboratories as in independent laboratories (158 778 versus 74 904, P < .001). Independent laboratories had the highest test volume per person, which was twice as high as physician office laboratories (62 228 versus 30 102, P < .001). Hospital and independent laboratories comprised 34% of all CoA and CoC laboratories but performed the largest portion of testing (81%). Physician office laboratories, accounting for 44% of all CoA and CoC laboratories, performed a comparatively low proportion of total tests (9%).

Conclusions.—: Numbers of testing personnel vary considerably by laboratory type and across states. These data can provide valuable insight when assessing laboratory workforce training needs and planning for public health emergencies.