Effect of a fever control protocol-based strategy on ventilator-associated pneumonia in severely brain-injured patients

During the two study periods, brain-injured patients were similarly managed according to identical local protocols based on international guidelines [17],[18]. The two major objectives were as follows: 1) to infuse neurosedation combining midazolam and opioids; and 2) to maintain cerebral perfusion pressure ≥70 mmHg. All the patients were orotracheally intubated; VAP prevention measures included hand washing with alcohol based antiseptic, a 30° semi-recumbent patient position, MV with positive end-expiratory pressure >5 mmHg, and oral care based on local application of povidone-iodine solution six times daily [15]. Selective digestive decontamination or subglottic drainage was not performed, and ulcer prophylaxis was not systematically administered. The core body temperature was continuously monitored with a rectal thermal probe (Mon-a-Therm™, Covidien, Mansfield, MA, USA).

Fever control intervention was applied in cases in which the body core temperature was >38.2°C [12]. This protocol was established based on previously published studies [4],[19],[20] and successively included the following: external cooling, an infusion of 500 mL of cooled (4°C) saline serum and enteral administration of acetaminophen. A neuromuscular blocker (cisatracurium) was also used if shivering occurred. The objective was to reach a core body temperature of 36°C to 38.2°C.

Gender, age, body mass index, the reason for ICU admission, a history of smoking, a positive alcohol test at admission and CT scan-identified lung contusions involving more than one lobe were recorded for each patient at the time of admission to the ICU. The severity of the brain injury was assessed at admission using the Glasgow Coma Scale score and the Simplified Acute Physiological Score II (SAPS II) [21].

We collected the daily minimum (Tmin) and maximum (Tmax) body core temperature and calculated the mean daily temperature as (Tmax + Tmin)/2 during the first eight days of ICU hospitalization. We reported the delay between ICU admission and the initiation of the fever control protocol, the length of application of the fever control protocol and mortality at ICU discharge. Antibiotic prophylaxis, neuromuscular blocker agents, mannitol and thiopental use was reported for all patients. Additionally, acute respiratory distress syndrome (ARDS), defined as previously [22], and microbiological findings for VAP were reported.

VAP was defined as new and persistent pulmonary infiltrates on a chest radiograph occurring after 48 hours of MV, combined with at least two of the following criteria: 1) purulent tracheal secretions and/or body core temperature >38°C and/or leukocytosis >10,000/mm³ or leukopenia <3,000/mm³; and 2) microbiological confirmation via an endotracheal aspirate quantitative culture growing ≥10⁶ colony forming units (cfu)/mL [23]. The occurrence of VAP was reported and expressed as the density of incidence, defined as the number of VAP cases per 1,000 ventilator-days. We also defined early-onset VAP as VAP that occurred ≤5 days after admission and late-onset VAP as VAP diagnosed >5 days, according the American Thoracic Society criteria [24].

UTIs and CRBIs were defined according to the CDC guidelines [25],[26].

The categorical parameters are expressed as number and frequency. The quantitative parameters are expressed as the mean (standard deviation) if there was a normal distribution or the median (interquartile range) otherwise. The categorical parameters were compared using the Chi-square test or, when necessary, Fisher's exact test. The quantitative parameters were compared using the two-tailed t-test (when normally distributed) or with the Mann–Whitney U test. The temperature value was recorded on the date of the VAP diagnosis. To determine risk factors for VAP occurrence, we used a competing risk model, as proposed by Fine and Gray [27]. VAP before the 28th day of hospitalization in the ICU was an event of interest but could be precluded by death before the 28th day of hospitalization. Patients were therefore observed from the day of hospitalization to the occurrence of VAP before the 28th day of hospitalization in the ICU, to death before the 28th day of hospitalization, or to censoring (hospital discharge while living, without VAP occurrence at day 28 or before). We first compared patient characteristics according to whether the fever control protocol was used to identify potential confounding factors existing before VAP occurrence. A univariate analysis was first performed. Variables with P-values <0.20 were then included in the multivariate competing risk model, and a backward selection was applied. The model was adjusted for the risk factors selected according to use of the fever control protocol. By stratifying the duration of fever control, we analyzed the effect of the duration of fever control on VAP occurrence. The adjusted hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs) were calculated, and the cumulative incidence curves for VAP were determined using the Nelson-Aalen non-parametric estimator. By stratifying the fever control duration, we could also describe the impact of the fever control duration on VAP based on the cumulative incidence curve with the Nelson-Aalen non-parametric estimator. A two-sided P-value <0.05 was considered statistically significant. The statistical models were built using SAS 9.3 (SAS Institute, Inc., Cary, NC, USA) and R Core Team (2012) software (R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria).

All patients in the intervention group were treated with the fever control protocol when their fever was >38.2°C. The delay before its effective application was less than 24 hours, and its median duration was four days (range, two to six days). The daily mean temperature curves for each group are displayed in Figure 2.

The density of the incidence of VAP was significantly increased in the intervention group beginning on ICU day 5 (26.1 versus 12.5 VAP cases per 1,000 days of ventilation; P <0.001). The incidence of VAP was 38% (n = 37) versus 12% (n = 11) for the intervention group and historical control group, respectively. The mean delay for VAP occurrence was similar in the two groups (8 ± 4 days). The mean length of MV was significantly increased in the intervention group (15 ± 11 versus 10 ± 7 days, P <0.001). Methicillin-sensitive Staphylococcus aureus (MSSA), Streptococcus pneumoniae and Haemophilus influenzae comprised 80% of the microbiological findings. The microbiological findings are detailed in Table 3. Considering the VAP-onset delay, we reported 10 early-onset VAP cases in the intervention group compared with 5 in the historical control group and 27 late-onset VAP cases in the intervention group compared with 6 in the historical control group.

The densities of incidence of UTIs and CRBIs were not significantly different between the intervention and the historical control groups (9.7 versus 14.7 UTIs per 1,000 urinary catheter-days, respectively; P = 0.9; 3.9 versus 1.6 CRBIs per 1,000 catheter-days, respectively; P = 0.19). The mortality rate also did not differ between the two groups (34% in the intervention group versus 23% in the historical control group; P = 0.107).

In the univariate analysis (Table 4), the patients who developed VAP were significantly younger, used tobacco more frequently and had more extensive lung injury at admission to the ICU. Application of the fever control protocol was a major risk factor for developing VAP (HR 3.06; 95% CI (1.58, 5.94)). In the multivariate analysis, the competing risk model was built after adjustment of the HR for a history of smoking, age, the SAPS II score, pentothal use and the use of neuromuscular blocker agents. The use of the fever control protocol was the only significant independent risk factor for VAP occurrence (HR 2.73, 95% CI (1.38, 5.38)) (Table 4). The cumulative incidence function curves for VAP are displayed in Figure 3. The risk of VAP appeared to be higher when the duration of the fever control protocol was >3 days (Figure 4).