Artificial intelligence in digital pathology: a systematic review and meta-analysis of diagnostic test accuracy.

Ensuring diagnostic performance of artificial intelligence (AI) before introduction into clinical practice is essential. Growing numbers of studies using AI for digital pathology have been reported over recent years. The aim of this work is to examine the diagnostic accuracy of AI in digital pathology images for any disease. This systematic review and meta-analysis included diagnostic accuracy studies using any type of AI applied to whole slide images (WSIs) for any disease. The reference standard was diagnosis by histopathological assessment and/or immunohistochemistry. Searches were conducted in PubMed, EMBASE and CENTRAL in June 2022. Risk of bias and concerns of applicability were assessed using the QUADAS-2 tool. Data extraction was conducted by two investigators and meta-analysis was performed using a bivariate random effects model, with additional subgroup analyses also performed. Of 2976 identified studies, 100 were included in the review and 48 in the meta-analysis. Studies were from a range of countries, including over 152,000 whole slide images (WSIs), representing many diseases. These studies reported a mean sensitivity of 96.3% (CI 94.1-97.7) and mean specificity of 93.3% (CI 90.5-95.4). There was heterogeneity in study design and 99% of studies identified for inclusion had at least one area at high or unclear risk of bias or applicability concerns. Details on selection of cases, division of model development and validation data and raw performance data were frequently ambiguous or missing. AI is reported as having high diagnostic accuracy in the reported areas but requires more rigorous evaluation of its performance.

Associations between AI-Assisted Tumor Amphiregulin and Epiregulin IHC and Outcomes from Anti-EGFR Therapy in the Routine Management of Metastatic Colorectal Cancer.

PURPOSE: High tumor production of the EGFR ligands, amphiregulin (AREG) and epiregulin (EREG), predicted benefit from anti-EGFR therapy for metastatic colorectal cancer (mCRC) in a retrospective analysis of clinical trial data. Here, AREG/EREG IHC was analyzed in a cohort of patients who received anti-EGFR therapy as part of routine care, including key clinical contexts not investigated in the previous analysis. EXPERIMENTAL DESIGN: Patients who received panitumumab or cetuximab ± chemotherapy for treatment of RAS wild-type mCRC at eight UK cancer centers were eligible. Archival formalin-fixed paraffin-embedded tumor tissue was analyzed for AREG and EREG IHC in six regional laboratories using previously developed artificial intelligence technologies. Primary endpoints were progression-free survival (PFS) and overall survival (OS). RESULTS: A total of 494 of 541 patients (91.3%) had adequate tissue for analysis. A total of 45 were excluded after central extended RAS testing, leaving 449 patients in the primary analysis population. After adjustment for additional prognostic factors, high AREG/EREG expression (n = 360; 80.2%) was associated with significantly prolonged PFS [median: 8.5 vs. 4.4 months; HR, 0.73; 95% confidence interval (CI), 0.56-0.95; P = 0.02] and OS [median: 16.4 vs. 8.9 months; HR, 0.66 95% CI, 0.50-0.86; P = 0.002]. The significant OS benefit was maintained among patients with right primary tumor location (PTL), those receiving cetuximab or panitumumab, those with an oxaliplatin- or irinotecan-based chemotherapy backbone, and those with tumor tissue obtained by biopsy or surgical resection. CONCLUSIONS: High tumor AREG/EREG expression was associated with superior survival outcomes from anti-EGFR therapy in mCRC, including in right PTL disease. AREG/EREG IHC assessment could aid therapeutic decisions in routine practice. See related commentary by Randon and Pietrantonio, p. 4021.

Optimized Systematic Review Tool: Application to Candidate Biomarkers for the Diagnosis of Hepatocellular Carcinoma.

This review aims to develop an appropriate review tool for systematically collating metabolites that are dysregulated in disease and applies the method to identify novel diagnostic biomarkers for hepatocellular carcinoma (HCC). Studies that analyzed metabolites in blood or urine samples where HCC was compared with comparison groups (healthy, precirrhotic liver disease, cirrhosis) were eligible. Tumor tissue was included to help differentiate primary and secondary biomarkers. Searches were conducted on Medline and EMBASE. A bespoke "risk of bias" tool for metabolomic studies was developed adjusting for analytic quality. Discriminant metabolites for each sample type were ranked using a weighted score accounting for the direction and extent of change and the risk of bias of the reporting publication. A total of 84 eligible studies were included in the review (54 blood, 9 urine, and 15 tissue), with six studying multiple sample types. High-ranking metabolites, based on their weighted score, comprised energy metabolites, bile acids, acylcarnitines, and lysophosphocholines. This new review tool addresses an unmet need for incorporating quality of study design and analysis to overcome the gaps in standardization of reporting of metabolomic data. Validation studies, standardized study designs, and publications meeting minimal reporting standards are crucial for advancing the field beyond exploratory studies.

Decreasing formalin concentration improves quality of DNA extracted from formalin-fixed paraffin-embedded tissue specimens without compromising tissue morphology or immunohistochemical staining.

Genomic technologies are increasingly used clinically for both diagnosis and guiding cancer therapy. However, formalin fixation can compromise DNA quality. This study aimed to optimise tissue fixation using normal colon, liver and uterus (n=8 each) by varying neutral buffered formalin (NBF) concentration (1%-5% w/v) and fixation time (24-48 hours). Fixation using 4% NBF improved DNA quality (assessed by qPCR) compared with routine (4% unbuffered formal saline-fixed) specimens (p<0.01). Further improvements were achieved by reducing NBF concentration (p<0.00001), whereas fixation time had no effect (p=0.110). No adverse effects were detected by histopathological or QuPath morphometric analysis. Immunohistochemistry for multicytokeratin and α-smooth muscle actin revealed no changes in staining specificity or intensity in any tissue other than on liver multicytokeratin staining intensity, where the effect of fixation time was more significant (p=0.0004) than NBF concentration (p=0.048). Thus, reducing NBF concentration can maximise DNA quality without compromising tissue morphology or standard histopathological analyses.

The utility of biomarkers in hepatocellular carcinoma: review of urine-based 1H-NMR studies - what the clinician needs to know.

Hepatocellular carcinoma (HCC) is the fifth most common malignancy, the third most common cause of cancer death, and the most common primary liver cancer. Overall, there is a need for more reliable biomarkers for HCC, as those currently available lack sensitivity and specificity. For example, the current gold-standard biomarker, serum alpha-fetoprotein, has a sensitivity of roughly only 70%. Cancer cells have different characteristic metabolic signatures in biofluids, compared to healthy cells; therefore, metabolite analysis in blood or urine should lead to the detection of suitable candidates for the detection of HCC. With the advent of metabonomics, this has increased the potential for new biomarker discovery. In this article, we look at approaches used to identify biomarkers of HCC using proton nuclear magnetic resonance (1H-NMR) spectroscopy of urine samples. The various multivariate statistical analysis techniques used are explained, and the process of biomarker identification is discussed, with a view to simplifying the knowledge base for the average clinician.