Introduction Percutaneous endoscopic gastrostomy (PEG) insertion is a well-established technique for providing long-term enteral nutrition. However concerns have been raised regarding the high 30-day mortality associated with PEG, and the related ethical implications of patient selection. Accordingly, a previous predictive tool was developed using age and serum albumin level but was created on a relatively small cohort. External validation of the score was performed in the same region but has not been outside of this area. This study aimed to externally validate this previous scoring system and also try to identify any further predictors of 30-day mortality in a larger cohort.

Methods Retrospective review of all gastroscopy reports documenting PEG insertions between January 2001 and January 2012 in our centre was undertaken. Hospital electronic systems were used to determine patient demographics, laboratory results and outcome at 30 days. In patients with newly inserted PEG tubes, the scoring system was applied and assessed using receiver operating curve analysis to determine the discriminative capacity. Furthermore, univariate and multivariate binary logistic regression analyses were performed using the current database to identify additional predictors of 30-day mortality.

Results The PEG database included 1373 patients, of which 808 were new PEG insertions and suitable for analysis. For each increasing gradation of the scoring system, mortality rose with 4% of those scoring zero dying compared to 50% scoring three. An area under the ROC curve of 0.686 (95% confidence interval 0.635–0.737) indicated reasonable discriminative capacity. Multivariate analysis demonstrated that age ≥60 years (OR = 2.097 [p = 0.016]), serum albumin levels of 25–34 g/l (OR = 2.447 [p = 0.001]) or < 25 g/l (OR = 6.769 [p < 0.001]), C-Reactive Protein ≥10 mg/l (OR = 2.713 [p = 0.009]) and lymphocyte count of <1.5 × 10⁹/l (OR = 2.016 [p = 0.004]) increased the odds of 30-day mortality, whilst inpatient PEG placement decreased the risk of death (OR = 0.529 [p = 0.005]).

Conclusion The previous scoring system demonstrated reasonable predictive proficiency but the area under the ROC curves were not >0.8. The recognition of further predictors of 30-day mortality allows for remodelling of the score which may improve the accuracy. However, future prospective, multicentre studies with defined outcomes are necessary to improve data collection. Additionaly, more information is needed about cause of 30-day mortality and importantly quality of life following PEG insertion.

Disclosure of Interest None Declared.