Thread by @WellingtonICU

Our first @WellingtonICU Twitter journal club was prepared by @ETDETDETDETD & looks at SUP-ICU, a multicentre RCT examining the impact on mortality of
pantoprazole vs placebo for ulcer prophylaxis: nejm.org/doi/full/10.10…

PPIs are a common feature on ICU drug charts for ulcer prophylaxis. The authors point out that Pantoprazole isn’t licensed for prophylaxis by the FDA. Medsafe has approved it for prevention of NSAID-induced ulcers in NZ

There have been concerns raised regarding risks of PPI therapy, notably infections (particuarly C.Diff and pneumonia) and also myocardial ischaemia

SUP-ICU enrolled non-elective ICU admissions who were deemed at risk of GI bleeding (shock, anticoagulation, RRT, mechanical ventilation >24h, any history of liver disease and coagulopathy) were randomised. 3298 patients in total were randomised

The intervention arm received 40mg of IV pantoprazole daily. The primary outcome was death by 90 days post randomisation, with secondary endpoints of GI bleeding, pneumonia, C.Diff infection or acute myocardial ischaemia – or a composite of one or more of these events.

The authors’ hypothesis was that pantoprazole would reduce GI bleeding, but at the cost of increased nosocomial infection and myocardial ischaemia. The primary outcome however was mortality by 90 days post randomisation.

31.1% of patients in the intervention group, and 30.4% of patients in the control group had died at 90 days. (RR 1.02; 95% CI: 0.91-1.13, p=0.76)

Subgroup analyses suggested higher mortality for sicker patients (SAPS scores over vs. under 53), who received pantoprazole

Other subgroup analyses showed there was no difference in response to treatment for patients with and without hepatic dysfunction, shock vs. no shock, & ventilated vs. not

In the intervention arm, 2.5% of patients had significant GI bleeding, compared with 4.2% in the control group. (NNT 59)

The rates of infectious events were virtually identical between groups
(16.8% and 16.9%). C. Diff occurred in 1.5% in the PPI group and 1.2% in placebo group.

The rates of myocardial ischaemia are in the supplementary index. In the intervention arm there were 77 events (4.7%) compared with 66 (4.0%) in the placebo group.

A 2018 meta-analysis of observational studies showed an association between PPI use and cardiac events, but that this was not shown in meta-analysis of seven RCTs. onlinelibrary.wiley.com/doi/abs/10.111….

The SUP-ICU findings are broadly consistent with another 2018 meta-analysis of 57 trials, examining PPIs, H2RBs and sucralfate. link.springer.com/article/10.100…

This meta-analysis found that PPIs were found to be most effective, mortality was felt to be equal across groups (moderate quality), but PPIs ‘probably’ did increase pneumonia

Diagnostic endoscopy wasn’t mandated so we don’t know whether GI bleeding was from a peptic ulcer and erosion or some other source less likely to have been prevented by pantoprazole. If anything this may dilute the effect of the intervention on ulcer prevention.

It would have been interesting to know what mortality was from GI bleeding that developed in ICU. My guess would be none.

The SUP ICU authors in a 2015 observational study of 1034 patients found that clinically significant GI bleeding in ICU was not associated with 90 day mortality. (This study did help identify risk factors however).
link.springer.com/article/10.100…

On that note, Josh Farkas at @emcrit has an articulate description of the difficulties of having mortality as a primary endpoint, for a condition that has low mortality. emcrit.org/pulmcrit/sup-i…

Perhaps the more useful information in SUP ICU is in the secondary outcomes. It is reassuring there weren’t higher rates of infection or myocardial ischaemia, and consistent with other studies. Pantoprazole appears to halve rates of GI bleeding.

As an aside, there appears to be a healthy literature examining PPI use in alpacas and llamas. The efficacy and pharmacokinetics of pantoprazole in alpacas was described in 2010. doi.org/10.1111/j.1939…

It turns out that gastric ulcers are a common cause of mortality in New World camelids. But can you imagine how hard it must have been to intubate those alpacas??

On that note, experts in the field emphasise that ‘camelid intubation typically requires additional personnel compared with other species.’ ncbi.nlm.nih.gov/pmc/articles/P…

Perhaps it’s all that Andean coffee. This too has received attention: 3mg/kg of IV caffeine for your average adult Llama doesseems a hefty amount, but as it turns out it does ‘not induce signs of clinical excitement.’ Or ulcers. doi.org/10.2460/ajvr.6…

side note from @DogICUma
Under the assumption that all subgroup pairs have the same response to treatment, you can calculate the chance you will see one P of 0.05 or less easily.

Imagine you are going to toss a coin once. What is the chances of getting a head? [hint: this is not a trick question]

What are the chances of getting two heads if you toss a coin twice?

The answer is 1/2 to the power of the number of coin tosses (in this case two)

For a trial the number of comparisons can be thought of as the number of coin tosses. In this case, there were 6 subgroup comparisons.

Under the assumption of no difference, with one comparison, the chance of a P value of 0.05 or less is 5%. This is the same as a coin that has a 5% chance of landing on one side (the P<0.05 side) and a 95% chance of landing on the other (the P>0.05 side).

The chances of landing on the 'no difference' side 6 times is 0.95^6 (the ^ symbol is the lazy way of writing to the power of) which is 0.74 (74%).

Even if there is no difference, there is 26% chance of one P value of 0.05 of less.

side note from @DogICUma
1. some patients who had GI bleeds that met the definition of clinically significant might have had coincident reasons for transfusion or increase NA (i.e. they might not have been clinically significant bleeds at all

2. most GI bleeds were treated by just giving PPIs (i.e. a lot of PPI was given to prevent needed to give a PPI)

3. a reasonable position IMOP is that PPIs appear to be well tolerated & GI bleeding events can be distressing for families so giving PPIs is reasonable for this reason alone

@threadreaderapp unroll please