BackgroundTwo reversible tyrosine-kinase inhibitors of the epidermal growth factor receptor (EGFR-TKI) (A and B) were approved for the treatment of EGFR-mutant advanced non-small cell lung cancer (aNSCLC), with similar activity and results. Pharmacovigilance detected efficacy differences between A and B in our centre.PurposeTo define and possibly correct the cause of this finding.Material and methodsDrug A was considered our standard treatment for EGFR-mutant aNSCLC from April 2011 to March 2013, and was replaced with B from April 2013 to the present. EGFR-mutant patients were sequentially diagnosed in two different external platforms (PA and PB) during the same periods of time. We retrospectively reviewed the medical charts of TKI-treated EGFR-mutant aNSCLC patients from April 2011 to March 2014. Progression free survival (PFS) was analysed in any, first, and second line of treatment by Kaplan-Meier curves and Cox regression. The finding of significant differences in PFS between A and B led to the retrospective review of all EGFR-analysed aNSCLC patients.ResultsFifteen EGFR-mutant aNSCLC patients were treated with A (7 s line), and 16 with B (10 s line). Mean age of the series was 65 years (44–82), and 74.2% were women. PFS benefited A in any (11.43 vs. 4.96 months; p = 0.000), first (13.3 vs. 3.98 months; p = 0.014), and second line of treatment (9.5 vs. 5.53 months; p = 0.023). PA analysed 108 aNSCLC, patients detecting 12.1% EGFR-mutant and 15% non-analysable tissue samples. PB analysed 85 aNSCLC, patients finding 20% EGFR-mutant and 3.5% non-analysable tissue samples.ConclusionThe lower PFS of B-treated aNSCLC patients was attributed to an excessive sensitivity of PB in detecting EGFR-mutation. Re-calibration of the technique modified the current percentage of EGFR-mutant and non-analysable samples to 15.5% and 10% of 129, respectively. Periodic monitoring of selected drugs helps to correct protocol defects, to improve quality of treatments, and to reduce costs.References and/or acknowledgementsConflict of interest

were combined to produce the risk level for each situation. Only 2 Delphi rounds were necessary. Results After the first round a consensus was reached for 8 situations. Experts agreed on the level of risk associated with 48 out of 52 modelled situations. A high or extreme consensus risk level is determined for 45 modelled situations. These situations represent a variety of drug-related problems. Overdosing was the most frequent situation [12 (22%)]. Cardiovascular, Psychiatric and Endocrinological drug classes are the most common involved in respectively [25 (45%)], [7 (13%)] and [5 (9%)] situations. Conclusion and RelevanceThe symbolic artificial intelligence to detect drug-related problems in patients' medications will be much more shared if pharmaceutical algorithms including the clinical risk are defined through consensus.

structure and mechanism of action, may play a unique role in patients who have limited or no alternative treatment options. Aim and objectives The aim of this study was to describe the first cases of prescriptions of cefiderocol used in the 5 months following its availability in Italy, and the hospital pharmacist interventions in assisting clinicians from microbiology select a safe and appropriate antibiotic treatment. Material and methods A standardised prescription form was sent to the infectious disease specialist to collect patients' characteristics, infection type, reasons for cefiderocol use, doses and duration of treatment (concomitant treatments, adverse events and outcome). A susceptibility testing kit (30 mg cefiderocol disc) was provided to the microbiology specialist in order to reserve this new antibiotic for patients with cefiderocol-susceptible isolates. A retrospective study was performed to collect the data of adult patients who received cefiderocol. Results A total of 30 patients with mean age of 56 (23-90) years received cefiderocol (9 females, 21 males). Of these, 19 patients were treated in intensive care units, with the most common regimen of 2 g three times/day tid (n=6), while 3 patients with acute renal failure required a regimen of 750 mg twice daily. The main sites of infection were respiratory tract (n=16), urinary tract (n=3), intra-abdominal (n=4) and bloodstream (n=5). 5 patients had multisite infections.The duration of therapy was in the range 6-16 days. The most common pathogens were Acinetobacter baumannii (n=13), Klebsiella pneumoniae (n= 8), Pseudomonas aeuroginosa (n=10) and Enterobacter spp (n=5). 10 patients had superinfections. The most concomitant therapy was colistin (n=9). No severe adverse events were reported. 7 patients with septic shock died. Conclusion and relevanceOur study describes real-life experience of the use of cefiderocol as a salvage option in critical patients, providing additional data on its benefit, safety and limits in both empirical and targeted treatment of multidrugresistant Gram-negative bacteria (MDR-GNB) infections, and it confirms the need for a multidisciplinary team.

BackgroundPaediatric patients involve certain complexities that make them vulnerable to medication errors and adverse patient outcomes. Most of the medication errors occur at the stage of physician ordering and they are often dosing errors. Computerised Physician Order Entry (CPOE) results in legible, structured and complete prescriptions. Furthermore, there is an improvement in the communication between physicians, nurses and pharmacists compared with handwritten orders.PurposeThe objective of this study was to evaluate the impact of CPOE on the frequency of errors in the medication ordering process in a paediatric unit.Material and methodsA prospective observational study was conducted in a 30-bed paediatric unit of a tertiary teaching hospital. The physician’s orders were reviewed for 2 months before and 2 months after CPOE implementation. Medication errors were identified and classified into errors of: dosing, interval, units, route of administration, treatment duration, schedule, wrong drug, incomplete order and rule violation.ResultsA total of 1164 orders of 212 patients were reviewed. Before implementation, medication errors occurred at a rate of 3.3 per 100 orders (n=20): 35% (n=7) were dosing errors, 25% (n=5) incomplete orders and 20% (n=4) unit errors. After implementation, the rate was increased to 6.6 per 100 orders (n=37): 24.3% (n=9) were dosing errors, 18.9% (n=7) rule violations, 18.9% (n=7) wrong treatment duration, 13.5% (n=5) schedule errors, 20% (n=4) unit errors and 8.1% (n=3) interval errors.ConclusionThe implementation of CPOE resulted in an increase in the number of medication errors, but the type of them was clearly different. While handwritten errors were the result of calculation errors, missing information or confusion in writing, CPOE errors were mainly due to the inexperience of using the program. The consequences of the CPOE errors were less harmful than handwritten prescription errors.No conflict of interest

No abstract