Feb 232019

Jon-Emile S. Kenny MD [@heart_lung]

The truth is balance, but the opposite of truth, which is unbalance, may not be a lie.”

-Susan Sontag


A 49 year old man presents with two days of hemoptysis, right-sided pleuritic chest pain and a few hours of ‘confusion’ according to his teammate in a pick-up hockey league.  One week prior, he had experienced respiratory symptoms and myalgias with fever, but he improved over 5 days.  He has no known medical history and takes no medicine.  On presentation, the patient’s heart rate is regular at 155 beats per minute; his saturation without supplemental oxygen is 84% and he is with notable tachypnea.  His blood pressure is 82/29 mmHg.  He is warm to the touch with rapid capillary refill noted by the medical student.  The patient’s CXR demonstrates a right lower lobe cavity with surrounding infiltrate and large effusion and his blood work is notable for acute kidney injury, leukocytosis and a venous lactate of 7.0 mmol/L; the resident in the emergency department starts the patient on high-flow nasal cannula, immediately orders linezolid and places a right-sided subclavian triple lumen catheter.  The patient’s CVP is 6 mmHg.


There may be a tendency to cement guideline recommendations into our collective minds as inalienable and sacrosanct – as if carved onto stone tablets and brought down to us from the heights of Mount Sinai.  Thoughtful medicine, however, demands that we resist this inclination.

Within the 2016 Surviving Sepsis Guidelines lies the following recommendation: ‘normalize lactate in patients with elevated lactate levels as a marker of tissue hypo-perfusion.’  This, however, is graded as a weak recommendation, low quality of evidence.  Interestingly, buried within the text, the guideline authors declare – correctly – that ‘serum lactate is not a direct measure of tissue perfusion.’  Could there exist a better indicator of tissue starvation in septic shock?

Almost 40 years ago, Champion introduced the idea of monitoring capillary refill time [CRT] for shock resuscitation.  This low-tech metric fell out of favour during the zenith of invasive monitoring and oxygen delivery optimization in the 1980s and ‘90s.  However, recent data have shown that CRT: 1. is potentially a more responsive resuscitation sign in sepsis than biochemical guides, 2. spares excessive intravenous fluids and their inherent harms and 3. predicts mortality in septic patients in the ED.

Given the above, the ANDROMEDA-SHOCK trial was released last week at the 48th Critical Care Congress and the caterwauling was recorded immediately across multiple time zones.  Do we dare worship this false idol?  This Golden Calf called ‘capillary refill time?’

What They Did

This was a multi-centre, open-label randomized controlled study. 424 adult septic shock patients were enrolled [i.e. lactate ≥2.0 mmol/L, requiring vasopressors to maintain a MAP of 65 mm Hg after a fluid bolus of at least 20 mL/kg] to an 8 hour resuscitation protocol.  One group [212 patients] had their hemodynamic targets driven by normalization of capillary refill time [CRT] to less than 3 seconds; the other group [212 patients] were resuscitated based upon serum lactate reduction [20% per 2 hours].  In addition to the primary goal of either CRT normalization or lactate reduction, the other stopping points for fluid resuscitation were: a rise in CVP by at least 5 mmHg and/or the cessation of ‘fluid responsiveness.’  The latter was determined based on pulse pressure/stroke volume variation if the patient was adapted with a mechanical ventilator, end-expiratory occlusion test if invasively-ventilated with spontaneous efforts or with passive leg raise and beat-to-beat stroke volume monitoring if the patient was not intubated.

What They Found

Those randomized to lactate-guided resuscitation received an excess of 400 mL of crystalloid over the 8 hour resuscitation period as compared to the CRT group, yet had more organ dysfunction at 72 hours [SOFA score increased by 1 point] including a clinically [but not statistically] significant increase in renal replacement therapy.  All-cause 28-day mortality was 34.9% in the CRT group and 43.4% in the lactate-guided group.


There is something truly sophisticated and wonderful about resurrecting a simple bedside test to guide hemodynamics in the ICU – most of us lost in the numbers and gadgets of resource-rich medicine – reminding us that the physical exam has not lost its place.

Yet beyond my overwrought exhortation of bedside medicine above, there is the glaring 8.5% absolute risk reduction in 28-day mortality observed in patients resuscitated by CRT as compared to lactate.  While clinically very significant, this did not quite reach statistical significance [i.e. hazard ratio 0.75 [(95% confidence interval of 0.55 to 1.02); p=.06].  I will leave the statistical wizardry to others and instead wonder, how could this result – biologically, medically – be?

Possible Mechanism: gut protection?

On first blush, it is tempting to regard the excess mortality in the lactate group as a side-effect of ‘over-resuscitation.’  To be sure, in a small, proof-of-concept study, these authors observed exactly that – patients randomized to CRT-perfusion received 1.8 fewer litres of resuscitation fluid in the first 6 hours [6 L versus 4.2 L] as well as fewer days in the hospital and lower SOFA scores.  However, there are a number of problems with this ‘excessive resuscitation’ theory in ANDROMEDA SHOCK.  First, the absolute fluid volume difference between the CRT and lactate groups was only 400 mL.  Second, the total fluid over 8 hours in both groups was relatively consistent with – if not more conservative than – the first 6 hours of fluids administered in ProCESS trial [see table 1].

Table 1: Comparison of APACHE II score and fluids administered in the early phase of ProCESS and ANDROMEDA-SHOCK; note that total fluids are over 6 hours in ProCESS and 8 hours in ANDROMEDA. The last two hours of IVF - intravenous fluid - infusion rate from ProCESS are given to estimate 8 hours of resuscitation.  EGDT is Early Goal Directed Therapy.  In the ProCESS trial, the non-EGDT protocol group received 1.1 more litres of fluid in the first 6 hours than the 'usual care' group but had the same mortality rate.  Thus, the 400 mL absolute difference in ANDROMEDA-SHOCK seems unlikely to be the cause of the mortality difference.

Finally, and most interestingly, the difference in the evolution of fluid non-responders over 8 hours in ANDROMEDA-SHOCK does not speak to the lactate group being more aggressively pushed to the flat portion of the Starling curve as compared to the CRT group [see figure 1].

Figure 1: y-axis is absolute number of fluid non-responsive patients evolving over the first 8 hours between the CRT group [blue] and lactate group [red]; each group contained 212 patients. The x-axis represents time zero, 2 hours, 4 hours and 8 hours. Below the graph is the percentage of all patients at these time points who could not have their fluid-responsiveness determined 'unknown status'.  Of those who did, the percentage of non-responsive patients who evolved during the resuscitation is listed below - these percentages reflect all 424 patients.  Data adapted from supplemental material.

If the mortality difference observed in ANDROMEDA-SHOCK is true, the mechanism is not discernably clear.  Perhaps it’s less about the absolute amount of fluid and more about its timing; this concept has been explored in operative goal-directed therapy.  Is it possible that fingertip perfusion acts as a window to the visceral organs and, in particular, the gut?  Indeed, peripheral perfusion indices like the CRT have been associated with visceral arterial resistance as measured by pulsed-wave Doppler.  The splanchnic circulation responds to adrenergic tone as the skin does, by decreasing its perfusion and discharging blood into systemic circulation.  Transfer of blood volume from the splanchnic to systemic circulations may transiently support hemodynamics, but over time may lead to gut endo- and epithelial integrity failure.  In turn, translocation of gut bacteria – and their fragments – potentiate the inflammatory response.

So if sluggish CRT warns of visceral tissue starvation, then intravenous volume is one potentially beneficial intervention.  Monge-Garcia and colleagues showed that septic patients who do not increase their MAP, but do augment their cardiac output in response to fluids have a fall in their arterial elastance; more simply, fluids act as ‘vasodilators’ in fluid responsive/MAP-non-responsive septic patients.  This physiology is potentially helpful when the critical closing pressure [Pcc – i.e. ‘arteriolar tone’] in the GI tract is high [as gauged by prolonged CRT], but much less so – if not harmful – when the critical closing pressure is low-normal and/or when the heart is fluid intolerant.

Usual Practice Arm

One potential criticism of ANDROMEDA-SHOCK is that there was not a ‘usual care’ arm.  Presumably, this would entail a gestalt approach drawn largely from the current Surviving Sepsis Guidelines.  In other words, resuscitation based on lactate and/or mixed venous oxygen saturation with fluids provided based on the CVP or some non-invasive surrogate.  Indeed, based on the recent FENICE survey, right atrial pressure continues to be the predominant metric for assessing ventricular fluid tolerance.  Thus a high lactate and low CVP would likely trigger intravenous fluid.  Yet, it is known that 25% of patients intolerant of further fluid loading have a CVP below 5 mmHg; this percentage rises to ~40-45% of patients with a CVP below 8 mmHg!  Lastly, because lactate takes much longer to normalize than CRT, it seems quite likely that ‘standard practice’ would favour over-resuscitation though, as above, the fluids administered in the ‘usual care’ arm of ProCESS are strikingly similar to ANDROMEDA.  This comparison is tenuous, however, given the sicker cohort [only septic shock] enrolled in ANDROMEDA.

A Physiological Summation

Tissue perfusion, very simply, requires balance between arterial tone and volume relative to its ‘back pressure.’  Notably, the authors of ANDROMEDA-SHOCK used abnormal rise in CVP as a metric of excessive venous back pressure – others are using a markedly pulsatile portal vein.  Regardless, when the clinician is confident that neither excessive venous nor interstitial pressures are retarding perfusion, then capillary refill time [CRT] provides a simple and elegant measure of the balance between arterial tone and volume.  For example, an abnormally prolonged CRT suggests that the pre-capillary critical closing pressure [Pcc – i.e. ‘arteriolar tone’] is too high relative to the current cardiac output.  Put simply, the Pcc – rather than downstream venous pressure – now acts as a disproportionate ‘back pressure’ for tissue perfusion.  Faced with this, augmenting flow into the arterial tree [i.e. raising the MAP relative to the Pcc] with preload [in a responsive heart] or with inotropes [in a non-responsive heart] is a parsimonious tactic to correct the imbalance between arterial volume and tone.

Let us not forget what Susan Sontag reminded us; truth is balance.

Return to Case

The patient tolerates high flow oxygen well, however, as his work of breathing falls so too does his blood pressure and CVP; his capillary refill time remains brief.  Having recently led a journal club on ANDROMEDA-SHOCK, the ED resident confidently initiates the patient on norepinephrine.  She notes near normalization of the patient's mean arterial pressure; one hour later his lactate falls to 4.0 mmol/L and his urine output rises despite her holding further IV fluids. 



Dr. Kenny is the cofounder and Chief Medical Officer of Flosonics Medical; he also the creator and author of a free hemodynamic curriculum at heart-lung.org




Get our weekly email update, and explore our library of practice updates and review articles.

PulmCCM is an independent publication not affiliated with or endorsed by any organization, society or journal referenced on the website. (Terms of Use | Privacy Policy)


Should We Stop Trending Lactate in Septic Shock? ANDROMEDA-SHOCK Published