A randomized trial of supplemental parenteral nutrition in underweight and overweight critically ill patients: the TOP-UP pilot trial

Worldwide, there is considerable controversy about the optimal amount and feeding route in critically ill patients [1]. Nutrition practice guidelines in Europe, Canada, and the United States endorse enteral nutrition (EN) for patients who are critically ill and hemodynamically stable [2‐4]. To evaluate the success of EN delivery in the intensive care unit (ICU), a recent observational cohort study of nutrition practices in 167 ICUs across 21 countries was conducted to evaluate worldwide nutrition practices in 2772 patients [5]. Despite multiple international guidelines recommending early initiation of EN in the ICU [2, 3, 6], the data revealed practitioners are only successfully delivering approximately 50% of prescribed daily calories from EN over the first 12 days in the ICU [5]. In addition, in some developed countries like the United States, it takes an average of >60 hours to initiate EN [5].

Because of this consistent and longstanding failure to deliver prescribed EN, parenteral nutrition (PN) has been utilized in up to 35–70% [5] of critically ill patients. However, current guidelines do not agree on when to initiate PN in the ICU [1]. For patients who are intolerant to or have other contraindications to EN, European Society for Clinical Nutrition and Metabolism (ESPEN) guidelines recommend initiating PN within 24–48 hours in patients not expected to receive full oral nutrition within 3 days, and initiating supplemental PN (SPN) if EN levels are not at goal in 48 hours [7]. New US (American Society for Parenteral and Enteral Nutrition [ASPEN]/Society of Critical Care Medicine [SCCM]) guidelines hesitate to recommend early PN in the ICU, with PN initiation advised only after 7 days in well-nourished patients [4]. Although, in patients found to be significantly malnourished via nutrition risk scores (i.e., Nutrition Risk in Critically Ill [NUTRIC] score [without IL-6] ≥5 or Nutrition Risk Score [NRS] ≥5) [8], total PN is recommended to start at ICU admission [4, 7].

Thus, current guidelines and even recent larger randomized trials are conflicting and do not provide clear guidance regarding the use of PN in the early phase of critical illness [1]. In our previous international, multicenter, observational study, we found a significant inverse linear relationship between the odds of mortality and total daily calories received [9]. Our key finding was that increased amounts of calories were associated with reduced mortality for the body mass index (BMI) <25 group and BMI >35 group, with no benefit of increased calorie intake for patients in the BMI 25– < 35 group. Independent of the route of delivery (either EN or PN), an additional 1000 kcals was associated with an almost 50% reduction of 60-day mortality in patients with a BMI of <25 or >35 [9]. These categories of patients have not been studied separately in large-scale prospective randomized controlled trials comparing two nutritional intake levels [10‐14].

Thus, we proposed a randomized trial of supplemental parenteral nutrition in underweight and overweight critically ill patients (the TOP-UP trial) as a multicenter study of critically ill underweight and obese patients with acute respiratory failure expected to require mechanical ventilation for >72 hours. In a future full trial, we proposed to address two questions: (1) the effect of SPN + EN compared with EN alone on 60-day mortality, and (2) the effect of early SPN protein and calorie intake on key quality-of-life (QoL) and functional outcomes. We estimated conservatively that a sample size of approximately 1000 patients/arm would be required to demonstrate a significant mortality effect, assuming an additional 1000 kcal/d would be associated with an approximately 29% relative risk reduction of mortality. (This was based on our pre-existing international nutrition survey data [9].) Prior to implementation of a large-scale definitive trial, we felt a multicenter pilot trial to evaluate the feasibility of a full trial was needed. The primary aim of the pilot trial reported herein was to ensure a clinically significant difference in calorie/protein intake (approximately 30% difference; or 600–1000 kcal/day and 20–30 g protein/day) between the two intervention groups was achievable. We also evaluated the feasibility of performing functional endpoints research in the ICU setting. All clinical endpoints proposed to be assessed in a future full trial were also collected and evaluated. We believe the results and experience gained from this multicenter pilot trial will allow for refinement and optimization of a full-scale multicenter trial to assess optimal methods for targeting SPN to nutritionally “at-risk” patients and inform current practice on the use of PN in the ICU.

This was an investigator-initiated, multicenter, randomized, controlled, pilot clinical study (ClinicalTrials.gov identifier NCT01206166). This trial was conducted between June 1, 2011, and January 20, 2015, in 11 ICUs in Canada, the United States, Belgium, and France. Local jurisdictional approval and institutional research ethics board approval was secured at each site, as described in declarations section below. Written informed consent was obtained from patients, family members, or their legal representatives before enrollment. Eligible patients were randomized within 72 hours of admission to the ICU. A centralized web-based randomization system at the Clinical Evaluation Research Unit (CERU) at Kingston General Hospital was used to randomly allocate patients to study groups. Randomization was stratified by site, presence of medical or surgical admission diagnosis, EN started before randomization, and BMI (<25 or >35). Patients were randomized in random block sizes of two, four, or eight within strata.

Over a 44-month recruitment period, 730 patients were screened, of whom 304 met enrollment criteria and 125 were randomized (Fig. 1). Screening periods at sites varied; enrollment was capped after 20 patients to allow for other sites to contribute. Overall, the average enrollment rate per site was 0.8 patients/month (range 0.3–1.9). Characteristics at baseline were similar in both groups (Table 2). Quality measures regarding protocol adherence and success in intervention delivery are reported in Table 3. Overall, patients in the SPN + EN group were randomized and initiated on study PN rapidly after ICU admission and had a median study intervention duration of 5.9 days (range 2.4–7.6). In the SPN + EN group, 13 patients (25%) received <80% of goal calories/day at some point during their enrollment in trial, which was reported as a protocol violation (Table 4). The reasons for these episodes of <80% of goal calories being delivered during a given day are reported in Table 5. In total, 16 patients (30.8%) in the SPN + EN arm received <72 hours of study PN, and 3 of these patients never received SPN because their nutritional goal was reached early by EN alone.

In this pilot trial of SPN + EN versus EN alone, we found SPN + EN significantly increased calorie/protein delivery over the first ICU week, nearly achieving the targeted 30% increase in caloric delivery. SPN + EN proved feasible to deliver with our prescribed protocol. As expected in this pilot trial, which was not powered for clinical outcomes, no significant outcome differences, including no difference in infection risk between groups, were observed. However consistent encouraging trends in hospital/ICU mortality, QoL, and functional endpoints in the SPN + EN group were observed. Signals of reduced mortality in the NUTRIC ≥5 and BMI <25 subgroups also indicate that SPN + EN may have a particular benefit in higher-nutritional-risk, lower-BMI patients.

Enrollment of critically ill patients meeting the BMI <25 or >35 criterion proved challenging. As the average BMI in recent North American and even European ICU nutrition trials has ranged from 26.5–30.1 [11, 16, 17], a limited number of patients were ultimately eligible for screening. As a result of funding constraints and eligibility challenges, enrollment was constrained to 125 total patients. Further, we block-randomized patients, stratifying by site, medical/surgical diagnosis, BMI, and baseline use of EN. Since the study had several small sites and a large number of strata (eight within each site), there was a high proportion of incomplete blocks, which undermined the effectiveness of the stratification and allowed for a large overall imbalance in the number of patients randomized to each arm. This increases the variance of the between-arm comparisons by 3%, compared with if we had the same number in both arms (see Hsieh et al. for the VIF formula of (k + 1)^2/(4 k) where k = 73/52) [18]. Or, stated another way, this imbalance results in a study with the same power and precision as a study with a total sample size that is 3% smaller but has even numbers in each arm. Thus, this imbalance may have caused a minimal reduction in power but does not meaningfully or statistically bias the estimates or interfere with results. Although this would be less of an issue for a much larger trial, it may be worth considering reducing the number of strata or using an alternative balancing method such as minimization [19]. Another limitation of this study is that all calorie prescriptions were determined using weight-based formulas. Compared with indirect calorimetry-determined nutrition targets, these prescriptions may lead to a greater risk of over- or under-feeding actual caloric need [20]. In the future, we hope for improved metabolic cart availability to allow for improved guidance of feeding targets in the ICU.

Compliance with pre-discharge and post-discharge functional and QoL measures proved challenging to collect in all patients. For the functional tests, this was most often due to patients’ inability to complete testing due to death or significant disability following ICU stay. For example, 60% of patients could not complete the hospital discharge 6-minute walk test due to either an inability to walk (40%) or death (20%). The rank-based analytic approach allowed the inclusion of decedents and patients too ill to perform functional testing. This challenge in obtaining functional outcomes post-ICU stay has been observed in similar trials, such as the EPaNIC trial, where only approximately 26% of enrolled patients were able to complete or provide data at the ICU discharge 6-minute walk test [10]. The ability for patients to complete functional endpoints and rigorous follow-up for QoL outcomes requires careful consideration when designing future trials. Collection of functional outcomes continues to be a challenge for ICU trials with many patients who are too debilitated to perform many of the functional outcome measures.

Another key issue in ICU pilot trials regarding compliance with new, more complex study procedures (such as handgrip strength and 6-minute walk testing) is that other critical care trials have demonstrated that enrollment of the first one to three patients in each site is effectively a “run-in period” that can be fraught with complexity [21, 22]. These data would indicate that after the second patient is randomized, site protocol violations decrease and treatment effect tends to increase (i.e., becomes more stable toward the true estimate of treatment effect). Thus, in a larger definitive trial, compliance may improve with larger patient numbers enrolled at each site, producing more complex functional and lean body mass outcomes.

Strengths of this study include that we were able to demonstrate a significant separation in the amount of delivered calories and protein between groups with early SPN, particularly in surgical ICU patients. Other key findings include that early SPN did not contribute to any increased risk of infection, as has been hypothesized by past trials [23]. Another strength is the utilization of a more modern, non-pure-soy-oil-based lipid formulation, which may have contributed to the lack of infection risk from SPN in this trial. Recent meta-analyses have shown that lipid formulations reducing soy-based lipid delivery via use of non-pure-soy-oil formulations have lower rates of infection in ICU patients [24].

A key goal of this trial was to attempt to identify a “high nutritional risk” group of ICU patients to target the use of more complex PN delivery and assess the potential benefits of SPN + EN given poor EN delivery worldwide. In this pilot study, encouraging trends toward reduced ICU and hospital mortality were observed only in the BMI <25 subgroup of the SPN + EN arm, and no trend was observed in the BMI >35 subgroup. Thus, it is possible that this strategy of early SPN delivery may have greatest efficacy in patients with lower BMIs and who may have the lowest lean body mass reserve. As neither BMI group was powered to meaningfully look at clinical outcomes, both BMI subgroups should be considered targets of future research and will require further study. In addition, subgroup analysis revealed that patients with the highest ICU admission nutrition risk, as defined by a NUTRIC score of ≥5, appeared to show the largest trend to benefit from SPN. As such, we believe that the future full TOP-UP trial should focus enrollment on patients with a NUTRIC score ≥5 to target, or personalize, early SPN therapy for patients most likely to benefit. Thus, we may have further learned that BMI is not the ideal indicator of nutrition risk in the ICU, but perhaps the NUTRIC score has promise as a better objective measure of nutritional risk [8, 25].

Additionally, a significantly greater increase in calorie delivery was achieved by SPN + EN over EN alone in the surgical ICU patients versus medical ICU patients. As has been previously described [15], surgical ICU patients in our study had a much poorer delivery of baseline EN than the medical ICU patients. Further, the targeted greater than 30% increase in calorie delivery by SPN was also able to be achieved in the surgical ICU group. It is possible these data suggest that a future SPN trial may also be optimally focused on a high-nutritional-risk surgical ICU group, as these patients demonstrate a greater deficit in EN calorie and protein delivery and thus may be more likely to benefit from additional SPN delivery.

Finally, over the last 10 years we have begun to reduce in-hospital mortality following severe sepsis in some countries worldwide [26]. However, the same data also reveal that we have tripled the number of patients going to rehabilitation settings [26]. We also know that up to 40% of mortality within the first year of ICU stay occurs after ICU discharge [27], often due to post-intensive care syndrome (PICS). As a result, many leading experts are calling for future ICU trials to not focus on mortality as the primary endpoint, but rather to focus on QoL [26]. As such, we strived to introduce functional and key QoL indicators in our outcomes, particularly as early protein/calorie delivery may be key in optimizing post-ICU lean body mass and QoL. Our pilot data reveal consistent trends in improvement of functional and QoL endpoints in the SPN + EN group versus EN alone. In particular, trends to improved hospital discharge handgrip strength, 6-minute walk test, Barthel Index, and SF-36 scores were observed in the SPN + EN group versus EN alone. This included a significant improvement in the vitality subscore at 6 months (p = 0.05). Overall, these data are consistent in the direction of benefit for functional and QoL outcomes in patients receiving early SPN, and we believe this deserves further study in the larger TOP-UP trial. Further, given the consistent signal seen in functional and QoL outcomes, we would propose considering a QoL or functional outcome be the primary outcome of a future full-scale SPN trial. For, as many have said, the epidemic of PICS is one that we must address with targeted trials as soon as possible [26, 28].

This pilot trial was undertaken to answer key questions on the feasibility of conducting a multinational, multicenter trial of SPN in low- and high-BMI patients, based on the concept that these patients would most likely benefit from additional calorie and protein delivery in the first week of ICU care. Additionally, compliance and patient ability to complete functional and QoL testing needed to be evaluated. Our data show that the provision of SPN + EN versus EN alone significantly increased calorie/protein delivery over the first ICU week versus EN alone. Further, consistent encouraging trends in hospital mortality, ICU mortality, and QoL and functional endpoints (with no increased infection risk from PN) indicates a full-scale trial of SPN in high-nutritional-risk ICU patients focused on those with a NUTRIC score ≥5 regardless of BMI is indicated and has the potential to change practice by clarifying an objective measure of malnutrition to guide optimal use of SPN. It may also be optimal to focus a future trial in the more poorly EN-fed surgical ICU setting. Assuming we can carefully select sites and address patient ability to complete follow-up functional and QoL data, we propose that this future trial focus on a functional and/or QoL endpoint rather than mortality as its primary outcome.