🤷‍♂️ Vitamins in Sepsis Again

The VITAMINS trial was not the only trial investigating the efficacy of the hydrocortisone/vitamin C/thiamine cocktail in severe sepsis – and will thus not be the last word on the matter, by any chance. This is one of what is sure to be a slow trickle of other relevant data – as much as anything can break through the cacophony of the coronavirus pandemic.

This is a small trial, just 137 patients, and they used ascorbic acid 1,500 mg q6h, thiamine 200 mg q12, and hydrocortisone 50 mg q6h or a matching saline placebo for a four day course of treatment. Open-label corticosteroids were permissible per the clinical judgement of treating clinicians. The primary outcome was – wait for it – resolution of shock and change in SOFA score. This is, unfortunately, a change from their original primary outcome – in-hospital mortality. The original primary outcome is not mentioned in their manuscript, however, and includes a power calculation based on their secondary outcome – and this power calculation serendipitously matches their original anticipated study size as seen on clinicaltrials.gov.

There are circumstances in which changing the primary outcome is reasonable based on new, outside information obtained during the course of a study. That is not the case here. Moreover, and less appetizing, the new primary outcome is conveniently the only outcome measured significantly favoring the intervention. The authors tracked WBC counts, platelet counts, lactates, SOFA scores, fluid balance, procalcitonin clearance, ICU length-of-stay, hospital mortality, ICU mortality, and more, ad nauseaum. It is clear their original primary outcome would not have reached statistical significance because mortality was far too low – 16% with the intervention, 19% for control – based on sample size. However, the correct thing to do is simply run the trial out and have a reasonable academic discussion of the observed findings, not change to a disease-oriented surrogate.

Because, after all this, the authors make a fair bit of hay of their observed difference in shock resolution, a finding favoring the intervention by nearly a day. However, there were differences[1] in the groups at baseline specifically regarding the initiation of vasopressors – and this probably trickles down to the duration of vasopressors, as well. This study probably mostly shows just how difficult it is to do a study in intensive care, and how robust a sample size is required. Giving patients vitamins is unlikely to cause specific harms, but it doesn’t seem to be all that helpful. Remember – reliably useful treatments give reliably positive results.

“Outcomes of Metabolic Resuscitation Using Ascorbic Acid, Thiamine, and Glucocorticoids in the Early Treatment of Sepsis: The ORANGES Trial”
https://www.sciencedirect.com/science/article/pii/S0012369220304554

  1. Differences reported with a p-value, no less. There is no reason to report p-values for baseline characteristics in a randomized trial. A p-value here describes the likelihood an observation occurred by chance alone, but, obviously, because it was randomized, the chance it occurred by chance alone is 100%.

Happy VITAMINS Day!

After a couple years wandering in the wilderness after the Marik report, and a few small or unrevealing trials, we finally have our first RCT evidence regarding the use of thiamine, vitamin C, and steroids in sepsis: the VITAMINS trial.

In this trial, 216 patients with septic shock received hydrocortisone 50mg every six hours, and were then randomized to usual care or open-label vitamin C 1.5g every six hours plus thiamine 200mg every twelve hours. The primary outcome was time alive and free of vasopressor administration up to day 7, along with pre-specified secondary mortality outcomes. Important exclusions in the eligibility screening process included onset of septic shock >24 hours, imminent death, and use of study drugs for other reasons. Few differences between the two groups were observed at baseline, although the control arm had median lactate level of 3.3 at enrollment compared with 4.2 in the intervention arm.

The primary outcome was: no different, 122.1 hours with the intervention and 124.6 hours in the control arm. The secondary mortality outcomes: at 28 days, no different, 22.6% vs. 20.4%; at 90 days, no different, 28.6% vs 24.5%. The remaining secondary outcomes, including ventilation-free days, renal replacement, and acute kidney injury were no different, but there was a small different in SOFA scores at day 3 favoring the intervention. The authors appropriately caution this observed improvement in SOFA score should be interpreted at your peril, as its significance is not adjusted for multiple comparisons and subject to both competing risks from early death and early discharge from the ICU.

So, the pure frequentist conclusion: this trial, repeated, would provide an estimate containing the true effect size with bounds crossing unity most of the time. The Bayesian conclusion, accounting for the weak, positive, prior evidence: there is a low probability of the true effect being positive. If you were holding off adopting the addition of thiamine and vitamin C, this trial reinforces your skepticism. If you’ve already adopted thiamine and vitamin C, it is unlikely harms were caused, but it is now more likely than not these treatments are not resulting in benefit. Additional trials will report results capable of further clarifying these observations.

The accompanying editorial by Andre Kalil sums up the interpretation of these results in context beautifully:

“Given that other studies are forthcoming, there appears to be no immediate justification for adoption of high-dose vitamin C, alone or in combination, as a component of treatment for sepsis. Moreover, use of high-dose vitamin C in combination or alone “just in case” or as a “measure of last resort,” aside from providing no survival benefits, could have several other potential consequences, including diverting funding from needed research to examine sepsis mechanisms and diagnostics; stifling the development of other sepsis therapies; perpetuating false hopes for patients, families, and clinicians; and delaying proven lifesaving therapies such as prompt initiation of antibiotic therapy.”

“Effect of Vitamin C, Hydrocortisone, and Thiamine vs Hydrocortisone Alone on Time Alive and Free of Vasopressor Support Among Patients With Septic Shock”
https://jamanetwork.com/journals/jama/fullarticle/2759414

The Vitamins in Sepsis Parade Begins

A couple years back, we saw one of the first reports describing the potential efficacy of treating sepsis with a cocktail of vitamin C, thiamine, and steroids. These observational findings have been viewed with a healthy dose of skepticism while awaiting prospective, randomized evidence with regard to their validity.

Well, here’s one of the first: a short communication from CITRIS-ALI, a randomized clinical trial that almost, sort of, not quite, addresses the question of interest. This study was designed and initiated well before the aforementioned observational report, and it examines vitamin C monotherapy in patients with sepsis and acute respiratory distress syndrome. Their primary outcome and goal was to see if vitamin C could reduce sequential organ failure assessment scores, along with biological markers of inflammation and vascular injury.

With 1,262 patients screened leading to 167 patients receiving their randomized interventions, the answer is: no. Neither modified SOFA scores, c-reactive protein, nor thrombomodulin were different between groups.

But, wait! There’s more! In fact, 46 additional pre-specified secondary outcomes – 43 of which showed no difference. These included both the esoteric – angiopoietin-2 levels, tissue factor pathway inhibitor – and patient-oriented. It is these patient-oriented outcomes that pique the most interest: at 28 days, mortality in the vitamin C group was 29.8%, as compared with 46.3% with placebo. Ventilator-free days and ICU-free days similarly favored the vitamin C cohort.

So, interesting data incapable of informing practice. Another small sample, designed (appropriately) around a different target and primary outcome, with a secondary outcome still falling into the realm of hypothesis-generating. This will likely influence no one. Anyone already giving vitamins in sepsis – cheap, likely harmless – will continue to do so, and those awaiting a more informative trial will, also, continue to do so.

“Effect of Vitamin C Infusion on Organ Failure and Biomarkers of Inflammation and Vascular Injury in Patients With Sepsis and Severe Acute Respiratory Failure”

https://jamanetwork.com/journals/jama/fullarticle/2752063

The Antibiotic Penalty on Blood Cultures

As the administrative team likes to remind us: blood cultures before antibiotics.

Blood cultures before antibiotics.

Blood cultures? Before antibiotics.

What’s the point, we say – aren’t the antibiotics the actual life-saving intervention? And the answer, when relevant, ties into identifying the specific susceptibility of the infective agent, such that antibiotics may ultimately be narrowed to the minimum necessary for cure. It’s a noble premise, at least.

But, so, what does happen when you give antibiotics first?

At least one recent retrospective study has pulled data from their health system showing a clear decrease in blood culture positivity following administration of antibiotics, but these results may be limited by potential differences between groups. In contrast, this clever little study looks at it prospectively: the same 325 Emergency Department patients with “severe manifestation of” sepsis – hypotensive or lactate >4 mmol/L – received blood culture draws both prior to, and just following, antibiotic administration.

Before antibiotics: 31.4% positive blood cultures.

After antibiotics: 19.4% positive blood cultures.

It is not a perfect study by any means, but a long story short: if you’re going to go to the trouble of drawing and processing blood cultures, draw them before you start antimicrobial treatment. But, clearly, the antimicrobials are doing their job – do it expeditiously such that your patient does not suffer from unwanted delay.

“Blood Culture Results Before and After Antimicrobial Administration in
Patients With Severe Manifestations of Sepsis”
https://annals.org/aim/fullarticle/2751453/blood-culture-results-before-after-antimicrobial-administration-patients-severe-manifestations

The United Colors of Sepsis

Here it is: sepsis writ Big Data.

And, considering it’s Big Data, it’s also a big publication: a 15 page primary publication, plus 90+ pages of online supplement – dense with figures, raw data, and methods both routine and novel for the evaluation of large data sets.

At the minimum, to put a general handle on it, this work primarily demonstrates the heterogeneity of sepsis. As any clinician knows, “sepsis” – with its ever-morphing definition – ranges widely from those generally well in the Emergency Department to those critically ill in the Intensive Care Unit. In an academic sense, this means the patients enrolled and evaluated in various trials for the treatment of sepsis may be quite different from one another, and results seen in one trial or setting may generalize poorly to another. This has obvious implications when trying to determine a general set of care guidelines from these disparate bits of data, and resulting in further issues down the road when said guidelines become enshrined in quality measures.

Overall, these authors ultimately define four phenotypes of sepsis, helpfully assigned descriptive labels using the letters of the greek alphabet. These four phenotypes of sepsis are derived from retrospective administrative data, then validated on additional retrospective administrative data, and finally the raw data from several prominent clinical trials in sepsis, including ACCESS, PROWESS, and ProCESS. The four phenotypes were derived by clustering and refinement, and are described by the authors as effectively: a mild type with low mortality; a cohort of those with chronic illness; a cohort with systemic inflammation and pulmonary disease; and a final cohort with liver dysfunction, shock, and high mortality.

We are quite far, however, from needing to apply these phenotypes in a clinical fashion. Any classification model is highly dependent upon the inputs, and in this study the inputs are the sorts of routine clinical data available from the electronic health record: vital signs, demographics, and basic labs. Missing data was common, including, for example, lactate levels, which was not obtained on 80% of patients in their model. These inputs then dictate how many different clusters you obtain, how the relative accuracy of classification diminishes with greater numbers of clusters, as well whether the model begins to overfit the derivation data set.

Then, this is a little bit of a fuzzy application in the sense these data represent as much different types of patients with sepsis, as it represents different types of sepsis. Consider the varying etiologies of sepsis, including influenza pneumonia, streptococcal toxic shock, or gram-negative bacteremia. These different etiologies would obviously result in different host responses depending on individual patient features. These phenotypes derived here effectively mash up causative agent with the underlying host, muddying clinical application.

If clinical utility is limited, then what might the best utility for this work? Well, this goes back to the idea above regarding translating work from clinical trials to different settings. A community Emergency Department might primarily see alpha-sepsis, a community ICU might see a lot of beta-sepsis, while an academic ICU might see predominantly delta-sepsis. These are important concepts to consider – and potentially subgroup-analyses to perform – when evaluating the outcomes of clinical trials. These authors do several simulations of clinical trials while varying the composition of phenotypes of sepsis, and note potentially important effects on primary outcomes. Pathways of care or resuscitation protocols could potentially be more readily compared between trial populations if these phenotypes were calculated.

This is a challenging work to process – but an important first step in better recognizing the heterogeneity in potential benefits and harms resulting from various interventions. The accompanying editorial does really a very excellent job of describing their methods, outcomes, and utility, as well.

“Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis”
https://jamanetwork.com/journals/jama/fullarticle/2733996

“New Phenotypes for Sepsis”
https://jamanetwork.com/journals/jama/fullarticle/2733994

The Precogs Take On Sepsis

It seems like every week there’s another publicized instance of our impending replacement by artificial intelligence. Big Data, they say, is going to free us of the cognitive burdens of complex thought while maximizing healthcare outcomes. This latest entry is the “AI Clinician”, which has been created as a demonstration for the treatment of sepsis.  Or, rather more narrowly, the AI Clinician tries to prescribe the balance of fluids and vasopressors.

In this predictive feat of strength, decision models were created based on retrospective data sets comprised of tens of thousands of patients meeting Sepsis-3 criteria. Each patient’s clinical trajectory was described by their receipt of intravenous fluids or vasopressors in four-hour blocks, and the ultimate outcome of 90-day survival designated as the reward or penalty for their model. It’s rather beyond the scope of my statistical expertise to precisely describe their value comparison between the AI and clinicians, but suffice to say their results favor their models.

We are rather far from this sort of software being validated as a management adjunct in sepsis, but what’s most interesting is their incidental description of how deviations from their model affected mortality. Effectively by definition, of course, they find patients receiving IV fluids or vasopressors in doses most similar to the AI model had the lowest mortality. Greater variance from these optimal doses tended to increase mortality – most prominently excesses of IV fluids, rather than restrictive IV fluids. Vasopressors, on the other hand, showed a more symmetric distribution of poor outcomes with deviation from the optimal model:

The implication here mostly ties into the oft-repeated concern that high-volume fluid resuscitation is not necessarily the magic bullet in sepsis, and there is likely a point at which returns diminish, or turn harmful. This is virtually the exact hypothesis addressed by the CLOVERS trial. It will be quite interesting to see if these model findings are validated by the trial.

“The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care”
https://www.nature.com/articles/s41591-018-0213-5

The Secret Ingredient Is: Thiamine

At least, when you’re thiamine deficient.

Of the magic cocktail of profound improvement in sepsis, it is not known the relative importance of the various ingredients, whether it is a synergistic effect, or, technically, whether the treatment is real or artifactual.  Thiamine deficiency, however, is frequently detected in patients with sepsis and septic shock. A small pilot study showed no overall effect of thiamine administration to a general population with sepsis, but a subgroup with documented thiamine deficiency suggested improvements in lactate clearance and mortality.

Following up these findings, these authors performed a retrospective review of outcomes of patients admitted to their single center with sepsis and septic shock, as defined by a lactate greater than 2 mmol/L and a need for vasopressors. They created two matched cohorts using Mahalanobis distance, and, lo: lactate clearance was improved with thiamine, as was overall mortality, with a Hazard ratio of 0.666 (95% CI, 0.490-0.905).

This is, again, retrospective data and statistical tomfoolery to match. But, it is consistent, at least, with other prospective and observational data. It seems quite reasonable to evaluate patients with septic shock for thiamine deficiency, with the expectation supplementation may improve outcomes.

“Effect of Thiamine Administration on Lactate Clearance and Mortality in Patients With Septic Shock”

https://www.ncbi.nlm.nih.gov/pubmed/30028362

All Sepsis Is Not the Same

This is a fairly dense informatics evaluation of sepsis, but it boils down to a general hypothesis with some face validity: all sepsis is not the same! This is abundantly obvious from the various clinical manifestations of response to infection, with a spectrum ranging from Group A Streptococcal pharyngitis to gram-negative bacteremia and distributive shock.

This analysis uses genetic expression sampling from whole blood to perform unsupervised machine learning analyses and clustering, and they identify three subtypes the authors term “Inflammopathic, Adaptive, and Coagulopathic”. Whether these are terribly illustrative of the underlying pathology is unclear, but, if you want to be in one of these clusters, you want to be in “Adaptive” with its 8.1% mortality – compared to 29.8% in Inflammopathic and 25.4% in Coagulopathic.

Validity of this specific analysis aside, it’s an interesting example of what may ultimately be a useful approach to treating sepsis – targeting the specific underlying genetic expressions associated with dysregulated immune response or underlying end-organ dysfunction. The best thing about this paper, however, are the acronyms reported for some of the statistical methods: “COmbined Mapping of Multiple clUsteriNg ALgorithms” or COMMUNUAL, and “COmbat CO-Normalization Using conTrols” or COCONUT.

“Unsupervised Analysis of Transcriptomics in Bacterial Sepsis Across Multiple Datasets Reveals Three Robust Clusters”
https://www.ncbi.nlm.nih.gov/pubmed/29537985

The qSOFA Story So Far

What do you do when another authorship group performs the exact same meta-analysis and systematic review you’ve been working on – and publishes first? Well, there really isn’t much choice – applaud their great work and learn from the experience.

This is primarily an evaluation of the quick Sequential Organ Failure Assessment, with a little of the old Systemic Inflammatory Response Syndrome thrown in for contextual comparison. These studies included those in the Intensive Care Unit, hospital wards, and Emergency Departments. Their primary outcome was mortality, reported in these studies mostly as in-hospital mortality, but also 28-day and 30-day mortality.

The quick synopsis of their results, pooling 38 studies and 383,333 patients, mostly from retrospective studies, and mostly from ICU cohorts:

  • qSOFA is not terribly sensitive, particularly in the settings in which it is most relevant. Their reported overall sensitivity of 60.8% is inflated by its performance in ICU patients, and in ED patients sensitivity is only 46.7%.
  • Specificity is OK, at 72.0% overall and 81.3% in the ED. However, the incidence of mortality from sepsis is usually low enough in a general ED population the positive predictive value will be fairly weak.
  • In their comparative cohort for SIRS, which is frankly probably irrelevant because SIRS is already well-described, the expected results of higher sensitivity and lower specificity were observed.

Their general conclusion, to which I generally agree, is qSOFA is not an appropriate general screening tool. They did not add much from a further editorial standpoint – so, rather than let our own draft manuscript for this same meta-analysis and systematic review languish unseen, here is an abridged version of the Discussion section of our manuscript written by myself, Rory Spiegel, and Jeremy Faust:

This analysis demonstrates qualitatively similar findings as those observed in the original derivation study performed by Seymour et al. We find our pooled AUC, however, to be lower than the 0.81 reported in their derivation and validation cohort, as well as the 0.78 reported in two external validation cohorts. The meaning of this difference is difficult to interpret, as the clinical utility of this instrument is derived from its use as a binary cut-off, rather than an ordinal AUC. Our sensitivity and specificity from our primary analysis, respectively, compare favorably to their reported 55% and 84%. We also found qSOFA’s predictive capabilities remained robust when exposed to our sensitivity analyses. When only studies at low risk for bias were included, qSOFA’s performance improved.

While our evaluation of SIRS is limited by restricting the comparison solely to those studies which contemporaneously reported qSOFA, our results are broadly consistent with results previously reported. The SIRS criteria at the commonly used cut-off benefits from superior sensitivity for mortality in those with suspected infection, while its specificity is clearly lacking due to its impaired capability to distinguish between clinically important immune system dysregulation and normal host responses to physiologic stress. The important discussion, therefore, is whether and how to incorporate each of these tools – and others, such as the Modified Early Warning Score or National Early Warning Score – into clinical practice, guidelines, and quality measures.

The current approach to sepsis revolves around the perceived significant morbidity and mortality associated with under-recognized sepsis, favoring screening tools whose purpose is minimizing missed diagnoses. Current sepsis algorithms typically rely upon SIRS, depending on its maximal catchment at the expense of over-triage. Such maximal catchment almost certainly represents a low-value approach to sepsis, considering the in-hospital mortality of patients in our cohort with ≥2 SIRS criteria is not meaningfully different than the overall mortality of the entire cohort. The subsequent fundamental question, however, is whether qSOFA and its role in the new sepsis definitions provides a structure for improvement.

Using qSOFA as designed with its cut-off of ≥2, it should be clear its sensitivity does not support its use as an early screening tool, despite its simplicity and exclusion of laboratory measures. However, in a cohort with suspected infection and some physiologic manifestations of sepsis, e.g., SIRS, the true value of qSOFA may be in prioritizing a subgroup for early clinical evaluation. In a healthcare system with unlimited resources, it may be feasible to give each patient uncompromising evaluation and care. Absent that, we must hew towards an idealized approach, where our resources are directed towards those highest-yield patients for whom time-sensitive interventions modify downstream outcomes.

Less discussed are the direct, patient-oriented harms resulting from falsely-positive screening tools and over-enrollment into sepsis bundles. Recent data suggests benefits from shorter time-to-antibiotics administration intervals are realized primarily in critically ill patients. As such, utilization of overly sensitive tools, such as the SIRS criteria, would lead to over-triage and over-treatment, leading to potential iatrogenic harms in excess of net benefits. These harms include effects on individual and community patterns of antibiotic resistance, as exposure to broad-spectrum antibiotics leads to induction of extended-spectrum beta-lactamase resistance in gram-negative pathogens or vancomycin- and carbapenem-resistance in enterococci. Unnecessary antibiotic exposures lead to excess cases of C. difficile infections. The aggressive fluid resuscitation mandated by sepsis bundles leads to metabolic derangement and potential respiratory impairment. Further research should assess the extent of these harms, and in what measure they counterbalance those benefiting from time-sensitive interventions.

This meta-analysis has several limitations. First, we were limited by the relative dearth of high quality prospective data; most of the studies included in our analysis were retrospective. Second, we restricted our prognostic analyses to mortality alone, rather than diagnosis of sepsis. We chose to analyze only mortality because of competing sepsis definitions among expert bodies and government-issued guidelines. Among them, however, mortality is a common feature, the most objective metric, and manifestly the most important patient-centered outcome. Our analysis would not capture other important sequelae of sepsis, including amputation, loss of neurologic and/or independent function, chronic pain, and prolonged psychiatric effects of substantial critical illness. Third, we do not know whether patients included in these studies were septic on presentation, or developed sepsis later in their hospitalization. This may degrade the accuracy assessment of both SIRS and qSOFA. Fourth, while we know that qSOFA alone may miss some cases of sepsis that SIRS might detect, we do not know how many would, in reality, have been deprived of antibiotics and other necessary treatments. In other words, the fate of “qSOFA negative” patients who were evaluated and treated by physicians qualified to detect and treat critical illness via clinical acumen is not known; nor it should not be presumed that all such patients would have necessarily been deprived of timely treatment. Our analysis and comparison of SIRS is definitively incomplete, and not the most reliable estimate of its diagnostic characteristics, but provided for incidental comparison.

The prudent clinical role for qSOFA, however, is as yet undefined, and these data do not offer insight regarding its superiority to clinician judgment for determining a cohort at greatest risk for poor outcomes. Compared with SIRS, at least, those patients identified by qSOFA likely better represent the subset of patients for whom aggressive early treatment confers a particular advantage, and may drive high-value care in the sepsis arena. Future research should assist clinicians in further individualizing initial treatment of sepsis for those stratified to differing levels of risk for poor outcome, as well as to account for the iatrogenic harms and system costs.

“Prognostic Accuracy of the Quick Sequential Organ Failure Assessment
for Mortality in Patients With Suspected Infection: A Systematic Review and Meta-analysis”
http://annals.org/aim/fullarticle/2671919/prognostic-accuracy-quick-sequential-organ-failure-assessment-mortality-patients-suspected

Is the Urinalysis Reliable in Young Infants?

The evaluation of the very young infant with a fever is complex, with multiple competing factors including the rarity of serious illness, the severity of serious illness, and the cost of the intensive evaluation frequently required. The most commonly identified bacterial source for fever is a urinary tract infection, and our bedside test in the Emergency Department is the urinalysis.

So, how reliable and accurate is that test?

This is an analysis of prospectively collected data from the PECARN network, looking at the evaluation of febrile infants fewer than 60 days of age. Of 4,147 patients enrolled, 289 patients had UTIs by a 50,000 CFUs/mL definition on the subsequent urine culture. Only 27 patients had bacteremia and a UTI. The news is generally mixed: using the 50,000 CFUs/mL cut-off, any abnormality on the UA was 94% sensitive for UTI and 91% specific, but was 100% sensitive for a UTI associated with bacteremia.

The authors also do analyses including different cut-offs for UTI based down to 10,000 CFUs/mL and, as you might expect, the sensitivity for any UTI diminishes. While the interpretation of the urine culture result is less applicable to the initial Emergency Department evaluation, the subsequent threshold for diagnosis is relevant to the ongoing follow-up care for the febrile infant, particularly if an initial decision involved observation without antibiotics and the infant remains symptomatic without another source.

Overall, it is reasonable to suggest – if the UA is negative, a serious bacterial illness is unlikely to be present. Some consideration should be made to the duration of illness, and natural course of delayed onset of development of cystitis or pyuria in the urine. A positive UA, however, despite the apparent high specificity, does not reliably indicate a true positive for UTI, owing to the low prevalence. This should also be taken into consideration regarding whether additional invasive evaluation is indicated.

“Accuracy of the Urinalysis for Urinary Tract Infections in Febrile Infants 60 Days and Younger”
http://pediatrics.aappublications.org/content/early/2018/01/12/peds.2017-3068