Does Delaying Early Intravenous Fat Emulsion during Parenteral Nutrition Reduce Infections during Critical Illness?

Patients and Methods

A retrospective review was performed after obtaining approval from our Institutional Review Board. Patients admitted to our 44-bed SICU at The Ohio State University Medical Center who received PN from October 2006 to June 2009 were identified by querying a prospectively collected computer database. Intravenous fat emulsions began to be delayed routinely in November 2007; thus, the “before” period was October 2006–2007, and the “after” period was November 2007–June 2009. Patients receiving PN prior to SICU admission (including home PN), propofol, or clevidipine or who received PN for more than seven days were excluded. Patients aged less than 18 years and patients who were pregnant or were prisoners also were excluded. The data collected included demographics, indication for PN, red blood cell (RBC) and fresh frozen plasma (FFP) transfusion(s), nutritional data (prealbumin and glucose concentrations), microbiology data (respiratory, blood, catheter, and urine cultures and sensitivities), duration of mechanical ventilation, and ICU and hospital lengths of stay. Acute Physiology and Chronic Health Evaluation (APACHE) II scores were calculated for all patients on admission to the SICU. All patients receiving PN were followed by the Nutritional Support Service, consisting of dieticians, clinical pharmacists, nutritional service nurses, and an attending surgeon, who approved the use of PN. Prior to initiation of PN, new central venous catheters with dedicated ports were placed. These ports were accessed only for PN infusions. Nutritional measures, including total calories, grams of amino acids, amount of glucose, and use of IVFE, were monitored by the Nutritional Support Service. Weekly laboratory nutritional measurements including prealbumin were ordered. The clinical pharmacists and dietician monitored patients daily, and, in conjunction with the primary medical team and Nutritional Support attending surgeon, modified PN appropriately. The PN formulation was compounded in the Department of Pharmacy as a 3-in-1 mixture using commercially available IVFE made from soybean oil (Intralipid®, Baxter Healthcare Corp., Deerfield, IL).

Infectious complications were defined as follows. A diagnosis of pneumonia required at least one of the following: (1) Temperature > 100.4°F; (2) white blood cell count < 4,000 or > 12,000 per microliter; or (3) altered mental status in combination with at least two of the following: (a) New-onset purulent sputum or change in sputum character; (b) new or worsening cough, dyspnea, or tachypnea; (c) worsening gas exchange or increased oxygen requirements on ventilatory support; or (d) rales in combination with > 10,000 colony-forming units (cfu) of bacteria/mL from protected-catheter bronchoalveolar lavage quantitative cultures [10,11]. Blood stream infections (BSI) were defined as bacteremia or fungemia if two or more blood cultures grew the same organism and no apparent source could be identified for blood-borne pathogens [9]. The diagnosis of CRBSI required clinical evidence of sepsis and catheter and blood cultures yielding the same pathogen [9]. Diagnosis of catheter-associated UTIs required symptoms and > 100,000 cfu/mL in a urine culture [12].

The primary outcome was the development of any infectious complication between PN initiation and hospital discharge. Secondary outcomes were the duration of PN, lengths of stay and mechanical ventilation, and death. A sample size of 33 patients per group was calculated to be sufficient to detect a decrease in total infectious complications from 60% to 30% with an alpha of 0.05 and a beta of 20%. Statistical analyses included the Fisher exact test for nominal data, the Mann-Whitney U test for nonparametric or ordinal data, and the Student t-test for continuous data. Non-parametric or ordinal data are presented as medians (25%–75% intra-quartile ranges), and continuous data are presented as means ± standard deviations. A p value < 0.05 was considered significant.

Results

During the study period, 6,691 patients were admitted to the SICU, and PN was used in 163 (2.4%). Ninety-nine of these patients were excluded from the analysis for the following reasons: 56 were already receiving IVFE at ICU admission (42 PN, 14 propofol), 13 were prisoners, and 30 received PN for six or fewer days. Therefore, 64 patients, including three trauma patients, were included in the analysis; 30 received immediate IVFE, and 34 had IVFE withheld for seven to ten days. The groups had similar demographics and admitting services, although there was a trend toward older age in the patients receiving immediate IVFE (68 ± 13.4 years vs. 61.5 ± 14.2 years; p = 0.069; Table 1). Weights at admission and body mass indexes were similar in the two groups (Table 1). Importantly, the mean durations of PN, mechanical ventilation, and ICU stay and the hospital mortality rate were similar (Table 2).

Because transfusions have been associated with infectious complications, we evaluated receipt of RBCs or FFP. There was no significant difference in infectious complications in patients who received an RBC transfusion and those who did not (p = 0.59). Likewise, infectious complications were not different for those who received FFP (67.5%) and those who did not (62.5%; p = 0.79). The percentages of patients who required RBCs (70% immediate vs. 56% delayed IVFE; p = 0.31) or FFP transfusions (70% immediate vs. 56% delayed IVFE; p = 0.31) were similar in the two groups. For those who required RBCs, a median of 6 (range 3–13) units were transfused in those receiving immediate IVFE vs. 4 (range 2–7) units in the delayed IVFE group (p = 0.29). The median amounts of FFP for those transfused also were similar (5 [range 3–13] units in the immediate IVFE group vs. 4 [range 2–7] in the delayed IVFE group; p = 0.62).

The majority of patients receiving PN developed at least one infection (65.6%) with similar rates in the two groups (Table 3). The most common infectious complication was pneumonia, which occurred in 53.1% of patients, and UTIs were documented in approximately one-third of patients. Blood stream infections, including candidemia or CRBSI, likewise occurred at similar rates in the two groups. Patients who developed BSI or CRBSI had a significantly longer duration of PN (32.5 ± 30.3 days) than those who did not have such infections (14.7 ± 8.1 days; p = 0.013).

Parenteral nutrition was started after a mean of 9.6 ± 6.2 days in the immediate IVFE group and 9.2 ± 6.7 days in the delayed IVFE group (p = 0.86). The degree of malnutrition, as indicated by the baseline prealbumin concentration, was similar in the two groups (Table 2). Because IVFE is the most calorically dense macronutrient, those having IVFE delayed received fewer calories during the first seven to ten days (25.8 ± 3.3 kcal/kg immediate IVFE vs. 22.1 ± 3.9 kcal/kg delayed IVFE; p = 0.0004). Those who received immediate IVFE also received less glucose (2.57 ± 0.52 mg/kg/min vs. 2.97 ± 0.54 mg/kg/min; p = 0.0068) and less amino acids (1.54 ± 0.28 gm/kg vs. 1.76 ± 0.30 gm/kg; p = 0.005). All patients received insulin while on PN, with 61 patients receiving an intravenous insulin infusion on the day of initiation of PN (96.7% immediate IVFE vs. 94.1% delayed IVFE; p > 0.99) and all patients by the third day of PN. The mean glucose values during treatment with PN were similar in the two groups (127.5 ± 37.6 immediate IVFE vs. 133 ± 37 mg/dL delayed IVFE; p = 0.16).

Discussion

In this study, withholding IVFE for seven to ten days did not reduce infectious complications or the mortality rate in a mixed population of critically ill surgical patients. Importantly, patients in the two treatment periods had similar baseline prealbumin concentrations, transfusion requirements, glucose management, and mean glucose values, all of which influence infectious complications. These results are in substantial disagreement with currently available data showing reduction of infectious complications if IVFE is withheld for several days [2,9,10].

Some authors have hypothesized that provision of 100% of the estimated caloric requirements stimulates inflammation, cytokine production, and oxidative damage, increasing the risk of infectious complications [13,14]. Unfortunately, few studies comparing hypocaloric and eucaloric feedings describe infectious outcomes. Dickerson et al. compared hypocaloric (<20 kcal/kg) and eucaloric (≤20 kcal/kg) enteral feedings retrospectively in 40 critically ill obese patients [15]. They reported that the risk of pneumonia, intra-abdominal abscess, empyema, or sepsis did not differ in the two groups, although these results may have been confounded by differences in antibiotic therapy. Both Battistella et al. and McCowen et al. reported that delaying IVFE reduces infectious complications, but in both studies, patients having delayed IVFE received fewer calories, supporting the hypocaloric feeding influence hypothesis. In contrast, in the present report, patients who had delayed IVFE showed no reduction in infectious complications, despite receipt of fewer calories.

There are several possible influences that might explain these differences in outcome. First, and perhaps most important, is the temporal influence of IVFE infusion. Previous work by others suggested that IVFE administration patterns influence the development of infections secondary to effects on reticuloendothelial system (RES) function [7]. Specifically, Seidner et al. showed that a single 10-h infusion of IVFE does not influence RES function, whereas intermittent 10-h infusions repeated daily for three days cause significant RES impairment [16]. In contrast, continuous infusion of identical total quantities of IVFE over three days does not impair RES function [17]. The fact that Battistella et al. administered their lipids over 10–12 h [9] whereas we administer IVFE as part of a mixture over 24 h could explain the significant differences in the outcomes of our studies.

Second, the association between hyperglycemia and infectious complications in critically ill patients is well accepted [18–22]. In the current report, serum glucose concentrations were reasonably well controlled and were not significantly different in the two groups (Table 2). Because the studies by both Battistella et al. [9] and McCowen et al. [10] were published before the appearance of the landmark article by van den Berghe et al. [23], glucose control presumably was not as aggressive in the earlier studies. Nonetheless, glucose control was similar in patients receiving immediate vs. delayed IVFE in both studies. We therefore conclude that the differences in the outcomes of our studies are not likely consequent to differences in the incidence or management of hyperglycemia in patients receiving delayed IVFE and fewer calories.

Another major contributor to infectious complications during critical illness is transfusion of either RBCs or FFP [24,25]. One strength of our study is inclusion of transfusion data, and importantly, there were no significant differences in transfusions of either RBCs or FFP in those receiving immediate vs. delayed IVFE. Although Battistella et al. report comparable RBC transfusions, it is unclear if their accounting includes FFP and what percentage of patients required a transfusion [9]. Unfortunately, McCowen et al. did not report transfusion data. It therefore is possible that differences in transfusion practices explain the differences between our reported outcomes [10].

Interestingly, our rates of infections and, in particular, pneumonia are higher than those previously reported, and this discrepancy could be attributable to any of several additional factors. First, there undoubtedly are selection biases inherent in each institution's decision to initiate PN. Second, we observed significant differences between the patients of Battistella et al. and those of McCowen et al., including in obesity, age, and overall patient mix (trauma vs. mixed medical/surgical vs. surgical). This is important because both obesity [26,27] and age [28] have been associated with a higher risk of ICU-acquired infections. Finally, the significantly higher rates of pneumonia in the current report might simply reflect differences in pneumonia definition or methodologic sensitivities between the sputum cultures used in previous studies and the quantitative cultures used in our study. Together, we believe that these differences explain the overall higher rate of infections in the current report.

The major limitations of this study include its retrospective nature and small sample size. Although it may have been insufficiently powered to detect minor differences between groups, the previously published differences in infection rates are not small. In fact, the studies of both Battistella et al. [9] and McCowen et al. [10] managed to show significant differences using similar sample sizes. In contrast, we saw no appreciable benefit of delaying IVPE, not even something we could consider a “trend.” This suggests that inherent differences between institutional practices or our patient populations must have contributed to these discrepant results. Finally, the use of prealbumin as a nutritional marker in the ICU is problematic, as concentrations can be influenced by inflammation [3]. The development of infectious complications such as pneumonia may be associated with inflammation, which could decrease the prealbumin concentration. Only one patient had C-reactive protein measured as an indicator of inflammation, and no patients had the erythrocyte sedimentation rate measured.

Conclusions

Withholding IVFE during the first seven to ten days of PN did not reduce infectious complications or the mortality rate in our multispecialty SICU patients requiring PN. These findings suggest that the previously demonstrated benefits of withholding IVFE may not be generalizable to all SICU patients. Given the myriad of differences in practices between ICUs and institutions, a multicenter randomized controlled trial will be needed to confirm the benefit of delaying IVFE in critically ill surgical patients.