Quick Hit Article #9: Thiamine for severe sepsis? Yes, No and Maybe?

Today’s quick hit article is probably the first in a long list of articles to come looking at Thiamine, vitamin C, or steroids for septic shock.

This article looks only at Thiamine without vitamin C or steroids. This study was a retrospective chart review looking at a cohort of patients with sepsis who were given thiamine and matched to a cohort without thiamine given. I’ll be honest it is not an straightforward read. They base it more on clearance of lactate and survival at 28 day mortality. Survival analysis is done with Cox hazard ratios and Kaplan-Meier Curves.  When you look at the raw data it doesn’t seem to show much difference however when they do regresssion analysis AND adjust for: Thiamine, age, sex, race, and clinical factors where clinical fractures are: Liver disease, Elixhauser comorbidity index, ICU service (medical vs surgical), Sequential Organ Failure Assessment score day 1 of ICU admission, hydrocortisone, and peak lactate… phew… THEN you get a Hazard Ratio of 0.666…. Funny isn’t that the devils number? I’m concerned there are a lot of statistical waterboarding going on here to make the data talk. None of the adjusted variables were mentioned a priori (like how would you predict women do better with Thiamine????) nor were the p-values listed for the multiple regression. Those would have been nice to see.

A “NERDY” aside: The authors state in the abstract that “Thiamine administration was also associated with a reduction in 28-day mortality (hazard ratio, 0.666; 95% CI, 0.490–0.905)”. This is technically not a correct way to report a hazard ratio. A hazard ratio is not the same as a risk reduction. A hazard ratio is a rate. A helpful analogy would be if we were driving in a car and you asked how long would it take to get to our destination. If I told you we are going a velocity of 20 mph the only way you would know how long to get there would be to know how far we have to travel. It doesn’t actually tell you how much more quickly we would get there. The proper way to report these results would be to say for example: “Those in the thiamine group (when adjusted for multiple factors like gender, liver disease, etc) reached a 28 day survival faster than the non thiamine group by 33% (HR = 0.666, CI 0.490-0.905)”. The 33% is calculated as (1-HR)x100. Confusing hazard ratios and risk reduction is not uncommon in medicine [1].

REFERENCES: