BMC Gastroenterol. 2019 Jun 27;19(1):109. doi: 10.1186/s12876-019-1022-0.

Interactive decision support for esophageal adenocarcinoma screening and surveillance.

Vaughan TL, Onstad L, Dai JY.

Abstract

BACKGROUND:

A key barrier to controlling esophageal adenocarcinoma (EAC) is identifying those most likely to benefit from screening and surveillance. We aimed to develop an online educational tool, termed  IC-RISC™, for providers and patients to estimate more precisely their  absolute  risk of developing EAC, interpret this estimate in the context  of risk  of dying from other causes, and aid in decision-making.

RESULTS:

U.S.  incidence and mortality data and published relative risk estimates from  observational studies and clinical trials were used to calculate  absolute risk of EAC over 10 years adjusting for competing risks. These  input parameters varied depending on presence of the key precursor, Barrett’s esophagus. The open source application works across common devices to gather risk factor data and graphically illustrate estimated risk on a single page. Changes to input data are immediately reflected  in the colored graphs. We used the calculator to compare the risk  distribution between EAC cases and controls from six population-based  studies to gain insight into the discrimination metrics of current  practice guidelines for screening, observing that current guidelines  sacrifice a significant amount of specificity to identify 78-86% of  eventual cases in the US population.

Figure 1.

Figure 1. Risk calculator tab. In example a) A 60-year-old white male with moderate physical activity, non-use of NSAIDs or statins, no family history of BE or EAC, weekly – daily reflux symptoms, a body mass index (BMI) of 28 (“overweight”), who never   smoked cigarettes, and has not been screened (BE status unknown) is   estimated to have a 10-year risk of developing EAC of 5.7 per 1000, or 1   in 175 people of similar characteristics. This is higher than the 10-year risk of dying from colon cancer and stroke for a 60-year-old  white male, but lower than that from injury or heart disease. In example b) this same individual has undergone an upper endoscopy and found to have a visible Barrett’s  segment length of 5 cm, but no evidence of dysplasia. His 10-year risk  is now estimated to be  34.0 per 1000 (1 in 29 people) which is approximately equal to his risk  of dying from heart disease.

Figure 2.

Fig. 2.  Metrics describing estimated 10-year risk in EAC cases and controls based on data from six population-based studies in the BEACON consortium. The  bottom panel  shows the distribution of 10-year risk estimates by case status. The  solid vertical lines represent examples of an individual for whom  current ACG guidelines suggest that screening endoscopy be  considered. The first (10-year risk=0.66/1000) is for a 40-year-old  male with  weekly-daily reflux and two additional risk factors (white and   BMI>25). The second (10-year risk=0.97/1000) is for a 50-year-old male with weekly-daily reflux and with two different additional risk factors (age?>?50 and positive cigarette smoking history). The top, second and third panels show how specificity, sensitivity and positive predictive value (ppv), respectively, vary according to possible thresholds for further action.

CONCLUSIONS:

This educational tool provides a simple and rapid means to graphically communicate risk of EAC in the context of other health risks, facilitates “what-if” scenarios regarding potential preventative actions, and can inform discussions regarding screening, surveillance  and treatment options. Its generic architecture lends itself to being  easily extended to other cancers with distinct pathways and/or  intermediate stages, such as hepatocellular cancer. IC-RISC™ extends  current qualitative clinical practice guidelines into a quantitative assessment, which brings the possibility of preventative actions being  offered to persons not currently targeted for screening and, conversely, reducing unnecessary procedures in those at low risk. Prospective validation and application to existing well-characterized cohort studies are needed.