Patient population
This retrospective before-after study was performed with a cohort of haematological patients treated from 2007-2014 at the haematology department of the UMCU. The intervention consisted of the introduction of a new local protocol that, in contrast to the previous local protocol, promoted early discontinuation of EAT with carbapenems and discouraged standard empirical treatment with vancomycin. This new protocol was started on January 1st 2011.
All adult patients (≥ 18 years) with the diagnoses AML or high risk MDS, treated between January 1st 2007 and December 31st2014 in the UMCU with at least one period of prolonged and profound neutropenia were included. Of these patients, only neutropenic episodes related to intensive chemotherapy (including one of the following cytostatic agents: idarubicin, cytarabine, daunorubicin, vincristine, adriamycin, mitoxantrone or etoposide) or allogenic SCT were included. Neutropenic episodes following other chemotherapeutic regimens or unrelated to chemotherapy or allogenic SCT were excluded. Neutropenic episodes between January 1st 2007 and October 1st 2010 (period I) were compared to neutropenic episodes between April 1st 2011 and December 31st 2014 (period II, after starting the new protocol). The period from October 1st2010 to April 1st 2011 was considered a transition period, therefore neutropenic episodes occurring within this interval were excluded. Complete patient data was not available before 2007, therefore the maximal period that could be studied before the intervention was 45 months (taken into account the transition period). A similar length was chosen for study period II, i.e. 45 months, making the total study period eight years (2007-2014).
Definition of profound and prolonged neutropenia
Neutropenia was defined as at least two consecutive neutrophil measurements of < 0.5x109cells/L within 90 days. A single neutrophil count above 0.5x109cells/L was ignored, if flanked by neutrophil counts below < 0.5x109 cells/L within one week. Only prolonged neutropenic episodes, with a duration of seven days or more, occurring within fourteen days after start of chemotherapy or conditioning for allogenic SCT were taken into account. These neutropenic episodes are frequently accompanied by mucositis. If chemotherapy or allogenic SCT was started during a period of pre-existing neutropenia, the ensuing neutropenia was considered to be (at least partially) treatment- related if the neutropenia continued for at least seven days after start of treatment.
Prophylaxis and antimicrobial treatment; old protocol
Data sources
A primary database containing patient data extracted from hospital electronic patient record systems was set up using the Research Data Platform in the UMCU. This database consisted of all patients linked to the diagnosis or treatment of AML or MDS. Data of prescribed antibacterial therapy and cell-count data used to identify neutropenic episodes, were derived from the Utrecht Patient Oriented Database (UPOD) (12). Data of administered cytostatic agents and data concerning allogenic SCT in our patient selection was collected from the in-hospital pharmacy department of the UMCU and the treatment files of the haematology department, respectively. Data of positive blood cultures was derived from the General Laboratory Information Management System (GLIMS). This study was performed in accordance with the ethical standards of our centre.
Outcome measurements
The primary outcome was carbapenem and vancomycin use within neutropenia following chemotherapy or conditioning for allogenic SCT in AML and high-risk MDS patients. Because neutropenia following chemotherapy or SCT was considered as the period that patients were most at risk for severe infections and the intervention involved a policy change regarding antibiotic use within this period of neutropenia, antibiotic use was expressed as total days of therapy (DOT) per 100 neutropenic days. Interrupted time series (ITS) analysis was performed to assess pre-existing trends, the immediate effect of intervention (step change) and the sustainability of this effect in period II. In addition, the overall use of carbapenems and vancomycin within period I, defined as the sum of DOT divided by the sum of neutropenic day, was compared to the overall use in period II.
As secondary outcome the use of other antibiotic agents was analysed, expressed as DOT per 100 neutropenic days. Total antibiotic use consisted of the sum of DOT of all antibiotics combined. If a patient used 2 (or more) different antibiotics on a particular day this was counted as 2 (or more) DOT. Cotrimoxazole was analysed when dosed in 960 mg BID, leaving out Pneumocystis jirovecii/Toxoplasma prophylaxis after allogenic SCT (480 mg QD).
Other secondary outcomes were 30-day mortality, ICU-admission within 30 days after start EAT and blood cultures positive for microorganisms sensitive to imipenem. These outcomes were measured for the neutropenic episodes in which EAT was started. Cases of mortality within 30 days after start of EAT with a carbapenem were reviewed separately by three of the authors (AN, AB and TVDB). Overall mortality, infection-associated mortality and carbapenem preventable mortality were distinguished. Infection-associated mortality was defined as clinical signs and/or microbiologically results compatible with infection at time of death. Carbapenem preventable mortality was defined as infection-associated mortality where continuing EAT possibly could have prevented the adverse outcome, because of a suspected or proven etiologic agent that was carbapenem sensitive.
Positive blood cultures within neutropenia drawn after discontinuation of EAT were analysed, because the goal was to study the possible adverse consequences of early discontinuation of carbapenems (e.g. infection/bacteraemia with carbapenem sensitive microorganisms).
Data analysis
Analysis was conducted at the level of the neutropenic episodes. Changes in carbapenem use and vancomycin use were analysed by ITS. Effect of intervention was represented by a step change (representing the immediate effect). A segmented linear mixed regression model was used with a random intercept for individual patients to correct for multiple neutropenic episodes per patient. Analysis was not corrected for confounders. The 95% confidence interval was derived through bootstrapping with 20000 iterations. For all antibiotics included in this study, the overall use of different antibiotics within period I was compared to period II, expressed in DOT/100 neutropenic days. Data-analysis was performed using SAS enterprise Guide 7.1. Statistical analysis was done using R version 3.5.1.