Metabolic resuscitation for sepsis is currently quite controversial. Marik et al. published a before-after study in 2017 describing the combination of hydrocortisone, ascorbate, and thiamine for septic shock.1 That study incited a media storm of surprising intensity. Currently, several multi-center RCTs are underway to answer this question more definitively. In the interim, some additional before/after trials may emerge. These studies likely won’t answer the question, but they may provide us with some clues about whether metabolic resuscitation is on the right track.
Brief rationale for metabolic resuscitation in severe community-acquired pneumonia
Steroid: Mounting evidence supports the use of steroid for severe community-acquired pneumonia. This evidence was explored on this blog in 2015 here. Consensus guidelines by the SCCM and ESICM societies subsequently recommended steroid for severe community-acquired pneumonia in 2017.2 Thus, the use of steroid for severe community-acquired pneumonia is supported by several RCTs as well as society guidelines.3
Ascorbate & Thiamine: The rationale for using ascorbate and thiamine in sepsis was previously explored on the blog here. Ascorbate is often deficient among patients with pneumonia, with some evidence suggesting that it might be beneficial in this population.4
Kim et al 2018: Combined vitamin C, hydrocortisone, and thiamine therapy for patients with severe pneumonia who were admitted to the intensive care unit: Propensity score-based analysis of a before-after cohort study.
This is a single-center before-after trial describing the use of metabolic resuscitation for patients admitted to the ICU with severe pneumonia.5 In June 2017, their hospital started routinely using hydrocortisone, ascorbate, and thiamine for severe pneumonia. This allowed them to perform a retrospective before/after study comparing patients admitted in two consecutive years over a matching timeframe (June-January).
Exclusion criteria included the following:
- Patients who only required conventional oxygen therapy
- Delayed ICU admission (>48 hours after hospitalization)
- Delayed metabolic resuscitation (>48 hours after hospitalization)
The primary outcome was hospital mortality. Secondary outcomes included pressor-free days, vent-free days, and ICU length of stay.
Primary unadjusted analysis
Baseline characteristics of both patient groups are shown above. Patients in the treatment group were somewhat sicker (with a greater rate of renal replacement therapy and vasopressor use). Differences in renal replacement use were significant (p=0.004). Nonetheless, patients in the treatment group had a lower hospital mortality:
Subgroup analysis was performed to evaluated patients with APACHE II score >28 and PaO2/FiO2 ratio <120. These cutoff values are both the median values among all patients, so these subgroups evaluate the sicker half of patients in each cohort. In both cases, metabolic resuscitation seemed to improve mortality:
Propensity matching is a statistical technique to account for differences in baseline variables (with the goal of eliminating confounding effects of these variables). This is accomplished roughly as follows:
- Differences in baseline variables between the treatment and control groups are analyzed. Baseline variables are combined to create a propensity score that describes the likelihood of the patient’s being assigned to the treatment group.
- Pairs of patients (one in the treatment and one in the control group) are identified whose propensity scores match. Combining numerous propensity-matched pairs generates two new groups (each of which is a subset of the original groups). These propensity-matched groups should ideally be better matched than the original groups.
Propensity-matching revealed two 36-patient groups. Compared to the original groups, these groups were better matched:
Propensity matching can’t magically convert a retrospective study into a prospective, double-blind RCT. Rather, propensity matching is a bit like multivariable statistical analysis: it ought to get you closer to the truth but there’s no guarantee (especially if numerous interacting covariates are present). Propensity matching works best if no medical judgement is used to determine which patients receive treatment, which is actually the case here:
Ultimately this analysis isn’t definitive for several reasons: small sample size, marginal p-value, and use of propensity matching. No matter how good propensity matching is, it cannot correct for non-random sources of bias (e.g. improvements in critical care between 2016 and 2017).
No increase in acute kidney injury or superinfection was detected. However, this study wasn’t powered to detect rare adverse events.
They are numerous:
- Small size
- Non-randomized, non-blinded
- Baseline imbalance in patient groups (which is partially addressed by propensity-matching analysis)
- 65% of patients in the control group received steroid. This suggests that most treatment differences will reflect the addition of ascorbate & thiamine treatment. However, some benefit due to steroid alone cannot be excluded.
- The use of steroid for community-acquired pneumonia is supported by several RCTs and currently recommended by SCCM/ESICM guidelines.2 However, metabolic resuscitation with hydrocortisone, ascorbate, and thiamine remains quite controversial.
- Kim 2018 performed a before/after study evaluating metabolic resuscitation for patients admitted to an ICU with severe pneumonia.5 Various analyses suggest a mortality benefit, particularly among the sickest patients. However, this study is far from definitive due to numerous limitations.
- A multi-center RCT is needed to address this question. In the interim, outcomes from this hospital suggest that metabolic resuscitation is currently a reasonable treatment option for severe community-acquired pneumonia admitted to the ICU.
Opening image courtesy of Dr Chris O'Donnell.