A Skeptic’s Guide to Vitamin C in Sepsis
The recent hydroxychloroquine saga in COVID19 has been an eye-opening example of how enthusiasm for a therapy can get far ahead of the evidence base. This is a phenomenon we’ve seen in critical care previously. Let’s consider another controversial ICU therapy: Vitamin C, or ascorbate, which has been heralded as the “metabolic cure” for sepsis.
Hydrocortisone, Ascorbate and Thiamine (also known as HAT therapy) was based on a small single-center before/after study that found a miraculous decrease in hospital mortality from 40.4% to 8.5% when given to ICU patients with sepsis.
Before-After studies have numerous methodological issues but can be hypothesis generating for more rigorous trials.
Though this effect size is quite large, it’s not completely implausible: There have been several ICU studies that achieved large reductions in mortality. Furthermore, several pre-clinical and observational studies suggest mechanisms whereby vitamin C could be beneficial. Biochemically, vitamin C is an essential cofactor for numerous enzymes involved in oxidative phosphorylation, collagen synthesis (necessary for wound healing), and in synthesis of endogenous catecholamines (necessary for blood pressure support). Vitamin C also limits reactive oxygen species (ROS) generation, reduces inflammation while enhancing neutrophil and lymphocyte function. Additionally, small observational studies have found that patients with sepsis have low levels of vitamin C compared to healthy controls.
Over the last 3 years, several randomized controlled trials of vitamin C in sepsis have been completed, and several more are still ongoing. These trials differ slightly in design, intervention, control, and primary endpoints.
| Study | Inclusion | Sites | Blinding | Control | n |
|---|---|---|---|---|---|
| CITRIS-ALI | Sepsis & ARDS | 7 ICUs | Double blind | Placebo | 167 |
| VITAMINS | Septic shock | 10 ICUs | Open label | Hydrocortisone | 216 |
| HYVCTTSSS | Sepsis & septic shock | 1 ICU | Single blind | Placebo | 140 |
| ORANGES | Sepsis & septic shock | 1 ICU | Double blind | Placebo | 140 |
| ATESS | Septic shock | 4 EDs | Double blind | Placebo | 116 |
| ACTS | Septic shock | 14 ICUs | Double blind | Placebo | 200 |
While all 6 studies used vitamin C, dosing differed slightly, and 5 of the 6 administered vitamin C in combination with other therapies, as shown:
| Study | Hydrocortisone | Ascorbic acid | Thiamine |
|---|---|---|---|
| CITRIS-ALI | - | 50 mg/kg q6 | - |
| VITAMINS | 50 mg q6 | 1.5 gm q6 | 100 mg BID |
| HYVCTTSSS | 50 mg q6 | 1.5 gm q6 | 200 mg BID |
| ORANGES | 50 mg q6 | 1.5 gm q6 | 200 mg BID |
| ATESS | - | 50 mg/kg q6 | 200 mg BID |
| ACTS | 50 mg q6 | 1.5 gm q6 | 100 mg BID |
5 of 6 studies were negative for their primary endpoint. One study (ORANGES) had two primary endpoints, and only one reached significance.
| Study | Primary Endpoint | Significant |
|---|---|---|
| CITRIS-ALI | ΔSOFA score | NO |
| VITAMINS | Vasopressor free days | NO |
| HYVCTTSSS | Hospital mortality | NO |
| ORANGES | ΔSOFA score, Time to resolution of shock | NO, YES |
| ATESS | ΔSOFA score | NO |
| ACTS | ΔSOFA score | NO |
For 4 of the 6 studies, change in SOFA score was a primary endpoint and all 6 studies reported the ΔSOFA score. For 4/6 there was no significant difference in SOFA with the intervention. The average difference in SOFA score with vitamin C treatment was 3.3 points compared to 2.6 points in the control group. This means that patients who received vitamin C had on average a 0.6 point greater drop in SOFA score compared to controls.
| Study | ΔSOFA timeframe | ΔSOFA (intervention) | ΔSOFA (control) | Significant ΔSOFA score |
|---|---|---|---|---|
| CITRIS-ALI | 96 hrs | 3 | 3.5 | NO |
| VITAMINS | 72 hrs | 2.0 | 1.0 | YES |
| HYVCTTSSS | 72 hrs | 3.5 | 1.8 | YES |
| ORANGES | 72 hrs | 3.4 | 2.3 | NO |
| ATESS | 72 hrs | 3.0 | 3.0 | NO |
| ACTS | 72 hrs | 4.7 | 4.1 | NO |
Despite the large decrease in mortality described in the original Marik et al article, 5/6 RCTs did not find a significant difference in mortality. In fact in 4/6 studies, mortality was actually higher in the vitamin C treated cohort than the control:
| Study | Mortality Timeframe | Mortality (intervention) | Mortality (control) | OR | Significant Δmortality |
|---|---|---|---|---|---|
| CITRIS-ALI | 28 day | 29.8% | 45.8% | 0.65 | YES |
| VITAMINS | 28 day | 22.4% | 20.2% | 1.11 | NO |
| HYVCTTSSS | 28 day | 27.5% | 16.9% | 1.63 | NO |
| ORANGES | 28 day | 16.2% | 18.8% | 0.86 | NO |
| ATESS | 28 day | 20.8% | 15.5% | 1.34 | NO |
| ACTS | 30 day | 34.7% | 29.3% | 1.18 | NO |
While there are methodological differences between these studies, it is reasonable to combine them into a meta-analysis.
I chose 28-day mortality as the primary outcome in the meta-analysis because (1) this is a patient-centered endpoint, (2) it was widely reported across studies, and (3) this is what Marik et al looked at when touting their cure.
Both of these Forrest plots examine mortality. The first plot looks at studies of HAT therapy; the second includes all vitamin C studies. Other than Marik et al, every study listed is a RCT.
Notice just how much of an outlier Marik’s original study is when placed alongside the more rigorous RCTs. The overall effect size is non-significant with an OR = 0.85 or 0.87. If we were to look only at RCTs (excluding Marik et al), the actual effect size would be closer to 1.0.
Mortality can be a difficult endpoint to achieve in individual ICU trials. While some important ICU studies have demonstrated improved survival (ABC, ARDSNet, PROSEVA, DEXA-ARDS, FLORALI, HYPERION, etc), many therapies we use routinely are not proven to reduce mortality. This is where multiple studies and meta-analysis can be helpful. If there were true clinical benefit with vitamin C it should be more apparent in meta-analysis. Instead we see the opposite trend.
Progress in critical care is often punctuated by setbacks and disappointment. I suspect that HAT therapy will join early goal directed therapy (EGDT), the pulmonary artery catheter (PAC), and activated protein C (APC) among the ranks of failed ICU therapies, but the story may not be over yet. More trials of vitamin C in sepsis are still ongoing. For example the massive 43 site, VICTAS trial has yet to be published (it was stopped early after enrolling n=501 patients of a planned 2000). I will update this analysis as more high quality RCTs are published.
But until more evidence is published, I conclude that HAT therapy is unlikely to be beneficial…
Summary:
Vitamin C (or HAT therapy) for sepsis once appeared promising based on a single-center, retrospective, before/after study but 6 high-quality, randomized controlled studies have failed to replicate the benefit.
5/6 studies were negative for their primary endpoints.
5/6 studies failed to demonstrate a significant decrease in mortality, in fact 4/6 studies actually show a non-significant increase in mortality with vitamin C treatment.
4/6 studies failed to demonstrate improvement in SOFA scores with treatment. The average decrease in SOFA score at 72-96 hours was less than 1 and is neither statistically or clinically significant.
My meta-analysis of the 6 published RCTs fails to show a mortality benefit for patients with sepsis treated with vitamin C.
Data/code availability:
You can download my raw data here.
The meta-analysis was done using R version 3.6.2 with the MetaFoR package.