Remifentanil discontinuation and subsequent intensive care unit-acquired infection: a cohort study

Introduction Recent animal studies demonstrated immunosuppressive effects of opioid withdrawal resulting in a higher risk of infection. The aim of this study was to determine the impact of remifentanil discontinuation on intensive care unit (ICU)-acquired infection. Methods This was a prospective observational cohort study performed in a 30-bed medical and surgical university ICU, during a one-year period. All patients hospitalised in the ICU for more than 48 hours were eligible. Sedation was based on a written protocol including remifentanil with or without midazolam. Ramsay score was used to evaluate consciousness. The bedside nurse adjusted sedative infusion to obtain the target Ramsay score. Univariate and multivariate analyses were performed to determine risk factors for ICU-acquired infection. Results Five hundred and eighty-seven consecutive patients were included in the study. A microbiologically confirmed ICU-acquired infection was diagnosed in 233 (39%) patients. Incidence rate of ICU-acquired infection was 38 per 1000 ICU-days. Ventilator-associated pneumonia was the most frequently diagnosed ICU-acquired infection (23% of study patients). Pseudomonas aeruginosa was the most frequently isolated microorganism (30%). Multivariate analysis identified remifentanil discontinuation (odds ratio (OR) = 2.53, 95% confidence interval (CI) = 1.28 to 4.99, P = 0.007), simplified acute physiology score II at ICU admission (1.01 per point, 95% CI = 1 to 1.03, P = 0.011), mechanical ventilation (4.49, 95% CI = 1.52 to 13.2, P = 0.006), tracheostomy (2.25, 95% CI = 1.13 to 4.48, P = 0.021), central venous catheter (2.9, 95% CI = 1.08 to 7.74, P = 0.033) and length of hospital stay (1.05 per day, 95% CI = 1.03 to 1.08, P < 0.001) as independent risk factors for ICU-acquired infection. Conclusions Remifentanil discontinuation is independently associated with ICU-acquired infection.


Introduction
Healthcare-associated infections are the most common complications affecting hospitalised patients [1]. Intensive care unit (ICU)-acquired infections represent the major part of these infections [2]. In a recent multicentre study conducted in 71 adult ICUs [3], 7.4% of the 9493 included patients had an ICU-acquired infection. ICU-acquired pneumonia (47%) and ICU-acquired bloodstream infection (BSI) (37%) were the most frequently reported infections. Another recent multicentre study was conducted in 189 ICUs [4]. Of the 3147 included patients, 12% had an ICU-acquired sepsis. ICU-acquired infections are frequently advocated as a significant contributor to mortality and morbidity [5,6]. Identifying risk factor for healthcare-associated infections could be helpful for future studies aiming at preventing these infections.
Sedative and analgesic medications are routinely used in mechanically ventilated patients to reduce pain and anxiety and to allow patients to tolerate invasive procedures in the ICU [7]. Mostly a combination of an opioid, to provide analgesia, and a hypnotic, such as benzodiazepines or propofol to provide anxiolysis, is used. A variety of opioids used by intrave-BSI: bloodstream infection; cfu: colony forming units; CI: confidence interval; ICU: intensive care unit; IL: interleukin; MDR: multidrug-resistant; OR: odds ratio; SAPS: simplified acute physiology score; TNF: tumour necrosis factor; VAP: ventilator-associated pneumonia. nous administration in adults are available for use in the ICU, including morphine, fentanyl, alfentanil, sufentanil and remifentanil [8]. The use of fentanil, alfentanil and sufentanil, as well as morphine, is always accompanied by concerns regarding drug accumulation. In contrast, remifentanil is a short-acting opioid which is characterised by a rapid and uniform clearance and a highly predictable onset and offset of effect [9]. Recently, negative effects of opioid withdrawal on the immune system were reported in opiate abusers [10,11]. In addition, animal studies suggested that morphine withdrawal induces immunosuppression resulting in an increased risk of infection [12,13]. We hypothesised that remifentanil discontinuation would be associated with a higher risk for subsequent ICU-acquired infection.

Study design
This prospective observational cohort study was conducted in a 30-bed medical and surgical university ICU from December 2006 to December 2007. In accordance with the French law, approval by the local Institutional Review Board and informed consent was not required, given that this observational study did not modify current diagnostic or therapeutic strategies. All patients hospitalised in the ICU for more than 48 hours were eligible for this study.

Study population
The infection control policy included isolation techniques, routine screening of multidrug-resistant (MDR) bacteria, written antibiotic treatment protocol and continuous surveillance of nosocomial infections. Isolation techniques were performed in all patients with colonisation or infection related to MDR bacteria and in all immunosuppressed patients [14]. These techniques included protective gowns and gloves usage associated with adequate hand hygiene using alcohol-based hand rub formulation before and after patient contacts. No selective digestive decontamination was performed.

Sedation protocol
Sedation was based on a written protocol including remifentanil with or without midazolam. Ramsay score was used to evaluate consciousness [15]. The target Ramsay score was determined by the physicians. The bedside nurse adjusted sedative infusion to obtain target sedation level. Remifentanil was first used to obtain the sedation target and infusion of remifentanil could be increased every five minutes. If maximal dose of remifentanil was insufficient to obtain the prescribed Ramsay score, midazolam infusion was started and adjusted by the bed-side nurse. For example, in a patient weighing 60 kg, midazolam was started if remifentanil was insufficient to obtain the target Ramsay score at 58 mg/day. No daily interruption of sedation was performed. Sedation discontinuation was at the physician's discretion. However, in patients who received a combination of remifentanil and midazolam, midazolam was always discontinued before or at the same time as remifentanil. Acute withdrawal was treated by levomepronazin and lorazepam. Propofol infusion was used to treat acute withdrawal refractory to these medications.

Data collection
All data were prospectively collected. At ICU admission, the following data were collected: age; gender; simplified acute physiology score (SAPS) II [16]; logistic organ dysfunction score [16]; McCabe score; admission category; presence of comorbidities, including chronic obstructive pulmonary disease [17], chronic heart failure, immunosuppression [14] and diabetes mellitus; presence of infection; prior antibiotic treatment; and length of hospital stay before ICU admission. During ICU stay, the following data were collected: central venous and arterial catheter use; urinary tract catheter use; mechanical ventilation; duration of use of catheters and mechanical ventilation; reintubation; tracheostomy; fibreoptic bronchoscopy; digestive tract endoscopy; antimicrobial treatment; duration of antimicrobial treatment; remifentanil and midazolam use; duration and doses of remifentanil and midazolam used; discontinuation of remifentanil and midazolam; neuromuscular blocking agent use; and acute withdrawal. Information on length of ICU stay and ICU mortality was also collected.

Definitions
Sedation discontinuation was defined as an interruption of all sedatives for at least 24 hours, except those medications given to treat acute withdrawal. Acute withdrawal was defined by the presence of at least five of the following criteria [18] during the six hours following sedation discontinuation: fever (> 38°C), tachycardia (> 100 beats/minute), hypertension (mean arterial pressure > 100 mmHg), sweating, mydriasis, diarrhoea, nausea/vomiting and restlessness. Infection was considered as ICU-acquired if it was diagnosed more than 48 hours after ICU admission. ICU-acquired infections occurring less than five days after ICU admission were considered as early onset. Late-onset ICU-acquired infections were defined as those infections diagnosed five days or more after ICU admission.
BSI was defined as onset of infection associated with one or more positive blood culture result unrelated to an infection incubating at ICU admission. Coagulase-negative Staphylococcus BSI was defined as two or more positive blood cultures on separate occasions within a 48-hour period, or at least one blood culture positive with clinical sepsis, no other infectious process and antibiotic treatment given by the attending physician [20]. Ventilator-associated tracheobronchitis was defined by all of the following criteria: fever (>38°C) with no other recognisable cause, purulent sputum production, positive (≥ 10 6 cfu/ml) endotracheal aspirate culture and no radiographic signs of new pneumonia [21]. All Other ICU-acquired infections were defined according to modified Centers for Disease Control and Prevention criteria [22]. Only infections confirmed by microbiological results were taken into account.
Incidence rate was defined as the number of ICU-acquired infections divided by the number of days at risk for these infections. Prior antibiotic treatment was defined as any antibiotic treatment during the four weeks preceding ICU-admission. MDR bacteria were defined as methicillin-resistant Staphylococcus aureus, ceftazidime or imipenem-resistant Pseudomonas aeruginosa, Acinetobacter baumannii and Stenotrophomonas maltophilia and extending spectrum βlactamase producing Gram-negative bacilli [19]. During the study period, no vancomycin-resistant Enterococcus was isolated in the ICU.
Statistical methods SPSS 11.5 software (SPSS, Chicago, IL, USA) was used for data analysis. Results are presented as number (percentage) for categorical variables, and mean ± standard deviation for quantitative variables. Distribution of quantitative variables was tested. All P values were two-tailed. The statistical significance was defined as P < 0.05.
Univariate analysis was used to determine factors associated with ICU-acquired infection. All the above cited variables were included in this analysis. Qualitative variables were compared using the Pearson chi-square test or the Fisher's exact test, as appropriate. Quantitative variables were compared using the Mann-Whitney U test or the Student's t-test, as appropriate. In patients with ICU-acquired infection, exposure to potential risk factors was taken into account until occurrence of the last ICU-acquired infection. For example, remifentanil discontinuation was considered a risk factor if it occurred before ICUacquired infection. In patients without ICU-acquired infection, exposure to potential risk factors was taken into account until ICU discharge. Similarly, length of hospital stay was taken into account until ICU discharge in patients without ICU-acquired infection, and until the last ICU-acquired infection in patients with ICU-acquired infection. Patients with several ICUacquired infections were considered at risk until the occurrence of the last infection.
Multivariate analysis was used to determine factors independently associated with ICU-acquired infection. All predictors showing a P < 0.1 association with ICU-acquired infection in univariate analysis were incorporated in the multivariate logistic regression analysis. Potential interactions were tested.
Odds ratio (OR) and 95% confidence interval (CI) were calculated, as well as the Hosmer-Lemshow goodness-of-fit.
In the subgroup of patients who received remifentanil for at least 96 hours before discontinuation, incidence rate of ICUacquired infection, duration of antimicrobial treatment and rate of patients with antimicrobial treatment discontinuation were compared between the two periods of 96 hours before and after remifentanil discontinuation. Further, incidence rate of ICU-acquired infection was compared between the 96 hours following remifentanil discontinuation and the whole ICU stay. Patients who died before remifentanil discontinuation were excluded from this analysis. For these comparisons, paired student's t test, and McNemar's test were used for quantitative and categorical variables, respectively. In this subgroup, characteristics of patients with or without ICU-acquired infection were compared at ICU admission.

Results
During the study period, 587 patients were hospitalised in the ICU for more than 48 hours, and were all included in the study. Patient characteristics at ICU admission, and during ICU stay are presented in Table 1 and Table 2, respectively.
Among the 394 patients who received sedation during the ICU stay, 90 patients died before remifentanil discontinuation. Remifentanil was discontinued in 304 patients, including 286 patients in which remifentanil was discontinued before ICUacquired infection and 18 patients in which remifentanil was discontinued after ICU-acquired infection. No significant difference was found in the reintubation rate between patients with remifentanil discontinuation and patients without remifentanil discontinuation (29 of 304 (9%) vs 22 of 283 (7%), P = 0.242). Mean time from remifentanil discontinuation to subsequent ICU-acquired infection was 4 ± 3 days.

Risk factors for ICU-acquired infection
Several factors were significantly associated with ICUacquired infection by univariate analysis at ICU-admission and during ICU stay (Tables 1 and 2).
Multivariate analysis identified remifentanil discontinuation, SAPS II at ICU admission, mechanical ventilation, tracheostomy, central venous catheter and length of hospital stay as independent risk factors for ICU-acquired infection (Table 3).

Incidence rate of ICU-acquired infection before and after remifentanil discontinuation
Characteristics of the subgroup of patients who received remifentanil for at least 96 hours before discontinuation (n = 266) are presented in Table 4. In this subgroup, incidence rate of ICU-acquired infection was significantly higher during the 96 hours following remifentanil discontinuation compared with the 96 hours preceding remifentanil discontinuation (94 ± 192 vs 29 ± 103 per 1000 ICU days, P < 0.001) and compared with the whole ICU stay (94 ± 192 vs 40 ± 46 per 1000 ICU days, P < 0.001). In these patients, the highest rate of ICUacquired infection was observed on day four after remifentanil discontinuation (Figure 1). Duration of antimicrobial treatment (3 ± 2 vs 2 ± 1 days, P = 0.182) and antimicrobial discontin- uation rate (51 of 266 (19%) vs 69 of 266 (25%), P = 0.121) were similar during the two periods of 96 hours before and after remifentanil discontinuation, respectively.

Discussion
Multivariate analysis identified remifentanil discontinuation, SAPS II at ICU admission, use of mechanical ventilation, tracheostomy, central venous catheter and length of hospital stay as independent risk factors for ICU-acquired infection. All these factors, except remifentanil discontinuation, were identified by previous studies as important risk factors for health care-associated infections [2,[23][24][25][26]. To our knowledge, our study is the first clinical study to evaluate the impact of opioid discontinuation on ICU-acquired infection [27].
Several potential explanations could be provided for the association between remifentanil discontinuation and ICUacquired infection. First, higher rates of reintubation could be observed after discontinuation of sedation. Reintubation is a well-known risk factor for aspiration and VAP [28]. However, no significant difference was found in reintubation rate between patients with remifentanil discontinuation compared Data are presented as mean ± SD or number (%).
In patients with ICU-acquired infection, exposure to potential risk factors was taken into account until occurrence of the last ICU-acquired infection. In patients without ICU-acquired infection, exposure to potential risk factors was taken into account until ICU-discharge. ICU, intensive care unit; OR = odds ratio; SD = standard deviation.
with patients without remifentanil discontinuation. In addition, reintubation was not independently associated with ICUacquired infection. Second, antibiotic discontinuation may have influenced our results. Several studies demonstrated that systemic antibiotic treatment was associated with reduced rates of early-onset ICU-acquired infections [29,30]. On the other hand, other studies demonstrated that antibiotic treatment was a risk factor for subsequent infections related to MDR bacteria [31,32]. However, in the subgroup of patients sedated for at least 96 hours before discontinuation, duration of antimicrobial treatment and rate of patients with antimicrobial treatment discontinuation were similar during the two periods of 96 hours preceding and following remifentanil discontinuation. Third, the higher rate of ICU-acquired infec-tion after remifentanil discontinuation could be related to immunosuppressive effects observed after opioid withdrawal.
Previous animal studies found morphine withdrawal to be associated with higher rates of infection. In mice exposed to morphine for 96 hours, the effect of morphine withdrawal on spontaneous sepsis and on oral infection with Salmonella enterica was recently examined [13]. Withdrawal significantly increased the Salmonella burden in various tissues of infected mice compared with animals who had placebo withdrawn and decreased the mean survival time. Elevated levels of proinflammatory cytokines were observed in the spleens of mice who had morphine withdrawn, compared with mice who had placebo withdrawn. The same authors [12] demonstrated a correlation between the suppression of IL-12 production and an increased susceptibility to Salmonella infection in mice undergoing withdrawal from morphine. Further, in another animal study, morphine withdrawal sensitised the animals to lipopolysaccharide lethality via increased production of TNF-α and nitric oxide [33].
In rats, immunomodulatory effects of morphine withdrawal were investigated alone and in the presence of the α-2-adrenergic agonist, clonidine [34]. Weight change was observed with peak decreases in weight occurring 24 hours after withdrawal. Rats withdrawn from morphine also exhibited a timedependent suppression of immune status with significantly altered proliferation of T-cells stimulated by concanavalin A, altered proliferation of splenic T-cells stimulated by toxic shock syndrome toxin-1, altered production of the interferon-γ by concanavalin A-stimulated splenocytes and significantly Lemeshow goodness-of-fit test, P = 0.86. CI = confidence interval; ICU = intensive care unit; OR = odds ratio; SAPS = simplified acute physiology score.

Figure 1
Distribution of ICU-acquired infections according to remifentanil discontinuation in patients who received remifentanil for 96 hours of more before discontinuation Distribution of ICU-acquired infections according to remifentanil discontinuation in patients who received remifentanil for 96 hours of more before discontinuation. Mean (standard deviation) length of intensive care unit (ICU) stay was 29 ± 28 days, including 17 ± 7 days before the first ICUacquired infection.
altered natural-killer cell activity. These immunomodulatory effects were most evident 12 hours following morphine withdrawal. In addition, clonidine prevented withdrawal-induced immunosuppression. Other recent animal studies [35,10] confirmed these results and suggested that abrupt cessation of morphine administration resulted in an activation of stressrelated pathways that may contribute to an increased susceptibility of infection during the initial withdrawal phase. These suppressive effects on the immune system were significant for up to 72 hours after withdrawal from chronic morphine. Our results are consistent with these findings because a peak of ICU-acquired infection was observed at day four after remifentanil discontinuation. Disturbances of the stress axis were reported to have a major impact on infections, including higher rates of postoperative pneumonia in long-term alcoholics [36]. A recent randomised controlled study, performed in long-term alcoholics, suggested that intervention at the level of the hypothalamus-pituitary-adrenal axis altered the immune response to surgical stress [37]. This resulted in decreased post-operative pneumonia rates and shortened ICU stay in these patients.
There are biological, as well as pathological, interactions between neuropeptide substance P, which is a modulator of neuroimmunoregulation, and opiates [38]. In an in vitro model, Wang and colleagues [39] investigated the relationship between morphine withdrawal and HIV infection of human T lymphocytes. They concluded that the interaction of opiates and neuropeptide substance P in human T lymphocytes was likely to a have a role in the immunopathogenesis of HIV disease among opiate abusers.
Prevalence of ICU-acquired infection and duration of sedation were high in our study. This could be explained by the high severity of our patients at ICU admission, the large proportion of patients with comorbidities and the high proportion of patients requiring mechanical ventilation. These factors are well known to be associated with prolonged ICU stay and higher risk for ICU-acquired infection [19]. A recent study was performed in 151 ICUs in France, including 30% of university ICUs [40]. During a six-month period, 20,632 patients were included in the study. Incidence rates of ICU-acquired infection were consistent with our findings, including 17 It could be argued that the higher incidence rate of ICUacquired infection during the 96 hours following sedation discontinuation, found by our study, could simply reflect the fact that the cumulative risk of ICU-acquired infection was higher during this period compared with the 96 hours preceding sedation discontinuation. However, although the cumulative risk for ICU-acquired infection was higher during the whole ICU stay compared with the 96 hours following sedation discontinuation, incidence rate of ICU-acquired infection was higher during the 96 hours following sedation discontinuation compared with the whole ICU stay. Additionally, remifentanil discontinuation was independently associated with ICUacquired infection.
Daily sedation interruption was not performed in this study. However, a nurse-adjusted sedation protocol was used. A recent randomised controlled study compared nursing-implemented sedation algorithm with daily interruption of sedation [41]. The authors found the sedation algorithm to be associated with reduced duration of mechanical ventilation and lengths of stay compared with daily interruption of sedation. In addition, a recent web-based survey was performed on members of the Society of Critical Care Medicine [42]. Daily interruption of sedation was only performed by 40% of the 904 responders. Another recent survey demonstrated that none of the 44 participating ICUs were conducting daily interruption of sedation [43].
Our study has some limitations. First, this was a single centre study and the results may not be generalisable to other ICUs. Second, our study was observational, and further randomised studies are needed to confirm these results. Third, remifentanil was used for sedation. Whether our results are applicable to patients sedated with other opioids is unknown. A recent in vitro study found remifentanil, but not sufentanil or alfentanil, to attenuate lipopolysaccharide-induced neutrophil responses and neutrophil-mediated inflammatory responses [44]. Fourth, midazolam was used in a large proportion of patients sedated with remifentanil. Therefore, midazolam discontinuation may have influenced the subsequent ICU-acquired infection. However, midazolam use and discontinuation were not independently associated with ICU-acquired infection. Fifth, our risk analysis did not account for the patient's condition on the day of remifentanil discontinuation. However, a different study design, such as a matched-controlled study, would be more appropriate to adjust for this potential confounder. Finally, although remifentanil discontinuation was an independent risk factor for ICU-acquired infection, acute withdrawal was not independently associated with ICU-acquired infection. One potential explanation is the fact that our definition of acute withdrawal was stringent resulting in possible underestimation of acute withdrawal frequency. However, no relationship between infection and clinical signs of abstinence has been reported by animal studies [12,13].

Key messages
• Remifentanil discontinuation is independently associated with a higher risk for ICU-acquired infections.
• Further randomised studies are needed to confirm these results.