Morgenstern, J. The RePHILL study: Is saline the fluid of choice in trauma?, First10EM, April 25, 2022. Available at:
Trauma patients aren’t bleeding salty water, so why would you give them intravenous salty water? It doesn’t make any sense. Obviously, if they require fluid resuscitation, it should be with the fluid they are losing. So when we compare blood products to normal saline in an RCT (the RePHILL study), we are bound to see a massive benefit, right? RIGHT?!
The RePHILL trial: Crombie N, Doughty HA, Bishop JRB, Desai A, Dixon EF, Hancox JM, Herbert MJ, Leech C, Lewis SJ, Nash MR, Naumann DN, Slinn G, Smith H, Smith IM, Wale RK, Wilson A, Ives N, Perkins GD; RePHILL collaborative group. Resuscitation with blood products in patients with trauma-related haemorrhagic shock receiving prehospital care (RePHILL): a multicentre, open-label, randomised, controlled, phase 3 trial. Lancet Haematol. 2022 Apr;9(4):e250-e261. doi: 10.1016/S2352-3026(22)00040-0. PMID: 35271808 ISRCTN62326938
This was a multi-center, open label, RCT from four prehospital critical care services in the United Kingdom.
Adult trauma patients with hypotension (systolic blood pressure less than 90 or lack of a radial pulse) believed to be due to traumatic hemorrhage.
Exclusions: transfusion of blood products before enrollment, refusal of blood products, isolated head injury, pregnancy, and prisoners.
Blood products as the primary resuscitation fluid. Up to 4 units total. 2 units of packed red blood cells and 2 units of lyophilised plasma. The protocol suggested alternating between PRBCs and plasma.
Normal saline as the primary resuscitation fluid. Up to 4 boluses of 250 mL.
In both groups, fluids were given until hospital arrival, or until hypotension resolved.
The primary outcome was a composite of inhospital mortality and lactate clearance at 2 hours.
The trial was stopped early because of COVID without hitting any of their prespecified endpoints. They enrolled 432 participants of a planned 490. Perhaps more concerning, the median follow up was only 8 days, when the protocol called for 30 day outcomes.
The participants were primarily white (62%) males (82%) with a median age of 38. Most (62%) were from motor vehicle collisions, with some stabbings (16%), falls (12%), but – and this might be surprising to American readers – almost no gunshots (2%). These were very sick trauma patients, with 12% presenting in traumatic cardiac arrest.
There was no difference in the primary outcome, which occurred in 64% of the blood group and 65% of the saline group (ARR 1.01, 95% CI 0.88-7.17, p=1.0).
There also wasn’t a difference in either of the components of the composite outcome. Mortality was 43% in the blood group and 45% in the saline group (p=0.57). Failure to clear lactate occurred in 50% of the blood group and 55% of the saline group (p=0.33). (Many patients are obviously counted twice in the primary outcome, which is yet another significant problem with using composite endpoints.)
Adverse events were the same in both groups. Transfusion related adverse events were also identical, at 7% in both groups.
I am never a fan of composite outcomes, but the composite outcome used here is truly ridiculous. They combined mortality with lactate clearance; a true patient oriented outcome with a lab value with very questionable associations with any outcome. Decisions like this are why peer review should occur before research starts, when changes can still be made, rather than after results are submitted to a journal.
Mortality was incredibly high in this cohort, with more than 40% of the patients dying. In my mind, the sickest patients are exactly the patients who would need blood instead of saline. (If mortality was <5%, you might have argued that these patients were going to do fine no matter what we gave them.) However, the flip side of that argument could be that this trial may have included a number of patients that were simply too sick to benefit from blood, which would also bias the trial towards the null hypothesis.
Although the comparison group received normal saline, it must be emphasized that this was not your old school trauma resuscitation where multiple liters of saline were dumped into the patient. The maximum in either group was a liter bolus. Even the most blood focused trauma doc is okay with one liter of saline. The small volume of saline may have mitigated against harmful impacts of saline (if such a thing exists). However, I think this probably actually biases the trial against the saline group. Crystalloid rapidly redistributes away from the intravascular space, so a liter of saline does not provide the same intravascular volume as a liter of blood product. Clinically, no one would ever consider 1 liter of saline to be the equivalent of 4 units of blood products. On the other hand, I wouldn’t want to give a patient 4 L of saline prehospital, so I am not sure how I would have designed the trial personally to make this a fair comparison.
On the other hand, if you think that saline is poisonous, the fact that both groups had received an average of 430 mL of saline before trial enrollment may be troubling to you. That is essentially exactly as much blood as these patients received (443). This crossover is made worse by the fact that once all 4 units of blood were given, the blood group was allowed to receive more saline. (I don’t see it stated anywhere in the manuscript how often this happened.)
The trial was unblinded, which is totally understandable given the nature of the intervention, but certainly raises the possibility that the groups were treated differently in some way that is not adequately recorded. However, allocation concealment was excellent, so there should not have been significant bias in selecting who was included in the trial based on which intervention they were expected to receive.
I have mentioned it numerous times in the past, but sample size calculations might be the least scientific part of almost every scientific paper. This entire trial is based around the assumption that they would see a 10% absolute benefit. Even when including the irrelevant lactate clearance in the primary outcome, we just never see benefits that big. I don’t think anyone working in emergency medicine would have guessed that using blood instead of saline in the prehospital environment was going to make a 10% difference. A 1-2% difference in mortality would be possible, and would still have been important, but of course would have required a much larger trial. (They do provide some post-hoc calculations, and state that even with extensive further recruitment, such as a sample size of 5,000, they would still expect to observe a null result). This highlights a fundamental issue with the way we fund medical research. Almost none of our studies are powered appropriately. To do so would be expensive, but not as expensive as running hundreds of trials that were doomed to fail because their methods were flawed from the outset.
Perhaps I shouldn’t be so critical about the search for a 10% reduction, considering the last major trial looking at prehospital fluids in trauma – the PAMPer trial – showed a 9.8% absolute reduction in mortality. (Sperry 2018) On the other hand, although those results were exciting, they also left many of us with the feeling that they were probably too good to be true. I said that results would need to be replicated. The RePHILL trial gave 2 units of plasma, exactly like the PAMPer trial, so unless you think the addition of packed red blood cells could result in a reversal of that 10% mortality benefit, this trial acts as a partial replication of PAMPer, and supports the viewpoint that the PAMPer results were too good to be true.
The rate of transfusion-related adverse events was identical between the two groups. That is, 7% of the saline group had a transfusion related adverse event. If placebo causes as many transfusion reactions as blood products, I shouldn’t have to get consent from my patients when transfusing blood. Either that, or like I have been preaching for years, clinical trials are just not very good at identifying harms. Obviously, transfusions cause adverse events, so the identical rate of transfusion reactions should provide us with a valuable lesson when interpreting other research. (One of the common arguments in favour of novel COVID therapies like Paxlovid is that we aren’t seeing more adverse events in the trials. That is unlikely to represent the true clinical impact of these drugs.)
Despite some very strong physiologic reasons that blood should be preferred in trauma patients, this RCT failed to demonstrate any benefit over normal saline. I still would never preferentially use saline if I thought a patient needed blood, but this trial should reassure you if you need to give a small crystalloid bolus while waiting for blood to arrive.
Crombie N, Doughty HA, Bishop JRB, Desai A, Dixon EF, Hancox JM, Herbert MJ, Leech C, Lewis SJ, Nash MR, Naumann DN, Slinn G, Smith H, Smith IM, Wale RK, Wilson A, Ives N, Perkins GD; RePHILL collaborative group. Resuscitation with blood products in patients with trauma-related haemorrhagic shock receiving prehospital care (RePHILL): a multicentre, open-label, randomised, controlled, phase 3 trial. Lancet Haematol. 2022 Apr;9(4):e250-e261. doi: 10.1016/S2352-3026(22)00040-0. PMID: 35271808
Sperry JL, Guyette FX, Brown JB, et al. Prehospital Plasma during Air Medical Transport in Trauma Patients at Risk for Hemorrhagic Shock. The New England journal of medicine. 2018; 379(4):315-326. PMID: 30044935