Cost-effectiveness of artificial intelligence for screening colonoscopy: a modelling study.

BACKGROUND: Artificial intelligence (AI) tools increase detection of precancerous polyps during colonoscopy and might contribute to long-term colorectal cancer prevention. The aim of the study was to investigate the incremental effect of the implementation of AI detection tools in screening colonoscopy on colorectal cancer incidence and mortality, and the cost-effectiveness of such tools. METHODS: We conducted Markov model microsimulation of using colonoscopy with and without AI for colorectal cancer screening for individuals at average risk (no personal or family history of colorectal cancer, adenomas, inflammatory bowel disease, or hereditary colorectal cancer syndrome). We ran the microsimulation in a hypothetical cohort of 100 000 individuals in the USA aged 50-100 years. The primary analysis investigated screening colonoscopy with versus without AI every 10 years starting at age 50 years and finishing at age 80 years, with follow-up until age 100 years, assuming 60% screening population uptake. In secondary analyses, we modelled once-in-life screening colonoscopy at age 65 years in adults aged 50-79 years at average risk for colorectal cancer. Post-polypectomy surveillance followed the simplified current guideline. Costs of AI tools and cost for downstream treatment of screening detected disease were estimated with 3% annual discount rates. The main outcome measures included the incremental effect of AI-assisted colonoscopy versus standard (no-AI) colonoscopy on colorectal cancer incidence and mortality, and cost-effectiveness of screening projected for the average risk screening US population. FINDINGS: In the primary analyses, compared with no screening, the relative reduction of colorectal cancer incidence with screening colonoscopy without AI tools was 44·2% and with screening colonoscopy with AI tools was 48·9% (4·8% incremental gain). Compared with no screening, the relative reduction in colorectal cancer mortality with screening colonoscopy with no AI was 48·7% and with screening colonoscopy with AI was 52·3% (3·6% incremental gain). AI detection tools decreased the discounted costs per screened individual from $3400 to $3343 (a saving of $57 per individual). Results were similar in the secondary analyses modelling once-in-life colonoscopy. At the US population level, the implementation of AI detection during screening colonoscopy resulted in yearly additional prevention of 7194 colorectal cancer cases and 2089 related deaths, and a yearly saving of US$290 million. INTERPRETATION: Our findings suggest that implementation of AI detection tools in screening colonoscopy is a cost-saving strategy to further prevent colorectal cancer incidence and mortality. FUNDING: European Commission and Japan Society of Promotion of Science.

Endoscopists' diagnostic accuracy in detecting upper gastrointestinal neoplasia in the framework of artificial intelligence studies.

BACKGROUND: Estimates on miss rates for upper gastrointestinal neoplasia (UGIN) rely on registry data or old studies. Quality assurance programs for upper GI endoscopy are not fully established owing to the lack of infrastructure to measure endoscopists' competence. We aimed to assess endoscopists' accuracy for the recognition of UGIN exploiting the framework of artificial intelligence (AI) validation studies. METHODS: Literature searches of databases (PubMed/MEDLINE, EMBASE, Scopus) up to August 2020 were performed to identify articles evaluating the accuracy of individual endoscopists for the recognition of UGIN within studies validating AI against a histologically verified expert-annotated ground-truth. The main outcomes were endoscopists' pooled sensitivity, specificity, positive and negative predictive value (PPV/NPV), and area under the curve (AUC) for all UGIN, for esophageal squamous cell neoplasia (ESCN), Barrett esophagus-related neoplasia (BERN), and gastric adenocarcinoma (GAC). RESULTS: Seven studies (2 ESCN, 3 BERN, 1 GAC, 1 UGIN overall) with 122 endoscopists were included. The pooled endoscopists' sensitivity and specificity for UGIN were 82 % (95 % confidence interval [CI] 80 %-84 %) and 79 % (95 %CI 76 %-81 %), respectively. Endoscopists' accuracy was higher for GAC detection (AUC 0.95 [95 %CI 0.93-0.98]) than for ESCN (AUC 0.90 [95 %CI 0.88-0.92]) and BERN detection (AUC 0.86 [95 %CI 0.84-0.88]). Sensitivity was higher for Eastern vs. Western endoscopists (87 % [95 %CI 84 %-89 %] vs. 75 % [95 %CI 72 %-78 %]), and for expert vs. non-expert endoscopists (85 % [95 %CI 83 %-87 %] vs. 71 % [95 %CI 67 %-75 %]). CONCLUSION: We show suboptimal accuracy of endoscopists for the recognition of UGIN even within a framework that included a higher prevalence and disease awareness. Future AI validation studies represent a framework to assess endoscopist competence.

Standalone performance of artificial intelligence for upper GI neoplasia: a meta-analysis.

OBJECTIVE: Artificial intelligence (AI) may reduce underdiagnosed or overlooked upper GI (UGI) neoplastic and preneoplastic conditions, due to subtle appearance and low disease prevalence. Only disease-specific AI performances have been reported, generating uncertainty on its clinical value. DESIGN: We searched PubMed, Embase and Scopus until July 2020, for studies on the diagnostic performance of AI in detection and characterisation of UGI lesions. Primary outcomes were pooled diagnostic accuracy, sensitivity and specificity of AI. Secondary outcomes were pooled positive (PPV) and negative (NPV) predictive values. We calculated pooled proportion rates (%), designed summary receiving operating characteristic curves with respective area under the curves (AUCs) and performed metaregression and sensitivity analysis. RESULTS: Overall, 19 studies on detection of oesophageal squamous cell neoplasia (ESCN) or Barrett's esophagus-related neoplasia (BERN) or gastric adenocarcinoma (GCA) were included with 218, 445, 453 patients and 7976, 2340, 13 562 images, respectively. AI-sensitivity/specificity/PPV/NPV/positive likelihood ratio/negative likelihood ratio for UGI neoplasia detection were 90% (CI 85% to 94%)/89% (CI 85% to 92%)/87% (CI 83% to 91%)/91% (CI 87% to 94%)/8.2 (CI 5.7 to 11.7)/0.111 (CI 0.071 to 0.175), respectively, with an overall AUC of 0.95 (CI 0.93 to 0.97). No difference in AI performance across ESCN, BERN and GCA was found, AUC being 0.94 (CI 0.52 to 0.99), 0.96 (CI 0.95 to 0.98), 0.93 (CI 0.83 to 0.99), respectively. Overall, study quality was low, with high risk of selection bias. No significant publication bias was found. CONCLUSION: We found a high overall AI accuracy for the diagnosis of any neoplastic lesion of the UGI tract that was independent of the underlying condition. This may be expected to substantially reduce the miss rate of precancerous lesions and early cancer when implemented in clinical practice.

Association of Polyps with Early-Onset Colorectal Cancer and Throughout Surveillance: Novel Clinical and Molecular Implications.

Early-onset colorectal cancer (EOCRC) is an increasing and worrisome entity. The aim of this study was to analyze its association with polyps concerning prognosis and surveillance. EOCRC cases were compared regarding the presence or absence of associated polyps (clinical and molecular features), during a minimum of 7 years of follow-up. Of 119 cases, 56 (47%) did not develop polyps (NP group), while 63 (53%) did (P group). The NP group showed a predominant location of the CRC in the rectum (50%), of sporadic cases (54%), and diagnosis at advanced stages: Only P53 and SMARCB1 mutations were statistically linked to this group. The P group, including mainly early-diagnosed tumors, was linked with the most frequent and differential altered chromosomal regions in the array comparative genomic hybridization. The two most frequent groups according to the follow-up were the NP group (40%), and patients developing polyps in the first 5 years of follow-up (P < 5FU) (34%) (these last groups predominantly diagnosed at the earliest stage and with adenomatous polyps (45%)). EOCRC with polyps that developed during the entire follow-up (PDFU group) were mainly located in the right colon (53%), diagnosed in earlier stages, and 75% had a familial history of CRC. Patients developing polyps after the first 5 years (P > 5FU) showed a mucinous component (50%). Our results show that the absence or presence of polyps in EOCRC is an important prognostic factor with differential phenotypes. The development of polyps during surveillance shows that it is necessary to extend the follow-up time, also in those cases with microsatellite-stable EOCRC.

Clinical and Molecular Comparative Study of Colorectal Cancer Based on Age-of-onset and Tumor Location: Two Main Criteria for Subclassifying Colorectal Cancer.

Our aim was to characterize and validate that the location and age of onset of the tumor are both important criteria to classify colorectal cancer (CRC). We analyzed clinical and molecular characteristics of early-onset CRC (EOCRC) and late-onset CRC (LOCRC), and we compared each tumor location between both ages-of-onset. In right-sided colon tumors, early-onset cases showed extensive Lynch syndrome (LS) features, with a relatively low frequency of chromosomal instability (CIN), but a high CpG island methylation phenotype. Nevertheless, late-onset cases showed predominantly sporadic features and microsatellite instability cases due to BRAF mutations. In left colon cancers, the most reliable clinical features were the tendency to develop polyps as well as multiple primary CRC associated with the late-onset subset. Apart from the higher degree of CIN in left-sided early-onset cancers, differential copy number alterations were also observed. Differences among rectal cancers showed that early-onset rectal cancers were diagnosed at later stages, had less association with polyps, and more than half of them were associated with a familial LS component. Stratifying CRC according to both location and age-of-onset criteria is meaningful, not only because it correlates the resulting categories with certain molecular bases, but with the confirmation across larger studies, new therapeutical algorithms could be defined according to this subclassification.

Clinical Outcomes of Golimumab as First, Second or Third Anti-TNF Agent in Patients with Moderate-to-Severe Ulcerative Colitis.

BACKGROUND: Golimumab efficacy data in ulcerative colitis (UC) are limited to anti-tumor necrosis factor α (TNF)-naive patients. The aim of this study was to assess the short-term and long-term efficacy of golimumab used as first, second, or third anti-TNF in UC in a real-life clinical setting. METHODS: This retrospective multicenter cohort study included patients with moderate-to-severe UC treated with golimumab. The primary efficacy endpoints were short-term partial Mayo score response, long-term golimumab failure-free survival, and colectomy-free survival. RESULTS: In 142 patients with UC, golimumab was administered as first (40%), second (23%), or third anti-TNF (37%). Ninety-two patients (65%, 95% confidence interval 56.6-73) achieved short-term clinical response. Forty-five patients (32%, 95% confidence interval 23.7-39.7) achieved clinical remission. Response rates for golimumab were 75% as first anti-TNF, 70% as second anti-TNF (ns versus first anti-TNF), and 50% as third anti-TNF (P = 0.007 versus first anti-TNF). After 12 months median follow-up (interquartile range 6-18), 60 patients (42%, 95% confidence interval 34-51) had golimumab failure, and 15 patients (11%) needed colectomy. Thirty-one patients (22%) needed golimumab dose escalation, and 71% of these regained response after escalation. Starting maintenance with 100 mg golimumab doses and short-term nonresponse were independent predictors of golimumab failure. CONCLUSIONS: In this real-life cohort of patients with UC, golimumab therapy was effective for inducing and maintaining clinical response. Although anti-TNF-naive patients had better outcomes, golimumab was also effective in anti-TNF-experienced patients. Only the patients given golimumab after previous failure of 2 anti-TNF agents had significantly worse outcomes. Golimumab dose escalation was beneficial and safe.