Not so fast with the capillary refill guided resuscitation (ANDROMEDA-SHOCK-2)

As far as I can tell, despite talking about the paper widely at conferences, I never included a write up of the original ANDROMEDA-SHOCK trial on First10EM. (Hernández 2019) (There is a massive file of all the topics I want to cover, and would cover if this was a job rather than a hobby. I assume it just got lost in there.) As a reminder, that trial showed that clinical management of septic shock patients based on capillary refill time was not statistically different from management guided by trending lactates, although there was some optimism because the point estimate for all cause mortality was actually 9% better. The ANDROMEDA-SHOCK-2 trial goes a step further, looking at the use of a bundle of interventions targeting capillary refill time, to see whether outcomes can be improved in septic shock. 

The paper

ANDROMEDA-SHOCK-2 Investigators for the ANDROMEDA Research Network, Spanish Society of Anesthesiology, Reanimation and Pain Therapy (SEDAR), and Latin American Intensive Care Network (LIVEN); Hernandez G, Ospina-Tascón GA, Kattan E, Ibarra-Estrada M, Ramasco F, Orozco N, Ramos K, Aldana JL, Ferri G, Hamzaoui O, De Backer D, Teboul JL, Vieillard-Baron A, Petri Damiani L, García-Gallardo GA, Morales S, Carmona Garcia P, Mendez R, Hernandez-Gilsoul T, Pérez-Nieto OR, Olea Vielba C, Ramos S, Dominguez D, Bruna M, David S, Wendel-Garcia PD, Galiana-Ivars M, Myatra SN, Messina A, Cecconi M, Pozo M, Amthauer M, Higuera E, Al Duhalib Z, Rico-Feijoo J, Ferrer-Gómez C, Pérez-Carbonell A, Martinez-Castro S, Redondo Calvo FJ, Vives M, Sanchez HF, Bilbao I, Fernandez P, Al-Fares A, Benitez-Cano A, Gonzalez C, Reyes LF, Rodriguez-Guillen JH, Cristino AV, Pendino JC, Ortiz G, Alonso-Gonzalez MC, Murias G, Aguirre-Ávalos G, Hernández L, Calderón Barajas ZA, Zarragoikoetxea I, Monnet X, Goury A, Mendonça Dos Santos T, Vallecilla L, Martins de Lima L, Sady E, Alegria L, Ostermann M, Bakker J, Biasi Cavalcanti A. Personalized Hemodynamic Resuscitation Targeting Capillary Refill Time in Early Septic Shock: The ANDROMEDA-SHOCK-2 Randomized Clinical Trial. JAMA. 2025 Oct 29:e2520402. doi: 10.1001/jama.2025.20402. Epub ahead of print. PMID: 41159835  NCT05057611

The question

In adult patients with septic shock, can death, duration of vital support, or hospital length of stay be improved by a “personalized hemodynamic resuscitation protocol targeting capillary refill time”?

The methods

This is an open label, multicentre RCT.

Patients

Adult patients with septic shock. (Suspected or confirmed infection, plus a lactate greater than 2 mmol/L, and the need for norepinephrine to maintain a MAP over 65 after at least a 1L IV fluid bolus). 

Intervention

A personalized hemodynamic resuscitation protocol targeting capillary refill time (CRT-PHR). 

Not to give away the conclusion, but given that I am not going to be using this protocol myself, especially in the emergency department, I don’t think it is worth going through in precise detail. Their protocol was focused on normalizing the capillary refill time, and used sequential bedside assessments of pulse pressure, diastolic pressure, fluid responsiveness, and echocardiography to guide fluids, vasopressors, and inotropes. 

Comparison

Usual care.

Outcome

The primary outcome was (I think unfortunately) a composite of all cause mortality at 28 days, duration of vital support (itself a composite of vasoactives, mechanical ventilation, and renal replacement therapy), and hospital length of stay. They made this a hierarchical outcome, but as I will discuss below, I am not sure that adds much. 

The results

They enrolled 1501 patients, of whom 1467 are included in the final analysis. The groups look pretty similar at baseline. Their CRT-PHR protocol did result in more patients with a corrected capillary refill within 6 hours (85.9% vs 61.7%). The CRT-PHR group received somewhat less IV fluid (595 mL vs 847 mL), but more dobutamine (12% vs 5%).

They analyzed the data using “win ratios”, and for the overall composite the CRT-PHR had more wins (48.9% vs 42.1%, p=0.04).

There was no difference in mortality (26.5% vs 26.6, p=0.91).

There was a statistical decrease of 1 day in mean vital support free days (16.5 vs 15.4)

There was no statistical difference in hospital length of stay (15.3 vs 16.2).

Of course “viral support free days” is itself a composite outcome. They don’t report on the components of this outcome in the manuscript, but based on the supplementary material, there doesn’t seem to be any real difference in the use of mechanical ventilation or vasopressors, and so the entire difference in this trial might come down to a 1 day difference in renal replacement therapy.

My thoughts

Although I would have designed the trial differently myself, this is an excellent effort, asking an interesting and important question, and I was excited to see it published (much like the original ANDROMEDA SHOCK trial). That is especially true when you realize that this trial is not funded by any large organization. Each researcher had to raise funds at their local hospital to be involved in the trial, so this is truly an effort of passion, and represents a tremendous accomplishment.

As an aside: I think there are many problems with the way that science is currently done in medicine. Perhaps the biggest is represented by the role that I play. Critically appraising science after the science is done sort of sucks. Obviously it is important to help clinicians understand the research that is being done, but the critique is happening at the wrong time. The best time for critical appraisal and peer review is during the design of the trial. In the ideal world, peer review would occur before the first patient is enrolled, and help to shape trials to eliminate the major critiques beforehand. Ideally, this pre-enrollment peer review would be accompanied by a guarantee that a journal will publish your results, whatever they are, because the quality of science is based on the methodology, not the results. This would prevent a lot of our problems with publication bias, and would probably help improve trials, rather than just using peer review to critique them after the fact. 

Although there will be some excitement for this trial, this data does not seem practice changing to me. You will hear a lot of differing opinions about this trial, and that is one of the problems with using a composite outcome: the results are often confusing and difficult to interpret. However, without getting into any fancy discussions about their methodology, I think the most important point to understand is that this was an unblinded trial, and the only objective outcome (which is also the most patient important outcome) was completely unchanged. There is no change in mortality. That is the real headline for this trial. 

Some people will undoubtedly think that a 1 day difference in vital support will warrant this more intensive 6 hour protocol. Although clearly important for systems and financial reasons, I am not sure a one day difference in vital support matters much to patients if function, disposition, and mortality are all identical. The fact that this is a composite outcome within a composite outcome makes it even harder to comment on. They don’t include the specifics in the paper, but based on the supplement, I think the only difference is in renal replacement therapy, which is probably the aspect of vital support patients care least about. (A one day difference in mechanical ventilation is much more clearly patient oriented.) Outside of financial reasons, I honestly don’t know if anyone should care about a one day difference in renal replacement therapy. I can imagine related clinical outcomes that we might care about. Is there a difference in long term renal outcomes; in the use of long term dialysis? Those are patient important outcomes, buy this trial doesn’t mention them.

One major problem with this paper, and composite outcomes in general, is how jumbled it makes the results. I have read the paper through twice, and it still isn’t perfectly clear to me which outcomes actually changed. How are we supposed to present this data to patients? How do we translate these win-ratios into information that the standard clinician can understand?

However, far more important than quibbling about whether the components of the composite within a composite are truly patient oriented is the question of whether they are real. This is an unblinded trial. Every component of the ‘vital support’ composite is entirely subjective. This composite doesn’t actually represent patient outcomes. It is a collection of treatment decisions being made by a treating team who is aware of the study and the group that the patient is assigned to. That results in an incredibly high risk of bias.

Some of that bias is easy to imagine. The CRT-PHR protocol required doctors to go to the bedside more often, and when you are at the bedside more often you are more likely to discontinue or change therapy sooner. Perhaps the protocol adds nothing at all to a simple rule that clinicians need to recheck their critically ill patients every 30 minutes? This obvious bias probably cannot fully explain these results, seeing as the protocol only extended for 6 hours and we are seeing a 24 hour difference in vital support free days. However, it provides an example of how easy it is to bias subjective outcomes in open label trials. 

The bias doesn’t mean that the reduction in ‘vital support’ is false, but it does dramatically reduce my confidence. If there had been a change in the more objective mortality, I could see changing practice while waiting for more research. However, when you combine the high risk of bias with the muddy composite outcome of questionable patient importance, the opportunity costs, and the potential for harm, I think it is very clear that this is not a practice changing finding. This is a finding that should be noted by researchers, and targeted for follow-up studies. 

The counter argument might be that these assessments are non-invasive and make physiologic sense based on previous research. You could argue that you don’t see a big risk of harm (although the protocol does use agents that can cause harm, and harms were not actually measured or reported in this trial). When it comes to sepsis, it is quite possible that the pendulum has swung too far. Manny Rivers taught us not to ignore these patients. However, after systemically examining all of the components of his bundle and finding no benefit, the impetus for aggressive sepsis care has somewhat diminished. (The failure of a bundle of aggressive sepsis care in future high quality research is perhaps a bit of an omen for this trial.) Now that we aren’t routinely placing central lines and flooding these patients with fluids, I can imagine that some of these patients are being subtly neglected. (Just the fact that you write fewer orders for sepsis patients means that nurses spend less time at the bedside). Therefore, I could imagine that people will look at this paper and argue that we need to get back to spending more time at the bedside of septic shock patients. That might be true, but it doesn’t mean that this specific protocol is the best method forward. The now abandoned Manny Rivers protocol had a massive mortality benefit, whereas all we are seeing here is a slight decrease in renal replacement therapy. Considering that we are still struggling with mistaken mandates from the Manny Rivers era, I think history should be a huge caution against being overly enthused about these results. 

I think I need to come back to that point about harms. It is impossible to judge the harms and benefits of an intervention when harms are not adequately measured. This is a well known issue in medical research. Harms are consistently under-reported. I do not see any measure of adverse events or harms listed among the outcomes in this trial’s protocol. It is impossible to compare harms and benefits when harms are not measured. 

We also need to consider cost. A 24 hour reduction in renal replacement therapy (if real) is obviously a financial win, but it is important to consider the upfront opportunity cost. Every minute you spend engaging in this aggressive bundle is a minute taken away from other patients in the department. If there is a proven benefit, that cost is obviously worth it. I am not sure we are there based solely on this study. 

Some of the interventions used make perfect sense, but a major problem with studies of bundles of care is that it is impossible to know which parts of the bundle matter and which don’t. It is impossible to know if you might get some benefit by only using some of the interventions. In fact, it is possible that some of the interventions are harmful, but just outweighed by benefits from other aspects of the bundle. (See the recent OPTRESS study for an example of potential harms from higher MAP targets in septic shock.) Again, turning to history, the last time we dissected one of these bundles of care for sepsis, we found that none of the individual components improved outcomes at all. If anything mattered, it was just paying attention to the patient. Of course, that didn’t stop us from dumping exorbitant amounts of time and money into the bundles. History cautions. (If you want to get a sense of the significant problems that RCTs of bundles present for interpretation, listen to the EMCrit interview with this study’s lead authors, because there was debate about almost every aspect of this bundle.)

I am obviously not that enthusiastic about these results, but there is one bias to consider that could mean this trial is underselling the importance of this personalized resuscitation protocol: contamination. This was a very enthusiastic group of researchers. They managed to pull off an incredibly complicated study without external funding. They obviously believed in what they were doing. The research involved a lot of training on their somewhat complicated 6 hour resuscitation protocol. As a result of this education and enthusiasm, physicians trained in these techniques are likely to use them even when their patient is randomized to “usual care”. The closer usual care was to the protocol, the less opportunity there is to see a benefit. I would be worried about overly nihilistic interpretations of this trial. I already perform regular assessments of capillary refill in my shock patients. I am already taking some of the actions described here. Rejection of this specific 6 hour protocol does not give one permission to ignore the septic shock patient. 

As for the more nerdy stats stuff, we have seen this type of hierarchical win ratio study before. The most recent I remember was the PEERLESS study. At the time, I wrote that the hierarchical ranking made some sense to me, but by the time I recorded the podcast on BroomeDocs I was more skeptical, and the longer I think about this the more I worry that this approach is misleading. It seems good at first glance, but when you think about it deeply, it makes far less sense. There is pretty wide agreement that composite outcomes are bad. My sense is that although this is a better approach than a standard composite outcome, we might just be polishing a turd.

At the end of the day, the primary outcome is just a composite. It is no different than any other composite. It has all the problems of any other composite. You get a single number, presented as a win ratio, but without the secondary outcomes, it would be impossible to know which competents of the composite were driving the outcome. It is described differently, but when you read the abstract and conclusion, this composite outcome is presented as a single outcome exactly like any other composite outcome. I think it is just as problematic. All of the clinical information from this study comes from secondary outcomes. (I expect Ken might add some nuance to this point when we discuss it on the SGEM, but I can’t see any real benefit of this hierarchical composite over a standard composite.) 

In my mind, there is really no reason to use a composite outcome. If you really believe that there is a hierarchy of important outcomes in your trial, look at the outcomes independently, and rank them hierarchically. You don’t need to combine them together. This hierarchy does not turn all 3 outcomes of the composite into co-primary outcomes. It is still just a composite with 3 secondary outcomes. The primary outcome is still just a composite that combines dramatically different outcomes, with different values, into a more difficult to interpret pile. 

I think the mistake being made is a misunderstanding about why we dislike secondary outcomes in trials. There are two main reasons that researchers and appraisers downplay the value of secondary outcomes: researcher degrees of freedom and statistical power.

The first problem with secondary outcomes is that studies usually just include a very long list, and often look behind the scenes at an even longer list of outcomes that aren’t registered. This expands researchers’ degrees of freedom. They get to cherry pick secondary outcomes to talk about (which are always the ones that happen to be statistically positive). However, this concern completely disappears if you just register all your outcomes in a clear hierarchy before the trial starts. For this trial, the primary outcome should have been morality. And then the single registered secondary outcome could have been vital support (although that is still a composite). And then the single registered tertiary outcome could have been hospital length of stay. You can continue down a list of as many outcomes as you please, without issue, as long as they are registered in order of importance. This completely eliminates any concerns about researcher degrees of freedom, because before the trial starts they are registering exactly which outcomes are most important. This accomplishes exactly what the researchers were trying to accomplish here, without the mess of the composite. 

The second problem with secondary outcomes is that when you are making multiple comparisons in a study, you are supposed to adjust your statistics to account for multiple comparisons. I think this is the biggest problem with the way they designed their composite here, because they are still using the same very lax p value of 0.05 for all comparisons. If instead of the hierarchical composite, you registered the primary, secondary, tertiary outcomes as I described above, you could then also appropriately adjust your statistics for multiple comparisons. In this case, the mortality comparison would use the standard p of 0.05, but the secondary outcome would use a more strict p value, and the third outcome would have to use an even more strict p value. This is not an approach I have seen used (perhaps because lax p values on secondary outcomes really helps research seem positive, which drives publication value in our broken medical science environment), but I think it is the best approach across the board.  

In my mind, there was no value in combining these outcomes into a single composite, even when you add the hierarchy. The composite only obscures the underlying truth. It confuses the data. It mixes subjective outcomes with objective ones in an unblinded trial. It mixes outcomes that patients really care about with outcomes that they probably don’t, and despite ranking them hierarchally, it still lumps all the data together into a single win ratio. My sense is that this statistical method does not help us get closer to the truth, and does not make the data easier to understand, and is therefore similar to all composite outcomes – hierarchical or not. 

The longer I sit with the stats, the more problematic they seem. Yes, there is a positive primary outcome, but it is a composite. The most important part of that composite (mortality) is unchanged. The component of the composite that is statistically different is itself a composite, and when you break that apart, the component that is probably most important (time on mechanical ventilation) is also unchanged. We have a composite embedded within a composite, and none of the most important outcomes are changed. The only change seems to be a secondary component of a secondary outcome. Add the fact that this is a subjective outcome in an unblinded trial and the clinical take away seems pretty clear: this is not a protocol that should be implemented without further study. 

Bottom line

This multicentre RCT demonstrated an increase in win ratios for a composite outcome, driven primarily by a reduction in renal replacement therapy, by using a personalized hemodynamic resuscitation protocol targeting capillary refill time. It is promising research, but because of its open label nature, the subjective nature of the changed outcomes, and the lack of change in the one objective and truly patient oriented outcome (mortality), it should not currently change practice.

Of course, whether something is “practice changing” depends greatly on your current practice. This trial does not give you permission to leave septic shock patients unobserved in a back corner. These patients clearly need resuscitation, and many of the components of the protocol used here are parts of what I would consider usual care. Your resuscitation needs some target, and I have been using a combination of cap refill and MAP since the original ANDROMEDA SHOCK paper. I am not giving a fixed quantity of fluid, but instead trying to monitor response to therapy and gauge fluid responsiveness. Most patients in shock are getting an echo at some point, especially if they are not rapidly responding to early resuscitation. I don’t think that this protocol needs to be followed, and I think there are harms and opportunity costs that could arise if it is pushed too aggressively, but it will be up to each clinician to think about their current approach to septic shock to see if there is anything they can learn from what the world’s experts are recommending.

Other FOAMed

EMCrit has 2 episodes out on this paper:

• EMCrit 411 – You Need to Understand the Andromeda-Shock-2 RCT for Septic Shock
• EMCrit Wee – ANDROMEDA-SHOCK-2 Explosion – A Discussion with the Lead Investigators

I will be recording an episode of the SGEM with Ken Milne soon, so keep your eyes out for that.

Evidence based medicine is easy

The EBM bibliography

Evidence based medicine resources

EBM deep dives

References

ANDROMEDA-SHOCK-2 Investigators for the ANDROMEDA Research Network, Spanish Society of Anesthesiology, Reanimation and Pain Therapy (SEDAR), and Latin American Intensive Care Network (LIVEN); Hernandez G, Ospina-Tascón GA, Kattan E, Ibarra-Estrada M, Ramasco F, Orozco N, Ramos K, Aldana JL, Ferri G, Hamzaoui O, De Backer D, Teboul JL, Vieillard-Baron A, Petri Damiani L, García-Gallardo GA, Morales S, Carmona Garcia P, Mendez R, Hernandez-Gilsoul T, Pérez-Nieto OR, Olea Vielba C, Ramos S, Dominguez D, Bruna M, David S, Wendel-Garcia PD, Galiana-Ivars M, Myatra SN, Messina A, Cecconi M, Pozo M, Amthauer M, Higuera E, Al Duhalib Z, Rico-Feijoo J, Ferrer-Gómez C, Pérez-Carbonell A, Martinez-Castro S, Redondo Calvo FJ, Vives M, Sanchez HF, Bilbao I, Fernandez P, Al-Fares A, Benitez-Cano A, Gonzalez C, Reyes LF, Rodriguez-Guillen JH, Cristino AV, Pendino JC, Ortiz G, Alonso-Gonzalez MC, Murias G, Aguirre-Ávalos G, Hernández L, Calderón Barajas ZA, Zarragoikoetxea I, Monnet X, Goury A, Mendonça Dos Santos T, Vallecilla L, Martins de Lima L, Sady E, Alegria L, Ostermann M, Bakker J, Biasi Cavalcanti A. Personalized Hemodynamic Resuscitation Targeting Capillary Refill Time in Early Septic Shock: The ANDROMEDA-SHOCK-2 Randomized Clinical Trial. JAMA. 2025 Oct 29:e2520402. doi: 10.1001/jama.2025.20402. Epub ahead of print. PMID: 41159835

Hernández G, Ospina-Tascón GA, Damiani LP, Estenssoro E, Dubin A, Hurtado J, Friedman G, Castro R, Alegría L, Teboul JL, Cecconi M, Ferri G, Jibaja M, Pairumani R, Fernández P, Barahona D, Granda-Luna V, Cavalcanti AB, Bakker J; The ANDROMEDA SHOCK Investigators and the Latin America Intensive Care Network (LIVEN); Hernández G, Ospina-Tascón G, Petri Damiani L, Estenssoro E, Dubin A, Hurtado J, Friedman G, Castro R, Alegría L, Teboul JL, Cecconi M, Cecconi M, Ferri G, Jibaja M, Pairumani R, Fernández P, Barahona D, Cavalcanti AB, Bakker J, Hernández G, Alegría L, Ferri G, Rodriguez N, Holger P, Soto N, Pozo M, Bakker J, Cook D, Vincent JL, Rhodes A, Kavanagh BP, Dellinger P, Rietdijk W, Carpio D, Pavéz N, Henriquez E, Bravo S, Valenzuela ED, Vera M, Dreyse J, Oviedo V, Cid MA, Larroulet M, Petruska E, Sarabia C, Gallardo D, Sanchez JE, González H, Arancibia JM, Muñoz A, Ramirez G, Aravena F, Aquevedo A, Zambrano F, Bozinovic M, Valle F, Ramirez M, Rossel V, Muñoz P, Ceballos C, Esveile C, Carmona C, Candia E, Mendoza D, Sanchez A, Ponce D, Ponce D, Lastra J, Nahuelpán B, Fasce F, Luengo C, Medel N, Cortés C, Campassi L, Rubatto P, Horna N, Furche M, Pendino JC, Bettini L, Lovesio C, González MC, Rodruguez J, Canales H, Caminos F, Galletti C, Minoldo E, Aramburu MJ, Olmos D, Nin N, Tenzi J, Quiroga C, Lacuesta P, Gaudín A, Pais R, Silvestre A, Olivera G, Rieppi G, Berrutti D, Ochoa M, Cobos P, Vintimilla F, Ramirez V, Tobar M, García F, Picoita F, Remache N, Granda V, Paredes F, Barzallo E, Garcés P, Guerrero F, Salazar S, Torres G, Tana C, Calahorrano J, Solis F, Torres P, Herrera L, Ornes A, Peréz V, Delgado G, López A, Espinosa E, Moreira J, Salcedo B, Villacres I, Suing J, Lopez M, Gomez L, Toctaquiza G, Cadena Zapata M, Orazabal MA, Pardo Espejo R, Jimenez J, Calderón A, Paredes G, Barberán JL, Moya T, Atehortua H, Sabogal R, Ortiz G, Lara A, Sanchez F, Hernán Portilla A, Dávila H, Mora JA, Calderón LE, Alvarez I, Escobar E, Bejarano A, Bustamante LA, Aldana JL. Effect of a Resuscitation Strategy Targeting Peripheral Perfusion Status vs Serum Lactate Levels on 28-Day Mortality Among Patients With Septic Shock: The ANDROMEDA-SHOCK Randomized Clinical Trial. JAMA. 2019 Feb 19;321(7):654-664. doi: 10.1001/jama.2019.0071. PMID: 30772908