PREDICTING THE NEED FOR INTERVENTION IN PATIENTS PRESENTING WITH ASYMPTOMATIC GALLSTONES: CONSTRUCTION AND VALIDATION OF A RISK STRATIFICATION TOOL
Gareth Morris-Stiff*, Shashank Sarvepalli, Ari Garber, Carol A. Burke, Niyati M. Gupta, Pooja Lal, Mounir Ibrahim, John McMichael, Michael Rothberg, Amit Bhatt, John J. Vargo, Maged Rizk
Cleveland Clinic Foundation, Cleveland, OH
Introduction
The optimal management of asymptomatic gallstones (AG) remains controversial. Although previous studies have evaluated the time to development of gallstone-related complications following identification of AG, factors associated with the need for surgical intervention in this population have not been documented. Our aim was to perform big data analysis to evaluate risk factors associated with intervention in AG, and to develop a risk stratification tool (RST) to aid in patient consultations.
Methods
A validated natural language algorithm was used to screen radiological reports to identify patients with a diagnosis of AG within the institutional data warehouse. For this population, each individual electronic medical record (EMR) was screened using a second algorithm to identify if patients subsequently underwent intervention in the form of cholecystectomy (C), endoscopic retrograde cholangiopancreatography (ERCP), or percutaneous cholecystostomy (PC). The indication for each intervention be it for chronic or acute cholecystitis, acute pancreatitis, cholangitis, or gallbladder cancer was recorded. These manifestations were collectively termed complex gallstone disease. Kaplan Meier curves were constructed to analyze time to intervention, and both cumulative incidence ratios and hazard estimates were calculated. Forward stepwise cox-regression identified factors associated with future intervention and were used to develop a RST. The area under the receiver operating characteristics (AUROC) curves were calculated and internally validated using 1000 boot-strapped resamples.
Results
During the period 1996-2016, 22,257 patients were identified with AG. The mean patient age was 61 ± 0.11 years, and there was a slight female predominance of 51.3% female. The median follow-up was 4.5 years (inter-quartile range [IQR]: 1.8-7.9 years) and a total of 112,112 years of observation. 1762 (7.9%) underwent intervention [C (n=1273); PC (n=66); ERCP (n=592)] within a median of 3.9 years (IQR 1.4-6.9 years) of initial presentation. The cumulative incidence of intervention at 15 years was 25% and increased in a linear manner from time of initial diagnosis of AG. Factors identified on regression analysis as predicating the development of symptoms following initial presentation with AG are summarized in Table 1. The AUROC of the RST was 0.66 (95%CI, 0.64-0.67) for the initial sample and a 1000 boot-strapped resample yielded an AUROC of 0.66 (95%CI, 0.64-0.67) indicating good internal validity.
Conclusions
The need for intervention due to development of symptomatic disease is common in patients initially diagnosed with AG, increasing over time from initial presentation. The use of an application-based RST can reliably identify these patients and can provide invaluable prognostic information for counseling.
Table 1.
| Variable | HR (95% CI) | p value |
| Age (increment of 5 years) | 0.94 (0.93, 0.96) | <0.001 |
| Male gender | 0.78 (0.71, 0.86) | <0.001 |
| Ethnicity (vs. White) | ||
| Black | 1 (0.87, 1.13) | 0.953 |
| Hispanic | 1.65 (1.17, 2.32) | 0.004 |
| Asian | 1.06 (0.66, 1.72) | 0.807 |
| Comorbidities | ||
| Tobacco | 1.16 (1.05, 1.27) | 0.003 |
| Obesity | 1.44 (1.31, 1.58) | <0.001 |
| Hemolytic disorder | 2.42 (1.57, 3.74) | <0.001 |
| Hyperlipidemia | 1.19 (1.07, 1.32) | 0.001 |
| Statin use | 0.67 (0.61, 0.75) | <0.001 |
| Gallstone characteristics | ||
| Large stone (>9mm) | 1.56 (1.33, 1.82) | <0.001 |
| Sludge | 1.46 (1.24, 1.72) | <0.001 |
| Polyp | 1.68 (1.31, 2.15) | <0.001 |
| Multiple stones | 1.69 (1.53, 1.86) | <0.001 |
| Thickened gallbladder wall | 2.1 (1.28, 3.44) | 0.003 |
Back to 2019 Abstracts
