The document summarizes findings from a study on the appropriateness of upper endoscopy referrals in Italy. Key findings include:
1) 22% of over 13,000 endoscopies were deemed inappropriate based on guidelines. Relevant findings were found in 51% of appropriate referrals and 32% of inappropriate ones.
2) Common relevant findings included esophagitis, gastric erosions, esophageal varices, and ulcers. New cancers were found in 1.6% of patients.
3) Guidelines had high sensitivity but low specificity for relevant findings and cancer. A simple rule using age and alarm symptoms had similar accuracy to guidelines.
4) Over 20% of endoscopy
26. Appropriateness of the indication for upper endoscopy: a meta-analysis Di Giulio E, Hassan C, Zullo A, et al. , DLD 2009 ASGE guidelines Sens. Spec. PPV NPV Relevant finding 85% 28% 49% 70%
28. 2.4% 0.13% NOT INDICATED INDICATED 13 856 PATIENTS 78% 22% CANCER CANCER
29. b) Appropriateness of the indication for upper endoscopy: a meta-analysis Di Giulio E, Hassan C, Zullo A, et al. , DLD 2009 ASGE guidelines Sens. Spec. PPV NPV Relevant finding 85% 28% 49% 70% Cancer 97% 22% 2% 99.8%
33. RESULTS Table 4. Multivariate analysis for the detection of relevant finding Clinical variable Relevant finding OR (95% CI) Bleeding 3.51 (2.9 – 4.2) Relevant finding at previous EGD 2.76 (2.5 – 3.1) Appropriateness 2.7 (2.4 – 3) Male sex 1.77 (1.6 – 1.9) Age > 45 1.55 (1.4 – 1.7) Alarm symptoms 1.39 (1.2 – 1.6) Weight loss 1.32 (1 – 1.6) Reflux 1.16 (1.05 – 1.3) PPI therapy 1.03 (0.93 – 1.1)
34. RESULTS Table 5. Estimates of accuracy of the different strategies in selecting EGD referrals for the detection of relevant findings b) Strategy Rate of EGDs indicated Sens. Spec. PPV NPV AUC ASGE guidelines 80% 88% 27% 51% 72% 0.55 Age > 45/alarm features 78% 82% 26% 49% 63% 0.52
35. RESULTS Figure 1. Receiver operating curve (ROC) for multivariate and ANN models for relevant findings. b)
36. RESULTS Table 4. Multivariate analysis for the detection of malignancy b) Clinical variable New malignancy OR (95% CI) Weight loss 15.2 (9.3 – 24.8) Dysphagia 9.3 (5.7 – 15.6) Alarm features 8 .78 (5.2 – 14.8) Age 8.2 (2.8 – 24) Age > 45 years 8 (2.6 – 23.8) Age > 45 years or alarm features 7.63 (2.3 – 24.7) Vomiting 5.64 (3.2 – 10.1) No previous EGD 7.5 (2 – 28) Anaemia 3.66 (2.2 – 6.1) Bleeding 1.91 (0.5 – 6.4) Family history for cancer 1.77 (0.5 – 6.7) Male sex 1.63 (1 – 2.6) Specialist 1.2 (1 – 2) NSAIDs/anti-COX2/aspirin 1 (0.5 – 2.2)
37. RESULTS Table 5. Estimates of accuracy of the different strategies in selecting EGD referrals for the detection of cancer b) Strategy Sens. Spec. PPV NPV NNT AUC ASGE guidelines 98% 20% 2% 99.8% 50 0.59 Age > 45/alarm features 97% 22% 2% 99.8% 50 0.59
38. RESULTS Figure 1. Receiver operating curve (ROC) for multivariate and ANN models for new cases of malignancy . b)
39.
40. Buri L, Hassan C,Bersani G , Anti M, Bianco MA, Cipolletta L, Di Giulio E, Di Matteo G, Familiari L, Ficano L, Loriga P, Morini S, Pietropaolo V, Zambelli A, Grossi E, Intraligi M, and the SIED Appropriateness Working Group.* Appropriateness guidelines and predictive rules to select patients for upper endoscopy: a nationwide, multicenter study on behalf of SIED
47. “ In conclusion, our study showed that a simple rule based on age and alarm features may be as accurate as the more complex ASGE guidelines in predicting endoscopic outcome in an unselected EGD population. The implementation of such predictive rule would immediately result in the exclusion of more than 20% of the patients from the EGD waiting list, with only a marginal loss of clinical information. Linear and ANN models may be useful to prioritize patients at higher risk of malignancy.”