A pragmatic multi-centre randomised controlled trial of fluid loading in high-risk surgical patients undergoing major elective surgery - the FOCCUS study

Fluid interventions

Patients were randomised to ward-based intravenous pre-operative fluid loading or no pre-operative fluid loading. The pre-operative fluid loading group patients were electively commenced on 25 ml/kg Ringer's lactate solution over the six-hour period in the ward setting before surgery [5]. In the standard fluid regimen no routine pre-operative fluid loading was given. All patients receiving bowel preparation were given 10 ml/kg Ringer's lactate solution in the 12- to 6-hour period before surgery irrespective of trial group allocation as this is deemed to be the best clinical practice. Fluids were not warmed during administration. This meant that fluid loading patients who received bowel preparation would receive 35 ml/kg in the 12 hours before surgery.

All non-protocol fluid prescriptions and other management decisions were made by the clinically responsible surgical team. We did not control or protocolise the time of discharge from hospital. Due to the nature of the interventions no blinding of the interventions was possible.

All participants were followed up daily for one week for major morbidity and mortality, then at hospital discharge and then one, three and six months after surgery for survival and quality of life. The decision to discharge the patient from hospital was made by the caring team with no involvement of the study personnel. Study outcomes measured during hospital stay were measured by study personnel not blinded to the intervention. Outcomes assessed after hospital discharge were measured using postal participant questionnaires.

Outcomes

The primary outcome was the number of days in hospital after surgery [9–13] Secondary outcomes included cost-effectiveness at six months, measured by the Net Benefit statistic, which is calculated using the following equation: ((λ * quality adjusted life year (QALY)) - costs) where λ indicates society's 'willingness to pay' (λ is typically set at £20,000) [24], QALY are calculated using EQ-5D scores and costs included both primary and secondary care costs [25]. Other secondary outcomes, including measures of health status, included changes in health status and quality of life over 6 months after surgery and quality of life at 48 hours, 1, 3 and 6 months after surgery, measured using SF-36 and EQ-5D [25–27]; health care costs including full hospital costs and primary care costs; mortality measured using time-to-event analysis; and the level of major morbidities in hospital using the Post-Operative Morbidity Survey at days 1, 3 and 7 after surgery (POMS) [28] Serious adverse events were assessed by clinical leads defined according to standard definitions, (that is, an untoward occurrence that results in death, is life-threatening, requires hospitalisation or prolongation of existing hospitalisation, results in persistent or significant disability or incapacity, consists of a congenital anomaly or birth defect, or is otherwise considered medically significant by the investigator) [29].

Sample size

The full power calculation is presented in the trial protocol paper [20]. For the fluid comparison, we aimed to be able to detect a 0.5 SD difference in the primary outcome of number of days in hospital following surgery, with 80% power and 5% significance. This resulted in a target sample size of 128 patients. Data on number of days in hospital were available for all patients in the trial.

Statistical analysis

The statistical analysis was based on all people randomised, irrespective of subsequent compliance to treatment allocation. Trial analysis was undertaken using standard methods for two-group comparisons for continuous, binary and time-to-event outcomes using intention to treat principles. All statistical analyses were pre-specified in a Statistical Analysis Plan which was agreed upon before the analysis was undertaken. The primary analysis was adjusted for the minimisation variables using analysis of covariance. A significance level of P < 0.05 was considered as evidence of statistical significance for the primary outcome and confidence intervals presented. An a priori secondary subgroup analysis was investigated through tests for interaction and included patients with high cardiac risk [21]; patients with high-grade functional limitation due to heart failure and type of surgery. Stricter levels of statistical significance (P < 0.01) were sought, reflecting the exploratory nature of these subgroup analyses.

Economic analysis

A within trial economic analysis was conducted from the UK National Health Services' perspective. Discounting was not used since the follow-up was only six months. Benefits in the economic analysis were reported in terms of QALYs estimated from the responses to the EQ-5D questionnaire at five time points - baseline (before surgery), 48 hours, 1 month, 3 months and 6 months after surgery. These responses were converted to utility values using the EQ-5D social tariff, which has been estimated from a representative sample of the UK population [25]. The utility scores obtained at these time points were transformed into QALYs using the area under the curve method by assuming linear extrapolation between subsequent data collection time points.

Patient specific data for the use of NHS resources were retrieved using patient case notes (post-operative inpatient length of stay and outpatient visits) as well as six months' questionnaires to participants (primary care contacts and medicines consumed). Unit costs were retrieved from a number of publicly available sources [30–32]. Intervention cost for base case analysis considered no extra time for pre-operative fluid loading, 30 minutes doctor's time and 42 minutes nurse's time.

Sensitivity analyses were performed for cost and QALYs by alternatively excluding from the analysis participants with 5% and 10% highest and lowest total cost or QALYs. Moreover, the intervention cost was adjusted by adding 12 hours inpatient time in the surgical ward. This scenario would be relevant if normal practice (in the absence of fluid loading) was to conduct surgery on the day of admission. In this situation, patients who receive the fluid intervention would be required to be admitted to hospital earlier to allow time for pre-operative fluid loading to take place. Furthermore, alternative bed-ridden EQ-5D scores were used as baseline utility scores for all participants (that is, -0.402). Finally, missing data were not regarded a priori as an issue for the base case analysis. However, this was tested by imputing mean EQ-5D and mean costs for those categories with missing data within each study group and this was further explored using multiple imputation techniques. A base case analysis and sensitivity analyses were performed. For every analysis, differences in mean total cost and mean QALY data were bootstrapped (1,000 repetitions) adjusting for minimisation variables as well as baseline EQ-5D score [33].

Trial oversight

The Multicentre Research Ethics Committee for Scotland approved the study (Ref 04/MRE10/76). The trial was registered in a public trials registry (registry number ISRCTN32188676). The trial was overseen by a TSC with an independent Chair and an independent DMC reviewed accruing data at regular intervals.

Role of the funding source

The study sponsor and funding source had no role in the collection, analysis and interpretation of data, or in the writing of the report, or in the decision to submit the paper for publication.

The fluid intervention

The fluid loading group received a median of 1,875 ml of Ringer's solution in the pre-operative period compared to a median of 0 ml in the control; this correlates well with the 25 ml/kg protocolised target pre-operative fluid load. This fluid was delivered over a six-hour period before surgery. With clinical practice changing towards day of surgery admission in many countries, future studies in this area may find it difficult to deliver this intervention and may choose to consider testing whether shorter periods of fluid administration are appropriate; for instance, delivery of the fluid intervention in a two-hour period rather than six. It is important to identify that our sensitivity analysis suggests that the intervention is still highly likely to be cost-effective even if patients, who would normally be admitted on the day of surgery, had to be admitted to hospital up to 12 hours earlier (that is, the night before surgery) to receive the intervention.

With regard to the choice of fluid used, we chose Ringer's solution due to our desire to select a crystalloid and use a balanced solution. We believe this was the appropriate choice and would be unlikely to change this selection in a future study. In this study we specifically documented cardiac adverse events as being potentially important morbidities related to the fluid intervention but there was no evidence of increased cardiac adverse events associated with fluid loading and, indeed, there was a reduction in overall adverse events in the fluid group. We do not have direct data to explain why there were fewer gastro-intestinal and infectious adverse events in the fluid group but this may be explained by an improvement in cardiac output and oxygen delivery that would be expected in patients having fluid loading [3–7]. This may improve tissue (including gut blood flow) and reduce anastomotic breakdown and reduce tissue infection. Some centres use cardiac output guided therapy to guide fluid therapy in such patients and it will be important to test this fluid intervention combined with, or compared to such peri-operative strategies. Despite the lack of clinical evidence of fluid related complications in this study the use of more invasive monitoring in future studies may allow a more full investigation of this issue.

What this study adds

We hypothesised that simple pre-operative fluid loading would represent a straightforward and cost-effective way to shorten stay and improve outcome after major high risk surgery. We designed the intervention to be as simple and pragmatic as possible by delivering a fixed "dose" (25 ml/kg) of intravenous Ringer's lactate solution in the six hours before surgery in the ward environment without the requirement to site, monitor or target the complex cardiovascular targets that cardiac output devices allow. This intervention was designed to be easily protocolized for clinical practice and be delivered by non-medical staff to further enhance its utility.

We have demonstrated a trend towards a reduction in hospital length of stay after surgery and that the fluid intervention is likely to be cost-effective. This is supported by concomitant observed reductions across both the high dependency and ward length of stay, both of which are reduced in the fluid loading group. We also demonstrated that fluid loading was associated with short term reductions in adverse events as well as improvements in longer term (six-month) outcomes, such as the effectiveness of care, reductions in health care costs and improved cost-effectiveness as well. All these secondary outcomes show effects in the direction of benefit for the intervention and help us understand the contributing factors that are associated with improvements in outcome.

The mechanisms by which this fluid intervention appears to reduce hospital length of stay are not entirely clear from the results of the study. We know that the fluid loading group received the pre-operative fluid intervention as per protocol and received a significantly greater amount of fluid before surgery commenced. This group was also found to have received more fluid by the end of surgery than the fluid control group. Therefore, these patients received more fluid, earlier, than controls and much of this fluid was administered before the surgical insult. Pre-operative optimisation studies often demonstrate an increase in cardiac index and oxygen delivery with fluid loading alone, which is further augmented by inotropic support to achieve supranormal oxygen delivery targets [3–7]. We would hypothesis that pre-operative fluid loading improves cardiac output and oxygen delivery, but to levels below "supranormal" levels [3, 5, 34], and this is associated with improved organ perfusion and function [3, 5, 34], fewer surgical complications [35] and fewer adverse events, lower post-operative morbidity, and these factors contribute to the shorter length of stay in hospital after surgery. The magnitude of reduction in hospital length of stay is similar to that seen in other optimisation studies [5, 9–11]. It should be noted that this study, and the pre-optimisation literature, appears to contradict the evidence base for intra-operative fluid restriction, which has appeared more recently [15–19]. However, the apparent discrepancy between these two bodies of evidence may be less difficult to reconcile than it appears. There are seven randomised studies in the literature on fluid restriction and of these only three show benefit [36]. Evidence from the first study of fluid restriction strategies suggests that restricting day of surgery fluid intake from approximately 6,200 ml (of which 5,388 ml were IV) to approximately 3,700 ml (of which 2,740 ml were IV) may be beneficial in terms of complications [15]. Whereas, the current study, and many of the studies cited in the pre-operative optimisation literature, utilise peri-operative fluid loads of 3,000 to 4,000 ml [5]. In studies of fluid restriction, a range of a "liberal intra-operative fluid regimens" from 2,750 to 5,388 ml compared with 998 to 2,740 ml for the "restrictive fluid regimen"[36]. This may suggest that "restrictive fluid regimens" may not actually differ that significantly from optimization strategies in terms of fluid volume. The difference that may explain these two apparently contradictory strategies may relate to either the timing of the fluid administration (early preoperative fluid loading being beneficial and late post-operative fluid overloading being harmful) or related to the achievement of "supranormal" targets. In the Noblett study, which utilised an intra-operative fluid optimisation regime, a significant majority of the fluid administration occurred in the first 40 minutes of surgery [9]. Therefore, there could be an argument that we should target early (pre- and early intra-operative) fluid loading/optimisation and then move to target late (end of surgery) active fluid restriction to avoid post-operative fluid overload and late complications. These bodies of evidence may, therefore, be complementary and not contradictory [37].

If this reduction in hospital length of stay can be replicated in a larger study then this finding will have a major impact on service delivery and resource allocation. From a patient's perspective this could allow them to get back to their home environment earlier. The intervention may also lead to less adverse events and other acute morbidities, which would clearly have direct patient benefit. From a clinical practice perspective, this intervention would be simple, inexpensive and is likely to be cost-effective as well as being relatively easy to implement in acute hospitals. There would be little educational support required for implementation as the knowledge and skills are already present in surgical teams in all acute hospitals. The intervention could be easily protocolized and nurse-led to allow reliable and reproducible delivery in practice. This reduction in length of stay would lead to significant direct savings in clinical budgets and allow the re-allocation of these resources. The simplicity and low cost of this intervention may make it more attractive than other optimisation strategies [3–14].

Strengths and weaknesses

One of the main strengths of this trial is the importance and simplicity of the clinical research question. We believe this is the first randomised controlled trial of fluid loading in high risk major surgery. The multi-centre nature of the study adds to the generalizability of the study results. The simplicity and low cost of the fluid intervention are key factors making this intervention comparatively simple to implement into surgical practice internationally. An integral part of the study was a prospective cost-effectiveness analysis that is unique in this clinical field and still uncommon in randomised clinical trials in acute care. The economic evaluation has been conducted using the best available methods, including an extensive and detailed costing approach. This analysis suggests that the fluid loading intervention is highly likely to be cost-effective, adding greatly to the importance and impact of the study. However, this study was not powered to detect a difference in cost-effectiveness between groups. Therefore, it is perhaps not surprising that the evidence on cost-effectiveness falls short of conventional levels of statistical significance. Consideration has, therefore, been given to the balance of probabilities when drawing conclusions about cost-effectiveness. There was some minor imbalance between groups with regard to baseline characteristics, such as age and number of patients undergoing abdominal surgery with bowel preparation. A slightly higher number of patients in the fluid intervention group received ICU care in the early post-operative period and this could be argued to introduce a bias in favour of the intervention group by improving the care delivered to this group. This difference is believed to have occurred by chance and not be driven by clinical issues, including no increase in the requirement for post-operative ventilation and no major differences in surgery performed. The definition of high risk status varies between studies and we chose to use one of the most widely used RCRI [21]. Despite the precise application of this score we still recruited patients with American Society of Anesthesiologists (ASA) grades of 2 and, therefore, apparent low to moderate risk patients. This does not represent a misclassification of patients but does mean the study group is moderate to high risk by some clinical risk criteria. We believe the external generalizability of this study is high. Such high risk patients undergo major surgery in all countries and they have a significant morbidity and mortality. This intervention is applicable and implementable in all developed countries but many countries have day of surgery admission. Since the loading dose of fluid is relatively modest, it does not seem unreasonable to think this intervention could be delivered in the two to three hours before surgery for patients admitted on the day of surgery. Further, our sensitivity analysis suggests that it would still be highly likely to be cost-effective to admit such patients up to 12 hours before surgery (the night before) to receive their fluid intervention. Our median total length of stay for this surgery seems to be broadly in line with UK averages and other studies [14, 38], although above those lengths of stays seen in "fast track surgery studies" [15, 18, 36]. A further significant weakness was that our final recruitment total was short of our desired sample size (111 instead of 128), which meant that we were slightly underpowered to detect our a priori proposed minimally important difference of 0.5 SD (111 patients gave 80% power to detect 0.54 SD change rather than 0.5 SD).