A randomised controlled trial examining a treatment for lung injury reports the following results:

"The primary outcome was change in SOFA score over 96 hours. The mean SOFA score from baseline to 96 hours decreased from 9.8 to 6.8 in the treatment group (3 points) and from 10.3 to 6.8 in the placebo group (3.5 points) (difference, -0.10; 95% CI, - 1. 23 to 1.03; P = 0.86).

There were 30 prespecified secondary outcomes. Twenty-nine were not significantly different between the treatment and the placebo group. In exploratory analyses that did not adjust for multiple comparisons, day 28 mortality was 46.3% in the placebo group vs 29.8% in the treatment group (P = 0.03; between-group difference, 16.58% [95% CI, 2% to 31.1%))."

a)  Interpret these results.    (30% marks)
 

[Click here to toggle visibility of the answers]

College answer

a)
The primary outcome does not demonstrate a significant difference between the two groups and so the overall result of the trial is negative. A secondary outcome of day 28-day mortality does show a significant difference in favour of the treatment – however as this is one of 30 secondary outcomes, with no adjustment for multiplicity of testing this is likely a false positive result and should be interpreted cautiously. (3 marks)
 

Discussion

Thee findings borrowed for this SAQ come from the CITRIS-ALI trial (Truwit et al, 2019), in case anybody cares. 

The primary outcome is not statistically significant because of the high p-value (0.86 is pretty terrible) and because the confidence interval crosses over 1.0

As to the secondary outcome. If you have thirty (and ultimately CITRIS-ALI had forty-six) some of them are bound to produce some sort of publishable information. Day 28 mortality difference was statistically significant (p= 0.03), but because this is a secondary outcome, it should be viewed as hypothesis-generating. On an unrelated note, mortality of 46% in sepsis or ARDS is so 1990s. 

References